Professional Documents
Culture Documents
K. Anders Ericsson, Neil Charness, Paul J. Feltovich, Robert R. Hoffman-The Cambridge Handbook of Expertise and Expert Performance (2006) PDF
K. Anders Ericsson, Neil Charness, Paul J. Feltovich, Robert R. Hoffman-The Cambridge Handbook of Expertise and Expert Performance (2006) PDF
K. Anders Ericsson, Neil Charness, Paul J. Feltovich, Robert R. Hoffman-The Cambridge Handbook of Expertise and Expert Performance (2006) PDF
Edited by
K. Anders Ericsson
Florida State University
Neil Charness
Florida State University
Paul J. Feltovich
Florida Institute for Human and Machine Cognition
Robert R. Hoffman
Florida Institute for Human and Machine Cognition
Cambridge, New York, Melbourne, Madrid, Cape Town, Singapore, So Paulo
- ---- hardback
- --- hardback
Acknowledgements page xi
Contributors xiii
pa r t i
INTRODUCTION AND PERSPECTIVE
1. An Introduction to The Cambridge Handbook of Expertise and Expert Performance:
Its Development, Organization, and Content 3
K. Anders Ericsson
p a r t ii
OVERVIEW OF APPROACHES TO THE STUDY OF EXPERTISE BRIEF
HISTORICAL ACCOUNTS OF THEORIES AND METHODS
4. Studies of Expertise from Psychological Perspectives 41
Paul J. Feltovich, Michael J. Prietula, & K. Anders Ericsson
5. Educators and Expertise: A Brief History of Theories and Models 69
Ray J. Amirault & Robert K. Branson
6. Expert Systems: A Perspective from Computer Science 87
Bruce G. Buchanan, Randall Davis, & Edward A. Feigenbaum
vii
viii contents
p a r t iii
METHODS FOR STUDYING THE STRUCTURE OF EXPERTISE
8. Observation of Work Practices in Natural Settings 127
William J. Clancey
9. Methods for Studying the Structure of Expertise: Psychometric
Approaches 147
Phillip L. Ackerman & Margaret E. Beier
10. Laboratory Methods for Assessing Experts and Novices Knowledge 167
Michelene T. H. Chi
11. Task Analysis 185
Jan Maarten Schraagen
12 . Eliciting and Representing the Knowledge of Experts 203
Robert R. Hoffman & Gavan Lintern
13. Protocol Analysis and Expert Thought: Concurrent Verbalizations of
Thinking during Experts Performance on Representative Tasks 223
K. Anders Ericsson
14. Simulation for Performance and Training 243
Paul Ward, A. Mark Williams, & Peter A. Hancock
p a r t iv
METHODS FOR STUDYING THE ACQUISITION AND MAINTENANCE OF
EXPERTISE
15. Laboratory Studies of Training, Skill Acquisition, and Retention of
Performance 265
Robert W. Proctor & Kim-Phuon L. Vu
16. Retrospective Interviews in the Study of Expertise and Expert
Performance 287
Lauren A. Sosniak
17. Time Budgets, Diaries, and Analyses of Concurrent Practice Activities 3 03
Janice M. Deakin, Jean Cote, & Andrew S. Harvey
18. Historiometric Methods 3 19
Dean Keith Simonton
pa r t v
DOMAINS OF EXPERTISE
p a r t v.a
PROFESSIONAL DOMAINS
19. Expertise in Medicine and Surgery 339
Geoff Norman, Kevin Eva, Lee Brooks, & Stan Hamstra
Contents ix
p a r t v.b
ARTS, SPORTS, & MOTOR SKILLS
2 6. Music 45 7
Andreas C. Lehmann & Hans Gruber
2 7. Expert Performance in Sport: A Cognitive Perspective 471
Nicola J. Hodges, Janet L. Starkes, & Clare MacMahon
2 8. Artistic Performance: Acting, Ballet, and Contemporary Dance 489
Helga Noice & Tony Noice
2 9. Perceptual-Motor Expertise 5 05
David A. Rosenbaum, Jason S. Augustyn, Rajal G. Cohen, & Steven A. Jax
p a r t v.c
GAMES AND OTHER TYPES OF EXPERTISE
30. Expertise in Chess 5 23
Fernand Gobet & Neil Charness
31. Exceptional Memory 539
John M. Wilding & Elizabeth R. Valentine
32 . Mathematical Expertise 553
Brian Butterworth
33. Expertise in History 5 69
James F. Voss & Jennifer Wiley
p a r t vi
GENERALIZABLE MECHANISMS MEDIATING EXPERTISE AND GENERAL
ISSUES
34. A Merging Theory of Expertise and Intelligence 5 87
John Horn & Hiromi Masunaga
x contents
Anders Ericsson wants to gratefully ack- (Clemson University), Alice F. Healy (Uni-
nowledge the financial support provided by versity of Colorado), Anastasia Kitsantas,
the John D. and Catherine T. MacArthur (George Mason University), Reinhold Kliegl
Foundation, Grant #3 2005 -0, which sup- (University of Potsdam, Germany), Ralf Th.
ported the planning and the invitation of Krampe (University of Leuven, Belgium),
handbook authors during his year as a Fel- Richard E. Mayer (University of Califor-
low at the Center for Advanced Study in the nia, Santa Barbara), Daniel Morrow (Uni-
Behavioral Sciences. He also would like to versity of Illinois at Urbana-Champaign),
credit the Conradi Eminent Scholar Endow- Kathleen Mosier (San Francisco State Uni-
ment at the Florida State Foundation for its versity), Gary D. Phye (Iowa State Univer-
support during the editing phase of the work sity), Mauro Pesenti (Universite Catholique
on the handbook. de Louvain, Belgium), Pertti Saariluoma
Neil Charness gratefully acknowledges (University of Jyvaskyla, Finland), Mike
support from the National Institutes of Saks (University of Lincoln, UK), John
Health / National Institute on Aging, Grants B. Shea (Indiana University), Dean Keith
R01 AG13 969 and 1P01 AG 17211, that per- Simonton (University of California, Davis),
mitted him both to edit and contribute to J. Michael Spector (Florida State Univer-
chapters in this handbook. sity), Janet L. Starkes (McMaster Univer-
Paul Feltovich and Robert Hoffman sity, Canada), Gershon Tenenbaum (Florida
would like to acknowledge the Florida Insti- State University), Oliver Vitouch (Univer-
tute for Human and Machine Cognition sity of Klagenfurt, Austria), and Richard K.
for support during the preparation of the Wagner (Florida State University) for their
handbook. full-length reviews of particular chapters,
We also want to thank M. Anne Britt along with the numerous authors of chap-
(Northern Illinois University), Jamie I. D. ters within the handbook itself, who pro-
Campbell (University of Saskatchewan, vided insightful comments and suggestions
Canada), Randall Davis (MIT), Leo Gugerty for other chapters in this volume.
xi
Contributors
xiii
xiv contributors
INTRODUCTION AND
PERSPECTIVE
CHAPTER 1
K. Anders Ericsson
A significant milestone is reached when a some special field (Websters New World
field of scientific research matures to a point Dictionary, 1968, p. 168), or someone widely
warranting publication of its first handbook. recognized as a reliable source of knowl-
A substantial body of empirical findings, edge, technique, or skill whose judgment is
distinctive theoretical concepts and frame- accorded authority and status by the pub-
works, and a set of new or adapted meth- lic or his or her peers. Experts have pro-
ods justify a unifying volume. The growth of longed or intense experience through prac-
this field is evident from the publication of a tice and education in a particular field
series of edited books on diverse sets of skills (Wikipedia, 2005 ). Expertise then refers to
and expertise from many domains during the the characteristics, skills, and knowledge that
last several decades (Anderson, 1981; Bloom, distinguish experts from novices and less
1985 a; Chase, 1973 ; Chi, Glaser, & Farr, 1988; experienced people. In some domains there
Ericsson, 1996a; Ericsson & Smith, 1991a; are objective criteria for finding experts,
Feltovich, Ford, & Hoffman, 1997; Hoffman, who are consistently able to exhibit supe-
1992; Starkes & Allard, 1993 ; Starkes & rior performance for representative tasks
Ericsson, 2003 ). And as in many other fields, in a domain. For example, chess masters
the name of a branch of scientific study, in will almost always win chess games against
our case expertise and expert performance, recreational chess players in chess tour-
often communicates the domain of studied naments, medical specialists are far more
phenomena. likely to diagnose a disease correctly than
advanced medical students, and professional
musicians can perform pieces of music
Expert, Expertise, and Expert in a manner that is unattainable for less
Performance: Dictionary Definitions skilled musicians. These types of superior
reproducible performances of representative
Encyclopedias describe an Expert as one tasks capture the essence of the respective
who is very skillful and well-informed in domains, and authors have been encouraged
3
4 the cambridge handbook of expertise and expert performance
wealthy or well-born he is, they still will ity of the craft and the age and prior experi-
have none of him, but jeer at him and create ence of the apprentice (Epstein, 1991). Once
an uproar, until either the would-be speaker an apprentice had served out their contract
is shouted down and gives up of his own they were given a letter of recommendation
accord, or else the police drag him away or and were free to work with other masters
put him out on the order of the presidents.
for pay, which often involved traveling to
(Plato, 1991, pp. 1112 )
other cities and towns they were there-
Aristotle relied on his own senses as the fore referred to as journeymen. When a jour-
primary source of scientific knowledge and neyman had accumulated enough additional
sought out beekeepers, fishermen, hunters, skill and saved enough money, he, or occa-
and herdsmen to get the best and most reli- sionally she, would often return to his home
able information for his books on science town to inherit or purchase a shop with tools
(Barnes, 2000). He even tried to explain and apply to become a master of the guild.
occasional incorrect reports from some of his In most guilds they required inspection of
informants about how offspring of animals the journeymans best work, that is, master
were generated. For example, some of them pieces, and in some guilds they administered
suggested that the ravens and the ibises special tests to assess the level of perfor-
unite at the mouth (Aristotle, 2000, p. 3 15 ). mance (Epstein, 1991). When people were
But Aristotle notes: It is odd, however, that accepted as masters they were held responsi-
our friends do not reason out how the semen ble for the quality of the products from their
manages to pass through the stomach and shop and were thereby allowed to take on
arrive in the uterus, in view of the fact that the training of apprentices (See Amirault &
the stomach concocts everything that gets Branson, Chapter 5 , and Chi, Chapter 2, on
into it, as it does the nourishment (pp. 3 15 the progression toward expertise and mas-
& 3 17). Similarly, those who assert that the tery of a domain).
female fishes conceive as a result of swallow- In a similar manner, the scholars guild
ing the males semen have failed to notice was established in the 12th and 13 th cen-
certain points (p. 3 11). Aristotle explains tury as a universitas magistribus et pupil-
that Another point which helps to deceive lorum, or guild of masters and students
these people is this. Fish of this sort take only (Krause, 1996, p. 9). Influenced by the
a very short time over their copulation, with University of Paris, most universities con-
the result that many fishermen even never ducted all instruction in Latin, where the
see it happening, for of course no fishermen students were initially apprenticed as arts
ever watches this sort of thing for the sake of students until they successfully completed
pure knowledge (p. 3 13 ). Much of Aristo- the preparatory (undergraduate) program
tles knowledge comes, at least partly, from and were admitted to the more advanced
consensus reports of professionals. programs in medicine, law, or theology. To
Much later during the Middle Ages, become a master, the advanced students
craftsmen formed guilds to protect them- needed to satisfy a committee of examin-
selves from competition. Through arrange- ers, then publicly defending a thesis, often
ments with the mayor and/or monarch they in the town square and with local grocers
obtained a monopoly on providing partic- and shoemakers asking questions (Krause,
ular types of handcraft and services with 1996, p. 10). The goal of the universities was
set quality standards (Epstein, 1991). They to accumulate and explain knowledge, and
passed on their special knowledge of how in the process masters organized the exist-
to produce products, such as lace, barrels, ing knowledge (See Amirault & Branson,
and shoes, to their students (apprentices). Chapter 5 ). With the new organization of
Apprentices would typically start at around the existing knowledge of a domain, it was
age 14 and commit to serve and study with no longer necessary for individuals to dis-
their master for around 7 years the length cover the relevant knowledge and methods
of time varied depending on the complex- by themselves.
6 the cambridge handbook of expertise and expert performance
Todays experts can rapidly acquire the all types of skills, such as tannery, carpentry,
knowledge originally discovered and accu- glassmaking, and ironworking (Pannabecker,
mulated by preceding expert practitioners 1994), along with descriptions of how to
by enrolling in courses taught by skilled sharpen a feather for writing with ink, as
and knowledgeable teachers using specially shown in Figure 1.1. His goal was to describe
prepared textbooks. For example, in the all the raw materials and tools that were nec-
13 th century Roger Bacon argued that it essary along with the methods of produc-
would be impossible to master mathematics tion. Diderot and his associate contributors
by the then-known methods of learning had considerable difficulties gaining access
(self-study) in less than 3 0 to 40 years to all the information because of the unwill-
(Singer, 195 8). Today the roughly equiva- ingness of the guild members to answer their
lent material (calculus) is taught in highly questions. Diderot even considered sending
organized and accessible form in every high some of his assistants to become apprentices
school. in the respective skills to gain access to all the
Sir Francis Bacon is generally viewed as relevant information (Pannabecker, 1994). In
one of the architects of the Enlightenment spite of all the information and pictures (dia-
period of Western Civilization and one of grams of tools, workspaces, procedures, etc.,
the main proponents of the benefits of gen- as is illustrated in Figure 1.2 showing one
erating new scientific knowledge. In 1620 of several plates of the process of printing)
he described in his book Novum Organum provided in the Encyclopedie, Diderot was
his proposal for collecting and organizing under no illusion that the provided informa-
all existing knowledge to help our civiliza- tion would by itself allow anyone to become
tion engage in learning to develop a bet- a craftsman in any of the described arts and
ter world. In it, he appended a listing of wrote: It is handicraft that makes the artist,
all topics of knowledge to be included in and it is not in Books that one can learn
Catalogus Historarium Particularium. It to manipulate (Pannabecker, 1994, p. 5 2).
included a long list of skilled crafts, such In fact, Diderot did not even address the
as History of weaving, and of ancillary higher levels of cognitive activity, such as
skills associated with it, History of dyeing, intuitive knowledge, experimentation, per-
History of leather-working, tanning, and of ceptual skills, problem-solving, or the anal-
associated ancillary skills (Rees & Wakely, ysis of conflicting or alternative technical
2004, p. 483 ). approaches (Pannabecker, 1994, p. 5 2).
The guilds guarded their knowledge and A couple of years after the French revo-
their monopoly of production. It is there- lution the monopoly of the guilds as elim-
fore not surprising that the same forces that inated (Fitzsimmons, 2003 ), including the
eventually resulted in the French revolu- restrictions on the practice of medicine and
tion were directed not only at the oppres- law. After the American Revolution and the
sion by the king and the nobility, but also creation of the United States of America
against the monopoly of services provided laws were initially created to require that
by the members of the guilds. Influenced by doctors and lawyers be highly trained based
Sir Francis Bacons call for an encyclopedic on the apprenticeship model, but pressure to
compilation of human knowledge, Diderot eliminate elitist tendencies led to the repeal
and DAlembert worked on assembling all of those laws. From 1840 to the end of the
available knowledge in the first Encyclopedie 19th century there was no requirement for
(Diderot & DAlembert, 196667), which certification to practice medicine and law
was originally published in 175 180. in the United States (Krause, 1996). How-
Diderot was committed to the creation of ever, with time both France and America
comprehensive descriptions of the mechan- realized the need to restrict vital medical
ical arts to make their knowledge available and legal services to qualified profession-
to the public and to encourage research and als and developed procedures for training
development in all stages of production and and certification.
Figure 1.1. An illustration for how to sharpen a goose feather for writing with ink from Plate IV
in the entry on Ecriture in the 23 rd volume of Encyclopedie ou dictionnare de raisonne des
sciences, des artes et des metier (Diderot & DAlembert, 196667).
Figure 1.2 . An illustration of the workspace of a printer with some of his type elements from
Plate I in the entry on Imprimerie in the 28th volume of Encyclopedie ou dictionnare de
raisonne des sciences, des artes et des metier (Diderot & DAlembert, 196667).
introduction 9
Over the last couple of centuries there the instruction and training of future pro-
have been several major changes in the rela- fessionals, has occurred, until quite recently,
tion between master and apprentice. For almost exclusively within each domain with
example, before the middle of the 19th cen- little cross-fertilization of domains in terms
tury children of poor families would often of teaching, learning methods, and skill-
be taken on by teachers in exchange for training techniques.
a contractual claim for part of the future It is not immediately apparent what is
dancers, singers, or musicians earnings as generalizable across such diverse domains of
an adult (Rosselli, 1991). Since then the expertise, such as music, sport, medicine,
state has gotten more involved in the train- and chess. What could possibly be shared
ing of their expert performers, even out- by the skills of playing difficult pieces
side the traditional areas of academia and by Chopin, running a mile in less than
professional training in medicine, law, busi- four minutes, and playing chess at a high
ness, and engineering. In the late 19th cen- level? The premise for a field studying
tury, public institutions such as the Royal expertise and expert performance is that
Academy of Music were established to pro- there are sufficient similarities in the the-
mote the development of very high levels of oretical principles mediating the phenom-
skill in music to allow native students com- ena and the methods for studying them in
pete with better trained immigrants (Rohr, different domains that it would be possi-
2001). In a similar manner during the lat- ble to propose a general theory of exper-
ter part of the 20th century, many countries tise and expert performance. All of these
invested in schools and academies for the domains of expertise have been created by
development of highly skilled athletes for humans. Thus the accumulated knowledge
improved success in competitions during the and skills are likely to reflect similarities
Olympic Games and World Championships in structure that reflect both human bio-
(Bloomfield, 2004). logical and psychological factors, as well
More generally, over the last century there as cultural factors. This raises many chal-
have been economic developments with lenging problems for methodologies used
public broadcasts of competitions and per- to describe the organization of knowledge
formances that generate sufficient revenue and mechanisms and reveals the medi-
for a number of domains of expertise, such as ating expert performance that generalizes
sports and chess, to support professional full- across domains.
time performers as well as coaches, train- Once we know how experts organize
ers, and teachers. In these new domains, their knowledge and their performance, is it
along with the traditional professions, cur- possible to improve the efficiency of learn-
rent and past expert performers continue ing to reach higher levels of expert perfor-
to be the primary teachers at the advanced mance in these domains? It should also be
level (masters), and their professional asso- possible to answer why different individ-
ciations have the responsibility of certifying uals improve their performance at differ-
acceptable performance and the permission ent rates and why different people reach
to practice. Accordingly, they hold the clout very different levels of final achievement.
in thus influencing training in professional Would a deeper understanding of the devel-
schools, such as law, medical, nursing, and opment and its mediating mechanisms make
business schools testing is the tail that it possible to select individuals with unusual
wags the dog (Feltovich, personal commu- potential and to design better developmen-
nication) as well as continuing education tal environments to increase the proportion
training (see Evetts, Meig, & Felt, Chapter 7 of performers who reach the highest levels?
on sociological perspectives on expertise). Would it be possible even to facilitate the
The accumulation of knowledge about the development of those rare individuals who
structure and acquisition of expertise in a make major creative contributions to their
given domain, as well as knowledge about respective domains?
10 the cambridge handbook of expertise and expert performance
the experts knowledge (Chi, Feltovich, & (1962, 1966) is generally recognized as the
Glaser, 1981; Chi, Glaser, & Rees, 1982) and first critic who saw that nonconscious and
on methods to extract the experts knowl- intuitive mediation limits the possibility of
edge to build computer-based models emu- eliciting and mapping the knowledge and
lating the experts performance (Hoffman, rules that mediates experts intuitive actions.
1992). These approaches have studied Subsequent discussion of the development
experts, namely, individuals who are socially of expertise by Dreyfus and Dreyfus (1986)
recognized as experts and/or distinguished and Benner (1984) has argued that the high-
by their extensive experience (typically over est levels of expertise are characterized by
10 years) and by knowledge of a particular contextually based intuitive actions that are
subject attained through instruction, study, difficult or impossible to report verbally.
or practical experience. The work of Robert Several chapters in this handbook propose
Glaser, Micheline Chi, and Paul Feltovich methods for uncovering tacit knowledge
examined the representations of knowledge about the successful development of exper-
and problem solutions in academic domains, tise (Cianciolo, Matthew, Wagner, & Stern-
such as physics (See Chi, Chapters 3 and 10). berg, Chapter 3 5 ), about methods of work
Of particular importance, Chi studied chil- through observation (Clancey, Chapter 8),
dren with extensive knowledge of chess and Concept Mapping (Hoffman & Lintern,
dinosaurs (See Chi, Chapter 10), and found Chapter 12), similarity judgment (Chi,
these children displayed many of the same Chapter 10), and traditional psychometric
characteristics of the knowledge representa- analyses of individual differences in perfor-
tion of adult experts. This work on exper- mance (Ackerman & Beier, Chapter 9) or
tise is summarized in Feltovich, Prietula, and simulated environments (Ward, Williams, &
Ericsson, Chapter 4, Chi, Chapter 10, and Hancock, Chapter 14). Other investigators
Hoffman and Lintern, Chapter 12, and in argue that expert performers often continue
a couple of edited volumes (Chi, Glaser, & to engage in deliberate practice in order to
Farr, 1988; Starkes & Allard, 1993 ). improve and that these performers have
In a parallel development in the com- to actively retain and refine their mental
puter science of the late 1970s and early representations for monitoring and control-
1980s, Ed Feigenbaum and other researchers ling their performance. This retained abil-
in the area of artificial intelligence and cog- ity to monitor performance allows them to
nitive science have attempted to elicit the give informative concurrent and retrospec-
knowledge of experts (Hoffman, 1992) and tive reports about the mediating sequences
to incorporate their knowledge in computer of thoughts (see Ericsson, Chapter 13 ).
models (c.f. expert systems) that seek to
replicate some of the decision making and
Expertise as Elite Achievement Resulting
behavior of experts (see Buchanan, Davis, &
from Superior Learning Environments
Feigenbaum, Chapter 6, and Hoffman &
Lintern, Chapter 12). There has been a long- There are other approaches to the study
standing controversy over whether highly of expertise that have focused on objec-
experienced experts are capable of articu- tive achievement. There is a long tradi-
lating the knowledge and methods that con- tion of influential studies with interviews
trol their generation of appropriate actions in of peer-nominated eminent scientists (Roe,
complex situations. 195 2) and analyses of biographical data on
The tradition of skill acquisition of Bryan Nobel Prize winners (Zuckerman, 1977)
and Harter (1899), Fitts and Posner (1967), (see Simonton, Chapter 18, 1994, for a
and Simon and Chase (1973 ) assumed that more extensive account). In a seminal study,
expert performance was associated with Benjamin Bloom and his colleagues (Bloom,
automation and was virtually effortless per- 1985 a) interviewed international-level per-
formance based on pattern recognition and formers from six different domains of exper-
direct access of actions. However, Polanyi tise ranging from swimming to molecular
introduction 13
genetics. All of the 120 participants had won sprinter will be able to reproduce superior
prizes at international competitions in their running performance on many tracks and
respective domains. They were all inter- even indoors in a large laboratory. Similarly,
viewed about their development, as were de Groot (1978) found that the ability to
their parents, teachers, and coaches. For select the best move for presented chess
example, Bloom and his colleagues collected positions is the best correlate of chess rat-
information on the development of athletes ings and performance at chess tournaments
who had won international competitions in a finding that has been frequently replicated
swimming and tennis. They also interviewed (Ericsson & Lehmann, 1996; van der Maas
artists who have won international competi- & Wagenmakers, 2005 ). Once it is possi-
tions in sculpting and piano playing and sci- ble to reproduce the reliably superior perfor-
entists who had won international awards in mance of experts in a controlled setting, such
mathematics and molecular biology. In each as a laboratory, it then becomes feasible to
of these six domains Bloom (1985 b) found examine the specific mediating mechanisms
evidence for uniformly favorable learning with experiments and process-tracing tech-
environments for the participants. Bloom niques, such as think aloud verbal reports
(1985 b) concluded that the availability of (see Ericsson, Chapter 13 , and Ericsson &
early instruction and support by their fam- Smith, 1991b). The discovery of represen-
ily appeared to be necessary for attaining tative tasks that measure adult expert per-
an international level of performance as an formance under standardized conditions in
adult. He found that the elite performers a controlled setting, such as a laboratory,
typically started early to engage in rele- makes it possible to measure and compare
vant training activities in the domain and the performance of less-skilled individuals
were supported both by exceptional teach- on the same tasks. Even more important,
ers and committed parents. One of the con- it allows scientists to test aspiring perform-
tributors to the Handbook, Lauren Sosniak ers many times during their development
(1985 a, 1985 b, 1985 c, 1985 d), describes in of expertise, allowing the measurement of
Chapter 16 the main findings from the orig- gradual increases in performance.
inal study (Bloom, 1985 a), along with more The new focus on the measurement
recent interview studies aimed to uncover of expert performance with standardized
the development of elite performers. tasks revealed that experts, that is, indi-
viduals identified by their reputation or
their extensive experience, are not always
Expertise as Reliably Superior (Expert)
able to exhibit reliably superior perfor-
Performance on Representative Tasks
mance. There are at least some domains
It is difficult to identify the many mediat- where experts perform no better than
ing factors that might have been responsi- less-trained individuals and that sometimes
ble for the elite performer to win an award experts decisions are no more accurate
and to write a groundbreaking book. When than beginners decisions and simple deci-
eminence and expertise is based on a sin- sion aids (Camerer & Johnson, 1991; Bolger
gular or small number of unique creative & Wright, 1992). Most individuals who
products, such as books, paintings, or musi- start as active professionals or as begin-
cal as compositions, it is rarely possible to ners in a domain change their behavior and
identify and study scientifically the key fac- increase their performance for a limited time
tors that allowed these people to produce until they reach an acceptable level. Beyond
these achievements. Consequently, Ericsson this point, however, further improvements
and Smith (1991b) proposed that the study appear to be unpredictable and the num-
of expertise with laboratory rigor requires ber of years of work and leisure experience
representative tasks that capture the essence in a domain is a poor predictor of attained
of expert performance in a specific domain performance (Ericsson & Lehmann, 1996).
of expertise. For example, a world-class Hence, continued improvements (changes)
14 the cambridge handbook of expertise and expert performance
in achievement are not automatic conse- ters by two of the pioneers of the study of
quences of more experience, and in those cognitive skill and expertise. Michelene Chi
domains where performance consistently (Chapter 2) describes two approaches to the
increases, aspiring experts seek out partic- study of expertise and Earl Hunt (Chapter 3 )
ular kinds of experience, that is, deliber- gives his general perspective on the princi-
ate practice (Ericsson, Krampe, & Tesch- pal factors related to expertise. In a recent
Romer, 1993 ). Such activities are designed, book Hunt (1995 ) has made a convincing
typically by a teacher, for the sole purpose case for the increasing importance of high
of effectively improving specific aspects of levels of skill in occupations of the future. He
an individuals performance. A large body argues that with the development of tech-
of research shows how deliberate prac- nology to automate less complex jobs the
tice can change mediating mechanisms and most important occupations of the future
that the accumulated amounts of deliber- will require creative design and planning that
ate practice are related to the attained level cannot be easily automated. He foresees a
of performance (see Ericsson, Chapter 3 8, rapidly increasing need to train students to
and Deakin, Cote, & Harvey, Chapter 17, even higher levels of expertise to continue
Zimmerman, Chapter 3 8, as well as the the development of our modern society. The
edited books by Ericsson [1996a] and Starkes key competitive differences between com-
& Ericsson [2003 ]). panies of the future may not have to do
with raw materials and monetary resources
General Comments but with human capital, namely, the abili-
ties of the employees. The Nobel Prize win-
In summary, there are a broad range of
ner Gary Becker has for a long time made
approaches to the study of the structure and
the case for the critical role of education
acquisition of expertise as well as expert per-
and human capital in our current industri-
formance. Although individual researchers
alized world, and especially the crucial role
and editors may be committed to one
of highly accomplished people. He (Becker,
approach over the others, this Handbook
2002) illustrated this claim by a quote
has been designed to fairly cover a wide
from Microsoft founder Bill Gates: Take our
range of approaches and research topics in
20 best people away and . . . Microsoft would
order to allow authors to express their dif-
become an unimportant company (Becker,
ferent views. However, the authors have
2002, p. 8).
been encouraged to describe explicitly their
The second section of the Handbook
empirical criteria for their key terms, such
contains reviews of the historical devel-
as experts and expert performance. For
opment of the study of expertise in four
example, the authors have been asked to
major disciplines, namely, psychology, edu-
report if the cited research findings involve
cation, computer science, and sociology.
experts identified by social criteria, criteria
Three pioneers in the psychological study of
of lengthy domain-related experience, or cri-
expertise, Paul Feltovich, Michael Prietula,
teria based on reproducibly superior perfor-
and Anders Ericsson, describe the develop-
mance on a particular set of tasks representa-
ment of the study of expertise in psychol-
tive of the individuals domain of expertise.
ogy (Chapter 4). One of the pioneers in the
development of instructional design, Robert
Branson, has together with Ray Amirault
General Outline of the Handbook (Chapter 5 ) described the role of exper-
tise in the historical development of educa-
The handbook is organized into six gen- tional methods and theories. Three of the
eral sections. First, Section 1 introduces the pioneers in the development of expert sys-
Handbook with brief accounts of general per- tems, Bruce Buchanan, Randall Davis, and
spectives on expertise. In addition to this Edward Feigenbaum (Chapter 6), describe
introductory chapter that outlines the orga- the role of expertise in shaping contempo-
nization of the handbook, there are chap- rary approaches in computer science and
introduction 15
artificial intelligence. Finally, Julia Evetts, chapter, Robert Proctor and Kim-Phuon
Harald Mieg, and Ulrike Felt (Chap- Vu (Chapter 15 ) describe how laboratory
ter 7) provide a description of the relevant methods for the study of skilled perfor-
approaches to the study of expertise from mance can inform research on expertise and
the point of view of sociology. expert performance. Lauren Sosniak (Chap-
The next two sections of the Handbook ter 16) discusses how she and her colleagues
review the core methods for studying the used retrospective interviews to describe
structure (Section 3 ) and acquisition (Sec- the development of expertise in the clas-
tion 4) of expertise and expert performance. sic studies led by Benjamin Bloom (1985 a),
Each of the chapters in Sections 3 and 4 along with some recent extensions of that
has been written by one of the pioneer- work. Janice Deakin, Jean Cote, and Andrew
ing researchers who have developed these Harvey (Chapter 17) use diaries and describe
methods and approaches for use in research different methods to study how expert per-
on expertise and expert performance. The formers spend their time and how experts
chapters consist of a historical background, allocate their practice time. In the final
a detailed description of the recommended chapter of this section, Dean Simonton
methodology with a couple of examples, (Chapter 18) reviews the methods of histo-
and a general review of the type of empir- riometrics and how data about the develop-
ical evidence that has been collected. In the ment of eminent performers can be collected
first chapter of Section 3 William Clancey and analyzed.
(Chapter 8) gives an overview of the ethno- Section 5 consists of fifteen chapters that
graphic observational methods for study- review our current knowledge about exper-
ing the behavior of experts. Philip Ack- tise and expert performance in particular
erman and Margaret Beier (Chapter 9) domains and represents the core of this
review the use of psychometric methods for Handbook. Each chapter has been written
studying expertise. Michelene Chi (Chap- by internationally respected experts on the
ter 10) describes how laboratory meth- associated areas of expertise and contains
ods have been used to assess the struc- a brief historic background followed by a
ture of knowledge. Jan Maarten Schraagen review and future directions. The chap-
(Chapter 11) describes how tasks presented ters in Section 5 have been broken down
to skilled and less-skilled individuals can into three subsections. The first subsec-
be analyzed and how a task analysis can tion is focused on different types of profes-
guide data analysis and theory construction. sional expertise, namely, medicine (Chap-
Robert Hoffman and Gavin Lintern (Chap- ter 19 by Geoff Norman, Kevin Eva, Lee
ter 12) review methods for how knowledge Brooks, and Stan Hamstra), transportation,
of experts can be elicited and represented by such as driving, flying, and airplane control
interviews, Concept Maps, and abstraction- (Chapter 20 by Francis Durso and Andrew
decomposition diagrams. Anders Ericsson Dattel), software design (Chapter 21 by
(Chapter 13 ) describes how the elicitation Sabine Sonnentag, Cornelia Niessen, and
of think-aloud protocols can allow inves- Judith Volmer), and writing (Chapter 22 by
tigators to trace the thought processes of Ronald Kellogg). There are two chapters on
experts while they perform representative various aspects of decision making, namely,
tasks from their domain. Finally, Paul Ward, judgments in dynamic situations (natu-
Mark Williams, and Peter Hancock (Chap- ral decision making, Chapter 23 by Karol
ter 14) review how simulated environments Ross, Jennifer Shafer, and Gary Klein) and
can both be used to measure experts rep- decision-making expertise (Chapter 24 by
resentative performance as well as be used Frank Yates & Michael Tschirhart), followed
for training. by Chapter 25 by Eduardo Salas, Michael
Section 4 contains chapters examining Rosen, Shawn Burke, Gerald Goodwin, and
methods for studying how skill, exper- Stephen Fiore on research on expert teams.
tise, and expert performance develop and The second subsection contains chapters
are acquired through practice. In the first that review expert performance in music
16 the cambridge handbook of expertise and expert performance
(Chapter 26 by Andreas Lehmann and for the key role of deliberate practice in caus-
Hans Gruber) and in sports (Chapter 27 ing physiological adaptations and the acqui-
by Nicola Hodges, Janet Starkes, and Clare sition of mechanisms that mediate expert
MacMahon), and expertise in other types performance. Finally, Barry Zimmerman
of arts, such as acting, ballet, and dance (Chapter 3 9) describes the importance of
(Chapter 28 by Helga Noice and Tony self-regulated learning in the development
Noice). The final chapter in this subsec- of expertise. The last three chapters review
tion reviews research on perceptual-motor general issues in expertise. Ralf Krampe
skills (Chapter 29 by David Rosenbaum, and Neil Charness (Chapter 40) review
Jason Augustyn, Rajal Cohen, and Steven the effects of aging on expert performance
Jax). The third and final subsection covers and how it might be counteracted. Harald
the findings in a diverse set of domains of Mieg (Chapter 41) reviews the importance
expertise, including games. The first chap- of social factors in the development of
ter (Chapter 3 0 by Fernand Gobet and expertise. Finally, Robert Weisberg (Chapter
Neil Charness) describes the pioneering and 42) discusses the relation between expertise
influential work on expertise in the game of and creativity.
chess. The next chapter (Chapter 3 1 by John
Wilding and Elizabeth Valentine) reviews
research on exceptional memory, in particu- Conclusion
lar for information that most people have
difficulty remembering, such as numbers, This Handbook has been designed to provide
names, and faces. The last two chapters researchers, students, teachers, coaches, and
review research on mathematical ability and anyone interested in attaining expertise with
expertise (Chapter 3 2 by Brian Butterworth) a comprehensive reference to methods, find-
and expertise in history (Chapter 3 3 by Jim ings, and theories related to expertise and
Voss and Jennifer Wiley) an example of a expert performance. It can be an essential
knowledge-based domain. tool for researchers, professionals, and stu-
In the last section of the Handbook we dents involved in the study or the training of
have invited some of the worlds lead- expert performance and a necessary source
ing researchers on general theoretical issues for college and university libraries, as well
that are cutting across different domains as public libraries. In addition, the Hand-
of expertise to review the current state of book is designed to provide a suitable text
knowledge. In the first chapter John Horn for graduate courses on expertise and expert
and Hiromi Masunaga (Chapter 3 4) dis- performance. More generally, it is likely that
cuss the relation between general intelli- professionals, graduate students, and even
gence and expertise. In the following chapter undergraduates who aspire to higher levels
Anna Cianciolo, Cynthia Mattew, Richard of performance in a given domain can learn
Wagner, and Robert Sternberg (Chapter 3 5 ) from experts pathways to superior perfor-
review the relation between expertise and mance in similar domains.
central concepts, such as practical intelli- Many researchers studying expertise and
gence and tacit knowledge. Mica Endsley expert performance are excited and person-
(Chapter 3 6) reviews evidence for situa- ally curious about the established research
tional awareness, namely, experts superior findings that most types of expertise require
ability to perceive and monitor critical at least a decade of extended efforts to attain
aspects of situations during performance. the mechanisms mediating superior perfor-
The next three chapters focus on aspects of mance. There is considerable knowledge
learning. Nicole Hill and Walter Schneider that is accumulating about generalizations
(Chapter 3 7) review the neurological evi- across many domains about the acquisition
dence on physiological adaptations result- and refinement of these mechanisms during
ing from the acquisition of expertise. Anders an extended period of deliberate practice.
Ericsson (Chapter 3 8) reviews the evidence The generalizable insights range from the
introduction 17
Chi, M. T. H., Glaser, R., & Farr, M. J. (Eds.) Ericsson, K. A., & Lehmann, A. C. (1996).
(1988). The nature of expertise. Hillsdale, NJ: Expert and exceptional performance: Evidence
Erlbaum. on maximal adaptations on task constraints.
Chi, M. T. H., Glaser, R., & Rees, E. (1982). Annual Review of Psychology, 47, 273 3 05 .
Expertise in problem solving. In R. S. Stern- Ericsson, K. A., Patel, V. L., & Kintsch, W.
berg (Ed.). Advances in the psychology of human (2000). How experts adaptations to represen-
intelligence (Vol. 1, pp. 175 ). Hillsdale, NJ: tative task demands account for the expertise
Erlbaum. effect in memory recall: Comment on Vicente
De Groot, A. (1978). Thought and choice in chess. and Wang (1998). Psychological Review, 107,
(Original by published in 1946). The Hague, 5 785 92.
the Netherlands: Mouton. Ericsson, K. A., & Smith, J. (Eds.) (1991a). Toward
Diderot, D., & DAlembert, J. L. R. (Eds.) (1966 a general theory of expertise: Prospects and limits.
67). Encyclopedie ou dictionnaie de raisonne des Cambridge: Cambridge University Press.
sciences, des artes et des metier [Encyclopedia or Ericsson, K. A., & Smith, J. (1991b). Prospects
dictionary of the sciences, arts, and occupations]. and limits in the empirical study of expertise:
(Originally published 175 180). Stuttgart-Bad An introduction. In K. A. Ericsson and J. Smith
Cannstatt, Gemany: Frommann. (Eds.), Toward a general theory of expertise:
Djakow, Petrowski, & Rudik (1927). Psychologie Prospects and limits (pp. 13 8). Cambridge:
des Schachspiels [Psychology of chess]. Berlin: Cambridge University Press.
Walter de Gruyter. Ernst, G. W., & Newell, A. (1969). GPS: A case
Doll, J., & Mayr, U. (1987). Intelligenz und study in generality and problem solving. New
Schachleistung eine Untersuchung an York: Academic Press.
Schachexperten. [Intelligence and achieve- Feltovich, P. J., Ford, K. M., & Hoffman, R. R.
ment in chess a study of chess masters]. (Eds.) (1997). Expertise in context: Human and
Psychologische Beitrage, 2 9, 270289. machine. Cambridge, MA: AAAI/MIT Press.
Dreyfus, H. L., & Dreyfus, S. E. (1986). Mind over Fitts, P., & Posner, M. I. (1967). Human perfor-
machine: The power of intuition and expertise in mance. Belmont, CA: Brooks/Cole.
the era of the computer. New York: The Free Fitzsimmons, M. P. (2003 ). The night that the Old
Press. Regime ended. University Park, PA: Pennsylva-
Epstein, S. A. (1991). Wage, labor, & guilds nia State University Press.
in medieval Europe. Chapel Hill, NC: North Galton, F., Sir (1869/1979). Hereditary genius:
Carolina University Press. an inquiry into its laws and consequences.
Ericsson, K. A. (Ed.) (1996a). The road to (Originally published in 1869). London: Julian
excellence: The acquisition of expert performance Friedman Publishers.
in the arts and sciences, sports, and games. Gobet, F., & Simon, H. A. (1996). Templates in
Mahwah, NJ: Erlbaum. chess memory: A mechanism for recalling sev-
Ericsson, K. A. (1996b). The acquisition of expert eral boards. Cognitive Psychology, 3 1, 140.
performance: An introduction to some of the Hoffman, R. R. (Ed.) (1992). The psychology of
issues. In K. A. Ericsson (Ed.), The road to expertise: Cognitive research and empirical AI.
excellence: The acquisition of expert performance New York: Springer-Verlag.
in the arts and sciences, sports, and games Howe, M. J. A., Davidson, J. W., & Sloboda, J. A.
(pp. 15 0). Mahwah, NJ: Erlbaum. (1998). Innate talents: Reality or myth? Behav-
Ericsson, K. A. (2004). Deliberate practice and ioral and Brain Sciences, 2 1, 3 99442.
the acquisition and maintenance of expert per- Hunt, E. B. (1995 ). Will we be smart enough? New
formance in medicine and related domains. York: Russell Sage Foundation.
Academic Medicine, 10, S70S81. Krause, E. A. (1996). Death of guilds: Professions,
Ericsson, K. A., & Kintsch, W. (1995 ). Long- states and the advance of capitalism, 193 0 to
term working memory. Psychological Review, the present. New Haven, CT: Yale University
102 , 211245 . Press.
Ericsson, K. A., Krampe, R. T., & Tesch-Romer, Lyons, R., Payne, C., McCabe, M., & Fielder,
C. (1993 ). The role of deliberate practice in the C. (1998). Legibility of doctors handwriting:
acquisition of expert performance. Psychologi- Quantitative and comparative study. British
cal Review, 100, 3 63 406. Medical Journal, 3 17, 863 864.
introduction 19
Newell, A., & Simon, H. A. (1972). Human prob- Sosniak, L. A. (1985 c). Phases of learning. In
lem solving. Englewood Cliffs, NJ: Prentice- B. S. Bloom (Ed.), Developing talent in young
Hall. people (pp. 40943 8). New York: Ballantine
Pannabecker, J. R. (1994). Diderot, the mechani- Books.
cal arts and the encyclopedie: In search of the Sosniak, L. A. (1985 d). A long-term commitment
heritage of technology education. Journal of to learning. In B. S. Bloom (Ed.), Developing
Technology Education, 6, 45 5 7. talent in young people (pp. 4775 06). New York:
Plato (1991). Protagoras (Translated by C. C. W. Ballantine Books.
Taylor). Oxford: Clarendon Press. Starkes, J. L., & Allard, F. (Eds.) (1993 ). Cogni-
Polanyi, M. (1962). Personal knowledge: Toward tive issues in motor expertise. Amsterdam: North
a post-critical philosophy (Corrected edition). Holland.
Chicago: The University of Chicago Press. Starkes, J., & Ericsson, K. A. (Eds.) (2003 ).
Polanyi, M. (1966). The tacit dimension (Reprinted Expert performance in sport: Recent advances
in 1983 ). Gloucester, MA: Peter Smith. in research on sport expertise. Champaign, IL:
Rees, G., & Wakely, M. (Eds.) (2004). The instau- Human Kinetics.
ratio magna, Part II: Novum organum and asso- Sternberg, R. J., & Grigorenko, E. L. (Eds.)
ciated texts. Oxford: Clarendon Press. (2003 ). Perspectives on the psychology of abil-
Roe, A. (195 2). The making of a scientist. New ities, competencies, and expertise. Cambridge:
York: Dodd, Mead & Company. Cambridge University Press.
Rohr, D. (2001). The careers and social status Taylor, I. A. (1975 ). A retrospective view of
of British musicians, 175 0185 0: A profession creativity investigation. In I. A. Taylor and
of artisans. Cambridge: Cambridge University J. W. Getzels (Eds.), Perspectives in creativity
Press. (pp. 13 6). Chicago, IL; Aldine Publishing Co.
Rosselli, J. (1991). Music & musicians in Van der Maas, H. L. J., & Wagenmakers, E. J.
nineteenth-century Italy. Portland, OR: (2005 ). A psychometric analysis of chess
Amadeus Press. expertise. American Journal of Psychology, 118,
2960.
Shanteau, J. (1988). Psychological characteristics
and strategies of expert decision makers. Acta Vicente, K. J., & Wang, J. H. (1998). An ecological
Psychologica, 68, 203 215 . theory of expertise effects in memory recall.
Psychological Review, 105 , 3 3 5 7.
Simon, H. A., & Gobet, F. (2000). Expertise
effects in memory recall: Comment on Vicente Websters New World Dictionary (1968).
and Wang (1998). Psychological Review, 107, Cleveland, OH: The World Publishing
5 93 600. Company.
Simon, H. A, & Chase, W. G. (1973 ). Skill in Wikipedia. http://en.wikipedia.org/wiki/Expert
chess. American Scientist, 61, 3 94403 . Zuckerman, H. (1977). Scientific elite: Nobel laure-
Simonton, D. K. (1994). Greatness: Who makes ates in the United States. New York: Free Press.
history and why. New York: Guilford Press.
Singer, C. (195 8). From magic to science. New
York: Dover. Author Notes
Sosniak, L. A. (1985 a). Learning to be a concert
pianist. In B. S. Bloom (Ed.), Developing talent This article was prepared in part with support
in young people (pp. 1967). New York: Ballan- from the FSCW/Conradi Endowment Fund of
tine Books. Florida State University Foundation. The author
Sosniak, L. A. (1985 b). Becoming an outstand- wants to thank Neil Charness, Paul Feltovich,
ing research neurologist. In B. S. Bloom (Ed.), Len Hill, Robert Hoffman, and Roy Roring for
Developing talent in young people (pp. 3 48408). their valuable comments on earlier drafts of this
New York: Ballantine Books. chapter.
CHAPTER 2
Michelene T. H. Chi
This chapter differentiates two approaches debated, such as pathologist Warren and gas-
to the study of expertise, which I call troenterologist Marshalls proposal that bac-
the absolute approach and the relative teria cause peptic ulcers (Chi & Hausmann,
approach, and what each approach implies 2003 ; Thagard, 1998; also see the chapters
for how expertise is assessed. It then summa- by Wilding & Valentine, Chapter 3 1, Simon-
rizes the characteristic ways in which experts ton, Chapter 18, and Weisberg, Chapter 42).
excel and the ways that they sometimes Several methods can be used to identify
seem to fall short of common expectations. someone who is truly an exceptional expert.
One method is retrospective. That is, by
looking at how well an outcome or prod-
Two Approaches to the Study uct is received, one can determine who is or
of Expertise is not an expert. For example, to identify a
great composer, one can examine a quanti-
The nature of expertise has been studied in tative index, such as how often his or her
two general ways. One way is to study truly music was broadcast (Kozbelt, 2004). A sec-
exceptional people with the goal of under- ond method may be some kind of concur-
standing how they perform in their domain rent measure, such as a rating system as a
of expertise. I use the term domain loosely result of tournaments, as in chess (Elo, 1965 ),
to refer to both informal domains, such as or as a result of examinations (Masunaga &
sewing and cooking, and formal domains, Horn, 2000), or just measures of how well
such as biology and chess. One could choose the exceptional expert performs his task. A
exceptional people on the basis of their third method might be the use of some inde-
well-established discoveries. For example, pendent index, if it is available. In chess,
one could study how Maxwell constructed for example, there exists a task called the
a quantitative field concept (Nersessian, Knights Tour that requires a player to move
1992). Or one could choose contemporary a Knight Piece across the rows of a chess
scientists whose breakthroughs may still be board, using legal Knight Moves. The time it
21
22 the cambridge handbook of expertise and expert performance
takes to complete the moves is an indication functioning probably because genetic inheri-
of ones chess skill (Chi, 1978). Although this tance does seem to be a relevant component
task is probably not sensitive enough to dis- for expertise in music and sports. In short,
criminate among the exceptional experts, a the tacit assumption is that greatness or cre-
task such as this can be adapted as an index of ativity arises from chance and unique innate
expertise. In short, to identify a truly excep- talent (Simonton, 1977). Lets call this type
tional expert, one often resorts to some kind of work in psychology the study of excep-
of measure of performance. The assessment tional or absolute expertise.
of exceptional experts needs to be accurate A second research approach to expertise
since the goal is to understand their supe- is to study experts in comparison to novices.
rior performance. Thus, this approach stud- This relative approach assumes that exper-
ies the remarkable few to understand how tise is a level of proficiency that novices can
they are distinguished from the masses. achieve. Because of this assumption, the
Though expertise can be studied in the definition of expertise for this contras-
context of exceptional individuals, there tive approach can be more relative, in the
is a tacit assumption in the literature that sense that the more knowledgeable group
perhaps these individuals somehow have can be considered the experts and the less
greater minds in the sense that the global knowledgeable group the novices. Thus
qualities of their thinking might be dif- the term novices is used here in a generic
ferent (Minsky & Papert, 1974, p. 5 9). For sense, in that it can refer to a range of non-
example, they might utilize more power- experts, from the naives to the journeymen
ful domain-general heuristics that novices (see Table 2.1 for definitions).
are not aware of, or they may be natu- Proficiency level can be grossly assessed
rally endowed with greater memory capacity by measures such as academic qualifications
(Pascual-Leone, 1970; Simonton, 1977). This (such as graduate students vs. undergrad-
line of reasoning is extended to cognitive uates), seniority or years performing the
two approaches to the study of experts characteristics 23
task, or consensus among peers. It can also research literature and discussed at some
be assessed at a more fine-grained level, in length (see edited volumes by Chi, Glaser, &
terms of domain-specific knowledge or per- Farr, 1988; Ericsson & Smith, 1991; Ericsson,
formance tests. 1996; Feltovich, Ford, & Hoffman, 1997;
One advantage of this second approach, Hoffman, 1992). Most of the research has
the study of relative expertise, is that we focused on how experts excel, either in
can be a little less precise about how to an absolute context or in comparison to
define expertise since experts are defined novices. However, it is equally important to
as relative to novices on a continuum. In understand how experts fail. Knowing both
this relative approach, a goal is to under- how they excel and how they fail will pro-
stand how we can enable a less skilled or vide a more complete characterization of
experienced persons to become more skilled expertise. This section addresses both sets of
since the assumption is that expertise can be characteristics.
attained by a majority of students. This goal
has the advantage of illuminating our under-
standing of learning since presumably the Ways in which Experts Excel
more skilled person became expert-like from I begin by very briefly highlighting seven
having acquired knowledge about a domain, major ways in which experts excel because
that is, from learning and studying (Chi & this set of findings have been reviewed
Bassok, 1989) and from deliberate practice extensively in the literature, followed by a
(Ericsson, Chapter 3 8; Ericsson, Krampe, & slightly more elaborate discussion of seven
Tesch-Romer, 1993 ; Weisberg, 1999). Thus, ways in which they fall short.
the goal of studying relative expertise is not
merely to describe and identify the ways in
which experts excel. Rather, the goal is to generating the best
understand how experts became that way so Experts excel in generating the best solu-
that others can learn to become more skilled tion, such as the best move in chess, even
and knowledgeable. under time constraints (de Groot, 1965 ), or
Because our definition characterizes the best solution in solving problems, or the
experts as being more knowledgeable than best design in a designing task. Moreover,
non-experts, such a definition entails sev- they can do this faster and more accurately
eral fundamental theoretical assumptions. than non-experts (Klein, 1993 ).
First, it assumes that experts are people
who have acquired more knowledge in a
domain (Ericsson & Smith, 1991, Table 2.1) detection and recognition
and that this knowledge is organized or Experts can detect and see features that
structured (Bedard & Chi, 1992). Second, novices cannot. For example, they can see
it assumes that the fundamental capacities patterns and cue configurations in X-ray
and domain-general reasoning abilities of films that novices cannot (Lesgold et al.,
experts and non-experts are more or less 1988). They can also perceive the deep
identical. Third, this framework assumes structure of a problem or situation (Chi,
that differences in the performance of Feltovich, & Glaser, 1981).
experts and non-experts are determined by
the differences in the way their knowledge
is represented. qualitative analyses
Experts spend a relatively great deal of time
analyzing a problem qualitatively, devel-
Manifestations of Experts Skills oping a problem representation by adding
and Shortcomings many domain-specific and general con-
straints to the problems in their domains
Numerous behavioral manifestations of of expertise (Simon & Simon, 1978; Voss,
expertise have been identified in the Greene, Post, & Penner, 1983 ).
24 the cambridge handbook of expertise and expert performance
monitoring opportunistic
Experts have more accurate self-monitoring Experts are more opportunistic than
skills in terms of their ability to detect novices; they make use of whatever sources
errors and the status of their own com- of information are available while solv-
prehension. In the domain of physics, ing problems (Gilhooly et al., 1997) and
experts were more accurate than novices in also exhibit more opportunism in using
judging the difficulty of a problem (Chi, resources.
Glaser, & Rees, 1982). In the domain of
chess, expert (Class B) chess players were cognitive effort
more accurate than novices in predicting
Experts can retrieve relevant domain knowl-
the number of pieces they thought they
edge and strategies with minimal cognitive
could recall immediately or the number of
effort (Alexander, 2003 , p. 3 ). They can also
times they thought they needed to view a
execute their skills with greater automatic-
chess position in order to recall the entire
ity (Schneider, 1985 ) and are able to exert
position correctly. Moreover, the experts
greater cognitive control over those aspects
were significantly more accurate in dis-
of performance where control is desirable
criminating their ability to recall the ran-
(Ericsson, Chapter 13 ).
domized (positions with the pieces scram-
bled) from the meaningful chess positions,
whereas novices thought they could recall Ways in which Experts Fall Short
equal number of pieces from the random- An equally important list might be ways in
ized as well as the meaningful positions which experts do not excel (Sternberg, 1996;
(Chi, 1978). Sternberg & Frensch, 1992). Because much
less has been written about experts handi-
strategies caps, I present a slightly more extensive dis-
cussion of seven ways in which experts do
Experts are more successful at choosing the
not surpass novices. This list also excludes
appropriate strategies to use than novices.
limitations that are apparent in experts, but
For example, in solving physics problems,
in fact novices would be subjected to the
the instructors tend to work forward,
same limitations if they have the knowledge.
starting from the given state to the goal
For example, experts often cannot articu-
state, whereas students of physics tend to
late their knowledge because much of their
work backwards, from the unknown to
knowledge is tacit and their overt intuitions
the givens (Larkin, McDermott, Simon, &
can be flawed. This creates a science of
Simon, 1980). Similarly, when confronted
knowledge elicitation to collaboratively cre-
with routine cases, expert clinicians diag-
ate a model of an experts knowledge (Ford &
nose with a data-driven (forward-working)
Adams-Webber, 1992). However, this short-
approach by applying a small set of rules
coming is not listed below since novices
to the data; whereas less expert clini-
would most likely have the same problem
cians tend to use a hypothesis-driven (back-
except that their limitation is less apparent
ward chaining) approach (Patel & Kaufman,
since they have less knowledge to explicate.
1995 ). Even though both more-expert and
the less-expert groups can use both kinds
of strategies, one group may use one kind domain-limited
more successfully than the other kind. Expertise is domain-limited. Experts do not
Experts not only will know which strat- excel in recall for domains in which they
egy or procedure is better for a situa- have no expertise. For example, the chess
tion, but they also are more likely than masters recall of randomized chess board
novices to use strategies that have more fre- positions is much less accurate than the
quently proved to be effective (Lemaire & recall for actual positions from chess games
Siegler, 1995 ). (Gobet & Simon, 1996), and the engineers
two approaches to the study of experts characteristics 25
attempt to recall the state of affairs of the surface features and overlook details. For
thermal-hydraulic processes that are not example, in recalling a text passage describ-
physically meaningful is much less success- ing a baseball game, individuals with high
ful than attempts to recall such states that baseball knowledge actually recalled fewer
are meaningfull (Vicente, 1992). There are baseball-irrelevant sentences than individ-
a number of demonstrations from various uals with low baseball knowledge (Voss,
other domains that show experts superior Vesonder, & Spilich, 1980), such as sentences
recall compared to novices for representa- containing information about the weather
tive situations but not for randomly rear- and the team. But high-knowledge indi-
ranged versions of the same stimuli (Ericsson viduals do recall information that is rele-
& Lehmann, 1996; Vicente & Wang, 1998). vant to the goal structure of the game, as
Thus, the superiority associated with their well as changes in the game states. Simi-
expertise is very much limited to a specific larly, in answering questions about computer
domain. programs, novices are better than experts
Of course there are exceptions. For exam- for concrete questions, whereas experts are
ple, expert chess players can display a reli- better than novices for abstract questions
able, but comparatively small, superiority of (Adelson, 1984).
memory performance for randomized chess In medical domains, after the presenta-
positions when they are briefly presented tion of an endocarditic case, 4th and 6th year
(see Gobet & Charness, Chapter 3 0), or medical students recalled more propositions
when the random positions are presented about the case than the internists (Schmidt
at slower rates (Ericsson, Patel, & Kinstch, & Boshuizen, 1993 ). Moreover, because the
2000). Nevertheless, in general, their exper- internists biomedical knowledge was bet-
tise is domain-limited. ter consolidated with their clinical knowl-
edge, resulting in short cuts, their expla-
overly confident nations thus made few references to basic
pathophysiological processes such as inflam-
Experts can also miscalibrate their capabil-
mation. In short, it is as if experts gloss over
ities by being overly confident. Chi (1978)
details that are the less relevant features of a
found that the experts (as compared to both
problem.
the novices and the intermediates) overesti-
mated the number of chess pieces they could
recall from coherent chess positions (see context-dependence within a domain
Figure 9, left panel, Chi, 1978). Similarly,
The first limitation of expertise stated above
physics and music experts overestimated
is that it is restricted to a specific domain.
their comprehension of a physics or music
Moreover, within their domain of expertise,
text, respectively, whereas novices were far
experts rely on contextual cues. For exam-
more accurate (Glenberg & Epstein, 1987).
ple, in a medical domain, experts seem to
It seems that experts can be overly confi-
rely on the tacit enabling conditions of a sit-
dent in judgments related to their field of
uation for diagnosis (Feltovich & Barrows,
expertise (Oskamp, 1965 ). Of course, there
1984). The enabling conditions are back-
are also domains, such as weather forecast-
ground information such as age, sex, previ-
ing, for which experts can be cautious and
ous diseases, occupation, drug use, and so
conservative (Hoffman, Trafron, & Roebber
forth. These circumstances are not necessar-
2005 ).
ily causally related to diseases, but physicians
pick up and use such correlational knowl-
glossing over edge from clinical practice. When expert
Although experts surpass novices in under- physicians were presented the complaints
standing and remembering the deep struc- associated with a case along with patient
ture of a problem, a situation, or a computer charts and pictures of the patients, they
program, sometimes experts fail to recall were 5 0% more accurate than the novices in
26 the cambridge handbook of expertise and expert performance
their diagnoses, and they were able to repro- text (To make a cake with 2 cups of flour
duce a large amount of context information you need 5 spoonfuls of water; how many
that was directly relevant to the patients spoonfuls do you need for 3 cups of flour?),
problem (Hobus, Schmidt, Boshuizen, & they did not adapt their mathematical strate-
Patel, 1987). The implication is that without gies. Instead, they used estimation strategies,
the contextual enabling information, expert resulting in only 20% accuracies.
physicians might be more limited in their
ability to make an accurate diagnosis. inaccurate prediction, judgment, and advice
Experts skills have been shown to be
Another weakness of experts is that some-
context-dependent in many other studies,
times they are inaccurate in their prediction
such as the failure of experienced waiters
of novice performance. For example, one
to indicate the correct surface orientation
would expect experts to be able to extrapo-
of liquid in a tilted container, despite their
late from their own task-specific knowledge
experience in the context of wine glasses
how quickly or easily novices can accomplish
(Hecht & Proffitt, 1995 ), and the inaccura-
a task. In general, the greater the expertise
cies of wildland fire fighters in predicting the
the worse off they were at predicting how
spread of bush fire when the wind and slope
quickly novices can perform a task, such as
are opposing rather than congruent, which is
using a cell phone (Hinds, 1999). In tasks
an unusual situation (Lewandowsky, Dunn,
requiring decision under uncertainty, such
Kirsner, & Randell, 1997).
as evaluating applicants for medical intern-
ships (Johnson, 1988) or predicting successes
inflexible in graduate school (Dawes, 1971), it has been
shown consistently that experts fail to make
Although Hatano and Inagaki (1986) have
better judgments than novices. Such lack
claimed that exceptional (versus routine)
of superior decision making may be limited
experts are adaptive, sometimes experts do
to domains that involve predicting human
have trouble adapting to changes in prob-
behavior, such as parole decisions, psychi-
lems that have a deep structure that devi-
atric judgment, and graduate school suc-
ates from those that are acceptable in the
cesses (Shanteau, 1984).
domain. For example, Sternberg and Frensch
An alternative interpretation of experts
(1992) found that expert bridge players
inaccuracies in making predictions is to pos-
suffered more than novice players when
tulate that they cannot take the perspectives
the games bidding procedure was changed.
of the novices accurately. Compatible with
Similarly, expert tax accountants had more
this interpretation is the finding that stu-
difficulty than novice tax students in trans-
dents are far more able to incorporate feed-
ferring knowledge from a tax case that dis-
back from their peers than from their expert
qualified a general tax principle (Marchant,
instructor in a writing task (Cho, 2004).
Robinson, Anderson, & Schadewald, 1991).
Perhaps the experts in these studies are rou-
tine experts; but they nevertheless showed bias and functional fixedness
less flexibility than the novices. Bias is probably one of the most serious
Inflexibility can be seen also in the use handicaps of experts, especially in the med-
of strategies by Brazilian street vendors who ical profession. Sometimes physicians are
can be considered experts in street math- biased by the probable survival or mor-
ematics (Schliemann & Carraher, 1993 ). tality rates of a treatment. Christensen,
When presented with a problem in a pric- Heckerling, Mackesy, Berstein, and Elstein
ing context, such as If 2 kg of rice cost 5 (1991) found that residents were more sus-
cruzeiros, how much do you have to pay ceptible to let the probable survival outcome
for 3 kg?, they used mathematical strate- determine options for treatment, whereas
gies with 90% accuracies. However, when novice students were not. Fortunately, expe-
presented with a problem in a recipe con- rienced physicians were not affected by
two approaches to the study of experts characteristics 27
the mortality rates either. In another study, excel, it is equally important to know ways in
however, my colleagues and I found the which their knowledge is limiting. The facil-
experienced physicians to manifest seri- itations and limitations of knowledge can
ous biases. We presented several types of provide boundary conditions for shaping a
cases to specialists, such as hematologists, theory of expertise.
cardiologists, and infectious disease spe-
cialists. Some were hematology cases and
others were cardiology cases. We found References
that regardless of the type of specialized
case, specialists tended to generate hypothe- Adelson, B. (1984). When novices surpass
ses that corresponded to their field of experts: The difficulty of a task may increase
expertise: Cardiologists tended to generate with expertise. Journal of Experimental Psy-
more cardiology-type hypotheses, whether chology: Learning, Memory, and Cognition, 10,
483 495 .
the case was one of a blood disease or an
infectious disease (Hashem, Chi, & Fried- Alexander, P. A. (2003 ). Can we get there from
here? Educational Researcher, 3 2, 3 4.
man, 2003 ). This tendency to generate diag-
noses about which they have more knowl- Bedard, J., & Chi, M. T. H. (1992). Expertise.
Current Directions in Psychological Science, 1,
edge clearly can cause greater errors. More-
13 5 13 9.
over, experts seem to be more susceptible
Chi, M. T. H. (1978). Knowledge structure
to suggestions that can bias their choices
and memory development. In R. Siegler
than novices (Walther, Fiedler, & Nickel,
(Ed.), Childrens thinking: What develops?
2003 ). (pp. 73 96). Hillsdale, NJ: Erlbaum.
Greater domain knowledge can also be
Chi, M. T. H., & Bassok, M. (1989). Learning from
deleterious by creating mental set or func- examples via self-explanations. In L. B. Resnick
tional fixedness. In a problem-solving con- (Ed.), Knowing, learning, and instruction: Essays
text, there is some suggestion that the more in honor of Robert Glaser (pp. 25 1282).
knowledgeable participants exhibit more fu- Hillsdale, NJ: Erlbaum.
nctional fixedness in that they have more Chi, M. T. H., Feltovich, P., & Glaser, R. (1981).
difficulty coming up with creative solutions. Categorization and representation of physics
For example, in a remote association task, problems by experts and novices. Cognitive
three words are presented, such as plate, bro- Science, 5 , 12115 2.
ken, and rest, and the subjects task is to come Chi, M. T. H., Glaser, R., & Farr, M. J. (Eds.)
up with a fourth word that can form a famil- (1988). The nature of expertise. Hillsdale, NJ:
iar phrase with each of the three words, such Erlbaum.
as the word home for home plate (a baseball Chi, M. T. H., Glaser, R., & Rees, E. (1982). Exper-
term), broken home, and rest home. A mis- tise in problem solving. In R. Sternberg (Ed.),
leading set of three words can be plate, bro- Advances in the Psychology of Human Intelligence
ken, and shot, in which the correct solution is ( Vol. 1, pp. 776). Hillsdale, NJ: Erlbaum.
glass. High baseball knowledge subjects were Chi, M. T. H., & Hausmann, R. G. M. (2003 ). Do
less able than low baseball knowledge sub- radical discoveries require ontological shifts?
In L. V. Shavinina (Ed.), International hand-
jects to generate correct solutions to the mis-
book on innovation (pp. 43 0444). New York:
leading type of problems because the first
Elsevier Science Ltd.
word plate primed their baseball knowledge
Cho, K. (2004). When experts give worse advice
so that it caused functional fixedness (Wiley,
than novices: The type and impact of feedback
1998). given by students and an instructor on student
In conclusion, the two sections above writing. Unpublished dissertation, University
each summarized seven ways in which of Pittsburgh.
experts excel and seven ways in which they Christensen, C., Heckerling, P. S., Mackesy, M. E.,
fall short. Although much more research has Berstein, L. M., & Elstein, A. S. (1991). Fram-
been carried out focusing on ways in which ing bias among expert and novice physicians.
experts greater knowledge allows them to Academic Medicine, 66 (suppl): S76S78.
28 the cambridge handbook of expertise and expert performance
Kozbelt, A. (2004). Creativity over the lifes- Schliemann, A. D., & Carraher, D. W. (1993 ).
pan in classical composers: Reexamining the Proportional reasoning in and out of school.
equal-odds rule. Paper presented at the 26th In P. Light & G. Butterworth (Eds.), Context
Annual Meeting of the Cognitive Science and cognition: Ways of learning and knowing
Society. (pp. 4773 ). Hillsdale, NJ: Erlbaum.
Larkin, J. H., McDermott, J., Simon, D. P., & Schmidt, H. G., & Boshuizen, P. A. (1993 ). On
Simon, H. A. (1980). Models of competence acquiring expertise in medicine. Educational
in solving physics problems. Cognitive Science, Psychology Review, 5 , 205 220.
4, 3 173 45 . Schneider, W. (1985 ). Training high performance
Lemaire, P., & Siegler, R. S. (1995 ). Four aspects skills: Fallacies and guidelines. Human Factors,
of strategic change: Contributions to childrens 27(3 ), 285 3 00.
learning of multiplication. Journal of Experi- Shanteau, J. (1984). Some unasked questions
mental Psychology: General, 124, 83 97. about the psychology of expert decision mak-
Lesgold, A., Rubinson, H., Feltovich, P., Glaser, ers. In M. El Hawaray (Ed.), Proceedings of the
R., Klopfer, D., & Wang, Y. (1988). Expertise 1984 IEEE conference on systems, man, and cyber-
in a complex skill: Diagnosing X-ray pictures. netics. New York: Institute of Electrical and
In M. T. H. Chi, R. Glaser, M. J. Farr (Eds.), Electronics Engineers, Inc.
The nature of expertise (pp. 3 113 42). Hillsdale, Simon, D. P., & Simon, H. A. (1978). Indi-
NJ: Erlbaum. vidual differences in solving physics prob-
Lewandowsky, S., Dunn, J. C., Kirsner, K., & lems. In R. Siegler (Ed.), Childrens thinking:
Randell, M. (1997). Expertise in the manage- What develops? (pp. 3 25 3 48). Hillsdale, NJ:
ment of bushfires: Training and decision sup- Erlbaum.
port. Australian Psychologist, 3 2(3 ), 171177. Simonton, D. K. (1977). Creative productiv-
Marchant, G., Robinson, J., Anderson, U., & ity, age, and stress: A biographical time-
Schadewald, M. (1991). Analogical transfer and series analysis of 10 classical composers.
expertise in legal reasoning. Organizational Journal of Personality and Social Psychology,
Behavior and Human Decision Making, 48, 3 5 , 791804.
272290. Sternberg, R. J. (1996). Costs of expertise. In
Masunaga, H., & Horn, J. (2000), Expertise and K. A. Ericsson (Ed.), The road to excellence:
age-related changes in components of intelli- The acquisition of expert performance in the arts
gence. Psychology & Aging, Special Issue: Vol. and sciences, sports, and games (pp. 3 473 5 4).
16(2), 293 3 11. Hillsdale, NJ: Erlbaum.
Minsky, M., & Papert, S. (1974). Artificial intelli- Sternberg, R. J., & Frensch, P. A. (1992). On
gence. Condon Lectures, Oregon /State System being an expert; A cost-benefit analysis. In R. R.
of Higher Education, Eugene, Oregon. Hoffman (Ed.), The psychology of expertise: Cog-
Nersessian, N. J. (1992). How do scientists think? nitive research and empirical AI (pp. 191203 ).
Capturing the dynamics of conceptual change New York: Springer Verlag.
in science. In R. N. Giere (Ed.), Cognitive mod- Thagard, P. (1998). Ulcers and bacteria I: Discov-
els of science: Minnesota studies in the philoso- ery and acceptance. Studies in History and Phi-
phy of science, Vol. XV (pp. 3 44). Minneapo- losophy of Science, 29, 10713 6.
lis, MN: University of Minnesota Press. Vicente, K. J. (1992). Memory recall in a process
Oskamp, S. (1965 ). Overconfidence in case-study control system: A measure of expertise and dis-
judgments. Journal of Consulting Psychology, 29, play effectiveness. Memory and Cognition, 20,
261265 . 3 5 63 73 .
Pascual-Leone, J. (1970). A mathematical model Vicente, K. J., & Wang, J. H. (1998). An
for the decision rule in Piagets developmental ecological theory of expertise effects in
stages. Acta Psychologica, 3 2, 3 013 45 . memory recall. Psychological Review, 105 ,
Patel, V. L., & Kaufman, D. R. (1995 ). Clinical 3 3 5 7.
reasoning and biomedical knowledge: impli- Voss, J. F., Greene, T. R., Post, T., & Penner, B. C.
cations for teaching. In J. Higgs & M. Jones (1983 ). Problem solving skill in the social sci-
(Eds.), Clinical reasoning in the health profes- ences. In G. Bower (Ed.), The psychology of
sions (pp. 117128). Oxford, UK: Butterworth- learning and motivation. (pp. 165 213 ). New
Heinemann Ltd. York: Academic Press.
30 the cambridge handbook of expertise and expert performance
Voss, J. F., Vesonder, G., & Spilich, H. (1980). Wiley, J. (1998). Expertise as mental set: The
Text generation and recall by high-knowledge effects of domain knowledge in creative
and low-knowledge individuals. Journal of Ver- problem solving. Memory and Cognition, 26,
bal Learning and Verbal Behavior, 19, 65 1667. 71673 0.
Walther, E., Fiedler, K., & Nickel, S. (2003 ).
The influence of prior knowledge on construc-
tive biases. Swiss Journal of Psychology, 62(4), Author Notes
21923 1.
Weisberg, R. W. (1999). Creativity and know- The author is grateful for support provided
ledge: A challenge to theories. In R. J. by the Pittsburgh Science of Learning Center.
Sternberg (Ed.), Handbook of Creativity Reprints may be requested from the author
(pp. 22625 0). Cambridge, UK: Cambridge or downloaded from the WEB, at www.pitt.
University Press. edu/ chi
CHAPTER 3
Earl Hunt
the role of intelligence. Operationally, intel- gence scale, and markedly lower at the high
ligence is usually defined by scores on tests end (Detterman & Daniel, 1989; Deary et al.,
of cognitive abilities. Based on the distri- 1996). This is important, as expertise is gen-
butions of test scores, modern psychome- erally associated with high levels of perfor-
tricians have largely agreed on a hierarchi- mance.
cal model of intelligence, originally due to Measures of Gf have substantial corre-
Cattell (1971), in which general intelligence lations with measures of the performance
g is inferred from positive correlations of working memory. A high-Gf person is
between sets of broadly applicable but dis- probably good at keeping track of several
tinct cognitive abilities. These include a gen- things at once and of concentrating his or her
eralized reasoning ability (fluid intelligence- attention in the face of distractions (Engle,
Gf ), the possession and use of knowledge Kane & Tulhoski, 1999; Kyllonen & Christal,
to solve problems (crystallized intelligence- 1990). These talents are good to have dur-
Gc), spatial-visual reasoning, a general abil- ing the learning phase of most psychomo-
ity to think quickly, and several other broad tor activities (e.g., skiing, riding a bicycle,
factors (Carroll, 1993 ). playing tennis). However, they are much less
The distinction between g, Gf, and Gc needed once an activity has been learned.
often drops out in discussions of the rela- Laboratory studies of how people learn to do
tion between intelligence and social out- psychomotor tasks have shown that intelli-
comes. This is unfortunate, for Gf and Gc gence is a reasonably good predictor of per-
are measured by different instruments. The formance early in learning but does not pre-
Wechsler Adult Intelligence Scale (WAIS) dict asymptotic levels of learning very well
confounds Gf and Gc (Horn, 1985 ). Two (Ackerman 1996; Fleishman, 1972).
group tests that are widely used in industrial An important study by Ackerman and
and academic settings, the Armed Services Cianciolo (2000) modifies this conclusion.
Vocational Aptitude Battery (ASVAB) and Ackerman and Cianciolo reasoned that if
the Scholastic Assessment Test (SAT) are a task taxes working memory after it has
essentially tests of Gc, based on the general been learned, the correlation with tests of
knowledge and problem-solving skills that reasoning should remain. They then trained
one expects an American high school grad- people on two different, greatly reduced
uate to have (Roberts et al., 2000). The best versions of an air traffic controllers task.
tests of Gf, by contrast, are tests in which an One could be solved by memorizing a not-
examinee must detect patterns in abstract too-complicated set of rules. To solve the
and unusual material (Jensen, 1998). more complicated task the participant had
The definition of Gc ensures that any Gc to develop orderly patterns of traffic in the
test is culture specific. Cattell (1971) antic- area near a terminal. Participants practiced
ipated this when he noted that within a the tasks for several days. The correlation
person Gc consists of two components, a between the first task and a measure of
general ability to use knowledge and the fluid intelligence decreased over practice
possession of specific knowledge. He even from .45 to .3 0. The correlations between
suggested that the proper evaluation of Gc the intelligence measure and performance
would require separate tests for every pro- increased from .40 to .5 5 over the training
fession. The same spirit can be found in the period.
research of Sternberg et al. (2000) on prac- There are obvious parallels between this
tical intelligence, which is evaluated by tests study and the general study of expertise.
of culture- or subgroup-specific knowledge. Some aspects of expertise, such as swing-
Gf and Gc are correlated, which makes it ing a golf club, require learning a con-
possible to speak reasonably about g. How- stant relationship between stimulus and
ever, the correlations between measures of response. Others aspects, such as the ana-
different types of cognitive abilities are high- logical reasoning typical of the law, involve
est toward the low end of the general intelli- varied mappings, the development of
expertise, talent, and social encouragement 33
mental models of a situation, and extensive discussion of the general issue of intelligence
knowledge. Demands on both Gf and Gc and workplace performance.)
never cease. Similar observations have been made in
A second important observation is based the civilian sector. Scores on tests of cog-
on studies of natural decision making. By nitive competence are related to workplace
definition, experts make better decisions performance, and the correlations are some-
than novices. However, this does not mean what higher during training than during per-
that experts become better decision makers formance after training (Schmidt & Hunter,
in the sense that they learn to avoid the mis- 1998).2
takes that have been documented in labora- The conclusions just offered were drawn
tory studies of decision making (Kahneman, from analyses of jobs that might be charac-
2003 ). Instead, experienced real-life deci- terized as blue collar or lower level white
sion makers rely on analogical reasoning and collar. Although the data base is more lim-
schematic techniques for selecting and mon- ited, the same thing seems to be true of
itoring a plan of action (Klein, 1998). This upper-level professional jobs. One large, par-
kind of decision making depends on two ticularly well-designed study of managers
things: having the experiences on which the found a correlation of .3 8 between cogni-
analogies can be based and encoding those tive scores obtained at the outset of employ-
experiences in a way that makes information ment and level of management reached after
accessible when needed. Gc again! more than fifteen years on the job (Howard
& Bray, 1988).
Evidently intelligence-as-reasoning and
Findings from Industrial- working memory are always important dur-
Organizational Psychology ing the early stages of learning, well before
the expert level is reached. A task analy-
Although laboratory studies offer the advan- sis is necessary to determine the extent to
tage of control, they cannot replicate the which performance depends on reasoning
very long periods of time over which exper- and working memory after the expert level
tise is acquired in the workplace. The appro- has been reached.
priate studies are the domain of indus- Specialized knowledge will always be
trial-organizational, rather than cognitive, important if expertise depends largely on
psychology. the execution of psychomotor sequences,
In the late 1980s the US military evalu- as in ball-striking in golf. The sports exam-
ated various predictors of the performance ple is obvious. Psychomotor sequences are
of enlisted men and women (Wigdor & important in other areas, including medicine
Green, 1991) in military occupations rang- and piloting high-performance aircraft. In
ing from artillerymen to cooks. Performance other cases (e.g., the law, physics), exper-
increased with experience, but appeared to tise requires the development of schema that
asymptote after about three years. Asymp- can guide problem solving. To some extent
totic level of performance was related to the use of such schema can reduce the bur-
scores on a test of mental skills, the Armed den on working memory, thus shifting the
Forces Qualifying Test (AFQT), taken at balance between the Gf-and Gc-aspects of
time of enlistment, but there was an intelligence.
interaction.1 Enlisted personnel with high Different types of expertise can be char-
scores reached asymptotic performance in a acterized by their location on the psychomo-
year, personnel with lower test scores took tor/mental-modeling-and problem-solving/
longer. Differences in performance could be use-of-experience dimensions. Almost every
related to the AFQT after more than three task in which expertise can be illustrated
years of service, but the differences were less contains some elements of each dimension.
than half those for personnel with only a Will we ever be able to test people at
years service. (See Hunt [1995 ] for a further the outset of their experience, say early in
34 the cambridge handbook of expertise and expert performance
high school, and predict who would become many others were attending some of the
experts solely on the basis of their talents? most prestigious graduate schools in the
Probably not, for we have not yet consid- country. Some had made substantial con-
ered the social-personality aspect of expert tributions outside of academia. The type
development. of achievement differed by group. Students
whose mathematics scores were higher than
their verbal scores at age 13 gravitated to
Why Become an Expert? mathematics and science courses in college,
students whose verbal scores were high-
Sternberg (1996) has observed that intelli- est gravitated toward the humanities and
gence is successful to the extent that it has social sciences, and students with a flat pro-
been used to meet ones goals. It does not file (very high scores everywhere) showed a
make sense to do the work that it takes to be more even distribution of interests. Prefer-
an expert unless you want to be one. In order ences appeared relatively early. Reports of
to understand expertise we have to under- favorite class in high school mirrored later
stand interests. professional specialization.
Ackerman and his colleagues (Ackerman Talents are channeled by interests. In gen-
& Beier, 2001; Ackerman & Rolfhus, 1999; eral, people are more interested in things
Rolfhus & Ackerman, 1999) have shown that they are good at than things they find diffi-
within American society interests fall into cult. The combination of talent and interest
three definable clusters:science and math- leads to specialized knowledge, and knowl-
ematics, intellectual and cultural activity, edge produces expertise. Society reacts to
and social activities. People have knowl- the combination of talent and interest by
edge bases that correspond to their interests. offering support, which leads to further
They also show markedly different personal- specialization.
ity profiles. Most important for our concerns
here, the amount of knowledge a person has
within his or her own interest area is best Social Encouragement and Expertise
predicted by measures of Gc, or the extent
to which a person has picked up knowledge Because the acquisition of expertise re-
of the society in general. quires substantial effort, the social sup-
Because intelligence is differentiated at port provided during the learning phase
the upper end, one would expect differen- is extremely important. Chess experts
tial patterns of ability to be particularly pre- begin early, often by participation in chess
dictive of career choices of the gifted. They clubs (Charness, Krampe, & Mayr, 1996).
are. Lubinski, Benbow, and their colleagues Lubinski et al.s gifted students made sub-
have conducted longitudinal studies of gifted stantial use of advanced placement courses
students who, at age 13 , were in the top in high school and other educational accel-
ten-thousandth of examinees on tests of ver- eration programs. If we look at individual
bal and mathematical skill (Lubinski, Webb, cases, the amount of social support can be
Morelock, & Benbow, 2001). They differ- dramatic. Gardners (1993 ) biographic study
entiated between students who had signif- of exceptional contributors to society, such
icantly higher verbal scores than mathemat- as Einstein and Picasso, stresses how these
ics scores, or the reverse, and students who great contributors were able to be single-
were high flat, that is, verbal and mathe- minded because they were supported by
matical scores were essentially the same. It family, friends, and colleagues, often at con-
is important to remember that in this group siderable expense. At a less earthshaking
a low score corresponds to above average level of expertise, the 2004 winner of the
performance in the general population. Wimbledon womans tennis tournament,
Overall the gifted students did very well. Maria Sharapova, received a scholarship to
Several had doctorates at age 23 or less; a tennis academy at age eight!
expertise, talent, and social encouragement 35
3.5
2
Economists,
mathematicians,
1.5
engineers
1 HSTeachers
0.5
0
10% 25% 50% 75% 90%
Percentile of income distribution
Figure 3.1. Differential reward indices as a function of type of occupation and
the percentile of the income distribution. Data derived from US Census 2000
income reports.
The differential reward index varied If we need experts in some field we must
markedly across occupations. Financial and encourage people to acquire appropriate
business advisors in the 90th percentile of expertise and reward them when they have
their profession earned 3 .5 times the median done so.
income for their profession, whereas those at
the 10th percentile earned half the median
income. A similar but not-so-drastic acceler- Closing Remarks
ation was shown for the physician-dentist-
lawyer group. The differential reward func- In order to understand the development of
tion for mathematicians and engineers was expertise we have to distinguish between
almost identical to that for high school expertise in perceptual-motor tasks and ex-
teachers. In all groups acceleration occurred pertise in cognitive activities. Perceptual-
at the top. The differential reward functions motor expertise requires automation in
were virtually linear from the 10th to the the literal sense. Cognitive expertise re-
5 0th percentile. quires experience, and probably depends to
These data suggest, but certainly do not some extent on automated nonconscious
prove, that our society encourages the devel- thought. It also depends very much on the
opment of expertise in business, law, and acquisition of knowledge.
the biomedical professions. The figures do Working memory and attention are gener-
not suggest very much encouragement for ally considered to be the intellectual bottle-
the development of expertise in mathemat- necks on human thought. These are the pro-
ics and engineering. It is of interest to note cesses most taxed in the early stages of either
that as of 2004 educators and policy makers perceptual-motor learning or knowledge
were deploring the dearth of American stu- acquisition. Therefore it is harder to become
dents in engineering and mathematics, the an expert than to be one!Nevertheless, in
biomedical fields were prosperous, and busi- some areas of expert performance work-
ness schools were booming. ing memory demands, and hence demands
Bleske-Recheck, Lubinski, and Benbow for high fluid intelligence, appear to extend
(2004) make a related point. They observed beyond the learning period.
that extremely gifted mathematics students This conclusion does not deny the impor-
reported liking Advanced Placement classes tance of practice. Becoming an expert in
because it gave them an opportunity to almost anything requires literally years of
study with, and proceed at the pace of, their work. People will do this only if they have
academic peers. Bleske-Recheck et al. then some initial success, enjoy the work, and are
asked whether a well-documented trend supported by the social climate. Expertise is
toward opening up Advanced Placement not solely a cognitive affair.
classes to a greater range of students, in order
to encourage participation by students from
Footnotes
a wider spectrum of society, might actu-
ally make these classes less attractive to the
1. The AFQT is a subset of the ASVAB, and
very gifted, and therefore channel talented
therefore a test of Gc.
individuals away from the areas where they
2. Schmidt and Hunter refer to tests of general
might maximize their contributions. This is
cognitive competence. However, the tests that
not the place to debate the overall social
they list appear to be tests mainly of Gc.
merits of opening up opportunities to non-
3 . My claim is not that expertise is the sole deter-
traditional students versus offering special
miner of income. That would be silly. I do
nurturance to the gifted. (Bleske-Recheck claim, however, that expertise is one of the
et al. acknowledge that such benefits exist.) determinants of income. Therefore the differ-
What is relevant here is that experiences ential distribution of income within an occupa-
relatively early in adolescence do motivate tion partly reflects payment for expertise and
students to make particular career choices. partly reflects other features, such as seniority.
expertise, talent, and social encouragement 37
acquisition of expert performance in the arts Sloboda, J. A. (1996). The acquisition of musi-
and sciences, sports, and games (pp 167188). cal performance expertise: Deconstructing the
Mahwah, NJ: Erlbaum. Talent account of individual differences in
Roberts, R. D., Goff, G. N., Anjoul, F., Kyllonen, musical expressivity. In K. A. Ericsson (Ed.),
P. C., Pallier, G., & Stankov, L. (2000). The The road to excellence: The acquisition of expert
Armed Services Vocational Aptitude Battery performance in the arts and sciences, sports, and
(ASVAB): Little more than acculturated learn- games (pp 107126). Mahwah, NJ: Erlbaum.
ing (Gc)!? Learning and Individual Differences, Sternberg, R. J. (1996). Successful intelligence: How
12 , 81103 . practical and creative intelligence determine suc-
Rolfhus, E. L., & Ackerman, P. L. (1999). cess in life. New York: Simon & Schuster.
Assessing individual differences in knowledge: Sternberg, R. J., Forsythe, G. B., Hedlund, J.,
Knowledge, intelligence, and related traits. Horvath, J. A., Wagner, R. K., Williams, W. M.,
Journal of Educational Psychology, 91, 5 115 26. Snook, S. A., & Grigorenko, E. L. (2000). Prac-
Schmidt, F. L., & Hunter, J. E. (1998). The valid- tical intelligence in everyday life. New York:
ity and utility of selection methods in personnel Cambridge University Press.
psychology: Practical and theoretical implica- Sternberg, R. J., & Grigorenko, E. L. (Eds.)
tions of 85 years of research findings. Psycho- (2001). The psychology of abilities, competencies,
logical Bulletin, 12 4, 262274. and expertise. Cambridge: Cambridge Univer-
Schneider, W., & Shiffrin, R. M. (1977). Con- sity Press.
trolled and automatic processing: I. Detection, Wigdor, A. K., & Green, B. F., Jr. (1991).
search, and attention. Psychological Review, 84, Performance assessment in the workplace.
166. Washington, DC: National Academy Press.
Part II
OVERVIEW OF APPROACHES
TO THE STUDY OF
EXPERTISE BRIEF
HISTORICAL ACCOUNTS OF
THEORIES AND METHODS
CHAPTER 4
with theoretical models of human thought LT was able to create some novel proofs. The
processes that could reproduce the observ- heuristics from LT were later generalized
able performance. Even though there was into a model that could solve problems in
important earlier work on expertise, this many different domains, the General Prob-
was the period when a number of forces lem Solver (Ernst & Newell, 1969). There
came together to provide enough traction were also other computer models that were
for the field to take off. There were three built, not as simulations of human prob-
main sources to this impetus: artificial intel- lem solving, but based on effective computa-
ligence, psychology, and education. We will tion designed to represent artificial methods
survey these briefly. for producing intelligent action. For exam-
Early computer models developed by ple, Samuels (195 9) checker-playing pro-
Herbert Simon and Allen Newell demon- gram was able to challenge and beat excel-
strated that it is relatively easy for com- lent human checker players. These early,
putational devices to do some things wor- along with subsequent, successes spawned
thy of being considered intelligent. This some themes regarding expertise pertinent
breakthrough at Carnegie-Mellon was based to the present chapter.
on the confluence of two key realizations First, the idea that computation could
that emerged from the intellectual milieu support intelligent behavior reinforced the
that was developing between Carnegie and growing idea that computers and their pro-
Rand at the time (Prietula & Augier, 2005 ). grams could stand as formal models of
First, they (Al Newell, Cliff Shaw, and Herb human cognition. This grew into a perva-
Simon) envisioned that computers could sive stance toward human and machine cog-
be used to process symbols and symbol nition, the information processing model
structures. To explore this, they necessarily that is still widely held. Cognitive psychol-
developed what was to become the first list- ogy and computer science merged into a
processing computer language, IPL, which very close collaboration (along with linguis-
afforded them the ability to create arbitrarily tics and a few other fields) that was later
complex list structures and manipulate them named Cognitive Science. These computa-
recursively. Second, they incorporated the tional models and theories provided at least
concept of levels of abstraction in articu- alternatives to the behaviorist (stimulus-
lating their theories and, consequently, their response, no internal mental mechanisms)
programs. These allowed them to address approaches that had dominated psychol-
two critical technical problems: the spec- ogy for the prior half a century (more
ification problem, in which the compo- on this in our treatment of psychology
nents and processes of the target system are and expertise below). Newell and Simon,
sufficiently specified to capture the char- two pioneers of the information-processing
acteristics of interest, and the realization viewpoint, asserted this forcefully:
problem, in which the specification can be
implemented in an actual physical system As far as the great debates about the empty
to enable synthesis (Newell & Simon, 195 6). organism, behaviorism, intervening vari-
The seeds of viewing humans and machines ables, and hypothetical constructs are con-
as complex information-processing systems cerned, we take these simply as a phase in
the development of psychology. Our theory
had been sown.
posits internal mechanisms of great extent
During these early years, the first artifi- and complexity, and endeavors to make
cial intelligence program, called the Logic contact between them and the visible evi-
Theorist (Newell & Simon, 195 6), was writ- dences of problem solving. That is all there
ten. The Logic Theorist (LT) was coded in is to it. (Newell & Simon, 1972 , pp. 910)
IPL. Significantly, it was able to prove the-
orems in the predicate calculus in a manner As we will address in our treatment of psy-
that mimics human adults (Newell & Simon, chological influences, it is quite difficult to
1972). Of particular relevance to expertise, imagine what a field of studying expertise
studies of expertise from psychological perspectives 43
could have looked like if behaviorism had timony to the efficacy of strong methods.
continued to hold sway. As related to this chapter, this is impor-
The second theme has to do with alter- tant because a similar progression unfolded
native basic approaches to achieving intel- in other kinds of investigations of expertise,
ligence in a computational device, what including those in psychology (see later sec-
have been termed weak and strong meth- tions in this chapter on Expertise Is Limited
ods (Newell, 1973 ). The earliest success- in Its Scope and Elite Performance Does Not
ful AI programs utilized weak reasoning and Transfer; and Knowledge and Content Mat-
problem-solving methods that were draw- ter Are Important to Expertise).
ing on descriptions of human thought pro- Behaviorism was the school of psychol-
cesses. Indeed, at one point Newell termed ogy that eschewed resorting to unobserv-
artificial intelligence the science of weak able mental constructs, structural or process,
methods, at least as one characterization of any kind. Only the observable environ-
of AI (Newell, 1973 , page 9). Weak meth- ment (the stimulus) and an organisms overt
ods are highly portable, generalizable meth- reaction (the response) were considered
ods that do not depend on the particular the legitimate purview of a psychological
content of the domain of problem solving science. Behaviorism had dominated psy-
but, in being so, are less capable of find- chology for much of the first half of
ing solutions. Examples are generate and the twentieth century. During the reign
test (produce and apply all possible known of behaviorism, considerable success was
next steps, and see if any of them yields obtained in analyzing complex skills in terms
success) and means-ends analysis (repre- of acquired habits, that is, as a large collec-
sent the goal state, what you are trying to tion of stimulus-response pairs in the form
achieve; represent where your progress has of learned reactions associated to specific
brought you right now; and try to find some situations. The principle difficulties of this
currently available computational operator approach were associated with explaining
that can decrease some aspect of the distance the acquisition of abstract rules, creative use
between these. Repeat until done. Strong of language, general mental capacities, and
methods are more heavily dependent on rich logical reasoning in unfamiliar domains. It
knowledge of the problem-solving area and was around the middle of the century that
an understanding of what kinds of operations this hold on the field began to lo0sen. There
are likely to be successful in encountered was both a push side and a pull side to this
situations. They are domain specialists, not development.
generalists. On the push side, as we have noted,
When early AI was being applied in stimulus-response models were facing great
relatively simple and well-structured areas, difficulty in trying to account for complex
such as elementary games like checkers, human processes such as language, reason-
weak methods fared fairly well. As the ing, and abstractions that were indepen-
field developed and researchers started to dently coming under increasing investiga-
address richer, complex, and knowledge- tion. In this respect, the work of the linguist
laden task environments, such as medicine Chomsky (195 7) was critical. The findings
(Pauker, Gorry, Kassirer, & Schwartz, 1976; and theorizing out of linguistics were affect-
Shortliffe, 1976) and chemical spectral anal- ing psychology, in exposing what seemed to
ysis (Buchanan & Feigenbaum, 1978), the be significant inadequacies in accounting for
need for ever-stronger methods became complex psychological processes. A notable
clear. Portability across task domains had volume (Jakobovits & Miron, 1967), not
to be sacrificed in favor of capability, but surprisingly focusing on language, brought
narrowly restricted capability. The highly the camps head to head in their explana-
successful expert systems industry that tory systems for complex human activity.
eventually developed (Buchanan, Davis, & The Herculean effort by Osgood (1963 ),
Feigenbaum, Chapter 6) is in large part tes- reprinted in that volume, to save S-R theory
44 the cambridge handbook of expertise and expert performance
in the face of discoveries about language, just conduct such studies (Charness, 1976, 1979,
served in its cumbersomeness to prove the 1981; Chi, 1978; Chi, Feltovich, & Glaser,
inadequacies of S-R theories to account for 1981; Elstein, Shulman, & Sprafka, 1978;
language. Larkin, McDermott, Simon, & Simon,
On the pull side, theories, mechanisms, 1980), and the rest, as they say, is history.
and constructs were arising that showed The existence of this Cambridge Handbook
promise for providing an infrastructure is its own best evidence for the subsequent
to support a new kind of psychology. development and tremendous expansion
These included the development of the of the field of Expertise Studies into its
information-processing viewpoint in psy- current myriad forms.
chology, along with the platform to sup- It is interesting to think about whether a
port it, the computer. Electrical engineer field of expertise studies could have emerged
Newell and economist/philosopher Simon at all and if so, what it could possi-
believed that what they were doing was psy- bly have looked like if alternatives to
chology (see earlier quote)! In fact, they behaviorism had not emerged. For instance,
predicted in 195 8 that within ten years would we have discovered that experts do
most theories in psychology will take the not just complete tasks and solve problems
form of computer programs, or of quali- faster and better than novices, but often
tative statements about the characteristics attain their solutions in qualitatively differ-
of computer programs and discussed the ent ways? Would we have discovered that
nature of heuristic search and ill-structured experts frequently spend a greater propor-
problems (Simon & Newell, 195 8, p. 7). tion of their time in initial problem evalua-
In his landmark volume titled Cognitive tion compared to novices (e.g., Glaser & Chi,
Psychology, Ulric Neisser (1967) engaged 1988, regarding Experts spend a great deal
information-processing language and the of time analyzing a problem qualitatively;
computer metaphor as advances that helped Lesgold et al., 1988; see also Kelloggs Chap-
enable the creation of a cognitive psychol- ter 22 on planning by professional writers
ogy, and he acknowledged the contributions and Noice & Noices Chapter 28 on the deep
of Newell, Shaw, and Simon in this regard encoding by professional actors as they study
(Neisser, 1967, pp. 89). their lines)?
Additionally, and often not indepen- We will, of course, never know, but
dently, researchers were progressively there was considerable interest in complex
encroaching the realm of the mental, study- thought processes among some of the behav-
ing such things as planning (Miller, Galanter, iorists. For example, John B. Watson (1920)
& Pribram, 1960), thinking (Bartlett, 195 8; was the first investigator to study problem
Bruner, Goodnow, & Austin, 195 6), and solving by instructing a participant to think
mental structures and their functioning aloud while the participant figured out the
(Bartlett, 193 2; Miller, 195 6). Not surpris- function of an object (Ericsson, Chapter 13 ).
ingly, groundbreaking progress in this regard Neo-behaviorists, such as Berlyne (1965 ),
came from the information-processing proposed stimulus-response accounts for
camp in their studies of problem solving complex goal-directed thought and cogni-
(Newell & Simon, 1972), especially in their tive development. Today, behavior analysts
studies (following de Groot, 1946, 1965 ) recommend the collection of think-aloud
of expertise in chess (Chase & Simon, protocols to better understand complex per-
1973 a, 1973 b; See also Gobet & Charness, formance (Austin, 2000). Given the broad
Chapter 3 0). The clear, surprising, and even divide in the theoretical mechanisms used
enchanting findings (two people looking by cognitive and behavioral researchers, it
at the very same stimulus can see totally is interesting that researchers are converg-
different things, even things that are not ing on methods of collecting observable
actually there!) arising from this research process indicators and have mutual inter-
about the cognitive differences between est in large, reproducible differences in
experts and novices stimulated others to performance.
studies of expertise from psychological perspectives 45
The last peg in the story of expertise stud- and theorizing about mediating processes
ies that we consider is education and edu- and structures that support, and can be
cational psychology. There are at least two developed to produce, these superior per-
dimensions in the evolution of education formances (see later sections in this chap-
that are related to expertise studies, and that ter on Expertise Involves Larger and More
we have also seen in the other influences Integrated Cognitive Units; and Expertise
we have considered. First, like psychology, Involves Functional, Abstracted Representa-
educational theory and practice was under tions of Presented Information). Interestingly,
the influence of behaviorism in and around however, current theorizing about the criti-
the mid century (Skinner, 1960; Watson, cal role of deliberate practice in the devel-
1913 ). Both learning and teaching centered opment of expertise emphasizes mecha-
around establishing appropriate stimulus- nisms not incompatible with these earlier
response connections. Programmed learn- theories, in particular the need for clear
ing and teaching machines were in vogue. goals, repeated practice experiences, and
A representative example is the landmark the vital role of feedback about the qual-
volume co-edited by Robert Glaser (Lums- ity of attempts (Ericsson, Krampe, & Tesch-
daine & Glaser, 1960), who would go on Romer, 1993 ). In addition, it is possible that
to play a central role in newer incarna- discoveries from behaviorist research about
tions of educational and psychological the- different schedules of reinforcement (e.g.,
ory and practice. Essentially, a teaching Ferster & Skinner, 195 7), and their relation
machine, in doing programmed learning, to sustaining motivation and effort over long
would present questions or problems to periods of time, might contribute to our
learners, one by one, and depending on the understanding of how some people manage
students response either reinforce a correct to persevere through the very long periods
response or note an incorrect one (and per- of practice and experience, involving both
haps also provide some remedial guidance). successes and inevitably many failures, that
This process was believed to establish sta- we now know are so essential to the devel-
ble connections between problematic situa- opment of expert levels of skill. How to scaf-
tions and appropriate situational responses. fold sustained, consistent, purposeful effort,
What would expertise look like under such over very long periods of time and despite
a worldview? It is interesting in this regard inevitable setbacks, appears at this time to
to examine a statement about this made by be one of the great puzzles to be solved in
one of Behaviorisms founders: developing a science of human excellence
(see Hunt, Chapter 3 , for a discussion).
Mathematical behavior is usually regarded
With the emergence of the cognitive
not as a repertoire of responses involving
numbers and numerical operations, but as turn in psychology and educational psychol-
evidence of mathematical ability or the ogy, a new role for expertise studies also
exercise of the power of reason. It is true that emerged. Expert cognition was conceived
the techniques which are emerging from as the goal state for education, the cri-
the experimental study of learning are not terion for what the successful educational
designed to develop the mind or to fur- process should produce, as well as a mea-
ther some vague understanding of math- sure by which to assess its progress. In this
ematical relationships. They are designed, regard, advanced methods have now been
on the contrary, to establish the very behav- developed for eliciting and representing
iors which are taken to be evidences of such the knowledge of experts (see Hoffman &
mental states or processes. (Skinner, 1960,
Lintern, Chapter 12) and for observing and
pp. 111)
describing experts work practices in natu-
In this view, it seems expertise would be a ral settings (see Clancey, Chapter 8). Novice
matter of responding well in challenging sit- cognition (as well as that of various lev-
uations. Although modern views of exper- els of intermediates) could serve as initial
tise retain this criterion of superior perfor- states, as models of the starting place for the
mance, there is also considerable interest educational process. In a sort of means-ends
46 the cambridge handbook of expertise and expert performance
analysis, the job of education was to deter- A second theme related to expertise stud-
mine the kinds of operations that could ies that also appears in education, as well as
transform the initial conditions into the in the other contributors we have discussed,
desired more expertlike ones (Glaser, 1976). is related to weak and strong methods
Although it is tempting to believe that upon (Amirault & Branson, Chapter 5 ). As long
knowing how the expert does something, as there has been education, there has been
one might be able to teach this to novices controversy about what constitutes an edu-
directly, this has not been the case (e.g., Klein cated person, what such a person should
& Hoffman, 1993 ). Expertise is a long-term know and be able to do, and how to bring
developmental process, resulting from rich such a person about. Examination of the
instrumental experiences in the world and history of education as it relates to exper-
extensive practice. These cannot simply be tise (Amirault & Branson, Chapter 5 ) reveals
handed to someone (see the later section in the ebb and flow between understanding the
this chapter on Simple Experience Is Not Suf- object of education (expertise) to be the
ficient for the Development of Expertise). generalist (sound reasoning, broad knowl-
One venue in which expertise as goal edge, critical thinking) or the specialist (one
state has gained considerable use is who has undergone a great amount of train-
intelligent computer-based education, for ing and experience in a limited domain of
example, intelligent tutoring systems. (e.g., activity and has acquired a vast knowledge
Clancey & Letsinger, 1984; Forbus & base specifically tailored for that activity).
Feltovich, 2001; Sleeman & Brown, 1982). As with the development of artificial intel-
Such systems often utilize an expert ligence, our modern educational and psy-
model, a representation of expert compe- chological conception of expertise seems to
tence in a task, and a student model, a favor the specialist and specialized skills,
representation of the learners pertinent cur- honed over many years of extensive train-
rent understanding. Discrepancy between ing and deliberate practice (Ericsson, Chap-
the two often drives what instructional inter- ter, 3 8). The notion of an expert gen-
vention is engaged next. Another educa- eralist is difficult to capture within the
tional approach is to build tools for enhanc- current explanatory systems in expertise
ing and accelerating experience (e.g., Klein studies (e.g., Feltovich, Spiro, & Coulson,
& Hoffman, 1993 ; Spiro, Collins, Thota, & 1997; see also the discussion of the prepara-
Feltovich, 2003 ), and this is closely related tion for creative contributions by Weisberg,
to methods for analyzing the representa- Chapter 42).
tive tasks to be mastered (see Schraagen,
Chapter 11).
Some early research on the difference
between experts and novices led directly Toward Generalizable Charactistics
to the creation of new methods of instruc- of Expertise and Their Theoretical
tion. This is particularly true in medical Mechanisms
education, where early expert-novice stud-
ies (Barrows, Feightner, Neufeld, & Norman, From the kinds of beginnings just discussed,
1978; Elstein, Shulman, & Sprafka, 1978) expertise studies have become a large and
led to the creation of problem-based learn- active field. Fortunately, periodic volumes
ing (Barrows & Tamblyn, 1980). Over a have served to capture its state of devel-
long period of time, PBL (and variants) opment over time (Anderson, 1981; Bloom,
has come to pervade medical education, as 1985 ; Chase, 1973 ; Chi, Glaser, & Farr, 1988;
well as making significant inroads into all Clancey & Shortliffe, 1984; Ericsson, 1996a;
types of education, including K-12, univer- Ericsson & Smith, 1991a; Feltovich, Ford, &
sity, and every sort of professional educa- Hoffman, 1997; Hoffman, 1992; Starkes &
tion (see Ward, Williams, & Hancock, Chap- Allard, 1993 ; Starkes & Ericsson, 2003 ).
ter 14, for a review of the use of simulation The remainder of the current chap-
in training). ter attempts to crystallize the classic and
studies of expertise from psychological perspectives 47
enduring findings from the study of exper- is also characteristic of expertise involving
tise. It will draw on generalizable character- perceptual-motor skills (e.g., Fitts & Posner,
istics of expertise identified in earlier reviews 1967; Rosenbaum, Augustyn, Cohen, & Jax,
(Glaser & Chi, 1988; Chi, Chapter 2) and Chapter 29), as exemplified in many chap-
discuss them and other aspects in the light ters in this handbook, but in particular in
of the pioneering research that uncovered perceptual diagnosis and surgery (Norman,
them. We will also discuss the original theo- Eva, Brooks, & Hamstra, Chapter 19), sports
retical accounts for these findings. However, (Hodges, Starkes, & MacMahon, Chap-
where pertinent, we will also present more ter 27), and music (Lehmann & Gruber,
recent challenges and extensions to these Chapter 26).
classic accounts, including pertinent findings Some of the most solid early evidence for
and theoretical treatments reviewed in the specificity in expertise came from expert-
chapters of this handbook. novice difference studies in medicine, inves-
tigating the clinical reasoning of practition-
ers (Barrows et al., 1978; Elstein et al., 1978).
Expertise Is Limited in Its Scope and Elite
These studies showed that the same physi-
Performance Does Not Transfer
cian can demonstrate widely different pro-
There is a general belief that talented peo- files of competence, depending on his or her
ple display superior performance in a wide particular experiential history with differ-
range of activities, such as having superior ent types of cases. Indeed, in modern med-
athletic ability and superior mental abilities. ical education, where assessment of clinical
However, if we restrict the claims to individ- skill is often evaluated by performance on
uals who can perform at very high levels in real or simulated cases, it has been found
a domain, then it is clear that people hardly that because of the case-specificity of clin-
ever reach an elite level in more than a sin- ical skill, a large number of cases (on the
gle domain of activity (Ericsson & Lehmann, order of fourteen to eighteen) are needed
1996). This has proven to be one of the most to achieve an acceptably reliable assess-
enduring findings in the study of expertise ment of skill (Petrusa, 2002; Norman et al.,
(see Glaser & Chi, 1988, Characteristic 1). Chapter 19).
There is little transfer from high-level profi-
ciency in one domain to proficiency in other
Knowledge and Content Matter Are
domains even when the domains seem,
Important to Expertise
intuitively, very similar.
For example, in tasks similar to those used In and around the late 1960s and the
in the Simon and Chase chessboard stud- 1970s, maintaining a traditional distinction
ies, Eisenstadt and Kareev (1979) studied the between domain-specific skills and general
memory for brief displays for expert GO cognitive abilities was becoming less ten-
and Gomoko players. Even though these two able. In research studies, knowledge was no
games are played on the same board and use longer seen as a nuisance variable but as a
the same pieces, GO players showed quite dominant source of variance in many human
poor performance on Gomoko displays, and tasks. In particular, Newell and Simon (1972)
vice versa. In tasks involving political sci- found that problem solving and skilled per-
ence, for example, devising plans for increas- formance in a given domain were primar-
ing crop production in the Soviet Union, ily influenced by domain-specific acquired
Voss and colleagues (Voss, Greene, Post, & patterns and associated actions. Domain-
Penner, 1983 ; Voss, Tyler, & Yengo, 1983 ) specific skills and knowledge were also found
found that experts in chemistry (chem- to influence even basic cognitive abilities.
istry professors) performed very much like For example, Glaser and others (Pellegrino
novices in political science, in comparison & Glaser, 1982a,1982b) investigated basic
to political science experts (see Voss & foundations of intelligence, including induc-
Wiley, Chapter 3 3 , and Endsley, Chapter 3 6, tion, and found evidence that even these
for more recent examples). Task specificity were strongly influenced by a persons
48 the cambridge handbook of expertise and expert performance
knowledge in the operative domain (for rectly the implications of a simple, single-
example, a persons conceptual knowledge conditional logical expression when it was
about numbers in number analogy and num- stated in terms of a realistic setting (e.g.,
ber series tasks). a work setting: every purchase over thirty
Acquired knowledge in a domain was dollars must be approved by the regional
found to be associated with changes in manager) versus when the expression was
fundamental types of cognitive processing. stated in an understandable but less mean-
For example, drawing on the expert-novice ingful form (e.g., every card with a vowel
paradigm, Chi (1978) compared experi- on the front must have an integer on
enced chess-playing children with other the back).
children in their performance on memory These kinds of studies in the psychology
and learning tasks related to chess. The dif- of learning and reasoning were mirrored by
ferences in experience, knowledge, and skill developments within artificial intelligence.
in chess produced differences, in favor of There was an evolution from systems in
the chess players, in such basic learning pro- which knowledge (declarative) and reason-
cesses as the spontaneous use of memory ing (procedural) were clearly separated, to
strategies (like grouping and rehearsal), the systems in which these components were
ability to use such strategies even under indistinct or at least strongly interacted. For
experimental prompting, and the amount of example, early computer systems, such as
information that could be held in short-term Greens QA3 (Green, 1969) and Quillians
memory (Chi, 1978). TLC (Quillian, 1969), utilized databases of
Voss and colleagues (Chiesi, Spilich, & declarative knowledge and a few general-
Voss, 1979; Spilich, Vesonder, Chiesi, & purpose reasoning algorithms for operating
Voss, 1979) extended this kind of research on those knowledge bases. Such systems
into other forms of learning. Studying high- were progressively supplanted by ones in
and low-knowledge individuals with regard which the separation between knowledge
to the game of baseball, they found that, and reasoning was not nearly as distinct, and
compared to the low-knowledge individu- in which general reasoning algorithms gave
als, high-knowledge ones exhibited superior way to more narrowly applicable reasoning
learning for materials from that and only strategies, embedded in procedures for oper-
that particular domain. In particular, high- ating within specific domains of knowledge
knowledge individuals had greater recogni- (e.g., Norman, Rumelhart, & LNR, 1979;
tion and recall memory for new material, Sacerdoti, 1977; VanLehn & Seely-Brown,
could make useful inferences from smaller 1979; Winograd, 1975 ).
amounts of partial information, and were It was within this kind of context that
better able to integrate new material within studies of expertise and expert-novice differ-
a coherent and interconnected framework ences, along with the growth of knowledge-
(organized, for instance, under a common intense expert systems in artificial intel-
goal structure). ligence (e.g., Shortliffe, 1976; Buchanan,
Some studies showed reasoning itself to Davis, & Feigenbaum, Chapter 6), began also
be dependent on knowledge. Wason and to emphasize the criticality of knowledge.
Johnson-Laird (1972) presented evidence This was evident in the progression in AI
that individuals perform poorly in testing from weak to strong methods, and within
the implications of logical inference rules psychology in the growing recognition of the
(e.g., if p then q) when the rules are stated role in expertise of such knowledge-based
abstractly. Performance greatly improves for features as perceptual chunking, knowledge
concrete instances of the same rules (e.g., organization, knowledge differentiation, and
every time I go to Manchester, I go by effective perceptual-knowledge coupling.
train). Rumelhart (1979), in an extension This research clearly rejects the classi-
of this work, found that nearly five times cal views on human cognition, in which
as many participants were able to test cor- general abilities such as learning, reasoning,
studies of expertise from psychological perspectives 49
problem solving, and concept formation cor- Expertise Involves Larger and More
respond to capacities and abilities that can Integrated Cognitive Units
be studied independently of the content
domains. In fact, inspired by the pioneer- With increased experience and practice,
ing work by Ebbinghaus (1885 /1964) on most people cognitively organize the per-
memory for nonsense syllables, most lab- ceptually available information in their
oratory research utilized stimulus materi- working environment into larger units. This
als for which the prior experience of par- is a classic and one of the best-established
ticipants was minimized, in order to allow phenomenon in expertise (Glaser & Chi,
investigators to study the cognitive processes 1988, Characteristic 2). It is supported by
of learning, reasoning, and problem solv- a long line of research, but was first discov-
ing in their purest forms. This kind of ered in the game of chess (see also Gobet &
research, some examples of which were dis- Charness, Chapter 3 0).
cussed earlier in this section, showed that In the 1960s and early 1970s, de Groot
participants, when confronted with unfa- (1965 ) and Chase and Simon (1973 a, 1973 b)
miliar materials in laboratory tasks, demon- studied master-level and less-accomplished
strated surprisingly poor performance. In chess players. In the basic experimental task,
contrast, when tested with materials and participants were shown a chess board with
tasks from familiar domains of everyday pieces representing game positions from real
activity, people exhibited effective reason- games. Participants were shown the posi-
ing, learning, and problem solving. Simi- tions for only five seconds, and they were
larly, the performance of experts is supe- then asked to reproduce the positions they
rior to novices and less-skilled individuals had seen.
primarily for tasks that are representative After this brief glance, an expert was able
of their typical activities in their domain to reproduce much more of the configura-
of expertise the domain specificity of tion than a novice. In the studies by Chase
expertise (see the earlier section Expertise and Simon (1973 a, 1973 b) noted earlier, the
Is Limited). expert recalled four to five times the num-
In the expert-performance approach to ber of pieces recalled by the novice. In the
expertise, researchers attempt to identify similar studies by de Groot, the recall per-
those tasks that best capture the essence of formance by world-class players was nearly
expert performance in the corresponding perfect (for 25 -piece boards). In contrast,
domain, and then standardize representa- novices were able to reproduce about five
tive tasks that can be presented to experts pieces, or about the number of items that can
and novices. By having experts repeat- be maintained in short-term memory exclu-
edly perform these types of tasks, exper- sively by rehearsal.
imenters can identify, with experimental The original, classical explanation by
and process-tracing techniques, those com- Chase and Simon (Simon & Chase, 1973 ;
plex mechanisms that mediate their supe- Chase & Simon, 1973 a, 1973 b) for expert
rior performance (Ericsson, Chapter 13 and superiority involved chunking in percep-
Chapter 3 8). The experts superior perfor- tion and memory. With experience, experts
mance on tasks related to their domain of acquire a large vocabulary, or memory
expertise can be described by psychometric store, of board patterns involving groups of
factors (expert reasoning and expert working pieces, or what were called chunks. A chunk
memory) that differ from those general abil- is a perceptual or memory structure that
ity factors used to describe the performance bonds a number of more elementary units
of novices (Horn & Masunaga, Chapter 3 4, into a larger organization (e.g., the individual
and see Ackerman & Beier, Chapter 9, for a letters c, a, and r into the word car).
review of individual differences as a function When experts see a chess position from a real
of level of expertise). In short, knowledge game, they are able to rapidly recognize such
matters (Steier & Mitchell, 1996). familiar patterns. They can then associate
50 the cambridge handbook of expertise and expert performance
these patterns with moves stored in mem- being able to recall multiple chess boards
ory that have proven to be good moves in presented in rapid succession (see Ericsson,
the past. Novices do not have enough expo- Chapter 13 , and Gobet & Charness, Chap-
sure to game configurations to have devel- ter 3 0, for an extended discussion of new
oped many of these kinds of patterns. Hence theoretical mechanisms accounting for the
they deal with the board in a piece-by-piece experts expanded working memory). The
manner. Similarly, when experts are pre- experts superior ability to encode repre-
sented with chess boards composed of ran- sentative information from their domain of
domly placed pieces that do not enable the expertise and store it in long-term memory,
experts to take advantage of established pat- such that they can efficiently retrieve mean-
terns, their advantage over novices for ran- ingful relations, provides an alternative to
dom configurations amounts to only a few the original account of superior memory in
additional pieces. terms of larger chunks stored in STM. There
These basic phenomena attributed to is another, similar characteristic of expertise.
chunking were replicated many times, in It has to do with the nature and organiza-
chess but also in other fields (e.g., the tion of the perceptual encoding and mem-
games of bridge, Engle & Bukstel, 1978; ory structures experts develop and use. This
GO, Reitman, 1976; and electronics, Egan & is discussed next.
Schwartz, 1979). In many such studies, it is
the chunk size that is larger for experts. Both
Expertise Involves Functional, Abstracted
the novice and the expert are constrained
Representations of Presented Information
by the same limitations of short-term (or
working) memory (Cowan, Chen, & Rouder, Some studies, utilizing methods similar to
2004; Miller 195 6). However, expert chunks the Simon and Chase chessboard paradigm,
are larger. A chess novice sees a number of examined the nature of expert and novice
independent chess pieces; the expert recog- cognitive units, such as chunks or other
nizes about the same number of larger units. knowledge structures. Chase and Simon
For example, one chunk of chess pieces for (1973 a, 1973 b) themselves analyzed the
an expert might be a king defense configu- characteristics of the chess pieces their
ration, composed of a number of individual experts grouped together as they repro-
chess pieces. duced a chess position after a brief presenta-
As we have just discussed, it was orig- tion. Expert configurations of chess pieces
inally believed that experts develop larger were based largely on strategic aspects of
chunks and that these enable the expert to the game, for example, configurations repre-
functionally expand the size of short-term senting elements of threat or opportunity. It
or working memory. However, in the mid- was not clear how novice units were orga-
1970s, Charness (1976) showed that expert nized. Glaser and Chi (1988) identified a
chess players do not rely on a transient short- related general characteristic, namely, that
term memory for storage of briefly presented Experts see and represent a problem in their
chess positions. In fact, they are able to recall domain at a deeper (more principled) level
positions, even after the contents of their than novices; novices tend to represent a
short-term memory have been completely problem at a superficial level (p. xviii). Our
disrupted by an interfering activity. Subse- characterization for expert representations,
quent research has shown that chess experts functional and abstracted as elaborated
have acquired memory skills that enable next, simply seeks to provide a bit more
them to encode chess positions in long- insight into the nature of deep (see Chi,
term working memory (LTWM, Ericsson & Chapter 10, for a review of research on
Kintsch, 1995 ). The encoding and storage of assessments of experts cognitive represen-
the chess positions in LTWM allow experts tations).
to recall presented chess positions after dis- Early studies involving bridge (Charness,
ruptions of short-term memory, as well as 1979, Engle & Bukstel, 1978) and electronics
studies of expertise from psychological perspectives 51
(Egan & Schwartz, 1979), patterned after the Reitman, Reuter, and Hirtle (1981) pre-
Chase and Simon procedure, showed simi- sented a list of 21 commands in the ALGOL
lar results. In the bridge studies, experts and language to ALGOL experts, students after
novices were briefly presented depictions one ALGOL course, and students at the
of four-handed bridge deals, and they were beginning of an ALGOL course. Participants
required to reproduce these deals. Experts were given 25 recall trials after they ini-
reproduced the cards by suit, across hands. tially learned the list. The organization of
They remembered cards of the same suit the recalled items by pre-ALGOL students
from three hands and inferred the fourth; was by surface features of commands (e.g.,
this is an organization useful in playing the commands with the same beginning letter or
game of bridge. Novices recalled the cards same length of command name) and groups
by order of card rank within hands, an orga- of commands forming natural language seg-
nization not useful to supporting strategic ments (e.g., STRING IS NULL BITS) that
aspects of the game. In electronics, sub- have no conceptual meaning within the lan-
jects were shown an electronic circuit dia- guage. Experts, in contrast, grouped com-
gram, which they were then to reproduce. mands that formed mini ALGOL algorithms
Experts grouped individual diagram com- (e.g., formation of loops) or constituted
ponents into major electronic components types of ALGOL data structures. Students,
(e.g., amplifiers, filters, rectifiers). Novice after an ALGOL course, produced group-
organization was based largely on the spa- ings that were a mixture of surface-related
tial proximity of symbols appearing in the and meaningful ALGOL organizations.
diagram. In a similar study, Adelson (1981) pre-
Similar results have been shown from sented a list of programming commands,
yet other fields, using somewhat different constituting three intact computer pro-
methodologies that compared the perfor- grams, scrambled together and out of order,
mance of groups of adults who differ in to expert and novice programmers. Partici-
their knowledge about a given domain. For pants were required to recall the list. Over
example, Voss and colleagues (Spilich et recall trials, experts reconstructed the orig-
al., 1979) studied ardent baseball fans and inal three algorithms. The organization of
more casual baseball observers. Participants novice recall was by syntactic similarities in
were presented a colorful description of a individual command statements, regardless
half-inning of baseball and were then asked of the embedded source algorithms. Sonnen-
to recall the half-inning. Expert recall was tag, Niessen, and Volmer (Chapter 21) pro-
structured by major goal-related sequences vide a review of the more recent research
of the game, such as advancing runners, scor- on knowledge representations and superior
ing runs, and preventing scoring. Novices performance of software experts.
recall contained less integral components, Other pertinent findings came from early
for example, observations about the weather work in physics (Chi et al., 1981) and
and the crowd mood. Novice recall did medicine (Feltovich, Johnson, Moller, &
not capture basic game-advancing, sequen- Swanson, 1984; Johnson et al., 1981). In the
tial activity nearly as well. More recent basic task from the physics study, problems
research on fans that differ in their knowl- from chapters in an introductory physics
edge about soccer and baseball has found text were placed on individual cards. Expert
that comprehension and memory for texts (professors and advanced graduate students)
describing games from these sports is more and novice (college students after their first
influenced by relevant knowledge than by mechanics course) physics problem solvers
verbal IQ scores (see Hambrick & Engle, sorted the cards into groups of problems
2002, for a recent study and a review of they would solve in a similar manner.
earlier work). The finding was that experts created groups
Two early studies of computer program- based on the major physics principles (e.g.,
ming produced similar results. McKeithen, conservation and force laws) applicable in
52 the cambridge handbook of expertise and expert performance
the problems solutions. Novice groupings awareness). This general theoretical frame-
were organized by salient objects (e.g., work can account for the slow acquisition of
springs, inclined planes) and features con- abstract representations that support plan-
tained in the problem statement itself. Sim- ning, reasoning, monitoring, and evaluation
ilarly, in studies of expert and novice diag- (Ericsson, Patel, & Kintsch, 2000). For exam-
noses within a subspecialty of medicine, ple, studies of expert fire fighters have shown
expert diagnosticians organized diagnostic that experts interpret any scene of a fire
hypotheses according to the major patho- dynamically, in terms of what likely pre-
physiological issue relevant in a case (i.e., ceded it and how it will likely evolve. This
constituting the Logical Competitor Set kind of understanding supports efforts to
of reasonable alternatives for the case, e.g., intervene in the fire. Novices interpret these
lesions involving right-sided heart volume scenes in terms of perceptually salient char-
overload), whereas novice hypotheses were acteristics, for example, color and inten-
more isolated and more dependent on par- sity (Klein, 1998, and see Ross, Shafer, &
ticular patient cues. Klein, Chapter 23 ). Studies of expert sur-
Zeitz (1997) has reviewed these and geons have shown that some actions within
more recent studies of this type, investigat- a surgery have no real value for imme-
ing what she calls experts use of Mod- diate purposes, but are made in order to
erately Abstracted Conceptual Representa- make some later move more efficient or
tions(MACRs), which are representational effective (Koschmann, LeBaron, Goodwin,
abstractions of the type just discussed. She & Feltovich, 2001). The research on expert
proposes numerous ways in which such chess players shows consistent evidence for
abstraction aids the efficient utilization of extensive planning and evaluation of conse-
knowledge and reasoning by experts. These quences of alternative move sequences (See
include: (a) the role of abstracted repre- Ericsson, Chapter 13 , and Gobet & Charness,
sentations in retrieving appropriate material Chapter 3 0). Furthermore, there is consid-
from memory (e.g., Chi et al., 1981); (b) the erable evidence pertaining to experts elab-
schematic nature of MACRs in integrating orated encoding of the current situation,
information and revealing what information such as in situational awareness (Endsely,
is important, (c) providing guidance for a Chapter 3 6), mental models (Durso & Dat-
line of action and supporting justification tel, Chapter 20), and LTWM (Noice &
for such a line of approach (e.g., Phelps & Noice, Chapter, 28).
Shanteau, 1978; Schmidt et al., 1989; Voss In summary, research conducted in the
et al., 1983 ); (d) aiding productive analogi- last thirty or so years indicates that expert
cal reasoning (e.g., Gentner, 1988); and (e) performers acquire skills to develop com-
providing abstract representations that sup- plex representations that allow them imme-
port experts reasoning and evaluation of diate and integrated access to information
diagnostic alternatives (e.g., Patel, Arocha, & and knowledge relevant to the demands of
Kaufman, 1994). action in current situations and tasks. These
The functional nature of experts rep- acquired skills can account also for the their
resentations extends to entire activities or superior memory performance when they
events. Ericsson and Kintsch (1995 ) pro- are given a task, such as recalling a briefly
posed that experts acquire skills for encod- presented chess position, as in the stud-
ing new relevant information in LTWM to ies by Chase and Simon (1973 a, 1973 b).
allow direct access when it is relevant and to Novices, on the other hand, lack such knowl-
support the continual updating of a men- edge and associated representations and
tal model of the current situation akin skills, and thus perform these tasks with the
to the situational models created by readers only knowledge and skills they have avail-
when they read books (see Endsley, Chap- able. They try to impose organization and
ter 3 6, on the experts superior ability to meaningful relations, but their attempts are
monitor the current situation situational piecemeal and less relevant to effectively
studies of expertise from psychological perspectives 53
functioning in the task domain, organized, activity in a certain part of the brain, but
for example, by items named in a situation, not a shift in anatomical loci (Jansma, Ram-
current salient features, proximity of entities sey, Slagter, & Kahn, 2001; Hill & Schneider,
to others, or superficial analogies. Chapter 3 7).
There are many examples in the early
expertise-related literature of the effects
Expertise Involves Automated
of practice on dual-task performance of
Basic Strokes
experts. For example, expert typists can type
Most people considered to be experts are and recite nursery rhymes at the same time
individuals with extreme amounts of prac- (Shaffer, 1975 ). Skilled abacus operators can
tice on a circumscribed set of tasks in answer routine questions (What is your
their work environment. For example, some favorite food?) without loss of accuracy or
expert radiologists estimated they had ana- speed in working with the abacus (Hatano,
lyzed more than half a million radiographs Miyake, & Binks, 1977). After six weeks of
(X-rays) in their careers (Lesgold et al., practice (one hour per day), college students
1988). Such experience, appropriately con- could read unfamiliar text while simulta-
ducted, can yield effective, major behav- neously copying words read by an experi-
ioral and brain changes (Hill & Schneider, menter, without decrement in reading speed
Chapter 3 7). or comprehension (Spelke, Hirst, & Neisser,
Research on the effects of practice has 1976).
found that the character of cognitive oper- Automaticity is important to expertise. It
ations changes after even a couple of hours appears it has at least two main functions.
of practice on a typical laboratory task. The first has to do with the relationship
Operations that are initially slow, serial, between fundamental and higher-order cog-
and demand conscious attention become nitive skills, and the second has to do with
fast, less deliberate, and can run in parallel the interaction between automaticity of pro-
with other processes (Schneider & Shiffrin, cesses and usability of available knowledge.
1977). With enough practice, one can learn With regard to the first, in complex skills
how do several tasks at the same time. with many different cognitive components,
Behavioral studies of skill acquisition have it appears that some of the more basic ones
demonstrated that automaticity is central (e.g., fundamental decoding, encoding of
to the development of expertise, and prac- input) must be automated if higher-level
tice is the means to automaticity (Posner skills such as reasoning, comprehension, inf-
& Snyder, 1975 , see also Proctor & Vu, erence, monitoring, and integration are ever
Chapter 15 ). Through the act of prac- to be proficient (e.g., Logan, 1985 ; Endsley,
tice (with appropriate feedback, monitor- Chapter 3 6). For example, in a longitudi-
ing, etc.), the character of cognitive opera- nal study, Lesgold and Resnick (1982) fol-
tions changes in a manner that (a) improves lowed the same group of children from their
the speed of the operations, (b) improves initial exposure to reading in kindergarten
the smoothness of the operations, and (c) through third and fourth grade. They found,
reduces the cognitive demands of the oper- for example, that if basic reading skills do
ations, thus releasing cognitive (e.g., atten- not become automated, such as the decoding
tional) resources for other (often higher) and encoding of letters and words, compre-
functions (e.g., planning, self-monitoring; hension skills will not substantially develop.
see also Endsley Chapter 3 6). Automatic Furthermore, the relationship seems to be
processes seem resistant to disruption by causal; that is, speed increases in word skills
reduced cognitive capacity and, to a lim- predict comprehension increases later on,
ited degree, are largely resource insensi- whereas increases in comprehension do not
tive (Schneider & Fisk, 1982). Interestingly, predict increases in word facility. However,
fMRI studies have demonstrated that shifts subsequent pertinent research has accentu-
to automaticity reveal a shift (decrease) in ated the complex-nature of the relationship
54 the cambridge handbook of expertise and expert performance
between automated basic processes and vant contexts are encountered. In this model
higher-order deliberate ones and point to of the experts working memory storage in
the need for continued research (Hill & LTM, the large capacity of LTM allows the
Schneider, Chapter 3 7). expert to preserve access to a large body of
There is also a possible interaction relevant information without any need to
between automaticity of processes and the actively maintain the information in a lim-
usability of available knowledge. Investiga- ited general capacity STM (Ericsson, Chap-
tors (e.g., Feltovich et al., 1984; Jeffries ter 13 ; Gobet & Charness, Chapter 3 0; Noice
et al., 1981) have suggested that a major & Noice, Chapter 28; Wilding & Valentine,
limitation of novices is their inability to Chapter 3 1).
access knowledge in relevant situations, even
when they can retrieve the same knowledge
Expertise Involves Selective Access
when explicitly cued by the experimenter.
of Relevant Information
Problems in knowledge usability may be
associated with overload or inefficiency in Within the classical expertise framework
using working (or short-term) memory. The based on chunking, questions about access
usable knowledge of experts may, in turn, to task-relevant information are important
result from the subordination of many task issues, and a critical aspect of intelligence
components to automatic processing, which (Sternberg, 1984). Given the functional
increases capability for controlled manage- nature of expert representations, how are
ment of memory and knowledge application they properly engaged in the context of solv-
(cf. Perfetti & Lesgold, 1979). ing a problem? To what kind of problem fea-
An alternative proposal about usability tures do experts attend? How are these fea-
of knowledge has subsequently been made tures linked up to the significant concepts
by Ericsson and Kintsch (1995 ), in which in memory? In a sense, having a trace laid
experts acquire skills that are designed to down in memory is not a sufficient condi-
encode relevant information in long-term tion for use. Extant traces must be accessed
memory (LTM) in a manner that allows and important non-extant traces must be
automatic retrieval from LTM when later inferred or otherwise computed.
needed, as indicated by subsequent acti- This characteristic of expertise addresses
vation of certain combinations of cues in the critical problem of accessing knowledge
attention. They argued that experts acquire structures. This development overcomes (at
LTWM skills that enable them, when they least) two difficulties for expertise as a big
encounter new information (such as a switch (Newell, 1973 ) between the recog-
new symptom during an interview with a nition of familiar events and application
patient), to encode the relevant associations of experience associated with those events
such that when yet other related information (see also Ross, Shafer, & Klein, Chapter 23 ,
is encountered (such as subsequent infor- recognition-primed decision making). The
mation reported by the patient), the expert first of these is related to the variability
will automatically access relevant aspects in events; one cannot step into the same
of the earlier information to guide encod- river twice. The useful utilization of events
ing and reasoning. The key constraint for as familiar requires a degree of appropri-
skilled encoding in LTM is that the expert ate abstraction, both in the event features
be able to anticipate potential future con- utilized and in the memory organization
texts where the encountered information imposed on the memory models them-
might become relevant. Only then will the selves. The former adaptation is reflected in
expert be able to encode encountered infor- expert utilization of abstracted features for
mation in LTWM in such a way that its problem classification, features whose loci
future relevance is anticipated and the rel- in a problem statement are not apparent
evant pieces of information can be automat- (Chi et al., 1981). The latter adaptation is
ically activated when the subsequent rele- reflected in the development of hierarchical
studies of expertise from psychological perspectives 55
organizations, which characterize expert or students are given early exposure to rep-
experienced memory (e.g., Feltovich et al., resentative clinical situations. This type of
1984; Patil, Szolovits, & Schwartz, 1981). training forces learners to develop mental
Critical to this characteristic is selectivity. representations and an LTWM that support
Selectivity is based on the attribution of dif- medical reasoning under real-time, represen-
ferential importance or, broadly conceived, a tative constraints (Norman, Eva, Brooks, &
separation of signal from noise either in the Hamstra, Chapter 19; Ericsson, Chapter 13 ;
features extracted from events or on inter- Endsley, Chapter 3 6).
nal cognitive processes themselves (see also
Hill & Schneider, Chapter 3 7). Selectivity,
Expertise Involves Reflection
as a means of task adaptation, is assumed
to be forced on the human based on their Another challenge to the traditional infor-
limited cognitive capacity. With regard to mation processing view, with its severe con-
events, selectivity involves the abstraction of straints on cognitive capacity, concerns the
invariances of the discriminating cues that experts ability not just to perform effec-
define types of situations or are otherwise tively but also to be able to reflect on
integral to a task. Expertise, then, involves their thought processes and methods (Glaser
learning which information is most useful & Chi, 1988, Characteristic 7 (see also
and which is tangential or superfluous (e.g., Zimmerman, Chapter 3 9). Metacognition is
Chi et al., 1981; Hinsley et al., 1978; Patel & knowledge about ones own knowledge and
Groen, 1991; Spilich et al., 1979). In certain knowledge about ones own performance
types of stable environments, the impor- (Flavell, 1979). It is what an individual
tant invariance is well defined and the task knows about his or her own cognitive pro-
is sufficiently constrained so that the mech- cesses. Its relevance to expertise is derived,
anisms linking selectivity and performance in part, from the observation that experts
can be explicated. For example, as consistent are graceful in their reasoning process. As
with the LTWM hypothesis, skilled typists Bartlett (195 8) notes, Experts have all the
appear to achieve subordination, usability, time in the world. There is an element
and access by developing integrated repre- of unencumbered elegance in expert per-
sentations of letters and key presses that formance, the underpinnings of which are
facilitate translation between perception and based on the efficient management and con-
response (Rieger, 2004). trol of the adaptive processes. A source for
This theme of expertise also reflects the this might be in abstracted layers of control
general problem of knowledge inversion; and planning.
that is, the notion of moving from a concept- The traditional (classical) account of
centered mode of reasoning to a mode that metacognition within the information-
must somehow scan the problem features processing model is that abstract descrip-
for regularities, incorporate abstraction, tions of plans and procedures enable an
integrate multiple cues, and accept natural individual to operate on or manipulate
variation in patterns to invoke aspects of the problem-solving operations, for example, to
relevant concept. We find this in many fields. modify and adjust them to context. They
For example, medical students acquire much also provide a general organizational struc-
specific disease-centered knowledge ture that guides and organizes the details
given disease X, this is the underlying patho- of application, so that a general line of
physiology, these are the variations, and reasoning can be maintained despite low-
these are the classic manifestations. When level (detailed) fluctuations and variations.
faced with a patient, however, they are Novice physics problem solvers, in con-
presented with just the opposite situation: trast to experts, have no abstract or meta-
Given a patient, what is the disease? Recent level descriptions for their basic problem-
developments in medical education focus solving operators, which for them are
on case-oriented learning in which medical physics equations (Chi et al., 1981). Rather,
56 the cambridge handbook of expertise and expert performance
operators are tied directly to problem details, nosis for a patient, their rationale for pos-
show little modifiability, and can only orga- sible alternatives is generally incompatible
nize problem-solving activity locally (i.e., at with the symptoms presented. Experts fail
the level of isolated problem components gracefully; novices crash. Vimla Patel and
present). her colleagues (Groen & Patel, 1988; Patel
In addition to abstraction in control and & Groen, 1991) have found that medical
planning, there must also be mechanisms for experts are able to explain their diagnoses
maintaining information to allow efficient by showing how the presented symptoms
back-tracking or starting over when lines of are all explained by the proposed integrated
reasoning need to be modified or abandoned. disease state, whereas less advanced medical
Largely, the traditional view proposes that students have a more piece-meal represen-
experts deal with the severe working- tation that is less well integrated.
memory demands required by backtracking Metacognition, then, is important for
by minimizing the need for it. For exam- people to test their own understanding and
ple, experts can attempt to withhold deci- partial solutions to a problem. This kind of
sions until they are sufficiently constrained monitoring prevents blind alleys, errors, and
to restrict the options. In other cases when the need for extensive back-up and retrac-
decisions are under-constrained, experts can tion, thus ensuring overall progress to a goal.
rely on abstract solution descriptions and In addition, these same kinds of monitoring
conditions for solution (constraints) that behaviors are critical throughout the pro-
both guide the search for solutions and help cess of acquiring knowledge and skills on
eliminate alternatives. which expertise depends. The mental rep-
The traditional information procesing resentations developed by aspiring experts
view has difficulties in accounting for the have multiple functions. They need to allow
possibility that experts might be disrupted efficient and rapid reactions to critical sit-
or otherwise forced to restart their plan- uations, and they need to allow modifia-
ning. More recent research has shown that bility, mechanisms by which a skilled per-
experts are far more able to maintain large former, for instance, adjusts his performance
amounts of information in working memory. to changed weather conditions, such as a
For example, chess masters are able to play tennis player dealing with rain or wind, or
chess games with a quality that approaches adjusts to unique characteristics of the place
that of normal chess-playing under blind- of performance, such as musicians adjust-
folded conditions in which perceptual access ing their performance to the acoustics of
to chess positions is withheld (for a review the music hall. Furthermore, these repre-
see Ericsson et al., 2000; Ericsson & Kintsch, sentations need to be amenable to change
2000). Chess masters are able to follow so aspiring expert performers can improve
multiple games when they are presented aspects and gradually refine their skills and
move by move and can recall the locations their monitoring representations.
of all pieces with high levels of accuracy. Experts, for the most part, work in the
Chess masters are also able to recall a series realm of the familiar (familiar for them, not
of different chess positions when they are for people in general) and may often be
briefly presented (5 seconds per position). In able to generate adequate actions by rapid
studies of expert physicians (e.g., Feltovich, recognition-based problem solving (Klein,
Spiro, & Coulson, 1997), it was found that 1998). The same experts are also the indi-
when experts do not know the correct diag- viduals called on to address the subtle,
nosis for a patient, they often can give a plau- complicated, and novel problems of their
sible description of the underlying patho- field. They need to recognize when the task
physiology of a disease; that is, they are able they are facing is not within their normal,
to reason at levels that are more fundamen- routine domain of experience and adjust
tal and defensible in terms of the symptoms accordingly (Feltovich et al., 1997); this
presented. When novices fail to reach a diag- is just one of many pertinent aspects of
studies of expertise from psychological perspectives 57
mediate between long-term memory and mits expanding the functional size of work-
the environment (Baddeley, 2000, 2002). ing memory and increasing the efficiency of
When focusing attention on making a par- search. This phenomenon has been experi-
ticular decision or solving a particular prob- mentally demonstrated across a remarkably
lem, three types of events occur that wide variety of domains.
are critical for effective reasoning: (1) we The role and function of automaticity
seek (and perceive) data from the envi- within expertise is important in this regard
ronment, (2) we bring relevant knowledge also. Automaticity seems to be entwined
to bear from our long-term memory to with functional organization, chunking, and
working memory and, by reviewing the conditions of application. They work in con-
data in the presence of relevant knowledge cert to adapt to the demands of the task,
retrieved from long-term memory, (3 ) we under the constraints of both the task and
draw inferences about what is going on their own capabilities to make appropriate
which may lead to seeking more data and use of our memory. Automaticity, then, is
activating more knowledge. intricately bound with the overall adaptation
Finally, there is a limit of long-term mem- of the system through knowledge reorgani-
ory access. To what extent we truly forget zation and refinement.
things is uncertain, so there may not actually The general argument is that expert
be an arbitrary size constraint on this aspect knowledge structures and procedures are
of our long-term memory. That, however, is reorganized in directions that enable effec-
not the issue. What is certain is that we lose tive application to task demands of a working
access to (or the power to evoke) the knowl- environment. As we have discussed, most
edge stored. A typical demonstration of this of these changes are adaptations that enable
is the tip of the tongue phenomenon utilization of large amounts of information
in which you know that you know some- in the context of limited internal-processing
thing, but cannot retrieve it (Brown 1991; resources (in particular those imposed by
Brown & McNeill 1966). Therefore, even the small capacity of short-term or work-
though we may scan the right data in an ing memory). Grouping or chunking on
analysis, there is no guarantee that we will information structures and procedure com-
be able to trigger the appropriate knowledge ponents functionally increases the size of
in long-term memory to allow us to make working memory and its efficiency. More
correct inferences from those data. In prac- information can be considered for each
tice, a large part of expert problem solving is unit in working memory. Expert selec-
being able to access relevant knowledge, at tivity, discrimination, and abstraction (dis-
the right time, for use in working memory. cussed earlier) insure that only the most use-
This traditional approach to expertise was ful information is thrown into competition
founded on the powerful theoretic assump- for resources. Automaticity is a means of
tion that experts cognitive processes, such as restructuring some procedures so that work-
generating, representing, and using knowl- ing memory is largely circumvented, free-
edge, had to conform to these severe lim- ing resources for other cognitive chores. It
its. This theory proposed many mecha- is a tension between high information load
nisms by which experts would be able to and limited internal resources that encour-
functionally adapt to these constraints to ages the development of strategies for the
produce superior performance. The expert efficient use of knowledge and processing.
chunking mechanism, for example, permits This pioneering theory of expertise
a vocabulary that is much more robust (Simon & Chase, 1973 ) has been and remains
and complex than the novice can invoke. very influential and has been extended
Although both the expert and the novice with additional mechanisms to explain
have the same working-memory constraints, experts greatly expanded working memory
the expert sees the world in larger and (Gobet & Simon, 1996; Richman, Gobet,
more diverse units. In effect, chunking, per- Staszewski, & Simon, 1996). At the same
studies of expertise from psychological perspectives 59
time there have been many arguments raised The second general criticism of the tra-
against the claims that the computational ditional theory of expertise comes from a
architecture remains fixed and thus presents rejection of the premise that expertise is
an invariant constraint on skilled and expert an extension of the processes observed in
processing. everyday skill acquisition (Fitts & Posner,
One of the most general criticisms is 1967). According to this model, the acqui-
that the laboratory produced empirical sition of skill proceeds in stages, and
evidence for capacity constraints of atten- during the first stage people acquire a
tion and STM are based on an operational cognitive representation of the task and
definition of chunks in terms of indepen- how to react in typical situations so they
dent pieces of information, no matter how can avoid gross errors. During the subse-
small or large the individual chunks (see quent stages, the performance of sequences
Cowan, 2001, and Ericsson & Kirk, 2001). of actions becomes smoother and more
It is relatively easy to design experimen- efficient. In the final stage, people are
tal materials for memory experiments that able to perform with a minimal amount
are made of independent pieces and mea- of effort, and performance runs essen-
sure experts and novices memory in terms tially automatically without active cogni-
of chunks. However, when one analyzes tive control. In an edited book on gen-
the information processed by experts when eral theories of expertise (Ericsson & Smith,
they perform representative tasks in their 1991a), several researchers raised concerns
domain of expertise, then all the heeded about explaining expertise as an extension
and relevant information has relations to the of this general model (Ericsson & Smith,
task and other pieces of information. If the 1991b; Holyoak, 1991; and Salthouse, 1991).
encountered information can be encoded Ericsson and Smith (1991b) found evidence
and integrated within a model of the cur- that experts maintain their ability to con-
rent context, then how many independent trol their performance and are able to give
chunks are stored or maintained in atten- detailed accounts of their thought processes
tion and working memory? Similarly, when that can be validated against other observ-
experts encounter representative tasks sit- able performance and process data. Ericsson
uations where beginners perceive several and Smith reviewed evidence that complex
independent tasks, the aspiring experts are cognitive representations mediate the per-
able to develop skills and encodings that formance and continued learning by experts,
allow them to integrate the different tasks which has been confirmed by subsequent
into a more general task with more diver- reviews (Ericsson, 1996b, 2003 , Chapter 13 ).
sified demands. More recent research has The third and final type of criticism comes
shown how in laboratory studies, partici- from the emerging evidence that extended
pants performing dual tasks that are believed focused practice has profound effects on,
to contain immutable bottlenecks of pro- and can influence virtually every aspect of,
cessing can, after training, perform them the human body, such as muscles, nerve
without any observable costs of the dual task systems, heart and circulatory system, and
(Meyer & Kieras, 1997; Schumacher et al., the brain. Several chapters in this hand-
2001, but see Proctor & Vu, Chapter 15 , for book review the structural changes result-
an alternative account). If the definition of ing from practice, such as Butterworth,
chunks and tasks requires independence for Chapter 3 2, on mathematical calculation;
imposing limits on information processing, Ericsson, Chapter 3 8; Lehmann and Gruber,
then it seems that the acquisition of exper- Chapter 26, on music performance; Proctor
tise entails developing integrated represen- and Vu, Chapter 15 , on adaptations in skill
tations of knowledge and coordination of acquisition; and Hill and Schneider, Chapter
initially separate tasks that make the funda- 3 7, with an overview of changes in the struc-
mental information-processing limits inap- ture and function of the brain with extended
plicable or substantially attenuated. practice and the development of expertise.
60 the cambridge handbook of expertise and expert performance
Simple Experience Is Not Sufficient there are many domains where performance
for the Development of Expertise decreases as a function of the number of
years since graduation from the training
Most everyday skills are relatively easy to institution (Ericsson, Chapter 3 8).
acquire, at least to an acceptable level. Several research methods have been
Adults often learn to drive a car, type, developed to describe the development
play chess, ski, and play (bad) golf within paths of expert performers, such as analy-
weeks or months. It is usually possible to sis of the historical record of eminent per-
explain what an individual needs to know formers (Simonton, Chapter 18), retrospec-
about a given skill, such as rules and proce- tive interviews (see Sosniak, Chapter 16),
dures, within a few hours (see also Hoffman and diary studies of practice (See Deakin,
& Lintern, Chapter 12). Once individuals Cote, & Harvey, Chapter 17). Research with
have learned the underlying structure of the these methods has shown that additional
activity and what aspects they must attend experience appears to make performance
to, they often focus on attaining a func- less effortful and less demanding, but to
tional level of performance. This is often improve performance it is necessary to seek
attained in less than 5 0 hours of practice. At out practice activities that allow individu-
this point, an acceptable standard of perfor- als to work on improving specific aspects,
mance can be generated without much need with the help of a teacher and in a protected
for more effortful attention and execution environment, with opportunities for reflec-
of the everyday activity has attained many tion, exploration of alternatives, and prob-
characteristics of automated performance lem solving, as well as repetition with infor-
(Anderson, 1982, 1987; Fitts & Posner, 1967; mative feedback.
Shiffrin & Schneider, 1977) and requires only In this handbook several chapters dis-
minimal effort. cuss the effectiveness of this type of delib-
In their seminal paper, Simon and Chase erate practice in attaining elite and expert
(1973 ) pointed to similarities between the levels of performance (Ericsson, Chap-
decade-long mastery of ones first language ter 3 8; Zimmerman, Chapter 3 9), in soft-
and the need for extended experience to ware design (Sonnentag, Niessen, & Volmer,
master complex domains of expertise, such Chapter 21), in training with simulators
as chess and sports. They made a strong (Ward, Williams, & Hancock , Chapter 14),
argument for a long period of immersion in maintaining performance in older experts
in active participation in activities in the (Krampe & Charness, Chapter 40), and in
domain, making the claim that even the best creative activities (Weisberg, Chapter 42).
chess players needed to spend over ten years Other chapters review evidence on the rela-
studying chess before winning at the inter- tionship between deliberate practice and
national level. The necessity for even the the development of expertise in particu-
most talented performers to spend ten years lar domains, such as professional writing
working and practicing was later converted (Kellogg, Chapter 22), music performance
into an equivalence, namely, that ten years of (Lehmann & Gruber, Chapter 26), sports
experience in a domain made somebody an (Hodges, Starkesi & MacMohan, Chap-
expert. However, for chess, tennis, and golf, ter 27), chess (Gobet & Charness, Chap-
everyone knows examples of excited recre- ter 3 0), exceptional memory (Wilding &
ational players who regularly engage in play Valentine, Chapter 3 1), and mathematical
for years and decades, but who never reach calculation (Butterworth, Chapter 3 2).
a very skilled level.
Reviews of the relation between the
amount of experience and the attained
level of performance show consistently that Concluding Remarks
once an acceptable level is attained, there
are hardly any benefits from the com- The theoretical interest in expertise
mon kind of additional experience. In fact, and expert performance is based on the
studies of expertise from psychological perspectives 61
assumption that there are shared psycho- tunities for repetition and gradual refine-
logical constraints on the structure and ment.
acquisition of expert performance across The discovery of the complex structure of
different domains. The theory of Simon and the mechanisms that execute expert perfor-
Chase (1973 ) proposed that the invariant mance and mediate its continued improve-
limits on information processing and STM ment has had positive and negative impli-
severely constrained how expert skill is cations. On the negative side, it has pretty
acquired and proposed a theory based on much dispelled the hope that expert per-
the accumulation through experience of formance can easily be captured and that
increasingly complex chunks and pattern- the decade-long training to become an
action associations. This theory emphasized expert can be dramatically reduced. All
the acquired nature of expertise and focused the paths to expert performance appear to
on the long time required to reach elite require substantial extended effortful prac-
levels and the learning processes sufficient tice. Effortless mastery of expertise, magical
to gradually accumulate the large body of bullets involving training machines, and dra-
requisite patterns and knowledge. This view matic shortcuts, are just myths. They cannot
of expertise offered the hope that it would explain the acquisition of the mechanisms
be possible to extract the accumulated and adaptations that mediate skilled and
knowledge and rules of experts and then expert performance. Even more important,
use this knowledge to more efficiently train the insufficiency of the traditional school
future experts and, thus, reduce the decade system is becoming apparent. It is not rea-
or more of experience and training required sonable to teach students knowledge and
for elite performance. Efforts were made rules about a domain, such as programming,
even to encode the extracted knowledge in medicine, and economics, and then expect
computer models and to build expert sys- them to be able to convert this material
tems that could duplicate the performance into effective professional skills by additional
of the experts (Bachanan et al., Chapter 6). experience in the pertinent domain. Schools
Subsequent research on extended train- need to help students acquire the skills and
ing revealed that it is possible to acquire skills mechanisms for basic mastery in the domain,
that effectively alter or, at least, circumvent and then allow them gradually to take over
the processing limits of attention and work- control of the learning of their professional
ing memory. Studies of expertise focused skills by designing deliberate practice activi-
initially on the experts representation and ties that produce continued improvement.
memory for knowledge. As research started On the positive side, the discovery of
to examine and model experts superior per- effective training methods for acquiring
formance on representative tasks, it became complex cognitive mechanisms has allowed
clear that their complex representations investigators to propose types of training
and mechanisms that mediate performance that appear to allow individuals to acquire
could not be acquired by mere experience levels of performance that were previously
(Ericsson, Chapter 3 8). Research on what thought to be unobtainable, except for the
enabled some individuals to reach expert elite group of innately talented. The study of
performance, rather than mediocre achieve- the development of expert performers pro-
ment, revealed that expert and elite per- vides observable paths for how they modi-
formers seek out teachers and engage in spe- fied or circumvented different types of psy-
cially designed training activities (deliberate chological and physiological constraints. It
practice). The future expert performers should be possible for one type of expert in
need to acquire representations and mech- one domain, such as surgery, to learn from
anisms that allow them to monitor, control, how other experts in music or sports, for
and evaluate their own performance, so they instance, have designed successful training
can gradually modify their own mechanisms procedures for mastering various aspects of
while engaging in training tasks that provide perceptual-motor procedures, and to learn
feedback on performance, as well as oppor- the amount of practice needed to reach
62 the cambridge handbook of expertise and expert performance
specified levels of mastery. If someone is Anderson, J. R. (Ed.) (1981). Cognitive skills and
interested, for instance, in whether a certain their acquisition. Hillsdale, NJ: Erlbaum.
type of perceptual discrimination can ever Anderson, J. R. (1982). Acquisition of cognitive
be made reliably, and how much and what skill. Psychological Review, 89, 3 69406.
type of training would be required to achieve Anderson, J. R. (1987). Skill acquisition: Compi-
this, then one should in the future be able to lation of weak-method problem situations. Psy-
turn to a body of knowledge of documented chological Review, 94, 192210.
expert performance. Our vision is that the Austin, J. (2000). Performance analysis and per-
study of expert performance will become a formance diagnostics. In J. Austin & J. E. Carr
science of learning and of the human adap- (Eds.), Handbook of applied behavior analysis
tations that are possible in response to spe- (pp. 3 213 49). Reno, NV: Context Press.
cialized extended training. At the same time Baddeley, A. (2000). Short-term and working
that our understanding of the real constraints memory. In E. Tulving & F. Craik (Eds.),
on acquiring high levels of performance in The Oxford handbook of memory (pp. 7792).
Oxford, UK: Oxford University Press.
any domain becomes clearer, and the simi-
larities of those constraints across many dif- Baddeley, A. (2002). Is working memory
ferent domains are identified, the study of still working? European Psychologist, 7(2 ),
85 97.
the acquisition of expert performance will
offer a microcosm for how various types of Barrows, H. S., Feightner, J. W., Neufeld, V. R.,
& Norman, G. R. (1978). Analysis of the clinical
training can improve human performance
methods of medical students and physicians. Final
and provide insights into the potential for Report, Ontario Department of Health Grants
human achievement. ODH-PR-273 & ODH-DM-226. Hamilton,
The study of expert performance is not Ontario: McMaster University.
concerned only with the ultimate limits of Barrows, H. S., & Tamblyn, R. M. (1980).
performance, but also with earlier stages of Problem-based learning: An approach to medical
development through which every future education. New York: Springer.
performer needs to pass. There is now Barsalou, L. W. (1999). Perceptual symbol sys-
research emerging on how future expert per- tems. Behavioral and Brain Sciences, 22, 5 77
formers will acquire initial and intermedi- 660.
ate levels of performance. Attaining these Barsalou, L. W., Simmons, W. K., Barbey, A. K.,
intermediate levels may be an appropriate & Wilson, C. D. (2003 ). Grounding concep-
goal for people in general and for systems of tual knowledge in modality-specific systems.
general education (e.g., recreational athletes, TRENDS in Cognitive Sciences, 7(2), 8491.
patrons of the arts). However, knowing how Bartlett, F. (193 2). Remembering. Cambridge, UK:
to achieve certain goals is no guarantee that Cambridge University Press.
people will be successful, as we know from Bartlett, F. (195 8). Thinking. New York: Basic
studies of dieting and exercise. On the other Books.
hand, when the goal is truly elite achieve- Berlyne, D. E. (1965 ). Structure and direction in
ment, the study of expert performance offers thinking. Oxford, UK: Wiley.
a unique source of data that is likely to help Bloom, B. (Ed.) (1985 ). Developing talent in young
us understand the necessary factors for suc- people. New York: Ballentine.
cess, including the social and motivational Brown, A. (1991). A review of the tip-of-the-
factors that push and pull people to per- tongue experience. Psychological Bulletin, 109,
severe in the requisite daunting regimes of 204223 .
training. Brown, R. W., & MacNeill, D. (1966). The tip
of the tongue phenomenon. Journal of Verbal
Learning and Verbal Behavior, 5 , 3 25 3 3 7.
References Bruner, J. S., Goodnow, J. J., & Austin, G. A.
(195 6). A study of thinking. New York: John
Adelson, B. (1981). Problem solving and the devel- Wiley and Sons.
opment of abstract categories in programming Buchanan, B. G., & Feigenbaum, E. A. (1978).
languages. Memory and Cognition, 9, 42243 3 . DENDRAL and MetaDENDRAL: Their
studies of expertise from psychological perspectives 63
applications dimension. Artificial Intelligence, task: A logical sequel to Miller (195 6). Psycho-
11(1), 5 24. logical Science, 15 (9), 63 4640.
Charness, N. (1976). Memory for chess positions: de Groot, A. (1946). Het denken van den schaker.
Resistance to interference. Journal of Experi- Amsterdam: Noord-Hllandsche Uit. Mij.
mental Psychology: Human Learning and Mem- de Groot, A. (1965 ). Thought and choice in chess,
ory, 2 , 64165 3 . 1st Edition. The Hague: Mouton.
Charness, N. (1979). Components of skill in Duncker, K. (1945 ). On problem solving. Psycho-
bridge. Canadian Journal of Psychology, 3 3 , 1 logical Monographs, 5 8 (Whole No. 270), 1113 .
5 0. Ebbinghaus, H. (1885 /1964). Memory: A contribu-
Charness, N. (1981). Search in chess: Age and skill tion to experimental cognitive psychology (Henry
differences. Journal of Experimental Psychology: A. Ruger & Clara E. Bussenius, Trans.). New
Human Perception and Performance, 7 , 467 York: Dover.
476. Egan, D. E., & Schwartz, B. J. (1979). Chunking in
Chase, W. G. (Ed.) (1973 ). Visual information pro- recall of symbolic drawings. Memory and Cog-
cessing. New York: Academic Press. nition, 7 , 14915 8.
Chase, W. G., & Simon, H. A. (1973 a). The minds Eisenstadt, M., & Kareev, Y. (1979). Aspects of
eye in chess. In W. G. Chase (Ed.), Visual human problem solving: The use of internal
information processing. New York: Academic representations. In D. A. Norman & D. E.
Press. Rumelhart (Eds.), Exploration in cognition. San
Chase, W. G., & Simon, H. A. (1973 b). Perception Francisco: W. H. Freeman.
in chess. Cognitive Psychology, 1, 3 3 81. Elstein, A. S., Shulman, L. S., & Sprafka, S. A.
Chi, M. T. H. (1978). Knowledge structures and (1978). Medical problem solving. Cambridge,
memory development. In R. S. Siegler (Ed), MA: Harvard University Press.
Childrens thinking: What develops? (pp. 73 Engle R. W., & Bukstel, L. (1978). Memory pro-
96). Hillsdale, NJ: Erlbaum. cesses among bridge players of differing exper-
Chi, M. T. H., Feltovich, P., & Glaser, R. (1981). tise. American Journal of Psychology, 91, 673 89.
Categorization and representation of physics Ericsson, K. A. (Ed.) (1996a). The road to
problems by experts and novices. Cognitive Sci- excellence: The acquisition of expert performance
ence, 5 , 12115 2. in the arts and sciences, sports, and games.
Chi, M. T. H., Glaser, R., & Farr, M. J. (Eds.) Mahwah, NJ: Erlbaum.
(1988). The nature of expertise. Hillsdale, NJ: Ericsson, K. A. (1996b). The acquisition of expert
Erlbaum. performance: An introduction to some of the
Chiesi, H. L., Spilich, G. J., & Voss. J. F. (1979). issues. In K. A. Ericsson (Ed.), The road to
Acquisition of domain-related information in excellence: The acquisition of expert performance
relation to high and low domain knowledge. in the arts and sciences, sports, and games (pp. 1
Journal of Verbal Learning and Verbal Behavior, 5 0). Mahwah, NJ: Erlbaum.
18, 25 7274. Ericsson, K. A. (2003 ). The acquisition of expert
Chomsky, N. (195 7). Syntactic structures. The performance as problem solving: Construction
Hague: Mouton. and modification of mediating mechanisms
Clancey, W. J. & Letsinger, R. (1984). through deliberate practice. In J. E. Davidson &
NEOMYCIN: Reconfigure a rule-based expert R. J. Sternberg (Eds.), Problem solving (pp. 3 1
system for application to teaching. In W. J. 83 ). New York: Cambridge University Press.
Clancey & E. H. Shortliffe, (eds.), Readings Ericsson, K. A., & Kintsch, W. (1995 ). Long-
in medical artificial intelligence (pp. 3 613 81). term working memory. Psychological Review,
Readings Addision. 102 , 211245 .
Clancey, W. J., & Shortliffe, E. H. (1984). Read- Ericsson, K. A., & Kintsch, W. (2000). Shortcom-
ings in medical artificial intelligence: The first ings of generic retrieval structures with slots of
decade. Reading, MA: Addison-Wesley. the type that Gobet (1993 ) proposed and mod-
Cowan, N. (2001). The magical number 4 in eled. British Journal of Psychology, 91, 5 715 88.
short-term memory: A reconsideration of men- Ericsson, K. A., & Kirk, E. P. (2001). The search
tal storage capacity. Behavioral and Brain Sci- for fixed generalizable limits of pure STM
ences, 2 4, 87185 . capacity: Problems with theoretical proposals
Cowan, N., Chen, Z., & Rouder, J. (2004). Con- based on independent chunks. Behavioral and
stant capacity in an immediate serial-recall Brain Sciences, 2 4, 120121.
64 the cambridge handbook of expertise and expert performance
Ericsson, K. A., Krampe, R. T., & Tesch-Romer, Gentner, D. (1988). Metaphor as structure-
C. (1993 ). The role of deliberate practice in the mapping: The relational shift. Child Develop-
acquisition of expert performance. Psychologi- ment, 5 9, 475 9.
cal Review, 100, 3 63 406. Glaser, R. (1976). Cognition and instructional
Ericsson, K. A., & Lehmann, A. C. (1996). design. In D. Klahr (Ed.), Cognition and instruc-
Expert and exceptional performance: Evidence tion (pp. 3 03 3 15 ). Hillsdale, NJ: Erlbaum.
on maximal adaptations on task constraints. Glaser, R., & Chi, M. T. H. (1988). Overview. In
Annual Review of Psychology, 47 , 273 3 05 . M. T. H., Chi, R., Glaser, & M. J., Farr (Eds.),
Ericsson, K. A., Patel, V. L., & Kintsch, W. (2000). The nature of expertise (pp. xv-xxviii). Hillsdale,
How experts adaptations to representative NJ: Erlbaum.
task demands account for the expertise effect Gobet, F., & Simon, H. A. (1996). Templates in
in memory recall: Comment on Vicente and chess memory: A mechanism for recalling sev-
Wang (1998). Psychological Review, 107 , 5 78 eral boards. Cognitive Psychology, 3 1, 140.
5 92. Green, C. (1969). The application of theorem
Ericsson, K. A., & Smith, J. (Eds.) (1991a). Toward proving to question-answering systems. Doc-
a general theory of expertise: Prospects and limits. toral dissertation Dept. of Electrical Engineer-
Cambridge, MA: Cambridge University Press. ing, Stanford University. Also Standford Arti-
Ericsson, K. A., & Smith, J. (1991b). Prospects ficial Intelligence Project Memo AI-96, June
and limits in the empirical study of expertise: 1969.
An introduction. In K. A. Ericsson and J. Smith Groen, G. J., & Patel, V. L. (1988). The rela-
(Eds.), Toward a general theory of expertise: tionship between comprehension and reason-
Prospects and limits (pp. 13 8). Cambridge, ing in medical expertise. In M. T. H. Chi, R.
MA: Cambridge University Press. Glaser, & M. J. Farr (Eds.), The nature of exper-
Ernst, G., & Newell, A. (1969). GPS: A case-study tise (pp. 2873 10). Hillsdale, NJ: Erlbaum.
in generality and problem solving. New York: Hambrick, D., & Engle, R. (2002). Effects of
Academic Press. domain knowledge, working memory capacity,
Feltovich, P., Ford, K. M., & Hoffman, R. R. and age on cognitive performance: An investi-
(Eds.) (1997). Expertise in context: Human and gation of the knowledge-is-power hypothesis.
Machine. Menlo Park, CA: AAAI Press. Cognitive Psychology, 44, 3 3 93 87.
Feltovich, P. J., Johnson, P. E, Moller, J., & Swan- Hatano, G., Miyake, Y., & Binks, M. B. (1977).
son, D. (1984). The role and development Performance of expert abacus operators. Cog-
of medical knowledge in diagnostic reasoning. nition, 5 , 5 771.
In. W. Clancey & E. Shortliffe (Eds.), Read- Hinsley, D., Hayes, J., & Simon, H. A.(1978).
ings in medical artificial intelligence: The first From words to equations: Meaning and rep-
decade (pp. 275 3 19). Reading, MA: Addison resentation in algebra word problems. In M.
Wesley. Just & P. Carpenter (Eds.), Cognitive processes
Feltovich, P., Spiro, R., & Coulson, R. (1997). in comprehension. Hillsdale, NJ: Erlbaum.
Issues of expert flexibility in contexts charac- Hoffman, R. R. (Ed.) (1992). The psychology of
terized by complexity and change. In P. Fel- expertise. New York: Springer-Verlag.
tovich, K. Ford, & R. Hoffman (Eds.), Expertise Holyoak, K. J. (1991). Symbolic connectionism:
in context: Human and machine. Menlo Park, toward third generation theories of expertise.
CA: AAAI Press. In K. A. Ericsson & J. Smith (Eds.), Toward
Ferster, C. B., & Skinner, B. F. (195 7). Schedules a general theory of expertise: Prospects and lim-
of reinforcement. New York: Appleton-Century- its (pp. 3 013 3 5 ). Cambridge, MA: Cambridge
Crofts. University Press.
Fitts, P. M., & Posner, M. I. (1967). Human per- Jakobovits, L. A. & Miron, M. S. (1967). Readings
formance. Belmont, CA: Brookes Cole. in the Psychology of language. Englewood, cliffs,
Flavell, J. (1979). Metacognition and cogni- NJ: Prentice-Hall.
tion monitoring: A new area of cognitive- Jansma, J., Ramsey, N., Slagter, H., & Kahn,
developmental inquiry. American Psychologist, R. (2001). Functional anatomical correlates of
3 4, 906911. controlled and automatic processing. Journal of
Forbus, K. D. & Feltovich, P. J. (2001), Smart Cognitive Neuroscience, 13 (6), 73 0743 .
machines in education. Menlo PK, CA: AAAI/ Jeffries, R., Turner, A. A., Polson, P. G, & Atwood,
MIT Press. M. E. (1981). The processes involved in
studies of expertise from psychological perspectives 65
software design. In J. R. Anderson (Ed.), Cog- Lumsdaine, A. A., & Glaser, R. (1960). Teach-
nitive skills and their acquisition. Hillsdale, NJ: ing machines and programmed learning: A source
Erlbaum. book. Washington, DC: National Education
Johnson, P. E., Duran, A. S., Hassebrock, F., Assoc.
Moller, J., Prietula, M. J., Feltovich, P. J., & McKeithen, K. B., Reitman, J. S., Reuter, H. H.,
Swanson, D. B. (1981). Expertise and error in & Hirtle, S. C., (1981). Knowledge organization
diagnostic reasoning. Cognitive Science, 5 , 23 5 and skill differences in computer programmers.
283 . Cognitive Psychology, 13 , 3 073 25 .
Klein, G. A. (1998). Sources of power: How Meehl, P. E. (195 4). Clinical versus statistical
people make decisions. Cambridge, MA: MIT prediction: A theoretical analysis and a review
Press. of the evidence. Minneapolis: University of Min-
Klein, G. A., & Hoffman, R. R. (1993 ). Seeing nesota Press.
the invisible: Perceptual-cognitive aspects of Meyer, D. E., & Kieras, D. E. (1997). A com-
expertise. In M. Rabinowitz (Ed.), Cognitive putational theory of executive cognitive pro-
science foundations of instruction. Hillsdale, NJ: cesses and multiple-task performance: Part 2.
Erlbaum. Accounts of psychological refractory-period
Koschmann, T. D., LeBaron, C., Goodwin, C., phenomena. Psychological Review, 104, 749
& Feltovich, P. J. (2001). Dissecting common 791.
ground: Examining an instance of reference Miller, G. A. (195 6). The magical number seven
repair. In Proceedings of the 2 3 rd Conference plus or minus two: Some limits on our capac-
of the Cognitive Science Society. Mahwah, NJ: ity for processing information. Psychological
Erlbaum. Review, 63 , 8197.
Kruger, J., & Dunning, D. (1999). Unskilled and Miller, G. A., Galanter, E., & Pribram, K. H.
unaware of it: How difficulties in recognizing (1960). Plans and the structure of behavior. New
ones own incompetence lead to inflated self- York: Holt, Rinehart & Winston.
assessments. Journal of Personality and Social Neisser, U. (1967). Cognitive psychology. New
Psychology, 77, 1121113 4. York: Appleton-Century-Crofts.
Kuiper, R., & Pesut, D. (2004). Promoting cog- Newell, A. (1973 ). Artificial intelligence and the
nitive and metacognitive reflective reasoning concept of mind. In R. C. Schank & K. M.
skills in nursing practice: Self-regularities learn- Colby (Eds.), Computer models of language
ing theory. Journal of Advanced Nursing, 45 , and thought (pp. 160). San Franciso: W. H.
3 813 91. Freeman.
Larkin, J., McDermott, J., Simon, D. & Simon, Newell, A., & Simon, H. A. (195 6). The logic
H. A. (1980). Expert and novice performance theory machine: A complex information pro-
in solving physics problems. Science, 2 08, 13 3 5 cessing system. IRE Transactions on Information
13 42. Theory, Vol IT-2, No. 3 , 6179.
Lamme, V. (2003 ). Why visual attention and Newell, A., & Simon, H. A. (1972). Human
awareness are different. Trends in Cognitive Sci- problem solving. Englewood Cliffs, NJ: Prentice
ences, 7 (1), 1218. Hall.
Lesgold, A. M., & Resnick, L. (1982). How read- Newell, A., & Simon, H. A. (1976). Com-
ing difficulties devlop: Perspectives from a lon- puter science as empirical inquiry: Symbols
gitudinal study. In J. Das, R. Mulcahey, & A. and search. Communications of the ACM, 19(3 ),
wall (Eds.), Theory and research in learning 113 126.
disabilities (pp. 15 5 187). New York: Plenum Norman, D. A., Rumelhart, D. E., & the LNR
Press. group (1979). Explorations in cognition. San
Lesgold, A. M., Rubinson, H., Feltovich, P. J., Francisco: W. H. Freeman.
Glaser, R., Klopfer, D., & Wang, Y. (1988). Osgood, C. E. (1963 ). On understanding and cre-
Expertise in a complex skill: Diagnosing x-ray ating sentences. American Psychologist, 18, 73 5
pictures. In M. T. H. Chi, R. Glaser, & M. J. Farr 75 1.
(Eds.), The nature of expertise (pp. 3 113 42). Paris, S., & Winograd, P. (1990). How metacog-
Hillsdale, NJ: Erlbaum. nition can promote academic learning and
Logan, G. (1985 ). Skill and automaticity: Rela- instruction. In B. Jones & L. Idol (Eds.), Dimen-
tions, implications and future directions. Psy- sions of thinking and cognitive instruction (pp. 15
chological Review, 95 , 4925 27. 5 1). Hillsdale, NJ: Erlbaum.
66 the cambridge handbook of expertise and expert performance
Patel, V. L., Arocha, J. F., & Kaufman, D. R. Prietula, M., & Simon, H. (1989). The experts in
(1994). Diagnostic reasoning and medical your midst. Harvard Business Review, JanFeb,
expertise. In D. Medin (Ed.), The psychology of 120124.
learning and motivation. Vol. 3 0 (pp. 18725 1). Quillian, R. (1969). The teachable language
New York: Academic Press. comprehender: A Simulation program and the-
Patel, V. L., & Groen, G. J. (1991). The general ory of language. Communications of the ACM,
and specific nature of medical expertise: A crit- 12 45 9476.
ical look. In K. A. Ericsson & J. Smith (Eds.), Reder, L., & Shunn, C. (1996). Metacognition
Toward a general theory of expertise (pp. 93 does not imply awareness: Strategy choice is
125 ). Cambridge, MA: Cambridge University governed by implicit learning and memory. In
Press. L. Reder (Ed.), Implicit memory and metacogni-
Patil, R. S., Szolovitz, P., & Schwartz, W. B. (1981). tion. Mahwah, NJ: Erlbaum.
Causal understanding of patient illness in med- Reitman, J. (1976). Skilled perception in GO:
ical diagnosis. In Proceedings of the seventh inter- Deducing memory structure from inter-
national conference on artificial intelligence. Vol. response times. Cognitive Psychology, 8, 3 3 6
2 (pp. 893 899). Los Altos, CA: William Kauf- 3 5 6.
man.
Reitman, W. R. (1965 ). Cognition and thought.
Pauker, S. G., Gorry, G. A., Kassirer, J. P., & New York: Wiley.
Schwartz, W. B. (1976). Towards simulation of
Richman, H. B., Gobet, F., Staszewski, J. J., &
clinical cognition: Taking a present illness by
Simon, H. A. (1996). Perceptual and mem-
computer. American Journal of Medicine, 60,
ory processes in the acquisition of expert
981996.
performance: The EPAM model. In K. A. Eric-
Pellegrino, J. W., & Glaser, R. (1982a). Improving sson (Ed.), The road to excellence: The acquisi-
the skills of learning. In D. K. Detterman & tion of expert performance in the arts and sciences,
R. J. Sternberg (Eds.), How much and how can sports, and games (pp. 167187). Mahwah, NJ:
intelligence be increased. Norwood, NJ: Ablex. Erlbaum.
Pellegrino, J. W., & Glaser, R. (1982b). Analyzing Rieger, M. (2004). Automatic keypress activa-
aptitudes for learning: Inductive reasoning. In tion in skilled typists. Journal of Experimental
R. Glaser (Ed.), Advances in instructional psy- Psychology: Human Perception and Performance,
chology. Vol. 2 (pp. 2693 45 ). Hillsdale, NJ: 3 , 5 5 5 5 65 .
Erlbaum.
Rumelhart, D. E. (1979). Analogical processes
Perfetti, C. A., & Lesgold, A. M. (1979). Coding and procedunal representations. CHIP Report #
and comprehension in skilled reading. In L. B. 81. Center for Human Information Processing,
Resnick & P. Weaver (Eds.), Theory and practice Univ. California, San Diego: February, 1979.
of early reading. Hillsdale, NJ: Erlbaum.
Sacerdoti, E. D. (1977). A structure for plans and
Phelps, R. H., & Shanteau J. (1978). Livestock behavior. New York: Elsevier-North Holland
judges: How much information can an expert Publishing.
use? Organizational Behavior and Human Per- Salthouse, T. A. (1991). Expertise as the circum-
formance, 2 1, 209219. vention of human processing limitations. In
Petrusa, E. R. (2002). Clinical performance K. A. Ericsson and J. Smith (Eds.), Toward a
assessments. In G. R. Norman, C. P. M. van general theory of expertise: Prospects and limits
der Vleuten, & D. I. Neuble (Eds.), Interna- (pp. 2863 00). Cambridge, MA: Cambridge
tional handbook of research in medical education. University Press.
Dordrecht: Kluwer Academic Publishers. Samuel, A. L. (195 9). Some studies in machine
Posner, M., & Snyder, C. (1975 ). Attention and learning using the game of checkers. IBM
cognitive control. In R. L. Solso (Ed.), Informa- Journal of Research and Development, 3 , 210
tion processing and cognition: The Loyola sympo- 229.
sium. Hillsdale, NJ: Erlbaum. Schneider, W., & Fisk, A. (1982). Concurrent
Prietula, M., & Augier, M. (2005 ). Adventures automatic and controlled visual search: Can
in software archeology: Seeking (ABTOF) the- processing occur without resource cost? Jour-
ory in the code. Proceedings of the 2 005 Annual nal of Experimental Psychology, 8, 261278.
Academy of Management Meeting, August 5 10, Schneider, W., & Shiffrin, R. M. (1977). Con-
Honolulu HI. trolled and automatic human information
studies of expertise from psychological perspectives 67
processing: 1. Detection, search and attention. Starkes, J., & Ericsson, K. A. (Eds.) (2003 ).
Psychological Review, 84, 166. Expert performance in sport: Recent advances
Schmidt, J. A., McLaughlin, J. P., & Leighton, in research on sport expertise. Champaign, IL:
P. (1989). Novice strategies for understanding Human Kinetics.
paintings. Applied Cognitive Psychology, 3 , 65 Steier, D., & Mitchell, T. (Eds) (1996). Knowledge
72. matters: A tribute to Allen Newell. Mahwah, NJ:
Schumacher, E. H., Seymour, T. L., Glass, J. M., Erlbaum.
Fencsik, D. E., Lauber, E. J., Kieras, D. E., & Sternberg, R. (1984). Toward a triarchic theory of
Meyer, D. E. (2001). Virtually perfect time human intelligence. Behavioral and Brain Sci-
sharing in dual-task performance: Uncorking ences, 7, 269287.
the central cognitive bottleneck. Psychological Tulving, E. (1994). Forward. In J. Metcalfe &
Science, 12 , 101108. A. Shimamura (Eds.), Metacognition: Knowing
Shaffer, L. H. (1975 ). Multiple attention in con- about knowing. Cambridge, MA: MIT Press.
tinuous verbal tasks. In P. M. A. Rabbitt & S. Van Gelder, T., & Port, R. (1995 ). Its about time:
Dornic (Eds.) Attention and performance V. New An overview of the dynamical approach to cog-
York: Academic Press. nition. In T. van Gelder & R. Port (Eds.), Mind
Shiffrin, R. M., & Schneider, W. (1977). Con- as motion (pp 143 ). Cambridge, MA: MIT
trolled and automatic human information Press.
processing: II. Perceptual learning, automatic VanLehn, K., & Brown, J. S. (1979). Planning nets:
attending and a general theory. Psychological A representation for formalizing analogies and
Review, 84, 127189. semantic models of procedural skills. In R. E.
Shipp, S. (2004). The brain circuitry of attention. Snow & W. E. Montague (Eds.), Aptitude, learn-
Trends in Cognitive Sciences. 8, 223 23 0. ing and instruction: Cognitive process analyses.
Shortliffe, E. H. (1976). Computer-based medi- Hillsdale, NJ: Erlbaum.
cal consultations: MYCIN. New York: Ameri- Voss, J. F., Greene, T. R., Post, T. A., & Penner,
can Elsevier. B. C. (1983 ). Problem solving skill in the social
Simon, H. A., & Chase, W. G. (1973 ). Skill in sciences. In G. H. Bower (Ed.), The psychology
chess. American Scientist, 61, 3 94403 . of learning and motivation: Advances in research
Simon, H. A., & Newell, A. (195 8). Heuristic theory. Vol. 17 (pp. 165 213 ). New York: Aca-
problem solving: The next advance in opera- demic Press.
tions research. Operations Research, 6, 110. Voss, J. F., Tyler, S., & Yengo, L. (1983 ). Individ-
Skinner, B. F. (1960). The science of learning and ual differences in the solving of social science
the art of teaching. In A. A. Lumsdaine & problems. In R. Dillon & R. Schmeck (Eds.),
R. Glaser (Eds.), Teaching machines and pro- Individual differences in cognition (pp. 205 23 2).
grammed learning: A source book (pp. 99113 ). New York: Academic Press.
Washington, DC: National Education Assoc. Wason, P. M., & Johnson-Laird, P. N. (1972). Psy-
Sleeman, D., & Brown, J. S. (1982). Intelligent chology of reasoning: Structure and content. Cam-
tutoring systems. New York: Academic Press. bridge, MA: Harvard University Press.
Spelke, E., Hirst, W., & Neisser, U. (1976). Skills Watson, J. B. (1913 ). Psychology as the behaviorist
of divided attention. Cognition, 4, 215 3 0. views it. Psychological Review, 2 0, 15 877.
Spilich, G. J., Vesonder, G. T., Chiesi, H. L., & Watson, J. B. (1920). Is thinking merely the action
Voss, J. F. (1979). Text processing of domain- of language mechanisms? British Journal of Psy-
related information for individuals with high chology, 11, 87104.
and low domain knowledge. Journal of Verbal Winograd, T. (1975 ). Frame representations and
Learning and Verbal Behavior, 14, 5 065 22. the declarative procedural controversy. In D. G.
Spiro, R. J., Collins, B. P. Thota, J. J., & Fel- Bobrow & A. M. Collins (Eds.), Representation
tovich, P. J. (2003 ). Cognitive flexibility theory: and understanding. New York: Academic Press.
Hypermedia for complex learning, adaptive Zeitz, C. M. (1997). Some concrete advantages
knowledge application, and experience accel- of abstraction: How experts representations
eration. Educational technology, 44, 5 10. facilitate reasoning. In P. J. Feltovich, K. M.
Starkes, J. L., & Allard, F. (Eds.) (1993 ). Cogni- Ford, & R. R. Hoffman (Eds.), Expertise in
tive issues in motor expertise. Amsterdam: North context: Human and machine (pp. 43 65 ).
Holland. Menlo PK, CA: AAAI/MIT Press.
CHAPTER 5
two and a half millennia from Socrates 3 . The changing role of the instructor, jux-
(ca. 400 BC) to Gagne (ca. 1970 AD) is taposing at various points in history
an increasingly constrained view toward the between subject matter expert and expert in
study and development of expertise. The earli- educational techniques (reflecting current
est recorded educators, including Plato and views on how to best achieve learning in
Socrates, often viewed expertise in what can students), and
be described as a whole man approach, a 4. A shift in skills assessment from informal
holistic view that included aspects of knowl- and oral assessment in the ancient context
edge, skills, and morality to achieve virtue to formal, objective, and measurable assess-
in the learner. Medieval European educa- ment in the Twentieth century (reflecting
tors, describing new educational programs the increasing desire to objectively mea-
such as the trivium and the quadrivium, and sure expertise).
implementing those paradigms within novel
institutions such as the cathedral schools These trends, all of which can be seen in
and the university, constrained the focus seed form in the ancient context, laid the
of skills development to more specialized foundation for, and sometimes the bound-
areas, including geometry and the Latin aries circumscribing, later attempts to study
language, resulting in greater codification the nature and development of highly skilled
of educational systems and their attendant individuals.
instructional techniques. In the most recent We commence our review by looking first
period, twentieth-century educational psy- at the ancient views of skill building and
chologists, working in scientifically based expertise. We then move on to examine
learning paradigms, further constrained the the evolution of these views through the
focus of skills development and expertise, Early and High Middle Ages. We then exam-
specifying detailed and empirically based ine some of the modern salient influential
models for the acquisition of the highest theories of learning and skills building that
levels of skill within highly specific domain affect theories of expertise, culminating with
areas (e.g., concert violin performance, pro- the most recent attempt to quantify and
fessional golfing, and tournament-level chess objectively measure skills in specific domain
competition). This trend, broad and imper- areas, Ericssons expert performance model
fect as it may be, will nevertheless serve (Ericsson, 1996; Chapter 3 8).
nicely to trace the general outlines of our
history.
It is beneficial at the outset of our review The Ancient Context
to make note of some key historical trends
that will be presented in this chapter and Education as a discipline has never suffered
that have impacted educators views of a shortage of divergent views, and it comes
expertise throughout the centuries. Among as little surprise that we immediately wit-
these, we will see ness in the ancient period an early demarca-
tion between two positions on its purpose:
1. The progression from largely individual- one that focused on the holistic develop-
ized instruction in the ancient context to ment of the individual, and one that focused
mass education in later periods (finding on applied skills building. These two early
culmination in the mass production edu- philosophies of education played a direct
cational techniques of the nineteenth and role in the manner in which expertise was
twentieth centuries), defined and measured. Regardless of the
2. The progression of a model of education position, however, the assumption was that
for the few in ancient times to education the instructor should be an expert in the area
for the many after the Industrial Revolu- in which he taught. This placed the teacher
tion (a function of the decreasing cost of at the focal point of all education, with
educating a learner via mass production students building expertise via transmission
techniques), from the expert, the instructor himself.
educators and expertise: theories and models 71
Socrates, Plato, and the Sophists an early belief in the presence of natural
ability based on some form of genetic
Socrates (4693 99 BC), one of historys ear- endowment, a prototypical concept fore-
liest educators, was born in Athens of a shadowing all the way to Sir Francis Galtons
stonemason, but grew to become one of nineteenth-century attempts to measure
the most influential educators of his time. a generalized, inheritable intelligent quo-
His student, Plato (4283 47 BC), was the tient, g (Galton, in Ericsson, 2004; Horn &
recorder of Socrates words and shared many Masunaga, Chapter 3 4).
of Socrates philosophical positions. Much of Socrates and Plato did not seem strongly
what we know about Socrates spoken words concerned with the development of applied
comes from the Platonic writings. Plato has skills and actually seemed to demonstrate
been often cited as producing the most long- an aversion to practical skills training when
lived and influential views impacting west- devoid of what they viewed as the deeper
ern education, and his beliefs are still refer- meanings of education (Johnson, 1998). Fur-
enced and debated today. ther, neither viewed educations primary
Socrates did not promote a formalized role as the transmission of information to
educational system consisting of schools that students: education was viewed as inherently
delivered and assessed learning outcomes; valuable as an intellectual exploration of the
rather, he viewed education as a process soul. This position therefore provided a defi-
of developing the inner state of individu- nition of expertise as a general set of inner eth-
als through dialogue and conversation (Jeffs, ical and knowledge-based traits that was infor-
2003 ). Now referred to as the Socratic mally and orally assessed by the instructor.
method, the teacher employing this method In notable opposition to the whole man
would not transmit knowledge or practical educational approach advocated by Plato
skills (techne), but would engage the stu- and Socrates were the Sophists. The name
dent in a dialogical process that brought Sophist itself implies the orientation: the
out knowledge believed to be innate within Greek word sophia denotes skill at an applied
the student (Gardner, 1987). Instruction in craft (Johnson, 1998), and Sophist educa-
the Socratic context, therefore, was con- tors focused on the development of spe-
ducted by means of interactive questioning cific applied skills for individuals studying to
and dialog, without concern for fixed learn- become lawyers or orators (Elias, 1995 ).
ing objectives, and with the goal of devel- Much of what we know about the
oping virtue and achieving truth (Rowe, Sophists anticipates todays professional
2001). Socrates similarly assessed his stu- or vocational training movement. Sophists
dents via informal, dialectic questioning, his were freelance teachers who charged a fee
quest always to find some person who knew for their services and were generally in com-
more than he (Rowe, 2001). petition with one another for teaching posi-
Plato, generally sharing Socratic views, tions (Saettler, 1968). Sophists taught arete,
had some specific recommendations con- a skill at a particular job, using a carefully
cerning the education of younger learners, prepared lecture/tutorial approach in what
which can be found in his classic work, could be conceived as an early attempt at
The Republic. For example, Plato states that mass instruction (Johnson, 1998). Sophist
future Guardians of the State should pur- instructional methodologies were system-
sue math-related studies for a period of ten atic, objective in nature, and made use of
years before advancement to subjects such as student feedback (Saettler, 1968). It was the
rhetoric or philosophy (Cooper, 2001). Plato applied skills-building aspect of Sophism
also emphasized the importance of abstract that Plato rejected, accusing the Sophists of
disciplines, such as philosophy and mathe- turning education into a marketable bag of
matics, but also believed that only a very tricks (Johnson, 1998).
few individuals possessed the natural abil- Sophists would likely have defined exper-
ities required for pursuit of such subjects tise as the presence of highly developed and
(Cooper, 2001). Thus, we witness in Plato comprehensive rhetorical and applied skills
72 the cambridge handbook of expertise and expert performance
that spanned the knowledge base of the era, The Medieval Context
a definition quite distinct from the notions
of Socrates or Plato. Central to Sophism Medieval Educational Structures
was the belief that there was a single base
Much of the knowledge from the ancient
skill rhetoric that once learned could be
school was carried through to the medieval
transferred to any subject (Johnson, 1998).
context, but medieval institutions increas-
Rhetoric therefore proved to be the chief
ingly codified and delineated that knowl-
subject of Sophist instruction, the educa-
edge. Subject matter was also acquiring an
tional goal being the development of what
increasingly practical application that would
today we might call a polymath, an individ-
serve the medieval church: geometry was
ual who had mastered many subjects and
required to design new church buildings,
whose knowledge was complete (Saettler,
astronomy was required for calculating the
1968). Sophist methods attempted to trans-
dates for special observances, and Latin was
fer rhetorical skill into all types of sub-
required for conducting religious services
ject domains, including geography, logic, his-
and interpreting ancient texts (Contreni,
tory, and music, through the acquisition
1989). Latin was the central focus of nearly
of cognitive rules for each domain (Saet-
all education, and mastery of the language
tler, 1968). The systematic nature of sophist
was required in order for one to be deemed
instructional techniques ensured that stu-
educated.
dents clearly understood learning objec-
A key event in the development of educa-
tives and assisted learners in gauging their
tional practice in medieval Europe occurred
own progress in achieving those objec-
with the ascent of the Frank leader Charle-
tives (Saettler, 1968). This approach, then,
magne (742814 AD), who established the
moved educational assessment slightly more
Frankish Empire, later to evolve into the
towards an objective standard than the
Holy Roman Empire, across a large por-
informal, oral techniques of Socrates and
tion of Europe. Charlemagne had a deep
Plato.
and abiding interest in education, imple-
menting educational reform in law through
Summary: Expertise in the Ancient Period a device called the capitularies, a collec-
tion of civil statues (Cross & Livingstone,
We witness in the ancient context two
1997). Charlemagnes motivation for edu-
unfolding views toward expertise, each
cation centered around two concerns: he
vested in a philosophical view of the nature
felt an educated populace was necessary for
and purpose of education. If one subscribed
the long-term well-being of the empire, but
to the notion that education held innate
also understood that the medieval church
worth and that its goal was the develop-
required highly trained individuals to con-
ment of the inner man (as did Plato and
duct all facets of the institutions business,
Socrates), then expertise could be seen
both secular and religious (LeGoff, 2000).
as the attainment of a general set of inner
Charlemagne set in motion a movement
traits that made one wise, virtuous, and in
toward formalized education that was to
harmony with truth. If one subscribed to
shape education in western Europe for cen-
the value of applied skills development (as
turies (Rowland-Entwistle & Cooke, 1995 ).
did the Sophists), then expertise could be
viewed as the attainment of a set of compre-
The University
hensive practical abilities. Regardless of the
position, the emphasis on rhetorical skills The emphasis Charlemagne placed on for-
and the individualized nature of instruction malized education in continental Europe
in this period proscribed a generally informal was both long-lived and influential. By
assessment of expertise based on the judg- the thirteenth century, the university had
ment of the teacher, not strictly on objec- become a focal point for intellectual devel-
tively defined performance measures. opment, and with it came a systematized
educators and expertise: theories and models 73
curriculum called the seven liberal arts medieval students initial course of study.
(Cross & Livingstone, 1997). The curricu- Instead, students engaged in oral discussion
lum was divided into an initial four-to-seven and debate with and between themselves
year period of study in Latin, rhetoric, and and the master for purposes of improving
logic called the trivium (leading to the bac- intellectual and rhetorical skills, as well as
calaureate), followed by a second period of weeding out students. After a period of some
advanced studies in arithmetic, astronomy, five years, a committee formed by the uni-
geometry, and music called the quadrivium versity presented the student with a prelim-
(leading to the masters or doctorate). It inary examination consisting of two parts:
was by progression through these curricula a series of private questions (responsio) fol-
that students acquired expertise and status lowed by a public dispute (determinatio). If
as a master. University courses were deliv- the student successfully defended both parts
ered in traditional didactic manner, with the of the exam, he was awarded baccalarii sta-
instructor presenting material that the stu- tus and was able to function with a mas-
dents would assimilate, grammatically ana- ter as an assistant teacher. Should a bac-
lyze, and restate to the instructor via written calarii decide to continue studies under the
and oral dialogue (Contreni, 1989). guidance of a master, the would-be doc-
The increased formalization and struc- toral candidate would, after many years of
ture of the medieval university amplified the additional study, be presented an exam-
performance demands placed on students. ination by the chancellor of the univer-
Students who sat on the floor while tak- sity. Completion of this examination, which
ing notes from the masters lectures were included reports on the moral character
forced to develop a battery of mnemonic of the student, led to the awarding of the
devices to remember lecture material, much doctoral degree. A newly awarded master
of which was extemporaneously delivered would then give his inaugural lecture (incep-
because of prohibitions against a master tio), which was also called commence-
reading from notes (Durant, 195 0). Fur- ment at Cambridge University (Durant,
ther adding to the demand placed on stu- 195 0).
dents was the fact that many students could Expertise and specialization among
not afford textbooks. Still handmade at this teaching faulty was also a salient element
point, books were rare and costly artifacts, of the university system. The abbey of St-
making oral lectures the primary source of Victor of Paris, for example, was well known
information (Durant, 195 0). for a series of highly acclaimed teaching
Formalization of medieval educational faculty, and John of Salisbury, teaching at
structures also affected the amount of time the cathedral school of Chartres, was held in
required to achieve a degree. It could, for esteem for his knowledge of political theory
example, take up to 16 years to achieve (Jordan, 2003 ). This trend evolved to a
the doctorate in theology or philosophy at point where institutions themselves devel-
the University of Paris, and as little as five oped reputations for excellence in specific
percent of students ever reached this level areas based on the faculty: among many
(Cantor, 2004). Most students left the sys- others, Bologna for law, Paris for theology,
tem in far shorter time (usually five to ten and Cambridge for natural philosophy and
years), taking lesser degrees that allowed theology (Jordan, 2003 ). The expertise
them to function successfully as cathedral represented by such institutions drew a
canons (Cantor, 2004). steadily increasing number of students,
The assessment techniques applied to many having costs defrayed by scholarships
medieval university students is described (Durant, 195 0). This trend reflected both
in detail in volume five of Durants clas- the extent to which expertise was valued
sic 11-volume text, The Story of Civilization within the education community, as well
(195 0). Durants history reveals that no for- as the increasing domain specialization
malized examinations took place during a of instructors, whose domain knowledge
74 the cambridge handbook of expertise and expert performance
was regarded as a critical component for Seville (5 6063 6 AD) applied grammatical
mastering any topic. rules to a wide variety of fields of study in
an attempt to view all knowledge through
the lens of language and its structure (Con-
Medieval Instructional Techniques
treni, 1989). But the techniques demon-
The medieval period saw the birth of a num- strate how an expert teacher, with highly-
ber of teaching techniques that were applied developed domain knowledge, sought to
at the various universities, cathedral schools, inculcate that knowledge in students and,
and monasteries throughout Europe. It was over time, develop highly proficient indi-
through these techniques that learners were viduals who would someday take over the
expected to master grammar, rhetoric, and teaching task.
language, all of which were the bedrock
requirement for mastery of higher educa-
The Craft Guilds
tion, and reflected the continuing impor-
tance of communication skills carried over A fascinating parallel to the formalized aca-
from the ancient context. demic systems found in medieval schools
Typical of such techniques was Scholasti- and universities were the craft guilds (Icher,
cism, an eleventh-century innovation greatly 1998) that targeted development of the high-
influenced by the questioning techniques est levels of expertise in their members.
of Abelard (10791142 AD) and later fully Begun around the tenth century, the craft
described by Aquinas (1225 1274 AD). guilds represented an applied-skills move-
Scholasticism was a syllogistic learning and ment that eventually covered a wide range
teaching technique that investigated appar- of building and manufacturing trades. The
ent contradictions in ancient texts (Cross & example of the European cathedral builders
Livingstone, 1997). Assessment under the reveals such trades as masons, stone cut-
Scholastic method was conducted by the ters, carpenters, plasterers, and roofers (Icher,
masters review of student responses to such 1998). By the thirteenth century, a total of
apparently contradictory source material: 120 craft corporations were catalogued with
the student was required to apply the rules of over 5 ,000 members. This number swelled
logic in an exacting technique, with the goal to 70,000 members in 165 0, consisting of
of being able either to defend the contradic- 20,000 masters and the rest apprentices and
tory statements as not actually containing journeyman (Cole, 1999).
contradiction, or to otherwise craft a con- In contrast to the general intellectually
vincing statement positing the human inabil- oriented emphasis of the medieval univer-
ity to resolve the contradiction (i.e., the sity, craftsmen progressed through a hands-
contradiction was a divine truth). These on apprenticeship of some seven-to-ten
interactions followed a set ritual (scholas- years within a specialized area. Craftsmen
tica disputatio), whereby a masters question were defined, even within groups, as supe-
required from the student first a negative rior and less important based on abilities
answer and defense, followed by a positive (Icher, 1998). The craft guilds movement
answer and defense, and finally a reasoned emphasized exacting performance within
set of responses to any objections1 (Durant, each discipline, all under the watchful eyes
195 0). Thus, it can be seen that the ancient of a hierarchy of fellow artisans who both
topics of rhetoric and oratory still held pow- formally and informally critiqued ongoing
erful sway in the medieval curriculum. work. The rule for being a master crafts-
There were also other instructional man was Whosoever wishes to practice the
approaches employed in the medieval craft as master, must know how to do it in all
university: Comenius (15 921670 AD), for points, by himself, without advice or aid from
example, taught by using visuals embed- another, and he must therefore be examined by
ded within instructional texts, such as his the wards of the craft (Cole, p. 5 0). In many
Orbus Pictus (Saettler, 1968), and Isidore of respects, because of the emphasis placed
educators and expertise: theories and models 75
on the development of specific skills, the ments (Roberts, 1997). European transfor-
extended period of training, and the repro- mation from an agrarian society into towns
ducibility of performance, an argument can with seemingly innumerable factories and
be made that these medieval craftsmen con- mills placed new demands on existing edu-
formed loosely to our modern understanding cational systems. Prior to this, education
of expert performance (cf. Ericsson & Smith, was restricted to privileged groups, including
1991; Ericsson, Chapter 3 8). males, religious clerics, nobility, and those
with the means to afford it. Even the craft
Summary: Expertise in the Middle Ages guilds often charged large sums of money
for admittance, restricting membership to
Three primary factors characterized the a select pool of potential apprentices. This
development of expertise in the medieval left education at a fundamental disconnect
period, all continuing to proffer the notion of with the common classes, leaving them to
teacher as expert. First, the formalization and learn what they could outside formal sys-
systemization of educational structures such tems (Contreni, 1989).
as the university and cathedral schools helped The role of privilege and gender as it per-
to strengthen and codify knowledge that tained to education greatly diminished with
could then be studied and mastered by topic. the Industrial Revolution. As a countrys
Second, the implementation of new instruc- economic situation improved because of the
tional techniques, typified by the Scholastic new industries, the demand for supplying a
method, moved educators away from ad hoc continual stream of skilled industry work-
instruction into analyzing learning processes ers forced educational structures to evolve in
in a more systematic manner and estab- order to keep pace with that demand. This
lishing sequences of instruction to improve contributed in part to the ever-increasing
learning outcomes. Finally, the appearance enrollment rates that were seen in many
of the craft guilds established a skills-based, European schools, eventually resulting in
performance-assessed, and domain-specific essentially universal enrollment in portions
learning community that mastered the arti- of Europe (Craig, 1981). In some countries,
san trades under the direct guidance and free, state-based education became compul-
supervision of experts. sory (Roberts, 1997), and basic primary edu-
Medieval assessment continued to make cation became available to both girls and
use of informal, rhetorically based tech- boys (Davies, 1996).
niques. Although medieval educational str- The postindustrial educational model
uctures increasingly moved assessment therefore represents a significant shift in the
toward formalization, informal assessment development of human skill. In ancient
nevertheless continued to prevail. The craft times, learners spent time with the instruc-
guilds were the exception, where skills were tor on an individual basis, engaging in inter-
developed and assessed to a high level of active dialogue and questioning. Later, in
specificity and were routinely measured and medieval times, although educational for-
formally assessed by the guild masters. malization was increasingly present, stu-
dents still moved to the location of their
master of choice, working with the master
to achieve educational goals.
The Modern Context
After the industrial revolution, however,
mass-production techniques from industry
Impact of Modernization on Education
were applied to the educational world,
One of the most significant historical events employing large instructional classes and
to impact education was the Industrial Rev- many teachers. In such an environment, the
olution, a period commencing in Britain in upper limit construct, the upper performance
the eighteenth century as a result of a vari- bounding of such a massed, classroom-based
ety of economic and technological develop- learning environment, began to come into
76 the cambridge handbook of expertise and expert performance
play (cf. Branson, 1987). Learners were now Education Becomes a Science
taught basic skills such as reading, writ-
ing, and basic math, and the goal was the By the late nineteenth century, the subject of
development of competent industry workers Education became institutionalized in the
(Madsen, 1969). Removing learners even fur- universities as a distinct field, no longer the
ther from the one-on-one and personalized forte of the various specialized disciplines.
instruction of the ancient context, and no Universities at this point were transitioning
longer focused on the development of exper- into research institutions, and calls for the
tise in any one particular area, this indus- application of a science-based approach to
trial education model can be seen in many education were becoming increasingly com-
settings until the current day (e.g., in liberal mon (Lagemann, 2000). Harvard professor
arts vs. engineering education). Josiah Royce wrote his influential piece, Is
Further, the notion of the development There a Science of Education?, in which he
of a true polymath, an educational goal said there was no universally valid science of
tracing its roots all the way back to the pedagogy (Royce, in Lagemann, p. ix). John
ancient context (and later revived in the Dewey (185 9195 2) was another of the early
Renaissance in the concept of a Renaissance players in the attempt to apply science to
man), became increasingly disregarded in education, writing a 1929 work, The Sources
the Industrial context. Indeed, the move of a Science of Education (Lagemann, 2000).
toward industrialization was not the only Much of the subsequent work in education
factor at play: as the amount of avail- was spearheaded by psychologists who had
able knowledge exploded with the Renais- recently undergone the division of their field
sance, it became increasingly apparent that from philosophy, and the discipline of edu-
no one person would ever master in toto cational psychology soon came into existence.
such a collection of knowledge. Specializa- Carrying on from the pioneering work of
tion by field was now becoming the dominant the Wundt laboratory at Leipzig in 1879 and
paradigm when moving beyond the basic subsequent work by Ebbinghaus and others,
skills demanded by industry. Van Doren learning was to be scientifically and empirically
(1991) notes that investigated as a distinct psychological process
(Boring, 195 0).
The failure of the Renaissance to produce The impact of this extraordinary shift in
successful Renaissance men did not go approach can hardly be overstated: every
unnoticed. If such men as Leonardo, Pico, aspect of the learning process, includ-
Bacon, and many others almost as famous ing learner characteristics, instructional
could not succeed in their presumed dream
methodologies, psychological processes, and
of knowing all there was to know about
even physiological factors were now to be
everything, then lesser men should not pre-
sume to try. Thealternative became self- scrutinized, quantified, and aggregated into
evident: achieve expertise in one field what would eventually become learning
while others attained expertise in theirs. theories. This approach was also highly
(Van Doren, 1991) (emphasis added) significant in that it threatened to remove
teaching from the exclusive control of
Thus, the goal of developing expertise in domain experts: the field of education
all fields had been fully abandoned by the would now seek to develop generalized
time of the Industrial Revolution. If any per- scientific approaches for teaching and
son was to become an expert, that recogni- learning any subject, and the joint efforts
tion was likely to be gained in a single field of educators and psychologists would
or domain of study.2 (See Feltovich et al., develop these approaches (Lagemann,
Chapter 4, for a comparision of this trend 2000).
from generality to specificity in the concept These investigations would play a domi-
of expertise.) nant role in the manner in which researchers
educators and expertise: theories and models 77
viewed expertise. If a science-based and during the 1960s, including use in the U.S.
empirically validated theory of learning military (Gardner, 1987).
could describe the psychological process of
learning, then the development of exper-
World War II, the Military, and
tise, that is, learning taken to its ultimate
Performance
realization, would similarly be described. It
was often assumed that expertise was devel- World War II, much like the Indus-
oped by successive application of the pre- trial Revolution, brought new performance
scriptive methods built from each theory demands to the educational establishment.
until the specified performance level was The requirements for consistent and com-
achieved. Some of the more prominent of petent performance under battle conditions
these theories and models are now briefly required that new theories and techniques
presented. be applied within military educational struc-
tures. In reaction to these demands, gen-
eral systems theory was applied to a vari-
ety of practical military problems. Because
Programmed Instruction
many psychologists were involved in mili-
and Teaching Machines
tary training and selection programs, these
Programmed instruction was one of the individuals began to adopt systems think-
first technologies of instruction that used ing in approaching military-related human
a psychological theory (i.e., behaviorism) resources issues.
as a rationale for its technique (Saettler, Robert B. Miller (1962) first formal-
1968; Feltovich et al., Chapter 4). Sidney L. ized the relationships among various psy-
Pressey, a psychologist at Ohio State Univer- chological approaches to create expertise
sity, developed the technique in the 1920s, in military jobs. Both the Armys Human
though presaged much earlier by Comenius, Resources Research Organization (Hum-
Montessori, and others. Skinner, seeing the RRO) and the Air Force Human Resources
educational potential of the approach, pop- Laboratory were major contributors to this
ularized the technique a few decades later, effort (Ramsberger, 2001). Much of the the-
even using the method to teach his own ory surrounding this history is captured in
classes at Harvard University in the 195 0s Gagnes (1966) Psychological Principles in
(Saettler, 1968). System Development, which contains mate-
The technique used a mechanical or rial by the leading advocates of performance
paper-based device, called a teaching development through application of systems
machine, to control the presentation of a thinking.
programmed sequence of highly structured In the late 1960s, the Armys Continental
questions to the learner. The learners Army Command (now TRADOC) issued
understanding was shaped by providing a regulation that set forth the major func-
immediate feedback as the learner answered tions of training design, development, and
questions embedded in the material, implementation (The Systems Engineering
branching to appropriate places based of Training, 1968). Beginning in the early
on learner response (Garner, 1966). This 1970s, the term systems engineering was
allowed students to perform self-assessment gradually replaced by instructional systems
through the instructional sequence, branch- development in the training community, and
ing either forward or backward in the all branches of the military service formally
sequence depending on the correctness of adopted that term with the publication of
particular responses. The methodology was the Interservice Procedures for Instructional
found to be highly effective in a number Systems Development (Branson, 1978).
of cases, prompting a large programmed Because a substantial number of military
instruction movement in the United States tasks were highly consequential, a demand
78 the cambridge handbook of expertise and expert performance
for expertise in a wide variety of jobs was Questions surrounding the ability of learn-
both desirous and necessary. ers to organize information and solve prob-
Such military-related jobs had histori- lems, for example, seemed to be left unad-
cally been trained through standard school- dressed by raw behavioral theory (Tuckman,
house platform instruction. The introduc- 1996). This led to the development of a num-
tion of increasingly effective simulators and ber of new learning theories that pointedly
part-task trainers, coupled with a complete included mental operations as part of the
instructional design process based on sys- description of learning. Among these were
tems theory, made training more efficient the information processing theory of Atkin-
and effective across all jobs. son and Shiffrin (Matlin, 2002) and the cog-
Fundamental, too, for success in mili- nitive approach of Robert M. Gagne (1989).
tary training was the research effort that
supported the development of new prac- Learning Hierarchies
tices and the continuing commitment to
Robert M. Gagne (19162002), a leading
use systems- and evidence-based approaches
educational psychologist at the University of
to training. Increasingly complex jobs and
California at Berkeley and subsequently at
missions required increasingly sophisticated
Florida State University, conducted exten-
training approaches, and the implemen-
sive investigations into the nature of learn-
tation of highly capable simulators made
ing as a psychological process, leading him
possible the practice necessary for success.
to the development of a concept he termed
Through the process of distributed interac-
learning hierarchies. As implied by the name,
tive simulation, full missions could now be
a learning hierarchy is a set of specified
rehearsed in advance until a criterion perfor-
abilities having an ordered relationship to
mance level was met. The capabilities con-
one another, generally depicted in graphi-
tained in such simulated systems included a
cal format (Gagne, 1989). The learning hier-
full range of what-if scenarios and set the
archy, then, depicts a skill and its com-
early stage for the introduction of comput-
ponent subskills requisite for performing
erized simulators with real-time, software-
the skill. Gagne simultaneously categorized
based programming in the later part of
skills with regard to their placement within
the twentieth century (Ward, Williams, &
a learning outcome taxonomy consisting of
Hancock, Chapter 14).
psychomotor skills, verbal information, intel-
lectual skills, cognitive strategies, and attitudes
The Rise of Cognitivism (Gagne, Briggs, & Wager, 1992). The hierar-
chy is often constructed in conjunction with
Behaviorism had posited that learners were
a task analysis, a detailed specification of
essentially blank slates, changed by their
the mental and physical processes involved
environment and learning through the
in task performance (Smith & Ragan, 1999;
mechanisms of stimulus and response. In
Schraagen, Chapter 11).
this view the learner was a passive recipi-
ent of knowledge, waiting for imprinting of
Carrolls Model of School Learning
new learning. Over time, the new stimulus-
response patterns would be strengthened Harvard University professor John B. Car-
and become automatic: learning was then roll (19162003 ) in 1963 proposed his model
said to have occurred. Expertise could of school learning (Carroll, 1963 ). Carrolls
be viewed as the development of many model, although not a learning theory per se,
automatized stimulus-response patterns fully nevertheless demonstrated a practical equa-
imprinted within the learner. tion for how individual task mastery is
By the mid-twentieth century, however, attained and also challenged traditional
a number of theorists were raising questions notions of student aptitude (Guskey, 2001).
about the ability of behaviorism to explain Carrolls system used five variables, three
all learning and psychological processes. internal to the learner (amount of time
educators and expertise: theories and models 79
required for learning the task, ability to affective entry characteristics, and quality of
understand instruction, and time willing to instruction could account for up to 90% of
be spent on learning) and two external (time achievement variation in students (Guskey,
allowed for learning and quality of instruc- 2001). Noting the inadequacies of tradi-
tion). Carroll combined these five elements tional instructional methods, Bloom inves-
into a ratio that results in the degree of tigated two types of processes: the processes
learning: degree of learning = (time actually involved in pairing students with excel-
spent on learning) / (time needed for learn- lent tutors, and the practices and strate-
ing) (Carroll, 1963 ). Challenging the tradi- gies of academically successful students.
tional notion of student aptitude as ability Blooms resultant instructional methodol-
(see also Horn & Masunaga, Chapter 3 4), ogy included two principal components:
Carroll said that aptitude was more accu- first, feedback, corrective, and enrichment
rately a measure of learning rate, or the time processes; and second, instructional align-
an individual student requires to master a ment (Guskey, 2001). These were combined
new learning task. Carrolls model depicted into a self-paced, individualized learning sys-
acquisition of expertise, therefore, as pri- tem to give each student the specific instruc-
marily a function of time: time required for tion and adequate time needed for mas-
learning (called aptitude), time willing to be tery of the instructional task (Kulik et al.,
spent on learning (called perseverance), and 1990). Numerous studies have confirmed the
time allowed for learning (called opportu- efficacy of the mastery learning model (see
nity). Carrolls work influenced a number of Kulik et al., 1990, for a meta-analysis), and
individualized instruction methodologies, the technique remains in use today.
including Individually Prescribed Instruc- The relationship of Blooms model to
tion (Glaser, 1966), Individually Guided the development of expertise lies within
Education (Klausmeier, 1971), and others the theorized percentage of students master-
(Guskey, 2001). ing a topic when using the method: Bloom
claimed that the outcome of such a mastery
approach could alter the normal perfor-
Mastery Learning
mance curve frequently witnessed in edu-
One of the key results of Carrolls model cational settings. Such normal performance
was the interest it stirred for Benjamin curves, Bloom posited, were actually what
Bloom (19211999) in suggesting methods might be witnessed with no instructional
to improve school outcomes. Bloom was intervention present, and with student apti-
an educational research and policy analyst tude alone determining learning outcomes
from the University of Chicago interested in (Smith & Ragan, 1999). The implication was
improving the effectiveness and quality of that expertise in this model lay well within
educational methods. Bloom believed that the grasp of a majority of students, not sim-
virtually all students could excel and master ply a small percentage of those with natural
most any subject, given the proper learning aptitude.
conditions (Bloom, 1968). Bloom had pre-
dicted that, given such conditions, 90% of
Objectives
students could perform to levels previously
only reached by the top 10% (Kulik, Kulik, In the early 1960s, Mager sought a method
& Bangert-Drowns, 1990). that would enable teachers and trainers
Carrolls work stimulated Bloom to to operationalize their instructional inten-
extend the work to encompass a new model tions. Magers widely read book, Prepar-
of teaching and learning called mastery learn- ing Objectives for Programmed Instruction
ing. Bloom laid out the theory in his 1976 (1962), influenced the systematic design of
work, Human Characteristics and School instruction probably more than any other
Learning, in which he theorized that the text. Mager intended that all instructors
combination of cognitive entry behaviors, should state in advance the precise behaviors
80 the cambridge handbook of expertise and expert performance
that they intended to cause and then mea- come. Most domains contain multiple out-
sure the accomplishment of those behav- comes.
iors with criterion-referenced procedures. Considering expertise from another per-
Instructional outcome specifications and spective, one can think of a student being
measurement to absolute standards gradu- an expert third-grader. For hierarchically
ally became the norm for training critical ordered intellectual skills such as mathe-
tasks in a variety of domains. matics, learners must achieve behavioral flu-
ency at one level before they can successfully
Learning Outcomes progress to the next level (Binder & Watkins,
1990). Binder and Watkins argue that behav-
In the mid-1960s, Gagne published his mon- ioral fluency is similar to automaticity and
umental work, The Conditions of Learning that the best dependent variables for assess-
(1965 ), in which he integrated research and ing learning are the response time required
theory to provide a comprehensive approach to recall and use any fact or relationship
to instruction. Although Gagne was princi- (e.g., solving equations) and the accuracy of
pally focused on the learning of school sub- the response. Thus, those students with the
jects, his work was widely used in other are- shortest times and highest accuracy scores
nas. It was Gagnes interest in school sub- are the experts. Binder and Watkins have
jects that led him to conceptualize the con- made a strong case that instruction should
struct of the learning hierarchy. He recog- be designed to cause behavioral fluency in
nized three major relationships between ini- all students (Binder & Watkins, 1990).
tial and subsequent learning:
r The learning of A was independent of Constant Time of Exposure Model vs.
the learning of B, and no sequence was Criterion Referenced Instruction
implied, As in the ancient context, the twentieth
r The learning of A facilitated the learning century witnessed the development of two
of B, thus suggesting a favored sequence, distinct educational philosophies and their
and related instructional practices that were in
r The learning of B could not occur unless tension with one another. The vast major-
A had first been learned, thus requiring a ity of public schools, universities, and many
correct sequence. military schools applied the traditional con-
stant time of exposure model. The constant-
These relationships are substantially time model produces learner results that
incorporated into Gagnes Nine Events of vary much as the normal curve, establish-
Instruction (1965 ). From the research liter- ing the basis for grading students and causing
ature, Gagne defined five possible learning winners and losers. Because many situations
outcomes: and occupations require constant, compe-
r Motor skills encompass the totality of the tent performance, the constant time model
does not meet the requirements for many
perceptual-motor domain. learners.
r Verbal information is declarative knowl-
Consistent with the work of Carroll and
edge about a domain. Bloom, who demonstrated that providing
r Intellectual skills are required to apply different amounts of time to learn pro-
principles and rules to a domain. duced a much larger proportion of students
r Cognitive strategies are higher-order pro- that reached criterion, Glaser and Klaus
cesses for learning and problem solving. (1962) elaborated the practice of criterion
r Attitudes are choices that a learner makes referenced testing. For any level of expertise,
about future behavior. subject matter experts developed criterion per-
formances that could be reliably judged by
Designing instruction appropriate for the those proficient in the domain. Instruction was
domain could facilitate each learning out- designed to accommodate a distribution of
educators and expertise: theories and models 81
time-to-completion measures that resem- to allow the process to adjust and improve
bled the normal curve of performance scores (Rothwell & Kazanas, 1992). The result is an
found in traditional settings. The intention objective, tightly controlled, and research-
was to identify the level of performance grounded process that is easily applied to a
that was required by the authentic situation wide variety of learning situations.
and measure the performance of individuals
compared to the standard or criterion. It was
deemed particularly important to use crite- Instructional Systems and Experts
rion performances to judge competence in A significant characteristic of the ISD
highly consequential tasks. approach relating to expertise is the man-
In the majority of education and training ner in which it makes use of domain experts.
environments, the goal is usually to develop Because the use of learning hierarchies pred-
competence in large numbers of people. The icates an understanding of the skills and sub-
overlap between competence and exper- skills required for task performance, ISD
tise might be illustrated by comparing naval employs domain experts as a source of accu-
aviators and concert violinists. The Naval rate information on how specific tasks should
Aviation flight training community strives be conducted. In addition, the use of domain
to have every cadet become competent at experts to inform instructional decisions in
carrier landings. Thus, each landing must the ISD approach means that the desired
be made according to defined standards for outcome for instruction is aligned with
approach and touchdown. Each landing is how experts, not novices, perform a task
highly consequential and when done incor- (Feltovich et al., Chapter 4, expertise as
rectly, the result is immediately and pub- goal state for learning). Although ISD is
licly known. For the world-class concert vio- frequently used to train novices in a subject
linist, only a small portion of the audience area and the stated goal is often described as
would ever know that the performance given competency, ISDs use of criterion-aligned
was not up to the violinists high expecta- performance against an expert standard
tions. Rarely would an average performance implies that the goal is to develop learners
be consequential to an expert violinist. who do not stop at competence, but con-
tinue on the path to expertise. The ISD edu-
Instructional Systems cational approach is therefore unique in its
A number of the aforementioned theory and view of experts: domain experts are now seen
research efforts coalesced with the Instruc- as a source of information for informing the
tional Systems Development (ISD) move- learning process, but not necessarily as design-
ment in the late twentieth century. Making ers of instruction.
simultaneous use of performance objectives
(Mager), the events of instruction (Gagne),
Curriculum Reform: Using Domain
instruction with feedback (Skinner), criterion-
Experts to Design Instruction
referenced instruction (Mager), and learn-
ing hierarchies (Bloom, Gagne), the systems Interestingly, there was a concurrent series
approach to instructional design is a method- of curriculum reform efforts in the United
ology rooted in both educational research States alongside the Instructional Systems
and applied experience, whose goal is the movement that applied an opposite approach
development of effective, quality instruc- toward the use of experts in education.
tional materials. The ISD methodology is The movement included such now-famous
differentiated from others in that it applies efforts as the new math, and encom-
basic concepts from systems theory (Katz & passed a wide variety of disciplines includ-
Kahn, 1978) to the design of instruction. ing physics, history, and mathematics. This
Each stage of the ISD process is viewed movement is documented in Lagemanns An
as input to another stage as well as output Illusive Science: The Troubling History of Edu-
from a previous stage, with feedback loops cational Research (2000). These curricular
82 the cambridge handbook of expertise and expert performance
reforms were aimed at reinfusing discipline- now represented by the Society for the Anal-
based scholarship into the design of edu- ysis of Behavior.
cational materials in reaction to what was Around 1970, cognitive psychologists
considered the poor results of educators, began to provide data and theory suggest-
who had assumed responsibility for teach- ing that humans were subject to acquiring
ing such subjects with the rise of educa- behavior that was best explained from an
tion as a discipline. The claim was that Information processing perspective. Cogni-
domain experts, including physicists, math- tion was again considered a legitimate source
ematicians, and historians, would be able of data, depending on the experimental
to bring academic rigor to their subjects, methodology that established it. Sensation
and improve classroom materials. This heav- and perception, as well as other functions of
ily funded movement lasted more than a the nervous system, were important areas of
decade, and with the results of the approach study.
contested from all sides, produced no clear Another approach appeared on the scene
consensus of its impact (Lagemann, 2000). with the advent of constructivism. Beginning
around 1985 although some would argue
that the date was much earlier a number
Constructivism
of educational researchers began to elaborate
As implied by our history, each generation the tenets of constructivism. Based primarily
has found ways to reject prior wisdom and on the study of school subjects, as construc-
strike out on a new direction. Psychology tivist literature is almost exclusively tied to
has seen many such excursions in which the the development of learning environments
current fad or fashion is considered to be within school settings (Tobin, 1993 ), con-
the truth. There were the structuralists, the structivists posited that students could learn
behaviorists, and then the radical behavior- only if they effectively mapped new infor-
ists, each group vociferously marking out mation onto prior knowledge and experi-
intellectual territory. Parallel to these posi- ence. Stated another way, learners were said
tions was a generic empirical psychology that to construct their own knowledge, which
sought to find answers to basic questions may or may not map to what others consider
from a theory and research base. This group objective reality.
included those who sought empirical meth- Limiting the bounds of constructivism to
ods to improve military training and ways the study of school subjects is a productive
of increasing performance. In that group can effort. As previously mentioned, systems
be found Gagne (1989) and Glaser (1966), psychologists recognize that the traditional
among others. model of schooling long ago reached the
Many of these viewpoints had their own upper limit of its capability. Therefore,
research agenda and methods of collecting the design of constructivist learning envi-
data. Few would challenge the empirical ronments in schools can be a significant
findings of Skinner and his colleagues who step forward. Early research (Scardamalia &
detailed the results of schedules of reinforce- Bereiter, 1994) indicates that students can
ment, that were primarily described from greatly improve their knowledge acquisi-
animal research and were demonstrated to tion skills using technologies and constructs
apply in the same manner to rats, pigeons, based on information processing.
and humans. For decades, psychologists have Students advancing their learning in con-
known that a stimulus event, followed by a structivist learning environments represent
behavior, followed by a consequence would one level of achievement. However, they do
lead to a change in the probability of that not represent promising options for develop-
behavior occurring at the next presentation ing the two kinds of expertise mentioned ear-
of the cue stimulus. Skinner and his stu- lier. Earlier in this chapter, we attempted to
dents and colleagues refined this generaliza- classify development that leads to expertise
tion over the years. This school of thought is into two major categories: instruction that
educators and expertise: theories and models 83
memory, sports, chess, and music. Ericssons as the point at which undergraduate univer-
model, with its emphasis on objective and sity programs totally abandoned the idea of a
verifiable assessment of skill levels, remains a liberal education, at which point . . . the lib-
leading empirical explanation of the acquisi- eral curriculum was abandoned almost every-
tion of expert performance in a wide variety where, and the departmental organization of
the educational establishment was installed at
of performance domains.
all levels below the university, even in many
elementary schools (p. 142).
Conclusion
References
The historic evolution of views concerning
skills development, commencing with the Binder, C., & Watkins, C. L. (1990). Precision
informal and individualized instruction of teaching and direct instruction: Measurably
Socrates and Plato and continuing to the superior instructional technology in schools.
empirically measured and formally assessed Performance Improvement Quarterly, 3 (4),
instruction of today, has resulted in the mod- 7496.
ern attempt at building a common, empiri- Bloom, B. S. (1968). Learning for mastery. Evalu-
cally based understanding of the attainment ation Comment, 1(2), 15 .
of expertise and expert performance. To Boring, E. G. (195 0). A history of experimental psy-
achieve this goal, a transdisciplinary group of chology. New York: Appleton-Century-Crofts,
scholars, including educational researchers, Inc.
cognitive psychologists, domain experts, and Branson, R. K. (1978). The interservice proce-
many others, work together to build a shared dures for instructional systems development.
understanding of high-performance phe- Educational Technology, 2 6(3 ), 1114.
nomena. The standard, too, is now higher: Branson, R. K. (1987). Why the schools cant
empirical performance measures, repro- improve: The upper limit hypothesis. Journal
of Instructional Development, 10(4), 15 26.
ducibility of results within and between
learners, and theoretical models that with- Cantor, N. F. (2004). The last knight: The twilight
of the middle ages and the birth of the modern era.
stand the rigors of experimental validation
New York: Free Press.
are all a part of this quest. Advances in
Carroll, J. B. (1963 ). A model of school learning.
these fields provide evidence that empiri-
Teachers College Record, 64, 723 73 3 .
cally verifiable models, encompassing all of
Chi, M. T. H., & Glaser, R. (1982). Expertise
the variables surrounding the phenomenon
in problem solving. In R. S. Sternberg (Ed.),
of learning, are still a worthy goal for
Advances in the psychology of human intelligence
educators. (Vol. 1, pp. 175 ). Hillsdale, NJ: Erlbaum.
Cole, R. (1999). A travellers history of Paris (3 rd
Footnotes ed.). Gloucestershire: The Windrush Press.
Contreni, J. J. (1989). Learning in the early middle
ages: New perspectives and old problems. The
1. This instructional technique bore loose resem-
International Journal of Social Education (offi-
blance to an older teaching methodology
cial journal of the Indiana Council for the Social
traced as far back as the ninth century called
Studies), 4(1), 925 .
the Quaestio method, in which the master
embedded questions to be answered by the Cooper, D. E. (2001). Plato. In J. A. Palmer (Ed.),
scholar at selected points within the instruc- Fifty major thinkers on education from Confucius
tional material (Contreni, 1989). to Dewey (pp. 1014). London: Routledge.
2. Van Doren posits that the segmentation of the Craig, J. E. (1981). The expansion of education.
university into departments helped facilitate Review of Research in Education, 9, 15 1213 .
the shift towards specialization, the uni in Cross, F. L., & Livingstone, E. A. (Eds.). (1997).
the term university having been abandoned The Oxford dictionary of the Christian church.
(p. 141). Van Doren points to World War II New York: Oxford University Press.
educators and expertise: theories and models 85
Davies, N. (1996). Europe: A history. New York: Glaser, R. & Klaus, D. J. (1962). Proficiency
Oxford University Press. measurement: Assessing human performance.
Durant, W. (195 0). The age of faith. New York: In R. M. Gagne (Ed.), Psychological princi-
MJF Books. ples in systems development. New York: Holt,
Elias, J. L. (1995 ). Philosophy of education: Clas- Rinehort & Winston.
sical and contemporary. Malabar, FL: Krieger Guskey, T. R. (2001). Benjamin S. Blooms con-
Publishing Company. tributions to curriculum, instruction, and school
Ericsson, K. A. (1996). The acquisition of expert learning. Paper presented at the Annual Meet-
performance: An introduction to some of the ing of the American Educational Research
issues. In K. A. Ericsson (Ed.), The road to Association, Seattle, WA.
excellence: The acquisition of expert performance Hannafin, M. J., & Hill, J. R. (2002). Epis-
in the arts and sciences, sports, and games. temology and the design of learning envi-
Mahwah, NJ: Erlbaum. ronments. In R. A. Reiser (Ed.), Trends and
Ericsson, K. A. (2004). Deliberate practice and issues in instructional design and technology.
the acquisition and maintenance of expert per- Upper Saddle River, NJ: Merrill/Prentice-
formance in medicine and related domains. Hall.
Academic Medicine, 79(10), S70S81. Icher, F. (1998). Building the great cathedrals (A.
Ericsson, K. A., & Delaney, P. F. (1998). Long- Zielonka, Trans.). New York: Harry N. Abrams,
term working memory as an alternative to Inc.
capacity models of working memory in every- Jeffs, T. (2003 ). Quest for knowledge begins with
day skilled performance. In A. Miyake & P. a recognition of shared ignorance. Adults Learn-
Shah (Eds.), Models of working memory: Mech- ing, 14, 28.
anisms of active maintenance and executive Johnson, S. (1998). Skills, Socrates and the
control (pp. 25 7297). Cambridge, England: Sophists: Learning from history. British Journal
Cambridge University Press. of Educational Studies, 46(2), 201213 .
Ericsson, K. A., & Smith, J. (1991). Prospects Jordan, W. C. (2003 ). Europe in the high mid-
and limits of the empirical study of expertise: dle ages (Vol. 3 ). New York: Penguin Putnam
An introduction. In K. A. Ericsson & J. Smith Inc.
(Eds.), Toward a general theory of expertise: Katz, D., & Kahn, R. (1978). The social psychology
Prospects and Limits (pp. 13 8). New York: of organizations (2nd ed.). New York: Wiley.
Cambridge. Klausmeier, H. J. (1971). Individually guided edu-
Fitts, P., & Posner, M. I. (1967). Human perfor- cation in the multi-unit school: Guidelines
mance. Belmont, CA: Brooks/Cole. for implementation. Phi Delta Kappan, 5 3 (3 ),
Gagne, R. M. (1965 ). The conditions of learning. 181184.
New York: Holt, Rinehart & Winston. Kulik, C.-L. C., Kulik, J. A., & Bangert-Drowns,
Gagne, R. M. (1966). Psychological principles in R. L. (1990). Effectiveness of mastery learn-
system development. New York: Holt, Rinehart ing programs: A meta-analysis. Review of Edu-
& Winston. cational Research, 60(2), 265 299.
Gagne, R. M. (1989). Studies of learning: 5 0 years Lagemann, E. C. (2000). An elusive science: The
of research. Tallahassee, FL: Learning Systems troubling history of education research. Chicago:
Institute. The University of Chicago Press.
Gagne, R. M., Briggs, L. J., & Wager, W. W. LeGoff, J. (2000). Medieval civilization: 400
(1992). Principles of instructional design (4th 15 00: Barnes & Noble Books.
ed.). Fort Worth: Harcourt Brace Jovanovich. Madsen, D. (1969). History and philosophy of
Gardner, H. (1987). The minds new science. New higher education. In H. E. Mitzel (Ed.), Ency-
York: Basic Books. clopedia of educational research (5 th ed., Vol. 2,
Garner, W. L. (1966). Programmed instruction. pp. 795 803 ). New York: Macmillan Publish-
New York: Center for Applied Research in ing Co, Inc.
Education. Mager, R. F. (1962). Preparing objectives for pro-
Glaser, R. (1966). The program for individually pre- grammed instruction. Belmont, CA: Fearon.
scribed instruction. Pittsburgh, PA: University of Matlin, M. W. (2002). Cognition: Harcourt Col-
Pittsburgh. lege Publishers.
86 the cambridge handbook of expertise and expert performance
Miller, R. B. (1962). Task description and analysis. nities. Journal of the Learning Sciences, 3 (3 )
In R. M. Gagne (Ed.), Psychological principles in 265 283 .
system development (pp. 3 5 3 3 80). New York: Smith, P. L., & Ragan, T. J. (1999). Instruc-
Holt, Rinehart & Winston. tional design (2nd ed.). Upper Saddle River, NJ:
Ramsberger, P. F. (2001). HumRRO: The first Merrill/Prentice Hall.
5 0 years. Alexandria, VA: Human Resources Spiro, R. J., Feltovich, P. J., Jacobson, M. J.,
Research Organization. & Coulson, R. L. (1991). Cognitive flexibility,
Roberts, J. M. (1997). A history of Europe. New constructivism, and hypertext: Random access
York: Allen Lane/The Penguin Press. instruction for advanced knowledge acquisition
Rothwell, W. J., & Kazanas, H. C. (1992). Master- in ill-structured domains. Educational Technol-
ing the instructional design process: A systematic ogy, 11(5 ), 243 3 .
approach. San Francisco: Jossey-Bass, Inc. The Systems Engineering of Training. (1968). Army
Rowe, C. J. (2001). Socrates. In J. A. Palmer (Ed.), Continental Army Command, U.S. Army.
Fifty major thinkers on education from Confucius Tobin, K. (1993 ). The practice of constructivism
to Dewey (pp. 5 10). London: Routledge. in science education. American Association for
Rowland-Entwistle, T., & Cooke, J. (1995 ). Great Advancement of Science.
rulers of history: A biographical history. Barnes Tuckman, B. W. (1996). Theories and applications
& Noble Books. of educational psychology. New York: McGraw-
Saettler, P. (1968). A history of instructional tech- Hill, Inc.
nology. New York: McGraw-Hill. Van Doren, C. (1991). A history of knowledge:
Scardamalia, M., & Bereiter, C. (1994). Com- Past, present, and future. New York: Ballantine
puter support for knowledge-building commu- Books.
CHAPTER 6
Expert systems are computer programs that this coincides with the development of
exhibit some of the characteristics of exper- the digital stored-program computer, com-
tise in human problem solving, most notably puter programming, artificial intelligence
high levels of performance. Several issues (AI) research, and information-processing
are described that are relevant for the study models of human cognition (Feltovich,
of expertise and that have arisen in the Prietula, & Ericsson, Chapter 4). The lan-
development of the technology. Moreover, guage of cognitive information processing
because expert systems represent testable was developed by the same AI researchers
models that can be manipulated in labora- and cognitive psychologists that had adopted
tory situations, they become a new method- computation as the basis for models of
ology for experimental research on exper- thought (Anderson, 1982; Feigenbaum &
tise. The main result from work on expert Feldman, 1963 ; Newell & Simon, 1972;
systems has been demonstrating the power VanLehn, 1996).
of specialized knowledge for achieving high AIs scientific goal is to understand intel-
performance, in contrast with the relatively ligence by building computer programs that
weak contribution of general problem solv- exhibit intelligent behavior and can be
ing methods. viewed as models of thought. One core
of the AI science is concerned with the
concepts and methods of symbolic infer-
AI and Expert Systems: ence, or reasoning, by a computer, and
Foundational Ideas how the knowledge used to make infer-
ences will be represented inside the com-
A science evolves through language and puter. The term intelligence covers many cog-
tools that express its concepts, mechanisms, nitive skills, including the ability to solve
and issues. The science of studying exper- problems, perceive, learn, and understand
tise evolved largely in the second half of language. AI scientists study and model all
the 20th century. It is not accidental that of those.
87
88 the cambridge handbook of expertise and expert performance
problem areas that were acknowledged to the quest to make programs into smart
require intelligence and skill. For example, in and useful aids to humans. For example,
195 65 7, Newell, Shaw, and Simons (195 7) this was done in the domains of symbolic
Logic Theory Program found two novel and algebra and calculus (e.g., the work of
interesting proofs to theorems in White- Hearn [Hearn, 1966]; and of Moses and
head and Russells Principia Mathematica; an MIT team [Moses, 1971]). Knowledge
in 195 75 8, Gelernters Geometry Theo- of mathematical specialists was sought
rem Proving Program showed superb per- and used, though no attempt was made to
formance in the New York State Regents separate the mathematical knowledge base
Examination in Plane Geometry; and by from the inference methods.
1963 Samuels Checker Playing program
had beaten one of the best checker play- The Emergence of the Expert Systems
ers in the United States (Samuel, 195 9). Focus in AI Research in the Period
Samuels work is especially interesting, given 196575
the expert-systems work that was to come,
Beginning in 1965 , the DENDRAL research
because he chose the components of the fea-
project (Lindsay et. al., 1980) at Stanford
ture vector used to evaluate the goodness of
University was exploring several big ques-
a board position by extensively interviewing
tions in AI using the experimental method of
master checker players.
modeling-by-programming. The aim of the
Some researchers have pursued general
project was to emulate the analytic expertise
methods of cognition that were relatively
of world-class chemists who could hypothe-
knowledge-free (e.g., Newell and Simons
size organic chemical structures from spec-
General Problem Solver, McCarthys advo-
tral data. The AI questions were similar to
cacy of methods from mathematical logic,
those mentioned earlier:
and Robinsons Resolution Theorem Prov-
ing Method that advanced that part of the 1. Could the methods-based approach of
science). In this line of research (weak meth- earlier AI work be augmented by domain-
ods), expertise is seen to reside in the specific knowledge to model human
power of reasoning methods such as search, expertise in difficult tasks of hypothesis
means-end analysis, backtracking, and ana- induction?
logical reasoning. Others experimented with 2. For programs that achieved expert levels
knowledge-rich programs (strong methods) of performance, what was the source of
in a quest for powerful behavior. With their power? Relatively speaking, was the
knowledge-rich programs, expertise is seen power in the knowledge used, or in the
to lie in the domain-specific and common- reasoning method used?
sense facts, assumptions, and heuristics in a 3 . How could the domain-specific knowl-
programs knowledge base: in the knowledge edge be represented in a way that was
lies the power. The reasoning methods in modular, easily understandable to both
these programs are quite simple, often lit- system-builders and end-users, efficient
tle more than modus ponens (If A, and A at the engineering stage of knowledge
implies B, then B). acquisition, and efficient at run time
Theorem proving was a major focus in when reasoning programs were using the
AI in the 1960s. It appeared to be a univer- knowledge?
sal method for solving problems in any task
4. Were there any new AI methods, or com-
domain. To some, it seemed that the main
binations of old methods, to discover in
problem of creating intelligent computers
relation to the induction task?
had been solved (Nilsson, 1995 ) because in
all of this early work, expert-level perfor- The task of analyzing data from a mass
mance was considered to be due more to the spectrometer on an unknown chemical sam-
methods than to the knowledge. ple was unusual in AI at the time because
However, the research focused on it was recognized to require expertise not
knowledge-based methods continued, in held by the programmer: it was performed
expert systems: a perspective from computer science 91
of good guessing (Polya, 195 4). Although according to the relations. One unit can also
Polanyi (1962) and others have asserted that represent knowledge that is a special case
much expertise relies on tacit knowledge of another unit, or some units can be parts
that cannot be articulated, the working view of another unit. Structured objects are espe-
of knowledge engineering is that tacit knowl- cially convenient for representing taxonomic
edge is explicable. knowledge and knowledge of prototypical
The knowledge representation formalizes cases.
and organizes the knowledge. One widely The problem solving model (or framework,
used representation is the production rule, or problem solving architecture, or paradigm)
simply rule. A rule consists of an antecedent organizes and controls the steps taken to
(IF part) and a consequent (THEN part), solve the problem. These problem solving
also called a condition and an action. The IF methods are built into program modules we
part lists a set of conditions in some logical earlier referred to as inference engines (or
combination. When the IF part of the rule is inference procedures) that use knowledge in
satisfied, the THEN part can be concluded, the knowledge base to form a line of reason-
or its problem solving action taken. ing. Whereas human experts probably use
A production rule is a somewhat broader combinations of these, and more, expert sys-
concept than a conditional sentence in logic. tems have been successful following each of
First, it may carry a degree of certainty these strategies singly.
that allows the program to draw a plausi- One common but powerful paradigm
ble conclusion that is less than certain from involves the chaining of IF-THEN rules to
premises that are themselves uncertain. Sec- form a line of reasoning. If the chaining
ond, both the condition and action parts may starts from a set of conditions and moves
name functions which may be primitive toward some conclusion, the method is
concepts in the task domain but complex called forward chaining. If the conclusion is
functions from an information-processing known (for example, a goal to be achieved)
point of view. These allow the program to but the path to that conclusion is not
check whether the result of the condition known, then reasoning backwards is called
function is true (or true enough) and, if for, and the method is backward chaining.
it is, to execute the function in the action Data interpretation problems tend to call
rather than merely assert the truth of a state- for forward chaining. Diagnostic problems,
ment in logic. Expert systems whose knowl- however, often call for backward chaining
edge is represented in rule form are called because goals (and subgoals) direct the col-
rule-based systems (Buchanan & Shortliffe, lection of relevant data.
1984). The blackboard model of reasoning
Another widely used representation, (Engelmore & Morgan, 1988; Erman et al.,
called the structured object (also known as 1980) is opportunistic in that the order of
frame, unit, schema, or list structure) is based inferences in problem solving is dictated by
on a more passive view of knowledge. Such a the items that seem most relevant in the
unit is an assemblage of associated symbolic problem description, in the partial solution,
knowledge about an entity to be represented or in the knowledge base. This model can be
(Minsky, 1981) including its place in a taxo- used effectively to combine the judgments
nomic hierarchy, its most common proper- of multiple expert systems with specialized
ties, and its defining criteria. Typically, a unit knowledge in different parts of the problem.
consists of a list of properties of the entity Still another paradigm, which empha-
and associated values for those properties. sizes the power of experiential knowledge,
Since every task domain consists of many is case-based or analogical reasoning (Kolod-
entities that stand in various relations, the ner, 1993 ; Leake, 1996). In a case-based rea-
properties can also be used to specify rela- soning system, previously solved problems
tions, and the values of these properties are (cases) are stored in memory. A new prob-
the names of other units that are linked lem is matched against those and the closest
expert systems: a perspective from computer science 93
matches are retrieved to suggest solutions for themselves, without the assistance of knowl-
the new problem. edge engineers.
Two other properties of expert systems
The Tools Used are important and are commonly built into
the shell system: reasoning with uncertainty,
Today there are presently two ways to build and explanation of the line of reasoning.
expert systems. They can be built from Facts about people and things in the
scratch, or built using a piece of develop- world are almost always incomplete and
ment software known as a tool or a shell. uncertain. Expertise must include knowl-
edge and methods for dealing with facts
programming languages that are uncertain or missing altogether. An
expert, or expert system, may fill in reason-
The fundamental working hypothesis of
able defaults by looking at prototypes or by
AI is that intelligent behavior can be pre-
inferring plausible features from others that
cisely described as symbol manipulation and
are known. Or it may be possible to ignore
can be modeled with the symbol-processing
the missing information and deal just with
capabilities of the computer. In the late
available data. Knowing how to treat incom-
195 0s, special programming languages were
plete descriptions is a small, but important,
invented that facilitate this kind of mod-
part of high performance and expertise.
eling. The most prominent is called LISP
Inference is also typically uncertain few
(LISt Processing) and has been extensively
inferences outside of mathematics are abso-
used in expert-systems development. In the
lutely true. To deal with uncertain inference,
early 1970s another AI programming lan-
a rule may have associated with it a confi-
guage, called PROLOG (PROgramming in
dence factor or a weight. The set of meth-
LOGic), was invented in France. LISP has
ods for using uncertain knowledge in com-
its roots in one area of mathematics (lambda
bination with uncertain data in the reason-
calculus), PROLOG in another (first-order
ing process is called reasoning with uncer-
predicate calculus).
tainty. One important method for reason-
ing with uncertainty combines probability
shells, tools statements using Bayes Theorem to infer
Only a small number of AI methods have the probabilities associated with events or
been developed in enough detail to be outcomes of interest. Whereas Tversky and
useful in building expert systems. Currently, Kahneman (1974) have shown that even
there are only a handful of ways in which to expert decision makers fail to combine prob-
represent knowledge, to make inferences, or ability statements rationally (i.e., according
to generate explanations. As a consequence, to Bayes Theorem and other laws of proba-
software infrastructure can be built that con- bility), a program makes no such errors. This
tains these useful methods and formalisms; helps emphasize the point that expert sys-
then the domain-specific knowledge tems are normative models of human exper-
model can be added. Such software tools tise as it ought to be applied, not descriptive
are known as shells, or simply AI tools (e.g., computational models of observed human
CLIPS, 2004). performance (with all of its foibles).
Building expert systems by using shells
offers significant advantages. A system can
be built to perform a unique task by enter- The Applications of Expert Systems
ing into a shell all the necessary knowl-
edge about a task domain. The inference Expert systems are widely used today as
engine is itself part of the shell. If the pro- surrogate experts and decision-making assis-
gram is not very complicated and if experts tants in business, manufacturing, and ser-
have had some training in the use of a vice industries, in health care, education,
shell, the experts can enter the knowledge finance, science, space exploration, and
94 the cambridge handbook of expertise and expert performance
defense (see also, Chapters 193 3 ). The ben- areas to which expert-systems technol-
efits of expert systems derive from a few ogy was applied (see e.g., Shortliffe 1976,
basic facts: Norman et al., Chapter 19), but diagno-
sis of engineered systems quickly became
1. Computers process information faster important commercialy. There are proba-
and more reliably than people. bly more diagnostic applications of expert
2. Computer software can be replicated systems (including telephone help desks
cheaply and easily. and equipment troubleshooting) than any
3 . Expertise is scarce. other type. The diagnostic problem can be
stated in the abstract as: given the evi-
Because of these facts, the major benefits
dence presenting itself, what is the under-
become:
lying problem/reason/cause?
1. A speed-up of human professional or
semi-professional work typically by a Planning and Scheduling
factor of ten and sometimes by a factor
Systems that fall into this class analyze a
of a hundred or more.
set of one or more potentially complex and
2. Improved quality of decision making. In interacting goals in order to determine a set
some cases, the quality or correctness of of actions to achieve those goals, and/or pro-
decisions evaluated after the fact show a vide a detailed temporal ordering of those
ten-fold improvement. actions, taking into account personnel,
3 . Major internal cost savings. Savings materiel, and other constraints (see also
within companies can result from quality Durso & Dattel, Chapter 20). This class has
improvement or more efficient produc- great commercial impact. Examples involve
tion of goods and information, and pro- airline scheduling of flights, personnel, and
vide a major motivation for using expert gates; cargo loading and unloading for mul-
systems. tiple ships in a port; manufacturing job-
4. Preservation of expertise. Expert systems shop scheduling; and manufacturing process
are used to preserve scarce know-how in planning.
organizations, to capture the expertise of
individuals who are retiring, and to pre- Configuration of Manufactured Objects
serve corporate know-how so that it can from Subassemblies
be widely distributed to other factories,
offices, or service centers of the company Configuration is historically one of the most
(Hoffman & Lintern, Chapter 12). important of expert system applications and
involves synthesizing a solution to a prob-
5 . Wide spectrum of applications. Applica-
lem from a set of elements related by a set
tions of expert systems find their way into
of constraints. Configuration applications
most areas of knowledge work and are as
were pioneered by computer companies as
varied as helping salespersons sell modu-
a means of facilitating the manufacture of
lar factory-built homes to helping NASA
semi-custom minicomputers (McDermott,
plan the maintenance of a space shuttle
1982). The technique has found its way into
in preparation for its next flight. In gen-
use in many different industries, for exam-
eral terms, the applications tend to cluster
ple, modular home building, telecommu-
into the following six major classes.
nications, manufacturing, and other areas
involving complex engineering design and
Diagnosis and Troubleshooting manufacturing.
of Devices and Systems
Financial Decision Making
This class comprises systems that diagnose
faults and suggest corrective actions for a The financial services industry has been a
malfunctioning device or process. Medical vigorous user of expert-system techniques.
diagnosis was one of the first knowledge Early applications were in credit card fraud
expert systems: a perspective from computer science 95
detection software. Advisory programs have substantial results (e.g., Chi et al., 1988;
been created to assist bankers in determin- Feltovich et al., 1997), whereas for others
ing whether to make loans to businesses and these issues are driving new research.
individuals. Insurance companies have used
expert systems to assess the risk presented by
Knowledge Representation
a customer and to determine a price for the
insurance. In the financial markets, foreign- In knowledge representation, the key top-
exchange trading is an important expert- ics are concepts, languages, and standards
system application. for knowledge representation (see also Chi,
Chapter 10; Hoffman & Lintern, Chap-
Knowledge Publishing ter 12). There are many issues involved in
scaling up expert systems: defining the prob-
The primary function of the expert system lems encountered in the pursuit of large
is to deliver knowledge that is relevant to knowledge bases; developing the infrastruc-
the users problem, in the context of that ture for building and sharing large knowl-
problem. Two widely distributed expert sys- edge bases; and actually accumulating a large
tems are in this category: an advisor that body of knowledge, for example, common-
counsels a user on appropriate grammatical sense knowledge or engineering and techni-
usage in a text, and a tax advisor that accom- cal knowledge. Moreover, expertise involves
panies a tax preparation program and advises an efficient organization of knowledge: a dis-
the user on tax strategy, tactics, and individ- parate collection of unrelated facts does not
ual tax policy. Note that in both cases the constitute expertise.
role of the system is to find and then present As with human experts, problem solving
the user with knowledge relevant to a deci- by computer requires an efficacious repre-
sion the user has to make. sentation of a problem and of the knowl-
edge needed to solve it (Davis et al., 1993 ).
Process Monitoring and Control IF-THEN rules, for example, seem natu-
Systems in this class analyze real-time data ral for stating the inferential knowledge
from physical devices with the goal of notic- needed to diagnose the causes of many med-
ing anomalies, predicting trends, and con- ical problems or for classifying loan appli-
trolling for both optimality and failure cor- cants into levels of credit risk. However,
rection. Examples of real-time systems that work on expert systems has shown that a
actively monitor processes can be found in single representation can be insufficient for
steel making, oil refining, and even the con- different tasks in the same domain (Clancey,
trol of space probes for space exploration 1985 ). For example, teaching about a sub-
(Nayak & Williams, 1998). ject domain requires different knowledge
and skills from solving problems in the
domain. Moreover, work has shown that dif-
ferent representations may be used equally
Issues about Expertise Arising from well for the same task in a domain (Aikins,
Work on Expert Systems 1983 ). Diagrams are known to be useful for
human problem solving (Polya, 195 4) but
As one would expect, the two main areas their use by computer is still only partially
for research on expert systems are also understood.
central issues in AI: knowledge representa- Experts knowledge is not homogeneous
tion and reasoning. In addition, three other and can be categorized along at least two
major lines of work take on extra impor- dimensions: formal versus informal knowl-
tance in dealing with expert systems: knowl- edge, and public versus private (Forsythe,
edge acquisition, explanation, and valida- Osheroff, Buchanan, & Miller, 1991). Knowl-
tion. Within each of these areas many issues edge encoded in textbooks and journals is
have been explored in both psychology and formal and public, heuristics shared among
AI; for some of them there have been members of a lab tend to be informal and
96 the cambridge handbook of expertise and expert performance
private. Paradoxically, when some private were narrow. Hence, their performance
knowledge (e.g., of how to get around an was brittle: when the boundary of a sys-
institutions rules) is made public, it loses tems knowledge was traversed, the sys-
its value (because administrators change tems behavior went from extremely com-
the rules). petent to incompetent very quickly (Davis,
Strategic knowledge is important because 1989a, 1989b). To overcome such brittleness,
of its power: experts use more efficient prob- researchers are now focusing on reasoning
lem solving strategies than novices. This from models, principles, and causal mecha-
capability is replicated to some extent in nisms. Thus, a knowledge-based system will
an expert system through meta-level knowl- not have to know everything about an area,
edge (Hayes-Roth et al., 1983 ). For example, as it were, but can reason with a broader base
MYCINs diagnostic strategy was predomi- of knowledge by using the models, the prin-
nantly backward chaining: starting with the ciples, and the causal mechanisms.
goal of recommending therapy for a patient As mentioned above, experts and expert
with an infection, MYCIN works backward systems must be able to reason under uncer-
to what it needs to know to do that recur- tainty (Tversky & Kahneman, 1974). Several
sively until the answers to what it needs to methods have been introduced for assessing
know can be found by asking a doctor or the strength of evidence and of the conclu-
nurse. This conveys a sense of purpose to the sions it supports within expert systems (e.g.,
doctor or nurse using the program. However, Zadeh, 1965 ; Pearl, 2002; Buchanan & Short-
MYCIN was also given meta-knowledge to liffe, 1984; Weiss et al., 1978; and Gordon &
direct the lines of reasoning even further, Shortliffe, 1985 ). One of the lessons learned
For example, to indicate the order in which from these investigations is that rough esti-
to pursue different goals. Meta-knowledge mates of uncertainty often support expert-
in the program, as with experts, also told level performance. Moreover, rough esti-
MYCIN whether enough information was mates do not create the illusion of knowing
available on a case to warrant a conclusion facts with more precision than they are actu-
or whether it had enough knowledge rel- ally known.
evant to a case to attempt solving it at all As a result of the line of research starting
(Davis, 1980). with the classic study by Simon and Chase
(1973 ), it is now recognized that expertise
is truly task specific and does not trans-
Knowledge Use
fer from one domain to another (see also
Knowledge, once codified, should be use- Feltovich et al., Chapter 4). Expertise depe-
ful for solving different kinds of prob- nds on well-organized, specialized knowl-
lems within different reasoning paradigms. edge much more than on either superior
Research on knowledge use, or problem memory skills (which would transfer) or
solving, involves the development of new general problem solving ability (which also
methods for different kinds of reasoning, would transfer).
such as causal models, analogical reason-
ing, reasoning based on probability theory
Knowledge Acquisition (KA)
and decision theory, and reasoning from case
examples. At present, each of these rea- Experience is a prerequisite to human exper-
soning paradigms uses a specialized repre- tise (Ericsson, Chapter 3 8). In expert sys-
sentation of knowledge even for the same tems, expertise gained through experience
problem domain. As with human prob- can be codified in the knowledge base
lem solvers, communication is difficult when (as rules and heuristics, definitions, tax-
programs are working in different concep- onomies, prototypes, etc.), in the statistics of
tual frameworks. prior associations, in a library of previously
The first generation of expert systems solved cases, and in a library of prototypi-
was characterized by knowledge bases that cal cases.
expert systems: a perspective from computer science 97
And they allow users to query and browse from MYCIN agreed or disagreed with
the knowledge base in order to see the scope their own recommendation for these cases.
and limits of what the system knows. Based on the number of times the outside
However, expertise in a program rests experts said that MYCINs recommendation
on implicit assumptions (Clancey, 1985 ). In was acceptable, compared with numbers
some contexts, for example, training, it is of acceptable recommendations among the
important to be able to convey the assump- practitioners, MYCINs performance was
tions and strategies, and even describe the found to be indistinguishable from that of
mechanisms on which they rest. Each task the local experts, and better than the perfor-
and problem solving context dictates the mance of the competent practitioners and
amount of detail that has to be made novice.
explicit. But there will always be unstated In this and other task domains for which
assumptions. there is no gold standard of correctness, using
acknowledged experts to judge the relative
expertise of an expert system has become a
Evaluation
widely used method of evaluation. Numer-
It became obvious that measuring a systems ous other methods for judging the appropri-
level of expertise by the size of its knowl- ateness and correctness of knowledge bases
edge base was misleading because the grain have also been proposed (see the bibliogra-
size of the primitive concepts used can vary phy in Buchanan, 1995 ).
widely. For example, in MYCIN the con-
cept of degree of sickness could either be
a primitive, whose value would be filled Future Directions and Main Result
in by a physician or nurse, or it could be Regarding Expertise
inferred from rules using the values of several
other primitives such as temperature and Although work on expert systems has eluci-
heart rate. dated many issues regarding expertise and,
Rather than measuring the size of the perhaps most important, has provided tools
knowledge base, MYCINs level of exper- for building testable models, can we say what
tise was measured through a series of eval- some important future directions are, and
uations that compared its performance to what the most important thing we have
that of humans (Buchanan & Shortliffe, learned is, from all of these experiments?
1984). Because expertise in medical diag-
nosis, as in all other areas, is not pre- Future Directions
cisely defined, MYCINs performance was
ranked by a panel of acknowledged out- Except for Internist (Pople et al., 1975 ) and
side experts against the performance of sev- a few other programs, most expert systems
eral persons, called the practitioners, whose have been narrow in the scope of their
presumed expertise ranged from novice (a domain because knowledge acquisition has
medical student) to competent practitioner been difficult and costly. A consequence is
(physicians without subspecialty training) to that continuing knowledge maintenance of
local expert (faculty providing the subspe- an expert system is also difficult and costly.
cialty training). The practitioners were asked Research directions in expert systems and,
to look at descriptions of randomly selected more generally, in AI are seeking to widen
cases of infections and provide therapy rec- the scope and size of KBs and facilitate
ommendations. MYCIN was given infor- knowledge acquisition.
mation from the same descriptions. Then
the panel of the outside experts was asked very large knowledge bases
whether each of the recommendations If knowledge is the source of power for intel-
from the practitioners and (anonymously) ligent systems (as we have argued), then
expert systems: a perspective from computer science 99
their organisations and their customers. systems depends on breaking the know-
Knowledge management systems (Smith & ledge-acquisition bottleneck and on codi-
Farquhar, 2000) enable the distribution of fying and representing a large knowledge
corporate-wide information and knowledge infrastructure (Chi, Chapter 10; Hoffman &
efficiently and effectively. Lintern, Chapter 12).
Berg, C. A. & Sternberg, R. J. (1992). Adults Davis, R. (1989b). Expert systems: How far
conception of intelligence across the adult life can they go? Part II. AI Magazine, 10, 65
span. Psychology and Aging, 7, 2 2 12 3 1. 67 .
Berners-Lee, Hendler, T., J., & Lassila, O. (2001). Davis, R. & King, J. (1984). The origin of rule-
The semantic web. Scientific American, 2 84, based systems in AI. In B. G. Buchana & E.
3 443 . H. Shortliffe (Eds.), Rule-based expert sys-
Borron, J., Morales, D., & Klahr, P. (1996). Devel- tems: The MYCIN experiments of the stanford
oping and deploying knowledge on a global heuristic programming project reading. MA:
scale. AI Magazine, 17 , 65 76. Addison-Wesley.
Boose, J. H. (1989). A survey of knowledge acqui- Davis, R., Shrobe, H. E., & Szolovits, P. (1993 ).
sition techniques and tools. Knowledge Acqui- What is a knowledge representation? AI Mag-
sition, 1, 3 95 8. azine, 14, 173 3 .
Buchanan, B. G. (1989). Can machine learn- Deep Blue (2005 ). Web site http://www.
ing offer anything to expert systems? Machine research.ibm.com/deepblue/.
Learning, 4, 2 5 12 5 4. Elstein, A. S., Shulman, L. S., & Sprafka, S. A.
(1978). Medial problem solving: An analysis of
Buchanan, B. G. (1994). The role of experimen-
clinical reasoning. Cambridge, MA: Harvard
tation in artificial intelligence. Philosophical
University Press.
Transactions of the Royal Society, 3 49,
15 3 166. Engelmore, R. & Morgan, T. (1988). Blackboard
systems. Reading, MA: Addison-Wesley.
Buchanan, B. G. (1995 ). Verification and vali-
dation of knowledge-based systems: A repre- Erman, L., Hayes-Roth, F., Lesser, V., &
sentative bibliography. http://www.quasar.org/ Reddy, D. R. (1980). The Hearsay-II speech-
21698/tmtek/biblio.html. understanding system: Integrating knowledge
to resolve uncertainty, ACM Computing Sur-
Buchanan, B. G. & Shortliffe, E. H. (Eds.) (1984).
veys, 12 , 2 13 2 5 3 .
Rule-based expert systems: The MYCIN exper-
iments of the Stanford heuristic programming Feigenbaum, E. A. & Feldman, J. (1963 ). Comput-
project. Reading, MA: Addison-Wesley. ers and thought. New York: McGraw-Hill.
Buchanan, B. G., Smith, D. H., White, W. C., Feltovich, P. J., Ford, K. M., & Hoffman, R. R.
Gritter, R. J., Feigenbaum, E. A., Lederberg, (Eds.) (1997). Expertise in context: Human and
J., & Djerassii, C. (1976). Application of arti- machine. Menlo Park, CA and Cambridge, MA:
ficial intelligence for chemical inference XXII: AAAI Press/MIT Press.
Automatic rule formation in Mass Spectrom- Forsythe, D. E., Osheroff, J. A., Buchanan, B. G.,
etry by means of the Meta-DENDRAL pro- & Miller, R. A. (1991). Expanding the concept
gram. Journal of the American Chemical Society, of medical information: An observational study
98, 6168. of physicians needs. Computers & Biomedical
Buchanan, B. G. & Wilkins, D. C. (1993 ). Read- Research, 2 5 , 1812 00.
ings in knowledge acquisition and learning. San Forsythe, D. E. & Buchanan, B. G. (1992) Non-
Mateo, CA: Morgan Kaufmann. technical problems in knowledge engineering:
Implications for project management. Expert
Chi, M., Glaser, R., & Farr, M. J. (Eds) (1988).
Systems with Applications, 5 , 2 03 2 12 .
The nature of expertise. Hillsdale, NJ: Erlbaum.
Goldstein, I. & Papert, S. (1977). Artificial intel-
Clancey, W. J. (1985 ). Heuristic classification.
ligence, language and the study of knowledge.
Artificial Intelligence, 2 7 , 2 893 5 0.
Cognitive Science, 1, 8412 3 .
CLIPS web site. (2004). http://www.ghg.net/
Gordon, J. & Shortliffe, E. H. (1985 ). A method
clips/CLIPS.html.
for managing evidential reasoning in a hierar-
Davis, R. (1979). Interactive transfer of expertise: chical hypothesis space. Artificial Intelligence,
Acquisition of new inference rules. Artificial 2 6, 3 2 3 3 5 7 .
Intelligence, 12 , 12 115 7 .
Hammond, T. & Davis, R. (2004). Automati-
Davis, R. (1980). Meta-rules: Reasoning about cally transforming symbolic shape descriptions
control. Artificial Intelligence, 15 , 1792 2 2 . for use in sketch recognition. Proceedings of
Davis, R. (1989a). Expert systems: How far can the Nineteenth National Conference on Artificial
they go? Part I. AI Magazine, 10, 6167 . Intelligence, USA, 45 045 6.
102 the cambridge handbook of expertise and expert performance
Hearn, A. C. (1966). Computation of algebraic Newell, A. & Simon, H. (1972). Human prob-
properties of elementary particle reactions lem solving. Englewood Cliffs, NJ: Prentice-
using a digital computer. Communications of the Hall.
ACM, 9, 5 73 5 77 . Nilsson, N. J. (1995 ). Eye on the prize. AI Maga-
Hoffman, R. R., Shadbolt, N. R., Burton, A. M., zine, 16, 917 .
& Klein, G. (1995 ). Eliciting knowledge from Pauker, S. P. & Szolovits, P. (1977). Analyz-
experts: A methodological analysis. Organiza- ing and simulating taking the history of the
tional Behavior and Human Decision Processes, present illness: Context formation. In W.
62 , 12 915 8. Schneider and A. L. Sagvall-Hein (Eds.), IFIP
Kolodner, J. (1993 ). Case-based reasoning. San working congress on computational linguistics in
Mateo, CA: Morgan Kaufmann. medicine (pp. 109118). Amsterdam: North-
Larkin, J., McDermott, J., Simon, D. P., & Simon, Holland.
H. A. (1980). Expert and novice performance Pazzani, M. J., & Brunk, C. A. (1991). Detect-
in solving physics problems. Science, 2 08, 13 3 5 ing and correcting errors in rule-based expert
13 42 . systems: An integration of empirical and
Leake, D. B. (Ed.) (1996). Case-based reasoning: explanation-based learning. Knowledge Acqui-
Experiences, lessons, and future directions. Menlo sition, 3 , 15 7173 .
Park, CA: AAAI Press /MIT Press. Pearl, J. (2002). Reasoning with cause, effect. AI
Magazine, 2 3 , 95 112 .
Lenat, D. & Feigenbaum E. A. (1987). On the
thresholds of knowledge. Proceedings of the Polanyi, M. (1962). Personal knowledge. Chicago:
Tenth International Joint Conference on Artificial University of Chicago Press.
Intellingence, Italy, 1173 1182 . Polya, G. (195 4). Mathematics and plausible rea-
Lenat, D. B. & Guha, R. V. (1990). Building soning (Vols. I & II). Princeton: Princeton Uni-
large knowledge-based systems: Representation versity Press.
and inference in the Cyc project. Reading, MA: Pople, H. E., Myers, J., & Miller, R. (1975 ).
Addison-Wesley. DIALOG: A model of diagnostic logic for inter-
Lindsay, R. K., Buchanan, B. G., Feigenbaum, nal medicine. Proceedings of the Fourth Interna-
E. A., & Lederberg, J. (1980). Applications of tional Joint Conference on Artificial Intellingence,
artificial intelligence for chemical inference: The USSR, 84885 5 .
DENDRAL project. New York: McGraw-Hill. Richards, D. & Compton, P. (1998). Taking up the
McDermott, J. (1982). A rule-based configurer situated cognition challenge with ripple down
of computer systems. Artificial Intelligence, 19, rules. International Journal of Human Computer
3 988. Studies, 49, 895 92 6.
Michie, D. (Ed.) (1979). Expert systems in Rutledge, G., Thomsen, G. E., Farr, B. R., Tovar,
the micro-electronic age. Edinburgh: Edinburgh M. A., Polaschek, J. X., Beinlich, I. A. Sheiner,
University Press. L. B., & Fagan, L. M. (1993 ). The design and
implementation of a ventilator-management
Minsky, M. (1981). A framework for represent- advisor. Artificial Intelligence in Medicine, 5 ,
ing knowledge. In J. Haugland (Ed.), Mind 6782 .
design: Philosophy, psychology, artificial intelli-
gence (pp. 95 128). Montgomery, VT: Bradford Samuel, A. (195 9). Some studies in machine
Books. learning using the game of checkers. IBM Jour-
nal of Research and Development, 3 (3 ) 211
Mitchell, T. Machine learning. New York: 229. Reprinted in (Feigenbaum & Feldman,
McGraw Hill, 1997. 1963 ).
Moses, J. (1971). Symbolic integration: The Scott, A. C., Clayton, J. E., & Gibson,
stormy decade. Communications of the ACM, E. L. (1991). A practical guide to knowledge
14, 5 485 60. acquisition. Boston: Addison-Wesley Long-
Nayak, P. & Williams, B. C. (1998). Model- man.
directed autonomous systems. AI Magazine, 19, Shanteau, J. (1988). Psychological characteristics
12 6. and strategies of expert decision makers. Acta
Newell, A., Shaw, J. C., & Simon, H. A. (195 7). Psychologica, 68, 2 03 2 15 .
Empirical explorations with the logic theory Shortliffe, E. H. (1976). Computer-based medi-
machine. Reprinted in Feigenbaum & Feldman cal consultation. MYCIN. New York: American
(1963 ). Elsevier.
expert systems: a perspective from computer science 103
Simon, H. A. & Chase, W. G. (1973 ). Skill in Weiss, S. M., Kulikowski, C. A., Amarel, S., &
chess. American Scientist, 62 1, 3 94403 . Safir, A. (1978). A model-based method for
Smith, R. & Farquhar, A. (2000). The road ahead computer-aided medical decision making. Arti-
for knowledge management: An AI perspec- ficial Intelligence, 11, 145 172 .
tive. AI Magazine, 2 1, 1740. Wilkins, D. C., Clancey, W. J., & Buchanan, B. G.
Tversky, A. & Kahneman, D. (1974). Judgment (1988). Knowledge base refinement by moni-
under uncertainty: Heuristics and biases. Sci- toring abstract control knowledge. International
ence, 185 , 112 4113 1. Journal of Man-Machine Studies, 2 7 , 2 812 93 .
VanLehn, K. (1996). Cognitive skill acquisition. Zadeh, L. (1965 ). Fuzzy sets. Information and
Annual Review of Psychology, 47 , 5 13 5 3 9. Control, 8, 3 3 83 5 3 .
CHAPTER 7
professions includes the analysis of how associated with the sociologies of work and
the discourse on professionalism operates occupation, where Anglo-American sociol-
at occupational/organizational (macro) and ogists began to differentiate particular occu-
individual/employee (micro) levels. pations (such as law and medicine) in terms
Section Two of this chapter is concerned of their aspects of service orientation and
with the sociology of science. Scientists are moral community, and hence their contri-
regarded as experts par excellence, and sci- bution to the stability and civility of social
ence is the expert system par excellence. systems. In Europe generally, the influence
From a sociological perspective, science as an of the study of work and occupations on
expert system is based on specific practices of the analysis of professions has been strong.
knowledge production that have gained social The focus has been wide, including occu-
and cultural authority. pational identity and socialization (Dubar,
Section Three deals with the relation- 2000), but also the analysis of professional
ship between experts and elites. Notions of elites or cadres (Gadea, 2003 ) and the
elite imply not only power, prestige, and consideration of the professions as employ-
privileges as key components, but also the ment in public sector organizations (Svens-
idea of excellence in a field of activity that son, 2003 ).
may be seen as an intersection with notions The study of the sociology of organiza-
of expert. From a political point of view, tions is also strongly influencing analysis of
expert power (Turner, 2001) is a problem professions because even the traditional pro-
because it violates the equality conditions fessions of law, and particularly medicine,
presupposed by democratic accountability. increasingly involve employment in work
We will have to ask: What role do experts organizations; hence, the differences in the
play in the formation and functioning of professional practitioners employment rela-
elites, and what role does expertise play in tions (compared with other employees) are
the acquisition of legitimacy and the estab- reducing or disappearing. Indeed it is some-
lishment of elite positions? times claimed that professions, as a special
As we will see, the golden thread run- (privileged) category of service-sector occu-
ning through the sociological discussion on pations, are in decline. Professions, as a cat-
experts is social closure (Murphy, 1988): egory, have been criticized as not being a
professions, sciences, and expert elites are generic occupational type (Crompton, 1990)
forms of exclusion, separating experts from and have been perceived as under threat
nonexperts. Sociology studies the structure, from organizational, economic, and politi-
culture, discourse, and action dimensions cal changes (e.g., Greenwood & Lachman,
underlying this process of social closure. 1996; Reed, 1996). Professions are portrayed
as experiencing a reduction in autonomy and
dominance (Freidson, 1988; Mechanic, 1991;
Allsop & Mulcahy, 1996; Harrison, 1999;
Professional Expertise: The Sociology Harrison & Ahmad, 2000); a decline in their
of Professional Groups abilities to exercise the occupational control
of work (Freidson, 1994); and a weakening of
One way of operationalizing and analyzing their abilities to act as self-regulating occu-
the concept of expertise in sociology is by pational groups (MacDonald, 1995 ), able
means of its formation and utilization in to enter into regulative bargains (Cooper,
different professional occupational groups. Lowe, Puxty, Robson, & Willmott, 1988)
This will be addressed in this section where with states.
the focus is the history, concepts, and theo- Many other researchers, often from
ries of the sociology of professional groups. non-Anglo-American societies, have argued
This intellectual field has a long and com- that knowledge-based occupations are the
plex history. It is clearly linked and closely expanding employment categories and the
professionalization, scientific expertise, and elitism 107
growth sectors of labor markets in developed tal bureaucracies. Marshall (195 0) empha-
(Lyotard, 1984; Perkin, 1988; Reed, 1996; sized altruism or the service orientation
Frenkel, Korczynski, Donoghue, & Shire, of professionalism and how professionalism
1995 ), transitional (Buchner-Jeziorska, 2001; might form a bulwark against threats to sta-
Buchner-Jeziorska & Evetts, 1997) and ble democratic processes.
developing societies (Hiremath & Guda- The best-known, though perhaps the
gunti, 1998; Sautu, 1998). This interpreta- most frequently misquoted, attempt to
tion has focused on the expansion of occupa- clarify the special characteristics of pro-
tions based on knowledge (Murphy, 1988), fessionalism and its central normative and
whether or not the concept of profession functional values was that of Parsons (195 1).
is used, and the growing capacity of higher Indeed, Dingwall has claimed (Dingwall &
education systems in most societies to pro- Lewis, 1983 ) that research in the sociology
duce workers who are educated and trained. of the professions is largely founded on
It is also the case in Europe that in the com- the contributions of Parsons, as well as the
mon market of the European Union (EU) work of Hughes. Parsons tried to clarify
there are changes in the political and eco- the importance of professionalism through
nomic environment for professions. There a theoretical base in the sociology of
are attempts both to harmonize professional knowledge, in terms of a socially-grounded
service provision, on the one hand, and to normative order (Dingwall & Lewis, 1983 ,
deregulate, on the other. In 2003 , the EU p. 2). Parsons recognized, and was one of
Commission invited the European profes- the first theorists to show, how the capitalist
sional federations to take part in the process economy, the rational-legal social order (of
of defining vocational qualifications for their Weber), and the modern professions were
members on a European level (Evetts, 2001; all interrelated and mutually balancing in
Evetts & Dingwall, 2002), which could refo- the maintenance and stability of a fragile
cus the emphasis on knowledge work as the normative social order. He demonstrated
new wealth of nations. The sociology of pro- how the authority both of the professions
fessional groups, however, has its own intel- and of bureaucratic organizations rested
lectual history. on the same principles (for example, of
functional specificity, restriction of the
power domain, application of universalistic,
The Early Years: Professionalism as
impersonal standards). The professions,
a Normative and Functional Value
however, by means of their collegial orga-
The earliest analyses and interpretations of nization and shared identity, demonstrated
professional groups tended to focus on and an alternative approach to the hierarchy
to utilize the concept of professionalism, of bureaucratic organizations, towards the
and for the most part these analyses referred shared normative end.
to professionalism as providing a normative Whereas Parsons distinguished between
value and emphasized its meanings and func- professions and occupations, Hughes
tions for the stability and civility of social regarded the differences between pro-
systems. fessions and occupations as differences
Durkheim (1992) assessed professional- of degree, rather than kind, in that all
ism as a form of moral community based occupational workers have expertise (Mieg,
on occupational membership. Tawney (1921) Chapter 41 relative experts). For Hughes
perceived professionalism as a force capable (195 8), professions and occupations not
of subjecting rampant individualism to the only presume to tell the rest of their society
needs of the community. Carr-Saunders and what is good and right for it, they also
Wilson (193 3 ) saw professionalism as a force determine the ways of thinking about prob-
for stability and freedom against the threat lems that fall in their domain (Dingwall
of encroaching industrial and governmen- & Lewis, 1983 , p. 5 ). Professionalism in
108 the cambridge handbook of expertise and expert performance
occupations and professions implies the our understanding of the power of particu-
importance of expertise but also trust in lar occupations (such as law and medicine,
economic relations in modern societies historically) or of the appeal of being a
with an advanced division of labor. In other professional in all occupational groups. It
words, lay people must place their trust no longer seems important to draw a hard
in professional workers (electricians and line between professions and occupations.
plumbers as well as lawyers and doctors) Instead, sociologists regard both as similar
and, as a result, some professionals acquire social forms that share many common char-
confidential knowledge. Professionalism acteristics.
requires professionals to be worthy of that Researchers now handle the definitional
trust, that is, to maintain confidentiality problem in different ways. Some avoid giv-
and to protect private knowledge and not ing a definition of profession and instead
exploit it for self-serving purposes. In return offer a list of relevant occupational groups
for this professionalism in relations with (e.g., Hanlon, 1998, claimed to be follow-
clients, professionals are granted authority, ing Abbott, 1988). Others have used the
rewards, and high status. disagreements and continuing uncertainties
about precisely what a profession is to dis-
miss the separateness of professions as a
Professions as Institutions:
field, although not necessarily to dispute
Defining the Field
the relevance of current analytical debates.
For a period in the 195 0s and 1960s, Crompton (1990), for example, considered
researchers shifted focus to the concept of how paradoxes and contradictions within
profession as a particular kind of occupation, the sociological debates about professions
or as an institution with special characteris- actually reflected wider and more general
tics. The difficulties of defining these spe- tensions in the sociologies of work, occupa-
cial characteristics and clarifying the differ- tions, and employment.
ences between professions and occupations Hence, professions can be analyzed as
have long troubled analysts and researchers. a generic group of occupations based on
For a period the trait approach occu- knowledge and expertise, both technical
pied sociologists who struggled to define the and tacit. Professions are essentially the
special characteristics of professional (com- knowledge-based category of occupations
pared with other occupational) work. For that usually follow a period of tertiary edu-
example, Greenwood (195 7) and Wilensky cation and vocational training and experi-
(1964) argued that professional work had ence. Another way of differentiating these
a number of characteristics: it required a occupations is to see professions as the
long and expensive education and training structural, occupational, and institutional
in order to acquire the necessary knowledge arrangements for dealing with work associ-
and skill; professionals were autonomous ated with the uncertainties of modern lives
and performed a public service; they were in risk societies. Professionals are extensively
guided in their decision making by a profes- engaged in dealing with risk, and with risk
sional ethic or code of conduct; they were assessment, and, through the use of expert
in special relations of trust with clients; and knowledge, in enabling customers and
they were altruistic and motivated by univer- clients to deal with uncertainty (also Mieg,
salistic values. In the absence of such charac- Chapter 41). To paraphrase and adapt a list
teristics, the label occupation was deemed in Olgiati, Orzack, and Saks (1998), profes-
to be more appropriate, and for occupations sions are involved in birth, survival, phys-
having some but not all of the character- ical and emotional health, dispute resolu-
istics, the term semi-profession was sug- tion and law-based social order, finance and
gested (Etzioni, 1969). credit information, educational attainment
The trait approach is now seen largely and socialization, physical constructs and
as inadequate in that it did nothing to assist the built environment, military engagement,
professionalization, scientific expertise, and elitism 109
peacekeeping and security, entertainment section, however. Larson asked why and how
and leisure, and religion and our negotiations a set of work practices and relations that
with the next world. characterized medicine and law become a
rallying call for a whole set of knowledge-
based occupations in very different employ-
Professionalization: The
ment conditions. This question points to
Professional Project
the importance of the appeal and attraction
During the 1970s and 1980s, when soci- of the concept of professionalism to skilled
ological analysis of professions was domi- workers in all types of modern society.
nated by various forms of professionalism Another version of the professional-
as ideological theorizing and by the influ- ization as market closure has been the
ence of Marxist interpretations, one con- notion of professions as powerful occupa-
cept that became prominent was the profes- tional groups who not only close markets and
sional project. The concept was developed dominate and control other related occupa-
by Larson (1977) and included a detailed and tions, but also capture states and negotiate
scholarly historical account of the processes regulative bargains (Cooper et al., 1988)
and developments whereby a distinct occu- with states in the interests of their own prac-
pational group both sought a monopoly in titioners. Again, this was an aspect of theo-
the market for its service as well as status rizing about professions in Anglo-American
and upward mobility (collective as well as societies that began in the 1970s (e.g.,
individual) in the social order. The idea of a Johnson, 1972), was influenced by Marxist
professional project was developed in a dif- interpretations, and focused on medicine
ferent way by Abbott (1988), who exam- and law. It has been a particular feature
ined the carving out and maintenance of a of analyses of the medical profession (e.g.,
jurisdiction through competition, as well as Larkin, 1983 ), where researchers have inter-
the requisite cultural and other work that preted relations among health professionals
was necessary to establish the legitimacy of as an aspect of medical dominance as well as
a monopoly practice. gender relations (e.g., Davies, 1995 ).
Larsons work is still frequently cited, Since the mid-1980s, the flaws in the
and MacDonalds textbook on professions more extreme versions of the professional
(1995 ) continues to use and to support Lar- project have become apparent. Annandale
sons analysis in the examination of the (1998) has investigated aspects of medi-
professional field of accountancy. The out- cal dominance and has linked this with
come of the successful professional project diversity, restratification, and growing hier-
was a monopoly of competence legitimized archy within the medical profession itself
by officially sanctioned expertise, and a namely, only some doctors can become dom-
monopoly of credibility with the public inant, along with some nurses and some
(Larson, 1977, p. 3 8). Larsons interpreta- midwives. More generally, it has turned
tion has not gone unchallenged. Freidson out that governments could successfully
(1982) preferred market shelters to com- challenge the professions. Professions do
plete monopolies in characterizing the sometimes initiate projects and influence
provision of professional service, which indi- governments, but as often professions are
cated the incomplete nature of most market- responding to external demands for change,
closure projects. It is also the case that which can be political, economic, cultural,
Larsons careful analysis has been oversim- and social. This has resulted in a reap-
plified by enthusiastic supporters, such that praisal of the historical evidence, which is
some researchers talk about the professional still incomplete. One line of development
project, as if professions and professional has been the view that the demand-led the-
associations do nothing else apart from pro- ory of professionalization needs to be com-
tecting the market monopoly. One aspect of plemented by an understanding of the sup-
Larsons work is of particular interest in this ply side (Dingwall, 1996). Instead of the
110 the cambridge handbook of expertise and expert performance
question How do professions capture The claim is now being made (e.g.,
states? it is suggested that the central ques- Freidson, 1994, 2001) that professionalism
tion should be Why do states create pro- is a unique form of occupational control of
fessions, or at least permit professions to work that has distinct advantages over mar-
flourish? This has resulted in a renewed ket, organizational, and bureaucratic forms
interest in the interpretation of profession- of control. In assessing the political, eco-
alism as providing normative and functional nomic, and ideological forces that are exert-
values. It has also spawned new interest in ing enormous pressure on the professions
the historical evidence about the parallel today, Freidson (1994) has defended pro-
processes of the creation of modern nation- fessionalism as a desirable way of pro-
states in the second half of the 19th cen- viding complex, discretionary services to
tury and the development of modern pro- the public. He argues that market-based
fessions in the same period. It is suggested, or organizational and bureaucratic meth-
for example, that professions might be one ods impoverish and standardize the quality
aspect of a state founded on liberal princi- of service to consumers and provide disin-
ples, one way of regulating certain spheres centives to practitioners. Thus, professions
of risky life without developing an oppres- might need to close markets in order to be
sive central bureaucracy. able to endorse and guarantee the educa-
tion, training, expertise, and tacit knowl-
Professionalism: As Discourse edge of licensed practitioners, but once
of Occupational Control achieved, the profession might then con-
centrate more fully on developing service-
In the 1990s researchers began to reassess the oriented and performance-related aspects
significance of professionalism and its posi- (Halliday, 1987; Evetts, 1998). The process
tive (as well as negative) contributions for of occupational closure will also result in
customers and clients, as well as for social monopoly in the supply of the expertise and
systems. To an extent this indicates the same the service, and probably also to privileged
return to professionalism as normative and access to salary and status. However, as has
functional value, but in addition there are been noted, the pursuit of private interests
new directions in the analysis. is not always in opposition to the pursuit of
the public interest, and indeed both can be
reappraisal developed simultaneously (Saks, 1995 ).
One result of this return and reappraisal is In general, then, some recent Anglo-
a more balanced assessment of professional- American analyses of professions have
ism as providing normative value. In addi- involved the reinterpretation of the concept
tion to protecting their own members mar- of professionalism as a normative and func-
ket position through controlling the license tional value in the socialization of new work-
to practice and protecting their elite posi- ers, in the preservation and predictability of
tions, professionalism might also represent normative social order in work and occu-
a distinctive form of decentralized occupa- pations, and in the maintenance and sta-
tional control that is important in civil soci- bility of a fragile normative order in state
ety (see Durkheim, 1992). It has been argued and increasingly international markets. The
also that the public interest and professional result is now a more balanced and cautious
self-interest are not totally at odds and that appraisal in which, for example, a possible
the pursuit of self-interest may be compati- benefit is recognized in some professional
ble with advancing the public interest (Saks, groups wanting to promote professionalism
1995 ). Professionalism might work also to as normative value. This latest interpretation
confer distinct professional values or moral involves a reevaluation of the importance of
obligations that restrain excessive competi- trust in client/practitioner relations (Karpik,
tion and encourage cooperation (Dingwall, 1989), of discretion (Hawkins, 1992), of the
1996). importance of risk management (Grelon,
professionalization, scientific expertise, and elitism 111
1996), and of the value of expert judg- the analysis of how the discourse operates
ment (Milburn, 1996; Trepos, 1996). It also at both occupational/organizational (macro)
includes a greater valuing of quality of ser- and individual worker (micro) levels.
vice and of professional performance in the The occupational, organizational, and
best interests of both customers (in order to worker changes entailed by this new con-
avoid further standardization of service pro- ception have been summarized by Hanlon
vision) and practitioners (in order to protect (1999, p. 121), who stated that in short the
discretion in service work decision making) state is engaged in trying to redefine pro-
(Freidson, 1994). fessionalism so that it becomes more com-
mercially aware, budget-focused, manage-
new directions rial, entrepreneurial and so forth. Hanlon
A different interpretation of the concept of emphasized the state because he was dis-
professionalism is also developing, and this cussing the legal profession. When this
involves examination of professionalism as a analysis is applied to the use of the dis-
discourse of occupational change and con- course of professionalism in other occupa-
trol. This interpretation would seem to have tional groups, the state might be less directly
greatest relevance in the analysis of occupa- involved, and the service company, firm,
tional groups in organizations where the dis- organization, and perhaps pertinent regula-
course is increasingly applied and utilized. tory bodies would probably be the construc-
There is now extensive use of the concept tors, promoters, and users of the professional
of professionalism in an increasingly wide discourse.
range of work, occupational, organizational, It is necessary to clarify and operational-
and institutional contexts. It is used as a mar- ize the concept of discourse. Here dis-
keting slogan in advertising to appeal to cus- course refers to the ways in which work-
tomers (Fournier, 1999) and in campaigns to ers themselves are accepting, incorporating,
attract prospective recruits. It is used in com- and accommodating the concepts of profes-
pany mission statements and organizational sion, and particularly professionalism, in
aims and objectives to motivate employees, their work. It will also become apparent that
and also in policy procedures and manuals. in the case of many, if not most, occupational
It is an appealing prospect for an occupa- groups the discourse of professionalism is in
tion to be identified as a profession and for fact being constructed and used by the man-
occupational workers and employees to be agers, supervisors, and employers of workers,
labeled as professionals. The concept of pro- and it is being utilized in order to bring about
fessionalism has entered the managerial lit- occupational change and rationalization, as
erature and CPD (Continuing Professional well as to (self-) discipline workers in the
Development) procedures. The discourse conduct of their work. It is argued that this
of professionalism is also claimed by both use of the discourse is very different from the
sides in disputes and political and policy earlier (historical) constructions and uses of
arguments, and in disagreements between professionalism within medicine and law
practitioners and governments particularly from where the discourse originated.
with respect to proposed changes in fund- At the level of individual actors, the
ing and organizational and administrative appeal to professionalism can be seen as
arrangements within health and education a powerful motivating force of control at
(Crompton, 1990). a distance (Miller & Rose, 1990; Burchell,
In trying to account for such wide-ranging Gordon, & Miller, 1991). At the level of
appeal and attraction of the discourse of systems, such as occupations, the appeal
professionalism, a different interpretation is to professionalism can be seen also as a
required. It is suggested that professionalism mechanism for promoting social change. In
is being used as a discourse to promote and these cases, however, the appeal is to a
facilitate particular occupational changes in myth or an ideology of professionalism that
service work organizations. This includes includes aspects such as exclusive ownership
112 the cambridge handbook of expertise and expert performance
of an area of expertise, autonomy and dis- ple the medical profession historically) even
cretion in work practices, and occupational self-regulation or the occupational control of
control of work. However, the reality of work (Freidson, 1994).
the new professionalism is very different. The reality of professionalism in most ser-
The appeal to professionalism most often vice and knowledge-based occupational con-
includes the substitution of organizational texts is very different, however, and even
for professional values; bureaucratic, hierar- medicine and law in Anglo-American social
chical, and managerial controls rather than systems are no longer exempt. Fiscal crises
collegial relations; budgetary restrictions and have been features of most states, and such
rationalizations; and performance targets, crises have been explained by governments
accountability, and increased political con- as resulting from the rising costs of wel-
trol. In this sense, then, it can be argued that fare states and particularly social service pro-
the appeal to professionalism is in effect a fessionalism. Remedial measures to contain
mechanism of social control at micro, meso, the fiscal crises have been taken (sometimes
and macro levels. motivated, as in the UK, by a New Right
ideology), and these have included cut backs
in funding and increases in institutional effi-
The Sociology of Professional Groups:
ciency measures, as well as the promotion of
Theories and Results
managerialist/organizational cultures in the
When returning to the question of the professional public service sector (including
appeal of professionalism, it is necessary medicine).
to understand how professionalism as a Accountability and audit, targets, and per-
discourse is now being increasingly used formance indicators have now become fun-
in modern organizations, institutions, and damental parts of the new professionalism
places of work as a mechanism to facil- (Evetts, 2003 ). Professionals of all kinds and
itate and promote occupational change. the institutions in which they work are sub-
Why, and in what ways, have a set of ject to achievement targets to justify their
work practices and relations that historically receipt of public expenditure. These, in turn,
characterized medicine and law in Anglo- enable the performance of particular orga-
American societies resonated first with engi- nizations (such as schools, universities, and
neers, accountants, and teachers, and now hospitals), and the professionals who work
with pharmacists, social workers, care assis- in them, to be measured, assessed, and com-
tants, computer experts, and law enforce- pared. Accountability has been operational-
ment agencies in different social systems ized as audit. Work organizations specify
around the world? such targets and sometimes, by means of
The discourse of professionalism that is so devolved budgets, are requiring all bud-
appealing to occupational groups and their getary units to clarify and maximize income
practitioners includes aspects such as exclu- streams while controlling expenditures.
sive ownership of an area of expertise and It is also important to consider the appeal
knowledge, and the power to define the of professionalism as a discourse of disci-
nature of problems in that area, as well plinary control at the micro level. Fournier
as the control of access to potential solu- (1999, p. 290) has demonstrated how the
tions. It also includes an image of collegial reconstitution of employees as profession-
work relations, of mutual assistance and sup- als involves more than just a process of
port rather than hierarchical, competitive, relabeling, it also involves the delineation
or managerialist control. Additional aspects of appropriate work identities and poten-
of the discourse and its appeal are auton- tially allows for control at a distance by
omy in decision making and discretion in inscribing the disciplinary logic of profes-
work practices, decision making in the pub- sionalism within the person of the employee
lic interest fettered only marginally by finan- so labelled. In new and existing occupa-
cial constraints, and in some cases (for exam- tional and organizational contexts, service
professionalization, scientific expertise, and elitism 113
and knowledge workers and other employ- however, the discourse is being used to
ees are having to, and indeed choosing promote and facilitate occupational change
to, reconstitute themselves in organizational and as a disciplinary mechanism used by
and occupational forms that incorporate autonomous subjects to ensure appropriate
career development for the self-managing conduct. The discourse is grasped by the
and self-motivated employee (Grey, 1994; occupational group since it is perceived to
Fournier, 1998). In other words, those who be a way of improving the occupations sta-
as workers act like professionals are self- tus and rewards collectively and individu-
controlled and self-motivated to perform in ally. However, the realities of professional-
ways the organization defines as appropri- ism from above are very different.
ate. In return, those who achieve the targets When professionalism is constructed and
will be rewarded with career promotion and demanded from within, and it corresponds
progress. with a (supply-side) states willingness and
In trying to understand how the dis- perception that the delegation of profes-
course is used differently between occupa- sional powers is in the states best interest,
tional groups, it might be useful to turn to then the aspects of normative and functional
McClellands categorization (1990, p. 170) values of professionalism can be paramount
of professionalization from within (suc- in the discourse. The professional group con-
cessful manipulation of the market by the structs and controls the discourse that it con-
group) and from above (domination of tinues to use in its own as well as in the
forces external to the group). This catego- publics interest. The historically powerful
rization was intended to differentiate Anglo- professions of medicine and law have some-
American and German forms of profes- times demonstrated opposition to moral
sionalization, but instead it might be used conduct and appropriate behavior mech-
to indicate and explain the various usages anisms, however, particularly in their devel-
of, and indeed the appeal of, professional- opment of alternative interpretations of the
ism in different occupational groups. Where public interest.
the appeal to professionalism is made and The willingness by states to concede pro-
used by the occupational group itself, from fessional powers and regulatory responsibil-
within, then the returns to the group can ities (and for occupational groups to con-
be substantial. In these cases, historically, the struct and demand professionalism from
group has been able to use the discourse in within) is now universally in decline. The
constructing its occupational identity, pro- consequence of this is still diversity in the use
moting its image with clients and customers, and construction of the discourse between
and in bargaining with states to secure and different occupational groups although
maintain its (sometimes self-) regulatory this diversity might be in decline. The legal
responsibilities. In these instances the occu- profession now (in contrast to medicine)
pation is using the discourse partly in its own is perhaps the best example of an occupa-
occupational and practitioner interests, but tional group in a relatively privileged posi-
sometimes also as a way of promoting and tion and still able to construct profession-
protecting what it would claim to be the alism from within. There are, however,
public interest. numerous occupational groups within the
In the case of most contemporary ser- profession of law, and in general it is those
vice occupations, however, professionalism occupations categorized as social service law,
is being imposed from above, and for the rather than entrepreneurial law professions
most part this means from the employ- (Hanlon, 1999), who are publicly funded.
ers and managers of the service organiza- Hence, the discourse is constructed and con-
tions in which these professionals work. trolled by others. The medical professions
Here the discourse of dedicated service and are similarly highly stratified and differen-
autonomous decision making are part of the tially powerful in the sense of being able
appeal of professionalism. In these cases, to construct and demand professionalism
114 the cambridge handbook of expertise and expert performance
from within. It is also interesting to observe expertise. Scientists are regarded as experts,
that the professional groups who are becom- and science is the prime example of an
ing powerful in international markets (for expert system with its own checks, valida-
example some accountancy and legal profes- tion procedures, recognition and authority
sions) might be in a better position to con- processes, and hence claims to legitimacy.
struct and demand professionalism from From a sociological perspective, science as an
within. expert system is based on specific practices
In summary, the sociological analysis of of knowledge production that have gained
expertise has always been closely linked with social and cultural authority. The sociologi-
the analysis of professions and professional- cal analysis of science as an institutionalized
ism. However, unlike in the past, it seems form and social practice has varied over time
that increasingly the discourse of profes- and offers different (sometimes contrasting)
sionalism is being used to convince, cajole, interpretations of expertise.
and persuade employees, practitioners, and
other workers to perform and behave in Scientific Experts: The Social
ways that the organization or the institution Study of Science
deem to be appropriate, effective, and effi-
cient. And professional workers are very
When trying to trace and understand the cre-
keen to grasp and lay claim to the norma-
ation and performance of scientific exper-
tive values of professionalism. But profes-
tise and the role of scientific experts from
sional expertise now needs to be measured,
the perspective of social studies of science, a
assessed, regulated, and audited. From a dis-
look from at least three different comple-
course controlled and constructed by practi-
mentary angles seems necessary. First, the
tioners, professionalism is now increasingly
question of the construction and protection
used in work organizations and occupations
of the boundaries between the science sys-
as an instrument of managerial control and
tem both as a knowledge system and as a
occupational change. The discourse includes
social territory and other forms of expertise
normative elements, but not in the sense
present in society need to be addressed. In a
of increasing occupational powers. Organi-
second step the social conditions and prac-
zational professionalism is very different in
tices of knowledge production on a more
control and relationship terms from the his-
micro-sociological level have to be consid-
torical and idealized image of the indepen-
ered as being an important manifestation of
dent, semi-autonomous practitioner of the
what gets defined as, or is (to be) under-
liberal professions very different from the
stood as, scientific expertise. Finally, the pic-
third logic analyzed by Freidson (2001).
ture will be completed by taking a gender-
It becomes even more important, therefore,
sensitive approach to the question of the
for sociologists to understand the appeal of
construction of expertise and experts.
professionalism in new and old occupations,
and how the discourse is being used to pro-
The Shaping of Scientific Expertise: A
mote and facilitate occupational change and
Historical Perspective
social control.
Professional groups have been one main In trying to understand the place of modern
form of the institutionalization of expertise science in contemporary societies, a closer
in industrialized countries, and the socio- look has to be taken at the processes and
logical analysis of professions has provided procedures that play a part in constructing
different, and sometimes contrasting, inter- the boundary around the territory labeled
pretations of professionalism and expertise science (Gieryn, 1995 ). This demarca-
over time. We now turn to the sociology tion is meant not only to delimit science
of science and consider the processes and from other forms of cultural knowledge-
procedures in science as an alternative form producing activities, but also to secure the
(to professions) of the institutionalization of authority of scientific expertise in the larger
professionalization, scientific expertise, and elitism 115
collective endeavor become visible in the partly still is, claimed has been put in ques-
laboratory. It is the individual that con- tion by feminists from the 1980s onwards.
tributes its creativity and intellectual capac- In their view, behind the very concept of
ities, while at the same time the collective is objectivity lies the idea of the sacrifice of
the setting in which research has to be real- the self for the collective, thereby, deliv-
ized, procedures and outcomes have to be ering knowledge that would go far beyond
negotiated, and results validated. the individual standpoint and could make
more powerful and far-reaching claims for
validity. However, the fact was overlooked
Scientific Expertise From a Gender
that these collectives represented possi-
Perspective
ble standpoints only in a rather selective
The third perspective from which the cre- way, namely, by excluding female actors to
ation and performance of scientific exper- a large degree (Keller, 1985 ). Recent exam-
tise/experts has to be considered is that of ples, such as the case study on the way grants
gender relations. Two aspects appear to be of were attributed by a medical research coun-
particular relevance. The first concerns the cil in Sweden, as well as the MIT report
question of scientific careers (Zuckerman, on the difficult position of women scientists
Cole, & Bruer, 1991) and the fact that women in this elite institution, clearly suggest the
even though they now have had access multiple and subtle mechanisms and values
to academic institutions for more than 100 that implicitly define not only who is to be
years are still largely underrepresented in regarded as a scientific expert, but also what
the group of scientific experts, at higher lev- kind of scientific expertise is worthy of sup-
els in particular. This fact holds even though port (Wenneras & Wold, 1997; Members of
numerous actions have been taken on the the first and second Committees on Women
policy level, both nationally and internation- Faculty in the School of Science, 1999).
ally, over recent years in order to improve
the situation. Without wanting to claim that
women can be regarded as a homogeneous Experts, Elites, and Political Power
group, one that would necessarily act and
need to be considered in standardized ways,
The existence of experts and expertise plays
it has so far remained unclear in what insti-
an important role in the constitution and
tutional environment working conditions,
functioning of elites. There is no standard
daily practices, and policies women could
definition of elite in social science. But cur-
attain significant opportunities to perform in
rent definitions generally have some core fea-
scientific careers. Drawing on studies of the
tures in common:
historical dimension of this exclusion pro-
cess (Schiebinger, 1989), it becomes obvi- r Elites are small groupings of persons who
ous how strongly scientific expertise and are endowed with a high degree of poten-
the expert role was and is intertwined with tial power.
power relationships within society (Rose, r This power may be due to the tenure of
1994; Haraway, 1989). In that sense keep- a formal position within an organization,
ing women (or any other group) out of sci- or it may be due to the charisma of a
ence would also mean keeping the power person.
over those societal domains where scientific r Being a member of an elite entails suc-
expertise plays an important and shaping cessfully passing through a process of
role. selection (Carlton, 1996; Dogan, 1989).
Second, gender has an impact also on the
epistemic level and thus on what counts The notion of elite has been introduced
as expertise and how it takes shape. The by the three Italian classics of sociology,
very way in which the universality and Pareto (193 5 ), Mosca (193 9), and Michels
objectivity of scientific knowledge was, and (1915 ), as an alternative concept to Marxist
118 the cambridge handbook of expertise and expert performance
egalitarian concepts. With reference to the tive field of activity by founding new profes-
ancient idea of aristocracy ( o = the sions. Professionalization in this strong sense
best), Pareto defined elites as those who are means that a group of experts claims juris-
most capable in any area of activity (193 5 , diction over the skills needed to be duly
2026 et seqq.). This ideal type defini- qualified to practice in the respective field.
tion of elite has a direct link to the notion In cooperation with state authorities, they
of expertise and experts. But in present- aim to transform their claims into a legal
day definitions of elite, the emphasis is restriction of access to the respective field
placed on power, not on excellence (Etzioni- of activity for people who have undergone a
Halevy, 2001). The power of elites is based certain vocational training, accounted for
on the possession and/or control of various by formal credentials. In short, groups of
resources or capitals. As Bourdieu (1984) experts strive to install mechanisms of social
puts it, inclusion and exclusion to protect certain
privileges against potential competitors.
economic capital: money; any tradable Rationalization thus brings about a shift
property; means of production. from collective mechanisms of social clo-
social capital: tenure of leading posi- sure that is, social exclusion on the basis of
tions in organizations; being inter- race, gender, religion, ethnicity, or language
locked in social networks supplying to individual mechanisms of social inclusion
informal support (Granovetter, 1973 ); and exclusion as a result of individual per-
(privileged) access to institutions of formance in standardized competitions on
training, sources of information, etc.; the basis of formal equality of opportunities
reputation. (Murphy, 1988). This shift towards a prin-
human capital: any esteemed knowl- ciple of merit changes the rules of repro-
edge and ability; charisma, ambition, duction of social standing within families:
stamina, etc. Parents who hold an elite position due to
professional expertise cannot bequeath this
We know three main historic mechanisms status directly to their children. They can
of transferring elite positions from one gen- only provide cultural capital that matches
eration to the next: heredity, charisma, and the requirements of the educational system
merit (Weber, 1979). Charisma ( ) and also mobilize financial and social capital
means gift out of (divine) favor and thus to improve the starting conditions of their
a qualification that cannot be generated sys- offspring. Statistically, these mechanisms of
tematically by training. Mainly in the sphere reproduction of elite positions due to exper-
of politics, it remains a source of legitimiza- tise are still quite successful (Bourdieu, 1984,
tion alternative to expertise, but in mod- and plenty of subsequent studies based on
ern democracies its function is restricted to this classic), but in many cases they fail the
being a (populist) ferment in the process of link between the social standing of parents
political decision making. In the course of and that of their children is no longer deter-
history the complexity of societies increased, ministic as it was to a large extent in pre-
and the skills needed for adequate gover- modern societies, but it has grown stochas-
nance and economic success grew more and tic with culturally specific biases. Thus, the
more demanding and specialized. Hence, for safeguarding of privileges usually associated
lack of selectivity towards skills, the princi- with elite positions based on expertise has
ple of heredity in elites became increasingly become two-stage: families try to reproduce
inappropriate and has largely been replaced access to any field of distinguished exper-
by a principle of merit, mainly based on tise in their children, and groups of experts
expertise (Elias, 1982). try to establish and have legally protected
In the course of rationalization of gov- privileges by placing emphasis on the func-
ernmental and economic functions, experts tional importance of their services for clients
try to monopolize the access to their respec- and society as a whole. It is characteristic
professionalization, scientific expertise, and elitism 119
for modern societies that these two contexts whole respond more or less to the demands
of reproduction are completely independent of the public, discontent with the state of
of each other. This disentanglement brings society remains a phenomenon of individu-
about an increase in societal rationality. It als and marginalized anomic groups and does
makes it possible for children of experts, who not give rise to upheavals apt to overthrow
want to reproduce the parental expert sta- the social order (Etzioni-Halevy, 1999). In
tus, not to be forced to choose the same order to maintain the capability to meet
field of expertise as their parents, but the public demands and an equilibrium between
field they are most talented for. This, on the public and particular interests, elites have to
other hand, allows for higher selectivity in admit talented members of the nonelite and
the staffing of elite positions. It fuels compe- to dispose of those doing damage to their
tition among aspirants and thus aggravates reputation by incapability, violation of pub-
the problem of reproduction of high social lic morality, and excessive parasitism on pub-
status in families. lic goods. This process of self-purification
In order to get a deeper insight into these is called circulation of elites (Kolabinska,
structures, it is adequate to go back to some 1912), and it is particularly characteristic for
postulates of the Enlightenment and the American elites (Lerner, Nagai, & Rothman,
French Revolution that amplified the func- 1996).
tional importance of expertise, namely, The emergence of counter-elites: Soci-
eties have to cope with unintended unde-
r Perfectability of societies: social struc-
sirable side effects of human activities and
tures are not an inalterable fate, but may with newly emerged natural dangers. In
be the subject of well-directed moulding many cases single experts or small groups
through progress in each field of human of experts first anticipate or perceive such
activity. a problem, make research on it, and try to
r Democracy: as a precondition of demo- initiate public discussions. But as it is hard
cratic control, governments and bureau- to call public attention to displeasing things,
cracies are accountable to the public for such problem awareness normally remains
their settlement of public affairs. confined for a long time to small circles
r Merit principle: privileges need legit- of specifically interested or heavily affected
imization through outstanding achieve- people. The issue will grow into a matter
ments in a field of activity. of public concern and an item of the polit-
r Equal opportunities: children should ical agenda only if it is picked up by the
have equal access to all educational insti- mass media most frequently on the occa-
tutions regardless of their social back- sion of an event apt to be scandalized. Seek-
ground, and their advancement within ing means for dissemination of their pre-
these institutions should depend exclu- monitions, experts may try and indeed
sively on their achievements. have often tried to incite and lead a social
movement or to support an already existing
As social developments since the 18th social movement by supplying expertise and
century show, there is a conflict between expert respectability. In the case in which an
the first two postulates: The idea of per- issue passes successfully through such a pro-
fectability of human affairs was a stimulus cess, marginalized views become common
inter alia for exceeding expansion and dif- sense, and formerly nameless or even ill-
ferentiation of expertise and its application. reputed experts may grow respectable and
But as expertise is not easily comprehensi- gain fame. During periods of controversy
ble for lay citizens, democratic control of the the apologist experts constitute a counter-
expanding activities of experts in contempo- elite to established elites that are still reluc-
rary societies is more questionable than ever tant to recognize the issue as a problem or
(Feyerabend, 1978; Etzioni-Halevy, 1993 ). the solutions recommended. Counter-elites
On the other hand, as long as elites as a play a decisive role in the generation of
12 0 the cambridge handbook of expertise and expert performance
cultural change in modern societies and as in the evolution of modern state societies:
an element of their checks and balances. The institutional frame of modern societies
If the issue as a matter of public concern has grown so widely ramified and differen-
gets undisputable and its solutions standard- tiated that only experts can overview its
ized, the counter-elite becomes a new estab- parts in full complexity each expert able
lished elite, and parts of old elites may be to focus on only one small section and
forced to resign (Imhof & Romano, 1994). propose advancements. Therefore, contem-
A prominent example of the long latency of porary states are forced to cede a great por-
a matter in circles of experts is ecology and tion of institutional change to experts. This
the green issue. The environmental move- makes democratic engagement and control
ment and subsequent social change created difficult (Turner, 2001), and in parts of the
the demand for environmental expertise to society it entails alienation and disorienta-
grow rapidly and provided a basis for a new tion that may result in social unrest. It is
elite of risk professionals (Dietz & Rycroft, impossible to foretell what this kind of social
1987). evolution will bring about in the long run,
On a global level the interplay between in particular with respect to the stability of
political, bureaucratic, and military elites societies, an area where the notion of exper-
of different states and the economic elites tise has its roots. But there is little doubt
of transnational corporations promotes pro- that experts and expertise will be highly in
cesses of globalization and the increase demand as long as modern societies keep
of societal complexity (Bornschier, 1989, evolving towards higher complexity, as they
1996). The prosperity of the members of have done in recent centuries despite the
political, bureaucratic, and military elites fact that there have always been antimod-
depends on the success of the economies ernist movements that challenge the role of
in their countries, and this success in turn expertise and experts in society by trying to
depends on the degree of legitimacy of the revert to simpler structures of understanding
social order, that is, on the extent to which and control, such as faith-based ones.
the social order meets the needs of the citi-
zens and fosters or hampers their vocational References
capabilities and achievement motivation. A
prominent example of a pact of bureau-
Abbott, A. (1988). The system of professions: An
cratic and economic elites is the constitution
essay on the division of expert labor. Chicago:
of the unified market in Europe (Nollert, University of Chicago Press.
2000): The plans for this giant project were
Allsop, J., & Mulcahy, L. (1996). Regulating
generated chiefly by expert bureaucrats of medical work: Formal and informal controls.
Brussels, that is, by an expert elite whose Buckingham: Open University Press.
members have no or only very indirect Annandale, E. (1998). The sociology of health and
democratic legitimization, whereas the role medicine. Cambridge: Polity Press.
of the democratically legitimized European Bazerman, C. (1989). Shaping written knowledge.
Council remained more or less confined to Madison: University of Wisconsin Press.
formal approval of already elaborated plans. Bornschier, V. (1989). Legitimacy and compara-
The coherence of plans and public tive success at the core of the world system.
projects may increase if they are worked European Sociological Review, 5 (3 ), 215 23 0.
out by experts without the fear of being Bornschier, V. (1996). Western society in transition.
voted out of their position. A disadvan- New Brunswick, NJ: Transaction.
tage of purely expert-driven projects is their Bourdieu, P. (1984). Distinction. Cambridge, MA:
tendency to evolve too far from common Harvard University Press.
sense and to jeopardize themselves through Buchner-Jeziorska, A. (2001). Price of knowledge:
insufficient responsiveness to public con- the market of intellectual work in Poland in the
cerns and objections. This example may be 90s. Paper presented in Professions Network of
taken as illustration for a general problem SASE Conference, Amsterdam, June 28July 1.
professionalization, scientific expertise, and elitism 12 1
Buchner-Jeziorska, A., & Evetts, J. (1997). Regu- Etzioni-Halevy, E. (1999). Inequality and the
lating professionals: The Polish example. Inter- quality of democracy in ultramodern society.
national Sociology, 12 (1), 6172. International Review of Sociology, 9(2), 23 925 0.
Burchell, G., Gordon, C., & Miller. P. (Eds.). Etzioni-Halevy, E. (2001). Elites: Sociological
(1991). The Foucault effect: Studies in gov- aspects. In N. Smelser (Ed.), International ency-
ernmentality. Hemel Hempstead: Harvester clopedia of the social & behavioral sciences
Wheatsheaf. (pp. 44204424). Amsterdam: Elsevier.
Carlton, E. (1996). The few and the many: A typol- Evetts, J., & Dingwall, R. (2002). Professional
ogy of elites. Brookfield, VT: Scolar Press. occupations in the UK and Europe: Legitima-
Carr-Saunders, A. M., & Wilson, P. A. (193 3 ). The tion and governmentality. International Review
professions. Oxford: The Clarendon Press. of Sociology, 12 (2), 15 9171.
Cooper, D., Lowe, A., Puxty, A., Robson, K., Evetts, J. (1998). Analysing the projects of profes-
& Willmott, H. (1988). Regulating the U.K. sional associations: National and international
accountancy profession: Episodes in the rela- dimensions. Unpublished paper presented at
tion between the profession and the state. ISA Congress, Montreal, 26 July 1 August.
Paper presented at ESRC Conference on Evetts, J. (2001). Professions in European and
Corporatism at the Policy Studies Institute, UK markets: The European professional fed-
London, January 1988. erations. International Journal of Sociology and
Crompton, R. (1990). Professions in the current Social Policy, 2 0 (11/12), 13 0.
context. Work, Employment and Society, Special Evetts, J. (2003 ). Professionalization and
Issue, 14766. professionalism: Explaining professional per-
formance initiatives. In H. A. Mieg & M.
Daston, L. J. (1992). Objectivity and the escape
Pfadenhauer (Eds.), Professionelle Leistung
from perspective. Social Studies of Science, 2 2 ,
Professional Performance: Positionen der Profes-
5 97618.
sionssoziologie (pp. 4969). Konstanz: UVK.
Davies, C. (1995 ). Gender and the professional
Feyerabend, P. (1978). Science in a free society.
predicament in Nursing. Buckingham: Open
London: NLB.
University Press.
Fournier, V. (1998). Stories of development and
Dietz, T. M., & Rycroft, R. W. (1987). The risk pro-
exploitation: Militant voices in an enterprise
fessionals. New York: Russell Sage Foundation.
culture. Organization, 5 (1), 5 5 80.
Dingwall, R., (1996). Professions and social order
Fournier, V. (1999). The appeal to professional-
in a global society. Plenary presentation at ISA
ism as a disciplinary mechanism. Social Review,
Working Group 02 Conference, Nottingham,
47 (2), 2803 07.
1113 September.
Freidson, E. (1982). Occupational autonomy and
Dingwall, R., & Lewis, P. (Eds.). (1983 ). The soci-
labor market shelters. In P. L. Steward & M. G.
ology of the professions: Doctors, lawyers and oth-
Cantor (Eds.), Varieties of work (pp. 3 95 4).
ers. London: Macmillan.
Beverley Hills: Sage.
Dogan, M. (1989). Pathways to power: Selecting
Freidson, E. (1988 [1970]). Profession of medicine:
rulers in pluralist democracies. Boulder, CO:
A study in the sociology of applied knowledge.
Westview Press.
Chicago: University of Chicago Press.
Dubar, C. (2000). La crise des identites: Freidson, E. (1994). Professionalism reborn:
Linterpretation dune mutation. Paris: Presses Theory, prophecy and policy. Cambridge: Polity
Universitaires de France. Press.
Durkheim, E. (1992). Professional ethics and civic Freidson, E. (2001). Professionalism: The third
morals. London: Routledge. logic. London: Polity.
Elias, N. (1982). The civilizing process, Vol. II: Power Frenkel, S., Korczynski, M., Donoghue, L., &
and civility. New York: Pantheon Books. Shire, K. (1995 ). Re-constituting work: Trends
Etzioni, A. (1969). The semi-professionals and their towards knowledge work and info-normative
organization: Teachers, nurses and social workers. control. Work, Employment and Society, 9(4),
New York: Free Press. 773 796.
Etzioni-Halevy, E. (1993 ). The elite connection: Gadea, C. (Ed.). (2003 ). Sociologie des cadres
Problems and potential of western democracy. et sociologie des professions: Proximites et
Cambridge, MA: Polity Press. paradoxes. Knowledge, Work & Society, 1, 1.
12 2 the cambridge handbook of expertise and expert performance
Michels, R. (1915 ). Political parties. New York: Sautu, R. (1998). The effect of the marketiza-
Hearsts International Library. tion of the Argentine economy on the labor
Milburn, P. (1996). Les territoires professionnels market: Shifts in the demand for university
et la negociation experte du reel: Compati- trained professionals. Paper presented at ISA
bilite des modeles theoriques. Paper presented Congress, Montreal, July 26Aug 1.
at ISA Working Group 02 Conference, Schiebinger, L. (1989). The mind has no sex?:
Nottingham, 1113 September. Women in the origins of modern science. Boston:
Miller, P., & Rose, N. (1990). Governing Harvard University Press.
economic life. Economy and Society, 19(1), 13 1. Shapin, S., & Schaffer, S. (1985 ). Leviathan
Mosca, G. (193 9). The ruling class. New York: and the air-pump. Princeton, NJ: Princeton
McGraw-Hill. University Press.
Mulkay, M. (1976). Norms and ideology in sci- Svensson, L. (2003 ). Market, management and
ence. Social Science Information, 15 , 63 765 6. Professionalism: Professional work and chang-
Murphy, R. (1988). Social closure: The theory ing organisational contexts. In H. A. Mieg &
of monopolization and exclusion. Oxford: M. Pfadenhauer (Eds.), Professionelle Leistung
Clarendon Press. Professional Performance: Positionen der Profes-
sionssoziologie (pp. 3 13 3 5 5 ). Konstanz: UVK.
Nollert, M. (2000). Lobbying for a Europe of big
business: The European roundtable of indus- Tawney, R. H. (1921). The acquisitive society.
trialists. In V. Bornschier (Ed.), State-building New York: Harcourt Bruce.
in Europe: The revitalization of Western Euro- Trepos, J. (1996). Une modelisation des juge-
pean integration (pp. 187209). Cambridge: ments dexperts: Categories et instruments
Cambridge University Press. de mesure. Paper presented at ISA Working
Olgiati, V., Orzack, L. H., & Saks, M. (Eds.). Group 02 Conference, Nottingham, 1113
(1998). Professions, identity and order in com- September.
parative perspective. Onati: The International Turner, S. (2001). What is the problem with
Institute for the Sociology of Law. experts? Social Studies of Science, 3 1(1), 123 149.
Pareto, V. (193 5 ). The mind and society: A treatise Weber, M. (1979). Economy and society (trans.
on general sociology (ed. by A. Livingstone). by G. Roth & C. Wittich). Berkeley, CA:
New York: Harcourt, Brace & Co. University of California Press.
Parsons,T. (195 1). The social system. New York: Wilensky, H. L. (1964). The professionalization
Free Press. of everyone? The American Journal of Sociology,
Perkin, H. (1988). The rise of professional society. 70(2), 13 715 8.
London: Routledge. Wenneras C., & Wold, A. (1997). Nepotism and
Reed, M. (1996). Expert power and control sexism in peer review. Nature, 3 87 , 3 413 43 .
in late modernity: An empirical review and Wynne, B. (1995 ). Public understanding of
theoretical synthesis. Organization Studies, science. In S. Jasanoff, G. E. Markle, J. C.
17 (4), 5 73 5 97. Petersen, & T. Pinch (Eds.), Handbook of
Rose, H. (1994). Love, power and knowledge: science and technology studies (pp. 3 613 88).
Towards a feminist transformation of the Thousand Oaks: Sage.
sciences. Cambridge: Polity Press. Zuckerman, H., Cole, J., & Bruer, J. (Eds.).
Saks, M. (1995 ). Professions and the public interest. (1991). The outer circle: Women in the scientific
London: Routledge. community. New York: Norton.
Part III
CHAPTER 8
William J. Clancey
Keywords: Ethnography, Workplace Study, ing activities over time in different cir-
Practice, Participant Observation, Ethnome- cumstances, we can study and document
thodology, Lived Work work practices, including those of proficient
domain practitioners. This chapter intro-
duces and illustrates a theoretical framework
Introduction as well as methods for observing work prac-
tices in everyday (or natural) settings in a
Expertise is not just about inference applied manner that enables understanding and pos-
to facts and heuristics, but about being a sibly improving how the work is done.
social actor. Observation of natural settings In the first part of this chapter, I explain
begins not with laboratory behavioral tasks the notion of work practices and the his-
problems fed to a subject but with how torical development of observation in nat-
work methods are adapted and evaluated by ural settings. In the middle part, I elabo-
experts themselves, as situations are expe- rate the perspective of ethnomethodology,
rienced as problematic and formulated as including contrasting ways of viewing people
defined tasks and plans. My focus in this and workplaces, and different units of analy-
chapter is on socially and physically located sis for representing work observations. In the
behaviors, especially those involving conver- final part, I present methods for observation
sations, tools, and informal (ad hoc) interac- in some detail and conclude with trends and
tions. How an observer engages with practi- open issues.
tioners in a work setting itself requires exper-
tise, including concepts, tools, and methods
for understanding other peoples motives What are Work Practices in
and problems, often coupled with methods a Natural Setting?
for work systems design.
By watching people at work in everyday Every setting is natural for the people who
settings (Rogoff & Lave 1984) and observ- frequent it. A laboratory is a natural work
12 7
12 8 the cambridge handbook of expertise and expert performance
setting for some scientists, whereas expe- a day in the life of the people and that
dition base camps are natural for others. means 24 hours, not the nominal work day.
The framework provided here is intended Thus, a study of work practices is actu-
to be applicable to any setting, includ- ally a study of a setting; this context makes
ing school playgrounds, churches, interstate the observed behavior understandable. For
highways, and so on. But we focus on example, consider understanding clowns:
workplaces, where people are attempting to
get some work done, for which they have If we had a film of a clown doing som-
been prepared, and have sufficient expe- ersaults, and nothing else (i.e., we knew
rience to be acknowledged as experts by nothing about circuses, about the history of
other people with whom they interact. This clowns and so on), then the film would not
can be contrasted with studies of every- tell us what we need to know to make sense
of what the clown was doing. . . . One would
day people being expert at everyday things
need to know something about how they
(e.g., jumping rope, car driving) or events are part and parcel of circuses, and how
purposely arranged by a researcher in a their somersaulting is viewed [by many
laboratory. observers] as a kind of sentimental self-
In studying natural settings, one views mockery. (Harper 2 000, pp. 2 442 45 ;
them broadly: Consider a teacher in a school attributed to Gilbert Ryle)
within a community, not just a classroom.
Seek to grasp an entire place, with its nested To understand a setting, it is useful to
contexts: Rather than focusing on a physi- view all workers (not just performers on
cian in a patient exam room, study the clinic, stage) as social actors. When we say that
including the waiting room. work is socially recognized as requiring spe-
Heuristically, one can view an experts cial skills or knowledge derived from exten-
performance as a play, identifying the stage, sive experience (Hoffman, 1998, p. 86), we
the acts, roles, and the audience. But also mean that people are visibly demonstrat-
view the play as having a history, whose ing competency, in how they make inter-
nature is changing in todays performance: pretations, conduct business, and produce
What are the actors long-term motives? results that are recognizably accountable
How is this performance challenging or (of agreeable quality) to institutional and
influenced by the broader community of prac- public audiences (Heritage 1984; Dourish &
tice (Wenger 1998) (e.g., other clinics and Button 1998). This perspective has the dual
nurses)? effect for expertise studies of considering the
Also inquire more locally about the worker as an agent who, with other agents,
chronology and flow of a performance: How coconstructs what constitutes a problem to
do people prepare, who assists them (think be solved and how the product will be eval-
of actors), how do they get information uated. Methods for applying this theoreti-
about todays work, when and where do cal perspective, called ethnomethodology, are
they review and plan their work, how are presented in this chapter.
events scheduled? Look for informal places Observing people in a natural setting is
and off-stage roles backrooms and prepara- commonly called fieldwork. Besides watch-
tion areas, dispatchers, janitors, and support ing and recording and asking questions, field-
personnel. All of this is part of the exper- work may include interviewing, studying
tise of getting a job done, and multiple parts documents, and meeting with the people
and contributions need to be identified if the being studied to analyze data together and
fundamental question about work is to be present findings (Forsythe 1999, p. 128).
answered: What affects the quality of this Fieldwork is most often associated with the
performance? What accounts for its success? broader method of study called ethnogra-
As a heuristic, to capture these contextual phy (Spradley 1979; Fetterman 1998 Harper
effects, one might frame a study as being 2000, p. 23 9), literally, the written study
observation of work practices in natural settings 12 9
1991). The work of studying knowledge and provided to subjects to read, to judge, to
learning moved to everyday settings such as type, or learn from. Expertise was viewed
supermarkets (Lave 1988), insurance offices, not about competence in settings (i.e.,
and weather bureaus (Hoffman et al. 2000). situated action), but decision making, rea-
The discipline of human-computer interac- soning, memory retrieval, pattern match-
tion (HCI) became a large, specialized sub- ing predominantly aspects of the assumed
field, a consortium of graphics artists, social internal, mental activity occurring in the
theorists, psychology modelers, and software brain. For example, a study of restaurant
engineers (Nardi 1996; Blum 1996; Kling & waiters (p. 27) was reduced to a study of
Star 1998). memory, not the lived work of being a
Broadly speaking, HCI research has pro- waiter. A study of typing concerned timing
gressed from viewing people as computer of finger movements, nothing about office
users that is, asking questions such as What work. Of the twelve studies of experts,
happens if people are in the loop? to view- only one included naturalistic observation
ing people, computers, documents, facilities, to fashion a relatively naturalistic task
and so on as a total system, and understanding (Lesgold et al. 1988; p. 3 13 ), namely, dictat-
the processes holistically. In some respects, ing X-ray interpretations.
this approach began with socio-technical This said, one of the most influential anal-
systems analysis in the 195 0s1970s (Corbet yses of the contextual aspects of behav-
et al. 1991, p. 9ff). Hutchins (1995 ) provides ior, Suchmans (1987) Plans and Situated
especially well-developed examples of how Actions, also did not involve the study of
tools, interfaces, and distributed group inter- practice. Suchman studied two people work-
actions constitute a work system. ing together who had never used a photo-
copier before (p. 115 ) a form of puzzle solv-
ing in which a predefined task is presented
Expertise in Context: Learning to See
in the real world (p. 114). Suchmans study
Observing and systematically studying a is an example of ethnomethodological analy-
work place is sometimes treated as easy by sis because it focuses on mutual, visible con-
non-social-scientists, who might perform the struction of understanding and methods, but
work sketchily or not actually analyze prac- it is not carried out using the ethnographic
tices (Forsythe 1999). The spread of the method (Dourish and Button 1998, p. 406)
anthropological and social perspectives to because this was not a study of established
cognitive science was at first limited, at best practices in a familiar setting.
shifting the analysis to include the social In summary, a participatory design pro-
context. For example, only one chapter in ject uses ethnography to study work prac-
Expertise in Context (Feltovich et al. 1997) tice, which may be analyzed from an
explicitly involved an observational study of ethnomethodological perspective (Heritage
a natural setting (Shalins video analysis in a 1984). More generally, ethnography may
hospital). Ericsson and Charness used diaries involve many other analytic orientations,
for studying violinists, without investigating emphasizing different phenomena, topics,
their home setting. Other researchers con- and issues (Dourish & Button 1998, p. 404).
sidered experts as socially selected (Agnew Ethnographic observation involves a rigor-
et al. 1997) and more broadly serving and ous commitment to confronting the worlds
part of market, organization, or commu- of people as they experience everyday life,
nity networks (Stern 1997); or viewed exper- to understand how problematic situations
tise as part of cultural construction (Collins actually arise and are managed. Workplace
1997, Clancey 1997). studies, contrasted with the study of knowl-
An edited volume from a decade ear- edge and experts in the 1970s and 1980s (Chi
lier, The Nature of Expertise by Chi et al. et al. 1988), signify a dramatic change in how
(1988), focused even more narrowly on expertise is viewed and studied, often with
mental processing of text: Documents were entirely different motivations, methods, and
132 the cambridge handbook of expertise and expert performance
process, and the interactions of work prac- in the mid-195 0s of jurors methodological
tice between scientists and rover engineers. issues:
Over time many scientists and managers
became informally involved in assisting the . . . such as the distinction between fact
observation and documentation process and and opinion, between what were enti-
refining design recommendations. tled to say, what the evidence shows, and
The researchers developed a naming con- what can be demonstrated . . . These dis-
tinctions were handled in coherently orga-
vention and ontology for objects on Mars,
nized and agree-able ways and the jurors
a prioritization scheme for planning rover
assumed and counted on one anothers
activities, and a method of documenting abilities to use them, draw appropriate
the scientific intent of telerobotic opera- inferences from them and see the sense of
tions to facilitate communication between them . . . common-sense considerations that
operations shifts and mission disciplines. anyone could see. (Heritage 1984, p. 4)
They trained MER scientists in these pro-
cedures and associated tools during simu- Ethnomethodology thus emphasizes the
lated missions. During the mission in 2004, commonsense knowledge and practices of
two researchers moved to Pasadena; six ordinary members of society, as they make
researchers rotated to cover the shifts that sense of, find their way about in, and act on
moved forty minutes later each day in syn- the circumstances in which they find them-
chrony with local Mars time. The team selves (p. 4). However, formalizing these
then developed operations concepts for an assumptions, values, and resulting proce-
extended mission phase, during which dures is not necessarily easy for people with-
scientists worked from their home cities, out training (Forsythe 1999).
and rover planning was compressed and Ethnomethodology has led researchers
simplified to reduce work on nights and to reconceive how knowledge and action
weekends. are framed, wresting . . . preoccupation with
Overall, this work systems design project the phenomenon of error prevalent in
helped define and enhance the telerobotic human factors research (Heritage 1984). The
scientific process and related mission sur- focus shifts to how people succeed, how
face operations, including design of facili- they construct the inherent intelligibility
ties for science meetings. Researchers con- and accountability of social activity, placing
tributed to the design of four computer new emphasis on the knowledge people use
systems used for rover planning and scien- in devising or recognizing conduct (p. 5 ).
tific collaboration that were being developed Button and Harper (1996) provide a cogent
simultaneously by NASA Ames colleagues example about how Decisions about what
and JPL. The MER work systems design and crimes are reported by police are intimately
the methods employed are influencing oper- tied up with questions of what is practical for
ations concepts and system architectures for the reporting officer and what is in the inter-
subsequent missions. ests of the police organization as a whole
(p. 275 ).
Contrasted with technical knowledge
(Schon 1987), this aspect of work methods
Ethnomethodologys Analytic is reflective and social, concerning how ones
Perspective behavior will be viewed, through under-
stood norms and social consequences. Eth-
In this section, I explain how the method- nomethodology thus provides a kind of
ology being studied in a workplace is not logical, systemic underpinning to how activ-
just a technical process for accomplishing a ity becomes coordinated how the actors
task, but incorporates social values and cri- come to share a common appraisal of their
teria for judging the quality of the work. empirical circumstances (Heritage 1984,
This idea originated in Garfinkels discovery p. 3 05 ) that is, the process by which they
134 the cambridge handbook of expertise and expert performance
come to cooperate and their methods for cal, setting-determined know-how in being
resolving conflicts. a recognizably competent social actor. Eth-
The idea of intelligibility and account- nomethodology reveals the reflective work
ability means that the work activity is of constructing observable (nonprivate) cat-
organized so that it can be rationalized egorizations (e.g., deciding which Mars rocks
(Dourish & Button 1998, p. 415 ), that is, to investigate). Thus, an essential task for
so that it appears rational. For example, the outside observer is to learn to see the
the Mars Exploration Rovers (MER) oper- ordered world of the community of practice:
ations (Squyres et al. 2004) were planned Human activity exhibits a methodical
and orchestrated by the science team so the orderliness . . . that the co-participants can
exploration could be recognizable to oth- and do realize, procedurally, at each and
ers in perpetuity as being science, especially every moment. . . . The task for the analyst
through the method of justifying instrument is to demonstrate just how they do this
applications in terms of hypothesis testing. (Whalen et al. 2004, p. 6). The following
In practice, geologists will often just strike section provides some useful frameworks.
a rock to see what is inside. In MER, the
application of the rock abrasion tool was
often explained within the group and to What People Do: Contrasting
the public as looking for something spe-
Frameworks
cific. As the mission continued on for many
months, the need for such rationalization
diminished, but as the scientists were bound Social-analytic concepts for understanding
at the hip, with one rover to command (at human behavior in natural settings are con-
each site), they continued to justify to each trasted here with information processing
other why they would hit a particular rock concepts that heretofore framed the study of
and not another something that would be knowledge and expertise (Newell & Simon
inconceivable in their activity of physically 1972).
walking through such a site with a hammer
and hand lens. Thus, the practice of geology Practice vs. Process
changed during the MER mission to adapt
to the circumstances of a collective, his- Practice concerns work as experienced by
torical, public, time-pressured activity; and those who engage in it (Button & Harper
production of accounts of what should be 1996, p. 264), especially, how recognizable
demonstrably scientific action were adapted categories of work are assembled in the real-
to fit this situation (cf. Dourish & Button time actions and interactions of workers
1998, p. 416). (p. 264), memorably described by Wynn
One must avoid a misconception that (1991):
technical knowledge is just being selectively
applied in social ways. Rather, what counts as The person who works with information
expertise the knowledge required to iden- deals with an object that is more diffi-
tify and solve problems reflectively devel- cult to define and capture than information
ops within the setting, which Collins calls the flow charts would have us imagine. These
mutual constitution of the social and con- show information in little blocks or trian-
ceptual (Collins 1997, p. 296). During the gles moving along arrows to encounter spe-
MER mission, a cadre of scientists and engi- cific transformations and directions along
the diagram. In reality, it seems, all along
neers capable of doing science with rovers
the arrows, as well as at the nodes, that
has developed new expertise and meth- there are people helping this block to be
ods of working across disciplines in a time- what it needs to be to name it, put it
pressured way. under the heading where it will be seen as
In summary, expertise is more than facts, a recognizable variant, deciding whether to
theories, and procedures (e.g., how to be a leave it in or take it out, whom to convey it
geologist or policeman); it includes practi- to. (pp. 5 65 7).
observation of work practices in natural settings 135
Button and Harper (1996, p. 265 ) give the organization, and so on. Scenarios used for
example of people analyzing interviews: studying expertise often represent an exper-
The coders would resort to a variety of prac- imenters idealized notion of the inputs,
tices to decide what the coding rules actu- and thus working a scenario may be more
ally required of them and whether what they like solving a contrived puzzle than inter-
were doing was actually (or virtually) in cor- acting with the flow of events that an actor
respondence with those rules. naturally experiences.
Practice is also called lived work what
work consists of as it is lived as part of orga-
Invisible vs. Overt Work
nizational life by those who do it (Button
& Harper 1996, p. 272). Practice is to be Observing work is not necessarily as easy
contrasted with formal process specification as watching an assembly line. Work may be
of what work is to be done. In the work- invisible (Nardi & Engestrom 1999, p. 2)
place itself, processes are often idealized and to an observer because of biases, because it
constitute shared values crimes should be occurs back stage, or because it is tacit,
reported to the bureau as soon as possible even to the practitioners. These three aspects
(p. 277). Narratives that people record or are discussed here.
present to authorities cater to these avowed First, preconceptions and biased methods
policies or preferences, creating an inherent may prevent the ethnographer from seeing
conflict in the work system between what what workers accomplish. For example, in a
people do and what they say they do. The study of telephone directory operators, the
point is not just that the documents and researchers a priori notion of the canon-
behavior may disagree, but rather, for exam- ical call rendered the variability of actual
ple, the records may reveal workers under- calls invisible and led to a poor design for
standing of how their practices must be rep- a partially automated directory assistance
resented to appear rational. system (p. 3 ). A related presumption is
Two fundamental concepts related to the that people with authority are the experts
practiceprocess distinction are behavior (Jordan 1992). For example, the Mycin pro-
function and activitytask. Process mod- gram (Buchanan & Shortliffe 1984) was
els (e.g., information processing diagrams) designed in the 1970s to capture the exper-
are idealized functional representations of tise of physicians, but no effort was made
the tasks that people in certain roles are to understand the role of nurses and their
expected to do. Practice concerns chrono- needs; in the study of medical expertise,
logical, located behaviors, in terms of every- nurses were non-persons (Goffman 1969;
day activities, for example, reading email, Star & Strauss 1999, p. 15 ).
meeting with a client, and sorting through The second aspect of invisibility arises
papers. Activities are how people chunk because many performers athletes, musi-
their day, how they would naturally describe cians, actors, and arguably, scientists keep
what I am doing now. Tasks are discovered, the arduous process of preparation for pub-
formulated, and carried out within activities, lic display well behind the scenes (Star &
which occur on multiple levels in parallel Strauss 1999, p. 21), which Goffman called
(Clancey 2002). back stage. One must beware of violat-
Putting these ideas together, one must ing autonomy or not getting useful informa-
beware of identifying a formalized scenario tion because of members strategic filtering
(cf. Feltovich et al. 1997, p. 117) with the or hiding of behavior (Star & Strauss 1999,
physical, interactive, social context in which p. 22).
work occurs. The work context is fun- The third form of invisible work is tacit
damentally conceptual (i.e., it cannot be articulation work work that gets things
exhaustively inventoried in descriptions or back on track in the face of the unexpected,
diagrams) and dynamically interpreted, in and modifies action to accommodate unan-
which the actor relates constraints of loca- ticipated contingencies. (Star & Strauss
tion, timing, responsibility, role, changing 1999, p. 10). These may be steps that people
136 the cambridge handbook of expertise and expert performance
take for granted, such as a phone call to a foam during launch) didnt classify foam
colleague, which they wouldnt necessarily damage on the prior mission, STS-112 in
elevate to being a method. December 2002, as an in-flight anomaly
Participatory design handles the various the established practice. Doing so could have
forms of invisible work by using ethnogra- delayed a subsequent mission in February
phy to identify stakeholders and then involv- that she would manage (CAIB 2003 , p.
ing them in the work systems design project 13 813 9). Thus, a recurrent consideration
(e.g., see the examples in Greenbaum & in how work is managed is what-this-will-
Kyng 1991). mean-for-me-later-on besides what-can-I-
do-about-it-now. The organizational con-
Members Documentation text of work, not just the facts of the case,
vs. Literal Accounts affects reporting a mishap event (Button &
Harper 1996, p. 277).
The production of documentation is part
In summary, the view of rationality as
of the lived work of most business, gov-
applying knowledge can be adapted to fit
ernment, and scientific professions. To deal
natural settings, but the goal of analysis must
with the nonliteral nature of documentation
include broad organizational factors that
mentioned previously, one should study the
include role, identity, values, and long-term
activity of reporting involved in sustaining
implications. The expert as agent (actor,
an account of the work as a formal sequen-
someone in a social setting) is more than a
tial operation (Button & Harper 1996,
problem solver, but also an expert problem
p. 272) as a situated action with social func-
finder, avoider, delegator, prioritizer, refor-
tions. For example, in Mars habitat simula-
mulater, communicator, and so on.
tions (Clancey 2002, in press), one can learn
We have now considered several con-
from daily reports what the crew did. But
trasts between information processing and
one must also inquire how the reporting
social-analytic concepts for understanding
was accomplished (e.g., contingencies such
human behavior in natural settings. But
as chores, fatigue, power failure, etc. that
how does one apply an analytic perspective
made reporting problematic), what account-
systematically?
ability concerned the crew (e.g., the public
image of the Mars Society; hence what was
emphasized or omitted), and why report- Unit of Analysis: The Principle of
ing was given such priority (e.g., to adhere Multiple Perspectives
to scientific norms). What people write may
not be what they actually did, and interviews A fundamental aspect of ethnography is to
may present yet another perspective on why triangulate information received from dif-
the reports even exist. ferent sources at different times, including
reinterpreting ones own notes in light of
Managing Inherent Conflicts later events, explicitly related to previous
vs. Applying Knowledge studies and analytic frameworks (Forsythe
One view of expertise is that people apply 1999, pp. 127128). In conventional terms,
knowledge to accomplish goals (Newell & to make a study systematic, one gathers data
Simon 1972). Yet, goals are not simply the to model the work from several related dif-
local statement of a task, but relate to long- ferent perspectives:
term social-organizational objectives, such r Flows: Information, communication,
as later work load and responsibilities
(Button & Harper 1996, p. 277). For exam- product
r Independent variables: Time, place, per-
ple, the chair of NASAs Mission Manage-
ment Team during the Columbia mission son, document
(which was destroyed on re-entry by wing r Process influences: Tool, organization/
tiles damaged by broken tank insulation role, facility, procedure
observation of work practices in natural settings 137
To provide a suitable social framing and orga- copy machine technicians] shares few cul-
nization of these data categories, this sec- tural values with the corporation; techni-
tion suggests the following units of analysis: cians from all over the country are much
activity system, temporality, and collectives. more alike than a technician and a sales-
person from the same district (Orr 1996,
Activity System p. 76).
How is the study of a collective related
Activity theory (Leontev 1979) provides to individual expertise? Lave (1988) con-
essential analytic concepts for understand- trasts the view that culture is a collection of
ing what is happening in a natural setting value-free factual knowledge with the view
(Lave 1988; Nardi 1996; Engestrom 2000). that society and culture shape the partic-
Psychologically, activity theory suggests how ularities of cognition and give it content
motives affect how people conceptually (p. 87). Thus, the study of culture is insepa-
frame situations and choose problem-solving rable from a study of how expertise is identi-
methods (Schon 1979). People broadly fied, developed, exploited, organized, and so
understand what they are doing as identity- on. Orrs study reveals that The technicians
related activities (e.g., exploring an Arctic are both a community and a collection of
crater as if we were on Mars). Career, social, individuals, and their stories celebrate their
or political motives and identities may influ- individual acts, their work, and their individ-
ence how procedures are interpreted and ual and collective identities (p. 143 ), such
tasks enacted. Engestrom (1999) provides an that storytelling has a social-psychological
exemplary activity theory analysis of a hos- function with many practical and institu-
pital setting. tional effects.
Rather, the observational work must be a sys- what, where, and how to study the setting
tematic investigation, with some sequential (Harper 2000, p. 248). For example:
order (though often dynamically replanned) r Map out the key processes of the organi-
that covers a related set of roles, places,
situations, and timelines. For example, in zation.
r Understand the diversities of work.
studying MER rover operations mentioned
previously, the researchers were confronted r Understand how different sets of persons
with a 24-hour operation in three floors depend on one another.
of a building, involving three shifts of dis- r Determine salient junctures in the infor-
tinct engineering and scientist teams. Given mation life cycle.
access constraints, the group focused on one
room at first, where the scientists met three A plan will specify particular kinds of records
times during the day and worked out from kept over a certain period, and how they
that group and place to understand how will be created, as described in subsequent
instructions were prepared for the rovers sections.
next-day operations and then how Mars data
was received and stored for the scientists Person, Object, Setting, Activity,
to access. Time-oriented Records
To stimulate inquiry and make learning To be systematic, the observer must deliber-
progressive, the observer should keep a jour- ately adopt a perspective and keep records
nal and review it periodically for issues to organized accordingly. Jordan (1996) sug-
revisit. Another method is to review pho- gests the perspectives person, object (e.g.,
tographs and ask about every object, What documents), setting, and task or process.
is that? What is it for? Who owns it? Where More generally, an activity-oriented record
is it used and stored? This can be done includes any recurrent behavior, including
effectively via email with colleagues who both predefined work tasks (e.g., processing
are not at the study site, encouraging them an order) and behaviors that may not be part
to ask questions about what they see in of a job description (e.g., answering a phone
the photos. call). Time is an orthogonal dimension. For
The ideal in participatory design is to find example, one could check to see what peo-
at least one person in the setting who can ple in a work area are doing every 15 minutes
be a champion for the inquiry, explaining or observe a given setting at the same time
the study to others, getting access, and mak- every day. Time-lapse video can be used to
ing the observational activity legitimate. By record when people enter and leave a partic-
this conception, people in the workplace are ular place (Clancey 2001).
partners in a cooperative activity, and not Anthropologists make a distinction
referred to as subjects, users, or opera- between two kinds of data: Emic categories
tors (Wynn 1991, p. 5 4). Probably no other (after phonemic) are used by participants;
philosophical stance is more fundamental to etic categories (after phonetic) are formal
the observers success. Data are discussed distinctions from an analysts perspective
with the workers (in appropriate forums); (Jordan 1996). The basic systematic units
report outlines are circulated for comment; mentioned in this section are etic: activities,
related local expertise responsible for mod- roles, objects, persons, places, durations,
eling the workplace is solicited for advice; etc.; in Western European and North
documents about the work may even be American business settings these often fit
coauthored with organizational champions. emic distinctions.
logistics and natural rhythm of the setting, interference. An audio mixer with several
technical complexity, and the studys pur- microphones enables combining different
pose. Generally speaking, long-term involve- sources (e.g., computer speech output,
ment is preferable to follow the devel- ambient remarks, radio or telephone con-
opment of work practice. However, a versations). Typically, observation reveals
few months of regular observation can be settings where interpersonal interaction
sufficient; a few weeks of daily participation occurs, from which one chooses hot spots
usually enables a proficient analyst to form (Jordan 1996) for systematic video record-
an understanding that can be a launching ing. The following methods are suggested:
point for more focused interviews and design Use a tripod and wide angle lens, and
sessions. Indeed, one aspect of a study is to multiple cameras for different view points
identify periodicities and historical develop- if possible. Take systematic photographic
ments, that is, to locate observations within records (e.g., the same place each day, such
overarching cycles and trends. as a whiteboard) or take a rapid sequence to
create a film strip that captures changing
postures and positions as people interact
Recording Methods and Logistics
with materials and each other. Interviews
Data from natural settings is recorded using can be audio recorded, but video (on a
tools varying from paper and pen to elec- tripod off to the side) provides more
tronic tracking devices. The standard media information.
are texts (e.g., field notes, documents found Written records can include a pocket
in the setting), video and audio recordings, notebook (for jotting down phrases or not-
photographs, and computer models (e.g., the ing things to do), a daily journal (often
Brahms work practice simulation system, handwritten) that describes ones personal
Clancey et al. 1998; Sierhuis 2001). Record- experience, and field notes (perhaps using
ing has enabled repeated and detailed an outline-based note-taker), with different
examination of the events of an interac- sections to elaborate on observations, raise
tion . . . permits other research to have direct questions, and interpret what is happening.
access to the data about which claims are Surveys given before, during, and after obser-
being made . . . can be reused in a variety vation are recommended. View a survey as a
of investigations and can be re-examined in way to prompt conversations and to encour-
the context of new findings (Heritage 1984, age people to reflect with you on what is
p. 23 8). Having a body of such data is the sine important, including their sense of account-
qua non for being a researcher who studies ability and how they evaluate their own per-
natural settings. formance (see Clancey, in press). Finally, if
Recordings must be labeled, indicating the circumstances of privacy and intellectual
at least the setting, date, and time. Expe- property allow, one may learn a great deal
rienced ethnographers suggest the follow- from documents found in garbage cans.
ing procedures: Collect photographs in a
computer catalog, where they can be sorted
Data Analysis
by categories into folders. Transcribe field
notes (not necessarily journals) in an elec- Experienced researchers suggest flexible use
tronic form so that they can be shared of computer tools for representing work
and searched. Organize computer files in (Engestrom 1999, pp. 85 90). Analysis
folders, separating preparatory/logistic infor- methods are detailed in the handbooks cited
mation, miscellaneous graphics, documents above. Key pointers are provided here.
acquired, photographs, field notes, presen- First, video data must be inventoried
tations and reports, press stories, email, and or will probably never be analyzed. Use a
so on. spreadsheet or outline to list the general con-
When recording outdoors, wireless tent for each recording, and as you watch
microphones can be used to avoid wind loosely transcribe material of special inter-
observation of work practices in natural settings 141
one reason or another failed to depict the their work the ideas presented here. The
core practices of the occupational worlds MER Human-Centered Computing ethnog-
which they have studied (Heritage 1984, raphy team from NASA/Ames included
p. 3 00). For example, the MER mission Roxana Wales, Charlotte Linde, Zara
study was limited in practice by the num- Mirmalek (University of California, San
ber of observers and their stamina. Out- Diego), Chin Seah, and Valerie Shalin
side the defined workflow of mission oper- (Wright State University). Topic sugges-
ations, the scientists were also participating tions and editorial advice for this chapter
in parallel activities of grant writing, pub- were also provided by Robert Hoffman,
lic affairs, paper preparation, and so forth. Patty Jones, Brigitte Jordan, Mike Shafto,
Some of these unobserved activities directly Maarten Sierhuis, Marilyn Whalen, and
affected science operations (e.g., prepar- Judith Orasanu. This work has been
ing a comparative graph for a conference supported in part by NASAs Comput-
presentation might require additional data ing, Communications, and Information
from Mars). Technology Program.
The role of simulation for driving obser-
vation and formalizing data is unclear
(Sierhuis 2001; Seah et al. 2005 ). As one
Footnote
delves into individual behaviors of spe-
cialists, are approaches recurrent or just
idiosyncratic? To what extent does a col- 1. The description of this project has been pro-
vided by Peggy Szymanski at PARC.
lective have uniform methods? Should a
simulation be broad (e.g., several weeks)
or deep (e.g., modeling computer system
interfaces)? References
Finally, social scientists, like other work-
ers, may find it difficult to articulate their Bernard, H. R. (Ed.). (1998). Handbook of meth-
own methods: There is no stable lore ods in cultural anthropology. Walnut Creek, CA:
of tried and trusted procedures through AltaMira Press.
which, for example, taped records . . . can be Beyer, H. & Holtzblatt, K. (1998). Contextual
brought to routine social scientific descrip- design: Defining customer-centered systems. San
tion (Heritage 1984, p. 3 01). Researchers Francisco: Morgan-Kaufmann.
often have different disciplinary interests, so Birdwhistell, R. (195 2). Introduction to kinesics.
a group of ethnographers at one site might Louisville: University of Louisville Press.
not collaborate until they write a report Blomberg, J., Giacomi, J., Mosher, A. & Swenton-
for the host organization. At this point, Wall, P. (1993 ). Ethnographic field methods
and their relation to design. In D. Schuler &
the problem of indexing and sharing data
A. Namioka (Eds.), Participatory design: Princi-
becomes visible, both within the group and ples and practices, (pp. 123 15 5 ). Hillsdale, NJ:
to others seeking to better understand a Erlbaum.
study. Effectively, in documenting observa- Blum, B. I. (1996). Beyond programming: To a
tional studies, the work practice researcher new era of design. New York: Oxford Univer-
is caught up in all the familiar issues of sity Press.
lived work, accountability, and contingent Buchanan, B. G., & Shortliffe, E. H. (Eds.).
methods. (1984). Rule-based expert systems: The MYCIN
experiments of the Heuristic Programming
Project. Reading, MA: Addison Wesley.
Acknowledgments Button, G., & Harper, R. (1996). The relevance of
work-practice for design. Computer Supported
I am especially indebted to my colleagues Cooperative Work, 4, 263 280.
at the Institute for Research on Learning Chi, M. T. H., Glaser, R., & Farr, M. J. (1988).
(19871997) for demonstrating through The nature of expertise. Hillsdale, NJ: Erlbaum.
observation of work practices in natural settings 143
Jordan, B. (1993 ). Ethnographic workplace stud- PPS (Project for Public Spaces). [December
ies and computer supported cooperative work. 2004] http://www.pps.org/info/placemaking
Proceedings of the interdisciplinary workshop on tools/placemakers/wwhyte.
informatics and psychology. Scharding, Austria, Rogoff, B., & Lave, J. (1984). Everyday cognition.
June. Amsterdam: North Holland. Cambridge: Harvard University Press.
Jordan, B. (1996). Ethnographic workplace stud- Rasmussen, J., Pejtersen, A. M., & Goodstein, L.
ies and computer supported cooperative work. P. (1994). Cognitive systems engineering. New
In D. Shapiro, M. Tauber, & R.Traunmuller York: Wiley.
(Eds.), The design of computer-supported cooper- Scheflen, A. E. (1972). Body language and the
ative work and groupware systems (pp. 1742). social order. Englewood Cliffs: Prentice-Hall.
Amsterdam: North Holland/Elsevier Science.
Schon, D. A. (1979). Generative metaphor: A
Jordan, B. (1997). Transforming ethnography perspective on problem-setting in social polity.
reinventing research. Cultural Anthropology In A. Ortony (Ed.), Metaphor and thought
Methods, 9, 1217, October. (pp. 25 4283 ). Cambridge: Cambridge Univer-
Jordan, B., & Henderson, A. (1995 ). Interaction sity Press,
analysis: Foundations and practice. The Journal Schon, D. A. (1987). Educating the reflective
of the Learning Sciences, 4, 3 9103 . practitioner: Toward a new design for teach-
Kling, R., & Star, S. L. (1998). Human centered ing and learning in professions. San Francisco:
systems in the perspective of organizational Jossey-Bass.
and social informatics. Computers and Society, Seah, C., Sierhuis, M., & Clancey, W. J. (2005 ).
28, 2229, March. Multi-agent modeling and simulation approach
Lave, J. (1988). Cognition in practice. Cambridge: for design and analysis of MER mission oper-
Cambridge University Press. ations. Proceedings of 2 005 international con-
Leontev A. N. (1979). The problem of activity in ference on human-computer interface advances
psychology. In Wertsch, J. V. (Ed.), The concept for modeling and simulation (SIMCHI05 ),
of activity in soviet psychology, Armonk, NY: M. ISBN 1-5 65 5 5 -288-1.
E. Sharpe, (pp. 3 771). Senge, P. M. (1990). The fifth discipline: The art
Lesgold, A., Rubinson, H., Feltovich, P., Glaser, and practice of the learning organization. New
R., Klopfer, D., & Wang, Y. (1988). Expertise York: Doubleday.
in a complex skill: Diagnosing x-ray pictures. In Sierhuis, M. (2001). Modeling and simulating
M. T. H. Chi, R. Glaser, and M. J. Farr, (Eds.), work practice. Ph.D. thesis, Social Science and
The nature of expertise. (pp. 3 113 42) Hillsdale, Informatics (SWI), University of Amsterdam,
NJ: Erlbaum, The Netherlands, ISBN 90-6464-849-2.
Lorenz, K. (195 2). King Solomons ring. New York: Spradley, J. P. (1979). The ethnographic interview.
Crowell. New York: Holt, Rinehart and Winston.
Luff, P., Hindmarsh, J., & Heath, C. (Eds.). Spradley, J. P. (1980). Participant observation. Fort
(2000). Workplace studies: Recovering work Worth: Harcourt Brace College Publishers.
practice and informing system design. Squyres, S. W., Arvidson, R. E., Bell III, J. F.,
Cambridge: Cambridge University Press. et al. (2004). The Spirit Rovers Athena sci-
Nardi, B. (1995 ). Context and consciousness: ence investigation at Gusev Crater, Mars, Sci-
Activity theory and human-computer interaction. ence, 3 05 , 794800.
Cambridge, MA: MIT Press. Star, S. L., & Strauss, A. (1999). Layers of silence,
Nardi, B., & Engestrom, Y. (Eds.) (1999). A web arenas of voice: The ecology of visible and
on the wind: The structure of invisible work. invisible work. Computer Supported Coopera-
Computer Supported Cooperative Work, 8(12), tive Work, 8, 93 0.
18. Suchman, L. A. (1987). Plans and situated
Newell, A., & Simon, H. A. (1972). Human prob- actions: The problem of human-machine com-
lem solving. Englewood Cliffs, NJ: Prentice- munication. Cambridge: Cambridge University
Hall, Inc. Press.
Orr, J. E. (1996). Talking about machines. An Vicente, K. J. (1999). Cognitive work analysis:
ethnography of a modern job. Cornell University Toward safe, productive, and healthy computer-
Press. based work. Mahwah, NJ: Erlbaum.
observation of work practices in natural settings 145
Websters Third New International Dictionary, Discourse and technology: Multimodal discourse
Unabridged. Merriam-Webster, 2002. http:// analysis. Washington, DC: Georgetown Uni-
unabridged. merriam-webster. com (13 Dec. versity Press (pp. 208229).
2004). Whyte, W. H. (1980). The social life of small
Wenger, E. (1998). Communities of practice: urban spaces. Washington, DC: The Conserva-
Learning, meaning, and identity. New York: tion Foundation.
Cambridge University Press. Wynn, E. (1991). Taking practice seriously.
Whalen, M., Whalen, J., Moore, R., Raymond, In J. Greenbaum & M. Kyng (Eds.),
G., Szymanski, M., & Vinkhuyzen, E. (2004). Design at work: Cooperative design of
Studying workscapes as a natural observational computer systems. Hillsdale, NJ: Erlbaum,
discipline. In P. LeVine & R. Scollon (Eds.), pp. 45 64.
CHAPTER 9
Psychometrics refers to the scientific dis- intellectual expertise. Finally, we will discuss
cipline that combines psychological inquiry challenges for future investigations.
with quantitative measurement. Though Before we start, some psychological
psychometric theory and practice pertain to terms need to be defined. The first terms are
all aspects of measurement, in the current traits and states. Traits refer to relatively
context, psychometric approaches to exper- broad and stable dispositions. Traits can be
tise pertain to the measurement and pre- physical (e.g., visual acuity, strength) or
diction of individual differences and group psychological (e.g., personality, interests,
differences (e.g., by gender, age) and, in intelligence). In contrast to traits, states
particular, high levels of proficiency includ- represent temporary characteristics (e.g.,
ing expertise and expert performance. The sleepy, alert, angry). The second set of
scientific study of expertise involves sev- terms to be defined are interindividual
eral important psychometric considerations, differences and intraindividual differ-
such as reliability and validity of measure- ences. Interindividual differences refer
ments, both at the level of predictors (e.g., in to differences between individuals, such
terms of developing aptitude measures that as the difference between the heights of
can predict which individuals will develop students in a classroom or the speed of
expert levels of performance), and at the different runners in a race. Intraindividual
level of criteria (the performance measures differences refer to differences within
themselves). We will discuss these basic individuals, such as the difference between
aspects of psychometric theory first, and the typing speed of an individual measured
then we will provide an illustration of psy- at the beginning of typing class and that
chometric studies that focus on the predic- same individuals typing speed at the end
tion of expert performance in the context of a year of practice in typing. Studies of
of tasks that involve the development and the development of expertise during skill
expression of perceptual-motor skills, and training can focus on interindividual dif-
tasks that involve predominantly cognitive/ ferences (e.g., the rank ordering of a group
147
148 the cambridge handbook of expertise and expert performance
ability, and skill assessments. This problem as baseball skill. There are three differ-
is learning that is, examinees often learn ent aspects of a test that need to be con-
either explicitly or implicitly during the sidered in evaluating validity: content valid-
test. Tests ranging in difficulty from simple ity, construct validity, and criterion-related
arithmetic problems to complex simulations validity. Content validity refers to the under-
typically show significant and sometimes lying content of the trait under considera-
substantial improvements in performance tion. For baseball, the content of the skill
from one occasion to the next, either be- would include batting, running, fielding, and
cause examinees have learned the correct other aspects. A test of baseball skill that
responses, or they have become more skilled focused on all relevant components of these
at performing the basic operations required tasks to the same degree that they are impor-
by the test. Under these conditions, a more tant would have high content validity. Gen-
appropriate method of assessing the reliabil- erally, content validity is established through
ity of the test is to use what is known as an judgments of subject-matter experts and
alternate form. An alternate form is typi- is not directly assessed in a quantitative
cally a test that is designed in a very similar fashion.
fashion to the first test, but one that differs Construct validity refers to the rela-
in terms of the actual items presented to the tionship between a measure of a particu-
examinee. When fatigue is not an important lar trait or state and the underlying the-
consideration, alternate forms of a test can oretical concept or construct. Establishing
be administered one right after the other. the construct validity of a measure usually
Otherwise, alternate forms can be adminis- involves evaluating the correlation between
tered after a delay, just like in the test-retest the measure and other assessments of the
procedure described above. same or similar constructs. Generally, in
A final type of reliability that is relevant to order for a measure to have high construct
the study of expertise is inter-observer relia- validity, it must correlate substantially with
bility. This is an index of agreement between other measures of the same or similar con-
different judges, when performance can- structs (this is called convergent validity),
not be objectively evaluated (e.g., gymnas- and the measure should not correlate sub-
tics, diving, art, music). When judges have stantially with measures of different con-
high agreement in rank-ordering individuals, structs (this is called discriminant valid-
there is high inter-observer reliability; but ity). For example, a measure of general
when there is little agreement, reliability of baseball skill might be expected to have
the judgements is low. high correlations with a measure of base-
Reliability of a measure is the first hur- ball strategy knowledge (convergent valid-
dle that must be passed for it to be scien- ity), but low correlations with a measure
tifically or practically useful. Without relia- of football strategy knowledge (discriminant
bility, a test has little or no utility. But, just validity).
having a consistent rank-ordering of individ- Especially important in terms of the
uals on a test says nothing about whether application of psychometric measures is
or not the test actually measures what it sets criterion-related validity. The key to cri-
out to measure. For that assessment, we have terion-related validity is prediction; it refers
the concept of validity. to the degree to which the measure can pre-
dict individual differences in some criterion
measure. For an intelligence test, criterion
Validity
validity is frequently demonstrated by the
Validity is a property of an instrument that degree to which scores on the intelligence
refers to whether it measures what it sets test correlate with a criterion of academic
out to measure. Thus, a test of baseball performance, such as grade point average or
skill is valid to the degree it actually pro- academic promotion from one grade to the
vides a measurement of the trait defined next. The typical application-oriented goal
150 the cambridge handbook of expertise and expert performance
phenomenon, the expected correlation performers who have the least to gain from
between initial performance and later per- task practice or training.
formance will be negative (simply as a func- It is important to emphasize that these
tion of the regression to the mean). An substantial changes in interindividual vari-
unsuspecting researcher might be tempted ability are typically found only for tasks
to conclude that a training program has the that are within the capabilities of nearly all
effect of leveling individual differences in learners. When tasks are complex or incon-
performance, in that the poor performers sistent in information-processing demands,
get relatively better and the good performers interindividual variability may not change
get relatively worse (McNemar, 1940). Ulti- over task practice (Ackerman, 1987, 1992;
mately, it is a bad idea to attempt to measure Ackerman & Woltz, 1994), or there may
individual differences in learning by corre- be a Matthew effect (e.g., see Stanovich,
lating initial performance with performance 1986). A Matthew effect refers to the phe-
after practice or training (e.g., see Cronbach nomenon of the rich getting richer, essen-
& Furby, 1970). tially a positive association between initial
A second issue related to measuring indi- standing and the amount of learning. (The
vidual differences in learning is the nature term derives from Jesus Parable of the Tal-
of interindividual variability during learn- ents, Matthew, XXV:29, For unto every
ing or skill acquisition. Because the mag- one that hath shall be given, and he shall
nitude of interindividual variability is asso- have abundance: but from him that hath
ciated with changes in the reliability and not shall be taken away even that which
validity of predictor measures, it is impor- he hath.) Such effects have been found in
tant to take account of factors that might reading skills and in other cognitive or intel-
lead to changes in interindividual variabil- lectual tasks (Lohman, 1999). For example,
ity. For skills that can be acquired by all expertise in mathematics is likely to show a
or nearly all learners, interindividual vari- Matthew effect, as many learners effectively
ability tends to decline with task practice drop out of the learning process at different
or training (e.g., see Ackerman, 1987). Fre- stages along the way to developing exper-
quently, the changes in variability can be tise (e.g., at the level of acquiring skill at
substantial. For tasks with substantial motor algebra, at calculus, or beyond). Across nor-
or perceptual-motor components, such as mal development, the differences between
typing or golf, there are extremely large experts and non-experts in mathematics will
interindividual differences in initial perfor- become more pronounced, which will be
mance, but after extensive training, per- manifest as larger interindividual variability
formance variability is much smaller. One in performance after practice or education.
reason for this is that there are physical lim- A few laboratory-based examples may
itations on performance at high levels of help illustrate the nature of the development
expertise. The most expert typist can type of expert performance with the context of
only as rapidly as one keystroke every 100 changing mean performance and changes
ms, and the most expert golfer is likely to in interindividual variability (as expressed
perform a handful of strokes under par. In in the between-individual standard devia-
contrast, there are few limits at the other tion of performance). Figure 9.1ac shows
end of the performance continuum there three different tasks, which are reasonably
are many more ways that an individual can well defined, but differ in the nature of the
perform a task poorly than there are ways task demands and the effects of practice on
that a task can be performed at an expert task performance. The first graph is from a
level. Thus, when it comes to comparing skill-acquisition experiment with a simpli-
the learning rates of a group of individu- fied air traffic controller (ATC) task (the
als, it is the poorest performing learners that Kanfer-Ackerman Air Traffic Controller
have the most to gain, and the highest initial task; Ackerman & Cianciolo, 2000; for a
152 the cambridge handbook of expertise and expert performance
Figure 9.1ac. Performance means and between-individual standard deviations over task practice.
Panel a. Kanfer-Ackerman Air Traffic Controller task (data from Ackerman & Cianciolo, 2000);
Panel b. Noun-Pair Lookup task (data from Ackerman & Woltz, 1994); Panel c. Terminal Radar
Approach Control task (data from Ackerman & Kanfer, 1993 ).
more extensive description of the task, see how mean performance increases quickly in
Kanfer & Ackerman, 1989). The task is diffi- the early sessions of practice, but becomes
cult for participants when they first perform asymptotic as most learners develop high
it. Cognitive/intellectual abilities (discussed levels of skills. Between-individual standard
below) are substantially related to individ- deviations start off high (when there are
ual differences in initial task performance. large differences between those learners who
However, the task has only seven rules, and easily grasp the task demands early in prac-
all task operations can be accomplished with tice, and those learners who must struggle to
just four different keys on the computer key- keep up), then decline as the slower learn-
board. As a result, within five or six hours of ers ultimately acquire the skills necessary to
practice, nearly all of the learners become perform the task at an expert level. At the
expert performers. Figure 9.1 Panel a shows end of six hours of practice, the magnitude of
psychometric approaches 153
between-individual standard deviations has trial level, but very quickly these individ-
changed from 9.08 to 6.5 5 , a reduction of uals get quite a bit faster than their scan-
about 28%. ner counterparts because they can retrieve
In the second example, a more high- the word pairs from memory much faster
fidelity air traffic control task was used. than it takes to scan the display. Thus, we
In contrast to the previous task, this called these individuals retrievers. Retriev-
one (called TRACON, for Terminal Radar ing the items from memory can be very
Approach Control; see Ackerman & Kanfer, fast, and expert retrievers performed about
1993 ) involves sustained and focused atten- twice as fast as the best scanners (e.g., about
tion, continuous sampling of the visual 5 00 msec/trial). Eventually (i.e., after sev-
radar screen, short-term memory, problem- eral hundred task trials), the scanners learn
solving abilities, and spatial visualization. at least some of the word pairs, almost inci-
There are many more commands to learn, dentally, and get faster at the task.
and each 3 0-minute task trial involves novel When one looks at the overall per-
configurations of airplanes that the learner formance and interindividual variability
must handle in real-time. Few participants (Figure 9.1c), there is a general mean im-
perform very well on the first few trials, provement in the speed of responding, as
and in general, it takes much longer to would be expected. However, there is an
acquire skills on TRACON than it does for initial increase in between-individual vari-
the simpler Kanfer-Ackerman ATC task. For ability as practice proceeds, since the retriev-
TRACON, many learners do not reach ers are getting much better at task perfor-
expert levels of performance, even after mance, and the scanners are profiting much
extensive task practice. Figure 9.1b shows less from each practice trial. Eventually,
that while mean performance markedly between-individual variability decreases, as
increases over 18 hours of task practice, there even the scanners begin to memorize the
is a slow rise in between-individual standard word pairs. That is, an initial SD level of
deviations in performance. From the initial 421 msec increases to 5 21 msec (an increase
to final practice sessions, standard deviations of 24%) before declining to 406 ms. At the
have changed from 3 .19 to 4.49, an increase end of 1,3 5 0 task trials, there has been an
of 41%. overall decline in between-individual SD of
The third example illustrates what hap- only 4%.
pens when learners adopt different learn-
ing or performance strategies. The task for
Restriction of Range
this example is a simple lookup task, where
the learner is presented with nine pairs of When interindividual variability declines
nouns on the upper part of a the computer with the development of expertise, a group
display, and a test probe (which either has of expert performers can be expected to
one of the matching pairs of words, or has show much smaller differences between
two words that do not match) on the lower them than do novices. The psychometric
part of the computer display (Ackerman & problem associated with such a restriction
Woltz, 1994). What happens in this task is in range of performance is that correlations
that some individuals simply look up the with measures of limited variability atten-
words on each task trial, which is a strategy uate (i.e., they get closer to zero). This
that minimizes effort on the individual trial can make it very difficult to find tests that
level. We called these individuals scanners. can predict individual differences in perfor-
Performing with this strategy rarely can be mance, simply because there is relatively lit-
accomplished in less than about 1 sec/trial tle variance to account for by the predictor
(1000 msec). Other individuals, however, measures. Of course, this makes betting on
work to memorize the word pairs while they the winner of competitive sports competi-
are also looking up the words for each trial. tions a highly speculative activity, whether
Their efforts are greater at the individual one is wagering money on the outcome
154 the cambridge handbook of expertise and expert performance
of the competition or predicting the rank scientist may depend on how many other
order of individuals when validating an abil- researchers are working toward the same
ity or personality test. In the final analysis, it goal getting to the goal a few days or
is much easier to predict which individuals months ahead of the competition may sig-
will develop expertise in a task that shows a nal the difference between fame and fortune
Matthew effect with practice than it is to on the one hand, and relative obscurity on
predict which individuals will develop the other hand (e.g., see the discussion by
expertise in a task that evidences a decline Watson, 2001, on the race to discover the
in interindividual variability. structure of DNA).
When expert performance is interdepen-
Base Rate Issues dent with the performance of others, the
ideal measurement of an individuals per-
In addition to restriction of range in perfor-
formance would be an average of multi-
mance, there is a more fundamental prob-
ple measures taken with as divergent a
lem for developing valid predictors of per-
set of other performers as possible. For
formance, namely, the problem of extreme
some types of expert performance, a round-
base rates (i.e., the rate at which a behavior
robin type tournament would be one means
is exhibited in the population). It has been
toward accomplishing this goal; however,
shown (e.g., Meehl & Rosen, 195 5 ) that as a
this kind of procedure is not practical in
behavior becomes less likely to occur (such
many different domains. Race car drivers
as when only 1 in 100 college athletes ulti-
do not compete in cars from all competi-
mately end up playing professional sports),
tor manufacturers, football players cannot
a test to predict the likelihood of reaching
be assigned willy-nilly to different teams
the professional teams must have extremely
every week, and research professors cannot
high validity to be practically useful. Thus,
be easily moved around from one institu-
when expert performance on a task is a rare
tion to another. When random assignment
phenomenon, it may not be practically feasi-
is not possible, more complicated statistical
ble to develop a selection measure that pro-
designs are needed to attempt to disentangle
vides valid inferences for which individual is
the effects of the team or other performers
going to succeed.
on the performance of the individual. Some-
times, however, this is simply impossible
Interdependence of Performance
to accomplish. Under these circumstances,
Another difficulty that arises in the study the only acceptable solution is to create
of expert performance in some tasks is that an artificial environment (such as a labora-
performance is not solely dependent on the tory experiment with simulations) in which
efforts of the individual performer. In many the individuals performance can be evalu-
occupations, ranging from sports (such as ated in the absence of other performers (Ed:
team efforts, or when there are other indi- see Ward, et al. Chapter 14, this volume).
vidual competitors, as in tennis or auto rac- Although these procedures can provide the
ing) to scientific discovery or technologi- needed experimental control, the risk is that
cal research and development, performance the performance measurements taken under
success depends to a nontrivial degree on the artificial laboratory conditions may not be
actions or behaviors of others, or depends on valid representations of the actual real-world
environmental influences outside of the con- task (e.g., see Hoffman, 1987).
trol of the performer (Ed: see Salas, et al.,
Chapter 25 , this volume). Thus, a baseball
players batting performance is dependent Trait Predictors of Expertise
on the skill level of the pitcher, perhaps
nearly as much as it does on the skill of the One of the most universal findings regarding
batter. Or, the scientific contribution of a individual differences in task performance
psychometric approaches 155
over practice or education is that as the their validity for predicting individual differ-
time between measurements increases, the ences in expert performance, or in the devel-
correlations between measurements attenu- opment of expertise.
ate, though the correlations rarely drop all
the way to zero. Sometimes, when the task
Cognitive Traits
is simple and skills are rapidly acquired, the
decline in correlations between initial task Perhaps the most pervasive evidence for
performance and performance on later task the validity of psychological measurements
trials is extremely rapid (e.g., see Ackerman, in predicting individual differences in the
1987; Fleishman & Hempel, 195 5 ; Jones, development of expertise is found for mea-
1962). When tasks are more complex, there sures of cognitive or intellectual ability (e.g.,
is still a pattern of declining correlations, but see Jensen, 1998; Terman, 1926). Cognitive
it is much less steep. Intelligence test perfor- ability measures can be very general (such
mance for children older than about age 5 , as IQ, or general intelligence); they can
for example, is very stable from one occasion be broad (such as verbal, numerical, and
to the next. Test-retest correlations with a spatial ability); or they can be quite spe-
lag of less than a year for an omnibus IQ test cific (such as verbal fluency, computational
are in the neighborhood of .90. Correlations math, or spatial visualization). From the first
with a lag of a long time, such as age 6 to introduction of the modern test of intelli-
age 18, are indeed lower (r = .80 or so, see gence (Binet & Simon, 1973 ), it has been
Honzik, MacFarlane, & Allen, 1948), but are clearly demonstrated that IQ measures can
still substantial. provide a highly reliable and highly valid
The important aspect of this general phe- indicator of academic success or failure. In
nomenon (which is called a simplex-like fact, over the past 100 years, IQ testing is
effect, after Guttman, 195 4; see Humphreys, probably the single most important applica-
1960) is that when the correlations are low tion of psychological research in the west-
between initial task performance and per- ern world. IQ tests have the highest validity
formance after extensive practice, the deter- for the purpose for which they were devel-
minants of initial task performance cannot be oped namely, prediction of academic per-
the same as the determinants of final task formance of children and adolescents. They
performance. The critical questions, from a provide significant and substantial predictive
psychometric perspective, are what are the validity here, but somewhat less so for pre-
trait predictors of initial task performance, dictions of adult academic and occupational
what are the trait determinants of expert performance. Narrower tests, such as ver-
level performance, and what is the difference bal, numerical, and spatial-content abilities,
between the two? when properly matched with the task to be
From the time of Immanuel Kant (e.g., predicted, can have somewhat higher validi-
1790/1987), philosophers and psychologists ties for adults than general intelligence.
have referred to three major families of However, the general pattern found
traits: cognitive, affective, and conative. across many different investigations is that
Cognitive traits refer to abilities, such as general and broad measures of cogni-
intelligence, or domain-specific knowledge tive/intellectual abilities are the most impor-
and skills. Affective traits refer to personal- tant predictors of performance early in train-
ity characteristics (such as impulsivity, con- ing or learning. When tasks are within the
scientiousness, extroversion). Conative traits capabilities of most performers, and declin-
refer to motivation, interests, or more gener- ing interindividual variability is observed,
ally will. In addition, there are other traits broad ability measures tend to show lower
that do not fit neatly into the tripartite break- validity for predicting performance over task
down, such as self-concept or self-efficacy. practice and instruction (e.g., see Ackerman,
We briefly discuss these families of traits and 1988; Barrett, Alexander, & Doverspike,
156 the cambridge handbook of expertise and expert performance
1992). That is, what appears to limit per- correlations with performance after exten-
formance early in task practice (i.e., for sive practice. If training is a long, expen-
novices) are the same abilities that are sive process (such as training individuals to
tapped by broad measures of intelligence, fly airplanes), it may make more sense to
such as memory and reasoning. Individual focus on using broad measures of cogni-
differences in these abilities can determine tive/intellectual abilities for selection pur-
how well a learner understands what is poses, so as to minimize the number of
required in the task situation, and how trainees that wash out of a training pro-
effective the learner is in forming strate- gram. If training is less involved (such as
gies and plans for task accomplishment. in the selection of fast-food service work-
But, as we mentioned earlier in connec- ers or grocery-store checkout clerks), it may
tion with interindividual variability, a learner be more effective for the organization to
who quickly grasps the essence of the task base selection on perceptual speed and psy-
has an advantage early in practice that dimin- chomotor measures to maximize the num-
ishes as slower learners eventually begin to ber of expert performers in the long run.
catch up over time. Skills such as driving a More elaborate selection procedures can be
car provide a good example of this kind of used, such as a multiple-hurdle approach.
learning situation. Some learners grasp the This procedure would provide tests of both
procedures of scanning the various instru- cognitive and psychomotor measures, and
ments and operating the controls quickly, applicants would be selected only if they
and others more slowly, but after a few pass a threshold score on both measures.
months of training and practice, the role of Such a procedure maximizes both the like-
reasoning and memory in determining indi- lihood of training success and the likelihood
vidual differences in performance is substan- of high levels of expert job performance.
tially diminished. When the tasks are not within the capa-
There has been some evidence to sug- bilities of many performers, or the task
gest that when tasks are relatively simple is highly cognitively demanding even after
and highly dependent on speed of percep- extensive task practice, general and broad
tion and response coordination, there is an content ability measures may maintain high
increase in the predictive validity of per- levels of validity for predicting individ-
ceptual speed and psychomotor abilities for ual differences in performance long after
task performance as expertise is developed training (e.g., for a discussion and examples,
(e.g., see Ackerman, 1988, 1990; Ackerman see Ackerman, 1992; Ackerman, Kanfer, &
& Cianciolo, 2000). That is, after extensive Goff, 1995 ). Most real-world jobs that are
practice where most individuals become rea- highly cognitively demanding have substan-
sonably skilled at the task (such as driv- tial domain knowledge prerequisites (e.g.,
ing a car or typing), performance is limited the jobs of air traffic controller, neurosur-
by more basic and narrow abilities (such as geon, software developer). One aspect that
visual acuity and manual dexterity). Under differentiates these tasks from other kinds
these circumstances, the best ability predic- of knowledge work are the strong demands
tors for individual differences in expert per- of the tasks in handling novel information.
formance may be those measures that are Expert performance on these tasks is thus
associated with the limiting determinants of jointly influenced by individual differences
performance, rather than those abilities that in domain knowledge and by broad intellec-
are associated with reasoning and problem tual abilities (both general and content abil-
solving. ities, such as spatial abilities for air traffic
When attempting to select applicants controllers).
for training or for job performance, one Although domain knowledge can partly
needs to take account of both the cogni- compensate for ability shortcomings when
tive/intellectual ability correlations with ini- memory and reasoning abilities decline with
tial task performance and the narrow ability age, world-class performance for such tasks
psychometric approaches 157
generally remains the province of relatively ogy (e.g., schizophrenia, endogenous depres-
younger adults (e.g., see Simonton, 1994). sion) is negatively correlated with develop-
In contrast, for jobs that are predominantly ment of expertise (all other things being
associated with domain knowledge rather equal), ceteris paribus, simply because these
than the ability to deal with novelty, domain patterns of personality are associated with
knowledge and skills appear to be relatively the ability to manage oneself in society. It
more influential than current levels of gen- is noteworthy, though, that there are many
eral and broad content abilities (e.g., see Chi, counterexamples of experts who have had
Glaser, & Rees, 1982). Such jobs include serious psychopathology (such as the Nobel
author, lawyer, radiologist, and so on. In Laureate mathematician, John Forbes Nash
such cases, the additional domain knowl- Jr., the Russian dancer Vaslav Nijinsky, Sir
edge obtained through experience more Isaac Newton, Robert Schumann, and many
than compensates for declines in general others). The unanswered, and perhaps unan-
abilities with age, at least into middle age, swerable question, is whether these and
and sometimes into early old age. Predictors other such individuals would have devel-
of expert performance in these jobs appear oped their respective levels of world-class
to be those measures that tap the breadth expertise if they had not suffered from these
and depth of relevant domain knowledge affective disorders.
and skills (e.g., see Willingham, 1974). For In the realm of normal personality traits,
a classification of job types along these lines, one of the most promising constructs for
see Warr (1994). predicting expertise has been need for
In general, across both motor-dependent Achievement (nAch), proposed by Murray
tasks and knowledge or cognitive tasks, the et al. (193 8). McClelland and his colleagues
key ingredient in maximizing the correla- (McClelland & Boyatzis, 1982; see Spangler,
tions between predictors and criteria is the 1992 for a review), performed several studies
concept of Brunswik Symmetry (Wittmann that provided various degrees of validation
& Su, 1999) named after Egon Brunswiks for nAch in predicting successful perfor-
Lens Model (Brunswik, 195 2). That is, the mance in a variety of different occupations,
content and especially the breadth of both but especially in the domain of manage-
predictor and criterion need to match. When rial success. Other personality traits (e.g.,
a criterion is relatively narrow (e.g., specific openness to experience, conscientiousness,
task performance or a component of task extroversion) have been moderately linked
performance), the best ability predictors will to success in several different occupations
be those that are matched in both content (e.g., see meta-analyses by Barrick & Mount,
(e.g., spatial, verbal, numerical, perceptual- 1991). However, in contrast to cognitive-
motor) and breadth (in this example, a rela- ability predictors of expertise, the direction
tively narrow criterion would merit develop- and magnitude of personality trait measure
ment of a relatively narrow ability battery for correlations with success appear to be more
prediction purposes). Thus, predicting the highly dependent on the occupational con-
typing speed of a typist is much more likely text. That is, even when some ability traits
to be better predicted from a dexterity test are not directly relevant to a particular job
(narrow) than an IQ test (broad). or task, correlations between ability pre-
dictors and criterion performance measures
are almost always positive, even if not par-
Affective Traits
ticularly substantial. In contrast, for exam-
Affective, or personality, traits represent an ple, extroversion may be reliably higher
area of great promise for prediction of the among experts in jobs that require inter-
development and expression of expertise, personal skills and leadership (e.g., politics,
but this area has little substantive evidence senior management), but the same trait may
to date. Generally speaking, one can readily be relatively lower for experts in domains
predict that serious affective psychopathol- that require intensive individual efforts (e.g.,
158 the cambridge handbook of expertise and expert performance
mathematician, chess player). As a result, cient for predicting which individuals will
one perhaps might not expect that there develop expertise.
will be particular personality traits that are The concept of occupational level
associated with expertise across divergent (Holland, 1997), which represents how
domains. much challenge an individual desires in the
task, is probably at least as important as
the direction of interests is to the predic-
Conative Traits
tion of expertise. Occupational level is con-
Some researchers have argued that the need sidered to represent a complex function of
for achievement (nAch) falls more in the both an individuals abilities and his/her self-
domain of conation or will instead of per- concept, which is the individuals estimation
sonality, but this issue illustrates one of the of his/her own abilities. There is probably
more enduring issues in the field of person- more to this construct than self-concept and
ality research and theory the problem of objective ability, in the sense that some indi-
parsing the sphere of individual traits, when viduals have both high aptitude for attain-
they do not really exist in isolation. nAch ing expertise, and have high estimation of
and many other conative traits, such as voca- their own aptitude, but lack the motivational
tional interests, have clear and sometimes drive to develop expertise. Kanfer (1987)
substantial overlap with personality traits has referred to this last component as the
(e.g., see Ackerman & Heggestad, 1997 for utility of effort, that is, the individuals
a review). desired level of effort expenditures in a work
In the 195 0s, vocational-psychology context.
researchers converged on a set of core
interest themes on which individuals relia- self-concept and self-efficacy
bility differ (e.g., see Guilford et al., 195 4; Self-concept is a relatively broad set
Holland, 195 9; Roe, 195 6). Perhaps the of constructs that parallel abilities (e.g.,
most widely adopted framework from this general intelligence, verbal, spatial, numer-
research has been Hollands RIASEC ical abilities, etc.). Self-efficacy refers to
model which is an acronym for six task-specific confidence in ones abilities
major vocational interest themes, namely: to accomplish particular levels of perfor-
Realistic, Investigative, Artistic, Social, mance (Bandura, 1977). From a Brunswik
Enterprising, and Conventional (e.g., see Symmetry perspective (Wittmann & Su,
Holland, 1997). It is possible to match these 1999), predictions of expert performance
vocational interest themes with characteris- from self-efficacy measures are likely to
tics of jobs, so that individuals can be guided show higher criterion-related validity for
by vocational counselors to occupations performing specific tasks, mainly because of
that best match their underlying interests. a closer match of breadth of predictor to
It is possible that one could identify areas breadth of criterion. For example, a self-
of expert performance within each of these efficacy measure might ask a golfer to pro-
different interest themes. There is a body vide a confidence estimate for making a spe-
of research that would support the notion cific putt, whereas a self-concept measure
that if individuals and jobs are matched might ask the golfer to provide an estimate
on these themes, individuals are more of his/her competence in putting, overall.
likely to develop expertise (along with job There is also a motivational component to
satisfaction) than if there is a mismatch self-efficacy that entails what an individ-
between the individuals interests and ual will do in a task, in addition to what
the job characteristics (e.g., see Dawis & the individual can do. Existing data sug-
Lofquist, 1984; Super, 1940). However, a gest that when the task is well defined, and
match between the direction of interests when individuals have some experience with
and job characteristics is not in itself suffi- a task, self-efficacy measures can provide
psychometric approaches 159
expert knowledge is obtained when there for practical predictive purposes. Although
is a confluence of high intellectual abili- the trait complex approach has yet to be
ties and high levels of affective and cona- explored in terms of predicting individ-
tive traits that are aligned with the par- ual differences in expertise within a sin-
ticular knowledge domain. When abilities gle job classification or task performance,
are moderate or low, but personality and this approach appears to have promise both
interests are well aligned with the knowl- for improving understanding of what fac-
edge domain, some compensation is possi- tors determine ultimate expert performance
ble through investments of greater time and achievement and for providing a small
effort. However, even when suitable abili- number of predictors that could be used
ties are high for a particular domain, lower diagnostically in expertise development
levels of matching personality and interests contexts.
will likely tend to preclude development of
expert levels of performance.
Classification Issues
In several studies (e.g., Ackerman, 2000;
Ackerman, Bowen, Beier, & Kanfer, 2001; One of the fields of psychometric appli-
Ackerman & Rolfhus, 1999) these trait cations that has been less explored out-
complexes (and a few others) have been side of vocational counseling and large-scale
shown to be useful predictors of individ- selection (e.g., military placement) is the
ual differences in domain knowledge among concept of classification. Whereas occu-
college students and middle-aged adults. pational/educational selection starts with a
Such results support the broader tenets of larger number of applicants than positions
the PPIK investment approach, but they to fill, and focuses on which candidates will
also show that the panoply of possible be the most likely to succeed, classification
trait predictors across cognitive, affective, starts with the assumption that most, if not
and conative variables could very well be all, of the applicants will be selected, and
reduced to a manageable set of complexes the goal is to match the applicant with the
psychometric approaches 161
domains. Ultimately, the classification goal abilities are often just as effective, if
is to maximize the congruence between the not more so, in predicting expert per-
individuals characteristics and the charac- formance after extensive task practice.
teristics of the job or educational program. When the tasks are straightforward and
accessible to most learners, cognitive
abilities generally show lower predictive
Discussion and Challenges validity as expertise develops.
for Future Research 2. For tasks that are predominantly based
on domain knowledge and skills, we
In this chapter we have reviewed how psy- reviewed some of the findings for vari-
chometrics plays an important role in mea- ous trait predictors of expertise. It appears
suring the development of expertise and that a heuristically useful approach
the prediction of individual differences in to understanding and predicting indi-
expert performance. Concepts of reliabil- vidual differences in the development
ity and validity are central to all aspects and expression of expertise in domain-
of quantitative psychological research, but knowledge tasks is one that focuses on
these concepts are too often implicit in the long-term investment of cognitive
experimental research, often to the detri- (intellectual) resources, through a small
ment of the usefulness of the research. In number of trait complexes (made up of
the study of individual differences, reliability cognitive, affective, and conative traits),
and validity are explicitly considered as inte- leading to differences in the breadth and
gral to both theory and application. Special depth of domain knowledge and skills.
issues of measuring change and the problems Two trait complexes (science/math and
associated with restriction of range in per- intellectual/cultural) appear to be facili-
formance were reviewed, as these present tative in the development of knowledge
challenges to many studies of expert perfor- about different domains, whereas other
mance. trait complexes (e.g., Social) may impede
Over the past century, there have been the development of traditional domains
hundreds of studies that have focused on of expert knowledge (e.g., academic and
predicting individual differences in the per- occupational knowledge).
formance of laboratory tasks, achievement in
educational settings, and occupational per- 3 . In addition to these two types of expert
formance. We have described only a few performance, there is another one that
illustrative examples of theory and empiri- has not received anywhere near the same
cal results from these investigations, as they level of attention namely, expertise in
relate to cognitive, affective, and conative interpersonal tasks. As noted by Hunt
traits. Two general sets of findings are noted (1995 ), in the United States, there has
below, along with a third domain of exper- been an increase in the number of jobs
tise that presents both opportunities and fur- that are highly dependent on interper-
ther challenges to theory and application, sonal skills mostly in the service indus-
as follows: tries (e.g., child care worker, customer
service representative) an increase that
1. For tasks that require significant percep- has been concomitant with declines in
tual and motor components, most of the the manufacturing and traditionally blue-
existing literature focuses on the effec- collar jobs. To date, there is too little avail-
tiveness of ability predictors of individ- able information on even how to iden-
ual differences in the development and tify and describe expert performance in
expression of expert performance. Gen- this domain. We can speculate that there
eral and broad cognitive abilities are most are affective and conative traits that may
effective in predicting success with novel be effective predictors of expertise in this
tasks, but perceptual and psychomotor domain. There is both historical (e.g.,
psychometric approaches 163
Ferguson, 195 2) and current research (e.g. overlapping traits. Psychological Bulletin, 12 1,
Barrick & Mount, 1991) that is consis- 219245 .
tent with this speculation. There is much Ackerman, P. L., & Rolfhus, E. L. (1999). The
more theory and research needed on the locus of adult intelligence: Knowledge, abil-
criterion side of the equation, along with ities, and non-ability traits. Psychology and
a need for additional predictors on the Aging, 14, 3 143 3 0.
cognitive/predictor side of the equation, Ackerman, P. L., & Kanfer, R. (1993 ). Integrat-
before it will be possible to evaluate how ing laboratory and field study for improving
well we can predict expertise in the inter- selection: Development of a battery for pre-
personal domain. dicting air traffic controller success. Journal of
Applied Psychology, 78, 413 43 2.
Ackerman, P. L., Kanfer, R., & Goff, M. (1995 ).
Cognitive and noncognitive determinants and
References consequences of complex skill acquisition.
Journal of Experimental Psychology: Applied, 1,
Ackerman, P. L. (1987). Individual differences in 2703 04.
skill learning: An integration of psychometric
Ackerman, P. L., & Woltz, D. J. (1994). Deter-
and information processing perspectives. Psy-
minants of learning and performance in an
chological Bulletin, 102 , 3 27.
associative memory/substitution task: Task
Ackerman, P. L. (1988). Determinants of individ- constraints, individual differences, and voli-
ual differences during skill acquisition: Cogni- tion. Journal of Educational Psychology, 86, 487
tive abilities and information processing. Jour- 5 15 .
nal of Experimental Psychology: General, 117 ,
Adams, J. A. (1987). Historical review and ap-
2883 18.
praisal of research on the learning, retention,
Ackerman, P. L. (1990). A correlational analy- and transfer of human motor skills. Psychologi-
sis of skill specificity: Learning, abilities, and cal Bulletin, 101, 4174.
individual differences. Journal of Experimental
Psychology: Learning, Memory, and Cognition, Bandura, A. (1977). Self-efficacy: Toward a uni-
16, 883 901. fying theory of behavioral change. Psychological
Review, 84, 191215 .
Ackerman, P. L. (1992). Predicting individual dif-
ferences in complex skill acquisition: Dynamics Barrett, G. V., Alexander, R. A., & Doverspike, D.
of ability determinants. Journal of Applied Psy- (1992). The implications for personnel selec-
chology, 77 , 5 98614. tion of apparent declines in predictive validi-
ties over time: A critique of Hulin, Henry,
Ackerman, P. L. (1996). A theory of adult intellec- and Noon. Personnel Psychology, 45 , 601
tual development: Process, personality, inter- 617.
ests, and knowledge. Intelligence, 2 2 , 22925 9.
Barrick, M. R., & Mount, M. K. (1991). The Big
Ackerman, P. L. (2000). Domain-specific knowl- Five personality dimensions and job per-
edge as the dark matter of adult intelligence: formance: A meta-analysis. Personnel Psychol-
gf/gc, personality and interest correlates. Jour- ogy, 1991, 44, 126.
nal of Gerontology: Psychological Sciences, 5 5 B,
P69P84. Binet, A., & Simon, Th. (1973 ). The development of
intelligence in children.Trans. by Elizabeth Kite.
Ackerman, P. L., Bowen, K. R., Beier, M. B.,
New York: Arno Press.
& Kanfer, R. (2001). Determinants of indi-
vidual differences and gender differences in Brunswik, E. (195 2). The conceptual framework
knowledge. Journal of Educational Psychology, of psychology. Chicago: University of Chicago
93 , 797825 . Press.
Ackerman, P. L., & Cianciolo, A. T. (2000). Chi, M. T. H., Glaser, R., & Rees, E. (1982). Exper-
Cognitive, perceptual speed, and psychomo- tise in problem solving. In R. J. Sternberg (Ed.),
tor determinants of individual differences dur- Advances in the psychology of human intelligence
ing skill acquisition. Journal of Experimental (Vol. 1, pp. 776). Hillsdale, NJ: Erlbaum.
Psychology: Applied, 6, 25 9290. Cronbach, L. J., & Furby, L. (1970). How we
Ackerman, P. L., & Heggestad, E. D. (1997). Intel- should measure change or should we? Psy-
ligence, personality, and interests: Evidence for chological Bulletin, 74, 6880.
164 the cambridge handbook of expertise and expert performance
Dawis, R. V., & Lofquist, L. H. (1984). A psycho- mechanisms, processes, and determinants. Jour-
logical theory of work adjustment: An individual nal of Social and Clinical Psychology, 5 , 25 1
differences model and its applications. Minneapo- 278.
lis, MN: University of Minnesota Press. Kanfer, R., & Ackerman, P. L. (1989). Moti-
Feltz, D. L. (1982). Path analysis of the causal vation and cognitive abilities: An integrative/
elements in Banduras theory of self-efficacy aptitude-treatment interaction approach to
and anxiety-based model of avoidance behav- skill acquisition. Journal of Applied Psychology
ior. Journal of Personality and Social Psychology, Monograph, 74, 65 7690.
42 (4), 764781. Kant, I. (1790/1987). Critique of judgment. Trans.
Ferguson, L. W. (195 2). A look across the years by Werner S. Pluhar. Indianapolis, IN: Hackett
1920 to 195 0. In L. L. Thurstone (Ed.), Applica- Publishing Co.
tions of psychology (pp. 117). New York: Harper Lohman, D. F. (1999). Minding our ps and
& Brothers. qs: On finding relationships between learn-
Fleishman, E. A., & Hempel, W. E., Jr. (195 5 ). ing and intelligence. In P. L. Ackerman, P. C.
The relation between abilities and improve- Kyllonen, & R. D. Roberts (Eds.), Learning and
ment with practice in a visual discrimination individual differences: Process, trait, and content
reaction task. Journal of Experimental Psychol- determinants (pp. 5 5 76). Washington, DC:
ogy, 49, 3 013 12. American Psychological Association.
Guilford, J. P., Christensen, P. R., Bond, N. A., Jr., McClelland, D. C., & Boyatzis, R. E. (1982). Lead-
& Sutton, M. A. (195 4). A factor analysis study ership motive pattern and long-term success in
of human interests. Psychological Monographs, management. Journal of Applied Psychology, 67 ,
68, (4, Whole No. 3 75 ). 73 7743 .
Guttman, L. (195 4). A new approach to fac- McNemar, Q. (1940). A critical examination of
tor analysis: The radex. In P. F. Lazarsfeld the University of Iowa studies of environmental
(Ed.), Mathematical thinking in the social sci- influences upon the IQ. Psychological Bulletin,
ences (pp. 25 83 48). Glencoe, Illinois, The Free 3 7, 63 92.
Press. Meehl, P. E., & Rosen, A. (195 5 ). Antecedent
Hoffman, R. R. (1987, Summer). The problem of probability and the efficiency of psychometric
extracting the knowledge of experts from the signs, patterns, or cutting scores. Psychological
perspective of experimental psychology. The AI Bulletin, 5 2 , 194216.
Magazine, 8, 5 3 67. Murray, H. A. et al. (193 8). Explorations in
Holland, J. L. (195 9). A theory of vocational personality: A clinical and experimental study of
choice. Journal of Counseling Psychology, 6, fifty men of college age. New York: Oxford Uni-
3 5 45 . versity Press.
Holland, J. L. (1997). Making vocational choices: A Roe, A. (195 6). The psychology of occupations.
theory of vocational personalities and work envi- New York: Wiley & Sons.
ronments (3 rd Edition). Odessa, FL: Psycholog-
Simonton, D. K. (1994). Greatness: Who makes
ical Assessment Resources.
history and why. New York: Guilford Press.
Honzik, M. P., MacFarlane, J. W., & Allen, L.
Snow, R. E. (1989). Aptitude-treatment interac-
(1948). The stability of mental test perfor-
tion as a framework for research on individ-
mance between two and eighteen years. Journal
ual differences in learning. In P. L. Ackerman,
of Experimental Education, 17 , 3 093 24.
R. J. Sternberg, & R. Glaser (Eds.), Learning
Humphreys, L. G. (1960). Investigations of the and individual differences: Advances in theory
simplex. Psychometrika, 2 5 , 3 13 3 23 . and research (pp. 13 5 9). New York: W. H.
Hunt, E. (1995 ). Will we be smart enough?: A cogni- Freeman.
tive analysis of the coming workforce. New York: Spangler, W. D. (1992). Validity of question-
Russell Sage Foundation. naire and TAT measures of need of achieve-
Jensen, A. R. (1998). The g factor: The science of ment: Two meta-analyses. Psychological Bul-
mental ability. Westport, CT: Praeger. letin, 112 (1), 14015 4.
Jones, M. B. (1962). Practice as a process of sim- Stanovich, K. E. (1986). Matthew effects in read-
plification. Psychological Review, 69, 274294. ing. Some consequences of individual differ-
Kanfer, R. (1987). Task-specific motivation: An ences in the acquisition of literacy. Reading
integrative approach to issues of measurement, Research Quarterly, 2 1, 3 60406.
psychometric approaches 165
Super, D. E. (1940). Avocational interest patterns: Vol. 4 (pp. 485 5 5 0). Palo Alto, CA: Consult-
A study in the psychology of avocations. Stanford, ing Psychologists Press.
CA: Stanford University Press. Watson, J. D. (2001). The double helix: A personal
Terman, L. M. (1926). Genetic studies of genius: account of the discovery of the structure of DNA.
Mental and physical traits of a thousand gifted New York: Simon & Schuster.
children. Stanford, CA: Stanford University Willingham, W. W. (1974). Predicting success in
Press. graduate education. Science, 183 , 273 278.
Thorndike, E. L. (1908). The effect of prac- Wittmann, W. W., & Su, H.-M. (1999). Inves-
tice in the case of a purely intellectual func- tigating the paths between working mem-
tion. American Journal of Psychology, 19, 3 74 ory, intelligence, knowledge, and complex
3 84. problem-solving performances via Brunswik
Thorndike, R. L. (1949). Personnel selection. New symmetry. In P. L. Ackerman, P. C. Kyllonen,
York: John Wiley & Sons. & R. D. Roberts (Eds.), Learning and indi-
Warr, P. (1994). Age and employment. In H. C. vidual differences: Process, trait, and content
Triandis, M. D. Dunnette, et al. (Eds), Hand- determinants (pp. 77108). Washington, DC:
book of industrial and organizational psychology, American Psychological Association.
C H A P T E R 10
Michelene T. H. Chi
167
168 the cambridge handbook of expertise and expert performance
briefly summarize what these methods can not clear how and/or what domain knowl-
uncover about differences in the knowledge edge influences the solution of insight prob-
representations of experts and novices. lems.
A next generation of problem-solving
research explored both knowledge-lean
A Brief History on Representation (puzzle-like) problems (such as the Tower
in the Study of Expertise of Hanoi) as well as knowledge-rich prob-
lems (such as in chess). Even though chess
The studies of representation in expertise is arguably more knowledge-rich than the
have historically been intimately related to Tower of Hanoi problem, it shares similar-
the type of problems being used. In early ities with puzzles and other toy domains
research on problem solving, the study of in that the understanding phase of the rep-
representation was carried out in the context resentation had been assumed to be straight-
of insight-type problems, such as Dunckers forward (But see Ericsson, Chapter 13 , and
(1945 ) candle problem. The goal of this Gobet and Charness, Chapter 3 0). That is,
problem is to mount three candles at eye for a domain such as chess, the understand-
level on a door. Available to use for this prob- ing phase of the representation needs to
lem are some tacks and three boxes. Partic- include the chess pieces, the permissible
ipants were presented with the tacks either operators (or moves) for each kind of chess
contained in the three boxes or outside of piece, and the goal state of checking and win-
the boxes so that the boxes were empty. The ning. In short, the understanding phase of
solution requires that one re-represents the the representation had been assumed to not
function of the boxes not as a container but clearly discriminate experts from novices.
as a platform that can be mounted on a wall If understanding is not the phase that
to hold a candle. All the participants pre- affects the choice of efficient moves, then
sented with the empty boxes could solve the what is? One obvious answer is how effec-
problem, whereas less than half of the par- tively a solver can search for a solution. The
ticipants given the full boxes could solve it. classical contribution by Newell and Simon
The key to all of these kinds of insight (1972) put forth the idea that what differen-
problems is to re-represent the problem in tiates experts from novices is the way they
a way to either release a constraint that is search through problem spaces. A problem
commonly assumed, or to think of some new space includes not only the elements, the
operator, that is again not the conventional operators, but also all the possible or permis-
one. So in the case of the candle problem, sible states created by the application of
one could say that the conventional func- operators to the elements, which are entailed
tional attribution that one applies to boxes by the permissible strategies for guiding the
is use as a container. Solving the problem search through this problem space. In this
requires thinking of a new function or affor- perspective, a representation is a model of
dance for boxes, in this case, as objects that the search performance of a solver on a spe-
can hold things up rather than hold certain cific problem (Newell & Simon, 1972). Thus,
kinds of things inside. a problem representation consists of:
Although insight problems investigated
the role of representation in the under- 1. An understanding phase the phase in
standing phase of problem solving (i.e., how which information about the initial state,
the elements, constraints, and operators of the goal state, the permissible operators,
a problem are encoded and interpreted), and the constraints is represented (so for
insight problems did not lend themselves chess, that would be the pieces and their
well to the study of expertise. That is, since positions on the chess board, the moves
expertise is defined as the accumulation of a allowed and disallowed for each kind of
large storehouse of domain knowledge, it is chess piece, etc.), and
representations of experts and novices knowledge 169
2. A search phase the phase in which depth of search between masters and less-
a step-by-step search path through the experienced (but proficient) players. What
problem space is represented. he did find was that the masters were able
to reconstruct a chess position almost per-
Because the understanding phase had been fectly after viewing it for only 5 seconds. This
assumed to be straightforward, differences ability could not be attributed to any supe-
between experts and novices are assessed via rior general memory ability, for when the
comparing differences in the search phase. chess positions were randomized, the mas-
A variety of different search heuristics have ters performed just about as poorly as the
been identified, such as depth-first versus less-experienced players. This finding sug-
breadth-first searches, backward versus for- gests that the masters superior performance
ward searches, exhaustive versus reduced with meaningful positions must have arisen
problem-space searches, and so forth. from their ability to perceive structure in
This view that differences in search such positions and encode them in chunks.
strategies or heuristics accounted for dif- The findings that chess experts can per-
ferences in expertise was also applied ceive coherent structures in chess positions
to knowledge-rich domains for which the and rapidlly come up with an excellent
understanding phase may not be so straight- choice of moves suggest that the understand-
forward. A perfect example is the work ing phase must be more than merely the
of Simon and Simon in the domain of straightforward encoding of the elements
physical mechanics. In this research, Simon and permissible operators to apply to the ele-
and Simon (1978) compared the problem- ments. Moreover, the application of differ-
solving skills of an expert and a novice by ent search heuristics cannot be the character-
representing their solution paths in terms ization that differentiates the experts from
of a sequence of equations (a set of pro- the novices in the search phase. Thus, what
ductions or condition-action rules) that they differentiated the experts and the novices
used to solve a physics problem. Based problem representation is determined by the
on this sequencing, the experts representa- representation of their domain knowledge,
tion was characterized as a forward-working of chess in this case. This recognition led
search (working from initial state toward the Chase and Simon (1973 a, b) to the identi-
desired end state in a series of steps), whereas fication and characterization of the struc-
the novices representation was character- tures or chunks of meaningful chess pat-
ized as a backward-working search (work- terns in memory. Thus, the work of de Groot
ing from the desired end state back to the (1966) and Chase and Simon (1973 a, b) rep-
initial state). Thus, the postulated repre- resented a first attempt at representing not
sentational difference between the expert just a problem solution, but knowledge of
and the novice was restricted to the search the domain. Subsequent work on expertise
phase, even though the understanding phase attempted to focus on how domain knowl-
may be a more crucial component for this edge is represented in a way that leads to
knowledge-rich domain. better solutions.
The revelation that search may not be For example, we have shown that expert
the entire story came from the work of de physicists representation of their domain is
Groot (1966). He found that world-class more principle based, whereas novices rep-
chess players did not access the best chess resentations are more situation or formula
moves from an extensive search; rather, they based (Chi, Feltovich, & Glaser, 1981). Thus,
often latched onto the best moves immedi- the expertise work in the 80s reemphasized
ately after the initial perception of the chess the understanding phase of representation,
positions. For example, de Groot could not but it differed from the earlier work on
find any differences in the number of moves insight and other knowledge-lean problems
considered, the search heuristics, or the in that the focus was on the structure and
170 the cambridge handbook of expertise and expert performance
organization of domain knowledge, and not their familiar intrinsic tasks. Contrived tasks
merely the structure of the problem. serve different purposes so that there is a
The next challenge for researchers is to continuum of contrived tasks, based on the
combine the understanding phase and the degree of modifications to the familiar task
search phase of a representation in order in order to bring the world into the labo-
to understand how it differentiates experts ratory, as it were (Hoffman et al., 1995 ).
from novices. In addition, new challenges However, there is a set of standard tasks that
are also presented when expertise is being are commonly undertaken in psychological
investigated in real-world domains. Many laboratories, such as recall. Recall of chess
complexities are involved when one stud- positions, for example, can be considered a
ies expertise in real-world domains, where contrived task since chess experts primary
problems are complex and dynamic, so that skill is in the selection of the best moves, not
the space is constantly changing with con- in recalling chess patterns. Although experts
textual dependencies and contingencies. In do recall games for a number of reasons (e.g.,
this kind of real-world scenarios, the space- knowledge sharing), asking them to recall
search model of problem solving does not chess patterns can be thought of as a con-
always apply as an explanatory mechanism. trived task.
It is also essentially mute about problem It is often the case that asking experts to
finding, which is a main phenomenon in real- perform in their familiar intrinsic tasks will
world problem-solving (see Klein, Pliske, show only that they are faster, more error
Crandall, & Woods, 2005 ). free, and in general better in all ways than
the novices. Their efficiency and speed can
often mask how their skills are performed.
Asking experts to perform contrived tasks,
Empirical Methods to Uncover on the other hand, can have several advan-
Representational Differences tages. First, a contrived task is often one that
can be undertaken just as competently by a
The nature of expertise can be ascertained in novice as an expert. Thus, it is not merely
two general ways. One way is to see how they the completion, efficiency, or correctness
perform in tasks that are familiar or intrinsic of performance at a contrived task that is
to their domain of expertise. For example, being evaluated, but rather, what the perfor-
selecting the best chess move, generating the mance reveals about the knowledge struc-
optimal blueprint, or detecting a cancerous ture of the individual, whether an expert
mass on X-rays are tasks that are intrinsic or a novice. More importantly, a contrived
to the domains of chess playing, on being task can shed light on experts shortcomings
an expert architect, and on being an expe- (see Chi, Chapter 2), whereas an intrinsic
rienced radiologist. This has been referred task will not, by definition of expertise. A
to as the study of performance at familiar key limitation of contrived tasks, however, is
tasks (Hoffman, 1987; Hoffman, Shadbolt, that if the contrived task departs too much
Burton, & Klein, 1995 ). Although these from the familiar task (e.g., lacks ecological
tasks might be abridged or in many ways validity and/or representativeness), then the
adapted for empirical investigation under model of performance that comes out may
conditions of experimental control and be a model of how the person adapts to the
the manipulation of variables, they are task, not a model of their expertise.
nevertheless more-or-less representative of In this section, I describe four contrived
what the domain experts do when they are tasks that have been used most extensively in
doing their jobs. laboratory studies of expertise with the goal
Alternatively, one can use contrived tasks of uncovering representational differences.
(Hoffman, 1987; Vicente & Wang, 1998) that The four methods are: recalling, perceiving,
are likely to be either unfamiliar to the prac- categorizing, and verbal reporting. Studies
titioner, or that depart more radically from using these four methods are grouped on
representations of experts and novices knowledge 171
the basis of the tasks that were presented to blitz, but the class-C players showed a big
the participants, and not the responses that difference (Calderwood, Klein, & Crandall,
they gave. For example, one could present 1988).
a perceptual task and ask for verbal reports These findings suggest that it is not
as responses. However, such a task would the experts superior search strategies that
be classified here as a perceptual task and helped them find the best move. Neither
not a verbal reporting task. Clearly there are can the master players superior recall be
many combinations of methods and many attributed to any differences in the mem-
optional ways to classify a task used in a spe- ory capacities of the master and less experi-
cific study. The choice here reflects only the enced players, since masters can only recall a
organization of the presentation in this chap- couple more pieces when the pieces are ran-
ter. Moreover, many studies use a combina- domly placed on the chess board (Chase &
tion of several methods. Simon, 1973 a).
This same pattern of results was also
recall obtained when Go (or Gomoku) play-
One of the most robust findings in exper- ers were asked to recall briefly presented
tise studies comes from using the method of Gomoku (or Go) board patterns. Both Go
free recall. Experts excel in recalling mate- and Gomoku utilize the same lattice-like
rials from their domain of expertise, such board with two different colored stones, but
as better, faster, and more accurate recall, the object of the two games is very different:
in domains ranging from static chess posi- In Go the goal is to surround the opponents
tions (Chase & Simon, 1973 a) to dynamic stone and in Gomoku it is to place five stones
computer-simulated thermal-hydraulic pro- in a row (Eisenstadt & Kareev, 1975 ). The
cess plant (Vicente, 1992). The classic study success of players in recalling board configu-
by de Groot (1966) in the domain of rations suggests that it is the meaningfulness
chess involved presenting chess players with of the configurations that enables the strong
meaningful chess boards for a brief inter- players better recall.
val, such as 5 seconds, to see how many In order to understand how experts and
pieces they could recall by reproducing the novices might organize their knowledge to
arrangements of the pieces on a blank board. result in differential recall, Chase and Simon
Chess masters were able to recall the posi- (1973 a,b) incorporated two additional pro-
tions almost perfectly (consisting of around cedures in conjunction with their recall
25 pieces). Less experienced players, on the procedure, both aimed at segmenting the
other hand, typically recall only about 5 to sequence in which players place the chess
7 pieces (Chase & Simon, 1973 a). However, pieces during recall. The first procedure
when de Groot (1966) asked the players to tape-recorded players as they reproduced
find the best move, the masters and the less chess pieces from memory and used the
experienced players did not differ signifi- pauses in their placement of pieces to seg-
cantly in the number of moves they searched ment the sequence of placements. The sec-
nor the depth of their search, even though ond procedure was to modify the task from
the masters were always able to find and a recall to a visual memory task. In this
select the best move. Likewise, Klein, Wolf, modified visual task, players were simply
Militello, and Zsambok (1995 ) found that asked to copy chess positions. The head
the first move that expert chess players con- turns they made to view the positions in
sider is significantly better than chance. Fur- order to reproduce the chess positions were
thermore, chess experts do not differ from used to segment the sequence of placements,
class-C players in the percentage of blunders that is, to reveal how the game arrays were
and poor moves during regulation games, chunked. The results showed that players
but do differ during blitz games. In fact, the recalled positions in rapid bursts followed
experts showed very little increase in rate by relatively longer pauses (i.e., > 2 sec-
of blunders/poor moves from regulation to onds), and they reproduced a meaningful
172 the cambridge handbook of expertise and expert performance
cluster of pieces after a head turn. Because The chunking of patterns into a hierarchi-
the master players recalled and reproduced a cal representation applies not only to games
greater number of pieces before a long pause and architecture, but to other domains, such
and a head turn, respectively, these two as circuit fault diagnosis. Egan and Schwartz
results, together, suggest that chess experts (1979) found that expert circuit technicians
had many more recognizable configurations chunk circuit elements together according
of chess patterns in their knowledge base, to the function, such as chunking resistors
and these configurations (based on power and capacitors because together they per-
in controlling regions of the board) were form the function of an amplifier. Here too,
comprised of a greater number of pieces. chunking leads to superior recall for experts
The representational differences between as compared to non-experts. Moreover, the
the masters and less proficient players were skilled electronic technicians pattern recall
that the masters had a greater number of was faster and more accurate, again suggest-
recognizable patterns (or chunks) in mem- ing that the local patterns formed higher-
ory, and each pattern on average contained order patterns.
a greater number of pieces. The recall superiority of experts can be
More important, when memory perfor- captured not only in visual tasks, but also in
mance was reanalyzed in terms of experts verbal tasks. Looking at a practical domain,
and non-expert chunks, the number of Morrow, Mernard, Stine-Morrow, Teller, and
chunks recalled by experts and non-experts Bryant (2001) asked expert pilots and some
were now about the same, implying that non-pilots to listen to Air Traffic Control
their basic memory capacity is not that dif- messages that described a route through an
ferent after all, validating the finding of the air space. Participants were then asked to
depressed expert-recall performance for ran- read back each message and answer a probe
domized board arrangements. The findings question about the route. Expert pilots were
of equivalent recall for randomized positions more accurate in recalling messages and in
and equivalent recall in terms of number of answering the question than non-experts.
patterns, together, confirm that both expert In sum, several different types of recall-
and non-expert players are subject to the related contrived tasks provide some insight
same short-term memory capacity limita- into the experts and non-experts represen-
tions, but the limitation is not the point. The tation of their domain, such as patterns of
point is how people come to create meaning- familiar chunks, clusters of circuit elements
ful chunks. with related function, and hierarchical orga-
The recalled chess patterns (as deter- nization of chunks.
mined by segregated pauses and head turns),
when analyzed in detail, showed that they perceiving
tended to consist of commonly occurring Perception tasks address the issue of what
patterns that are seen in regular routine experts versus non-experts perceive in a
playing of chess, such as clusters in attack given amount of time (Chase & Chi, 1981).
and defense positions. It seems obvious that A good example of a perceptual task is
such local patterns may be used to form examining X-ray films. Although the goal
representations at a higher level of familiar of examining X-ray films is usually to diag-
global patterns. Direct evidence of such a nose disease, one can also determine what
hierarchical representation can be seen also experts and novices see (literal stimulus fea-
in the domain of architecture. Using the tures) and perceive (meanings of the fea-
same recall procedure, looking at pauses, tures or patterns of features). Lesgold et al.
Akin (1980) uncovered a hierarchical repre- (1988) asked four expert radiologists with
sentation of blueprints, with such things as 10 or more years of experience after res-
doors and walls at the lowest level and rooms idency, and eight first-to-fourth year resi-
at a higher level, and clusters of room at the dents to examine X-ray films for as long as
highest level. they wished, commenting on what they saw
representations of experts and novices knowledge 173
as well as verbally expressing their diagnoses. tures (the tumors) as general lung haziness.
Although diagnosis is the familiar intrin- That is, they interpreted the hazy spots in the
sic task, the participants were also asked to lungs as indicating fluid in the lungs, suggest-
undertake a more contrived task, which was ing congestive heart failure, whereas experts
to draw contours on the films showing what saw multiple tumors. Residents also saw the
they believed to be the problematic areas, heart as enlarged, while the experts did not.
as a way of identifying the relevant features Residents also interpreted the cues or fea-
they saw. (The films showed diseases such as tures they saw rather literally. For example,
multiple tumors or collapsed lung.) Two of a large size heart shadow implied an enlarged
the four experts, but only one of the eight heart, whereas experts might adjust their
residents, diagnosed the collapsed lung film evaluation of the heart to other possibilities,
accurately. Did they see the features in the such as a curvature in the spine.
films differently? Both experts and residents The results of this study show basically
saw the main feature, which was the col- that experts perceive things differently from
lapse of the middle lobe, producing a dense non-experts. There are many other studies
shadow. However, this feature can lead only that show the same kind of results (see Klein
to a tumor diagnosis; the correct diagnosis of & Hoffman, 1992). This includes the percep-
collapsed lung must require seeing the dis- tion tasks of reproducing chess board pat-
placed lobe boundaries or hyperinflation of terns as discussed earlier. Reitman (1976)
the adjacent lobes. Residents did not see the also replicated the Chase and Simon (1973 a)
more subtle cues and the relations among study for the game of Go. In addition to
the cues. asking participants to reproduce patterns
In addition to the accuracy of the diag- of Go stones as quickly and accurately as
noses, the researchers looked at two kinds possible while the stimulus board pattern
of coding of the protocols. The first coding remained exposed throughout the trial, she
was the diagnostic findings, which referred also asked the Go experts to draw circles (on
to the attribution of specific diagnostic prop- paper transcriptions of the real game posi-
erties in the film. For example, one finding tions) showing stones that were related, and
might be spots in the lungs. The second if appropriate, to indicate which groups of
coding was the meaningful clusters. A clus- stones were related on yet a higher strategic
ter is a set of findings that had a meaningful level. The results showed that the experts
path or reasoning chain from each finding partitioned the patterns not into a strictly
to every other finding within the set. That nested hierarchy, but rather into overlap-
is, the participants would relate the features ping subpatterns, as one might expect given
logically to entail a diagnostic explanation. the nature of Go a given stone can par-
For example, if the participants commented ticipate in, or play a strategic role in, more
that such spots might be produced by blood than one cluster of stones. Although there
pooling, which in turn could have been pro- were no novice data on penciled partition-
duced by heart failure, then such a reasoning ing, the experts partitioning into overlap-
chain would relate the findings into a cluster. ping structures suggests this more interre-
The results showed that the experts identi- lated lattice-like (versus strictly hierarchical)
fied around three more findings per film, and representation.
had about one more cluster than the resi- The perceptual superiority of experts
dents. This suggests that the experts not only applies to dynamic situations as well, such as
saw more critical features on a film than the perception of satellite infrared image loops
residents, but perceived more interrelations in weather forecasting (Hoffman, Trafton,
among the features. & Roebber, 2005 ), or watching a video-
Moreover, experts had finer discrimina- tape of classroom lesson (Sabers, Cushing,
tions. For example, the tumor film showed a & Berliner, 1991). For example, when expert
patient with multiple tumors. For this tumor and novice teachers were asked to talk out
film, residents tended to merge local fea- loud while watching a videotaped classroom
174 the cambridge handbook of expertise and expert performance
lesson that showed simultaneous events can be driven top down by more devel-
occurring throughout the classroom, the oped schemas (rather than a better devel-
experts saw more patterns by inferring what oped acuity) since greater experience with
must be going on (such as the students films means they have more familiarity with
note taking indicates that they have seen both under- and overexposed films. To dis-
sheets like this . . . ), whereas the non-expert prove the second interpretation that per-
teachers saw less, saying that I cant tell ceptual superiority is necessarily visual one
what they are doing. They are getting ready can show that experts can excel in per-
for class. In short, the explanations experts ception even if the materials are not pre-
and non-experts can give reveal the fea- sented visually, as in the case of chess masters
tures and meaningful patterns they saw and playing blindfolded chess (Campitelli &
perceived. Gobet, 2005 ) and expert counselors form-
A related task is detection of the presence ing an accurate model of a client from lis-
of features or events accompanied by mea- tening to a transcript of a counseling session
surement of reaction times. For example, (Mayfield, Kardash, & Kivlighan, 1999).
Alberdi et al. (2001) asked some more- and In sum, this section summarized percep-
some less-experienced physicians to view tion tasks and related contrived tasks such as
traces on a computer screen showing five asking experts and novices to circle Go pat-
physiological measurements, such as heart terns or draw contours of X-ray films. The
rate, transcutaneous oxygen, etc. The traces point of these studies is not merely to show
represented both key events, such as devel- whether experts are superior in performing
oping pneumothorax, as well as more sec- these kinds of tasks, but to uncover their
ondary but still clinically noteworthy events. underlying representations and skills that
Although the less-experienced physicians derive from practice and perceptual learn-
were almost as good in detecting and identi- ing, such as more interrelated clustering of
fying the key events, they were significantly findings on X-ray films and their representa-
worse than the more-experienced physi- tion of secondary events.
cians in detecting the secondary events. The
more-experienced physicians were also sig- categorizing
nificantly better at detecting artifacts. This Sorting instances according to categories is
suggests that they were not only better at a simple and straightforward task that can
detecting secondary events, but that they be readily undertaken by experts and non-
also made finer discriminations between experts. One procedure is to ask participants
meaningful events versus literal stimulus to sort problem statements (each problem
features. typed on a 3 5 card) into categories on
It should perhaps be pointed out that the basis of similarities in the solution or
such results do not arise from experts hav- some other functional categories. Chi et al.
ing better visual acuity. Nor do the results (1981) solicited the participation of physics
mean that the experts perceptual superior- graduate students (who technically would
ity is necessarily visual (vs. analytical). That be apprentices or perhaps journeymen on
is, expertise involves perceiving more, not the proficiency scale, but probably not fully
just seeing more. To deny the first interpre- expert) and undergraduate students (who
tation, one can show that novices visual had completed a semester of mechanics with
acuity is just as good as experts in some an A grade, making them initiates and not
other domain for which they have no exper- really novices). They were asked to sort 24
tise. However, expertise can enhance sensi- physics problems twice (for consistency),
tivity to critical cues, features, and dimen- and also to explain the reasons for their sort-
sions. Snowden, Davies, and Roling (2000) ing. One would not necessarily expect quan-
found expert radiologists to be more sen- titative differences in the sortings produced
sitive to low contrast dots and other fea- by the two skill groups, such as the num-
tures in X-rays. This increased sensitivity ber of groups, or the number of problems in
representations of experts and novices knowledge 175
the groups since anyone could sort prob- Much research followed that replicated
lems on any of a nearly boundless number the basic finding of shallow versus deep
of dimensions or criteria. The real interest representations for novices versus experts.
lies in the nature of the sortings. Based on For example, when expert and novice pro-
analyses of both the problems that the par- grammers were asked to sort programming
ticipants categorized into the same groups problems, the experts sorted them accord-
as well as their explanations for the sort- ing to the solution algorithms, whereas the
ings, it became apparent that the under- novices sorted them according to the areas
graduates grouped problems very differently of applications, such as creating a list of
from the graduate students. The undergrad- certain data types (Weiser & Shertz, 1983 ).
uates were more likely to base their sorting Similarly, when expert and novice coun-
on literal surface features, such as the pres- selors were asked to categorize client state-
ence of inclined planes or concepts such as ments from a counseling script as well as
friction, whereas the graduate students were to map the relationships among the cate-
much more likely to base their sorting on gories, novices tended to categorize and map
domain principles that would be critical to on the basis of superficial details, such as
the solutions (e.g., such as problems that the temporal order of the client statements
involve Newtons Second Law or the laws (Mayfield et al., 1999), whereas the expert
of thermodynamics such as conservation of counselors tended to categorize and map on
energy). This finding was further replicated the basis of more abstract, therapeutically
by a specially designed set of problems that relevant information. Similarly, Shafto and
had either the same surface features but dif- Coley (2003 ) found that commercial fish-
ferent deep principles, or different surface ermen sorted marine creatures according to
features but the same deep principles. The commercial, ecological, or behavioral fac-
same results emerged, namely, that under- tors, whereas undergraduates sorted them
graduates sorted according to the surface according to the creatures appearance.
features and graduates tended to sort accord- Many variations of the sorting task have
ing to the deep principles. also been used. One variation is to ask par-
One interpretation of such results is that ticipants to subdivide their groups further,
the undergraduates schemas of problems to collapse groups, or to form multiple and
are based on physical entities and literal for- differing sortings in order to shed light on
mulas, whereas experts schemas are more the hierarchical structure of their knowledge
developed and organized around the prin- representations (Chi, Glaser, & Rees, 1982).
ciples of mechanics. This means that the For example, by asking a young dinosaur
explicit words or terminologies and dia- expert to collapse his initial categories
grams used in the problem statements are formed about different types of dinosaurs,
connected (in experts reasoning) to the the child would collapse them into two
basic principles. However, that connection major superordinate categories meat-eaters
is not necessarily direct. For instance, an and plant-eaters (Chi & Koeske, 1983 ) sug-
inclined plane per se does not by itself indi- gesting that the superordinate categories are
cate a Newtons-Second-Law problem for an somewhat well defined.
expert physicist. An additional study asking Another variation is a speeded category-
participants to cite the most important fea- verification task. In such a task, a cate-
tures in a problem statement showed that gory name appears first, followed by a pic-
the words in the problem statements are ture. Participants press true if the picture
mediated by some intermediate concepts, matched the word, such as a picture of a
such as a before and after situation. Thus, dog with the term animal, and false if
the words in a problem interact to entail con- it does not match, and reaction latencies
cepts, and experts solutions may be based on can be measured. Moreover, the words can
these higher-level concepts (Chi et al, 1981; refer to a superordinate category such as
Chi & Ohlsson, 2005 ). animals, a basic-object-level category such
176 the cambridge handbook of expertise and expert performance
or retrospective verbal report is part of the ward from the goal (as one would predict
familiar intrinsic task (e.g., coroners audio on the basis of studies described earlier in
record during autopsies; and during weather this chapter). The same forward-backward
forecasting briefings, forecasters think aloud search patterns were obtained also in the
as they examine weather data). domain of genetics with experts and novices
The most difficult aspect of verbal report (Smith & Good, 1984).
methods is data analysis. That is, how does In a different kind of domain and task,
one code and analyze verbal outputs? Again Wineburg (1991) asked historians and his-
there are many methods; they can only be tory students to give think-aloud proto-
alluded to here (see Chi, 1997; Ericsson & cols while they constructed understanding
Simon, 1984, for explicit techniques, and of historical events from eight written and
Ericsson, Chapter 13 ). Typically, think-aloud three pictorial documents. The participants
protocols are analyzed in the context of the task was to decide which of the three pic-
cognitive task, which requires a cognitive tures best depicted what happened dur-
task analysis in order to know the functional ing the Battle of Lexington at the start
problem states that are to be used to cat- of the Revolutionary War, the event pre-
egorize individual statements. The goal of sented in the documents. Statements in
protocol analysis then is to identify which the participants picture-evaluation proto-
sequence of states a particular participant cols were coded into four categories: descrip-
progresses through, and perhaps a computa- tion, reference, analysis, and qualification.
tional model is built to simulate those steps Both experts and students provided descrip-
and the solution procedures. For explana- tive statements, but the experts made more
tions, coding methods involve segmenting statements that fell into the other three
and judging the content of the segments in categories. This is not surprising since the
terms of issues such as whether it is substan- experts obviously had more to say, being
tive or non-substantive (Chi, Siler, Jeong, more knowledgeable. What is more interest-
Yamauchi, & Hausmann, 2001), principle ing is to identify the first category for which
oriented (deep) or entity oriented (shallow) both the experts and novices described the
(Chi et al., 1981). Note that an analysis of ver- picture using the same number of state-
bal data means that the content of the data is ments. The quality of those descriptions was
not always taken literally or word-for-word. different. Historians noted 25 of the 5 6 pos-
That is, we are not asking experts and novices sible key features in the paintings that had
their subjective assessment of how they per- a bearing on the historical accuracy of the
formed, or how they have performed. This paintings, whereas the students noted only
is because much of expert knowledge is not four features on average. Moreover, in select-
explicit nor subject to introspection. ing the most accurate painting, historians
How people perform can be captured did so on the basis of the correspondence
by the coding scheme. A study by Simon between the visual representations and the
and Simon (1978) provides a good exam- written documents, whereas the students
ple. They collected concurrent protocols often chose on the basis of the quality of the
from an expert and a novice as they were artwork, such as its realism and detail. This
solving physics problems. The researchers suggests that the experts representations
coded only the equation-related parts of were much more meaningfully integrated.
the protocols. By examining what equa- Interviewing techniques can include both
tions were articulated, and when, the open-ended questions and more direct ques-
researchers were able to model (using a tions. For example, Hmelo-Silver and Pfef-
production-system framework) each partic- fer (2004) asked experts and students both
ipants problem-solving procedure and strat- direct questions about aquaria, such as
egy. The researchers showed that the expert What do fish do in an aquarium? and
solved the problems in a forward-working open-ended questions, such as thinking out
strategy, whereas the novice worked back- loud while attempting to Draw a picture
178 the cambridge handbook of expertise and expert performance
of anything you can think is in an aquar- be measured in terms of some units. Without
ium. Since biological systems and devices being precise, a bit of knowledge can be a
often can be characterized by their struc- factual statement, a chunk/familiar pattern,
ture, behavior, or function (Gellert, 1962; a strategy, a procedure, or a schema. Chase
Chi, 2000, p. 183 ; Goel et al., 1996), the and Simon (1973 a, b) estimated an expert
protocols were coded according to state- chess (master-level) player to know between
ments relating to those three categories. 10,000 and 100,000 chunks or patterns,
There were no differences between the whereas a good (Class-A) player has around
experts and the novices in the number of 1000 chunks; and Miller (1996, pp. 13 613 8)
statements referring to the structures, but estimated college-educated adults to know
there were predictable and significant dif- between 40,000 to 60,000 words. Hoff-
ferences in the number of statements refer- man et al., (in press; Hoffman, Trafton, &
ring to behaviors and functions. The novices Roebber, 2006) estimate that it would take
often did not offer additional behavioral or thousands of propositions to capture the
functional information even when probed. expert weather forecasters knowledge just
This suggests that the experts represent the about severe weather in one particular cli-
deeper features (i.e., behavior and function), mate. Regardless of how one wishes to
whereas novices think in terms of literal fea- quantify it, clearly, one can expect experts
tures (i.e., the structure). to know more than non-experts (includ-
In sum, the goal of these verbal reporting ing journeymen and especially compared to
methods is to capture the underlying repre- apprentices, initiates, and novices). Indeed,
sentations of the experts and novices, such this is one definition of expertise. The recall
as whether their searches are forward ver- task summarized earlier also revealed how
sus backward, whether their understanding the number of chunks and the chunk sizes
of pictures and text are integrated versus lit- differ for experts versus non-experts.
eral, or whether their understanding mani- Aside from the sheer number of bits
fest deep (behavioral and functional) versus (however these are defined) in their knowl-
shallow (structural) features. edge base, a related concept to the dimen-
sion of size is completeness. Completeness
has a different connotation than the idea of
merely greater amount or extent of knowl-
Representational Differences
edge. In real-world domains knowledge is
always expanding. Any notion of complete-
If the difference in representation (reflect- ness becomes very slippery.
ing the organization of knowledge and not In terms of frame theory, one can con-
just the extent of knowledge) is one key ceive of completeness in terms of the avail-
to understanding the nature of expertise, ability or number of slots, or necessary
then in what ways do the representations slots. For example, a tree expert might have
of experts and novices differ? In this sec- slots for susceptibility to different diseases
tion, I briefly address dimensions of repre- with knowledge about potential diseases
sentational differences, as captured by the (values) for each kind of trees, whereas a
empirical tasks of recalling, perceiving, cat- novice might not have such slots at all.
egorizing, and verbal reporting described The earlier-described finding from a per-
above. Each of these tasks has revealed ception task showed that the more- (but
ways in which representations of experts and not the less-) experienced physicians were
novices differ. able to recognize secondary events on traces
of physiological measurements (Alberdi et
knowledge extent al., 2001), can be interpreted to indicate
An obvious dimension of difference is that that the more-experienced physicians had
experts have more knowledge of their more complete frames or schemas. Greater
domain of expertise. More knowledge must amount of knowledge might also refer to
representations of experts and novices knowledge 179
in several domains, such as birds (Tanaka & problems at the level of principles, whereas
Taylor, 1991), faces (Tanaka, 2001), dinosaurs novices represent them at the concrete
(Chi et al., 1989), and geological and archae- level of entities or superficial features (Chi
ological classification (Burton et al., 1987, et al., 1981), or that landscaping experts sort
1988, 1990). trees into goal-derived categories (e.g., shade
Just as the notion of knowledge extent trees, fast-growing trees, etc.), whereas tax-
can be slippery (because knowledge is never onomists sort trees according to biological
static), so too the notion of hierarchi- taxa (Medin, Lynch, Coley, & Atran, 1997).
cal memory organization can be slippery. Such differences can be revealed also
For example, instead of conceiving of non- in perception tasks. For example, a patient
experts memory representation as missing putting his hands on his chest and leaning
the intermediate levels, another view is that forward as he walks slowly is interpreted by
their representations are more like lattices novices merely as someone having back pain
than hierarchies (Chi & Ohlsson, 2005 ). (a literal interpretation), whereas a more
(Technically, a lattice would involve cross expert physician might interpret the same
connections that would be category viola- observation as perhaps suggesting that the
tions in a strict hierarchy or is-a tree.) It patient has some unspecified heart problem
is valuable to look at an extreme, that is, (Patel & Arocha, 2001). Differences can also
domains where everything can be causally be revealed in a verbal reporting task, such as
related to everything else, and neither hierar- explaining the behavior/function of fish in an
chies, lattices, nor chains suffice to represent aquarium versus explaining the structure of
either the world or knowledge of the world, fish (Hmelo-Silver & Pfeffer, 2004). Differ-
such as the weather forecasters understand- ences can be revealed in a task that involves
ing of atmospheric dynamics (e.g., thunder- explaining causal relationships a novices
storms cause outflow, which in turn can trig- explanations might focus on the time
ger more thunderstorms). We do not yet and place of an historical event, whereas
have a clear understanding of how dynamic an experts explanations might focus on
systems are represented (Chi, 2005 ). On the using the time to reconstruct other events
other hand, for a domain such as terrain (Wineburg, 1991).
analysis in civil engineering, much of the In short, all four of the task types
experts knowledge is very much like a hier- reviewed here can reveal differences bet-
archy, highly differentiated by rock types, ween experts and novices representations
subtypes, combinations of layers of subtypes, in terms of depth.
types of soils, soil-climate interactions, etc.
(Hoffman, 1987). consolidation and integration
In sum, although any inferences about A fourth dimension of representational dif-
knowledge representation need to be ferences between experts and non-experts
anchored in the context of a specific is that the experts representation may be
domain, contrived tasks such as recalling, more consolidated, involving more efficient
perceiving, and categorizing can allow us to and faster retrieval and processing. A related
differentiate the ways experts and novices way to characterize it might be the integrat-
knowledge is organized. edness or coherence of a representation, that
is, the degree to which concepts and prin-
depth of knowledge ciples are related to one another in many
Representational differences can be char- meaningful ways (e.g., Falkenhainer, Forbus,
acterized not only by extent and orga- & Gentner, 1990; Schvaneveldt et al., 1985 ).
nization, but also by dimensions such as One interpretation of integratedness is the
deep versus shallow, abstract versus con- interaction of features. Evidence for this
crete, function versus structure, or goal- interpretation can be seen in physics experts
directed versus taxonomic. Such differences and non-experts representations (Chi et al.,
have been revealed using a sorting task, to 1981), in which they identify features
show, for example, that physicists represent that are combined or integrated to form
representations of experts and novices knowledge 181
for the design of computerized monitoring in problems by experts and novices. Cognitive Sci-
neonatal intensive care. International Journal of ence, 5 , 12115 2.
Human-Computer Studies, 5 5 , 191216. Chi, M. T. H., Glaser, R., & Rees, E. (1982). Exper-
Burton, A. M., Shadbolt, N. R., Hedgecock, tise in problem solving. In R. J. Sternberg (Ed.),
A. P., & Rugg, G. (1987). A formal evaluation Advances in the psychology of human intelligence,
of knowledge elicitation techniques for expert Vol. 1 (pp. 775 ). Hillsdale, NJ: Erlbaum.
systems: Domain 1. In D. S. Moralee (Ed.), Chi, M. T. H., Hutchinson, J., & Robin, A. F.
Research and development in expert systems, Vol. (1989). How inferences about novel domain-
4. (pp. 3 5 46). Cambridge: Cambridge Univer- related concepts can be constrained by struc-
sity Press. tured knowledge. Merrill-Palmer Quarterly, 3 5 ,
Burton, A. M., Shadbolt, N. R., Rugg, G., & 2762.
Hedgecock, A. P. (1988). Knowledge elicita- Chi, M. T. H., & Koeske, R. (1983 ). Network
tion techniques in classification domains. In Y. representation of a childs dinosaur knowledge.
Kodratoff (Ed.), ECAI-88: Proceedings of the Developmental Psychology, 19, 293 9.
8th European Conference on Artificial Intelligence Chi, M. T. H., & Ohlsson, S. (2005 ). Com-
(pp. 85 93 ). London: Pittman. plex declarative learning. In K. J. Holyoak, &
Burton, A. M., Shadbolt, N. R., Rugg, G., & R. G. Morrison, (Eds.), The Cambridge hand-
Hedgecock, A. P. (1990). The efficacy of book of thinking and reasoning (pp. 3 713 99).
knowledge elicitation techniques: A compari- Cambridge: Cambridge University Press.
son across domains and levels of expertise. Jour- Chi, M. T. H., Siler, S. A., Jeong, H., Yamauchi,
nal of Knowledge Acquisition, 2 , 167178. T., & Hausmann, R. G. (2001). Learning from
Calderwood, R., Klein, G. A., & Crandall, B. W. human tutoring. Cognitive Science, 2 5 , 471
(1988). Time pressure, skill, and move quality 533.
in chess. American Journal of Psychology, 101, Chiesi, H. L, Spilich, G. J., & Voss, J. F. (1979).
481493 . Acquisition of domain-related information in
Campitelli, G., & Gobet, F. (2005 ). The minds relation to high and low domain knowledge.
eye in blindfold chess. European Journal of Cog- Journal of Verbal Learning and Verbal Behavior,
nitive Psychology, 17 , 23 45 . 18, 25 7273 .
Chase, W. G., & Chi, M. T. H. (1981). Cognitive Cooke, N. J. (1994). Varieties of knowledge
skill: Implications for spatial skill in large-scale elicitation techniques. International Journal of
environments. In J. Harvey (Ed.), Cognition, Human-Computer Studies, 41, 801849.
social behaviors, and the environment (pp. 111 Cullen, J., & Bryman, A. (1988). The knowl-
13 6). Hillsdale, NJ: Erlbaum. edge acquisition bottleneck: A time for
Chase, W. G., & Simon, H. A. (1973 a). Perception reassessment? Expert Systems, 5 , 216225 .
in chess. Cognitive Psychology, 4, 5 5 81. De Groot, A. (1966). Perception and memory ver-
Chase, W. G., & Simon, H. A. (1973 b). The sus thought: Some old ideas and recent find-
minds eye in chess. In W. G. Chase (Ed.), ings. In B. Kleinmuntz (Ed.), Problem solving:
Visual information processing (pp. 215 281). Research, method, and theory (pp. 195 0). New
New York: Academic Press. York: Wiley.
Chi, M. T. H. (1997). Quantifying qualitative Duncker, K. (1945 ). On problem-solving.
analyses of verbal data: A practical guide. Jour- (L. S. Lees, Trans.). Psychological monographs,
nal of the Learning Sciences, 6, 2713 15 . 5 8 (Whole No. 270). (Original work published,
Chi, M. T. H. (2000). Self-explaining expository 193 5 .)
texts: The dual processes of generating infer- Egan, D. E., & Schwartz, B. J. (1979). Chunking
ences and repairing mental models. In R. Glaser in recall of symbolic drawings. Memory & Cog-
(Ed.), Advances in instructional psychology nition, 7 , 14915 8.
(pp. 16123 8). Hillsdale, NJ: Erlbaum. Eisenstadt, M., & Kareev, Y. (1975 ). Aspects
Chi, M. T. H. (2005 ). Common sense conceptions of human problem solving: The use of inter-
of emergent processes: Why some misconcep- nal representations. In D. A. Norman & D.
tions are robust. Journal of the Learning Sciences, E. Rumelhart (Eds.), Exploration in cognition
14, 161199. (pp. 3 083 46). San Francisco: Freeman.
Chi, M. T. H., Feltovich, P., & Glaser, R. (1981). Ericsson, K. A., Delaney, P. F., Weaver, G.,
Categorization and representation of physics & Mahadevan, R. (2004). Uncovering the
representations of experts and novices knowledge 183
structure of a memorists superior basic mem- eration in chess. Organizational Behavior and
ory capacity. Cognitive Psychology, 49, 19123 7. Human Decision Processes, 62 , 63 69.
Ericsson, K. A., & Simon, H. A. (1984). Protocol Lesgold, A., Rubinson, H., Feltovich, P., Glaser,
analysis. Cambridge, MA: MIT Press. R., Klopfer, D., & Wang, Y. (1988). Expertise
Falkenhainer, B., Forbus, K. D., & Gentner, D. in a complex skill: Diagnosing X-ray pictures.
(1990). The structure-mapping engine: Algo- In M. Chi, R. Glaser, & M. Farr (Eds.), The
rithm and examples. Artificial Intelligence, 41, nature of expertise (pp. 3 113 42). Hillsdale, NJ:
163 . Erlbaum.
Gellert, E. (1962). Childrens conception of the Mayfield, W. A., Kardash, C. M., & Kivlighan,
structure and function of the human body. D. M. (1999). Differences in experienced and
Genetic Psychology Monographs, 65 , 193 405 . novice counselors knowledge structures about
Goel, A. K., Gomez de Silva Garza, A., Grue, N., clients: Implications for case conceptualiza-
Murdock, J. W., Recker, M. M., & Govinderaj, tion. Journal of Counseling Psychology, 46, 5 04
T. (1996). Towards designing learning envi- 5 14.
ronments. In C. Frasson, G. Gauthier, & A. Means, M. L., & Voss, J. F. (1985 ). Star Wars:
Lesgold (Ed.), Intelligent tutoring systems: Lec- A developmental study of expert and novice
ture notes in computer science (pp. 493 5 01). knowledge structures. Journal of Memory and
Berlin: Springer-Verlag. Language, 2 4, 74675 7.
Hmelo-Silver, C. E., & Pfeffer, M. G. (2004). Medin, D. L., Lynch, E. B., Coley, J. D., & Atran,
Comparing expert and novice understanding S. (1997). Categorization and reasoning among
of a complex system from the perspective of tree experts: Do all roads lead to Rome? Cog-
structures, behaviors, and function. Cognitive nitive Psychology, 3 2 , 4996.
Science, 2 8, 12713 8. Miller, G. A. (1996). The science of words. New
Hoffman, R. R. (1987, Summer). The problem of York: McGraw-Hill.
extracting the knowledge of experts from the Morrow, D. G., Menard, W. E., Stine-Morrow,
perspective of experimental psychology. The AI E. A. L., Teller, T., & Bryant, D. (2001). The
Magazine, 8, 5 3 67. influence of expertise and task factors on age
Hoffman, R. R., Coffey, J. W., Ford, K. M., & differences in pilot communication. Psychology
Novak, J. D. (in press). A method for elicit- & Aging, 16, 3 146.
ing, preserving, and sharing the knowledge of Newell, A., & Simon, H. A. (1972). Human prob-
expert forecasters. Weather & Forecasting. lem solving. Englewood Cliffs, NJ: Prentice-
Hoffman, R. R., Shadbolt, N. R., Burton, A. M., Hall.
& Klein, G. (1995 ). Eliciting knowledge from Nisbett, R. E., & Wilson, T. D. (1977). Telling
experts: A methodological analysis. Organiza- more than we can know: Verbal reports on
tional Behavior and Human Decision Processes, mental processes. Psychological Review, 84, 23 1
62 , 12915 8. 25 9.
Hoffman, R. R., Trafton, G., & Roebber, P. (2006). Patel, V. L., & Arocha, J. F. (2001). The nature
Minding the weather: How expert forecasters of constraints on collaborative decision-making
think. Cambridge, MA: MIT Press. in health care settings. In E. Salas, & G. A.
Johnson, K., & Eilers, A. T. (1998). Effects of Klein (Eds.), Linking expertise and naturalistic
knowledge and development on subordinate decision making (pp. 3 83 405 ). Mahwah, NJ:
level categorization. Cognitive Development, 13 , Erlbaum.
5 15 5 45 . Reitman, J. S. (1976). Skilled perception in
Klein, G., Pliske, R. M., Crandall, B., & Woods, Go: Deducing memory structures from inter-
D. (2005 ). Problem detection. Cognition, Tech- response times. Cognitive Psychology, 8, 3 3 6
nology, and Work, 7, 1428. 3 5 6.
Klein, G. A., & Hoffman, R. R. (1992). Seeing Rosch, E., Mervis, C. B., Gray, W. D., Johnson,
the invisible: Perceptual-cognitive aspects of D. M., & Boyes-Braem, P. (1976). Basic objects
expertise. In M. Rabinowitz (Ed.), Cognitive in natural categories. Cognitive Psychology, 8,
science foundations of instruction (pp. 203 226). 3 8243 9.
Mahwah, NJ: Erlbaum. Sabers, D. S., Cushing, K. S., & Berliner,
Klein, G., Wolf, S., Militello, L., & Zsambok, C. D. C. (1991). Differences among teach-
(1995 ). Characteristics of skilled option gen- ers in a task characterized by simultaneity,
184 the cambridge handbook of expertise and expert performance
Task Analysis
were applied, why they were applied, and 2. The age of oil, automobiles, and mass
what their strengths and weaknesses were. production. Oil and gas allow massive
As the field is evolving constantly, I will end motorization of transport, civil economy,
with a discussion of some open avenues for and war. Leading branches of the econ-
further work. omy are automobiles, aircraft, refineries,
trucks, and tanks. Radio, motorways, air-
ports, and airlines constitute the trans-
port and communication infrastructure.
Historical Overview A new mode of control emerged: super-
visory control, characterized by monitor-
Task analysis is an activity that has always ing displays that show the status of the
been carried out more by applied researchers machine being controlled. The upswing
than by academic researchers. Academic in this period lasted from 1941 until 1973
psychology often involves research in which (Oil Crisis). The downswing of this era
the experimenters create the tasks. Con- is still continuing.
versely, applied researchers look into their
3 . The age of information and telecommu-
world to investigate the tasks that peo-
nications. Computers, software, telecom-
ple perform in their jobs. Indeed, task
munication equipment, and biotech-
analysis originated in the work of the
nology are the leading branches of
very first industrial psychologists, includ-
the economy. The internet has become
ing Wundts student Hugo Munsterberg
the major communication infrastructure.
(see Hoffman & Deffenbacher, 1992). For
Equipment is cognitively controlled, in
instance, early research conducted by the so-
the sense that users need to draw on
called psychotechnicians (Munsterberg,
extensive knowledge of the environment
1914) involved studies of the tasks of railway
and the equipment. Automation grad-
motormen, and for that research, one of the
ually takes on the form of intelligent
very first simulators was created.
cooperation. This period started around
The applied focus and origins may be
1970 with the emergence of cognitive
because the ultimate goal of task analysis
engineering, and still continues.
is to improve something be it selection,
training, or organizational design. Given the
Each of these periods has witnessed its typi-
applied nature of task analysis, one may
cal task-analysis methods, geared toward the
hypothesize that there is a close connec-
technology that was dominant during that
tion between the focus of task analysis and
period. In the historical overview that fol-
current technological, economical, political,
lows, I will use the breakdown into three
and cultural developments. One fairly com-
periods discussed above.
mon characterization of the past 100 years
is the following breakdown in three periods
(Freeman & Louca, 2001; Perez, 2002): The Age of Steel
1. The age of steel, electricity, and heavy Around 1900, Frederick Winslow Taylor
engineering. Leading branches of the observed that many industrial organizations
economy are electrical equipment, heavy were less profitable than they could be
engineering, heavy chemicals, and steel because of a persistent phenomenon that
products. Railways, ships, and the tele- he termed soldiering, that is, deliberately
phone constitute the transport and com- working slowly (Taylor, 1911/1998). Workers
munication infrastructure. Machines are in those days were not rewarded for work-
manually controlled. This period, dur- ing faster. Therefore, there was no reason to
ing which industrial psychology emerged do ones best, as Taylor noted. Workers also
(e.g., Viteles, 193 2), lasted from approxi- developed their own ways of working, lar-
mately 1895 1940. gely by observing their fellow workers.
task analysis 187
machine, perhaps even down to the level of ing categories: cues initiating action, con-
individual keystrokes (e.g., Card, Moran, & trols used, response, feedback, criterion of
Newell, 1983 ). Job analysis and task analy- acceptable performance, typical errors. The
sis may use the same methods, for instance, method was of general applicability, but
interviews, work observation, and critical was specifically designed for use in planning
incidents. However, as mentioned above, for training and training equipment. Miller
these methods address different levels of adopted a systems approach to task analysis,
aggregation. viewing the human as part of the systems
linkages from input to output functions.
The Age of Oil In his task-analysis phase, Miller included
cognitive concepts such as goal orienta-
It was not until after World War II that task tion and set, decisions, memory storage,
analysis and human factors (the preferred coordinations, and anticipations. These
term in North America) or ergonomics factors in task structure, as he called the
(the preferred term in Europe) began to concepts, are, to different degrees, inevitable
take on a decidedly more cognitive form. parts of every task. The task analyst needs
This was initiated by the development of to translate the set of task requirements
information-processing systems and com- listed in the task description into task-
puting devices, from the stage of manual structure terms. The next step would be
control to the stage of supervisory control to translate the task-structure terms into
(Hollnagel & Cacciabue, 1999). Although selection procedures, training procedures,
Tayloristic approaches to task analysis were and human engineering. Take, for instance,
still sufficient in most of the work con- the task of troubleshooting. Miller provided
ducted in the first half of the twentieth cen- some classical suggestions on how to train
tury (when machines were manually con- the problem-solving part of troubleshooting.
trolled), the development of instrumented One suggestion was to indoctrinate by con-
cockpits, radar displays, and remote process cept and practice to differentiate the func-
control forced the human into a supervi- tion from the mechanism that performs the
sory role in which knowledge and cognition function (Miller, 1962, p. 224). Although
were more important than manual labor, and too general to be useful as a concrete train-
conditional branchings of action sequences ing suggestion, this example predates later
were more important than strictly linear concepts such as the abstraction hierarchy
sequences of actions. Experience in World introduced by Jens Rasmussen in 1979 (see
War II had shown that systems with well- Vicente, 2001).
trained operators were not always working.
Airplanes with no apparent mechanical fail-
ures flew into the ground, and highly moti- flanagan: critical incident technique
vated radar operators missed enemy con- The applied area of human-factors engineer-
tacts. Apparently, the emphasis on testing ing was less reluctant to adopt cognitive ter-
and training had reached its limits, as had minology than mainstream North American
Taylors implicit philosophy of designing the academic psychology, which at that time
human to fit the machine. Now, experimen- was still impacted by behaviorism. We have
tal psychologists were asked to design the already seen how Millers (195 3 ) approach
machine to fit the human. to task analysis included cognitive con-
cepts. In 195 4, Flanagan published his crit-
miller: task description and task analysis ical incident technique (Flanagan, 195 4).
In 195 3 , Robert B. Miller had developed This is a method for collecting and ana-
a method for task analysis that went lyzing observed incidents having special sig-
beyond merely observable behavior (Miller, nificance. Although the modern-day reader
195 3 ; 1962). Miller proposed that each may associate incidents with severe disasters,
task be decomposed into the follow- this was not Flanagans primary definition.
task analysis 189
During World War II, he and his cowork- significant cognitive activity such as plan-
ers studied reasons for failure in learning ning, diagnosis, and decision making.
to fly, disorientation while flying, failures of In the 195 0s and 1960s, manual-control
bombing missions, and incidents of effective tasks had been taken over by automation.
or ineffective combat leadership. After the Operators became supervisors who were
war, the method was also applied to non- supposed to step in when things went
military jobs, such as dentistry, bookkeeping, wrong. The interesting and crucial parts of
life insurance, and industry. These incidents supervisory-control tasks do not lie with
were collected by interviewing hundreds of the observable behavior, but rather with
participants, resulting in thousands of inci- unobservable cognitive activities such as
dent records. Alternative methods of data state recognition, fault finding and schedul-
collection were group interviews, question- ing of tasks during start-up and shutdown
naires, and written records of incidents as sequences. Training for these tasks therefore
they happened. These incidents were then needed to be based on a thorough exami-
used to generate critical job requirements, nation of this cognitive activity. Annett and
which in turn were used for training pur- Duncan felt the existing methods for task
poses, job design, equipment design, mea- analysis (such as time and motion study
sures of proficiency, and to develop selec- and Millers method) were inadequate to
tion tests. Flanagan (195 4) did not provide address these issues. Also, they were more
much detail on the reliability and valid- clear about the need for task descriptions to
ity of his technique, although he empha- involve hierarchies (i.e., conditional branch-
sized the importance of the reporting of ings versus linear sequences.) Hence hier-
facts regarding behavior rather than resting archical task analysis. Complex systems are
solely on subjective impressions. His tech- designed with goals in mind, and the same
nique demonstrates the importance of using goals may be pursued by different routes.
domain experts as informants about any Hence, a direct listing of activities may be
behavior that makes a significant contribu- misleading (they may be sufficient for rou-
tion to the work that is carried out. tine repetitive tasks, though). The analyst
therefore needs to focus on the goals.
Goals may be successively unpacked to
hierarchical task analysis reveal a nested hierarchy of goals and sub-
Although R. B. Miller had used cognitive goals. For example, thirst may be the condi-
concepts in his method for task analysis, his tion that activates the goal of having a cup of
task descriptions were still tied very much tea, and subgoals are likely to include obtain-
to actual humanmachine interaction. His ing boiling water, a teapot with tea, and so
task descriptions would therefore basically on. We may answer the question why we
be lists of physical activities. His concept need boiling water by referring to the top-
of user goals had more to do with the cri- level goal of having a cup of tea. The analyst
teria of system performance that the user needs to ask next how to obtain boiling water.
had to meet, than with a nested set of inter- Whether the analyst needs to answer this
nal goals that drives user performance. A question is dependent on the purpose of the
method for task analysis that began by iden- analysis. If the purpose is to train someone
tifying the goals of the task was developed in who has never before made a cup of tea, then
the 1960s by Annett and Duncan under the the subgoal of obtaining boiling water itself
name of Hierarchical Task Analysis (HTA) needs to be unpacked further, for instance:
(Annett & Duncan, 1967). In accordance pour water in container, heat water, look for
with the dominant industries during this bubbles.
period (the Age of Oil), HTA was origi- Since a general purpose of HTA is to iden-
nally developed for training process-control tify sources of actual or potential perfor-
tasks in the steel and petrochemical indus- mance failure, Annett and Duncan (1967)
tries. These process-control tasks involved formulated the following stop rule: stop with
190 the cambridge handbook of expertise and expert performance
the analysis when the product of the prob- cover possible sources of performance fail-
ability of failure (p) and the cost of failure ure and solutions can be hypothesized.
(c) is judged acceptable. In the example For instance, one may discover that heat-
above, if we needed to train a child in making ing water with an electrical boiler in fact
a cup of tea, we might judge the product of p requires fairly extensive knowledge about
and c to be acceptable for the subgoals of electricity and the hazards associated with
pouring water in the container and looking the combination of water and electricity.
for bubbles. However, we may have some Based on current literature on training, and
doubts about the subgoal of heating the in particular training children, the analyst
water: a child may not know how to oper- may finally suggest some ways of educat-
ate the various devices used for boiling water ing children in the dangers of using electrical
(probability of failure is high); moreover, the boilers when making a cup of tea.
cost of failure may be high as well (burning To take a more complex example than
fingers and worse). The analyst will therefore that of making a cup of tea, and illustrat-
decide to further decompose this subgoal, ing the output of HTA in a graphical for-
but not the other subgoals. By successively mat, consider part of the HTA in Figure 11.1
decomposing goals and applying the p c for operating a continuous-process chemical
criterion at each step, the analyst can dis- plant (after Shepherd, 2001).
task analysis 191
This example is deliberately simplified in by George Miller and Jerome Bruner (see
that it does not show the order in which Gardner, 1985 ; Hoffman & Deffenbacher,
subgoals are pursued. A typical HTA would 1992; Newell & Simon, 1972 for histori-
include a plan that does specify that order. cal overviews). Neissers Cognitive psychol-
HTA may best be described as a generic ogy had appeared in 1967, and the scien-
problem-solving process. It is now one tific journal by the same name first appeared
of the most familiar methods employed in 1970. It took one more decade for this
by ergonomics specialists in the United approach to receive broader methodolog-
Kingdom (Annett, 2004). However, evalu- ical justification and its practical applica-
ation studies have shown that HTA can be tion. In 1984, Ericsson and Simon (1984)
very time intensive compared to other meth- published Protocol analysis: Verbal reports
ods such as observation and interview. HTA as data. This book reintroduced the use
is certainly far from simple and takes both of think-aloud problem-solving tasks, which
expertise and practice to administer effec- had been relegated to the historical dust-
tively (Annett, 2003 ). There is also a good bin by behaviorism even though it had
deal of variability in the application of HTA. some decades of successful use in psychol-
The reader may have had different thoughts ogy laboratories in Germany and elsewhere
than the writer of this chapter when read- in Europe up through about 1925 . In 1983 ,
ing about the particular decomposition of Card, Moran, and Newell published The psy-
the subgoal of obtaining boiling water: why chology of humancomputer interaction. This
not describe a particular procedure for a book helped lay the foundation for the
particular way of boiling (e.g., pour water field of cognitive science and presented the
in pan, put pan on stove, turn on stove, GOMS model (Goals, Operators, Methods,
wait until water boils)? One obvious reply and Selection rules), which was a family of
would be that this description is less gen- analysis techniques, and a form of task analy-
eral than the one offered above because that sis that describes the procedural, how-to-do-
description talks about containers in gen- it knowledge involved in a task (see later sec-
eral. Furthermore, the actions are less precise tion and Kieras, 2004, for a recent overview).
(does one need to set the stove to a partic- Task analysis profited a lot from the
ular setpoint?), and the conditions indicat- developments in artificial intelligence, par-
ing goal attainment are vague (how does one ticularly in the early 1980s when expert
see that the water boils?). If there can be systems became commercially interesting
disagreement with such a simple example, (Hayes-Roth, Waterman, & Lenat, 1983 ).
imagine what problems an analyst can run Since these systems required a great deal
into when dealing with a complex process- of expert knowledge, acquiring or elicit-
control task, such as the example above of ing this knowledge became an important
the chemical plant. topic (see Hoffman & Lintern, Chapter 12).
One of the pitfalls in applying HTA is the Because of their reliance on unstructured
fact that one may lose sight of the problem- interviews, system developers soon viewed
solving nature of the task analysis itself. This knowledge elicitation as the bottleneck
is not a critique of HTA as such, but rather in expert-system development, and they
a cautionary note that analysts need to keep turned to psychology for techniques that
the purpose of the study in sight throughout helped elicit that knowledge (Hoffman,
the analysis. 1987). As a result, a host of individual tech-
niques was identified (see Cooke, 1994, for
a review of 70 techniques), but no single
The Age of Information Processing
overall method for task analysis that would
In the early 1970s, the word cognitive guide the practitioner in selecting the right
became more acceptable in American aca- technique for a given problem resulted from
demic psychology, though the basic idea had this effort. However, the interest in the
been established at least a decade earlier knowledge structures underlying expertise
192 the cambridge handbook of expertise and expert performance
experts and existing systems with experi- studies of expert and novice operators in a
enced users to be analyzed. However, new paper mill, found evidence for a structured
systems for which there are no experts are approach to troubleshooting by experts. She
being developed with greater frequency, and presented experts and novices with realis-
urgency. tic alarms on paper and asked them to think
These issues have been taken up by the aloud. Consider the following protocol by
cognitive systems engineering approach. At its a novice when confronted with the alarm:
core, cognitive systems engineering seeks conveyor belt of pulper 1 broke down:
to understand how to model work in ways
directly useful for design of interactive sys- I would . . . I would stop the pulper to start
tems (Eggleston, 2002, p. 15 ). Egglestons with and then I would halt the whole cycle
afterwards and then try to repair the con-
useful overview of the field distinguishes
veyor belt . . . but you have to halt the whole
three phases in the development of cog- installation, because otherwise they dont
nitive systems engineering: (1) a concep- have any stock anymore.
tual foundations period that occurred largely
in the 1980s, (2) an engineering prac- An expert confronted with the same prob-
tice period that dominated the 1990s, and lem reacted as follows:
(3 ) an active deployment period that started
around 2000. Cognitive task analysis fig- OK. Conveyor belt of pulper 1 broke
ures prominently in the engineering practice down . . . conveyor belt of pulper 1 . . . if that
one breaks down . . . yeah . . . see how long
period of cognitive systems engineering.
that takes to repair . . . not postponing the
However, whereas traditional cognitive decision for very long, to ensure we dont
task analysis focuses primarily on under- have to halt the installation.
standing the way people operate in their
current world, cognitive systems engineer- The novice starts repairs that are not nec-
ing focuses also on understanding the essary at all given the situation, whereas
way the world works and the way in the expert first judges the seriousness of
which new envisioned worlds might work the problem. These and similar statements
(Potter, Roth, Woods, & Elm, 2000). led to the inclusion of the category judg-
With the discussion of cognitive task anal- ing the seriousness of the problem in the
ysis and cognitive systems engineering, we experts task structure of the diagnostic task.
have reached the present-day status of task As novices rarely showed this deliberation,
analysis. The next section will describe a this category did not appear in their task
number of case studies that exemplify the structure.
use of task analysis methods. The complete task structure is as follows
(see Figure 11.2).
Experts in a paper mill first start by mak-
Case Studies ing a judgment about the seriousness of the
problem. If the problem is judged to be seri-
ous, the operator will immediately continue
In this section, I will describe various case
with the application of a global repair, fol-
studies on task analysis, with the aim, first, to
lowed by an evaluation whether the prob-
provide the reader with some ideas on how
lem has been solved. This process may be
to carry out a task analysis, and second, to
followed by a more thorough diagnosis in
note some of the difficulties one encounters
order to determine the correct local repair,
when carrying out a task analysis in complex
ensuring a solution once and for all. If the
domains.
problem is not a very serious one, the expert
will consider possible faults one by one and
Improving the Training of Troubleshooting
test them, until a likely one is found. This is
The first case study is in the domain of then followed by a determination of repairs,
troubleshooting. Schaafstal (1993 ), in her their consequences, an ordering of repairs
194 the cambridge handbook of expertise and expert performance
symptom
yes
Judgment: serious problem?
Possible faults
no
Testing: is this the fault?
Determination of repairs
Consequences of repairs
Ordering of repairs
no no
Evaluation: problem solved?
EXIT
serious problem
(if necessary), application of repairs, and an he may realize that he has not yet spotted the
evaluation whether the problem has been actual fault, and therefore the problem has
solved. If the problem has not been solved, not been solved. In case no possible faults
the expert might do two things: either try are left, or the operator cannot think of any
another repair, or back up higher in the tree other faults than the ones he already tested,
task analysis 195
he will be inclined to use a global repair to used both experts and novices (trainees
alleviate the problem. who had just finished a course) in order
Inexperienced operators show a far more to uncover differences in the knowledge
simple diagnostic strategy. They dont judge and strategies employed. Our task-analysis
the seriousness of the problem, they dont method was to have technicians think aloud
consider the consequences of repairs, and while troubleshooting two malfunctions in a
they dont evaluate whether the problem has radar system. The resulting verbal data were
been solved. Also, novices jump much more analyzed by protocol analysis, that is, by iso-
quickly to repairs without realizing whether lating and categorizing individual proposi-
a certain repair actually is right for a certain tions in the verbal protocol.
situation. The categories we used for classifying the
We applied this expert task structure to propositions were derived from the expert
another area of troubleshooting (Schaafstal, task structure as shown in Figure 11.2. The
Schraagen, & Van Berlo, 2000). Around radar study showed that a theory instructor
1990, complaints started to emerge from the who was one of our participants had dif-
Dutch fleet concerning the speed and accu- ficulties troubleshooting this radar system.
racy of weapon engineers, who carry out This turned our attention to a gap between
both preventive and corrective maintenance theoretical instruction and practice. We also
onboard naval vessels. There were a num- observed virtually no transfer of knowledge
ber of reasons for the suboptimal perfor- from one radar system to the other, as wit-
mance of the troubleshooters. First, exper- nessed by the unsuccessful troubleshoot-
tise was not maintained very well. Engineers ing attempts of two participants who were
shifted positions frequently, left military ser- experienced in one radar system but not in
vice for more lucrative jobs in the civilian the radar system we studied. This turned
world, or were less interested in a technical our attention to the content of the train-
career in the first place. Second, a new gener- ing courses, which were component oriented
ation of highly integrated systems was intro- instead of functionally oriented. Finally, the
duced, and this level of integration made verbal protocols showed the typical unsys-
troubleshooting more demanding. Third, tematic approach to troubleshooting by the
the training the troubleshooters received novice participant in our study.
seemed inadequate for the demands they These studies provided a glimpse of what
encountered onboard ships. was wrong with the courses in troubleshoot-
We conducted a field study with real ing. They were highly theoretical, compo-
faults in real systems that showed that nent oriented, with little practice in actual
military technical personnel who had just troubleshooting. On the basis of our obser-
completed a course and passed their exam vations and experiments, we decided to
diagnosed only 40% of the malfunctions cor- change the courses. Basically, we wanted to
rectly. We also obtained scores on a knowl- teach the students two things: (1) a sys-
edge test, and found that the junior techni- tematic approach to troubleshooting, (2) a
cians scored only 5 5 % correct on this test. Of functional understanding of the equipment
even more importance was the low correla- they have to maintain. In our previous study
tion (0.27) between the scores on the knowl- (Schraagen & Schaafstal, 1996), we had
edge test and the actual troubleshooting per- found that the systematic approach to trou-
formance. This cast doubt on the heavy bleshooting could not be taught indepen-
emphasis placed on theory in the training dently of a particular context. In order to be
courses. able to search selectively in the enormous
Our suspicions about the value of theory problem space of possible causes, it is essen-
in the training courses were further raised tial that the representation of the system be
after having conducted a number of obser- highly structured.
vational studies (see Schraagen & Schaafstal, One candidate for such a structuring
1996: Experiment 1). In these studies, we is a functional hierarchical representation,
196 the cambridge handbook of expertise and expert performance
much like Rasmussens (1986) abstraction course). These results were most satisfy-
hierarchy (see Hoffman & Lintern, Chap- ing, especially considering the fact that the
ter 12). For a course on a computer system, new course lasted four weeks instead of
we decomposed the system into four levels, six weeks.
from the top-level decomposition of a com- The Naval Weapon Engineering School,
puter system into power supply, central pro- convinced by the empirical results, subse-
cessor, memory, and peripheral equipment, quently decided to use this method as the
down to the level of electrical schemata. We basis for the design of all its function courses.
stopped at the level of individual replace- We have helped them with the first few
able units (e.g., a printed circuit board). In courses, and subsequently wrote a man-
this way, substantial theoretical background ual specifying how to develop new courses
that was previously taught could be elimi- based on our philosophy, but gradually the
nated. We replaced this theory with more naval instructors have been able to mod-
practice in troubleshooting itself. Students ify courses on their own. Course length for
were instructed to use a troubleshooting more than 5 0 courses has on average been
form as a job aid. This form consisted sim- reduced by about 3 0%.
ply of a sheet of paper with four different As far as task analysis is concerned, the
steps to be taken in troubleshooting (prob- reader may have noted that little mention
lem description, generate causes, test causes, has been made of any extensive task decom-
repair and evaluate). These four steps were a position. Yet, this project could not have
high-level abstraction of the diagnostic task been as successful as it was without a cog-
structure previously identified in the paper- nitive task analysis of troubleshooting on
mill study by Schaafstal (1993 ). In this way, the part of highly proficient domain prac-
the systematic approach to troubleshooting titioners. Troubleshooting is first and fore-
was instilled in the practice lessons, while most a cognitive task. Little can be observed
at the same time a functional understand- from the outside, just by looking at behavior.
ing of the system was instilled in the the- Observations and inferences are all knowl-
ory sessions. Theory and practice sessions edge based. We therefore almost always
were interspersed such that the new theo- used think-aloud problem solving followed
retical concepts, once mastered, could then by protocol analysis as the data analysis
be readily applied to troubleshooting the method.
real system.
We demonstrated to the Navy the suc-
Shore-based Pilotage
cess of this approach in a one-week add-
on course: the percentage of problems In 1992, Rotterdam Municipal Port Manage-
solved went up from 40% to 86%. Sub- ment, with cooperation by the Rotterdam
sequently, we were asked to completely Pilots Corporation, ordered an investigation
modify the computer course according to into the possibilities of extending shore-
our philosophy. Again, we evaluated this based pilotage (Schraagen, 1993 ). Normally,
new course empirically by having students a pilot enters a ship and conns the ship
think aloud while solving four problems, from the sea to the harbor entrance. The
and rating their systematicity of reason- expertise of a pilot lies in his or her exten-
ing and level of functional understanding. sive knowledge of the particular conditions
Results were highly favorable for our mod- of a specific harbor or port. The expertise
ified course: 95 % of malfunctions were cor- of the ships captain lies primarily in his or
rectly solved (previously 40%), and experts her extensive knowledge of a specific ship.
rated the students systematicity and level Because of rough seas, situations sometimes
of functional understanding each at 4.8 arise where it is too dangerous for the pilot
on a 15 scale (whereas these numbers to board the ship himself. In those situa-
were 2.6 and 2.9, respectively, for the old tions, either the pilot is brought on board
task analysis 197
by a helicopter, or the ship is piloted by a cooperated during the trips. During the trips
pilot ashore. The latter is called shore-based onboard ships, an extensive array of mea-
pilotage. Extending shore-based pilotage surements was taken:
would reduce costs for shipping companies
since shore-based pilotage, at least at the (a) Pilots were instructed to talk aloud; their
time of the study, was cheaper than being verbalizations were tape-recorded.
served by a helicopter or waiting in a har- (b) Pilot decisions and communication were
bor for better weather. However, cost reduc- timed with a stopwatch.
tion should be weighed against decreased (c) Pilots were asked about their plans
levels of safety as a result of the pilot not before entering the ship and were inter-
being on the bridge himself. In particular, viewed afterwards about the trip made.
in bad weather conditions, the captain has (d) The ships movements and positions
no overview of the traffic image in relation were recorded.
to the environment. Sometimes, large drift (e) Recordings were made of the move-
angles are seen by the captain. These angles ments of ship traffic (via radar).
are sometimes not accepted by the captain
(f) Photographs of the view ahead were
and advice given by the pilot is not followed
taken from the vantage point of the
up. The problem is that the captain may not
helm: photographs were taken every five
be familiar with the local situation.
minutes in order to obtain a running
One important element in the study was
record of the pilots perceptual input.
to specify the extra information required
by pilots if shore-based pilotage would be After the trips had been made, seven
extended to ships exceeding 170 m in length. pilots participated in follow-up sessions. The
Ideally, a simulator study would be required primary aim was to obtain more detailed
in which one could systematically vary the information on pilot information usage than
information available to the pilot, variables had been possible during the trips (detailed
such as ship length and height, traffic den- interviews were impossible, of course, since
sity, wind, current, and visibility conditions, pilots were doing their jobs). A secondary
and the quality of interaction between pilot benefit was that data could be compared, in
and captain. However, this kind of study ex- contrast to the trip data that differed in terms
ceeded the projects budget, so we actually of ship, weather, and traffic conditions. In the
undertook a literature study, a comparison follow-up session, pilots were asked to carry
with air traffic control, a study of vessel- out the following tasks:
based systems, and a task analysis. The pur-
pose of the task analysis was to find out what (a) Reconstruct the exact rudder advice
information pilots used onboard the ships. given during a trip using fragments of
The selection of expert pilots that par- video recordings as input (fragments of
ticipated in the task analysis was largely curved situations lasting four to ten min-
determined by the Pilots Corporation, based utes were presented on a TV monitor).
on a few constraints that we put forward. (b) Indicate on a map of the trajectory they
First, we needed pilots with at least ten were familiar with at what points course
years of practical experience. Second, we and speed changes were made.
needed pilots who were proficient commu- (c) Draw on a map the course over ground
nicators, so they could explain what they together with acceptable margins under
were doing. various wind and current conditions for
The task analysis consisted of two parts the entrance into Hook of Holland.
(see Schraagen, 1994, for more details): (1)
observation and recording of pilot activities These tasks represent what Hoffman
on eleven trips on ships, (2) paper and pen- (1987) has called limited-information tasks.
cil tasks given to seven pilots who had also Limited-information tasks are similar to
198 the cambridge handbook of expertise and expert performance
the task the expert is familiar with, but the their trips, so that they could convince us of
amount or kind of information that is avail- the added value of having experts on board.
able to the expert is somehow restricted. In another project that was to have started
For example, the video recordings were a year later, their knowledge of how to conn
taken from a fixed point in the middle of the a ship into the harbor was required for pro-
bridge, whereas pilots would normally stand ficiency testing, that is, training captains of
at the righthand side of the bridge looking ships to conn their own ships into the har-
at the starboard side of the river. Although bor without the assistance of a pilot. Pilot
experts may initially feel uncomfortable participation in this project was withdrawn
with limited-information tasks, the tasks after a few months and the entire project
can be informative in revealing practitioner was cancelled. In the end, this project may
reasoning (Hoffman, Shadbolt, Burton, & have been used by the Port Authorities to
Klein, 1995 ). pose a threat to the pilots: if you dont lower
The task analysis yielded a wealth of infor- your rates for helicopter assistance (the heli-
mation on the kinds of information pilots copter was leased by the pilots), we will
actually use when navigating. The most extend shore-based pilotage. It seems that
important references used were pile moor- this threat worked. Shore-based pilotage was
ings, buoys, and leading lines. Several unex- not extended, hence the results of this study
pected results emerged. First, pilots develop were not implemented.
highly individualized references both to ini- By spending time with the pilots, it would
tiate course and speed changes and to check be easy for the researchers to develop loy-
against the ships movements. Although all alties with them and their organization,
pilots rely on pile moorings, buoys, and lead- rather than with the Port Authorities who
ing lines, which of these they use differs remained at a distance. In general, the task
greatly among them. This is perhaps due analyst who is studying expertise in context
to their individualized way of training. Sec- needs to be aware of these agenda issues.
ond, one might hypothesize that the deci- A second lesson learned is that obtaining
sion points mentioned by pilots on paper too many data can be counterproductive. In
constitute only a fraction of, or are differ- this project, for instance, the video record-
ent in nature from, the decision points used ings that were made of the synthetic radar
during actual pilotage onboard a ship. This image in the Harbor Coordination Center
latter possibility turned out not to be the were never analyzed afterwards, although it
case. Decision points used in actual prac- seemed potentially valuable at the time we
tice were all covered by the decision points started. Second, the timing of the pilot deci-
mentioned on paper. This implies that this sions with a stopwatch, although laudable
kind of knowledge is not tacit or difficult from a quantitative point of view, was really
to verbalize. unnecessary given the focus of the project on
More interesting than the precise results the qualitative use of categories of informa-
of the task analysis, at least for this chap- tion. Hindsight is always 20/20, but the gen-
ters purposes, are the lessons learned. First, eral lesson for task analysts is to think twice
this was a politically very sensitive project. before recording massive amounts of infor-
It turned out that the sponsor, Rotterdam mation just because the gathering of certain
Port Authorities, had a different political data types might be possible. Information
agenda than the Rotterdam pilots. The Port gathering should be driven by the goals of
Authorities wanted to reduce shipping costs the research.
in order to increase total amount of cargo Finally, the paper and pencil tasks were
handling. The pilots, on the other hand, received with downright hostility by the
publicly said they were afraid of jeopardiz- pilots. They had been forced to spend their
ing safety in case shore-based pilotage was spare time on this exercise, and when they
extended. They therefore offered their full noted certain inadequacies in the infor-
assistance by allowing us to follow them on mation provided to them on paper, they
task analysis 199
became uncooperative and very critical. This tems design is still a critical one. Recently,
required some subtle people-handling skills interesting work has been carried out
on the part of the task analyst. In retro- on integrating task analysis with standard
spect, it would have been better to first talk software-engineering methods such as Uni-
through the materials with an independent fied Modeling Language (UML) (see Diaper
pilot in order to remove the inadequacies. & Stanton, 2004, Part IV).
This confirms a conclusion already drawn A second issue regarding the gulf between
in 1987 by Hoffman that experts do not task analysis and design concerns the devel-
like it when you limit the information that opment of systems that do not yet exist. Task
is available to them ( . . . ). It is important analyses generally work well when experts
when instructing the expert to drive home can be interviewed who are experienced
the point that the limited-information task is with current systems. However, with novel
not a challenge of their ego or of their exper- systems, there are no experts. If introduc-
tise (Hoffman, 1987, p. 5 6). ing new technology changes tasks, the anal-
In another cognitive task analysis, geared ysis of a current task may be of limited use
toward discovering the search strategies in the design of new sociotechnical systems
employed by forensic analysts, we also used (Woods & Dekker, 2000). Therefore, a some-
limited-information tasks, this time with- what different set of techniques is required
out encountering resistance (Schraagen & for exploring the envisioned world, includ-
Leijenhorst, 2001). This may have been due ing storyboard walkthroughs, participatory
to the fact that the cases that were presented design, and high-fidelity simulations using
to the experts were developed in close coop- future scenarios.
eration with a forensic analyst who was not
part of the study participants. Also, their
familiar task, by definition, involves working References
with limited information.
Annett, J. (2000). Theoretical and pragmatic
influences on task analysis methods. In J. M.
Conclusions and Future Work Schraagen, S. F. Chipman, & V. L. Shalin (Eds.),
Cognitive task analysis (pp. 25 3 7). Mahwah,
Where do we stand? Although it may be too NJ: Lawrence Erlbaum Associates.
early to tell, we may have shifted from the Annett, J. (2003 ). Hierarchical task analysis. In
age of information to the age of knowledge E. Hollnagel (Ed.), Handbook of cognitive task
sharing or innovation. Task analysis now has design (pp. 173 5 ). Mahwah, NJ: Lawrence
a focus of understanding expert knowledge, Erlbaum Associates.
reasoning, and performance, and leveraging Annett, J. (2004). Hierarchical task analysis. In
that understanding into methods for training D. Diaper & N. Stanton (Eds.), The handbook
and decision support, to amplify and extend of task analysis for humancomputer interaction
(pp. 6782). Mahwah, NJ: Lawrence Erlbaum
human abilities to know, perceive, and col-
Associates.
laborate. To do this, we have an overarch-
Annett, J., & Duncan, K. D. (1967). Task analy-
ing theory macrocognition (Klein, et al.,
sis and training design. Occupational Psychology,
2003 ), and a rich palette of methods, with 41, 211221.
ideas about methods strengths and limita-
Card, S. K., Moran, T. P., & Newell, A. (1983 ).
tions, and methods combinatorics. Task anal- The psychology of humancomputer interaction.
ysis, and cognitive task analysis in particular, Hillsdale, NJ: Lawrence Erlbaum Associates.
are both useful and necessary in any investi- Chipman, S. F., Schraagen, J. M., & Shalin, V.
gation of expertise in the wild. L. (2000). Introduction to cognitive task anal-
Despite this generally positive outlook, ysis. In J. M. Schraagen, S. F. Chipman, &
there are several lingering issues that deserve V. L. Shalin (Eds.), Cognitive task analysis
future work. First, the issue of bridging (pp. 3 23 ). Mahwah, NJ: Lawrence Erlbaum
the gulf between task analysis and sys- Associates.
2 00 the cambridge handbook of expertise and expert performance
Cooke, N. J. (1994). Varieties of knowledge tional Behavior and Human Decision Processes,
elicitation techniques. International Journal of 62 , 12915 8.
HumanComputer Studies, 41, 801849. Hoffman, R. R., Crandall, B. W., & Shadbolt,
Diaper, D. (2001). Task Analysis for Knowl- N. R. (1998). A case study in cognitive task
edge Descriptions (TAKD): A requiem for a analysis methodology: The critical decision
method. Behavior and Information Technology, method for elicitation of expert knowledge.
2 0, 199212. Human Factors, 40, 25 4276.
Diaper, D. (2004). Understanding task analysis Hoffman, R. R., & Woods, D. D. (2000). Study-
for humancomputer interaction. In D. Diaper ing cognitive systems in context: Preface to the
& N. Stanton (Eds.), The handbook of task anal- special section. Human Factors, 42 , 17 (Special
ysis for humancomputer interaction (pp. 5 47). section on cognitive task analysis).
Mahwah, NJ: Lawrence Erlbaum Associates. Hollnagel, E. & Cacciabue, P. C. (1999). Cogni-
Diaper, D., & Stanton, N. (2004). Wishing on a tion, technology & wrok: An introduction. Cog-
sTAr: The future of task analysis. In D. Diaper & nition, Technology & Work, 1(1), 16.
N. Stanton (Eds.), The handbook of task analysis Hollnagel, E. (2003 ). Prolegomenon to cognitive
for humancomputer interaction (pp. 603 619). task design. In E. Hollnagel (Ed.), Handbook of
Mahwah, NJ: Lawrence Erlbaum Associates. cognitive task design (pp. 3 15 ). Mahwah, NJ:
Drury, C. G., Paramore, B., Van Cott, H. P., Grey, Lawrence Erlbaum Associates.
S. M., & Corlett, E. N. (1987). Task analysis. In Kieras, D. (2004). GOMS models for task anal-
G. Salvendy (Ed.), Handbook of human factors ysis. In D. Diaper & N. A. Stanton (Eds.),
(pp. 3 71401). New York: John Wiley & Sons. The handbook of task analysis for human
Eggleston, R. G. (2002). Cognitive systems engi- computer interaction (pp. 83 116). Mahwah, NJ:
neering at 20-something: Where do we stand? Lawrence Erlbaum Associates.
In M. D. McNeese & M. A. Vidulich (Eds.), Kirwan, B., & Ainsworth, L. K. (Eds.). (1992). A
Cognitive systems engineering in military aviation guide to task analysis. London: Taylor & Francis.
environments: Avoiding cogminutia fragmentosa!
Klein, G. (1993 ). A recognition-primed deci-
(pp. 15 78). Wright-Patterson Air Force Base,
sion (RPD) model of rapid decision making.
OH: Human Systems Information Analysis
In G. Klein, J. Orasanu, R. Calderwood, &
Center.
C. E. Zsambok (Eds.), Decision making in
Ericsson, K. A., & Simon, H. A. (1984). Proto- action: Models and methods (pp. 13 8147).
col analysis: Verbal reports as data. Cambridge, Norwood, NJ: Ablex.
MA: MIT Press.
Klein, G. A., Calderwood, R., & Macgregor, D.
Flanagan, J. C. (195 4). The critical incident tech- (1989). Critical decision method for eliciting
nique. Psychological Bulletin, 5 1, 3 273 5 8. knowledge. IEEE Transactions on Systems, Man,
Freeman, C., & Louca, F. (2001). As time goes and Cybernetics, 19, 462472.
by: From industrial revolutions to the information
Klein, G., Ross, K. G., Moon, B. M., Klein, D. E.,
revolution. Oxford: Oxford University Press.
Hoffman, R. R., & Hollanagel, E. (May/June
Gardner, H. (1985 ). The minds new science: A his- 2003 ). Macrocognition. IEEE Intelligent Sys-
tory of the cognitive revolution. New York: Basic tems, pp. 8185 .
Books.
Lesgold, A. (2000). On the future of congni-
Hayes-Roth, F., Waterman, D. A., & Lenat, D. B. tive task analysis. In J. M. Schraagen, S. F.
(Eds.). (1983 ). Building expert systems. Reading, Chipman, & V. L. Shalin (Eds.), Cognitive task
MA: Addison-Wesley Publishing Company. analysis (pp. 45 1465 ). Mahwah, NJ: Lawrence
Hoffman, R. R. (1987, Summer). The problem Erlbaum Associates.
of extracting the knowledge of experts from Luczak, H. (1997). Task analysis. In G.
the perspective of experimental psychology. AI Salvendy (Ed.), Handbook of human fac-
Magazine, 8, 5 3 67. tors and ergonomics (2nd ed.) (pp. 3 40416).
Hoffman, R. R., & Deffenbacher, K. (1992). A New York: John Wiley & Sons.
brief history of applied cognitive psychology. Medsker, G. J., & Campion, M. A. (1997).
Applied Cognitive Psychology, 6, 148. Job and team design. In G. Salvendy (Ed.),
Hoffman, R. R., Shadbolt, N. R., Burton, A. M., Handbook of human factors and ergonomics
& Klein, G. (1995 ). Eliciting knowledge from (2nd ed.) (pp. 45 0489). New York: John
experts: A methodological analysis. Organiza- Wiley & Sons.
task analysis 2 01
Meister, D. (1999). The history of human factors Schaafstal, A. M., Schraagen, J. M., & van Berlo,
and ergonomics. Mahwah, NJ: Lawrence Erl- M. (2000). Cognitive task analysis and innova-
baum Associates. tion of training: The case of structured trou-
Militello, L. G., & Hoffman, R. R. (2006). bleshooting. Human Factors, 42 , 75 86.
Perspectives on cognitive task analysis. Schraagen, J. M. C. (1993 ). What information
Cambridge: MIT Press. do river pilots use? In Proceedings of the Inter-
Militello, L. G., & Hutton, R. J. B. (1998). national Conference on Marine Simulation and
Applied cognitive task analysis (ACTA): A Ship Manoeuvrability MARSIM 93 (Vol. II,
practitioners toolkit for understanding cog- pp. 5 095 17). St. Johns, Newfoundland:
nitive task demands. Ergonomics, 41, 1618 Fisheries and Marine Institute of Memorial
1641. University.
Miller, R. B. (195 3 ). A method for man Schraagen, J. M. C. (1994). What infor-
machine task analysis. Dayton, OH: Wright Air mation do river pilots use? (Report TM
Development Center (Technical Report 5 3 1994 C-10). Soesterberg: TNO Institute for
13 7). Human Factors.
Miller, R. B. (1962). Task description and analysis. Schraagen, J. M. C., & Leijenhorst, H. (2001).
In R. M. Gagne (Ed.), Psychological principles in Searching for evidence: Knowledge and search
system development (pp. 187228). New York: strategies used by forensic scientists. In E. Salas
Holt, Rinehart and Winston. & G. Klein (Eds.), Linking expertise and natural-
Munsterberg, H. (1914). Psychotechnik. Leipzig: istic decision making (pp. 263 274). Mahwah,
J. A. Barth. NJ: Lawrence Erlbaum Associates.
Nemeth, C. P. (2004). Human factors methods for Schraagen, J. M. C., & Schaafstal, A. M.
design: Making systems human-centered. Boca (1996). Training of systematic diagnosis: A case
Raton: CRC Press. study in electronics troubleshooting. Le Travail
Humain, 5 9, 5 21.
Newell, A., & Simon, H. A. (1972). Human prob-
lem solving. Englewood Cliffs, NJ: Prentice- Schraagen, J. M., Chipman, S. F., & Shalin, V. L.
Hall. (2000). Cognitive task analysis. Mahwah, NJ:
Perez, C. (2002). Technological revolutions and Lawrence Erlbaum Associates.
financial capital: The dynamics of bubbles and Seamster, T. L., Redding, R. E., & Kaempf, G. L.
golden ages. Cheltenham: Edward Elgar. (1997). Applied cognitive task analysis in avia-
Potter, S. S., Roth, E. M., Woods, D. D., & tion. London: Ashgate.
Elm, W. C. (2000). Bootstrapping multiple Shepherd, A. (2001). Hierarchical task analysis.
converging cognitive task analysis techniques London: Taylor & Francis.
for system design. In J. M. Schraagen, S. F. Taylor, F. W. (1998). The principles of scientific
Chipman, & V. L. Shalin (Eds.), Cognitive task management (unabridged republication of the
analysis (pp. 3 173 40). Mahwah, NJ: Lawrence volume published by Harper & Brothers, New
Erlbaum Associates. York and London, in 1911). Mineola, NY: Dover
Rasmussen, J. (1986). Information processing and Publications.
humanmachine interaction: An approach to cog- Vicente, K. J. (2001). Cognitive engineering
nitive engineering. Amsterdam: Elsevier. research at Ris from 19621979. In E. Salas
Rouse, W. B. (1988). Adaptive aiding for human/ (Ed.), Advances in human performance and cog-
computer control. Human Factors, 3 0, 43 1 nitive engineering research (Vol. 1, pp. 15 7).
443 . New York: Elsevier.
Schaafstal, A. M. (1991). Diagnostic skill in pro- Viteles, M. S. (193 2). Industrial psychology. New
cess operation: A comparison between experts York: W. W. Norton.
and novices. Unpublished dissertation. Univer- Woods, D. D., & Dekker, S. (2000). Antic-
sity of Groningen, The Netherlands. ipating the effects of technological change:
Schaafstal, A. M. (1993 ). Knowledge and strate- A new era of dynamics for human factors.
gies in diagnostic skill. Ergonomics, 3 6, 13 05 Theoretical Issues in Ergonomics Science, 1,
13 16. 272282.
C H A P T E R 12
2 03
2 04 the cambridge handbook of expertise and expert performance
Where This Topic Came From (Hoffman, 1992), but serious problems soon
arose. For example, software brittleness
(breakdowns in handling atypical cases)
The Era of Expert Systems
and explanatory insufficiency (a printout
The era of expert systems can be dated from of cryptic procedural rules fails to clearly
about 1971, when Edward Feigenbaum and express to non-programmers the reasoning
his colleagues (Feigenbaum, Buchanan, & path that was followed by the software)
Lederberg, 1971) created a system in which were quickly recognized as troublesome (for
a computable knowledge base of domain reviews that convey aspects of the history of
concepts was integrated with an inference this field, see David, Krivine, & Simmons,
engine of procedural (if-then) rules. This 1993 ; Raeth, 1990). At the same time, there
expert system was intended to capture was a burgeoning of interest in expertise on
the skill of expert chemists regarding the the part of cognitive psychologists.
interpretation of mass spectrograms. Other
seminal systems were MYCIN (Shortliffe,
Expertise Studies in Psychology
1976), for diagnosing bacterial infections and
PROSPECTOR (Duda, Gaschnig, & Hart, The application of cognitive science and
1979), for determining site potential for geo- the psychology of learning to topics in
logical exploration. instructional design led to studies of the
It seemed to take longer for computer basis for expertise and knowledge organi-
scientists to elicit knowledge from experts zation at different stages during acquisi-
than to write the expert system software. tion of expertise (Lesgold, 1994; Means &
This knowledge acquisition bottleneck Gott, 1988). In the early 1970s, a group
became a salient problem (see Hayes-Roth, of researchers affiliated with the Learn-
Waterman, & Lenat, 1983 ). It was widely ing Research and Development Center at
discussed in the computer science commu- the University of Pittsburgh and the Psy-
nity (e.g., McGraw & Harbison-Briggs, 1989; chology Department at Carnegie-Mellon
Rook & Croghan, 1989). An obvious sugges- University launched a number of research
tion was that computer systems engineers projects on issues of instructional design in
might be trained in interview techniques both educational contexts (e.g., elementary-
(Forsyth & Buchanan, 1989), but the bottle- school-level mathematics word problems;
neck also spawned the development of auto- college-level physics problems) and techni-
mated knowledge acquisition shells. These cal contexts of military applications (e.g.,
were toolkits for helping domain experts problem solving by electronics techni-
build their own prototype expert systems cians) (e.g., Chi, Feltovich, & Glaser, 1981)
(for a bibliography, see Hoffman, 1992). Lesgold et al., 1981. The research empha-
By use of a shell, experts entered their sized problem-solving behaviors decom-
expert knowledge about domain concepts posed as learning hierarchies (Gagne &
and reasoning rules directly into the com- Smith, 1962), that is, sequences of learn-
puter as responses to questions (Gaines & ing tasks arranged according to difficulty and
Boose, 1988). Neale (1988) advocated elim- direction of transfer.
inating the knowledge engineer and getting Interest in instructional design quickly
the expert to work directly with the com- became part of a larger program of inves-
puter (p. 13 6) because human-on-human tigation that generated several foundational
KE methods (interviews, protocol analysis) notions about the psychology of expertise
were believed to place an unjustified faith (see Glaser, 1987). A number of researchers,
in textbook knowledge and what experts say apparently independently of one another,
they do (p. 13 5 ). began to use the term cognitive task anal-
The field of expert systems involved ysis both to refer to the process of iden-
literally thousands of projects in which tifying the knowledge and strategies that
expert knowledge was elicited (or acquired) make up expertise for a particular domain
eliciting and representing the knowledge of experts 2 05
and task as well as to distinguish the pro- Cacciabue, 1996), although there are rem-
cess from so-called behavioral task analysis nants of debate about the utility and sci-
(e.g., Glaser et al., 1985 ; see Schraagen, this entific foundations of research that is con-
volume). A stream of psychological research ducted in uncontrolled or non-laboratory
evolved that shifted emphasis from stud- contexts (e.g., Banaji & Crowder, 1989;
ies with naive, college-aged subjects who Hoffman & Deffenbacher, 1993 ; Hoffman &
participated in artificial tasks using artificial Woods, 2000).
materials (in service of control and manipu-
lation of variables) to studies in which highly
The Early Methods Palette
skilled, domain-smart participants engaged
in tasks that were more representative of Another avenue of study involved attempts
the complexity of the real world in which to address the knowledge-acquisition bot-
they practiced their craft (Chi, Glaser, & tleneck, the root cause of which lay in the
Farr, 1988; Hoffman, 1992; Knorr-Cetina & reliance on unstructured interviews by
Mulkay, 1983 ; Shanteau, 1992). the computer scientists who were build-
Investigators began to shift their atten- ing expert systems (see Cullen & Bryman,
tion from cataloging biases and limitations 1988). Unstructured interviews gained early
of human reasoning in artificial and sim- acceptance as a means of simultaneously
ple problems (e.g., statistical reasoning puz- bootstrapping the researchers knowledge
zles, syllogistic reasoning puzzles) to the of the domain, and establishing rapport
exploration of human capabilities for mak- between the researcher and the expert. Nev-
ing decisions, solving complex problems, and ertheless, the bottleneck issue encouraged a
forming mental models (Gentner & Stevens, consideration of methods from psychology
1983 ; Klahr & Kotovsky, 1989; Klein & that might be brought to bear to widen the
Weitzenfeld, 1982; Scribner, 1984; Sternberg bottleneck, including methods of structured
& Frensch, 1991). The ethnographic research interviewing (Gordon & Gill, 1997). Inter-
of Lave (1988) and Hutchins (1995 ) revealed views could get their structure from pre-
that experts do not slavishly conduct tasks planned probe questions, from archived test
or adhere to work rules or work proce- cases, and so forth.
dures but instead develop informal heuris- In addition to interviewing, the researcher
tic strategies that, though possibly inefficient might look at expert performance while the
and even counterintuitive, are often remark- expert is conducting their usual or famil-
ably robust, effective, and cognitively eco- iar task and thinking aloud, with their
nomical. One provocative implication of this knowledge and reasoning revealed subse-
work is that expertise results in part from a quently via a protocol analysis (see Chi
natural convergence on such strategies dur- et al., 1981; Ericsson & Simon, 1993 , Chap-
ing engagement with the challenges posed ter 3 8, this volume). In addition, one
by work. could study expert performance at con-
Studies spanned a wide gamut of top- trived tasks, for example, by withholding
ics, some of which seem more traditional certain information about the case at hand
to academia (e.g., physics problem solving), (limited-information tasks), or by manip-
but many that would traditionally not be fair ulating the way the information is pro-
game for the academic experimental psy- cessed (constrained-processing tasks). In the
chologist (e.g., expertise in manufacturing method of tough cases the expert is asked
engineering, medical diagnosis, taxicab driv- to work on a difficult test case (perhaps
ing, bird watching, grocery shopping, natural gleaned from archives) with the idea that
navigation). Mainstream cognitive psychol- tough cases might reveal subtle aspects of
ogy took something of a turn toward appli- expert reasoning, or particular subdomain
cations (see Barber, 1988), and today the or highly specialized knowledge, or aspects
phrase real world seems to no longer of experts metacognitive skills, for exam-
require scare quotes (see Hollnagel, Hoc, & ple, the ability to reason about their own
2 06 the cambridge handbook of expertise and expert performance
Table 12 .2 . A Sample of a Coded CDM Protocol (Adapted from Klein et al., 1989)
Appraisal This is going to be a tough fire,
Cue and we may start running into heat exhaustion problems.
Cue It is 70 degrees now and it is going to get hotter.
Action The first truck, I would go ahead and have them open the roof up,
Action and the second truck I would go ahead and send them inside and
Action have them start ventilating, start knocking the windows out and working
Elaboration with the initial engine crew, false ceilings, and get the walls opened up.
Action As soon as I can, order the second engine to hook up to supply and pump to
engine 1.
Anticipation I am assuming engine 2 will probably be there in a second.
Cue-deliberation I dont know how long the supply lay line is,
Anticipation but it appears we are probably going to need more water than one supply
line is going to give us.
Metacognition So I would keep in mind,
Contingency unless we can check the fire fairly rapidly.
Contingency So start thinking of other water sources.
Action-Deliberation Consider laying another supply line to engine 1.
The Critical Decision Method in the case have been placed into a timeline
and coded into the categories indicated in
The Critical Decision Method (CDM) the leftmost column. As in all methods for
involves multi-pass retrospection in which coding protocols, multiple coders are used
the expert is guided in the recall and and there is a reliability check.
elaboration of a previously experienced Given its focus on decision making, the
case. The CDM leverages the fact that strength of the CDM is its use in the cre-
domain experts often retain detailed mem- ation of models of reasoning (e.g., deci-
ories of previously encountered cases, espe- sions, strategies). Detailed presentations of
cially ones that were unusual, challenging, the method along with summaries of stud-
or in one way or another involved critical ies illustrating its successful use can be found
decisions. The CDM does not use generic in Crandall, Klein, and Hoffman (2006)
questions of the kind Tell me everything and Hoffman, Crandall, and Shadbolt
you know about x, or Can you describe (1998).
your typical procedure? Instead, it guides
the expert through multiple waves of re-
telling and prompts through the use of Work Domain Analysis
specific probe questions (e.g., What were
you seeing?) and what-if queries (e.g., Unlike the CDM, which focuses on the
What might someone else have done in reasoning and strategies of the individual
this circumstance?). The CDM generates practitioner, Work Domain Analysis (WDA)
rich case studies that are often useful as builds a representation of an entire work
training materials. It yields time-lined sce- domain. WDA has most frequently been
narios, which describe decisions (decision used to describe the structure of human-
types, observations, actions, options, etc.) machine systems for process control, but it
and aspects of decisions that can be easy is now finding increasing use in the analy-
or difficult. It can also yield a list of deci- sis and design of complex, systems (Burns
sion requirements and perceptual cues the & Hajdukiewicz, 2004; Chow & Vicente,
information the expert needs in order to 2002; Lintern, Miller, & Baker, 2002; Naikar
make decisions. & Sanderson, 2001).
An example of a coded CDM transcript An Abstraction-Decomposition matrix
appears in Table 12.2. In this example, events represents a work domain in terms of
2 10 the cambridge handbook of expertise and expert performance
levels of abstraction, where each level text, down to social collectives (teams),
is a distinctive type of constraint. Fig- down to individual worker or individual
ure 12.1 presents matrices for two systems, component (e.g., software package residing
one designed predominantly around phys- on a particular workstation).
ical laws and the other designed predom- Typically, a work-domain analysis is ini-
inantly around social values. The library tiated from a study of documents, although
example grapples with a pervasive social once an Abstraction-Decomposition matrix
issue, the need for individual identification is reasonably well developed, interviews
balanced against the desire for personal con- with domain experts will help the ana-
fidentiality. These tutorial examples demon- lyst extend and refine it. Vicente (1999)
strate that the Abstraction-Decomposition argues that the Abstraction-Decomposition
format can be used with markedly differ- matrix is an activity-independent represen-
ent work domains. In both of these matrices, tation and should contain only descrip-
entries at each level constitute the means tions of the work domain (the tutorial
to achieve ends at the level above. The examples of Figure 12.1 were developed
intent is to express means-ends relations with that stricture in mind). However,
between the entries of adjacent levels, with Vicentes advice is not followed universally
lower levels showing how higher-level func- within the community that practices WDA;
tions are met, and higher levels showing some analysts include processes in their
why lower-level forms and functions are Abstraction-Decomposition matrices (e.g.,
necessary. Burns & Hajdukiewicz, 2004).
Work domains are also represented in It is possible to add activity to the repre-
terms of a second dimension: levels of sentation yet remain consistent with Vicente
decomposition, from organizational con- (1999) by overlaying a trajectory derived
eliciting and representing the knowledge of experts 2 11
for making Concept Maps can be found in sus formation). Teams can be structured in
Canas et al., 2004; and Crandall, Klein, & a variety of ways and can make and share
Hoffman, 2006.) Figure 12.4 is a Concept Concept Maps over the world-wide web (see
Map that lays out expert knowledge about Canas et al., 2004).
the role of cold fronts in the Gulf Coast The ability to hyperlink digital reso-
(Hoffman, Coffey, & Ford, 2000). urces such as text documents, images,
Although Concept Maps can be made by video clips, and URLs is another signifi-
use of paper and pencil, a white board, or cant advantage provided by computerized
Post-Its, the Concept Maps presented here means of developing Concept Maps (Cmap-
were created by use of CmapTools, a soft- Tools indicate hyperlinks by the small icons
ware suite created at the Institute for Human underneath concept nodes). Hyperlinks can
and Machine Cognition (free download at connect to other Concept Maps; a set
http://ihmc.us). In the KE procedure involv- of Concept Maps hyperlinked together is
ing an individual expert, one researcher regarded as a knowledge model. Figure 12.5
stands at a screen and serves as the facilitator shows a screen shot from the top-level Con-
while another researcher drives the laptop cept Map in the System To Organize Rep-
and creates the Concept Map that is pro- resentations in Meteorology (STORM), in
jected on the screen. The facilitator helps which a large number of Concept Maps
the domain expert build up a representa- are linked together. In Figure 12.5 , some of
tion of their domain knowledge, in effect the resources have been opened for illus-
combining KE with knowledge representa- trative purposes (real-time satellite imagery,
tion. (This is one reason the method is rela- computer weather forecasts, and digital
tively efficient.) Concept Mapping can also video in which the domain expert pro-
be used by teams or groups, for purposes vides brief explanatory statements for some
other than KE (e.g., brainstorming, consen- of the concepts throughout the model).
eliciting and representing the knowledge of experts 2 13
All of the STORM Concept Maps and lists dozens of methods. Although not all
resources can be viewed at http://www. of them are methods that would be use-
ihmc.us/research/projects/STORMLK/ ful as knowledge-elicitation or knowledge-
Knowledge models structured as Con- representation procedures, it is clear that the
cept Maps can serve as living repositories roster of tools and methods available to cog-
of expert knowledge to support knowledge nitive engineers has expanded considerably
sharing as well as knowledge preservation. over the past two decades. We look now to
They can serve also as interfaces for intelli- core ideas and tidbits of guidance that have
gent systems where the model of the experts stood the test of time.
knowledge becomes the interface for a per-
formance support tool or training aid. (Ford
et al., 1996). Where the Rubber Meets the Road
(1). In eliciting expert knowledge one can: (a)
Ask people questions, and (b) Observe per-
Methodological Concepts and Issues formance. Questions can be asked in the
great many forms and formats for inter-
Research and various applied projects con- viewing, including unstructured interviews,
ducted since the seminal works on KE the CDM procedure, and Concept Map-
methodology have left some ideas standing ping, as well as many other techniques (e.g.,
and have led to some new and potentially Endsley & Garland, 2000). Performance can
valuable ideas. One recent review of CTA be observed via ethnographic studies of pat-
methods (Bonacteo & Burns, forthcoming) terns of communication in the workplace,
Figure 12 .4. A Concept Map about cold fronts in Gulf Coast weather.
2 14 the cambridge handbook of expertise and expert performance
Figure 12 .5. A screen shot of a Concept Map with some opened resources.
modest infusion of supporting technologies edge that had not been articulated up to
might have positive results (e.g., redesign that point.
of interfaces, redesign of the workspace (8). Good knowledge elicitation procedures
layout, creation of new functionalities for are effective scaffolds. Although there may
existing software, and ideas about entirely be phenomena to which one could legiti-
new software systems.) Again we can use mately, or at least arguably, append the des-
the project on expert weather forecasting ignation tacit knowledge, there is no indi-
as an example (Hoffman, Coffey, & Ford, cation that such knowledge lies beyond the
2000). That project compared a number reach of science in some unscientific nether-
of alternative knowledge-elicitation meth- world of intuitions or unobservables. Over
ods including protocol analysis, the CDM, and over again, the lesson is not that there
the Knowledge Audit (Militello & Hutton, is knowledge that experts literally cannot
1998; see Ross, Shafer and Klein, this vol- articulate, nor is it the hangover issue of
ume), an analysis of Standard Operating whether verbalization interferes with rea-
Procedures documents, the Recent Case soning. Rather, the issue is whether the KE
Walkthrough method (Militello & Hutton, procedure provides sufficient scaffolding to
1998), a Workspace and Workpatterns anal- support the expert in articulating what they
ysis (Vicente, 1999), and Concept Mapping. know. Support involves the specifics of the
All methods yielded data that spoke to prac- procedure (e.g., probe questions), but it also
titioner knowledge and reasoning and all also involves the fact that knowledge elicitation
identified leverage points. is a collaborative process. There is no sub-
(7). Tacit knowledge is not a salient prob- stitute for the skill of the elicitor (e.g., in
lem. Without getting into the philosophi- framing alternative suggestions and word-
cal weeds of what one means by kinds ings). Likewise, there is no substitute for
of knowledge, another concern has to do the skill of the participating practitioner.
with the possibility that routine knowledge Some experts will have good insight, but
about procedures or task activities might others will not. Though it might be pos-
become tacit, that is, so automatic as to sible for someone to prove the existence
be inexpressible via introspection or ver- of knowledge that cannot be uncovered,
bal report. This hangover issue from the knowledge engineers face the immediate,
heyday of Behaviorism remains to this day practical challenges of designing new and
a non-problem in the practical context of better sociotechnical systems. They accom-
eliciting knowledge from experts. For one plish something when they uncover useful
thing, it has never been demonstrated that knowledge that might have otherwise been
there exists such a thing as knowledge that missed.
cannot be verbalized in principle, and the
burden of proof falls on the shoulders of
New Ideas
those who make the existence claim. Again
sidestepping the philosophical issues (i.e., if Recent research and application efforts have
it cannot be articulated verbally, is it really also yielded some new ideas about the
knowledge?), we maintain that the empiri- knowledge elicitation methods palette.
cal facts mitigate the issue. For instance, in (1). The (hypothetical) problem of differ-
Concept-Mapping interviews with domain ential access has given way to a practi-
experts, experience shows that almost every cal consideration of differential utility. Any
time the expert will reach a point in making a given method might be more useful for
Concept Map where s/he will say something certain purposes, might be more applica-
like, Well, Ive never really thought about that, ble to certain domains, or might be more
or thought about it in this way, but now that useful with experts having certain cogni-
you mention it . . . , and what follows will tive styles. In other words, each knowledge-
be a clear specification on some procedure, elicitation method has its strengths and
strategy, or aspect of subdomain knowl- weaknesses. Some of these are more purely
eliciting and representing the knowledge of experts 2 17
the new weather radar systems and the use tise Studies, which is supported by the
of the many computer models for weather Sciences of Learning Program of the
forecasting (Hoffman, Coffey, & Ford, National Science Foundation, and his par-
2000). A current project on knowledge re- ticipation in The Advanced Decision Archi-
covery that involves reclamation of expert tectures Collaborative Technology Alliance,
knowledge about terrain analysis from exist- which is sponsored by the US Army
ing documents such as the Terrain Analysis Research Laboratory under cooperative
Database (Hoffman, 2003 b) has generated agreement DAAD19-01-2-0009.
over 15 0 Concept Maps containing more
than 3 ,000 propositions.
Although knowledge elicitation on such a
scale is daunting, we now have the technolo-
References
gies and methodologies to facilitate the elic-
itation, preservation, and sharing of expert Adelman, L. (1989). Management issues in
knowledge on a scale never before possible. knowledge elicitation. IEEE Transactions on
Systems, Man, and Cybernetics, 19, 483 488.
This is a profound application of cognitive
science and is one that is of immense value Ausubel, D. P., Novak, J. D., & Hanesian, H.
(1978). Educational psychology: A cognitive view,
to society.
2 nd ed. New York: Holt, Rinehart and Winston.
Banaji, M. R., & Crowder, R. G. (1989). The
Practice, Practice, Practice bankruptcy of everyday memory. American
Psychologist, 44, 1185 1193 .
No matter how much detail is provided Barber, P. (1988). Applied cognitive psychology.
about the conduct of a knowledge-elicitation London: Methuen.
procedure, there is no substitute for prac- Becerra-Fernandez, I., Gonzalez, A., & Sabher-
tice. The elicitor needs to adapt on the fly wal, R. (2004). Knowledge management: chal-
to individual differences in style, personal- lenges, solutions, and technologies. Upper Saddle
ity, agenda, and goals. In breaking the ice River, NJ: Prentice-Hall.
and establishing rapport, the elicitor needs Bonaceto, C., & Burns, K. (forthcoming). Map-
to show good intentions and needs to be ping the mountains: A survey of cognitive engi-
sensitive to possible concerns on the part of neering methods and uses. In R. R. Hoffman
(Ed.), Expertise out of context: Proceedings of the
the expert that the capture of his/her knowl-
Sixth International Conference on Naturalistic
edge might mean the loss of their job (per-
Decision Making. Mahwah, NJ: Erlbaum.
haps to a machine). To be good and effective
Burns, C., & Hajdukiewicz, J. R. (2004). Ecologi-
at knowledge-elicitation, one must attempt
cal interface design. Boca Raton, FL: CRC Press.
to become an expert apprentice expe-
Burns, C. M., Bryant, D., & Chalmers, B.
rienced at, skilled at, and comfortable with
(2001). Scenario mapping with work domain
going into new domains, boostrapping effi- analysis. In Proceedings of the Human Factors
ciently and then designing and conducting and Ergonomics Society 45 th Annual Meeting
a series of knowledge-elicitation procedures (pp. 424428). Santa Monica, CA: Human Fac-
appropriate to project goals. The topic of tors and Ergonomics Society.
how to train people to be expert apprentices Canas, A. J., Hill, G., Carff, R., Suri, N., Lott, J.,
is one that we hope will receive attention Eskridge, T., Gomez, G., Arroyo, M., & Carva-
from researchers in the coming years (see jal, R. (2004). CmapTools: A Knowledge Mod-
Militello & Quill, forthcoming). eling and Sharing Environment. In A. J. Canas,
J. D. Novak, & F. M. Gonzalez (Eds.), Concept
Maps: Theory, Methodology, Technology, Proceed-
Acknowledgments ings of the 1st International Conference on Con-
cept Mapping (pp. 125 13 3 ). Pamplona, Spain:
The senior Authors contribution to this Universidad Publica de Navarra.
chapter was supported through his partici- Chase, W. G., & Simon, H. A. (1973 ). Perception
pation in the National Alliance for Exper- in chess. Cognitive Psychology, 4, 5 5 81.
eliciting and representing the knowledge of experts 2 19
Chi, M. T. H., Feltovich, P. J., & Glaser, R. (1981). Societe de Gens de lettere. Paris: Le Breton.
Categorization and representations of physics Compact Edition published in 1969 by The
problems by experts and novices. Cognitive Sci- Readex Microprint Corporation, New York.
ence, 5 , 12115 2. Translations of selected articles can be found
Chi, M. T. H., Glaser, R., & Farr, M. L. (Eds.) at http://www.hti.umich.edu/d/did/.
(1988). The nature of expertise. Hillsdale, NJ: Duda, J., Gaschnig, J., & Hart, P. (1979). Model
Erlbaum. design in the PROSPECTOR consultant sys-
Chow, R., & Vicente, K. J. (2002). A field study tem for mineral exploration. In D. Michie
of emergency ambulance dispatching: Impli- (Ed.), Expert systems in the micro-electronic age
cations for decision support. In Proceedings of (pp. 15 3 167). Edinburgh: Edinburgh Univer-
the Human Factors and Ergonomics Society 46th sity Press.
Annual Meeting (pp. 3 13 3 17). Santa Monica, Endsdley, M. R., & Garland, D. L. (2000). Situa-
CA: Human Factors and Ergonomics Society. tion awareness analysis and measurement. Hills-
Clancey, W. J. (1993 ). The knowledge level dale, NJ: Erlbaum.
reinterpreted: Modeling socio-technical sys- Ericsson, K. A., & Simon, H. A. (1993 ). Protocol
tems. In K. M. Ford & J. M. Bradshaw (Eds.), analysis: Verbal reports as data, 2nd ed. Cam-
Knowledge acquisition as modeling (pp. 3 3 49, bridge, MA: MIT Press.
Pt. 1). New York: Wiley. Evans, A. W., Jentsch, F., Hitt, J. M., Bow-
Cooke, N. M., & McDonald, J. E. (1986). A for- ers, C., & Salas, E. (2001). Mental model
mal methodology for acquiring and represent- assessments: Is there a convergence among dif-
ing expert knowledge. Proceedings of the IEEE ferent methods? In Proceedings of the Human
74, 1422143 0. Factors and Ergonomics Society 45 th Annual
Cooke, N. M., & McDonald, J. E. (1987). The Meeting (pp. 293 296). Santa Monica, CA:
application of psychological scaling techniques Human Factors and Ergonomics Society.
to knowledge elicitation for knowledge-based Feigenbaum, E. A., Buchanan, B. G., & Leder-
systems. International Journal of Man-Machine berg, J. (1971). On generality and prob-
Studies, 2 6, 5 3 3 5 5 0. lem solving: A case study using the DEN-
Crandall, B., Klein, G., & Hoffman, R. R. (2006). DRAL program. In B. Meltzer & D. Michie
Working Minds: A practitioners guide to cognitive (Eds.), Machine intelligence 6 (pp. 165 190).
task analysis. Cambridge, MA: MIT Press. Edinburgh: Edinburgh University Press.
Cullen, J., & Bryman, A. (1988). The knowledge Ford, K. M., & Adams-Webber, J. R. (1992).
acquisition bottleneck: Time for reassessment? Knowledge acquisition and constructive epis-
Expert Systems, 5 , 216225 . temology. In R. Hoffman (Ed.), The cognition of
David, J.-M., Krivine, J.-P., & Simmons, R. (Eds.) experts: Psychological research and empirical AI
(1993 ). Second-generation expert systems. Berlin: (pp. 12113 6). New York: Springer-Verlag.
Springer Verlag. Ford, K. M., Coffey, J. W., Canas, A., Andrews,
Davenport, T. H., & Prusak, L. (1998). Work- E. J., & Turne, C. W. (1996). Diagnosis and
ing knowledge: How organizations manage what explanation by a nuclear cardiology expert sys-
they know. Cambridge, MA: Harvard Business tem. International Journal of Expert Systems, 9,
School Press. 4995 06.
De Keyser, V., Decortis, F., & Van Daele, Forsyth, D. E., & Buchanan, B. G. (1989). Knowl-
A. (1998). The approach of Francophone edge acquisition for expert systems: Some pit-
ergonomy: Studying new technologies. In V. falls and suggestions. IEEE Transactions on Sys-
De Keyser, T. Qvale, & B. Wilpert (Eds.), tems, Man, and Cybernetics, 19, 3 45 442.
The meaning of work and technological options Gaines, B. R., & Boose, J. H. (Eds.) (1988). Knowl-
(pp. 147163 ). Chichester, England: Wiley. edge acquisition tools for expert systems. London:
Dekker, S. W. A., Nyce, J. M., & Hoffman, R. R. Academic Press.
(MarchApril 2003 ). From contextual inquiry Gagne, R. M., & Smith, E. C. (1962). A study
to designable futures: What do we need to get of the effects of verbalization on problem
there? IEEE Intelligent Systems, pp. 7477. solving. Journal of Experimental Psychology, 63 ,
Diderot, D. (with dAlembert, J.) (Eds.) (175 1 1218.
1772). Encyclopedie ou Dictionnaire raisonne Gentner, D., & Stevens, A. L. (Eds.) (1983 ). Men-
des sciences, des arts et des metiers, par une tal models. Mahwah, NJ: Erlbaum.
220 the cambridge handbook of expertise and expert performance
Hollnagel, E., & Woods, D. D. (1983 ). Cognitive Lesgold, A., Feltovich, P. J., Glaser, R., & Wang,
systems engineering: New wine in new bottles. M. (1981). The acquisition of perceptual diag-
International Journal of Man-Machine Studies, nostic skill in radiology. Technical Report No.
18, 5 83 600. PDS-1, Learing Research and Development
Johnston, N. (2003 ). The paradox of rules: Proce- Center, University of Pittsburgh, Pittsburgh,
dural drift in commercial aviation. In R. Jensen PA.
(Ed.), Proceedings of the Twelfth International Lintern, G. (2003 ). Tyranny in rules, autonomy
Symposium on Aviation Psychology (pp. 63 0 in maps: Closing the safety management loop.
63 5 ). Dayton, OH: Wright State University. In R. Jensen (Ed.), Proceedings of the Twelfth
Hutchins, E. (1995 ). Cognition in the wild. Cam- International Symposium on Aviation Psychology
bridge, MA: MIT Press. (pp. 719724). Dayton, OH: Wright State Uni-
versity.
Klahr, D. (Ed.) (1976). Cognition and instruction.
Hillsdale, NJ: Erlbaum. Lintern, G., Diedrich, F. J. & Serfaty, D. (2002).
Engineering the community of practice for
Klahr, D., & Kotovsky, K. (Eds.) (1989). Complex
maintenance of organizational knowledge. Pro-
information processing: The impact of Herbert A.
ceedings of the IEEE 7th Conference on Human
Simon. Mahwah, NJ: Erlbaum.
Factors and Power Plants (pp. 6.76.13 ). New
Klein, G. (1992). Using knowledge elicitation to York: IEEE.
preserve corporate memory. In R. R. Hoff-
Lintern, G., Miller, D., & Baker, K. (2002). Work
man (Ed.), The psychology of expertise: Cogni-
centered design of a USAF mission planning
tive research and empirical AI (pp. 170190).
system. In Proceedings of the 46th Human Fac-
Mahwah, NJ: Erlbaum.
tors and Ergonomics Society Annual Meeting
Klein, G. A., Calderwood, R., & MacGregor, D. (pp. 5 3 15 3 5 ). Santa Monica, CA: Human Fac-
(1989). Critical decision method for eliciting tors and Ergonomics Society.
knowledge. IEEE Transactions on Systems, Man,
McDonald, N., Corrigan, S., & Ward, M. (2002).
and Cybernetics, 19, 462472.
Cultural and organizational factors in system
Klein, G., Orasanu, J., Calderwood, R., & Zsam- safety: Good people in bad systems. Proceedings
bok, C. E. (Eds.) (1993 ). Decision making in of the 2 002 International Conference on Human-
action: Models and methods. Norwood, NJ: Computer Interaction in Aeronautics (HCI-Aero
Ablex Publishing Corporation. 2 002 ) (pp. 205 209). Menlo Park, CA: Ameri-
Klein, G., & Weitzenfeld, J. (1982). The use of can Association for Artificial Intelligence Press.
analogues in comparability analysis. Applied McGraw, K. L., & Harbison-Briggs, K. (1989).
Ergonomics, 13 , 99104. Knowledge acquisition: Principles and guidelines.
Knorr-Cetina, K. D. (1981). The manufacture of Englewood Cliffs, NJ: Prentice Hall.
knowledge. Oxford: Pergamon. Means, B., & Gott, S. P. (1988). Cognitive task
Knorr-Cetina, K. D., & Mulkay, M. (1983 ). Sci- analysis as a basis for tutor development:
ence observed. Berkeley Hills, CA: Sage. Articulating abstract knowledge representa-
Koopman, P., & Hoffman, R. R. (2003 / tions. In J. Psotka, L. D. Massey, & S. A.
NovemberDecember). Work-arounds, make- Mutter (Eds.), Intelligent tutoring systems:
work, and kludges. IEEE: Intelligent Systems, Lessons learned (pp. 3 5 5 7). Hillsdale, NJ:
pp. 7075 . Erlbaum.
LaFrance, M. (1992). Excavation, capture, col- Mieg, H. (2000). The social psychology of expertise.
lection, and creation: Computer scientists Mahwah, NJ: Erlbaum.
metaphors for eliciting human expertise. Militello, L. G., & Hutton, R. J. B. (1998). Applied
Metaphor and Symbolic Activity, 7 , 13 5 15 6. Cognitive Task Analysis (ACTA): A practi-
Lave, J. (1988). Cognition in practice: Mind, math- tioners toolkit for understanding cognitive task
ematics, and culture in everyday life. Cambridge: demands. Ergonomics, 41, 16181641.
Cambridge University Press. Militello, L., & Hoffman, R. R. (forthcoming).
Lesgold, A. M. (1984). Acquiring expertise. In Perspectives on cognitive task analysis. Mahwah
J. R. Anderson & S. M. Kosslyn (Eds.), Tuto- NJ: Erlbaum.
rials in learning and memory: Essays in honor of Militello, L., & Quill, L. (2006). Expert appren-
Gordon Bower (pp. 3 160). San Francisco, CA: tice strategies. In R. R. Hoffman (Ed.), Expertise
W. H. Freeman. out of context. Mahwah, NJ: Erlbaum.
222 the cambridge handbook of expertise and expert performance
Naikar, N., & Sanderson, P. M. (2001). Evaluat- Rook, F. W., & Croghan, J. W. (1989). The
ing system design proposals with work domain knowledge acquisition activity matrix: A sys-
analysis. Human Factors, 43 , 5 295 42. tems engineering conceptual framework. IEEE
Naikar, N., & Saunders, A. (2003 ). Crossing Transactions on Systems, Man, and Cybernetics,
the boundaries of safe operation: A technical 19, 5 865 97.
training approach to error management. Cog- Schmidt, L., & Luczak, H. (2000). Knowledge
nition Technology and Work, 5 , 171180. representation for engineering design based on
Neale, I. M. (1988). First generation expert a cognitive model. In Proceedings of the IEA
systems: A review of knowledge acquisition 2 000/HFES 2 000 Congress (vol. 1) (pp. 623
methodologies. Knowledge Engineering Review, 626). Santa Monica, CA: Human Factors and
3 , 105 146. Ergonomics Society.
Novak, J. D. (1998). Learning, creating, and Scribner, S. (1984). Studying working intelli-
using knowledge: Concept maps as facilitative gence. In B. Rogoff & S. Lave (Eds.), Every-
tools in schools and corporations. Mahweh, NJ: day cognition: Its development in social context
Lawrence Erlbaum Associates. (pp. 940). Cambridge: Harvard University
Press.
Orr, J. (1996). Talking about machines: An ethnog-
raphy of a modern job. Ithaca, NY: Cornell Uni- Shadbolt, N. R., & Burton, A. M. (1990). Knowl-
versity Press. edge elicitation techniques: Some experimen-
tal results. In K. L. McGraw & C. R. West-
Potter, S. S., Roth, E. M., Woods, D. D., and
phal (Eds.), Readings in knowledge acquisition
Elm, W. C. (2000). Bootstrapping multiple
(pp. 213 3 ). New York: Ellis Horwood.
converging cognitive task analysis techniques
for system design. In J. M. Schraagen and S. F. Shanteau, J. (1992). Competence in experts:
Chipman (Eds.), Cognitive task analysis The role of task characteristics. Organizational
(pp. 3 173 40). Mahwah, NJ: Erlbaum. Behavior and Human Decision Processes, 5 3 ,
25 2266.
Prerau, D. (1989). Developing and managing expert
systems: Proven techniques for business and Shortliffe, E. H. (1976). Computer-based medical
industry. Reading, MA: Addison Wesley. consultations: MYCIN. New York: Elsevier.
Raeth, P. G. (Ed.) (1990). Expert systems. New Simon, H. A., & Hayes, J. R. (1976). Under-
York: IEEE Press. standing complex task instructions. In D. Klahr
(Ed.), Cognition and instruction (pp. 5 180).
Rasmussen, J. (1985 ). The role of hierarchical
Hillsdale, NJ: Erlbaum.
knowledge representation in decision-making
and system-management. IEEE Transactions on Spradley, J. P. (1979). The ethnographic interview.
Systems, Man, and Cybernetics, SMC-15 , 23 4 New York: Holt, Rinehart and Winston.
243 . Stein, E. W. (1997). A look at expertise from a
Rasmussen, J. (1986). Information processing social perspective. In P. J. Feltovich, K. M. Ford,
and human-machine interaction: An approach & R. R. Hoffman (Eds.), Expertise in context
to cognitive engineering. Amsterdam: North- (pp. 181194). Cambridge: MIT Press.
Holland. Sternberg, R. J., & Frensch, P. A. (Eds.) (1991).
Rasmussen, J. (1992). Use of field studies for Complex problem solving: Principles and mecha-
design of workstations for integrated manufac- nisms. Mahwah, NJ: Erlbaum.
turing systems. In M. Helander & N. Naga- Suchman, L. (1987). Plans and situated actions:
machi (Eds.), Design for manufacturability: A The problem of human-machine communication.
systems approach to concurrent engineering and Cambridge: Cambridge University Press.
ergonomics (pp. 3 173 3 8). London: Taylor and Vicente, K. J. (1999). Cognitive work analysis:
Francis. Towards safe, productive, and healthy computer-
Rasmussen, J., Petjersen, A. M., & Goodstein, based work. Mahwah, NJ: Erlbaum.
L. P. (1994). Cognitive systems engineering. New Vicente, K. J., Christoffersen, K., & Pereklita,
York: John Wiley. A. (1995 ). Supporting operator problem solv-
Rasmussen, J., & Rouse, W. B. (Eds.) (1981). ing through ecological interface design. IEEE
Human detection and diagnosis of system failures. Transactions on Systems, Man, and Cybernetics,
New York: Plenum Press. SMC-2 5 , 5 295 45 .
C H A P T E R 13
K. Anders Ericsson
The superior skills of experts, such as accom- capable of explaining the nature and struc-
plished musicians and chess masters, can be ture of their exceptional performance. Some
amazing to most spectators. For example, pioneering scientists, such as Binet (1893 /
club-level chess players are often puzzled by 1966), questioned the validity of the experts
the chess moves of grandmasters and world descriptions when they found that some
champions. Similarly, many recreational ath- experts gave reports inconsistent with those
letes find it inconceivable that most other of other experts. To make matters worse,
adults regardless of the amount or type of in those rare cases that allowed verifica-
training have the potential ever to reach tion of the strategy by observing the perfor-
the performance levels of international com- mance, discrepancies were found between
petitors. Especially puzzling to philosophers the reported strategies and the observations
and scientists has been the question of the (Watson, 1913 ). Some of these discrepancies
extent to which expertise requires innate were explained, in part, by the hypothe-
gifts versus specialized acquired skills and sis that some processes were not normally
abilities. mediated by awareness/attention and that
One of the most widely used and simplest the mere act of engaging in self-observation
methods of gathering data on exceptional (introspection) during performance changed
performance is to interview the experts the content of ongoing thought processes.
themselves. But are experts always capable These problems led most psychologists in
of describing their thoughts, their behaviors, first half of the 20th century to reject all
and their strategies in a manner that would types of introspective verbal reports as valid
allow less-skilled individuals to understand scientific evidence, and they focused almost
how the experts do what they do, and per- exclusively on observable behavior (Boring,
haps also understand how they might reach 195 0).
expert level through appropriate training? In response to the problems with the
To date, there has been considerable contro- careful introspective analysis of images and
versy over the extent to which experts are perceptions, investigators such as John B.
223
224 the cambridge handbook of expertise and expert performance
Watson (1920) and Karl Duncker (1945 ) been applied to a number of domains of
introduced a new type of method to elicit expertise, such as memory experts, chess
verbal reports. The subjects were asked masters, and medical specialists. The chap-
to think aloud and give immediate ver- ter is concluded with a broad overview of
bal expression to their thoughts while they the issues of applying protocol analysis to
were engaged in problem solving. In the the study of expert performance.
main body of this chapter I will review evi-
dence that this type of verbal expression
of thoughts has not been shown to change Historical Development of Verbal
the underlying structure of the thought pro- Reports on Thought Processes
cesses and thus avoids the problem of reac-
tivity, namely, where the act of generat- Introspection or looking inside to uncover
ing the reports may change the cognitive the structure of thinking and its mental
processes that mediate the observed per- images has a very long history in philos-
formance. In particular, I will describe the ophy. Drawing on the review by Ericsson
methods of protocol analysis where verbal and Crutcher (1991), we see that Aristotle
reports are elicited, recorded, and encoded is generally given credit for the first system-
to yield valid data on the underlying thought atic attempt to record and analyze the struc-
processes (Ericsson & Simon 1980, 1984, ture of sequences of thoughts. He recounted
1993 ). an example of series of thoughts mediat-
Although protocol analysis is generally ing the recall of a specific piece of infor-
accepted as providing valid verbalizations mation from memory. Aristotle argued that
of thought processes (Simon & Kaplan, thinking can be described as a sequence of
1989), these verbal descriptions of thought thoughts, where the brief transition periods
sequences frequently do not contain suffi- between consecutive thoughts do not con-
cient detail about the mediating cognitive tain any reportable information, and this has
processes and the associated knowledge to never been seriously challenged. However,
satisfy many scientists. For example, these such a simple description of thinking was
reports may not contain the detailed proce- not sufficiently detailed to answer the ques-
dures that would allow cognitive scientists tions about the nature of thought raised by
to build complete computer models that are philosophers in the 17th, 18th, and 19th cen-
capable of regenerating the observed perfor- turies (Ericsson & Crutcher, 1991).
mance on the studied tasks. Hence, inves- Most of the introspective analysis of
tigators have continued to search for alter- philosophers had been based on self-analysis
native types of verbal reports that generate of the individual philosophers own thought.
more detailed descriptions. Frequently sci- In the 19th century Sir Francis Galton along
entists require participants to explain their with others introduced several important
methods for solving tasks and to give detailed innovations that set the groundwork for
descriptions of various aspects. These alter- empirical studies of thinking. For example,
native reporting methods elicit additional Galton (1879, see Crovitz, 1970) noticed
and more detailed information than is spon- repeatedly that when he took the same walk
taneously verbalized during think aloud. through a part of London and looked at
The desire for increased amounts of reported a given building on his path, this event
information is central to the study of exper- triggered frequently the same or similar
tise, so I will briefly discuss whether it is thoughts in memory. Galton recreated this
possible to increase the amount reported phenomenon by listing the names of the
without inducing reactivity and change of major buildings and sights from his walk on
performance. The main sections of this chap- cards and then presented a card at a time
ter describe the methods for eliciting and to observe the thoughts that were triggered.
analyzing concurrent and retrospective ver- From this self-experiment Galton argued
bal reports and how these methods have that thoughts reoccur with considerable
protocol analysis and expert thought 225
frequency when the same stimulus is and illustrates the problems and limits of
encountered. introspection.
Galton (1883 ) is particularly famous for In a similar manner Bryan and Harter
the innovation of interviewing many peo- (1899) interviewed two students of telegra-
ple by sending out a list of questions about phy as they improved their skill and found
mental imagery said to be the first ques- evidence for an extended plateau for both
tionnaire. He had been intrigued by reports as they reached a rate of around 12 words
of photographic memory and asked ques- per minute. Both reported that this arrest in
tions of the acuity of specific memories, such development was associated with attempts
as the clarity and brightness of their mem- to move away from encoding the Morse
ory for specific things such as their break- code into words and to encode the code into
fast table. He found striking individual dif- phrases. Subsequent research (Keller, 195 8)
ferences in the clarity or vividness, but no has found that this plateau is not a necessary
clear superiority of the eminent scientists; step toward expert levels of performance
for example, Darwin reported having weak and referred to it as the phantom plateau.
visual images. Now a hundred years later it In parallel with the interviews and the
is still unclear what these large individual informal collection of self-observations
differences in reported vividness of memory of expertise in everyday life, laboratory
images really reflect. They seem almost com- scientists attempted to refine introspective
pletely unrelated to the accuracy of memory methods to examine the structure of think-
images and there is no reproducible evidence ing. In the beginning of the 20th century,
for individuals with photographic or eide- psychologists at the University of Wurzburg
tic memory (McKelvie, 1995 ; Richardson, presented highly trained introspective
1988). observers, with standardized questions and
In one of the first published studies on asked them to respond as fast as possible.
memory and expertise Binet (1893 /1966) After reporting their answers, the observers
reported a pioneering interview of chess recalled as much as possible about the
players and their ability to play blind- thoughts that they had while answering the
folded without seeing a chess board. Based questions. They tried to identify the most
on anecdotes and his interviews Binet con- basic elements of their thoughts and images
cluded that the ability required to main- to give as detailed reports as possible. Most
tain chess position in memory during blind- reported thoughts consisted of visual and
fold play did not appear to reflect a basic auditory images, but some participants
memory capacity to store complex visual claimed to have experienced thoughts with-
images, but a deeper understanding of the out any corresponding imagery imageless
structure of chess. More troubling, Binet thoughts. The principle investigator, Karl
found that the verbal descriptions on the Buhler (1907), argued that the existence of
visual images of the mental chess positions imageless thoughts had far-reaching theoret-
differed markedly among blindfold chess ical implications and was inconsistent with
players. Some claimed to see the board the basic assumption of Wilhelm Wundt
as clearly as if it were shown perceptu- (1897) that all thoughts were associated
ally with all the details and even shad- with particular neural activity in some part
ows. Other chess players reported seeing of the brain. Buhlers (1907) paper led to
no visual images during blindfold play and a heated exchange between Buhlers intro-
claimed to rely on abstract characteristics spective observers, who claimed to have
of the chess position. Unfortunately, there observed them, and Wundt (1907), who
was no independent evidence to support or argued that these reports were artifacts of
question the validity of these diverse intro- inappropriate reporting methods and the
spective reports. Binets (1893 /1966) classic theoretical bias of the observers. A devastat-
report is a pioneering analysis of blindfold ing methodological conclusion arose from
chess players opinions and self-observations this controversy: the existence of imageless
226 the cambridge handbook of expertise and expert performance
thoughts could not be resolved empirically ized that many types of complex cogni-
by the introspective method. This finding tive processes, such as problem solving, cor-
raised fundamental doubts about analytic responded to ongoing processes that were
introspection as a scientific method. inherently complex and were mediated by
The resulting reaction to the crisis was reportable thoughts. In fact, Watson (1920)
to avoid the problem of having to trust was the first investigator to publish a study
the participants verbal reports about inter- where a participant was asked to think
nal events. Instead of asking individuals to aloud while solving a problem. Accord-
describe the structure of their thoughts, par- ing to Watson, thinking was accompanied
ticipants were given objective tests of their by covert neural activity of the speech
memory and other abilities. More gener- apparatus that is frequently referred to as
ally, experimental psychologists developed inner speech. Hence, thinking aloud did
standardized tests with stimuli and instruc- not require observations by any hypothetical
tions where the same pattern of performance introspective capacity, and thinking aloud
could be replicated under controlled con- merely gives overt expression to these sub-
ditions. Furthermore, the focus of research vocal verbalizations. Many other investiga-
moved away from complex mental pro- tors proposed similar types of instructions
cesses, such as experts thinking, and toward to give concurrent verbal expression of ones
processes that were assumed to be unaf- thoughts (see Ericsson & Simon, 1993 , for a
fected by prior experience and knowledge. more extended historical review).
For example, participants were given well- The emergence of computers in the 195 0s
defined simple tasks, such as memorization and 1960s and the design of computer pro-
of lists of nonsense syllables, e.g., XOK, grams that could perform challenging cog-
ZUT, where it is easy to measure objective nitive tasks brought renewed interest in
performance. In addition, experimenters human cognition and higher-level cogni-
assumed that nonsense syllables were com- tive processes. Investigators started study-
mitted to memory without any reportable ing how people solve problems and make
mediating thoughts, and the interest in col- decisions and attempted to describe and
lecting verbal reports from participants vir- infer the thought processes that mediate
tually disappeared until the cognitive revo- performance. They proposed cognitive the-
lution in the late 195 0s. ories where strategies, concepts, and rules
In one of the pioneering attempts to were central to human learning and problem
apply this approach to the study of exper- solving (Miller, Galanter, & Pribram, 1960).
tise, Djakow, Petrowski, and Rudik (1927) Information-processing theories (Newell &
tested the basic abilities of world-class chess Simon, 1972) sought computational models
players and compared their abilities to other that could regenerate human performance
adults. Contrary to the assumed importance on well-defined tasks by the application of
of superior basic cognitive ability and mem- explicit procedures. Much of the evidence
ory, the international players were only supe- for these complex mechanisms was derived
rior on a single test a test involving memory from the researchers own self-observation,
for stimuli from their own domain of exper- informal interviews, and systematic ques-
tise, namely, chess positions. A few decades tioning of participants.
later de Groot (1946/1978d) replicated chess Some investigators raised concerns almost
players superior memory for chess positions immediately about the validity of these
and found that correct recall was closely data. For example, Robert Gagne and his
related to the level of chess skill of the player. colleagues (Gagne & Smith, 1962) demon-
Many investigators, including the famous strated that requiring participants to ver-
behaviorist and critic of analytic introspec- balize reasons for each move in the Tower
tion, John Watson, are very critical of the of Hanoi improved performance by reduc-
accuracy of verbal descriptions of skilled ing the number of moves in the solutions
activities, such as where one looks dur- and improving transfer to more difficult
ing a golf swing (Watson, 1913 ). He real- problems as compared to a silent control
protocol analysis and expert thought 227
condition. Although improvements are wel- to verbalize their thoughts in a manner that
come to educators, the requirement to does not alter the sequence and content
explain must have changed the sequences of thoughts mediating the completion of a
of thoughts from those normally gener- task and therefore should reflect immedi-
ated. Other investigators criticized the valid- ately available information during thinking.
ity and accuracy of the retrospective ver-
bal reports. For instance, Verplanck (1962) Elicitation of Non-Reactive Verbal
argued that participants reported that they Reports of Thinking
relied on rules that were inconsistent with
their observed selection behavior. Nisbett Based on their theoretical analysis, Ericsson
and Wilson (1977) reported several exam- and Simon (1993 ) argued that the clos-
ples of experiments in social psychology, est connection between actual thoughts and
where participants gave explanations that verbal reports is found when people verbal-
were inconsistent with their observed behav- ize thoughts that are spontaneously attended
ior. These findings initially led many inves- during task completion. In Figure 13 .1 we
tigators to conclude that all types of verbal illustrate how most thoughts are given a ver-
reports were tainted by similar methodolog- bal expression.
ical problems that had plagued introspec- When people are asked to think aloud
tion and led to its demise. Herb Simon (see Ericsson and Simon, 1993 , for complete
and I showed in a review (Ericsson and instructions), some of their verbalizations
Simon, 1980) that the methods and instruc- seem to correspond to merely vocalizing
tions used to elicit the verbal reports had inner speech, which would otherwise have
a great influence on both the reactivity of remained inaudible. Nonverbal thoughts can
the verbal reporting and on the accuracy of also be often given verbal expression by brief
the reported information. We developed a labels and referents. Laboratory tasks stud-
parfticular type of methodology to instruct ied by early cognitive scientists focused on
participants to elicit consistently valid non- how individuals applied knowledge and pro-
reactive reports of their thoughts that I will cedures to novel problems, such as men-
describe in the next section. tal multiplication. When, for example, one
participant was asked to think aloud while
mentally multiplying 3 6 by 24 on two test
Protocol Analysis: A Methodology occasions one week apart, the following pro-
for Eliciting Valid Data on Thinking tocols were recorded:
Theories of human cognition (Anderson, task analysis may not perfectly correspond
1983 ; Newell & Simon, 1972; Newell, 1990) to the verbalizations. Inconsistencies may
proposed computational models that could result from instances where, because of
reproduce the observable aspects of human acquired skill, the original steps are either
performance on well-defined tasks through not generated or not attended as distinct
the application of explicit procedures. One steps. However, there is persuasive evidence
of the principle methods applied by these for the validity of the thoughts that are ver-
scientists is an analysis of the cognitive task balized, that is, that the verbalizations can
(see Chapter 11 by Schraagen for a discussion reveal sequences of thoughts that match
of the methods referred to as cognitive task those specified by the task analysis (Ericsson
analysis), and it serves a related purpose in & Simon, 1993 ). Even if a highly skilled par-
the analysis of verbal protocols. Task analy- ticipants think-aloud report in the multipli-
sis specifies the range of alternative proce- cation task only consisted of 144 and 720,
dures that people could reasonably use, in the reported information would still be suf-
the light of their prior knowledge of facts and ficient to reject many alternative strategies
procedures, to generate correct answers to a and skilled adaptations of them because
task. Moreover, task analysis can be applied these strategies do not involve the generation
to the analysis of think-aloud protocols; for of both of the reported intermediate prod-
example, during a relatively skilled activity, ucts. The most compelling evidence for the
namely, mental multiplication, most adults validity of the verbal reports comes from the
have only limited mathematical knowledge. use of task analysis to predict a priori a set
They know the multiplication tables and of alternative sequences of concurrently ver-
only the standard pencil and paper proce- balized thoughts that is associated with the
dure taught in school for solving multiplica- generation of the correct answer to the pre-
tion problems. Accordingly, one can predict sented problem.
that they will solve a specific problem such Furthermore, verbal reports are only one
as 3 6 24 by first calculating 4 3 6 = 144, indicator of the thought processes that occur
then adding 20 3 6 = 720. More sophisti- during problem solving. Other indicators
cated adults may recognize that 24 3 6 can include reaction times (RTs), error rates, pat-
be transformed into (3 0+6)(3 06) and that terns of brain activation, and sequences of
the formula (a+b)(ab) = a2 b2 can be eye fixations. Given that each kind of empir-
used to calculate 3 6 24 as 3 0 2 6 2 = 900 ical indicator can be separately recorded
3 6 = 864. and analyzed, it is possible to examine the
When adults perform tasks while think- convergent validity established by indepen-
ing aloud the verbalized information must dent analyses of different types of data.
reflect information generated from the cog- In their review, Ericsson and Simon (1993 )
nitive processes normally executed during found that longer RTs were associated with
the task. By analyzing this information, the a longer sequence of intermediate reported
verbalized sequences of thoughts can be thoughts. In addition, analyses show a close
compared to the sequence of intermediate correspondence between participants ver-
results required to compute the answer by balized thoughts and the information that
different strategies that are specified in a they looked at in their environment (see
task analysis (Ericsson & Simon, 1993 ). The Ericsson & Simon, 1993 , for a review).
sequence of thoughts verbalized while mul- Finally, the validity of verbally reported
tiplying 24 3 6 mentally (reproduced in the thought sequences depends on the time
protocol examples above) agrees with the interval between the occurrence of a thought
sequence of intermediate thoughts specified and its verbal report, where the highest
by one, and only one, of the possible strate- validity is observed for concurrent, think-
gies for calculating the answer. aloud verbalizations. For tasks with relatively
However, the hypothesized sequence of short response latencies (less than 5 to 10 sec-
intermediate products predicted from the onds), people are typically able to recall their
2 30 the cambridge handbook of expertise and expert performance
outcomes rather than test their knowledge tions that he had analyzed for a long time and
for medicine and memory of encountered established an informal task analysis. Based
patients. It is, however, difficult to com- on this analysis he could evaluate the relative
pare different individuals levels of naturally merits of different moves and encode the
occurring performance in a domain because thoughts verbalized by chess players while
different individuals tasks will differ in dif- they were selecting the best move for these
ficulty and many other aspects. For exam- positions.
ple, for medical doctors who primarily treat The verbal protocols of both world-
patients with severe and complex problems class and skilled club-level players showed
but with a relatively low frequency of full that both types of players first familiarized
recovery, is their performance better than themselves with the position and verbally
the performance of doctors who primarily reported salient and distinctive aspects of
treat patients with milder forms of the same the position along with potential lines of
disease with uniform recovery? Unless all attack or defense. The players then explored
doctors encounter patients with nearly iden- the consequences of longer move exchanges
tical conditions, it will be nearly impossible by planning alternatives and evaluating the
to compare the quality of their performance. resulting positions. During these searches
The problem of comparing performers per- the players would identify moves with the
formance for comparable tasks is a general best prospects in order to select the single
challenge for measuring and capturing supe- best move.
rior performance in most domains. De Groots (1946/1978) analysis of the
For example, chess players rarely, if ever, protocols identified two important differ-
encounter the same chess positions during ences in cognitive processes that explained
the middle part of chess games (Ericsson the ability of world-class players to select
& Smith, 1991). Hence, there are no nat- superior moves compared to club play-
urally occurring cases where many chess ers. De Groot noticed that the less-skilled
players select moves for the identical com- players didnt even verbally report think-
plex chess position such that the quality ing about the best move during their move
of their moves can be directly compared. selection, implying that they did not, in fact,
In a path-braking research effort, de Groot think about it. Thus, their initial inferior rep-
(1946/1978) addressed this problem by iden- resentation of the position must not have
tifying challenging situations (chess posi- revealed the value of lines of play starting
tions) in representative games that required with that move. In contrast, the world-class
immediate action, namely, the selection of players reported many strong first moves
the next move. De Groot then presented even during their initial familiarization with
the same game situations to chess players the chess position. For example, they would
of different skill levels and instructed them notice weaknesses in the opponents defense
to think aloud while they selected the next that suggested various lines of attack and
chess move. Subsequent research has shown then examine and systematically compare
that this method of presenting representa- the consequences of various sequences of
tive situations and requiring generation of moves. During this second detailed phase
appropriate actions provides the best avail- of analysis, these world-class players would
able measure of chess skill that predicts per- often discover new moves that were superior
formance in chess tournaments (Ericsson, to all the previously generated ones.
Patel, & Kintsch, 2000; van der Maas &
Wagenmakers, 2005 ).
mechanisms mediating chess expertise
De Groots analysis revealed two different
the pioneering studies of chess expertise mechanisms that mediate the world-class
In his pioneering research on chess expertise, players superiority in finding and selecting
de Groot (1946/1978) picked out chess posi- moves. The first difference concerns the best
protocol analysis and expert thought 2 33
players ability to rapidly perceive the rele- 2003 ). Furthermore, highly skilled players
vant structure of the presented chess posi- have been shown to be superior in mentally
tion, thus allowing them to identify weak- planning out consequences of sequences of
nesses and associated lines of attack that the chess moves in experimental studies. In fact,
less-accomplished players never reported chess masters, unlike less-skilled players, are
noticing in their verbal protocols. These pro- able to play blindfold, without a visible
cesses involve rapid perception and encod- board showing the current position, at a rel-
ing, and thus only the end products of these atively high level (Chabris & Hearst, 2003 ;
encoding processes are verbalized. There has Karpov, 1995 ; Koltanowski, 1985 ). Experi-
been a great deal of research attempting to ments show that chess masters are able to
study the perceptual encoding processes by mentally generate the chess positions associ-
recording and analyzing eye fixations dur- ated with multiple chess games without any
ing brief exposures to reveal the cognitive external memory support when the experi-
processes mediating perception and memory menter reads sequences of moves from mul-
of chess positions (see Gobet & Charness, tiple chess games (Saariluoma, 1991, 1995 ).
Chapter 3 0). However, most of this research In sum, the analyses of the protocols along
has not studied the task of selecting the best with experiments show that expert chess
move but has used alternative task instruc- players ability to generate better moves
tions, namely, to recall as many chess pieces cannot be completely explained by their
as possible from briefly presented positions, more extensive knowledge of chess pat-
or to find specific chess pieces in pre- terns. Recognition of patterns and retrieval
sented postions. These changes in the tasks of appropriate moves that they have stored
appear to alter the mediating cognitive pro- in memory during past experiences of chess
cesses, and the results cannot therefore be playing is not sufficient to explain the
directly integrated into accounts of the rep- observed reasoning abilities of highly skilled
resentative expert performance (Ericsson & players. As their skill increases, they become
Kintsch, 2000; Ericsson & Lehmann, 1996; increasingly able to encode and manipulate
Ericsson et al., 2000). internal representations of chess positions to
The second mechanism that underlies the plan the consequences of chess moves, dis-
superior performance of highly skilled play- cover potential threats, and even develop
ers concerns a superior ability to generate new lines of attack (Ericsson & Kintsch,
potential moves by planning. De Groots 1995 ; Saariluoma, 1992). (For a discussion
protocols showed that during this planning of the relation between the superior mem-
and evaluation process, the masters often ory for presented chess positions and the
discovered new moves that were better than memory demands integral to selecting chess
those perceived initially during the familiar- moves, see Ericsson et al., 2000, and Gobet
ization phase. In a subsequent study Char- & Charness, Chapter 3 0.)
ness (1981) collected think-aloud protocols
on the planning process during the selection
of a move for a chess position. Examples of medicine and other domains
an analysis of the protocols from a club-level The expert-performance approach has been
and an expert-level chess player are given applied to a wide range of domains, where
in Figure 13 .2. Consistent with these exam- skilled and less-skilled performers solve rep-
ples, Charness (1981) found that the depth of resentative problems while thinking aloud.
planning increased with greater chess skill. In When the review is restricted to studies
addition, there is evidence that an increase in in domains that show reproducibly supe-
the time available for planning increases the rior performance of experts, the think-aloud
quality of the moves selected, where move protocols reveal patterns of reports that are
selection during regular chess is superior to consistent with those observed in chess.
that of speed chess with its limited time for For example, when expert snooker play-
making the next move (Chabris & Hearst, ers are instructed to make a shot for a
2 34 the cambridge handbook of expertise and expert performance
Figure 13.2 . A chess position presented to chess players with the instruction to select the best next
move by white (top panel). The think-aloud protocols of a good club player (chess rating = 165 7) and
a chess expert (chess rating = 2004) collected by Charness (1981) are shown in the bottom panel to
illustrate differences in evaluation and planning for one specific move, P-c5 (white pawn is moved
from c4 to c5 ), which is the best move for this position. Reported considerations for other potential
moves have been omitted. The chess expert considers more alternative move sequences and some of
them to a greater depth than the club player does. (From Ericsson, K. A., & Charness, N., 1994,
Expert performance: Its structure and acquisition. American Psychologist, 49(8), 725 747, Figure 13 .2
copyright American Psychological Association).
given configuration of pool balls, they ver- 2004) reveal a more complete and superior
balize deeper plans and more far-reaching representation of the current game situa-
exploration of consequences of their shots tion that allow them to prepare for future
than less-skilled players (Abernethy, Neal, & immediate actions better than less-skilled
Koning, 1994). Similarly, athletes at expert players in the same domains. In domains
levels given protocols from dynamic situa- involving perceptual diagnosis, such as in the
tions in baseball (French, Nevett, Spurgeon, interpretation of Electrocardiograms (ECG)
Graham, Rink, & McPherson, 1996) and (Simpson & Gilhooly, 1997) and micro-
soccer (Ward, Hodges, Williams, & Starkes, scopic pathology (Crowley, Naus, Stewart,
protocol analysis and expert thought 2 35
& Friedman, 2003 ), verbal protocols reveal exceptional memory (Wilding & Valentine,
that the experts are able to encode essential Chapter 3 1), mathematical expertise (But-
information more accurately and are more terworth, Chapter 3 2), and historical exper-
able to integrate the information into an tise (Voss & Wiley, Chapter 3 3 ).
accurate diagnosis. The evidence reviewed in this section has
Most of the research on medical diag- been based primarily on findings that are
nosis has tried to minimize the influence based on averages across groups of experts.
of perceptual factors and has relied primar- In the next section we will search for evi-
ily on verbal descriptions of scenarios and dence on the validity of reported thoughts of
patients. This research on medical exper- individual experts as well as individual dif-
tise has shown that the process of generat- ferences between different experts.
ing a diagnosis becomes more efficient as
medical students complete more of their
Individual Differences and Validity
medical training. The increase in efficiency
of Verbal Reports from Expert Performance
is mediated by higher levels of represen-
tation that is acquired to support clinical It is well established that to be successful
reasoning (Boshuizen & Schmidt, 1992; in competitions at the international level,
Schmidt & Boshuizen, 1993 ). When stud- experts need to have engaged in at least
ies present very challenging medical prob- ten years of intensive training a finding
lems to specialists and medical students, that applies even to the most talented
the experts give more accurate diagnoses individuals (Ericsson Krampe, & Tesch-
(Ericsson, 2004; Norman, Trott, Brooks, & Romer, 1993 ; Simon & Chase, 1973 ). Conse-
Smith, 1994). The specialists are also more quently, researchers have not been surprised
able to give complete and logically sup- that verbal reports of experts and, thus,
ported diagnoses (Patel & Groen, 1991) that the corresponding sequences of reported
appear to reflect higher-level representa- thoughts, differ between expert performers
tions that they have acquired to support rea- at least at the level of detailed thoughts. In
soning about clinical alternative diagnoses the previous section I showed how protocols
(Ericsson & Kintsch, 1995 ; Ericsson et al., uncover many higher-level characteristics of
2000; Patel, Arocha, & Kaufmann, 1994). expert performers mediating mechanisms,
There are also studies showing differ- such as skills supporting the expanded work-
ences in knowledge between experts and ing memory (Ericsson & Kintsch, 1995 ). In
less-accomplished individuals that mediate this section I will discuss attempts to exper-
successful task performance in experimen- imentally validate the detailed structure of
tal design of experiments in psychology the reported cognitive processes of individ-
(Schraagen, 1993 ) and detection of fraud ual expert performers.
in financial accounting (Johnson, Karim, & The complexity of the knowledge and
Berryman, 1991). The work on account- acquired skills of expert performers in
ing fraud was later developed into a gen- most domains, such as chess and medicine,
eral theory of fraud detection (Johnson, makes it virtually impossible to describe
Grazioli, Jamal, & Berryman, 2001). In the complete structure of the expertise
this handbook there are discussions of of an individual expert. For example,
the applications of verbal report method- Allen Newell (personal communication)
ology to study thinking in several differ- described a project in which one of his
ent domains of expertise, such as medicine graduate students in the 1970s tried to elicit
(Norman, Eva, Brooks, & Hamstra, Chap- all the relevant knowledge of a stamp col-
ter 19), software design (Sonnentag, Niessen, lector. After some forty hours of interviews,
& Volmer, Chapter 21), professional writ- Newell and his student gave up, as there
ing (Kellogg, Chapter 22), artistic perfor- was no sight of the end of the knowledge
mance (Noice & Noice, Cahpter 28), chess that the expert had acquired. As it may be
playing (Gobet & Charness, Chapter 3 0), difficult, perhaps impossible, to describe all
2 36 the cambridge handbook of expertise and expert performance
the knowledge and skills of experts, scien- 200 hours of practice was able to recall over
tists should follow the recommendations 80 digits in the digit-span task. During this
of the expert-performance approach. extended training period SF gave retrospec-
Namely, they should focus on the repro- tive reports on his thought processes after
ducible structure of the experts mecha- most memory trials. As his memory per-
nisms that mediate their superior perfor- formance started to increase he reported
mance on representative tasks (Ericsson, segmenting the presented lists into 3 -digit
1996b). Consequently, I will focus on groups and, whenever possible, encoding
selected domains of expertise in which them as running times for various races
regularities in the verbal reports of different because SF was an avid cross-country runner.
trials with representative tasks have been For example, SF would encode 3 5 8 as a very
analyzed. fast mile time, 3 minutes and 5 8 seconds, just
In the early applications of protocol anal- below the 4-minute mile. The central ques-
ysis there were several studies that col- tion concerning verbal reports is whether we
lected protocols from experts solving repre- can trust the validity of these reports and
sentative problems while thinking aloud. For whether the ability to generate mnemonic
example, Clarkson and Metzler (1960) col- running-time encodings influences memory.
lected protocols from a professional investor To address that issue Bill Chase and I
constructing portfolios of investments. Sim- designed an experiment to test the effects of
ilar detailed analyses of individual experts mnemonic encodings and presented SF with
from different domains have been briefly special types of lists of constrained digits. In
described in Ericsson and Simon (1993 ) and addition to a list of random digits we pre-
Hoffman (1992). These analyses were not, sented other lists that were constructed to
however, formally evaluated, and the pro- contain only 3 -digits groups that could not
posed mechanisms were not demonstrated be encoded as running times, such as 3 64 as
to account for reproducibly superior perfor- three minutes and sixty four seconds, in a list
mance on representative tasks. (3 64 895 481 . . . ). As predicted his perfor-
The most extensive applications of the mance decreased reliably. In another exper-
expert-performance approach using proto- iment we designed digit sequences where
col analysis to study individual experts have all 3 -digit groups could be encoded as run-
examined people with exceptional memory ning times (412 63 7 5 24 . . . ) with a reli-
(Ericsson & Lehmann, 1996). In the intro- able increase in his associated performance.
duction of this chapter I mentioned Binets In over a dozen specially designed experi-
(1894) pioneering work studying individu- ments it was possible to validate numerous
als with exceptional memory for numbers. aspects of SFs acquired memory skill (Chase
Several subsequent studies interviewed peo- & Ericsson, 1981, 1982; Ericsson, 1988b).
ple with exceptional memory, such as Lurias Other investigators, such as Wenger and
(1968) Subject S and Hunt and Loves (1972) Payne (1995 ), have also relied on protocol
VP (see Wilding and Valentine, 1997, Chap- analysis and other process-tracing data to
ter 3 1 for a review). However, the first study assess the mechanisms of individuals who
to trace the development of exceptional increased their memory performance dra-
memory from average performance to the matically with practice on a list-learning
best memory performance in the world (in task.
some memory tasks) was conducted in a More generally, this method has been
training study by Chase and Ericsson (1981, extended to any individual with exceptional
1982; Ericsson, Chase, & Faloon, 1980). We memory performance. During the first step,
studied a college student (SF) whose initial the exceptional individuals are given mem-
immediate memory for rapidly presented ory tasks where they could exhibit their
digits was around 7, in correspondence exceptional performance while giving con-
with the typical average (Miller, 195 6), but current and/or retrospective verbal reports.
he eventually acquired exceptional perfor- These reports are then analyzed to iden-
mance for immediate memory and after tify the mediating encoding and retrieval
protocol analysis and expert thought 2 37
human intelligence, Vol. 4 (pp. 13 7179). Hills- processes: A response to Howe. New Ideas in
dale, NJ: Erlbaum. Psychology, 9, 5 771.
Ericsson, K. A. (Ed.) (1996a). The road to Ericsson, K. A., Delaney, P. F., Weaver, G., &
excellence: The acquisition of expert performance Mahadevan, R. (2004). Uncovering the struc-
in the arts and sciences, sports, and games. Mah- ture of a memorists superior basic memory
wah, NJ: Erlbaum. capacity. Cognitive Psychology, 49, 19123 7.
Ericsson, K. A. (1996b). The acquisition of expert Ericsson, K. A., & Kintsch, W. (1995 ). Long-
performance: An introduction to some of the term working memory. Psychological Review,
issues. In K. A. Ericsson (Ed.), The road to 102 , 211245 .
excellence: The acquisition of expert performance Ericsson, K. A., & Kintsch, W. (2000). Shortcom-
in the arts and sciences, sports, and games (pp. 1 ings of article retrieval structures with slots of
5 0). Mahwah, NJ: Erlbaum. the type that Gobet (1993 ) proposed and mod-
Ericsson, K. A. (1998) Protocol analysis. In W. eled. British Journal of Psychology, 91, 5 715 88.
Bechtel & G. Graham (Eds.), A companion Ericsson, K. A., Krampe, R. Th., & Tesch-Romer,
to cognitive science (pp. 425 43 2). Cambridge, C. (1993 ). The role of deliberate practice in the
MA: Basil Blackwell. acquisition of expert performance. Psychologi-
Ericsson, K. A. (2001). Protocol analysis in psy- cal Review, 100(3 ), 3 63 406.
chology. In N. Smelser & P. Baltes (Eds.), Inter- Ericsson, K. A., & Lehmann, A. C. (1996).
national Encyclopedia of the Social and Behav- Expert and exceptional performance: Evidence
ioral Sciences (pp. 1225 612262). Oxford, UK: on maximal adaptations on task constraints.
Elsevier. Annual Review of Psychology, 47, 273 3 05 .
Ericsson, K. A. (2002). Toward a procedure for Ericsson, K. A., Patel, V. L., & Kintsch, W.
eliciting verbal expression of nonverbal expe- (2000). How experts adaptations to represen-
rience without reactivity: Interpreting the ver- tative task demands account for the expertise
bal overshadowing effect within the theoretical effect in memory recall: Comment on Vicente
framework for protocol analysis. Applied Cog- and Wang (1998). Psychological Review, 107,
nitive Psychology, 16, 981987. 5 785 92.
Ericsson, K. A. (2003 a). Valid and non-reactive Ericsson, K. A., & Polson, P. G. (1988a). Memory
verbalization of thoughts during performance for restaurant orders. In M. Chi, R. Glaser, &
of tasks: Toward a solution to the central prob- M. Farr (Eds.), The nature of expertise (pp. 23
lems of introspection as a source of scientific 70). Hillsdale, NJ: Erlbaum.
data. Journal of Consciousness Studies, 10, 118. Ericsson, K. A., & Polson, P. G. (1988b). An
Ericsson, K. A. (2003 b). Exceptional memori- experimental analysis of a memory skill for din-
zers: Made, not born. Trends in Cognitive Sci- ner orders. Journal of Experimental Psychology:
ences, 7, 23 3 23 5 . Learning, Memory, and Cognition, 14, 3 05 3 16.
Ericsson, K. A. (2003 c). The development of Ericsson, K. A., & Simon, H. A. (1980). Verbal
elite performance and deliberate practice: An reports as data. Psychological Review, 87, 215
update from the perspective of the expert- 25 1.
performance approach. In J. Starkes & K. A. Ericsson, K. A., & Simon, H. A. (1984). Proto-
Ericsson (Eds.), Expert performance in sport: col analysis: Verbal reports as data. Cambridge,
Recent advances in research on sport exper- MA: Bradford books/MIT Press.
tise (pp. 4981). Champaign, IL: Human Ericsson, K. A., & Simon, H. A. (1993 ). Protocol
Kinetics. analysis: Verbal reports as data (revised edition).
Ericsson, K. A. (2004). Deliberate practice and Cambridge, MA: Bradford books/MIT Press.
the acquisition and maintenance of expert per- Ericsson, K. A., & Smith, J. (1991). Prospects
formance in medicine and related domains. and limits in the empirical study of expertise:
Academic Medicine, 10, S70S81. An introduction. In K. A. Ericsson & J. Smith
Ericsson, K. A., Chase, W., & Faloon, S. (1980). (Eds.), Toward a general theory of expertise:
Acquisition of a memory skill. Science, 2 08, Prospects and limits (pp. 13 8). Cambridge:
11811182. Cambridge University Press.
Ericsson, K. A., & Crutcher, R. J. (1991). Intro- French, K. E., Nevett, M. E., Spurgeon, J. H.,
spection and verbal reports on cognitive pro- Graham, K. C., Rink, J. E., & McPherson,
cesses two approaches to the study of thought S. L. (1996). Knowledge representation and
2 40 the cambridge handbook of expertise and expert performance
problem solution in expert and novice youth Miller, G. A., Galanter, E., & Pribram, K. H.
baseball players. Research Quarterly for Exercise (1960). Plans and the structure of behavior. New
and Sport, 67, 3 863 95 . York: Holt, Rinehart, and Winston.
Gagne, R. M., & Smith, E. C. (1962). A study of Neuman, Y., & Schwarz, B. (1998). Is self-
the effects of verbalization on problem solving. explanation while solving problems helpful?
Journal of Experimental Psychology, 63 , 1218. The case of analogical problem-solving.
Galton, F. (1879). Psychometric experiments. British Journal of Educational Psychology, 68,
Brain, 2 , 148162. 15 24.
Galton, F. (1883 ). Inquiries into human faculty and Neuman, Y., Leibowitz, L., & Schwarz, B. (2000).
its development. New York: Dutton. Patterns of verbal mediation during prob-
lem solving: A sequential analysis of self-
Green, A. J. F. (1998). Using verbal protocols in lan-
explanation. Journal of Experimental Education,
guage testing research: A handbook. Cambridge,
68, 197213 .
UK: Cambridge University Press.
Newell, A. (1990). Unified theories of cognition.
Henderson, R. D., Smith, M. C., Podd, J., &
Cambridge, MA: Harvard University Press.
Varela-Alvarez, H. (1995 ). A comparison of
the four prominent user-based methods for Newell, A., & Simon, H. A. (1972). Human prob-
evaluating the usability of computer software. lem solving. Englewood Cliffs, NJ: Prentice-
Ergonomics, 3 9, 203 02044. Hall.
Hoffman, R. R. (Ed.) (1992). The psychology of Nielsen, S. (1999). Regulation of learning strate-
expertise: Cognitive research and empirical AI. gies during practice: A case study of a single
New York: Springer-Verlag. church organ student preparing a particular
work for a concert performance. Psychology of
Hughes, J., & Parkes, S. (2003 ). Trends in the use
Music, 2 7, 218229.
of verbal protocol analysis in software engineer-
ing research. Behaviour & Information Technol- Nisbett, R. E., & Wilson, T. D. (1977). Telling
ogy, 2 2 , 127140. more than we can know: Verbal reports on
mental processes. Psychological Review, 84,
Hunt, E., & Love, T. (1972). How good can mem-
23 125 9.
ory be? In A. W. Melton & E. Martin (Eds.),
Coding processes in human memory (pp. 23 7 Norman, G. R., Trott, A. D., Brooks, L. R., &
260). New York: Holt. Smith, E. K. M. (1994). Cognitive differences
in clinical reasoning related to postgraduate
Johnson, P. E., Karim, J., & Berryman, R. G.
training. Teaching and Learning in Medicine, 6,
(1991). Effects of framing on auditor deci-
114120.
sions. Organizational Behavior & Human Deci-
sion Processes, 5 0, 75 105 Patel, V. L., Arocha, J. F., & Kaufmann, D. R.
(1994). Diagnostic reasoning and medical
Johnson, P. E., Grazioli, S., Jamal, K., & Berry-
expertise. In D. Medin (Ed.), The psychology of
man, R. G. (2001). Detecting deception: Adver-
learning and motivation, Vol. 3 0 (pp. 18725 1).
sarial problem solving in a low base-rate world.
New York: Academic Press.
Cognitive Science, 2 5 , 3 5 5 3 92.
Patel, V. L., & Groen, G. J. (1991). The general
Karpov, A. (1995 ). Grandmaster musings. Chess
and specific nature of medical expertise: A crit-
Life, November, pp. 3 23 3 .
ical look. In K. A. Ericsson & J. Smith (Eds.),
Keller, F. S. (195 8). The phantom plateau. Journal Toward a general theory of expertise (pp. 93
of the Experimental Analysis of Behavior, 1, 113 . 125 ). Cambridge, MA: Cambridge University
Koltanowski, G. (1985 ). In the dark. Coraopolis, Press.
PA: Chess Enterprises. Pressley, M., & Afflerbach, P. (1995 ). Verbal pro-
Luria, A. R. (1968). The mind of a mnemonist. New tocols of reading: The nature of constructively
York: Avon. responsive reading. Hillsdale, NJ: Erlbaum.
McKelvie, S. J. (1995 ). The VVIQ and beyond: Renkl, A. (1997). Learning from worked-out
Vividness and its measurement. Journal of Men- examples: A study on individual differences.
tal Imagery, 19, 19725 2. Cognitive Science, 2 1, 129.
Miller, G. A. (195 6). The magical number seven, Richardson, J. T. E. (1988). Vividness and
plus or minus two: Some limits of our capac- unvividness: Reliability, consistency, and valid-
ity for processing information. Psychological ity of subjective imagery ratings. Journal of
Review, 63 , 8197. Mental Imagery, 12 , 115 122.
protocol analysis and expert thought 2 41
Saariluoma, P. (1991). Aspects of skilled imagery Von Eckardt, B. (1998). Psychology of intro-
in blindfold chess. Acta Psychologica, 77, 65 spection. In E. Craig (Ed.), Routledge encyclo-
89. pedia of philosophy (pp. 842846). London:
Saariluoma, P. (1992). Error in chess: The Routledge.
apperception-restructuring view. Psychological Ward. P., Hodges, N. J., Williams, A. M., &
Research/Psychologische Forschung, 5 4, 1726. Starkes, J. L. (2004). Deliberate practice and
Saariluoma, P. (1995 ). Chess players thinking. expert performance: Defining the path to
London: Routledge. excellence. In A. M. Williams & N. J. Hodges
Schmidt, H. G., & Boshuizen, H. (1993 ). On (Eds.), Skill acquisition in sport: Research, the-
acquiring expertise in medicine. Educational ory and practice (pp. 23 125 8). London, UK:
Psychology Review, 5 , 205 221. Routledge.
Schraagen, J. M. (1993 ). How experts solve a Watson, J. B. (1913 ). Psychology as the behaviorist
novel problem in experimental design. Cogni- views it. Psychological Review, 2 0, 15 877.
tive Science, 17, 285 3 09. Watson, J. B. (1920). Is thinking merely the action
Simon, H. A., & Chase, W. G. (1973 ). Skill in of language mechanisms? British Journal of Psy-
chess. American Scientist, 61, 3 94403 . chology, 11, 87104.
Simon, H. A., & Kaplan, C. A. (1989). Founda- Wenger, M. J., & Payne, D. G. (1995 ). On the
tions of cognitive science. In M. J. Posner (Ed.), acquisition of mnemonic skill: Application of
Foundations of cognitive science (pp. 147). Cam- skilled memory theory. Journal of Experimental
bridge, MA: MIT Press. Psychology: Applied, 1, 194215 .
Wilding, J., & Valentine, E. (1997). Superior mem-
Simpson, S. A., & Gilhooly, K. J. (1997). Diagnos-
ory. Hove, UK: Psychology Press.
tic thinking processes: Evidence from a con-
structive interaction study of electrocardio- Wundt, W. (1897). Outlines of psychology (Trans-
gram (ECG) interpretation. Applied Cognitive lated by C. H. Judd). Leipzig: Wilhelm
Psychology, 11, 5 43 5 5 4. Engelmann.
Starkes, J., & Ericsson, K. A. (Eds.) (2003 ). Wundt, W. (1907). Uber Ausfrageexperimente
Expert performance in sport: Recent advances und uber die Methoden zur Psychologie des
in research on sport expertise. Champaign, IL: Denkens [On interrogation experiments and
Human Kinetics. on the methods of the psychology of thinking].
Philosophische Studien, 3 , 3 013 60.
Sudman, S., Bradburn, N. M., & Schwarz,
N. (Eds.) (1996). Thinking about answers:
The application of cognitive processes to survey
methodology. San Francisco, CA: Jossey-Bass.
van der Maas, H. L. J., & Wagenmakers, E. J. Author Notes
(2005 ). A psychometric analysis of chess exper-
tise. American Journal of Psychology, 118, 2960. This article was prepared in part with sup-
Verplanck, W. S. (1962). Unaware of wheres port from the FSCW/Conradi Endowment Fund
awareness: Some verbal operants-notates, of Florida State University Foundation. The
moments and notants. In C. W. Eriksen author wants to thank Robert Hoffman, Katy
(Ed.), Behavior and awareness a symposium Nandagopal, and Roy Roring for their valu-
of research and interpretations (pp. 13 015 8). able comments on an earlier draft of this
Durham, NC: Duke University Press. Chapter.
C H A P T E R 14
(e.g., representative real-world tasks recre- mance (e.g., Chi, Glaser, & Farr, 1980;
ated using mechanical, video, or computer Ericsson, 1996; Ericsson & Smith, 1991;
technology) (for a review, see Gray, 2002). Hoffman, 1992; Salas & Klein, 2001; Starkes
Although these technologies have been & Ericsson, 2003 ). These researchers and
developed primarily for purposes other than others have arrived at different assump-
understanding complex performance, they tions about expert cognition and have, thus,
can be put to that purpose. relied on divergent paradigms and meth-
In this chapter, we consider a narrow ods to assess performance. Following de
bandwidth of studies that fall under the Groots (1978/46) original work in chess,
general rubric of simulated task environ- proponents of the early expertise approach
ments, as well as virtual reality, primar- designed classic structured and unstructured
ily because they have specifically addressed recall experiments to capture the mem-
issues related to expert performance and ory feats of expert chess players (Chase
skill development. Each study varies with & Simon, 1973 ). This work was motivated
respect to the degree of physical fidelity by the assumption that experts could cir-
and ecological representativeness (Hoffman cumvent short-term memory (STM) limita-
& Deffenbacher, 1993 ). However, psycho- tions by storing chunks in STM. However,
logical fidelity the degree to which the subsequent research questioned the STM
system captures the real-world demands of storage assumption, revealing that experts
the task, as well as the way in which it is stored domain-specific information in acces-
implemented as an assessment and training sible form in long-term memory (LTM)
tool is likely to be of greater importance (e.g., Charness, 1976). Moreover, superior
(e.g., Entin, Serfaty, Elliot, & Schiflett, 2001; memory recall is likely to be an inciden-
Salas, Bowers, & Rhodenizer, 1998; Williams tal by-product of their memory organiza-
& Ward, 2003 ). tion as opposed to a representative per-
Our aim is to provide an overview of formance metric (see Chase & Ericsson,
simulation tasks, environments, and tech- 1982). De Groot (1978/46) noted that other
nologies used to assess expert performance activities that better simulated the task
and train sport, medical surgery, and avia- requirements (e.g., selection of next best
tion skills. First, we describe the current state move) were actually better predictors of per-
of expert-performance research conducted formance than memory recall (de Groot,
in a simulated task environment and high- 1978/1946).
light some factors constraining the develop- As an alternative to the original expertise
ment of paradigms and methods. Next, we approach, and to counter the assumption
summarize the development of simulation that the knowledge elicited from so-called
in each domain and emphasize pertinent experts could account for expert-novice per-
issues in nurturing skill acquisition. We also formance differences (see Fischhoff, 1989),
address some misconceptions about simula- Ericsson and Smith (1991) advocated the
tion training by reviewing available proce- expert-performance approach in which
dures used to successfully train individuals researchers first identify tasks that truly cap-
under simulated conditions. We begin with ture expertise. Representative tasks can then
a synopsis of expertise research and its devel- be recreated in the laboratory where reliably
opment and the role of simulation in assess- superior performance can be assessed, exper-
ing expert performance. imental control maintained, and underlying
mechanisms identified via the use of process-
tracing methods. Although the expert-
Simulation for Performance: Assessing performance approach has been adopted
the Superior Performance of Experts in domains such as sport, music, games,
and medicine (see Ericsson, 1996; Starkes &
A bounty of research now exists on the Ericsson, 2003 ), few researchers have fully
nature of expertise and expert perfor- embraced it or used simulation as a means
simulation for performance and training 2 45
of recreating the task. What follows is an instance, methods were devised to measure
overview of relevant expertise research using an individuals skill in anticipating an oppo-
simulated task environments in sports, avia- nents intentions (e.g., Haskins, 1965 ; Jones
tion, and surgery. & Miles, 1978). Although this research was
innovative for its time, participants were
required to respond in a different modality
Expert Performance in Simulated
(e.g., pen and paper, joystick) to that used
Sports Tasks
during actual task performance, or informa-
The recall paradigm (Chase & Simon, 1973 ) tion was presented in a fundamentally dif-
was used by early researchers interested ferent manner (e.g., via static images, Xs
in experts who engaged in perceptually- and Os representing offense and defense).
demanding sports. Allard and colleagues For instance, using a video-based simula-
(Allard, Graham, & Paarsalu, 1980; Allard tion and paper-and-pen response, Williams
& Starkes, 1980) investigated whether skill and Burwitz (1993 ) examined a key com-
groups differed in their ability to recall ponent of soccer goalkeeping anticipat-
patterns of play in basketball and volley- ing the direction of a penalty kick. Through
ball, respectively. Varsity and intramural bas- the use of a temporal occlusion paradigm
ketball players were presented with static, (a technique used to temporally limit the
structured game-play and non-game sce- availability of visual information on which
narios and asked to recall player positions a decision could be made), they found that,
under time pressure. In line with the chess compared to inexperienced players, experi-
findings, varsity players recalled more posi- enced players could more accurately predict
tions than less-skilled players in the struc- shot destination only when the simulation
tured game condition only. In volleyball, no was occluded prior to striking the ball. No
differences were found between national- skill-based differences were observed there-
and intramural-level players in either condi- after (i.e., at/post foot-ball contact), indi-
tion. Borgeaud and Abernethy (1987) modi- cating that only skilled players could antici-
fied Allard and Starkes (1980) study by pate the future consequences of action based
using dynamic film sequences in volleyball. on advance information available from key
They found distinct differences in expert and contextual cues, such as their opponents
novice recall and concluded that simulating posture.
real-world perceptual characteristics of the Such tasks may capture different, or at
task is likely to be a more informative way of least ancillary, cognitive processes instead of
studying expert-novice differences in sport. those used during actual task performance.
The dynamic task used by Borgeaud and A recent meta-analysis of sports-expertise
Abernethy (1987) and others (e.g., Williams, research suggests that increasing the ecolog-
Davids, Burwitz, & Williams, 1993 ) recre- ical representativeness of the action compo-
ated the typical viewpoint experienced dur- nent resulted in a larger effect size (Thomas,
ing a game. However, the experimental task 2003 ). Counter to this intuition, however,
still required individuals to invoke a pro- research from our laboratories shows that
cess (i.e., memory recall) that they otherwise when the aim of the simulation is to recreate
may not have used, at least explicitly, during perceptual-cognitive demands of the task,
a typical game. Accordingly, recall tasks may participants need not necessarily be placed
provide only limited insight into the mech- under associated perceptual-motor demands
anisms underlying actual performance com- (Williams, Ward, Allen, & Smeeton, 2004).
pared to more representative tasks that sim- Consequently, even when part-task simula-
ulate real-world constraints. tion is used, the crucial aspects of perfor-
In parallel with the memory-recall mance (i.e., ecological salience, see Hoffman
research, paradigms emerged that more & Deffenbacher, 1993 ) may still be cap-
closely simulated the actual perceptual- tured, if not the essence of the task itself.
cognitive demands placed on an athlete. For Advances in measurement and simulation
2 46 the cambridge handbook of expertise and expert performance
Anticipation-
Projector
simulation
system
Life-size
screen
Eye
tracking
system
suggesting that skilled players were more to more superficial form cues, to direct their
adept at picking up earlier-occurring and response.
more-informative movement cues. Ward Although sports researchers have used
et al. (2002) presented the same information a number of minimally interactive video-
(i.e., movements of the tennis opponent) based simulations to examine issues in
in point-light form to determine the nature expert performance, few have adopted
of the perceptual information extracted alternative or, arguably, more-advanced
during simulated anticipatory performance. interactive simulation. In a rare study
Novices shifted their attention solely toward Walls, Bertrand, Gale, and Saunders (1998)
the racket, whereas skilled players continued assessed dinghy sailing performance in com-
to extract information from the torso. While petitive helmsmen. The simulator was com-
a performance decrement was observed in prised of a physical laser dinghy deck piv-
both groups under point-light conditions, oting between two supports, dynamically
the skill-based differences remained across controlled by a computer-operated pneu-
conditions. The results suggested experts matic arm. Helming, sheeting, tacking, and
used the relative motion information avail- boat trim were represented virtually using
able from the joint kinematics, as opposed computer graphics. An illustration of the
2 48 the cambridge handbook of expertise and expert performance
Figure 14.3. Graphic depiction of the dinghy-sailing simulator used by Walls et al. (1998).
simulator is provided in Figure 14.3 . Partic- ment, or maintain situation awareness (SA)
ipants had to sail upwind, tacking and then under challenging conditions has become
maneuvering the boat round a buoy while synonymous with skilled performance (e.g.,
continually monitoring conditions to main- Endsley, 1995 ). Aviation research has high-
tain control of the dinghy. Time to com- lighted the need to increase pilot SA to
plete the simulated course was highly cor- lessen the risk of accidents, reduce fatality
related with rankings of performance. The rates in landing and take-off, and improve
best sailors did not sail shorter distances, they performance (see Durso and Dattel, Chap-
were simply faster at completing the course. ter 20, this volume). However, our under-
The implication is that more-skilled sailors standing of the cognitive mechanisms and
were better at transforming the available processes that constitute or facilitate SA is
visual information into necessary actions limited. Moreover, although aviation simu-
that allowed them to maintain control and lation is perhaps more advanced than any
maximize speed of the dinghy. other domain, relatively few attempts have
A number of virtual realities have been been made to use simulated task envi-
created that simulate the sporting envi- ronments to aid our understanding of the
ronment, such as EasyBowling, a virtual perceptual-cognitive or perceptual-motor
bowling game machine; PCCAVEmash, bases of expert pilot performance.
an immersive table tennis game (see As in the sports literature, a number
www.cad.zju.edu.cn/division/vrmm.html; of researchers have used both static tasks
Zheijiang University, China); and the and dynamic simulations to examine perfor-
Virtual Football Trainer, a CAVE-based mance. Doane, Sohn, and Jodlowski (2004)
American football simulation (see www- examined expert and novice pilots ability to
vrl.umich.edu/project/football/index.html; anticipate the consequences of flight actions
University of Michigan). However, sim- (an integral aspect of long-term working
ulations using this media have typically memory and a higher-level component of
not been embraced by the sports science SA; see Ericsson & Kintsch, 1995 ) using a
community to examine expert performance. static, simulated task environment. The task
An important question to ask is whether was to determine whether a change state-
the increased physical fidelity and cost of ment (i.e., the resultant main or side effects
such systems increases their benefit to per- of a control movement) was consistent with
formance compared to video- or PC-based the application of control movements on
simulations (see Salas et al., 1998). a simulated cockpit depiction of the cur-
rent flight situation. Experts were typically
quicker and marginally more sensitive to
Assessing the Skills of Expert Aviation
whether trials were consistent. Differences
Pilots using Simulation
were significantly amplified when two con-
The ability to recognize situations as famil- trol movements interacted with each other
iar, make an appropriate strategic assess- compared to when they were independent
simulation for performance and training 2 49
or when single control movements were pre- flight yoke and throttle, and a modified ver-
sented. sion of Microsoft Flight Simulator, was used
Doane et al. (2004) argued that experts to display a typical instrument panel. Par-
functionally increase their working memory ticipants flew seven simulated flight seg-
capacity, enabling them to process simul- ments within predetermined bounds (i.e.,
taneously the interactive effects of main 5 0 feet, 5 knots; 5 ). On the next
and side effects of control movements on day, they flew the same seven plus an addi-
the current situation more effectively, in tional two segments involving a partial vac-
contrast to processing control information uum failure that affected the attitude indi-
independently or in sequence. In a vari- cator. The failure was announced on the final
ation of this study, Jodlowski and Doane segment only.
(2003 ) examined whether long-term work- Expert pilots with an average of over 2200
ing memory skill (LT-WM; defined by these total hours of flight time were 20% more suc-
authors as difference scores between recall cessful at staying within the specified bounds
on structured and unstructured flight dis- compared to apprentice pilots who had 89
plays) or working memory (WM) capacity hours of total flight time. When the failure
(e.g., reading span and spatial orientation) was announced, skill groups did not signifi-
predicted performance on a similar static cantly differ from each other (although we
simulation. Expert instructor and student calculated a moderate effect size in favor of
pilots WM memory scores did not differ experts; Cohens d = 0.5 ) and their perfor-
(but see Sohn & Doane, 2003 ), but experts mance was comparable to that in routine
did attain higher LT-WM scores than stu- flight. When the failure was unannounced,
dents, and this was a good predictor of their, both groups scores were reduced by around
but not novices, performance level. Results 25 %. The lack of an expert advantage for
suggest that explanations based on LT- non-routine or unfamiliar situations implies
WM skill in recognizing meaningful displays that only routine, as opposed to adap-
were more informative than those based on tive, expertise was acquired (see Hatano &
WM capacity. Ignaki, 1986). Although some researchers
Although the definition of LT-WM used have intimated that experts acquire flex-
was somewhat restrictive, the results are ibility rather than rigidity with increased
consistent with Ericsson and Kintschs skill (see Feltovich, Spiro, & Coulson,
(1995 ) LT-WM theory. Experts can over- 1997), only rarely has this distinction been
come short-term working memory con- empirically tested.
straints by acquiring and applying superior To examine expert pilots attentional
indexing skills at encoding that result in flexibility and monitoring skills, Bellenkes,
a more accessible domain-specific retrieval Wickens, and Kramer (1997) assessed per-
structure. The static simulation tasks used formance using a PCATD (e.g., see http://
in these experiments, however, are likely to www.flyelite.com/faa-approved.php). The
have omitted key dynamic aspects of perfor- simulator displayed an on-screen virtual
mance that are important for both captur- instrument panel and was controlled via
ing the ecologically salient aspects of expert a right side arm-mounted joystick. Partici-
performance and assessing experts ability pants flew similar segments to those men-
to build an accurate situation model (see tioned in Jodlowski et al. (2003 ). Skilled
van Dijk & Kintsch, 1983 ; Kintsch, 1988; flight instructors demonstrated superior
Ericsson & Kintsch, 1995 ). tracking accuracy in the vertical and longi-
Jodlowski, Doane, and Brou (2003 ) tudinal axes but not on the lateral axis com-
extended their work to a dynamic flight pared to student pilots. Less deviation from
simulation and examined pilots ability to the desired flight path on each axis by flight
adapt to the changing constraints of routine instructors seemed to be a result of greater
and non-routine instrument flight situations. responsiveness to the changing constraints of
A personal computer-based aviation train- the flight situation and was largely a func-
ing device (PCATD), comprised of a Cirrus tion of greater flexibility in attention and
2 50 the cambridge handbook of expertise and expert performance
control strategy. This conclusion was sup- Assessing Expert Skill via Surgical
ported by the eye-movement data. Flight Simulation
instructors sampled relevant instruments
more frequently than student pilots who, in Surgery is one of the most demanding of all
turn, distributed their time scanning each performance tasks since, by definition, life-
instrument. threatening circumstances are encountered
In a comparison of United States in almost every incidence (see Norman,
Air Force pilots (experts) and Academy et al., Chapter 19, this volume). Mistakes,
cadet pilots (apprentices), Kasarskis, Ste- which in other domains could be considered
hwien, Hickox, Aretz, and Wickens (2001) negligible, are not simply dangerous here.
assessed participants eye-movement behav- They can be fatal. Given the costs of failure
ior during simulated flight. Participants flew in this domain, the attainment and assess-
approaches and landings using a PCATD, ment of expert skills is vital. Objective per-
starting on a 45 turn at 1000ft with the formance assessment in the real-world is
descent lasting around three minutes. Fly! problematic for obvious reasons. Simulation
software (Terminal Reality) was used to allows such limitations to be circumvented
depict both flight instruments and external by reproducing real-world task demands in
environment, and a flight yoke was used to a standardized setting.
control pitch, roll, and airspeed. Experts Initial research examining performance
landing performance was less variable than on relatively low-tech simulators indicated
the apprentices, and they employed signifi- that although basic skills transfer from one
cantly more fixations of shorter duration on simulated task to another using the same
the runway aim point and airspeed indica- system, there is little evidence that these
tor. The authors suggested that this strategic skills improve surgical performance (Rosser,
monitoring difference between the groups Rosser, & Savalgi, 1997; Strom, Kjellin,
afforded greater control and precision of Hedman, Wredmark, & Fellander-Tsai,
the aircraft by experts. The experts shorter 2004). However, with the introduction of
dwell times were thought to reflect their skill more advanced minimal access simulators,
level; they simply needed less time to extract researchers have been able to recreate
relevant information. traditional clinical tasks used in educational
In summary, the research suggests that contexts. Haluck, et al. (2001), for instance,
expert pilots are better able to anticipate used a laparoscopic simulator (Laparoscopic
the consequences of the current situation. Impulse Engine, Immersion Corporation,
The higher visual search rate during impor- see http://www.immersion.com/medical/
tant aspects of flight, and marginally more products/laparoscopy/) to assess the ability
effective use of flight controls, indicates of skilled surgical staff (experts) and medi-
that experts develop superior monitoring cal students to navigate a virtual operating
and control skills that allow them to main- volume and identify six randomly placed
tain awareness of the situation and adapt to arrows a laparoscopic procedure consis-
the dynamic constraints of the environment. tent with those used by the Royal College
However, in situations that are not necessar- of Surgeons (RCS) (see Torkington, Smith,
ily routine the reduction of expert perfor- Rees, & Darzi, 2001). Experts identified
mance to student levels indicates that exper- more arrows within the allotted timeframe
tise is highly context specific, not only to and made fewer tracking errors than stu-
the domain itself but to particular aspects of dents. Although this study demonstrates
the task. experts superior perceptual-motor skills
From aviation, we now turn to research during laparoscopic-type procedures, it
on surgical expertise using simulated tasks is difficult to determine whether the
or virtual environments and highlight perceptual-cognitive elements of the task,
potential mechanisms implicated in expert or the task as a whole, truly reflected the
performance. demands faced during surgical conditions,
simulation for performance and training 2 51
or simply whether some other skill set was iarization on the Xitact LS5 00 laparoscopic
being assessed. cholecystectomy simulator and hands-on
Verner, Oleynikov, Holtmann, Haider, task instruction, participants performed
and Zhukov (2003 ) used the da Vinci the task three times, attaining a sum skill
robotic surgical system (see http://www. score for each trial. Skilled surgeons were
intuitivesurgical.com/products/da vinci approximately 13 % better than novices for
html) to examine the coordination patterns the second and final trials, and experts com-
of surgeons during a simulated procedure. pleted the task in half the time (approx. 100
Grip position and three-dimensional posi- v 210 s). Although these findings demon-
tion and trajectory of the laparoscopic strate the utility of simulation for medical
instrument were assessed as participants performance assessment, the question is
picked up and placed a bead onto a peg raised: by what specific mechanisms are
(another RCS-type training task) and then experts able to consistently out-perform
returned the surgical tool to its original novice participants on such tasks? The next
position. Experts were quicker than novices challenge for surgical simulation research is
at performing the task (approx. 7 v 11s, to employ process-tracing methodologies
respectively), mainly due to the fact that to determine the mechanisms implicated
experts spent less time during points of tran- in superior performance (e.g., see Patel &
sition (i.e., picking up and placing the bead). Groen, 1986).
Kinematic data indicated that novices were In an attempt to trace perceptual-
much slower and more variable. Experts cognitive processes during performance
were 15 % and 5 0% faster than novices in a simulated environment, Law, Atkins,
with their dominant and non-dominant Kirkpatrick, Lomax, and Mackenzie (2004)
hand, respectively. The lower variability assessed expert surgeons and novice college
of experts, and lack of difference between students eye-movements while performing
hands, is consistent with the literature a PC-based, laparoscopic simulation (http://
on motor-skill acquisition and deliberate www.immersion.com/medical/products/
practice (see, Williams & Hodges, 2004; laparoscopy/; Immersion Corporation
Ericsson, Krampe, & Tesch-Romer, 1993 ). Laparoscopic Impulse Engine). In a sim-
As individuals refine their skills over time, plified version of a laparoscopic task,
their coordination improves, significantly participants guided a virtual tool with one
reducing the variability with which an hand toward a specified target. Expert
action is executed, allowing performance to surgeons were approximately 100s quicker
be more reliably reproduced with greater than novices at reaching the target location,
control. although this difference diminished to
A few studies have begun to make use around 25 s with practice. Expert surgeons
of simulation technologies that closely rep- visually fixated on the target far more
resent the stimuli, procedures, and actions and tracked the laparoscopic tool far less
that would be present or performed in than their novice counterparts. Experts
the real-world. Schijven and Jakimowicz performance was facilitated by centering
(2003 ) compared surgeons with experience their point of gaze around the target earlier
of performing over 100 clinical laparoscopic in the tool movement, allowing the tool
cholecystectomies (i.e., surgical excision of to be tracked in the periphery. In contrast,
the gallbladder) with novice surgical resi- the novices tended to track the tool using
dents and interns who had no experience the fovea and alternated the gaze more
in this procedure on a laparoscopic clip- frequently between the tool and target.
and-cut-cystic-artery-and-duct task (for The available surgical-simulation
examples of the typical viewpoint during research suggests that experts demon-
a simulated clip-and-cut task, see http:// strate greater precision and speed through
www.simbionix.com/LAP Laparoscopic an enhanced ability to control and track
Instruments.html). After a period of famil- laparoscopic tools. Experts also acquire a
2 52 the cambridge handbook of expertise and expert performance
different perceptual or attention strategy tion and feedback were used to guide the
to that of novices, which facilitates their mode of delivery in a simulated task envi-
superior level of control. Simulation has ronment (Staszewski & Davison, 2000).
only recently been used as a medium to This method considerably improved sol-
examine expert performance in surgery. diers mine-detection performance, and
Initial research suggests that simulated most encouragingly, the greatest improve-
task environments are likely to be very ment occurred on the most threatening
conducive to improving our understanding types of mines when retested in the real
of skilled performance. world.
We now turn to an overview of the use of Training based on methods similar to
simulation for training perceptual-cognitive those proposed by Stazsewski and others
and perceptual-motor skills and provide a (e.g., Williams et al., 2002) is limited, partic-
summary of the development of simulation ularly when it has taken place in a simulated
and training research in aviation, medicine, environment. However, recent innovations
and sport. have been employed that, when coupled
with effective training procedures, offer sup-
port for the idea that simulation is an effec-
Simulation for Training: Development tive training tool. The next section provides
and Research an overview of a selection of studies from
each domain. The history of simulation lies
Historically, training strategies have been largely in the aviation domain, and so we
based on intuition and emulation rather begin with a summary of the development
than on evidence-based practice (Williams & of flight simulation and its application to
Ward, 2001; 2003 ). Researchers have often training.
relied on mere exposure to simulation as
opposed to using simulation as a means to
The Development of Flight Simulation
deliver instruction. Today, however, empir-
and its Application to Training
ically grounded methods exist that have
demonstrated improvement. Staszewski and Flight simulation can be traced back almost
Davison (2000) developed a method they as far as powered flight itself. One of the ear-
termed Cognitive Engineering Based on liest simulators, the Link Trainer, was pro-
Expert Skill (CEBES) that applies the the- duced by Ed Link in 1929 and was first
ory, principles, and methods of cognitive used by the Army Air Corps in 193 4 (for
science to developing effective training. In a more detailed history of the Link Trainer,
this approach, an expert model is first see http://www.link.com/history.html). The
derived from empirical evidence of how an Link Trainer was used to reduce the num-
expert performs a particular task, which ber of pilot fatalities in the first few days
in turn serves as a blueprint for training. of service; events that were attributed to a
Staszewski (1999) used this approach to lack of experience in instrumented flight,
derive an expert model of mine-detection night operations, and inclement weather
clearance operations. Using an information- (see Allerton, 2000). Development was typ-
processing analysis of an expert while per- ically spurred by technological advances and
forming the task, the way in which an specific motives, such as the desire to famil-
expert searched for, discriminated between, iarize military and commercial pilots with
and accurately detected mines was decom- flying missions without actually having to
posed into specific equipment manipula- fly, and to create affordable training environ-
tions and performance-specific perceptual- ments to prepare apprentice pilots.
information, knowledge, and thought pro- In the decades following introduction of
cesses. This information provided the basis the Link Trainer, technological advancement
for the content on which participants were was the primary motivation for improve-
trained, and established methods of instruc- ment. Systems moved from pneumatic to
simulation for performance and training 2 53
hydraulic, analog to digital, from the exclu- 1990; Smode, Hall, & Meyer, 1966). Typ-
sion of visual displays entirely to the intro- ically, however, simulation-trained groups
duction of simple light projections, and have spent more time in training overall
finally toward high-end digital image gen- compared to those trained in traditional air-
eration. Although these advances have sur- craft training. After factoring in the typ-
passed any other area of simulation, such ical cost of traditional training methods,
development has occurred at a high finan- the current results suggest that significant
cial cost. State-of-the-art simulations are savings would be made by using simula-
now beyond the budget of most researchers. tion, making this a fiscally viable, albeit not
Furthermore, high-fidelity systems are rarely necessarily time-efficient method of train-
developed with sophisticated measurement ing. Roscoe and colleagues (e.g. Roscoe,
in mind. Where systems can be adapted, the 1971; Povenmire & Roscoe, 1971, 1973 )
costs are often inordinate and such adapta- pointed out that the efficiency of flight-
tions lend themselves to only a handful of simulation training is greater during the ini-
research questions (see Gray, 2002; Hays, tial periods of learning. As training con-
Jacobs, Prince, & Salas, 1992). tinues and performance improves, transfer
Hays et al. (1992) conducted a meta- gain significantly reduces in a negatively
analysis of experiments published between decelerating manner. To determine at what
195 7 and 1986 to assess the effectiveness of point simulation becomes ineffective, or
flight simulation for improving trainee-pilot at least less cost-effective, Povenmire and
performance. From 247 simulation articles, Roscoe (1971, 1973 ) suggested that incre-
only 19 on jet and seven on helicopter sim- mental transfer functions need to be deter-
ulation were retained in the analysis. Over mined.
90% of the effects supported the joint use of Taylor, Talleur, Emanuel, Rantanen, Brad-
simulation and aircraft training over aircraft shaw, and Phillips (2001, 2002) examined
training alone. Small but positive effects three different periods of PCATD simu-
were observed for jet, but not helicopter, lation training; five hours (PCATD 5 ), 10
simulation training when contrasted with hours (PCATD 10), and 15 (PCATD 15 )
aircraft training alone, and these effects were hours, and compared to a control group.
particularly pronounced for certain types of Transfer-effectiveness ratios (see Povenmire
task, such as takeoffs, approaches, and land- & Roscoe, 1973 ) showed that PCATD train-
ings. This finding is encouraging given the ing was generally effective and resulted in
hazardous nature of landing and takeoff (see fewer trials in the airplane compared to
Kasarskis et al., 2001; Khatwa & Helmreich, the control group who received no PCATD
1999, see also Higgins, Chignell, & Hancock, training. Incremental transfer-effectiveness
1989). ratios suggested that the greatest amount of
Hays and colleagues analyses indicated positive transfer was found in the PCATD
that self-paced training to criterion was more 5 group. The additional training received
effective than when practice was simply by the PCATD 15 group failed to save any
blocked. However, given the trend to train additional trials in the airplane compared
to criterion rather than assessing comparable to the PCATD 10 group. Overall, only lim-
degrees of different types of training under ited additional time/trials were saved by the
simulation (e.g., explicit vs. implicit instruc- PCATD 10 group compared to the PCATD
tion), it is unclear to what extent partici- 5 group. The authors concluded that lit-
pants improvement is simply a function of tle additional benefit was found for PCATD
the amount of time invested rather than the simulation training beyond five hours. This
nature of training employed. In a number finding questions the common conception
of reviews, researchers have concluded that that more is necessarily better (see Salas
simulation reduces the number of air train- et al., 1998). In the future, researchers
ing hours needed to attain criterion profi- need to consider the relative performance
ciency (Lintern, Roscoe, Koonce, & Segal, improvement over time with training. When
2 54 the cambridge handbook of expertise and expert performance
missed) of performance was measured pre- ical advancement was once the primary
and post-training on the virtual trainer and focus of medical simulation, the emphasis
an inanimate model. Eight hours of train- has shifted toward the use of simulation
ing were provided, including video instruc- in clinical learning environments such that
tion on the use of the simulator, instruction domain-specific knowledge and perceptual-
on tracheobronchial anatomy and flexible motor skills can be acquired in unison rather
fiberoptic bronchoscopy techniques, super- than in isolation. Caution is warranted until
vised instruction, and unsupervised practice such systems have been effectively eval-
in the simulator. Trainees speed and time in uated, standards have been derived, and
red out did not improve from pre- to post- measures of performance and methods of
test, but they missed fewer segments and training have been refined and appropri-
made fewer contacts with the bronchial wall. ately validated. Although some evidence of
Post-training performance approached that transfer to the operating room has been
of a control group of skilled surgeons. reported, the mechanisms by which perfor-
The absence of a placebo group and/or mance improves and the nature of instruc-
similarly skilled control group often makes tion and feedback provided have received
it difficult to objectively determine whether only limited attention.
the observed improvements merely reflect
increased task familiarization, or are the
Using Simulation to Train
result of a placebo effect. There have also
Perceptual-Cognitive Skills in Sport
been few attempts to determine whether
simulation-trained skills transfer to the oper- In sport, and in many other activities, train-
ating room. In a recent study, fourth-year ing is typically the sole responsibility of
surgical residents were trained in laparo- the coach (see Section V.II, this volume).
scopic cholecystectomy procedures using Training methods are passed down from
standard methods as well as in the MIST coach to coach, and are usually based on
(Seymour, et al., 2002; see also Gallagher & tradition rather than scientific evidence.
Satava, 2002). Participants were trained in a Although coaches typically invest much
diathermy task; a medical technique used to time in field-based training, many subscribe
generate heat in tissue through electric cur- to the belief that some players are endowed
rent. Training lasted approximately one hour with innate talent. This doctrine discour-
until expert criterion levels were attained. ages coaches from explicitly investing time
Although comparative pre- or post-tasks in the types of training that could be con-
were not employed to assess the absolute sidered intangible (i.e., perceptual-cognitive
change in performance, transfer to the oper- skills such as anticipation and decision mak-
ating room was subjectively assessed by ing). The research on training perceptual-
two attending surgeons. The transfer task cognitive skills using simulation, however,
required participants to perform a real surgi- suggests that such skills are highly amenable
cal gallbladder excision. Simulation trained to practice and instruction. Moreover, the
participants were six minutes (29%) faster in research suggests that such skills are vital
this procedure than residents who received to successful performance (e.g., Helsen &
standard training. In addition, the simulation Starkes, 1999; Ward & Williams, 2003 ).
group made fewer errors than the standard There have been a number of recent reviews
training group (1.19 v 7.3 8, respectively) and on perceptual-cognitive skill training (e.g.,
were much less likely to cause injury to the Abernethy, Wann, & Parks, 1998; Williams &
gallbladder or burn non-target tissue. Ward, 2003 ). We provide a brief summary of
In a time when medical error is under this literature and of the evolution of sports
close scrutiny (see Senate of Surgery report, simulation.
1998; Kohn, Corrigan, & Donaldson, 1999), In an early attempt to create a sim-
these findings are likely to impact the ulated training environment to enhance
future training of medical students. Knee- perceptual-cognitive skill, film-based sim-
bone (2003 ) noted that where technolog- ulation and flash card training were used
2 56 the cambridge handbook of expertise and expert performance
group completed just the pre- and post- ate aspects of the task. At the other end,
test. Only the experimental groups signif- salient task demands have been captured
icantly improved their response time from through video-based simulations, desktop
pre- to post-test, and the greatest improve- simulators, and virtual reality environments.
ment was made when these groups trans- Eye-movement technologies and experi-
ferred to the field. The superior perfor- mental manipulations have been used dur-
mance of the experimental groups beyond ing simulation to help identify processes and
that of a placebo or control indicate that strategies that experts use to maintain supe-
improvement was not a result of task famil- riority over less-skilled counterparts. How-
iarity. Results suggest that the perceptual- ever, differing results have promoted alter-
cognitive skills-training method employed native interpretations. In sport, for instance,
by these authors in a simulated task environ- depending on the situation, soccer experts
ment was effective in improving real-world can exhibit either a high or low number of
performance (see also Williams, Ward, & fixations of short and long duration, respec-
Chapman, 2003 ). tively, but also use an attention strategy
Williams and colleagues (2002) also that makes benefit of peripheral information
examined whether the nature of the deliv- extraction. In tennis, experts tend to use a
ery of the training content would differ- search strategy with fewer fixations of longer
entially affect performance improvement. duration compared to novices. This diversity
Two experimental training groups, explicit- across simulations is likely to indicate that
instruction and guided-discovery, were con- experts flexibly employ effective strategies
trasted. In addition to the perceptual simula- across divergent scenarios to extract mean-
tion training and on-court practice received ingful information. In micro-game simula-
by both experimental groups, the explicit tions of team sports (e.g., 1 v 1 in soccer) and
group received instruction with regard to in individual sports simulations (e.g., tennis),
relationships between important informa- expert superiority appears to lie in the abil-
tion cues and eventual shot placement. ity to pick up postural cues that are pre-
Explicit feedback and opportunities for dictive of future events (e.g., hip/shoulder
practice as well as error detection and rotation in tennis), whereas in more macro-
correction were provided throughout. The game team situations (e.g., 11 v 11 soccer,
guided-discovery group, on the other hand, 5 v 5 basketball), experts are also likely to
was directed to focus on potential areas integrate option selection and pattern recog-
of interest, to work out the relationships nition strategies into their skill repertoire
between key cues and shot outcome, and (Ward & Williams, 2003 ).
given equivalent opportunity for practice. The research on assessing expert-novice
No differences were found between the two differences in medical simulation indicates
experimental groups, although both groups that this medium has been useful in identify-
significantly improved their performance ing superior perceptual-motor skills. Experts
beyond that exhibited by the placebo and typically demonstrate less movement vari-
control groups (also see Smeeton, Williams, ability with fewer positioning errors dur-
Hodges, & Ward, 2005 ). ing task execution. Novice tool manipula-
tion is slower and more variable. In line with
the sports research, performance is aided by
employing an attention strategy that cen-
Summary
ters the point of fixation on the target ear-
lier in the movement, using peripheral visual
Simulation for Performance
information to track and guide the tool. In
Many forms of simulation have been used to contrast, novices use a foveal strategy to aid
study experts. At one end of the spectrum, aiming, focus on the tool throughout move-
static slide presentations and mannequin- ment, and, as they approach the target, alter-
based simulators have been used to recre- nate fixations between the two, proving to
2 58 the cambridge handbook of expertise and expert performance
be a costly strategy in terms of speed and that simulator. There is little evidence to
accuracy. support this viewpoint. The research find-
In aviation, experts superiority during ings suggest that whereas increasing ecolog-
simulated performance is typically accom- ical representativeness with respect to the
panied by a search strategy that uses more action component may increase the size of
fixations of shorter duration (cf. Sport). The the effect, relatively lower-level simulations
suggestion is that experts required less time that capture the salient characteristics of the
per fixation than non-experts to pick up task are far more versatile for measuring and
meaningful information and to monitor and are very effective at improving performance,
control the airspeed indicator. This was par- particularly for specific skill sets that are
ticularly evident during approach as air- perceptual-cognitive in nature.
speed changed, affording greater precision Technological advances in simulation
during touchdown. In general, experts typ- have outpaced research that could con-
ically viewed their instruments more fre- tribute to our understanding of how skilled
quently than novices and flexibly adapted and less-skilled individuals learn or how
their search as the task constraints change. training should be implemented using this
However, expert strategies, including the medium. The cost-effectiveness and effi-
ability to anticipate future consequences of ciency of advanced simulation and virtual
the situation, may be limited to relatively reality systems will remain elusive until
routine operations (i.e., on tasks or under research is conducted that systematically
conditions in which performance is well addresses the content and delivery of the
practiced) and may not extend to unex- training program used in high-fidelity sim-
pected events or high-uncertainty situations. ulation training and comparisons are made
Although much of this research has been to similar training programs in lower-fidelity
conducted on relatively low- to moderate- systems. Research that has addressed the
fidelity simulators, the results suggest that nature of the content and method of delivery
simulation is an extremely useful tool for of training in a simulated task environment
assessing real-world, expert performance indicates that there is much promise in using
under standardized conditions. both explicit and implicit-type training pro-
grams (Williams et al., 2002).
A key question to ask, given the simu-
Simulation for Training
lated nature of the environment, is whether
Researchers examining simulation training training under simulated conditions is actu-
have made only moderate use of expert ally useful in improving real-world perfor-
empirical data as a basis for determining mance. The results on transfer of training
training content and delivery (see Staszewski from a simulated to the actual environment
& Davison, 2000; Williams et al., 2002). Tra- suggest that simulation can be very effective
ditionally, mere exposure and time spent in at improving performance on the criterion
a simulator has been equated with effec- task. Structured programs that have trained
tive (i.e., deliberate) practice, but the doc- individuals for as little as one hour have
trine that simulation is all you need for sometimes shown dramatic improvements
learning to occur has recently been shown in performance. However, as Salas et al.
to be inaccurate. The way in which the (1998) pointed out, more is not necessar-
simulation is implemented during training ily always better, and transfer effective-
is of greater importance than the simula- ness may actually reduce with additional
tion itself. Salas and colleagues (1998) high- training time (Povenmire & Roscoe, 1973 ).
lighted a number of misconceptions about Although performance improvement may
simulation and training that have been be less pronounced after the first few hours
implicitly addressed in this chapter. One of of simulation training (see Taylor et al.,
these suggests that greater financial invest- 2001, 2002), performance improvement is
ment in a simulator facilitates learning on typically a monotonic function of practice.
simulation for performance and training 2 59
When a sustained investment in deliber- psychology of learning and motivation (pp. 15 8).
ate practice is maintained, performance will New York: Academic Press.
likely continue to improve (see Ericsson, Chase, W. G., & Simon, H. A. (1973 ). Perception
2003 ; Ericsson et al., 1993 ). The task for in chess. Cognitive Psychology, 4, 5 5 81.
the scientist working in simulation training is Chi, M. T. H., Glaser, R., & Farr, M. (1980). The
to identify the training content and delivery nature of expertise. Hillsdale, NJ: Erlbaum.
methods that will continue to improve the Colt, H. G., Crawford, S. W., & Galbraith, O.
trainees performance and move them closer (2001). Virtual reality bronchoscopy simula-
to excellence. tion. Chest, 12 0, 13 3 3 13 3 9.
Day, L. J. (1980). Anticipation in junior tennis
players. In J. Groppel & R. Sears (Eds.), Pro-
ceedings of the International Symposium on the
References Effective Teaching of Racquet Sports (pp. 107
116). Champaign, IL: University of Illinois.
Abernethy, B., Thomas, K. T., & Thomas, J. R. Doane, S. M., Sohn, Y. W., & Jodlowski, M. T.
(1993 ). Strategies for improving understanding (2004). Pilot ability to anticipate the conse-
of motor expertise (or mistakes we have made quences of flight actions as a function of exper-
and things we have learned!). In J. L. Starkes tise. Human Factors, 46, 92103 .
& F. Allard (Eds.), Cognitive issues in motor
Endsley, M. (1995 ). Towards a theory of situation
expertise (pp. 3 173 5 6). Amsterdam: Elsevier
awareness. Human Factors, 3 7, 3 264.
Publishing.
Entin, E. B., Serfaty, D., Elliott, L. R., & Schiflett,
Abernethy, B., Wann, J., & Parks, S. (1998). Train-
S. G. (2001). DMT-RNet: An internet-based
ing perceptual motor skills for sport. In B.
infrastructure for distributed multidisciplinary
Elliott (Ed.), Training in sport: Applying sport
investigations of C2 performance. In Proceed-
science (pp. 15 5 ). London: Wiley Publications.
ings of the 6th International Command and Con-
Allard, F., Graham, S., & Paarsalu, M. L. (1980). trol Research and Technology Symposium, June
Perception in sport: Basketball. Journal of Sport 2001. Annapolis, MD.
Psychology, 2 , 1421.
Ericsson, K. A. (1996). The road to excellence:
Allard, F., & Starkes, J. L. (1980). Perception in The acquisition of expert performance in the arts
sport: Volleyball. Journal of Sport Psychology, 2 , and science, sports, and games. Mahwah, NJ:
225 3 . Erlbaum.
Allerton, D. J. (2000). Flight simulation: Past, Ericsson, K. A. (2003 ). How the expert-
present and future. Aeronautical Journal, 104, performance approach differs from traditional
65 1663 . approaches to expertise in sports: In search
Bellenkes, A. H., Wickens, C. D., & Kramer, A. F. of a shared theoretical framework for study-
(1997). Visual scanning and pilot expertise: ing expert performance. In J. Starkes & K. A.
The role of attentional flexibility and mental Ericsson (Eds.), Expert performance in sport:
model development. Aviation, Space, and Envi- Recent advances in research on sport exper-
ronmental Medicine, 68, 5 695 79. tise (pp. 3 71401). Champaign, IL: Human
Brehmer, B., & Dorner, D. (1993 ). Experiments Kinetics.
with computer-simulated microworlds: Escap- Ericsson, K. A., & Kintsch, W. (1995 ). Long term
ing both the narrow straits of the laboratory and working memory. Psychological Review, 102 ,
the deep blue sea of the field study. Computers 211245 .
in Human Behavior, 9, 171184. Ericsson, K. A., Krampe, R. Th., & Tesch-Romer,
Borgeaud, P., & Abernethy, B. (1987). Skilled per- C. (1993 ). The role of deliberate practice in the
ception in volleyball defense. Journal of Sport acquisition of expert performance. Psychologi-
Psychology, 9, 400406. cal Review, 100, 3 63 406.
Charness, N. (1976). Memory for chess positions: Ericsson, K. A., & Smith, J. (1991). Prospects
Resistance to interference. Journal of Experi- and limits of the empirical study of expertise:
mental Psychology: Human Learning and Mem- An introduction. In K. A. Ericsson & J. Smith
ory, 2 , 64165 3 . (Eds.), Toward a general theory of expertise:
Chase, W., & Ericsson, K. A. (1982). Skill and Prospects and limits (pp. 13 8). New York: Cam-
working memory. In G. H. Bower (Ed.), The bridge University Press.
2 60 the cambridge handbook of expertise and expert performance
Farrow, D., Chivers, P., Hardingham, C., & for aerospace applications. In P. A. Hancock
Sacuse, S. (1998). The effect of video-based & M. H. Chignell (Eds.), Intelligent interfaces:
perceptual training on the tennis return of Theory, research, and design. North Holland:
serve. International Journal of Sport Psychology, Elsevier.
2 9, 23 1242. Hoffman, R. R. (1992). The psychology of expertise:
Feltovich, P., Spiro, R., & Coulson, R. (1997). Cognitive research and empirical AI. New York:
Issues of expert flexibility in contexts char- Springer-Verlag.
acterized by complexity and change. In P. J. Hoffman, R. R., & Deffenbacher, K. A. (1993 ).
Feltovich, K. M. Ford, & R. R. Hoffman An analysis of the relations of basic and applied
(Eds.), Expertise in context. Menlo Park, CA: science. Ecological Psychology, 5 , 3 15 3 5 2.
AAAI/MIT Press. Hubbard, J. P., Levitt, E. J., Schumacher, C. F., &
Fischhoff, B. (1989). Eliciting knowledge for ana- Schnabel, T. G. (1965 ). An objective evaluation
lytical representation. IEEE Transactions on Sys- of clinical competence: New techniques used
tems, Man, and Cybernetics, 19, 448461. by the National Board of Medical Examiners.
Gallagher, A. G., & Satava, R. M. (2002). Virtual New England Journal of Medicine, 2 72 , 13 21
reality as a metric for the assessment of laparo- 13 28.
scopic psychomotor skills. Surgical Endoscopy, Jodlowski, M. T., & Doane, S. M. (2003 ). Event
16, 1746175 2. reasoning as a function of working memory
Goldiez, B., Ahmad, A. M., & Hancock, capacity and long term working memory skill.
P. A. (2005 ). Wayfinding and navigation in In Proceedings of the 2 5 th Annual Meeting of
augmented reality. Manuscript submitted for the Cognitive Science Society, July 3 1August 2.
publication. Boston, MA: Cognitive Science Society.
Gray, W. (2002). Simulated task environments: Jodlowski, M. T., Doane, S. M., & Brou, R. J.
The role of high-fidelity simulations, scaled (2003 ). Adaptive expertise during simulated
worlds, synthetic environments, and laboratory flight. In Proceedings of the Human Factors and
tasks in basic and applied cognitive research. Ergonomics Society 47th Annual Meeting, Octo-
Cognitive Science Quarterly, 2 , 205 227. ber 1017. Denver, Colorado: HFES.
de Groot, A. (1978/1946). Thought and choice in Jones, C. M., & Miles, T. R. (1978). Use of advance
chess. The Hague: Mouton. cues in predicting the flight of a lawn tennis
Haskins, M. J. (1965 ). Development of a ball. Journal of Human Movement Studies, 4,
response-recognition training film in tennis. 23 15 .
Perceptual and Motor Skills, 2 1, 207211. Kasarskis, P., Stehwien, J., Hickox, J., Aretz, A., &
Hatano, G, & Ignaki, K. (1986). Two courses of Wickens, C. (2001). Comparison of expert and
expertise. In H. Stevenson, H. Azuma, & K. novice scan behaviors during VFR flight. Paper
Hakuta (Eds.), Child development and education presented at the 11th International Symposium
in Japan. New York: W. H. Freeman. on Aviation Psychology. Columbus OH: The
Hays, R. T., Jacobs, J. W., Prince, C., & Salas, E. Ohio State University.
(1992). Flight simulator training effectiveness: Khatwa, R., & Helmreich, R. (1999). Analysis of
A meta-analysis. Military Psychology, 4, 63 74. critical factors during approach and landing in
Haluck, R. S., Webster, R. W., Snyder, A. J., accidents and normal flight. Flight Safety Digest.
Melkonian, M. G., Mohler, B. J., Dise, M. L., Alexandria, VA: Flight Safety Foundation.
& Lefever, A. (2001). A virtual reality surgi- Kintsch, W. (1988). The use of knowledge in dis-
cal simulator trainer for navigation in laparo- course processing: A construction-integration
scopic Surgery. In J. D. Westwood et al. model. Psychological Review, 95 , 163 182.
(Eds.), Medicine meets virtual reality 2 001: Kneebone, R. (2003 ). Simulation in surgical
Outer space, inner space, virtual space (pp. 171 training: Educational issues and practical appli-
177). Amsterdam: IOS Press. cations. Medical Education, 3 7, 26777.
Helsen, W. F., & Starkes, J. L. (1999). A multidi- Kohn, L. T., Corrigan, J. M., & Donaldson, M.
mensional approach to skilled perception and (1999). To err is human: Building a safer health
performance in sport. Applied Cognitive Psy- system. Washington, DC: Institute of Medicine.
chology, 13 , 127. Law, B., Atkins, M. S., Kirkpatrick, A. E., Lomax,
Higgins, T. J., Chignell, M. H., & Hancock, P. A. A. J., & Mackenzie, C. L. (2004). Eye gaze
(1989). Knowledge-based supervisory control patterns differentiate novice and experts in a
simulation for performance and training 2 61
virtual laparoscopic surgery training environ- of a randomized, double blinded study. Annals
ment. Proceedings of the eye tracking research & of Surgery, 2 3 6, 45 8463 .
applications symposium on eye tracking research Singer, R. N., Cauraugh, J. H., Chen, D., Stein-
& applications (pp. 4148). New York: ACM berg, G. M., Frehlich, S. G., & Wang, L.
Press. (1994). Training mental quickness in begin-
Lintern, G., Roscoe, S. N., Koonce, J. M., & Segal, ning/intermediate tennis players. The Sport Psy-
L. (1990). Transfer of landing skills in beginning chologist, 8, 3 05 3 18.
flight training. Human Factors, 3 2 , 3 193 27. Smeeton, N. J., Williams, A. M., Hodges, N. J.,
Londerlee, B. R. (1967). Effect of training with & Ward, P. (2005 ). The relative effectiveness
motion pictures versus flash cards upon foot- of various instructional approaches in devel-
ball play recognition. Research Quarterly, 3 8, oping anticipation skill. Journal of Experimental
202207. Psychology: Applied, 11, 98110.
McCloy, R., & Stone, R. (2001). Science, Smode, A. F., Hall, E. R., & Meyer, D. E. (1966).
medicine, and the future: Virtual reality in An assessment of research relevant to pilot train-
surgery. British Medical Journal, 3 2 3 , 912 ing (Vol. 11) (Technical Report No. AMRL-TR-
915 . 66-196). Wright Patterson Air Force Base, OH:
McMorris, T., & Hauxwell, B. (1997). Improv- Aerospace Medical Research Laboratory.
ing anticipation of soccer goalkeepers using Sohn, Y. W., & Doane, S. M. (2003 ). Roles of
video observation. In T. Reilly, J. Bangsbo, & M. working memory capacity and long-term work-
Hughes (Eds.), Science and football III (pp. 290 ing memory skill in complex task performance.
294). London: E. & F. N. Spon. Memory & Cognition, 3 1, 45 8466.
Patel, V. L., & Groen, G. (1986). Knowledge- Starkes, J. L., & Ericsson, K. A. (2003 ). Expert
based solution strategies in medical reasoning. performance in sports: Advances in research
Cognitive Science, 10, 91116. on sports expertise. Champaign, IL: Human
Povenmire, H. K., & Roscoe, S. N. (1973 ). Incre- Kinetics.
mental transfer effectiveness of a ground-based Starkes, J. L., & Lindley, S. (1994). Can we hasten
general aviation trainer. Human Factors, 15 , expertise by video simulations? Quest, 46, 211
5 3 45 42. 222.
Roscoe, S. N. (1971). Incremental transfer effec-
Staszewski, J. (1999). Information processing
tiveness. Human Factors, 13 , 5 615 67.
analysis of human land mine detection skill.
Rosser, J. C., Rosser, L. E., & Savalgi, R. S. (1997). In T. Broach, A. C. Dubey, R. E. Dugan, & J.
Skill acquisition and assessment of laparoscopic Harvey, (Eds.), Detection and remediation tech-
surgery. Archives of Surgery, 13 2 , 200 204. nologies for mines and mine-like targets IV. Pro-
Salas, E., Bowers, C. A., & Rhodenizer, L. (1998). ceedings of the Society for Photo-Optical Instru-
It is not how much you have but how you use mentation Engineers 13 th Annual Meeting, SPIE.
it: Toward a rational use of simulation to sup- Vol. 3 710, 766777.
port aviation training. The International Journal Staszewski, J., & Davison, A. (2000). Mine detec-
of Aviation Psychology, 8, 197208. tion training based on expert skill. In A. C.
Salas, E., & Klein, G. (2001). Linking expertise and Dubey, J. F. Harvey, J. T. Broach, & R. E. Dugan
naturalistic decision making (Expertise: Research (Eds.), Detection and remediation technologies for
and applications). Mahwah, NJ: Erlbaum. mines and mine-like targets V. Proceedings of Soci-
Schijven, M., & Jakimowicz, J. (2003 ). Construct ety of Photo-Optical Instrumentation Engineers
validity: Experts and novices performing on the 14th Annual Meeting, SPIE Vol. 403 8, 90101.
Xitact LS5 00 laparoscopy simulator. Surgical Strom, P., Kjellin, A., Hedman, L., Wredmark, T.,
Endoscopy, 17, 803 810. & Fellander-Tsai, L. (2004). Training in tasks
Senate of Surgery (1998). Response to the gen- with different visual-spatial components does
eral medical council determination on the Bristol not improve virtual arthroscopy performance.
Case. London: Senate Paper 5 , The Senate of Surgical Endoscopy, 18, 115 120.
Surgery of Great Britain and Northern Ireland. Taffinder, N., Sutton, C., Fishwick, R. J.,
Seymour, N. E., Gallagher, A. G., Roman, S. A., McManus, I. C., & Darzi, A. (1998). Validation
OBrien, M. K., Bansal, V. K., Anderson, D. K., of virtual reality to teach and assess psycho-
& Satava, R. M. (2002). Virtual reality training motor skills in laparoscopic surgery: Results
improves operating room performance: Results form randomized controlled studies using the
2 62 the cambridge handbook of expertise and expert performance
MIST VR laparoscopic simulator. Studies in Williams, A. M., & Burwitz, L. (1993 ). Advance
Health Technology and Informatics, 5 0, 12413 0. cue utilisation in soccer. In T. Reilly, J. Clarys,
Tayler, M. A., Burwitz, L., & Davids, K. (1994). & A. Stibbe (Eds.), Science and football II
Coaching perceptual strategy in badminton. (pp. 23 9244). London: E. & F. N. Spon.
Journal of Sports Sciences, 12 , 213 . Williams, A. M., & Davids, K. (1998). Visual
Taylor, H. L., Talleur, D. A., Emanuel, T. W., search strategy, selective attention, and exper-
Rantanen, E. M., Bradshaw, G. L., & Phillips, tise in soccer. Research Quarterly for Exercise
S. I. (2001). Incremental training effective- and Sport, 69, 111128.
ness of personal computers used for instru- Williams, A. M., Davids, K., Burwitz, L., &
ment training. Paper presented at the 11th Williams, J. G. (1993 ). Cognitive knowledge
International Symposium on Aviation Psy- and soccer performance. Perceptual and Motor
chology. Columbus, OH: The Ohio State Skills, 76, 5 795 93 .
University. Williams, A. M., Davids, K., Burwitz, L., &
Taylor, H. L., Talleur, D. A., Emanuel, T. W., Williams, J. G. (1994). Visual search strategies
Rantanen, E. M., Bradshaw, G. L., & Phillips, of experienced and inexperienced soccer play-
S. I. (2002). Incremental training effective- ers. Research Quarterly for Exercise and Sport,
ness of personal computers used for instrument 65 , 12713 5 .
training: Basic instruments. Interim Technical Williams, A. M., & Hodges, N. J. (2004) (Eds.).
Report ARL-02 -4/NASA-02 -2 . Moffett Field, Skill acquisition in sport: Research, theory and
CA: NASA Ames Research Center (Contract practice. London: Routledge.
NASA NAG 21282). Williams, A. M., & Ward, P. (2001). Develop-
Thomas, J. R. (2003 ). Meta analysis of motor ing perceptual skill in sport: The need for
expertise. Invited Motor Development Scholar evidence-based practice. In A. Papaioannou,
Address presented at the North American Soci- M. Goudas, & Y. Theodorakis (Eds.), Pro-
ety for the Psychology of Sport and Physi- ceedings of the 10th World Congress of Sport
cal Activity Conference. NASPSPA: Savannah, Psychology: Vol. 3 . In the dawn of the new mil-
GA. lennium (pp. 15 9161). Skiathos, Hellas: Inter-
Torkington, J., Smith, S. G. T., Rees, B. I., & Darzi, national Society of Sport Psychology.
A. (2001). Skill transfer from virtual reality to Williams, A. M., & Ward, P. (2003 ). Percep-
a real laparoscopic task. Surgical Endoscopy, 15 , tual expertise: Development in sport. In J. L.
10761079. Starkes & K. A. Ericsson (Eds.), Expert per-
van Dijk, T. A., & Kintsch, W. (1983 ). Strategies of formance in sport: Advances in research on
discourse comprehension. New York: Academic sport expertise, (pp. 220249). Champaign, IL:
Press. Human Kinetics.
Verner, L., Oleynikov, D., Holtmann, S., Haider, Williams, A. M., Ward, P., Allen, D., & Smee-
H., & Zhukov, L. (2003 ). Measurements of the ton, N. J. (2004). Developing perceptual skill
level of surgical expertise using flight path anal- in tennis using on-court instruction: Perception
ysis from da Vinci Robotic Surgical System. versus perception and action. Journal of Applied
Medicine Meets Virtual Reality, 11, 3 73 3 78. Sport Psychology, 16, 3 5 03 60.
Walls, J., Bertrand, L., Gale, T., & Saunders, N. Williams, A. M., Ward, P., & Chapman C. (2003 ).
(1998). Assessment of upwind dinghy sailing Training perceptual skill in field hockey: Is
performance using a virtual reality dinghy sail- there transfer from the laboratory to the field?
ing simulator. Journal of Science and Medicine Research Quarterly for Exercise and Sport, 74,
in Sport, 1, 6171. 98103 .
Ward, P., & Williams, A. M. (2003 ). Percep- Williams, A. M., Ward, P., Knowles, J., & Smee-
tual and cognitive skill development in soccer: ton, N. (2002). Anticipation skill in a real-
The multidimensional nature of expert perfor- world task: Measurement, training, and trans-
mance. Journal of Sport and Exercise Psychology, fer in tennis. Journal of Experimental Psychology:
2 5 , 93 111. Applied, 8, 25 9270.
Ward, P., Williams, A. M., & Bennett, S. J. (2002). Yates, J. F. (2001). Outsider: Impressions of nat-
Visual search and biological motion perception uralistic decision making. In E. Salas & G. Klein
in tennis. Research Quarterly for Exercise and (Eds.), Linking expertise and naturalistic decision
Sport, 73 , 107112. making (pp. 93 3 ). Mahwah, NJ: Erlbaum.
Part IV
C H A P T E R 15
For most investigations of expertise, conclu- to provide detailed information about the
sions are drawn about the knowledge repre- changes in information processing and per-
sentations and strategies of experts through formance that occur as the skill develops and
observing their performance of tasks in natu- the conditions that optimize acquisition and
ral or artificial settings, analyzing verbal pro- retention of these skills.
tocols that they provide while performing Learning and retention have been stud-
the tasks, and using knowledge-elicitation ied extensively in the laboratory since the
methods. A major component of research earliest days of psychology. For a large part
on expertise involves comparing the per- of the 20th century, much of the efforts
formance of experts to that of novices on of experimental psychologists were centered
specific tasks in the laboratory. Level of on studying animal learning and human
expertise is a subject variable for which the verbal learning (e.g., Leahey, 2003 ). This
prerequisite training and experience of the research resulted in an extensive database
experts has occurred prior to task perfor- and numerous facts and principles con-
mance. Investigations of experts in a vari- cerning acquisition and retention, which
ety of domains with these methods have are summarized in numerous sources (e.g.,
provided invaluable information about the Bower & Hilgard, 1981; Crowder, 1976).
nature of expert performance and knowl- There is also a long history of research
edge, and the ways in which they differ on skill acquisition and retention in labo-
from those of novices. However, because ratory settings (e.g., Bilodeau & Bilodeau,
the acquisition of the experts skills is com- 1969), which is the main focus of this chap-
pleted prior to the investigation, issues con- ter. Laboratory studies of skill acquisition
cerning how this expertise was acquired and offer the advantage of being able to con-
how it is maintained outside of the labora- trol the conditions of training and test-
tory can be investigated only through self- ing so that effects of independent variables
reports. Although self-reports can yield sub- can be isolated and causal relations estab-
stantial data, they are limited in their ability lished. This method allows evaluation of
2 65
2 66 the cambridge handbook of expertise and expert performance
alternative hypotheses and theories concern- simple tasks, performance asymptotes after
ing the acquisition, retention, and transfer of relatively little practice and is retained at
skill. It is possible also to determine factors that level for a long period, implying that
that influence the speed with which a skill a durable skill has been acquired. Ericsson
can be acquired and its generalizability to and Smith (1991) noted, It is clear that the
other tasks and environments. learning mechanisms that mediate increas-
Various methods can be used to eval- ing improvements from repeated practice
uate the nature of skill acquisition and trials must play important roles in the acqui-
retention in the laboratory. Functions relat- sition of expertise (p. 27). But, they regard
ing performance to amount of practice can as shortcomings that the learning mecha-
be measured, allowing implications to be nisms can account only for making the ini-
drawn about the rate of skill acquisition and tial cognitive processes more efficient and
the changes in processing that accompany ultimately automatic (p. 27) and do not
development of skill. Different schedules of take into account the acquisition of new cog-
practice and feedback can be compared to nitive structures, processes that are prereq-
evaluate factors that influence immediate uisites for the unique ability of experts to
performance and learning. Retention tests plan and reason about problem situations
can be conducted after delays of minutes, (p. 28). Although we agree that a major con-
days, weeks, months, or years to establish tribution of the laboratory studies is to show
that the differences evident during acqui- how performance improves with practice,
sition reflect differences in learning and to which is an important part of expert perfor-
establish the durability of the acquired skill. mance, we describe several studies that also
Psychophysiological and brain-imaging tech- reveal development of new cognitive struc-
niques can be employed to assess neuro- tures and changes in strategy.
physiological changes that accompany skill
acquisition.
Perhaps the most widely used technique Phases of Skill Acquisition
is that of transfer designs (e.g., Speelman &
Kirsner, 2001), in which participants practice For virtually any task, performance improves
a task and subsequently are transferred to with practice, with the greatest improve-
another task that shares some features with ment occurring early in training. One issue
the first task. Positive transfer is an indication is whether the improvement in perfor-
that the skills acquired in the practice task mance reflects only quantitative changes
are applicable to the transfer task, whereas (i.e., increased processing efficiency) or qual-
negative transfer implies that their appli- itative changes (i.e., changes in processing
cation cannot be prevented even though it mode). In an early study, Bryan and Harter
interferes with performance. Through the (1899) characterized improvement in per-
use of transfer designs it is possible to deter- formance at telegraphy as the development
mine exactly which processes have been of a hierarchy of habits, reflecting increas-
affected by practice and the nature of the ingly higher-order chunking and automa-
changes that have occurred. tization as the telegrapher became more
A generally accepted rule is that a min- skilled. They stressed the importance of
imum of ten years of deliberate practice is automatization for expert performance, stat-
required to attain expert performance in ing, Only when all the necessary habits,
many domains (Ericsson & Smith, 1991). high and low, have become automatic, does
Although the amount of practice in labo- one rise into the freedom and speed of the
ratory studies of skill acquisition is neces- expert (p. 3 5 7).
sarily considerably less than ten years, lab- The distinction between attention-
oratory studies nevertheless can illuminate demanding controlled processes early in
many aspects of skill acquisition and reten- practice and automatic processes later in
tion. One reason why is that for many practice is evident in many formulations of
training, skill acquisition, and retention 2 67
skill acquisition, including the influential response time (RT) to amount of practice.
one of Schneider and Shiffrin (1977). It is Accounts that suggest changes in modes
customary to distinguish three phases of of processing or strategies seem to imply
skill acquisition, which Fitts (1964) referred that the learning curves will be discon-
to as cognitive, associative, and autonomous tinuous, although such accounts can gen-
and Anderson (1982) called declarative, erate smooth functions. Bryan and Har-
knowledge compilation, and procedural. In ter (1899) reported discontinuities, called
the first phase, task instructions are encoded plateaus, which they attributed to acquisi-
in declarative representations and held tion of lower-order habits prior to higher-
in working memory. General interpretive level habits. However, until recently, the pre-
procedures use these representations to vailing view has been that the function is
generate behavior appropriate to the task. continuous. Newell and Rosenbloom (1981)
In the transitional phase, procedures specific concluded that across a variety of tasks, the
to the task or skill are acquired that no reduction in RT with practice can be cap-
longer require the interpretive procedures. tured by a power law, as first proposed by
In the final phase, further progression of Snoddy (1926):
skilled performance is achieved through a
gradual strengthening of the procedures and RT = A + B N ,
tuning of the conditions that will trigger
where N is the number of practice trials, B
them. In this phase, some skills can become
is performance time on the first trial, is
automatized.
the learning rate, and A is the asymptotic
Whereas Andersons (1982) account of
RT after learning has been completed.
skill acquisition, like most others, attributes
The power law of practice has become
it to development of procedures, or asso-
widely accepted as a benchmark that must
ciations, that become strengthened with
be generated by any theory of skill acquisi-
practice, an alternative view is that skill
tion (e.g., Logan, 1988). However, it recently
acquisition reflects a change in processes.
has been challenged as being applicable only
Logan (1988) proposed an instance the-
to averaged acquisition functions. Heath-
ory of automatization, according to which
cote, Brown, and Mewhort (2000) fit power
execution of two processes an algorithm
and exponential functions to the data from
based on task instructions and retrieval from
individual participants for 40 data sets. They
memory of previously encoded instances
found that the power function, for which
occurs in parallel. An assigned task is per-
the learning rate is a hyperbolically decreas-
formed initially using an appropriate algo-
ing function of practice, did not fit the
rithm. With practice, specific instances of
individual-participant data as well as the
stimuli and their responses are encoded into
exponential function, for which the learning
memory, and performance can instead be
rate is constant at all levels of practice. Con-
based on retrieval of a prior instance. At first,
sequently, Heathcote and colleagues pro-
retrieval is slow, but with practice under con-
posed a new exponential law of practice:
sistent conditions, it becomes much faster,
resulting in a mix of trials on which per- RT = A + Be N,
formance is algorithm based and ones on
which it is retrieval based. With sufficient where is the rate parameter.
experience, the retrieval process comes to Whereas both the power and exponen-
be used on all trials. Thus, according to the tial laws assume that the acquisition func-
instance theory, increasing automatization tions are continuous, several authors have
with practice is a consequence of a grad- reported evidence that for some tasks, such
ual shift from algorithm-based to memory- as mental arithmetic problems, the func-
based performance. tions for individual participants show abrupt
One way to evaluate models of skill acqui- changes (e.g., Haider & Frensch, 2002;
sition is to examine the functions relating Rickard, 2004). Much of this research has
2 68 the cambridge handbook of expertise and expert performance
been conducted within the framework of ceptual learning is important not only for
Logans (1988) instance theory and has been basic theory about skill acquisition but also
interpreted as indicating that the change for the acquisition and training of real-world
from an algorithmic process to a retrieval skills that have substantial perceptual com-
process is discrete, rather than being a grad- ponents, such as identification of abnormal
ual shift in dominance of the parallel exe- features in X-ray images (Sowden, Davies,
cution of algorithm and search strategies as & Roling, 2000), wine tasting (Brochet &
proposed by Logan (Rickard, 2004). More Dubourdieu, 2001), and discriminating the
generally, Delaney, Reder, Staszewski, and sex of baby chicks (Biederman & Shiffrar,
Ritter (1998) concluded that acquisition of 1987).
skill and expertise is characterized by multi- Goldstone, Schyns, and Medin (1997)
ple strategy shifts that produce discontinu- specify five mechanisms involved in percep-
ities in individual learning curves. They pro- tual learning: (1) Attention weighting con-
vided evidence that individual improvement cerns shifts of attention from less-relevant
in solution time for mental arithmetic prob- to more-relevant dimensions. This is accom-
lems with practice is characterized better by plished in part by (2) detector creation, for
separate power functions for each specific which functional units are established, each
strategy than by a single power function for of which respond selectively to a specific
the task. type of input, and (3 ) dimensionalization, or
the creation of ordered detector sets that
represent objects by their distinct dimen-
Basic Information-Processing Skills sions. These processes act to enable effi-
cient selective attention to specific stimulus
dimensions. (4) Unitization involves acqui-
Research on the acquisition and retention of
sition of higher-level functional units that
basic information-processing skills has been
can be activated by complex configurations
conducted using a variety of tasks. Although
of features, thus allowing stimuli to be pro-
there is not a clean separation between
cessed as a whole. The final mechanism of
skills involving perception, response selec-
perceptual learning identified by Goldstone
tion, and motor control, it is convenient to
et al. is (5 ) contingency detection: The contin-
organize studies around this distinction. (See
gencies between parts of stimuli are learned
also Rosenbaum, Augustyn, Cohen & Jax,
to allow more efficient extraction of infor-
Chapter 29.)
mation by, for example, changing scanning
patterns.
Perceptual Skill
Ahissar (1999) notes, The extent of adult
improvement in not only complex but feature identification
also simple perceptual tasks is remarkable A major aspect of perceptual skill is learning
(p. 124). Perceptual learning has been stud- to identify features that distinguish alterna-
ied since the 1800s, with much of the tive stimuli or classes of stimuli. For exam-
work in the second half of the 20th cen- ple, Sowden et al. (2000) note, regard-
tury conducted from the ecological per- ing perception of medical X-ray images,
spective (Gibson & Pick, 2000). Accord- The expert film reader apparently perceives
ing to this perspective, perceptual learning features present in X-ray images that go
involves the individual becoming tuned to unnoticed by the novice (p. 3 79). Numer-
pick up information afforded by the envi- ous studies have shown that training that
ronment. Interest in perceptual learning has emphasizes distinctive features is highly
increased considerably in the last decade beneficial. Gibson (1947) reported an exper-
(e.g., Fahle & Poggio, 2002), with an empha- iment in which cadets received 3 0 hours of
sis on examining the underlying cognitive training for distinguishing among slides of
and neural mechanisms. Understanding per- 40 different aircraft, using instructions that
training, skill acquisition, and retention 2 69
emphasized a set of distinctive features or the target items and another if it does not.
the total form of each plane. Cadets who In visual search tasks, the displays contain
received the distinctive-feature instructions one or more stimuli, and the participant is
performed better on a subsequent recog- to indicate whether an item held in memory
nition test than did those who received is in the display (or, to identify which of two
the total-form instructions. Biederman and possible targets is in the display). For hybrid
Shiffrar (1987) found that novices could memory-visual search tasks, the sizes of both
be trained to perform at a similar level the memory and the display sets are varied.
to experts at classifying chicks as male or Schneider and Shiffrin (1977) and Shiffrin
female, a skill that typically takes years to and Schneider (1977) established that a crit-
acquire, by using instructions and diagrams ical factor influencing the benefit of prac-
that emphasized the critical features for dif- tice in search tasks is whether the target
ferentiating male and female chicks. items in the memory set on one trial are
Although the tasks above required never distractors on other trials (consistent
relatively complex discriminations, many mapping, CM), or whether the same items
simple perceptual tasks such as grating can be targets on some trials and distractors
waveform discrimination and motion dis- on others (varied mapping, VM) (see Fig-
crimination show substantial learning as well ure 15 .1). Initially, RT is slow, and it increases
(Karni & Bertini, 1997). Transfer techniques substantially as a linear function of the tar-
have been used to determine the conditions get set size. Practice with CM results in
to which each skill generalizes and to provide a large decrease in RT and elimination of
evidence about the neuronal mechanisms the set-size effect. In contrast, practice with
involved. Karni and Bertini note that an VM produces little improvement in per-
important characteristic of perceptual learn- formance. These results suggest that auto-
ing is a lack of broad transfer. Karni (1996) maticity develops only when the mapping
has proposed a model for which the cen- of stimuli to target and nontarget categories
tral idea is that the acquisition of a skill is remains consistent. Shiffrin and Schneider
at the earliest level of the stream for pro- concluded that automatic processes are fast
cessing the sensory information in which and operate in parallel, whereas controlled
the relevant stimulus parameter can be rep- processes are slow and operate serially.
resented. Karni and Bertini conclude that Schneider and Chein (2003 ) list five
top-down mechanisms control perceptual additional phenomena for search behavior
learning because repeated exposure to a that reflect differences in controlled and
stimulus is not sufficient for learning to automatic processing. (1) Controlled search
occur, but they note that perceptual learn- requires considerable effort, whereas auto-
ing has often been reported in the absence of matic search does not. (2) Controlled pro-
explicit performance feedback (e.g., Fahle & cessing is more sensitive to stressors, such
Edelman, 1993 ). as fatigue, than is automatic processing.
(3 ) Controlled processes can be modified
easily, but automatic processes cannot.
automaticity and unitization (4) Controlled processing results in explicit
Research on skill acquisition and transfer has learning of task characteristics, whereas
been conducted using search tasks for which automatic processing does not. (5 ) Auto-
participants must indicate whether a probe matic attraction of attention to a stimulus
item is a member of a target set. For a mem- is determined by the priority of the stimu-
ory search task, the participant receives a lus alone and not the context in which the
memory set of one to four target items (e.g., stimulus occurs.
letters) and then one or more displays con- Letters and digits are already highly uni-
sisting of probe items that may or may not tized when used as stimuli in experiments.
include the target. One response key is to To examine the acquisition of unitized rep-
be pressed if the probe item matches any of resentations, Shiffrin and Lightfoot (1997)
2 70 the cambridge handbook of expertise and expert performance
Figure 15.1. Illustration of a search task with consistent or varied mapping. With
consistent mapping (left column), stimuli in the memory set for one trial are
never distractors on other trials. With varied mapping (right column), stimuli in
the memory set for one trial can be distractors on other trials (as illustrated by
the letter J changing from a target on trial N to a distractor on trial N + 1).
used a visual search task with novel charac- slope ratios suggestive of controlled search.
ters composed of line segments for which Thus, no automatic attraction of attention
no single feature could be used to identify developed under CM conditions for these
the target (see Figure 15 .2), comparing learn- novel stimuli, in contrast to the results
ing for CM and VM tasks. At the begin- obtained with familiar letter and digit stim-
ning of training the slopes of the set-size uli. Shiffrin and Lightfoot concluded that
functions for both tasks averaged approxi- the changes with practice were due to a uni-
mately 100 ms for positive (target present) tization process that allowed a holistic rep-
responses and 200 ms for negative (target resentation for the stimuli to develop.
absent) responses. The VM task received as
much benefit of practice as the CM task,
Response-Selection Skill
and both tasks continued to show relatively
high slope values and 2:1 negative-to-positive Response selection refers to processes
involved in determining which response to
make to a stimulus. The phenomena that can
be attributed primarily to response-selection
processes are stimulus-response compatibil-
ity (SRC) effects (Sanders, 1998), which
are differences in performance as a func-
tion of the mapping of individual stim-
uli to responses and the overall relation
Figure 15.2 . Examples of the novel, between the stimulus and response sets
conjunctively defined stimuli used in Shiffrin (e.g., whether physical stimulus locations
and Lightfoots (1997) visual search study. are mapped to keypresses or vocal location
training, skill acquisition, and retention 2 71
Figure 15.3. Left panel: Illustration of Displays A and C, and the response arrangement, used in Fitts
and Seegers (195 3 ) study. Right panel: Mean reaction time as a function of practice session and
display type.
names). In the simplest form of SRC study, compatible mapping but not for the incom-
left and right stimuli are mapped to left patible mapping. In network models, the
and right responses. RT is shorter with the controlled processing is represented as short-
mapping of right stimulus to right response term stimulus-response associations defined
and left stimulus to left response than with by task instructions, and the automatic pro-
the opposite mapping. This element-level cessing as long-term stimulus-response asso-
SRC effect is larger when the stimulus and ciations that are overlearned through years
response sets are both visuospatial or both of experience (Zorzi & Umilta, 1995 ).
verbal than when one set is visuospatial and
the other verbal (Proctor & Wang, 1997), due
to the higher compatibility of physical loca- practice with various tasks and mappings
tions with manual responses and of location- Most accounts of skill acquisition imply that
words with vocal responses. SRC effects should disappear with practice,
Most explanations attribute the bene- but numerous studies have shown that they
fit for the compatible mapping at least in do not. Fitts and Seeger (195 3 ) had partic-
part to intentional, controlled processing ipants respond to eight possible stimuli by
being faster for that mapping than for the moving a stylus to one of eight response loca-
incompatible mapping. According to some tions, arranged in a circle. Within each of 26
accounts, the compatible mapping also ben- sessions, three displays were used that dif-
efits from automatic activation of the cor- fered in their compatibility with the circular
responding response (e.g., Kornblum & Lee, response array. In the first session, RT was
1995 ), which is the correct response for the 45 0 ms longer for the least compatible
2 72 the cambridge handbook of expertise and expert performance
Display C than for the most compatible with two new members added to each cat-
Display A (see Figure 15 .3 ). RT decreased egory. Perfect transfer was obtained for the
across sessions, with the asymptotic differ- new category members, indicating that par-
ence between Displays C and A stabiliz- ticipants had learned category-to-response
ing at about 80 ms between sessions five associations and not specific stimulus-to-
and ten. response associations. Also, a change in the
Dutta and Proctor (1992) had partici- hand used for responding resulted in com-
pants practice 1,200 trials with either a com- plete transfer, indicating that participants
patible or incompatible mapping of left- were not learning to make specific motor
right stimuli to left-right response keys. responses to the categories.
The SRC effect was 72 ms initially and
decreased to 46 ms at the end of prac-
tice. Dutta and Proctor also showed that practice and the simon task
two other types of SRC effects, one for When stimulus location is irrelevant to the
orthogonal stimulus-response arrangements task, performance is better when stimu-
and another for two-dimensional symbolic lus and response locations correspond than
stimuli, remained present across the same when they do not (the Simon effect; Simon,
amounts of practice. 1990). For example, when instructed to
Proctor and Dutta (1993 ) examined respond to a green stimulus with a left
whether the benefit of practice in two- keypress and a red stimulus with a right
choice tasks arises from participants learn- keypress, responses to the green stimulus
ing associations between stimulus-response are faster when it occurs in a left loca-
locations or stimuli and effectors. Partici- tion than when it occurs in a right loca-
pants practiced with a compatible or incom- tion, and vice versa for responses to the
patible spatial mapping over three days, half red stimulus. The Simon effect is typically
with their hands in the natural adjacent posi- attributed to automatic activation of the cor-
tions and half with them crossed so that the responding response produced via the long-
right hand operated the left key and the left term stimulus-response associations. As for
hand the right key. When transferred to one SRC proper, the Simon effect is reduced
of the other mapping/placement conditions, but persists with practice. Simon, Craft, and
positive transfer was evident only if the spa- Webster (1973 ) found that the Simon effect
tial mapping was the same in the transfer for high or low pitch tones presented in
session as in the practice sessions and not if the left or right ear decreased from 60 ms
hand position was the same but spatial map- initially to 3 5 ms after 1,080 trials. Sim-
ping different. With an incompatible spatial ilarly, Prinz, Aschersleben, Hommel, and
mapping, the practice benefit was evident Vogt (1995 ) had a single person perform
even when participants switched periodi- 210 trials of an auditory Simon task in each
cally between crossed and uncrossed hand of 3 0 sessions. The Simon effect decreased
placements. These results imply that the from 5 0 ms in the first three sessions to 20
improvement with practice in two-choice ms over the last 20 sessions.
spatial tasks primarily involves faster selec- Although the Simon effect is not elim-
tion of a location code. inated with practice, it can be eliminated
Pashler and Baylis (1991) obtained sim- or reversed by prior practice with an
ilar benefits of practice on the speed of incompatible spatial mapping. Proctor
response selection for a task in which par- and Lu (1999) had participants perform a
ticipants made keypresses with three fin- two-choice visual SRC task with an incom-
gers of one hand in response to stimuli from patible mapping for 900 trials. When a
three categories (two letters, digits, and non- Simon task was performed on the next day,
alphanumeric symbols). RT decreased by the Simon effect reversed, yielding better
15 0 ms over 75 0 practice trials. Each par- performance on noncorresponding trials
ticipant then performed 5 0 additional trials, than on corresponding trials. Tagliabue,
training, skill acquisition, and retention 2 73
Figure 15.4. Data from Willingham et al.s (1989) study showing mean of
median reaction time as a function of practice block for groups with full, some,
or no knowledge of the repeating sequence. From R.W. Proctor and A. Dutta,
Skill acquisition and human performance, p. 173 . Copyright 1995 , Sage
Publications. Reprinted with permission.
(RT < 100 ms) were removed, performance sion instructions, the proportion of the train-
for the full-knowledge group was not signif- ing sequence that was generated was above
icantly different from that for the some- and chance for both RSI groups. With exclusion
no-knowledge groups. This finding implies instructions, the proportion of chunks gener-
that sequence learning did not depend on ated by participants in the 25 0-ms RSI group
explicit awareness, although awareness led was less than that under inclusion instruc-
to an anticipation strategy. tions and did not differ from chance, indi-
Researchers have questioned whether the cating that their learning was explicit. In
generation task is an adequate measure of contrast, the proportion generated by par-
explicit-implicit knowledge, noting that it ticipants in the 0-ms RSI group did not dif-
suffers from several methodological prob- fer from that under inclusion instructions.
lems (e.g., Perruchet & Amorim, 1992). Consequently, Destrebecqz and Cleeremans
To circumvent the problems of this and concluded that learning of the sequence was
other measures of explicit-implicit learn- implicit when there was not sufficient time
ing, Destrebecqz and Cleeremans (2001) between trials to prepare for the next event.
adapted the process-dissociation procedure However, Wilkinson and Shanks (2004)
originally devised by Jacoby (1991) to study reported three experiments using only a
implicit memory. This procedure uses one 0-ms RSI in which they were unable
generation test in which participants are to to replicate Destrebecqz and Cleeremanss
attempt to generate what they have been (2001) findings. In all experiments, partic-
exposed to previously (inclusion instruc- ipants were able to avoid generating the
tions) and another in which they are to sequence presented in the practice session
attempt to avoid doing so (exclusion instruc- under exclusion instructions. Wilkinson and
tions). After practice on a serial RT task Shanks suggest that a lack of power may have
using a repeating 12-element sequence, with prevented this difference from being signifi-
a constant response-stimulus interval (RSI) cant in Destrebecqz and Cleeremanss study.
of 0 or 25 0 ms, participants were asked to Regardless of whether sequence learning is
generate the sequence of trials. With inclu- truly implicit, it is clear that people are able
training, skill acquisition, and retention 2 75
to extract the sequential structure of events mapping for which the sequence of stimulus
to which they are exposed without necessar- locations was the same.
ily intending to do so. Given that the learned representation
Another issue in sequence learning does not seem to be perceptual or specific
concerns whether the representation that is to particular effectors, what is its nature?
learned is perceptual, motor, or stimulus- Willingham, Wells, Farrell, and Stemwedel
response associations. Several studies (2000) proposed that a sequence of response
obtained results suggesting that the learning locations is learned. They tested this propo-
is not purely motor. Cohen, Ivry, and Keele sition by configuring the four stimulus loca-
(1990) trained participants on a serial RT tions in a lopsided diamond arrangement
task with repeating sequences for which the (see Figure 15 .5 ) that allowed them to have a
responses were keypresses of one of three left-to-right order along the horizontal axis.
fingers from a single hand. The sequence The arrangement was mapped compatibly
learning transfered well to a situation in to one of two keyboards on which partic-
which a single finger was used to press ipants responded using the index finger of
each of the three keys. Willingham et al. their preferred hand. One keyboard was con-
(1989) dissociated stimulus and response figured in a diamond shape and the other
sequences by conducting a four-choice task had the keys linearly arranged in a row.
in which one of four keypresses was made Participants who switched from one key-
to a stimulus in one of four positions, but board in the training phase to the other in
the relevant stimulus dimension was color. the transfer phase showed no benefit from
Learning was evident when the responses the repeating sequence. In another experi-
(and colors) followed a repeating sequence, ment, all participants used horizontal stim-
but this learning did not transfer to a task ulus and response arrangements and per-
for which stimulus location was relevant. formed with the hands in normal positions
Willingham et al. (1989) also found no in the transfer phase. Participants who per-
evidence of learning when stimulus location formed with their hands crossed during the
(which was irrelevant) followed a repeat- training phase, such that the left hand oper-
ing sequence, suggesting that the sequence ated the right keys and the right hand the left
learning was not perceptual. Other stud- keys, showed no cost in the transfer phase
ies have suggested a perceptual basis to relative to participants who performed only
sequence learning (e.g., Howard, Mutter, with the hands uncrossed, as long as the
& Howard, 1992), but Willingham (1999) sequence of response locations remained
noted that participants in some of those unaltered. In contrast, when the sequence
studies showed substantial explicit knowl- of finger movements was unaltered but the
edge and that artifacts associated with eye sequence of response locations changed, no
movements could have been responsible transfer was evident.
for the learning in others. He reported
three experiments that provide strong evi-
dence against a perceptual basis for sequence
learning: Mere observation of a repeat-
ing stimulus sequence produced no bene-
fit when participants who showed explicit
knowledge were removed; transfer was
robust from a training condition in which the
stimuli were the digits 14 (mapped left-to-
right to the responses) to one in which the
stimuli were spatial locations; a less compat-
ible spatial mapping in training (press the Figure 15.5. Illustration of the stimulus (top)
key to the right of the stimulus location) and response (bottom) configurations used in
showed no transfer to a spatially compatible Willingham et al.s (2000, Experiment 1) study.
2 76 the cambridge handbook of expertise and expert performance
1998, and Lien & Proctor, 2002). In a typ- can be divided in different amounts across
ical PRP study, participants perform two two tasks, but at reduced efficiency com-
tasks (T1 and T2), each of which requires pared to when the entire capacity is devoted
a response to a stimulus. The stimulus (S1) to one task alone (e.g., Tombu & Jolicur,
for T1 is presented, followed after a variable 2003 ). The second challenge, which is more
interval (stimulus onset asynchrony; SOA) germane to present concerns, is whether
by the stimulus (S2) for T2. Instructions are the bottleneck is a structural limitation of
to respond as rapidly as possible for both the cognitive architecture, as both the bot-
tasks, sometimes with an emphasis on mak- tleneck and resource accounts assume, or
ing the response to T1 (R1) prior to that whether it reflects a strategy adopted by par-
for T2 (R2). The PRP effect is that RT for ticipants to satisfy the task demands (e.g.,
T2 (RT2) is longer at short SOAs than at Schumacher et al., 2001).
long SOAs, with the function relating RT2 If the PRP effect is due to a structural
to SOA often having a slope of 1 until a limitation, the effect should still be evi-
critical SOA, after which it asymptotes. dent after extended practice. Several stud-
Most accounts attribute the PRP effect ies have found the PRP effect to persist
to a response-selection bottleneck (Pashler across extended practice (e.g., Gottsdanker
& Johnston, 1998): Stimulus identification & Stelmach, 1971). However, despite the
for T1 and T2 can occur in parallel, as can tendency for the PRP effect to persist, sev-
response programming and execution, but eral recent studies have reported conditions
response selection can be performed only for under which the effect is small or possibly
one task at a time. Consequently, selection of even absent after practice (e.g., Hazeltine,
R2 cannot begin until selection of R1 is com- Teague, & Ivry, 2002; Levy & Pashler,
pleted. According to the response-selection 2001; Schumacher et al., 2001). Van Selst,
bottleneck model, variables that increase the Ruthruff, and Johnston (1999) noted that
time to identify S2 should interact underad- most of the studies showing little reduction
ditively with SOA (i.e., their effects should in the PRP effect with practice used man-
be smaller at short SOAs than long SOAs). ual responses for both tasks, which creates
This underadditive interaction is predicted output interference between the two tasks.
because in the easiest identification condi- Consequently, they conducted an exper-
tions there is slack at short SOAs, dur- iment in which there was little overlap
ing which the system is waiting to begin between the stimuli and responses for the
response selection for T2 after having iden- tasks: T1 used four tone pitches, with the
tified S2. Consequently, much of the addi- two lowest mapped to the vocal response
tional time for identification of S2 in the low and the two highest to high, and
difficult condition can be absorbed by the T2 used visually presented numbers and let-
slack without increasing RT2. In contrast, ters mapped to keypresses, made with the
variables that have their influence on selec- four fingers of the right hand. Six partici-
tion of R2 should have additive effects with pants performed the tasks for 3 6 sessions of
SOA because the time for these processes 400 trials each. A PRP effect of 3 5 3 ms was
cannot be absorbed into the slack. Results evident in the first session, and this effect
have generally conformed to these and other was reduced to 40 ms by practice but elim-
predictions of the bottleneck model (Pashler inated entirely for only one subject. Van
& Johnston, 1998). Selst et al. concluded that several aspects
Recently, the response-selection bottle- of their results indicated that a bottleneck
neck model has been challenged on two was still present after extensive practice, and
fronts. One challenge is whether response showed that the substantial decrease in the
selection is restricted to only one task at a PRP effect could be attributed entirely to a
time, as the bottleneck model assumes, or reduction in RT1.
whether response selection is better charac- Ruthruff, Johnston, and Van Selst (2001)
terized as a limited-capacity resource that and Ruthruff, Johnston, Van Selst, Whitsell,
2 78 the cambridge handbook of expertise and expert performance
and Remington (2003 ) presented additional tion is whether training should focus on the
evidence that the reduction of the PRP individual subcomponents first (part-whole
effect with practice is primarily due to a training) or on the entire task (whole-task
decrease in the duration of the bottleneck training). The logic behind part-whole train-
stage for T1. Ruthruff et al. (2001) tested ing is that a higher level of skill can be
the five participants from Van Selst et al.s attained if participants are able to practice
(1999) study for whom the PRP effect had and master the individual components prior
not been eliminated, changing T1 so that to integrating them in the whole task con-
it required same-different responses for text. However, because participants are not
pairs of tones instead of pitch judgments. exposed to the integrated task, the skills that
The idea was that introducing a new T1 they acquire while practicing the subcom-
would slow R1, increasing the magnitude of ponents may not transfer to it. Briggs and
the PRP effect. Consistent with this pre- Naylor (1962) noted that the type of train-
diction, the initial PRP effect was 194 ms, ing used should depend on the nature of
a value considerably larger than the 5 0-ms the task complexity and organization. Part-
effect obtained for those participants in the whole training is most beneficial when the
last session of the previous study. In contrast, complexity of the whole task is high but the
when switched to a new T2, pressing one of organization is low. That is, when the sub-
two response keys to a visual letter X or Y, components are not highly integrated, then
the PRP effect was only 98 ms. part-whole training allows the participant to
Ruthruff et al. (2003 ) conducted further focus on mastering the skill for each subcom-
investigations of the participant from Van ponent without being distracted by other
Selst et al.s (1999) study for whom the PRP subcomponents.
effect had been eliminated. Their hypothesis Frederiksen and White (1989) provided
was that the bottleneck was latent for that one notable demonstration of a benefit
participant, that is, the PRP effect was absent for part-whole training using the Space
because the operations of the bottleneck Fortress game. In that game, participants
stage for T1 were completed prior to that operate a spaceship with the goal of destroy-
stage being needed for T2. To support this ing a fortress by shooting missiles at it
hypothesis they showed that a PRP effect while protecting the spaceship from dan-
was evident for that participant when neg- ger (e.g., avoiding shells fired at the space-
ative SOAs were introduced for which T2 ship from the fortress and navigating the
preceded T1 by up to 216 ms and when a new ship to avoid mines). Frederiksen and
T1 (judging whether the third tone in a rapid White performed analyses regarding the
series of three was higher or lower in pitch skills/knowledge needed to perform certain
than the first) was paired with the old T2. individual components of the game and had
On the whole, the experiments of Van Selst one group of participants engage in part-
et al. (1999) and Ruthruff et al. (2001, 2003 ) whole training (practice with these individ-
suggest that the bottleneck is not eliminated ual components alone for three days and
by practice but only hidden. More recently, with the whole game on the fourth day),
though, Ruthruff, Van Selst, Johnston, and and another group engage in whole-task
Remington (in press) have found evidence training (practice with the whole game dur-
that a few participants show evidence of ing the four days). After the practice ses-
bypassing the bottleneck after extensive sion, both groups played the whole game
practice under specific circumstances. for nine successive games, and mean game
scores were obtained for each group (see
Figure 15 .6). For the first game, participants
Part-Whole Transfer
who received whole-task training scored a
Many complex tasks can be decomposed little over 5 00 more points than participants
into distinct subtasks that are integrated in the part-whole training group. However,
when performing the whole task. One ques- beginning with Game 2, the participants in
training, skill acquisition, and retention 2 79
Figure 15.6. Data from Frederiksen and Whites (1989) study showing mean
game score as a function of part-whole training or whole-game training across
successive game blocks. From R. W. Proctor and A. Dutta, Skill acquisition and
human performance, p. 274. Copyright 1995 , Sage Publications, Reprinted with
permission.
the part-whole group performed much bet- within a block. When presented with both
ter than those in the whole-task training types of discriminations, participants who
group. By Game 8, the part-whole training first practiced difficult discriminations out-
group scored about 1,3 00 more points that performed those who first practiced easy
the whole-task training group. discriminations. In subsequent experiments,
The benefit of part-whole training sug- Doane et al. showed that this superior per-
gests that higher-level skills can be attained formance for the difficult-first group was
with training of the easier component tasks due to both stimulus-specific knowledge and
first. However, several studies have shown strategic knowledge.
that participants who engage in learning Although Doane et al. (1996) found a
more difficult discriminations first show bet- benefit for practicing the more difficult dis-
ter transfer than those who engage in learn- criminations first, Clawson, Healy, Ericsson,
ing easier discriminations first to subsequent and Bourne (2001) showed that if the initial
tasks containing both easy and difficult items subset of stimuli is very difficult to learn, it
(e.g., Doane, Alderton, Sohn, & Pellegrino, will not yield superior performance on the
1996). Doane et al. had participants deter- full set. Clawson et al. gave participants 12
mine whether two polygons were identi- Morse codes to receive and translate into
cal or not. The complexity of the polygons their corresponding letters. Six of the code-
was defined by the number of vertices from letter pairs were difficult items and six were
which they were made, and the difficulty easy. Clawson et al. evaluated the difference
of the task was determined by the degree in performance accuracy for all items dur-
of similarity of the two polygons in the ing acquisition immediately after the part-
pair (see Figure 15 .7). Doane et al. (1996, stimulus training and after a retention inter-
Experiment 1) had one group of partici- val of four weeks. In contrast to the findings
pants practice first with easy discriminations of Doane et al., training with the difficult
and another group with difficult discrimi- stimuli first did not result in better perfor-
nations. After practice, the groups received mance with the full set of stimuli. Further-
both the easy and difficult discriminations more, after the four-week delay, the group
2 80 the cambridge handbook of expertise and expert performance
that initially received the difficult stimuli answers are table-related errors that would
showed a decrease in accuracy, whereas the be correct for another problem that shares
group that initially received the easy stim- an operand with the current problem (e.g.,
uli did not. Clawson et al. attributed this Campbell & Graham, 1985 ). Campbell
decrement in performance for the difficult- (1987) showed that such errors can be
first training group to the fact that the Morse primed by prior problems. He had partici-
code reception task was extremely difficult, pants say the answers to single-digit multi-
preventing participants from mastering the plication problems (e.g., 7 9) as quickly
task during training. as possible. Incorrect retrieval of a product
(e.g., 7 9 = 5 6) was more likely when
the answer had been the correct product for
Solving Arithmetic Problems
a recent problem (7 8), and this prim-
Basic arithmetic problems consist of an oper- ing of incorrect responses was restricted pri-
ator (add, subtract, multiply, divide) and marily to the table-related errors that would
operands (the numbers to which the opera- be most frequent without the prior prim-
tor is applied). Skilled performance on basic ing event. Retrieval times for correct answers
arithmetic tasks is not due primarily to more were also longer if strong false associates
efficient application of the operator to the were activated by way of other problems.
operands, but instead to retrieval of spe- Campbells study not only provided strong
cific facts (e.g., 2 + 2 = 4) from mem- evidence of retrieval of answers via associa-
ory (Ashcraft, 1992). Under speed stress, tions, it also showed no evidence for a con-
for example, 70% 90% of the incorrect tribution of general algorithmic procedures
training, skill acquisition, and retention 2 81
based on the operator. Specifically, partici- Rickard and Bourne interpreted as indicating
pants were first pretested on all problems, that modality-independent representations
then trained on half of the problems for sev- of the type hypothesized by the identical-
eral sessions, and finally retested on all prob- elements model and modality-specific per-
lems. Only those problems that were prac- ceptual representations were involved. One
ticed showed a benefit on the posttest, with possibility they suggested is that number-
performance being worse for the unprac- fact retrieval is mediated by an abstract
ticed problems on the posttest than it had representation in most cases, but practice
been on the pretest. on a limited number of problems in the
Rickard, Healy, and Bourne (1994) had same input format may allow a more direct
participants practice multiplication and divi- retrieval route to develop.
sion problems for 40 trial blocks, each of When unfamiliar pseudoarithmetic rules
which included the 3 6 problems in the are learned in the laboratory, participants
practice set. Immediately afterward, a test presumably cannot directly retrieve the
was given that included those problems as answers but must perform algorithmic pro-
well as others that differed in the order of cedures. Sohn and Carlson (1998) pro-
the operands and the format in which they posed a procedural framework hypothesis
were presented. Performance on the test was according to which goals to apply opera-
good if the operator and operands were the tors can be instantiated abstractly to provide
same as on a practiced problem, even if the a framework for processing the operands.
order of the operands was changed, and was The hypothesis thus predicts that perfor-
considerably poorer if any element other mance will be best when the operator is
than operand order was changed. Rickard encoded first. To test this hypothesis, Sohn
et al. interpreted their results as support- and Carlson had participants apply four
ing an identical-elements model, accord- Boolean rules (AND, NAND, OR, NOR)
ing to which distinct knowledge represen- in a pseudoarithmetic format [e.g., #(0,1),
tations exist for each unique combination of where # indicates the AND rule and the
the digits that constitute a problem, arith- correct response is 0]. Each participant per-
metic operator, and answer. Because there formed six blocks of 48 trials during the
were only small effects of operand order and acquisition phase and six blocks of 64 tri-
of which arithmetic symbol was presented, als in the test phase. Within each block, the
they concluded that this representation is operator and operands could appear simulta-
relatively abstract. Problems with the same neously, the operator could appear first, the
elements are presumed to access the same operands first, or the operator could appear
knowledge representation. in the interval between the two operands.
Rickard and Bourne (1996) conducted In both phases, all of the sequential pre-
some additional tests of the identical- sentation conditions showed faster responses
elements model. Experiment 1 was similar than the simultaneous condition, but the
to that of Rickard et al. (1994), only with operator-first display produced the largest
new problems added at the test phase. Con- benefit.
sistent with the identical-elements model,
problems with the same operators and
operands were faster than those with Conclusion
changes or those that were completely new,
with the latter conditions not differing In his chapter, Laboratory experimentation
significantly. In Experiment 2, in addi- on the genesis of expertise, Shiffrin (1996)
tion to the Arabic numeral format used notes, Laboratory studies of the develop-
in the previous experiments, a written ment of expertise have a history as old
verbal format was used (e.g., four as experimental psychology (p. 3 3 8). He
seven). Transfer tests showed both format- points out that although most current views
nonspecific and specific components, which of the development of expertise place an
2 82 the cambridge handbook of expertise and expert performance
Behavioral and neurophysiological stud- Bower, G. H., & Hilgard, E. R. (1981). Theo-
ies of practice and transfer have converged ries of learning (5 th ed.). Englewood Cliffs, NJ:
to show that learning occurs not only at cen- Prentice-Hall.
tral levels of information processing but also Briggs, G. E., & Naylor, S. C. (1962). The rela-
at early and late levels. Regarding perceptual tive efficiency of several training methods as a
learning, Fahle and Poggio (2002) empha- function of transfer task complexity. Journal of
size, We now know for sure that even the Experimental Psychology, 64, 5 05 5 12.
adult primary sensory cortices have a fair Brochet, F., & Dubourdieu, D. (2001). Wine
amount of plasticity to perform changes of descriptive language supports cognitive speci-
ficity of chemical senses. Brain and Language,
information processing as a result of train-
77 , 187196.
ing. . . . All cortical areas seem to be able to
Bryan, W. L., & Harter, N. (1899). Studies on the
change and adapt their function both on a
telegraphic language: The acquisition of a hier-
fast and a slow time scale (p. xiii). Sim- archy of habits. Psychological Review 6, 3 45
ilarly, Sanes (2003 ) notes, The historical 3 75 .
and recent record provide incontrovertible
Campbell, J. I. D. (1987). Network interference
evidence that many neocortical regions, and mental multiplication. Journal of Experi-
including the motor-related areas, exhibit mental Psychology: Learning, Memory, and Cog-
plasticity and are likely to contribute to nition, 13 , 109123 .
motor-skill learning (p. 225 ). An impli- Campbell, J. I. D., & Graham, D. J. (1985 ). Men-
cation of these findings is that acquiring tal multiplication skill: Structure, process, and
expertise in a domain is likely to involve acquisition. Canadian Journal of Psychology, 3 9,
fundamental cognitive and neural changes 3 3 83 66.
throughout most of the information-pro- Clawson, D. M., Healy, A. F., Ericsson, K. A., &
cessing system. Bourne, L. E., Jr. (2001). Retention and trans-
fer of Morse code reception skill by novices:
Part-whole training. Journal of Experimental
Acknowledgment Psychology: Applied, 7 , 129142.
Clegg, B. A., DiGirolamo, G. J., & Keele, S. W.
The authors thank Alice F. Healy for helpful (1998). Sequence learning. Trends in Cognitive
Sciences, 2 , 275 281.
comments on an earlier draft of this chapter.
Cohen, A., Ivry, R. I., & Keele, S. W. (1990).
Attention and structure in sequence learning.
Journal of Experimental Psychology: Learning,
References Memory, and Cognition, 16, 173 0.
Crowder, R. G. (1976). Principles of learning and
Ahissar, M. (1999). Perceptual learning. Cur- memory. Hillsdale, NJ: Erlbaum.
rent Directions in Psychological Science, 8, 124 Delaney, P. F., Reder, L. M., Staszewski, J. J., &
128. Ritter, F. E. (1998). The strategy-specific nature
Anderson, J. R. (1982). Acquisition of cognitive of improvement: The power law applies by
skill. Psychological Review, 89, 3 69406. strategy within task. Psychological Science, 9,
Ashcraft, M. H. (1992). Cognitive arithmetic: A 17.
review of data and theory. Cognition, 44, 75 Destrebecqz, A., & Cleeremans, A. (2001).
106. Can sequence learning be implicit? New evi-
Biederman, I., & Shiffrar, M. M. (1987). Sexing dence with the process dissociation proce-
day-old chicks: A case study and expert sys- dure. Psychonomic Bulletin & Review, 8, 3 43
tems analysis of a difficult perceptual-learning 3 5 0.
task. Journal of Experimental Psychology: Doane, S. M., Alderton, D. L., Sohn, Y. W., &
Human Perception and Performance, 13 , 640 Pellegrino, J. W. (1996). Acquisition and trans-
645 . fer of skilled performance: Are visual discrimi-
Bilodeau, E. A., & Bilodeau, I. M. (Eds.) (1969). nation skills stimulus specific? Journal of Exper-
Principles of skill acquisition. New York: Aca- imental Psychology: Human Perception and Per-
demic Press. formance, 2 2 , 12181248.
2 84 the cambridge handbook of expertise and expert performance
Dutta, A., & Proctor, R. W. (1992). Persistence Gottsdanker, R., & Stelmach, G. E. (1971).
of stimulus-response compatibility effects with The persistence of psychological refractori-
extended practice. Journal of Experimental ness. Journal of Motor Behavior, 3 , 3 01
Psychology: Learning, Memory, and Cognition, 3 12.
18, 801809. Haider, H., & Frensch, P. A. (2002). Why aggre-
Ericsson, K. A., & Smith, J. (1991). Prospects gated learning follows the power law of prac-
and limits of the empirical study of expertise: tice when individual learning does not: Com-
An introduction. In K. A. Ericsson & J. ment on Rickard (1997, 1999), Delaney et al.
Smith (Eds.), Toward a general theory of (1998), and Palmeri (1999). Journal of Experi-
expertise: Prospects and limits (pp. 13 8). mental Psychology: Learning, Memory, and Cog-
Cambridge, UK: Cambridge University nition, 2 8, 3 92406.
Press. Hazeltine, E., Teague, D., & Ivry, R. B. (2002).
Fahle, M., & Edelman, S. (1993 ). Long-term Simultaneous dual-task performance reveals
learning in vernier acuity: Effects of stimu- parallel response selection. Journal of Experi-
lus orientation, range and of feedback. Vision mental Psychology: Human Perception and Per-
Research, 3 3 , 3 97412. formance, 2 8, 5 275 45 .
Fahle, M., & Poggio, T. (Eds.) (2002). Perceptual Healy, A. F., Wohldmann, E. L., & Bourne,
learning. Cambridge, MA: MIT Press. L. E., Jr. (2005 ). The procedural reinstatement
Fitts, P. M. (1964). Perceptual-motor skill learn- principle: Studies on training, retention, and
ing. In A. W. Melton (Ed.), Categories of human transfer. In A. F. Healy (Ed.), Experimental cog-
learning (pp. 243 285 ). New York: Academic nitive psychology and its applications (pp. 5 9
Press. 71). Washington, DC: American Psychological
Fitts, P. M., & Seeger, C. M. (195 3 ). S-R Association.
compatibility: Spatial characteristics of stimu- Heathcote, A., Brown, S., & Mewhort, D. J. K.
lus and response codes. Journal of Experimental (2000). The power law repealed: The case for
Psychology, 46, 199210. an exponential law of practice. Psychonomic
Fendrich, D. W., Gesi, A. T., Healy, A. F., & Bulletin & Review, 7 , 185 207.
Bourne, L. E., Jr. (1995 ). The contribution Howard, J. H., Mutter, S. A., & Howard, D. V.
of procedural reinstatement to implicit and (1992). Serial pattern learning by event obser-
explicit memory effects in a motor task. In A. F. vation. Journal of Experimental Psychology:
Healy & L. E. Bourne, Jr. (Eds.), Learning and Learning, Memory, and Cognition, 18, 1029
memory of knowledge and skills: Durability and 103 9.
specificity (pp. 6694). Thousand Oaks, CA: Jacoby, L. L. (1991). A process dissociation
Sage. framework: Separating automatic from inten-
Frederiksen, J. R., & White, B. Y. (1989). An tional uses of memory. Journal of Memory &
approach to training based upon principled Language, 3 0, 5 13 5 41.
task decomposition. Acta Psychologica, 71, 89 Karni, A. (1996). The acquisition of perceptual
146. and motor skills: A memory system in the adult
Gibson, E. J., & Pick, A. D. (2000). An ecological human cortex. Cognitive Brain Research, 5 , 3 9
approach to perceptual learning and development. 48.
Oxford, UK: Oxford University Press. Karni, A., & Bertini, G. (1997). Learning percep-
Gibson, J. J. (1947) (Ed.). Motion picture tual skills: Behavioral probes into adult plastic-
research and testing (Army Air Forces Aviation ity. Current Opinion in Neurobiology, 7 , 5 3 0
Psychology Program Research Reports, Report 535.
No. 7). Washingon, DC: U.S. Government Kornblum, S., & Lee, J.-W. (1995 ). Stimulus-
Printing Office. response compatibility with relevant and irrel-
Goldstone, R. L., Schyns, P. G., & Medin, D. L. evant stimulus dimensions that do and do not
(1997). Learning to bridge between percep- overlap with the response. Journal of Experi-
tion and cognition. In R. L. Goldstone, D. L. mental Psychology: Human Perception and Per-
Medin, & P. G. Schyns (Eds.), Perceptual learn- formance, 2 1, 85 5 875 .
ing (The psychology of learning and motivation, Leahey, T. H. (2003 ). Cognition and learning.
Vol. 3 6, pp. 114). San Diego, CA: Academic In D. K. Freedheim (Ed.), History of psychol-
Press. ogy (pp. 10913 3 ). Volume 1 in I. B. Weiner
training, skill acquisition, and retention 2 85
the central cognitive bottleneck. Psychological Tagliabue, M., Zorzi, M., Umilta, C., & Bassig-
Science, 12 , 101108. nani, F. (2000). The role of LTM links and
Shiffrin, R. M. (1996). Laboratory experimenta- STM links in the Simon effect. Journal of Exper-
tion on the genesis of expertise. In K. A. Eric- imental Psychology: Human Perception and Per-
sson (Ed.), The road to excellence: The acquisi- formance, 2 6, 648670.
tion of expert performance in the arts and sciences, Tagliabue, M., Zorzi, M., & Umilta, C.
sports, and games (pp. 3 3 73 45 ). Mawah, NJ: (2002). Cross-modal re-mapping influences
Erlbaum. the Simon effect. Memory & Cognition, 3 0,
Shiffrin, R. M., & Lightfoot, N. (1997). Percep- 1823 .
tual learning of alphanumeric-like characters. Tombu, M., & Jolicur, P. (2003 ). A cen-
In R. L. Goldstone, D. L. Medin, & P. G. Schyns tral capacity sharing model of dual-task per-
(Eds.), Perceptual learning (The psychology of formance. Journal of Experimental Psychology:
learning and motivation, Vol. 3 6, pp. 45 81). San Human Perception and Performance, 2 9, 3 18.
Diego, CA: Academic Press. Van Selst, M. A., Ruthruff, E., & Johnston, J. C.
Shiffrin, R. M., & Schneider, W. (1977). Con- (1999). Can practice eliminate the psycholog-
trolled and automatic human information ical refractory period effect? Journal of Exper-
processing: II. Perceptual learning, automatic imental Psychology: Human Perception and Per-
attending, and a general theory. Psychological formance, 2 5 , 12681283 .
Review, 84, 127190. Vu, K.-P. L. (2006). Influences on the Simon
Simon, J. R. (1990). The effects of an irrelevant effect of prior practice with spatially incompatible
directional cue on human information process- mappings: Transfer within and between horizon-
ing. In R. W. Proctor & T. G. Reeve (Eds.), tal and vertical dimensions. Manuscript submit-
Stimulus-response compatibility: An integrated ted for publication.
perspective (pp. 3 186). Amsterdam: North- Vu, K.-P. L., Proctor, R. W., & Urcuioli, P.
Holland. (2003 ). Effects of practice with an incompat-
Simon, J. R., Craft, J. L., & Webster, J. B. (1973 ). ible location-relevant mapping on the Simon
Reaction toward the stimulus source: Analysis effect in a transfer task. Memory & Cognition,
of correct responses and errors over a five-day 3 1, 1146115 2.
period. Journal of Experimental Psychology, 101, Wilkinson, L. & Shanks, D. R. (2004). Intentional
175 178. control and implicit sequence learning. Journal
Snoddy, G. S. (1926). Learning and stability: A of Experimental Psychology: Learning, Memory,
psychophysiological analysis of a case of motor & Cognition, 3 0, 3 5 43 69.
learning with clinical applications. Journal of Willingham, D. B. (1999). Implicit motor
Applied Psychology, 10, 13 6. sequence learning is not purely perceptual.
Sohn, M.-H., & Carlson, R. A. (1998). Procedural Memory & Cognition, 2 7 , 5 615 72.
frameworks for simple arithmetic tasks. Journal Willingham, D. B., Nissen, M. J., & Bulle-
of Experimental Psychology: Learning, Memory, mer, P. (1989). On the development of pro-
and Cognition, 2 4, 105 21067. cedural knowledge. Journal of Experimental
Sowden, P. T., Davies, I. R. L., & Roling, P. (2000). Psychology: Learning, Memory, & Cognition, 15 ,
Perceptual learning of the detection of features 10471060.
in X-ray images: A functional role for improve- Willingham, D. B., Wells, L. A., Farrell, J. M.,
ments in adults, visual sensitivity? Journal of & Stemwedel, M. E. (2000). Implicit motor
Experimental Psychology: Human Perception and sequence learning is represented in response
Performance, 2 6, 3 793 90. locations. Memory & Cognition, 2 8, 3 663 75 .
Speelman, C. P., & Kirsner, K. (2001). Predicting Zorzi, M., & Umilta, C. (1995 ). A computa-
transfer from training performance. Acta Psy- tional model of the Simon effect. Psychological
chologica, 108, 247281. Research, 5 8, 193 205 .
C H A P T E R 16
Lauren A. Sosniak
If we want to know how people become (Bloom, 1985 ), work directed by Benjamin
extraordinary adults, we can start with some Bloom at the University of Chicago in the
of the latter . . . and then try to find out how 1980s and carried out by a large research
they came to do it (Gruber, 1982, p. 15 ). team. This series of studies has received
That is the premise for retrospective inter- widespread attention and to some degree has
views in the study of expertise and expert become emblematic of retrospective studies
performance. of expert performance.
As this Handbook makes quite clear, there Ironically, when we conducted the Devel-
is a large body of work that coalesces around opment of Talent Project we did not think of
what Gruber (1986), again, calls an interest ourselves as studying expertise. That was not
in . . . human beings . . . at their best (p. 248). the language we spoke. Apparently, it was
Only a very small portion of that work uses not yet the language of the day. According to
the method of retrospective interview. But Glaser and Chi (1988), The topic of exper-
findings from retrospective interview stud- tise first appears in major textbooks in cogni-
ies have been important in their own right tive psychology in 1985 , in John Andersons
and of significant use to others studying second edition of Cognitive Psychology and
expert performance in other ways. My task Its Implications (p. xvii). Of course that
in this chapter is to speak about studies was the very same year we published the
using retrospective interviews, to highlight major account of the Development of Talent
what we have learned from them, to note Project.
their strengths and limitations, and to con- In 1985 , when expertise was becoming
sider where we might head next both in this known by that label and was becoming an
tradition and as a result of work from this important area of study in psychology, we,
tradition. working from a department of education,
In some respects I am well qualified to wrote instead about the development of
tackle this task. I served as research coordi- talent, about exceptional accomplishment,
nator for the Development of Talent Project about reaching the limits of learning. Others,
2 87
2 88 the cambridge handbook of expertise and expert performance
previously and since, have used the lan- cussion of potentially fruitful directions for
guage of giftedness, human extraordinari- future work.
ness, genius, prodigious performance, and
so on.
The particular language we use has con- The Development of Talent Project
sequences for what we investigate, how we
investigate, and what we report as findings. The Development of Talent Project (Bloom,
But, as a set, the various words do repre- 1985 ) was a study of groups of individu-
sent an interest in human performance at its als who, though relatively young (under the
very best. And the words frequently are used age of 3 5 ), had realized exceptional lev-
interchangeably, irrespective of the theoret- els of accomplishment in one of six fields:
ical orientation of the author. Consider, for concert piano, sculpture, swimming, tennis,
example, the opening sentences in Michael mathematics, and research neurology (two
Posners (1988) chapter on What is it to be artistic fields, two psychomotor activities,
an expert?: How do we identify a person and two academic/intellectual fields). The
as exceptional or gifted? One aspect is truly project explored the lives of 120 talented
expert performance in some domain (p. xxix, individuals in all, approximately 20 in each
all italics added). The various words used field. The plan was to search for regulari-
by different authors have more that unites ties and recurrent patterns in the educational
them than divides them, and so this chapter histories of groups of clearly accomplished
includes them collectively. individuals, hoping that such consistencies
In this chapter I also walk a fine line might shed light on how the development
regarding method. Retrospective interview of high levels of talent is achieved.
studies are, inherently, biographical studies. We conducted retrospective, semi-
They mine life experiences. Yet they do not structured, face-to-face interviews with
rely on standard biographical material. They the individuals who met criteria of excep-
rely on interviews, on allowing an individual tional achievement set by experts in their
to tell his or her life story shaped within the respective fields. We interviewed parents as
theoretical framework of the interviewer. well, by telephone, for corroborative and
Yet unlike most interviews done in the ser- supplementary information. Data analysis
vice of investigating expertise, these focus was a process analogous to superimposing
on more distant and less proximal events the unique histories one on the other, and
and experiences. In this chapter I ignore identifying the patterns that were common
both biographical studies and the studies across most cases. We did this first within
using retrospective verbal reports of near- each field and later across the fields.
term activities. These forms of research are One of the important findings from this
tackled most ably elsewhere in this Hand- study has to do with what we did not
book (e.g., Simonton, Chapter 18; Deakin, find. At the start of the project Bloom
Cote & Harvey, Chapter 17). expected that the individuals we studied
My plan for the rest of this chapter is would be initially identified as possessing
as follows: first, I will outline several of special gifts or qualities and then provided
the retrospective interview studies that have with special instruction and encouragement
informed research on expertise, beginning (Bloom, 1982, p. 5 20). But data from the
with the one I know best. Given, then, a study made it clear that this initial assump-
body of information about expertise derived tion of early discovery followed by instruc-
from retrospective interviews, I will dis- tion and support was wrong. The individ-
cuss the strengths and weaknesses of the uals in the sample for the Development
methodology and call attention to mecha- of Talent Project typically did not show
nisms researchers have used to strengthen unusual promise at the start. And, typi-
their studies. I will move then to high- cally, there was no early intention of working
light the convergence of certain findings toward a standard of excellence in a particu-
across studies of expertise and to a dis- lar field. Instead of early discovery followed
retrospective interviews 2 89
by development, we found that the individ- the sample for the Development of Talent
uals were encouraged and supported in con- Project, the development of talent was a pro-
siderable learning before they were identi- cess of learning that grew from and threaded
fied as special and then accorded even more around their everyday lives. It was both for-
encouragement and support. More time and mal and informal, structured and casual, self-
interest invested in the talent field resulted conscious and matter-of-fact, special and
in further identification of special qualities ordinary, all at the same time. Time spent
that in turn were again rewarded with more learning had a vertical dimension as well as
encouragement and support. Aptitudes, atti- a horizontal one, and an affective dimension
tudes, and expectations grew in concert with as well as a cognitive one.
one another, and were mutually confirming We also came to appreciate that the
(Bloom, 1985 ; Sosniak, 1987). long-term process of developing talent was
As Bloom once told a reporter: We not simply a matter of becoming quan-
were looking for exceptional kids and what titatively more knowledgeable and skilled
we found were exceptional conditions over time, or of working more intensely for
(Carlson, 1985 , p. 49). The exceptional longer hours. It was, predominantly, a mat-
conditions we found can be summarized ter of qualitative and evolutionary transfor-
under the headings of opportunity to learn, mations. The individuals were transformed,
authentic tasks, and exceptionally support- the substance of what was being learned
ive social contexts (Sosniak, 2003 ). was transformed, and the manner in which
Some of our findings about time (oppor- individuals engaged with teachers and field-
tunity to learn) are well known. We con- specific content was transformed. Students
firmed what many researchers had suspected progressively adopted different views of who
and some already had demonstrated: devel- they were, of what their fields of expertise
oping exceptional abilities takes a lot of were about, and of how the field fitted into
time. We found, for example, that interna- their lives (Sosniak, 1987). These transfor-
tionally recognized concert pianists worked mations generally followed a pattern rem-
for an average of 17 years from their first iniscent of Whiteheads (1929) rhythms of
formal lessons to their first international learning phases of romance, precision, and
recognition; the quickest in the group went generalization.
from novice to tyro in a dozen years. For The tasks the individuals were engaged
Olympic swimmers, 15 years elapsed, on with over time also seem to be particularly
average, between the time they began swim- notable. Although tasks obviously changed
ming just for fun and the time they earned a over time, consistent with the amount of
place on an Olympic team. experience the individuals had with their
Although the sheer number of years the field, they were stable in some key ways.
individuals spent acquiring knowledge and Typically they were tasks that represented
skill was impressive, the way time was dis- the field itself in its contemporary social con-
tributed also was striking. We took note, for struction. They were genuine, they were real
example, of the small amount of time the in an everyday commonsense way. Children
individuals spent in formal instruction per- and youth did things they knew other peo-
haps once a week, for many of their years of ple of various ages from various settings also
learning. And we were frequently jolted by were doing and had been doing for years.
the multiple and overlapping arenas for time Children and youth used materials that
spent learning: children and youth not only are part of our social technologies they
received instruction in a talent area, they played pianos that adults used also, swam
also played at the talent quite informally, in Olympic-sized pools, read field-specific
they read about and watched or listened to books and magazines that were created for
others working at the talent area, and they the consuming public. Both the tasks the
demonstrated their involvement with the youth engaged in, and the materials they
talent area to others both publicly and used to pursue their tasks, were connected
privately. For the individuals who formed to tasks valued by significant portions of
2 90 the cambridge handbook of expertise and expert performance
society. And youth knew these tasks and before the Development of Talent Project,
materials were valued because they saw by Anne Roe (195 2), who used retrospec-
them being displayed by others in their tive interviews (in part) to study the making
family, in their community, and in ever larger of scientific talent. The central question that
arenas. guided her work The making of a scientist
Finally, and as a logical extension, we asked what kinds of people do what kinds
found that the development of talent of scientific research and why, and how, and
appears to require enormously supportive when (p. 1). Woven among the more spe-
social contexts. One of the lessons we cific questions she worked with were: Are
learned from the project is that no one scientists different from other people? Is one
develops talent on his or her own, without kind of scientist characteristically different
the support, encouragement, advice, insight, from another? (p. 13 ).
guidance, and goodwill of many others. The Roe studied sixty-four of the countrys
many years of work on the way to interna- leading scientists, divided approximately
tional recognition involved increasing expo- equally among biologists, physicists, and
sure to and participation in communities of social scientists. Her focus was on a sam-
practice for their respective talent fields. ple representing the best men in each field
Communities of practice are groups of (p. 20). (Yes, Roe studied men only. Appar-
people who share a substantive focus that ently there was only one woman who, but for
is important to their lives, and who share a her gender, would have qualified to be part
willingness to invest time and effort around of the sample in biology.) Roe interviewed
work in that substantive area. Communi- the scientists and administered several psy-
ties of practice offered models for develop- chological tests the Rorschach, the The-
ment, and they offered resources for support, matic Apperception Test, and a set of timed
inspiration, and sustenance. Communities of intelligence tests verbal, spatial, and math-
practice created standards for work for ematical.
work by the novice, the knowledgeable lay- Roe concluded her major report of her
man, and for the expert. For participants in study with findings that emphasized per-
the Development of Talent Project, com- sonal and psychological characteristics. Her
munities of practice modeled and inspired leading scientists largely came from families
excellence, they defined and gave meaning with professional fathers. They came from
to significant educative tasks, and they sup- homes with select religious backgrounds
ported and sustained work over the long (none were from Catholic homes). What
periods necessary for the development of seems to be important in the home back-
talent. The youth and young adults in the ground is the knowledge of learning, and
Development of Talent Project were fortu- the value placed on it for its own sake, in
nate to be welcomed into, or to find their way terms of the enrichment of life, and not just
into, communities that shaped and inspired for economic and social rewards (p. 23 1).
their work. They had many varied oppor- Roe noted that [m]ore than is usual, these
tunities to see themselves as a member of men were placed on their own resources
field-specific communities, to come to know (p. 23 1), as a result of losing a parent early
the commitments, and to watch and live out in life, suffering serious physical problems,
for themselves the process of a community being eldest sons.
renewing itself. Roe also found that [m]ost of them were
inveterate readers, and most of them enjoyed
school and studying (p. 23 1). They were,
Other Retrospective Interview Studies principally, products of public, not private,
of Expertise and Expert performance schools. Their early interests typically were
consistent with the fields they would later
Another of the well-known series of studies take on vocationally, with physical scien-
in this genre was conducted three decades tists involved in gadgeteering, for example,
retrospective interviews 2 91
and the biologists interested in natural his- work. Even for those who focus more on
tory. Their [l]evel of intellectual function- the individual and less on the context there
ing (p. 23 3 ) was very high, with different is much here that supplements and comple-
patterns for the different scientific fields. ments work done by Roe, Bloom, and others.
Roe also addressed the scientists gen- Zuckerman, like Roe before her, found
eral personality structure at some length. that her elite scientists largely were prod-
She highlighted their curiosity, their gen- ucts of middle-class families with pro-
eral need for independence, for autonomy, fessional fathers. The laureates religious
for personal mastery of the environment origins also dovetailed with Roes scien-
(p. 23 5 ). tists, with just 1 percent of the laure-
Harriet Zuckermans (1977/1996) work ates from a Catholic background and with
as reported, for example, in Scientific elite: Jewish backgrounds somewhat overrepre-
Nobel laureates in the United States sented. Zuckerman reveals little about the
represents a third important example of ret- laureates earliest years, but makes significant
rospective interviews in the study of exper- contribution to the making of . . . scien-
tise. In this instance, the central focus was tific expertise beginning with the informa-
on the sociology of science the stratifi- tion she reports about college enrollment
cation system in science, its shape, mainte- and continuing through her discussion of the
nance and consequences (p. xxxii) rather graduate school experience.
than a focus on the individuals for the pur- Zuckermans account of masters and
pose of learning more about their personal- apprentices (chapter 4) is particularly
ities or their educational experiences. important reading for a better understand-
Zuckerman began her investigations by ing of a significant stage in the development
interviewing forty-one of the fifty-six lau- of expertise. In this chapter Zuckerman pro-
reates then at work in the United States vides important details about finding places
(p. 4). Subsequently, she collected consider- to study, working with expert teachers, and
able additional information, including some becoming socialized into the world and the
additional interviews, for a larger sample of ways of the scientific elite.
ninety-two, including all American Nobel There have been other retrospective
laureates from the first, Albert A. Michelson, interviews of expertise and expert perfor-
who won the award in physics in 1907, to mance that might be included in this chap-
those who had won prizes by 1972 (p. 4). ter if we maintain a very broad approach to
Zuckermans research focus expanded over the language of expertise. For example, in
time, and ultimately centered on questions an interesting twist on retrospective studies
of how [American laureates] are educated, of expertise and expert performance, Sub-
recruited, sustained, and what contributions otnik et al. (1993 ) studied the adult lives
they have made to the advancement of sci- of grown-up high-IQ children from Hunter
ence (p. 5 ). College Elementary School (p. vii). Hunter
Zuckermans work is finely detailed and was a school for the intellectually gifted
engagingly written. Her primary focus on (p. 20) in New York City, serving students
the world of elite science rather than on from nursery school to grade six who entered
the development of the scientists themselves with an IQ score of 13 0 or above (mean
means that some significant parts of the book IQ 15 7). The Subotnik volume speaks to
might not seem at first glance to inform scholars interested in gifted education as it
studies of expertise. But those whose work was conceived and practiced in the middle
emphasizes expertise as a co-construction of the 20th century. However, for the pur-
between individuals and domains (cf., pose of better understanding expertise, the
Csikszentmihalyi & Robinson, 1986; Feld- chief contribution this retrospective study
man, 1986) will find much in the way of makes is to remind us how difficult it is
detail that both coalesces with other work to identify youth who will realize excep-
and that can inform hypotheses for future tional adult accomplishment, and how poor
2 92 the cambridge handbook of expertise and expert performance
proposition that we do not yet have the rather than an individual, researchers can
appropriate markers to know whom to fol- harvest what is essential for the development
low longitudinally. As Goldsmith (2000) of expertise and leave behind that which is
writes: idiosyncratic.
The process for identifying the specific
By and large, the children I have studied
people who would become part of the
have not gained national attention for their
work, although they are still young enough studied group by virtue of their excep-
that they may yet do so in the future. Some tional accomplishment also was quite sim-
have long since given up their original areas ilar across the studies: it relied on other
of achievement, so if they are to develop experts, specific to the fields of study. People
national visibility, it will be in some other with broad and deep knowledge of each field
domain of accomplishment. (p. 115 ) made decisions about what was most impor-
tant in and to each field. For the Zuckerman
The question, then, is not whether we
study, for example, people Zuckerman did
need to use the method of retrospective
not know, who had served on Nobel Prize
interview in the study of expertise, but,
selection committees, defined for her the
rather, how best to use the method. The
talented individuals she would study (every
Roe, Zuckerman, and Bloom studies offer
American who had been honored with the
insights into mechanisms for enhancing reli-
Nobel award in one of the sciences was
ability and validity in retrospective interview
included in the sample). For the Bloom
studies.
project, older experts who were known for
knowing about their fields defined the cri-
Defining a Sample
terion measures for exceptional performance
The Roe, Zuckerman, and Bloom studies all that would capture a sample of the younger
chose samples of (a) recognized experts (b) talented individuals in the study; everyone
in numbers sufficient to allow for a study of who met the criteria set (winning certain
groups rather than individuals. The quality competitions or awards, and so on) was
of the people interviewed the transparency invited to participate.
of their expertise mattered a great deal to Some researchers argue against rely-
all of the researchers. As Roe (195 2) put it, ing on social judgments and/or relying
in a sentiment shared across these studies, on a relativistic approach of looking for
if there were particular factors in the lives the best among a group to charac-
or personalities of men which were related terize expertise. According to Salthouse
to their choice of vocation, these factors (1991), [c]onsensual judgments of exper-
should appear and possibly would appear tise should . . . be avoided, because they can
most clearly in the men who had been most be influenced by a variety of characteris-
successful at the vocation (pp. 2021). For tics other than true competence, such as
Zuckerman (1977/1996), It seemed sensi- popularity or reputation (p. 287). Sloboda
ble to assume that Nobel prize-winners con- (1991) argues that we have to find a char-
stituted a small sample of the most accom- acterization of expertise that will allow
plished American scientists making up the any number of people (up to and includ-
scientific elite (p. 3 ). ing all) to be expert in a particular area
Extreme instances of accomplishment (p. 15 4). Generally, however, examining
were thought to provide the clearest and extreme cases as defined normatively by
most compelling findings. And relying on people who should be qualified to make such
data from a set of expert performers, rather distinctions is widely accepted as a reason-
than focusing on individuals, also was consid- able strategy, at least for certain domains of
ered important to the design. The assump- expertise where other measures of compe-
tion underlying the preference for studies tence might not be available. Both Bloom
of groups rather than the individual-case- and Zuckerman acknowledged that their
study approach is that by studying a group samples may have excluded others who
2 94 the cambridge handbook of expertise and expert performance
were similarly exceptional except for meet- The Issue of Control / Comparison
ing a very particular criteria (e.g., a Nobel Groups
Prize, one of a defined set of piano com-
petitions). But both Bloom and Zuckerman Retrospective interviews of such elite groups
argued that although some talented people as these pose the ultimate challenge to
might have been excluded, there could be the issue of control or comparison groups.
no doubt that those who were included were Against whom should the small number
exceptional. of people reaching the highest rungs of a
All three of the sets of studies tweaked domain ladder be compared? Do we com-
the definition of sample in small ways. Both pare elite concert pianists, for example, with
the Roe and Bloom studies set age ceil- people who never took music lessons at
ings for the individuals they would invite all? With people who studied music briefly,
to participate. In the Bloom studies the age but abandoned their studies before they left
ceiling was set much lower (approximately childhood? With a random sample of Amer-
age 3 5 ) than in the Roe study (61 years). icans stratified by age, geography, socioeco-
Because the Bloom studies were focused nomic status? With people who almost
on the educational experiences of the sam- made the elite and thus might be expected
ple, rather than personality and intelligence to have the closest relation in experiences to
characteristics of the sample, the age of the the elite group?
people interviewed had special importance. What difference does it make if we have
The aim was to identify a sample that had no comparison group at all? A great deal, per-
reached the age of demonstration of tal- haps. Absent a control or comparison group,
ent, rather than merely indication of poten- it is impossible to ascertain which of the
tial, but at the same time the individuals findings, or to what extent any of the find-
should not be so far along in a career ings, tells anything about the development
sense that they would be unlikely to remem- of exceptional talent. Absent a control or
ber their early experiences or might shade comparison group it is possible that anything
those recollections significantly in relation and everything reported as findings about
to subsequent events. The Bloom studies the development of talent could be so for
also included interviews with parents of the many people who have never demonstrated
talented individuals in a further effort to exceptional talent of any sort. Yet a random
capture early experiences and to help tri- sample control group makes no sense for a
angulate the information provided by the purposefully chosen elite study group, and
talented individuals themselves; the rela- the possibility of a matched sample on cer-
tively young age of the talented individ- tain key criteria gets weighed down with the
uals increased the likelihood that parents question of what to match for a long-term
might be able and willing to contribute to experience.
the research. Roe, Zuckerman, and Bloom took over-
Finally, all three of the sets of studies lapping approaches to the challenge of con-
focused on American experts. Although this trol or comparison groups. Roe created a
might seem a concession to the costs and subsidiary study to check on how closely
other difficulties associated with scanning eminent men resembled other men in the
the world for the clearest instances of exper- same fields (p. 214); for this study she
tise, the Bloom study articulated a differ- administered a group Rorschach to more
ent rationale: the focus on Americans only than 3 82 scientists at fourteen universities.
would reduce the variation that might be Roe also made use of the naturally occur-
associated with differences in the home and ring divisions in science to study separately
educational cultures in different parts of the biologists (who were involved with basic
world. I will have more to say about the research into normal life processes), phys-
American nature of these studies later in ical scientists (both theoretical and experi-
this chapter. mental), and social scientists (psychologists
retrospective interviews 2 95
and anthropologists), and then compared relied on what we were told in the course of
her experts in one domain with experts in our interviews without any predetermined
another domain. Zuckerman did not create intentional probing.
a formal subsidiary study but did take into Freemans larger point is the possibility
account in her discussions a group of sci- for comparison groups created by some of
entists who are said to hold the forty-first the key findings in the studies. In this regard,
chair. . . . uncrowned laureates who are the as will become clearer shortly, another likely
peers of prize-winners in every sense except comparison group would be all of the indi-
that of having the award (p. 42). viduals who studied with the same final
The Bloom studies took advantage of master teacher (or in the same graduate
the approach of comparing experts in one school program) identified as so important
domain with experts in another domain across many studies.
and added an additional methodological ele- The bigger error I think we may have
ment. Bloom created a spectrum of talents: made in the Bloom study (and this could be
psychomotor activities, artistic fields, aca- said as well for the Roe and Zuckerman stud-
demic/intellectual fields, and fields empha- ies also) was that we purposefully chose to
sizing interpersonal relations. Each portion study only Americans. Although this made
of the spectrum was to be represented conceptual and procedural sense, it did deny
by two separate investigations, with fields us the chance of using as a comparison group
that clearly belonged together in the spec- people who had realized exceptional levels of
trum but diverged in identifiably signifi- talent under different conditions (e.g., other
cant ways. Olympic swimmers and tennis countries best pianists, best swimmers, best
players represented psychomotor activi- mathematicians, and so on).
ties, concert pianists and sculptors repre-
sented the arts, and mathematicians and
Other Issues of Validity
research neurologists represented the aca-
demic/intellectual talents. (The last portion We need to recognize that whether we
of the spectrum interpersonal relations use retrospective interviews or any other
ultimately was abandoned; we found our- method to study a long-term process like
selves unable at the time to specify appro- the development of expertise, we cannot
priate talent fields and criteria for recog- collect every bit of information that might,
nizing individual achievement.) Thus the ultimately, be of value. As Freeman (2000)
Bloom studies involved six groups of tal- points out, we have to recognize that we
ented individuals: each group would have its can never identify and measure the full con-
own story, pairs of fields should be related text of anyones life, even in the present, and
by many common considerations, and all six interpretation of data can only be as well
groups might share at least some significant informed as possible (p. 23 6).
elements of the development of talent (or so The Roe, Zuckerman, and Bloom studies
it was hoped). were theory driven. Of course, each relied
Freeman (2000) points out that the on different bodies of knowledge for its the-
Bloom studies (and this could be said as well oretical framework. Theory-driven work is
for the Roe and Zuckerman studies) failed important because only in this way can we
to make comparisons with siblings even as clearly focus our attention for data collection
we concluded that family influences were and analysis, and make reasonable efforts to
significant. Freeman is correct. Although in look for both confirming and disconfirming
the Bloom studies we did make mention evidence. Given the limits of data collec-
of siblings from time to time, in relation tion in all instances, working atheoretically
to the talented individuals own compar- would not help us build increasing bodies
isons with their siblings or parents compar- of knowledge and would be disrespectful of
isons for their own children, we did not set the time and trust that the talented indi-
out to study siblings, and our comparisons viduals invest in our studies. Nevertheless,
2 96 the cambridge handbook of expertise and expert performance
as social science theories evolve, prior work interview experts as young as is reasonable
may come to look increasingly less meaning- to do so; if the consequences of certain later-
ful and ridden with significant holes. Perhaps stage experiences are the research interest,
one way to compensate for this is to encour- then of course we want to interview when
age the collection and reporting of data at the enough time has lapsed to be able to see a
most descriptive and least inferential levels wide range of possible consequences.
possible; others, then, might be able to take Because people have many more mem-
significant advantage of previous work even ories than can be captured in interviews,
as theoretical frameworks change. and especially because in this body of work
One further dilemma associated with no single memory nor even any single par-
studying experts and working backward is ticipant is essential to data collection and
that we run the risk of confusing what analysis, the issue of recollective memory
was the case for the experts with what seems less problematic than other consid-
needs to be the case or might be the case erations for retrospective interview studies.
in other circumstances. Retrospective inter- For example, working with a coherent sam-
views focused on experiences from many ple and then trying to collect supplementary
years back are particularly sensitive to chal- and collaborative information (from parents,
lenges from social moments, or cohort from previously conducted and published
effects. In other words, retrospective inter- interviews, and from other data about peo-
views conducted with adults who were ples lives) seem to be important method-
youth in, say, the 195 0s, run the risk of con- ological considerations. Similarly, the care
fusing elements important to the develop- taken in describing and analyzing data within
ment of expertise with the circumstances of cases and across cases, and looking thought-
the times. For example, retrospective inter- fully for negative evidence for developing
views that highlight intact families with cer- arguments, will distinguish studies that will
tain child-rearing characteristics might tell hold up or at least continue to be useful over
as much about the social times as they tell time. And it is important that studies be use-
about what it takes to become an expert. ful over time, be repeated and repeatable
Again, this calls attention to the importance in some fashion over time, to help separate
of multiple studies of the development of out important elements of the development
expertise under different conditions. of expertise from specific cohort or context
One specific aspect of a cohort effect effects.
seems not to be particularly problematic: Of course all methods for studying exper-
how much people are able to recollect about tise have strengths and weaknesses. There
their life experiences across their lifespans. are no methodologies inherently better than
Considerable research on autobiographical the others except in relation to the particular
or recollective memory (e.g., Rubin, 1986, research question(s) proposed. Ochse (1990,
1996; Rubin & Schulkind, 1997) indicates pp. 3 745 ) offers a succinct summary of dif-
first that people have many more memories ferent methodologies for studying people
than can be captured in any set of interviews. who have achieved excellence (p. 3 7), cites
Interestingly, the distribution of memories various studies done in each tradition, and
across the lifespan seems quite orderly: fol- notes the strengths and weaknesses of the
lowing a period of childhood amnesia for different forms of study. Ochse argues for
a persons earliest years, there seems to be the importance of convergence of findings
simple retention for the most recent 20 to across different methodologies:
3 0 years of a persons life, and, for people
If consistent findings emerge from stud-
older than 3 5 , a bump in the number of
ies in which different methods were
memories from ages 10 to 3 0. Of course employed, one may with some confidence
the age of interviewees most appropriately ascribe the correspondence to principles
should reflect the data the researchers are governing natural underlying regularities
trying to collect. If data about early years rather than repeated methodological errors.
matter a great deal, then we would want to (p. 3 7)
retrospective interviews 2 97
Common Patterns across Studies and the near term. My own puzzlement has to do
Some Implications for Future Work with how individuals negotiate the transition
points. How, for example, do children make
Notwithstanding all of the limitations of ret- the move from enjoying playful experiences
rospective interviews in the service of study- with a field to becoming more deliberate,
ing expertise and expert performance, and precise, and intense in their involvement?
the significant differences in the theoret- Can we learn to account for how and why
ical frameworks for the Roe, Zuckerman, some people make this transition and oth-
and Bloom studies, there are findings that ers are left behind or drop out entirely?
appear again and again in these studies and Similarly, there seems to be a critical tran-
not infrequently in studies using other meth- sition between developing serious compe-
ods of investigation. The most obvious of tence and then moving further toward the
the common findings relate to time. Multiple limits of expertise. This seems to be the tran-
studies report specific, continued, long-term sition between precision and generalization
experience with a field before a person real- in Blooms work, the move to study with
izes exceptional accomplishment (see, for a master teacher in Zuckermans work, and
example, Simon & Chase, 1973 ; Ericsson, the crisis that Bamberger (1982) identifies
Krampe, & Tesch-Romer, 1993 ). This find- in her work with prodigies. How and why do
ing has led many of us, even from quite dif- students of a field get so very far in develop-
ferent theoretical orientations, to hypoth- ing competence and move no further?
esize that the central challenge of helping An alternative point of view about the
people develop exceptional abilities is that phases, however, might be that they are
of creating and maintaining the motivation stunting growth, unnecessarily prolonging
necessary to stay with a field for the many the development of expertise, and that
years it takes to develop expertise (Sosniak, students of a field might benefit by being
1987; Posner, 1988). See also Zimmerman, introduced early to the final phase, which
Chapter 3 9. represents work as experts engage it. I am
We do not yet seem to have research probably misinterpreting Bereiter and Scar-
addressing how to help create and main- damalia (1986) when they write relative
tain long-term investments in learning. We experts are not merely better at doing the
do have evidence, however, that the long- same things that others do; they do things
term experience is not all of one kind, but, differently, and the same differences appear
rather, involves a series of phases of quali- in various domains (p. 16), but I read into
tatively different experiences (Sosniak, 1987). their discussion of the qualitative differences
Motivation undoubtedly needs to be under- the implication that we should, then, teach
stood as both an individual quality and as people from the start how to work as experts
socially promoted, embedded in tasks and do. Ochse (1990) makes just this leap
teaching and learning interactions, public from what experts know and do to what
performances, and so on. And motivation we should teach novices when she sup-
undoubtedly needs to be understood as it ports teaching thinking styles rather than
likely changes over time in relation to activi- facts early in a students experience based
ties and experiences. Changes in motivation on work indicating that Nobel Prize winners
over time are suggested also in the work of testified that learning thinking styles was the
Ericsson, Tesch-Romer, and Krampe (1990) most influential part of their work with mas-
and Ericsson and Charness (1994), who used ter teachers (see p. 25 9).
a modified version of the phases from the It may indeed make sense to begin a
Bloom studies to pursue investigations into students education with what the most
the influence of practice on the development advanced learners teach us about their
of expert performance. knowledge, skill, and ways of working. Cer-
The retrospective interviews that point to tainly, it might make sense to focus on the
qualitatively different phases suggest addi- underlying principles and processes of a field
tional research that might be profitable in in even the earliest instruction. But it might
2 98 the cambridge handbook of expertise and expert performance
also be the case that learners need signifi- development, inevitably lead to a predom-
cantly different activities and aims, different inant point of view about the importance of
feedback and correctives, different standards an early start. I have used early start in
of excellence, before they can make sense of quotation marks because, although it sum-
and make use of the expectations and strate- marizes findings within studies, it does not
gies associated with expert performance. necessarily mean the same thing across stud-
Although I have argued elsewhere about ies. In the Bloom studies, for example, early
the necessity of distinctly different phases of start would refer to experiences that people
learning, this is, of course, a testable question who eventually became outstanding concert
that is as yet untested. pianists or Olympic swimmers had before
There does seem to be considerable agree- they were old enough even to enroll in ele-
ment across retrospective interview stud- mentary school. In the Zuckerman study,
ies and other investigations that people however, early start would mean earning
who demonstrate exceptional accomplish- ones doctorate at a younger age than the
ment have almost always had the experi- average scientist, and producing scientific
ence of studying with a master teacher a publications in ones twenties.
teacher who has considerable standing in The value of an early start undoubt-
the field and who has helped prepare others edly is intricately linked with the demands
who are known for their accomplishments. and expectations of a particular domain of
There is evidence also that work with a mas- expertise. In domains that call for consid-
ter teacher is significantly different from ear- erable physiological development, a young
lier educational experiences. Zuckermans start might be particularly beneficial in order
reports of Nobel laureates experiences with to catch developing bodies and minds at the
master teachers dovetail with reports from most malleable times. In domains that call
the Bloom studies. for significant advanced education, including
Zuckerman writes about the match doctoral studies and post-doctoral research
between a master teacher and a student as experiences, an early start might logically be
a jointly operating process of self-selection linked with choice of college, which then
by the future scientists and selective recruit- becomes a feeder into graduate programs
ment by the academic institutions (p. 107). and beyond.
How the master and the apprentice get The age at which one must begin work
together, what they do together, and how in a field surely is linked with moments that
some survive this experience while others the field uses to recognize and honor its par-
fall by the wayside, would seem to be fruit- ticipants. And following from this, the age
ful arenas for investigation. at which one must begin work in a field
Investigations into experiences with mas- also likely is linked to the coherence and
ter teachers should be mindful of the fact history of the field. Newly emerging fields
that considerable learning has already taken (like research neurology when the Bloom
place, and the only way to understand what studies were conducted) might allow for
happens with a master teacher is to know or even require different processes for the
well what prepared the student for this development of expertise than more firmly
teacher. In other words, students come to established and perhaps more closed fields.
a master teacher with no small measure of For reasons I still do not understand well,
expertise already. The type of expertise they the development of exceptional research
come with, and how they arrived at it, may neurologists (in the Bloom studies) did seem
be as important in defining the mechanisms to be different in important ways from, say,
of developing exceptional talent as the learn- the concert pianists experiences.
ers experiences with the master teachers. One further related finding that appears
Discussion about the long-term nature of again and again across studies but may be
developing talent, and the phases that may an artifact of the domain or the larger cul-
or may not be important aspects of that ture has to do with the importance of family
retrospective interviews 2 99
influences on providing early experiences and studies across a greater range of domains.
motivating learning over the long term. It Albert (1983 ) called attention to this last
seems reasonable to wonder if, in fields for issue some time ago. What do we know
which there is a strong societal press (cer- about the long-term development of exper-
tain sports, perhaps, or maybe even televi- tise in business, finance, law, public service,
sion news or entertainment reporting), the or social service? What do we know about
societal press may well have more power the long-term development of expertise as
than anything parents might do or say. Or in playwrights or poets, teachers or preachers,
circumstances where young children spend engineers or architects or statisticians?
more time outside of the home than with
Since what we know obviously depends
parents, people outside the home might
upon who is available for study as well as
prove more influential than parents. It might our techniques, interests, and values, this
be wise to think about family influences as raises the possibility that the model we con-
a proxy measure for activities that dominate struct concerning creativity and eminence
the lives of young children and for the ways may be of limited generality. (Albert, 1983 ,
children experience early activities. p. viii)
In this regard it may become as impor-
Ultimately, where we might head in
tant to know about the society in which the
future research will depend a great deal on
development of expertise takes place or
our intentions and our theoretical orienta-
maybe especially the local community for
tions. Researchers interested in the long-
the young child as to know about the home
term development of expertise likely will
in which a learner may get an initial start.
choose different domains to study and cer-
Scholars like Feldman (1986) and Csikszent-
tainly will choose different methods of inves-
mihalyi and Robinson (1986), arguing the
tigation than will researchers interested in
case for the coincidence or co-construction
the structure and characteristics of expert
of expertise, point us in important directions
performance. Absent an interest in long-
for future research. For educators, especially,
term development (or at least aspects of the
separating the family from the knowledge,
long-term experience), retrospective inter-
skills, and dispositions that might develop
views will not be the method of choice. It
early, and attending more to the youngsters
will become crucial, then, to ensure repeated
communities, would help us think about
syntheses of studies across domains and
new ways of helping more people reach
across methods of investigation in order that
greater levels of expertise in more fields. It
we do not end up following a trail that was
would do this by reducing the importance
created artificially by the fields we chose to
of being born into the right household and
study and the methods we chose to use for
increasing the likelihood of choice and fit
our studies.
for children and the areas in which they
I am hardly impartial in the discussion
might go on to exceptional accomplishment.
of what fields to study. As an educator
Again, these are testable questions that call
I resonate with Slobodas (1991) argument
for studies of talent development under dif-
regarding the purpose for our studies:
ferent conditions.
One of the principal reasons for studying
expertise is practical. Given that it would
Further Challenges in Thinking about be socially desirable for certain manifesta-
Future Research tions of expertise to be more widespread
than they are, we want to know what we
can do to assist people to acquire them.
For a great many reasons, then, it seems
(p. 15 6)
important that we conduct studies of the
long-term development of expertise across To this end, I am interested not only in the
a greater range of conditions and cultures. long-term experience of learning but also
It seems important, also, that we conduct in learning in domains that have particular
300 the cambridge handbook of expertise and expert performance
social value. Given the enormous investment view studies will allow us to frame and test
an individual must make in the development meaningful opportunities for advancing the
of expertise, and the considerable invest- development of talent, however far, for ever-
ment others make in the success of a sin- expanding numbers of individuals and, of
gle individual, I would argue for keeping our course, for society.
eye on fields in which expertise could serve
simultaneously the individual and society.
Finally, I would like to suggest that an
important challenge we face as we continue References
studying expertise, especially the develop-
ment of expertise over the long term, is Albert, R. S. (Ed.) (1983 ). Genius and eminence:
to beware of labeling levels of development The social psychology of creativity and exceptional
below the ultimate as failures. Students who achievement. Oxford: Pergamon Press.
at some point in their development aban- Bamberger, J. (1982). Growing up prodigies:
don a field in which they may have been The midlife crisis. In D. Feldman (Ed.), New
directions for child development: Developmental
selected as a potential elite or may have
approaches to giftedness and creativity, no. 17.
already demonstrated considerable expertise
San Francisco: Jossey-Bass.
are not abandoning their talent. They may
Bereiter, C. & Scardamalia, M. (1986). Educa-
be choosing to devote their energies in other
tional relevance of the study of expertise. Inter-
directions, but the talent they have devel- change, 17 (2), 1019.
oped already does not disappear. It is not
Bloom, B. S. (1982, April). The role of gifts and
lost and it is not wasted. In choosing to move markers in the development of talent. Excep-
in other directions, the individuals still carry tional Children, 48(6), 5 105 21.
with them both what they have learned and Bloom, B. S. (Ed.) (1985 ). Developing talent in
the experience of learning. And the ways young people. New York: Ballantine.
they may use their knowledge and experi- Carlson, B. (1985 , Fall). Exceptional conditions,
ence of learning create various other avenues not exceptional talent, produce high achievers.
of investigation, depending on the theoreti- The University of Chicago Magazine, 78(1), 18
cal orientation of a researcher. 19, 49.
Expertise is not an endpoint, it is a con- Csikszentmihalyi, M. & Robinson, R. E. (1986).
tinuum. Although retrospective interviews Culture, time, and the development of talent.
may concentrate on the ultimate expression In R. J. Sternberg & J. E. Davidson (Eds.), Con-
of expertise because this allows us to study ceptions of giftedness (pp. 264284). Cambridge:
expertise most clearly, and although some Cambridge University Press.
scholars studying expertise might be inter- Ericsson, K. A. & Charness, N. (1994). Expert
ested only in that ultimate demonstration, performance: Its structure and acquisition.
we must remember not to devalue the many American Psychologist, 49(8), 725 747.
other levels of expertise that serve individ- Ericsson, K. A., Krampe, R. Th., & Tesch-Romer,
uals and society well. There are limits on C. (1993 ). The role of deliberate practice in the
the number of people who can enter an acquisition of expert performance. Psychologi-
cal Review, 100, 3 63 406.
Olympic contest or who can be considered
the very most talented of living musicians. Ericsson, K. A., Tesch-Romer, C., & Krampe,
R. Th. (1990). The role of practice and moti-
But there are no limits on the number of
vation in the acquisition of expert-level perfor-
people who can value music and sports and
mance in real life: An empirical evaluation of a
so many other avenues for expression in their theoretical framework. In M. J. A. Howe (Ed.),
lives, who can value the opportunity to con- Encouraging the development of exceptional skills
tinue advancing their own personal bests, and talents (pp. 10913 0). Leicester, UK: The
and who can make significant contributions British Psychological Society.
to their communities through their consider- Feldman, D. H. (with Goldsmith, L. T.) (1986).
able, although perhaps not exceptional, tal- Natures gambit: Child prodigies and the devel-
ent. In the best of worlds, retrospective inter- opment of human potential. NY: Basic Books.
retrospective interviews 301
Freeman, J. (2000). Teaching for talent: Lessons Rubin, D. C. & Schulkind, M. D. (1997). Distri-
from the research. In C. F. M. van Lieshout & bution of important and word-cued autobio-
P. G. Heymans (Eds.), Developing talent across graphical memories in 20-, 3 5 -, and 70-year-
the life span (pp. 23 1248). East Sussex: Psy- old adults. Psychology and aging, 12 (3 ), 5 24
chology Press. 535.
Glaser, R. & Chi, M. T. H. (1988). Overview. In Salthouse, T. A. (1991). Expertise as the circum-
M. T. H. Chi, R. Glaser, & M. J. Farr (Eds.), The vention of human processing limitations. In
nature of expertise (pp. xvxxvii). Hillsdale, NJ: K. A. Ericsson & J. Smith (Eds.), Toward a
Erlbaum. general theory of expertise: Prospects and limits
Goldsmith, L. T. (2000). Tracking trajectories (pp. 2863 00). Cambridge: Cambridge Uni-
of talent: Child prodigies growing up. In versity Press.
R. C. Friedman & B. M. Shore (Eds.), Talents Simon, H. A. & Chase, W. G. (1973 ). Skill in
unfolding: Cognition and development (pp. 89 chess. American Scientist, 61, 3 94403 .
122). Washington, DC: American Psychologi- Sloboda, J. (1991). Musical expertise. In K. A.
cal Association. Ericsson & J. Smith (Eds.), Toward a general
Gruber, H. E. (1982). On the hypothesized theory of expertise: Prospects and limits (pp. 15 3
relation between giftedness and creativity. In 171). Cambridge: Cambridge University Press.
D. H. Feldman (Ed.), Developmental approaches Sosniak, L. A. (1987). The nature of change
to giftedness and creativity (pp. 729). San in successful learning. Teachers College Record,
Francisco: Jossey-Bass. 88(4), 5 195 3 5 .
Gruber, H. E. (1986). The self-construction Sosniak, L. A. (2003 ). Developing talent: Time,
of the extraordinary. In R. J. Sternberg & task and context. In N. Colangelo & G. Davis
J. E. Davidson (Eds.), Conceptions of gifted- (Eds.), Handbook of gifted education (3 rd ed.)
ness (pp. 247263 ). Cambridge: Cambridge (pp. 24725 3 ). New York: Allyn & Bacon.
University Press. Subotnik, R., Kassan, L., Summers, E., and
Ochse, R. (1990). Before the gates of excellence: Wasser, A. (1993 ). Genius revisited: High IQ
The determinants of creative genius. Cambridge: children grown up. Norwood, NJ: Ablex.
Cambridge University Press. Terman, L. M., & Oden, M. H. (195 9). Genetic
Posner, M. I. (1988). Introduction: What is it to be studies of genius: The gifted group at mid-life.
an expert? In M. T. H. Chi, R. Glaser, & M. J. Stanford, CA: Stanford University Press.
Farr (Eds.), The nature of expertise (pp. xxix Wagner, R. K. & Stanovich, K. E. (1996). Exper-
xxxvi). Hillsdale, NJ: Erlbaum. tise in reading. In K. A. Ericsson (Ed.), The
Roe, A. (195 2). The making of a scientist. NY: road to excellence: The acquisition of expert perfor-
Dodd, Mead & Company. mance in the arts and sciences, sports, and games
Rubin, D. C. (Ed.) (1986). Autobiographical (pp. 189225 ). Mahwah, NJ: Erlbaum.
memory. Cambridge: Cambridge University Whitehead, A. N. (1929). The aims of education.
Press. New York: Macmillan.
Rubin, D. C. (Ed.) (1996). Remembering our Zuckerman, H. (1977/1996). Scientific elite: Nobel
past: Studies in autobiographical memory. laureates in the United States. New Brunswick:
Cambridge: Cambridge University Press. Transaction Publishers.
C H A P T E R 17
activities and will result in the coding of dif- the examination of leisure time and activi-
ferent episodes. The decision regarding the ties. As the foregoing suggests, such studies
level of analysis that one wants to investigate can be used to study both activities and pop-
will determine a method of data collection ulation groups.
that would be most appropriate. The diary The Multinational Time Use Study in
method may be more appropriate to investi- the mid-1960s directed by Alexander Szalai
gate time-use data at a macro level, whereas (1972) stands as a landmark in cross-national
systematic observation may be more appro- survey research and was unquestionably the
priate for the micro analysis of selected most significant time-diary undertaking in
activities. After providing an historical per- the last century. The study examined the
spective of time-use research and time- use of quantitative political, social, and cul-
budget methods used in expertise research, tural data from thirteen countries and six-
this chapter will focus on the use of diary teen different survey sites. This study had
and observation methods as they relate to a profound effect on all subsequent collec-
the investigation of expert performance. tion of time-use data. Specifically, the cod-
ing scheme used in that study shaped those
of all subsequent national time studies, and
Historical Perspective the wide range of analyses using the Multi-
national Study data has broadened the scope
Two of the earliest published accounts of of data collectors and data analysts.
time use are How Working Men Spend Their The United States, which had never
Time (Bevans, 1913 ) and Round about a Pound collected official national time-use data,
a Week (Pember-Reeves, 1913 ). Time-use launched, through the Bureau of Labor
research emerged in Europe in conjunction Statistics, an ongoing study in January of
with early studies of living conditions of the 2003 . The major national studies in the
working class in response to pressures gen- United States have been undertaken by the
erated by the rise of industrialization. Stud- Institute of Social Research (ISR) at the
ies examined activities such as paid work, University of Michigan (Juster, 1985 ) and
household work, personal care, leisure, and by the Survey Research Center at the Uni-
so on, in the daily, weekly, or yearly time versity of Maryland (Robinson & Godbey,
budget of the population. They also exam- 1997). The first general population survey
ined how the time use varied among pop- conducted in Canada was undertaken in
ulation groups such as workers, students, Halifax, Nova Scotia in 1971. The Halifax
and housewives, and in the use of leisure study was a time-space study that captured
time. not only what people were doing, but also
Most pre-World War II diaries originated where they were, coded to a one-tenth kilo-
in the Soviet Union, Great Britain, and the meter grid (Elliott et al., 1976). The fist
United States. One of the earliest land- nationwide time use study in Canada was
mark studies was undertaken in the Soviet conducted in 1981(Kinsley & ODonnell,
Union (1924) by S. G. Strumilin (1980) for 1983 ). As a part of that study, over 45 0
use in governmental and communal plan- respondents to the 1971 Halifax study com-
ning. In the United States, home economists pleted diaries, thus providing a ten-year
started using time-use studies in the 1920s to panel of time use (Harvey & Elliott, 1983 ).
study farm and rural women (Avery, Bryant, Statistics Canada, as part of its General
Douthitt, & McCullough, 1996; Kneeland, Social Survey program, collected diaries for
1929). Work began at Cornell during the approximately 9,000 Canadians in 1986,
1920s on a program to study household 1992, and 1998 (Fredrick, 1995 ; Harvey,
output in terms of time use (Walker & Marshall, & Frederick, 1991).
Woods, 1976). In the 193 0s Lundberg and In the last decade, there have been signif-
Komarovsky (193 4) launched a new era in icant advances in time-use methodologies,
time budgets, diaries, and analyses of concurrent practice activities 305
including innovative applications to non- expert performance has given way to the per-
traditional topics of inquiry, and a variety vasive belief that practice and other forms
of analysis strategies. The literature on time of preparation are essential prerequisites for
use have been remarkable in reflecting the the development of expertise. Understand-
interests of many different fields, inclu- ing practice as a major independent vari-
ding economics (Juster & Stafford, 1991; able in the acquisition of skill requires a
Goldschmidt-Clermont, 1987), business ad- concomitant understanding of time use in
ministration (Das, 1991; Grossin, 1993 a, that process.
1993 b; McGrath & Rotchford, 1983 ), geron- The study of expert performers affords
tology (Harvey & Singleton, 1989, Moss us the opportunity to determine the charac-
& Lawton, 1982), urban planning (Chapin, teristics that underlie their superior perfor-
1974; Gutenschwager, 1973 ), political sci- mance and examine the process by which
ence and occupational therapy (Larson, those characteristics of performance are
1990; McKinnon, 1992; Pentland, Harvey, acquired. The uniqueness of this opportu-
& Walker, 1998), nursing and medicine nity was eloquently stated by Starkes (2003 )
(Frankenberg, 1992), recreation and phys- when she noted that few human endeavors
ical and health education (Rosenthal & exist to which people dedicate so much time,
Howe, 1984; Ujimoto, 1985 ), sociology and energy, resources and effort all with the
anthropology (Andorka, 1987; Elchardus & goal of becoming quite simply the best they
Glorieux, 1993 , 1994; Garhammer, 1995 ), can be. Although she was speaking specifi-
psychology (Block, 1990; Lawton, Moss, & cally of sport-related expertise, these charac-
Fulcomer, 1987), and expert performance teristics are uniformly applicable to experts
(Ericsson, Krampe, & Tesch-Romer, 1993 ; across a vast variety of domains, includ-
Deakin & Cobley, 2003 ; Starkes, Deakin, ing chess (de Groot, 1946/1978; Chase &
Allard, Hodges, & Hayes, 1996). Simon, 1973 ; Simon & Chase, 1973 ), physics
Time-use methodologies provide hard, (Chi, Glaser, & Rees, 1982), music (Ericsson
replicable data that are the behavioral et al.,1993 ), and sports (Deakin & Cobley,
output of decisions, preferences, attitudes, 2003 ; Starkes et al. 1996).
and environmental factors. It can be used Although time use per se was not the
to examine, describe, and compare role per- focus of the classic work on chess expertise
formance (Ross, 1990), cultures and life- (Chase & Simon, 1973 ), it was heavily impli-
styles (Chapin, 1974), poverty (Douthitt, cated in their hypothesis that chess-specific
1993 ), household and community eco- pattern-recognition processes underlay the
nomies (Knights & Odih, 1995 ), as well superior memory recall performance of the
as social indicators such as quality of life Grand Master player. They claimed that
and well-being (Japan, Ministry of Economic a minimum of ten years of preparation is
Planning, 1975 ). required to develop and organize the nec-
essary repertoire of domain-specific infor-
mation to attain an international level of
Time Use and Expertise chess skill. Further, they suggested that a
similar timeframe would be required in
Time and its use are central to contem- other domains. Subsequent studies crossing
porary discussions on the acquisition and many domains have substantiated this claim
retention of expert performance. Early attri- (e.g., Bloom, 1985 ; Ericsson et al., 1993 ;
butions of superior performance to a set Hayes, 1981; Helsen, Starkes, & Hodges,
of general inherited capacities (e.g., Galton, 1998; Krogius, 1976; Starkes et al., 1996).
1869; Cattell & Drevdahl, 195 5 ) diminished However, specific relationships between
the need to evaluate the contribution of the experience with activities in a domain and
use of time to the attainment of eminence. acquired performance have been uniformly
This predominantly genetic explanation of weak (Ericsson et al., 1993 ). This suggests
306 the cambridge handbook of expertise and expert performance
that maximal levels of performance are not estimates as a way of validating the retro-
attained merely as a function of extended spective data.
experience, but rather by deliberate effo- Ericsson et al. (1993 ) concluded that
rts to improve, further implicating the solitary practice was the only activity that
importance of elucidating the compo- met their definition of deliberate practice.
nents of domain-related experience that Although both the best and good violin-
are critical to performance improvement ists spent a similar amount of time engaged
over time. in music-related activities, the only distin-
Ericsson and colleagues (1993 ) presented guishing activity was solitary practice, where
a theoretical framework for the acquisition the best violinists spent four hours per day
of expert performance that implicated delib- each day of the week. When they consid-
erate practice a highly structured set of ered next the estimates of time spent in soli-
activities, with the explicit goal of improv- tary practice since the beginning of involve-
ing performance (p. 3 68) as the central ment from the retrospective recall data, they
factor in the determination of acquired per- reported that by age eighteen, the best vio-
formance. They defined deliberate practice linists had accumulated an average of about
as specific tasks rated very high on rele- 7,5 00 hours of practice, which was reli-
vance for performance, high on effort and ably different from the good violinists, with
comparatively low on inherent enjoyment 5 ,3 00 hours of practice. Both of these
(p. 3 73 ). Their proposition that the expert groups had accumulated more hours than
could be differentiated from both less expert the music-teacher group, who had accumu-
and novice performers solely on the extent lated 3 ,400 hours of practice at the same age.
of their involvement in deliberate practice They concluded that the differential skill
formed the basis of an elegant series of stud- levels seen between these groups of violin-
ies undertaken with three groups of violin ists could be accounted for by the accumu-
students (best, good, and music-teacher can- lated hours of solitary practice. Though the
didates) from the Music Academy of West relationship between the amount of solitary
Berlin. Ericsson et al. (1993 ) collected ret- practice and acquired performance has been
rospective reports on when the musicians demonstrated in chess (Charness, Krampe
first became involved with the violin, the & Mayr, 1996) and music (Ericsson et al.,
beginning of practice with the instrument, 1993 ), it has been more problematic to iden-
the number of music teachers they had stud- tify domain-related activities that meet the
ied under, and their involvement in competi- definition of deliberate practice.
tions. Participants were asked to estimate the Retrospective reports and diaries of time
number of hours per week they had devoted use have been used extensively to exam-
to practicing alone with the violin for each ine the development of expert performance
year since they started to practice. The par- in sport. Starkes et al. (1996) first used
ticipants also completed daily diaries for a the methodologies to examine the prac-
seven-day period, using ninety-six fifteen- tice activities of skilled wrestlers and figure
minute intervals. All activities were then skaters with a view to validating the def-
encoded by the participant using a thirty- inition of deliberate practice in the sport
item preestablished taxonomy. The purpose environment, and to determine whether
of the diary work was multidimensional. time spent in deliberate practice was mono-
First, use of the taxonomy facilitated the cal- tonically related to acquired performance.
culation of total time spent in any one cate- Similar to Ericsson et al.s (1993 ) protocol,
gory of activity by addition across days. Sec- a taxonomy of sport-related and everyday
ond, it enabled the concurrent rating of each activities was compiled by asking partici-
activity category on the defining attributes of pants to think back and report what activities
deliberate practice (relevance, effort, enjoy- they had participated in during their most
ment). Finally, the diary data could be con- recent typical week. They then kept a diary
trasted with the retrospective data on time for seven days, with each day divided into
time budgets, diaries, and analyses of concurrent practice activities 307
fifteen-minute intervals. Daily entries were their musicians in that they too overesti-
made at the end of each day, and all activ- mated practice hours, suggesting that retro-
ities were encoded using the sport-specific spective estimates reflected the amount of
taxonomy. practice participants aspire to, rather than
On the issue of identifying practice activi- what was actually attained. Although the
ties that met the requirement of being rated international wrestlers overestimated the
as highly relevant but not enjoyable, none amount of time they spent in some prac-
were found. For both the wrestlers and figure tice activities involving others, there was no
skaters, all of the activities that were rated difference in predicted time spent on other
highest for relevance were also rated high for elements of practice. For example, correla-
enjoyment. The second requirement being tions between the retrospective recall and
high on relevance and high on effort was diary data for the international wrestlers on
apparent in both sets of data, with the top strength training with others (r = 0.98),
two activities for relevance also being rated strength training alone (r = 0.96), and time
high on concentration. Although, strictly spent attending wrestling practice (r = 0.76)
speaking, no activities in these studies met were uniformly high and superior to those of
the definition of deliberate practice, the the club-level wrestlers. Although these cor-
strong positive relationship between rele- relations support the validity of both time-
vance and concentration seen across domains use methodologies, diary data allows for an
was evident in these data. Interestingly, if accurate assessment of time use in activities
the high-relevance/ high-effort definition of on a finer level of detail than does retrospec-
deliberate practice is used, the top two tive reports.
deliberate-practice activities for each of the The question of differential accuracy in
violinists, wrestlers, and figure skaters are the estimation of time use across skill level
perfectly consistent and include an activity led Deakin and Cobley (2003 ) to include
that is identical to what must be done in the a separate evaluation of time use through
actual performance and work with a teacher direct observation of practice in figure skat-
or coach (Starkes et al., 1996). ing, in addition to the retrospective-recall
The availability of both retrospective time and diary techniques. They reasoned that the
estimates and diary data made it possible inclusion of an observational assessment of
to examine the reliability of these time- practice sessions would provide data on the
use methodologies. For those wrestlers who extent to which those elements of practice
took part in both the retrospective recall consistently rated as high on relevance and
and diary studies, time-use estimates from concentration or effort where represented
their most recent year (retrospective recall) in actual on-ice practice sessions. The study
were compared to the time-use attributions involved the recording of on-ice activities of
of the diary week. For international wrestlers three groups of skaters (n = 24) differing
the correlation between a typical week in skill level (elite or national team mem-
and the diary week was 0.66 for wrestling bers, competitive skaters, and test skaters;
related activities. Although their retrospec- see Deakin & Cobley, 2003 for details on
tive reports indicated that in a typical week participants and methodology). A log of the
they would have spent over seventeen hours practice activities including the number and
in practice with others, the data from time spent in jump attempts, spins, lessons,
the diary week indicated that they actu- program run-through, and rest time was
ally spent under eleven hours in that activ- recorded for each of three taped sessions per
ity. Similarly, the less-expert club wrestlers skater. In addition, a seven-day diary and a
reported spending the same amount of time series of questionnaires were completed to
on this activity in a typical week, but actu- provide estimates of time during a typical
ally recorded spending under ten hours in week spent in a variety of activities, includ-
this activity during the diary week. Ericsson ing skating-related activities, sleep, educa-
et al. (1993 ) reported similar findings for tion, and non-skating-related activities. Each
308 the cambridge handbook of expertise and expert performance
skater was asked to rate the activities on a ten more problematic for the interpretation of
point scale in terms of their relevance to skill retrospective reports than for diary data,
development. another example of the bias toward the over-
Of central interest was whether time use estimation of practice elements raises simi-
during practice reflected the relative impor- lar concerns for diary data. The relationship
tance assessed to the element by the skaters, between what skaters say they practice and
and whether the estimates of time use pro- what they actually practice was examined
vided for a typical week were consistent by asking each skater, immediately prior to
with what was observed in the on-ice ses- beginning their practice session, the num-
sions. The practicing of jumps and spins ber and type of jumps and spins they were
were rated as being highly relevant to perfor- going to undertake in that particular session.
mance improvement by all groups of figure Without exception skaters overestimated
skaters, and all groups reported spending the the number and difficulty-level of the jumps
highest proportion of their practice time on they undertook in practice. The examination
these elements. However, despite the lack revealed that all skaters spent considerably
of group differences in the assessment of more time practicing jumps that existed in
relevance to improvement, the time-motion their repertoire and less time on jumps they
analyses revealed that the elite and compet- were attempting to learn. Whereas it appears
itive skaters spent 68% and 5 9% of their they aspired to work on increasingly diffi-
sessions practicing jumps, whereas the test cult elements, they instead opted to execute
group was engaged in those activities for elements that required less effort on their
only 48% of their on-ice time. Further, when part for successful completion. For exam-
rest time was expressed as a percentage of ple, the elite skaters estimated attempting
total session length by group, an inverse rela- seven double jumps and twenty triple jumps
tionship with skill level emerged. Specifi- per session, whereas they actually attempted
cally, the elite group spent an average of 14% an average of thirty doubles and six triples.
of their total on-ice practice time on rest, This difference amounts to a three- to four-
the competitive group spent 3 1%, and the fold discrepancy between what they intend
test skaters, 46%. The analysis of rest time to do and what they are observed to do. The
indicated that the elite skaters utilized their results of our investigation corroborate those
on-ice time more effectively than the other of Ericsson et al. (1993 ) on this matter and
groups by practicing the critical elements for highlight the question of the extent to which
a higher proportion of their on-ice practice diary data on practice activities themselves
time and by resting less. may be influenced by aspired versus actual
Inclusion of an observational time-motion practice time.
analysis of practice in figure skating has pro- The assessment of time use across con-
vided another dimension to the determina- current activities by both direct observa-
tion of time use and to the applicability tion and diary methodologies should elu-
of retrospective estimation of the quantity cidate the complex relationships between
of practice. Any interpretation of skaters time, activities, and the attainment of expert
reports of cumulative practice must be made performance. In the following section we
cautiously. Despite skaters ability to accu- provide specific details of diary and obser-
rately recall scheduled hours of practice, the vational methodologies for the evaluation of
relationship between scheduled hours and how time is spent.
actual hours of practice is far from clear.
Although the three groups of skaters in this
study had spent a similar number of years A Macro Analysis of Time Use:
practicing, the actual active practice time The Diary Method
would be in the order of 13% to 46% lower
than the reported hours of scheduled prac- The time diary provides one of the most
tice. Although at first glance this might seem comprehensive and accurate means of
time budgets, diaries, and analyses of concurrent practice activities 309
collecting data about time spent on specific episodes, what activities are done, and when.
categories of activities. The diary captures Thus, it is possible to derive an understand-
the sequence and duration of activities ing of how an individuals day is organized.
engaged in by an individual over a speci- Forms of the stylized activity log have been
fied period. All activities during the period used in the examination of expert perfor-
are recorded in sequence, including their mance to investigate how expert performers
start and completion times. One of the key maximize the quality of their multiple prac-
strengths of the diary methodology is that it tice sessions through managing their time in
can exhaust all activities and all time during daily cycles (see Starkes et al., 1996).
the selected period. A broad range of subjec- Time-use studies report the state of se-
tive and contextual data can also be collected lected details at each successive time point
at the same time, providing rich context for or period. The basic unit in a time-use study
the individuals, the activities, and individual is the episode, a single diary entry with
behaviour. all attendant dimensions sought by the
Diaries may be presented in a num- researcher. Although episodes are meaning-
ber of ways, each with different implica- ful for analysis at one level, they are less
tions for implementation and data genera- useful at another. One may be interested
tion. The simplest is a Stylized Activity List. in each episode of lessons, jump attempts,
It provides an abbreviated list of activities. or program run-throughs during an on-ice
Respondents are then asked to estimate how training session for expert figure skaters. This
much time they spent on each, say, dur- would be accompanied by information on
ing the previous day or week. Two partic- when, where, and with whom the practice
ular problems are inherent in this approach. occurred, as well as the overall allocation
First, responses are heavily dependent on of time to skating over the study period.
the respondents ability to recall and accu- Thus, it is necessary to aggregate identical or
mulate time on each activity listed with similar episodes during the study period of
essentially no external clues. Second, gen- interest into higher-level categories for more
erally, by using selected activities it lacks the aggregated analysis.
period of time constraint that aids reporting The focal aggregation is typically an activ-
accuracy. However, such an instrument can ity. What are relevant activities? How are
get participation and duration for the listed they organized? Episode and activity orga-
activities. nization is spelled out in a coding scheme.
A Stylized Activity Log improves on the All such schemes have an implicit, if not
above by capturing episodes, the basic build- explicit, theoretical or heuristic base. At the
ing blocks of time use. Sometimes known most fundamental level interest centres
as the quick diary, the log typically provides on the actual amount of time allocated to
a vertical activity list and a horizontal time specific activities, such as those related to
referent covering the twenty-four hours of the domain under investigation (practice
the day, broken into ten- to thirty-minute alone, lessons with a coach or teacher),
units. Respondents can then indicate their those related to the demands of daily living
activities over the day by drawing a line (paid work, housework, childcare, educa-
beside the activity and under the relevant tion, rest), and other activities meaningful
time period. This approach provides much to the particular interests being examined.
enhanced control over the quality of the The activity list used in a time-use study
time-use data since it requires the respon- can range from a few to hundreds of
dent to think through the day and to identify activities. The greater detail provides for an
transitions from activity to activity. The styl- elaboration of activities, such as the type of
ized activity log is capable of providing sig- practice undertaken, work done, television
nificantly more information on time use than show watched, the type of book read, and so
is provided by the stylized activity list. As on. Although on the surface more detail is
indicated, it provides information on activity better, this is true only if there are sufficient
310 the cambridge handbook of expertise and expert performance
episodes of an identified activity for analysis. as a vector that may or may not be of fixed
Thus, there are tradeoffs between detail and length, dependent on the number of charac-
usability. teristics collected with a diary entry. Typ-
A diary survey typically consists of two ically, studies collect what is being done,
parts. One is socio-demographic and in- where, with whom, and sometimes what else
cludes other information useful in classify- is being done at the time (secondary/parallel
ing and interpreting the material collected. activity). A sample of a partial diary consist-
This information can be treated as a fixed- ing of a time period from 7 am to 9 pm in
length vector in a manner similar to that one hour blocks is presented in Figure 17.1.
done in ordinary surveys. The second, time It provides only one example of the types
diaries, collect information on how time is of information that can be surveyed using a
being used and is most frequently collected diary technique.
time budgets, diaries, and analyses of concurrent practice activities 311
as does the activity. Ideally, data should tion than another, and how much effort is
be collected reflecting social contact with required in order to complete certain activ-
at least the following categories: 1) alone, ities relative to others?
2) coach/instructor, 3 ) training partners, The diary method allows researchers to
4) other participants, 5 ) parents, 6) siblings, collect multi-dimensional data related to the
7) friends/relatives outside household, and use of time. At a macro level, diaries provide
8) others outside the household. However, a general picture of how experts conduct
there is wide variety in the detail used to cap- their daily activities; however, it provides lit-
ture with whom. One important aspect of tle information about the structure and con-
the with whom coding is that a distinction text of a specific activity. For example, two
must be made during capture and/or coding people can engage in the same activity for
between being in the presence of and act- the same amount of time, but the way they
ing with. The data should at a minimum experience the activity can be very differ-
permit the researcher to identify the time ent and ultimately produce a different out-
that the respondent was acting with oth- come. Systematic observation of experts and
ers. It is suggested that respondent instruc- non-experts in various structured settings
tion is important for accurate reporting of can provide a wealth of information about
with whom. how teachers or coaches structure an activity
and how participants experience the activ-
duration ity. Demarco, Mancini, Wuest, and Schempp
At what time does an activity start and end? (1996) suggested that various descriptive-
At any given time (morning, 12 noon, 5 pm), analytic instruments of behavior have made
what is being done? Many differences significant contributions to the quality of
can occur regarding when and how long teaching and coaching in education, physi-
an activity is performed, and these can cal education, and sport. Although system-
have important consequences on expertise atic observation methods have not been used
development. extensively in expertise research, it is a crit-
ical part of any study interested in human
behaviors and issues that revolve around the
location
use of time. The possibility of using system-
Though not central in the use of diary data atic observation in expertise research is lim-
or the examination of the development of ited by the fact that observation can occur
expert performance, the collection of loca- only in particular physical settings where tar-
tion data in general provides valuable infor- geted behaviors can be formalized, observed,
mation for use in coding the diaries, as well as and analyzed.
providing important analytical information.
A strong case can be made for the collec-
tion of geographic location information, if
its collection is feasible (Elliott, Harvey, & Micro-Analysis of Time Use:
Procos, 1976). Systematic Observation
they occur. This section will focus on two place. For example, an expertise researcher
issues related to the use of systematic obser- could observe the settings of various prac-
vation in expertise research: 1) what should tice activities and the physical facilities
be observed, and 2) how should behavior be that could constrain or enhance expertise
observed? development.
The first decision to make in a systematic- The most probable useful focus of exper-
observation study involves selecting the tar- tise research is on actors (i.e., performer
get information to observe. Johnson and or teacher). Once the actors have been
Sackett (1998) suggest three broad cate- selected, their actions and the setting in
gories of information common in activity which they practice or perform becomes
studies: 1) actors (participants), 2) actions the content of the observation. More rarely,
(activities), and 3 ) settings (location). The the focus of the observation would be on a
actors include the subject(s) of interest and specific action or setting. The selection of
other participants with whom he or she actor(s), behavior(s), and setting(s) will be
interacts. In expertise research, this would directly linked to the research issue under
include making decision about observing investigation. If researchers are interested in
the performer(s), the instructor, teacher, or how performers spend their time in practice,
coach, or the interactions between these they could develop an observation code that
actors. would allow them to account for the various
Actions refer to the specific behaviors or activities that a performer can get involved
activities of interest. In an expertise study, in when training. Another researcher could
performers behaviors could range from spe- be interested in the behaviors of expert
cific practice activities, such as practicing a teachers and could use a behavior code that
new piece in music, to focusing on warm- would include categories of behaviors such
up, technical, or tactical activities in a sport as instruction, feedback, modeling, and so
practice. Specific actions of a teacher or forth. In sum, the selection of behaviors to
coach that could be linked to the develop- be recorded is dependent on the research
ment of expertise include providing tech- questions asked and the overall purpose of
nical instruction, strategies, feedback, and the study. Operationalization of the tar-
so forth. The theoretical or conceptual per- get behaviors will determine the quantita-
spective of the study will most often deter- tive measures to be gathered. Researchers
mine the choice of behavior(s) to record. (Foster, Laverty-Finch, Gizzo, Osantowski,
According to Hawkins (1982), behaviors that 1999; Hawkins, 1982) have suggested vari-
have the greatest functional validity should ous alternatives that can be summarized into
be the focus of observation. When observ- four main choices: 1) frequency, 2) duration,
ing experts, researchers have to make deci- 3 ) latency, and 4) quality measures.
sions about behaviors that truly perform a
function for the individuals development of
skills and performance. A useful approach
Frequency
to selecting behaviors that have high func-
tional validity is to conduct a task analy- When focusing on frequency of behaviors,
sis (Sulzer-Azaroff & Reese, 1982). In task the researcher records each instance of the
analysis, a performers practice activities or targeted behaviors. For instance, a researcher
performance would be broken down into may keep tallies of a performers practice of
behavioral components so that each behav- various skills or interaction with an instruc-
ior usually exhibited in a practice can be tor during practices. According to Foster
observed and assessed. et al. (1999), frequency counts are appropri-
Finally, the settings include the location ate when a) rates of behavior are important,
of action and the details of the physical b) the behavior occurs with a lot to medium
space in which the observation is taking frequency, and c) the target behavior is a
314 the cambridge handbook of expertise and expert performance
discrete event with an easily identifiable ings it is important to provide the perform-
beginning and end (p. 427). ers with clear anchors that trigger a mem-
orable past experience that in turn acts as
Duration reference point (Cote et al., in press; Foster
et al., 1999). For instance, when rating con-
When focusing on duration, the researcher centration, performers can be asked to iden-
records how long each instance of a targeted tify the most mentally demanding activity
behavior lasts using a stopwatch or a timer. they have ever been involved in during a
For example, a researcher may be interested practice and consider this level of concentra-
in recording the amount of time devoted to tion to correspond to 100% concentration.
the practice of skills in a specific environ- Then performers can be asked to identify
ment. Recording this dimension of a behav- an activity during practice where their con-
ior is useful when the amount of time each centration level had been non-existent or at
episode lasts is more important than the its lowest level to correspond to 0% con-
actual occurrence of the episode. centration. Using these two points of refer-
ence, performers can rate their concentra-
Latency tion level for all their observed behaviors
Response latency or temporal location refers during a given practice.
to time elapsed from the moment one In addition to determining the behaviors
has the opportunity to perform a behavior to be observed and the dimensions of behav-
until the same person actually executes the ior to be assessed, a researcher must select
behavior. For example, latency measures of a procedure by which behavior is recorded.
expertise could entail measuring the time This issue focuses on the how of systematic
period between the beginning of a practice observation.
and the actual involvement of the performer
in skill-development activities, or recording Coding Strategies
the time intervals between involvement in Whether observing experts behaviors live
effortful activities during practices. Latency during practices or performances or from
measures can provide expertise researchers a videotape, the researcher must choose a
with valuable information regarding the procedure for converting observed behav-
optimal structure of practices. iors or events into quantitative data. There
are numerous strategies by which behav-
Quality iors can be coded. Four procedures are
A specific behavior can be observed also in common use: event recording, duration
in terms of its quality or intensity. For recording, interval recording, and momen-
example, a specific practice behavior can be tary time sampling (Darst, Zakrajsek, &
assessed in terms of its effort, concentra- Mancini, 1989; Foster et al., 1999; Hawkins,
tion, or enjoyment. The challenge in assess- 1982).
ing more subjective dimensions of a behav-
ior, such as enjoyment and effort, involves event recording
finding observable indicators that can be This procedure consists of a tally of each
used as valid measures of these psycholog- occurrence of a defined behavior through-
ical states. For example, one can infer a out the observation session. Event record-
lower level of effort when observing ath- ing provides the researchers with data on
letes in practices that sit down to rest, joke the frequency of occurrence of a discrete
around, or observe others train (Cote et behavior. Behaviors that could be measured
al., in press). Alternatively, subjective rat- through event recording include the num-
ings can be used during practices or at the ber of times a performer has the opportunity
end of a practice to evaluate the quality of to do a specific drill or an activity during
specific activities. When using subjective rat- a practice, the number of times a teacher
time budgets, diaries, and analyses of concurrent practice activities 315
presents verbal instruction during a lesson, upon a signal that is emitted at constant time
or how often a performer arrives late to a intervals. Contrary to interval recording,
scheduled practice. Overall, event recording where observation starts at the beginning of
provides a numerical account of the occur- the interval and continues throughout the
rence of behaviors. entire interval, the observation in momen-
tary time sampling occurs at the end of each
duration recording interval. Darst et al. (1989) suggest using
This procedure is needed when the focus is intervals ranging from one to ten minutes.
on the amount of time spent in a particu- The length of the intervals depends on the
lar activity or on the response latency to a duration of the observation session and the
specific stimulus. This recording procedure number of sample behavior, needed. With
is useful, for example, to collect data on time momentary time sampling the researcher
spent by performers on relevant tasks during records the targeted behavior(s) when sig-
a practice, the time spent by a teacher or a naled. In other words, when receiving a sig-
coach to explain the technical aspect of a nal, the observer scans the observed set-
particular skill, or the amount of time spent ting and records the presence or absence
between involvements in relevant learning of an individuals behavior, or, if observing
activities during practice. In sum, duration a group of people, records the number of
recording provides a temporal account of the people engaged in the specified behavior.
observed behaviors. Momentary time sampling procedure could
be useful in expertise research to observe
interval recording categories of behaviors such as effort, par-
This procedure refers to observing behav- ticipation in an activity, or number of indi-
ior for short time periods (intervals) and viduals on a team or in a class that are
making a decision as to what behavior best involved in a given activity. Data recorded
characterizes that time period. Darst et al. from a momentary time sampling procedure
(1989) suggest interval lengths that usually are reported as percentage of total intervals
vary between six and thirty seconds. Interval or as percentage of individuals in a group
data provides neither frequency nor duration that are engaged in a specific behavior.
information; however, it can be used to esti- Observation techniques in expertise stud-
mate both. A problem with interval record- ies allow researchers to investigate practices
ing is when the observation system includes at a micro level of analysis that focuses
multiple categories and the intervals are so on temporal matters such as frequency and
long that several behaviors can occur within duration of specific behaviors. Questions rel-
the same interval; this situation forces the evant to expertise researchers may concern
researcher to make a decision as to which how often a teacher provides instruction
behavior should be recorded. On the other during a class (i.e., frequency), how long
hand, the advantage of this procedure is that before an athlete shows signed of tiredness
it can record behavior that occurs frequently during a practice (i.e., latency), how much
and has starting and stopping points that are concentration is required to learn a new
difficult to detect (e.g., social interactions, piece of music (i.e. quality), or how long
teacher/coach observing during a practice). a chess player spends between moves dur-
The data collected from interval recording ing a game (i.e., duration). The temporal
are expressed as a percentage of intervals in dimensions that are descriptive of an expert
which the behavior(s) occurred. learning environment can be uncovered with
proper observation methods. Nevertheless,
momentary time sampling it is important that expertise researchers tai-
This procedure is used to gather periodic lor their observation and quantification sys-
data on an individuals behavior or the tem to the unique properties and behav-
behavior of a group of people. With momen- iors of the expertise environment under
tary time sampling, the behavior is recorded examination.
316 the cambridge handbook of expertise and expert performance
Diary and systematic observation are mation processing (pp. 215 281). San Diego, CA:
complementary methodologies in the exam- Academic Press.
ination of expert performance. The data Chi, M. T. H., Glaser, R., & Rees, E. (1982). Exper-
each technique provides informs our under- tise in problem solving. In R. S. Sternberg (Ed.),
standing of both the macro- and microstruc- Advances in the psychology of human intelligence
ture of daily activities and their relevance (Vol. 1, pp. 175 ). Hillsdale, NJ: Erlbaum.
to the advancement of expertise. Further, Cote, J., Baker, J., & Abernethy, B. (2003 ). From
the multiple levels of analysis allow for play to practice: A developmental framework
direct and robust assessments of the valid- for the acquisition of expertise in team sports.
In J. Starkes & K. A. Ericsson (Eds.), Expert per-
ity and reliability of the diary data that
formance in sports: Advances in research on sport
relates specifically to the activities thought expertise (pp. 89110). Champaign, IL: Human
to be relevant to the development of Kinetics.
expertise. Cote, J., Ericsson, K. A., & Beamer, M. (in-
press). Tracing the development of athletes
using retrospective interview methods: A pro-
References
posed interview and validation procedure for
reported information. Journal of Applied Sport
Andorka, R. (1987). Time budgets and their uses. Psychology.
Annual Review of Sociology, 13 , 149164. Darst, P. W., Zakrajsek, D. B., Mancini, V. H.
Avery, R. J., Bryant, W. K., Douthitt, R. A., & (Eds) (1989). Analyzing physical education
McCullough, J. (1996). Lessons from the past. and sport instruction. Champaign, IL: Human
Directions for the future. In R. J. Avery (Ed.). Kinetics.
Household time use: Research in the 21st cen- Das, T. L. (1991). Time: The hidden dimension in
tury [Special Issue] Journal of Family and Eco- strategic planning. Long Range Planning, 2 4(3 ),
nomic Issues, 17(3 /4), 409418. 495 7.
Bevans, G. E. (1913 ). How working men spend Deakin, J. M., & Cobley, S. (2003 ). An exami-
their spare time. New York: Columbia Univer- nation of the practice environments in figure
sity Press. skating and volleyball: A search for deliber-
Block, R. A. (1990). Models of psychological ate practice. In J. Starkes & K. A. Ericsson
time. In R. A. Block (Ed.), Cognitive models (Eds.), Expert performance in sports: Advances in
of psychological time (pp. 13 5 ). Hillsdale, NJ: research on sport expertise (pp. 90113 ). Cham-
Erlbaum. paign, IL: Human Kinetics.
Bloom, B. S. (1985 ). Developing talent in young Demarco, G., Mancini, G. M., Wuest, V. H., &
people. New York: Ballintine Books. Schempp, P. G. (1996). Becoming reacquainted
Cattell J. M., & Drevdahl, J. E. (195 5 ). A com- with a once familiar and still valuable tool:
parison of the personality profile of eminent Systematic observation methodology revisited.
researchers with that of eminent teachers and International Journal of Physical Education, 3 2
administrators, and of the general population. (1), 1726.
British Journal of Psychology, 46, 248261. Douthitt, R. (1993 ). The inclusion of time
Chapin, R. S. (1974). Human activity patterns in availability in Canadian poverty measures.
cities: Things people do in time and in space. New In Instituto Nazionale di Statistica (ISTAT)
York: Wiley. (Ed.), Time use methodology: Toward consensus
Charness, N., Krampe, R. T., & Mayr, U. (pp. 8892). Rome: ISTAT.
(1996). The role of practice and coaching in Elchardus, M., & Glorieux, I. (1993 ). Towards
entrepreneurial skill domains: An international a semantic taxonomy: Classifying activities
comparison of life-span chess skill acquisition. on the basis of their meaning. In Instituto
In K. A. Ericsson (Ed.), The road to excellence: Nazionale di Statistica (ISTAT) (Ed.), Time use
The acquisition of expert performance in the arts methodology: Toward consensus (pp. 25 0276).
and sciences, sports, and games (pp. 5 180). Rome: ISTAT.
Mahwah, NJ: Erlbaum. Elchardus, M., & Glorieux, I. (1994). The search
Chase, W. G., & Simon, H. A. (1973 ). The minds for the invisible 8 hours: The gendered use
eye in chess. In W. G. Chase (Ed.), Visual infor- of time in a society with a high labour force
time budgets, diaries, and analyses of concurrent practice activities 317
participation of women. Time and Society, 3 (1), to study behavior (pp. 213 5 ). San Francisco:
5 27. Jossey-Bass.
Elliott, D. H., Harvey, A. S., & Procos, D. (1976). Hayes, J. R., (1981). The complete problem solver.
An overview of the Halifax time-budget study. Philadelphia, PA: Franklin Institute Press.
Society and Leisure, 3 , 145 15 9. Helsen, W. F., Starkes, J. L., & Hodges, N. J.
Ericsson, K. A., Krampe, R. T., & Tesch-Romer, (1998). Team sports and the theory of delib-
C. (1993 ). The role of deliberate practice in the erate practice. Journal of Sport and Exercise Psy-
acquisition of expert performance. Psychologi- chology, 2 0, 260279.
cal Review, 100, 3 63 406. Japan Ministry of Economic Planning (1975 ). An
Foster, S. L., Laverty-Finch, C., Gizzo, D. P., & analysis of structure of living time: The pattern of
Osantowski, J. (1999). Practical issues in self use of time and the quality of life. Tokyo: Authro
observation. Psychological Assessment, 11, 426 (in Japanese).
43 8. Johnson, A., & Sackett, R. (1998). Direct sys-
Frankenberg, R. (1992). Time, health and medicine. tematic observation of behavior. In H. R.
London: Sage. Bernard (Ed.), Handbook of methods in cultural
Fredrick, J. A. (1995 ). As time goes by, time use anthropology (pp. 3 013 3 2). Walnut Creek,
of Canadians: General social survey. Ottawa: CA: Sage.
Statistics Canada; Housing, Family and Social Juster, F. T. (1985 ). A note on recent changes
Statistics Division. in time use. In F. T. Juster & F. P. Stafford
Galton, F., Sir (1869). Hereditary genius: An (Eds.), Time, goods, and well-being (pp. 3 13
inquiry into its laws and consequences. London: 3 3 2). Michigan: Institute for Social Research.
Julian Friedman Publishers. Juster, F. T., & Stafford, F. P. (1991). Comment.
Garhammer, M. (1995 ). Changes in working Public Opinion Quarterly, 5 5 , 3 5 73 5 9.
hours in Germany: The resulting impact on Kinsley, B., & ODonnell, T. (1983 ). Marking time.
everyday life. Time and Society, 4, 167203 . Ottawa: Canada Employment and Immigra-
Goldschmidt-Clermont, L. (1982). Unpaid work tion Commission.
in the household: A review of economic evaluation Kneeland, H. (1929). Womens economic con-
methods. Geneva: International Labour Office. tribution in the home. Annals of the Ameri-
Grossin, W. (1993 a). Le temps de la vie quotidi- can Academy of Political and Social Science, 143 ,
enne. Paris: Mouton. 3 3 40.
Grossin, W. (1993 b). Technology evolution, Knights, D., & Odih, P. (1995 ). Its about time!:
working time and remuneration. Time and The significance of gendered time for financial
Society, 2 , 15 7167. services consumption. Time and Society, 4(2),
205 23 1.
Gutenschwager, G. A. (1973 ). The time budget:
Activity systems perspective in urban research Krogius, N. (1976). Psychology in chess. New York:
and planning. Journal of the American Institute RHM Press.
of Planners, 3 9(6), 3 783 87. Larson, K. B. (1990). Activity patterns and life
Harvey, A. S., & Elliott, D. H. (1983 ). Time and changes in people with depression. American
time again. In M. C. Casserly (Ed.) & B. L. Kins- Journal of Occupational Therapy, 44(10), 902
ley (Vol. Ed.), Explorations in time use series 906.
(Vol. 6, p. 88). Ottawa: Canada Employment Lawton, M. P., Moss, M., & Fulcomer, M. (1987).
and Immigration Commission. Objective and subjective uses of time by older
Harvey, A. S., Marshall, K., & Frederick, J. A. people. International Aging and Human Devel-
(1991). Where does time go? Ottawa: Statistics opment, 2 4(3 ), 171188.
Canada; Housing, Family and Social Statistics Lundberg, G. A., & Komarovsky, M. (193 4).
Division. Leisure: A suburban study. New York:
Harvey, A. S., & Singleton, S. (1989). Cana- Columbia University Press.
dian activity patterns across the life span: A McGrath, J. E., & Rotchford, N. L., (1983 ).
time budget perspective. Canadian Journal on Time and behaviour in organization. Research
Aging, 8(3 ), 268285 . in Organizational Behaviour, 5 , 5 7101.
Hawkins, R. P. (1982). Developing a behavior McKinnon, A. L. (1992). Time use for self-
code. In D. P. Hartmann (Ed.), Using observers care, productivity, and leisure among elderly
318 the cambridge handbook of expertise and expert performance
Historiometric Methods
ratio scale. In this sense, historiometrics does has obtained with respect to expertise acqui-
not differ from psychometrics except that sition and expert performance.
the measurement techniques are applied to
historical individuals. Second, the measure-
ments are subjected to statistical analyses History
using the full panoply of tools available for
drawing inferences from correlational data. Although historiometrics is not as well
These two features set historiometrics apart known as other techniques, it can be con-
from psychobiography and psychohistory, sidered the first scientific method that was
an approach to the psychological study of applied to the objective and quantitative
historical figures that entails qualitative study of expertise and expert performance.
rather than quantitative analysis (Elms, To be specific, the first bona fide histo-
1994; Runyan, 1982). riometric investigation was conducted by
3 . Nomothetic hypotheses concern general Adophe Quetelet back in 183 5 . Quetelet
laws or regularities of human behavior. For is best known for his pioneering applica-
example, a considerable research literature tions of statistics and probability theory to
has grown around whether expert perfor- social phenomena. Much of this work con-
mance in various domains is described by centrated on establishing the normal curve
a distinctive age curve (Simonton, 1988a). as descriptive of the distribution of human
Thus, in the case of creative expertise, traits around some central value represent-
some have argued for a monotonic func- ing the homme moyen (or average person).
tion like a learning curve (Ohlsson, 1992), Yet his empirical investigations were by no
whereas others have proposed a nonmono- means confined to establishing the ubiq-
tonic, single-peaked function with a mid- uity of this symmetric distribution. He was
career optimum (Lehman, 195 3 ; Simonton, also intrigued with the question of how
1997). This aspect of historiometrics con- creative productivity is distributed across
stitutes another characteristic that sepa- the career. To address this issue, he scruti-
rates this method from psychobiography and nized the lifetime output of eminent French
psychohistory. The latter approach favors and English dramatists. By tabulating their
the idiographic rather than the nomothetic, dramatic works into consecutive age peri-
that is, it attempts to explain the distinc- ods, he was able to discern the character-
tive attributes of eminent personalities. I istic longitudinal distribution. In addition,
should also point out that the historiometric Quetelet directly examined the empirical
emphasis on the nomothetic almost invari- relation between quantity and quality of cre-
ably requires that the hypotheses be tested ative output. Not only was he the first to
on large samples of historical individuals. apply quantitative and objective techniques
Only in this way can the investigator be con- to biographical and historical data, but he
fident that the findings extend beyond the also did so with a methodological sophisti-
idiosyncrasies of any single research subject. cation that was not to be surpassed for nearly
As a consequence, single-case or N = 1 stud- a century (Simonton, 1997).
ies are rare in historiometric research (but Unfortunately, Quetelets (183 5 ) histo-
see Simonton, 1989a, 1998a). riometric inquiry was buried in a larger
Now that the method has been defined, work dealing with different topics, and
I would like to accomplish three tasks in so this particular contribution was largely
the remainder of this chapter. First, I pro- ignored (Simonton, 1988a). Therefore, the
vide a brief history of the method. Second, first behavioral scientist to publish a truly
I offer an overview of the diverse method- influential historiometric study was not
ological issues involved in carrying out histo- Quetelet but rather Francis Galton. The spe-
riometric research. Third, I will review the cific work was Hereditary Genius: An Inquiry
main empirical findings that this approach into Its Laws and Consequences, which was
historiometric methods 32 1
published in 1869, albeit as a significant claimed that the approach has some spe-
expansion of an earlier historiometric study cial value for research on the psychology
published as an article four years earlier of genius. Somewhat surprisingly, Woodss
(Galton, 1865 ). The main goal of Galtons own historiometric inquiries seldom dealt
investigation was to establish a biological with this issue directly. Instead, his most
basis for natural ability, the capacity under- important publications using this method
lying exceptional achievement in all its were on the inheritance of intelligence and
forms, including creativity, leadership, and morality in royalty (Woods, 1906) and the
sports. To accomplish this task, he collected influence of monarchs on their nations wel-
extensive biographical data on the fam- fare (Woods, 1913 ). Hence, subsequent his-
ily pedigrees of eminent creators, leaders, toriometric research most germane to exper-
and athletes, and then subjected these data tise and expert performance was conducted
to quantitative analysis. Unlike Quetelets by others.
study, Galtons investigation had both a In this later work one investigation stands
short- and a long-term impact, thereby out well above the others: Catharine Coxs
becoming one of the classics in psychol- (1926) The Early Mental Traits of Three Hun-
ogy (Simonton, 2003 b). In the short term, dred Geniuses. This study forms the sec-
the work sparked a controversy among his ond volume of Termans (1925 195 9) clas-
contemporaries (e.g., Candolle, 1873 ) that sic work Genetic Studies of Genius. Although
led Galton (1874) to formulate the nature- the other four volumes concern a longitu-
nurture issue, one of the critical questions dinal study of over a thousand intellectu-
concerning the development of expertise ally gifted children, Coxs is retrospective.
(Ericsson, 1996; Simonton, 1999b). In the Rather than collect data on gifted children
long term, the work inspired other histori- and follow them into adulthood to see if they
ometric inquiries into the role of genetics displayed world-class expertise, Cox gath-
in exceptional achievement (e.g., Bramwell, ered a sample of unquestionable geniuses
1948). In addition, Galtons research pro- Napoleon, Luther, Newton, Descartes,
vided the impetus for James McKeen Voltaire, Michelangelo, Beethoven, and so
Cattells innovations in the area of the mea- Forth to determine whether they showed
surement of differential expertise, as gauged any signs of precocious intellect in child-
by achieved eminence in a domain (e.g., hood. After compiling a list of early intellec-
Cattell, 1903 ). tual achievements and applying the opera-
Despite the fact that the research con- tional definition of the intelligence quotient
ducted by Quetelet, Galton, and Cattell was as the ratio of mental to chronological age
clearly historiometric, it was not identified (times 100), she was able to obtain reason-
as such by these or any other researcher at ably reliable estimates of IQ scores for nearly
the time. Instead, the investigations were all those sampled. Significantly, Cox identi-
labeled as empirical, scientific, or quan- fied her study as an example of historiomet-
titative. The method was not actually given rics. Not only was it an example, but it soon
a formal name until 1909, when Woods became an exemplar of the technique.
(1909) published an article in Science on In fact, Coxs (1926) publication repre-
A New Name for a New Science. There sents the climax of the early period of his-
he defined the technique as encompassing toriometric research. Subsequent investiga-
those investigations in which the facts of tions were seldom as ambitious, and few
history of a personal nature have been sub- came anywhere close to the same level
jected to statistical analysis by some more of methodological sophistication (see, e.g.,
or less objective method (p. 703 ). This Raskin, 193 6). The only work to come close
definition was followed by a 1911 article in to the same level was Harvey Lehmans
the same journal on Historiometry as an (195 3 ) book Age and Achievement, which
Exact Science, (Woods, 1911) in which he dealt with the same issue first investigated by
32 2 the cambridge handbook of expertise and expert performance
Quetelet (183 5 ) over a century earlier (see be essentially equivalent to any other. His-
also Lehman, 195 8, 1960, 1962, 1963 , 1966a, toriometric research, in contrast, depends
1966b). Nonetheless, historiometrics under- on what has been called significant samples
went something of a revival in the 1960s (Simonton, 1999a). In this case, the indi-
and 1970s. As a consequence, the first book viduals in the studies have known identi-
summarizing historiometric findings with ties, and their identities are such that they
respect to genius, creativity, and leadership cannot be said to be interchangeable with
came out in 1984 (Simonton, 1984c), and in other participants. In particular, the partici-
1990 the first book totally devoted to expli- pants are persons who have made a name
cating the methodological issues entailed in for themselves by attaining eminence in
historiometric research appeared (Simon- some domain that presumes special exper-
ton, 1990). Other publications on historio- tise. Examples include famous or world-class
metrics have appeared in the Annual Review creators, leaders, athletes, and performers
of Psychology (Simonton, 2003 c) and Psy- (e.g., Elo, 1965 ; Oleszek, 1969; Schulz &
chological Methods (Simonton, 1999a), sug- Curnow, 1988; Simonton, 1975 a; Simonton,
gesting that the approach has become an 1977a; Zusne, 1976). Moreover, these lumi-
accepted, even if relatively rare, methodol- naries have attained sufficient distinction
ogy in psychological science. to have substantial information about them
readily available in archival sources, such
as encyclopedias, histories, biographical dic-
Methodological Issues tionaries, autobiographies, and biographies.
Accordingly, unlike what holds for any
other general research method, historiomet-
As a scientific technique, historiometrics
ric samples can include individuals who are
departs appreciably from other methods in
deceased. Indeed, it is not uncommon for a
the behavioral sciences (Simonton, 1999a,
historiometric investigation to be confined
2003 c). It certainly differs from experi-
to eminent achievers who have already fin-
mental approaches, whether laboratory or
ished out their life spans (e.g., Cox, 1926;
field, insofar as it depends on correlational
Raskin, 193 6; Simonton, 1975 b). This capac-
data analyses. In this sense it has a close
ity has critical implications for the study
affinity with psychometrics. Even so, histo-
of expertise acquisition and expert perfor-
riometrics and psychometrics dramatically
mance because it becomes thereby possible
differ on several key methodological param-
to examine exceptional achievement across
eters. As a result, it is necessary to treat
the entire life, from birth to death. It should
some of the technical concerns that are espe-
be noted, too, that because the samples often
cially prominent in this distinctive approach
consist of deceased celebrities, they cannot
(see also Simonton, 1990). These issues per-
properly be called participants, as is the
tain to sampling procedures, variable defini-
current convention, but rather they must
tions, research designs, and methodological
be referred to by the older term subjects
artifacts.
(Simonton, 1999a).
Given the distinctive nature of historio-
Sampling Procedures
metric samples, the next question is how to
Most psychological research relies on assemble the individuals who will become
research participants who are anonymous the research subjects. Sometimes the sam-
and inherently replaceable. This is especially ple will be defined according to membership
the case for investigations that draw their in well-defined groups of eminent achiev-
samples from college undergraduates who ers, such as all Nobel laureates in the sci-
sign up as research participants in order ences (e.g., Manniche & Falk, 195 7; Stephan
to fulfill a course requirement. The spe- & Levin, 1993 ). The only limitation may be
cific identity of the participant is not rel- that some subjects will have to be deleted
evant, and one participant is presumed to owing to the lack of necessary biographical
historiometric methods 32 3
data (see, e.g., Cox, 1926). Other times the productivity or the output of highly influen-
domain of achievement is not so specifically tial works (Murray, 2003 ; Simonton, 1977c,
defined, such as the expert performance dis- 1991b, 1992b), (d) objective scoring systems
played by great generals, illustrious scien- such as those used to rate chess players and
tists, outstanding artists, or famous com- athletes (e.g., Elo, 1986; Schulz & Curnow,
posers. In such instances eligibility for the 1988), (e) the attainment of high offices
sample is more open-ended. The most com- and positions, such as president, prime min-
mon procedure in this case is to sample those ister, pontiff, patriarch, or company CEO
individuals who attain the highest degree (Lehman, 195 3 ; Sorokin, 1925 ,1926), and
of eminence in the targeted domain of (f) subjective assessments based on surveys
achievement (e.g., Simonton, 1977c, 1984a, of scholars and other experts (Simonton,
1980, 1991a). For example, in Coxs (1926) 1977b, 1987c, 1992b). Occasionally, investi-
study the sample was derived from the most gators have gauged historic individuals with
eminent historical personalities on Cattells respect to their display of multiple compe-
(1903 ) list, where eminence was based on tencies, usually under the variable category
the amount of space devoted to each person of versatility (Cassandro, 1998; Simonton,
in standard reference works. Because most 1976a; White, 193 1).
domains of expertise are not well defined, However, sometimes historiometricians
sampling according to eminence is a very will adopt a more fine-grained analysis by
common procedure. However, it does have taking particular achievements or events as
one major disadvantage: By selecting only the units of analysis. Performance is then
those subjects who attain the highest degree gauged according to the differential impact
of distinction, the investigator necessarily or success of those units. Examples include
truncates the amount of variance that will the critical evaluations bestowed on motion
be exhibited by many relevant variables. This pictures (e.g., Simonton, 2004b; Zickar &
variance truncation will reduce the expected Slaughter, 1999), the frequency that an
correlations that can be obtained between opera appears in the worlds major opera
performance criteria and various predictor houses (Simonton, 2000), and whether a
variables. battle resulted in victory or defeat for a
particular general (Simonton, 1980). Finally,
Variable Definitions sometimes the analysis of singular acts
of exceptional performance will be aggre-
Historiometric inquiries into expertise gen- gated into consecutive periods of a career,
erally must include two types of assessments. such as decades. For instance, investiga-
The first type concerns measures of actual tors might examine how the magnitude of
performance and the second concerns indi- performance changes as a function of age
cators of acquisition. (Simonton, 1977a, 1984d, 1985 ). Alterna-
tively, researchers might study how exper-
performance measures tise in separate domains must be distributed
Most commonly expertise is viewed as an across the career course so as to maxi-
attribute of individuals, and accordingly the mize impact or influence (Root-Bernstein,
assessment of expert performance is car- Bernstein, & Garnier, 1993 ; Simonton,
ried out at the level of individuals. In this 1992b).
case, individual attainment can be gauged in
terms of (a) eminence as recorded by space
allotted in reference works (Cattell, 1903 ; acquisition indicators
Cox, 1926; Galton, 1869; Simonton, 1976a), Expert performance has numerous predic-
(b) the receipt of major honors such as the tors, but certainly among the most crucial is
Nobel Prize or Olympic medals (Clark & the acquisition of the necessary competence
Rice, 1982; Berry, 1981; Manniche & Falk, in the first place. This acquisition has been
195 7; Zuckerman, 1977), (c) total lifetime accessed several ways. The easiest is to use an
32 4 the cambridge handbook of expertise and expert performance
and decline in the capacity for exceptional estimated across multiple time series repre-
achievement. This approach has two main senting more than one career. This approach
forms: individual and aggregated. Individual was first applied to the study of ten top
designs scrutinize the performance of a sin- classical composers (Simonton, 1977a), but
gle expert over the course of his or her career. was later applied to the careers of ten emi-
An example is Ohlssons (1992) demonstra- nent psychologists (Simonton, 1985 ) as well
tion that Isaac Asimovs output of books can as to the reigns of absolute monarchs in
be described according to a standard learn- Europe (Simonton, 1984d) and Great Britain
ing curve. See also Weisberg, chapter 42, (Simonton, 1998a). A far more sophisti-
for a case study approach. Such single-case cated procedure takes advantage of the lat-
designs have also been applied to such his- est advances in multi-level designs. An excel-
toric figures as Napoleon (Simonton, 1979), lent example is Zickar and Slaughters (1999)
Shakespeare (Simonton, 1986b); Simanton, use of hierarchical linear modeling to assess
1989b Beethoven (Simonton, 1987b), and the creative performance of distinguished
Edison (Lehman, 195 3 ). The obvious draw- film directors. The method permits the esti-
back to this approach is that the observed mate of a typical career trajectory while at
fluctuations in performance may be idiosyn- the same time obtaining estimates for each
cratic to that particular person. There is expert making up the sample.
no guarantee that the longitudinal results
would generalize to a larger collection of
Methodological Artifacts
experts drawn from the same domain. As
a consequence, the vast majority of longi- By the very nature of the method, his-
tudinal studies utilize an aggregated design toriometric research is correlational rather
(Simonton, 1988a). In this case, performance than experimental. As a necessary repercus-
is averaged across multiple experts, produc- sion, such research lacks the random assign-
ing an overall career trajectory in which indi- ment and variable manipulations required
vidual idiosyncrasies cancel out. The num- for the secure causal inferences found in
ber of cases making up the aggregated anal- laboratory and field experiments. Instead,
ysis may run into the hundreds. This mode controls must be implemented statistically,
of analysis was first introduced by Quetelet most often via a multiple-regression analysis
(183 5 /1968) and was most extensively used (Simonton, 1990). That is, in addition to the
by Lehman (195 3 ). It is difficult to iden- substantive variables that are directly rele-
tify a form of world-class expertise to which vant to the hypothesis at hand, the inves-
this design has not been applied. Exam- tigator must include one or more control
ples include creativity, leadership, sports, variables that permit statistical adjustment
and chess. for potential artifacts. In particular, statistical
controls help the researcher avoid the intru-
mixed designs sion of spurious associations. These controls
Although aggregated longitudinal designs include such variables as birth year, life span,
are widely used and have a long history, gender, nationality, and domain of exper-
they suffer from a number of methodolog- tise. For instance, one historiometric inquiry
ical problems (see later discussion). As a examined the ages at which scientists pro-
consequence, historiometricians have more duce their first major contribution, their sin-
recently applied mixed designs that integrate gle best contribution, and their last major
individual and aggregate levels of analysis. contribution (Simonton, 1991a). The specific
The first mixed design was cross-sectional issue was whether the location of these three
time-series analysis (Simonton, 1977b). career landmarks varied according to the
Here the performance data are tabu- specific scientific discipline. However, such
lated across consecutive units of an indi- interdisciplinary contrasts are contaminated
viduals career, producing individual-level by the fact that life expectancies are not
time series, but then the age functions are constant across domains. Mathematicians in
32 6 the cambridge handbook of expertise and expert performance
particular tend to live less long than scientists presumed to be a function of three effects:
in other disciplines. Accordingly, the inter- (a) history, defined as the particular point in
disciplinary differences had to be estimated time (T); (b) cohort, defined as the individ-
after first controlling for life expectancy. uals year of birth (B); and (c) age, defined as
The specific source of spuriousness the persons chronological age (T B). Yet it
depends on the nature of the substantive should be obvious that these three effects are
question. What may be an essential control not linearly independent. Specifically, if his-
variable in one study may prove irrelevant tory and cohort are given, then age is fixed.
in another. Nonetheless, certain research This linear dependence can introduce sub-
designs are especially vulnerable to artifac- tle problems in data analysis. For example,
tual results. A case in point is longitudi- if a study is looking at the number of cita-
nal designs that aggregate results across the tions a scientist receives to work published
careers of multiple cases. These designs are in a given time period, and if variables are
particularly susceptible to what has been introduced for the scientists age and year
called the compositional fallacy (Simonton, when the publications appeared, then it is
1988a). This is a specific form of aggrega- impossible to also include a control for a sci-
tion error (Hannan, 1971). That is, statistics entists birth year. This means that any vari-
that are aggregated across individuals may ation across cohorts in output levels must
produce age curves that are not character- be ignored. Because this limitation is math-
istic of any individual making up the sam- ematical rather than empirical it cannot be
ple. To illustrate, suppose that a sample of overcome by any statistical method.
creators is used to tabulate the number of
creative products produced in consecutive
decades. Let us also assume that the sam- Empirical Findings
pled creators vary appreciably in life span.
Then the total count of products in the Since Quetelets (183 5 /1968) pioneering
later decades will be smaller than the total study, historiometric research has come up
count in the earlier decades simply because with an impressive inventory of empirical
there are fewer creators still alive in the later results. These can be grouped into two cate-
decades. Thus, even if there is not age decre- gories, namely, those concerning the acquisi-
ment in performance at the individual level, tion of expertise and those regarding expert
there will still appear an artifactual decre- performance. In each category I will begin
ment in the aggregated data. Furthermore, by giving an overview of some of the central
even if an individual-level age decline exists, findings and then end by describing a partic-
that decline will be exaggerated at the aggre- ular historiometric study that addresses an
gate level. This is a recurrent problem with issue in that category.
many of the empirical findings reported in
Lehmans (195 3 ) classic work. Fortunately, Expertise Acquisition
methods exist to circumvent this bias. For
instance, statistical adjustment of the totals overview
might be implemented based on the number Early historiometricians were often inter-
of individuals alive each period (Quetelet, ested in the developmental antecedents of
183 5 /1968), or the sample might be confined exceptional achievement (e.g., Bowerman,
to individuals who lived unusually long lives 1947; Candolle, 1873 ; Cox, 1926; Ellis,
(Dennis, 1966; Lindauer, 1993 ). 1926; Galton, 1869; Raskin, 193 6). Indeed,
Finally, it is important to recognize one of the original arguments for the
another methodological difficulty inherent method was that it could provide important
in studies of the relation between age and insights into the origins of genius and tal-
expert performance (Adam, 1978; Schaie, ent (Woods, 1911). Furthermore, many of the
1986). The expected performance of an indi- early empirical findings have been replicated
vidual at a particular point in time is often and extended in more recent historiometric
historiometric methods 32 7
work (e.g., Goertzel, Goertzel, & Goertzel, a multitude of forms, including formal
1978; Simonton, 1976a, 1984a; Walberg, education, private instruction, coaching or
Rasher, & Parkerson, 1980). The most crit- mentoring, exposure to domain-specific
ical of these results concern the following role models, and various forms of self-
three factors (Simonton, 1987a). education, such as omnivorous reading
First, world-class expertise tends to (Raskin, 193 6; Simonton, 1976a, 1984a,
emerge from a distinctive family background. 1986a, 1992a; Walberg, Rasher, & Parkerson,
As already noted, Galtons (1869) classic 1980). Nonetheless, historiometric research
inquiry was dedicated to showing that emi- also has pinpointed some important qual-
nent achievers tended to come from distin- ifications and complications regarding this
guished family pedigrees (see also Bramwell, 10-year rule (Simonton, 1996a, 2000). To
1948). Although a significant portion of this begin with, the specific nature of the instruc-
tendency may reflect the influence of nur- tion and training depends greatly on the
ture rather than nature, the relevance of type of expertise being acquired (Goertzel,
genetic endowment cannot be totally dis- Goertzel, & Goertzel, 1978; Raskin, 193 6;
missed (Simonton, 1999b). For instance, Simonton, 1986a). For example, outstanding
notable individuals in certain domains of leaders require different educational expe-
achievement are also more prone to come riences than do exceptional creators, and,
from family lineages in which the inci- within eminent creators, distinguished sci-
dence rate of psychopathology is above entists need distinct educational experiences
the population average (Jamison, 1993 ; than do illustrious artists (Simonton, 2004a).
Juda, 1949; Karlson, 1970). Nevertheless, In fact, there is evidence that in some
numerous historiometric investigations have domains of achievement it is possible to have
documented how certain family circum- too much formal education or scholastic
stances serve as environmental factors that success to be successful (Simonton, 1976a,
influence the acquisition of extraordinary 1986a). This can be viewed as a form of
expertise (Simonton, 1987a). These factors overtraining (Simonton, 2000). In addi-
include socioeconomic class, early traumatic tion, substantial individual differences exist
experiences, and birth order (Eisenstadt, in the amount of time used to master the
1978; Goertzel, Goertzel, & Goertzel, 1978; domain-specific skills and knowledge that
Raskin, 193 6; Silverman, 1974; Sulloway, are needed for exceptional accomplishments
1996). Interestingly, these factors work pri- (Cox, 1926; Raskin, 193 6; Simonton, 1991b,
marily by channeling a young person into 1992a). In particular, those who attain the
a particular form of expertise. For instance, highest levels of achievement are more likely
scientific creators, in comparison to artistic to have undergone expertise acquisition at
creators, are more likely to grow up in stable an accelerated rate.
and conventional homes with highly edu- Third and last, expertise of the highest
cated parents who pursue professional occu- order is most likely to appear in a particu-
pations (Simonton, 2004a). lar sociocultural context. This reality is indi-
Second, genius and exceptional talent cated by the fact that genius and talent
are associated with distinctive education are not randomly distributed across space
and training. Galtons (1869) assertion that and time but rather tend to cluster into
exceptional achievement is born rather than particular geographical locations (Candolle,
made is plain wrong. The literature on 1873 ; Charness & Gerchak, 1996; Yuasa,
expertise acquisition suggests that it requires 1974) and historical periods (Kroeber, 1944;
around a decade of committed training Murray, 2003 ; Simonton, 1988b, 1996b).
and practice to attain world-class exper- The underlying causes of such clustering
tise (Ericsson, 1996), an idea that has involve a host of cultural, social, political,
received endorsement from historiometric economic, and cultural factors (Simonton,
research as well (Hayes, 1989; Simonton, 2003 a). For instance, a large portion of the
1991b). This acquisition process can take temporal clustering of exceptional creators
32 8 the cambridge handbook of expertise and expert performance
and leaders can be attributed to the availabil- Next, two sets of substantive variables were
ity of domain-specific role models (Murray, defined.
2003 ; Simonton, 1975 b, 1988b, 1992a). In First, for each composer determinations
particular, the number of great achievers wre made of the total lifetime output, the
in one generation is a positive function of maximum annual output rate, the age at
the number in the preceding generation. In maximum output, the age at first hit, the
the specific area of creativity, some political age at best hit, and the age at last hit,
environments tend to nurture creative devel- where a hit was a work that had obtained
opment whereas others tend to discourage a secure place in the standard repertoire.
creative development (Simonton, 2003 a). These measures were all assessed two ways,
For example, exceptional creators are less namely, complete works (or compositions)
likely to develop during times of political and the individual themes (or melodies)
anarchy but are more likely to develop dur- making up those works. Two alternative def-
ing periods of political fragmentation, when initions were used to take into considera-
a civilization is divided into numerous inde- tion that works vary greatly in the magni-
pendent states (Simonton, 1975 b, 1976b). tude of achievement, such as the contrast
Of course, another critical factor underly- between a song and an opera. This con-
ing the appearance of certain forms of high trast might be better captured by assessing
achievement is the value or importance that the total amount of melodic material going
a particular culture assigns to that activity into each work, a song having much fewer
at a given point in time (Candolle, 1873 ; themes than an opera.
Charness & Gerchak, 1996; Murray, 2003 ). Second, for each composer the age at
Potential talent will not become fully real- first formal music lessons was gauged, as
ized in a milieu that discourages the corre- well as the age at which composition began,
sponding domain of achievement. including any juvenilia. These two measures
were then combined with the assessment
of age at first hit to create another set of
illustration variables: (a) musical preparation, or the age
To provide a better idea of how historio- at first hit minus the age at first lessons and
metric investigations can contribute to our (b) compositional preparation, or the age
scientific understanding of expertise acqui- at first hit minus the age at which compo-
sition, I will describe in somewhat more sition was initiated. The first variable con-
detail a specific study in this area (Simonton, cerns how many years transpired between
1991a). A major goal of the inquiry was to the onset of lessons and the first composi-
discern how individual differences in exper- tional success, whereas the second concerns
tise acquisition are correlated with indi- how many years passed between the initi-
vidual differences in expert performance. ation of composition and the first success.
The particular domain under scrutiny was Because age at first hit was defined two dif-
the composition of classical music. A sam- ferent ways (works and themes), there were
ple of 120 eminent composers was obtained actually two alternative indicators of musical
by taking those who had entries in two and compositional preparation.
distinct reference works. The differential Finally, two control variables were also
eminence of these composers was then included, namely, the composers birth year
assessed using six different sources, includ- and the life span. The former allows adjust-
ing their performance frequencies in concert ment for any historical trends whereas the
halls, their ranking by musicologists, and the latter permits adjustment for how long the
space devoted to them in various reference composer lived, a factor that places an obvi-
works. This composite eminence measure ous constraint on lifetime output as well as
was shown to reflect a high degree of consen- the age at last hit.
sus regarding the creative achievements of These measures were subjected to a series
these composers (see also Simonton, 1991c). of correlation and regression analyses, with
historiometric methods 32 9
chess. With that restriction in mind, we It should be pointed out that many of
see that the historiometric work on the the above results apply to other domains of
age-creativity relationship has arrived at outstanding achievement. For instance, the
the following four empirical generalizations curvilinear function seen in creative domains
(Simonton, 1988a, 1997): has a very similar form in leadership, sports,
1. The output of creative products in and chess, where single-peaked functions
consecutive age periods is described by a are commonplace (Elo, 1965 ; Lehman, 195 3 ;
curvilinear function. The output first rapidly Schulz & Curnow, 1988). At the same time,
increases to a single peak in the 3 0s or the specific form of the curve, including
40s and then gradually declines, produc- the location of the peak and the post-peak
ing an age decrement that approaches the decline, also varies from domain to domain.
zero output rate asymptotically. Most typ- For example, the age for top performance
ically productivity in the final years of the in sports depends on the specific event
career is about half the rate seen at the (Lehman, 195 3 ; Schulz & Curnow, 1988).
career peak. On the other hand, some of the findings for
2. The specific location of the peak as well creative achievement do not necessarily hold
as the magnitude of the post-peak decline for other forms of world-class expert perfor-
varies according to the particular domain of mance. A case in point regards the distinc-
creative achievement. In some fields such tion between career and chronological age.
as lyric poetry and mathematics the peak To the extent that performance depends on
arrives relatively early and the age decre- physiological rather than psychological vari-
ment is usually large, whereas in fields like ables, the longitudinal curves will be defined
history and geology the peak comes later and in terms of chronological age. This qualifi-
the ensuing decline is minimal. cation holds for some types of leadership
3 . Properly speaking, the age functions and virtually all forms of sports (Schulz &
just described are based on career age rather Curnow, 1988; Simonton, 1998b).
than chronological age. In any given field
there will always appear considerable indi- illustration
vidual differences in the age at career onset A good example of historiometric research
(e.g., age at receiving a doctorate). Those on expert performance is a recent study
with an accelerated onset (early bloomers) of top movie directors (Zickar & Slaughter,
will have their career peak occur earlier 1999). The specific goal was to determine the
in chronological age, whereas those with a age-achievement function for 73 directors
delayed onset (late bloomers) will have their who made at least 20 feature-length movies
career peak appear later. The latter temporal for the Hollywood film industry. Because
shift is commonplace for those creators who the investigators used a hierarchical linear
exhibited a mid-life career change. model, the units of analysis existed at two
4. The age-performance curves are the levels, directors and films. The films for each
same for both quantity and quality of out- director were evaluated according to the
put. That is, the production of total works ratings they received from film critics, as
independent of creative impact follows the recorded in two movie guides. A composite
same longitudinal form as the production of evaluation constituted the dependent vari-
just those works that manage to exert some able for the investigation. The independent
influence on the domain of creativity. As a variable was the order of the film in the
consequence, a creators best work is more directors career. This order was introduced
likely to appear in those periods in which the into the predictions equation in both linear
most total work appears. In fact, the ratio and quadratic functions in order to test for
of hits to total output per age period does the curvilinear single-peak function found
not systematically change over time, a find- in the literature on creative performance. In
ing known as the equal-odds rule (Simonton, addition, acting quality was inserted as a con-
1997). trol variable. This was gauged by the number
historiometric methods 331
of Academy Awards that each film received mathematical models that can account for
in the acting categories (two points for each the fine structure of careers (Simonton,
lead role and one point for each suppor- 1984b, 1989a, 1997). Of course, there remain
ting role). considerable gaps in our knowledge. This is
In general, the cinematic performance particularly apparent in those domains out-
of the 73 directors was described by a side creative achievement most notably
quadratic function. That is, the ratings the the diverse forms of leadership. In addition,
films received from critics first increased to even in domains pertaining to creativity we
a single peak and thereafter declined. The have much more to learn about the devel-
most typical outcome was for a significant opmental correlates of expertise acquisition.
decline in performance to set in after the Indeed, no matter what the domain, much
tenth film. Nevertheless, there were substan- more is known about expert performance
tial differences across directors in the spe- than about how experts acquire the capacity
cific form of this curve. For instance, those to perform at world-class levels.
directors who had higher rates of cinematic Admittedly, historiometric methods have
output tended to reach higher levels in peak certain features that militate against its wide
performance, an effect found for virtually usage in empirical research. As a correla-
all creative domains (Simonton, 1997). More tional method, it lacks the power of causal
surprising was the finding that directors who inference that is enjoyed by experimental
launched their careers with an exceptionally methods. Historiometrics also has to rely
successful film were most likely to exhibit on biographical and historical data that is
a linear decline in performance rather than sometimes of questionable reliability. Fur-
rise to a yet higher peak. This subgroup thermore, historiometric inquiries focus on
probably represents directors whose initial a subject pool or research participants
performance was due to luck, and thus the that depart significantly from the norm.
subsequent decrement can be attributed to Rather than anonymous college undergrad-
regression toward the mean. uates earning extra credit points in intro-
ductory psychology classes, historiometric
samples invariably consist of individuals
Conclusions whose achievements have earned them a
place in the annals of history (Simonton,
Although historiometrics cannot be consid- 1999a, 2003 c). No wonder that historiomet-
ered a mainstream method in the behav- ric research is rare in any area of psycholog-
ioral sciences, the earliest research on expert ical research, expertise or otherwise.
performance used this technique. Further- Nevertheless, the latter characteristic of
more, the number of studies that have historiometrics must also be viewed as one
accumulated since 183 5 is truly impressive of its great assets. Any theory of expertise
(Simonton, 1988a). The resulting literature acquisition and expert performance must
has produced an impressive body of empir- ultimately be able to account for those
ical findings, particularly in those domains persons whose expertise reaches the high-
that entail creative expertise (Simonton, est possible levels. For example, a theory
1996a). Not only do we know a lot that explains how students solve science
about the factors that underlie the acqui- problems in laboratory experiments but
sition of world-class expertise (Simonton, not how real scientists earn Nobel Prizes
1987a), but also we have learned even more must be considered woefully incomplete.
about how that expertise manifests itself in Although other methods exist that permit
adulthood performance (Simonton, 1988a). the direct examination of eminent achiev-
In fact, the cumulative results regarding ers, these have their own methodological
the age-performance function in creative limitations. For instance, Nobel laureates can
domains has become sufficiently rich and certainly be subjected to intensive inter-
robust to provide the basis for complex view and assessment techniques (Roe, 195 3 ;
332 the cambridge handbook of expertise and expert performance
Zuckerman, 1977). Yet these methods Clark, R. D., & Rice, G. A. (1982). Family con-
depend on the willingness of such nota- stellations and eminence: The birth orders of
bles to participate. It should also be recog- Nobel Prize winners. Journal of Psychology, 110,
nized that some of the liabilities of histori- 281287.
ometric studies are becoming progressively Cox, C. (1926). The early mental traits of three
removed. In the first place, the historical hundred geniuses. Stanford, CA: Stanford Uni-
record is becoming far richer and more com- versity Press.
plete, in addition to becoming more available Dennis, W. (1966). Creative productivity
in electronic form on the Internet and com- between the ages of 20 and 80 years. Journal of
Gerontology, 2 1, 18.
puter storage media. Even more importantly,
the statistical techniques suitable for the Eisenstadt, J. M. (1978). Parental loss and genius.
American Psychologist, 3 3 , 211223 .
analysis of correlational data are becoming
more sophisticated and powerful, thereby Ellis, H. (1926). A study of British genius (rev. ed.).
mitigating some of the problems in nonex- Boston: Houghton Mifflin.
perimental causal inference. These statistics Elms, A. C. (1994). Uncovering lives: The uneasy
include structural equation and hierarchical alliance of biography and psychology. New York:
Oxford University Press.
linear models as well as time-series analy-
ses. By applying these tools to the lives and Elo, A. E. (1965 ). Age changes in master chess
careers of the leading figures in the major performance. Journal of Gerontology, 2 0, 289
299.
achievement domains, it should be possi-
ble to enhance our scientific understand- Elo, A. E. (1986). The rating of chessplayers, past
and present (2nd ed.). New York: Arco.
ing of both expertise acquisition and expert
performance. Ericsson, K. A. (1996). The acquisition of expert
performance: An introduction to some of the
issues. In K. A. Ericsson (Ed.), The road to
expert performance: Empirical evidence from the
References arts and sciences, sports, and games (pp. 15 0).
Mahwah, NJ: Erlbaum.
Adam, J. (1978). Sequential strategies and Ericsson, K. A. (Ed.). (1996). The road to expert
the separation of age, cohort, and time-of- performance: Empirical evidence from the arts
measurement contributions to developmental and sciences, sports, and games. Mahwah, NJ:
data. Psychological Bulletin, 85 , 13 0913 16. Erlbaum.
Berry, C. (1981). The Nobel scientists and the ori- Galton, F. (1865 ). Hereditary talent and charac-
gins of scientific achievement. British Journal of ter. Macmillans Magazine, 12 , 15 7166, 3 18
Sociology, 3 2 , 3 813 91. 3 27.
Bowerman, W. G. (1947). Studies in genius. New Galton, F. (1869). Hereditary genius: An inquiry
York: Philosophical Library. into its laws and consequences. London:
Bramwell, B. S. (1948). Galtons Hereditary Macmillan.
Genius and the three following generations Galton, F. (1874). English men of science: Their
since 1869. Eugenics Review, 3 9, 14615 3 . nature and nurture. London: Macmillan.
Candolle, A. de (1873 ). Histoire des sciences et des Goertzel, M. G., Goertzel, V., & Goertzel, T. G.
savants depuis deux siecles. Geneve: Georg. (1978). 3 00 eminent personalities: A psychosocial
Cassandro, V. J. (1998). Explaining premature analysis of the famous. San Francisco: Jossey-
mortality across fields of creative endeavor. Bass.
Journal of Personality, 66, 805 83 3 . Hannan, M. T. (1971). Problems of aggregation.
Cattell, J. M. (1903 ). A statistical study of emi- In H. M. Blalock (Ed.), Causal models in the
nent men. Popular Science Monthly, 62 , 3 5 9 social sciences (pp. 473 5 08). Chicago: Aldine-
3 77. Atherton.
Charness, N., & Gerchak, Y. (1996). Participation Hayes, J. R. (1989). The complete problem solver
rates and maximal performance: A log-linear (2nd ed.). Hillsdale, NJ: Erlbaum.
explanation for group differences, such as Jamison, K. R. (1993 ). Touched with fire: Manic-
Russian and male dominance in chess. Psycho- depressive illness and the artistic temperament.
logical Science, 7, 465 1. New York: Free Press.
historiometric methods 333
Juda, A. (1949). The relationship between high- Roe, A. (195 3 ). The making of a scientist. New
est mental capacity and psychic abnormalities. York: Dodd, Mead.
American Journal of Psychiatry, 106, 2963 07. Root-Bernstein, R. S., Bernstein, M., & Garnier,
Karlson, J. I. (1970). Genetic association of gifted- H. (1993 ). Identification of scientists mak-
ness and creativity with schizophrenia. Hered- ing long-term, high-impact contributions, with
itas, 66, 177182. notes on their methods of working. Creativity
Kroeber, A. L. (1944). Configurations of cul- Research Journal, 6, 3 293 43 .
ture growth. Berkeley: University of California Runyan, W. M. (1982). Life histories and psychobi-
Press. ography. New York: Oxford University Press.
Lehman, H. C. (1946). Age of starting to con- Schaie, K. W. (1986). Beyond calendar definitions
tribute versus total creative output. Journal of of age, time, and cohort: The general develop-
Applied Psychology, 3 0, 460480. mental model revisited. Developmental Review,
Lehman, H. C. (195 3 ). Age and achievement. 6, 25 2277.
Princeton, NJ: Princeton University Press. Schulz, R., & Curnow, C. (1988). Peak perfor-
Lehman, H. C. (195 8, May 23 ). The chemists mance and age among super athletes: Track and
most creative years. Science, 12 7, 1213 1222. field, swimming, baseball, tennis, and golf. Jour-
nal of Gerontology, 43 , 113 120.
Lehman, H. C. (1960). Age and outstanding
achievement in creative chemistry. Geriatrics, Silverman, S. M. (1974). Parental loss and scien-
15 , 193 7. tists. Science Studies, 4, 25 9264.
Lehman, H. C. (1962). More about age and Simonton, D. K. (1975 a). Age and literary
achievement. Gerontologist, 2 , 141148. creativity: A cross-cultural and transhistorical
survey. Journal of Cross-Cultural Psychology, 6,
Lehman, H. C. (1963 ). Chronological age versus 25 9277.
present-day contributions to medical progress.
Simonton, D. K. (1975 b). Sociocultural context
Gerontologist, 3 , 7175 .
of individual creativity: A transhistorical time-
Lehman, H. C. (1966a). The most creative years series analysis. Journal of Personality and Social
of engineers and other technologists. Journal of Psychology, 3 2 , 1119113 3 .
Genetic Psychology, 108, 263 270.
Simonton, D. K. (1976a). Biographical determi-
Lehman, H. C. (1966b). The psychologists most nants of achieved eminence: A multivariate
creative years. American Psychologist, 2 1, 3 63 approach to the Cox data. Journal of Personality
3 69. and Social Psychology, 3 3 , 218226.
Lindauer, M. S. (1993 ). The span of creativity Simonton, D. K. (1976b). Philosophical emi-
among long-lived historical artists. Creativity nence, beliefs, and zeitgeist: An individual-
Research Journal, 6, 23 123 9. generational analysis. Journal of Personality and
Manniche, E., & Falk, G. (195 7). Age and the Social Psychology, 3 4, 63 0640.
Nobel prize. Behavioral Science, 2 , 3 013 07. Simonton, D. K. (1977a). Creative productivity,
Murray, C. (2003 ). Human accomplishment: age, and stress: A biographical time-series anal-
The pursuit of excellence in the arts and ysis of 10 classical composers. Journal of Person-
sciences, 800 B.C. to 195 0. New York: ality and Social Psychology, 3 5 , 791804.
HarperCollins. Simonton, D. K. (1977b). Cross-sectional time-
Ohlsson, S. (1992). The learning curve for writing series experiments: Some suggested statisti-
books: Evidence from Professor Asimov. Psy- cal analyses. Psychological Bulletin, 84, 489
chological Science, 3 , 3 803 82. 5 02.
Oleszek, W. (1969). Age and political careers. Simonton, D. K. (1977c). Eminence, creativity,
Public Opinion Quarterly, 3 3 , 100103 . and geographic marginality: A recursive struc-
Quetelet, A. (1968). A treatise on man and the tural equation model. Journal of Personality and
development of his faculties. New York: Franklin. Social Psychology, 3 5 , 805 816.
(Reprint of 1842 Edinburgh translation of 183 5 Simonton, D. K. (1979). Was Napoleon a military
French original.) genius? Score: Carlyle 1, Tolstoy 1. Psychological
Raskin, E. A. (193 6). Comparison of scientific and Reports, 44, 2122.
literary ability: A biographical study of eminent Simonton, D. K. (1980). Land battles, generals,
scientists and men of letters of the nineteenth and armies: Individual and situational determi-
century. Journal of Abnormal and Social Psychol- nants of victory and casualties. Journal of Per-
ogy, 3 1, 203 5 . sonality and Social Psychology, 3 8, 110119.
334 the cambridge handbook of expertise and expert performance
DOMAINS OF EXPERTISE
P a r t V. A
PROFESSIONAL DOMAINS
C H A P T E R 19
was little change in the process with increas- There are several possible explanations
ing experience. Experts had higher diag- for this apparent anomaly. Reading a page
nostic accuracy, not because of a dif- of text takes much longer than the five sec-
ferent process, but because they knew ond exposure to a mid-game chess posi-
more and organized their knowledge dif- tion. Indeed, Schmidt and Boshuizen (1993 )
ferently, which enabled them to generate showed that superior recall of experts
and test more accurate diagnostic hypothe- emerges with short exposure of as little as
ses (Feltovich et al., 1984; Neufeld et al., thirty seconds. Second, in other domains,
1981). This observation is remarkably consis- differences with expertise may occur sim-
tent with studies of chess expertise (Simon & ply because being able to remember all
Chase, 1973 ). But the most surprising finding the details is an adaptive strategy for chess
was that success on one problem was a poor and for medical students (who are often
predictor of success on a second problem; expected to recite details of clinical cases),
the correlation across problems was typically but synthesizing all the details into a brief
of the order of 0.1 to 0.3 , (Elstein et al., but coherent problem formulation, ignor-
1978), and the factors leading to variation in ing extraneous details, is a better description
performance appear to be multiple. Collec- of medical experts (Eva, Norman, Neville,
tively, these findings spelled the death knell Wood, & Brooks, 2002).
for the idea of a general problem-solving pro- There are, however, areas of medicine
cess and led to a change in direction and a where it is necessary to keep all the data
closer examination of the role of knowledge in mind. Nephrology requires an under-
in expertise. standing of abnormal kidney function at
a physiological level, which can be under-
stood by examining the relations among as
many as twenty numerical laboratory val-
Medical Expertise as Amount ues. Norman, Brooks, and Allen (1989) were
of Knowledge able to show a positive relation between
expertise and recall; however, apparently
The new tradition began by adopting strate- this superiority occurs only when experts
gies that had proved successful in other do not have clinical information available
domains (de Groot, 1978) based on mem- and hence can not infer possible diagnoses
ory for typical cases (for example, chess mas- from clinical patterns (Verkoeijn et al.,
ters shown a mid-game position for five sec- 2004).
onds can recall about 80% of the positions). Regardless, it is not clear what implica-
Instead of actually doing diagnosis, clinicians tions are to be derived from these findings.
in these studies read written cases, typically We may conclude, as did Simon and Chase
about a page long, then recalled what they (1973 ), that experts have large memory
had read. chunks and this superior memory (when
Surprisingly, this method, which worked it does occur) may reflect that expertise in
well elsewhere, led to few successes in medicine, as in chess (Burns, 2004), is an
the medical domain. Given unlimited time, index of rapid pattern recognition related to
novices could recall as much as experts experience with many cases. But it would
(Muzzin et al., 1983 ), although experts appear that such observations yield little
appeared to acquire information more effi- direction for improving the education of
ciently and attended to more critical infor- medical students.
mation (Coughlin & Patel, 1987). However, It is not surprising therefore that, in view
intermediates (final-year medical students) of the failure of recall measures to charac-
appear to consistently recall more informa- terize knowledge, the focus again changed
tion than novices or experts (Schmidt & this time to an examination of the type and
Boshuizen, 1993 ). organization of knowledge.
342 the cambridge handbook of expertise and expert performance
Medical Expertise and the edge is a relatively small but active area of
Organization of Knowledge cognition, associated with researchers like
Kim and Ahn (2003 ). Analytical knowledge,
In this section, we will review a number as we have described it, might be aligned
of perspectives on how knowledge is rep- with semantic memory in the classical
resented. But before we begin, a caution- view of memory types, and with prototype
ary note about research method is in order. theories of categorization (Rosch & Mervis,
The standard approach to these studies is 1976). Experiential knowledge can be seen as
to assemble groups of experts and novices, a kind of episodic memory, from the classical
engage them in some task, and examine the view, and is closely associated with exemplar
data for differences between the two groups. models of categorization (Medin, Altom &
When differences are found, then the claim Murphy, 1984; Brooks, 1978).
is made that therein lies the essence of exper-
tise. There are fundamental problems with
Causal Knowledge: The Role
this inference. Experts differ from novices
of Basic Science
in many ways. The fact that a difference is
found in some domain does not by itself Medical students spend the first half of their
justify the conclusion that the domain is a time in school studying aspects of the basic
central cause of their superior performance. mechanisms of disease, and the last half
The measure under scrutiny may be a con- in wards and clinics with patients learn-
sequence of success rather than a cause, or ing two kinds of knowledge: first, the for-
some other variable may be causing differ- mal clinical knowledge of signs and symp-
ences on both this variable and better per- toms, predictive value of tests, and preferred
formance. Second, it is unlikely that experts management approaches, and second, the
represent knowledge in any single form, or experiential knowledge of specific cases.
even that any representation is, in some One prominent research agenda has been
sense, more basic than any other. Experts, to investigate how these various knowl-
when interrogated, can provide everything edge types contribute to the acquisition of
from the probability that a child with ven- expertise.
tricular septal defect has growth retarda- One of the first attempts to character-
tion, to the colour of the hair of the last ize the structure of knowledge was the pro-
child they saw with VSD (Hassebrock et al., lific research program of Patel and Groen,
1988). It is likely that both forms of knowl- beginning in the mid-1980s. Experts and
edge, and many more, are available to the students were given a written case and
expert. were then asked to explain the case in
With these reservations, what have terms of processes or mechanisms. They may
we discovered about representations of have access to a relevant basic science text,
knowledge? First, investigations have exam- before or after the case (Patel & Groen,
ined three broad kinds of knowledge: 1986), or they may just generate the ele-
causal knowledge (essentially, understand- ments from memory. The resulting verbal
ing basic mechanisms), analytical knowledge think-aloud protocols were then analyzed
(the formal relation between diagnoses and using the propositional-analysis methods of
features signs and symptoms of vari- Kintsch (1974). Their conclusions were that
ous conditions), and experiential knowledge experts showed greater diagnostic accuracy
(the accumulation of a storehouse of prior than novices, had more coherent explana-
cases that comes with experience) (Schmidt, tions for the problems, were selective in
Norman, & Boshuizen, 1990; Gruppen & the use of findings, and made more infer-
Frohna, 2002). ences from the data and fewer literal inter-
Although this classification evolved with- pretations. However, experts used less basic
in medical expertise, it also has equivalents science in their explanations than medical
in psychology more broadly. Causal knowl- students, and experts made greater use of
expertise in medicine and surgery 343
forward reasoning (for example, The the role that basic science plays in the
patient has retrosternal chest pain with radi- acquisition of expertise. Woods, Brooks,
ation down the left arm and diaphoresis. and Norman (2005 ) have shown experi-
It is likely a heart attack, as opposed to mentally that students who learn a causal
backward reasoning like It might be a heart explanation for signs and symptoms are no
attack because the patient has crushing chest more proficient in diagnosis of straightfor-
pain. Or it could be a fever, because shes ward cases than a comparable group who
sweating). simply learned the signs and symptoms of
Schmidt and his colleagues (Boshuizen a disease. However, when the task is made
& Schmidt, 1992) used similar methods more difficult by imposing a delay in the
and arrived at similar conclusions, primar- test, changing the specific descriptions, or
ily that expert clinicians actually make lit- adding extraneous details students who
tle use of biomedical science in routine rea- understand mechanisms show improved
soning. However, their investigations have diagnostic performance, suggesting that
gone further. In order to explain both the understanding mechanisms may add coher-
absence of science in clinicians protocols ence to the relation between signs and
and the intermediate effect, where inter- symptoms and diagnoses.
mediates recall more from case descriptions Of course, mechanistic basic science
than experts or novices, they postulated that knowledge is not the only kind of knowledge
this knowledge is encapsulated but avail- that is learned in medical school. Although
able in response to specific probe ques- medical students begin their learning by
tions (Schmidt & Boshuizen,1993 ). That is, an immersion into the basic sciences, they
although experts may not mention the basic then move to a consideration of diseases and
mechanisms in their case explanations, they spend endless hours learning the 29 causes
have the information available and can recall of anemia or the signs and symptoms of
it on demand. Hashimotos disease. It is reasonable, there-
A more microscopic look at expertise, fore, to assume that one dimension of exper-
within the causal framework, can reveal how tise is the acquisition of elaborate rules relat-
concepts themselves are learned. Coulson, ing signs and symptoms to diseases. Such
Spiro, and Feltovich (1997) did this by exam- knowledge may be of a simple list-like form,
ining how misconceptions arise in medical or may be more of a narrative form, such
practitioners. They identified a large number as an illness script (Feltovich & Barrows,
of factors that contribute to misconceptions, 1984; Custers, Boshuizen, & Schmidt,
many related to the fact that practitioners 1996).
might never have learned the pertinent basic Another possibility is that the knowl-
science very well at all. edge may be in more formal structures
Perhaps the most intriguing finding of all like schemas, which may resemble a men-
of these studies is that the role of basic sci- tal decision tree from chief complaint to
ence in expertise appears to be minimal. diagnosis. An observational study (Coderre,
However, this may be misleading. Schmidts Mandin, Harasym, & Fick, 2003 ) examined
studies show that the knowledge is encapsu- the clinical reasoning of medical students
lated but available with specific probes (Rik- and experts, categorizing the self-reports
ers, Loyens, & Schmidt, 2004). Coulson and as schema induction, patter-recognition,
colleagues (1997) showed that a subgroup and hypothetico-deductive. The former
of physicians will approach management two processes were more strongly associated
using an understanding of basic science. Nor- with diagnostic success. Of course, the cau-
man et al. (1994) showed empirically that, tion we expressed earlier should be exer-
when faced with difficult diagnostic prob- cised. It would seem unlikely that anyone
lems, experts revert to basic science. used one approach exclusively. Moreover, it
Nevertheless, descriptive studies of is not really possible to identify cause and
experts and novices may underestimate effect.
344 the cambridge handbook of expertise and expert performance
Altom, & Murphy, 1984; Brooks, 1978). The images themselves are sometimes difficult
basic notion is that every learned category to decompose into features. For that rea-
is accompanied by a number of examples son, another study was conducted on elec-
acquired through experience, and that these trocardiographic (ECG) diagnosis (Hatala,
examples are still individually retrievable Norman, & Brooks, 1999). With ECGs, the
and provide support for the categorization of features, though still part of an overall image,
new cases that are similar to at least one prior are much more separable (for example, an
example. Retrieval occurs rapidly and with- ST segment elevated more than three mm.).
out any conscious application of rules. Since The experimental manipulation was actu-
the process is not analytical and conscious, ally conducted on verbal information the
it may be stimulated by similarity based on age, gender, and occupation of the patient
features that are objectively irrelevant to the which was objectively irrelevant to the diag-
category. nosis. The test materials were designed to be
Clearly, since the process is not amenable ambiguous with two likely diagnoses. How-
to introspection, the usual think-aloud ever, if the nonrelevant historical informa-
strategies cannot be used. Instead, it is nec- tion were used to recall a prior example, this
essary to manipulate, in a limited way, prior would lead to the incorrect diagnosis. Accu-
experience and then examine its impact racy of residents dropped from 46% to 23 %
on subsequent problem solving. Commonly, when the test-case historical data matched
therefore, the studies conducted in this tra- the prior (incorrect) example. The fact that
dition all rely on a practice phase, which the effect was observed with manipulation
involves exposure to particular exemplars, of information that was not diagnostically
followed by a test phase where the influence relevant suggests that the process was not
of these specific examples is examined. This available to critical reflection.
two-step process, with a requirement for Whereas the studies to date provide a
multiple examples both during learning and convincing case for an exemplar-based form
testing, imposes real time constraints, and of knowledge organization in the domains
the studies commonly use visual domains studied, it remains less certain how com-
such as dermatology or electrocardiology, mon this mechanism may be in other areas of
which can be learned more quickly than medicine. However, there is no shortage of
written cases. anecdotes, and some evidence (Hassebrock,
Initial studies in dermatology (Brooks, Bullemer, & Johnson, 1988; van Rossum &
Norman, & Allen, 1991), using resident Bender, 1990), that individual cases can be
physicians in family medicine and slides of highly memorable.
common dermatologic conditions, showed These findings suggest that it takes many
an increase of diagnostic accuracy of about examples, and not just formal knowledge, to
10% when preceded by a similar slide, both become an expert (a finding consistent with
on immediate test and two weeks later. Sub- the 10,000-hours practice of chess masters
sequent studies, where more care was taken [Simon & Chase, 1973 ]). But there is as yet
in matching the cases similarity, show even very little evidence about how these experi-
larger effects: an increase in accuracy of ences should be structured to enhance the
about 40% with residents (Regehr et al., efficiency of learning. A few studies have
1994) and 28 to 44% with medical stu- begun to examine the effectiveness of dif-
dents (Kulatanga-Moruzi, Brooks, & Nor- ferent sequences of examples, both within
man, 2001). medicine and elsewhere. There is a clear
Although these results are impressive, it advantage for starting from prototypes and
is difficult to generalize beyond dermatol- moving to ambiguous examples (Avrahami,
ogy for several reasons. Dermatologic diag- Kareev, Bogot, Caspi et al., 1997). Further,
nosis is highly empirical, with little science mixed practice, where examples from differ-
to clearly and precisely explain why one ent confusable categories are interspersed,
lesion is pink and another red. Further, the leads to large gains in efficiency over practice
346 the cambridge handbook of expertise and expert performance
blocked into categories (Hatala, Norman, & arrives at a conclusion, presumably based on
Brooks, 2003 ). the weighting of the features against inter-
The implications of these findings for nalized rules. Every so often, more likely for
instruction are not obvious. When one tough or unusual cases, the physician may
argues that expertise is critically dependent well go back to the basics and reason things
on the organization of knowledge, then it out from basic science principles.
is tempting to speculate that the teacher The question, then, is whether these dif-
should spend time teaching strategies of ferent strategies are invoked in response
knowledge organization, thereby encourag- to different degrees of problem complex-
ing students to become compiled learners, ity, or alternatively, whether all are used to
to strive for global coherence, or and to elicit additional evidence and hence decrease
use forward reasoning. In one experiment uncertainty. There is some evidence that
where this strategy was tried (Nendaz & Bor- physicians revert to reasoning based on basic
dage, 2002), it emerged that the organization science principles when confronted with
may be a consequence of knowledge, not a particularly difficult cases (Norman et al.,
causal factor in the formation of knowledge 1994). On the other hand, Patel and Groen
or knowledge representations. And although (1986) claimed that experts in another disci-
forward reasoning has become a hallmark pline used more backward reasoning, but did
of expertise, there is some evidence that this not use more basic science when confronted
represents a methodological artifact (Eva, with difficult cases.
Norman, & Brooks, 2002). The situation differs when we examine
Moreover, it is unlikely that a search for analytical, feature-based knowledge versus
the single representation (whether mental exemplar knowledge. In a series of studies in
or theoretical) that fits all is appropriate. dermatology (Regehr et al., 1994; Kulatanga-
Far more likely, there are multiple forms of Moruzi et al., 2001), independent effects
expert knowledge, and each may be used to of rule-based and exemplar-based reasoning
greater or lesser degree depending on the sit- was examined using a clever experimental
uation. Indeed, recent experimental manip- design, so that performance was assessed on
ulations instructing novices to utilize multi- typicalsimilar, typicaldissimilar, atypical
ple forms of knowledge do appear to enable similar, and atypicaldissimilar slides. Since
gains in diagnostic accuracy (Ark, Brooks, typicality amounts to the presence of a large
& Eva, 2004). Experts (and novices) may number of specified, individual features, an
invoke causal knowledge, rules relating fea- effect of typicality would be evidence of
tures to diagnoses, or prior examples to solve application of a rule, whereas an effect of
the problem. So a better question is How similarity would result from recognition of a
are these various forms of knowledge used in prior exemplar.
solving clinical problems? For that, we turn The results for both of the studies were
to a limited body of research on the coordi- similar: an effect of similarity amounting to
nation of knowledge. an increase in accuracy of about 3 0 to 40%,
and an effect of typicality of about 15 %.
There was no evidence of an interaction,
Coordination of Causal, Analytical,
suggesting that the two processes are inde-
and Exemplar-based Processes
pendent, or additive. The effect of similar-
At one level, it is almost self-evident that ity was higher in residents than medical stu-
multiple processes must be operating in dents, and the effect of typicality was higher
medical reasoning, if not simultaneously, at in medical students, perhaps because med-
least within the same problem. Although a ical students rely more on analytical rules,
physician may recognize a patients problem whereas with increasing experience, there
within seconds or minutes, she then com- is greater use of exemplar-based reason-
monly goes through a more-or-less system- ing. A subsequent study (Kulatanga-Moruzi,
atic search for additional data before she Brooks, & Norman, 2004) tested expert
expertise in medicine and surgery 347
dermatologists, general practitioners, and appears to be consistent with our earlier sec-
residents with a series of dermatology lesions tions, as well as previous writing (Ericsson,
in three forms a verbal description, a 2004). We now have some evidence that sur-
description followed by a photo, and a pho- gical expertise is acquired and highly local.
tograph alone. Resident performance was The ability to perform one task derives from
best with the verbal description with or specific practice with that task and does not
without the photograph, and worst with generalize to other, even apparently similar,
just the photograph. However, general prac- surgical tasks.
titioners and dermatologists did best with Recent transfer-of-learning research in-
just the photograph, and worst with the volves training within a given task and has
verbal description. Again, this suggests that been driven by developments in surgical
greater experience results in greater reliance simulators, ranging from simple inanimate
on exemplar-based knowledge, and less on bench models to computer-based virtual
formal rules. reality systems. Several studies (Anastakis
et al., 1999; Matsumoto et al., 2002; Grober
et al., 2004) have now shown that technical
Expertise and the Acquisition skills acquired on low-fidelity bench models
of Technical Skills transfer to improved performance on higher-
fidelity models (such as human cadavers), as
well as live patients in the operating room,
Although the discussion so far has focused
both in laparoscopic surgery (Scott et al.,
largely on expertise in terms of diagnos-
2000) and anaesthesia (Naik et al., 2001). In
tic ability, a significant component of med-
these studies, the authors identified essential
ical practice also involves technical skill.
constructs inherent in the relevant techni-
However, there has been very little work
cal tasks and developed low-fidelity bench
on expertise in surgery and related techni-
models to facilitate transfer of these con-
cal fields. There are numerous reports of a
structs to the clinical setting. One interpre-
positive relation between surgeon volume
tation of this is that transfer occurs because
and patient outcome (e.g., Halm, Lee, &
the low-fidelity models preserve the func-
Chassin, 2002), but there are virtually no
tional (process) aspects of training, and this
studies that address the direct assessment of
appears to be more important than struc-
expert surgical performance. Almost all of
ture, at least in the case of surgery and
the related research has involved educational
related technical disciplines; that is, trainees
and training issues, such as the development
are asked to suspend disbelief about the
of valid measures for curriculum evaluation
physical structure and focus on the pro-
or identification of trainees for remediation
cess of the task. Once they learn the pro-
(e.g., Moorthy et al., 2004). These mea-
cess components, it is apparently relatively
sures evolved from validation studies involv-
trivial to transfer the task across physical
ing gross differences in relative performance
structures.
(e.g., senior residents vs. novices), and thus
However, transfer of learning across sur-
are relatively insensitive to distinctions of
gical tasks is a different story. For exam-
expertise among practicing clinicians. But
ple, Wanzel et al. (2002) had novice trainees
perhaps we can gain some clues about the
learn a simple two-flap Z-plasty (a proce-
process of becoming an expert by examining
dure to rotate skin flaps along a Z-shaped
this literature, including transfer of learning
line) and found that some trainees had dif-
and correlates of performance.
ficulty in transferring to a more spatially-
complex surgical task on the same physical
Learning and Transfer of Surgical Skills
model, again highlighting the specificity of
A popular notion is that surgeons possess process learning. An interesting additional
innate talent; however, the limited body finding was that those who had scored signif-
of research touching on surgical expertise icantly lower on tests of visuo-spatial ability
348 the cambridge handbook of expertise and expert performance
had more difficulty in the transfer task, lead- ipants were retested on the Z-plasty, and
ing to more questions about the correlates of those with low visuo-spatial test scores per-
surgical performance. formed as well as the higher-scoring group.
This suggests, at most, that the learning
curves might be different. It appears that
Correlates of Surgical Performance
individuals with low visuo-spatial test scores
Several studies have examined characteris- may have more difficulty initially in per-
tics of surgical trainees in an attempt to pre- forming a spatially complex surgical task,
dict performance in technical skills. By and but that they can learn the task with minimal
large, demographic information (e.g., age, practice and training. In a follow-up study,
gender) (Schueneman et al., 1985 ; Risucci expert craniofacial surgeons who perform
et al., 2001), medical school grades (Keck spatially complex surgical tasks on a regu-
et al., 1979; Papp et al., 1997), and manual lar basis were found to have visuo-spatial
dexterity (Schueneman et al., 1985 ; Squire test scores and manual dexterity around the
et al., 1989; Steele et al., 1992; Francis et al., norm (Wanzel et al., 2003 ), suggesting that
2001) fail to correlate with surgical ability. It their expertise is related less to complex spa-
is not surprising that variables such as medi- tial abilities or manual skills than repeated
cal school grades yield no relation to techni- practice under the carefully controlled
cal skill since these assessments by and large conditions of training during residency
measure cognitive ability, which may have and fellowship.
very little to do with technical competence. It is conceivable that intimate knowl-
But why not the manual dexterity test? edge of and experience with surgical tasks,
A popular lay notion is that surgery requires when combined with competent intraoper-
fine psychomotor control. However, among ative judgment, may overshadow any advan-
surgeons, it has been hypothesized that tage afforded by superior visuo-spatial abil-
visuo-spatial ability is significant (Grace, ity. The ultimate skill may not be related to
1989; Cuschieri, 1995 ). Recent work on the same mechanisms that mediate the ini-
the relative importance of visuo-spatial and tial performance. For novices, innate abili-
psychomotor ability has found strong cor- ties may help in acquiring technical skills,
relations between scores on higher-level whereas for experts, experience alone may
visuo-spatial tests (i.e., the ability to men- significanctly determine the acquisition of
tally manipulate and rotate complex three- technical skills independent of or perhaps
dimensional objects) and efficiency of hand in spite of innate abilities.
motion in novice surgical trainees, with
Pearson correlations ranging from 0.40 to
0.5 8 (Wanzel et al., 2002; Wanzel et al., Aging and Medical Expertise
2003 ). Surprisingly, manual dexterity did
not correlate with hand motion, suggesting Despite three decades of research focused
that efficient hand motion during surgery on the development and nature of medical
may be more closely related to planning expertise, little attention has been paid to
and preoperative visualization than precise the dynamic relationship between age and
motor control during subtasks, thus calling medical expertise. Krampe and Charness
into question the importance of a steady (Chapter 40) address the relationship in
hand. other domains, but as indicated at the
Is surgical expertise therefore a matter beginning of this chapter, medical expertise
of innate visuo-spatial abilities? We suspect has proven to be sufficiently unique that
not. Wanzel et al. (2002) found that resi- the impact of age in this specific context
dents scores on selected visuo-spatial tests must be considered. As the average age of
correlated strongly with performance on the medical practitioners increases along with
Z-plasty. However, following ten minutes the population, delineation of expertise in
of supervised practice and feedback, partic- this older and more experienced subgroup
expertise in medicine and surgery 349
and diagnostic success. Medical Education, 3 7, Feltovich, P. J., Johnson, P. E., Moller, J. H., &
695 703 . Swanson, D. B. (1984). The role and devel-
Coughlin, L. D., & Patel, V. L. (1987). Processing opment of medical knowledge in diagnostic
of critical information by physicians and med- expertise. In W. J. Clancey & E. H. Shortliffe
ical students. Journal of Medical Education, 62 , (Eds.), Readings in medical artificial intelligence:
818828. The first decade. Reading, MA: Addison Wesley.
Coulson, R. L., Feltovich, P. J., & Spiro, R. J. Feltovich, P. J., & Barrows, H. S. (1984). Issues
(1997). Cognitive flexibility in medicine: An of generality in medical problem solving. In
application to the recognition and understand- H. G. Schmidt & M. L. DeVolder (Eds.),
ing of hypertension. Advances in Health Sci- Tutorials in problem-based learning. Assen, the
ences Education, 2 , 141161. Netherlands: Van Gorcum.
Cuschieri, A. (1995 ). Visual displays and visual Feltovich, P. J., Coulson, R. L., & Spiro, R. J.
perception in minimal access surgery. Seminars (1997). Issues of expert flexibilty in contexts
in Laparoscopic Surgery, 2 , 209214. characterized by complexity and change. In
Custers, E. J., Boshuizen, H. P., & Schmidt, H. G. P. J. Feltovich, K. D. Ford, & R. R. Hoff-
(1996). The influence of medical expertise, case man (Eds.), Expertise in context: Human and
typicality and illness script component on case machine. Menlo Park, CA: AAAI / MIT Press.
processing and disease probability estimates. Francis, N. K., Hanna, G. B., Cresswell, A. B.,
Memory and Cognition, 2 4, 3 843 99. Carter, F. J., & Cuschieri, A. (2001). The per-
de Groot, A. D. (1978). Thought and choice in formance of master surgeons on standard apti-
chess. The Hague: Mouton. tude testing. American Journal of Surgery, 182 ,
3 03 3 .
Elstein, A. S., Shulman, L. S., & Sprafka, S. A.
(1978). Medical problem solving: An analysis Gobet, F., & Simon, H. A. (2000). Five seconds
of clinical reasoning. Cambridge, MA: Harvard or sixty? Presentation time in expert memory.
University Press. Cognitive Science, 2 4, 65 1682.
Ericsson, K. A. (2000). How experts attain and Grace, D. M. (1989). Aptitude testing in surgery.
maintain superior performance: Implications Canadian Journal of Surgery, 3 2 , 3 963 97.
for the enhancement of skilled performance in Grober, E. D., Hamstra, S. J., Wanzel, K. R.,
older individuals. Journal of Aging and Physical Reznick, R. K., Matsumoto, E. D., Sidhu, R. S.,
Activity, 8, 3 463 5 2. & Jarvi, K. A. (2004). The educational impact
Ericsson, K. A. (2004). Deliberate practice and of bench model fidelity on the acquisition
the acquisition and maintenance of expert per- of technical skill: The use of clinically rele-
formance in medicine and related domains. vant outcome measures. Annals of Surgery, 2 40,
Academic Medicine, 79, S70S81. 3 743 81.
Eva, K. W. (2002). The aging physician: Changes Gruppen, L. D., & Frohna, A. Z. (2002). Clini-
in cognitive processing and their influence cal reasoning. In G. R., Norman, D. I., New-
on medical practice. Academic Medicine, 77 , ble, & C. P. M, van der Vleuten (Eds.), Interna-
S1S6. tional Handbook of Research in Medical Educa-
Eva, K. W., Brooks, L. R., & Norman, G. R. tion (pp. 205 23 0). Dordrecht: Kluwer.
(2001). Does shortness of breath = dyspnea: Halm, E. A., Lee. C., & Chassin, M. R. (2002). Is
The biasing effect of feature instantiation in volume related to outcome in health care? A
medical diagnosis. Academic Medicine, 76, S11 systematic review and methodological critique
S13 . of the literature. Annals of Internal Medicine,
Eva, K. W., Norman, G. R., & Brooks, L. R. 13 7 , 5 115 20.
(2002). Forward reasoning as a hallmark of Hashem, A., Chi, M. T. H., & Friedman, C. P.
expertise in medicine: logical, psychological, (2003 ). Medical errors as a result of special-
phenomenological inconsistencies. Advances in ization. Journal of Biomedical Informatics, 3 6,
Psychology Research, 8, 25 40. 6169.
Eva, K. W., Norman, G. R., Neville, A. J., Hassebrock, F., Bullemer, P., & Johnson, P. E.
Wood, T. J., & Brooks, L. R. (2002). Expert / (1988). When less is more: Selective memory
novice differences in memory: A reformula- of problem-solving experts. Paper presented at
tion. Teaching and Learning in Medicine, 14, the annual meeting of the American Educa-
25 7263 . tional Research Association, New Orleans.
352 the cambridge handbook of expertise and expert performance
Hatala, R. M., Norman, G. R., & Brooks, L. R. testinal simulator and cross validation using
(1999). Influence of a single example upon structured assessment of individual perfor-
subsequent electrocardiogram interpretation. mance with video playback. Surgical Endoscopy,
Teaching and Learning in Medicine, 11, 110117. 18, 3 283 3 3 .
Hatala, R. M., Norman, G. R., & Brooks, L. R. Muzzin, L. J., Norman, G. R., Feightner, J. W., &
(2003 ). Practice makes perfect: The critical Tugwell, P. (1983 ). Expertise in recall of clin-
role of deliberate practice in the acquisition of ical protocols in two specialty areas. Proceed-
ECG interpretation skills. Advances in Health ings of the 2 2 nd Conference on Research in Med-
Sciences Education, 8, 1726. ical Education (pp. 122127). Washington, DC:
Higgins, M. P., & Tully, M. P. (2005 ). Hospital Association of American Medical Colleges.
doctors and their schemas about appropriate Naik, V. N., Matsumoto, E. D., Houston, P. L.,
prescribing. Medical Education, 3 9, 184193 . Hamstra, S. J., Yeung, R. Y. M., Mallon, J. S.,
Hobus, P. P., Schmidt, H. G., Boshuizen, H. P., & Martire, T. M. (2001). Fiberoptic orotra-
& Patel, V. L. (1987). Contextual factors in cheal intubation on anesthetized patients: Do
the activation of first diagnostic hypotheses: manipulation skills learned on a simple model
Expert-novice differences. Medical Education, transfer into the operating room? Anesthesiol-
2 1, 471476. ogy, 95 , 3 43 3 48.
Keck, J. W., Arnold, L., Willoughby, L., & Nendaz, M. R., & Bordage, G. (2002). Promot-
Calkins, V. (1979). Efficacy of cognitive/ ing diagnostic problem representation. Medical
noncognitive measures in predicting resident- Education, 3 6, 760766.
physician performance. Journal of Medical Edu- Neufeld, V. R., Norman, G. R., Barrows, H. S., &
cation, 5 4, 75 965 . Feightner, J. W. (1981). Clinical problem solving
Kim, N. S., & Ahn, W. (2002). The influence of by medical students: A longitudinal and cross-
nave causal theories an lay concepts of mental sectional analysis. Medical Education, 15 , 3 15
illness. American Journal of Psychology, 115 , 3 3 3 22.
65 . Norcini, J. J., Lipner, R. S., & Kimball, H. R.
Kintsch, W. (1974). The representation of meaning (2002). Certifying examination performance
in memory. Hillsdale, NJ: Erlbaum. and patient outcomes following acute myocar-
Kulatanga-Moruzi C., Brooks L. R., & Norman, dial infarction. Medical Education, 3 6, 85 3
G. R. (2001). Coordination of analytical 85 9.
and similarity based processing strategies and Norman, G. R., Brooks, L. R., & Allen, S. W.
expertise in dermatological diagnosis. Teaching (1989). Recall by experts and novices as a
and Learning in Medicine, 13 , 110116. record of processing attention. Journal of Exper-
Kulatanga-Moruzi, C., Brooks, L. R., & Norman, imental Psychology: Learning, Memory and Cog-
G. R. (2004). The diagnostic disadvantage of nition, 15 , 11661174.
having all the facts: Using comprehensive fea- Norman, G. R., Trott, A. L., Brooks, L. R., &
ture lists to bias medical diagnosis. Journal of Smith, E. K. M. (1994). Cognitive differences
Experimental Pychology: Learning, Memory and in clinical reasoning related to postgraduate
Cognition, 3 0, 5 63 5 72. training. Teaching and Learning in Medicine, 6,
Matsumoto, E. D., Hamstra, S. J., Radomski, S. B., 114120.
& Cusimano, M. D. (2002). The effect of bench Papp, K. K., Polk, H. C. J., & Richardson, J. D.
model fidelity on endourologic skills: A ran- (1997). The relationship between criteria used
domized controlled study. Journal of Urology, to select residents and performance during res-
167 , 1243 1247. idency. American Journal of Surgery, 173 , 3 26
Medin, D. L., Altom, M. W., & Murphy, 3 29.
T. D. (1984). Given versus induced category Patel, V. L., & Groen, G. J. (1986), Knowledge
representations: Use of prototype and exem- based solution strategies in medical reasoning.
plar information in classification. Journal of Cognitive Science, 10, 91115 .
Experimental Psychology: Learning, Memory and Patel, V. L., Evans, D. A., & Groen, G. J. (1989).
Cognition, 10, 3 3 3 3 5 2. Biomedical knowledge and clinical reasoning.
Moorthy, K., Munz, Y., Jiwanji, M., Bann, S., In D. A. Evans & V. L. Patel (Eds.), Cognitive sci-
Chang, A., & Darzi, A. (2004). Validity and ence in medicine: Biomedical modelling (pp. 127
reliability of a virtual reality upper gastroin- 166). Mahwah, NJ: Erlbaum.
expertise in medicine and surgery 353
Regehr, G., Cline, J., Norman, G. R., & Brooks, manual dexterity in physicians and surgeons.
L. R. (1994). Effect of processing strategy Canadian Journal of Surgery, 3 2 , 467470.
on diagnostic skill in dermatology. Academic Steele R. J., Walder, C., & Herbert, M. (1992).
Medicine, i, S3 4S3 6. Psychomotor testing and the ability to per-
Rikers, R. M., Loyens, S. M., & Schmidt, H. G. form an anastomosis in junior surgical trainees.
(2004). The role of encapsulated knowledge British Journal of Surgery, 79, 1065 1067.
in clinical case representations of medical stu- Sternberg, R. J., & Frensch, P. A. (1992). On
dents and family doctors. Medical Education, being an expert: A cost-benefit analysis. In
3 8, 103 5 1043 . R. Hoffman (Ed.), The psychology of expertise:
Risucci, D., Geiss, A., Gellman, L., Pinard, B., Cognitive research and empirical artificial intelli-
& Rosser, J. (2001). Surgeon-specific factors in gence. New York: Springer-Verlag.
the acquisition of laparoscopic surgical skills. Tamblyn, R. M., Abrahamowicz, M., Dauphi-
American Journal of Surgery, 181, 289293 . nee, W. D., Hanley, J. A., Norcini, J., Girard,
Rosch, E. H., & Mervis, C. B. (1976). Family N., GrandMaison, P., & Brialovsky, C. (2002).
resemblances: Studies in the internal structure Association between licensure examination
of categories. Cognitive Psychology, 7, 5 73 605 . scores and practice in primary care. Journal of
Schueneman, A. L., Pickleman, J., & Freeark, the American Medical Association, 2 88, 3 019
R. J. (1985 ). Age, gender, lateral dominance, 3 026.
and prediction of operative skill among general Turnbull, J., Carbotte, R., Hanna, E., Norman,
surgery residents. Surgery, 98, 5 065 15 . G. R. et al. (2000). Cognitive difficulty in
Schmidt, H. G., & Boshuizen, H. P. A. (1993 ). physicians. Academic Medicine, 75 , 177181.
On the origin of intermediate effects in clini- Van Rossum, H. J. M., & Bender, W. W. (1990).
cal case recall. Memory and Cognition, 2 1, 3 3 8 What can be learned from a boy with acute
3 5 1. appendicitis? Persistent effects of a case pre-
Schmidt, H. G., Norman, G. R., & Boshuizen, sentation on the diagnostic judgment of family
H. P. (1990). A cognitive perspective on med- doctors. Presented at the Fourth Ottawa Con-
ical expertise: Theory and implications. Aca- ference, Ottawa, Ontario.
demic Medicine, 65 , 611621. Verkoeijn, P. P., Rikers, R. M., Schmidt, H. G., van
Schuwirth, L., Gorter, S., van der Heijde, D., de Wiel, M. W., & Kooman, J. P. (2004). Case
Rethans, J. J., Brauer, J., Houben, H., van der representation by medical experts, intermedi-
Linden, S., van der Vleuten, C., & Scherpbier, ates and novices for laboratory data presented
A. (2005 ). The role of a computerized case- with or without a clinical context. Medical Edu-
based testing procedure in practice perfor- cation, 3 8, 617627.
mance assessment. Advances in Health Sciences Wanzel, K. R., Hamstra, S. J., Anastakis, D. J.,
Education, 10, 145 15 5 . Matsumoto, E. D., & Cusimano, M. D. (2002).
Scott, D. J., Bergen, P. C., Rege, R. V., Laycock, Effect of visual-spatial ability on learning of
Z. R., Tesfay, S. T., & Valentine, R. J., Euhus, spatially-complex surgical skills. Lancet, 3 5 9,
D. M., Jeyarajah, D. R., Thompson, W. M., & 23 023 1.
Jones, D. B. (2000). Laparoscopic training on Wanzel, K. R., Hamstra, S. J., Caminiti, M. F.,
bench models: Better and more cost effective Anastakis, D. J., Grober, E. D., & Reznick,
than operating room experience? Journal of the R. K. (2003 ). Visual-spatial ability correlates
American College of Surgeons, 191, 272283 . with efficiency of hand motion and successful
Simon, H. A., & Chase, W. G. (1973 ). Skill at surgical performance. Surgery, 13 4, 75 075 7.
chess. American Scientist, 61, 3 94403 . Woods, N. N., Brooks, L. R., & Norman, G. R.
Squire, D., Giachino, A. A., Profitt, A. W., & (2005 ). The value of basic science. Medical
Heaney, C. (1989). Objective comparison of Education, 3 9, 107112.
C H A P T E R 20
There are more expert drivers in the United Instead, we chose to look at relative dif-
States than any other type of expert. A ferences in experience. Table 20.1 details the
3 5 -year-old Los Angelino who commutes participant characteristics for a number of
an hour to work and travels only minimally the studies reviewed here, along with one
on weekends will have spent over 10,000 modern-day classification scheme (Hoff-
hours behind the wheel, a number some- man, 1996). Expertise is usually defined by
times held up as a threshold for exper- number of years operating the vehicle, or
tise (Chase & Simon, 1973 ). If, however, miles driven, or hours flown by the operator.
instead of a time-based definition, one takes Table 20.2 shows a variety of comparisons in
as the criterion for expertise performing a relative experience.
task better than someone with less expe- We begin by briefly considering the
rience, then our opening assertion about nature of transportation tasks and how
expert drivers in the United States is less experience affects performance in those
apparent. tasks. From there, the chapter carves trans-
In the transportation domain, defining portation into three underlying psychologi-
expertise in any absolute sense is nontrivial. cal components, attention, perception, and
Self-evaluations, as is often the case, place knowledge.
most drivers as above average (Waylen,
Horswill, Alexander, & McKenna, 2004) and
The Transportation Domain
are at best weakly correlated with evalua
tions of a driving instructor (Groeger & The sheer number of operators controlling
Grande, 1996). If we searched the litera- vehicles warrants a better understanding of
ture for highly experienced operators back these human-technical systems. Also, trans-
to the turn of the last century, we would portation offers a crucible in which basic
find a plethora of transportation studies, but laboratory findings can be tested and from
they would not inform modern notions of which insights can be gleaned to inform
expertise. basic theory.
355
356 the cambridge handbook of expertise and expert performance
Table 2 0.1. Classification of participants used in studies cited in the chapter. The experience label
borrows from Hoffman (1996)
Criteria from selected
publications with a Summary Sample publications considered
Experience label inclusion rule in the chapter
Naivette: no experience or No driving experience; no past Mourant & Rockwell (1972)
knowledge of the domain drivers ed. Nonpilots Tsang & Voss (1996)
Novice: A beginner; initial Student pilots (average of Doane & Sohn (2004)
instruction 3 2 flight hours)
Apprentice: Undergoing ATC developmental Seamster et al. (1993 )
instruction <100 hours cross country Wiggins & OHare (1995 )
60 hours Sohn & Doane (2003 )
3 61 hours Schvaneveldt et al. (2001)
in class OR under 400 hours Schvaneveldt et al. (1985 )
flight time
Advanced apprentice/ Junior 1 to 5 2 wk post-license Duncan et al. (1991)
journeyman: 1 to 1.5 years driving Quenault & Parker (1973 )
Newly licensed with little <1 year FPL controller Shinar et al. (1998)
experience 101 to 1000 cross country Seamster et al. (1993 )
3 months post-license; Wiggins & OHare (1995 )
1047 miles Chapman & Underwood (1998)
<5 0 hours instrument Stokes et al. (1997)
907 flight hours Sohn & Doane (2003 )
2,000 km Wikman et al. (1998)
postlicense & (<2 year OR
2 000 km) OR
<1000 hours
Journeyman/Expert 20 years licensed driver Duncan et al. (1991)
5 5 00 flight hours Tsang & Voss (1996)
25 years licensed driver Shinar et al. (1998)
5 years as FPL controller Seamster et al (1993 )
8000 miles for 5 years Mourant & Rockwell (1972)
200,000 km Wikman et al. (1998)
>1000 hours cross country Chapman & Underwood (1998)
5 to 10 yrs; 79K miles Williams & ONeill (1974)
drivers license; random Stokes et al. (1997)
licensed drivers Schvaneveldt et al. (1985 )
Commercial license, instrument Schvaneveldt et al. (2001)
rating, & Quenault & Parker (1973 )
15 00 hours
43 00 hours
>5 years OR 4000 hours OR
(1000 hrs + other feature)
OR 40,000 miles
Experts/Exemplary Instructor pilots Schvaneveldt et al. (1985 )
journeymen trained to observe Duncan et al. (1991)
race car drivers license Williams & ONeill (1974)
Special feature plus many
years experience
Table 2 0.2 . Half-matrix showing relative comparisons extractable from articles considered in the chapter.
Nave/Novice Apprentice Advanced appren/ Jr Journeyman Journeyman/ Expert
Nave/novice
Apprentice
Advanced Seamster et al. (1993 )
Apprentice/ Junior Wiggins & OHare (1995 )
journeyman Sohn & Doane (2003 )
Journeyman/ Expert Mourant & Rockwell (1972) Seamster et al. (1993 ) Duncan et al. (1991)
Tsang & Voss (1996) Wikman et al. (1998) Seamster et al. (1993 )
Schvaneveldt et al. (2001) Shinar et al. (1998)
Schvaneveldt et al. (1985 ) Chapman & Underwood (1998)
Quenault & Parker (1973 ) Stokes et al. (1997)
Expert/Master Schvaneveldt et al. (1985 ) Duncan et al. (1991) Williams & ONeill (1974)
Duncan et al. (1991)
Schvaneveldt et al. (1985 )
357
358 the cambridge handbook of expertise and expert performance
The most notable feature of the crucible is of growing older (poorer performance) and
that transportation takes place in a dynamic growing more expert (better performance).
environment. Many studies of expertise take In fieldwork, alternative hypotheses can be
place in static or less than fully dynamic envi- eliminated, if not from each study, at least
ronments. Studies of expertise in chess, for from the confluence of studies.
example, are at best semi-dynamic. How- Overall, studying outcome variables, like
ever, in transportation, the environment will safety, in transportation tasks (especially in
change whether or not the operator takes an the wild) will likely suggest multiple causes,
action. Expertise in modern dynamic envi- only one of which will be experience. For
ronments (see Cellier, Eyrolle, & Marine, example, drivers differ in their styles and
1997, for a review) has been of interest since willingness to accept risk. They also are able
the 1970s with process control (DeKeyser & to engage in compensatory behaviors, allow-
Piette, 1970), aviation (Bisseret, 1971), and ing more distance by driving farther back
driving (Mourant & Rockwell, 1972) draw- when they believe their speed of reaction
ing the attention of numerous researchers. is slowed (e.g., Groeger, 2000). Thus, the
Transportation offers complex embeddings relationship between experience and rep-
of cognitive factors in an operator who is resentative performance can be weak (see
using technology that is embedded in a Ericsson, Chapter 3 8). Finally, the rela-
dynamic workplace that in turn is embedded tionship between experience and a variable
in a social setting of a rich culture. This terroir like safety raises the question: What kind
is certain to add complexity to our theoriz- of experience (McKinney & Davis, 2003 )?
ing about expertise. For example, in domains Expertise in commercial flying is a function
like chess, it is not surprising that expertise of the aircraft; an expert flying a regional
presents itself as heavily cognitive. In more jet may not be an expert flying a 75 7. On
dynamic domains with perceptual invariants the other hand, train engineers are required
in the environment and choices among tasks additional training and certification for spe-
for the operator, the analyses may offer new cific geographical areas rather than type of
insights into expertise. train. For our Los Angelena, the routine
The complexity of transportation also car- experience of her daily commute may be of
ries a price: Variables of interest, such as little help if she must swerve to miss the
expertise for this volume, are entangled with statue of Jabba the Hutt that fell off of the
other factors, such as aging. So, the expert truck in front of her.
driver and the experienced pilot found in the
literature are typically older than the less-
Experience in Transportation
experienced ones. Of course, efforts can be
made to control or assess confounding vari- The value of experience in transportation
ables, and studies reported here often make tasks has been documented for a number of
attempts to equate groups on age either years and is reflected in everyday life in acci-
by matching or statistical control. How- dent records and insurance rates. Research in
ever, these are not totally satisfactory, not transportation dates back to the early 20th
only because the matching is rarely pre- century, with selection and aptitude tests
cise, but also because matching on age, for of military (Henmon, 1919) and transport
instance, can sometimes introduce other dif- pilots (Snow, 1926). Research on experience
ferences between experts and nonexperts. and driving can be traced back at least to
For example, other confounding variables the work of Allgaier (193 9), who showed
may be introduced when people with ten that although reaction time and visual acuity
years of driving experience are matched on were best for the younger driver, acci-
age with drivers newly licensed. Thus, some dent rates and traffic violations decreased
studies forego methodological fixes for the with experience. World War II brought the
confounding with age, interpretation relying need for more sophisticated and depend-
instead on the anticipated opposite effects able equipment. The new field of human
expertise and transportation 359
factors placed greater emphasis on the skills were observers from Britains Institute of
of the operators and their interaction with Advanced Motorists and had held their
the equipment. That focus on the relation- license for an average of 18 years; journey-
ship between experience and performance man drivers had held a license for about
skills continued. the same, 20 years; apprentices had received
For example, in a field study reported in their license within the year. Mean ages were
1973 , Quenault and Parker studied appren- similar, between 3 5 and 41. Of interest to
tice drivers matched for age; the more- the current chapter are the findings that
experienced apprentices (as much as a year apprentices and experts were sometimes rel-
of practice) showed better car control than atively similar to each other, with the nor-
the less-experienced ones (as little as a mal journeyman drivers performing most
week of post-license practice). Further, all poorly: Apprentices and experts checked
of these apprentices drove more poorly mirrors comparably frequently and applied
(e.g., near accidents) than older, experienced brakes comparably early, whereas journey-
drivers. Advantages of experience continue man drivers with 20 years experience infre-
to accumulate; for example, when Recarte quently checked the mirrors and braked sig-
and Nunes (1996) gave their participants nificantly later. This disconnection between
the opportunity to control the accelerator experience and performance can manifest
to obtain a target speed, the group with no for many reasons. Driving accident rates for
driving experience had a greater mean error those learning to drive have been reported
than the groups with driving experience. at less than one in one hundred, whereas
However, experienced operators do not about one in five apprentices one year post
always outperform less experienced ones license are involved in an accident (Forsyth,
(Duncan, Williams, & Brown, 1991; see Maycock, & Sexton, 1995 ). Driving under
Shanteau, 1992, for other examples). Every supervision, or driving with less arrogance, or
year, the public is baffled by incidents both may explain these results. Williams and
attributable to errors made by experienced ONeill (1974) showed that licensed race-
airline pilots. For example, in March 2000, a car drivers had more on-the-road accidents
Southwest Airlines flight overran the run- than did controls. Racecar drivers may be
way in Burbank, California stopping just more risk tolerant, or the type of experience
feet from a gas pump. The captain had gained during a race does little for on-the-
11,000 flight hours. Although weather condi- road safety.
tions were VFR (visual flight rules) and the Our approach to understanding these
winds were relatively calm at the surface, patterns of data relating experience is to con-
the pilot flew the approach at 5 0 mph above sider how experience affects underlying psy-
the recommended approach speed for the chological processes. Along the way, we stop
conditions and the type of airplane (NTSB, to point out how basic research on expertise
2004). can be informed by applied work.
More than anecdotes suggest that the
relationship between experience and perfor-
mance can be complicated in the domain Attention
of transportation. The literature on exper-
tise and transportation also includes some An excellent example of basic research being
studies that reach the interesting conclu- informed by applied work is the study of
sion that recognized expert drivers do not expert pilots in low-altitude flight. That
necessarily have fewer accidents on public work forces us to revisit our beliefs about the
roads than would our hypothetical LA role of attention and resource management
commuter. Duncan et al. (1991) compared of skilled performance. Haber and Haber
experts, journeyman drivers, and advanced (2003 ) rightly view low-altitude air combat
apprentices in an instrumented car over real- as at the very edge of human perceptual and
world urban and rural routes. The experts attentional capabilities (p. 21). The pilot is
360 the cambridge handbook of expertise and expert performance
flying close to the speed of sound, so near or because of more efficient management of
the ground that controlled flight into terrain controlled processes.
(CFIT) is the primary concern, while try- Although low-altitude combat is an
ing to complete the mission and avoid being extreme form of transportation, it does
shot down. rely heavily on attentional components that
Haber and Haber (2003 ) review possi- are found in all transportation tasks. We
ble sources of information that pilots could cover two of them here, namely, perform-
use to determine their altitude above ground ing tasks simultaneously (dual-task perfor-
level (AGL) and find that both preflight mance) and finding information in the envi-
information (such as familiar size, experi- ronment (visual search).
ence with the route) and automatic percep-
tual processes (such as optic flow, vestibu-
Dual-Task Performance
lar information) are insufficient or even
misleading in the determination of AGL. Tsang and Voss (1996) looked at exper-
Instead, contrary to the reasonable expecta- tise in time-sharing by comparing pilots
tion that such high-speed processing would (who were presumed to be experts in
rely on automatic processing, Haber and time-sharing) with nonpilots. Pilots did, in
Haber effectively argue for the necessity fact, show a smaller decrement in perfor-
of focused, controlled attention in these mance in a dual task (horizontal tracking
extreme situations. plus some other task (e.g., vertical track-
Expertise in low-altitude military flying ing) than did nonpilots, and this was espe-
comes from explicit training and deliberate cially true for elderly participants (over
practice not only on the maneuvers unique 60 years). Thus, the results support the con-
to these environments, but also on task man- clusion that experience can mitigate age
agement and prioritization. For example, decrements in dual-task performance. In
pilots learn about free time (time available addition, if we accept pilot experience as a
for tasks before a CFIT is inevitable) learn surrogate for time-sharing expertise, the data
about times required by tasks, and learn to support the notion that resource manage-
abandon tasks failed during the free time ment is part of what develops with expertise.
so that ground clearance tasks regain pri- Other transportation tasks also require man-
ority. This training and resulting expertise aging more than one subtask. An often-used
on resource management is necessary before classroom example is, of course, changing
a pilot can be effective on an operational gears in a manual-shift car. Apprentices do
mission. indeed take longer to downshift than experts
We know from the literature that expert (Duncan et al., 1991). Experienced drivers
performance on domain-relevant tasks is less have been shown not only to shift faster,
constrained by resource limits than is novice but also to be unaffected by an additional
performance (e.g., Chi, Glaser, & Farr, 1988; cognitive load (randomly generating letters
Ericsson & Smith, 1991). Most researchers while driving; Duncan, Williams, Nimmo-
would argue that resources are less con- Smith, & Brown, 1992). Shinar, Meir, and
straining because tasks have become auto- Ben-Shoham (1998) looked at experience
mated. Thus, in tasks more mundane than with automatic and manual-shift cars in a
low-altitude combat, we see that exper- field setting. The study took place in busy
tise is a factor in resource allocation, but downtown Tel Aviv, presumably insuring a
not necessarily by developing sophisticated considerable number of gear changes. The
techniques for performing tasks serially as dependent variable was success at detect-
Haber and Habers pilots seem to do. In ing designated traffic signs (i.e., Slow
dynamic environments, freedom from con- Children on the Road and No Stopping).
straints accompanies expertise, and this free- Experienced drivers, those with a median
dom may occur because of automatization driving experience of nine years, were
expertise and transportation 361
instruments (e.g., Bellenkes, Wickens, & greater percentage of extreme (short or long)
Kramer, 1997). glances from the road to the shared task.
Experienced and inexperienced drivers Short looks are likely ineffective, and long
have been thought to differ in some aspects looks dangerous. Forty percent of the male
of their eye-movement patterns. For exam- apprentices looked away from the road for
ple, some evidence exists that apprentices over three seconds; experts never took their
concentrate their search in a smaller area, eyes off the road for this length of time. Dis-
closer to the front of the vehicle, than turbing is the possibility that this tendency
do journeymen who look at the focus of to take longer fixations may be especially
expansion (e.g., Mourant & Rockwell, 1972; true in dangerous situations (Chapman &
cf. Wikman, Nieminen, & Summala, 1998). Underwood, 1998).
Beginners are thought to use foveal vision to Thus, experts and apprentices seem to
keep the car on the road, whereas experi- differ in their visual search patterns, but it is
enced drivers make better use of peripheral not always the case that expertise differences
vision (Wikman et al., 1998). will be found in eye-movement patterns.
Although reliable scanning differences Part of this difficultly may rest with the vari-
are sometimes uncovered, scanning differ- ability produced in the behaviors of novices,
ences are often swamped by the variability but this variability may be more than an
in novice or apprentice data. This difference inconvenience to statistical tests. This vari-
is not surprising if one assumes that novices ability may reflect the processes that novices
have yet to develop the underlying cognitive employ, the controlled nature of those pro-
representations that allow for good supervi- cesses, and may prove to be a fundamental
sory control. Greater variability in novices characteristic of novice cognition in dynamic
is often taken as an indication that auto- environments.
mated processes have not been developed.
Wikman et al. (1998), for example, suggest
that automated performance is more stereo- Perception
typed and consistent. More variable does not
imply, however, more flexible. Crundall and Although experienced and inexperienced
Underwood (1998) found less-experienced operators may scan their environments dif-
drivers showed no differences when driving ferently, there is also evidence that experts
on a quiet rural road as compared to driving and nonexperts perceive the world dif-
a busier multi-lane road. With experience, ferently. It is likely that experts have
drivers develop particular scanning strategies learned to interpret not just informational
for particular situations. cues, but also cue configurations and struc-
The difference in expert and novice scan- tural invariants in their dynamic environ-
ning strategies can also bear on understand- ments that apprentices have not learned to
ing resource management. If one views perceive: The blurring of the passing land-
supervisory control as related to resource scape can substitute nicely for a glance at
management, then we again observe support the speedometer. In an experiment where
for a resource management deficit in nonex- drivers viewed video clips photographed
perts. Consider Wikman et al. (1998), who from inside a minibus, experienced drivers
asked young apprentices and older, experi- seemed to have no problem judging distance
enced drivers to drive an instrumented com- when switching back and forth between tar-
pact car on a 126 km route in three hours. gets in a convex mirror (when the angle of
During this time the participants were occa- convexity is appropriate) and objects in the
sionally asked to timeshare with another task direct visual field (Fisher & Galer, 1984).
(e.g., dial their own number on a cellphone, Thus, when it is said that experts and novices
search for soft music on the radio). Of see the world differently, it can mean several
interest here is the finding that apprentices things. An expert might notice things that
glances were more variable, and included a an apprentice does not (e.g., an oil spot); an
expertise and transportation 363
expert might interpret an object as a threat skill and allows more precise focus on hazard
(e.g., to braking), whereas the apprentice detection. Provided that the experienced
may see the same object but not see it as operators are substantially skilled (say ten
threatering; or, both might see the object years of experience), it is typical to find an
as a threat, but the expert sees also the expertise advantage. For example, McKenna
solution (e.g., swerve). See Klein (1997) for and Cricks (1991) experienced drivers (more
a discussion of recognition-primed decision than ten years of experience) reacted faster
making. to hazards than did drivers with less than
Perceiving and interpreting patterns can three years of experience.
be considered from a number of perspec- We noted above that experts seem to be
tives. Here we consider two overlapping less affected by load in a number of tasks,
areas of research: how well an operator can presumably because those tasks had become
detect a problem (hazard detection), and automatic, although efficient resource man-
how well an operator understands the sit- agement of controlled processes was also
uation (situation awareness). sometimes implicated. There seems to be
at least some evidence that hazard per-
ception is not automatic, but rather is a
Hazard Detection
controlled, effortful process. McKenna and
If experts have freed resources by automat- Farrand (1999) studied drivers who com-
ing some constituent processes or by supe- pleted a random letter-generation task dur-
rior resource management, what might they ing a hazard-perception test. Experienced
do with those freed resources? The trans- drivers showed more interference on the
portation literature suggests that one thing hazard-perception task than did inexperi-
they do is apply the resources to hazard enced drivers. If low-altitude military flight
detection (Horswill & McKenna, 2004). requires controlled processing (Haber &
Horswill and McKenna (2004) point out Haber, 2003 ) as we discussed earlier, it may
that despite evidence from driving research be because hazard detection requires it. Pro-
that experience leads to automatic per- cesses responsible for skilled control of the
formance, this automaticity does not pre- plane itself may develop automaticity, but
dict safer driving (e.g., Williams & ONeill, the processing of hazards may not. Perhaps
1974). In fact, it is difficult to find studies there is simply not enough consistent map-
that suggest that driver skill, or better vehi- ping between some contextualized hazards
cle control skills, predict driver safety on and the appropriate response.
public roads. Perhaps it is driver style the Hazard detection is a special case of gath-
desire to speed and take risks (Wasielewski, ering and interpreting cues from the envi-
1984) rather than driver skill. Elander, ronment. In this sense, expertise differences
West, and Frenchs (1993 ) review of the lit- are ubiquitous. For example, Wiggins and
erature admits a place for driver style but OHare (1995 ) note evidence that experi-
also argues for the speed of hazard detection enced pilots acquire weather-related data
as a factor in driver performance. Perhaps from a menu-driven display more efficiently
the vehicle-control skills develop quickly, (less time, fewer returns to same informa-
leaving only the more cognitive aspects tion) than do apprentice or junior journey-
of driving, like hazard detection, with man pilots. The groups also tend to dif-
enough variance to account for differences in fer in the weather cues they view as useful
performance. (Wiggins & OHare, 1995 ).
In a prototypical hazard-detection study, Chapman and Underwood (1998) con-
the participant sits in front of a monitor and ducted a study of apprentice and jour-
presses a button when the situation presents neyman automobile drivers. Participants
a danger. The participant typically is not con- watched a video from the drivers perspec-
trolling the vehicle, a methodological nicety tive and were to indicate as soon as they
that eliminates differences in vehicle-control saw any hazardous event, while reaction
364 the cambridge handbook of expertise and expert performance
time and eye movements were recorded. and matching the dialogue into a preexisting
No difference in latency between groups schema.
was found; however, apprentice drivers had The schema notion also appears in
longer fixations than experienced drivers Randel, Pugh, and Reed (1996), who stud-
during periods of hazardous events. They ied electronic-warfare technician operators
also found the experienced drivers had on navy ships. The technician operator mon-
greater knowledge than apprentice drivers itors a screen that shows radar emitters of
did about the potential location of threat- hostile and friendly radar targets. He or she
related information. must monitor hostile and friendly traffic
It seems that experienced drivers have to keep a constant update of the chang-
a better awareness than less-experienced ing environment. This experiment used the
drivers of where potential hazardous sit- same software that is in the real fleet and
uations may appear. Therefore, when a simulated nearly identical displays. Situa-
hazardous event is encountered, the expe- tion awareness was measured by requiring
rienced driver is able to react to the the technician operators to reconstruct the
event accordingly without significant inter- locations of the hostile and friendly tar-
ference of other tasks of driving. Conversely, gets after the information on the screen was
although the apprentice driver responds to removed from view, a constrained process-
the hazardous event as fast as the experi- ing task. The participants reconstructed the
enced driver, the apprentices longer fixa- targets one-fourth of the way through the
tion is thought to reflect efforts at trying to scenario and at the end of the scenario.
understand the hazard. Consequently, this Those who scored high on a job-relevant
extra cognitive time will interfere with other test recalled more targets than intermedi-
driving tasks. ates, and intermediates recalled more targets
than low-scoring operators did. A higher per-
centage of hostile targets were recalled than
Situation Awareness (Understanding)
the percentage of friendly targets. Skilled
It is clear that detecting hazards is a criti- operators also outperformed intermediates
cal part of understanding the environment and novices when verbally responding to
in which one is operating. The understand- situation-awareness questions about threats
ing of a dynamic environment has come to in the scenario. Randel et al. concluded that
be studied under the rubric of situation experts repertoires of electronic-warfare sit-
awareness (SA; Durso & Gronlund, 1999; uations, again a schema notion, can assist in
Endsley, 1995 ; Endsley, Chapter 3 6). determining the correct assessment of a new
Stokes, Kemper, and Kite (1997) found situation.
that experienced pilots (> 15 00 hours) out- In a similar vein, Wiegmann, Goh, and
performed apprentice pilots (< 5 0 hours of OHare (2002) found that experienced
instrument flight) in a situation-recognition pilots were better at integrating conflict-
task. The task required participants to lis- ing information than less-experienced pilots
ten to recorded dialogue between air traffic when flying a Frasca 142 flight simulator.
controllers and pilots. Participants listened More-experienced pilots recognized dete-
to the recordings and then selected from riorating weather conditions faster than
an array of diagrams the situation that best did less-experienced pilots. Essentially, the
represented the scenario being discussed. more-experienced pilots were more effi-
The experienced pilots were twice as likely cient than less-experienced pilots in modi-
to generate a plausible conceptualization of fying a VFR model to acquire an Instrument
the dialogue. Experienced pilots were more Meteorological Conditions (IMC) model
adept than less-experienced pilots at creat- when encountering discrepancies. Addition-
ing an assessment of the situation, which ally, Underwood, Crundell, and Chapman
Stokes et al. theorized was by integrating (in press) argue in their literature review
expertise and transportation 365
misdefined (both over- and underdefined) Chapter 3 6). Thus, a mental model can
concepts. Expert structures were not neces- be viewed as a mental outline of expected
sarily more complex than those of appren- events for a dynamic situation. By filling
tices. In fact, the instructor pilots had a quite in particular information from the environ-
simple, elegant Pathfinder network. Thus, ment into the mental model, one mental
expertise seems not be accompanied by a model of a problem type can generate an
more complex knowledge structure, but by endless number of specific situation mod-
a better organized one. els. For example, an air traffic controller has
There are numerous knowledge struc- a mental model of the activity that occurs
tures that researchers have proposed as can- in her low-altitude sectors. She also knows
didates for the better organized one. We the procedures for arrivals and departures,
have already referred above to the schema including procedures for handing off air-
notion. Like schemata, most of the knowl- planes to other controllers and accepting air-
edge structures attributed to experts seem planes from other controllers; and this gen-
to have a large top-down processing com- eral schematic information is represented in
ponent. Even when no difference between the mental model. A situation model about
experienced and less experienced pilots are todays traffic in the west departure sector
found on a declarative knowledge test, expe- for DFW can be spawned from the mental
rienced pilots were more able to apply their model. Departure or arrival procedures for
declarative knowledge when presented in the specific airport that day, procedures for
open-ended scenarios in combination with instantiating temporary flight restrictions,
procedural knowledge (Stokes et al., 1997). and expectation of a flight trajectory of a
Thus, Stokes et al. suggested that appren- particular flight that flies to the same desti-
tice pilots cannot apply knowledge learned nation every day may be represented in the
in ground school to real-time situations situation model along with all of the rele-
as well as could the journeyman pilots, vant information inherited from the parent
despite sufficient knowledge on a multiple- mental model.
choice test. A highly developed mental model assists
We turn now to a particular schematic the operator in discriminating information
knowledge structure that has played a key that is relevant from that which is irrele-
role in thinking about dynamic environ- vant. The operator who has a well-developed
ments, the mental model, and its progeny, mental model should be more efficient at
the situation models. processing information because he or she
knows the information to attend to and the
information to ignore or discard. For exam-
Mental and Situation Models
ple, in Stokes et al. (1997) where pilots listen
In a dynamic environment such as trans- to ATC radio communications, expert pilots
portation, it is thought that the opera- recalled twice the number of concept words,
tor develops mental models that aid in but recalled less filler words than appren-
planning, strategies, and decision making. tices did. Stokes et al. take this as support
Mental models of dynamic environments for the idea that [experts] are better able to
are schematic representations of the pro- make practical use of situational schemata
cesses operating in the world (Gentner & to impose form on sensory data in real time
Stevens, 1983 ). They contain information (p. 191).
about causality and can give rise to judg- Mental models also assist experts in antic-
ments of probability. Durso and Gronlund ipating what will happen next. For exam-
(1999) argue that mental models of real- ple, in an experiment where participants
world dynamic systems are difficult to mod- (with a student pilots certificate or better)
ify once they are developed, but together viewed displays of cockpit instruments and
with the particular situation, they do give text, Doane and Sohn (2004) showed that
rise to situation models (see Endsley, novices were especially poor at predicting
expertise and transportation 367
the result of multiple, meaningfully related tion. The scenario was presented to another
control activities. Finally, superior mental member of the group who had to explain
models should generate superior situation each step in solving the problem. The goals
models, and they do. For example, in Stokes of the controllers were categorized into areas
et al. (1997), while listening to the transcrip- of importance. The category to receive the
tions, the pilots were instructed to build highest priority was a violation of minimum
a mental picture of the situation and then separation standards. The next highest pri-
select from a set of diagrams that best rep- ority was deviations from standard operat-
resented the situation. Expert pilots outper- ing procedures, followed by situations that
formed the apprentice pilots in matching the could lead to increased workload, and exe-
correct diagram with the dialogue. cuting unnecessary requests of pilots.
Seamster et al. (1993 ) found that develop-
mental controllers initially focused on solv-
Strategies
ing potential airspace-violation problems.
Knowledge differences also manifest as dif- Only after these problems were solved did
ferences in strategies. Experts seem to have they begin to concentrate on solving devi-
more strategies overall and more abstract ations. Conversely, experts did not focus on
strategies from which to choose. This dif- solving violation problems to the extent that
ference in turn allows experts to show developmentals did. The experts were much
more flexibility in their decision making more likely to solve violations and devia-
and planning. For example, Wiggins and tions alternately, and developmentals were
OHare (1995 ) found that when making more likely to solve violations sequentially
diversion decisions, less-experienced pilots followed by deviations. Seamster et al. sug-
returned to the starting point and the gested that apprentices attention to the vio-
more-experienced pilots were more likely lations may be more cognitively demanding
to choose safe alternative routes. Strate- for them compared to experts and therefore
gies also come faster for those experts. The reduced their cognitive resources and pre-
participants in that study had a series of vented them from attending to other prob-
menu items from which to choose. Each lems and goals.
menu item contained categorized informa- Seamster et al. (1993 ) also found that
tion about the scenario (e.g., weather infor- experts used fewer strategies than appren-
mation, landing-performance information). tices did when solving problem scenarios.
More-experienced pilots seemed to have The scenarios were rated at 65 % complexity,
a strategy developed before searching the a complexity level that should not overload
menu. For example, they searched less of apprentices. Despite Seamster et al.s finding
the menu items than the less-experienced that experts use fewer strategies, the experts
pilots. In addition to searching more menu used a greater variety of strategies than did
items, the less-experienced pilots were more apprentices. Presumably, a smaller but more
likely to repeat reviews of the menu items. differentiated set of strategies allows experts
The less-experienced pilots seemed to be to solve a wide range of problems.
developing their strategy by searching the Contributing to greater differentiation is
menu, as if the information in the menu the fact that experts employ more workload-
were directing their strategy. management strategies than do apprentices.
The CTA of air traffic controllers con- In fact, within experts, workload-manage-
ducted by Seamster et al. (1993 ) also showed ment strategies were more frequent than
that apprentice, intermediate, and expert planning strategies. Apparently, their expe-
controllers differed substantially on the rience obviated the need to rely on planning
problem-solving techniques and strategies techniques and allowed more management
they employed. Seamster et al. had mem- strategies. Seamster et al. suggest that expe-
bers from each group develop an air traffic rience moderates the need to create planning
problem and generate the most optimal solu- strategies (see also Zsambok & Klein, 1997).
368 the cambridge handbook of expertise and expert performance
the highway. Most typical experiences are Cellier, J. M., Eyrolle, H., & Marine, C.
already likely to be easily and automatically (1997). Expertise in dynamic environments.
handled by the unimpaired experienced Ergonomics, 40, 285 0.
driver. Explicit training on pattern detection Chapman, P. R., & Underwood, G. (1998). Visual
and interpretation of hazards, the develop- search of driving situations: Danger and expe-
ment of strategies, and the management of rience. Perception, 2 7 , 95 1964.
resources are likely targets for such deliber- Chase, W. G., & Simon, H. A. (1973 ). Perception
ate practice (e.g., Ericsson, 2003 ; Ericsson, in chess. Cognitive Psychology, 4, 5 5 81.
Krampe, & Tesch-Romer, 1993 ). Simulators, Chi, M. T. H., Glaser, R., & Farr, M. J. (1988).
though once prohibitive, are well within The nature of expertise. Hillsdale, NJ: Erlbaum.
the technological and financial constraints Crundall, D., & Underwood, G. (1998). The
of the personal computer. Given that there effects of experience and processing demands
is little reason to believe that the culprit on visual information acquisition in drivers.
Ergonomics, 41, 44845 8.
for inferior driving is vehicle control, the
PC is a fine simulator for training hazard Dattel, A. R., & Sohn, Y. W. (2003 ). Novice-
expert differences in flight situation awareness
detection, strategy selection, and resource
and action planning as a function of complex-
management. Thus, we believe that a cog- ity. Proceedings of the 15 th Triennial International
nitive perspective combined with appropri- Ergonomics Association Meeting. Seoul, Korea:
ate practice specific and deliberate can The Ergonomics Society of Korea.
improve driving skills and make the drive for DeKeyser, V., & Piette, A. (1970). Analyse de
our LA commuter a little safer. lactivite des operateurs au tableau synop-
tique dune chaine dagglomeration. Le Travail
Humain, 3 3 , 3 413 5 2.
Doane, S. M., & Sohn, Y. W. (2004). Pilot ability
Footnote to anticipate the consequences of flight actions
as a function of expertise. Human Factors, 46,
1. We are grateful to Kate Bleckley, Ray King, 92103 .
Pat DeLucia, Anders Ericsson, and especially Doane, S. M., & Sohn, Y. W. (2000). ADAPT: A
Robert Hoffman for comments on earlier ver- predictive cognitive model of user visual atten-
sions of this chapter. tion and action planning. User Modeling and
User-Adapted Interaction, 10, 145 .
Duncan, J., Williams, P., & Brown, I. (1991). Com-
ponents of driving skill: Experience does not
References mean expertise. Ergonomics, 3 4, 91993 7.
Duncan, J., Williams, P., Nimmo-Smith, I., &
Brown, I. D. (1992). The control of skilled
Allgaier, E. (193 9). Experience counts: It takes
behavior: Learning, intelligence and distrac-
more than physical ability to make a good
tion. In D. E. Meyer & S. Kornblum (Eds.),
driver. Public Safety, 15 , 21.
Attention and performance XIV (pp. 3 23 3 41).
Beilock, S. L., Carr, T., MacMahon, C., & Starkes, Cambridge, MA: MIT Press.
J. L. (2002). When paying attention becomes Durso, F. T., & Gronlund, S. D. (1999). Situa-
counterproductive: Impact of divided versus tion Awareness. In F. T., Durso, R. Nickerson,
skill-focused attention on novice and experi- R. Schvaneveldt, S. Dumais, S. Lindsay, & M.
enced performance of sensorimotor skills. Jour- Chi (Eds.), The handbook of applied cognition,
nal of Experimental Psychology: Applied, 8, 616. (pp. 283 3 14). New York: John Wiley & Sons.
Bellenkes, A. H., Wickens, C. D., & Kramer, A. F. Elander, J., West, R., & French, D. (1993 ).
(1997). Visual scanning and pilot expertise: Behavioral correlates of individual differences
The role of attentional flexibility and mental in road-traffic crash risk: An examination of
model development. Aviation, Space, and Envi- methods and findings. Psychological Bulletin,
ronmental Medicine, 68, 5 695 79. 113 , 279294.
Bisseret, A. (1971). Analysis of mental processes Endsley, M. R. (1995 ). Measurement of situation
involved in air traffic control. Ergonomics, 14, awareness in dynamic systems. Human Factors,
5 65 5 70. 3 7 , 65 84.
370 the cambridge handbook of expertise and expert performance
Ericsson, K. A. (2003 ). The search for general on the road. In S. Banbury & S. Tremblay (Eds.),
abilities and basic capacities: Theoretical impli- A cognitive approach to situation awareness: The-
cations from the modifiability and complexity ory, measurement, and application. (pp. 15 5
of mechanisms mediating expert performance. 175 ) Hampshire, UK: Ashgate.
In R. J. Sternberg (Ed.), Psychology of abilities, Klein, G. (1997). An overview of naturalistic deci-
competencies, and expertice (pp. 93 125 ). New sion making applications. In C. E. Zsambok
York: Cambridge University Press. & G. Klein (Eds), Naturalistic decision making
Ericsson, K. A., Krampe, R. T., & Tesch-Romer, (pp. 4960). Mahwah, NJ: Erlbaum.
C. (1993 ). The role of deliberate practice in the McKenna, F. P., & Crick, J. L. (1991). Hazard
acquisition of expert performance. Psychologi- perception in drivers: A methodology for testing
cal Review, 100, 3 63 406. and training. Final Report. Behavioural Studies
Ericsson, K. A., & Smith, J. (1991). Toward a Unit. Transport and Road Research Laboratory,
general theory of expertise. Cambridge, MA: Crowthorne, UK.
Cambridge University Press. McKenna, F. P., & Farrand, P. (1999). The role of
Fisher, J. A., & Galer, I. A. R. (1984). The effects automaticity in driving. In G. B. Grayson (Ed.),
of decreasing the radius of curvature of convex Behavioural research in road safety IX. Transport
external rear view mirrors upon drivers judg- and Road Research Laboratory, Crowthorne,
ments of vehicles approaching in the rearward UK.
visual field. Ergonomics, 2 7 , 12091224. McKinney, E. H., & Davis, K. J. (2003 ). Effects
Forsyth, E., Maycock, G., & Sexton, B. (1995 ). of deliberate practice on crisis decision perfor-
Cohort study of learner and novice drivers: Part 3 , mance. Human Factors, 45 , 43 6444.
Accidents, offences and driving experience in the Mourant, R. R., & Rockwell, T. H. (1972). Strate-
first three years of driving. Transport and Road gies of visual search by novice and experienced
Research Laboratory, TRRL Report LR 111, drivers. Human Factors, 14, 3 25 3 3 5 .
Crowthorne, UK. NTSB (2004). Retrieved April 3 0, 2004 from the
Gentner, D., & Stevens, A. L. (1983 ). Mental mod- National Transportation Safety Board web cite:
els. Hillsdale, NJ: Erlbaum. www.ntsb.gov.
Glaser, R. (1987). Thoughts on expertise. In C. Randel, J. M., Pugh, H. L., & Reed, S. K. (1996).
Schooler & W. Schaie (Eds.), Cognitive func- Differences in expert and novice situation
tioning and social structure over the life course awareness in naturalistic decision making. Inter-
(pp. 8194). Norwood, NJ: Ablex. national Journal of Human-Computer Studies,
Groeger, J. A. (2000). Understanding driving: 45 , 5 795 97.
Applying cognitive psychology to a complex every- Recarte, M. A., & Nunes, L. M. (1996). Percep-
day task. Hove, UK: Psychology Press. tion of speed in automobile: Estimation and
Groeger, J. A., & Grande, G. E. (1996). Self- production. Journal of Experimental Psychology:
preserving assessments of skills? British Journal Applied, 2 , 2913 04.
of Psychology, 87 , 6179. Schvaneveldt, R. W., Beringer, D. B., & Lamon-
Haber, R. N., & Haber, L. (2003 ). Perception and ica, J. A. (2001). Priority and organization of
attention during low-altitude high-speed flight. information accessed by pilots in various phases
In P. S. Tsang & M. A. Vidulich (Eds.), Principles of flight. The International Journal of Aviation
and practice of aviation psychology (pp. 2168). Psychology, 11, 25 3 280.
Mahwah, NJ: Erlbaum. Schvaneveldt, R. W., Durso, F. T., Goldsmith,
Henmon, V. A. C. (1919). Air service tests of apti- T. E., Breen, T. J., Cooke, N. M., Tucker, R. G.,
tude for flying. Journal of Applied Psychology, 3 , & DeMaio, J. C. (1985 ). Measuring the struc-
103 109. ture of expertise. The International Journal of
Hoffman, R. (1996). How can expertise be Man-Machine Studes, 2 3 , 699728.
defined? Implications of research from cogni- Seamster, T. L., Redding, R. E., Cannon, J. R.,
tive psychology. In R. Williams, W. Faulkner, & Ryder, J. M., & Purcell, J. A. (1993 ). Cognitive
J. Fleck (Eds.), Exploring expertise (pp. 81100). task analysis of expertise in air traffic control.
Edinburgh, Scotland: University of Edinburgh International Journal of Aviation Psychology, 3 ,
Press. 25 7283 .
Horswill, M. S., & McKenna, F. P. (2004). Drivers Shanteau, J. (1992). Competence in experts:
hazard perception ability: Situation awareness The role of task characteristics. Organizational
expertise and transportation 371
Keywords: software design, programming, tion of the domain and give an overview
computer program, debugging, commu- of tasks in software development. Next, we
nication, knowledge representation, Pro- briefly describe the expertise concept and
fessionals. distinguish between a conceptualization of
expertise as years of experience and exper-
tise as high performance. The third main
Introduction section is the core part of this chapter. In
this section, we review empirical research
In this chapter, we review research evi- on expertise in tasks such as software
dence on expertise in software design, design, programming, program comprehen-
computer programming, and related tasks. sion, testing, and debugging. Moreover, we
Research in this domain is particularly inter- describe how expert performers differ from
esting because it refers both to rather non-experts with respect to knowledge as
general features and processes associated well as communication and cooperation pro-
with expertise (e.g., knowledge represen- cesses. In the final section, we present direc-
tation, problem-solving strategies) and to tions for future research and discuss some
specific characteristics of high performers practical implications.
in an economically relevant real-world set-
ting. Therefore, in this chapter we draw on
literature from various fields, mainly from Historical Context
cognitive psychology, but also from work
and organizational psychology and from the Extensive research on expertise on software
software-design literature within computer design and programming started in the early
science. 1980s. For example, Jeffries, Turner, Polson,
Our chapter is organized as follows: In the and Atwood (1981) as well as Adelson (1984)
first main section we provide a brief descrip- published influential studies on how experts
373
374 the cambridge handbook of expertise and expert performance
differ from novices. These studies stimulated the functions specified during software
subsequent research and were often cited design are translated into a computer pro-
also in more general publications on exper- gram, that is, a set of grammatical rules that
tise (e.g., Ericsson & Smith, 1991; Sternberg, refer to calculations, procedures, or objects.
1996). Generally, research activity in the A main challenge for software design is
domain of software design and programming to ensure that a computer program meets
has been very lively and intensive (Hoffman, the requirements and runs without errors.
1992). For many years, research on exper- This asks for extensive testing and debugging
tise in this domain was an important topic (i.e., error correction).
in the workshops on Empirical Studies of Traditionally, software design and pro-
Programmers (Olson, Sheppard, & Soloway, gramming has been conceptualized as a
1987; Soloway & Iyengar, 1986). Since the rather linear process, starting from require-
1990s, research interest seems to have shifted ment analysis and moving then to testing and
to investigations in more complex organi- debugging, with design and programming as
zational settings (e.g., Campbell, Brown, & distinct intermediate phases. However, such
DiBello, 1992; Sonnentag, 1995 ; Turley & a linear approach was not feasible in prac-
Bieman, 1995 ). tice because often not all requirements can
Studies on expertise in software design be specified in advance. Therefore, nowadays
and programming do not aim only at sci- a more iterative view dominates the field
entific insight into the processes associated (Sommerville, 2001).
with expertise in a complex domain. Nowa- From a more psychological perspective,
days, as many countries economy and peo- many software-design tasks can be described
ples life largely depend on information as ill-defined problems (Simon, 1973 ). Ill-
technology, the development and mainte- defined problems imply that problem spec-
nance of high-quality software systems are ifications are incomplete and have to be
of crucial relevance. Therefore, also with decided on during the design process. Thus,
respect to practical implications it is impor- ill-defined problems have no single correct
tant to examine how expert performance solution. Software design and programming
in software design and programming is draws largely on syntactic, semantic, and
achieved. schematic knowledge (Detienne, 2002). A
broad range of different design and program-
ming strategies have been distinguished,
The Field of Software Design such as top-down versus bottom-up, forward
and Programming versus backward development, or breadth-
first versus depth-first (Detienne, 2002).
The domain of software design and pro- Therefore, psychological research on soft-
gramming comprises distinct tasks, such as ware design and programming has focused
requirement analysis, software design, pro- mainly on designers and programmers cog-
gramming, testing, and debugging. It is the nitive processes.
main goal of requirement analysis to spec- However, it has been argued that pro-
ify the demands a new computer program fessional software design and programming
or software system should meet. Ideally, (programming in the large) takes place in
future users of the computer system or teams and other cooperative settings and,
their representatives should be involved therefore, an exclusive focus on the cogni-
in requirement analysis because they often tive processes of individuals can fall short
have detailed knowledge about the tasks and when describing software design and pro-
processes the computer system should sup- gramming (Curtis, 1986; Riedl, Weitzenfeld,
port. Software design aims at the description Freeman, Klein, & Musa, 1991). To under-
of the basic features of the future computer stand fully how high-quality software is
system and prescribes the functions the sys- produced research should also address com-
tem should perform. During programming, munication and cooperation processes.
expertise in software design 375
Table 2 1.1. Comparison between experts and non-experts in software design and programming
Experienced persons as High performers as
compared to opposed to
inexperienced persons moderate performers
Requirement analysis and Spend more time and effort on Spend less time on problem
design problem comprehension comprehension
Spend more time on clarifying Adequate problem
program requirements representation early in the
design process
Decompose design problems in
small components in a
top-down breadth-first manner
Program comprehension and Pursuit of abstract Pursuit of abstract
programming programming goals programming goals
Program comprehension based Cross-referencing strategy
on abstract concepts
Testing and debugging Test for inconsistency Active search for problems
Hypothesis-driven procedure
Knowledge Knowledge organized in Broader and more detailed
greater and more meaningful knowledge base
chunks
Knowledge of abstract
concepts
Superior meta-cognitive
knowledge
Communication and <not studied> Spend more time on
cooperation communication and
cooperation
a solution (Koubek, Salvendy, Dunsmore, & 2003 ; Malhotra et al., 1980). Similar to find-
LeBold, 1989; Malhotra, Thomas, Carroll, ings from early studies in physics problem
& Miller, 1980). Early studies showed that solving (e.g., Chi, Feltovich, & Glaser, 1981),
there are considerable individual differences studies have indicated that experienced soft-
in software design and that activities tend ware designers spend more time on clari-
to vary according to the design experi- fying the program requirements compared
ences of programmers (e.g. Jeffries et al., to students (Batra & Davis, 1992; Jeffries
1981; Weinberg & Schulman, 1974). Studies et al., 1981). Moreover, within the web-
focused on the comparison of inexperienced design domain, professionals included more
programmers (e.g., students) and expe- requirements in their web design than less-
rienced programmers (e.g., professionals; experienced designers (Chevalier & Ivory,
Agarwal, Sinha, & Tanniru, 1996), rather 2003 ). Even when specific clients require-
than on the comparison of specific perfor- ments were not explicitly stated, profession-
mance levels (e.g., Sonnentag, 1998). als inferred more requirements of the client
In real-world design situations, an impor- on the basis of their prior experience. In a
tant part of the design process consists of study with performance level as expertise
analysing the requirements the new software criterion, Sonnentag (1998) found no differ-
system should comply with. These require- ences between high and moderate perform-
ments are often ill defined and not clearly ers in analysing requirements in early phases
stated by clients or potential users. Thus, the of the design process. Moderate perform-
first task of the software designer is to set and ers spent even more time analyzing require-
refine the design goals (Chevalier & Ivory, ments in later phases of the design process
expertise in software design 377
compared to high performers. High per- indicated that design strategies were influ-
formers might have developed an adequate enced toward the specific language experi-
problem representation early in the design ences (Agarwal et al., 1996). Though experts
process, whereas moderate performers had tried to obtain a deep understanding of a
to do additional requirement analysis on the new language, they still used analogies to
later stages of the design process. languages they knew when generating a pro-
A lot of research has focused on the gram within a new framework (Campbell
ability to decompose the design problem et al., 1992; Scholz & Wiedenbeck, 1992).
into smaller components. Here, a consis- Furthermore, research showed that when
tent result across many studies, albeit with participants were experienced in one lan-
very small samples, is that experts decom- guage (in object-oriented programming),
posed the design problem in a top-down and complex plans were developed on the basis
breadth-first manner (Adelson & Soloway, of the deep structure of this programming
1985 ). For example, Jeffries et al. (1981) language. Inexperienced designers in object-
described that experts decomposed a prob- oriented programming developed plans on
lem in major parts through a number of iter- the basis of functional similarity and showed
ations. Also, advanced novices relied on such more plan revisions (Detienne, 1995 ).
a top-down and breadth-first solution strat- It thus seems that the knowledge base of
egy, but with fewer iterations and therefore experts is highly language dependent. How-
less detail. ever, experts also have abstract, transferable
A common assumption is that the decom- knowledge and skills. Agarwal et al. (1996)
position process is guided by a knowl- compared 22 computer scientists (more than
edge representation, which is built up with two years of experience in process-oriented
experience and stored in long-term mem- approaches) and 24 business students (lim-
ory (Jeffries et al., 1981). When such a ited knowledge in process-oriented mod-
knowledge representation is not available, eling). When solving a design problem
experts develop plans through a bottom- within a new non-process-oriented program-
up, backward strategy (Rist, 1991). Accord- ming approach (here, object-oriented pro-
ing to Rist (1986), experts have stored a gramming), experts did not differ from
broad range of programming plans that com- novices in the quality of solutions (struc-
prise knowledge of abstract and specific, ture and behavior of the program), and they
language-dependent code fragments. This showed even poorer performance on a finer
idea is supported by the study of Sonnentag level (processing aspects). Also Collani and
(1998). When software designers were asked Schomann (1995 ) examined the acquisition
about strategies they would recommend to of a new programming language. In con-
an inexperienced programmer working on trast to the study of Agarwal et al. (1996),
the same design problem, high performers the criterion for expertise was the pro-
reported twice as many strategies as moder- grammers performance. Their results sup-
ate performers. The studies of Davies (1991) ported the hypothesis that high performers
showed that experienced programmers did possess language-independent abstract pro-
not necessarily have more plans available, gramming skills and knowledge that can be
but focused on the most salient parts of the transferred faster and with less errors to the
plan and flexibly switched between plans. In new programming framework.
contrast, novices preferred a strategy with The contrary results of the studies that
which plans are implemented in a more lin- defined expertise in terms of the length
ear fashion. of experience versus performance can be
With the emergence of new programming explained by the different kinds of expertise:
approaches (e.g., object-oriented program- routine and adaptive expertise (Hatano
ming), studies examined the role of knowl- & Inagaki, 1986). Individuals character-
edge in the design activity of experts using a ized by routine expertise show superior
new programming approach. These studies performance in well-known, often highly
378 the cambridge handbook of expertise and expert performance
routinized tasks, whereas individuals char- Most studies on programming have com-
acterized by adaptive expertise are able to pared relatively inexperienced students with
master novel tasks and are successful in more-experienced programmers (Bateson,
transferring existing knowledge and skills Alexander, & Murphy, 1987; Davies, 1990;
to unknown tasks or settings. For exam- Soloway & Ehrlich, 1984). These studies
ple, participants in the study of Agarwal clearly showed that more advanced pro-
et al. (1996) could have developed routine grammers outperformed the relatively inex-
expertise through extended practice with perienced students with respect to perfor-
process-oriented design, which has rigidify- mance quality (Bateson et al., 1987; Soloway
ing effects on learning new skills and knowl- & Ehrlich, 1984) and solution time (Davies,
edge. On the other hand, the upper perfor- 1990). When comparing high performers
mance group in the study of Collani and with moderate performers within a group
Schomann (1995 ) might be characterized of professional programmers, the high per-
by the concept of adaptive expertise. Such formers followed more abstract goals but did
experts have a deeper conceptual under- not differ with respect to other process fea-
standing of the domain that allows for a tures (Koubek & Salvendy, 1991).
transfer of knowledge and skills to novel Program comprehension is a necessary
tasks (Hatano & Inagaki, 1986; Kimball & prerequisite for successfully completing pro-
Holyoak, 2000). This might also lead to gramming tasks. Moreover, also testing and
a faster in-depth requirement analysis that debugging tasks cannot be satisfactorily
the high performers in the above mentioned accomplished without understanding the
study of Sonnentag (1998) have shown. program. In addition, program comprehen-
Most studies have focused on one aspect sion is of crucial importance in the context of
of the design activity. One exception is the professional software maintenance and soft-
study of Sonnentag (1998) that analysed ware reuse. Nearly all empirical studies that
a broad range of design activities: require- have examined program comprehension
ment analysis, feedback processing, plan- have focused on the comparison between
ning, task focus, visualizations, and knowl- inexperienced and more-experienced per-
edge. According to the results, expertise in sons (Adelson, 1984; Guerin & Matthews,
software design was characterized by more 1990; Rist, 1996; Widowski & Eyferth, 1985 ;
local planning activities, more feedback pro- Wiedenbeck, Fix, & Scholtz, 1993 ). These
cessing, less task-irrelevant cognitions, more studies showed that experienced persons
solution visualizations, and more knowl- perform better on program comprehension
edge about design strategies experts would tasks and appear to be superior with respect
recommend to an imagined inexperienced to specific aspects of comprehension. For
colleague. example, Adelson (1984) found that experi-
After the requirements have been ana- enced programmers provided better answers
lyzed and the software design has been pro- to abstract questions than did students. Sim-
duced, the concepts have to be implemented ilarly, Rist (1996) reported that experienced
in computer programs. Therefore, in the graduates categorize programs according to
next paragraphs we review research findings programming plans. Thus, program com-
on differences between experts and non- prehension of more-experienced persons is
experts with respect to programming and based on more-abstract concepts.
program comprehension. Pennington (1987) compared highly and
poorly performing professional program-
mers. When trying to understand a program,
Programming and Program
high performers showed a cross-referencing
Comprehension
strategy characterized by systematic alter-
Programming refers to the process of trans- ations between systematically studying the
lating specifications of a calculation or computer program, translating it to domain
a procedure into a computer program. terms, and subsequently verifying domain
expertise in software design 379
terms back in program terms. In contrast, have a better representation of the program
poorer performers exclusively focused on and potential problems, which helps them
program terms or on domain terms with- to actively search for problems and to eval-
out building connections between the two uate information. In the next paragraphs we
worlds. Thus, it seems that connecting var- discuss experts knowledge and knowledge
ious domains and relating them to each other representations in greater detail.
is crucial for arriving at a comprehensive
understanding of the program and the under-
Knowledge and Knowledge
lying problem.
Representation
Designing and programming software is
only half of the story of software develop- Knowledge and adequate knowledge rep-
ment. Finding and correcting errors in the resentation are crucial in all phases of
software before introducing the software to software development. Researchers have
the market is crucial. In the next paragraphs investigated knowledge and knowledge rep-
we will therefore summarize research evi- resentation by using recall, problem-sorting,
dence in the area of testing and debugging. and other tasks. For example, study par-
ticipants have been asked to recall pro-
gram lines, either presented in a mean-
Testing and Debugging
ingful or in an arbitrary order (Barfield,
Software testing aims at identifying incon- 1986; McKeithen, Reitman, Rueter, & Hirtle,
sistencies and errors within computer pro- 1981). Again, most studies have contrasted
grams. Debugging refers to the process of inexperienced with more-experienced per-
removing errors, so-called bugs, from a com- sons (Etelapelto, 1993 ; McKeithen et al.,
puter program. Empirical studies on test- 1981; Weiser & Shertz, 1983 ; Ye & Salvendy,
ing and debugging have focused on com- 1994).
parisons between relatively inexperienced There is consistent evidence that novices
and more-experienced persons (Nanja & and more-advanced students or profes-
Cook, 1987; Teasley, Leventhal, Mynatt, & sionals differ with respect to knowledge
Rohlman, 1994; Weiser & Shertz, 1983 ). and knowledge representation. In recall
Not surprisingly, these studies illustrated tasks, more-experienced persons recalled
that experienced students and professionals more program lines, often only when
found more errors and detected the errors the lines were presented in a meaning-
more quickly than did novices (Law, 1998; ful order (Barfield, 1986, 1997; McKeithen
Weiser & Shertz, 1983 ). With respect to the et al., 1981), but sometimes also when pre-
testing processes, studies showed that expe- sented in an arbitrary order (Bateson et al.,
rienced persons tested more for inconsis- 1987; Guerin & Matthews, 1990). Thus, it
tency (Teasley et al., 1994) and proceeded seems that experienced persons have orga-
in a more hypothesis-driven and concept- nized their knowledge in greater and more
oriented way (Krems, 1995 ). Thus, as stu- meaningful chunks (Adelson, 1984; Ye &
dents programming experience increased, Salvendy, 1994). However, differences in
they seemed to work in a more system- chunking do not seem to be sufficient to
atic way. explain experienced persons better perfor-
Few studies have compared highly mance when recalling program lines pre-
performing software professionals with sented in an arbitrary order.
those who perform at a moderate level. Research has shown that professionals
A thinking-aloud study showed that high also possess more meta-cognitive knowl-
performers needed less time for debug- edge than do students (Etelapelto, 1993 ).
ging, made fewer errors, searched more Etelapelto assessed students and profes-
intensively for problems, and showed more sional programmers meta-cognitive know-
information-evaluation activity (Vessey, ledge during an interview that focused on
1986). Thus, it seems that high performers a specific computer program and on the
380 the cambridge handbook of expertise and expert performance
respondents more general working strate- in project teams, suggesting that expertise
gies. Professional programmers meta- in software design is not only a matter
cognitive knowledge was much more of knowledge and task strategies but also
detailed and more comprehensive than requires social and communicative skills.
students knowledge. For example, a pro- Field studies suggest that high per-
fessional programmer provided a comment formers show better communication and
such as The main program is not very well cooperation competencies than moderate
divided; the logic has been scattered over performers (e.g. Curtis, Krasner, & Iscoe,
different parts of the program; the updating 1988; Kelley & Caplan, 1993 ; Sonnentag,
paragraph has too much in it. The control 1995 ). In a field study with 17 software-
function of the program should be imple- development projects, Curtis et al. (1988)
mented in one paragraph; now the program found that exceptional software designers
has much depth in itself and scanning is showed superior communication skills. In
difficult (p. 248f.), whereas the students fact, much of the exceptional designers
comments were more general and diffuse. design work was accomplished while inter-
In addition, professional programmers acting with other team members. High
expressed a clear idea about a good strategy performers spent a lot of time educating
of reading and comprehending computer other team members about the applica-
programs. Thus, experienced persons know tion domain and its mapping into com-
more about how to proceed when solving putational structures. Similarly, Sonnentag
tasks in their domain. (1995 ), using a peer-nomination method in
When it comes to knowledge repre- a study with 200 software professionals,
sentation in highly versus moderately per- showed that experts did not spend more
forming software professionals, high per- time on typical software-development activ-
formers superior knowledge seems not ities such as design, coding, or testing, but
restricted to knowledge about how to solve were more often engaged in review meetings
rather narrowly defined software-design and spent more time in consultations than
tasks (Sonnentag, 1998). Highly perform- did other team members. In a study at the
ing software professionals also possess more Bell Laboratories, Kelley and Caplan (1993 )
detailed knowledge about how to approach compared top performers to average per-
cooperation situations (Sonnentag & Lange, formers and found that differences could not
2002). Thus, it seems that as work tasks be attributed to cognitive ability and that top
become more comprehensive in real-world performers considered interpersonal net-
settings, high performers develop a more work abilities as highly important. Further-
comprehensive representation of the entire more, top performers possessed better func-
work task, including necessary features of tioning interpersonal networks compared to
cooperation with coworkers. average performers. Riedl et al. (1991) also
In the next paragraphs we describe the observed highly developed interpersonal
role of communication and cooperation in skills in expert software engineers. In a study
expert performance in more detail. on behavior in team meetings, Sonnentag
(2001a) found that highly performing
software professionals involvement in coop-
Communication and Cooperation
eration situations differed, depending on
Most research on experts has focused on spe- situational demands. In highly structured
cific activities in software design, such as meetings, no differences between highly
design, programming, or testing in individ- and average performing software profes-
ual task settings. Nevertheless, some studies sionals were found, whereas differences
have addressed the question whether there were observed in poorly structured meet-
are differences between high- and average- ings. Thus, experts showed high adapta-
performing software professionals in coop- tion to specific situational constraints and
erative work settings. Modern professional displayed cooperation competencies, partic-
software development takes place mainly ularly in difficult situations. Interestingly,
expertise in software design 381
skills. Until now, motivational and self- or sports (Davids, 2000; Ericsson et al., 1993 ;
regulatory issues that might be highly rel- Krampe & Ericsson, 1996) and also in more
evant for high performance were beyond classical work settings such as insurance
the research interest of most scholars in the companies (Sonnentag & Kleine, 2000). We
field (But see Ericsson, Chapter 3 8). Empir- assume that deliberate practice might play
ical studies from other areas have shown also a core rule in the development of expert
that self-efficacy and the pursuit of diffi- performance in software design and pro-
cult and specific goals are closely related to gramming.
high performance levels (Latham, Locke, & An issue closely related to the devel-
Fassina, 2002; Locke & Latham, 1990; Sta- opment of expertise refers to the distinc-
jkovic & Luthans, 1998). Moreover, it has tion between routine and adaptive exper-
been found that high performers verbalize tise (Hatano & Inagaki, 1986). In areas such
less task-irrelevant cognitions during soft- as the software domain, where new tools
ware design and focus more on the task at and methodologies continuously emerge and
hand (Sonnentag, 1998). One interpretation where existing knowledge can become obso-
is that high performers have superior self- lete very quickly, more research on the
regulatory skills that inhibit task-irrelevant development of adaptive expertise is partic-
thoughts. It would be a highly interesting ularly needed.
avenue for future research to examine how To adequately investigate the develop-
motivation and self-regulation affect expert ment of expertise in professional domains,
performance. longitudinal field studies are needed.
Although retrospective studies can offer
some interesting insights, only longitudinal
(5) the question of how expertise develops studies that cover several years will allow
The question of how expertise develops is conclusive answers about how expertise
both highly interesting and practically rele- develops and unfolds over time.
vant. However, our knowledge about how
expertise in the software domain devel-
ops is very limited. Of course, as students Practical Implications
progress from a novice level to a graduate
student or even professional level, perfor- When it comes to practical approaches to
mance normally increases. However, within promote expert performance in work set-
the group of software professionals, results tings, two major approaches have to be
on the empirical relationship between years considered: (1) personnel selection and (2)
of experience and performance are inconclu- training. Both approaches are based on the
sive. Although some studies have found pos- assumption that the characteristics typical
itive relationships between years of experi- for experts causally contribute to their supe-
ence and performance (Koubek & Salvendy, rior performance. Most generally, person-
1991; Turley & Bieman, 1995 ), others have nel selection refers to organizational pro-
not (Sonnentag, 1995 , 1998; Vessey, 1986). cedures and specific methods in order to
There is some evidence that specific aspects select those individuals who may be well
of experience, such as its breadth and suited for a specific job. Based on the liter-
variety, are related to expert performance ature review provided in this chapter, per-
(Sonnentag, 1995 ). Ericsson et al. (1993 ) sonnel selection procedures at least after
have argued that deliberate practice (i.e., some domain-specific training is accom-
regularly pursued purposeful and effortful plished should include measures of knowl-
learning and practice activities) is crucial for edge and knowledge organization and of
the achievement and maintenance of expert domain-specific problem-solving strategies
performance. Empirical studies have shown that focus on abstract concepts and goals. In
that deliberate practice indeed is related to addition, assessments of communication and
high performance in domains such as music cooperation skills should be part of the
384 the cambridge handbook of expertise and expert performance
personnel-selection process. In addition Adelson, B., & Soloway, E. (1985 ). The role of
to more standardized tests (Stevens & domain experience in software design. IEEE
Campion, 1999), group discussions or other Transactions on software engineering, 11, 13 5 1
teamwork assignments within an assessment 13 60.
center might be used. With respect to expe- Agarwal, R., Sinha, A. P., & Tanniru, M. (1996).
rience as a predictor in personnel selection, The role of prior experience and task character-
the situation is more complicated. Of course, istics in object-oriented modeling: An empir-
ical study. International Journal of Human-
for most jobs one would prefer to select indi-
Computer Studies, 45 , 63 9667.
viduals with professional experience over
Barfield, W. (1986). Expert-novice differences for
beginning students. However, when having
software: Implications for problem-solving and
to select individuals from a group of pro-
knowledge acquisition. Behaviour & Informa-
fessionals which would be more often the tion Technology, 5 , 15 29.
case length of experience does not seem
Barfield, W. (1997). Skilled performance on soft-
to be a very reliable predictor of expert per- ware as a function of domain expertise and pro-
formance. Other aspects of experience, such gram organization. Perceptual and Motor Skills,
as experience variety, might be more helpful 85 , 14711480.
(Sonnentag, 1995 ). Bateson, A. G., Alexander, R. A., & Murphy,
In addition to personnel selection, train- M. D. (1987). Cognitive processing differences
ing is a promising approach to the man- between novice and expert computer program-
agement of expert performance (Hesketh & mers. International Journal of Man-Machine
Ivancic, 2002). Training should help indi- Studies, 2 6, 649660.
viduals to develop adequate mental mod- Batra, D., & Davis, J. G. (1992). Conceptual data
els of typical problems in the domain and modelling in database design: Similarities and
to choose the most appropriate working differences between expert and novice design-
strategy. Our literature review suggests that ers. International Journal of Man-Machine Stud-
training should focus on domain-specific ies, 3 7 , 83 101.
and meta-cognitive knowledge as well as Campbell, D. J., & Gingrich, K. F. (1986). The
on abstract planning and evaluation strate- interactive effects of task complexitiy and par-
gies. Effective communication and cooper- ticipation on task performance: A field exper-
iment. Organizational Behavior and Human
ation skills should be taught and deepened
Decision Processes, 3 8, 162180.
through practical exercises. In our view, it
Campbell, R. L., Brown, N. R., & DiBello, L. A.
is important that training in communica-
(1992). The programmers burden: Developing
tion and cooperation begins at the univer-
expertise in programming. In R. R. Hoffman
sity because at present training in computer (Ed.), The psychology of expertise (pp. 2693 62).
science and programming often relies exclu- New York: Springer.
sively on technical knowledge and skills Chevalier, A., & Ivory, M. Y. (2003 ). Web
and neglects communication and coopera- site designs: Influences of designers expertise
tion skills that seem to be particularly impor- and design constraints. International Journal of
tant for expert performance in professional Human-Computer Studies, 5 8, 5 787.
settings. Chi, M. T. H., Feltovich, P. J., & Glaser, R. (1981).
Categorization and representation of physics
problems by experts and novices. Cognitive Sci-
ence, 5 , 12115 2.
References Collani, G. V., & Schomann, M. (1995 ). The pro-
cess of acquisition of a new programming lan-
Adelson, B. (1984). When novices surpass guage (LISP): Evidence for transfer of expe-
experts: The difficulty of a task may increase rience and knowledge in programming. In
with expertise. Journal of Experimental K. F. Wender, F. Schmalhofer & H.-D. Boecker
Psychology: Learning, Memory, and Cognition, (Eds.), Cognition and computer programming
10, 483 495 . (pp. 169191). Norwood, NJ: Ablex.
expertise in software design 385
Curtis, B. (1986). By the way, did anyone study Ferris, G. R., Witt, L. A., & Hochwarter, W. A.
any real programmers? In E. Soloway & S. (2001). Interaction of social skill and general
Iyengar (Eds.), Empirical studies of programmers mental ability on job performance and salary.
(pp. 25 6262). Norwood, JN: Ablex. Journal of Applied Psychology, 86, 1075 1082.
Curtis, B., Krasner, H., & Iscoe, N. (1988). A field Fujigaki, Y., & Mori, K. (1997). Longitudinal
study of the software design process for large study of work stress among information system
systems. Communications of the ACM, 3 1, 1268 professionals. International Journal of Human-
1287. Computer Interaction, 9, 3 693 81.
Davids, K. (2000). Skill acquisitaion and the the- Glass, R. L. (1997). The ups and downs of pro-
ory of deliberate practice: It aint what you do grammer stress. Communications of the ACM,
its the way you do it International Journal of 40(4), 1719.
Sport Psychology, 3 1, 461466. Guerin, B., & Matthews, A. (1990). The effects
Davies, S. P. (1990). Plans, goals and selection of semantive complexity on expert and novice
rules in the comprehension of computer pro- computer program recall and comprehension.
grams. Behaviour & Information Technology, 9, The Journal of General Psychology, 117 , 3 79
201214. 3 89.
Davies, S. P. (1991). The role of notation and Hatano, G., & Inagaki, K. (1986). Two courses
knowledge representation in the determina- of expertise. In H. Stevenson, A. Azuma, &
tion of programming strategy: A framework for K. Hakuta (Eds.), Child development and edu-
integrating models of programming behavior. cation in Japan (pp. 262272). San Francisco:
Cognitive Science, 15 , 5 475 72. Freeman.
Detienne, F. (1995 ). Design strategies and knowl- Hesketh, B., & Ivancic, K. (2002). Enhancing per-
edge in object-oriented programming: Effects formance through training. In S. Sonnentag
of experience. Human-Computer Interaction, (Ed.), Psychological management of individual
10, 129169. performance (pp. 247265 ). Chichester: Wiley.
Detienne, F. (2002). Software design. Cognitive Hoffman, R. R. (1992). Bibliography: Expertise
aspects. London, UK: Springer. in programming. In R. R. Hoffman (Ed.),
Ericsson, K. A., Krampe, R. T., & Tesch-Romer, The psychology of expertise: Cognitive research
C. (1993 ). The role of deliberate practice in the and empirical AI (pp. 3 5 93 62). Hillsdale, NJ:
acquisition of expert performance. Psychologi- Erlbaum.
cal Review, 100, 3 63 406. Jeffries, R., Turner, A. A., Polson, P. G., &
Ericsson, K. A., & Lehmann, A. C. (1996). Atwood, M. E. (1981). The processes involved
Expert and exceptional performance: Evidence in designing software. In J. R. Anderson (Ed.),
of maximal adaptation to task constraints. Cognitive skills and their acquisition (pp. 25 5
Annual Review of Psychology, 47 , 273 3 05 . 283 ). Hillsdale, NJ: Erlbaum.
Ericsson, K. A., & Smith, J. (1991). Prospects Kelley, R., & Caplan, J. (1993 ). How Bell Labs cre-
and limits of the empirical study of expertise: ates star performers. Harvard Business Review,
An introduction. In K. A. Ericsson & J. Smith 71, 12813 9.
(Eds.), Toward a general theory of expertise: Kimball, D. R., & Holyoak, K. J. (2000). Trans-
Prospects and limits (pp. 13 8). Cambridge: fer and expertise. In E. Tulving & F. I. M.
Cambridge University Press. Craik (Eds.), The Oxford handbook of mem-
Etelapelto, A. (1993 ). Metacognition and the ory (pp. 109122). Oxford: Oxford University
expertise of computer program comprehen- Press.
sion. Scandinavian Journal of Educational Koubek, R. J., & Salvendy, G. (1991). Cogni-
Research, 3 7 , 243 25 4. tive performance of super-experts on com-
Feltovich, P. J., Spiro, R. J., & Coulson, R. L. puter program modification tasks. Ergonomics,
(1997). Issues of expert flexibility in contexts 3 4, 1095 1112.
characterized by complexity and change. In Koubek, R. J., Salvendy, G., Dunsmore, H. E.,
P. J. Feltovich, K. M. Ford, & R. R. Hoffman & LeBold, W. K. (1989). Cognitive issues in
(Eds.), Expertise in Context: Human and the process of software development: Review
Machine (pp. 125 146). Menlo Park, CA: MIT and rappraisal. International Journal of Man-
Press. Machine Studies, 3 0, 171191.
386 the cambridge handbook of expertise and expert performance
Krampe, R. T., & Ericsson, K. A. (1996). Main- Soloway & R. Iyengar (Eds.), Empirical studies
taining excellence: Deliberate practice and of programmers. Norwood, NJ: Abley.
elite performance in young and older pianists. Rist, R. S. (1991). Knowledge creation and
Journal of Experimental Psychology: General, retrieval in program design: A comparison of
12 5 , 3 3 13 5 9. novices and intermediate student program-
Krems, J. F. (1995 ). Expert strategies in mers. Human-Computer Interaction, 6, 146.
debugging: Experimental results and a com- Rist, R. S. (1996). System structure and design.
putational model. In K. F. Wender, F. In W. D. Gray & D. A. Boehm-Davis (Eds.),
Schmalhofer, & H.-D. Boecker (Eds.), Cogni- Empirical studies on programmers: Sixth work-
tion and computer programming (pp. 24125 4). shop. Norwood, NJ: Ablex.
Norwood, NJ: Ablex. Scholtz, J., & Wiedenbeck, S. (1992). Learning
Latham, G. P., Locke, E. A., & Fassina, N. E. new programming languages: An analysis of the
(2002). The high performance cycle: Standing process and problems encountered. Behaviour
the test of time. In S. Sonnentag (Ed.), Psy- & Information Technology, 11(4), 199215 .
chological management of individual performance Simon, H. A. (1973 ). The structure of ill-
(pp. 201228). Chichester: Wiley. structured problems. Artificial Intelligence, 4,
Law, L.-C. (1998). A situated cognition view 181204.
about the effects of planning and authourship
Soloway, E., & Ehrlich, K. (1984). Empirical stud-
on computer program debugging. Behaviour &
ies of programming knowledge. IEEE Transac-
Information Technology, 17 , 3 25 3 3 7.
tions on Software Engineering, 10, 5 95 609.
Locke, E. A., & Latham, G. O. (1990). A theory
Soloway, E., & Iyengar, S. (Eds.) (1986). Empirical
of goal setting and task performance. Englewood
studies of programmers. Norwood, NJ: Ablex.
Cliffs, NJ: Prentice Hall.
Sommerville, I. (2001). Software engineering (6th
Malhotra, A., Thomas, J. C., Carroll, J. M., &
edition). Harlow, UK: Addison-Wesley.
Miller, L. A. (1980). Cognitive processes in
design. International Journal of Man-Machine Sonnentag, S. (1995 ). Excellent software pro-
Studies, 12 , 119140. fessionals: Experience, work activities, and per-
McKeithen, K. B., Reitman, J. S., Rueter, H. H., ceptions by peers. Behaviour & Information
& Hirtle, S. C. (1981). Knowledge organization Technology, 14, 289299.
and skill differences in computer programmers. Sonnentag, S. (1998). Expertise in professional
Cognitive Psychology, 13 , 3 073 25 . software design: A process study. Journal of
Nanja, M., & Cook, C. R. (1987). An analysis of Applied Psychology, 83 , 703 715 .
the on-line debugging process. In G. M. Olson, Sonnentag, S. (2000). Expertise at work: Expe-
S. Sheppard, & E. Soloway (Eds.), Empiri- rience and excellent performance. In C. L.
cal studies of programmers. Second workshop Cooper & I. T. Robertson (Eds.), International
(pp. 172184). Norwood, NJ: Ablex. Review of Industrial and Organizational Psychol-
Olson, G. M., Sheppard, S., & Soloway, E. (Eds.) ogy (pp. 223 264). Chichester: Wiley.
(1987). Empirical studies of programmers: Second Sonnentag, S. (2001a). High performance and
workshop. Norwood, NJ: Ablex. meeting participation. An observational study
Pennington, N. (1987). Comprehension strate- in software design teams. Group Dynamics:
gies in programming. In G. M. Olson, S. Theory, Research and Practice, 5 , 3 18.
Sheppard, & E. Soloway (Eds.), Empirical stud- Sonnentag, S. (2001b). Using and gaining expe-
ies of programmers: Second workshop (pp. 100 rience in professional software development.
113 ). Norwood, NJ: Ablex. In E. Salas & G. Klein (Eds.), Linking expertise
Riedl, T. R., Weitzenfeld, J. S., Freeman, J. T., and naturalistic decision making (pp. 275 286).
Klein, G. A., & Musa, J. (1991). What we Mahwah, NJ: Erlbaum.
have learned about software engineering exper- Sonnentag, S., & Kleine, B. M. (2000). Deliber-
tise. Proceedings of the Fifth Software Engineer- ate practice at work: A study with insurance
ing Institute Conference on Software Engineering agents. Journal of Occupational and Organiza-
Education (pp. 261270). New York: Springer. tional Psychology, 73 , 87102.
Rist, R. S. (1986). Plans in programming: Defi- Sonnentag, S., & Lange, I. (2002). The relation-
nition, demonstration and development. In E. ship between high performance and knowledge
expertise in software design 387
Ronald T. Kellogg
Keywords: planning, translating, reviewing, the final section, the acquisition of writing
deliberate practice, ten-year rule, flow states, skill will be discussed, with comparisons and
working memory, long-term working mem- contrasts to other kinds of expertise high-
ory, domain-specific knowledge, verbal abil- lighted. Much remains to be learned, but
ity, concrete language, strategies, rituals, the lessons from the state-of-the-art research
work environment, work schedule. literature can be helpful to aspiring profes-
sional writers.
of the list of professional writers. Many pro- the author intends to eventually be read
fessionals often devote considerable time and understood by others. Several linguistic
and effort to writing, even though they processes, largely nonconscious, operate on
define their job in other ways (e.g., pro- images and propositions to generate a sen-
fessors, scientists, engineers, business man- tence, such as selecting words and assign-
agers, government bureaucrats, and diplo- ing them to syntactic roles (Bock & Levelt,
mats). The jobs of career writers, those who 1994). Besides translating ideas into a sen-
compose on technical and professional sub- tence, writers must also generate cohesive
jects for a living, certainly differ from those links among sentences, establishing coher-
of engineers and administrators in an organi- ence at both local and global levels in the text
zation (Couture, 1992). The scientific litera- (McCutchen, 1984). For written output,
ture on professional writing samples from all graphemic representations must be specified
of these domains, albeit too sparsely to draw for spelling each word (Caramazza, 1991).
many conclusions about how they differ. The motor transcription of words into writ-
The focus of this chapter, therefore, is on ten characters often closely follows sentence
the common elements of professional writ- generation in time and is difficult to distin-
ing expertise, while recognizing that much guish through natural observation alone. It
more research in this area is possible. can, therefore, be viewed as a subprocess of
translating (Berninger & Swanson, 1994).
Reviewing the text involves reading and
Cognitive Demands of Writing
editing operations that detect faults at mul-
Hayes and Flower (1980) distinguished three tiple levels of text structure, ranging from
basic processes of text production: plan- local matters of diction, spelling, or punc-
ning ideas, translating ideas into text, and tuation to the coherence of the whole text.
reviewing ideas and text. As will be dis- It further involves editing ideas and other
cussed, these are not linear phases of text products of planning (e.g., lists, outlines,
production, starting with planning and end- diagrams) prior to their conversion to sen-
ing with reviewing. Instead, each process can tences and extended text. The reading and
be and often is invoked throughout all phases editing processes are complex because so
of text development, from prewriting to a many things can go wrong in a plan or a
final draft. text at so many different levels of struc-
Planning includes generating concepts, ture (Hayes, Flower, Schriver, Stratman, &
organizing them, and setting goals to be Carey, 1987). At the word level of a text,
achieved in the structure of the text. The for example, the writer may detect prob-
products of planning fall along a contin- lems with the graphemes used for spelling
uum ranging from nonverbal imagery to or with the shade of meaning conveyed.
abstract propositions and word images that At the sentence level, grammatical errors,
are more readily translated to written text semantic ambiguities, alternative interpre-
(Flower & Hayes, 1984). Sequences of word tations, and problems of reference must be
images or mental pre-texts are what Witte evaluated. The writer must further address
(1987) called the last cheap gas before the faulty logic, errors of fact, and other incon-
major cost of concretizing plans into written sistencies with world knowledge, on the
text. The only visible products of a writers one hand, and problems with text structure,
planning consist of diagrams, outlines, lists, incoherence, and disorganization, on the
and other notebooks of the mind, to other. As if this were not enough, the writer
borrow John-Steiners (1985 ) description. must also consider the needs of the audi-
These externalized plans are typically cryp- ence, looking for the right tone and degree
tic and intended only for the writers of complexity.
private use. As noted earlier, the basic processes of
The translation of ideas into sentences writing can be distinguished from the tem-
and paragraphs yields a draft of a text that poral phases of composing a text. The
professional writing 391
research literature clearly shows that plan- struggle concurrently with two separate
ning, translating, and reviewing do not occur problems (Bereiter & Scardamalia, 1987).
in a linear sequence in text production; On the one hand, there is the problem of
rather, they occur and reoccur in complex what to say. The writer mentally represents
patterns through prewriting, first draft, and and manipulates beliefs and facts about the
subsequent draft or revision phases of com- topic within the content problem space.
position (Kellogg, 1994). The interactions On the other hand, there is the different
among planning, translating, and review- problem of how to say it. In the rhetori-
ing are responsible for the development of cal problem space, the writer finds ways
sophisticated descriptions, narratives, and to achieve the goals of the composition.
arguments. During a prewriting phase, for The importance of finding a good way
example, the writer may plan, review the to represent content and rhetorical intent
ideas and mental pre-text, and then plan can be seen in interviews with six profes-
some more to produce, say, an outline of top- sional essayists (Dowdy, 1984). For all of
ics to be included in a first draft. Although them, deciding on the topic, researching the
prewriting can be brief, experts approach- topic, and constructing an integrating theme
ing a serious writing assignment may spend came first. Having a clear, concrete theme
hours, days, or weeks thinking about the task was important because it allowed the essay-
before initiating a draft. ists to attach their ideas to it, thus intercon-
During the first draft phase, writers con- necting them. A necessary skill for a pro-
tinue to plan, generate sentences, read the fessional writer, therefore, is the ability to
developing text, and edit ideas and text. represent content and rhetorical problems,
Although some writers try to produce a per- to explore these representations for satisfac-
fect first draft that requires only minor cor- tory solutions during composition, and to
rections, the first draft is typically revised update them as new insights are achieved
through one or perhaps many subsequent (Scardamalia & Bereiter, 1991).
drafts. Fitzgerald (1987) defined the revi-
sion phase of reworking the first draft as
Language Use
identifying discrepancies between intended
and instantiated text, deciding what could Professional writers use language to acti-
or should be changed in the text and how vate extensive associations among verbal
to make the desired changes, and . . . making and imaginal representations in the reader
the desired changes (p. 484). In revision, as (Sadoski & Pavio, 2001). It is the reader who
in the original drafting of a text, the writer finds a text meaningful, coherent, interest-
may engage again in planning and translating ing, or persuasive. The words crafted by the
as well as further reviewing. author can potentially trigger the readers
long-term memory representations from the
bottom up and motivate constructive pro-
Writing Expertise cesses from the top down. That is to say, the
text both relates to what a reader already
knows and stimulates new thinking at the
We now turn to the specific skills that pro-
same time. Deep comprehension of a text
fessional writers bring to their job. This is not
calls on the reader to construct a mental
intended as an exhaustive list but rather as
representation of what the text says and
a selection of the key characteristics of pro-
how this information integrates with other
fessional writers and their writing skills.
knowledge about the world (Kintsch, 1998).
A shallow level of comprehension is frag-
Problem Solving
mentary by comparison. Professional writers
Aspiring professional writers must be good aim to engage the reader in such deep com-
at thinking through ill-structured problems prehension. How can their skill in doing so
(Hayes & Flower, 1980). Experienced writers be characterized?
392 the cambridge handbook of expertise and expert performance
chess experts evaluating multiple moves in imagination, as a way of composing the text
the middle stages of a game (Britton & (Tomlinson, 1986). But again, listening to
Tessor, 1982). the characters dictate the story is not an
A variety of cognitive strategies help the essential feature of successful creative writ-
writer to cope with the cognitive demands ers, though many do find daydreaming and
of text production. Prewriting, drafting, and even dreaming at night conducive to their
revising strategies are essential strategies. work (Epel, 1993 ).
Preparing an outline as a prewriting strat-
egy enables the writer to focus attention on
Domain Specificity
translating ideas into text while drafting and
can enhance productivity (Kellogg, 1988). As noted earlier, writing expertise is highly
Individuals who write in organizations as domain dependent. In an important sense,
part of their job responsibilities frequently then, the professional writer may be a rare
report the need to plan a lot mentally (98%), individual indeed, when conceived as a gen-
create notes and lists (95 %), and prepare eralist capable of writing across any num-
outlines (73 %), especially for serious writ- ber of domains (Carter, 1996). An outstand-
ing tasks that do not fit a routine schema ing journalist might be lost if they had
(Couture & Rymer, 1993 ). to compose a scientific report. A scientist
Highly successful writers have reported may be equally adrift in trying to write for
a wide range of prewriting strategies that the general public. Even within a profes-
can be expressed as styles (Cowley, 195 8; sion, a writer often specializes in a spe-
Plimpton, 1963 ; 1989). Beethovians engage cific rhetorical context. For example, the
in few prewriting activities and prefer to journalist who specializes in movie reviews
compose rough first drafts immediately to for the New Yorker might well perform
discover what they have to say (Bridwell- more like a novice in writing market anal-
Bowles, Johnson, & Brehe, 1987). Their yses for the Wall Street Journal. There are
drafting necessarily involves many rounds indeed diverse, highly specific skills required
of revision. By contrast, Mozartians delay by journalism, exposition, creative fiction,
drafting for lengthy periods of time in order business writing, or technical writing. Pro-
to allow time for extensive reflection and fessionalism in any one of these kinds of
planning. They may also plan mental pre- writing requires a progression beyond gen-
text that is later recalled and written down as eral writing skills, such as outlining, flu-
a first polished draft. A variety of notational ent sentence generation, and effective topic
methods are used to externalize plans dur- sentences, to domain-specific knowledge
ing prewriting, including tree diagrams, flow and skills.
charts, boxes, arrows, doodles, and scrib- For example, the expert must master the
bles, as well as lists and outlines (John- rhetorical style required in a given domain.
Steiner, 1985 ). Psychologists, for example, learn the rhetor-
Preparing a rough draft can help writers ical style mandated by the American Psy-
to reduce the number of processes juggled chological Association Publication Manual,
simultaneously by dissociating the author which contrasts with the Chicago Manual
from an editing mindset (Glynn, Britton, of Style used in the humanities (Madigan,
Muth, & Dogan, 1982). However, profes- Johnson, & Linton, 1995 ). Psychologists
sional writers who report attempting to pro- conclusions must be hedged so as to avoid
duce a perfect first draft are just as produc- stepping beyond the limits of the data, cre-
tive as those who report starting with a rough ating an air of uncertainty that is atypical in
draft (Kellogg, 1986). An intriguing strategy the humanities. Disagreements are couched
reported by some fiction writers is to disso- in terms of differences about appropriate
ciate the author into multiple characters, so methods or interpretations of data, never in
to speak. The writer observes what the char- the personal terms at times seen in literary
acters are doing and saying, in the writers criticism.
394 the cambridge handbook of expertise and expert performance
Managing the Emotional Challenges poets and 29 fiction writers about the con-
ditions that lead to flow. Entry into flow,
Professional writers, as with other practi-
she concluded, is chiefly facilitated by having
tioners of the creative arts, must be self-
a strong motivation to write. Believing that
motivated to commit long hours to a lonely
the writing task is intrinsically important is
task of working with ideas and language
sufficient for many professional writers, but
rather than with other people. The emo-
Perry found also that pay and other extrinsic
tional demands of writing are just as chal-
motivators stimulate flow. By contrast, some
lenging as the cognitive demands (Brand,
evidence suggests that a writers creativ-
1989). Experienced writers learn to self-
ity can be diminished by extrinsic rewards
regulate their emotions and behavior to stay
(Amabile, 1985 ).
on task and complete the work by los-
The positive affect of flow states can char-
ing themselves in their work, engineering
acterize the work of professional writers,
their work environment, adhering to a work
but they also face frustrations, anxieties, and
schedule, and practicing motivational ritu-
other negative emotions when the demands
als. Breakdowns in their efforts to regulate
of the task temporarily exceed their capac-
their emotions and behavior can, in seri-
ities. Writers block can be one conse-
ous cases, result in writers block (Boice,
quence whereby production is sharply cur-
1994).
tailed or stopped altogether (Boice, 1994).
Perry (1996) documented that some writ-
flow states ers force themselves during these periods to
A common experience of experts engaged produce line after line . . . and the product of
in their craft is feeling lost in the work. The such reluctant writing sessions is often not
writer looses a sense of ego and an awareness up to what the writers consider their stan-
of the immediate surroundings. Hours can dard (p. 275 ). Thus, judging from Perrys
pass quickly as a skilled writer is absorbed interviews, professionals manage to tolerate
in thought and language. Csikszentmihalyi negative emotional states as well as enjoy-
(1990) referred to such absorption and the ing the positive experience of flow. It is also
positive affect associated with it as a flow important to recognize that procrastination
state of consciousness. Flow can be produced can be beneficial when it reflects a real need
when an individuals aptitude and skills are to take more time (Murray, 1978). Exten-
well matched by the demands of any task sive prewriting activities can be a way for
and writing is no exception (Larsen, 1988). professional writers to digest and understand
Clarity of purpose, a sense of mastery, and their subject thoroughly before beginning a
high intrinsic motivation further character- first draft.
ize flow. When the demands exceed ones
capacity, on the other hand, then the con- environments, schedules, and rituals
sequence is anxiety and frustration. A mis- Where, when, and how writers work show
match in the opposite direction produces enormous variability. However, many writ-
boredom. ers develop habitual ways of approaching
Schere (1998) contrasted university pro- their work such that they become nec-
fessors of creative writing, fine arts, and engi- essary conditions for effective knowledge
neering on a drawing versus writing task. use. Perrys (1996) interviews with estab-
As predicted, the creative writers reported lished creative writers and surveys of uni-
a higher level of positive mood after com- versity faculty who produce scholarly pub-
pleting the writing task compared with the lications (Boice & Johnson, 1984; Hartley &
drawing task. Brand and Leckies study of Branthwaite, 1989; Kellogg, 1986) converge
professional writers (1989) also concluded on three points. All of the idiosyncratic
that the opportunity to engage a task that habits of professional writers (1) focus
fits the writers capabilities enhances pos- attention inward by eliminating distractions,
itive mood. Perry (1996) interviewed 3 3 (2) may alter consciousness to facilitate
396 the cambridge handbook of expertise and expert performance
entry in a flow state, and (3 ) help regulate of writing is important for writers to avoid
the writers emotional state to keep at burning out and perhaps ending their career
the task. prematurely. Self-regulation through daily
For example, whereas some writers report writing, brief work sessions, realistic dead-
that they find a quiet environment useful, lines, and maintaining low emotional arousal
others preferred background music or the help professional writers stay with the task
bustle of a cafe. Whereas some must write for the weeks, months, and years required
with a word processor, others prefer long- (Boice, 1994; Zimmerman & Risemberg,
hand or a typewriter. Many choose to write 1997). Binge writing hypomanic, euphoric,
at the same time each working day, but marathon sessions to meet unrealistic dead-
individuals differ from morning, afternoon, lines is generally counterproductive and
evening, late night, to early morning pref- potentially a source of depression and block-
erences. Work sessions of one to two hours ing (Boice, 1997).
correlate with productivity in scientific writ- To summarize, there is a wide variety of
ing, but the relationship is weak (r = .22) skills and qualities that professional writers
and the variability is large, with some indi- bring to their daily work. We now turn to
viduals writing four or more hours at a time what is known about the acquisition of those
(Kellogg, 1986). Successful poets also typ- writing skills.
ically write for one or two hours, whereas
most novelists typically report longer ses-
sions of two to three or even four to six Skill Acquisition
hours (Perry, 1996). Running or walking help
some writers think through problems while
Literacy is a fundamental goal of school-
away from the writing table (Oates, 2003 ;
ing in contemporary societies worldwide.
Kellogg, 1986). Others use meditation, cof-
Thus, unlike medical diagnosis, finance, and
fee, cigarettes, alcohol, or other drugs to
other skills for which practice begins in early
alter consciousness in the service of writing
adulthood for a select few, writing devel-
(Piirto, 2002).
opment starts in early childhood in indus-
blocking trialized nations. The foundations of liter-
acy are established in toddlers with letter
Writers block, defined as a persistent inabil-
and word recognition and scribbling. By the
ity to put thoughts on paper, can impede,
time a child enters school, practice in hand-
if not end, the career of a professional
writing is well underway, and children as
writer. Boice (1985 ) found the thoughts
young as four years of age distinguish writ-
of blocked academics were characterized
ing from drawing (Lee & Karmiloff-Smith,
by procrastination, dysphoria (e.g., burnout,
1996).
panic, obsessive worries), impatience (e.g.,
thoughts of achieving more in less time
Deliberate Practice
or imposing unrealistic deadlines), perfec-
tionism (e.g., thoughts reflecting an inter- Ericsson, Krampe, and Tesch-Romer (1993 )
nal critic who allows no errors), and evalua- defined the characteristics of deliberate prac-
tion anxiety (e.g., fears of being rejected). Of tice as (1) effortful exertion to improve per-
interest, unblocked as well as blocked writ- formance, (2) intrinsic motivation to engage
ers fretted about how difficult and demand- in the task, (3 ) practice tasks that are within
ing the task was for them. reach of the individuals current level of
Thus, professional writers not only know ability, (4) feedback that provides knowl-
how to enter flow, but how to endure the edge of results, and (5 ) high levels of rep-
intense negative feelings that often accom- etition. They contrasted deliberate practice
pany the early stages of writing a text, when with work activities for pay or other external
all can seem hopelessly incoherent. Learn- rewards and play activities that are enjoyable
ing to manage the emotional ups and downs but have no goal for improving performance.
professional writing 397
Deliberate practice can only be sustained for writer, and I wrote . . . I learned to write by
a limited amount of time each day because writing. As I once calculated, I must have
of the effort involved. written more than a half a million words
Signs of deliberate practice in writing can before I came to The Naked and the Dead.
be seen in work habits and practice tech- (pp. 13 14)
niques. Successful writers often schedule In his autobiography, Benjamin Franklin
only a few hours per day for composing, and explained how he rewrote admired texts
avoid binges that lead to exhaustion (Boice, that he wished to emulate (Lemay & Zall,
1994; 1997). High levels of practice can be 1981). He took careful notes on a text and
seen in the daily work schedules of pro- then several days later used these as retrie-
lific writers (Cowley, 195 8; Kellogg, 1982; val cues to reconstruct the original in his
Plimpton, 1963 ). own words.
It has long been known that composi- Evidence of deliberate practice emerges
tion instructors can coach writers through from several ethnographic studies of pro-
feedback. Apprenticeships in creative writ- fessional writers (Henry, 2000; MacKinnon,
ing programs and schools of journalism 1993 ; Paradis, Dobrin, & Miller, 1985 ).
rely on this method of practice. The Iowa Learning on the job, practicing to perfec-
Writers Workshop, the earliest creative tion, and viewing oneself as still develop-
writing degree program in the United States, ing in skill over the course of ones career
has from the outset emphasized learning are typical traits of these writers. Success-
by doing and immersing writers in feed- ful expository writers who meet weekly
back from others (Adams, 1993 ). As early and monthly deadlines continue to develop
as 1890, writers clubs formed in Iowa City, their skills through immersion in a habit-
where students and faculty could read and ual, predictable task environment (Root,
critique each others work and practice their 1983 ).
craft. Journalists, too, have always learned by
doing, first in newspaper offices and later in
Reading
formal university programs. From its begin-
ning in 1908, the University of Missouris Extensive reading is a powerful predictor of
School of Journalism has run a daily city the amount of general knowledge that an
newspaper for students to practice their individual accumulates in long-term mem-
skills with instruction and feedback from ory (Stanovich & Cunningham, 1993 ). A
the faculty. composite measure of print exposure is
Well-known writers have reported valued strongly correlated with a composite mea-
practice techniques. As a college student, sure of general knowledge (r = .85 ). This
Joyce Carol Oates would write a novel in relationship remains statistically significant
longhand, then turn the pages over, writ- even after the effects of cognitive ability are
ing another novel on the flipside. Both nov- removed. How much one knows depends on
els would then be tossed in the trash. Since how much one reads.
high school she began consciously train- Because a wide range of knowledge is
ing myself by writing novel after novel and so important for writers (Kellogg, 1994),
always throwing them out when I completed it is not surprising that professional writ-
them (Plimpton, 1989; p. 3 78). Her practice ers report reading extensively, even compul-
books would be modeled after specific works sively. Louis LAmour, the prolific writer of
by authors she admired, such as Heming- novels about the American West, recounted
ways In Our Time. Norman Mailer (2003 ) in Education of a Wandering Man (1989) that
credited his eventual success as a writer to he read hundreds of books as he traveled
self-motivated practice. the world. Mark Singer, a staff writer at the
New Yorker magazine, reads and re-reads the
I think from the time I was seventeen, I masters of nonfiction, such as E. B. White,
had no larger desire in life than to be a Calvin Trillin, and John McPhree (Pearson,
398 the cambridge handbook of expertise and expert performance
1998). Truman Capote, when asked whether telling (Bereiter & Scardamalia, 1987). An
he read a great deal, responded: idea is retrieved from long-term memory and
then told to the reader by writing it down.
Too much. And anything, including labels
This process continues until the writer has
and recipes and advertisements. I have a
passion for newspapers read all the New no more ideas to retrieve and communi-
York dailies everyday, and the Sunday edi- cate. The heavy demands made on working
tions, and several foreign magazines too. memory by handwriting alone make it dif-
The ones I dont buy I read standing at news ficult for young writers to do much more
stands. (Capote quoted in Cowley, 195 8; than retrieve and translate (Bourdin & Fayol,
p. 2 93 ) 1994; McCutchen, 1996). They can main-
tain a representation of what they intend to
Piirto (2002) concluded from her case
write, but no more than this.
studies that professional writers often begin
By contrast, Bereiter and Scardamalia
reading early in childhood and read compul-
(1987) characterized adult writing as knowl-
sively throughout their lifespan. The read-
edge transforming. What the writer thinks
ing begins as a way to learn about the world.
is transformed by the act of composition.
The reading later becomes more focused and
The adult writer reads the evolving text
intellectually challenging during the course
and develops a representation of what it
of formal education, according to Piirto.
says to the author. Reflection on the exist-
As writers develop professionally, they read
ing text can prompt the writer to restruc-
within the genre in which they work. For
ture the ideas stored in long-term memory,
example, science fiction writers read sci-
elaborating and reorganizing what the writer
ence fiction, whereas romance writers read
knows about the topic. Whereas knowledge-
romances.
telling often produces a string of assertions
linked by and, knowledge-transforming
The Ten-Year Rule
yields complex argument structures and the
Studies of chess players (Simon & Chase, use of cohesive devices to link clauses in a
1973 ), musical composers (Hayes, 1985 ), paragraph. The adult writer, therefore, main-
and other domains (Ericsson et al., 1993 ) tains representations of the authors cur-
have suggested a rule of thumb that it takes rent intent and what the text says from the
at least a decade of intensive practice to authors perspective.
achieve excellence. In the case of writing, After mastering handwriting and achiev-
the clock starts early, since spoken language ing written fluency at ages 12 to 14,
and scribbling are developed in preliterate approximately a decade of practice is need-
children (Lee & Karmiloff-Smith, 1996). By ed to progress from knowledge-telling
the age of 12 to 14 years, children have spent to knowledge-transforming. Bereiter and
ten years mastering the mechanics of hand- Scardamalia (1987) turned to graduate-
writing and spelling. Approximately during student writing to illustrate knowledge-
this same time frame, they advance from transforming. It is unknown how long it
thinking in term of concrete events in the takes to advance further to knowledge-
here and now to thinking in hypothetical, crafting, whereby professionals can mentally
abstract terms. From studies that tracked the represent the text as it appears to the reader
cohesion of texts produced by children at and respond to their needs adroitly. But sev-
different ages, Scinto (1986) concluded that eral years are probably needed to acquire the
this advancement from concrete to abstract domain-specific rhetorical skills and practice
thinking is essential for writing cohesive at crafting knowledge for a specific audience
texts. Although speech is proficient by the (Rymer, 1988).
age of six, written fluency does not catch up Wishbow (1988) examined the biogra-
until around the age of 12 years (Bereiter & phies of 66 poets listed in the Norton Anthol-
Scardamalia, 1987). ogy of Poetry, locating their approximate
The written production strategy of chil- starting date for reading and writing poetry.
dren is a simple one of egocentric knowledge The earliest work to appear in Nortons came
professional writing 399
during the ten years after this date or later many specific kinds of professional writers,
for 83 % of the sample. The poets began but some features apply to them all. Exten-
reading and writing poetry in their early sive reading, high verbal ability, the skilled
teens to early twenties. Thus, ten to twenty use of concrete language, and the ability to
years of writing seemed to span their first envision and respond to the readership are
scribbling as a toddler to their first master- hallmarks of the writers craft.
piece. T. S. Elliot wrote his masterpiece, The The other common features of profes-
Waste Land, in his early 3 0s, about a decade sional writers also overlap with different
after composing his first published poem kinds of expertise and figure prominently
(Gardner, 1993 ). in a general theory of expertise (Ericsson
Not surprisingly, the earlier the writer & Smith, 1991). Strategies for managing
starts the better. Childhood story writing the cognitive load on working memory are
was so commonly mentioned in Henrys important for the writer as well as for profes-
(2002; p. 3 7) ethnographies that people sionals in computer programming, for exam-
who were attracted to writing after child- ple. Knowledge about the domain permits
hood may even refer to themselves as late the rapid retrieval of information from long-
bloomers. A study of 986 creative writ- term memory and minimizes the demands
ers found a significant correlation between on short-term working memory, such as is
the age of first publication and the num- observed in master chess players. The ten-
ber of works published in total for poets year rule of expert skill acquisition applies
(Kaufman & Gentile, 2002). Replicating and to writers and, if anything, underestimates
extending Wishbows findings, both poets the number of years of deliberate prac-
and fiction writers developed mechanics and tice required to reach professional levels
cognitive writing skills for 15 to 20 years of achievement. Writers, as with surgeons,
before first publishing. Only 10% of the musicians, and athletes, perform best in a
sample published prior to the age of 21. flow state of consciousness. They all try to
About half (49%) first published in their shape their surroundings, work schedules,
twenties and the remainder in their thirties and rituals in ways that foster flow. Self-
or later. regulation of the emotional demands of writ-
Isaac Asimovs prolific career illustrates ing in these ways is necessary for a produc-
how productivity depends on decades of tive career, a feature shared in common with
preparation and practice (see the Asimov many professions. Much remains to be dis-
web site, http:www.asimovonline.com/). covered about the skills of professional writ-
His first book appeared in 195 0. He pub- ers and longitudinal studies of their develop-
lished one, two, or three books a year for the ment would be particularly informative, but
next six years, but then gradually began to those aspiring to the role can learn from the
publish more than five books a year on a con- findings reviewed here.
sistent basis from 1962 on. It took Asimov
19 years to publish his first 100 books, but
only ten years to publish the second 100. His References
productivity then accelerated after approx-
imately 3 0 years of practice at the craft. It Adams, K. H. (1993 ). A history of professional
took him only five additional years to bring writing instruction in American colleges. Dallas:
his total to 3 00 books. Southern Methodist Press.
Amabile, T. M. (1985 ). Motivation and creativity:
Effects of motivational orientation on cre-
ative writers. Journal of Personality and Social
Conclusion Psychology, 48, 3 93 3 99.
Asimov web site. (n.d.). Retrieved December 6,
Professional writing reflects not just general 2003 from http:www.asimovonline.com/.
writing ability but also expertise in a partic- Berninger, V. W., & Swanson, H. L. (1994). Mod-
ular genre and domain. Thus, there are really ifying Hayes and Flowers model of skilled
400 the cambridge handbook of expertise and expert performance
Geisler, C., Rogers, E. H., & Haller, C. R. (1998). early start in productive and successful creative
Disciplining discourse: Discourse practice in writing. High Ability Studies, 13 (2), 115 123 .
the affilitated professions of software engi- Kellogg, R. T. (1986). Writing method and pro-
neering design. Written Communication, 15 , ductivity of science and engineering faculty.
3 24. Research in Higher Education, 2 5 , 147163 .
Givon, T. (1995 ). Coherence in the text and Kellogg, R. T. (1988). Attentional overload and
coherence in the mind. In M. A. Gernsbacher writing performance: Effects of rough draft
& T. Givon (Eds.), Coherence in spontaneous text and outline strategies. Journal of Experimental
(pp. 13 9160). Amsterdam: John Benjamins. Psychology: Learning, Memory, and Cognition,
Glynn, S. M., Britton, B. K., Muth, D., & 14, 3 5 5 3 65 .
Dogan, N. (1982). Writing and revising persua- Kellogg, R. T. (1994). The psychology of writing.
sive documents: Cognitive demands. Journal of New York: Oxford University Press.
Educational Psychology, 74, 5 5 75 67. Kellogg, R. T. (2001). Long-term working mem-
Grobe, C. (1981). Syntactic maturity, mechanics, ory in text production. Memory & Cognition,
and vocabulary as predictors of quality ratings. 2 9, 43 5 2.
Research in the Teaching of English, 15 , 75 85 . Kintsch, W. (1998). Comprehension: A paradigm
Hartley, J. (2000). Legal ease and legalese. Psy- for cognition. Cambridge: Cambridge Univer-
chology, Crime, and Law, 6, 120. sity Press.
Hartley, J., & Branthwaite, A. (1989). The psy- Larsen, R. (1988). Flow and writing. In M. Csik-
chologist as wordsmith: A questionnaire study szentmihalyi & I. S. Csikszentmihalyi (1988).
of the writing strategies of productive British Optimal experience: Psychological studies of flow
psychologists. Higher Education, 18, 423 45 2. in consciousness (pp. 15 0171). Cambridge:
Hayes, J. R. (1985 ). Three problems in teaching Cambridge University Press.
general skills. In S. F. Chipman, J. W. Segal, & Lee, K., & Karmiloff-Smith, A. (1996). The devel-
R. Glaser (Eds.), Thinking and learning skills: opment of external symbol systems: The child
Vol. 2 . Research and open questions (pp. 3 91 as notator. In R. Gelman & T. Kit-Fong (Eds.),
405 ). Hillsdale, NJ: Erlbaum. Perceptual and cognitive development (185 211).
Hayes, J. R., & Flower, L. S. (1980). Identify- San Diego, CA: Academic Press.
ing the organization of writing processes. In Lemay, J. A. L., & Zall, P. M. (Eds.) (1981). The
L. W. Gregg & E. R. Steinberg (Eds.), Cognitive autobiography of Benjamin Franklin: A genetic
processes in writing (pp. 3 3 0). Hillsdale, NJ: text. Knoxville: Univerisity of Tennessee Press.
Erlbaum. LAmour, L. (1989). Education of a wandering
Hayes, J. R., Flower, L. S., Schriver, K. S., man. New York: Bantam.
Stratman, J., & Carey, L. (1987). Cogni- Madigan, R., Johnson, S., & Linton, P. (1995 ). The
tive approaches in revision. In S. Rosenberg language of psychology: APA style as episte-
(Ed.), Advances in applied psycholinguistics: mology. American Psychologist, 5 0, 42843 6.
Vol. 2 . Reading, writing, and language pro- Mailer, N. (2003 ). The spooky art: Some thoughts
cessing (pp. 176240). New York: Cambridge on writing. New York: Random House.
University Press.
MacKinnon, J. (1993 ). Becoming a rhetor: Devel-
Hayes, J. R., & Flower, L. S. (1986). Writing oping writing ability in a mature, writing inten-
research and the writer. American Psychologist, sive organization. In R. Spilka (Ed.), Writ-
41, 11061113 . ing in the workplace: New research perspectives
Henry, J. (2000). Writing workplace cultures: An (pp. 415 5 ). Carbondale, IL: Southern Illinois
archeology of professional writing. Carbondale, University Press.
IL: Southen Illinois University Press. McCutchen, D. (1984). Writing as a linguistic
Hyland, K. (2001). Bringing in the reader: problem. Educational Psychologist, 19, 22623 8.
Addressee features in academic articles. Writ- McCutchen, D. (1996). A capacity theory of
ten Communication, 18, 5 495 74. writing: Working memory in composition. Edu-
John-Steiner, V. (1985 ). Notebooks of the mind: cational Psychology Review, 8, 2993 25 .
Explorations of thinking. Albuquerque: Univer- McCutchen, D. (2000). Knowledge, processing,
sity of New Mexico Press. and working memory: Implications for a the-
Kaufman, J. C., & Gentile, C. A. (2002). The will, ory of writing. Educational Psychologist, 3 5 ,
The wit, The judgment: The importance of an 13 23 .
402 the cambridge handbook of expertise and expert performance
Murray, D. M. (1978). Write before writing. Col- or context availability? Reading Research Quar-
lege Composition and Communication, 2 9, 3 75 terly, 3 0, 278288.
3 81. Scardamalia, M., & Bereiter, C. (1991). Liter-
Oates, J. C. (2003 ). The faith of a writer: Life, craft, ate expertise. In K. A. Ericsson & J. Smith
and art. New York: Harper Collins. (Eds.), Toward a general theory of expertise:
Paradis, J., Dobrin, D., & Miller, R. (1985 ). Prospects and limits (pp. 172194). Cambridge:
Writing at Exxon ITD: Notes on the writ- Cambridge University Press.
ing environment of an R&D organization. In Schere, J. J. (1998). Effect of engaging in cre-
L. Odell & D. Goswami (Eds.), Writing in ative activity on the mood of artists and writers:
nonacademic settings (pp. 2813 07). New York: An empirical test of flow theory. (Doctoral
Guilford. Dissertation, The California School of Profes-
Pearson, M. (1998). Mr. Personalities: A conver- sional Psychology). Dissertation Abstracts Inter-
sation about the writing process with Mark national, 5 9 (06), 3 074B.
Singer. Creative Nonfiction, 9, 118125 . Scinto, L. F. M. (1986). Written language and
Perry, S. K. (1996). When time stops: How cre- psychological development. Orlando: Academic
ative writers experience entry into the flow Press.
state. Dissertation Abstracts International, 5 8 Simon, H. A. (1973 ). The structure of ill-
(8), 4484B. (UMI No. 9805 789) structured problems. Artificial Intelligence, 4,
Piirto, J. (2002). My teeming brain: Understand- 181210.
ing creative writers. Cresskill, NJ: Hampton Simon, H. A., & Chase, W. G. (1973 ). Skill in
Press. chess. American Scientist, 61, 3 94403 .
Plimpton, G. (Ed.) (1963 ). Writers at work: The Stanovich, K. E., & Cunningham, A. E. (1993 ).
Paris Review interviews, second series. New York: Where does knowledge come from? Specific
Penguin. associations between print exposure and infor-
Plimpton, G. (Ed.) (1989). Women writers at mation acquisition. Journal of Educational Psy-
work: The Paris Review interviews. New York: chology, 85 , 211229.
Penguin. Tomlinson, B. (1986). Characters as co-authors:
Root, R. L. (1983 ). The composing processes Segmenting the self, integrating the composing
of professional expository writers. Paper pre- process. Written Communication, 3 , 421448.
sented at the 3 4th annual meeting of the Traxler, M. J., & Gernsbacher, M. A. (1992).
Conference on College Composition and Com- Improving written communication through
munication. Detroit. (ERIC Document Repro- minimal feedback. Language and Cognitive
duction Service No. ED23 215 7) Processes, 7 , 122.
Rymer, J. (1988). Scientific composing processes: Wishbow, N. A. (1988). Studies of creativity in
How eminent scientists write journal arti- poets. (Doctoral Dissertation, Carnegie Mel-
cles. In D. A. Jollife (Ed.), Advances in lon University). Dissertation Abstracts Interna-
writing research, Volume 2 : Writing in aca- tional, 5 1, 0491A.
demic disciplines (pp. 21125 0). Norwood, NJ: Witte, S. P. (1987). Pre-text and composing.
Ablex. College Composition and Communication, 3 8,
Sadoski, M., & Paivio, A. (2001). Imagery and 3 97425 .
text: A dual coding theory of reading and writ- Zimmerman, B. J., & Risemberg, R. (1997).
ing. Mahwah, NJ: Erlbaum. Becoming a self-regulated writer: A social cog-
Sadoski, M., Goetz, E. T., & Avila, E. (1995 ). nitive perspective. Contemporary Educational
Concreteness effects in text recall: Dual coding Psychology, 2 2 , 73 101.
C H A P T E R 23
This chapter looks at expertise from the Making in Action: Models and Methods (G. A.
perspective of the community of prac- Klein, Orasanu, Calderwood, & Zsambok,
tice known as Naturalistic Decision Mak- 1993 ).1
ing (NDM). We provide an overview of the Zsambok (1997) described how this pub-
emergence of NDM, the underlying theoret- lication brought together the contextual fac-
ical orientation, and key NDM research. We tors that defined NDM:
discuss the impact NDM has had on one par-
ticular domain of expert judgment, military The identification of key contextual fac-
tors that affect the way real-world deci-
decision making. We conclude by discussing
sion making occurs, in contrast to their
applications and continuing research issues
counterparts in the traditional decision
in NDM. research paradigm, evolved as a major
contribution of the 1989 NDM conference
(Orasanu & Connolly, 1993 ). They are:
1. Ill-structured problems (not artificial,
Emergence of Naturalistic well-structured problems). 2 . Uncertain,
Decision Making dynamic environments (not static, simu-
lated situations). 3 . Shifting, ill-defined,
The focus of NDM research is on expert or competing goals (not clear and stable
practitioners trying to figure out what to goals). 4. Action/feedback loops (not one-
shot decisions). 5 . Time stress (as opposed
do under difficult circumstances. The need
to ample time for tasks). 6. High stakes
to understand decision making in the con-
(not situations devoid of true consequences
text of time pressure, uncertainty, ill-defined for the decision maker). 7. Multiple play-
goals, and high personal stakes was a major ers (as opposed to individual decision mak-
impetus for the emergence of NDM. The ing). 8. Organizational goals and norms
coalescence of NDM as a field of study (as opposed to decision making in a vac-
was marked by the publication of Decision uum). (p. 5 )
403
404 the cambridge handbook of expertise and expert performance
The primary theme of the Klein et al. alternatives; there is an input-output orien-
(1993 ) volume was how traditional decision- tation, a comprehensive information search,
making research did not provide much that and a formal development of an abstract,
was relevant to understanding professional context-free model amenable to quantita-
judgment and decisions in the field. To a tive testing (Lipshitz, 2001). The formal-
great extent, the initial motivation for the empiricist paradigm focused on behavioral
NDM paradigm, as sometimes happens in testing of formal models, not on understand-
scientific change, involved a reaction against ing cognitive processes.
the currently dominant paradigm, which in The rationalist paradigm overtook the
this case was the paradigm of Judgment classic decision making paradigm, but
& Decision Making (JDM) and the various retained the essential characteristics in terms
models and approaches that fell under that of normative (prescriptive) models as the
banner (Hoffman, 1995 , p. 3 3 ). standard for evaluating decision quality. The
A new view of decision making and judg- rationalist paradigm emphasized the con-
ment arose because NDM researchers were cept of errors due to bias in unaided deci-
describing what they were seeing in stress- sion making. The rationalist paradigm also
ful field conditions as professionals, includ- asserted that discrepancies in performance
ing experts, applied their judgment. Early are the fault of the decision maker, not the
NDM research discovered that expert pro- model. Earlier efforts in classic decision mak-
fessional judgment was largely based on ing had sought to modify the model when
a process in which experts expend effort discrepancies were found (Cohen, 1993 ).
on situation assessment (figuring out the What was the motivation for the change in
nature of the problem), then evaluate sin- the paradigm? According to Kahneman and
gle options through mental simulation, and Tversky (1982a), the goal of the rationalist
then arrive at a satisfactory answer or action. paradigm was to make the research more
Qualitative analysis of professional judg- cognitive, that is, to expose intellectual lim-
ment under stress in field conditions was a itations, reveal psychological processes, and
major departure from traditional research map the use of intuition.
in decision making. Mainstream models of Lipshitz, Klein, Orasanu, and Salas (2001)
decision making at that time were not much asserted that the naturalistic or NDM
help in understanding these new findings. paradigm places the expert at the cen-
Specifically, utility theory mandates a pro- ter of the research focus. Comprehensive
cedure for good judgment in which the choice was replaced by matching, input-
decision maker lays out all of the alternative output orientation was replaced by process
decision paths and iteratively evaluates each orientation, and context-free formal model-
for costs and benefits. The successful profes- ing was replaced by context-bound informal
sional judgment being observed in the field modeling . . . researchers within the NDM
was radically different from the prescriptive framework embarked on the construction
processes of good decision making found of descriptive models of proficient decision
in the literature at the time. makers in natural contexts without rely-
To put NDM in perspective, Cohen ing on normative choice models as starting
(1993 ) provided a discussion of three basic points (Lipshitz, 2001, p. 3 3 3 ).
paradigms of decision-making research: the The NDM paradigm liberates the study
formal-empiricist paradigm (also known of proficiency from reliance on traditional,
as classic decision making), the rational- decision-analytic theories of problem-
ist paradigm, and the naturalistic paradigm. solving and the related normative pre-
The formal-empiricist paradigm lasted until scriptions concerning proper decision-
the late 1960s. Its essential characteristic making methods and training techniques
was that it was a normative (prescriptive) (Hoffman, 1995 , p. 14). The essentials of
model of rational behavior. The decision NDM have remained the same since its
maker chooses among concurrently available emergence: proficient decision makers,
professional judgments and naturalistic decision making 405
situation-action matching decision rules, ideal for a given domain. Situated cogni-
context-bound informal modeling, process tion research also examines the nature of
orientation, and empirical-based prescrip- expertise, in works such as those of Suchman
tion (Lipshitz et al., 2001). Lipshitz et al. (1987) and Lave and Wenger (1991), which
point out, however, that the emphasis in along with Winograd and Flores (1986)
NDM has also changed in certain respects emphasize the contextual aspects of learn-
over time. In early work, the emphasis ing and performance and have provided the
was on shaping features of the field setting foundation for a plethora of studies in edu-
contexts, and expertise was a secondary cation and computer applications.
factor. By the time of the second NDM When NDM researchers study experts,
conference in 1994, the decision maker was they mean individuals who have achieved
the distinguishing focus of NDM. Expertise exceptional skill in one particular domain,
had become a core of NDM research. and the NDM research has focused on
understanding the process of developing
and applying that expertise in context.
Researchers have defined a number of vari-
Expertise from the NDM Perspective ables of expertise that are important to
NDM researchers (adapted from Phillips,
Expertise has been investigated from a num- Klein, & Sieck, 2004):
ber of theoretical perspectives, such as in
the JDM literature. JDM has studied heuris- Perceptual skills Experts have the abil-
tics and biases in an attempt to demonstrate ity to make fine discriminations. They
biases among expert populations (Tversky see more in a situation than a novice by
& Kahneman, 1982). Experts do not per- noticing cues a novice does not (see, for
form well when the tasks (e.g., probabil- example, Klein & Hoffman, 1993 ).
ity juggling tasks) depart from the experts
Mental models Experts have rich inter-
familiar tasks, and recommendations for
nal representations of how things work
debiasing do not fit within the confines of
in their domain of practice (Rouse &
their job. The JDM approach differs from
Morris, 1986). These mental models
an NDM approach, which emphasizes the
allow them to learn and to understand
domain specificity of expertise. Many stud-
situations more rapidly (Ross, Battaglia,
ies have found that when experts perform
Phillips, Domeshek, & Lussier, 2003 ).
in their domain in their natural context, bias
is alleviated and experts yield good judg- Sense of typicality and associations
ments (e.g., Borstein, Emler, & Chapman, Experts have a large repertoire of pat-
1999; Cohen, 1993 ; Keren, 1987; Shanteau, terns. They recognize what is typical in
1989; Smith & Kida, 1991). a situation (Ericsson & Simon, 1993 )
The development of computer applica- and they recognize complex patterns.
tions generated a huge literature devoted They also recognize when things are
to capturing expertise. With the advent of not going as expected, that is, when
expert systems, hundreds of projects sought there is an anomaly or something is
to capture expertise and embed it in decision missing.
aids across many domains. Hoffman (1995 ) Routines Experts know how to get
cites a number of studies that elicited exper- things done (Anderson, 1983 ). They
tise to create decision aids (Boose, 1986; have a wide repertoire of tactics. They
Bramer, 1985 ; Coombs, Dawes, & Tver- dont just know about things; they
sky, 1970; Keller, 1987; Waterman, 1986; know how to do things.
Weiss & Kulikowski, 1984). This approach Declarative knowledge Experts know
views expertise differently from the NDM more facts and details and have
paradigm in that it seeks to capture an more tacit knowledge than novices
objective expertise model as if there is one do. Tacit knowledge is the operational
406 the cambridge handbook of expertise and expert performance
simulation to assess the worth of a particu- them with expectations . . . . serial option
lar COA. Satisficing also occurs in this step; evaluation model. Here, an option is gen-
all the expert needs is a COA that will work erated, and then either implemented or
to meet his goals and fit the constraints of rejected. If rejected, a second option is
the situation. If the first COA evaluated is considered, and so forth. This may be
described as a serial model of decision mak-
found wanting, the expert generates a sec-
ing, because although one or more options
ond and so on, evaluating each in turn but
are considered, only one option is examined
never comparing options against each other. at a time. (G. A. Klein et al., 1986, p. 17)
The Origin of RPD RPD has been the subject of much study
and refinement. Below we describe the key
The RPD Model was first expressed in studies conducted by Klein and his collabo-
a study of fireground commanders (fire- rators that contributed to the further devel-
ground comanders) (G. A. Klein et al., opment of the RPD Model. One of the
1986). Klein et al. wanted to understand how first replications, in the study of neonatal
decisions are really made. To do this, they intensive care nurses, was also a test of the
eschewed the less meaningful, less time- Critical Decision Method (CDM), a knowl-
pressured decision tasks of most laboratory edge elicitation technique that was devel-
studies and elected to study real decision oped in tandem with and in order to study
makers who performed in naturalistic set- the RPD Model. The CDM has been found
tings. They selected fireground commanders to be reliable and effective in uncovering
as their research domain because fireground expert knowledge and reasoning strategies
decision making was a task with high time (Hoffman, Crandall, & Shadbolt, 1998; see
pressure, high stakes, and a wide range of Hoffman & Lintern, Chapter 12).
contexts or situations. In order to understand
how fireground commanders made their
decisions, they interviewed the commanders Neonatal Intensive Care Nurses
about nonroutine or command-challenging Crandall and Calderwood (1989) studied
past experiences (critical incidents). Since highly experienced neonatal intensive care
the interview was structured and not in real unit (NICU) nurses in order to test the
time, this was a quasi-naturalistic approach. emerging CDM. They used the method
They collected detailed retrospections on 3 2 to elicit challenging incidents. Each nurse
critical incidents from 29 interviews with 26 supplied two incidents, one to assess the
officers. The officers interviewed were lieu- CDM and one to examine the content
tenants, captains, and chiefs, with an aver- of expert knowledge. This second incident
age of 23 .2 and no less than 12 years of always involved identification of sepsis (a
experience. Data analysis found that approx- serious condition in neonates), before the
imately 80% of the commanders decisions infant became critically ill. When Crandall
were recognition-based. In fact, some inter- and Calderwood analyzed the data, they
viewees said that they never made deci- found that experienced nurses relied heav-
sions at all. Those statements were the ily on the recognition of perceptual cues and
catalyst for the development of the RPD patterns of cues to identify the early stages
Model: of sepsis. Many of these cues did not occur
in the medical literature, and many of the
Their ability to handle decision points
medical literatures cues did not appear in
appeared to depend on their skill at rec-
the nurses accounts. In addition, 16 ran-
ognizing situations as typical, as instances
of general prototypes that they had devel- domly chosen cases were assessed for judg-
oped through experience. The prototypes ment processes. Of the 16 cases, ten were
provided them with an understanding of coded as recognition-primed, two as a blend
the causal dynamics at work, suggested of recognitional and analytic strategies, and
promising courses of action, and provided two as analytic.
408 the cambridge handbook of expertise and expert performance
The MDMP and MCPP are highly proce- other options on a set of abstract evaluation
duralized and cumbersome to employ. Mil- dimensions as the MDMP or MCPP would
itary commanders and some officers who have them do. These findings have been gen-
have also become NDM researchers have eralized to a wide variety of tasks and spe-
reported that they tried hard to make these cialties and have been replicated a number
processes work in the field, but the process of times (Fallesen & Pounds, 2001; G. Klein,
kept failing them. One way that military 1998; G. Klein, 2004; R. Pascual & Hender-
decision makers deal with the prescribed son, 1997).
process when they go into the field is to Further, G. Klein et al. (1995 ) showed
abbreviate it. However, there is little guid- that skilled decision makers generated a good
ance on how to achieve this abbreviation, COA as the first one they considered. John-
since the current U.S. Army Field Manual son and Raab (in press) have recently repli-
(U.S. Army, 1997) includes general sugges- cated this finding and extended it, showing
tions but no complete abbreviated process. that when skilled decision makers aban-
Research into the cognition of decision doned their initial COA in favor of one they
making provides an understanding of why generated subsequently, the quality of that
and how the formal, analytic process must subsequent COA was significantly lower
be modified in field practice. G. Klein (1997) than their initial COA. Johnston, Driskell,
and Lipshitz (1993 ) have shown that expe- and Salas (1997) showed that recognitional
rienced decision makers working in their processes, now sometimes referred to as
domain do not analyze a situation in terms of intuitive decision making, resulted in higher
its components or generate a series of differ- performance than analytical processes.
ent courses of action as is prescribed in mil- These findings call into question the very
itary decision making. Instead, experts use rationale of the MDMP and MCPP, which
their domain knowledge to recognize the sit- attempt to ensure good decision making by
uation or aspects of the situation and retrieve having planners generate three COAs and
a plausible, basic COA almost immediately evaluate them analytically in order to find
(G. A. Klein, 1987; G. A. Klein et al., 1986). the best one possible. Officers with more
The RPD Model, which was developed expertise often had to work around the pre-
as a result of command and control work scribed process to achieve their objectives.
sponsored by the U.S. Army (G. A. Klein, The staff who generate three COAs for their
1987), helps us understand why a prescribed commander to review have often reported
analytic process doesnt help military deci- anecdotally that there is one main COA they
sion makers in the field. The RPD Model are working on and two that are more like
describes how decision makers can come straw men just to satisfy the process. In
up with a plausible COA as the first one reality, the commander and staff naturally
they consider. Their knowledge base, train- produce a good basic COA after an initial
ing, and experience generally render them assessment and work to make sure it is sat-
able to satisfactorily assess a situation, even isfactory for the purpose. An optimal plan is
if it is not exactly the same as previous just not necessary and may, in fact, be coun-
situations encountered. When the typical terproductive if we are to believe the exper-
aspects of a situation are recognized, a plau- tise of General George S. Patton, who said:
sible COA usually comes to mind. That A good plan, violently executed now, is bet-
initial COA is based on the experts recog- ter than a perfect plan next week.
nition of aspects or patterns in the situa- Based on findings from research on the
tion, and at times the entire situation. That RPD Model and on several studies of mili-
recognition brings to mind associated, typ- tary planning exercises, Schmitt and Klein
ical actions that are likely to work in that (1999) developed the Recognitional Plan-
circumstance. Experienced decision makers ning Model (RPM). The purpose was to
then assess that course of action by mentally codify the informal and intuitive planning
wargaming it, rather than contrasting it to strategies already used in the field by skilled
professional judgments and naturalistic decision making 411
planning teams observed in the Army and Ross et al. (2004) also compared the
the Marine Corps in order to support the RPM with the description of an abbrevi-
natural cognitive process of the experienced ated MDMP provided in the U.S. Army
commander rather than force him to use Field Manual FM 1015 (U.S. Army, 1997).
a procedure that disrupts the power of his Although the RPM and abbreviated MDMP
expertise. both rely on a single COA, this is seen as
Rather than trying to replace the MDMP, a highly degraded strategy for the abbrevi-
Schmitt and Klein sought to codify the way ated MDMP, and the commander and staff
planners actually work. As a result, the RPM are assumed to have followed all of the
does not feel awkward or unnatural to plan- MDMP steps, although some may be per-
ners. Rather, a typical comment is were formed automatically. Thus, for the abbre-
already doing this. And that was exactly the viated MDMP, using a single COA is the last
intent: to codify the existing and effective resort, whereas for the RPM it is the first
practices and give the military a set of proce- resort. If more time becomes available, the
dures that reflect their best practices as these abbreviated MDMP would restore some of
have evolved over decades. The RPM strat- the steps that were skipped or slighted. In
egy is for the commander to identify his pre- the RPM, additional time would be used to
ferred COA so the staff can work on detail- do more wargaming or to enable subordinate
ing and improving it. Of course, for a strong units to increase their preparation.
COA the commander must bring experience The RPM has stimulated interest in the
to the mission to achieve good situational military ever since it was first described.
understanding specifically in relation to his On their own initiative, individual U.S.
goals. As the officer role-playing the com- Army and Marine battalion commanders
mander in one experiment put it, The RPM have experimented with the RPM and
is built around the early identification of a found it useful. The British military has
base COA thats improved over time (Ross been conducting experiments with the RPM
et al., 2003 , p. 5 ). (Blendell, Molloy, Catchpole, & Henderson,
Comparison of the MDMP with the RPM in preparation; R. G. Pascual, Blendell, Mol-
yields several key distinctions (Ross, Klein, loy, Catchpole, & Henderson, in prepara-
Thunholm, Schmitt, & Baxter, 2004; Ross tion), so far demonstrating its face validity.
et al., 2002). First, the rationales behind The most stringent research on the RPM
the two models are completely different (an was performed by Thunholm (in press),
analytic rationale called multi-attribute util- who contrasted performance for division-
ity analysis for MDMP versus a recognition- level planning groups in the Swedish Army
primed decision rationale for RPM). Second, who used either a variant of the RPM or
the research support for the basic assump- the Swedish Armys version of the MDMP.
tion of the MDMP that comparing sev- Thunholm found that the RPM permitted
eral COAs will result in a superior COA an increase in tempo of about 20% and
compared to a simplified process is weak. plans that were of equivalent or higher qual-
Third, the RPM is designed to build on expe- ity than those generated using the MDMP.
rience and expertise whereas MDMP tries Plans were scored blindly by independent
to use analytical procedures that can actu- raters. Thunholm also observed that the
ally prevent or hamper experienced plan- RPM plans tended to be somewhat bolder
ners from using their ability to quickly assess and better adapted to situational demands
a situation and come up with a plausi- than those emerging from the MDMP,
ble COA. Fourth, time pressure degrades which tended to be more constrained by
the MDMP whereas the RPM is specifically an over-compliance with current doctrinal
adapted to time-constrained planning. The templates. Currently, the Swedish Army
MDMP is rarely fully implemented in the has adopted its variant of the RPM in the
field, whereas the RPM describes a natural draft of the new field manual and the
strategy. National Defence College provides training
412 the cambridge handbook of expertise and expert performance
Decision-Centered Design process not only Much of the early work in NDM con-
supports more usable designs for expert trasted traditional decision-making research
use, but allows expertise to develop and the paradigm underlying NDM. The
naturally. debate has matured, but still continues as
The process of Decision-Centered the quantitatively oriented decision research
Design involves (1) domain familiarization, challenges the qualitatively oriented NDM
(2) knowledge elicitation, (3 ) data analysis, research to meet more rigorous standards.
(4) knowledge representation, (5 ) system Even within the community, researchers
development, and (6) system evaluation. assert that NDM needs to focus on devel-
The core of the process is step (2), knowl- oping theory built on sound findings, tools,
edge elicitation to determine what experts and principles based on more empirical stud-
know, think, and do in performing their ies using more rigorous (i.e., experimental)
tasks. Decision-Centered Design yields methodology (Lipshitz et al., 2001). Recom-
a clear understanding of the cognitive mendations have been made to balance qual-
challenges involved in the task of interest, itative field studies with traditional experi-
including the requirements and the context mental work and controlled observation. It is
of the tasks, and uses information from difficult for qualitative researchers to accept
experts about those challenges to create the strictures of the quantitative paradigm.
systems that support cognitive performance The limitations of quantitative research
(Stanard, Uehara, & Hutton, 2003 ). often create blinders. It is certainly beyond
the scope of this chapter to address the dif-
ferences between qualitative and quantita-
tive research rigor. But that argument needs
Future Research to be addressed in the literature, and the pos-
sibility of NDM being more rigorous with-
The future of NDM research and applica- out becoming pseudo-quantitative should
tion is wide open. Research is becoming be investigated. An appropriate application
broader in the cognitive processes under of mixed qualitative-quantitative research
study, making the term decision making designs may be an optimal solution.
seem limited as a title for this commu- Training applications based on the NDM
nity of practice. Current research focuses perspective have the advantage and strength
on many of the cognitive elements that are of potentially moving the trainee much more
found in the execution of expertise, such as quickly along the path toward expertise.
the use of mental models, managing uncer- Training is a multimillion-dollar business in
tainty, and sensemaking, or the process of the commercial and military sectors, and the
assessing situations over time (see Hoffman, pursuit of physical fidelity is expensive. To
2006). These cognitive elements are being understand what truly makes training a suit-
referred to as macrocognition (which also able surrogate experience is an important
includes cognitive elements such as identifi- research goal both for creating the resource
cation of leverage points, mental simulation, of expertise and using our training resource
adaptation or replanning, and maintaining dollars wisely.
common ground) to distinguish them from We have come to believe that the key
the type of microcognitive processes (such issue for developing effective training is
as attention and short-term memory) stud- not physical fidelity, which seems centered
ied in the laboratory and also to maintain around the engineer or researchers view of
the emphasis on naturalistic field research development, but the establishment of what
(G. Klein, 2004). In addition, NDM practi- we call cognitive authenticity (Ross, Halter-
tioners confront many issues that need fur- man, Pierce, & Ross, 1998; Ross & Pierce,
ther study to refine and build on existing 2000). By this term we mean the inclu-
research. sion of the features that an expert would
professional judgments and naturalistic decision making 415
Chi, M. T. H., Glaser, R., & Farr, M. J. (Eds.). Glaser, R. (1996). Changing the agency for
(1988). The nature of expertise. Mahwah, NJ: learning: Acquiring expert performance. In
Lawrence Erlbaum Associates. K. A. Ericsson (Ed.), The road to excellence
Cohen, M. S. (1993 ). Three paradigms for viewing (pp. 3 03 3 11). Mahwah, NJ: Erlbaum.
decision biases. In G. A. Klein, J. Orasanu, R. Hoffman, R. R. (1995 ). A review of naturalistic
Calderwood, & C. E. Zsambok (Eds.), Decision decision making research on the critical decision
making in action: Models and methods (pp. 3 6 method of knowledge elicitation and the recog-
5 0). Norwood, NJ: Ablex. nition priming model of decision-making, with a
Cohen, M. S., Freeman, J. T., & Thompson, focus on implications for military proficiency (Pre-
B. B. (1997). Training the naturalistic deci- pared for the University of Nottingham under
sion maker. In C. E. Zsambok & G. Klein sponsorship of the Defense Research Agency-
(Eds.), Naturalistic decision making (pp. 25 7 Farnborough and Epistemics, Ltd.).
268). Mahwah, NJ: Lawrence Erlbaum Asso- Hoffman, R. R. (Ed.). (2006). Expertise out of
ciates. context: Proceedings of the Sixth International
Coombs, C. H., Dawes, R. M., & Tversky, A. Conference on Naturalistic Decision Making.
(1970). Mathematical psychology. Englewood Mahwah, NJ: Lawrence Erlbaum Associates.
Cliffs, NJ: Prentice Hall. Hoffman, R. R., Crandall, B. W., & Shadbolt,
Crandall, B., & Calderwood, R. (1989). Clinical N. R. (1998). Use of the critical decision
assessment skills of experienced neonatal intensive method to elicit expert knowledge: A case
care nurses (Contract 1 R43 NR0191101 for The study in cognitive task analysis methodology.
National Center for Nursing, NIH). Fairborn, Human Factors, 40, 25 4276.
OH: Klein Associates Inc. Hutton, R. J. B., Miller, T. E., & Thordsen, M. L.
Crandall, B. W., Kyne, M., Militello, L., & Klein, (2003 ). Decision-centered design: Leveraging
G. A. (1992). Describing expertise in one-on-one cognitive task analysis in design. In E. Holl-
instruction (Contract MDA903 -91-C-005 8 for nagel (Ed.), Handbook of cognitive task design
U.S. Army Research Institute, Alexandria, VA). (pp. 3 83 416). Mahwah, NJ: Erlbaum.
Fairborn, OH: Klein Associates Inc. Jacobson, M. J., & Spiro, R. J. (1994). A frame-
Endsley, M. R. (1995 ). Toward a theory of sit- work for the contextual analysis of technology-
uation awareness in dynamic systems. Human based learning environments. Journal of Com-
Factors, 3 7, 3 264. puting in Higher Education, 5 , 3 3 2.
Ericsson, K. A. (1996). The acquisition of expert Johnston, J., Driskell, J. E., & Salas, E. (1997). Vig-
performance: An introduction to some of the ilant and hypervigilant decision making. Jour-
issues. In K. A. Ericsson (Ed.), The road to nal of Applied Psychology, 82 , 614622.
excellence: The acquisition of expert performance Kahneman, D., & Tversky, A. (1972). Subjective
in the arts and sciences, sports, and games probability: A judgment of representativeness.
(pp. 15 0). Mahwah, NJ: Lawrence Erlbaum Cognitive Psychology, 3 , 43 045 4.
Associates. Kahneman, D., & Tversky, A. (1982a). On the
Ericsson, K. A., & Charness, N. (1994). Expert study of statistical intuitions. Cognition, 11, 123
performance: Its structure and acquisition. 141.
American Psychologist, 49, 725 747. Kahneman, D., & Tversky, A. (1982b). On the
Ericsson, K. A., & Simon, H. A. (1993 ). Protocol study of statistical intuitions. In D. Kahneman,
analysis: Verbal reports as data (2nd ed.). Cam- P. Slovic, & A. Tversky (Eds.), Judgment under
bridge, MA: MIT Press. uncertainty: Heuristics and biases. Cambridge,
Fallesen, J. J., & Pounds, J. (2001). Identifying MA: Cambridge University Press.
and testing a naturalistic approach for cog- Keller, R. (1987). Expert systems technology: Devel-
nitive skills training. In E. Salas & G. Klein opment and applications. Englewood Cliffs, NJ:
(Eds.), Linking expertise and naturalistic deci- Prentice Hall.
sion making. Mahwah, NJ: Lawrence Erlbaum Keren, G. (1987). Facing uncertainty in the game
Associates. of bridge: A calibration study. Organizational
Flin, R., Slaven, G., & Stewart, K. (1996). Emer- Behavior and Human Decision Processes, 3 9,
gency decision making in the offshore oil 98114.
and gas industry. Human Factors, 3 8, 262 Klein, G. (1997). The current status of the nat-
277. uralistic decision making framework. In R.
professional judgments and naturalistic decision making 417
Flin, E. Salas, M. E. Strub, & L. Martin Kobus, D. A., Proctor, S., Bank, T. E., & Hol-
(Eds.), Decision making under stress: Emerging ste, S. T. (2000). Decision-making in a dynamic
themes and applications (pp. 1128). Aldershot: environment: The effects of experience and infor-
Ashgate. mation uncertainty (No. 183 2). Space and Naval
Klein, G. (1998). Sources of power: How people Warfare Systems Center.
make decisions. Cambridge, MA: MIT Press. Larkin, J. H. (1983 ). The role of problem rep-
Klein, G. (2004). The power of intuition. New resentation in physics. In D. Gentner & A. L.
York: A Currency Book/Doubleday. Stevens (Eds.), Mental models. Mahwah, NJ:
Klein, G., & Wolf, S. (1998). The role of leverage Lawrence Erlbaum Associates.
points in option generation. IEEE Transactions Lave, J., & Wenger, E. (1991). Situation learning-
on Systems, Man and Cybernetics: Applications legitimate peripheral participation. Cambridge,
and Reviews, 2 8, 15 7160. UK: Cambridge University Press.
Klein, G., Wolf, S., Militello, L., & Zsambok, C. E. Lipshitz, R. (1993 ). Converging themes in the
(1995 ). Characteristics of skilled option gen- study of decision making in realistic settings.
eration in chess. Organizational Behavior and In G. A. Klein, J. Orasanu, R. Calderwood,
Human Decision Processes, 62 , 63 69. & C. E. Zsambok (Eds.), Decision making
Klein, G. A. (1987). Applications of analogical in action: Models and methods (pp. 103 13 7).
reasoning. Journal of Metaphor and Symbolic Norwood, NJ: Ablex.
Activity, 2 , 201218. Lipshitz, R. (2001). Puzzle seeking and model-
Klein, G. A. (1989). Recognition-primed deci- building on the fire ground: A discussion of
sions. In W. B. Rouse (Ed.), Advances in man- Karl Weicks keynote address. In E. Salas & G.
machine systems research (Vol. 5 , pp. 4792). Klein (Eds.), Expertise and naturalistic decision
Greenwich, CT: JAI Press, Inc. [also published making (pp. 3 3 73 45 ). Mahwah, NJ: Lawrence
as DTIC No. ADA24065 9]. Erlbaum Associates.
Klein, G. A., & Brezovic, C. P. (1986). Design Lipshitz, R., Klein, G., Orasanu, J., & Salas, E.
engineers and the design process: Decision (2001). Focus article: Taking stock of naturalis-
strategies and human factors literature. Proceed- tic decision making. Journal of Behavioral Deci-
ings of the Human Factors and Ergonomics Soci- sion Making, 14, 3 3 13 5 2.
ety 3 0th Annual Meeting, 2 , 771775 . Lipshitz, R., & Strauss, O. (1997). Coping with
Klein, G. A., Calderwood, R., & Clinton-Cirocco, uncertainty: A naturalistic decision making
A. (1986). Rapid decision making on the fire- analysis. Organizational Behavior and Human
ground. Proceedings of the Human Factors and Decision Processes, 66, 149163 .
Ergonomics Society 3 0th Annual Meeting, 1, Orasanu, J., & Connolly, T. (1993 ). The rein-
5 765 80. vention of decision making. In G. A. Klein,
Klein, G. A., & Crandall, B. W. (1995 ). The role J. Orasanu, R. Calderwood, & C. E. Zsam-
of mental simulation in naturalistic decision bok (Eds.), Decision making in action: Mod-
making. In P. Hancock, J. Flach, J. Caird, & K. els and methods (pp. 3 20). Norwood, NJ:
Vicente (Eds.), Local applications of the ecolog- Ablex.
ical approach to human-machine systems (Vol. Pascual, R., & Henderson, S. (1997). Evidence of
2, pp. 3 243 5 8). Mahwah, NJ: Lawrence Erl- naturalistic decision making in military com-
baum Associates. mand and control. In C. E. Zsambok & G. Klein
Klein, G. A., & Hoffman, R. (1993 ). Seeing (Eds.), Naturalistic decision making (pp. 217
the invisible: Perceptual/cognitive aspects of 226). Mahwah, NJ: Lawrence Erlbaum Asso-
expertise. In M. Rabinowitz (Ed.), Cognitive ciates.
science foundations of instruction (pp. 203 226). Pascual, R. G., Blendell, C., Molloy, J. J., Catch-
Mahwah, NJ: Lawrence Erlbaum Associates. pole, L. J., & Henderson, S. M. (in prepa-
Klein, G. A., Orasanu, J., Calderwood, R., & ration). An investigation of alternative com-
Zsambok, C. E. (Eds.). (1993 ). Decision mak- mand planning processes (Defense Education
ing in action: Models and methods. Norwood, and Research Agency paper prepared for the
NJ: Ablex. United Kingdom Ministry of Defence).
Klinger, D. W., & Klein, G. (1999). Emergency Phillips, J. K., Klein, G., & Sieck, W. R. (2004).
response organizations: An accident waiting to Expertise in judgment and decision making: A
happen. Ergonomics In Design, 7, 2025 . case for training intuitive decision skills. In D. J.
418 the cambridge handbook of expertise and expert performance
Koehler & N. Harvey (Eds.), Blackwell hand- Schmidt, R. A., & Wulf, G. (1997). Con-
book of judgment & decision making (pp. 297 tinuous concurrent feedback degrades skill
3 15 ). Victoria, Australia: Blackwell Publishing. learning: Implications for training and simula-
Pliske, R. M., McCloskey, M. J., & Klein, G. tion. Human Factors, 3 9, 5 095 25 .
(2001). Decision skills training: Facilitating Schmitt, J. F., & Klein, G. (1996). Fighting in
learning from experience. In E. Salas & G. Klein the fog: Dealing with battlefield uncertainty.
(Eds.), Linking expertise and naturalistic decision Marine Corps Gazette, 80, 6269.
making (pp. 3 75 3 ). Mahwah, NJ: Lawrence Schmitt, J. F., & Klein, G. (1999). How we plan.
Erlbaum Associates. Marine Corps Gazette, 83 (10), 1826.
Randel, J. M., Pugh, H. L., & Reed, S. K. (1996). Schuler, J. W. (1994). AN/SLQ-3 2 Operator
Differences in expert and novice situation Training: Development of Performance Assess-
awareness in naturalistic decision making. Inter- ment Instrument (NPRDC-TN-94-13 ). San
national Journal of Human-Computer Studies, Diego, CA: Navy Personnel Research and
45 , 5 795 97. Development Center.
Ross, K. G., Battaglia, D. A., Phillips, J. K., Serfaty, D., MacMillan, J., Entin, E. E., & Entin,
Domeshek, E. A., & Lussier, J. W. (2003 ). E. B. (1997). The decision-making expertise
Mental models underlying tactical thinking of battle commanders. In C. E. Zsambok &
skills. In Proceedings of the Interservice/Industry G. Klein (Eds.), Naturalistic decision making
Training, Simulation, and Education Conference (pp. 23 3 246). Mahwah, NJ: Lawrence Erl-
2 003 . Arlington, VA: National Training Sys- baum Associates.
tems Association.
Shanteau, J. (1989). Cognitive heuristics and
Ross, K. G., Halterman, J. A., Pierce, L. G., biases in behavioral auditing: Review, com-
& Ross, W. A. (1998). Preparing for the ments, and observations. Accounting Organiza-
instructional technology gap: A constructivist tions and Society, 14, 165 177.
approach. Paper presented at the 1998 Interser-
vice/Industry Training, Simulation, and Educa- Simon, H. A. (195 7). Models of man: Social
tion Conference, Orlando, FL. and rational. New York: John Wiley & Sons,
Inc.
Ross, K. G., Klein, G., Thunholm, P., Schmitt,
J. F., & Baxter, H. C. (2004). The recognition- Smith, J. F., & Kida, T. (1991). Heuristics and
primed decision model. Military Review, biases: Expertise and task realism in auditing.
LXXIV, 610. Psychological Bulletin, 109, 472489.
Ross, K. G., Lussier, J. W., & Klein, G. (2005 ). Stanard, T., Uehara, M. A., & Hutton, R. J. B.
From recognition-primed decision making to (2003 ). Decision-centered design: Principles &
decision skills training. In S. Haberstroh & processes (Year one Report prepared through
T. Betsch (Eds.), Routines of decision making participation in the Advanced Decision Archi-
(pp. 3 273 41). Mahwah, NJ: Lawrence Erl- tectures Collaborative Technology Alliance
baum Associates. [Contract #02TA1-SP3 -RT2] sponsored by the
U.S. Army Research Laboratory under Coop-
Ross, K. G., McHugh, A., Moon, B. M., Klein,
erative Agreement DAAD19-01-2-0009). Fair-
G., Armstrong, A. A., & Rall, E. (2002). High-
born, OH: Klein Associates Inc.
level cognitive processes in field research (Year
One Final Report under Contract 02TA2-SP1- Suchman, L. A. (1987). Plans and situated actions:
RT1 for U.S. Army Research Laboratory under The problem of human-machine communication.
Cooperative Agreement DAAD19-01-2-0009). Cambridge, England: Cambridge University
Fairborn, OH: Klein Associates Inc. Press.
Ross, K. G., & Pierce, L. G. (2000). Cognitive Thunholm, P. (in press). Military decision making
engineering of training for adaptive battlefield under time-pressure: To evaluate or not to eval-
thinking. In IEA 14th Triennial Congress and uate three options before the decision is made?
HFES 44th Annual Meeting (Vol. 2, pp. 410 Organizational Behavior and Human Decision
413 ). Santa Monica, CA: Human Factors and Processes.
Grgononics Society. Tversky, A., & Kahneman, D. (1982). Causal
Rouse, W. B., & Morris, N. M. (1986). On look- schemas in judgments under uncertainty. In D.
ing into the black box: Prospects and limits on Kahneman, P. Slovic, & A. Tversky (Eds.), Judg-
the search for mental models. Psychological Bul- ment under uncertainty: Heuristics and biases.
letin, 100, 3 493 63 . Cambridge, MA: Cambridge University Press.
professional judgments and naturalistic decision making 419
U.S. Army. (2003 ). Field Manual 6-0, Mis- Winograd, T., & Flores, F. (1986). Understanding
sion command: The command and con- computers and cognition: A new foundation for
trol of army forces. Washington, DC: U.S. design. Norwood, NJ: Ablex.
Army. Zsambok, C. E. (1997). Naturalistic decision
U.S. Army. (1997). Field Manual 101-5 , Staff making: Where are we now? In C. E. Zsambok
organization and operations. Washington, DC: & G. Klein (Eds.), Naturalistic decision making
Department of the Army. (pp. 3 16). Mahwah, NJ: Erlbaum.
U.S. Marine Corps. (2001). MCWP 5 -1, Marine Zsambok, C. E., Klein, G., Kyne, M. M., &
Corps planning process. Quantico, VA: U.S. Klinger, D. W. (1992). Advanced team deci-
Marine Corps. sion making: A developmental model (Con-
Waterman, D. A. (1986). A guide to expert systems. tract MDA903 -90-C-0117 for U.S. Army
Reading, MA: Addison-Wesley. Research Institute for the Behavioral and
Weiss, S., & Kulikowski, C. (1984). A practical Social Sciences). Fairborn, OH: Klein Asso-
guide to designing expert systems. Totowa, NJ: ciates Inc. [also published as DTIC No. 25 95 12,
Rowman and Allanheld. http://www.dtic.mil].
C H A P T E R 24
Decision-Making Expertise
Picture basketball coach Don Smith, who wasted trying to make a silk purse from a
desperately needs at least a couple of out- sows ear?
standing free throw shooters. Coach Smiths Coach Smiths situation is by no means
team has lost game after game in the wan- unique. In some form or another, it is repli-
ing seconds. That is because opposing teams cated in countless circumstances where the
easily regain possession of the ball by com- focus is on making decisions rather than
mitting intentional fouls since they know shooting free throws. Thus, instead of Coach
that Smiths players make few of their free Smith, our protagonist might be the man-
throws. Coach Smith is pondering how to aging partner in a medical practice, where
recruit a new player who has already demon- the need is for new staff physicians who
strated his proficiency at the line. He is also make excellent treatment decisions for their
trying to figure out how to train a current patients. It could be the head of a market-
player to elevate his free throw performance ing division who is recruiting brand man-
to the level of an expert. Related to this sec- agers, ones who will make choices that
ond approach, Coach Smith also has what improve market share. It could be a police
some would call scientific concerns. He department official who wants to assure that
wonders why none of his players is already an the officers retained after their probationary
excellent free throw shooter. What exactly periods make wise choices in the heat of the
is involved in foul shooting expertise? How moment. In all these cases, the fundamental
does one normally develop the skill? Why is questions are similar:
it that some players acquire the ability but
r Identification: How can we find people
others do not? Are there certain constitu-
tional factors, physical or psychological, that with high degrees of decision-making
limit a players potential at the free throw expertise?
line? If so, what are they? And how can r Explanation: How can we explain the
one detect them, so that resources are not presence or absence of such expertise?
42 1
42 2 the cambridge handbook of expertise and expert performance
r Development: How can we develop this in particular. Consider, for instance, the sug-
kind of expertise? gestion that scholarship on expertise in prob-
lem solving has been more productive than
Our original aims for this chapter were
that on decision making (e.g., Orasanu &
to survey, analyze, and summarize what the
Connally, 1993 ). What can one make of such
decision-making literature has to offer for
propositions if, as is often the case, decision
questions like those above, with a special
making and problem solving are not clearly
emphasis on the explanation question. A
distinguished. (To anticipate, we view deci-
portion of the literature does indeed directly
sion making as a special case of problem
address expertise, but it is small. Moreover,
solving.)
our examination of the literature forced
To avoid such difficulties, it would help
us to conclude that most research explic-
if investigators adhered consistently to a
itly focused on decision-making expertise
convention about what decisions are. Such
is incapable of providing answers that sat-
a convention actually exists, although, as
isfy all major constituencies. Put another
implied in the previous discussion, it is
way, scholarship on this variety of expertise
not universally followed. The convention is
appears to be much less well developed than
embodied in the following definition, which
one might like or expect, a perception that
we assume for the remainder of the chapter
is not unique to us (cf. Phillips, Klein, &
(Yates, 2003 , p. 24):
Sieck, 2004).
We therefore modified our aims. Thus, A decision is a commitment to a course of
in the chapter, we first offer our analysis action that is intended to yield results that
of significant impediments to progress in are satisfying for specified individuals.
research on decision-making expertise. We
There are several major decision varieties:
then describe a means of overcoming those
choices, which entail the selection of a
impediments, an overarching way of think-
subset from a larger collection of discrete
ing about decision making that makes clearer
alternatives (e.g., a class of ten new gradu-
what decision-making expertise must entail.
ate students from a pool of 100 applicants);
In the context of describing that perspec-
acceptances/rejections, which are special
tive, we interpret some of the few major
cases of choices in which only one specific
conclusions from earlier studies on decision-
option (e.g., potential marriage partner) is
making expertise that seem firmly defensi-
acknowledged and must be accepted or not;
ble. We also identify findings about aspects
evaluations, which are statements of worth
of decision behavior that have implications
that are backed up by commitments to act
for expertise, although this is usually unrec-
(e.g., a $3 10,000 binding bid on a house); and
ognized. The analysis also identifies specific
constructions, which are attempts to create
questions whose answers derived in future
ideal decision problem solutions given avail-
work should significantly accelerate progress
able resources (e.g., a departments budget
toward a deep understanding of decision-
or a plan for fighting a fire).
making expertise.
The present definition is a synthesis of
how the decision idea has been understood
implicitly in most scholarship on decision
Decisions making, for example, in psychology, educa-
tion, marketing, politics, operations, and the
It is obvious what a free throw is in a bas- military. Individual investigators are some-
ketball game. But things are murkier in the times inconsistent and imprecise in their use
case of decision making; there is not univer- of the word decision. But in our expe-
sal agreement about what constitutes a deci- rience, when pressed, they almost never
sion. The resulting fuzziness has been bur- claim that the definition here differs sig-
densome for decision scholarship generally nificantly from how they understand the
and for studies on decision-making expertise decision concept.
decision-making expertise 42 3
Several key features are packed into the example, when a physician is choosing
decision definition: a pain reliever to prescribe for a par-
r Action: Decisions ultimately are about ticular patient. (The decider is nearly
people doing things taking actions. always among the intended beneficiaries
Thus, when Jane Davis decides to pur- of a decision, since people rarely seek to
chase General Motors stock, the perti- decide contrary to their own interests.)
nent action is exchanging her money for The specification of beneficiaries is criti-
GM shares. cal, implicating what is arguably the sin-
r gle feature of decision problems that dis-
Commitment: Decisions need to be dis-
tinguishes them most sharply from more
tinguished from the actions they impli-
general problems differences among peo-
cate because, for instance, not all deci-
ple in the values they attach to decision
sions are eventually executed. Hence, we
results. The correct solution to an algebra
recognize that a decision has been made
problem is the correct solution for every-
as soon as there is a commitment to act in
body. But, whereas Joe Payne regards the
a particular way, as when Davis resolves
cobalt blue Phantom as the most beau-
to buy 100 GM shares.
r tiful car he has ever seen, Lew Walters
Intention: People cannot decide by acci- sees it as hideous and therefore cringes
dent, even though they sometimes decide at the mere idea of buying one. It is
haphazardly. That is, decision making is noteworthy that the implicit subjectiv-
intentional behavior; it has a purpose. ity represents a significant and challeng-
This is not to say, however, that unin- ing departure from most expertise schol-
tended consequences do not sometimes arship, which prizes unambiguous perfor-
(often?) occur or that deciders are always mance criteria.
fully aware of how they arrive at their
decisions, as when they decide according
to intuition.
r Satisfying results: Decision making is Decision Quality and
about achieving results that are experi- Decision-Making Expertise
enced as satisfying (e.g., with high util-
ity), not merely correct in some purely Even more than for the decision concept
logical sense. The intentional charac- itself, ambiguity about notions of decision
ter of decisions implicates their sought- quality and expertise is a major hindrance
after results (e.g., relief from arthritic to scholarship on decision-making expertise.
pain in making a choice of medication). Here we describe common views and their
But other results that people care about difficulties and propose an escape route.
are usually involved, too, ones recog-
nized as side effects (e.g., liver dam-
age from a medicine) or process costs The Satisfying-Results Perspective
(e.g., the time it takes to choose among It is plain for all to see when a basketball
alternative medications or the anxi- player is an expert free throw shooter. A
ety experienced when deliberating such successful shot is one that goes through the
choices). hoop. And an expert shooter is one who
r Specified individuals: People make deci- makes a high percentage of successful shots,
sions to serve the interests of particu- even when the game is in the balance and
lar individuals. Sometimes the targeted the tension is palpable. In principle, there
beneficiary is the decider alone, for is no reason that decision quality and, corre-
example, when one is dining solo and spondingly, decision-making expertise could
choosing between the Szechuan chicken not be conceptualized the same way. Thus,
and the egg foo young. But at other recalling that the aim of any decision is
times, the beneficiaries include others, for results that satisfy particular people, from a
42 4 the cambridge handbook of expertise and expert performance
whether the process used to arrive at a deci- key reason is statistical. It is nearly impos-
sion mimics the maximization of expected sible for people to defend statements like
utility: No principle other than maximiz- Decider As tendency for producing deci-
ing SEU deserves a moments consideration sions with satisfying results is better than
(p. 7). And in numerous laboratory studies Decider Bs on the basis of large samples of
of decision-making where the uncertainty observed cases. Instead, indications are that,
is not explicitly acknowledged (e.g., Payne, when expertise beliefs are at all based on
Bettman, & Johnson, 1988), the coherence observations of results, those beliefs rest on
standard is often taken to be equivalence to fewer observations than statistical principles
additive utility or value maximization. Such would demand. By default, then, the beliefs
expertise yardsticks are, incidentally, the must be driven by other considerations.
ones implicit in several well-known decision- An example: Extensive research has
analytic methodologies. That is, the promise shown that peoples judgment behavior
is that deciders (or, more importantly, their is susceptible to strong primacy effects
beneficiaries) would be well served by real- and that, under particular conditions, these
izing such standards (e.g., Morris, 1977). effects result from attention decrement
(e.g., Yates & Curley, 1986). Thus, people
are inclined to observe a small number of
Implications for Expertise Questions
cases, draw a conclusion on the basis of those
The satisfying results and the coherence per- cases, and then simply stop paying atten-
spectives are clearly different. This therefore tion to pertinent facts that present them-
poses a dilemma for addressing the exper- selves later. This phenomenon is plausibly
tise questions articulated at the outset of related to the one Tversky and Kahneman
this chapter, most obviously the identifica- (1971) dubbed the belief in the law of small
tion question. Suppose that one needs to numbers, which suggests that people feel
identify expert deciders, for example, to hire that attending to further cases beyond the
them or to study them. The alternative per- first few is unnecessary as well as burden-
spectives suggest two different ways of doing some. In the present instance, these mecha-
that, and there is every reason to expect nisms suggest that our presumptions about
that those approaches would sometimes lead a deciders expertise are dictated by the first
to different conclusions about who is actu- few decisions we happen to associate with
ally an expert. Which conclusions would that person. If those decisions turn out well,
be right? we apply the label expert. But if they turn
Interestingly, in terms of what happens out badly, the decider might well be called
in real life, this obvious question is actu- inept. And that opinion is cast in stone
ally moot. It is undoubtedly true that people since its underpinnings will not be revisited.
commonly believe that some deciders It is worth noting that these effects implicate
including professionals of various kinds the sense in which the outcome bias label is
are either more or less expert than others. sometimes justified. A process-control engi-
And those people almost certainly act on neer would never dismiss a single product
these beliefs, for example, in choosing to defect as irrelevant to conclusions about the
hire the services of individuals regarded as quality of the manufacturing process that
more expert than their peers. How do those produced it. In the same way, it is unrea-
beliefs arise and how are they justified? sonable to say that people should ignore
We are unaware of anyone pointedly try- the results of a single decision when trying
ing to appraise expertise in terms of coher- to infer the deciders expertise. The notion
ence. Nor have we seen full-fledged, legit- of bias enters the picture only when, in
imate attempts to assess decision-making violation of principles like the law of large
expertise according to satisfying results in numbers, people use nothing more than a
the kinds of complex, real-life situations for single case to draw definitive conclusions
which people typically seek expertise. A about long-term tendencies (Caplan, Posner,
42 6 the cambridge handbook of expertise and expert performance
& Cheney, 1991, provide a compelling pants inferred consulting candidates exper-
illustration). tise at probability judgment according to
There are indications that beliefs about their inclination to report extreme, that is,
decision-making expertise are in large mea- highly confident, opinions. They also found
sure social constructions, too. In research that many individuals demand that potential
on decision-making expertise, some investi- consultants be able to convincingly explain
gators consider as experts individuals who and justify their assessments; prospective
have extensive experience, training, and/or principals give short shrift to empirical evi-
professional and social standing, for exam- dence of consultants accuracy.
ple, practicing clinicians (Meehl, 195 9) or One worrisome possibility that the pre-
senior military personnel (Tolcott, Marvin, & sent analysis suggests is that our assumptions
Lehner, 1989). In part, this practice seems to about who is or is not a decision-making
reflect a tendency to infer decision-making expert might not be as good as we hope.
expertise from subject-matter expertise (This is not to say that they are, in fact, erro-
(e.g., the ability to recite standard psy- neous, only that we do not know.) To the
chiatric diagnostic categories). This is not extent that this is true, we are not enjoy-
unreasonable to the extent that subject- ing as many benefits of decision-making
matter expertise is necessary for decision- expertise as we might. There are trouble-
making expertise, which it no doubt often some research implications, too, concerning
is. For instance, it would be impossible for a the decision-making expertise explanation
layperson who knows nothing about the law and development questions. Some research
to consistently make decent legal decisions designs for approaching such questions are
on behalf of clients. Nevertheless, equating predicated on identifying and studying large
decision-making and subject-matter exper- numbers of documented expert and non-
tise effectively assumes that there is no expert deciders. Those designs are precluded
such thing as decision-making expertise per if certification is suspect, as it seems to be.
se. This also implies that studies of expert Prospects for those designs are also compro-
deciders in one arena (e.g., law) offer no mised by sample size restrictions. Studies of
insight for decision-making expertise else- decision-making expertise are conspicuous
where (e.g., medicine); generalizability is nil. for their small sample sizes, which causes
Other researchers operationally define some observers to not take the research seri-
expertise according to consensus among ously. Consider, for instance, the six empiri-
peers (Shanteau, 1992). Thus, experts are cal expertise studies reported in the impor-
people who already-acknowledged experts tant collection edited by Zsambok and Klein
say are experts. Of course, this begs the ques- (1997). The median number of participants
tion of how such self-sustaining impressions in those studies, ones considered experts and
of expertise arise initially. As suggested ear- otherwise, was only 20.5 . One likely reason
lier, these impressions plausibly are based for such small numbers is the difficulty of
at least partly on outcome observations, finding and persuading busy experts to par-
even if minimally. But they are also prob- ticipate in research. All in all, it is clear that a
ably affected by such validity-vulnerable good alternative to current perspectives and
factors as personal style, which charlatans practices is essential. We now sketch what
can easily fake. Shanteau (1992, p. 25 7) seems to be such an alternative.
put it nicely: In short, to be accepted as
an expert, it is necessary to act like one.
The Process-Decomposition Perspective
This involves things like exhibiting self-
confidence (bluster?) as well as good com- In the process-decomposition perspective,
munication and persuasion skills. This con- the overall process of making a decision
clusion agrees with the data of Yates, Price, is partitioned into elements. If each ele-
Lee, and Ramirez (1996). These investi- ment is executed well, this should con-
gators found that many of their partici- tribute significantly to the adequacy of the
decision-making expertise 42 7
resulting decision. (This is withstanding the some particular decision process element,
real likelihood of the kinds of interactions this is itself a partial explanation of that indi-
emphasized in complex systems research, viduals lack of overall decision-making pro-
e.g., Sterman, 2002.) Ideally, there should ficiency. It also points to very specific things
be empirical evidence that, statistically, suc- one can do to improve expertise, in the
cessful execution of an element does indeed spirit of the deliberate-practice approach
increase the odds of decisions with satisfy- (Ericsson, Krampe, & Tesch-Romer, 1993 ).
ing outcomes. The process-decomposition
perspective shares common ground with
both the coherence and the satisfying-results Decision Processes as Cardinal
perspectives. The standard coherence per- Decision Issue Resolution
spective focuses on just a couple of ele-
ments in normal decision processes. In that Decision-related writings in myriad fields
sense, it can be seen as a limited, special (e.g., psychology, health care, management,
case of the process-decomposition perspec- finance, engineering, law, operations, anthro-
tive. No one claims that coherence harms pology, counseling, politics, and marketing)
the odds of good decision outcomes. Nev- as well as analyses of hundreds of incidents
ertheless, there have never been convinc- indicate that a limited number of funda-
ing demonstrations that coherence is suffi- mental questions arise repeatedly in real-
cient to increase those odds substantially, life decision problems. That is why the term
which is what everyone wants decisions cardinal decision issues is used to describe
to do. Like the satisfying-results perspec- them (see Yates, 2003 , for a more complete
tive, the process-decomposition perspective treatment of the cardinal decision issue per-
insists on efficacy demonstrations, even if spective). In some instances, the issues go
they are only indirect or logical rather than unrecognized by deciders themselves and
empirical. It differs most sharply from the thus are resolved by default, by whatever
former point of view in its practicality. As happens to be natural in the given situa-
we have seen, it is exceedingly difficult tion. For instance, as we shall see, one car-
to appraise and study individuals overall dinal issue concerns detecting that there is
decision-making expertise. But it is often a decision problem in the first place. Sup-
feasible to isolate and examine how people pose, say, driver James Lawson never senses
deal with specific process elements. growing deterioration in his cars transmis-
Decision processes can be decomposed sion. Then he naturally makes no deliberate
into component processes in numerous dif- decision about how to deal with the impend-
ferent ways. But the particular decompo- ing disaster, and one day his car simply
sition to which we turn next has several refuses to move. Regardless of their recog-
attractions. Perhaps the most compelling, nition, the cardinal issues are indeed some-
for present purposes, is that it is com- how settled during the course of any decision
prehensive. That is, it appears to encom- episode. It therefore proves useful to charac-
pass a great many of the activities impli- terize decision processes themselves as the
cated in virtually any decision situation. means by which the various cardinal issues are
Thus, the following key inference makes addressed for the decision problem at hand.
sense: If a decider were to perform well on There are ten cardinal issues, which are
each element represented in the decompo- numbered for easy reference: 1) need, 2)
sition, we would be justified in expecting mode, 3 ) investment, 4) options, 5 ) pos-
true decision-making expertise, even in the sibilities, 6) judgment, 7) value, 8) trade-
satisfying-results sense. Moreover, studies of offs, 9) acceptability, and 10) implementa-
such dimensional performance are directly tion. Detailed statements and illustrations of
informative about explanation and develop- the issues appear below. But at this stage it
ment questions regarding decision-making is useful to put them in the context of the
expertise. After all, if a decider is poor at big picture of Figure 24.1, where the issues
42 8 the cambridge handbook of expertise and expert performance
Beneficiary Satisfaction
Other
Decision Contributors
are listed in the Decision Processes box. A ing matters such as the resources assigned
tour of that big picture is useful. to the decision task. Ordinarily, deciders
Figure 24.1 mimics the kind of chart have reached their decisions (saying, for
often sketched in root-cause analyses (see example, So this is what we are going to
Robitaille, 2004). It starts with Beneficiary do) once they have resolved the trade-
Satisfaction since that is the point of offs issue. In the aftermath, often before
decision-making. We are ultimately inter- anything else occurs, the acceptability and
ested in understanding (and sometimes implementation issues take center stage, for
influencing) contributors that, in cascade example, how various other parties feel
fashion, feed into that construct. The anal- about how the decision was made. It is
ysis acknowledges that, although decisions important to acknowledge that the stated
affect peoples satisfaction with the results order is indeed rough. In real-life decision
that ensue after a decision is made, so do episodes, deciders often tentatively resolve
Other Contributors, factors unrelated to issues but then revisit them later after con-
the decision. The decision is depicted as fol- fronting other issues (hence the double-
lowing from how the decider deals with each headed arrows). Suppose, for example, that
of the cardinal issues, Decision Processes as apartment seeker Linda Mathers, in address-
Cardinal Issue Resolution. ing the options issue, identifies two possi-
Observe that the issues are grouped ble places to rent. Later, in dealing with
into categories that roughly correspond to the tradeoffs issue, she realizes that pick-
when the issues arise in the typical deci- ing Apartment A over Apartment B requires
sion episode. At the core of the overall sacrifices that are too painful, and vice versa.
decision process are the options, possibili- She therefore returns to the options issue,
ties, judgment, value, and tradeoffs issues, attempting to find a third option that obvi-
matters that have received the most atten- ates such sacrifices.
tion in decision research. But before those The bottom portion of Figure 24.1
issues arise, usually the decider attends to acknowledges that a deciders decision pro-
several preliminaries, the resolution of the cesses derive from a host of other, sometimes
need, mode, and investment issues, concern- deep, contributors that are important to
decision-making expertise 42 9
understand in their own right, for exam- that pessimists (vs. optimists) exhibited
ple, unique personal experiences, constitu- especially rapid vigilance declines. Results
tional factors such as inherited dispositions such as these suggest a side of expertise that
or abilities, training, and culturally transmit- is seldom discussed in any context per-
ted local customs. The dotted arrow lead- sonality. That is, experts might perform par-
ing from Beneficiary Satisfaction to these ticular tasks exceptionally well not (solely)
contributors highlights the expectation that because of cognitive capabilities but (also)
decision results particularly bad results because of temperament.
are likely to instigate adjustments in how a How well a decider maintains vigilance
decider does things, in a feedback loop. when actually trying is one thing. Whether
Here we sketch and briefly illustrate the the decider happens to notice good rea-
issues, with each articulated in the voice sons for initiating decision-making efforts
of deciders confronted with it. We also in the midst of performing other tasks is
identify and discuss the implications of key quite another, and arguably a more signif-
expertise-significant ideas and findings asso- icant one. Consider, for instance, noticing
ciated with those issues. In addition, we subtle hints of customer taste changes in
identify specific and important addressable the normal course of commerce. We are
gaps in current understanding. unaware of research that has addressed dif-
Issue 1 Need: Why are we (not) decid- ferences in deciders inclinations for doing
ing anything at all? The need issue is about this, which should be an important aspect
whether and how decision problems are rec- of decision-making expertise. There is only
ognized at the outset. Suppose that a prob- a hint from the ancillary literature on prob-
lem goes unacknowledged. Then necessarily lem finding and creativity that suggests the
there is no decision of any kind, and default potential fruitfulness of decision-expertise
consequences occur, a particular variety of scholarship on the problem. Rostan (1994)
blindsiding. We already saw one illustra- found that a significant distinction between
tion, the one in which driver Lawson was acclaimed and merely competent artists
oblivious to the deteriorating transmission in and scientists was that, in an unstructured
his car. task, the former spent significantly more of
Vigilance is one tack that deciders some- their time finding problems.
times take to address the need issue. This Issue 2 Mode: Who (or what) will make
is the sustained attempt to monitor for sig- this decision, and how will they approach that
nals of threats and opportunities that war- task? Decision modes are qualitatively dis-
rant serious efforts to decide what, if any- tinct approaches to deciding. The concern is
thing, to do about them, as in the control with who decides and how those individuals
room of a power plant. Over the past 5 0+ address the various other cardinal issues that
years, researchers have learned much about will be resolved in arriving at a decision. The
human vigilance, for example, how surpris- mode issue per se is about how it is deter-
ingly rapidly it degrades (Howe, Warm, & mined which of those modes is applied to
Dember, 1995 ). Almost all vigilance research the decision problem at hand.
has examined the effects on performance A major part of the who mode dis-
of various aspects of the situation, not why tinction concerns whether and to whom
one person might more successfully moni- people with the authority for making a
tor signals than others. But some recent find- decision defer at least part of their work
ings have identified various correlates of vigi- to others, agents or consultants. Agents
lance expertise. For instance, Rose, Murphy, are parties (including possibly devices such
Byard, and Nikzad (2002) found that false as computers) who are granted the power
alarms in a vigilance task were associated to make a decision however they see fit,
with the Big Five dimensions of extraversion for example, a subordinate asked to make
and conscientiousness. And Helton, Dem- all hires in a particular unit. Consultants
ber, Warm, and Matthews (1999) observed are parties who provide input to a decision
430 the cambridge handbook of expertise and expert performance
procedure but who do not make the final (and not necessarily consciously), then a par-
decision, for example, a computer program ticular action spontaneously emerges, with
that searches for and ranks job candidates negligible effort, control, and self-insight.
but does not actually determine who gets a This mode is common in high-speed, fre-
job. No decider can possibly have the sub- quently repeated situations, for example,
stantive expertise required to make every operating a vehicle or playing a sport such as
kind of decision ideally. Thus, it seems appar- basketball. Intuitive decision making is said
ent that deciders who are good at deter- to occur when the decider selects a course of
mining when there are significant advan- action on the basis of considerations (feel-
tages of using agents and consultants for ings) that the decider cannot articulate, per-
particular purposes have an edge. That is, haps because of fundamental weaknesses in
their expertise at the who side of the the linkages between cognitive and affec-
mode issue should pay off. There is con- tive psychological processes (Sadler-Smith &
sistent evidence that people generally have Shefy, 2004).
extraordinary confidence in the quality of How decision modes differ most obvi-
their own decisions and that this confidence ously in their efficiency. Deliberately pon-
increases under particular conditions, for dering what to do, as in analytic decision-
example, when they explain their decisions making, obviously takes far more time
(e.g., Sieck & Yates, 1997). To the extent that than deciding via any of the other modes.
this amounts to overconfidence, this suggests And, to the degree that a good decision
that people generally defer to agents and maker is efficient, the application of vari-
consultants less often than they should. But ous how modes is a significant dimension
if our earlier conclusions about the difficulty of decision-making expertise. Moreover, to
of assessing true decision-making expertise the extent that expertise tends to (imper-
are correct, figuring out to whom one should fectly) increase with experience, we should
defer is a formidable task. We are unaware expect an association between mode use
of research that has even tried to identify, and measures of decision-making expertise
explain, or develop expertise on decision beyond efficiency. Such associations have
work deferral. indeed been inferred. Probably the best-
The how side of the mode issue is known example is embodied in research
about the nuts and bolts of how deciders related to Kleins (1993 ) recognition-primed
(as well as consultants) carry out their work. decision (RPD) model (see Ross et al., Chap-
Several broad categories of possibilities are ter 23 ). Such work has repeatedly demon-
recognized, including (but not limited to) strated that rule-based decision making is
analytic, rule-based, automatic, and intu- extremely common among highly expe-
itive decision-making. The analytic label rienced, recognized experts such as fire-
describes the broad category in which the ground commanders. Unfortunately, there is
decider effortfully seeks to figure out what not yet a body of research that systemat-
action makes sense to do in a given deci- ically addresses identification, explanation,
sion situation. That mode is presumed to be and development questions as they apply to
necessary when the situation is unfamiliar. individual differences in expertise at shifting
Most traditional decision scholarship (e.g., among how modes.
utility theory) addresses this mode. In rule- Issue 3 Investment: What kinds and
based decision-making, the decider applies amounts of resources will be invested in the
rules of the following form: If Conditions process of making this decision? The invest-
C hold, then take Action A. Such rules are ment issue is about how and how well
ubiquitous in business and medicine (e.g., it is determined whether the investment
If the patient displays signs and symptoms of resources in the process of making a
X, Y, and Z, then treat as follows: . . .). given decision is extensive or minimal, and
Automatic decision making is such that, if in which particular resource categories. As
the decider recognizes particular conditions implied in the mode issue discussion, if
decision-making expertise 431
two deciders are otherwise equivalent, it increasing the number of alternatives con-
clearly makes sense to say that the one sidered, a false assumption that many peo-
who routinely expends fewer resources is ple make. The ideal option consideration
more expert. As also suggested in that pre- set for a given problem consists of only a
vious discussion, one means for minimizing single alternative the best one. Recogniz-
decision-making costs as yet unstudied sys- ing others is wasteful, requiring the decider
tematically is deliberate, strategic reliance to expend precious resources vetting alter-
on cheap modes such as rule-based and natives that ultimately will (or should) be
intuitive decision making. rejected. But recent work has demonstrated
Conceivably, deciders who require lit- that the deliberation of large consideration
tle time to decide do so because decision sets can do more than simply waste time. It
chores demand little of the mental resources can also exact significant psychological costs,
they have available, compared to what oth- such as turmoil over the possibility of fail-
ers bring to the table. Such a possibil- ing to pick the very best alternative (e.g.,
ity has been given little credence in view Schwartz et al., 2002).
of results like those of Chase and Simon There has been surprisingly little research
(1973 ), who compared the memory per- that directly implicates expertise with
formance of chess masters and less-expert respect to the options issue. One of the
players. There was no indication that mas- few decision articles on the subject was by
ters had generally superior pure memory Dougherty and Hunter (2003 ), who found
skills. Instead, the inference drawn from such that working-memory capacity was reliably
work is that experts are especially good at related to the number of alternatives individ-
organizing or chunking pertinent informa- uals could recall in a laboratory decision task.
tion, which can facilitate effective memory The creativity literature is the major non-
performance in appropriate circumstances. decision literature most directly relevant to
Results like these as well as more recent option-issue expertise since, by definition,
research on the long-term working-memory highly creative individuals are especially
concept (Ericsson & Kintsch, 1995 ) suggest good at crafting new and useful alter-
a minimal role for capacity variations in natives. Assuming their validity, creativity
explaining expertise, decision making and measures should be helpful in identifying
otherwise. But other research has been more this particular variety of decision-making
favorable to the capacity view, finding reli- expertise. So should measures of person-
able connections between performance on ality characteristics, such as openness to
decision-related tasks, on the one hand, experience, that have been shown to be
and measures of working-memory capacity associated with creativity (e.g., Wolfradt &
(Dougherty & Hunter, 2003 ) and general Pretz, 2001). Unfortunately, there is little
cognitive facility (West & Stanovich, 2003 ), in the literature immediately informative
on the other. on the explanation and development exper-
Issue 4 Options: What are the different tise questions, for example, about where
actions we could potentially take to deal with creativity originates, how creative people
this problem we have? If a decider never rec- go about their work, or how their prac-
ognizes a certain alternative, then clearly the tices can be cultivated in others. Never-
decider cannot commit to that alternative. theless, some possibilities seem implicit in
Thus, the leaders of Ace Products cannot other research. For instance, numerous stud-
choose to market or not market Prod- ies have shown that creativity is enhanced
uct C if that prospect never reaches their by positive emotion (Fredrickson, 1998).
attention. The options issue is about how This suggests that creative individuals might
people come to see, if not create, prospec- achieve some of their advantage by strategi-
tive solutions to their decision problems. It cally arranging their work environments to
is essential to recognize that expert man- exploit phenomena such as the creativity-
agement of the options issue is not about emotion linkage.
432 the cambridge handbook of expertise and expert performance
Issue 5 Possibilities: What are the var- Issue 6 Judgment: Which of the things
ious things that could potentially happen if we that they care about actually would happen if
took that action things they care about? This we took that action? Usually, this issue log-
issue implicates another form of blindsid- ically and temporally follows the possibili-
ing, well illustrated by an actual case involv- ties issue. After the decider recognizes (accu-
ing a popular assistant school superinten- rately or otherwise) that some decision-
dent. Only by accident was she discovered relevant event can happen, the next task is
to have falsified her credentials (including judging whether it would happen. A judg-
degrees) when she was hired years earlier. ment is an opinion as to what was, is, or will
The officials responsible for her appoint- be the state of some decision-relevant aspect of
ment were stunned, saying things like: It the world. Accordingly, a judgment is accu-
simply never occurred to us that she might rate to the degree that there is a close cor-
not be who she said she was. More gener- respondence between that judgment and the
ally, the possibilities issue is about recogniz- actual pertinent state. Clearly, judgments and
ing outcomes of prospective actions that are decisions are distinct, but equally clearly,
capable of occurring and which, if they did, judgment accuracy imposes an upper bound
would matter greatly. Note that the concern on decision quality.1 Many considerations go
is not with whether those outcomes would into crafting a bid on a business contract. But
occur, only whether they could. Thus, in the the eventual deal cannot possibly be fully
superintendent case, if the mere possibil- satisfactory if it is predicated on false expec-
ity of false credentials had come to mind, tations about what the other party would do
the deciders might well have made routine during the contracts life.
checks that would have precluded an embar- Much research that has been described
rassing appointment. as concerning decision making has really
Clearly, a truly expert decider would been about judgment. Consider, for instance,
be good at anticipating possibilities. Yet, Shanteaus (1992) remarks on the good
the possibilities issue as such has gone decision performance of weather fore-
largely untouched in decision research. casters, who actually make predictions of
Nevertheless, work framed in other ways weather events rather than make decisions
arguably has implications for it. This predicated on such predictions. Making the
includes research demonstrating peoples judgment-versus-decision distinction is use-
difficulty even imagining the sometimes ful for several reasons, including the fact that
bizarre behaviors of common real-life non- decision problems demand the consideration
linear systems (Sterman, 2002). It also of much more than judgment problems. At
includes scholarship on stress. Numerous the same time, however, as we shall see, judg-
studies have shown that acute stress nar- ments of various kinds assume roles in the
rows the scope of attention. Thus, stress resolution of several other cardinal decision
should induce the neglect of possibilities. issues beyond the judgment issue per se.
But a decider who is stress resistant should Partly because its adequacy is often easy
be immune to such effects. This suggests to evaluate and analyze, judgment is the
that stress resistance should facilitate the aspect of decision behavior that has been
handling of the possibilities issue and, thus, studied more extensively than any other.
decision-making expertise, particularly in And the most consistent expertise conclu-
high-stress environments. There have been sion has been this: Subject matter experts
no direct tests of this proposition yet. But often exhibit much worse judgment accuracy
there are data consistent with it, such as indi- than most people expect. Meehl (195 9), for
cations that individuals who self-select and instance, found that clinical psychologists
succeed in becoming air traffic controllers diagnoses based on MMPI scores were less
tend to be more stress resistant than other accurate than those derived from simple lin-
people (see the review by Yates, Klatzky, & ear combinations of those scores. There have
Young, 1995 ). been many similar demonstrations since
decision-making expertise 433
then. One of the most recent was Onkal- judgments. Johnsons (1988) protocol data
Atay, Yates, Simga-Mugan, and Oztins provide direct evidence for this expecta-
(2003 ) finding that, although professionals tion. Substantive procedures are a power-
were generally more accurate at predict- ful but risky tool. The character of peoples
ing foreign exchange rates than amateurs, in personal theories about how events (e.g.,
many individual comparisons, the opposite criminal acts) literally occur has little if any
was true. place for uncertainty, which should encour-
Several contributors to such occurrences age overconfidence. Such theories also tend
have been hypothesized and, in some cases, to be convoluted, since they seek to account
documented. Some amount to artifacts of for every local nuance in the abundant and
the testing conditions, for example, as in interpretable information experts see. As
studies showing that auditors are minimally Camerer and Johnson (1991) argued, this fea-
susceptible to known biases when making ture should manifest itself in configurality
judgments in familiar professional contexts equivalent to interaction terms in statisti-
(Smith & Kida, 1991). Other contributors cal models, terms that tend to be unpredic-
are apparently real, including the fact that tive. This can actually harm the accuracy of
humans, unlike computers, necessarily can- experts judgments by making them noisy,
not perfectly reliably execute any intended laden with what amounts to random error, a
judgment policy (e.g., Dawes & Corrigan, phenomenon observed by Yates, McDaniel,
1974). An especially powerful contributor and Brown (1991) in securities forecasting.
seems to lie in the very character of human A significant challenge for future research
judgment. Yates (1990) observed that there is the development of human-device sys-
are two classes of judgment processes, for- tems that take effective account of the doc-
malistic and substantive. Formalistic proce- umented peculiarities as well as the com-
dures are similar to the application of rules plementary strengths of human experts and
such as those in probability theory or regres- judgment algorithms.
sion analysis. Significantly, such rules are Issue 7 Value: How much would they
indifferent to the content of judgment prob- really care positively or negatively if that
lems. Quite the opposite is true of sub- in fact happened? The value issue is a spe-
stantive procedures, which entail the per- cial case of the judgment issue, albeit an
son attempting to envision how nature exceptionally important special case. That
literally would (or would not) create the is because it centers on what makes deci-
event in question. Pennington and Hasties sion problems so distinctive and difficult
(1988) story model of juror decision mak- individual differences in what people like
ing is a good illustration. In that model, and dislike. In order for a decider to pur-
jurors judge the likelihood of a defendants sue actions that promise outcomes that
guilt by how easily evidence fits into a nar- the intended beneficiaries find satisfying
rative whereby the defendant committed the goal of any decision-making effort the
the crime in question. The mental simula- decider must know those persons tastes.
tion process in Kleins (1993 ) RPD model That is, an expert decider must be outstand-
is another illustrative substantive procedure. ing at making judgments for which the target
There are many indications in the liter- is how people feel about things. For instance,
ature that, as Yates (1990, pp. 209210) an expert designer, buyer, or casting director
argued, people resort to formalistic proce- must be excellent at anticipating how much
dures only when they cannot use substantive the typical potential customer would like or
ones, which are much more natural. dislike any given product.
By definition, subject-matter experts There is extraordinary interest in the
know more than other people about the value issue these days in decision research,
substance of a given area. This implies that under the rubric of hedonic forecasting.
they are also especially likely to apply sub- Specifically, the concern is with the consis-
stantive procedures when arriving at their tent errors people tend to make when trying
434 the cambridge handbook of expertise and expert performance
to predict their own feelings about future many finance tools prescribe how to trade off
occurrences, for example, winning a lottery, outcome value against time (Higgins, 1998).
suffering paraplegia, or getting promoted Behavioral decision research has been
(Loewenstein & Schakade, 1999). Unfortu- dominated by questions about deviations
nately, there has been much less interest and of peoples actual decision behavior from
systematic study of expertise in anticipating what is predicted or prescribed by rules
peoples values, particularly other peoples like the expected utility, additive utility, and
values. Nevertheless, work to date suggests discounting models. It is therefore perhaps
that subject-matter experts are not always as surprising that there is virtually no litera-
good at the value issue as one might expect ture on individual differences in adherence
or hope. Consider the study by Wilson to those standards. And, accordingly, there
et al. (1997), involving interns and attend- has been virtually no discussion of exper-
ing physicians caring for elderly patients. tise with respect to tradeoffs. This prob-
Only 4% of the interns knew their patients ably reflects an overly narrow conception
more than seven days, whereas almost half of how deciders tend to deal with trade-
the attending physicians had known their offs in real life. The dominant models just
patients more than six months. Neverthe- described presume a static, pick among
less, the attending physicians (despite their these stance by the decider. Yet, there is evi-
confidence) were no more accurate than the dence that a major tactic deciders use is, in
interns at predicting patients preferences effect, transforming tradeoffs problems into
for end-of-life care (e.g., lifelong tube feed- options problems (cf. Shafir, Simonson, &
ing). Determining why accuracy sometimes Tversky, 1993 ). Specifically, deciders some-
fails to improve with experience is an impor- times seek to avoid having to make an oner-
tant task on the research agenda. Part of the ous tradeoff altogether by finding or creating
problem is likely to reside in the lack of a new alternative that makes it unnecessary.
attention and feedback. Thus, instead of simply choosing between
Issue 8 Tradeoffs: All of our prospec- putting a new, untested product on the mar-
tive actions have both strengths and weak- ket and forgetting about it, a companys
nesses. So how should we make the tradeoffs deciders might seek other alternatives, such
that are required to settle on the action we as sharing the products risk with a partner.
will actually pursue? This issue concerns the Indeed, this approach seemed quite evident
fact that in virtually all decision situations, to Shapira (1995 ) in his study of risk tak-
deciders eventually arrive at this reality: ing among top executives. The executives
Every alternative has drawbacks. Investing appeared to believe that a major feature
a companys funds in a new, untested prod- of managerial expertise was the ability to
uct could provide unprecedented profits. On restructure risky alternatives such that they
the other hand, it could also push the com- were less hazardous to ones company.
pany into bankruptcy. To reach a decision, Issue 9 Acceptability: How can we get
the companys leaders must trade off the them to agree to this decision and this deci-
prospect of great profit against the risk of sion procedure? In perhaps most of the situa-
ruin. Mainstream decision scholarship his- tions where decision-making expertise mat-
torically has been preoccupied with the ters, the decider is not a free agent. Instead,
tradeoffs issue. Expected utility theory, the the decider must contend with the senti-
undisputed point of reference for the field, ments of many different people concerning
is at heart about the trading off of outcome two things, what is decided as well as how it
value and uncertainty (Yates, 1990, Chap- is decided, the province of the acceptability
ter 9). Multiattribute utility theory and its issue.
variants seek to explain or guide the trading Negotiations are the most familiar con-
off of values for some outcomes against those text where the acceptability issue figures
for others (Keeney & Raiffa, 1976). And the significantly. Unless both parties accept a
discounted cash-flow rules at the core of so given proposal, there is no deal. Many
decision-making expertise 435
deals fall through not only or even mainly as opposed to a single action that is virtually
because of the material aspects of a pro- synonymous with the decision itself (e.g.,
posal but because of the character of the purchasing a shirt). What sometimes occurs
deliberations (e.g., the parties perceptions is that the project proves to be difficult or
of one anothers integrity or respectful- even impossible to actually perform. A good
ness). Negotiation research and teaching are example is a company that agrees to gen-
vibrant enterprises these days (cf. Thomp- erous pension payments to its workers but
son & Leonardelli, 2004). This is likely ultimately discovers, years later, that it can-
tied to the immediacy, transparency, and not afford those payments.
obvious importance of negotiator effective- Disasters with respect to the implementa-
ness. As soon as a negotiation is over, each tion issue generally result from the mishan-
party seems clearly better or worse off dling of one or more of the other cardinal
than before, and apparently because of the issues. For instance, in the pension illus-
negotiators skills or lack thereof. We are tration, the deciders plausibly did a poor
unaware of extensive research on negotia- job with the judgment and options issues,
tor expertise per se. But there are cer- failing to anticipate increased employee
tainly popular assumptions that some nego- longevity or to create contract options
tiators (e.g., those entrusted with negotiating containing contingencies that take such
hostage releases) are more expert than oth- increases into account. There has been lit-
ers. And there is considerable work pointing tle systematic research aimed at under-
toward negotiator behaviors or characteris- standing how people, experts or other-
tics associated with their effectiveness and wise, address the implementation issue,
hence expertise, for example, their creativ- although this is changing (e.g., Dholakia
ity (Kurtzberg, 1998). & Bagozzi, 2002). But arguably relevant
The acceptability issue assumes signifi- scholarship framed differently has existed
cance beyond the realm of formal nego- for some time. This includes work high-
tiations. Striking illustrations have played lighting cultural differences in priorities
themselves out in courtrooms. American assigned to the issue. For instance, numerous
automakers have lost several major lawsuits Japanese businesses appear to have deliber-
because they mishandled the acceptability ately chosen to emphasize ease of decision
issue in design decisions. In one prominent implementation over decision-making speed
case, jurors were repelled by testimony that (cf. Liker, 2004).
the decision to limit costs on certain fea-
tures rested partly on a decision analysis
in which a dollar figure (based on actuar- Closing Remarks
ial records) was attached to lives that might
be lost in accidents linked to those fea- This review, built on the component anal-
tures. The jurors responded by forcing the ysis afforded by the cardinal issue perspec-
company to pay billions in punitive dam- tive, has cited numerous results suggesting
ages (Fix, 1999). Such examples dramati- specific behaviors that plausibly contribute
cally demonstrate the importance of exper- to decision-making expertise. Nevertheless,
tise with respect to the acceptability issue. the review has also made it painfully obvi-
Unfortunately, there appear to have been no ous how much remains unknown about such
systematic efforts to identify and understand behaviors. It also made clear how difficult
such expertise. it would be for any one person to con-
Issue 10 Implementation: Thats what sistently demonstrate excellence in resolv-
we decided to do. Now, how can we get it ing every one of the cardinal issues for any
done, or can we get it done, after all? This class of decision problem. This makes one
final cardinal issue arises in decision situa- suspect that true, across-the-board decision-
tions where the selected alternative entails making expertise is exceedingly rare. Yet, the
a nontrivial project that must be executed present analysis is optimistic in that it points
436 the cambridge handbook of expertise and expert performance
toward concrete and manageable questions Dougherty, M. R. P., & Hunter, J. E. (2003 ).
for future fundamental as well as develop- Hypothesis generation, probability judgment,
mental research, including efforts to create and individual differences in working memory
collectives and human-device systems that capacity. Acta Psychologica, 113 , 263 282.
decide with undeniable proficiency. Edwards, W., Kiss, I., Majone, G., & Toda, M.
(1984). What constitutes a good decision?
Acta Psychologica, 5 6, 5 27.
Acknowledgments Ericsson, K. A., & Kintsch, W. (1995 ). Long-
term working memory. Psychological Review,
102 , 211245 .
We are deeply grateful to Hannah Faye
Ericsson, K. A., Krampe, R. Th., & Tesch-Romer,
Chua, Georges Potworowski, Robert Hoff-
C. (1993 ). The role of deliberate practice in the
man, and Anders Ericsson for their invalu-
acquisition of expert performance. Psychologi-
able comments and suggestions. cal Review, 100, 3 63 406.
Fix, J. L. (1999, July 13 ). Memos key in $4.9-
billion verdict. Detroit Free Press, pp. 1Aff.
Footnote
Fredrickson, B. L. (1998). What good are posi-
tive emotions? Review of General Psychology, 2 ,
1. Confusingly, the term judgment is sometimes 3 003 19.
applied to evaluation decisions in the literature, Hammond, J. S., Keeney, R. L., & Raiffa, H.
as when authors distinguish judgment and (1999). Smart choices. Boston: Harvard Business
choice (e.g., Montgomery, Selart, Garling, School Press.
& Lindberg, 1994). The present use of the
Helton, W. S., Dember, W. N., Warm, J. S., &
expression is older, more firmly established,
Matthews, G. (1999). Optimism, pessimism,
and hence preferred.
and false failure feedback: Effects on vigilance
performance. Current Psychology: Developmen-
tal, Learning, Personality, Social, 18, 3 113 25 .
References Higgins, R. C. (1998). Analysis for financial man-
agement. Boston: Irwin McGraw-Hill.
Baron, J., & Hershey, J. C. (1988). Outcome bias Howe, S. R., Warm, J. S., & Dember, W. N. (1995 ).
in decision evaluation. Journal of Personality Meta-analysis of the sensitivity decrement
and Social Psychology, 5 4, 5 695 79. in vigilance. Psychological Bulletin, 117 , 23 0
Camerer, C. F., & Johnson, E. J. (1991). 249.
The process-performance paradox in expert Johnson, E. J. (1988). Expertise and decision
judgment: How can experts know so much and under uncertainty: Performance and process. In
predict so badly? In K. A. Ericsson & J. Smith M. T. H. Chi, R. Glaser, & M. J. Farr (Eds.), The
(Eds.), Toward a general theory of expertise: nature of expertise (pp. 209228). Mahwah, NJ:
Prospects and limits (pp. 195 217). New York: Erlbaum.
Cambridge University Press. Keeney, R. L., & Raiffa, H. (1976). Decisions with
Caplan, R. A., Posner, K. L., & Cheney, F. W. multiple objectives: Preferences and value trade-
(1991). Effect of outcome on physician judg- offs. New York: Wiley.
ments of appropriateness of care. Journal of the Klein, G. (1993 ). A recognition-primed decision
American Medical Association, 2 65 , 195 71960. (RPD) model of rapid decision making. In G.
Chase, W. G., & Simon, H. (1973 ). Perception in Klein, J. Orasanu, R. Calderwood, & C. Zsam-
chess. Cognitive Psychology, 4, 5 5 81. bok (Eds.), Decision making in action: Mod-
Dawes, R. M., & Corrigan, B. (1974). Linear mod- els and methods (pp. 13 8147). Norwood, NJ:
els in decision making. Psychological Bulletin, Ablex.
81, 95 106. Kurtzberg, T. R. (1998). Creative thinking, cogni-
Dholakia, U. M., & Bagozzi, R. P. (2002). Mus- tive aptitude, and integrative joint gain: A study
tering motivation to enact decisions: How deci- of negotiator creativity. Creativity Research
sion process characteristics influence goal real- Journal, 11, 283 293 .
ization. Journal of Behavioral Decision Making, Liker, J. K. (2004). The Toyota Way. New York:
15 , 167188. McGraw-Hill.
decision-making expertise 437
Loewenstein, G., & Schkade, D. (1999). Wouldnt Sadler-Smith, E., & Shefy, E. (2004). The intu-
it be nice? Predicting future feelings. In D. itive executive: Understanding and applying
Kahneman, E. Diener, & N. Schwarz (Eds.), gut feeling in decision-making. Academy of
Well-being: The foundations of hedonic psychol- Management Executive, 18, 7691.
ogy (pp. 85 105 ). New York: Russell Sage Foun- Schwartz, B., Ward, A., Monterosso, J.,
dation. Lyubomirsky, S., White, K., & Lehman,
Meehl, P. E. (195 9). A comparison of clinicians D. R. (2002). Maximizing versus satisficing:
with five statistical methods of identifying psy- Happiness is a matter of choice. Journal of
chotic MMPI profiles. Journal of Counseling Psy- Personality and Social Psychology, 83 , 11781197.
chology, 6, 102109. Shafir, E., Simonson, I., & Tversky, A. (1993 ).
Montgomery, H., Selart, M., Garling, T., Reason-based choice. Cognition, 49, 113 6.
& Lindberg, E. (1994). The judgment- Shanteau, J. (1992). Competence in experts:
choice discrepancy: Noncompatibility or The role of task characteristics. Organizational
restructuring? Journal of Behavioral Decision- Behavior and Human Decision Processes, 5 3 ,
Making, 7 , 45 5 5 . 25 2266.
Morris, W. T. (1977). Decision analysis. Colum- Shapira, Z. (1995 ). Risk taking: A managerial per-
bus, OH: Grid, Inc. spective. New York: Russell Sage Foundation.
Onkal-Atay, D., Yates, J. F., Simga-Mugan, C., Sieck, W., & Yates, J. F. (1997). Exposition effects
& Oztin, S. (2003 ). Professional vs. ama- on decision making: Choice and confidence in
teur judgment accuracy: The case of foreign choice. Organizational Behavior and Human
exchange rates. Organizational Behavior and Decision Processes, 70, 207219.
Human Decision Processes, 91, 169185 . Smith, J. F., & Kida, T. (1991). Heuristics and
Orasanu, J., & Connally, T. (1993 ). The reinven- biases: Expertise and task realism in auditing.
tion of decision making. In G. Klein, J. Orasanu, Psychological Bulletin, 109(3 ), 472489.
R. Calderwood, & C. Zsambok (Eds.), Decision Sterman, J. D. (2002). All models are wrong:
making in action: Models and methods (pp. 3 Reflections on becoming a systems scientist.
20). Norwood, NJ: Ablex. Systems Dynamics Review, 18, 5 015 3 1.
Payne, J. W., Bettman, J. R., & Johnson, E. J. Thompson, L., & Leonardelli, G. J. (2004,
(1988). Adaptive strategy selection in decision July/August). Why negotiation is the most
making. Journal of Experimental Psychology: popular business school course. Ivey Busi-
Learning, Memory, and Cognition, 14, 5 3 45 5 2. ness Journal, http://www.iveybusinessjournal.
Pennington, N., & Hastie, R. (1988). Explanation- com/.
based decision making: Effects of memory Tolcott, M. A., Marvin, F. F., & Lehner, P. E.
structure on judgment. Journal of Experimen- (1989). Expert decision making in evolving sit-
tal Psychology: Learning, Memory, & Cognition, uations. IEEE Transactions on Systems, Man,
14, 5 215 3 3 . and Cybernetics, 19, 606615 .
Phillips, J. K., Klein, G., & Sieck, W. R. (2004). Tversky, A., & Kahneman, D. (1971). The belief in
Expertise in judgment and decision making: A the law of small numbers. Psychological Bul-
case for training intuitive decision skills. In D. J. letin, 76, 105 110.
Koehler & N. Harvey (Eds.), Blackwell hand- West, R. F., & Stanovich, K. E. (2003 ). Is prob-
book of judgment and decision making (pp. 297 ability matching smart? Associations between
3 15 ). Malden, MA: Blackwell. probabilistic choices and cognitive ability.
Robitaille, D. (2004). Root cause analysis: Basic Memory & Cognition, 3 1, 243 25 1.
tools and techniques. Chico, CA: Paton Press. Wilson, I. B., Green, M. L., Goldman, L., Tsevat,
Rose, C. L., Murphy, L. B., Byard, L., & Nikzad, J., Cook, E. F., & Phillips, R. S. (1997). Is expe-
K. (2002). The role of the Big Five personality rience a good teacher? How interns and attend-
factors in vigilance performance and workload. ing physicians understand patients choices for
European Journal of Personality, 16, 185 200. end-of-life care. Medical Decision Making, 17 ,
Rostan, S. M. (1994). Problem finding, prob- 217227.
lem solving, and cognitive controls: An empir- Wolfradt, U., & Pretz, J. E. (2001). Individual dif-
ical investigation of critically acclaimed pro- ferences in creativity: Personality, story writing,
ductivity. Creativity Research Journal, 7 (2), and hobbies. European Journal of Personality, 15 ,
97110. 2973 10.
438 the cambridge handbook of expertise and expert performance
Yates, J. F. (1990). Judgment and decision making. and earnings: The hazards of nascent expertise.
Englewood Cliffs, NJ: Prentice Hall. Organizational Behavior and Human Decision
Yates, J. F. (2003 ). Decision management. San Processes, 49, 6079.
Francisco: Jossey-Bass. Yates, J. F., Price, P. C., Lee, J.-W., & Ramirez,
Yates, J. F., & Curley, S. P. (1986). Contingency J. (1996). Good probabilistic forecasters: The
judgment: Primacy effects and attention decre- consumers perspective. International Journal
ment. Acta Psychologica, 62 , 293 3 02. of Forecasting, 12 , 415 6.
Yates, J. F., Klatzky, R. L., & Young, C. A. (1995 ). Yates, J. F., Veinott, E. S., & Patalano, A. L.
Cognitive performance under stress. In R. S. (2003 ). Hard decisions, bad decisions: On deci-
Nickerson (Ed.), Emerging needs and oppor- sion quality and decision aiding. In S. L. Schnei-
tunities for human factors research (pp. 262 der & J. C. Shanteau (Eds.), Emerging perspec-
290). Washington, DC: National Academy tives on judgment and decision research (pp. 13
Press. 63 ). New York: Cambridge University Press.
Yates, J. F., McDaniel, L. S., & Brown, E. S. Zsambok, C. E., & Klein, G. (Eds.). (1997). Natu-
(1991). Probabilistic forecasts of stock prices ralistic decision making. Mahwah, NJ: Erlbaum.
C H A P T E R 25
The Making of a Dream Team: team, which had been seeded seventh in the
When Expert Teams Do Best 12-team tournament, went on to beat Fin-
land (42) for the gold medal.
The original use of the phrase Dream So what distinguishes these two teams
Team was in reference to the US basket- from other teams? Teamwork? Individ-
ball team that won the gold medal at the ual expertise? Both? What led the origi-
1992 Olympics in Barcelona. Team members nal Dream Team to dominate the 1992
included basketball greats (e.g., Michael Jor- Olympics? Conversely, what led the star
dan, Magic Johnson, and Larry Bird) as well Russian team to lose to a team they had dom-
as Charles Barkley and seven more NBA All- inated only a week before? The above exam-
Stars. This team of twelve proficient athletes ples illustrate that it takes more than a set of
who were at the top of their game seam- experts to make an expert team. Examples of
lessly blended their talents such that they teams composed of individuals highly skilled
dominated the Olympic competition, beat- in their task roles that have failed as teams,
ing their eight opponents by an average of sometimes with disastrous consequences are
44 points. not limited to sports, but abound within
On February 22, 1980 at the Olympic organizations (e.g., Hackman, 1990), indus-
Winter Games in Lake Placid a highly skilled try, aviation, and medicine as well as the
Russian hockey team, recognized as the military (e.g. Cannon-Bowers & Salas, 1998;
best hockey team in the world, lost 43 to Salas, Stagl, & Burke, 2004).
a young but skilled collegiate US hockey The lack of understanding that exists
team. The US victory over the undefeat- within organizations concerning the creation
able Russian team in the semi-finals, whom and management of expert teams poses a
they had just lost to 103 a week before in an challenge since, in recent decades, the cog-
exhibition match, put the US team in con- nitive complexity and demanding nature of
tention for the gold medal. The US hockey jobs has increased because of advances in
439
440 the cambridge handbook of expertise and expert performance
technology. This has caused organizations What Theories are Driving Expert
to increasingly adopt team-based systems Teams Research?
(Ilgen, 1994) in an effort to remain com-
petitive and handle the cognitive demands There have been several advances in the
placed on workers. In addition, the problem study of teams within the past 25 years (see
sets within organizations are often ambigu- Guzzo & Dickson, 1996; Kozlowski & Bell,
ous, unstructured and ill defined, causing 2003 ; Salas et al., 2004); however, the litera-
an increasing need for flexibility adap- ture often focuses on teams as a general topic
tive expert teams. Because expert teams, in and not expert teams specifically. By tak-
general, function in such dynamic, stressful, ing a multi-disciplinary approach and com-
and complex environments, research that bining advancements within the team liter-
examines teams situated in their natural ature with that on individual expertise, we
context has particular significance for under- can begin to understand, create, and manage
standing expert teams. Hence, this chap- adaptive expert team performance in com-
ter focuses primarily on studies of teams in plex environments.
complex environments functioning in such What has the literature told us so far?
areas as the military, business, aviation, and First, expert team members are able to com-
healthcare. bine their individual technical expertise and
The focus of this chapter is on current sci- coordinate their actions to achieve a com-
entific understanding of the performance of mon goal in such a manner that performance
expert teams what is it that these teams seems fluid; the team as a whole creates a
do, think, or feel that makes them expert. synergy greater than its parts (Salas et al.,
So, we present a brief review of the state- 2004). Second, expert teams need to pos-
of-the-art in the study of performance of sess routine expertise, that is, they need the
expert teams. We define an expert team as a ability to solve problems quickly and accu-
set of interdependent team members, each rately and understand problems in terms
of whom possesses unique and expert-level of principles and concepts (Chi, Feltovich,
knowledge, skills, and experience related & Glaser, 1981). Third, members must be
to task performance, and who adapt, coor- able to flexibly apply existing knowledge
dinate, and cooperate as a team, thereby structures such that when faced with a
producing sustainable and repeatable team novel situation, members can make predic-
functioning at superior or at least near- tions about system functioning and invent
optimal levels of performance. Expert teams new procedures based on these predictions
are primarily characterized by high levels of (Hatano & Inagaki, 1986). Fourth, expert
team and task outcomes, achieved via the teams seem to hold shared mental mod-
teams effective utilization of team mem- els of the task, the situation, their team-
ber task-related expertise and the mastery mates, and the equipment (Cannon-Bowers,
of team processes. To that end, this chap- Salas, & Converse, 1993 ; Orasanu & Salas,
ter addresses three questions. First, what are 1993 ), which promote implicit coordination.
the theories that are driving the research Finally, expert teams must possess adaptive
in the domain of expert teams? Second, expertise the ability to invent new pro-
what methods are being used to study expert cedures based on knowledge and to make
teams? Third, given these things, what do new predictions (Hatano & Inagaki, 1986).
we currently know about the performance Smith, Ford, and Kozlowski (1997) further
of expert teams? We hope that in briefly argue that the key to adaptive expertise is a
addressing these three questions we get a deep conceptual understanding of the target
glimpse at the making of a dream team domain such that declarative and procedural
what are the cognitions, behaviors, and atti- knowledge coalesce into strategic knowledge
tudes that we should strive for in high per- (i.e., why procedures are appropriate for cer-
forming teams. tain conditions). Hatano and Inagaki (1986)
the making of a dream team 441
argue that mindful processing and abstrac- Gully, & Salas, 1996). Taskwork skills are
tion are critical to the formation of adaptive those skills that members must understand
expertise. and acquire for actual task performance,
Given these characteristics, advances in whereas teamwork skills are the behavioral
theory that serve as drivers to understanding and attitudinal responses that members need
expert teams can be broken down into five to function effectively as part of an inter-
areas, those dealing with: (a) team effective- dependent team (Morgan et al., 1986). The
ness and teamwork, (b) team adaptability implication for the creation of expert teams
and decision making, (c) shared cognition, is that it is not sufficient that members
(d) team leadership, and (e) team affective be technical experts they must also be
states, such as collective efficacy and psycho- experts in the social interactions that lead to
logical safety. We briefly discuss these below. adaptive coordinated action (i.e., teamwork)
within the context of the technical expertise.
Figure 2 5.1. Team Adaptation Framework illustrating the relationship between input variables,
emergent states, and the multiple phases of the team adaptation cycle (adapted from Stagl, Burke,
Salas, & Pierce, 2006.)
A challenge in decision making within actions (e.g., Klimoski & Mohammad, 1994;
team environments occurs when team mem- Cooke et al., 2000; Mohammed & Dumville,
bers begin to experience stress and cant 2001), and (c) make accurate predictions not
easily diagnose the situation because of only about the world in which the team is
the performance-degrading effects of stress. operating but about the team functioning
Normally higher-status members are less that enables coordination (Rouse & Morris,
likely to take the advice of lower-status 1986). Shared cognition, in the form of com-
members. However, under stress three patible mental models, as well as mutual
things happen. First, higher-status mem- performance monitoring are necessary pre-
bers are more willing to accept input from cursors to effective team processes, such as
those with less expertise, but under these back-up behavior, because they form the
conditions low-status members generally foundation for decisions of when a team
arent vocal in their viewpoints (Driskell member must step in to provide back up,
& Salas, 1991). Second, attention tends who should step in, and what assistance is
to narrow, producing a form of tunnel needed.
vision (Salas, Cannon-Bowers, & Blickens-
derfer, 1993 ). Third, explicit communication
decreases as members become more focused team leadership
on their own respective roles (Kleinman The impact of leaders on individual, team,
& Serfaty, 1989). Despite these challenges and organizational effectiveness is substan-
expert teams are adaptive and able to main- tial (Zaccaro, Rittman, & Marks, 2001).
tain coordination levels and corresponding Researchers have increasingly taken a func-
effective decisions despite these conditions; tional perspective when examining team
they have behavioral and cognitive mecha- leadership (Fleishman, Mumford, Zaccaro,
nisms in place that allow them to maintain Levin, Korotkin, & Hein, 1991; Hackman,
high levels of performance. 2002; Zaccaro et al., 2001). From this per-
spective, leadership involves social problem
shared cognition solving that promotes coordinated, adaptive
Shared cognition has been used to refer team performance by facilitating goal defini-
to a number of related constructs (e.g., tion and attainment (Salas, Burke, & Stagl,
shared mental models, team situation aware- 2004, p. 3 43 ) and is composed of four classes
ness, common ground, team metacognition, of leader responses to social problems: infor-
transactive memory; Kelly, Badum, Salas, & mation search and structuring, information
Burke, 2005 ). Shared cognition has been use in problem solving, managing personnel
increasingly used as an explanation for how resources, and managing material resources
the members of expert teams are able to (Salas et al., 2004). Although theoretical
interact with one another and adapt com- work in this area is continuing and a large
munication and coordination patterns while leadership literature exists, research into the
under stress (e.g., Campbell & Kuncel, 2001; functional role of team leaders remains a glar-
Cannon-Bowers, Salas, & Converse, 1993 ; ing weakness (e.g. Salas et al., 2004). As the
Cooke, Salas, Cannon-Bowers, & Stout, complexity of the social problems faced by
2000; Ensley & Pearce, 2001; Entin & Serfaty, leaders and teams increases, so does the need
1999; Hinsz, Tindale, & Vollrath, 1997; for adaptation. Research into shared leader-
Klimoski & Mohammad, 1994; Orasanu, ship holds promise as a source for informing
1990). Shared cognition (in its various des- our understanding of the processes by which
ignations) has been argued to be the mech- team leadership can contribute to expert
anism that allows teams to: (a) coordinate team performance.
their action without explicit communica- Shared leadership is the transference
tion (Entin & Serfaty, 1999), (b) interpret of the leadership function among team
cues in a similar manner, make compatible members in order to take advantage of
decisions, and take coherent or convergent member strengths (e.g., knowledge, skills,
444 the cambridge handbook of expertise and expert performance
attitudes, perspectives, contacts, and time ual members feel at ease being themselves.
available) as dictated by either environmen- Using teams within a manufacturing com-
tal demands or the development stage of pany, Edmondson (1999) showed that high
the team (Burke, Fiore, & Salas, 2004, levels of psychological safety led teams to
p. 105 ). Pearce and Sims (2002) have shown view failure as a learning opportunity and to
that shared leadership can be more effective seek feedback from outside sources. Alter-
than traditional vertical leadership (i.e., a nately, low levels of psychological safety
rigid hierarchical authority structure). When led to an unquestioning acceptance of team
leadership is shared, the team can adapt goals for fear of reprisal from managers as
to situational demands by shifting leader- well as a disinclination to seek help. There-
ship functions (the four broad categories fore, the author argued that a teams engage-
of which are listed above), thereby more ment in learning behavior is strongly tied to
effectively moving toward the team goals. the teams level of psychological safety.
However, shared leadership does not pre- Thus far we have outlined the theoret-
suppose the absence of a formal hierarchi- ical drivers central to understanding adap-
cal leader. A formal leader can sometimes tive expert team performance. Research
most effectively lead by setting the climate into team effectiveness and teamwork, team
and team structure to facilitate the occur- adaptability and decision making, shared
rence of shared leadership. The success of cognition, team leadership, and collective
this shared leadership model depends on the efficacy and psychological safety serves to
fluidity with which leadership can be trans- inform us of the processes by which indi-
ferred a type of coordination itself. vidual and team competencies amalgamate
into adaptive expert team performance. The
team affective states: collective efficacy following section will review the methods
and psychological safety employed by researchers to examine expert
In addition to the cognitive and performance teams.
aspects of teams discussed in the previous
sections, recent research has highlighted the What Methods are Being Used
importance of a teams attitudes, percep- to Study Expert Teams?
tions, and beliefs and the roles that these fac-
tors play in team processes and outcomes. In order to exhibit expert performance,
Self-efficacy has long been known to be an expert team must be engaged in tasks
related to motivation and performance at the within their domain of expertise. There-
individual level (Bandura, 1977). Translated fore, observational field studies are the dom-
to the group level, it describes the teams inant research tool used to study expert
belief in the teams competence to handle teams, although methodologies incorporat-
specific environmental demands (Bandura, ing complex simulations and self-report sur-
1986). Zaccaro et al. (1995 ) define collective vey methods are used as well. In the fol-
efficacy as a sense of collective competence lowing sections, we briefly review these
shared among individuals when allocating, three methodological categories and present
coordinating, and integrating their resources exemplar studies from the expert teams
in a successful concerted response to specific research.
situational demands (p. 3 09).
In addition to collective efficacy, team observations in the field
psychological safety has been identified as Field observation studies are the main-
conducive to success when team learning is stay of expert teams research. A sam-
essential. Edmondson (1999) defined team pling of the methods used to research
psychological safety as a shared belief that expert teams include: retrospective analysis
the team is safe for interpersonal risk tak- of critical incidents (e.g. Carroll, Rudolph,
ing (p. 3 5 4). She argues that this con- Hatakenaka, Widerhold, & Boldrini, 2001),
struct comprises trust, but exceeds this to interviews (e.g. Kline, 2005 ), and field
include a team environment where individ- observations (e.g. Edmondson, Bohmer, &
the making of a dream team 445
Pisano, 2001), including video recording environmental cues and presents an oppor-
task performance (e.g. Omodei, Wearing, tunity for collecting a wider variety of quan-
& McLennon, 1997; McLennan, Pavlou, & titative and qualitative data than is nor-
Omodei, 2005 ). Observational studies are mally feasible in real-world field observation
necessary to access information about how (Pliske & Klein, 2003 ). To be valid, the sim-
teams operate in their environments; how- ulation must reach a level of functional qual-
ever, observational studies lack the con- ity that requires real-world expert teams to
trol imposed by experimental and quasi- use their expertise, regardless of the level of
experimental studies. See Lipshitz (2005 ) fidelity (Lipshitz et al., 2001; Pliske & Klein,
for a review of the issues of rigor in obser- 2003 ). That is, in order to generalize find-
vational studies. ings of simulation research back to expert
Patel and Arocha (2001) used observa- teams in a specific domain, the simulation
tional methods to study a medical and a must be engaging in such a way that it is rel-
surgical intensive care unit (MICU, SICU, evant to members of expert teams from that
respectively). Specifically, they examined domain.
how the task and environmental constraints Orasanu and Fischers (1997) study of
of the two units affected decision making. flight-crew decision-making performance is
The primary source of data in the study an example of the type of insights that can
was audiotapes of the morning rounds. The be gained into expert teams using simulation
verbatim transcripts of the audiotapes were methods. Their methodology involved using
divided into episodes based on the discus- a high-fidelity flight simulator to observe
sion topic; then each episode was divided real-world flight teams handling problem-
into a segment, or a particular aspect of atic and routine in-flight situations. The
care (e.g., lab tests, patient state) rele- scenario simulated several mechanical and
vant to the episode; segments were fur- weather conditions that required the crew
ther divided into propositions, or idea units. to perform several critical tasks: (a) decid-
The coded transcripts were then catego- ing whether to continue with a landing
rized into decision types: findings (i.e., deci- approach under risky conditions or per-
sion regarding patient-specific information), form a missed landing approach, (b) select-
actions (i.e., decision regarding future pro- ing an alternate airport to land at, and
cedures), and assessments (i.e., evaluation (c) coordinating extra functions during land-
of tradeoffs between different treatments). ing due to mechanical failures. By video-
Analysis of the data revealed that the MICU taping the sessions, the researchers had a
and SICU had markedly different com- record of expert team performance in action
munication patterns. The authors hypoth- during critical situations. They did not have
esized that this was due to differing goals to rely on retrospective reports of what
of the tasks performed in the two units. occurred and were therefore not reliant on
Similarly, by analyzing the transcripts coded the memories of team members. The authors
for type of decision making (i.e., forward- used ethnographic and cognitive engineering
or backward-driven inference), the research techniques to analyze the data and to derive
showed that there were differences between a set of decision strategies associated with
the units such that the MICU engaged more- and less-effective team performance.
in more deliberative decision making. This This work is representative of a growing
too was attributed to the differing envi- body of research into expert teams using
ronmental and task constraints in the two simulations (e.g. Roth, Woods, & Pople,
units. 1992; Cohen, Freeman, & Thompson, 1998;
Woods, 1993 ; Kanki, Lozito, & Foushee,
1989; Pascual & Henderson, 1997; Brun, Eid,
simulation Johnsen, Laberg, Ekornas, & Kobbeltvedt,
Simulation is an instrumental method for 2005 ; McLennan, Pavlou, & Omodei, 2005 ;
studying expert teams (Woods, 1993 ) in that Smith-Jensch et al., 1998; Stokes, Kemper,
it allows for experimental manipulation of & Kite, 1997).
446 the cambridge handbook of expertise and expert performance
for anticipating and reviewing the issues Expert Teams Have Strong
and problems of the members. Similarly, Team Leadership
expert teams deliberately self-diagnose ele-
Leaders of expert teams are not just techni-
ments of team effectiveness such as the
cally competent; they possess quality leader-
teams results, its processes, and vitality
ship skills. In expert teams, team members
issues such as morale, retention, and energy.
believe that the leaders care about them.
Smith-Jentsch, Zeising, Acton, and McPher-
Leaders of expert teams provide situation
son (1998) showed through a case study that
updates, foster teamwork, coordination, and
a US Navy combat information center (CIC)
cooperation, and self-correct first (see Salas
team realized high levels of performance
et al., 2004; Day, Gronn, & Salas, 2004;
by employing team self-correction and a
Pirola-Merlo, Hartel, Mann, & Hirst, 2002).
cycle of prebrief, perform/observe, diagnose
Chidester, Helmreich, Gregorich, and Geis
performance and debrief. The CIC team
(1991) showed that cockpit crews led by
was able to identify teamwork-related prob-
pilots who were highly motivated and task
lems, show immediate improvement on tar-
oriented performed better when confronted
geted goals, and generalize lessons learned,
with abnormal situations during a flight than
which resulted in sustained high levels of
did crews led by pilots with low motivation
performance.
and task orientation.
Expert Teams Have Clear Roles
and Responsibilities Expert Teams Develop a Strong Sense
of Collective, Trust, Teamness,
Expert teams are composed of individuals and Confidence
who manage their expectations by under-
standing each others roles and how they Members of expert teams are able to man-
work together to accomplish the team goals. age conflict appropriately by confronting
Expert teams have clarity of team mem- each other effectively. Expert team members
ber roles, but not to the point of excess have a strong sense of team orientation and
or rigidity in role definition. LaPorte and trust in the intentions of their fellow team
Consolini (1991) report on how air-traffic members (see Salas et al., 2004; Edmond-
controllers are able to self-organize shifts in son et al., 2001; Edmondson, 1999; Cannon
roles and responsibilities among themselves & Edmondson, 2001). They are confident in
to meet the evolving workload conditions the teams ability to succeed and develop
experienced throughout the day (see also collective efficacy. Edmondson (2003 ) found
Beauchamp, Bray, Eys, & Carron, 2002; Brun that team leaders that created a sense of trust
et al., 2005 ; Bliese & Castro, 2000). and minimized power differences were able
to realize higher levels of adaptive perfor-
Expert Teams Have a Clear, Valued, mance in interdisciplinary medical action
and Shared Vision teams.
shared mental models and task performance? at four nuclear power plants. In E. Salas &
In H. Montgomery, R. Lipshitz, & B. Klein, G. (Ed.), Linking expertise and natural-
Brehmer (Eds.), How professionals make deci- istic decision making (pp. 3 493 66). Mahwah,
sions (pp. 183 193 ). Mahwah, NJ: Erlbaum. NJ: Erlbaum.
Bunderson, J. S., & Sutcliffe, K. M. (2003 ). Man- Castka, P., Bamber, C., Sharp, J., & Belohoubek, P.
agement team learning orientation and busi- (2001). Factors affecting successful implemen-
ness unit performance. Journal of Applied Psy- tation of high performance teams. Team Perfor-
chology, 88, 5 5 25 60. mance Management, 7 , 123 13 4.
Burke, C. S., Fiore, S. M., & Salas, E. (2004). The Chi, M., Feltovich, P., & Glaser, R. (1981). Cate-
role of shared cognition in enabling shared lead- gorization and representation of physics prob-
ership and team adaptability. In C. L. Pearce & lems by experts and novices. Cognitive Science,
J. A. Conger (Eds.), Shared leadership: Refram- 5 , 12115 2.
ing the hows and whys of leadership (pp. 103 Chidester, T. R., Helmreich, R. L., Gregorich,
121). Thousand Oaks, CA: Sage. S. E., & Geis, C. E. (1991). Pilot personality
Burke, C. S., Stagl, K., Salas, E., Pierce, L., & and crew coordination: Implications for train-
Kendall, D. (in press). Understanding team ing and selection. The International Journal of
adaptation: A conceptual analysis and model. Aviation Psychology, 1, 25 44.
Journal of Applied Psychology. Cohen, M. S., Freeman, J. T., & Thompson, B.
Campbell, J. P., & Kuncel, N. R. (2001). Individ- (1998). Critical thinking skills in tactical deci-
ual and team training. In N. Anderson, D. S. sion making: A model and a training strategy.
Ones, H. K. Sinangil, & C. Viswesvaran (Eds.), In J. A. Cannon-Bowers and E. Salas (Eds.),
Handbook of industrial, work, and organiza- Making Decisions Under Stress: Implications
tional psychology. Thousand Oaks, CA: Sage for Individual and Team Training (pp. 15 5
Publications. 189). Washington, DC: American Psycholog-
Campion, M. A., Medsker, G. J. & Higgs, A. C. ical Association.
(1993 ). Relations between work group char- Cooke, N. J., Salas, E., Cannon-Bowers, J. A.,
acteristics and effectiveness: Implications for & Stout, R. J. (2000). Measuring team knowl-
designing effective work groups. Personnel Psy- edge. Human Factors, 42 , 15 1173 .
chology, 46, 823 847. Cooke, N. J., Salas, E., Kiekel, P. A., & Bell, B.
Cannon, M. D., & Edmondson, A. C. (2001). Con- (2004). Advances in measuring team cognition.
fronting failure: antecedents and consequences In E. Salas & S. M. Fiore (Eds.), Team cognition:
of shared beliefs about failure in organizational Understanding the factors that drive process and
work groups. Journal of Organizational Behav- performance. Washington, DC: American Psy-
ior, 2 2 , 161177. chological Association.
Cannon-Bowers, J. A., Salas, E., & Converse, S. Day, D. V., Gronn, P., & Salas, E. (2004). Lead-
(1993 ). Shared mental models in expert team ership capacity in teams. Leadership Quarterly,
decision making. In N. J. J. Castellan (Ed.), Indi- 15 , 85 7880.
vidual and group decision making (pp. 221246). DeShon, R. P., Kozlowski, W. J., Schmidt, A. M.,
Hillsdale, NJ: Erlbaum. Milner, K. R., & Wiechmann, D. (2004). A
Cannon-Bowers, J. A., Tannenbaum, S. I., Salas, multiple-goal, multilevel model of feedback
E., & Volpe, C. E. (1995 ). Defining compe- effects on the regulation of individual and team
tencies and establishing team training require- performance. Journal of Applied Psychology, 89,
ments. In R. Guzzon & E. Salas (Eds.), Team 103 5 105 6.
effectiveness and decision making in organiza- Driskell, J. E., & Salas, E. (1991). Group decision
tions. San Francisco, CA: Jossey-Bass. making under stress. Journal of Applied Psychol-
Cannon-Bowers, J. A., & Salas, E. (Ed.). (1998). ogy, 76, 473 478.
Making decisions under stress: Implications for Edmondson, A. C. (1999). Psychological safety
individual and team training. Washington, DC: and learning behavior in work teams. Adminis-
American Psychological Association. trative Science Quarterly, 44, 3 5 03 83 .
Carroll, J. S., Rudolph, J. W., Hatakenaka, S., Edmondson, A. C., Bohmer, R. M., & Pisano, G. P.
Widerhold, T. L., & Boldrini, M. (2001). Learn- (2001). Distrupted routines: Team learning and
ing in the context of incident investigation: new technology implementation in hospitals.
Team diagnoses and organizational decisions Administrative Science Quarterly, 46, 685 716.
the making of a dream team 451
Edmondson, A. C. (2003 ). Speaking up in the Kanki, B. G., Lozito, S., & Foushee, H. C. (1989).
operating room: How team leaders promote Communication indices of crew coordination.
learning in interdisciplinary action teams. Jour- Aviation, Space, and Environmental Medicine,
nal of Management Studies, 40(6), 1419145 2. 60, 5 660.
Ensley, M. D., & Pearce, C. L. (2001). Shared cog- Kayes, C. D. (2004). The 1996 Mount Everest
nition as a process and an outcome in top man- climbing disaster: The breakdown of learning
agement teams: Implications for new venture in teams. Human Relations, 5 7 , 1263 1284.
performance. Journal of Organizational Behav- Kelly, B. C., Badum, A., Salas, E., & Burke, C. S.
ior, 2 2 , 145 160. (2005 ). Shared cognition: Can we all get on the
Entin, E. E., & Serfaty, D. (1999). Adaptive same page? Paper presented at the 20th Annual
team coordination. Human Factors, 41, 3 21 Conference for the Society of Industrial Orga-
3 25 . nizational Psychology, Los Angeles, CA.
Fleishman, E. A., Mumford, M. D., Zaccaro, S. J., Klein, G. A. (1993 ). A recognition-primed deci-
Levin, K. Y., Korotkin, A. L., & Hein, M. B. sion (RPD) model of rapid decision making. In
(1991). Taxonomic efforts in the description G. A. Klein, J. Orasanu, R. Calderwood, & C.
of leader behavior: A synthesis and functional E. Zsambok (Eds.), Decision making in action:
interpretation. Leadership Quarterly, 4, 245 models and methods. Norwood, NJ: Ablex.
287. Klein, G. (1996). The effect of acute stressors on
Gersick, C. J. (1988). Time and transition in work decision making. In J. E. Driskell & E. Salas
teams: Toward a new model of group devel- (Eds.), Stress and human performance (pp. 49
opment. Academy of Management Journal, 3 1, 88). Mahwah, NJ: Erlbaum.
941. Klein, G. (1997). The current status of naturalis-
Guzzo, R. A., & Dickson, M. W. (1996). Teams in tic decision making framework. In R. Flin, E.
organizations: Recent research on performance Salas, M. E. Strub, & L. Martin (Eds.), Deci-
and effectiveness. Annual Review of Psychology, sion making under stress: Emerging themes and
47 , 3 073 3 8. applications (pp. 1128). Aldershot: Ashgate.
Hackman, J. R. (1983 ). A normative model of work Kleinman, D. L., & Serfaty, D. (1989). Team per-
team effectiveness (No. 2). New Haven, CT: Yale formance assessment in distributed decision-
University. making. Paper presented at the Symposium on
Hackman, J. R. (Ed.). (1990). Groups that work. Interactive Networked Simulation for Training,
San Francisco, CA: Jossey-Bass. Orlando.
Hackman, J. R. (2002). Leading teams: Setting the Klimoski, R., & Mohammad, S. (1994). Team
stage for great performances. Boston: Harvard Mental Model: Construct or Metaphor? Jour-
Business School Press. nal of Management, 2 0, 403 43 7.
Hatano, G., & Inagaki, K. (1986). Two courses Kline, D. A. (2005 ). Intuitive team decision mak-
of expertise. In H. W. Stevenson & H. ing. In H. Montgomery, R. Lipshitz, & B.
Azuma (Eds.), Child development and educa- Brehmer (Eds.), How professionals make deci-
tion in japan (pp. 262272). New York: W. H. sions (pp. 171182). Mahwah, NJ: Erlbaum.
Freeman. Kozlowski, S. W., Gully, S. M., & Salas, E. (1996).
Hinsz, V. B., Tindale, R. S., & Vollrath, D. A. Team leadership and development: Theory,
(1997). The emerging conceptualization of principles, and guidelines for training leaders
groups as information processors. Psychological and teams. In M. M. Beyerlein, & A. Dougals
Bulletin, 12 1, 43 64. (Eds.), Advances in interdisciplinary studies of
Ilgen, D. R. (1994). Jobs and roles: Accepting and work teams: Team leadership (Vol. 3 , pp. 25 3
coping with the changing structure of organiza- 291). Greenwich, CT: JAI Press.
tions. In M. G. Rumsey & C. B. Walker (Eds.), Kozlowski, S. W., & Bell, B. S. (2003 ). Work
Personnel selection and classification (pp. 13 3 2). groups and teams in organizations. In W. C.
Hillsdale, NJ: Erlbaum. Borman & D. R. Ilgen (Eds.), Handbook of
Jung, D. I., & Sosik, J. J. (2002). Transformational psychology: Industrial and organizational psy-
leadership in work groups: The role of empow- chology (Vol. 12, pp. 3 3 3 3 75 ). New York: John
erment, cohesiveness, and collective efficacy Wiley & Sons, Inc.
on perceived group performance. Small Group LaPorte, T. R., & Consolini, P. M. (1991). Work-
Research, 3 3 (3 ), 3 13 3 3 6. ing in practice but not in theory: Theoretical
452 the cambridge handbook of expertise and expert performance
challenges of high reliability organizations. Orasanu, J. (1990). Shared mental models and crew
Journal of Public Administration Research and decision making (No. 46). Princeton, NJ: Cogni-
Theory, 1, 1948. tive Sciences Laboratory, Princeton University.
Lipshitz, R. (2005 ). There is more to seeing Orasanu, J., & Salas, E. (1993 ). Team decision
that meets the eyeball: The art and science making in complex environments. In G. A.
of observation. In H. Montgomery, R. Lip- Klein, J. Orasanu, R. Calderwood, & C. E.
shitz, and B. Brehmer (Eds.), How Professionals Zsambok (Eds.), Decision making in action:
Make Decisions (pp. 3 65 3 78). Mahwah, NJ: Models and methods. Norwood, NJ: Ablex
Erlbaum. Publishing.
Lipshitz, R., & Pras, A. A. (2005 ). Not only for Orasanu, J., & Fischer, U. (1997). Finding deci-
experts: Recognition-primed decisions in the sions in natural environments. In C. E. Zsam-
laboratory. In H. Montgomery, R. Lipshitz, & bok & G. Klein (Eds.), Naturalistic deci-
B. Brehmer (Eds.), How professionals make deci- sion making (pp. 43 445 8). Hillsdale, NJ:
sions (pp. 91108). Mahwah, NJ: Erlbaum. Erlbaum.
Lipshitz, R., Klein, G., Orasanu, J., & Salas, E. Pascual, R., & Henderson, S. (1997). Evidence of
(2001). Rejoinder: A welcome dialogue and naturalistic decision making in military com-
the need to continue. Journal of Behavioral mand and control. In C. E. Zsambok & G. Klein
Decision Making, 14, 3 85 3 89. (Eds.), Naturalistic decision making (pp. 217
Marks, M. A., Mathieu, J. E., & Zaccaro, S. J. 226). Mahwah, NJ: Erlbaum.
(2001). A temporally based framework and tax- Patel, V. L., & Arocha, J. F. (2001). The nature
onomy of team processes. Academy of Manage- of constraints on collaborative decision mak-
ment Review, 2 6, 3 5 63 76. ing in health care settings. In E. Salas & G.
McLennan, J., Pavlou, O., & Omodei, M. M. Klein (Eds.), Linking expertise and naturalistic
(2005 ). Cognitive control processes discrimi- decision making (pp. 3 83 405 ). Mahwah, NJ:
nate between better versus poorer performance Erlbaum.
by fire ground commanders. In H. Mont- Pearce, C. L., & Ensley, M. D. (2004). A
gomery, R. Lipshitz, & B. Brehmer (Eds.), How reciprocal and longitudinal investigation of
professionals make decisions: Expertise, research the innovation process: The central role of
and applications (pp. 209221). Mahway, NJ: shared vision in product and process innovation
Erlbaum. teams. Journal of Organizational Behavior, 2 5 ,
Mohammed, S., & Dumville, B. C. (2001). 25 9278.
Team mental models in a team knowledge
Pearce, C. L., & Sims, H. P. (2002). Ver-
framework: Expanding theory and measure
tical versus shared leadership as predictors
across disciplinary boundaries. Journal of Orga-
of the effectiveness of change management
nizational Behavior, 2 2 (2), 89103 .
teams: An examination of aversive, directive,
Moreland, R. L. (1999). Transactive memory: transactional, transformational and empower-
Learning who knows what in work groups ing leader behaviors. Group Dynamics: Theory,
and organizations. In L. L. Thompson & J. M. Research, and Practice, 6, 172197.
Levine (Eds.), Shared cognition in organizations:
Pirola-Merlo, A., Hartel, C., Mann, L., & Hirst,
The management of knowledge (pp. 3 3 1). Mah-
G. (2002). How leaders influence the impact
wah, NJ: Erlbaum.
of affective events on team climate and perfor-
Morgan, B. B., Jr., Glickman, A. S., Woodward, mance in R&D teams. Leadership Quarterly, 13 ,
E. A., Blaiwes, A. S., & Salas, E. (1986). Mea- 5 615 81.
surement of team behaviors in a Navy environ-
ment (No. 86-014). Orlando, FL: Naval Training Pliske, R., & Klein, G. (2003 ). The naturalistic
Systems Center. decision-making perspective. In S. L. Schnei-
der & J. Shanteau (Eds.), Emerging perspectives
Omodei, M. M., Wearing, A. J., & McLennan,
on judgement and decision research (pp. 5 5 9
J. (1997). Head mounted video recording: A
5 85 ). New York: Cambridge University
methodology for studying naturalistic decision
Press.
making. In R. Flin, E. Salas, M. Strub, & L.
Martin (Eds.), Decision making under stress: Roth, E. M., Woods, D. D., & Pople, H. E. (1992).
Emerging themes and applications (pp. 161169). Cognitive simulation as a tool for cognitive task
Aldershot: Ashgate. analysis. Ergonomics, 3 5 , 1163 1198.
the making of a dream team 453
Rouse, W. B., & Morris, N. M. (1986). On look- Smith, E. M., Ford, J. K., & Kozlowski, S. W.
ing into the black box: Prospects and limits in (1997). Building adaptive expertise: Implica-
the search for mental models. Psychological Bul- tions for training design strategies. In M. A.
letin, 100, 3 493 63 . Quinones & A. Ehrenstein (Eds.), Training for a
Salas, E., Dickinson, T., Converse, S., & Tannen- rapidly changing workplace: Applications of psy-
baum, S. (1992). Toward an understanding of chological research (pp. 89118). Washington,
team performance and training. In R. Swezey DC: American Psychological Association.
& E. Salas (Eds.), Teams: Their training and per- Smith-Jentsch, K., Zeisig, R. L., Acton, B., &
formance. Norwood, NJ: Ablex Publishing. McPherson, J. A. (1998). Team dimensional
Salas, E., Cannon-Bowers, J. A., & Blickens- training: A strategy for guided team self-
derfer, E. L. (1993 ). Team performance and correction. In E. Salas & J. A. Cannon-Bowers
training research: Emerging principles. Jour- (Eds.), Making decisions under stress: Implica-
nal of the Washington Academy of Sciences, 83 , tions for individual and team training (pp. 271
81106. 297). Washington, DC: American Psychologi-
cal Association.
Salas, E., & Klein, G. (Eds.). (2001). Linking exper-
tise and naturalisitc decision making. Mahwah, Stagl, K. C., Burke, C. S., Salas, E., & Pierce,
NJ: Erlbaum. L. (2006). Team adaptation: Realizing team
Salas, E., & Cannon-Bowers, J. A. (2001). synergy. In C. S. Burke, L. Pierce, & E. Salas
The science of team training: A decade of (Eds.), Understanding adaptability: A prerequi-
progress. Annual Review of Psychology, 5 2 , 471 site for effective performance within complex envi-
499. ronments (pp. 117141). Oxford, UK: Elsevier
Science.
Salas, E., Burke, C. S., & Stagl, K. C. (2004).
Developing teams and team leaders: Strate- Stokes, A. F., Kemper, K., & Kite, K. (1997). Aero-
gies and principles. In D. Day, S. J. Zaccaro, nautical decision making, cue recognition, and
& S. M. Halpin (Eds.), Leader development expertise under time pressure. In C. E. Zsam-
for transforming organizations: Growing lead- bok & G. Klein (Eds.), Aeronautical decision
ers for tomorrow (pp. 3 25 3 5 5 ). Mahwah, NJ: making, cue recognition, and expertise under time
Erlbaum. pressure (pp. 183 196). Mahwah, NJ: Erlbaum.
Salas, E., Stagl, K. C., & Burke, C. S. (2004). Wong, S. S. (2004). Distal and local group
25 years of team effectiveness in organizations: learning: Performance trade-offs and tensions.
Research themes and emerging needs. Interna- Organization Science, 15 (6), 645 65 6.
tional Review of Industrial and Organizational Woods, D. D. (1993 ). Process-tracing methods for
Psychology, 19, 4791. the study of cognition outside of the experi-
Salas, E., Stagl, K. C., Burke, C. S., & Goodwin, mental psychology laboratory. In G. A. Klein,
G. F. (in press). Fostering team effectiveness J. Orasanu, R. Calderwood, & C. E. Zsambok
in organizations: Toward an integrative theo- (Eds.), Decision making in action: Models and
retical framework of team performance. In W. methods (pp. 22825 1). Norwood, NJ: Ablex.
Spaulding & J. Flowers (Eds.), Modeling com- Zaccaro, S. J., Blair, V., Peterson, C., & Zazanis,
plex systems: Motivation, cognition and social M. (1995 ). Collective efficacy. In J. E. Maddux
processes. Lincoln, NE: University of Nebraska (Ed.), Self-efficacy, adaptation, and adjustment:
Press. theory, research, and application. New York:
Schaafstal, A. M., Johnston, J. H., & Oser, R. L. Plenum.
(2001). Training teams for emergency manage- Zaccaro, S. J., Rittman, A. L., & Marks, M. A.
ment. Computers in Human Behavior, 17 , 615 (2001). Team leadership. Leadership Quarterly,
626. 12 , 45 1483 .
P a r t V. B
C H A P T E R 26
Music
Strauss families, which by their mere exis- prominently in everyday discourse. On the
tence suggest a strong heritability of musical other hand, much evidence exists that prac-
talents. Alas, the hope to identify heritabil- tice and other environmental factors have
ity of excellence in families is not justified a large impact on changes in many vari-
(see also discussion about Galton, 1979; e.g., ables related to music performance. Some
Simonton, Chapter 18). Genetic background researchers have expressed serious doubts
and environmental effects are mingled whether it is even possible to identify spe-
inevitably, and alternative explanations can- cific innate characteristics that mediate the
not be refuted. Older heritability explana- development of expertise (Ericsson, 2003 ).
tions failed to take into account the genetic However, the goal of this chapter is not
contribution of females in the genealogy to work out the nature-nurture debate for
or to the socio-historic fact that sons fre- music but to focus on the role of practice
quently followed in their fathers profes- for the attainment of expert performance.
sions (Farnsworth, 1969). Hence, there are In brief, we do not know whether practice is
different explanations for why many musi- a sufficient condition for high achievement,
cians have parents who are musically active but it is certainly a necessary one for invok-
(Gembris, 1998). ing the cognitive, physiological, and psycho-
The home environment is obviously motor adaptations observed in experts.
important for promoting musical excellence Although practice is omnipresent dur-
(Csikszentmihalyi, Rathunde, & Whalen, ing the development of expertise in music,
1993 ; Sosniak, 1985 ). Also, the socio- its role and manifestation is not identi-
economic conditions of a young musicians cal in all musical genres. Different musical
family constitute obvious factors that influ- styles are characteristic for specific musi-
ence the choice of a teacher, the quality cal cultures, and those different cultures
of the instrument played, and other possi- have their respective types of practice. Most
bilities awarded to the learner. An analysis research on musical expertise has been con-
of successful Polish musicians by Man- ducted in the classical conservatoire tradi-
turzewska (1995 ) revealed a common pat- tion also known as the Western art music
tern of attitudes, value systems, and family tradition. Investigations about expert per-
structure in the musicians families of origin. formance in jazz music, popular music, or
Families were emotionally stable, task orien- vernacular genres may yield somewhat dif-
ted, and careful in selecting their chil- ferent results (Berliner, 1994). For example,
drens friends, and they strongly supported whereas an early start of training is typi-
the musical activities. These attitudes gain cal for pianists and violinists in the classi-
importance as they translate into behavioral cal music domain, jazz guitarists start much
consequences in the daily lives of musicians. later (Gruber, Degner, & Lehmann, 2004),
For example, Csikszentmihalyi, Rathunde, and so do most singers today (Kopiez, 1998).
and Whalen (1993 ) demonstrated that fam- Despite some differences, important com-
ilies of high-achieving children changed monalities regarding phases of development
their lives to accommodate the needs of or deliberate practice should exist regardless
their talented offsprings, for example, by of the specific music style in question. These
exempting them from household chores will be addressed below.
to give them additional time to practice.
Biographies of famous musicians underline
such practices: the cellist Jacqueline du Pre
Increasing Performance
never did her own laundry and did not have
through Practice
any household responsibilities as an adoles-
cent (Easton, 1989).
Practice: Investing the Time
Taken together, it is difficult to obtain
clear evidence on the role of innate abili- The discussion about the role of training-
ties, despite the fact that giftedness features induced changes in performance was
music 459
triggered by Ericsson, Krampe, and Tesch- cessing to automatized processing was dis-
Romer (1993 ), who first introduced the cussed in terms of abilities versus practice,
concept of deliberate practice. See also one could argue that in early phases of skill
Ericsson, Chapter 3 8. Deliberate practice is development general abilities play an impor-
a set of structured activities that experts in tant role, which is reduced later, if con-
the domain consider important for improv- sistent task characteristics exist within the
ing performance; it is often strenuous and domain. These foster the development of
can therefore only be maintained for limited compilation processes that are heavily influ-
amounts of time per day without danger enced by practice. The more skilled a per-
of psychological or physiological burnout. son is, the more specific components of
At the Berlin Academy of Music, Ericsson information processing are relevant, so that
et al. (1993 ) investigated violin students the relation between general abilities and
from different degree programs that varied performance tends to disappear. The sub-
with regard to the instrumental proficiency jects in Ericsson et al.s (1993 ) study might
required. The students were interviewed already have compensated ability differences
retrospectively about their practice and through adaptation of practice. Thus, the
skill development. The amount of lifetime role of talent could not be judged ade-
accumulated practice up to the point of quately. Additionally, experts might be most
the interview (or even the point of entry competent in selecting proper practice. At
into the academy) was clearly related to the younger age levels, however, practice may
degree of level of performance attained. not be as efficient, and a smaller amount of
Less-proficient performers had practiced practice is accumulated. Therefore, practice
less than more-skilled ones. The lifetime time might be of less importance and abil-
trajectory of practice reported by the most ity of more importance in young musicians
promising group of students resembled (Lehmann, 1997b, for a review).
that of musicians currently employed in Sloboda, Davidson, Howe, and Moore
Berlin orchestras. The results underline (1996) addressed this problem by repli-
the predictive validity of the accumulated cating Ericsson et al.s (1993 ) study. Stu-
hours as an indicator of excellence, and dents aged eight to eighteen from a music
are hence at odds with the everyday belief school were rated by their teachers with
that some musicians the highly talented respect to musical achievement and promise.
ones need not practice as much as the less The students then were interviewed by the
talented, who have to compensate for lack researchers about their practice history. In
of talent with excessive practice. addition, many other data were collected,
Although it may be difficult to refute including a 42-week longitudinal record-
claims by famous musicians not to have prac- ing of practice diaries. The results clearly
ticed much or to hate practice (Mach, 1981), support the deliberate-practice assumptions
the empirical evidence regarding contem- obtained from the study of adults. The least-
porary musicians makes such claims rather proficient group of subjects had practiced
suspicious. Musicians are likely to engage less than the better-performing groups, and
in conscious impression management when the students who dropped out of music
belittling practice in the classical tradition lessons had practiced even less. In order to
or, in the case of rock and popular musi- proceed from one level of performance to
cians, to dismiss the role of formal instruc- the next, the best groups increase in practice
tion by emphasizing self-teaching, that is, was even larger than expected. Thus, already
autodidactic learning (Green, 2002). relatively early in instrumental music learn-
The relation between innate abilities and ing, the amount of practice is significantly
practice probably is a complex one. Accord- related with level of performance.
ing to Ackermans (1986, 1990) theory of Competing explanations could be that
ability determinants of skilled performance, talented children practice more in a rage
in which the change from controlled pro- to master the skill (Winner, 1996), or that
460 the cambridge handbook of expertise and expert performance
tangible progress and success keep children with an accompanist is an important practice
practicing. Although a proof in favor of one activity. A conductor has to become fami-
of these argumentations cannot easily be liarized with a piece first without the orches-
made, the latter is clearly better supported tra, silently reading and imaging the score
by research. The result that the amount of and the desired interpretation before work-
certain types of practice is related to level ing with the ensemble. Deliberate prac-
of performance was found in other domains tice is goal-directed, optimized practice, and
as well. responds to the typical demands imposed by
It is noteworthy that the number of the domain.
hours necessary for achieving particular lev-
els of performance is not a constant across
Investing the Effort
all musical instruments. Jrgensen (1997)
showed that different instrumental groups Ericsson et al. (1993 ) stressed that practice
practice very different numbers of hours. could lack inherent enjoyment because it
Pianists and violinists tend to be practice requires much mental and physical effort.
fanatics, logging the most hours, followed by Musicians may enjoy their own improve-
other strings, organ, woodwinds and brass, ment but dislike the actual practice activity.
closing with the singers at the bottom of the In a survey study on practicing, musicians
list. Such differences may result from differ- indicated that performing in front of an audi-
ent demands that instrumental performance ence was most enjoyable but highly effort-
imposes on the body. In the case of singers, ful (Lehmann, 2002), whereas learning new
different educational traditions may have an pieces and working on difficult spots was
influence as well (Kopiez, 1998). most effortful but not enjoyable. Appar-
Although duration of practice is predic- ently, activities that resemble the target
tive of long-term success, it might be not as activity of performing seem more enjoyable
indicative of performance in the short run, than activities of a preparatory nature, even
for example, when learning a specific piece though the latters relevance for improving
of music (Williamon & Valentine, 2000). skills is unquestioned. This result indicates
Here, a players prior knowledge with the that enhancing quality of practice requires
music might influence practice times. For substantial effort (Williamon, 2004, several
example, those who have not worked sys- chapters).
tematically on music by Bach may face prob- Ample advice from practitioners such as
lems that experienced Baroque performers master teachers is readily available in books.
do not (Lehmann & Ericsson, 1998). Although the suggestions are grounded in
Although time invested in practice is lifelong experience, some recipes appear
related to long-term level of performance, haphazard. Take, for example, the notion of
practice means different things for differ- slow practice. Playing a section very slowly
ent musical styles and sub-skills. For musi- is often recommended among music teach-
cians playing classical repertoire, a large ers as a remedy for all sorts of problems.
portion of practice is solitary practice, work- However, the piano teacher Matthay (1926),
ing on instrumental technique and acquiring who was knowledgeable about psychological
new pieces, assisted by more or less regu- research, remarked that slow practice with-
lar visits with a teacher. For jazz musicians, out actually imagining the upcoming note
in addition to solitary practice, a substantial is only a useless fetish (p. 12). This implies
part of practice is communal practice with that the quality of practice is not sufficiently
other musicians. Sitting in jam sessions, lis- defined by observable behavior (e.g., mere
tening to others play, and copying perfor- duration) but has to be judged by the co-
mances by famous musicians from record- occurrence of certain cognitive processes.
ings all constitute activities that improve Many researchers in the field use verbal-
performance (Gruber et al., 2004). For report methodologies to get at these
singers and some instrumentalists, working difficult-to-observe processes. For example,
music 461
Chaffin, Imreh, and Crawford (2002) pub- constraints such as starting at the jump-
lished an extensive case study about solitary points. If possible, the musician even
practice. In a naturalistic setting, they fol- practices under performance conditions,
lowed a performer practicing a new piece that is, in concert attire and in different
for performance and obtained retrospective locations. After some time, the returns of
and concurrent reports. The authors distin- such final polishing and preparation work
guished four stages of practice. are diminishing; further practice is con-
sidered to be maintenance work.
1. In the first, the musician tries to get the
4. The fourth stage, which sometimes
big picture of the piece. The first stage
extends over a long period of time
entails reading through the piece or more
between concerts or recordings, consti-
generally getting an aural representation
tutes the maintenance of the piece.
of the entire piece. Practice strategies vary
substantially: according to working habits Thus, practice is a systematic activity with
of the musician, sight-reading, analyzing, predictable stages and activities. They all
or listening to recordings are preferred. serve to establish a strong internal represen-
2. In the second stage, technical practice is tation of the piece and the conditions under
undertaken to master the piece. During which the performance will take place.
the second stage the piece is worked on in Practicing is an effortful activity and a
sections, which increase in length as prac- skill per se that has to be learned. Gruson
tice progresses. The length of the section (1988) demonstrated that experts differed
depends on the kind of problems encoun- from novices in their practice skill. A num-
tered and analyses undertaken in the first ber of studies revealed that (adult) supervi-
stage. Whether the artistic interpretation sion during practice is important for begin-
is developed during the course of learn- ning musicians (Davidson, Howe, Moore, &
ing to play the piece or during anticipa- Sloboda, 1996; Lehmann, 1997a; Sosniak,
tory analytic processes might be a matter 1985 ). In the simplest case, the adult or the
of individual preferences and habits (Hal- supervisor ensures that time is spent with
lam, 1995 ). During this elaborative stage the instrument. Preferably, goals and feed-
of practice, the motor programs become back are provided. Research suggests that
largely automatic and the piece is being not all parents or tutors necessarily have to
memorized. be musicians everyone can hear wrong
3 . Next, in the third stage the actual stage notes, encourage lovingly, or simply watch
performance is tried out. During this the clock. However, the mothers previous
stage, performance is prepared more experience with learning a musical instru-
directly by putting the pieces together ment may influence their ideas about how
and ironing out the seams between them. much practice is necessary and their abil-
Memory, which up to then was more ity to support the childs practice (McPher-
implicit, is now deliberately assisted by son & Davidson, 2002). Written procedures
creating an internal map of the piece, have been found also to be helpful in struc-
knowing the order of the parts as well as turing practice for beginners (Barry & Hal-
points where the performer could restart lam, 2001). After the musicians have devel-
in case of a memory lapse during per- oped metacognitive skills, they can take over
formance. During this stage the piece to regulate their practice themselves. A cru-
is polished by slow playing, playing for cial factor for doing so is the motivation to
an imagined or an informal real audi- invest effort and to engage in the process
ence, refining interpretation details, and of self-regulation (McPherson & Zimmer-
bringing all sections up to the correct man, 2002). Renwick and McPherson (2002)
tempo or even slightly above. As per- showed that children practicing by them-
formance approaches, memory is repeat- selves engaged in elaborated activities when
edly tried and tested using self-imposed they were motivated by the piece, but simply
462 the cambridge handbook of expertise and expert performance
played through the piece when disinterested cal households may be at an advantage.
in it. Similarly, when the goal is to master Then comes a phase during which formal
a certain piece or a specific difficulty, adults tuition is sought. This stage extends until the
may work hard and use more practice strate- young musician makes a full-time commit-
gies than when they want to enjoy them- ment to music in order to become a pro-
selves or relax rather than mastering the fessional. In a later phase, once a profes-
instrument (Lehmann & Papousek, 2003 ). sional status has been reached, the expert
is working at trying to make a lasting con-
tribution. For a musician, this would entail
making sound recordings for major record
The Development of Musical Expertise
labels, playing in prestigious concert halls
(e.g., Carnegie Hall), or winning certain
Stages and Phases
competitions (e.g., Frederic Chopin Interna-
Demarcating points along the time line of tional Piano Competition). Vitouch (2005 )
acquisition of a new skill allows us to bet- described in detail how parts of the exper-
ter conceptualize the process as a whole, tise may get lost in old age one could call
and several stage and phase models have this de-expertization and how experts
been proposed (see Proctor & Vu, Chap- like the piano soloist Horowitz possibly
ter 15 ). Fitts and Posner (1967), for example, compensate for it. Interestingly the psycho-
in their well-known model describe skills as motor adaptations do not decline inevitably
being first cognitive, then associative, and with old age but can be maintained for
later autonomous, in essence requiring less a long time through continuous practice.
and less cognitive mediation as skilled per- Krampe and Ericsson (1996) demonstrated
formance increases. Similar stages can be that older pianists were able to counter-
identified in Sudnows (1993 ) phenomeno- act losses in motor performance through
logical account of his learning to impro- practice, whereas non-pianists did not show
vise jazz on the piano. First, he had to this advantage. However, both groups suf-
decide consciously which chord to use next fered age-related declines in other cogni-
and then how to distribute the chordal tive domains (see also Krampe & Charness,
notes on the keyboard (voicing). Later his Chapter 40).
fingers seemed to find the right notes by The time needed for experts to develop
themselves. Much later, his aesthetic deci- sufficient skills for a professional career is
sion of what to play seemed to trigger sometimes estimated to be roughly a decade
the correct chord sequences with associ- (Ericsson & Crutcher, 1990). Hayes (1989)
ated voicings. Whereas the focus of atten- demonstrated that this 10-year rule also
tion in novices is directed toward technical, applies to composers in classical music,
low-level aspects, experts attend to higher- including Mozart. Works from Mozarts ear-
level, strategic or aesthetic issues, a finding liest phases were conspicuously underrepre-
also demonstrated for composing (Colley, sented in selected lists of his recordings. Sim-
Banton, & Down, 1992) and improvising ilarly, Weisberg (1999) demonstrated that it
(Hargreaves, Cork, & Setton, 1991). For took The Beatles approximately a decade
many musicians the earlier stages of skill to acquire international reputation. Prior to
development are successfully completed in writing their own songs they covered music
(early) childhood. by other bands. It is a futile effort to dwell
Fitts and Posners (1967) model is infor- on exact number of years, but it is impor-
mative with regard to skill development tant to note that even famous exponents of
of an individual, but it neglects the life- a domain take a long time to acquire their
span context. Bloom (1985 ) explicated how skills.
skills develop through life. First, the child In order to compete successfully for
is introduced to the domain in an informal scholarships, prizes, and media attention,
phase, and it is here that children in musi- instrumentalists in the classical music
music 463
domain have to master the most demanding & Kintsch, 1995 ). Chaffin and Imreh (2001)
repertoire as teenagers. This requires either showed convincingly how a concert pianist
an early start for highly competitive instru- developed sophisticated mental represen-
ments (e.g., violin) or the possibility of trans- tations with associated retrieval structures
fer of knowledge and skills from previously that lead to successful performance of the
played instruments onto those instrument rehearsed piece from memory even under
that do not allow such early start (e.g., high-stress conditions on stage. In addition,
string bass, oboe, trombone). Altogether, the knowledge is represented in an elaborated
development of expert performance can be format that allows quick access to relevant
seen as an adaptation to the typical task con- information and supports flexible reactions
straints of the domain (Ericsson & Lehmann, to domain-specific tasks, for example, in
1996), involving changes in cognitive, phys- medicine by encapsulation of knowledge in
iological, and perceptual-motor parameters procedural representations of earlier expe-
that facilitate superior performance. riences with cases (Boshuizen & Schmidt,
1992).
A particularly impressive effect of the
Cognitive Adaptations
impact of knowledge for musical perfor-
Among the cognitive adaptations are aspects mance was found in studies with autistic
of memory and problem solving. The for- savants. Despite the cognitive and commu-
mer can be seen in virtually all domains of nicative limitations that prevent them from
expertise (Ericsson & Lehmann, 1996). Even functioning normally in everyday contexts,
when memorization is not their explicit some autistic savants have exceptional musi-
goal, experts tend to have excellent long- cal skills and can play back music after only
term retention for domain-related material. a few hearings. It can be demonstrated that
For example, incidental memory for music these skills are based on knowledge-related
just played correlated moderately with generative processes (Miller, 1999; Sloboda,
accompanying ability in classical pianists Hermelin, & OConnor, 1985 ). When con-
(Lehmann & Ericsson, 1996). Kauffman fronted with atonal music, the savants fail to
and Carlsen (1989) showed that musicians imitate music but simply play haphazardly.
recalled musical material better than non- Obviously, familiarity with the material and
musicians, especially when the material was the genre mediates memory performance
structured according to rules of tonality (Charness, Clifton, & MacDonald, 1988).
(see our example concerning savants in The phenomenon of savants music mem-
later paragraph). Expert-novice differences ory demonstrates that making use of ones
decreased when tonality rules were violated knowledge about the structure of the stimu-
or when random note sequences had to be lus is a quick and automatic process. Special-
recalled. This skill-by-structure interaction, ized knowledge of musical timbre and pitch
demonstrated also in other domains, docu- even impacts early stages of perceptual pro-
ments that experts advantages are largely cessing that are not accessible to conscious-
due to their knowledge and how their mem- ness (Besson, 1997).
ory skills have adapted to the structure of The study of individuals cognitive repre-
the stimuli. sentation of musical structure is important
Acquired domain knowledge has been the for understanding how music performance
most prominent explanation for the supe- works (Palmer, 1997). It helps to understand
riority of expert performance. Studies in why certain mistakes happen, and in which
many different domains showed that the way a good use of the knowledge can be sup-
essential factor of development of exper- ported. But also from an educational point
tise is the accumulation of increasingly com- of view it is relevant to know how different
plex patterns in memory. It has been shown learning processes or learning methods may
that expert knowledge can be retrieved result in different representations (Gruhn &
quickly from long-term memory (Ericsson Rauscher, 2002).
464 the cambridge handbook of expertise and expert performance
Musical knowledge comprises not only ond study, Gruber et al. (1996) surprisingly
knowledge about musical pieces, but found that experts competence and control
the cognitive mechanisms to represent beliefs were weaker concerning musical per-
and manipulate the relevant knowledge. formance than regarding everyday life. This
Lehmann and Ericsson (1997) suggested a leads to an ambiguous situation. On the one
triangular model of mental representations side, early on experts practice effectively,
for musicians. In brief, musicians first look for challenging performance situations,
need to imagine their anticipated, desired and have aspirations. On the other side, once
outcome. Next, they have to represent their they work within their current community
currently ongoing performance in order to of experts, where they are only one among
compare it to their original plans. Finally, a many, they neither perceive themselves
mental representation is necessary of how a as outstanding nor do they have superior
particular plan can be implemented on the self-concepts.
instrument how it feels. Woody (1999)
investigated the connection between ongo-
Physiological Adaptations
ing and desired performance (see Woody,
2003 , concerning motor production repre- Everyone knows the minor physiological
sentations). Pianists were asked to imitate adaptations that happen in response to
the artistic, expressive features of a model habitual usage of our bodies in everyday life.
musical performance, and verbal reports These adaptations are specifically localized,
were recorded indicating which features such as the growth of muscle after a few days
they explicitly identified. Performance data of bike riding or the emergence of calluses
showed that they imitated more accurately on fingertips after starting to play the guitar
those features that they also correctly or working in the yard. Musicians undergo
identified in their verbal reports. Also, a number of less obvious but highly telling
researchers investigating African drummers adaptations. For example, Wagner (1988)
found that rhythms in triple meter such as found that degree of forearm rotation dif-
the Bolero-rhythm were notoriously diffi- fered systematically between pianists (larger
cult for experienced master drummers to extent of inward rotation), violinists (larger
imitate, who tried to assimilate the rhythms outward rotation), and controls. However,
to African rhythmic prototypes (Kopiez, the overall degree of rotation remained con-
Langner, & Steinhagen, 1999). Thus, even stant in all three groups but was shifted
seemingly automatic performance is medi- toward the respective habitual usages for
ated by complex cognition, even at high the instrumentalists. Singers and brass play-
levels of proficiency. ers were found to have significantly larger
In addition to knowledge-related cogni- vital and total lung capacities compared to
tive adaptations we can also observe changes controls (Sundberg, 1987). And the supe-
in the use of sophisticated metacognitive rior inhalation and expiration pressures in
and self-regulation skills in musical learners. trumpet players were found only after sev-
In a contrastive study addressing metacog- eral long notes were played (Fiz et al., 1993 ),
nitive components of expert performance, demonstrating the highly contextual speci-
Gruber, Weber, and Ziegler (1996) analyzed ficity of such changes.
top-level orchestra musicians and above- Additional links between training and cer-
average amateur musicians. Judging retro- tain adaptations were uncovered in recent
spectively, experts indicated higher levels of efforts to understand how the brain pro-
aspiration along with a more positive atti- cesses music, especially through the use of
tude toward performance situations com- imaging techniques (Munte, Altenmuller, &
pared to the amateurs. As regards their Jancke, 2002, for a review). The first study
current situation, experts rated themselves that received widespread attention was one
more effective in their learning behavior, that found that the cortical representation
but did not differ from the amateur play- of the fingers of the left hand in string play-
ers concerning their motivation. In a sec- ers was enlarged compared to that of the
music 465
thumb (Elbert, Pantev, Wienbruch, Rock- & Horowitz, 1987). However, the improved
stroh, & Taub, 1995 ). No changes occurred discrimination of timbre and tones by
with the representations of the fingers of the musicians does not transfer to speech
right hand (the bowing arm). And this corti- sounds (Munzer, Berti, & Pechmann, 2002).
cal reorganization was more pronounced for Musicians playing instruments that require
subjects who had started musical training at fine tuning of individual notes during
an earlier age. performance develop a more accurate dis-
Further studies, especially those compar- crimination for pitch height, whereas per-
ing experts with novices, showed that cor- cussionists, whose work relies heavily on
tical reorganization was not restricted to discriminating rhythms, show an improved
playing music but also occurred when lis- perception of auditory duration (Rauscher
tening. Pantev, Roberts, Schulz, Engelien, & Hinton, 2003 ). Likewise, pianists require
& Ross (2001) learned that larger areas of increased sensitive tactile discrimination,
the cortex were activated involuntarily when which proved to be related to the amount
musicians listened to tones of instruments of practice undertaken (Ragert, Schmidt,
they played. Or, differences in the volume of Altenmuller, & Dinse, 2004). Taken to-
gray matter in the motor as well as auditory gether, the increased acuity of the senses
and visuospatial brain regions were found and adaptations of the motor system are
when comparing professional musicians restricted to the stimuli musicians typically
(keyboard players) to amateur musicians and encounter when playing their respective
non-musicians (Gaser & Schlaug, 2003 ). We instruments. This indicates that the changes
can safely assume that music training and are highly specific, which makes the claim
practice leads to substantial functional and plausible that they are in large part linked to
structural changes in a persons brain and training and practice.
consequently alters processing capabilities.
at the top of the list. There may also exist an improved training and practice methods, and
upper limit for attainable performance with from the extrinsic rewards a society offers to
regard to the neuroplastic changes (Lim & those who try to make eminent contribu-
Altenmuller, 2003 ). Focal dystonia, a con- tions to the domain.
dition where, for instance, fingers start to Similar to the domain of sports, where
perform involuntary movements when other some disciplines are popular in certain coun-
fingers are activated, may be due to an tries but not in others, or where some coun-
overlap of expanded cortical representations tries provide incentives to reach the high-
(Elbert et al., 1998). The gradual enlarge- est levels of performance, music is affected
ment of the cortical representations in the by societal factors. China, for example,
somatosensory cortex during acquisition of has developed a highly competitive piano
expertise might reach a limit when the instruction system since the end of the Cul-
separation between adjacent areas becomes tural Revolution; playing the piano is now
blurred, resulting in uncontrollable coactiva- a valued cultural practice. Being proficient
tion of one finger through the use of another. at playing the piano affords girls the oppor-
It is obvious that research about physio- tunity to marry into better situated fami-
logical limits of musical performance and lies (similar to the situation in 19 th -century
about interventions to overcome the limits Germany), and men receive the possibil-
(or remedy existing problems) is still at its ity to make a career (as piano teachers).
very beginning. The same can be said about It is estimated that 5 0 million Chinese are
historical or societal constraints on musical seriously playing the piano. The large num-
performance. ber of highly qualified foreign students from
It is interesting to study in the history of a Eastern European and Far Eastern coun-
domain how the demands imposed on musi- tries entering performance degree programs
cians have changed over time (Lehmann & in music academies in Western countries
Ericsson, 1998). Everyone is aware of attests to this fact. At the same time, fewer
changed standards in sports, where records and fewer families in the West are will-
are kept about achievements that have to ing to surrender their children to a rigorous
be matched and surpassed by following gen- training starting in early childhood and to
erations of athletes. (Even if they are not accept personal and financial disadvantages.
as obvious as world records in athletics, Instead, broad ranges of competing activities
musical achievements offer similar trends and media use are offered to children. That
incidentally the young star pianist Yundi Li indicates that the cultural environment and
is making commercials for sportswear com- its value and reward systems promote the
pany Nike.) For example, the constraints development or neglect of skills in a certain
of performance are related to the develop- culture.
ment of instruments. When the piano was Research reported in this chapter pre-
invented in 1700, there was no specific way dominantly deals with the Western art music
of playing it, and a standard repertoire did tradition. To date, studies in non-European
not yet exist. Later refinements of the instru- music genres are rare but would be interest-
ment and the instrumental technique led to ing for many reasons. For example, Indian
more complex compositions. A number of musicians are likely to show interesting
pieces exist that were deemed unplayable problem-solving strategies because they per-
at the time of their composition, including form mainly improvised music as do
examples even from the 20th century (e. g., musicians in the Middle East. Or Balinese
Etudes for guitar by Villa-Lobos; Ham- musicians, who learn by ear an extensive
merklavier sonata for piano by Beethoven; repertoire, would offer insights into mem-
Etudes for piano by Ligeti; Caprices for ory processes that are not mediated by
violin by Paganini). Nowadays many of these music notation as would to a certain
pieces are standard fare for adolescent per- degree European vernacular musicians in
formers. Such historical increases in levels rock, popular, jazz, and folk music. New sub-
of performance result from specialization, domains emerge that constitute touchstones
music 467
for theories previously developed in the clas- Ackerman, P. L. (1990). A correlational analysis
sical music domain. For example, one should of skill specificity: Learning, abilities, and indi-
ask what constitutes practice for a DJ? Thus, vidual differences. JEP: Learning, Memory, and
research in music expertise needs to take a Cognition, 16, 883 901.
broader look at music. Barrington, D. (1770). Account of a very remark-
The fact that the Annual Review of Psy- able young musician. Philosophical Transactions
chology has, over the last 15 years, published of the Royal Society of London, 60, 464.
three articles on music (in 1991 one on per- Barry, N., & Hallam, S. (2001). Practice. In R.
ception, in 1997 one on performance, and Parncutt & G. McPherson (Eds.), Science and
psychology of music performance (pp. 15 1166).
in 2005 one on neuroscience) proves that
Oxford: Oxford University Press.
music is a domain with a high appeal for
Berliner, P. (1994). Thinking in jazz. Chicago:
studying a diversity of psychological topics.
Chicago University Press.
The combination of affective, perceptual,
Besson, M. (1997). Electrophysiological stud-
cognitive, and motor aspects in music mak-
ies of music processing. In I. Deliege &
ing, along with its high cultural value, make J. Sloboda (Eds.), Perception and cognition
it a prime candidate for the study of com- of music (pp. 21725 0). London: Taylor &
plex skills. Children are introduced to music Francis.
very early on in their lives earlier than in Billroth, T. (1895 ). Wer ist musikalisch? Nachge-
most other domains of expertise at a time lassene Schrift. [Who is musical?] (Ed. by E.
when their brains and bodies are malleable Hanslick). Berlin: Paetel.
and training can be most effective. There- Bloom, B. S. (1985 ). Generalizations about talent
fore, we observe differences between musi- development. In B. S. Bloom (Ed.), Developing
cal experts and novices of stupefying mag- talent in young people (pp. 5 075 49). New York:
nitude. The universal nature of music as a Ballantine.
grammar-based but non-semantic temporal Boshuizen, H. P. A., & Schmidt, H. G. (1992).
phenomenon theoretically allows studies in On the role of biomedical knowledge in clin-
all cultures and across time, adding to the ical reasoning by experts, intermediates and
appeal of music as a domain for expertise novices. Cognitive Science, 16, 15 3 184.
researchers. Finally, the potential connec- Boyle, J. D. (1992). Evaluation of music ability. In
tions to music education in and out of formal R. Colwell (Ed.), Handbook of research in music
learning contexts make expertise research teaching and learning (pp. 247265 ). New York:
a fruitful area of research for those whose Schirmer.
interests concern the effects of instruction Brandfonbrener, A., & Lederman, R. (2002).
and training. Performing arts medicine. In R. Colwell &
C. Richardson (Eds.), The new handbook
of research on music teaching and learning
(pp. 10091022). New York: Oxford University
Author Notes Press.
Chaffin, R., & Imreh, G. (2001). A comparison of
We thank two reviewers and R. H. Woody for practice and self-report as sources of informa-
their insightful comments on a previous version tion about the goals of expert practice. Psychol-
of the paper. The first author is greatly indebted ogy of Music, 2 9, 3 969.
to Anders Ericsson and Neil Charness for starting Chaffin, R., Imreh, G., & Crawford, M. (2002).
him out on this fascinating topic during a post- Practicing perfection: Memory and piano perfor-
doc at FSU. mance. Mahwah, NJ: Erlbaum.
Charness, N., Clifton, J., & MacDonald, L. (1988).
Case study of a musical mono-savant. In L.
References Obler & D. Fein (Eds.), The exceptional brain:
Neuropsychology of talent and special abilities
Ackerman, P. L. (1986). Individual differences (pp. 277293 ). New York: Guilford.
in information processing: An investigation of Colley, A., Banton, L., & Down, J. (1992). An
intellectual abilities and task performance dur- expert-novice comparison in musical composi-
ing practice. Intelligence, 10, 10113 9. tion. Psychology of Music, 2 0, 12413 7.
468 the cambridge handbook of expertise and expert performance
Csikszentmihalyi, M., Rathunde, K., & Whalen, Gaser, C., & Schlaug, G. (2003 ). Gray matter dif-
S. (1993 ). Talented teenagers: The roots of suc- ferences between musicians and nonmusicians.
cess or failure. Cambridge: Cambridge Univer- Annals of the New York Academy of Sciences,
sity Press. 999, 5 145 17.
Davidson, J. W., Howe, M. J. A., Moore, D. G., Gembris, H. (1998). Musikalische Begabung und
& Sloboda, J. A. (1996). The role of parental Entwicklung. [Musical ability and develop-
influences in the development of musical abil- ment] Augsburg: Wissner.
ity. British Journal of Developmental Psychology, Goolsby, T. W. (1994). Profiles of processing: Eye
14, 3 99412. movements during sightreading. Music Percep-
Easton, C. (1989). Jacqueline du Pre: A biography. tion, 12 , 97123 .
New York: Summit. Green, L. (2002). How popular musicians learn.
Elbert, T., Candia, V., Altenmuller, E. O., Rau, London: Ashgate.
H., Sterr, A., Rockstroh, B., Pantev, C., & Taub, Gruber, H., Degner, S., & Lehmann, A. C.
E. (1998). Alteration of digital representations (2004). Why do some commit themselves
in somatosensory cortex in focal hand dystonia. in deliberate practice for many years and
NeuroReport, 9, 3 5 713 5 75 . so many do not? Understanding the devel-
Elbert, T., Pantev, C., Wienbruch, C., Rockstroh, opment of professionalism in music. In M.
B., & Taub, E. (1995 ). Increased cortical repre- Radovan & N. Dordevic (Eds.), Current issues
sentation of the fingers of the left hand in string in adult learning and motivation (pp. 222
players. Science, 2 70, 3 05 3 07. 23 5 ). Ljubljana: Slovenian Institute for Adult
Ericsson, K. A. (2003 ). The search for general Education.
abilities and basic capacities. In R. J. Stern- Gruber, H., Weber, A., & Ziegler, A. (1996).
berg & E. L. Grigorenko (Eds.), The psychology Einsatzmoglichkeiten retrospektiver Befragun-
of abilities, competencies, and expertise (pp. 93 gen bei der Untersuchung des Expertiseer-
125 ). Cambridge: Cambridge University Press. werbs. [The use of retrospective inquiry in the
Ericsson, K. A., & Crutcher, R. J. (1990). The study of expertise acquisition] In H. Gruber &
nature of exceptional performance. In P. B. A. Ziegler (Eds.), Expertiseforschung. Theoretis-
Baltes, D. L. Featherman, & R. M. Lerner che und methodische Grundlagen (pp. 169190).
(Eds.), Life-span development and behavior Opladen: Westdeutscher Verlag.
(Vol. 10, pp. 187217). Hillsdale, NJ: Erlbaum. Gruhn, W., & Rauscher, F. (2002). The neuro-
Ericsson, K. A., & Kintsch, W. (1995 ). Long- biology of music cognition and learning. In R.
term working memory. Psychological Review, Colwell & C. Richardson (Eds.), The new hand-
102 , 211245 . book of research on music teaching and learning
Ericsson, K. A., Krampe, R. T., & Tesch-Romer, (pp. 445 460). New York: Oxford University
C. (1993 ). The role of deliberate practice in the Press.
acquisition of expert performance. Psychologi- Gruson, L. M. (1988). Rehearsal skill and musi-
cal Review, 100, 3 63 406. cal competence: Does practice make perfect?
Ericsson, K. A., & Lehmann, A. C. (1996). In J. A. Sloboda (Ed.), Generative processes in
Expert and exceptional performance: Evidence music (pp. 91112). Oxford: Clarendon.
of maximal adaptations to task constraints. Hallam, S. (1995 ). Professional musicians
Annual Review of Psychology, 47 , 273 3 05 . approaches to the learning and interpretation
Farnsworth, P. R. (1969). The social psychology of of music. Psychology of Music, 2 3 , 111128.
music. Ames: Iowa University Press. Hargreaves, D., Cork, C., & Setton, T. (1991).
Fitts, P. M., & Posner, M. I. (1967). Human per- Cognitive strategies in jazz improvisation:
formance. Belmont: Brooks/Cole. An exploratory study. Canadian Journal of
Fiz, J. A., Aguilar, J., Carreras, A., Teixido, A., Research in Music Education, 3 3 , 475 4.
Haro, M., Rodenstein, D., & Morera, J. (1993 ). Hayes, J. R. (1989). The complete problem solver.
Maximum respiratory pressures in trumpet (2nd ed.). Hillsdale, NJ: Erlbaum.
players. Chest, 104, 1203 1204. Houtsma, A. J., Durlach, N. I., & Horowitz, D. M.
Galton, F. (1979). Hereditary genius: An inquiry (1987). Comparative learning of pitch and
into its laws and consequences. London: Fried- loudness identification. Journal of the Acousti-
man. (Original published 1869.) cal Society of America, 81, 12913 2.
music 469
Jrgensen, H. (1997). Time for practising? In Lehmann, A. C., & Ericsson, K. A. (1997).
H. Jrgensen & A. C. Lehmann (Eds.), Does Research on expert performance and deliberate
practice make perfect? (pp. 123 140). Oslo: practice: Implications for the education of ama-
Norges Musikhogskole. teur musicians and music students. Psychomu-
Kauffman, W. H., & Carlsen, J. C. (1989). Mem- sicology, 16, 405 8.
ory for intact music works: The importance for Lehmann, A. C., & Ericsson, K. A. (1998).
musical expertise and retention interval. Psy- The historical development of domains of
chomusicology, 8, 3 19. expertise: Performance standards and innova-
Keele, S., Pokorny, R., Corcos, D., & Ivry, R. tions in music. In A. Steptoe (Ed.), Genius and
(1985 ). Do perception and motor production the mind: Studies of creativity and temperament
share a common timing mechanism? Acta Psy- in the historical record (pp. 6497). Oxford:
chologica, 60, 173 193 . Oxford University Press.
Kopiez, R. (1998). Singers are late beginners: Lehmann, A. C., & Papousek, S. (2003 ). Self-
Sangerbiographien aus Sicht der Exper- reported performance goals predict actual
tiseforschung. Eine Schwachstellenanalyse. practice behavior among adult piano begin-
[Singers biographies from the perspective ners. In R. Kopiez, A. C. Lehmann, I. Wolther,
of research on expertise. An analysis of & C. Wolf (Eds.), Proceedings of the 5 th Tri-
weaknesses] In H. Gembris, R. Kraemer, & ennial Conference of the European Society for
G. Maas (Eds.), Singen als Gegenstand der the Cognitive Sciences of Music (pp. 3 89
Grundlagenforschung (pp. 3 75 6). Augsburg: 3 92). Hannover: University of Music and
Wissner. Drama.
Kopiez, R., Langner, J., & Steinhagen, P. (1999). Lim, V., & Altenmuller, E. O. (2003 ). Musi-
Afrikanische Trommler (Ghana) bewerten und cians cramp: Instrumental and gender differ-
spielen europaische Rhythmen African dicum- ences. Medical Problems of Performing Artists, 18,
mers assess and play European rhythms. Musi- 2127.
cae Scientiae, 3 , 13 9160. Mach, E. (1981). Great pianists speak for them-
Krampe, R. T., & Ericsson, K. A. (1996). Main- selves. London: Robson.
taining excellence: Deliberate practice and Manturzewska, M. (1995 ). A biographical study
elite performance in young and older pianists. of the life-span development of professional
JEP: General, 12 5 , 3 3 13 5 9. musicians. In M. Manturzewska, K. Mik-
Lehmann, A. C. (1997a). Acquired mental rep- laszewski, & A. Bialkowski (Eds.), Psychology
resentations in music performance: Anecdo- of music today (pp. 3 113 3 7). Warsaw: Fryderyk
tal and preliminary empirical evidence. In H. Chopin Music Academy.
Jrgensen & A. C. Lehmann (Eds.), Does prac- Matthay, T. (1926). On memorizing and playing
tice make perfect? (pp. 141164). Oslo: Norges from memory and on the laws of practice gener-
Musikhogskole. ally. Oxford: Oxford University Press.
Lehmann, A. C. (1997b). Acquisition of exper- McPherson, G. E., & Davidson, J. W. (2002).
tise in music: Efficiency of deliberate practice Musical practice: Mother and child interactions
as a moderating variable in accounting for sub- during the first year of learning an instrument.
expert performance. In I. Deliege & J. A. Slo- Music Education Research, 4, 14115 6.
boda (Eds.), Perception and cognition of music McPherson, G. E., & Zimmerman, B. J.
(pp. 161190). London: Psychology Press. (2002). Self-regulation of musical learning:
Lehmann, A. C. (2002). Effort and enjoyment in A social cognitive perspective. In R. Col-
deliberate practice: A research note. In I. M. well & C. Richardson (Eds.), The new hand-
Hanken, S. G. Nielsen, & M. Nerland (Eds.), book of research on music teaching and learning
Research in and for music education. Festschrift (pp. 3 273 47). New York: Oxford University
for Harald Jrgensen (pp. 15 3 166). Oslo: Nor- Press.
wegian Academy of Music. Miller, L. K. (1999). The savant syndrome: Intel-
Lehmann, A. C., & Ericsson, K. A. (1996). Music lectual impairment and exceptional skill. Psy-
performance without preparation: Structure chological Bulletin, 12 5 , 3 146.
and acquisition of expert sight-reading. Psy- Munte, T. F., Altenmuller, E. O., & Jancke, L.
chomusicology, 15 , 129. (2002). The musicians brain as a model of
470 the cambridge handbook of expertise and expert performance
neuroplasticity. Nature Reviews: Neuroscience, ent in young people (pp. 1967). New York:
3 , 473 478. Ballantine.
Munzer, S., Berti, S., & Pechmann, T. (2002). Sudnow, D. (1993 ). Ways of the hand: The organ-
Encoding timbre, speech, and tones: Musicians isation of improvised conduct. London: Rout-
vs. Non-musicians. Psychologische Beitrage, 44, ledge & Kegan Paul.
187202. Sundberg, J. (1987). The science of the singing voice.
Palmer, C. (1997). Music performance. Annual De Kalb, IL: Northern Illinois University Press.
Review of Psychology, 48, 115 13 8. Vitouch, O. (2005 ). Erwerb musikalischer Exper-
Pantev, C., Roberts, L. E., Schulz, M. Engelien, tise. [Acquisition of musical expertise] In
A., & Ross, B. (2001). Timbre-specific enhance- T. H. Stoffer & R. Oerter (Eds.), Allgemeine
ments of auditory cortical representations in Musikpsychologie (pp. 65 7715 ). Gottingen:
musicians. Neuroreport 12 , 169174. Hogrefe.
Ragert, P., Schmidt, A., Altenmuller, E. O., & Wagner, C. (1988). The pianists hand: Anthro-
Dinse, R. (2004). Superior tactile performance pometry and biomechanics. Ergonomics, 3 1, 97
and learning in professional pianists: Evidence 13 1.
for metaplasticity in musicians. European Jour- Weisberg, R. W. (1999). Creativity and knowl-
nal of Neuroscience, 19, 473 478. edge. In R. J. Sternberg (Ed.), Handbook of cre-
Rauscher, F. H., & Hinton, S. C. (2003 ). Type ativity (pp. 22625 0). Cambridge: Cambridge
of music training selectively influences percep- University Press.
tual processing. In R. Kopiez, A. Lehmann, I. Wellek, A. (1970). Typologie der Musikbegabung
Wolther, & C. Wolf (Eds.), Proceedings of the im deutschen Volke [Typology of musical abil-
5 th Triennial Conference of the European Society ities in the German people]. Munchen: Beck.
for the Cognitive Sciences of Music (pp. 8992). (Original published 193 9.)
Hannover: University of Music and Drama. Williamon, A. (Ed.). (2004). Musical excellence:
Renwick, J., & McPherson, G. E. (2002). Interest Strategies and techniques to enhance perfor-
and choice: Student-selected repertoire and its mance. Oxford: Oxford University Press.
effect on practising behavior. British Journal of Williamon, A., & Valentine, E. (2000). Quantity
Music Education, 19, 173 188. and quality of musical practice as predictors of
Seashore (193 8/1967). The psychology of music. performance quality. British Journal of Psychol-
New York: Dover. ogy, 91, 3 5 3 3 76.
Shuter-Dyson, R. (1999). Musical ability. In Winner, E. (1996). The rage to master: The deci-
D. Deutsch (Ed.), The psychology of music sive role of talent in the visual arts. In K. A.
(2nd ed., pp. 62765 2). San Diego: Academic Ericsson (Ed.), The road to excellence (pp. 271
Press. 3 02). Mahwah, NJ: Erlbaum.
Sloboda, J. A., Davidson, J. W., Howe, M. J. A., & Woody, R. H. (1999). The relationship between
Moore, D. G. (1996). The role of practice in the advanced musicians explicit planning and their
development of performing musicians. British expressive performance of dynamic variations
Journal of Psychology, 87 , 2873 09. in an aural modeling task. Journal of Research
Sloboda, J. A., Hermelin, B., & OConnor, N. in Music Education, 47 , 3 3 13 42.
(1985 ). An exceptional musical memory. Music Woody, R. H. (2003 ). Explaining expressive
Perception, 3 , 15 5 170. performance: Component cognitive skills in
Sosniak, L. A. (1985 ). Learning to be a concert an aural modeling task. Journal of Research in
pianist. In B. S. Bloom (Ed.), Developing tal- Music Education, 5 1(1), 5 163 .
C H A P T E R 27
generalized theory of expert performance. that have arisen, and the relatively short
The goal now shifted from merely demon- length of this chapter, our focus is to provide
strating expert-novice differences in a sport the reader with a brief overview of key issues
to developing laboratory tasks to elucidate and methods, as well as the major findings to
the underlying mechanisms that afford date from a largely cognitive perspective.
consistent expert performance. As a result The chapter is comprised of three main
many studies over the past 15 years have sections. In the first section, we introduce
demonstrated concerted efforts to examine some unique issues that need to be consid-
the underlying mechanisms of expert per- ered when studying sport performance. The
formance in sport (see Starkes & Ericsson, second section begins with a review of the
2003 ). historical roots of this area of research. In
During this time Ericsson, Krampe, and this second section, we first present the lit-
Tesch- Romers (1993 ) model of deliber- erature from a cognitive perspective, outlin-
ate practice also had significant impact ing the different research paradigms such as
on research in sport. Over the past 12 anticipation, the identification of perceptual
years research on this model has been con- features, recall and recognition, and deci-
ducted in soccer, wrestling, figure skating, sion making. This is followed by a discus-
triathlon, swimming, netball, volleyball, and sion of perceptual training as one means of
basketball (see Ward, Hodges, & Starkes, improving performance. A competing the-
2004). The deliberate-practice model has oretical perspective is then presented with
been examined more often in sport than discussion of ecological psychology and the
in any other domain to date. See Ericsson, idea that perception is educated. The sec-
Chapter 3 8. ond section concludes with presentation of
The last few years have seen the emer- research on the influence of practice for
gence of different paradigms in what we sport expertise. In the final section of the
see as the third and most recent phase chapter we discuss and evaluate the first
in the development of research on expert meta-analysis of sport-expertise research
performance in sport. Ecological psychol- (Thomas, Gallagher, & Lowry, 2003 ).
ogy, dynamical systems theory, and associ-
ated techniques are expected to play a more
important role in the future in our under- Unique Features of Sport
standing of performance in sport (see Beek, as a Performance Area
Jacobs, Daffertshofer, & Huys, 2003 ; Huys,
Daffertshofer, & Beek, 2004). At present Sport performance demands proficiency in
there are only a few studies available on tasks that involve movement with severe
expert performance using these techniques. time constraints, and very often interaction
A major advantage of these paradigms is that with moving objects and opponents. Though
they view perception and action as inextri- sports differ from what are commonly per-
cably linked and emphasize the continuous, ceived as more cognitive tasks such as bridge
time-dependent (emerging) nature of sport- and chess playing, there are also tremen-
ing activities. This focus on movement (not dous differences among sports. Witness the
just cognition) as integral to performance is difference between a relatively slow-paced
particularly appealing when one considers introspective game of golf and the fast-paced
the level of movement skill inherent in world interceptive games of tennis or basketball.
class sport performances. As well, the unique combination of req-
Since the vast majority of existing uisite cognitive skill with movement skills
research on the topic of expertise in sport has may result in mismatches in the develop-
been approached from a cognitive perspec- ment of each area. For example, a young
tive, that is the primary focus of this chap- second baseman in baseball may understand
ter. Given the volume of cognitive research the necessity to throw a ball to cut off a
available on sport, the complexity of issues runner to home, but simply not be able to
expert performance in sport 473
make that throw (Nevett & French, 1997). closed sports), often characterized as those
Finally, sport demands may differ depend- in which athletes react to the movements
ing on the role occupied by the performer of their opponents, the timing of action is
as a player within a team, as a coach, or critical to success. Not only do perform-
referee. For example, anticipating a players ers have to deal with deciding when to per-
next offensive move is an important skill form a skill, the actual execution of the skill
for a player or a referee (who needs to get also has time constraints. For example, an
ahead of the play), but not very relevant to a ice hockey player may be presented with
coach. Adjusting a teams defensive strategy an opportunity to score when the oppos-
to deal with an opponents offensive struc- ing teams goalie is temporarily out of posi-
ture is most relevant to a coach and player tion. However, the shooter has a time win-
but inconsequential for a referee. Ones role dow of only milliseconds in which to select,
in a sport is an important factor in deter- prepare, and execute a successful shot. Once
mining the nature of skills that are critical. the hockey player has decided to take a shot,
Likewise, ones role is often quite different in the action must follow instantaneously. The
individual sports versus team sports. A foot- window of opportunity is wasted if the goalie
ball quarterbacks role and skills are quite is given the chance to prevent the shot, or
different from those required of a linebacker. the shooters intentions are telegraphed to
the goalie by a slow windup. One reason for
movement this pressure is the systemic lag time inter-
vening between an event and a decision to
The most salient feature of sport is the
move (i.e., reaction time) and between this
central role that movement plays. Athletes
decision and its actual initiation and com-
perform movements that vary from the
pletion (i.e., movement time). This pres-
seemingly simple, such as running, to the
sure on movement choice (response selec-
extremely complex such as a gymnastics
tion) and completion (response execution)
bar routine. Moreover, many sports involve
is illustrated most clearly in fast ball sports.
the coordination of movements between
In sports such as tennis, squash, and base-
two or more athletes. This is the case in
ball there is little or no time for a lag
sports such as pairs figure skating, row-
between movement choice and movement.
ing, or team synchronized swimming. Else-
In these sports, movement decisions must
where, a distinction has been made between
often be made based on early and incom-
interactive sports (basketball, soccer, ice
plete information. For example, a baseball
hockey) that involve many interdependen-
batter facing a 90-mph fastball must decide
cies between players, and coactive sports
whether to swing or not before the ball has
(bowling, archery, golf) that are performed
even left the pitchers hand. In this scenario,
independently (Cratty, 1983 ). This distinc-
there are constraints imposed by movement
tion has been useful in determining the
choice, movement timing, and coincidence
skill requirements and relative demands for
anticipation.
communication that a sport presents. Some
sports (rowing, swimming, track relays),
different abilities develop
however, demonstrate characteristics and at different rates
thus demands of both (see Eccles & Tenen-
The combination of movement demands,
baum, in press, for a review).
time constraints, and interaction with
moving objects creates a need for both
time constraints perceptual-cognitive and perceptual-motor
A critically important aspect that must be skill for high-level sport performance (e.g.,
taken into consideration in sport is the limit Starkes, Cullen, & MacMahon, 2004). Thus,
on performance imposed by inherent time an athletes performance level depends on
constraints in a game. For open sports (see the development of these two interactive
Poulton, 195 7, for a discussion of open vs. types of skills. This creates yet another
474 the cambridge handbook of expertise and expert performance
unique situation in sport where a performer ple or cooperating individuals these must be
may have mismatches in the level of devel- coordinated or integrated to achieve the best
opment of these two forms of skill. For performance. Thus, the team must not only
example, an athlete may know what to perform the task itself, but coordinate mem-
do, but not how to do it (French, Nevett, bers actions toward the task. In order to
Spurgeon, Graham, Rink, & McPherson, do so all team members must have shared
1996; Nevett & French, 1997). This mis- mental models about the necessary behav-
match can result in either poor motor per- iors of the team and its members, and have
formance or a movement choice that is less shared expectations. In addition there must
than optimal. be a general team knowledge of operations
that is shared by everyone, yet more specific
differing roles individual knowledge appropriate to certain
positions or roles is also necessary. Although
Although it has not received a great deal
sport performance shares many similarities
of attention within the literature, another
with industrial and military applications, to
unique aspect to sport is that individuals play
date the research on teams as process units is
different roles. Within a sport, the requisite
minimal. See Satas, Rosen, Burke, Goodwin,
skills for a soccer goalkeeper are quite dif-
and Fiore, Chapter 25 , for a review.
ferent from those of a forward. Within track
and field, a female sprinter has a very dif-
ferent skill set than a female javelin thrower. Historical Roots of the Expertise
Likewise, ones role within a sport may differ Approach in Sports
such that the requisite skills for a coach (e.g.,
Cote et al., 1995 ; Salmela & Moraes, 2003 )
One of the earliest studies of perceptual-
differ from those of judges and referees (e.g.,
motor expertise was that of Bryan and
Ste-Marie, 2003 ). Wheras a coach, athlete,
Harter (1897; 1899) in their now-classic
and referee may operate within the same
investigation of telegraphic skill (see Lee &
sub-domain or specific sport, there is evi-
Swinnen, 1993 ). The sending and receiving
dence that the different task demands result
of messages via Morse code required correct
in different skills and abilities (e.g., Allard,
production in the timing of signals and cor-
Deakin, Parker, & Rodgers, 1993 ; Williams
rect translation of incoming messages. Bryan
& Davids, 1995 ). To date, however, the vast
and Harter observed that experienced oper-
majority of the literature on expert perfor-
ators were more accurate and consistent in
mance in sport has dealt exclusively with
their productions than the less experienced
athletes. A second issue creates problems
and showed qualitatively different strategies
for any empirical test of role. Most often,
in receiving messages, delaying the copying
coaches and referees begin their career as
of the message until some idea of content
athletes and thus have subsumed various
and meaning was conveyed. In comparison,
roles throughout their athletic career.
the novice operators copied messages letter
by letter.
teams are more than a group of individuals Although Bryan and Harter did not
One aspect of team sports often ignored specifically use the term chunking to dis-
is that teams require successful processes cuss their findings (see Miller, 195 6), they
of coordination and communication well were one of the first to experimentally show
beyond the skills required of individuals. that skill-based differences were a result of
Yet it is only recently that teams have the (re)organisation of small units of infor-
been examined from the same social cog- mation, such as letters, into larger units, such
nition approach that is common in indus- as words and phrases. Perceptual chunking
trial and organizational psychology. Eccles ideas have been at the forefront of explana-
and Tenenbaum (2004) suggest that because tions for expert-novice differences in purely
team operations are performed by multi- cognitive domains (Chase & Simon, 1973 ;
expert performance in sport 475
Ericsson & Polson, 1988) and also in sport The cognitive nature of the expert advan-
(see, Starkes et al., 2000; Starkes, Cullen, tage in sport
& MacMahon, 2004; Tenenbaum & Bar-
Eli, 1993 ; Williams, Davids, & Williams,
1999). anticipation and decision making
The ability to quickly and efficiently pro- The exploitation of advance information
cess domain-specific information has since through highly developed internal stores and
been shown to be one of the defining effective organization of the motor system
features of expertise in sport, and hence has been proposed to underlie fast behav-
explanations for skill-based differences in ioral responses in the environment. This
motor skills have been heavily grounded in results in what Abernethy describes as the
cognitively-based theories of information- time paradox wherein skilled performers
processing activities (e.g., Fitts, 1965 ; Fitts operating under extreme time constraints
& Posner, 1967; Schneider & Shiffrin, 1977). appear to have all the time in the world
Accordingly, the performer was seen as (Abernethy, 1991). Recognition of familiar
an intelligent receiver and translator of scenarios and the chunking of perceptual
information resulting in various degrees of information into meaningful wholes and pat-
effective output or behavior. Time delays terns speeds up processes related to stimulus
between a stimulus and a response, that is, identification. Processing activities associ-
the RT (reaction time) interval, provided a ated with response selection can be reduced
critical index of processing efficiency and via knowledge and experience of previous
skill. Three somewhat independent process- stimulus-response situations and hence situ-
ing activities have been proposed to medi- ational probabilities (e.g., Alain & Proteau,
ate the reception of information and motor 1979, 1980; Nougier, Ripoll, & Stein, 1989;
behavior: stimulus identification, response Ward & Williams, 2003 ). Finally, as motor
selection, and response programming (see learning improves, processes associated with
Schmidt & Lee, 1998), such that skilled per- motor programming become more efficient
formance has been interpreted at a spe- and the degree of programming necessary for
cific level in terms of the type of pro- motor-skill execution is reduced. In this way
cessing activities engaged in at these vari- the whole action is merely parameterized
ous stages (e.g., Tenenbaum, 2003 ). At a or tuned based on prior experiences rather
more general level, differences in the pro- than constructed in terms of its individual
cessing activities of skilled and less-skilled components (see Schmidt, 1975 ; Schmidt &
performers in sport have been described on Lee, 1998).
the basis of the verbally-mediated, cogni- In tennis (Goulet et al., 1988, 1989) and
tive, and conscious-awareness nature of the baseball (Paull & Glencross, 1997) the expert
performance (e.g., Adams, 1971; Anderson, advantage has been evidenced through supe-
1983 ; Fitts & Posner, 1967). These exper- rior decisions and RTs in response to unfold-
imental findings, protocols, and theoreti- ing game scenarios. In cricket, squash, and
cal approaches have been the foundation badminton, Abernethy and Russell (1984)
of much of the laboratory-based experi- and Abernethy (1988, 1990) showed that
ments designed to examine the mecha- skilled performers made more accurate deci-
nisms responsible for the expert advantage in sions concerning stroke selection in cricket
sport, including the nature of the knowledge and shot direction in badminton and squash.
structures (e.g., McPherson, 1993 b) and con- Professional goalkeepers in soccer were bet-
trol processes (e.g., Beilock & Carr, 2001) ter able and faster at predicting shot-loca-
underpinning expert performance. Some of tion (Savelsburgh, Williams, van der Kamp,
the classic findings in this area will be pre- & Ward, 2002). Even as task complexity
sented next. See the reviews by Proctor increases, as with the 11 versus 11 scenario
and Vu, Chapter 15 , and Rosenbaum et al., in soccer, Williams et al. (1994) showed that
Chapter 29, for more detail. skilled soccer players were faster and more
476 the cambridge handbook of expertise and expert performance
accurate at verbalizing the future destination display, the decision accuracy of the experts
of the ball (see also Helsen & Pauwels, 1993 ). decreased to a level below that exhibited by
the novice performers, showing the racquet
identifying perceptual structure through to be a critical feature underlying the expert
occlusion studies and visual search advantage.
There have been a variety of methods A number of researchers have combined
employed in sport to examine the spe- eye-movement recording techniques with
cific nature of the information underlying temporal- or spatial-occlusion studies to gain
the expert advantage in both speed and a more precise picture of the nature of
accuracy (faster and more accurate deci- visual cues underlying the decision pro-
sions/responses). One of the most common cesses of experts (e.g., Helsen & Starkes,
methods is that of occlusion as first opera- 1999; Goulet et al., 1989; Savelsburgh et al.,
tionalized by Abernethy and Russell (1984). 2002). Goulet et al. (1989) found that
They determined that information could eye movements preceding decisions were
be occluded either temporally (by specific focused on the shoulder and trunk area for
periods of time in relation to ball contact skilled performers in tennis in comparison
for example) or spatially (via the removal to the head for the less-skilled perform-
of specific features or events within a dis- ers (see also Singer, Cauraugh, Chen, Stein-
play) (see Abernethy, 1988; Abernethy & berg, & Frehlich, 1996). Subsequent tempo-
Russell, 1984, 1987; Williams, et al., 1999; ral occlusion in a second experiment showed
Williams, Ward, & Smeeton, 2004) With that high accuracy levels could be main-
respect to temporal-occlusion studies, it has tained for the skilled performers even under
been shown that the expert advantage is situations where information was available
most clearly observed when a structured only from the preparatory stage of the
game clip, for example, in tennis (e.g., movement.
Goulet et al., 1988, 1989) or in goal-keeping One of the benefits of eye-movement
(e.g., Savelsburgh et al., 2002), is edited data is that a dynamic picture of the visual
prior to ball-racquet or ball-foot contact. search patterns of skilled performers is pro-
In these situations, experts are able to use vided as the action unfolds (Ward, Williams,
advance visual cues to predict shot-type and & Bennett, 2002). It has generally been
direction, whereas the less-skilful players do shown that skilled performers show rela-
not have this perceptual skill at their dis- tively fewer fixations than novice performers
posal. In this way, temporal-occlusion stud- (e.g., Abernethy, 1985 ), and that fixations are
ies help to elucidate generally on the type of qualitatively different, with experts directed
information used by skilled players to facil- to areas of the display that are believed
itate decision making. These findings have to be most informationally rich. A reduc-
also held up in real-world occlusion tasks tion in the number of fixations for skilled
where portions of a volleyball serve have rather than less-skilled performers is in keep-
been occluded for the service receiver on ing with proposals that experts extract more
a volleyball court (Starkes, Edwards, Dis- information from one fixation than novices
sanayake, & Dunn, 1995 ). Skilled volley- because of mechanisms of chunking (see
ball players extract more information from also Ripoll, Kerlizin, Stein, & Reine, 1995 ;
advance visual cues and are better able to Allard & Starkes, 1993 ). More recently the
predict the landing position of a serve. smaller number of fixations by experts has
More-specific information can be gleaned been linked to the idea of a perceptual pivot
through removal of various features of (see Huys & Beek, 2002; Williams & Davids,
the display, and this is often accomplished 1998; Williams & Elliott, 1999), where eye
through spatial occlusion of certain ele- position is anchored, thus enabling a wide
ments. For example, Abernethy and Russell field of search of peripheral features; and
(1987) showed that in badminton, when the second, to periods of quiet eye associated
arm and/or racquet was occluded during a with movement preparation and a reduction
expert performance in sport 477
of variability in the motor system (Vickers, ommended using this technique to gain a
1996). further understanding of the control pro-
However, perhaps not surprisingly, visual cesses underlying expert performance. In
search patterns can be relatively domain summary, if the question is to determine the
specific. Williams, Davids, Burwitz, and nature of the structural information underly-
Williams (1994) showed that in 11 versus ing the expert advantage, especially in high-
11 soccer situations expert players showed dimensional, team sports, a combination of
more fixations (i.e., an increased search rate) the above techniques is believed to help
than the less-skilled players, focusing on understand what information is processed,
peripheral aspects of the display, including in addition to how.
the position of other players, in comparison
to novice performers, who tended to track perceptual training
the ball. Helsen and Pauwels (1993 ) also
From a practical standpoint, there has been
showed a difference in visual search patterns
considerable interest in the implications of
depending on the defensive or offensive
the above findings for the training of per-
nature of the decision (for a detailed review
ceptual skill (see Williams & Ward, 2003 ;
of the eye-movement data see Cauraugh &
Williams et al., 2004). Williams, Ward,
Janelle, 2002; Williams et al., 1999; Williams,
Knowles, and Smeeton (2002) successfully
2002).
improved the response time of tennis play-
Although eye-movement fixations are
ers, both in the laboratory and in the field,
potentially useful sources of information for
after perceptual-skills training interventions.
determining the critical features underly-
Although this research has important impli-
ing expert decisions, the validity of these
cations for training and improving tactically-
methods has been questioned and the com-
based decisions, the need to maintain the
bination of multiple techniques to under-
perception-action links during training has
stand which cues afford the expert advan-
repeatedly been emphasized, in as much as
tage has been recommended (e.g., Williams
the effectiveness of a decision is dependent
et al., 1999). For example, Ward, Williams,
on the associated accuracy and speed of the
and Bennett (2002) removed all contextual
execution (i.e., motor efficiency). The need
cues from a tennis opponent by converting
to maintain a degree of flexibility in the
major joint centers into point light sources
nature of the perceptual information afford-
to determine whether the expert advantage
ing action has also been recommended, at
in tennis was dependent on these cues. In
the expense of specific, prescriptive instruc-
keeping with previous literature, the rela-
tion methods and techniques that fail to
tive motion information of the major joint
provide sufficient variation in practice (see
centers displayed by the point lights pro-
Beek et al., 2003 ). It is important to note
vided enough information to show differ-
that to date the effects of perceptual train-
ences as a function of skill, and visual search
ing have been assessed only in terms of
behaviors did not differ across normal video
immediate performance improvement and
and point-light displays. The conversion of
short-term retention. The long-term reten-
data into minimal directional units of x and
tion of performance improvements as a
y coordinates affords easy editing and also
result of perceptual training has not been
enables the application of statistical data-
determined.
reduction techniques, such as principal com-
ponent analysis to aid in uncovering the
common variance between the players and recall and recognition
the ball. This analysis technique has been Recall and recognition paradigms, in
used by Post, Daffertshofer, and Beek (2000) addition to verbal reports, have alerted
to examine change over time in the acquisi- researchers to the memory structures and
tion of three-ball juggling, and Beek, Jacobs, processing strategies underlying skilled
Daffertshofer, and Huys (2003 ) have rec- performance (i.e., anticipation and decision
478 the cambridge handbook of expertise and expert performance
making) in sport. These methods were & Starkes, 1999; Starkes, 1987; Ward &
based on the work of de Groot (1978), Williams, 2003 ). Indeed, Reilly, Williams,
Charness (1976, 1979), and Chase and Nevill, and Franks (2000) showed that
Simon (1973 ) in chess. Chase and Simon anticipatory skill, related to domain-specific
(1973 ) showed that experts in chess could experience, was one of four important pre-
be differentiated on their fast and accurate dictors of skill level among teenage soc-
ability to perceive and recognize structured cer players, in addition to speed and agility.
patterns of play, but not random placement Although physical skills quite understand-
of chess pieces. This finding highlighted the ably differentiate across skill in sport, fac-
importance of domain-specific experience, tors related to perceptual skill and cognitive
rather than natural abilities associated development are at least equally important.
with IQ and superior memory in general, Ward and Williams (2003 ) have shown also
underlying expertise. In sport, the historical that these cognitive skills are acquired some-
and common attribution of performance to what irrespective of cognitive development
indices associated with talent, rather than and maturational age (see also Abernethy,
experiential factors, meant that paradigms 1988). The degree of modification of cogni-
similar to that of Chase and Simons could tive abilities with practice has been the focus
illuminate on the various contributions of of much of the current expertise research in
domain-specific skills acquired as a result sports, as we detail later.
of practice. The ability to chunk information into
Even though sports are often not per- meaningful wholes or patterns of tacti-
ceived as being as highly cognitive in nature cal significance, via detailed and extensive
as chess, researchers have shown that strate- task-specific memory structures, has been
gic differences related to domain-specific proposed to underlie early and accurate
knowledge structures, which have been decision-making performance in sports (see
coined software features, rather than phys- Tenenbaum, 2003 ; Williams, et al. 1999),
ical hardware features, consistently dif- rather than being merely a consequence of
ferentiate across skill in sport (Starkes & experience with the game. For example,
Deakin, 1984). Recall of structured game Williams and Davids (1995 ) showed that
sequences is better for high-level rather only skilled players, not physically disabled
than low-level performers, across a vari- spectators matched for perceptual experi-
ety of sports and with a variety of medi- ence, showed superior recall on game struc-
ums (e.g., Abernethy, Neal, & Koning, 1994; tured scenarios. Likewise, Allard, Deakin,
Allard, Graham & Paarsalu, 1980; Borgeaud Parker, and Rodgers (1993 ) showed that
& Abernethy, 1987; Garland & Barry, 1991; coaches, athletes, and referees in basketball
Helsen & Pauwels, 1993 ; Nakagawa, 1982; were differentially more skilled on those
Starkes, 1987; Starkes & Deakin, 1984; cognitive tasks that more directly tapped
Williams, Davids, Burwitz, & Williams, their role in the sport (i.e., referees were
1993 ; Williams & Davids, 1995 ); inciden- superior on tasks related to recognition and
tal recognition tests of previously viewed naming of violations, coaches better at recog-
structured game plays are improved for nizing schematic plays, etc.). Nevertheless,
skilled rather than less-skilled performers anecdotal information from coaches and cer-
(e.g., Allard et al., 1980; Garland & Barry, tain other researchers (see Smeeton, Ward,
1991; Williams & Davids, 1995 ; Williams & Williams, 2004) suggests that some trans-
et al., 1993 ); and domain-specific percep- fer of perceptual skill is seen across sports
tual tests reliably differentiate expert and with similar perceptual demands. Finally,
novice performers, in comparison to percep- high correlations between decision accuracy
tual tests associated with physical features, and recall (e.g., Helsen & Starkes, 1999) sug-
such as static and dynamic visual acuity, sim- gest that the memory structures associated
ple visual RT, and central-peripheral aware- with each underpin the expert perceptual-
ness (e.g., Abernethy et al., 1994; Helsen cognitive advantage in sport.
expert performance in sport 479
the nature of knowledge and control the manipulation of thought. Although this
structures issue has been addressed in the work of
Performance differences between experts McPherson and colleagues (e.g., McPherson
and novices for cognitive and perceptual- & Thomas, 1989), whereby response selec-
motor skills have traditionally been tion has been differentiated from execution,
explained using terminology derived from the highly cognitive-based level of expla-
Andersons ACT (Active control of thought) nation for these expertise effects might be
or production system theory of skill acqui- over-stretched.
sition (1982, 1983 ; see also French & Production-based terminology and ideas
Thomas, 1987; McPherson & French, 1991; have also been incorporated in recent
McPherson, 2000; McPherson & Kernodle, research designed to examine notions of
2003 ; Starkes & Allard, 1991). Accord- automaticity. The procedural or automatic
ingly, early in practice, declarative rules stage of performing, whereby responses are
or knowledge structures underlie the slow executed without the need for problem solv-
and effortful decision making of novice ing and complex decisions, has been exam-
performers. With practice, these rules ined in relation to the nature of the control
become compiled into efficient produc- processes underlying skilled performance in
tions, such that certain conditions evoke golf and soccer (see Beilock et al., 2003 ;
actions without the necessity for interven- Beilock & Carr, 2001, for reviews). Beilock
ing processes associated with the bringing and colleagues have shown that manipu-
to mind of domain-specific, verbalizable lations, designed to encourage attention to
knowledge. Based primarily on analysis of aspects of performance that are believed
verbal protocols (see Ericsson & Simon, to have become proceduralized (e.g., the
1993 ) of children and adults in sports such as dominant foot in soccer dribbling), cause
tennis and basketball, McPherson, French, decrements in performance in skilled ath-
Thomas, and colleagues have shown how letes. These authors have argued that a con-
factual knowledge (i.e., what to do) devel- trol focus, characteristic of an earlier, less-
ops along with procedural knowledge (i.e., effective and more-declarative level of per-
how to do it) and that skilled performers formance, interferes with the procedural
plans, based on verbalizable goals, become skill and hence control structures governing
transformed into more specific if-then-do performance at high levels of skill. Indeed,
or condition-action rules. These rules are whereas skilled performers are affected by
then refined and improved as a function of this focus, novice performers are not.
extended practice, such that they become The interpretation of these effects in
specific to the task and more tactical in production-system terminology, however,
nature as skill improves. has been questioned. Perkins-Ceccato, Pass-
If knowledge does indeed become pro- more, and Lee (2003 ) showed interactions
ceduralized as a function of skill and of skill level with attentional focus in a golf
is thus supposedly non-verbalizable, then putting task, whereby skilled golfers became
issues are raised with respect to the valid- more variable in their performance under
ity of verbal reports for skilled performers focus manipulations designed to encourage
(see Abernethy, 1993 , 1994). Additionally, attention to their arms and the swing, but
whereas production-based terminology was not to the club, whereas the reverse was true
formulated based on observations of change for the novice performers. Rather than pro-
in cognitive-domains (such as learning a ceduralization of knowledge being respon-
computer language), in motor skills, produc- sible for these skill by attentional-focus
tions are no longer merely linked to the deci- interactions, another explanation might be
sion side but are highly dependent on exe- found from theories and approaches that
cution. Therefore, motor-based productions place less emphasis on cognitive processes
are likely to be of quite a different nature and more on self-organizing principles oper-
to the internal representations governing ating within the motor system. Bernstein
480 the cambridge handbook of expertise and expert performance
(1967/1996) originally discussed the motor that is important, rather than the general
system in terms of distinct levels of con- amount of variability per se (see Huys
trol that interact on many levels but change et al., 2004). The effective harnessing of
in their control function as practice pro- variables that are non-functional is a feature
gresses and skill develops. Accordingly, skill of expert performance within the motor
and associated notions of automaticity are system generally, not just with the external
linked to the devolvement of control pro- environment. In cycling, Bernasconi and
cesses to lower levels of the action sys- Kohl (1993 ) showed that distinct couplings
tem (e.g., muscular-articular synergies) that emerged between respiration and cycle
can operate somewhat independently from rate, which enabled the skilled performer
higher levels (e.g., action plans) and that do more economical and effective control of
not require cognitive involvement for effi- their movements. In runners, Diedrich and
cient and effective results. Breakdowns in Warren (1995 ) observed no correlation
performance, therefore, may be a result of between step cycle and breathing for novice
inappropriate levels of the motor system runners, but in expert runners there were
taking control of the movement (see also specific respiration/step ratios (1:4, 1:3 , 2:5 ,
Beek, 2000). 2:3 , 1:1) that were dependent on the tempo
and incline of running. Similar, within-
cognitive structures or the education system couplings have also been observed
of perception? in juggling between postural sway and arm
Highly cognitive information-processing movements (Huys et al., 2004). As with
explanations for the expert advantage cycling and running these system couplings
in sport, which rely heavily on internal afford greater efficiency and economy in
representations and cognitive processes performance.
mediating stimulus interpretation and
action choice, have also been questioned
by researchers influenced by the work of the importance of practice
Gibson (1979) in ecological psychology (see Whereas explanations for the nature of the
Beek et al., 2003 ; Huys, Daffertshofer, & expert advantage in sport might differ with
Beek, 2004). According to these researchers, respect to the role of cognition, there is
learning and hence skilled performance is no disagreement across researchers as to
seen as a process of educating attention, the necessity of years of task-specific prac-
whereby specific sources of information, tice to acquire skilled performance. Research
which inform action, are identified and in support of this viewpoint was originally
functionally coupled to movement as skill detailed by Ericsson et al. (1993 ). Individual
progresses. Following from observations of differences across different levels of perfor-
extremely fast modifications to responses on mance skill were shown to be closely related
the basis of vision during table-tennis and to deliberate practice hours and hence led
the long jump, for example (Bootsma & van the authors to conclude that many charac-
Wieringen, 1990; Lee, Lishman, & Thomson, teristics once believed to reflect innate tal-
1982), explanations for skilled performance ent are actually the result of intense prac-
were based on the tight couplings between tice extended over at least 10 years (p. 3 63 )
perception and action (i.e., compensatory and that the role of heritability in attainment
variability), rather than a reduction or of high levels of skill might be limited to
change in processing activities. Although a motivational factors. If one looks across the
reduction in the variability in movements range of sport studies on practice or deliber-
is a common distinguishing characteristic ate practice there is a correlation (sometimes
of experts when compared to intermediate high, sometimes low) between the amount
performers, it is the qualitative nature of of all types of reported practice and perfor-
this variability in relation to the task goal mance. This is consistent with the proposed
expert performance in sport 481
relation between deliberate practice, albeit theory. The issue of physical versus devel-
not equivalent (Ericsson, 2003 ). opmental causes underlying performance
Perhaps primarily because of enduring differences in sport clearly requires further
beliefs about the role of prior talent and analysis before more definitive conclu-
abilities in sport, this proposal has received sions can be drawn. Even though Hodges
vibrant interest from researchers working et al. (2004) have presented evidence that
within the sports field. Across a number gender differences persist when practice-
of sports, ranging from figure skating to related variables are controlled, there is
wrestling, from hockey to karate, sport- evidence that in other sports, particularly
specific practice has been shown to be a those that are less physically demand-
significant predictor of skill-based differ- ing (e.g., archery and bowling), gender
ences in sport (see Starkes, Deakin, Allard, differences are negligible once experience
Hodges, & Hayes, 1996; Ward, Hodges, & is controlled (e.g., Thomas, Schlinker, &
Starkes, 2004, for reviews). Though these Over, 1996). Similarly, Duffy, Baluch, and
results are encouraging for the theory and Ericsson (Submitted) failed to find that
speak to the role of practice in modify- differences in physical stature (i.e., reach)
ing physical attributes and skills, there have for darts players differentiated across
been reasons to question the ubiquitousness skill level. In their analyses of professional
of these findings, particularly when perfor- and amateur, male and female dart throwers
mance and practice differences are exam- Duffy, Baluch, and Ericsson (2004) found
ined at an individual level. that the magnitude of gender differences
Hodges, Kerr, Starkes, Weir, and Nanan- in darts performance remained the same
diou (2004) showed that the amount between professional and amateur throwers,
of variance that could be explained by thus explanations other than practice must
estimates related to practice was domain account for this (p. 243 ).
specific in swimming and triathlon, such Age-based interactions with practice have
that the distance of the event mediated the also been proposed to mediate skill perfor-
amount of variance in performance times mance such that Cote and colleagues (e.g.,
that could be explained by practice. In the Cote & Hay, 2002; Baker, Cote, &
100 m and 200 m sprint events in swimming, Abernethy, 2003 ) argue that it is only after a
only approximately 20% of the variance in decision to specialize (around 12 years of age)
times could be explained by practice-related that sport-specific practice becomes a crit-
variables, whereas gender accounted for ical component and predictor of exper-
approximately 40%. As for the 400 m, tise. Before this period, diversity in physical
1.5 km, and triathlon event comprising experience and play are presumed to
swimming, cycling, and running, these be important for later skill development.
findings were reversed. Gender no longer Despite these proposals, Hodges et al.
played an important role in accounting for (2004) showed a monotonic increase in the
performance differences, whereas practice yearly amounts of practice for competitive
accounted for approximately 3 5 to 40% of swimmers that was highly related to perfor-
the variance in performance times. It would mance times, even though their average start
seem that in the shorter, more-anaerobic age was around six to eight years (see also
events, physical factors related to height- Starkes et al., 1996; Ward, Hodges, & Starkes,
and-muscle and to body-fat ratios limit 2004). Interactions with development have
performance, somewhat independently also been explored as a function of aging.
of practice. This is the first empirical The main finding has been that age-related
evidence that non-practice specific factors declines in performance can be circum-
play significant roles in predicting expert vented by specific and sustained practice,
performance. It also appears to contradict as evidenced through longitudinal practice
strong interpretations of deliberate practice records in comparison to cross-sectional
482 the cambridge handbook of expertise and expert performance
comparisons (see Starkes, Weir, & Young, tional, and college-level athletes are better
2003 ; Young & Starkes, in press, for reviews). than developmental-level athletes or others.
Finally, greater ecological validity of the per-
ception/decision task produced higher effect
Meta-analysis of sport sizes for experts and novices, but it was a
expertise findings more important factor for higher levels of
performance. The more expert you are as
A recent landmark meta-analysis by a performer, the more important ecological
Thomas, Gallagher, and Lowry (2003 ) is the validity of the task becomes in assessing your
first to collate and examine sport-expertise performance.
research. The questions addressed were On the basis of this analysis a number of
whether experts and novices differ on recommendations can be made for future
perceptual versus decision-making aspects research. It is important that the research
of cognition, by level of expertise, by type of setting and technique focus on capturing the
sport (team vs. individual), across age levels, basis for superior performance, in terms of
according to levels of ecological validity either the process of acquisition or cognitive
of the test situation, according to levels of structure involved. It is very important in
internal validity, and by gender. In addition sport research to be specific and define the
the authors were interested in whether level of expertise/performance one is study-
the importance of the ecological validity ing, both in terms of years of experience
of the skill test varied by level of expertise and also in level of competition and perfor-
and whether the importance of perceptual mance attained. In terms of perceptual skill,
versus decision-making aspects of cognition adult high-performance athletes appear to
varied by type of sport. focus attention on specific informative areas
The authors located 66 published papers, of a game and as a result they exhibit bet-
of which 21 were eliminated because they ter recognition of game structure and better
did not include both expert and novice data, recall of game elements. In contrast, young
and another six studies lacked sufficient data athletes typically lack the knowledge to pro-
to calculate effect size. Thirty-nine studies duce quality solutions and are unable to reli-
published between 1987 and 2002, or 87% of ably separate relevant from irrelevant infor-
the literature available, were included. The mation. An interesting suggestion by the
meta-analysis included data on 1,112 experts authors is that those children who become
and 1,287 novices. Perceptual and cognitive experts at relatively young ages have clearly
skills were found to be equally important in derived more from practice than others. This
predicting expert performance in athletes. It is an important point worthy of future study.
appears that perception and decision mak- This chapter provides an overview of
ing are both critical aspects of individual the major approaches in sport expertise
sport; however, cognitive skill was slightly research. The complexity of this area is
more important for team-sport experts. The shown in the multiple research paradigms
same findings hold true regardless of gen- that have been applied, from visual search
der. In terms of design, adult expert-novice and decision making, to verbal protocol anal-
differences are typically larger than those ysis, and from an examination of the auto-
found in children and adolescents and per- matic nature of performance, to the role of
haps not unexpectedly level of experience cognition. We have also presented different
influences the extent of expert-novice dif- major theoretical approaches represented by
ferences. When the expert group is com- cognitive and ecological psychology, with
prised of international, national, or college- implications for skill training. The impor-
level athletes, differences across skill class are tance of practice to both of these approaches
more likely to be observed than when the is shown in the section reviewing research
elite group is comprised of developmental- on practice features in the acquisition of
level athletes, such that national, interna- sport expertise. Bringing all of this research
expert performance in sport 483
together, we discussed the meta-analysis of Abernethy, B., & Russell, D. G. (1987). Expert-
Thomas et al., (2003 ) which provides some novice differences in an applied selective atten-
directions for future research. Although the tion task. Journal of Sport Psychology, 9, 3 26
Thomas et al. meta-analysis demonstrates 3 45 .
great consistency in many of the findings Abernethy, B., Thomas, K. T., Thomas, J. T.
related to sport expertise, it is clear from (1993 ). Strategies for improving understanding
the issues outlined earlier that we have far of motor expertise (or mistakes we have made
and things we have learned!!!). In J. L. Starkes
to go in understanding skilled performance
& F. Allard (Eds.), Cognitive issues in motor
in sport. Certainly expert sport performance expertise (pp. 3 173 5 8). Amsterdam: North-
is not a unitary entity but a complex inter- Holland.
action of perception, decision-making, and
Abernethy, B. (1994). The nature of expertise in
movement skill as well as ones role and the sport. In S. Serpa, J. Alves, & V. Pataco (Eds.)
nature of the sport engaged in. The inherent International perspectives on sport and exercise
complexity of sport, coupled with its real- psychology (pp. 5 768). Morgantown, WV.: FIT
world application, is both an intriguing qual- Press.
ity and one that makes the domain ripe for Adams, J. A. (1971). A closed-loop theory of
research. motor learning. Journal of Motor Behavior, 3 ,
11115 0.
References Alain, C., & Proteau, L. (1979). Perception of
objective probabilities in motor performance.
In B. Kerr (Ed.) Human performance and behav-
Abernethy, B. (1985 ). Cue usage in open motor
ior (pp. 15 ) Alberta Canada: Banff.
skills: Areview of available procedures in Motor
Memory and Control. In D. G. Russell & B. Alain, C., & Proteau, L. (1980). Decision mak-
Abernethy (Eds.) The Otago Symposium (p. ing in sport. In C. H. Nadeau, W. R. Halliwell,
110122). Dunedin, NZ: Human Performance K. M. Newell, & G. C. Roberts (Eds.), Psychol-
Associates. ogy of motor behavior and sport (pp. 465 477).
Champaign, IL: Human Kinetics.
Abernethy, B. (1988). The effects of age and
expertise upon perceptual skill development in Allard, F., Deakin, J., Parker, S., & Rodgers, W.
a racquet sport. Research Quarterly for Exercise (1993 ). Declarative knowledge in skilled motor
and Sport, 5 9, 210221. performance: Byproduct or constituent? In J. L.
Starkes & F. Allard (Eds.), Cognitive issues
Abernethy, B. (1990). Expertise, visual search,
in motor expertise (pp. 95 107). Amsterdam:
and information pick-up in squash. Perception,
North-Holland.
19, 63 7.
Abernethy, B. (1991) Visual search strategies and Allard, F., Graham, S., & Paarsalu, M. L. (1980).
decision-making in sport. International Journal Perception in sport: Basketball. Journal of Sport
of Sport Psychology, 2 2 , 189210. Psychology, 2 , 1421.
Abernethy, B. (1993 ). Attention. In R. Singer, M. Allard, F., & Starkes, J. L. (1991). Motor-skill
Murphey, & L. Tennant (Eds.), Handbook of experts in sports, dance, and other domains.
research on sport psychology (pp. 127170). New In K. A. Ericsson & J. Smith (Eds), Toward a
York: McMillan. general theory of expertise: Prospects and limits
(pp. 12615 2). Cambridge: Cambridge Univer-
Abernethy, B., Burgess-Limerick, R., & Parks, S.
sity Press.
(1994). Contrasting approaches to the study of
motor expertise. Quest, 46, 186198. Anderson, J. R. (1982). Acquisition of cognitive
skill. Psychological Review, 89, 3 69406.
Abernethy, B., Neal, R. J., & Koning, P. (1994).
Visual-perceptual and cognitive differences Anderson, J. R. (1983 ). The architecture of cog-
between expert, intermediate, and novice nition. Cambridge, MA: Harvard University
snooker players. Applied Cognitive Psychology, Press.
18, 185 211. Baker, J., Cote, J., & Abernethy, B. (2003 ). Sport
Abernethy, B., & Russell, D. G. (1984). Advance specific training, deliberate practice and the
cue utilization by skilled cricket batsmen. Aus- development of expertise in team ball sports.
tralian Journal of Science and Medicine in Sport, Journal of Applied Sport Psychology, 15 , 12
15 , 210. 25 .
484 the cambridge handbook of expertise and expert performance
Beek, P. J. (2000). Toward a theory of implicit K. Davids, G. Savelsbergh, S. Bennett, & J. Van
learning in the perceptual-motor domain. Inter- der Kamp (Eds.), Interceptive actions (pp. 64
national Journal of Sport Psychology, 3 1, 5 47 89). London: E. & F. N. Spon.
5 5 4. Charness, N. (1976). Memory for chess positions:
Beek, P. J., Jacobs, D., Daffertshofer, A., & Huys, Resistance to interference. Journal of Experi-
R. (2003 ). Expert performance in sport: Views mental Psychology: Human Learning and Mem-
from the joint perspectives of ecological psy- ory, 2 , 64165 5 .
chology and dynamical systems theory. In J. L. Charness, N. (1979). Components of skill in
Starkes & K. A. Ericsson (Eds.), Expert per- bridge. Canadian Journal of Psychology, 3 3 , 1
formance in sports: Advances in research on 16.
sport expertise (pp. 3 213 44), Champaign, IL: Chase, W. G., & Simon, H. A. (1973 ). The minds
Human Kinetics. eye in chess. In W. G. Chase (Ed.), Visual
Beilock, S. L., & Carr, T. H. (2001). On the information processing (pp. 215 282). New York:
fragility of skilled performance: What governs Acasdemic Press.
choking under pressure? Journal of Experimen- Chiesi, H. L. O., Spilich, G. J., & Voss, J. F. (1979).
tal Psychology: General, 13 0, 701725 . Acquisition of domain related information in
Beilock, S., Carr, T., MacMahon, C., & Starkes, relation to high and low domain knowledge.
J. L. (2002). When paying attention becomes Journal of Verbal Learning and Verbal Behavior,
counterproductive: Impact of divided versus 18, 25 7273 .
skill-focused attention on novice and experi- Cote, J., & Hay, J. (2002). Childrens involvement
enced performance of sensorimotor skills. Jour- in sport: A developmental perspective. In J. M.
nal of Experimental Psychology: Applied, 8, 616. Silva & D. Stevens (Eds.), Psychological foun-
Bernasconi, P., & Kohl, J. (1993 ). Analysis of dations in sport (pp. 4845 02). Boston, MA:
co-ordination between breathing and exercise Merrill.
rhythms in man. Journal of Physiology, 471, Cote, J., Salmela, J. H., Trudel, P., Baria, A.,
693 706. & Russell, S. J. (1995 ). The coaching model:
Bernstein, N. A. (1967). The co-regulation of move- A grounded assessment of expert gymnastic
ments. Oxford: Pergamon. coaches knowledge. Journal of Sport and Exer-
Bernstein, N. A. (1996). On dexterity and its cise Psychology, 17 , 117.
development. In M. L. Latash & M. T. Cratty, B. J. (1983 ). Psychology in contempo-
Turvey (Eds.), Dexterity and its development rary sport: Guidelines for coaches and athletes
(pp. 3 244). Mahwah, NJ: Erlbaum. (2 nd Ed.). Englewood Cliffs, NJ: Prentice-
Bootsma, R. J., & van, Wieringen P. C. W. (1990). Hall.
Timing an attacking forehand drive in table ten- De Groot, A. (1978). Thought and choice in chess.
nis. Journal of Experimental Psychology: Human The Hague: Mouton. (Original work published
Perception and Performance, 16, 2129. in 1946).
Borgeaud, P., & Abernethy, B. (1987). Skilled per- Diedrich, F. J., & Warren, W. H. (1995 ).
ception in volleyball defense. Journal of Sport Why change gaits? Dynamics of the walk-run
Psychology, 9, 400406. transition. Journal of Experimental Psychology:
Bryan, W. L., & Harter, N. (1897). Studies in the Human Perception and Performance, 2 1, 183
physiology and psychology of the telegraphic 202.
language. Psychological Review, 4, 275 3 . Duffy, L., Baluch, B., & Ericsson, K. A. (2004).
Bryan, W. L., & Harter, N. (1899) Studies on Dart performance as a function of facets of
the telegraphic language. The acquisition of practice amongst professional and amateur
the hierarchy of habits. Psychological Review, 6, men and women players. International Journal
3 45 3 75 . of Sport Psychology, 3 5 , 23 2245 .
Carnahan, H. (1993 ). The role of three dimen- Duffy, L., Baluch, B., & Ericsson, K. A. (Submit-
sional analysis in the assessment of motor ted). Gender differences in dart performance:
expertise. In J. L. Starkes & F. Allard (Eds.), The effects of physical size and differential
Cognitive issues in motor expertise (pp. 3 5 5 3 ). recruitment rates. Research Quarterly for Exer-
Amsterdam: North-Holland. cise and Sport.
Cauraugh, J. H., & Janelle, C. M. (2002). Visual Eccles, D. W., & Tenenbaum, G. (2004). Why an
search and cue utilisation in racket sports. In expert team is more than a team of experts:
expert performance in sport 485
A cognitive conceptualization of team coor- vice au tennis. Canadian Journal of Sport Sci-
dination and communication in sport. Jour- ences, 13 , 79-87
nal of Sport and Exercise Psychology, 2 6, 5 42 Helsen, W. F., & Starkes, J. L. (1999). A multidi-
5 60. mensional approach to skilled perception and
Ericsson, K. A., Krampe, R. T., & Tesch-Romer, performance in sport. Applied Cognitive Psy-
C. (1993 ). The role of deliberate practice in the chology, 13 , 127.
acquisition of expert performance. Psychologi- Helsen, W., & Pauwels, J. M. (1993 ). The rela-
cal Review, 100, 3 63 406. tionship between expertise and visual infor-
Ericsson, K. A., & Polson, P. G. (1988). A cog- mation processing in sport. In J. Starkes & F.
nitive explanation of exceptional memory for Allard (Eds.), Cognitive issues in motor expertise
restaurant orders. In M. T. H. Chi, R. Glaser, (pp. 10913 4). Amsterdam: North-Holland.
& M. J. Farr (Eds), The nature of expertise (pp. Hodges, N. J., Kerr, T., Starkes, J. L., Weir,
23 70). Hillsdale, NJ: Erlbaum. P., & Nanandiou, A. (2004). Predicting per-
Ericsson, K. A., & Simon, H. A. (1993 ). Proto- formance from deliberate practice hours for
col analysis: Verbal reports as data. Cambridge, triathletes and swimmers: What, when and
MA: MIT Press. where is practice important? Journal of Experi-
Ericsson, K. A. (2003 ). Development of elite per- mental Psychology: Applied, 10, 21923 7.
formance and deliberate practice: An update Huys, R., & Beek, P. J. (2002). The coupling
from the perspective of the expert perfor- between point-of-gaze and ball movements
mance approach. In J. Starkes & K. A. Ericsson in three-ball cascade juggling: The effects of
(Eds.) Expert performance in sports: Advance in expertise, pattern and tempo. Journal of Sports
research on sport expertise. (pp. 4984). Cham- Sciences, 2 0(3 ), 171186.
paign. Ill.: Human Kinetics. Huys, R., Daffertshofer, A., & Beek, P. J. (2004).
Fitts, P. M. (1965 ). Factors in complex skill train- The evolution of coordination during skill
ing. In R. Glaser (Ed.), Training research and acquisition: The dynamical systems approach.
education 177197. New York: Wiley. In A. M. Williams & N. J. Hodges (Eds.), Skill
Fitts, P. M., & Posner, M. I. (1967). Human per- acquisition in sport: Research, theory and practice
formance. Belmont, CA: Brooks/Cole. (pp. 3 5 13 73 ). London: Routledge.
French, K. E., Nevett, M. E., Spurgeon, J. H., Lee, D. N., Lishman, J. R., & Thompson, J. A.
Graham, K. C., Rink, J. E., & McPherson, (1982). Regulation of gait in long jumping. Jour-
S. L. (1996). Knowledge representation and nal of Experimental Psychology: Human Percep-
problem solution in expert and novice youth tion and Performance, 8, 44845 9.
baseball performance. Research Quarterly for Lee, T. D., & Swinnen, S. P. (1993 ). Three lega-
Exercise and Sport, 66, 194201. cies of Bryan and Harter: Automaticity, vari-
French, K. E., & Thomas, J. R. (1987). The rela- ability and change in skilled performance. In
tion of knowledge development to childrens J. Starkes & F. Allard (Eds.), Cognitive issues
basketball performance. Journal of Sport Psy- in motor expertise (pp. 295 3 16). Amsterdam:
chology, 9, 15 3 2. North-Holland.
Garland, D. J., & Barry, J. R. (1991). Cognitive Miller, G. A. (195 6). The magical number seven,
advantage in sport: The nature of perceptual plus or minus two: Some limits on our capac-
structures. American Journal of Psychology, 104, ity for processing information. Psychological
211228. Review, 63 , 8197.
Gibson, J. J. (1979). An ecological approach McPherson, S. L. (1993 a). Knowledge represen-
to visual perception. Boston, MA: Houghton- tation and decision-making in sport. In J. L.
Mifflin. Starkes & F. Allard (Eds.), Cognitive issues
Goulet, C., Bard, M., & Fleury, C. (1989). Exper- in motor expertise (pp. 15 9188). Amsterdam:
tise differences in preparing to return a tennis North-Holland.
serve: A visual search information processing McPherson, S. L. (1993 b). The influence of player
approach. Journal of Sport and Exercise Psychol- experience on problem solving during batting
ogy, 11, 3 823 98. preparation in baseball. Journal of Sport and
Goulet, C., Flevry, M., Bard, C., Yerles, M., Exercise Psychology, 15 , 3 043 25 .
Michaud, D., & Lemine, L. (1988). Analyses McPherson, S. L. (2000). Expert-novice differ-
des indices visuels preleves en reception de ser- ences in planning strategies during collegiate
486 the cambridge handbook of expertise and expert performance
singles tennis competition. Journal of Sport and Salmela, J. H., & Moraes, L. C. (2003 ). Develop-
Exercise Psychology, 2 2 , 3 962. ment of expertise: The role of coaching, fam-
McPherson, S. L., & French, K. E. (1991). Changes ilies, and cultural contexts. In J. L. Starkes &
in cognitive strategy and motor skill in tennis. K. A. Ericsoon (Eds.), Expert performance in
Journal of Sport and Exercise Psychology, 13 , 26 sports: Advances in research on sport expertise
41. (pp. 275 294). Champaign, IL: Human Kinet-
McPherson, S. L. & Kernodle, M. (2003 ). Tac- ics.
tics, the neglected attribute of expertise: Prob- Savelsbergh, G. J. P., Williams, A. M., van der
lem representations and performance skills Kamp, J., & Ward, P. (2002). Visual search,
in tennis. In J. L. Starkes & K. A. Ericsson anticipation and expertise in soccer goalkeep-
(Eds.), Expert performance in sports: Advances in ers. Journal of Sport Sciences, 2 0, 279287.
research in sport expertise (pp. 13 7168). Cham- Schmidt, R. A. (1975 ). A schema theory of dis-
paign, IL: Human Kinetics. crete motor skill learning. Psychological Review,
McPherson, S. L., & Thomas, J. R. (1989). Rela- 82 , 225 260.
tion of knowledge and performance in boys Schmidt, R. A., & Lee, T. A. (1998). Motor con-
tennis: Age and expertise. Journal of Experimen- trol and learning: A behavioral emphasis. Cham-
tal Child Psychology, 48, 190211. paign, IL: Human Kinetics.
Nakagawa, A. (1982). A field experiment on Schneider, W., & Shiffrin, R. M. (1977). Con-
recognition of game situations in ball games: trolled and automatic information processing:
The case of static situation in rugby football. I. Detection, search, and attention. Psychologi-
Japanese Journal of Physical Education, 2 7 , 17 cal Review, 92 , 424428.
26. Singer, R. N., Cauraugh, J. H., Chen, O., Stein-
Nevett, M. E., & French, K. E. (1997). The devel- berg, G. M., & Frehlich, S. G. (1996). Visual
opment of sport-specific planning, rehearsal search, anticipation, and reactive comparisons
and updating of plans during defensive youth between skilled and beginning tennis players.
baseball game performance. Research Quarterly Journal of Applied Sport Psychology, 8, 926.
for Exercise and Sport, 68, 203 214. Smeeton, N. J., Ward, P., & Williams, A. M.
Nougier, V., Ripoll, H., & Stein, J. (1989). Ori- (2004). Do pattern recognition skills transfer
enting of attention with highly skilled athletes. across sports? A preliminary analysis. Journal
International Journal of Sport Psychology, 2 0, of Sports Sciences, 2 2 , 205 13 .
205 223 . Starkes, J. L. (1987). Skill in field hockey: The
Paull, G., & Glencross, D. (1997). Expert percep- nature of the cognitive advantage. Journal of
tion and decision making in baseball. Interna- Sport Psychology, 9, 146160.
tional Journal of Sport Psychology, 2 8, 3 5 5 6. Starkes, J. L. Motor experts: Opening thoughts.
Perkins-Ceccato, N., Passmore, S. R., & Lee, T. D. In J. L. Starkes & F. Allard (Eds.) Cognitive
(2003 ). Effects of focus of attention depend on issues in motor expertise (pp. 3 16). Amsterdam:
golfers skill. Journal of Sports Sciences, 2 1, 5 93 Elsevier.
600. Starkes, J. L., & Allard, F. (Eds.). (1993 ). Cogni-
Post, A. A., Daffertshofer, A., & Beek, P. J. (2000). tive issues in motor expertise. Amsterdam: North
Principal components in three-ball cascade jug- Holland.
gling. Biological Cybernetics, 82 , 143 15 2. Starkes, J. L., Cullen, J., & MacMahon, C. (2004).
Poulton, E. C. (195 7). On predicting skilled A model of skill acquisition and retention
movement. Psychological Bulletin, 5 4, 467 of perceptual-motor performance. In A. M.
478. Williams & N. J. Hodges (Eds.), Skill acquisition
Reilly, T., Williams, A. M., Nevill, A., & Franks, in sport: Research, theory and practice (pp. 25 9
A. (2000). A multidisciplinary approach to tal- 281). London: Routledge.
ent identification in soccer. Journal of Sports Sci- Starkes, J. L., & Deakin, J. (1984). Perception in
ences, 18, 668676. sport: A cognitive approach to skilled perfor-
Ripoll, H., Kerlirzin, Y., Stein, J. R., & Reine, mance. In W. F. Straub & J. M. Williams (Eds.),
B. (1995 ). Analysis of information processing, Cognitive sport psychology, (pp. 115 128). Lans-
decision making, and visual strategies in com- ing MI: Sport Science.
plex problem solving sport situations. Human Starkes, J. L., Deakin, J. M., Allard, F., Hodges,
Movement Science, 14, 92993 8. N. J., & Hayes, A. (1996). Deliberate practice
expert performance in sport 487
in sports: What is it anyway? In K. A. Eric- Ward, P., & Williams, A. M. (2003 ). Perceptual
sson (Ed.), The road to excellence: The acquisi- and cognitive skill development: The multidi-
tion of expert performance in the arts and sciences, mensional nature of expert performance. Jour-
sports, and games (pp. 81106). Mahwah, NJ: nal of Sport and Exercise Psychology, 2 5 , 93 111.
Lawrence Erlbaum. Ward, P., Williams, A. M., & Bennett, S. J. (2002).
Starkes, J. L., Edwards, P., Dissanayake, P., & Visual search and biological motion perception
Dunn, T. (1995 ) A new technology and fieldtest in tennis. Research Quarterly for Exercise and
of advance cue usage in volley ball. Research Sport, 73 , 107112.
Quarterly for Exercise and Sport, 66, 162167. Ward, P., & Williams, A. M. (2003 ). Percep-
Starkes, J. L., Helsen, W. F., & Jack, R. (2000). tual and cognitive skill development in Soccer:
Expert performance in sport and dance. In The multidimensional nature of expert perfor-
R. N. Singer, H. A. Hausenblas, & C. M. Janelle mance. Journal of Sport and Exercise Psychology,
(Eds.), Handbook of sport psychology (pp. 174 2 5 , 93 111.
201). New York: John Wiley. Ward, P., Hodges, N. J., & Starkes, J. L. (2004).
Starkes, J. L., Weir, P. L., & Young, B. (2003 ). Deliberate practice and expert performance:
Retaining expertise: What does it take for older Defining the path to excellence. In A. M.
expert athletes to continue to excel? In J. L. Williams & N. J. Hodges (Eds.), Skill acquisition
Starkes & K. A. Ericsson (Eds.), Expert perfor- in sport: Research, theory and practice (pp. 23 1
mance in sports: Advances in research on sport 25 8). London: Routledge.
expertise (pp 25 1272). Champaign, IL: Human Williams, A. M. (2002). Visual search behaviour
Kinetics. in sport. Journal of Sports Sciences, 2 0, 169170.
Ste-Marie, D. (2003 ). Expertise in sport judges Williams, A. M., & Davids, K. (1995 ). Declarative
and referees: Circumventing information- knowledge in sport: A byproduct of experience
processing limitations. In J. L. Starkes & K. A. or a characteristic of expertise? Journal of Sport
Ericsson (Eds.), Expert performance in sports: and Exercise Psychology, 7 , 25 9 275 .
Advances in research in sport expertise (pp. 169 Williams, A. M., & Elliott, D. (1999). Anxiety and
190). Champaign, IL: Human Kinetics. visual search strategy in karate. Journal of Sport
Tenenbaum, G. (2003 ). Expert athletes: An inte- and Exercise Psychology, 2 1, 3 623 75 .
grated approach to decision-making. In J. L. Williams, A. M., & Ward, P. (2003 ). Developing
Starkes & K. A. Ericsson (Eds.), Expert perfor- perceptual expertise in sport. In J. L. Starkes
mance in sport: Advances in research on sport & K. A. Ericsson (Eds.), Expert performance in
expertise (pp. 191218). Champaign, IL: Human sports: Advances in research on sport expertise
Kinetics. (pp. 219249). Champaign, IL: Human Kinet-
Tenenbaum, G., & Bar-Eli, M. (1993 ) Decision- ics.
making in sport: A cognitive perspective. Williams, A. M., Davids, K., & Williams, J. G.
In R. Singer, M. Murphey, & K. Tennant (1999). Visual perception and action in sport.
(Eds.), Handbook of research on sport psychol- London: E. & F. N. Spon.
ogy (pp. 171192). New York: Macmillan. Williams, A. M., Ward, P., & Smeeton, N. J.
Thomas, J. R., Gallagher, J., & Lowry, K. (2003 ). (2004). Perceptual and cognitive expertise in
Developing motor and sport expertise: Meta- sport: Implications for skill acquisition and per-
analytic findings. Paper presented at North formance enhancement. In A. M. Williams &
American Society for the Psychology of Sport N. J. Hodges (Eds.), Skill acquisition in sport:
and Physical Activity, Savannah, Georgia. Research, theory and practice (pp. 3 283 47).
Thomas, P. R., Schlinker, P. J., & Over, R. (1996). London: Routledge.
Psychological and psychomotor skills associ- Williams, A. M., Davids, K., Burwitz, L., &
ated wi0th prowess at ten-pin bowling. Journal Williams, J. G. (1993 ). Cognitive knowledge
of Sports Sciences, 14, 25 5 268. and soccer performance. Perceptual and Motor
Vickers, J. N. (1992). Gaze control in putting. Per- Skills, 76, 5 795 93 .
ception, 2 1, 11713 2. Williams, A. M., Davids, K., Burwitz, L., &
Vickers, J. N. (1996).Visual control while aim- Williams, J. G. (1994). Visual search strategies
ing at a far target. Journal of Experimental of experienced and inexperienced soccer play-
Psychology: Human Perception and Performance, ers. Research Quarterly for Exercise and Sport,
2 2 , 3 423 5 4. 5 , 12713 5 .
488 the cambridge handbook of expertise and expert performance
Williams, A. M., & Davids, K. (1998). Visual in tennis. Journal of Experimental Psychology:
search strategy, selective attention, and exper- Applied, 8, 25 9270.
tise in soccer. Research Quarterly for Exercise Young, B., & Starkes, J. L. (in press). Lifes-
and Sport, 69, 111128. pan analyses of track performance: Con-
Williams, A. M., Ward, P., Knowles, J. M., & tinued training moderates the age-related
Smeeton, N. J. (2002). Perceptual skill in a real- performance decline. Experimental Aging Re-
world task: Training, instruction, and transfer search.
C H A P T E R 28
that emphasize mans dehumanization. (Noice, A., 1995 ) reveals that the major-
However, the majority of performances in ity of training methods are rooted in the
todays theatre/film/television venues are in Stanislavski notion that good acting consists
line with Hamlets advice and are designed of determining the characters spine (the
to lure audiences into accepting the portray- life purpose that motivates everything the
als as being true to life. Indeed, this require- character says and does) and, more impor-
ment to make each performance uniquely tantly, each specific objective of the charac-
real lies at the heart of professional act- ter at every moment in the play (Stanislavski,
ing. Only peripheral skills such as vocal 193 6). These concepts are seminal to the
projection or bodily flexibility are acquired Stanislavski system, and succeeding gen-
through deliberate repetitive practice. The erations of teachers have devised innu-
essence of acting consists paradoxically of merable mental/physical exercises, explana-
successive attempts to refrain from repeating tions, and metaphors to explain the process
yesterdays rendition but to create the events (e.g., Adler, 1988; Guskin, 2003 ; Meisner &
anew at each moment of every performance Longwell, 1987; Morris, 1985 ). Even those
(Noice & Noice, 1997a). The components of academic conservatory programs devoted
this expertise, and the research into those primarily to classical theatre require this
components, constitute the subject of this truthful playing of character objectives but
chapter. place equal emphasis on technical matters
such as voice projection, bodily flexibility,
and use of heightened language.
Acquisition of Acting Expertise
However, some completely non-
In every era including the present, many out- Stanislavski techniques have also received
standing actors have learned solely by obser- fairly wide acceptance. Probably the best
vation and experience. In fact, the dominant known of these is theatre games (e.g.,
approach to present-day actor training Spolin, 1963 ), an approach that teaches
comes from the work of Russian actor- acting skills indirectly. For example, in a
director-theorist Constantine Stanislavski game called gibberish, the intent is to
(193 6), who said he invented nothing but remove the crutch of verbal meaning by
simply analyzed the acting of his day. He forcing the actors to communicate their
found that the worst actors were obviously desires in nonsense strings of sound that
feigning involvement in the dramatic situa- they spontaneously generate in response
tion but that the best seemed to be living in to a given situation. Thus, they must use
the present, experiencing what the charac- inflection, intonation, facial expression, and
ter would experience. Stanislavski spent the so forth to attain their goals without resort-
rest of his life trying to codify that approach ing to language, pantomime, or illustrative
into a teachable system of acting so that gestures typical of charades. The aim is to
what came naturally to the rare few could replace reliance on verbal meanings with
be reliably taught to others. The Stanislavski goal-directed human conduct so that, when
system and its offshoots form the basis of the dialogue is added, it will rest on a rich
the majority of training programs in todays behavioral structure.
colleges and universities. The term method A less frequently used actor-training
acting is often used to describe American approach employs neutral and character
variants of the Stanislavski system, but the masks. Wearing a neutral (i.e., expres-
phrase is subject to extreme interpretations, sionless) mask requires the actor to start
at times being applied only to director Lee from scratch when creating a new char-
Strasbergs view of Stanislavskis early writ- acter instead of unwittingly imposing his
ings, and at other times to any approach or her own eccentricities on that creation,
that results in truthful on-stage behavior whereas wearing a character mask poses
(Vineberg, 1991). Nevertheless, an examina- the challenge of generating real behav-
tion of todays most used acting textbooks ior that does not conflict with the fixed
artistic performance 491
expression of the mask (for a complete those five linked experiments was that actors
discussion of masks and actor-training, see could reliably access individual words from
Rolf, 1977). Although masks have been anywhere in their very lengthy roles when
used in performance from the beginnings given brief (one to four word) probes that
of theatre in ancient Greece to current provided little contextual information. (The
Japanese Noh performances, their use in authors of this chapter have informally repli-
actor training is relatively recent. Many his- cated this effect over a number of years and
torians credit Jacques Lecoq (19211999) for have yet to find a professional actor cur-
popularizing this instructional device (e.g., rently performing a play who could not sup-
Rudlin, 2000). ply the remainder of a line of dialogue given
In addition to the above, many theatre any unique word from the role. Retrieval is
practioners have devised methods suited to no faster when the cue is from the begin-
their particular production needs. Bertold ning, as opposed to the middle or end of the
Brecht believed that theatre must serve play.) Oliver and Ericsson (1986) also found
a didactic purpose, which is best accom- that the actors were successful at retrieval
plished when the audience is not caught (although slower) when the probes and the
up in the action. Therefore, he instructed targets crossed sentence boundaries (e.g., the
his actors to shun true emotional involve- first words of a sentence were used as probes
ment and to remain at a distance from and the actors task was to retrieve the last
the characters so that the audience would words of the immediately preceding sen-
not vicariously participate in the perfor- tence). Testing actors who were playing roles
mance and thus be free to think about the in two different plays revealed that identify-
issues raised (Barton, 1993 ). Musicals, ritual ing the roles from which the probes were
theatre, comedy improvisation, and many drawn affected response time but not accu-
other theatrical forms all have instructional racy. However, the methods by which this
approaches designed to target needs quite proficiency was acquired were not empiri-
different from those of traditional story- cally investigated.
telling in theatre and film. However, it is Intons-Peterson and Smyth (1987) did
essential to point out that whereas many investigate some aspects of learning strategy
college/professional training programs incor- but their research used non-professionals
porate games, or masks, or dozens of (student actors who were regarded as good
other alternate approaches, the overwhelm- memorizers). Two groups (acting students
ing majority of theatre education is still pri- and non-acting students) were placed either
marily Stanislavski-based (Noice, A., 1995 ). in a gist or rote condition. They performed
the same learn-by-repeating-aloud task with
Empirical Investigations a prose passage, with the gist participants
into the Acting Process being allowed to read the passage over
once before beginning and the rote partici-
retention and access to roles pants being asked to start memorizing aloud
Actors report that the question they hear immediately. These researchers found that
most often from theatregoers is, How do rote instructions encouraged closer atten-
you remember all those lines? (Noice & tion to exact wording, that most partici-
Noice, 2002a). Therefore, it is hardly sur- pants seemed to be using grammatical units
prising that when cognitive psychologists as landmarks, and that the acting students
began to investigate actors, they immedi- attempted to find optimal length units.
ately focused on their memory processes. In the pilot study for an ongoing series
Oliver and Ericsson (1986) performed of experiments, Noice (1992) collected self-
one of the earliest such investigations when reports of professional actors of various ages,
they tested actors currently performing backgrounds, and education. These reports
Shakespearean roles in a professional reper- were categorized and analyzed. Results indi-
tory company. One of the main findings of cated that all the actors, before they gave
492 the cambridge handbook of expertise and expert performance
any thought to memorization, stressed the time, the actor thought Im flattering him
notion of understanding the ideas behind (Noice, 1991).
the utterances, and the reasons the charac- For comparison purposes, novices were
ters used those words to express those ideas. given the same scenes to analyze. The over-
One of the actors summed up the process whelming majority of the novices verbal-
this way: izations did not concern the intentions of
the characters or the dramatic situation, but
We do things in reverse in the theater. We were either private (I wonder how long this
get the script, which is . . . at the end of the experiment will take?) or editorial (The
thought process: we have the lines there. plot thickens). From a quantitative stand-
Normally in life, you have an impulse and point, the actors retained significantly more
then a thought which you put into words. of a theatrical script than the novices, even
Well, I have the words, I get the words first when the participants were told not to mem-
in this finished script. And so I have to go
orize the material, but to read it only for
back and find out what the thought was, to
understanding. Yet when learning material
have you say those words. And more impor-
tantly, what was the impulse that created other than theatrical scripts, actors fared no
the thought that created the words, and better than novices.
usually it could be an emotional kind of The overall picture that emerged from
thing. What is the reason for that thought? these studies was that actors regard their pri-
Thats the way I have always thought of it. mary job as doing for real whatever the
(Noice, 1992 , pp. 42 042 1) character is doing. So, if the dramatic situa-
tion calls for character A to plead with char-
As intriguing as such statements were, they acter B, the actor playing character A actu-
nevertheless represented actors introspec- ally pleads with the actor playing character
tion into their own mental operations, a pro- B. He or she does NOT try to look and sound
cess that is always somewhat suspect. (See pleading, the plea must be genuine and
Ericsson, Chapter 13 , for discussion of pro- made at the moment of utterance (Noice &
tocol analysis.) Therefore, some follow-up Noice, 1994). Furthermore, the actor must
experiments employed the quantitative and pursue that intention truthfully at every per-
qualitative analysis of concurrent or retro- formance (or during every take in a TV or
spective verbal protocols (Noice, 1991, 1993 ; film studio), which requires that the actor
Noice & Noice, 1994, 1996). The underly- allow himself or herself to be influenced by
ing tasks in these studies consisted of the the behavior of the other actors. This keeps
actors being given a scene to prepare for the scene alive. The chemistry between the
an imaginary audition or performance. The actors will vary subtly from performance to
actors worked their way through the scene, performance and affect how the intentions
verbalizing all thoughts that occurred during are played.
the preparation phase, then performing free A further point that emerged from the
recall, cued recall, or summarization tasks. protocols was that the fulfillment of the
Analysis of the protocols revealed that, in intention had to be doable at every per-
almost every case, the primary concern was formance. That is, the actor had to choose
to extract from the literal text the true inten- intentions that were not only implied by the
tions of the assigned characters. Since most script, but that could be executed regardless
well-written plays employ subtlety and indi- of circumstances, when feeling sick or well,
rection, these thoughts rarely concerned the on opening night with a career possibly rid-
literal meaning of the words but rather an ing on the outcome, or at the 3 00th perfor-
underlying meaning inferred by the actor. mance of the role when torpor threatens.
For example, when a character said, Im An early step in role preparation is to seg-
sure you can trust Storey, the actor thought ment the script into goal-directed units that
Im trying to calm him down, and when actors call beats (Noice & Noice, 1993 ).
a character said He talks about you all the These beats are chunks of dialogue, devoted
artistic performance 493
to the fulfillment of a single intention. ical changes will occur, such as a possible
Most expertise investigations (e.g., Chase & tightening of the jaw or a more belligerent
Simon, 1973 ) show that experts generally stance. Each person will experience the sit-
form larger chunks of domain-specific infor- uation slightly differently, but, in every case,
mation than novices. However, this beat the changes will be multi-modal: cognitive,
study revealed that the experts chunks were affective, motoric. In addition to being a
far smaller and far more diverse. This was component of acting expertise, it has been
because novices, when asked to divide the shown that brief training in the use of active
script into segments for the purpose of experiencing can produce enhanced recall
learning the text, almost uniformly made in novice acting students and non-actors
divisions where the topic of conversation (Noice & Noice, 1997b).
changed. On the other hand, actors seg- Levenson, Ekman, and Friesen (1990)
mented the script into each separate inten- investigated the reverse of the active expe-
tion of the assigned character, a process that riencing principle. That is, instead of using
required extensive elaboration, perspective imagination to affect emotional and physi-
taking, self-referencing, self-generation, dis- ological changes, they coached participants
tinctiveness, and other cognitive principles to manipulate their faces into positions that
shown to benefit memory (see Noice & resembled the outer expression of the inner
Noice, 1997a). Therefore, it is hardly sur- experience. The researchers found that the
prising that, in the allotted 20 minutes, participants then experienced the emotion
actors who analyzed the script on a micro- consistent with that facial expression. How-
level remembered significantly more than ever, the assumed facial poses were so bla-
the novices who processed it on a topic-by- tant (amounting to grimaces) that they could
topic basis. not be used for most dramatic purposes.
The point that consistently emerged from Toward the end of his life, Stanislavski exper-
these protocol studies was that virtually all imented with a subtle form of this outside-
actors appear to go through a two-stage pro- in technique (usually referred to as the
cess. The first (analytical) stage consists of method of physical actions) in which phys-
examining the script in depth to determine ically doing what the character would do
the intentions of the characters. The sec- leads to actively experiencing the charac-
ond (rehearsal/performance) stage consists ters inner life. For example, a physical action
of what the researchers have called active such as stroking a loved ones face so gently
experiencing (Noice & Noice, 1997b). This as to not wake her, is apt to generate feel-
entails actually doing whatever the character ings of tenderness in the actor (Stanislavski,
is doing, whether it is taunting, interrogat- 1961).
ing, bullying, or any of the hundreds of com- Further confirmation of the memory ben-
mon human transactions that might occur in efits of the active experiencing principle
the performance of a role. Proper application comes from a recent study by Scott, Harris,
of this principle results in the actor experi- and Rothe (2001). Ninety-one female col-
encing the mental life of the character along lege students read the same long solo speech
with the concomitant feelings and appropri- (twice through) from a contemporary play.
ate physiological states. All participants were then given one of five
To demonstrate the active experiencing different 3 0-minute processing tasks from
principle, it is possible to carry out a thought answering questions about the character to
experiment. First, one would become imag- improvising a scene in which the character
inatively involved in an intention, such as might appear. Results showed greater recall
to demand obedience. Obviously cogni- (verbatim plus gist) for the improvisation
tive activity is present in terms of the spe- group. The authors concluded that a con-
cific nature of the demand, but also ones dition involving active experiencing at cog-
affect state will shift to a sense of strongly nitive, affective, and motoric levels leads to
felt determination, and simultaneous phys- better memory than other conditions that
494 the cambridge handbook of expertise and expert performance
affect states resulting from pursuing the craft tions between them. This finding casts doubt
of an actor who is portraying a character in on the claim made by a very small minor-
front of an audience. The task consisted of ity of actors who assert that acting consists
placing an X at either extreme or some- of totally becoming the character and expe-
where in between that best indicated the riencing only his or her feelings. However,
mix of the two sources of emotion during Konijn not only rejects this extreme position,
performance. she goes far beyond her data to theorize that
Results showed that not a single actor all on-stage affect consists solely of emotions
placed an X at either extreme. The average resulting from the task of performing in front
mark was nearer to the character side (40% of an audience. This would appear to con-
from the left end; SD 11.60), and the range of tradict the experience of almost all actors,
responses was skewed strongly to the right. not only those in our samples, but the hun-
That is, the greatest number of responders dreds who have described their processes in
indicated that their feelings on-stage tended the theatre literature (e.g.,Vilga, 1997). Even
to be more character influenced than perfor- the avowedly technical anti-method actor,
mance influenced. Laurence Olivier, reported that acting on-
The actors were also invited to offer stage is an emotional problem. Youve got
feedback. The comments varied greatly but to feel it, a great test for the imagination
almost all referred to multiple sources of (Bee & Bragg, 1982).
emotion, as in the following: Over a period of many years, Susana
Bloch, a neuroscientist, investigated the
I seem to alternate from one state to another. intentional generation of emotion by phys-
That is, I get immersed in the character side iological means. One study (Lemeignan,
and experience real situational feelings but Aguilera-Torres, & Bloch, 1992) showed
then I feel a rush of pleasure as an actor that, by duplicating the breathing patterns
sort of a self-congratulatory feeling that I am
that accompany six emotions, actors can
a good enough actor to actually experience
experience those emotions. Evidence for the
private character feelings during a public
performance. Then I regain some of my con- efficacy of the procedure was based on three
nection with the drama and get involved different measures: physiological arousal,
again until the pleasure of the involvement self-appraisal, and independent observation
forces itself on my consciousness once more. by concealed viewers.
the immune system. They found that imag- Implications of Acting Expertise
inatively produced affect states resulted in on Theoretical Issues
more variability in immunology than neutral
long-term working memory
states, with anxiety and happiness produc-
ing the highest variability and depression the The picture of actors memory processes
lowest. Such explorations may have some that arises from much of the research cited
interest to the expertise researcher inasmuch in this chapter would appear to be a perfect
as they demonstrate the ability of actors to fit with Ericsson and Kintschs (1995 ) long-
perform their tasks outside of a theatrical term working memory theory. That is, the
context even to the point of simultaneously actor possesses domain-specific information
experiencing invasive medical procedures. in LTM on human motivations and the myr-
iad specific intentions that are used to act
on those motivations. The text input is clas-
Applied Issues sified in terms of these intentions. (In this
The many experiments that shed light on section, Im trying to soften him up. Then,
actors processes prompted an intriguing in the next one, I go in for the kill.) These
question: Might actors learning strategies intentions then serve as efficient retrieval
benefit non-actors in terms of rapid assimila- cues and the entire succession of intentions
tion and efficient retrieval of material? The constitute a retrieval structure for the com-
first attempts to answer this question pro- plete role. Thus, residing in LTM are both
duced the surprising finding that memory the exact dialogue and the necessary per-
was enhanced far less by the preliminary ana- formance information (i.e., the reasons for
lytical work done on the script than by the uttering the dialogue). This theory would
rehearsal and performance components of explain the ability of actors to access any part
the actors role-learning process, referred to of the role with complete conviction given
as the active experiencing principle (Noice only minimal cues.
& Noice, 1997b). The efficacy of this princi-
ple was verified by findings with non-actors subject-performed tasks
under a wide variety of experimental pro- Researchers have frequently demonstrated
cedures (Noice & Noice, 2004). Soon this that phrases such as open the door or
inquiry became entwined with an entirely lift the pen are better remembered when
different branch of research: The effects actually performed (with real or imaginary
of effortful activities on healthy cognitive objects) than when acquired under standard
aging, and delayed onset of Alzheimers dis- verbal learning instructions (for reviews, see
ease (e.g., Friedland et al., 2001; Wilson Cohen, 1989; Engelkamp & Zimmer, 1990;
et al., 2002). This extensive body of research Nilsson, 2000). In these subject-performed
has shown the benefits of long-term cog- tasks (SPTs), the movements actually dupli-
nitive and social activities, but little atten- cate the verbal phrases. Recent experiments
tion has been devoted to short-term inter- with professional actors, student actors, and
ventions other than very narrowly targeted non-actors have shown that speeches accom-
ones (e.g., examining the effect of mnemon- panied by movements that do not dupli-
ics training on list learning). Recently, the cate their meaning are better recalled than
active experiencing components of acting speeches made while the participant is
techniques were taught to older adults (aver- standing or sitting in one place (Noice &
age age 73 years) over a four-week period. Noice, 1999, 2001; Noice, Noice, & Kennedy,
Participants were pretested on word recall, 2000). The authors refer to this as the non-
working memory, problem-solving ability, literal enactment effect and it occurs when
and quality of life issues. Significant gains the verbal material is related to the move-
were found over no-treatment controls and ment only at a higher-order level. For exam-
over a comparison art-study group, recruited ple, in a play, a husband might suspect that
to control for non-content-specific effects his wife had had lunch with her lover, so
(Noice, Noice, & Staines, 2004). he asks her where she ate that day while
artistic performance 497
he casually walks over to the bar and makes projection, great fluidity of movement, and
himself a drink. There is no literal connec- flawless retention of the longest scripts, but
tion between the stroll to the bar and the whose performances lack the juices of life,
question about lunch, but there is a goal- would probably never find employment in
directed one: They are both attempts to any professional theatre. On the other hand,
appear casual while trying to entrap the although a great dancer must be both an
wife. Despite this difference, an SPT-type artist and a technician, the technical com-
effect occurs, showing enhanced memory ponent is absolutely indispensable. An audi-
for verbal material accompanied by physical ence may greatly admire Barishnikovs inner
movement. fire, but it is his ability to leap to phe-
In a recent paper on SPTs, Koriat and nomenal heights and appear to remain sus-
Pearlman-Avnion (2003 ) argued that clus- pended in midair that truly dazzles view-
tering observed after verbal or enactive ers. Thus, the relatively greater contribution
encoding suggested that different encoding of practice-derived technical skill in dance
strategies resulted in different types of mem- becomes one defining difference between
ory organization, which in turn produced the two art forms.
different patterns of recall. However, that
study examined only action phrases typical
History
of SPT experiments (e.g., hammer a nail),
and the results would not appear applica- Although dance has been an important
ble to recall of non-literal dialogue that nei- aspect of our cultural heritage since the
ther describes nor implies the accompany- beginning of recorded history, it became
ing action (e.g., a character saying, I think a performance art with the birth of bal-
thats remarkable, Big Daddy while walking let in Europe during the Renaissance (Cass,
away from his father). However, the gener- 1993 ). In the 13 00s and 1400s, various forms
ation effect (Slamecka & Graf, 1978) might of dance performance emerged to enter-
help explain enhanced recall for both stan- tain the nobility. This movement reached
dard enactment and non-literal enactment its peak in France during the 1600s, and
because the participants necessarily generate by the next century, ballet had become a
the accompanying movements. recognized genre of art. Soon, dance grew
into the primary entertainment for much
theories of embodied cognition of Europe (Steeh, 1982). Ballet eventually
Art Glenberg and his associates (among oth- reached Russia, where, under the Czars, lav-
ers) have advanced a theoretical position in ish costumes, impressive scenery, and luxuri-
which cognitive activities (including speech ous theatres characterized the form. In addi-
acts) are viewed as evolutionary outcomes tion, dancers started to devote themselves
of our need to make our way in the physical to a strict schedule of rigorous rehearsals
world (e.g., Glenberg & Robertson, 1999). to achieve systematic technique (Burian,
The findings of the various experiments on 1963 ). Spearheaded by its preeminence in
the non-literal enactment effect appear com- Russia, ballet spread throughout the world.
pletely consistent with this theoretical view. In America, a Russian emigre, the late
(For a fuller discussion of the connections George Balanchine, founded The School of
between actors processes and embodied American Ballet to train outstanding dancers
cognition see Noice & Noice, 2001; see also for his New York City Ballet, currently one
Scott et al., 2001.) of the best-known and most respected com-
panies in the world (Gottlieb, 2004). Its neo-
classical style places strong emphasis on orig-
Dance inal choreography (Steeh, 1982).
America has made notable contributions
Acting and dance make both expressive and to two other contemporary dance forms. The
technical demands on the performer. How- first, modern dance (whose most famous
ever, an actor who exhibits mastery of voice exponent was Martha Graham) rejected
498 the cambridge handbook of expertise and expert performance
the restrictions of ballet and opted for an Moreover, expertise in dance consists of
approach where there are no rules other many components, some clearly defined and
than providing an aesthetically satisfying some highly elusive. The Danish ballet star
experience through movement, especially Eric Bruhn (1968) said,
spontaneous movement. Training for mod-
Behind the discipline and the form, behind
ern dance can be as rigorous as for bal-
whatever is the specific thing you are try-
let. Each company (Martha Graham, Hanya ing to express, there must be a mind . . . it
Holm, Merce Cunningham, Paul Taylor, isnt that the body is so different it didnt
Alvin Ailey, etc.) teaches original techniques fall out from Mars or from the moon. It
and methods designed to fulfill their origina- came from where everybody came from,
tors visions (Cheney, 1989). only somewhere along the road the mind
The newest dance discipline, jazz, is began to take it someplace. (p. 13 )
the dominant form on Broadway. Its expo-
nents usually study both ballet and mod-
Empirical Investigations
ern dance, then specialize in the choreo-
into the Dancing Process
graphic stylizations of Bob Fosse, Michael
Bennet, Susan Strohman, and other musical As with acting, much of the early scientific
theatre innovators. (The other well-known inquiry of dance focused on memory pro-
form of dance, tap, is more a sound-based cesses. Starkes, Deakin, Lindley, and Crisp
than a movement-based art. Indeed, its most (1987) published a seminal paper that antic-
expert exponents frequently appear at musi- ipated much of the investigation carried out
cal venues such as the Newport Jazz Festival during the following decades.1 The most
[Stearns & Stearns, 1968]). frequently cited finding was that, as with
other experts, ballet dancers demonstrated
superior memory for domain-specific infor-
Acquisition of Dance Expertise
mation. Participants were shown a video-
It is widely agreed that dance training (par- tape of either structured or unstructured
ticularly ballet) involves physically changing dance segments. The former was profession-
the body while it is still malleable, so most ally choreographed, whereas the latter com-
dancers start formal instruction between the bined the identical movements randomly.
ages of seven and nine. Approximately the The dancers were assigned to one of three
first ten years of training are devoted to conditions: verbal recall of the structured
physical preparation that provides balance, version, motor recall of the structured ver-
posture, control, elevation, and the abil- sion, and motor recall of the unstructured
ity to initiate and terminate movements to version. The researchers reported an advan-
match musical phrasing (De Mille, 1962). tage for experts in verbal or motor recall of
After that preparation phase, dancers gen- the structured version but found no differ-
erally seek out professional classes and ences in motor performance of the unstruc-
start auditioning for dance companies tured one, the usual finding in expertise
(Loren, 1978). studies.
Although ballet is often thought to be However, some exceptions to such
rigid and highly codified, there are actually domain-specificity in dance have been
many different training methods, including noted. Starkes, Caicco, Boutilier, and Sevsek
the Cecchetti (Italian), the Vaganova (Rus- (1990) reported that experienced modern
sian), the RAD (British), and the Balanchine dancers remembered all sequences better
(American). Each technique differs in its ter- than novices but did not display an advan-
minology, use of hands and arms, and meth- tage for structured sequences, suggesting
ods of flowing from one state to another. that modern dance (which includes far fewer
Only the five different foot positions are labeled structural movements than ballet)
common to almost all approaches (Craine requires the ability to recall any type of
& Mackrell, 2000). movement sequence. The authors opined
artistic performance 499
that this may have been accomplished by expert/novice experiments might not have
acquiring a flexible mental representation fully uncovered the details of dancers learn-
that incorporates the whole range of move- ing strategies because a test routine that was
ments used in modern dance. easy enough for a novice was too simple
Solso and Dallob (1995 ) examined the for an expert. Therefore, the investigators
possibility that experts acquire such a flexi- created different sets of dance sequences,
ble mental representation by extracting pro- judged by experienced dance instructors to
totypical information from a wide variety of clearly belong to the easy or difficult cate-
movements. The investigators constructed a gory. As predicted, the novice dancers lacked
simple, prototypical base dance that was the sophisticated strategies necessary to cope
used to generate a series of ten dances. (None with the difficult sequences and the expert
of the ten exactly matched the prototype.) dancers tended to abandon their usual strate-
Twenty professional dancers and 3 4 non- gies when faced with the simple ones. The
dancers learned these sequences to crite- researchers concluded that, in order to shed
rion (one rendition without mistakes). Then light on dancing expertise, stimuli would
the participants learned an additional set of have to be created in which the relative dif-
seven dances (three old, three new, and ficulty of each sequence would be approxi-
the original prototype). More (but not sig- mately the same when skill level was taken
nificantly more) experts falsely recognized into consideration.
the prototype as old. The researchers sug-
gested that the experts probably kinesthet- Mental Devices for Remembering
ically abstracted prototypical information Dance Patterns
from the original learning (but see also Jean,
Some dancers appear to encode steps by
Cadopi, & Ille, 2001).
a process called marking in which they
make small hand movements as they men-
Problems in Applying Expert/Novice tally rehearse a dance combination, result-
Procedures to Dancers ing in superior retention while conserving
energy (Starkes et al., 1987). Also, expert
An inherent problem with expert/novice
dancers employ multiple encoding devices,
research is that non-dancers are often inca-
including using verbal labels to represent a
pable of executing even rudimentary tech-
series of movements (e.g., step to the side, step
niques, thus making meaningful compar-
together, step to the side would be remem-
isons impossible. One way researchers have
bered simply as chasse; Poon & Rodgers,
circumvented this difficulty is by match-
2000).
ing two groups in terms of training, but
The latter investigators also reported
not of ability. For example Starkes et al.
a difficulty endemic to expertise research
(1987) compared the dancers from the
with professionals. During the study, their
local ballet schools with dancers from the
novice group was always available but their
National Ballet School, which recruits out-
expert group was cut in half by an unex-
standing students through worldwide audi-
pected rehearsal, resulting in the remaining
tions. Both groups had similar dance exp-
expert dancers having to perform twice as
erience (experts = 5 .1 years; novices = 4.7
many tasks as the novices in order for the
years).
researchers to complete their design.
Poon and Rodgers (2000) used a com-
parable approach, assigning dancers from
Use of Imagery and Proprioception
a University of Alberta Jazz Dance class
to the novice group (average training: In an attempt to see if the well known SPT
7.87 years) and selecting dancers from (subject-performed task) effect extended to
local professional programs for the expert simple dance movements, Foley, Bouffard,
group (average training: 9.25 years). How- Raag, and Disanto-Rose (1991, Experiment
ever, the researchers believed that previous 1) found that both common and uncommon
500 the cambridge handbook of expertise and expert performance
movements were remembered equally well dancers. The investigators referred to this as
as long as the participants (non-dancers) sensorimotor proprioception dominance. Fur-
performed them or imagined themselves ther evidence for proprioception comes
performing them. However, lower recall from a study by Smyth and Pendelton
was found for uncommon movements when (1994). They demonstrated that dancers
these non-dancers tried to learn them by memories for dance patterns were disrupted
observing or by imagining another person by concurrent motor tasks or concurrent ver-
performing. The researchers speculated that bal tasks, suggesting that both types of infor-
the participants who were imagining them- mation are implicated in performance.
selves performing might have experienced
greater enactive encoding of the sort respon- Music Cues
sible for the SPT effect (e.g., Engelkamp
Music and dance are integrally linked, with
& Zimmer, 1990). In Experiment 2, using
the former almost always accompanying the
both dancers and non-dancers, the investi-
latter. Starkes et al. (1987, Experiment 1)
gators found that neither the availability of
examined the role that music played in pro-
verbal descriptions nor stored motor repre-
ducing a serial position effect for dance
sentations were necessary for memory to be
movements. They had experts and novices
enhanced by enactment, but they did find
view dance sequences without accompany-
a domain-specific effect of imagery. That is,
ing music. In the subsequent recall phase, no
dancers reported greater ease in imagining
evidence was found for a recency effect, so
themselves performing dance movements
common in verbal-recall studies. In Exper-
than novices did, although dancers and non-
iment 2, they had experts view a sequence
dancers considered themselves equal in gen-
of eight ballet steps accompanied by music,
eral imagery ability (e.g., picturing a boat on
then administered a recall task either with
a lake).
or without music. Music exerted a signifi-
To investigate relative use of imagery
cant recall advantage especially on the most
among disciplines, Overby, Hall, and Haslam
recent step of the sequence.
(1998) sent questionnaires to dance teachers,
Poon and Rodgers (2000) reported that
figure skating coaches, and soccer coaches.
experts and novices differed in their han-
They found that, among the 44 dance teach-
dling of musical cues. The researchers col-
ers who responded, the majority employed
lected verbal protocols based on 20- to
both kinesthetic imagery and metaphorical
3 0-minute semi-structured interviews held
imagery (Imagine youre moving through
immediately following the learning of new
water). The teachers reported that such
dance routines. The findings revealed that
metaphorical imagery was highly effective in
expert dancers made far greater use of music
getting dancers to move slowly and smoothly
cues, attending to high and low notes, spe-
from one location to another. Kinesthetic
cific rhythmic phrases, and so on. The pro-
and metaphorical imagery were used far less
tocols showed also that experts used musi-
by sports coaches.
cal counts while creating fewer but larger
In an experiment to investigate depen-
chunks, which they learned separately, and
dence on proprioception, Golomer and
not necessarily in the original order. All
DuPui (2000) employed a precariously bal-
novices in the study tried to learn the first
anced seesaw attached to an accelerome-
segment, then add on the second, and so
ter (otherwise known as a stabilometer).
forth. Furthermore, the novices tended to
Dancers and non-dancers were assessed on
find the music distracting and either paid no
their ability to remain upright with eyes
attention or deliberately ignored it.
open (vision dependent balance) and eyes
closed (proprioceptive dependent balance).
The Expressive Aspects
Results showed that dancers (particularly
adult male dancers) depended more on pro- A dance performance is far more than a dis-
prioception and less on vision than non- play of technique, yet there appears to be
artistic performance 501
little consensus on methods of training for Bee, B. (Producer and Director), & Bragg, M.
the expressive components. In ballet com- (Editor). (1982). Laurence Olivier, a life. [Video
panies, young dancers frequently learn roles recording]. Chicago, IL: Public Media Video.
from older dancers who have previously per- Brockett, O. G. (1991). History of the theatre. (6th
formed them, a process that includes the ed.). Boston, MA: Allyn and Bacon.
interpretive aspects. Also, many dancers take Bruhn, E. (1968). Beyond technique. New York:
separate acting classes on their own initia- Johnson Reprint Corporation.
tive (B. Doyle-Wilch, personal correspon- Burian, K. V. (1963 ). The story of world ballet. Lon-
dence, July 16, 2004), but in response to a don, England: Allan Wingate.
query, the associate chair of one of the coun- Cass, J. (1993 ). Dancing through history. Engle-
trys larger college dance programs wrote, wood Cliffs, NJ: Prentice-Hall.
We do not offer or require any form of act- Chase, W. G., & Simon, H. A. (1973 ). Perception
ing classes within our dance programs (R. in chess. Cognitive Psychology, 4, 5 5 81.
Woodbury, personal correspondence, July Cheney, G. (1989). Basic concepts in modern
8, 2004). dance: A creative approach. Princeton, NJ: A
Dance Horizons Book.
Conclusion Cohen, R. (1994). Theatre. Mountain View, CA:
Mayfield Publishing.
Although many aspects of artistic perfor- Cohen, R. L. (1989). Memory for action events:
mance are being clarified through empirical The power of enactment. Educational Psychol-
inquiry, much remains to be done. For exam- ogy Review, 1, 5 780.
ple, great performing artists do not always Craine, D., & Mackrell, J. (2000). The Oxford
give emotionally moving performances and dictionary of dance. Oxford: Oxford University
lesser ones often do. Such differences can- Press.
not be ascribed to length or type of train- De Mille, A. (1962). To a young dancer. Boston:
ing, experience, motivation, or any other Little, Brown and Company.
tangible factor, but are a continuing part Diderot, D. (195 7). The paradox of acting. (W. H.
of the mystery of artistic endeavor. Indeed, Pollock, Trans.) New York: Hill and Wang.
understanding the true nature of these (Original work published 183 0.)
complex arts will probably call for many Engelkamp, J., & Zimmer, H. D. (1990). Mem-
more decades of interdisciplinary coopera- ory for actions events. A new field of research.
tion between actors/dancers and researchers Psychological Research, 5 1, 15 3 15 7.
in such disparate fields as expertise, memory, Ericsson, K. A., & Kintsch, W. (1995 ). Long-term
movement, emotion, and consciousness. working memory. Psychological Review, 102 (2),
211245 .
Foley, M. A., Bouffard, V., Raag, T., & Disanto-
Footnote Rose, M. (1991). The effects of enactive encod-
ing, type of movement, and imagined per-
1. There is also a large literature on movement spective on memory of dance. Psychological
in sports that is highly pertinent here (see Research, 5 3 , 25 125 9.
Hodges, Starkes, & MacMahon, Chapter 27). Friedland, R. P., Fritsch, T., Smyth, K. A., Koss, E.,
Lerner, A. J., Chen, C. H., et al. (2001). Patients
with Alzheimers disease have reduced activi-
ties in midlife compared with healthy control-
References
group members. Proceeding of the National
Academy of Sciences, 98 (6), 3 4403 445 .
Adler, S. (1988). The technique of acting. New Futterman, A. D., Kemeny, M. E., Shapiro,
York: Bantam Books. D., Polonsky, W., & Fahey, J. L. (1992).
Archer, W. (1888). Masks or faces? London: Long- Immunological variability associated with
mans, Green, & Co. experimentally-induced positive and negative
Barton, R. (1993 ). Style for actors. Mountain affective states. Psychological Medicine, 2 2 , 23 1
View, CA: Mayfield Publishing. 23 8.
502 the cambridge handbook of expertise and expert performance
Glenberg, A. M., & Robertson, D. A. (1999). tion A: Humanities & Social Sciences, 5 6 (5 -A),
Indexical understanding of instructions. Dis- 15 86.
course Processes, 2 8, 126. Noice, H. (1991). The role of explanations and
Golomer, E., & Dupui, P. (2000). Spectral analy- plan recognition in the learning of theatrical
sis of adult dancers sways: Sex and interaction scripts. Cognitive Science, 15 , 425 460.
vision proprioception. International Journal of Noice, H. (1992). Elaborative memory strategies
Neuroscience, 105 , 15 26. of professional actors. Applied Cognitive Psy-
Gottlieb, R. (2004). George Balanchine: The ballet chology, 6, 417427.
maker. New York: Harper Collins. Noice, H. (1993 ). Effects of rote versus gist
Guskin, H. (2003 ). How to stop acting. New York: strategy on the verbatim retention of theatri-
Faber and Faber. cal scripts. Applied Cognitive Psychology, 7,
Intons-Peterson, M. J., & Smyth, M. M. (1987). 75 84.
The anatomy of repertory memory. Journal Noice, H., & Noice, T. (1993 ). The effects of seg-
of Experimental Psychology: Learning, Memory, mentation on the recall of theatrical material.
and Cognition, 13 (3 ), 4905 00. Poetics, 2 2 , 5 167.
Jean, J., Cadopi, M., & Ille, A. (2001). How Noice, H., & Noice, T. (1994). An example of role
are dance sequences encoded and recalled by preparation by a professional actor: A think-
expert dancers? Current Psychology of Cogni- aloud protocol. Discourse Processes, 18, 3 45
tion, 2 0 (5 ), 3 25 3 3 7. 3 69.
Kelso, M. T, Weyhrauch, P., & Bates, J. (1993 ). Noice, H., & Noice, T. (1996). Two approaches
Dramatic presence. Presence, 2 (1), 115 . to learning a theatrical script. Memory, 4 (1),
Konijn, E. A. (2000). Acting emotions: Shap- 117.
ing emotions on stage. (B. Leach, Trans.) Noice, T., & Noice, H. (1997a). Expertise of pro-
Amsterdam: Amsterdam University Press. fessional actors: A cognitive view. Hillsdale, NJ:
Koriat, A., & Pearlman-Avnion, S. (2003 ). Mem- Erlbaum.
ory organization of action events and its rela- Noice, T., & Noice, H. (1997b). Effort and active
tionship to memory performance. Journal of experiencing as factors in verbatim recall. Dis-
Experimental Psychology: General, 13 2 (3 ), 43 5 course Processes, 2 3 , 5 169.
45 4. Noice, H., & Noice, T. (1999). Long-term reten-
Lemeignan, M., Aguilera-Torres, & Bloch, S. tion of theatrical roles. Memory, 7 (3 ), 3 5 7
(1992). Emotional effector patterns: Recogni- 3 82.
tion of expressions. European Bulletin of Cogni- Noice, H., & Noice, T. (2001). Learning dialogue
tive Psychology, 12 , 173 188. with and without movement. Memory and Cog-
Levenson, R. W., Ekman, P., & Friesen, W. V. nition, 2 9 (6), 820828.
(1990). Voluntary facial action generates Noice, T., & Noice, H. (2002a). A review of
emotion-specific autonomic nervous system recent research on the expertise of professional
activity. Psychophysiology, 2 7 (4). 3 63 3 94. actors. High Ability Studies, 13 , (1), 719.
Loren, T. (1978). The dancers companion. New Noice, T., & Noice, H. (2002b). Very long-term
York: Dial Press. recall and recognition of well-learned material.
Meisner, S., & Longwell, D. (1987). Sanford Meis- Applied Cognitive Psychology, 16, 25 9272.
ner on acting. New York: Vintage Books. Noice, T., & Noice, H. (2004). A cognitive learn-
Morris, E. (1985 ). Irreverent acting. New York: ing principle derived from the role acquisition
Putnam Publishing. strategies of professional actors. Cognitive Tech-
Nilsson, L-G. (2000). Remembering actions and nology, 9 (1), 3 43 9.
words. In E. Tulving & F. I. M. Craik (Eds.), Noice, H., Noice, T., & Kennedy, C. (2000).
The Oxford handbook of memory (pp. 13 7148). The contribution of movement on the
New York: Oxford University Press. recall of complex material. Memory, 8 (6),
Noice, A. A. (1995 ). The application of learn- 3 5 3 3 63 .
ing and memory principles of cognitive science Noice, H., & Noice, T., & Staines, G. (2004). A
to an actors ability to perform in the moment. short-term intervention to enhance cognitive
(Doctoral dissertation, Wayne State University, and affective functioning in older adults. Jour-
1995 ). Dissertation Abstracts International Sec- nal of Aging and Health, 16 (4), 124.
artistic performance 503
Perceptual-Motor Expertise
highlight the cornerstone concepts associ- intellectual skill obey different principles
ated with each approach. This selectivity than the systems underlying perceptual-
necessarily forces exclusion of useful find- motor skills. If that were true, the study
ings. Third and finally, we use the term of perceptual-motor expertise might bear
expertise in a way that is a bit unortho- little on the study of intellectual expertise
dox. Most investigators who study expertise and vice versa. However, the evidence from
argue that the study of the most accom- human performance reveals quite another
plished practitioners of a skill provides a picture. In fact, it is nearly impossible to
window into the factors that underlie the find meaningful differences between the fac-
skills development. We agree with this per- tors affecting the acquisition of perceptual-
spective but note that most tasks investi- motor skills and those affecting the acqui-
gated by students of perceptual-motor skill sition of intellectual skills (Rosenbaum,
are ones that can be mastered by virtually Carlson, & Gilmore, 2001; Schmidt & Bjork,
everyone (e.g., reaching for a glass of water). 1992). That this is so is perhaps not sur-
This makes their acquisition unremarkable, prising in view of the fact that the same
except for the fact that the apparent sim- neurophysiological principles support both
plicity of these skills frequently eludes their kinds of learning. Nevertheless, we can go
mimicry by robotic systems. Still, even mun- beyond a reductionist base to see common
dane skills such as reaching and grasping principles.
must be learned, and it seems reasonable to One similarity between intellectual and
suspect that mechanisms supporting learn- perceptual-motor skill is the short-term ben-
ing of commonplace skills will also support efit of massed over spaced practice. Prac-
acquisition of skills commonly viewed as ticing a task without breaks leads to bet-
being more sophisticated. ter performance immediately after practice
Our review of perceptual-motor skill than does practicing the same task with
acquisition begins with some evidence for breaks, and this is true both for intellec-
similarities between expertise in perceptual- tual (Glenberg, 1997) and perceptual-motor
motor skills and intellectual skills. Then tasks (Shea & Morgan, 1979). After several
we offer a brief review of neural plastic- hours of consolidation the relative benefit of
ity before focusing on three concepts that massed over spaced practice reverses when
have emerged from cognitive psychology: one assesses learning over the long term.
(1) Elements involved in the control of Then, spaced practice leads to better long-
perceptual-motor activities are linked to one term retention than does blocked or massed
another; (2) Actions are guided by predic- practice, and again this is true both for intel-
tion of sensory feedback; and (3 ) Attention lectual and perceptual-motor tasks. Magill
becomes less valuable as skill grows. Much and Hall (1990) provided a review of the rel-
of the classical and current work on the cog- evant evidence for perceptual-motor skills,
nitive substrates of perceptual-motor exper- and Melton (1970), Landauer and Bjork
tise can be traced to these ideas. Last, we dis- (1978), and Rea and Modigliani (1985 ) wrote
cuss the contributions from the ecological/ analogous reviews for intellectual skills.
dynamical systems approach. A similar interaction exists between
time of test and frequency of feedback.
If participants are given frequent feedback
about their performance on perceptual-
Similarities between Perceptual Motor motor tasks or on intellectual tasks, their per-
Skills and More Intellectual Functions formance in the short term is better than if
they are given frequent feedback. However,
The gulf between a computers ability to their performance in the long term is better
play chess and its inability to perform com- in both domains if they are given infrequent
mon acts of manual dexterity might lead rather than frequent feedback during train-
one to believe that the systems underlying ing. See Schmidt and Bjork (1992) for review.
perceptual-motor expertise 507
Figure 2 9.1. Adaptation to an artificial force field. (A). Overview of the subject
grasping a handle connected to two joints powered by torque motors. (B).
Normal handpaths in center-to-target or target-to-center movements. (C)
Artificial force field. (D) Initial distortions of movements to the artificial force
field. (E) Adapted movements, achieved five minutes later. From Brashers-Krug,
T., Shadmehr, R., & Bizzi, E. (1996). Consolidation in human motor memory.
Nature, 3 82 , 25 225 5 .
the everyday perceptual-motor task of tak- neural pools (Kaas, Merzenich, & Killackey,
ing hold of a vertical cylinder and moving it 1983 ). In contrast to older models of neu-
from one place to another. ronal processes, there is growing consensus
A final piece of evidence for the common- that such neural plasticity is a fundamental
ality of intellectual and perceptual-motor characteristic of brain organization.
skills comes from neuroscience. Early work Neural plasticity plays a role in
by Holmes (193 9) showed that damage to perceptual-motor skill learning. Brain-
the cerebellum often resulted in reduced imaging research has revealed enlargement
muscle tone, delayed movement initiation, of functionally relevant cortical areas in
motor planning errors, and tremor. These musicians who play stringed instruments
observations led to the view that the cere- such as the violin or cello (Elbert, Pantev,
bellum controls and coordinates movement. Wienbruch, Rockstroh, & Taub, 1995 ).
Other work has shown that the cere- These individuals have more brain tissue
bellum also subserves cognitive function- that responds to touch with the fingers of
ing (Fiez, 1996; Leiner, Leiner, & Dow, the left hand (the hand normally used for
1995 ). Cerebellar damage can lead to deficits depressing the strings) than brain tissue
in conditioning (Bracke-Tolkmitt, Linden, that responds to touch with the fingers
Canavan, & Diener, 1989) as well as dis- of the right hand (the hand that normally
ruptions in the analysis of temporal dura- bows). Furthermore, the difference between
tion (Ivry & Keele, 1989). Brain-imaging left- and right-hand somatosensory areas
studies have shown that the cerebellum is grows with practice. Similar changes in
active during performance of tasks as varied neural representation occur as monkeys
as word generation (Petersen, Fox, Posner, (Logothetis & Pauls, 1995 ) and humans
Mintun, & Raichle, 1989), tactile discrimi- (Gauthier, Skudlarski, Gore, & Anderson,
nation (Gao, Parsons, Bower, Xiong, 1996), 2000) develop perceptual expertise in
sequence learning (Jenkins, Brooks, Nixon, recognizing objects.
Frankowiak, & Passignham, 1994), and main- These results fit with a view of the ner-
tenance of information in working mem- vous system as a competitive arena in which
ory (Desmond, Gabrielli, Wagner, Binier, & perceptual-motor representations jockey for
Glover, 1997). Courchesne and Allen (1997) limited tracts of neural tissue. The findings
proposed that the cerebellum contributes to also accord with hypotheses from cogni-
the prediction and preparation of sequences, tive psychology that posit inhibition among
broadly construed. Insofar as sequences can psychological elements (Dagenbach & Carr,
consist of symbols (the sine qua non of intel- 1994).
lectual skills) or stimuli and responses (the
sine qua non of perceptual-motor skills),
the cerebellum may be viewed as an organ Elements of Motor Plans
subserving both perceptual-motor skills and
intellectual skills. The concept of associations among psycho-
logical elements has figured prominently in
the study of perceptual-motor skill acqui-
Neural Plasticity sition. The main questions about them are
how the elements are assembled and how
For any skill to be learned, changes must the assembly changes as skill develops.
occur in the nervous system. (see Hill and One hypothesis regarding assembly is that
Schneider, Chapter 3 7.) Classically, neuro- elements underlying successive behaviors
scientists have analyzed such changes at the are linearly arranged such that one element
level of the synapse (Hebb, 1949; Kandel, triggers the next. For early behaviorists, the
1981; Lynch & Baudry, 1984). A great deal of triggering signal was feedback: Performance
work has extended this analysis to the study of a movement caused by activation of ele-
of changes in the functional properties of ment i generated feedback that activated
perceptual-motor expertise 509
element i + 1, and so on. This idea was ings such as these cannot easily be accom-
refuted by Lashley (195 1), who noted that modated by a theory that holds that each
feedback loops take too long to explain movement is triggered by feedback from
rapid performance, as in keyboard sequences some previous movement. Instead, a plan is
of skilled pianists. He also observed, based needed to enable the effectors to move to
on clinical work, that disruption of feed- their targets with appropriate timing.
back loops does not prevent performance of Many of the insights into the nature
basic movement sequences. Lashley argued of perceptual-motor expertise were antic-
that close examination of behavior suggested ipated over a century ago by Bryan and
that successively executed movements can- Harter (1897, 1899), who suggested that
not be triggered by immediate, prior acti- perceptual-motor skill development relies
vation of their associated control elements. on the formation of ever more inclusive rou-
Rather, the elements are activated well in tines. Low-level control elements are joined
advance. The ensemble of such activated ele- together by higher-level control elements,
ments constitutes a plan for ensuing action. these higher-level control elements are in
By assuming plans for forthcoming action turn subsumed by still higher-order ele-
sequences, Lashley could account for the ments, and so on. The idea is familiar to com-
results just reviewed as well as errors in puter programmers, who know that routines
performance, such as slips of the tongue, call lower-level subroutines, which in turn
that reflect knowledge of what will be said call other, still lower-level subroutines. One
later. The spoonerism . . . queer old dean does not have to be a programmer nor sub-
instead of the presumably intended . . . dear scribe to the computer metaphor of mind to
old queen is an example. appreciate that nested structures are vital for
A rich set of data has been accumulated in the control of behavior. The concept of the
support of Lashleys hypothesis. Among the chunk introduced by Miller in his famous
relevant findings are changes in brain activity 195 6 paper embraced the concept of subrou-
that reliably predict distinguishing proper- tines while effectively sounding the death
ties of forthcoming actions (e.g., Jeannerod, knell for the prevailing computer metaphor
1988), changes in the time to initiate and of the time Shannon and Weavers (1949)
complete movement sequences as a function information theory. Chunking is the forma-
of their length and complexity (e.g., Klapp, tion of successively embedded data struc-
1977; Rosenbaum, 1987), changes in the tures used for storage of information or
way early features of movement sequences control of output. Shannon and Weavers
are performed depending on what move- information theory held that data content
ments will be performed later (e.g., Cohen could be meaningfully measured in terms of
& Rosenbaum, 2004), and analyses of errors the number of binary (yes/no) decisions
in speech (e.g., Dell, 1986; Fromkin, 1980), needed to uniquely identify an event. As
typewriting (Cooper, 1983 ), and other activ- Miller pointed out, however, the number of
ities (Norman, 1981; Reason, 1990). Modern items that can be held in immediate mem-
movement-recording techniques have also ory is not dictated by the number of such
shown that ongoing movements are suffi- binary decisions but instead is determined
ciently overlapped in time that it is diffi- by the number of meaningful elements or
cult to individuate successive movements, or chunks into which participant can orga-
even tell whether the order in which suc- nize presentations. People can store seven
cessive movements are completed comports plus or minus two nonsense syllables, but
with the order in which they were initiated. they can also store seven plus or minus two
In this connection, Grudin (1983 ) showed complex propositions comprised of scores of
that during typewriting, the fingers move syllables, provided the propositions fit with
toward their keys as soon as they can, such peoples experience. With proper training
that in skilled typists, the fingers move in a in memorizing random digits, people can
blur, not in a neatly sequenced chain. Find- boost their ostensive short-term memory
510 the cambridge handbook of expertise and expert performance
capacity by an order of magnitude (Chase & traversal routines are familiar to computer
Ericsson, 1981). users. To get to a directory, one opens the
The idea of chunking, proposed in the directory in which it is contained. Alterna-
late nineteenth century by Bryan and Harter tively, one could imagine that the terminal,
(1897, 1899) and revitalized by Miller or lowest-level, elements of the hierarchy
(195 6), can be shown to lead to the Power are simply read off in sequence without
Law of Learning (Newell & Rosenbloom, involving higher units. The available evi-
1981). If the time to form a higher-level dence suggests that this is not what happens.
chunk grows with the number of chunks it For example, when people recall hierarchi-
subsumes, and if the formation of a higher- cally organized keyboard sequences by key-
level chunk affords a reduction in the time ing them as rapidly as possible, the errors
to perform a task, then the time, T, to per- they make and the times between successive
form the task should decrease with the num- keystrokes are consistent with tree-traversal
ber of practice trials, P, in which the task is rather than rapid read-off of terminal nodes
practiced. Moreover, the rate at which the only (Povel & Collard, 1982; Restle, 1970;
time to perform the task decreases should in Rosenbaum, Kenny, & Derr, 1983 ; Simon,
turn decrease over practice trials. Expressed 1972). The same conclusion holds if the out-
mathematically, T = a Pb , where a and b put medium is speech rather than keyboard-
are nonnegative empirical constants. ing (Cooper & Paccia-Cooper, 1980).
If the psychological elements underlying The latter outcome reflects a similarity
perceptual-motor control are organized hier- between performance and perception. As
archically, they may be unpacked in a regular first shown by Sperling (1960), raw percep-
fashion for purposes of production. Accord- tual information is first entered in a limited-
ing to this view, higher-order elements are capacity buffer from which it is rapidly
unpacked into their immediate descendants, lost unless it is transferred to long-term
the descendants are in turn unpacked into storage. Unless the information is trans-
their descendants, and so on. Provided there ferred and allowed to make contact with
is a way of temporally ordering the unpack- already formed meaning-bearing representa-
ing of the descendants, one has a tree- tions, there is little chance it will be avail-
traversal process (see Figure 29.2). Tree- able later (Kroll & Potter, 1984). The data
perceptual-motor expertise 511
reviewed in the last paragraph lead to a sim- An important forerunner of current ideas
ilar conclusion for motor output. If one sup- about prediction was Helmholtz (1911), who
poses that there is a buffer for the output argued that perception relies, at least in part,
of motor commands, the buffer cannot be on unconscious inference: Given ambiguity
very large, for otherwise one would expect about the objective source of proximal stim-
motor commands for extended sequences of uli (e.g., images cast on the retina), one must
forthcoming motor acts to be read off seri- engage in a kind of unconscious problem
ally from the terminal elements of the hierar- solving to discern the objective source. See
chy without access to higher-level elements. Purves and Lott (2003 ) for a statistical ver-
Instead, a very small set of terminal elements sion of this theory. An instance of uncon-
is available for output. Between these out- scious inference that Helmholtz described
puts, retrieval is needed from higher-level was deciding whether the external environ-
representations. ment moves when ones eyes move. Eye
movements produce retinal image shifts, but
so do movements of the external environ-
Responding to Feedback ment when the eyes are still. How, then, does
one distinguish between these two events?
The second major concept in the cogni- Helmholtz and others following him (e.g.,
tive approach to perceptual-motor expertise von Holst & Mittelstaedt, 195 0) suggested
concerns prediction and feedback. When that copies of eye-muscle motor commands
one carries out a purposeful action, one typ- are sent to perceptual centers and these sig-
ically seeks to effect perceptible changes in nals are used to interpret subsequent visual
the environment. Whether the action is as input. One may think of this efference
mundane as lifting ones arm to stretch or copy as providing a basis for predicting the
as dramatic as raising ones arm to volun- perceptual consequences of imminent eye
teer for a dangerous mission, there is a cor- movements. In the case of initiating an eye
respondence between the plan underlying movement when the retinal image is station-
the action and the feedback one expects as a ary, the eye movement should cause a sweep
result of it. Of course, in the case of volun- of the image across the retina. When such
teering for a dangerous mission, feeling ones a predicted sweep occurs, the agent can be
arm rise up does not provide feedback about satisfied that the eye moved but the world
the success of the mission. By contrast, in did not.
the case of raising ones arm to stretch, feel- Forming predictions about the percep-
ing the arm ascend gives assurance that the tual consequences of voluntary actions helps
intended movement occurred as planned. A explain other phenomena for example,
challenge for theories of perceptual-motor why we dont see the world go dark when
skill is to understand how plans and feed- we blink our eyes (Volkman, Riggs, &
back are integrated over these very different Moore, 1980) and why we cant tickle our-
scales of experience. An equally important selves (Blakemore, Wolpert, & Frith, 1998).
question is how the interaction of plans and Other experimental demonstrations are sim-
feedback changes with practice. ilarly explained via prediction, including
For feedback to usefully guide behavior, motor imagery, imitation of others behav-
one must have a representation of what feed- ior, and the fact that reaction time is gen-
back is expected given the action one intends erally shorter for responses to stimuli that
to perform. In control theory, this represen- are similar to the perceptual effects the
tation is known as a reference condition. In responses engender than to stimuli that are
effect, a reference condition is a prediction not similar to the responses usual or learned
of what perceptual change will ensue as a perceptual effects. See Hommel, Musseler,
result of forthcoming action. The develop- Aschersleben, and Prinz (2001) for review.
ment of perceptual-motor expertise builds Many studies indicate that people learn
on the capacity to make such predictions. new perceptual-motor skills by forming ever
512 the cambridge handbook of expertise and expert performance
diminished in their ability to learn the motor plans (cf. Riley, Stoffregen, Grocki, &
sequence as compared to participants in the Turvey, 1999).
single-task condition. Cohen, Ivry, and Keele Both empirical and phenomenological
(1990) extended this finding by showing that data attest to the valuable role of atten-
dual-task interference with sequence learn- tion during early stages of perceptual-motor
ing scales with sequence complexity. Sim- skill acquisition. But what of later stages
ilarly, Leavitt (1979) showed that novice of expertise? The Fitts-Posner model and
hockey players were significantly impaired the controlled-automatic distinction sug-
in basic hockey skills when they simultane- gest that attention becomes unnecessary as
ously performed a visual shape-monitoring skill grows. Allport, Antonis, and Reynolds
task. By contrast, expert hockey players were (1972) found that skilled pianists could accu-
unaffected by the secondary task. In a simi- rately sight-read musical scores while repeat-
lar vein, Beilock, Weirenga, and Carr (2002) ing words presented through headphones.
had novice and expert golfers putt while In fact, too much attention can interfere
simultaneously monitoring a string of ver- with expert performance. The phenomenon
bally presented words. Not only did novices of choking is a well-known example of
putt less accurately during the secondary the way in which ones mind can get in
task, they also recognized fewer of the mon- the way of performance. When athletes are
itored words during a subsequent recogni- asked what causes them to choke, they
tion memory task as compared to the expert often report that thinking too hard about
golfers. These studies confirm the theoret- performance is what hurts them. Beilock,
ical insights of Fitts and Posner (1967) and Carr, MacMahon, and Starkes (2002) doc-
Shiffrin and Schneider (1977). umented the detrimental effects of such
If attention is necessary for unskilled per- attention. When their golfing subjects were
formance, to what should a skill-learner asked to attend to a specific element of their
attend? Wulf, Lauterbach, and Toole (1999) putting stroke, those participants were sig-
had novice golfers practice a pitching stroke nificantly less accurate than when they made
(in which the ball is lobbed a short dis- putts without such attention.
tance onto the putting green) under one
of two instructions. One instruction encour-
aged participants to focus on swinging their Ecological Psychology and the
arms, whereas the other instruction encour- Dynamical Systems Approach
aged participants to focus on swinging the
club. The latter instruction, which provided The second approach covered in this chapter
an external focus for the golfers attention, fits under the rubric of ecological psychology
was associated with more accurate pitch and dynamical systems analysis. These terms
shots than the former instruction. A sim- refer to perspectives that are traditionally
ilar result was obtained by Maddox, Wulf, linked, although their ancestries are quite
and Wright (1999), who found that novice different. The view associated with ecolog-
tennis players made more accurate forehand ical psychology stems from the thinking of
shots when they focused on the trajectory James Gibson (195 0, 1966, 1979), who con-
of the ball than when they focused on the tested Helmholtzs notion that sensory input
contact between the ball and the racket. is so impoverished that the observer must
These data suggest that skill acquisition is engage in unconscious inference to perceive
facilitated by attending to the consequences the environment. Gibson argued that owing
of ones actions rather than to the actions to the physical properties of objects and the
themselves. Wulf and Prinz (2001) reviewed way those objects emit or reflect energy, sen-
a number of studies that support this conclu- sory input is sufficiently rich and structured
sion and proposed that attending to the ele- to afford direct perception of the external
ments of a skill causes an over-regulation of world. Thus, the surface of a pond, if suf-
muscular degrees of freedom, which in turn ficiently smooth, is immediately perceived
limits ones ability to flexibly implement by a water spider to afford walking, whereas
514 the cambridge handbook of expertise and expert performance
Bracke-Tolkmitt, R., Linden, A., Canavan, Dell, G. S. (1986). A spreading activation theory
A. G. M., & Diener, H. C. (1989). The cere- of retrieval in sentence production. Psychologi-
bellum contributes to mental skills. Behavioral cal Review, 93 , 283 3 21.
Neuroscience, 103 , 442446. Desmond, J. E., Gabrielli, J. D., Wagner,
Brashers-Krug, T., Shadmehr, R., & Bizzi, E. A. D., Binier, B. L., & Glover, G. H.
(1996). Consolidation in human motor mem- (1997). Lobular patterns of cerebellar activa-
ory. Nature, 3 82 , 25 225 5 . tion of verbal working-memory and finger-
Bryan, W. L., & Harter, N. (1897). Studies in the tapping tasks as revealed by functional
physiology and psychology of the telegraphic MRI. Journal of Neuroscience, 17 , 9675
language. Psychological Review, 4, 275 3 . 9685 .
Bryan, W. L., & Harter, N. (1899). Studies on the Dizio, P., & Lackner, J. R. (1995 ). Motor adapta-
telegraphic language. Psychological Review, 6, tion to coriolis force perturbations of reaching
3 45 3 78. movements: Endpoint but not trajectory adap-
Calvin, S., Milliex L, Coyle, T., & Temprado, tation transfers to the nonexposed arm. Journal
J. J. (2004). Stabilization and destabiliza- of Neurophysiology, 74, 17871792.
tion of perception-action patterns influence Elbert, T., Pantev, C., Weinbruch, C., Rockstroh,
the self-organized recruitment of degrees of B., & Taub, E. (1995 ). Increased cortical repre-
freedom. Journal of Experimental Psychology: sentation of the fingers of the left hand. Science,
Human Perception and Performance, 3 0, 103 2 2 70, 3 05 3 07.
1042. Fajen, B. R., & Warren, W. H. (2003 ). Behav-
Carello, C., & Turvey, M. T. (2004). Physics and ioral dynamics of steering, obstacle avoidance,
psychology of the muscle sense. Current Direc- and route selection. Journal of Experimental
tions in Psychological Science, 13 , 25 28. Psychology: Human Perception and Performance,
Chase, W. G., & Ericsson, K. A. (1981). Skilled 2 9, 3 43 3 62.
memory. In J. R. Anderson (Ed.), Cognitive Fiez, J. A. (1996). Cerebellar contributions to cog-
skills and their acquisition (pp. 141189). Hills- nition. Neuron, 16, 13 15 .
dale, NJ: Erlbaum. Fitts, P. M., & Posner, M. I. (1967). Human perfor-
Cohen, A., Ivry, R. I., & Keele, S. W. (1990). mance. Belmont, CA: Brooks/Cole Publishing
Attention and structure in sequence learning. Company.
Journal of Experimental Psychology: Learning, Fromkin, V. A. (Ed.). (1980). Errors in linguistic
Memory, and Cognition, 16, 173 0. performance. New York: Academic Press.
Cohen, R. G., & Rosenbaum, D. A. (2004). Gao, J. H., Parsons, L. M., Bower, J. M., &
Where objects are grasped reveals how grasps Xiong, J. (1996). Cerebellum implicated in sen-
are planned: Generation and recall of motor sory acquisition and discrimination rather than
plans. Experimental Brain Research, 15 7 , 486 motor control. Science, 2 72 , 5 45 5 47.
495 . Gauthier, I., Skudlarski, P., Gore, J. C., & Ander-
Conditt, M. A., Gandolfo, F., & Mussa-Ivaldi, son, A. W. (2000). Expertise for cars and birds
F. A. (1997). The motor system does not learn recruit brain areas involved in face recognition.
the dynamics of the arm by rote memorization Nature Neuroscience, 3 , 191197.
of past experience. Journal of Neurophysiology, Gibson, E. J., & Walk, R. D. (1960). The visual
78, 5 5 45 60. cliff. Scientific American, 2 02 , 63 71.
Cooper, W. E. (Ed.). (1983 ). Cognitive aspects of Gibson, J. J. (195 0). Perception of the visual world.
skilled typewriting. New York: Springer-Verlag. Boston: Houghton-Mifflin.
Cooper, W. E., Paccia-Cooper, J. (1980). Syn- Gibson, J. J. (1966). The senses considered as per-
tax and speech. Cambridge: Harvard University ceptual systems. Boston: Houghton-Mifflin.
Press. Gibson, J. J. (1979). The ecological approach to
Courchesne, E., & Allen, G. (1997). Prediction visual perception. Boston: Houghton-Mifflin.
and preparation fundamental functions of the Glenberg, A. M. (1997). What memory is for.
cerebellum. Learning and Memory, 4, 13 5 . Behavioral and Brain Sciences, 2 0, 15 5 .
Dagenbach, D., & Carr, T. H. (Eds.). (1994). Goldfield, E. C., Kay, B. A., & Warren, W. H.
Inhibitory processes in attention, mem- (1993 ). Infant bouncing: The assembly and tun-
ory, and language. San Diego: Academic ing of action systems. Child Development, 64,
Press. 11281142.
518 the cambridge handbook of expertise and expert performance
Goodbody, S. J., & Wolpert, D. M. (1998). Tem- Kording, K. P., & Wolpert, D. M. (2004). Bayesian
poral and amplitude generalization in motor integration in sensorimotor learning. Nature,
learning. Journal of Neurophysiology, 79, 1825 43 7 , 244247.
183 8. Kroll, J. F., & Potter, M. C. (1984). Recognizing
Grudin, J. G. (1983 ). Error patterns in novice and words, pictures, and concepts: A comparison
skilled typists. In W. E. Cooper (Ed.), Cognitive of lexical, object, and reality decisions. Journal
aspects of skilled typewriting (pp. 121143 ). New of Verbal Learning and Verbal Behavior, 2 3 , 3 9
York: Springer-Verlag. 66.
Hebb, D. O. (1949). The organization of behavior: Landauer, T. K., & Bjork, R. A. (1978). Opti-
A neuropsychological theory. New York: John mal rehearsal patterns and name learning. In
Wiley. M. M. Gruneberg, P. E. Morris, & R. N. Sykes
Helmholtz, H. von (1911). Treatise on physiological (Eds.), Practical aspects of memory (pp. 625
optics. (J. P. Southall, Ed. & Trans.) (3 rd ed., 63 2). London: Academic Press.
Vols. 2 & 3 ). Rochester, NY: Optical Society of Lashley, K. S. (195 1). The problem of serial order
America. in behavior. In L. A. Jeffress (Ed.), Cerebral
Holmes, G. (193 9). The cerebellum of man. mechanisms in behavior (pp. 11213 1). New
Brain, 62 , 13 0. York: Wiley.
Hommel, B., Musseler, J., Aschersleben, G., & Leavitt, J. (1979). Cognitive demands of skating
Prinz, W. (2001). The theory of event cod- and stick handling in ice hockey. Canadian
ing (TEC): A framework for perception and Journal of Applied Sport Sciences, 4, 465 5 .
action planning. Behavioral and Brain Sciences, Leiner, H. C., Leiner, A. L., & Dow, R. S. (1995 ).
2 4, 84993 7. The underestimated cerebellum. Human Brain
Ivry, R. B., & Keele, S. W. (1989). Timing func- Mapping, 2 , 24425 4.
tions of the cerebellum. Journal of Cognitive Lewandowsky, S., Kalish, M., & Ngang, S. K.
Neuroscience, 1, 13 615 2. (2002). Simplified learning in complex
Jeannerod, M. (1988). The neural and behavioral situations: Knowledge partitioning in function
organization of goal-directed movements. Oxford: learning. Journal of Experimental Psychology:
Oxford University Press. General, 13 1, 163 193 .
Jenkins, I., Brooks, D., Nixon, P., Frackowiack, Logan, G. D. (1988). Toward an instance theory of
R., & Passignham, R. (1994). Motor sequence automatization. Psychological Review, 95 , 492
learning: A study with positron emission 5 27.
tomography. Journal of Neuroscience, 14, 3 775 Logan, G. D. (2002). An instance theory of atten-
3 790. tion and memory. Psychological Review, 109,
Kaas, J. H., Merzenich, M. M., & Killackey, H. P. 3 76400.
(1983 ). The reorganization of somatosensory Logothetis, N. K., & Pauls, J. (1995 ). Psychophys-
cortex following peripheral nerve damage in ical and physiological evidence for viewer-
adult and developing mammals. Annual Review centered object representations in the primate.
of Neuroscience, 6, 3 25 3 5 6. Cerebral Cortex, 5 , 270288.
Kandel, E. R. (1981). Nerve cells and behavior. Lynch, G., & Baudry, M. (1984). The biochem-
In E. R. Kandel & J. H. Schwartz (Eds.), Prin- istry of memory: A new and specific hypothesis.
ciples of neural science (pp. 1423 ). New York: Science, 2 2 4, 105 71063 .
Elsevier/North-Holland. Maddox, M. D., Wulf, G., & Wright, D. L. (1999).
Kerr, B. A. & Langgolf, G. (1977). The speed of The effect of an internal vs. external focus of
aiming moveents. Quarterly Journal of Experi- attention on the learning of a tennis stroke.
mental Psychology, 2 9, 475 481. Journal of Exercise Psychology, 2 1, S78.
Klapp, S. T. (1977). Reaction time analysis of pro- Magill, R. A., & Hall, K. G. (1990). A review of
grammed control. Exercise and sport sciences the contextual interference effect in motor skill
reviews, 5 , 23 125 3 . acquisition. Human Movement Science, 9, 241
Koh, K., & Meyer, D. E. (1991). Induction 289.
of continuous stimulus-response associations Melton, A. W. (1970). The situation with respect
for perceptual-motor performance. Journal of to the spacing of repetitions and memory. Jour-
Experimental Psychology: Learning, Memory, nal of Verbal Learning and Verbal Behavior, 9,
and Cognition, 17 , 81183 6. 5 96606.
perceptual-motor expertise 519
Meyer, D. E., Abrams, R. A., Kornblum, S., Rosenbaum, D. A. (1987). Successive approxima-
Wright, C. E., & Smith, J. E. K. (1988). Opti- tions to a model of human motor programming.
mality in human motor performance: Ideal In G. H. Bower (Ed.) Psychology of learning and
control of rapid aimed movements. Psychologi- motivation, Vol. 2 1 (pp. 15 3 182). Orlando, FL:
cal Review, 95 , 3 403 70. Academic Press.
Miller, G. A. (195 6). The magical number seven Rosenbaum, D. A. (1998). Is dynamical systems
plus or minus two: Some limits on our capac- modeling just curve fitting? Motor Control, 2 ,
ity for processing information. Psychological 101104.
Review, 63 , 8197. Rosenbaum, D. A., Carlson, R. A. & Gilmore,
Newell, A. M., & Rosenbloom, P. S. (1981). Mech- R. O. (2001). Acquisition of intellectual and
anisms of skill acquisition and the law of prac- perceptual-motor skills. Annual Review of Psy-
tice. In J. R. Anderson (Ed.), Cognitive skills chology, 5 2 , 45 3 470.
and their acquisition (pp. 15 5 ). Hillsdale, NJ: Rosenbaum, D. A., Kenny, S., & Derr, M. A.
Erlbaum. (1983 ). Hierarchical control of rapid move-
Newell, K. M., Liu, Y-T., & Mayer-Kress, ment sequences. Journal of Experimental
G. (2001). Time scales in motor learning Psychology: Human Perception and Performance,
and development. Psychological Review, 108, 9, 86102.
5 782. Rosenbaum, D. A., Meulenbroek, R. G.,
Nissen, M. J., & Bullemer, P. (1987). Attentional Vaughan, J., & Jansen, C. (2001). Posture-based
requirements of learning: Evidence from per- motion planning: Applications to grasping. Psy-
formance measures. Cognitive Psychology, 19, 1 chological Review, 108, 70973 4.
3 2. Sainburg, R. L. (2002). Evidence for a dynamic
Norman, D. A. (1981). Categorization of action dominance hypothesis of handedness. Experi-
slips. Psychological Review, 88, 115 . mental Brain Research, 142 , 24125 8.
Petersen, S. E., Fox, P. T., Posner, M. I., Mintun, Schmidt, R. A., & Bjork, R. A. (1992). New
M., & Raichle, M. E. (1989). Positron emission conceptualizations of practice: Common prin-
tomographic studies of the processing of single ciples in three paradigms suggest new con-
words. Journal of Cognitive Neuroscience, 1, 15 3 cepts for training. Psychological Science, 3 ,
170. 207214.
Povel, D-J., & Collard, R. (1982). Structural fac- Schmidt, R. A. & Lee, T. D. (1999). Motor control
tors in patterned finger tapping. Acta Psycho- and learning A behavioral emphasis (3 rd Ed.).
logica, 5 2 , 107124. Champaign, Illinois: Human Kinetics.
Proffitt, D. R., Bhalla, M., Gossweiler, R., & Schneider, W., & Chien, J. M. (2003 ). Controlled
Midgett, K. (1995 ). Perceiving geographical & automatic processing: Behavior, theory, and
slant. Psychonomic Bulletin & Review, 2 , 409 biological mechanisms. Cognitive Science, 2 7 ,
428. 5 25 5 5 9.
Purves, D., & Lott, R. B. (2003 ). Why we see what Shannon, C., & Weaver, W. (1949). The math-
we do: An empirical theory of vision. Sunderland, ematical theory of communication. Urbana, IL:
MA: Sinauer Associates. University of Illinois Press.
Rea, C. P., & Modigliani, V. (1985 ). The Shea, J. B., & Morgan, R. L. (1979). Contextual
effect of expanded versus massed prac- interference effects on acquisition, retention,
tice on the retention of multiplication and transfer of a motor skill. Journal of Experi-
facts and spelling lists. Human Learning, 4, mental Psychology: Human Learning and Mem-
1118. ory, 5 , 179187.
Reason, J. (1990). Human error. Cambridge: Shiffrin, R. M., & Schneider, W. (1977).
Cambridge University Press. Controlled and automatic human infor-
Restle, F. (1970). Theory of serial pattern mation processing. II. Perceptual learn-
learning: Structural trees. Psychological Review, ing, automatic attending, and a general
77 , 481495 . theory. Psychological Review, 84 (2), 127
Riley, M. A., Stoffregen, T. A., Grocki, M. J., & 190.
Turvey, M. T. (1999). Postural stabilization for Simon, H. A. (1972). Complexity and the rep-
the control of touching. Human Movement Sci- resentation of patterned sequences of symbols.
ence, 18, 795 817. Psychological Review, 79, 3 693 82.
52 0 the cambridge handbook of expertise and expert performance
Sperling, G. A. (1960). The information available von Holst, E., & Mittelstaedt, H. (195 0).
in brief visual presentation. Psychological Mono- Das Reafferenzprinzip. Die Naturwis-
graphs, 74, Whole No. 498. senschaften, 3 7, 464474. (English translation
Starkes, J. L. & Allard, F. (Eds.). (1993 ). Cognitive in P. C. Dodwell [Ed.], [1980], Perceptual
issues in motor expertise. Amsterdam: North- processing: Stimulus equivalence and pattern
Holland. recognition. New York: Appleton-Century-
Starkes, J. L., & Ericsson, K. A. (Eds.). (2003 ). Crofts.)
Expert peformance in sports: Advances in research Worringham, C. J. (1993 ). Predicting motor per-
on sport expertise. Champaign, IL: Human formance from variability measures. In K. M.
Kinetics. Newell & D. M. Corcos (Eds.), Variability
Sternad, D., Duarte, M., Katsumata, H., & and motor control (pp. 5 3 63 ). Champaign, IL:
Schaal, S. (2001). Bouncing a ball: Tuning Human Kinetics.
into dynamic stability. Journal of Experimental Wulf, G., Lauterbach, B., & Toole, T. (1999).
Psychology: Human Perception and Performance, Learning advantages of an external focus of
2 7 , 1163 1184. attention in golf. Research Quarterly for Exer-
Turvey, M. T. (1990). Coordination. American cise & Sport, 70, 120126.
Psychologist, 45 , 8, 93 895 3 . Wulf, G., & Prinz, W. (2001). Directing atten-
Vereijken, B., Whiting, H. T. A., & Beek, P. J. tion to movement effects enhances learning: A
(1992). A dynamic-systems approach to skill review. Psychonomic Bulletin & Review, 8 (4),
acquisition. Quarterly Journal of Experimental 648660.
Psychology Section A: Human Experimental Psy- Zanone, P. G., & Kelso, J. A. S. (1997). Coor-
chology, 45 , 3 23 3 44. dination dynamics of learning and transfer:
Volkmann, F. C., Riggs, L. A., & Moore, R. K. Collective and component levels. Journal of
(1980). Eyeblinks and visual suppression. Sci- Experimental Psychology: Human Perception
ence, 2 07 , 900902. And Performance, 2 3 , 145 41480.
P a r t V.C
C H A P T E R 30
Expertise in Chess
compared to a novice player. This dissocia- level rating scale enables fine-grained exam-
tion, a finding that has become a touchstone ination of the relation between expertise and
of the expert-performance approach in a variety of indicators of psychological pro-
many other domains, suggested that acqui- cesses. Measurement of expertise on this fine
red patterns, not innate abilities, accounted of a scale remains a central problem for many
for skill differences. On the basis of these other domains discussed in this volume.
data and those gathered in other exper- For instance, a psychometric approach to
iments and in simulation studies, Chase chess skill (e.g., Van der Maas & Wagenmak-
and Simon proposed their highly influential ers, 2005 ) can capitalize on the chess rat-
chunking theory of skilled performance in ing scale to examine how well it correlates
chess. That theory and its subsequent refine- with different markers of psychological pro-
ment has had a significant impact on exper- cesses such as measures of memory, prob-
tise research in general and that on games lem solving, and motivation. Early efforts at
in particular (see Gobet, de Voogt, & Rets- understanding skill in chess implicitly made
chitzki, 2004, for an extensive coverage on use of this correlates approach for measures
board games, and Charness, 1989, for a pre- of attention (Tikhomirov & Poznyanskaya,
sentation of the data on bridge). 1966), imagery (Milojkovic, 1982; Bachmann
& Oit, 1992), and personality (Charness,
Tuffiash, & Jastrzembski, 2004).
Brief Description of the Game In this chapter we focus on a process-
and the Rating System model approach to understanding exper-
tise in chess. Our goal is to shed light on
Chess is a game played by two opponents the process of how players choose the best
using an initial configuration, the starting moves to play in a chess game, starting from
position, consisting of 3 2 chess pieces placed early perceptual processes and tracing for-
on an eight-by-eight square chessboard. The ward to the search processes first described
rules of chess are sufficiently simple that by de Groot. We outline where skill dif-
children can be taught them at a very young ferences arise within such processes. We
age (four or five years old). Child prodi- describe computer simulation models that
gies are not uncommon and teenagers have capture some of the features of skilled per-
been able to compete at the highest level. formance by chess players. We also describe
(Nineteen-year-old Ruslan Ponomariov was how human players acquire the knowledge
crowned World Champion in 2002.) Chess is necessary to play chess expertly.
sufficiently difficult to play well that it took
about 40 years of effort by computer scien-
tists to program computers to compete on Information Processing Models
an equal level with the best human players. of Choosing a Good Move: The
Another important feature for chess is Trade-Offs Between Knowledge
the existence of a sophisticated measure- and Search
ment scale for evaluating chess skill based on
performance in chess tournaments. The Elo The goal of a chess player is to choose the
rating scale, available since the mid-1960s best possible move. Often, when playing
(Elo, 1965 ; Elo, 1986), is open-ended, start- through standard openings, the best move is
ing at a nominal value of zero and extending dictated by knowledge from published anal-
upward, with a nominal class interval (stan- yses, such as the Encyclopedia of Chess Open-
dard deviation) of about 200 rating points. ings. Sometimes, detecting the best move in
The worlds best players today hover above a sequence of exchanges of pieces is sim-
2800 rating points with Grandmaster level ple enough that even novices quickly find
at approximately 25 00 rating points, Inter- it. Much of the time, the best move is non-
national Master at about 2400 points, and obvious and the player must decide based on
Master at about 2200 points. This interval a search process that evaluates a candidate
expertise in chess 52 5
move in terms of potential future positions attention has shifted from investigating
reached via a branching tree of available search processes to understanding the role of
moves for the two sides. pattern-recognition processes in move selec-
Search is difficult because of the enor- tion. As de Groot noted, skilled players
mous number of possible moves stemming use their knowledge about chess configura-
from the opening position. Even Master tions to generate plausible moves for limited
players who can use well-tuned recognition searching. We now focus on understanding
processes to winnow the possible base moves the perceptual mechanisms that support this
down to three or four plausible alternatives rapid perception advantage.
at each point in the tree face a dilemma.
The number of possible moves, computed
Tracing Expertise Differences in
as breadth raised to a power equal to depth,
Perception and Attention with
is 4 76 moves, given that the average mas-
Eye-Tracking Techniques
ter game lasts about 3 8 moves or 76 plies
(moves for each side). So, both computers Jongman (1968) initiated work on percep-
and humans must search selectively among tual skill differences by examining eye move-
the alternatives using a variety of heuris- ments of expert and less-expert players,
tics (Newell & Simon, 1972) to decide when though his results became widely accessi-
a node reached in search can be properly ble with the re-analysis published by de
evaluated. Groot and Gobet (1996). These researchers
Given the relatively slow rate at which showed that in a memorizing task, where
moderately skilled players can generate anal- players were given a few seconds to exam-
ysis moves, estimated in Charness (1981b) ine an unfamiliar chess position, better play-
to be about four moves per minute, it is ers fixated more on the edges of squares
obvious that much of the time that human than weaker players did. Also, better players
players spend is not in generating all possi- were more likely to have greater distances
ble moves (perhaps taking a move per sec- between fixations, implying that they were
ond) but in generating moves selectively and able to encode more widely about a fixa-
using complex evaluation functions to assess tion than weaker players. Experts also made
their value. Computer chess programs can shorter duration fixations than did weaker
achieve high-level play by searching many players implying faster encoding.
moves using fast, frugal evaluation processes Reingold and colleagues confirmed the
that involve minimal chess knowledge to larger visual span for experts using a variety
evaluate the terminal positions in search. of tasks. Using a gaze-contingent paradigm
Deep Blue, the chess program that defeated that manipulated the number of visible
World Champion Garry Kasparov in a short squares, Reingold, Charness, Pomplun, and
match in 1997, searched hundreds of mil- Stampe (2001) showed that more-skilled
lions of positions per second. Todays leading players needed a larger area around fixation
microcomputer chess programs, which have to detect changes in successively displayed
drawn matches with the best human play- chess positions in order to match perfor-
ers, have sophisticated search algorithms and mance under unlimited view of the whole
attempt to use more chess knowledge but board. This was true only for structured,
still generate hundreds of thousands or mil- not random, chess positions. This result sug-
lions of chess moves per second. Generally, gested that better players had a larger visual
chess programs rely on search more heavily field from which they could extract chess
than knowledge; for humans it is the reverse. relationships. In a second experiment, the
Yet, each can achieve very high performance authors noted that when players made a sim-
levels because knowledge and search can ple determination of whether a King was
trade off (Berliner & Ebeling, 1989). in check by an attacking piece on a mini-
Because expert humans do so little search, mized chessboard (three-by-three squares),
yet still manage to find strong chess moves, experts required fewer fixations to decide,
52 6 the cambridge handbook of expertise and expert performance
and these fixations were more likely to be extracting, in parallel, chess relationships
between pieces (on empty squares), com- critical to choosing good moves. In later sec-
pared to intermediate-level players. (See also tions we outline how CHREST, a computer
Fisk & Lloyd, 1988, and Saariluoma, 1985 , simulation program, acquires and utilizes
for data on perceptual processes in simple such chess patterns (templates and chunks).
decision tasks.)
In a choose-a-move task using full chess- Memory Processes
boards, Charness, Reingold, Pomplun, and
Stampe (2001) demonstrated that experts As early as Binet (1894), and in particular in
made fewer fixations per trial and those fix- de Groots work, knowledge has been iden-
ations were more widely spaced out across tified as a key component of chess exper-
the board and again, more likely to be tise. In order to understand how knowl-
between than on chess pieces. More impor- edge (held in memory) mediates skill, a
tant, for the first five fixations, experts were substantial amount of research has been car-
more likely to fixate on relevant squares ried out on chess players memory. Domains
(rated as relevant by a strong International of interest include memory for static posi-
Master player). This very early advantage tions, memory for moves and sequences of
(within the first second of exposure to moves (discussed later in the section on
a new position, given that fixations aver- blindfold chess), and the structure and con-
age about 25 0 milliseconds each for both tents of long-term memory (LTM), includ-
experts and intermediate players) testifies ing the number of chunks necessary to reach
to the importance of pattern-recognition expert performance. In many cases, experi-
processes in providing a better representa- mentation has been carried out in concert
tional structure. Given this perceptual head with computational modeling.
start, experts also chose better moves and
did so more quickly than their less-expert memory recall for positions: chase
counterparts. and simons key results
Reingold, Charness, Schultetus and Although both Binet and de Groot high-
Stampe (2001) used a Stroop-like inter- lighted the key role of knowledge in chess
ference task within a five-by-five square expertise, one had to wait until Chase and
segment of a chessboard to demonstrate Simons work in 1973 to have a detailed the-
that expert players appear to extract chess ory of expert memory. Extending de Groots
relations in parallel, whereas weaker ones study showing a striking skill effect in the
appear to shift attention and encode the recall of game positions, Chase and Simon
same relations serially. In a two-attacker carried out detailed analyses to identify what
situation, supplying a cue about which were the building blocks of chess knowledge.
piece to attend to provided an advantage In a copy task, they analyzed the pattern of
in response time to less-skilled players, but eye fixations on the stimulus board, as well as
no advantage to expert players because the way pieces were grouped during recon-
the latter appeared to encode both attack struction. Comparing these results with
relations simultaneously. those obtained in a recall task, they inferred
In summary, experts rely on a rich net- that pieces placed within two seconds and
work of chess patterns stored in long-term sharing a number of semantic relations were
memory structures (or long-term working likely to belong to the same chunk. (These
memory, Ericsson & Kintsch, 1995 ) to give results were replicated by Gobet & Simon,
them a larger visual span when encoding 1998, and Gobet & Clarkson, 2004.) They
chess positions. They encode chess informa- proposed that skill did not reside in differ-
tion far more quickly and accurately than ences in short-term memory (STM) capac-
non-experts. Within the first second of expo- ity or encoding speed, but in the num-
sure to a new position, experts are exam- ber of chunks held in LTM memory. These
ining salient squares on the chessboard and chunks give access to information such as
expertise in chess 52 7
what move to play, what plan to follow, and a simulated eye. Each cognitive process
and what evaluation to give to (part of) the has a time cost; for example, it takes 5 0
position. Thus, their theory explained both milliseconds to place a chunk in STM, and
why masters choose better moves in spite eight seconds to create a new chunk. Dur-
of their selective search (because chunks ing the learning phase, the program auto-
enable them to identify the key features of matically acquires chunks and templates by
a position and guide search during look- scanning a database of positions taken from
ahead) and why they perform better in a masters games. During the testing phase,
memory task (because they can partition the it is placed in the same experimental situa-
position in relatively large groups of pieces, tion as human participants. The program has
unlike weaker players who have to use more simulated various characteristics of players
smaller groups, which overtax STM). Some eye movements (de Groot & Gobet, 1996),
of these ideas were implemented in a com- the details of reconstruction in recall experi-
puter program, MAPP (Simon & Gilmartin, ments (Gobet, 1993 ; Gobet & Simon, 2000;
1973 ), which simulated recall up to expert Gobet & Waters, 2003 ), as well as the way
level. novices learn to memorize chess positions
(Gobet & Jackson, 2002). Beyond chess,
problems with the chunking theory lead variants of the program have been applied
to the template theory to memory for computer programs, use of
A number of experiments have helped diagrammatic representation in physics, con-
refined Chase and Simons theory. Charness cept formation, and childrens acquisition of
(1976) showed that the presence of an inter- language (Gobet et al., 2001).
fering task reduced recall only marginally,
which runs counter to the assumption of a random positions
slow encoding in LTM. Several authors (Frey
& Adesman, 1976; Cooke et al., 1993 ; Gobet Whereas Chase and Simon (1973 a, 1973 b)
& Simon, 1996a) have shown that players found no skill difference in the recall of ran-
can remember multiple boards reasonably dom positions, Gobet and Simon (1996b)
well, which again highlighted a weakness of show that later studies did in fact find such
the original chunking theory. a difference, although the effect is rarely
These results, as well as the fact that significant because of the low statistical
verbal protocols reveal that masters use power within these studies. This skill effect
larger structures than the chunks identi- remains across a wide range of presenta-
fied by Chase and Simon (e.g., de Groot, tion times (from one second to 60 seconds;
1946; de Groot & Gobet, 1996; Freyhoff, Gobet & Simon, 2000) and with positions
Gruber, & Ziegler, 1992; Gobet, 1998a), where the location as well as the distri-
led Gobet and Simon (1996a, 2000) to bution of pieces is randomized (Gobet &
revise the chunking theory. Their template Waters, 2003 ). These results are consistent
theory aimed to remedy these weaknesses with the chunking and template theories,
while keeping the strengths of the origi- which predict that strong players, who have
nal chunking theory. It also aimed to show more chunks, are more likely than weaker
how high-level, schematic structures (tem- players to recognize some patterns even with
plates) can evolve from perceptual chunks. random positions. Indeed, computer simu-
As with the chunking theory, chunks and lations (Gobet & Simon, 1996b; Gobet &
now templates are crucial in explaining how Simon, 2000) confirmed these predictions.
players access relevant information by pat-
tern recognition. The computer program new estimates of the vocabulary
CHREST (Chunk Hierarchy and REtrieval of the master
STructures) implements aspects of the tem- Based on computer simulations with MAPP,
plate theory. CHREST consists of an STM, Simon and Gilmartin (1973 ) estimated that
an LTM indexed by a discrimination net, one needed to acquire from 10,000 to
52 8 the cambridge handbook of expertise and expert performance
100,000 patterns to reach master level in shelf microcomputers, they are of world-
chess. These estimates have led to sev- championship caliber. One could argue that
eral experiments, in part because Holding knowledge about human problem-solving
(1985 ) argued that a much smaller num- processes in chess has lagged the efforts in
ber was required if one assumed that the artificial intelligence, though steady progress
same chunk could encode the same constel- is evident. De Groot (1946/1965 ), Newell
lation of pieces placed at different locations and Simon (1972), and Wagner and Scurrah
of the board. Saariluoma (1994) modified (1971) have generated many of the explicit
positions by swapping quadrants, and Gobet models that describe the heuristics used by
and Simon (1996c) modified positions by humans to manage search.
taking their mirror images along various axes
of symmetry. Both found that these manipu- de groots study
lations affected recall, which runs counter to
De Groot (1946) asked his participants to
Holdings predictions but supports the origi-
think aloud when choosing their next move
nal chunking theory. Computer simulations
in a problem position. The quantitative and
with CHREST show that at least 3 00,000
qualitative measures he extracted from the
chunks are required to reach Grandmaster
verbal protocols provided important empir-
level, even with the presence of templates,
ical information about chess players think-
which were not part of the chunking theory.
ing. We will use some of the main phenom-
ena discussed by de Groot to organize this
recognition experiments
section.
In the past, few studies have been carried out
using a recognition paradigm (e.g., Goldin, macrostructure of search in chess
1978, 1979; Saariluoma, 1984). This pattern
has not changed in recent years, and we De Groot (1946) found few differences
could find only one study using this tech- in the macrostructure of search between
nique. McGregor and Howes (2002) pre- world-class grandmasters and relatively
sented positions for either nine or 3 0 seconds strong players (candidate masters). Surpris-
asked participants to evaluate them, and ingly, during their search, players from both
later carried out a recognition test. In one skill levels tended to search at similar depth,
experiment, the positions presented dur- to consider the same number of positions,
ing the recognition phase were distorted by and to propose similar numbers of candidate
shifting either all pieces or a single piece moves. But there were differences as well:
one square horizontally. Two further exper- the grandmasters chose better moves than
iments used a priming technique during the the candidate masters, they generated moves
recognition phase: a piece from the target faster, they reached a decision faster, and,
position was shown for two seconds; this during their search, they examined moves
was followed by a second piece, and par- and sequences of moves that tended to be
ticipants indicated whether they thought it more relevant.
was in the target position. The two pieces Holding (1985 ) argued that de Groots
shared either a relation of attack/defence or a (1946) small sample (five grandmasters and
relation of proximity. McGregor and Howes five candidate masters) may have con-
found that Class A players used information cealed existing skill differences. Supporting
about attack/defence more often that infor- Holdings view, some skill differences were
mation about the location of pieces. found with samples including weaker play-
ers (e.g., Charness, 1981b; Gobet, 1998b;
Saariluoma, 1992). Charness (1981a) sug-
Problem-Solving Processes
gested that depth of search increases up to
Todays chess-playing programs benefit from candidate master level, after which it stays
the enormous progress in refining com- uniform. Charness (1989) conducted a nine-
puter search algorithms. Running on off-the- year longitudinal investigation of a Canadian
expertise in chess 52 9
player who advanced (in power law fash- in thinking time fails to increase the num-
ion) from an average-level performance ber of blunders substantially, which counts
(1600 rating points) to International Mas- as direct support for theories emphasizing
ter level performance (23 00 rating points) pattern recognition.
and found no significant increase in depth of Proponents of search models often cite
search. However, international masters and Holding and Reynolds (1982) experiment as
grandmasters sometimes carry out shallower evidence that search and pattern recognition
search than masters (Saariluoma, 1990), per- can be dissociated. Holding and Reynolds,
haps indicating that they can tailor their who used semi-random positions as stim-
search mechanisms to the demands of the uli, first asked players to recall the position
position. Gobet (1997) carried out computer after an eight-second presentation, and then
simulations with the SEARCH model (see to choose what they thought would be the
below) and concluded that average depth of best move. They found that skill correlated
search keeps increasing with higher skill lev- with the quality of chosen move after a few
els, but with diminishing returns (i.e., it fol- minutes deliberation, but not with the recall
lows a power law). or the evaluation after brief presentation.
Schultetus and Charness (1999) extended
selective search, move generation, Holding and Reynolds (1982) experiment
and pattern recognition with a crucial addition: they asked players
De Groot found that all players were highly to recall the position at the end of problem
selective in their search, rarely visiting more solving. Like in the original study, stronger
than one hundred nodes before choosing players did not recall the position better after
a move. Recent results support this view. five seconds, but chose better moves. How-
Calderwood et al. (1988) showed that mas- ever, they also obtained better results in the
ters can make relatively good decisions even recall performance following problem solv-
under time pressure (about five seconds per ing. Schultetus and Charness (1999) argue
move). Gobet and Simon (1996d) found that that these results are consistent with the
world champion Kasparov, when playing hypothesis that pattern recognition under-
simultaneous games against teams consist- pins skill in chess. That is, in order to choose
ing of up to eight international masters and better moves, better players were able to
grandmasters, performed at a level that still form new relational patterns for the unusual
placed him among about the six best play- piece placements. These new chunks pro-
ers in the world. Gobet and Simon argued vided the recall advantage after problem
that, although Kasparovs performance was solving. Such results are also consistent with
weaker than in normal games and showed the Ericsson and Kintsch (1995 ) long-term
more variability, it was higher than theories working-memory perspective.
mainly based on search would predict (but
see Lassiter, 2000, and Chabris & Hearst,
2003 , for opposing views). Comparing the progressive deepening
quality of play of world-class grandmasters De Groot (1946) found that players were
in standard games (about three minutes per visiting the same branches of the search
move, on average) and rapid games (less than tree repeatedly, either directly or after visit-
3 0 seconds per move, on average), Chabris ing other branches. According to de Groot,
and Hearst (2003 ) found that this decrease this phenomenon of progressive deepen-
of thinking time by a factor of six only ing occurs both in order to compensate
marginally affected the number of blunders for limitations in memory and for propa-
per 1,000 moves (5 .02 in classical games vs. gating information from one branch of the
6.85 in rapid games). Although they took search tree to another (de Groot, 1946; de
this as evidence for the role of search, a Groot & Gobet, 1996). Gobet (1998b) found
more natural interpretation of these results that skill affects how progressive deepen-
is that they show that a substantial decrease ing is carried out. The maximum number
530 the cambridge handbook of expertise and expert performance
of immediate re-investigations (where the ten by Wilkins (1980) and Pitrat (1977).
same base move is analyzed directly in the Gobet and Jansen (1994) describe a program
next episode) was proportional to players that uses pure pattern recognition to select
strength, whereas the maximum number of moves, without carrying out any search. This
non-immediate re-investigations (where the is presented more as a first step toward
analysis of a base move and its reinvestiga- a full problem-solving program than as a
tion is interrupted by the analysis of at least theory of human problem solving. Gobet
one different move) was inversely propor- (1997) describes SEARCH, a probabilistic
tional to players strength. model that integrates pattern recognition
and search. This model, which is a direct
high-level knowledge and planning implementation of the template theory and
which incorporates insights from previous
De Groot (1946), who emphasized the role
theories (e.g., de Groot, 1946; Newell &
of conceptual knowledge in chess expertise,
Simon, 1972), does not play chess but com-
reported that players descriptions of games
putes several measures such as depth of
were centered on key positions; this finding
search, rate of search, and the level of fuzzi-
has been confirmed by recent research (e.g.,
ness in the minds eye as a function of the
Cooke et al., 1993 ; de Groot & Gobet, 1996;
skill level (i.e., number of chunks).
Saariluoma, 1995 ). The presence of these
When generating moves from the stimu-
key positions enables masters to acquire
lus position or later during look-ahead, the
what de Groot called a system of playing
model uses either fast pattern recognition
methods, many of which are stereotypical.
(chunks and templates) or slower heuristics.
By applying this routine knowledge, mas-
The same methods are used when the model
ters can often find good moves with min-
evaluates positions at the end of a sequence
imal look-ahead. Saariluoma (1990, 1992)
of moves. The generation of an episode
tested this hypothesis with tactical posi-
(sequence of moves) is stopped when the
tions and found that strong players tended
level of fuzziness in the minds eye is too
to choose stereotyped solutions and missed
high, an evaluation has been proposed, or
shorter (but non-typical) solutions.
no move or sequence of moves has been pro-
Saariluoma and Hohlfeld (1994) were
posed. It is assumed that information in the
interested in planning with strategic posi-
minds eye decays at a constant rate, which
tions. They found that null moves (missing
interferes with search. Finally, the model has
moves for one side) were common (about
a time cost for every cognitive operation; for
12% of all moves). This result is similar to the
example, it takes two seconds to carry out a
10% found in a previous study by Charness
move in the minds eye, and ten seconds to
(1981a). In a second experiment, Saariluoma
evaluate a position using heuristics.
and Hohlfeld (1994) changed the nature of
The program predicts that depth of search
positions by relocating a key piece so that a
follows a power law of skill. When simu-
combination possible before the transforma-
lating a small number of participants (as is
tion could not be carried out after. Eliminat-
typical in chess research), the program also
ing the combination produced an increase of
shows substantial variability, as was found
the number of null moves.
in Saariluomas (1992) study, where inter-
national masters and grandmasters searched
computational models of problem solving less than weaker masters.
Simon and his colleagues developed a num-
ber of process models of problem solving
Blindfold Chess
in chess (Baylor & Simon, 1966; Newell
& Simon, 1972), and two production sys- A number of studies have investigated blind-
tems at the boundary between cognitive sci- fold chess, to which Binet (1894) had already
ence and artificial intelligence were writ- devoted a lengthy study. In blindfold chess, a
expertise in chess 531
player carries out one or several games with- because they only slowly produce positions
out the view of the board and the pieces; the where it is not possible to recognize chunks.
moves are communicated through standard These results are in line with Chase and
chess notation. Simons (1973 b), who studied memory for
Ericsson and Oliver (1984; cited in Eric- moves with plain view of the board.
sson & Staszewski, 1989) investigated the In a series of experiments, Saarilu-
nature of the representation for chess posi- oma (1991) and Saariluoma and Kalakoski
tions dictated move by move with a player of (1997) systematically investigated memory
near Master strength. In a set of experiments for blindfold games. They presented one
they demonstrated that the retrieval struc- or several games aurally (dictating moves
ture utilized by the player was extremely using the standard algebraic chess nota-
flexible, permitting very fast responses about tion) or visually (presenting only the cur-
what piece was present (or if the square rent move on a computer screen). Only a
was unoccupied) to probes (square names few of their results can be presented here:
in algebraic notation) of each square on blindfold chess requires mainly visuo-spatial
the chessboard. In some cases the response working memory, and makes little use of
time of about two seconds was faster than verbal working memory; differences in LTM
that obtained when the player looked at a knowledge (e.g., number of chunks) rather
chessboard position and was probed with than differences in imagery ability underpin
the name of a square. They also found that skill differences; abstract representations are
when the player memorized two chess posi- essential (cf. also Binets, 1894); and there
tions, responses to probes of the squares is no difference between an auditory and a
of the chessboard improved with succes- visual presentation. Campitelli and Gobet
sive tests from the same board at a much (2005 ) used blindfold chess to study how
faster rate than in conditions where there perception filters out relevant from irrele-
were random probes or when probes alter- vant information. They found that irrelevant
nated across boards. In further experiments information affects chess masters only when
they demonstrated that the representation it changes during the presentation of the tar-
structure was different for memory retrieval get game.
versus perceptually available retrieval con- Problem solving has also been stud-
ditions. Such flexibility in the encoded rep- ied using blindfold chess (Saariluoma &
resentation is necessary for being able to Kalakoski, 1998). In a task consisting in
choose good moves when playing blind- searching for the best move, they found that
folded. (See Gobet, 1998a, for a further dis- players memorized pieces better when these
cussion of these results.) were functionally relevant. This difference
Saariluoma (1991) studied memory for disappeared in a task where players had to
move sequences using blindfold chess. He count the number of pieces. They also found
dictated one move every two seconds from that tactical combinations embedded in a
three types of sequences: moves actually game position were easier to solve than those
played in a game, random but legal moves, contained in a random position. As with nor-
and random and possibly illegal moves. He mal chess, visuo-spatial interfering tasks neg-
found that masters could recall almost per- atively affect problem-solving performance.
fectly the position for the moves taken from Finally, Chabris and Hearst (2003 ) found
actual games and legal random moves after that the number of blunders did not increase
15 moves, but performed poorly with ille- much when grandmasters played blindfold
gal random moves. With additional moves, games as compared to games with the view
the recall of legal random moves decreased of the pieces.
much faster than that of game moves. Saar- Campitelli and Gobet (2005 ) argue that
iluoma proposed that legal random moves most of the results found on blindfold chess
initially allow for a relatively good recall can be explained by the template theory.
532 the cambridge handbook of expertise and expert performance
of robust scientific research, but two well- higher than that in the general population
controlled studies (Frank & dHondt, 1979; (11%).
Christiaen & Verhofstadt-Deneve, 1981) Chabris and Hamilton (1992) carried out
found that a chess-playing group outper- a divided visual-field experiment with male
formed a control group in verbal ability and chess players and found that the right hemi-
school results, respectively. A limit of these sphere was better than the left at parsing pat-
two studies is that a large number of tests terns according to the default rules of chess
was used, which raises the possibility of type chunking, but that the left hemisphere was
I errors. more efficient at grouping pieces together
when these rules did not apply.
individual differences Several brain-imaging techniques have
been employed to study chess skill (Ather-
Data about individual differences do not
ton et al., 2003 ; Campitelli, 2003 ; Nichelli
offer a clear pattern. There is evidence that
et al.; 1994; Onofrj et al., 1995 ; Amidzic et
chess skill correlates with measures of intel-
al., 2001). Overall, these studies suggest that
ligence, both in children (Frank & dHondt,
frontal and posterior parietal areas, among
1979; Frydman & Lynn, 1992; Horgan &
other areas, are engaged in chess playing.
Morgan, 1990) and adults (Doll & Mayr,
These areas are known to be engaged in
1987). However, whereas Frank and dHondt
tasks requiring working-memory processes.
(1979) and Schneider, Gruber, Gold, and
There is also some evidence that chunks are
Opwis (1993 ) found that chess experts per-
encoded in temporal lobe areas, including
form better than control in non-chess visuo-
the fusiform gyrus, parahippocampal gyrus,
spatial tasks with children and teenagers,
and inferior temporal gyrus. In a different
Djakow, Petrowski, and Rudik (1927), Doll
line of research, Campitelli (2003 ) found
and Mayr (1987), and Waters, Gobet, and
that the left supramarginal gyrus and left
Leyden (2002) failed to find such differ-
frontal areas were involved in autobiograph-
ences with adults. Note that all the above
ical memory in two chess masters.
studies, with the exception of Frank and
dHondt (1979), who had an experimental
design, used quasi-experimental designs; The Role of Deliberate Practice
therefore, the results are based on correla- and Tournament Experience
tions, which are equivocal about the direc-
As appears to be true in other domains
tion of causality (is intelligence a prereq-
(see Ericsson, Chapter 3 8, in this volume),
uisite to chess skill, or does chess playing
skill acquisition in chess requires a consid-
improves ones intelligence?). The differing
erable investment. Few players reach Mas-
patterns between children and adults are
ter level performance with less than 1000
consistent with developmental theories that
hours of serious study (Charness, Krampe,
propose differentiation of abilities across
& Mayr, 1996). Relying on responses to ret-
time. Early in development all forms of prob-
rospective questionnaires, these investiga-
lem solving are dependent on fluid intel-
tors probed a large sample of tournament
ligence (search), but later in development
players from different countries focusing on
crystallized intelligence (knowledge: tem-
how much time they spent in serious study
plates and chunks) changes the way that
alone (deliberate practice) versus that spent
problem solving is carried out.
in tournament play and analysis of games
with others. Other predictors for current
neuroscience skill level included variables such as current
Based on the responses to a questionnaire age, starting age, age when serious about
sent to players rated in the US Chess Federa- chess, age when joining a chess club, pres-
tion ranking list, Cranberg and Albert (1988) ence of coaching, and size of chess library.
found that 18% of male chess players were The variables making independent contribu-
not right-handers. This percentage is reliably tions to explaining current chess rating were
534 the cambridge handbook of expertise and expert performance
serious study alone, size of chess library, and practice in maintaining performance across
current age. Tournament play was not sta- the life span (Krampe & Charness, Chapter
tistically significant after taking deliberate 40). Simulation work has proven useful in
practice time into account. describing how aging processes interact with
Not surprisingly, age was a negative pre- knowledge processes to predict memory per-
dictor (older players tended to have lower formance (Mireles & Charness, 2002).
ratings, averaging a loss of about five to six Nonetheless, many issues remain unre-
rating points per year), whereas deliberate solved. It is not yet clear how deliberate prac-
practice and size of chess library were strong tice and cognitive abilities jointly determine
positive predictors, accounting in combina- performance across the life span given the
tion for nearly 70% of explained variance in differing patterns seen in children, young
current rating. adults, and older adults. Tighter links still
In an enlarged version of the first sam- need to be drawn between perceptual pro-
ple and in a new sample, Charness, Tuffiash, cesses and search processes, particularly as
Krampe, Reingold, and Vasyukova (2005 ) a function of skill level. With the ready
showed a somewhat-different pattern of availability of modern tools (neuro-imaging,
relationships, with both coaching and tour- eye tracking, simulation) and in conjunction
nament play in addition to deliberate prac- with reliable older ones (think-aloud proto-
tice making independent predictions to cur- col analysis), the future seems bright indeed
rent chess rating. For predicting a players for expanding our knowledge of expertise in
peak rating, the two practice variables chess.
accounted for most of the variance. Of
course, this correlates approach suffers from
the weakness that causality is not identi- References
fiable. Longitudinal research is needed to
trace out how process variables covary with Amidzic, O., Riehle, H. J., Fehr, T., Wienbruch,
changes in rating. C., & Elbert, T. (2001). Pattern of focal gamma
bursts in chess players. Nature, 412 , 603 .
Atherton, M., Zhuang, J., Bart, W. M., Hu, X., &
Conclusion He, S. (2003 ). A functional MRI study of high-
level cognition. I. The game of chess. Cognitive
The combination of empirical and the- Brain Research, 16, 263 1.
oretical work has identified and success- Bachmann, T., & Oit, M. (1992). Stroop-like
fully characterized a rich range of phe- interference in chess players imagery: An
nomena from cortical activation patterns unexplored possibility to be revealed by
to eye-movement patterns and from mem- the adapted moving-spot task. Psychological
Research, 5 4, 273 1.
ory for static chess positions to memory
for sequences of moves (including blindfold Baylor, G. W., & Simon, H. A. (1966). A chess
mating combinations program. Proceedings of
chess). Many phenomena identified are cen-
the 1966 Spring Joint Computer Conference, 2 8,
tral to the concerns of psychology, particu-
43 1447.
larly to theories about individual differences,
Berliner, H., & Ebeling, C. (1989). Pattern knowl-
memory systems, developmental processes,
edge and search: The SUPREM architecture.
and theories in cognitive science. The dis- Artificial Intelligence, 3 8, 161198.
covery of the strong relation between skilled
Binet, A. (1894). Psychologie des grands calcu-
perception processes and skilled problem lateurs et joueurs dechecs. Paris: Hachette.
solving has influenced theory development [Republished by Slatkine Ressources, Paris,
in many other domains. For instance, chess 1981.]
research has been useful in characterizing Binet, A. (1966). Mnemonic virtuosity: A study of
the trade-offs seen between memory, per- chess players. Genetic Psychology Monographs,
ception, and problem-solving performance, 74, 127162. Translated from Revue des Deux
as well as in assessing the role of deliberate Mondes, (1893 ), 117 , 82685 9.
expertise in chess 535
Calderwood, R., Klein, G. A., & Crandall, B. W. of skilled performance in chess: Evidence from
(1988). Time pressure, skill, and move quality eye movements. Memory and Cognition, 2 9,
in chess. American Journal of Psychology, 100, 1146115 2.
481495 . Charness, N., Tuffiash, M., & Jastrzembski, T.
Campitelli, G. (2003 ). Cognitive and neuronal (2004). Motivation, emotion, and expert skill
bases of expertise. Unpublished doctoral disser- acquisition. In D. Dai & R. J. Sternberg
tation, University of Nottingham, UK. (Eds.), Motivation, emotion, and cognition: Inte-
Campitelli, G., & Gobet, F. (2005 ). The minds grative perspectives (pp. 2993 19). Mahwah, NJ:
eye in blindfold chess. European Journal of Cog- Erlbaum.
nitive Psychology, 17, 23 45 . Charness, N., Tuffiash, M., Krampe, R., Reingold,
Chabris, C. F., & Hamilton, S. E. (1992). Hemi- E. M., & Vasyukova, E. (2005 ). The role of
spheric specialization for skilled perceptual deliberate practice in chess expertise. Applied
organization by chessmasters. Neuropsycholo- Cognitive Psychology, 19, 15 1165 .
gia, 3 0, 475 7. Chase, W. G., & Simon, H. A. (1973 a). Perception
in chess. Cognitive Psychology, 4, 5 5 81.
Chabris, C. F., & Hearst, E. S. (2003 ). Visualiza-
tion, pattern recognition, and forward search: Chase, W. G., & Simon, H. A. (1973 b). The
Effects of playing speed and sight of the posi- minds eye in chess. In W. G. Chase (Ed.),
tion on grandmaster chess errors. Cognitive Sci- Visual information processing (pp. 215 281).
ence, 2 7, 63 7648. New York: Academic Press.
Charness, N. (1976). Memory for chess positions: Chi, M. T. H. (1978). Knowledge structures and
Resistance to interference. Journal of Experi- memory development. In R. S. Siegler (Ed.),
mental Psychology: Human Learning and Mem- Childrens thinking: What develops? (pp. 73
ory, 2 , 64165 3 . 96). Hillsdale, NJ: Erlbaum.
Charness, N. (1981a). Aging and skilled prob- Christiaen, J., & Verhofstadt-Deneve, L. (1981).
lem solving. Journal of Experimental Psychology: Schaken en cognitieve ontwikkeling. Neder-
General, 110, 213 8. lands Tijdschrift voor de Psychologie, 3 6, 5 61
5 82.
Charness, N. (1981b). Search in chess: Age
and skill differences. Journal of Experimental Cleveland, A. A. (1907). The psychology of chess
Psychology: Human Perception and Performance, and of learning to play it. American Journal of
7 , 467476. Psychology, 18, 2693 08.
Charness, N. (1989). Expertise in chess and Cooke, N. J., Atlas, R. S., Lane, D. M., & Berger,
bridge. In D. Klahr & K. Kotovsky (Eds.), R. C. (1993 ). Role of high-level knowledge in
Complex information processing: The impact of memory for chess positions. American Journal
Herbert A. Simon (pp. 183 208). Hillsdale, NJ: of Psychology, 106, 3 213 5 1.
Erlbaum. Cranberg, L., & Albert, M. L. (1988). The
Charness, N. (1992). The impact of chess research chess mind. In L. K. Obler & D. Fein (Eds.),
on cognitive science. Psychological Research, 5 4, The exceptional brain. Neuropsychology of talent
49. and special abilities (pp. 15 6190). New York:
Guilford Press.
Charness, N., & Gerchak, Y. (1996). Participation
rates and maximal performance: A log-linear de Groot, A. D. (1946). Het denken van den
explanation for group differences, such as schaker. Amsterdam: Noord Hollandsche.
Russian and male dominance in chess. Psycho- de Groot, A. D. (1965 ). Thought and choice in
logical Science, 7, 465 1. chess. The Hague: Mouton.
Charness, N., Krampe, R., & Mayr, U. (1996). The de Groot, A. D., & Gobet, F. (1996). Perception
role of practice and coaching in entrepreneurial and memory in chess. Assen (The Netherlands):
skill domains: An international comparison Van Gorcum.
of life-span chess skill acquisition. In K. A. Didierjean, A., Cauzinille-Marmeche, E., & Sav-
Ericsson (Ed.), The road to excellence: The acqui- ina, Y. (1999). Learning from examples: Case-
sition of expert performance in the arts and sci- based reasoning in chess for novices. Current
ences, sports, and games (pp. 5 180). Mahwah, Psychology of Cognition, 18, 3 3 73 61.
NJ: Erlbaum. Djakow, I. N., Petrowski, N. W., & Rudik, P. A.
Charness, N., Reingold, E. M., Pomplun, M., & (1927). Psychologie des Schachspiels. Berlin: de
Stampe, D. M. (2001). The perceptual aspect Gruyter.
536 the cambridge handbook of expertise and expert performance
Doll, J., & Mayr, U. (1987). Intelligenz und Gobet, F. (1998b). Chess players thinking revis-
Schachleistung eine Untersuchung an ited. Swiss Journal of Psychology, 5 7 , 183 2.
Schachexperten. Psychologische Beitrage, 2 9, Gobet, F., & Campitelli, G. (in press). Educa-
270289. tion and chess: A critical review. In T. Redman
Elo, A. E. (1965 ). Age changes in master chess (Ed.), Education and chess.
performances. Journal of Gerontology, 2 0, 289 Gobet, F., Campitelli, G., & Waters, A. J. (2002).
299. Rise of human intelligence: Comments on
Elo, A. E. (1986). The rating of chessplayers, past Howard (1999). Intelligence, 3 0, 3 03 3 11.
and present, (2nd ed.). New York: Arco chess. Gobet, F., & Clarkson, G. (2004). Chunks in
Ericsson, K. A., & Harris M. S. (1990, Novem- expert memory: Evidence for the magical
ber). Expert chess memory without chess number four . . . or is it two? Memory, 12 ,
knowledge: A training study. Paper presented 73 2747.
at the 3 1st Annual Meeting of the Psycho- Gobet, F., de Voogt, A., & Retschitzki, J. (2004).
nomics Society, New Orleans. Moves in mind: The psychology of board games.
Ericsson, K. A., & Kintsch, W. (1995 ). Long- New York: Psychology Press.
term working memory. Psychological Review, Gobet, F., & Jackson, S. (2002). In search of tem-
102 , 211245 . plates. Cognitive Systems Research, 3 , 3 5 44.
Ericsson, K. A., & Oliver, W. (1984, November). Gobet, F., & Jansen, P. J. (1994). Towards a chess
Skilled memory in blindfolded chess. Paper program based on a model of human memory.
presented at the Annual Meeting of the Psy- In H. J. van den Herik, I. S. Herschberg, & J. W.
chonomic Society, San Antonio, TX. Uiterwijk (Eds.), Advances in computer chess 7
Ericsson, K. A., & Staszewski, J. J. (1989). Skilled (pp. 3 5 60). Maastricht: University of Limburg
memory and expertise: Mechanisms of excep- Press.
tional performance. In D. Klahr & K. Kotovsky Gobet, F., & Jansen, P. J. (in press). Training
(Eds.), Complex information processing: The in chess: A scientific approach. In T. Redman
impact of Herbert A. Simon (pp. 23 5 267). (Ed.), Education and chess.
Hillsdale, NJ: Erlbaum A. Gobet, F., Lane, P. C. R., Croker, S., Cheng,
Fisk, A. D., & Lloyd, S. J. (1988). The role of P. C. H., Jones, G., Oliver, I., & Pine, J. M.
stimulus to role consistency in learning rapid (2001). Chunking mechanisms in human learn-
application of spatial rules. Human Factors, 3 0, ing. Trends in Cognitive Sciences, 5 , 23 6243 .
3 5 49. Gobet, F., & Simon, H. A. (1996a). Templates
Frank, A., & dHondt, W. (1979). Aptitudes et in chess memory: A mechanism for recalling
apprentissage du jeu dechecs au Zaire. Psy- several boards. Cognitive Psychology, 3 1, 140.
chopathologie Africaine, 15 , 8198. Gobet, F., & Simon, H. A. (1996b). Recall of
Frey, P. W., & Adesman, P. (1976). Recall memory rapidly presented random chess positions is a
for visually presented chess positions. Memory function of skill. Psychonomic Bulletin & Review,
& Cognition, 4, 5 415 47. 3 , 15 9163 .
Freyhoff, H., Gruber, H., & Ziegler, A. (1992). Gobet, F., & Simon, H. A. (1996c). Recall of ran-
Expertise and hierarchical knowledge repre- dom and distorted positions: Implications for
sentation in chess. Psychological Research, 5 4, the theory of expertise. Memory & Cognition,
3 23 7. 2 4, 493 5 03 .
Frydman, M., & Lynn, R. (1992). The general Gobet, F., & Simon, H. A. (1996d). The roles of
intelligence and spatial abilities of gifted young recognition processes and look-ahead search in
Belgian chess players. British Journal of Psychol- time-constrained expert problem solving: Evi-
ogy, 83 , 23 3 23 5 . dence from grandmaster level chess. Psycholog-
Gobet, F. (1993 ). Les memoires dun joueur ical Science, 7, 5 25 5 .
dechecs. Fribourg: Editions Universitaires. Gobet, F., & Simon, H. A. (1998). Expert chess
Gobet, F. (1997). A pattern-recognition theory of memory: Revisiting the chunking hypothesis.
search in expert problem solving. Thinking & Memory, 6, 225 25 5 .
Reasoning, 3 , 2913 13 . Gobet, F., & Simon, H. A. (2000). Five seconds
Gobet, F. (1998a). Expert memory: A comparison or sixty? Presentation time in expert memory.
of four theories. Cognition, 66, 115 15 2. Cognitive Science, 2 4, 65 1682.
expertise in chess 537
Gobet, F., & Waters, A. J. (2003 ). The role of ized tomography (SPECT). Neuroscience Let-
constraints in expert memory. Journal of Exper- ters, 198, 169172.
imental Psychology: Learning, Memory & Cogni- Pitrat, J. (1977). A chess combinations program
tion, 2 9, 10821094. which uses plans. Artificial Intelligence, 8, 275
Goldin, S. E. (1978). Memory for the ordinary: 3 21.
Typicality effects in chess memory. Journal of Reingold, E. M., Charness, N., Pomplun, M., &
Experimental Psychology: Human Learning and Stampe, D. M. (2001). Visual span in expert
Memory, 4, 605 616. chess players: Evidence from eye movements.
Goldin, S. E. (1979). Recognition memory for Psychological Science, 12 , 485 5 .
chess positions: Some preliminary research. Reingold, E. M., Charness, N., Schultetus, R. S.,
American Journal of Psychology, 92 , 193 1. & Stampe, D. M. (2001). Perceptual automatic-
Holding, D. H. (1985 ). The psychology of chess ity in expert chess players: Parallel encoding
skill. Hillsdale, NJ: Erlbaum. of chess relations. Psychonomic Bulletin and
Holding, D. H., & Reynolds, R. (1982). Recall or Review, 8, 5 045 10.
evaluation of chess positions as determinants of Saariluoma, P. (1984). Coding problem spaces in
chess skill. Memory and Cognition, 10, 23 7242. chess: A psychological study. Commentationes
Horgan, D. D., & Morgan, D. (1990). Chess exper- scientiarum socialium 23 . Turku: Societas Sci-
tise in children. Applied Cognitive Psychology, 4, entiarum Fennica.
109128. Saariluoma, P. (1985 ). Chess players intake of
Howard, R. W. (1999). Preliminary real-world task-relevant cues. Memory and Cognition, 13 ,
evidence that average human intelligence really 3 85 3 91.
is rising. Intelligence, 27, 23 5 25 0. Saariluoma, P. (1990). Apperception and restruc-
Jongman, R. W. (1968). Het oog van de turing in chess players problem solving. In K. J.
meester (The eye of the master). Assen (The Gilhooly, M. T. G. Keane, R. H. Logie, & G.
Netherlands): Van Gorcum. Erdos (Eds.), Lines of thought: Reflections on the
Lassiter, G. D. (2000). The relative contributions psychology of thinking (Vol. II, pp. 415 7). New
of recognition and search-evaluation processes York: Wiley.
to high-level chess performance: Comment on Saariluoma, P. (1991). Aspects of skilled imagery
Gobet and Simon. Psychological Science, 11, 172 in blindfold chess. Acta Psychologica, 77 , 65
173 . 89.
McGregor, S. J., & Howes, A. (2002). The role of Saariluoma, P. (1992). Error in chess: The
attack and defense semantics in skilled players apperception-restructuring view. Psychological
memory for chess positions. Memory & Cogni- Research, 5 4, 1726.
tion, 3 0, 707717. Saariluoma, P. (1994). Location coding in chess.
Milojkovic, J. D. (1982). Chess imagery in novice Quarterly Journal of Experimental Psychology,
and master. Journal of Mental Imagery, 6, 125 47A, 60763 0.
144. Saariluoma, P. (1995 ). Chess players thinking:
Mireles, D. E., & Charness, N. (2002). Com- A cognitive psychological approach. London:
putational explorations of the influence of Routlege.
structured knowledge on age-related cognitive Saariluoma, P., & Hohlfeld, M. (1994). Apper-
decline. Psychology and Aging, 17, 245 25 9. ception in chess players long-range planning.
Newell, A., & Simon, H. A. (1972). Human prob- European Journal of Cognitive Psychology, 6, 1
lem solving. Englewood Cliffs, NJ: Prentice- 22.
Hall. Saariluoma, P., & Kalakoski, V. (1997). Skilled
Nichelli, P., Grafman, J., Pietrini, P., Alway, D., imagery and long-term working memory.
et al. (1994). Brain activity in chess playing. American Journal of Psychology, 110, 177201.
Nature, 3 69, 191. Saariluoma, P., & Kalakoski, V. (1998). Appercep-
Onofrj, M., Curatola, L., Valentini, G. L., tion and imagery in blindfold chess. Memory, 6,
Antonelli, M., Thomas, A., & Fulgente, T. 6790.
(1995 ). Non-dominant dorsal-prefrontal acti- Saariluoma, P., & Laine, T. (2001). Novice con-
vation during chess problem solution evi- struction of chess memory. Scandinavian Jour-
denced by single photon emission computer- nal of Psychology, 42 , 13 7146.
538 the cambridge handbook of expertise and expert performance
Schultetus, R. S., & Charness, N. (1999). Recall of analyzing heuristics. Soviet Psychology, 5 ,
or evaluation of chess positions revisited: The 215 .
relationship between memory and evaluation Wagner, D. A., & Scurrah, M. J. (1971). Some
in chess skill. American Journal of Psychology, characteristics of human problem-solving in
112 , 5 5 5 5 69. chess. Cognitive Psychology, 2 , 45 4478.
Schneider, W., Gruber, H., Gold, A., & Opwis, Van der Maas, H. L. J., & Wagenmakers,
K. (1993 ). Chess expertise and memory for E.-J. (2005 ). A psychometric analysis of chess
chess positions in children and adults. Journal expertise. American Journal of Psychology, 118,
of Experimental Child Psychology, 5 6, 3 283 49. 2960.
Simon, H. A., & Gilmartin, K. J. (1973 ). A simu- Waters, A. J., Gobet, F., & Leyden, G. (2002).
lation of memory for chess positions. Cognitive Visuo-spatial abilities in chess players. British
Psychology, 5 , 2946. Journal of Psychology, 93 , 5 5 75 65 .
Tikhomirov, O. K., & Poznyanskaya, E. D. (1966). Wilkins, D. (1980). Using patterns and plans in
An investigation of visual search as a means chess. Artificial Intelligence, 14, 165 203 .
C H A P T E R 31
Exceptional Memory
roof of the building collapsed, killing every- memory championships has enabled the
one present. The bodies were unrecogniz- gathering of group data, but the contes-
able but Simonides was able to recall where tants at such competitions tend to rely heav-
each guest had been sitting and was thereby ily on strategies, hence these samples may
inspired to invent the mnemonic method be somewhat biased to a particular type of
of loci, which is still widely used. In this memorizer.
method each item to be retained is associ- Four main methods have been used in
ated with a location on a well-known route. the study of superior memory: performance
When recall of the items is required, a men- on experimental tasks, self-report on the
tal walk is taken along the route, the loca- methods used, psychometric analyses, and
tions serving as cues to recall the items brain-imaging. Most studies have combined
in order. the first two of these. Self-reports provide
The scientific study of superior memory unique insights into the methods employed
ability begins with Binets (1894) study of by expert memorizers, but their reliability
three experts in number recall. It was a fur- needs to be confirmed by objective perfor-
ther twenty years before a more detailed mance data. Psychometric data have been
study appeared, of Ruckle in Germany used to explore relationships between mem-
(Muller, 1911, 1913 ), and another twenty ory and other individual difference vari-
years before Susukita (193 3 , 193 4) under- ables such as intelligence, visual and ver-
took a lengthy study of Ishihara in Japan. bal imagery, and personality factors. Lastly,
From the 1960s onward many more reports the development of structural and functional
have appeared, some of them very detailed brain-imaging opens up the possibility of
and much more precise in their testing of the examining both possible differences in brain
processes involved. These include Lurias structure in memory experts and brain activ-
(1975 ) study of Shereshevskii (carried out ity during the process of memorizing, and
in the 193 0s but not available in English hence of investigating the neurological basis
until 1975 ), Hunters (1962, 1977) reports of memory expertise.
on the memory of Alexander Aitken, Hunt
and Loves (1972) account of VPs memory,
studies by Ericsson and his colleagues (e.g. Some Key Examples of Memory
Ericsson, 1985 ; Ericsson & Polson, 1988), Expertise
Thompson et al.s (1993 ) report on Rajan
Mahadevan, and our own studies of TE, Superiority in memory can be demonstrated
TM, and contestants at the World Memory in two main ways, firstly, by retrieval with-
Championships (Gordon et al., 1984; Wild- out error of a substantial amount of material,
ing & Valentine, 1985 , 1994a, 1994b). There compared with frequent errors in the case
is now therefore a substantial body of of normal memory ability, and secondly, by
data available collected under controlled retention of an unusually large proportion
conditions, together with self-reports from of the original material at some fixed inter-
experts on their methods. Theories have val afterwards, again in comparison with a
been developed and tested and a consensus control group. As pointed out earlier, this is
is emerging about underlying processes. known as episodic memory, in contrast to
One problem facing the study of excep- organized information in semantic memory.
tional memory performance is the loca- One example of memory expertise where
tion and identification of suitable individu- this distinction becomes blurred is the feat,
als. Expert performers are by definition rare. popular among memory experts, of learn-
Thus many of the studies in the literature ing the expansion of the number pi to thou-
(as seen above) are individual case studies. sands of places. This requires long, dedicated
These have the disadvantage that one can- exposure and the development of an organi-
not be sure how far the results are generaliz- zational framework, as shown by the ability
able to other cases. The rise of international of many memory experts to locate a specific
exceptional memory 541
part of the sequence or expand a fragment whereas reversing the order for backwards
of the original in response to a cue. recall necessitates additional operations and
We offer now some of the most strik- hence more recall time, exactly as observed.
ing examples of memory expertise, which Hence it seems likely that S was using some
serve to raise the most insistent ques- such method. Certainly he used the method
tions under investigation. Susukita pre- of loci described above with other material,
sented the Japanese expert Ishihara with basing it on Gorky Street in Moscow. Luria
random sequences of numbers in rows and argues that he adopted this method later
measured the time until he signaled he was only in order to organize the unusual spon-
ready to recall the list. For lists of 200 items taneous imagery that Luria assumed was the
Ishihara took an average of 1.7 seconds per ultimate basis of his superiority, but there
item and made about 1% errors. For the are indications that he may already have
longest list of 2400 items the time per item been using it when he first met Luria (see
was 8.2 seconds and the error rate was 2%, so Wilding & Valentine, 1997, p. 23 ). Ericsson
the list took some four hours to memorize. and Chase conclude that Ss superior mem-
Ishihara described his method in detail. First, ory depended entirely on such methods,
each numeral could be transformed into one but this may be an oversimplification. Sev-
of several syllables in Japanese and these eral observations made by Luria imply that
could be combined into words. Second, he there were a number of peculiarities about
had a database of 400 images of places he his memory, particularly the vivid imagery
knew. He associated the words with a suc- and synaesthesia that he experienced from a
cession of locations in one of these founda- very early age. However, synaesthesia is not
tions as he called them. We will be return- invariably associated with superior memory
ing frequently to these two basic processes (Wilding & Valentine, 1997, pp. 25 26). In
of recoding items and embedding them in an the case of S imagery was invariably evoked
already established structure. by surface characteristics of the input (par-
The Russian neuropsychologist Alexan- ticularly sounds) and impaired his ability
der Luria first met the memory expert Shere- to understand meaning or reference to a
shevskii (S) in the late 1920s and studied single object by different words (because
his memory intensively during the 193 0s. He the words evoked different images). Wild-
found that Ss ability to retain sequences ing and Valentine suggest that the domi-
of spoken items was virtually unlimited nance of surface characteristics of the input
and believed that this unusual memory was over meaning is reminiscent of some aspects
innate, not dependent on any technique. of Aspergers syndrome, a milder form of
One of Lurias main arguments was derived autism, which is characterised by difficulty
from Ss ability to memorize a number in interpreting meaning, social conventions,
matrix. Luria believed that S retained a and other minds.
photo-like image of the matrix. However, In 1972 Hunt and Love published a
Ericsson and Chase (1982) have noted that study of VP, who demonstrated superior
S took longer to recall columns than rows memory ability on a wide range of tasks.
and longer to recall the matrix backwards The authors obtained only general infor-
than forwards. If he had been reading an mation on his methods. With numbers he
image, there should have been little differ- searched for a meaningful association such
ence between these conditions. Many other as a date, then added further associations
experts tested on this task have shown a sim- about events occurring on that date. With
ilar pattern and their descriptions of their nonsense syllables he related them to one of
method make no reference to an image. many European languages he knew. How-
Instead they claim to code the matrix row ever, other evidence suggested that he pos-
by row, often recoding numbers into words. sessed some naturally superior ability. On a
Recalling by columns involves retrieving short-term memory-scanning task, he could
each row, then one item from each row, decide equally quickly, regardless of list
542 the cambridge handbook of expertise and expert performance
length, that a number had been present list. SF attained a span of 82 digits. Most of
in a recently presented list. Normally deci- these were held in a more permanent form
sion time increases with list length, imply- than fragile short-term memory, since sup-
ing serial search through the list, but VP pressing auditory rehearsal affected only the
appeared to be able to scan all items simulta- final three or four digits. Another subject,
neously. Anecdotal evidence also suggested DD, using dates or other meaningful recod-
early superior memory for rail and bus ings, attained a span of 101 digits. Wilding and
timetables and the city map. VP, however, Valentine (1997) have expressed some scep-
thought that his memory was the result of ticism that this method alone can account
intensive training in rote memorization at completely for SFs achievement. Running
school, so training and methods were prob- times would not provide unambiguous cues
ably also important. for recall, even if he had many variations (a
One of the most interesting examples of very poor mile time, an average mile time
superior memory is Alexander Aitken, Pro- for the marathon, etc.). A tape of SF recall-
fessor of Mathematics at Edinburgh Univer- ing at an early stage in his career refers to
sity, described by Hunter (1962, 1977) in 12 mile times and six two-mile times in a
a largely anecdotal account. Aitken had a single sequence. Some sequences could not
superior digit span (13 auditorily presented be coded as times (89.6 seconds would be
items and 15 visually presented items). He reported as 1 minute 29.6 seconds in ath-
stated that he allowed associations to groups letics), so SF used ages (a very old man)
of items to reveal themselves, based on his instead, but these codes are equally ambigu-
mathematical knowledge, and further struc- ous. Both the tape and a protocol provided
tured these groups by rhythm, but he used by Chase and Ericsson (1981) show many
no deliberate mnemonic methods. However, cases where no recoding occurred, imply-
his abilities were not solely dependent on ing that SF often fell back on rote recall.
his mathematical expertise. Hunter (1996) Finally, there is no indication of how SF
reports that Aitken read Bartletts (193 2) could avoid confusion between successive
War of the Ghosts story twice in 193 4 sequences (nor any data on the frequency
and recalled it in 1960. Hunt and Love gave of such confusions). SF may not, therefore,
this story to VP and counted the number of have been quite so average as the experi-
nouns and verbs recalled (62% at immediate menters believed. If nothing else, his drive
recall and 5 9% six weeks later). A control to succeed was exceptional, as is revealed
group with average memory ability recalled by his comments in the tape and protocol
45 % initially and 3 2% after six weeks. Aitken records, and might account for differences
recalled 5 8% after 26 years! He is thus the in performance across time for SF, DD, and
strongest candidate for a very superior, gen- for a woman who dropped out of digit-
eral, natural memory. span training without achieving notable
In contrast SF, an average undergraduate, improvement.
was selected by Ericsson and Chase (Erics- In 1981 Rajan Mahadevan recalled 3 1,811
son, Chase, & Faloon, 1980) to test whether digits of pi without errors. (But Hideaki
digit span, normally around seven items, Tomoyori subsequently outdid this record
could be improved by practice. SF was a by reciting 40,000 digits, using a story
long-distance runner and devised the idea mnemonic; the names of digits in Japanese
of encoding digit strings as running times have a second meaning, which can be used
such as a good time for the mile, thereby readily to build such a story.) As pointed
constructing units composed of three or out earlier, such performances could be the
four digits. In later developments these units result simply of sweat and tears, rather than
were combined into groups of three units, any superior memory ability or methods.
then pairs of these groups were combined However, Thompson et al. (1993 ) demon-
into super-groups and so forth, thereby strated that Rajan could recall 43 digits pre-
imposing a hierarchal structure on the whole sented auditorily and 28 (later 60) presented
exceptional memory 543
visually. His ability with letters was more during his learning of pi. When given other
modest (a span of 13 items) but still well symbols without this advantage, Rajans per-
above the norm. They argued that he had formance was much worse than with dig-
a natural ability to retain about 13 items, its, until he learned to convert these sym-
rather than the seven items of the normal bols into numbers. This conclusion may well
span. They supported this by measuring how be correct, but it leaves open the ques-
long Rajan took to begin recall after the tion of how Rajan had mastered pi without
end of list presentation, arguing that such any apparent recoding strategy such as those
a pause would remain constant as long as used by Ishihara and Tomoyori. At the very
the list was within the natural memory span, least, highly unusual persistence and motiva-
but would rise once some recoding of the tion were needed, and one has the suspicion
list became necessary because of the need that some superiority in normal memory is
to retrieve the additional codes. Normal necessary in any individual who can persist
subjects showed an increase in this mea- with a task that is so unrewarding for the
sure once list length exceeded seven items, normal memorizer.
the average digit span, but Rajan showed no Wilding and Valentine (1997) describe
increase until list length exceeded 13 items. a young female subject, JR, who showed
Rajan cannot have been maintaining the remarkable memory ability without employ-
items in a temporary short-term store, since ing any well-practiced methods. She per-
he could remember them after being dis- formed particularly well on immediate
tracted, or recite them backwards; some- memory for faces and for names to faces,
times he retained them for 24 hours or more. but her most impressive performance was
However, he did not describe any form of in delayed recall without warning a week
recoding strategy analogous to those we have later. On the four tasks tested (memory for a
encountered above. Thompson et al. claim story, faces, names, and words) she averaged
that he labelled each digit by its list position. 95 % retention of her already high scores at
When required to learn a 20 20 number immediate testing. A control group averaged
matrix he said that he encoded row by row, 5 9% and the group of experts tested by these
together with the first column in order to authors (Wilding & Valentine, 1994a) did
retain the order of the rows. This was con- only slightly better with 62% retention.
firmed by the time pattern of his recall; the We have now described the most intrigu-
first column took much less time than other ing cases of unusual memory that raise
columns. However, there was no evidence in some of the main questions. Other cases are
his time pattern when recalling the rows that described by Wilding and Valentine (1997).
he divided them up into smaller units. We will not discuss here the small number
Thompson et al. discuss in detail his of cases of superior memory of a very spe-
memory for pi. In the display from which he cific kind, which appear to be independent
learned, the numbers were laid out in rows of any strategy, such as for languages, music,
of 100, separated into blocks of ten, every ten visual imagery, and cases of autistic memory
rows being separated from the next set of ten (see Wilding & Valentine, 1997, pp 475 1 for
rows by a space; so it was possible to ask for details).
the 13 th item in the 5 th row, for example. He
was faster at locating the first five digits in a
block of ten than the second five, suggesting Theoretical Issues
that he had retained the display structure.
Thompson et al. suggest that the blocks of
We now specify more precisely the main
ten digits formed units like words, created
issues that have emerged in the course of
by sheer rote learning.
the above survey.
Recently Ericsson et al. (2004) have
argued that Rajans impressive digit span was r Is exceptional memory typically specific
primarily dependent on these units formed to a particular sub-process, type of task
544 the cambridge handbook of expertise and expert performance
or material, or does it generalize across types of input, there must be some cen-
these? tral process that integrates information from
r To what extent is exceptional mem- separate systems. The amnesic syndrome, for
ory superiority natural, or acquired by example, typically affects memory over a
training in mnemonic methods, or both? wide range of material. Additional evidence
If natural superiority occurs, are such in support of some overall memory abil-
cases interpretable as the tail of a nor- ity can be drawn from correlational studies.
mal distribution in memory ability or as Carroll (1993 ) has reviewed factor-analytic
qualitatively different from the normal studies of cognitive abilities and concludes
population? that there is a general memory ability that
r If superiority can be acquired by training, affects . . . performances in a wide variety
what is the nature of successful training? of tasks and behaviours involving memory
r (p. 3 02). He suggests that there may also
How does such superiority relate to
be separate abilities to learn over repeated
other individual differences, such as
exposure (to lay down long-term memories)
intelligence?
r and to retain information (Ingham, 195 2).
What is the neurological basis of excep- Wilding and Valentine (1997) also showed
tional memory? that memory for a wide range of material
r Does training in mnemonic methods have loaded on a single factor, whereas retention
practical benefits for everyday remember- showed some tendency to load on a sepa-
ing, development of expertise in other rate factor or factors. Wilding et al. (1999)
areas, remediation of memory impair- showed that self-rating of memory ability
ments, and so forth? correlated significantly, though not dramat-
ically, with performance on a wide range of
tasks and was also related to examination
General or Specific?
performance.
All memory requires material to be encoded, Thus it is possible, indeed likely, that
stored, and retrieved, so a full understanding memory efficiency may vary between indi-
of exceptional memory requires the contri- viduals either within specific systems or
bution of these different processes to be dis- overall, and hence exceptional ability may
tinguished. Furthermore, memory is widely be either specific or general. Examples of
regarded as comprising a number of sepa- unusual superiority in specific aspects of
rate systems. We have already referred to the memory are rare, as indicated above. In the
distinction between episodic memory and light of the preceding discussion individu-
semantic memory, pointing out that when als lying at the extreme end of the distribu-
considering memory expertise we are pri- tion of overall memory ability should show
marily concerned with the former, the ability wide-ranging superiority. Alternatively, neu-
to reproduce in some way (by speech, writ- ral or genetic abnormality might produce
ing, acting out, etc.) an earlier experience, performance that is so unusual as to appear
whereas semantic memory is more properly qualitatively different from the norm. Also
considered as a component of other forms in some cases superiority with several types
of expertise, such as knowledge in medicine, of material is clearly due to application of
the arts, sport, and so forth. the same technique. The limited data avail-
However, many other distinctions have able do not in general provide unequivocal
been made amongst subsystems in the brain evidence about the generality of the supe-
that serve memory, particularly those ded- riority exhibited since most of the studies
icated to processing specific inputs (e.g., tested mainly memory for numbers. Tests
visual, auditory) or materials (e.g., faces, using numbers are popular because num-
music). Wilding and Valentine (1997) have bers are universal, have the same significance
argued that, despite the evidence for sep- for all cultures, normally have no intrin-
arate memory systems that handle specific sic meaning when combined into sequences,
exceptional memory 545
and permit load to be varied to any required by already acquired knowledge and prac-
degree. tice, and Ishihara and SF depended primarily
In some of the cases we have discussed, on techniques, with the possibility of some
however, memory ability was demonstrated preexisting natural facility on which to base
over a wider range of material than numbers them.
alone, notably by VP and Aitken, and to a Wilding and Valentine (1994a), in their
lesser extent JR and S. Ishihara was profi- study of a group of memory experts, devel-
cient with both numbers and words (using oped some criteria for distinguishing those
a common strategy). On the other hand relying on techniques from those who
SFs superiority was restricted to numbers appeared to have a natural superiority.
because he had not devised a strategy for use The tasks used were divided into strate-
with letters; Rajans performance was daz- gic, those readily amenable to techniques
zling with numbers, less impressive with let- (on the basis of known methods, includ-
ters and other symbols, and otherwise quite ing those mentioned by the participants),
moderate (Biederman et al., 1992), so his and non-strategic, those where techniques
memory ability appears to have been rela- were inappropriate or not readily available
tively specific. Whether this is, as Ericsson et because of task novelty. Tests for faces,
al. (2004) believe, because it was dependent names for faces, word lists, and telephone
on his prior learning of pi (itself requiring numbers were assigned to the first category
explanation) or because he had some inborn and tests for a story, spatial and temporal
special facility with numbers, remains under positions of pictures, and snowflakes to the
debate. second. The eight participants in the study
Our conclusion to the first question is, who performed best on the immediate me-
therefore, that most often superior mem- mory tasks were divided into a group reliant
ory is quite general (as in the cases of VP on well-practiced mnemonic techniques
and Aitken), but in some cases (described (strategic memorizers; n = 5 ) and those
by Wilding & Valentine, 1997, p. 47 et seq.) denying any systematic use of such meth-
it can be highly specific. ods (non-strategic memorizers; n = 3 ).
Results are shown in Figure 3 1.1. Strategic
memorizers performed outstandingly well
Natural or Acquired?
on strategic tasks but little better than a
Many cases of memory superiority are control group on non-strategic tasks. Non-
clearly dependent on the use of special tech- strategic memorizers performed above aver-
niques, as already indicated. It does seem age and equally well on both types of task,
that highly impressive performance without being worse than strategic memorizers on
the use of techniques is very rare indeed. strategic tasks but better than them on non-
The more difficult question is whether there strategic tasks. However, the overall superi-
are any cases where techniques can be com- ority of non-strategic memorizers compared
pletely ruled out as an explanation. The with an average control group, though sub-
question also arises as to whether, where no stantial, was not outstanding, suggesting that
technique is apparent, the individual rep- very superior performance without using a
resents a case drawn from the extreme tail technique is rare. There were also dramatic
of a normal distribution of ability or has differences between the two groups in reten-
some qualitatively different ability from the tion over a week, as shown in Table 3 1.1.
norm. Of the cases we have discussed, S, like This occurred even though three of the tasks
autistic individuals, displays the most obvi- tested after a delay (faces, names, and words)
ous abnormalities in the quality as well as were strategic tasks (in the sense above)
the capacity of his memory. Aitken and JR and produced superior immediate perfor-
present the most persuasive cases of natu- mance in the strategic group. Thus, despite
ral all-round superiority, VP (and possibly the small number of cases, the differences
Rajan) demonstrated natural ability aided were clear and the most striking evidence of
546 the cambridge handbook of expertise and expert performance
2.5
2.0
MEAN Z SCORE
1.5
1.0
.5
TASK TYPE
STRATEGIC
0.0 NON-STRATEGIC
STRATEGIC MEMORIZERS NATURAL MEMORIZERS
superior natural memory seems to lie in 13 items, then rose. Wilding and Marshall
retention. took the same measure, but also recorded
Wilding and Marshall (2004) investigated times to produce each item in lists of dif-
the possibility of natural superiority in short- ferent length, testing four experts, all of
term memory, following the suggestion of whom had unusually long digit spans (14,
Thompson et al. (1993 ) that Rajan Mahade- 16, 18, and 21 items). Only one of them
van had a natural short-term memory span showed the same pattern as Rajan, with
of 13 items. As described earlier, the main delay before recall constant up to a list length
evidence for this claim was derived from of eight items, then an increase in this delay.
the delay before he started recall, which Hence Thompson et al.s measure was not
remained constant until list length exceeded a useful criterion for natural memory span.
Moreover, all four experts showed group-
ing of items, once list length exceeded about
Table 31.1. Percentage recalled/recognized at seven items, with groups separated by longer
immediate testing on four tasks and percentage pauses at recall. Groups were two, three or
of immediate recall/recognition retained over a four items in length in different cases, and
week for strategic and non-strategic memorizers in each case matched the recoding strategy
and a control group. described afterwards by the expert. Thus, in
Immediate test Delayed test all four cases superior digit span seemed to
depend on a mnemonic strategy; no evidence
Strategic 94 53 for any natural superiority was found.
memorizers We conclude that the most striking exam-
Natural 85 83
ples of superior memory are strategy depen-
memorizers
dent, but there is also evidence for a natu-
Control group 56 59
ral component. In the latter case, the ability
exceptional memory 547
that memory ability is independent of IQ, experience, as has been demonstrated in the
but the underlying reasons for unusual mem- case of musicians (e.g., Schlaug et al., 1995 ),
ory in such cases are as yet poorly understood Braille reading (Hamilton & Pascual-Leone,
and seem to involve narrowly focused obses- 1998), and navigational skills (Maguire
sion and motivation in some cases, rather et al., 2000). (In the case of the musicians,
than pure memory ability. On the other the structural changes were usually asso-
hand, experts in specific fields of attainment, ciated with early commencement of train-
as already pointed out, show impressive ing, e.g., before the age of seven. However,
knowledge of their field, but it is likely that Maguire et al.s (2000) study of taxi drivers
this stems largely from motivation, expo- provided evidence for brain plasticity dur-
sure, and ability to select and organize infor- ing adulthood.) The memory experts were
mation, and such abilities are undoubtedly much better than the controls on digits and
related to general intelligence. somewhat better on faces, however, there
A considerable body of evidence demon- was little difference on snow crystals. There
strates that the speed of processing in the were three main brain areas in which activity
brain is related to IQ. Salthouse (1985 ) sug- during memorizing occurred in the experts
gested that the decline in memory with only: the medial parietal cortex, retrosple-
age could be due to decreasing speed of nial cortex, and right posterior hippocam-
processing. On this view IQ and memory pus, areas known to be implicated in spatial
would in part depend on a single factor. memory and navigation (OKeefe & Nadel,
Rabbitt (1993 ) and his colleagues con- 1978; Maguire et al., 2003 b). Nine out of the
firmed the relation between performance ten memory experts, but none of the con-
on speeded processing and IQ but showed trols, reported that they used the method of
only weak relations between these measures loci. Hence the neurological data concurred
and age. Memory, however, was strongly with the self-reports. These differences were
related to age and only weakly to IQ and not explicable in terms of superior perfor-
processing speed. This is convincing evi- mance by the experts, since the differences
dence that memory and IQ are distinct between groups in brain activity were main-
constructs. See also Horn and Masunaga, tained across tasks, whereas the difference in
Chapter 3 4. performance varied across tasks, being neg-
ligible for snow crystals. Nor were the dif-
ferences explicable in terms of general cog-
Neurological Basis of Superior Memory nitive ability, since there was no significant
Although there is a fairly extensive liter- difference between the groups on tests of
ature on the neurological basis of normal verbal and non-verbal intelligence. The
memory, as yet few studies have investigated experts, however, were superior on tests of
the neurological basis of superior memory. working and long-term (though not visual)
Maguire et al. (2003 a) carried out structural memory.
and functional MRI on the brains of ten A question that remains to be answered is
memory experts and ten controls matched why there were no differences in any aspect
for sex, age, and handedness. Participants of the brain recordings to match the differ-
were scanned while they attempted to mem- ences in superiority of the experts for the
orize six-item sequences of three-digit num- different materials. Either the critical dif-
bers, faces, or snow crystals, the prediction ferences were too subtle to show in the
being that these types of material would measures taken, or some other difference
be differentially susceptible to the use of remains to be uncovered.
mnemonic strategies. Recognition was tested Nyberg et al. (2003 ) used PET scanning to
subsequently outside the scanner. A num- study brain activity in older and younger par-
ber of psychometric tests were also adminis- ticipants during training in acquisition and
tered. There were no structural brain differ- use of the method of loci. Younger partici-
ences between the groups, so there was no pants and older ones who improved showed
evidence for brain plasticity as a function of increased activity in occipito-parietal cortex
exceptional memory 549
and left retrosplenial cortex (see Maguire tors affecting instruction in a mnemonic
et al., 2003 b). Only younger participants to aid the learning of name-to-face asso-
showed increased activity in dorsal frontal ciations (selecting a prominent facial fea-
areas while using the method; these areas are ture and transforming the name into a con-
associated with working-memory processes crete word) in the elderly. Both interactive
such as organization and image generation imagery (forming an image to link the facial
from words. The authors suggest two reasons feature and the name) and semantic orient-
for age-related deficits: basic resources (cor- ing (judging the pleasantness of the image
responding to frontal processing deficiency) association formed) enhanced name recall;
and failure to engage in task-relevant cog- semantic orienting also led to decreased for-
nitive processing (associated with posterior getting in delayed recall.
production deficiency). Both are presum- Stigsdotter and Backman (1989; Stigsdot-
ably relevant for acquiring expertise. ter, Neely, & Backman, 1993 a, 1993 b) inves-
Tanaka et al. (2002) used fMRI to study tigated the effect of multifactorial mem-
brain activity in abacus experts while they ory training (training in the method of loci,
were retaining digits. Experts showed more attentional skills, and relaxation) on recall
activity in cortical areas relating to visuo- of concrete words in older adults. Improve-
spatial working memory (bilateral frontal ment was sustained in testing three and one-
sulcus and superior parietal lobule) than half years later. The fact that a unifacto-
in areas related to verbal working memory, rial group, trained only in the method of
whereas the reverse was true in non-experts. loci, performed as well as the multifacto-
Thus, once again, experts seemed to utilize rial group in one of their studies suggests
visuo-spatial representations more than non- that training in encoding may be sufficient to
experts in order to aid memory. (Also see produce long-term gains. The effects trans-
Hill & Schneider, Chapter 3 7.) ferred to recall of objects but not recall of
abstract words or subject-performed tasks,
suggesting that generalization may be lim-
ited (Neely & Backman, 1995 ).
Practical Applications of Memory Kliegl, Smith, and Baltes (1989, 1990;
Expertise Baltes & Kliegl, 1992) used improvement
in serial word recall following training in
Effectiveness of mnemonic techniques for the method of loci (reflecting cognitive
memory remediation has been reported in reserve capacity) to investigate the nature
the elderly (for reviews see Camp, 1998; of age differences in cognitive performance.
Verhaeghen & Marcoen, 1996; Verhaeghen, Though both younger and older participants
Marcoen, & Goossens, 1992) and in spe- showed substantial plasticity, differences
cial needs populations (e.g., Mastropieri & between the age groups were magnified as
Scruggs, 1989). a result of training. The new rank order of
Yesavage and Rose (1983 ) examined the participants after training remained stable
effect of training in the method of loci on across subsequent practice sessions, suggest-
immediate and delayed serial recall in the ing that training substantially affected the
elderly. Effects were enhanced when the way the task was performed. Also, consistent
mnemonic training was preceded by con- with the assumption of age-related decline in
centration training (techniques designed to developmental reserve capacity, the unique
improve selective and sustained attention) variance in serial word recall associated with
and transferred to paired-associate learning. age group became more salient as training
Anschutz et al. (1985 ) demonstrated that progressed. Fluid intelligence (as measured
training older adults in the method of loci in by digit symbol substitution) before training
a free-recall task transferred to the more eco- became a predictor of performance over the
logically valid task of memory for a grocery course of training.
list in real-life shopping situations. Yesavage, Although a number of memory interven-
Rose, and Bower (1983 ) investigated fac- tions have been developed for use with the
550 the cambridge handbook of expertise and expert performance
pathological elderly, for example, those suf- Baltes, P. B., & Kliegl, R. (1992). Further testing of
fering from dementia (see Camp et al. 1996), limits of cognitive plasticity: Negative age dif-
the existing evidence suggests that mnemon- ferences in a mnemonic skill are robust. Devel-
ics, which rely on internal processing rather opmental Psychology, 2 8, 121125 .
than external aids and are aimed at enhanc- Bartlett, F. C. (193 2). Remembering. Cambridge,
ing explicit rather than implicit memory, are UK: Cambridge University Press.
likely to be of limited use for such popula- Biederman, I., Cooper, E. E., Mahadevan, R. S.,
tions (Camp, 1998). & Fox, P. W. (1992). Unexceptional spatial
memory in an exceptional memorist. Journal
of Experimental Psychology: Learning, Memory
and Cognition, 18, 65 465 7.
Future Directions
Binet, A. (1894). Psychologie des Grands Calcula-
teurs et Jouers dEchecs. Paris: Libraire Hachette.
Although empirical work on memory exper- Camp, C. J. (1998). Memory interventions for
tise has continued, there has been little theo- normal and pathological older adults. Annual
retical advance beyond development of Eric- Review of Gerontology and Geriatrics, 18, 15 5
ssons (1985 ) framework, but more work 189.
is needed on the effectiveness and ease of Carroll, J. B. (1993 ). Human Cognitive Abilities: A
application of different methods for dif- Survey of Factor Analytic Studies. Cambridge,
ferent types of task. Patterns of improve- UK: Cambridge University Press.
ment in memory performance as a result Chase, W. G., & Ericsson, K. A. (1981). Skilled
of training in mnemonics, particularly the memory. In J. R. Anderson (Ed.), Cogni-
method of loci, have been used to examine tive Skills and Their Acquisition (pp. 141189).
mechanisms underlying cognitive plasticity Hillsdale, NJ: Lawrence Erlbaum Associates.
over the life span (Kliegl et al., 1989, 1990; Ericsson, K. A. (1985 ). Memory skill. Canadian
Verhaeghen & Marcoen, 1996). This basic Journal of Psychology, 3 9, 18823 1.
memory research, aimed at understanding Ericsson, K. A., & Chase, W. G. (1982). Excep-
the reasons underlying age differences, has tional memory. American Scientist, 70, 607
been extended to the development of inter- 615 .
ventions for the normal and pathological Ericsson, K. A., Chase, W. G., & Faloon, S. (1980).
elderly from a rehabilitation perspective, Acquisition of memory skill. Science, 2 08, 1181
with increasing emphasis on external valid- 1182.
ity and applicability to everyday life. Proba- Ericsson, K. A., Delaney, P. F., Weaver, G., &
bly the most striking advance in recent years Mahadevan, R. (2004). Uncovering the struc-
ture of a memorists basic memory capacity:
has been the development of brain-imaging
Toward a methodology for resolving theoreti-
techniques. It is likely that these will lead to
cal issues in the study of exceptional memory.
the most significant progress in the future. Unpublished paper.
The study by Nyberg et al. (2003 ) (reviewed
Ericsson, K. A., & Kintsch, W. (1995 ). Long-
above) combines two of these approaches by term working memory. Psychological Review,
examining the neural basis of facets of age- 102 , 211245 .
related changes in memory improvement as Ericsson, K. A., & Kintsch, W. (2000). Shortcom-
a function of mnemonic training. ings of generic retrieval structures with slots
of the type that Gobet (1993 ) proposed and
modelled. British Journal of Psychology, 91, 5 71
References 5 90.
Ericsson, K. A., & Polson, P. G. (1988). A
Anschutz, L., Camp, C. L., Markley, R. P., & and cognitive analysis of exceptional memory for
Kramer, J. J. (1985 ). Maintenance and general- restaurant orders. In M. Chi, R. Glaser,
ization of mnemonics for grocery shopping by & M. Farr (Eds.), The Nature of Expertise
older adults. Experimental Aging Research, 11, (pp. 23 70). Hillsdale, NJ: Lawrence Erlbaum
15 7160. Associates.
exceptional memory 551
Gordon, P., Valentine, E., & Wilding, J. (1984). Mastropieri, M. A., & Scruggs, T. E. (1989).
One mans memory: A study of a mnemonist. Constructing more meaningful relationships:
British Journal of Psychology, 75 , 114. Mnemonic instruction for special populations.
Hamilton, R. H., & Pascual-Leone, A. (1998). Educational Psychology Review, 1, 83 111.
Cortical plasticity associated with Braille learn- Muller, G. (1911). Zur Analyse der
ing. Trends in Cognitive Science, 2 , 168174. Gedachtnistatigkeit und des Vorstellungsver-
Hunt, E. B., & Love, T. (1972). How good can laufes. Zeitschrift fur Psychologie, Erganzungs-
memory be? In A. W. Melton & E. Martin band, 5 , 73 67.
(Eds.), Coding Processes in Human Memory Muller, G. (1913 ). Neue Versuche mit Ruckle.
(pp. 23 725 0). New York: Wiley. Zeitschrift fur Psychologie, 7, 193 213 .
Hunter, I. M. L. (1962). An exceptional talent for Neely, A. S., & Backman, L. (1995 ). Effects of
calculative thinking. British Journal of Psychol- multifactorial memory training in old age: Gen-
ogy, 5 3 , 243 25 8. eralizability across tasks and individuals. Jour-
Hunter, I. M. L. (1977). An exceptional memory. nal of Gerontology: Psychological Sciences, 5 0B,
British Journal of Psychology, 68, 15 5 164. P13 4P140.
Hunter, I. M. L. (1996). Remembering a story Nyberg, L, Sandblom, J., Jones, S., Neely, A. S.,
after 26 years. Unpublished paper. Petersson, K. M., Ingvar, M., & Backman, L.
Ingham, J. G. (195 2). Memory and intelligence. (2003 ). Neural correlates of training-related
British Journal of Psychology, 43 , 203 2. memory improvement in adulthood and aging.
Proceedings of the National Academy of Sciences,
Kliegl, R., Smith, J., & Baltes, P. B. (1989). Testing-
100, No. 23 , 13 72813 73 3 .
the-limits and the study of adult age differ-
ences in cognitive plasticity of a mnemonic OKeefe, J., & Nadel, L. (1978). The Hippocampus
skill. Developmental Psychology, 2 5 , 24725 6. as a Cognitive Map. Oxford: Oxford University
Press.
Kliegl, R., Smith, J., & Baltes, P. B. (1990). On
the locus and process of magnification of age Rabbitt, P. M. A. (1993 ). Does it all go together
differences during mnemonic training. Devel- when it goes? The Nineteenth Bartlett Memo-
opmental Psychology, 2 6, 894904. rial Lecture. Quarterly Journal of Experimental
Psychology, 46A, 3 85 43 4.
Kliegl, R., Smith, J., Heckhausen, J., & Baltes, P. B.
(1987). Mnemonic training for the acquisition Rubin, D. C. (1995 ). Memory in Oral Traditions.
of skilled digit memory. Cognition and Instruc- New York: Oxford University Press.
tion, 4, 203 223 . Salthouse, T. A. (1985 ). A Theory of Cognitive
Kalakoski, V., & Saariluoma, P. (2001). Taxi Aging. Amsterdam: North Holland.
drivers exceptional memory of street names. Schlaug, G., Jancke, L., Huang, Y. X., Staiger,
Memory and Cognition, 2 9, 63 463 8. J. F., & Steinmetz, H. (1995 ). Increased corpus-
Luria, A. R. (1975 ). The Mind of a Mnemonist. callosum size in musicians. Neuropsychologia,
Harmondsworth, Middlesex: Penguin. 3 3 , 1047105 5 .
Maguire, E. A., Gadian, D. G., Johnsrude, I. S., Stigsdotter, A., & Backman, L. (1989). Multi-
Good, C. D., Ashburner, J., Frackowiak, R. S. J., factorial memory training with older adults:
& Frith, C. D. (2000). Navigation-related How to foster maintenance of improved per-
structural change in the hippocampi of taxi formance. Gerontology, 3 5 , 260267.
drivers. Proceedings of the National Academy of Stigsdotter Neely, A., & Backman, L. (1993 a).
Sciences of the United States of America, 97, Maintenance of gains following multifac-
43 984403 . torial and unifactorial memory training in
Maguire, E. A., Valentine, E. R., Wilding, J. M., late adulthood. Educational Gerontology, 19,
& Kapur, N. (2003 a). Routes to remembering: 105 117.
The brains behind superior memory. Nature Stigsdotter Neely, A., & Backman, L. (1993 b).
Neuroscience, 6, 9095 . Long-term maintenance of gains from memory
Maguire, E. A., Spiers, H. J., Good, C. D., Hart- training in older adults: Two 3 1/2 year follow-up
ley, T., Frackowiak, R. S. J., & Burgess, N. studies. Journal of Gerontology, 48, 23 3 23 7.
(2003 b). Navigation expertise and the human Susukita, T. (193 3 ). Untersuchung eines ausseror-
hippocampus: A structural brain imaging anal- dentlichen Gedachtnisses in Japan (I). Tohoku
ysis. Hippocampus, 13 , 25 025 9. Psychologica Folia, 1, 113 4.
552 the cambridge handbook of expertise and expert performance
Susukita, T. (193 4). Untersuchung eines ausseror- Wilding, J., & Valentine, E. (1985 ). One mans
dentlichen Gedachtnisses in Japan (II). Tohoku memory for prose, faces and names. British
Psychologica Folia, 2 , 15 42. Journal of Psychology, 76, 215 219.
Tanaka, S., Michimata, C., Kaminaya, T., Honda, Wilding, J., & Valentine, E. (1994a). Memory
M., & Sadato, N. (2002). Superior digit champions. British Journal of Psychology, 85 ,
memory of abacus experts: An event-related 23 1244.
functional MRI study. Neuroreport, 13 , 2187 Wilding, J., & Valentine, E. (1994b). Mnemonic
2191. wizardry and the telephone directory But sto-
Thompson, C. P., Cowan, T., & Frieman, J. (1993 ). ries are another story. British Journal of Psychol-
Memory Search by a Memorist. Hillsdale, NJ: ogy, 85 , 5 015 09.
Lawrence Erlbaum Associates. Wilding, J., & Valentine, E. (1997). Superior Mem-
Verhaeghen, P., & Marcoen, A. (1996). On the ory. Hove: Psychology Press.
mechanisms of plasticity in young and older Wilding, J., Valentine, E., Marshall, P., & Cook, S.
adults after instruction in the method of loci: (1999). Individual differences in memory abil-
Evidence for an amplification model. Psychol- ity and GCSE performance. Educational Psy-
ogy and Aging, 11, 164178. chology, 19, 11713 1.
Verhaeghen, P., Marcoen, A., & Goossens, L. Yesavage, J. A., & Rose, T. L. (1983 ). Concentra-
(1992). Improving memory performance in the tion and mnemonic training in the elderly with
aged through mnemonic training A meta- memory complaints: A study of combined ther-
analytic study. Psychology of Aging, 7, 242 apy and order effects. Psychiatry Research, 9,
25 1. 15 7167.
Wilding, J., & Marshall, P. (2004). Superior Yesavage, J. A., Rose, T. L., & Bower, G. H.
digit span: Validation of reported strategies (1983 ). Interactive imagery and affective judg-
by the time structure of recall. Unpublished ments improve face-name learning in the
paper. elderly. Journal of Gerontology, 3 8, 197203 .
C H A P T E R 32
Mathematical Expertise
Brian Butterworth
equip someone to demonstrate repro- The Great Mental Calculators: The Psychol-
ducible superior performance, particularly ogy, Methods, and Lives of the Calculating
in calculation, since this is the only area of Prodigies (Smith, 1983 ). I shall be drawing
expertise in which there is much evidence. heavily on this volume along with Scripture,
There have been three extensive early Mitchell, and Binet for biographical details
reviews of exceptional mathematical abili- of calculating prodigies. Finally, one of the
ties. The first was E. W. Scripture, a psy- leading modern investigators has published a
chologist at Leipzig (Scripture, 1891). He review of expert calculators (Pesenti, 2005 ).
reviewed in some detail the lives of 12 arith- Calculators have attracted also the atten-
metical prodigies, including one mathe- tion of experimental psychologists usually
matician of note, Carl Friedrich Gauss. From focusing on a single case. The Polish calcula-
a psychological analysis of these lives, he tor Salo Finkelstein was studied by Weinland
identified five characteristics that seemed to and Schlauch (193 7), who used sophisti-
distinguish the prodigies: the accuracy and cated analyses of careful timing data from
rapidity of memory, arithmetical associa- various mathematical tasks, but with no
tion (knowing lots of arithmetical facts and control subjects! The British mathematician
procedures), inclination, mathematical pre- Alexander Aitken was studied by Hunter
cocity, and imagination (visual imagery). (1962), and a more recent case, Rudiger
The second review, like Scriptures, was Gamm, by Pesenti and colleagues (Pesenti
published in the American Journal of Psychol- et al., 2001). Only Binet (1894) seems to have
ogy, and was by Frank D. Mitchell of The explored the general phenomenon, carrying
Psychological Seminary of Cornell Univer- out experiments on two professional theatri-
sity (Mitchell, 1907). Also like Scripture, he cal calculators and comparing their results
reviews the lives of prodigies, the same ones with other groups of practiced and unprac-
as Scripture plus a further case, the author. ticed calculators. He took considerable pains
These are summarized in a table that lists to find the optimal way of timing the stimu-
the heredity, development, education, men- lus presentation and the response, in an age
tal calculation, and memory of each prodigy. when there were no voice-activated relays.
Mitchell takes the view that prodigious abil- However, he was less careful in designing the
ities grow out of counting, a verbal skill experiments so that all the subjects received
where the numbers are recited out loud ini- the same stimuli under the same conditions.
tially, and internally thereafter. This has an What is striking about all these students
implication for the memory type used by of exceptional mathematical abilities is their
prodigies. In his own case and that of most of conviction that there is nothing special about
the prodigies he has analyzed, the memory mathematics as a cognitive domain. This
is of the auditory type, reflecting this early may be contrasted with Gardners popu-
experience since most of us begin to learn lar theory of multiple intelligences, one
numbers through counting, which involves of which is logico-mathematical (Gardner,
spoken words. Only between two and four 1983 ).
of the prodigies he examined have memory The research on mathematical, especially
of the visual type. numerical, abilities in general strongly sug-
The third review, by Alfred Binet (Binet, gests the existence of a domain-specific
1894), included reaction-time tests of two capacity. There are two main arguments in
theatrical calculators of his day. support of this.
More recently, Barlow (195 2) has written First, specialized brain areas, especially
about mathematical prodigies in the con- the left angular gyrus and the intraparietal
text of other kinds of prodigy (Barlow, 195 2). sulci, are active when mathematical activi-
And there is an excellent reconsideration of ties are taking place (Donlan, 2003 ; Gruber,
the calculators studied by Scripture, Binet, Indefrey, Steinmetz, & Kleinschmidt, 2001;
and Mitchell, along with useful data on more Pesenti, Thioux, Seron, & De Volder, 2000).
recent calculators by Steven B. Smith in When these areas are damaged, selective
mathematical expertise 555
impairment of calculation frequently occurs other means on this basis. Smith (1983 ) com-
(Cipolotti & van Harskamp, 2001). Notice pares it to juggling. Any sufficiently dili-
that the specialized number areas are dis- gent nonhandicapped person can learn to
tinct from the areas active in reasoning, juggle, but the skill is actually acquired only
which are in the prefrontal cortex (e.g., Goel by a handful of highly motivated individu-
& Dolan, 2004). als (p. 6). Calculating prodigies themselves
Second, there is evidence for an innate often say that their abilities come from their
basis to this specialization. One strand of interest in numbers rather than from some
evidence supports the claim for numerical special gift.
capacity in infants, even in the first few So our central question in this chap-
months of life, when neither language nor ter is whether mathematical expertise, and
frontal-lobe functions such as reasoning have in particular calculation expertise, depends
developed. They are able to respond dis- on high cognitive abilities in non-numerical
criminatively on the basis of numerosity domains, such as reasoning and memory, or
(Antell & Keating, 1983 ; Starkey & Cooper, whether it can exist as a domain-specific
1980) and can even mentally manipulate achievement.
numerosities at six months by working out
what would happen when an object is added
or subtracted from an array (Wynn, 1992, What Makes for Mathematical
2000, 2002; Wynn, Bloom, & Chiang, 2002). Expertise?
These infant capacities are similar to ones
observed in monkeys (Hauser, MacNeilage, Francis Galton was quite clear that any kind
& Ware, 1996), which suggests ancestral ver- of eminence depended on natural ability,
sions that may have evolved because the which was by-and-large inherited. By nat-
ability to recognize the numerosity param- ural ability, I mean those qualities of intel-
eter in the environment offers advantages in lect and disposition, which urge and qualify
foraging, mating (Edwards, Alder, & Rose, a man to perform acts that lead to reputa-
2002), and also in conflict with conspecifics tion. I do not mean capacity without zeal,
(McComb, Packer, & Pusey, 1994). (See nor zeal without capacity, nor even a combi-
Butterworth, 1999, for a review.) Moreover, nation of both of them, without an adequate
developmental disorder can lead to selective power of doing a great deal of very labori-
deficits in the acquisition of even these sim- ous work.
ple numerical concepts when intelligence, Having compared divines, wrestlers, men
memory, and language are all at normal lev- of science, painters, poets, and others, he
els (Landerl, Bevan, & Butterworth, 2004). found that eminent people tended to have
However, even if it is accepted that we eminent parents and to produce eminent
humans have inherited a specialized capac- offspring. Of course, we are talking mostly
ity for representing numerosities, it does not about eminent men and their fathers, since
follow that this is normally distributed, such it was difficult for women to achieve emi-
that some people have it to a greater degree nence. Overall, 3 1% of eminent men had an
than others, like height or IQ. It may be eminent parent. Some 26% of scientific men
more like color vision either it is normal, had scientific fathers, but 60% of scientifi-
or it is defective in one of a small number of cally eminent fathers have eminent sons (vs.
ways. In the same way that ability as a col- 41% average). Galton speculates descen-
orist may require normal color vision, the dants [are] taught . . . not to waste [their]
range of colorist abilities is not determined powers on profitless speculation (p. 3 20).
by better or worse color discrimination, so it Although inherited characteristics are
is possible that calculating abilities require held to be the key, it may be well to add
a normal numerical starter kit, but the a few supplementary remarks on the small
abstract and complex skills that make expert effects of a good education on a mind of
mathematicians and calculators is built by the highest order. A youth of abilities G, and
556 the cambridge handbook of expertise and expert performance
X, is almost independent of ordinary school even better known as the co-editor with
education (p. 43 ). He gives as an example Denis Diderot of the Encyclopedie.
DAlembert, who was a foundling . . . put
out to nurse as a pauper baby to the wife
of a poor glazier. The childs indomitable Intelligence
tendency to the higher studies, could not
be repressed by his foster mothers ridicule Ability in mathematics is widely seen as
and dissuasion, nor by the taunts of his a marker for intelligence, and disability in
schoolfellows, nor by the discouragements mathematics in school is seen as a marker for
of his schoomaster, who was incapable of low intelligence. This is not the place to dis-
appreciating him, nor even by the reiter- cuss the general relationship between intel-
ated deep disappointment of finding that his ligence and mathematical ability but only to
ideas, which he knew to be original, were point out that severe disability can co-occur
not novel, but long previously discovered by with good to superior IQ scores (Landerl
others (pp. 43 44). He records many other et al., 2004).
eminent men with a comparably unpromis- In contrast to Galton, Mitchell (1907)
ing history. noted that Skill in mental calculation is,
He pays particular attention to Wran- owing to the isolation of mental arith-
glers, students of mathematics at Cam- metic already noted, independent of general
bridge University. Not only do these men education: the mathematical prodigy may be
have to pass the entrance examinations to illiterate or even densely stupid, or he may
Cambridge (not as difficult then as now, be an all-round prodigy and veritable genius
I believe), but they were ranked from the (p. 13 1). Many expert calculators achieved
highest (Senior) to the lowest strictly accor- eminence in a way that suggested excep-
ding to the marks obtained in an exam. tional cognitive abilities. These included
These ranged from less than 5 00 to over the mathematicians Euler, Gauss, Aitken,
75 00. He was thus able to compare the and DAlembert, and scientists and engi-
Senior with the Second Wrangler, and he neers such as Ampere, Bidder, and Mitchell
noted the enormous discrepancy often obse- himself.
rved, with the Senior frequently achieving Shakuntala Devi (born 1940) is in the
double the marks of the Second. Guinness Book of World Records for being
For Galton, this was a demonstration of able to multiply two 13 digit numbers in
the enormous range of inherited abilities. 28 seconds. She was tested formally by the
His tripartite theory of eminence capac- psychologist A. R. Jensen, who showed that
ity, zeal, and power to work hard is quite she was not much better than average on
general and can be applied to any career. By standard IQ tests, and was actually slower
capacity, he seems to have meant something than average on some tests of speed of men-
like what we would now call intelligence tal processing, which Jensen regards as a reli-
or g. There was thus no special capacity for able measure of intelligence (Jensen, 1990).
mathematics. He seeks to demonstrate this Dehaene comments that therefore Devis
by analyzing the subsequent careers of top calculation were obviously not due to a
Wranglers. He notes that several were also global speed up of her internal clock: Only
classical scholars, and a few were both Senior her arithmetic processor ran with lightning
Wranglers and the top classical prizemen speed (Dehaene, 1997).
of the year. Many achieved distinction in Moreover, other prodigious calculators
areas very different from mathematics, such seemed to have been men of ordinary or
as law, politics, or becoming headmasters of even very low cognitive ability. Dase, who
great schools. did calculations for Gauss and Schumacher,
His example of DAlembert (17171783 ) was unable to comprehend the first ele-
reinforces this point. Not only was he a ments of mathematics. One pair of twins
mathematician of great distinction, he was with prodigious abilities for calendrical
mathematical expertise 557
Table 32 .1.
Number of items
Step Numbers in memory to be maintained Calculation
1 3 5 8 464 2
2 120000 3 400 3 00
3 20000 4 400 5 0
4 140000 3 120000 + 20000
5 3 200 5 400 8
6 143 200 4 140000 + 3 200
7 18000 5 60 3 00
8 161200 4 143 200 + 18000
9 3 000 5 60 5 0
10 164200 4 161200 + 3 000
11 480 5 60 8
12 164680 4 164200 + 480
13 1200 5 4 3 00
14 165 880 4 164680 + 1200
15 200 5 4 50
16 166080 4 165 880 + 200
17 32 5 4 8
18 166112 4 166080 + 3 2
(adapted from Smith, 195 3 , p. 5 4)
seven steps and six intermediate results. calculators use and invent algorithms that
After some practice with the task, Gamm minimize the load on working memory. It
was taking around five seconds a problem has also been suggested that one of the
with a high degree of accuracy. (Two digit consequences of expertise is the ability to
squares, by contrast, took him just over a exploit the unlimited capacity of long-term
second because they were simply retrieved memory in the service of the current task
from memory.) Such a sequence of opera- (Ericsson & Kintsch, 1995 ). It is as if experts
tions and data handling would put a consid- develop an ability to use long-term episodic
erable strain on normal working memory, yet memory to maintain task-relevant materi-
all kinds of expertise show enormous gains als, rather as computers extend the capacity
in the temporary storage of task-relevant of RAM by using swap space on the hard
materials: musicians can recall tunes after a drive to create a larger virtual memory
single hearing, chess masters can recall posi- (Butterworth, 2001, p. 12).
tions after a single tachistoscopic presenta- Language processing is a more familiar
tion as well as the whole game that they have example of prodigious skill after years of
just played, expert waiters can keep in mind daily practice enabling retention of infor-
the precise orders for up to 20 people with- mation well beyond the span of short-term
out writing them down (at least until the working memory. We can effortlessly retain
customer has paid). Experts develop a kind meaningful sentences of 20 words or more,
of Long-term Working Memory (Ericsson well beyond the span for unrelated words
& Kintsch, 1995 ). (about six) or words not in our language
(about three). Several related accounts of
this phenonemon propose cues in work-
Long-term Working Memory
ing memory for retrieving well-organized
As we have seen, one of the barriers to domain-specific information in long-term
mental calculation is the limited capac- episodic memory (Butterworth, Shallice, &
ity of working memory. Many exceptional Watson, 1990). Pesenti and colleagues argue
mathematical expertise 559
that Gamm has learned to use this LTWM Three, and then use a hand to count
facility to maintain task-related mathemati- on the second addend: Four, five, six,
cal information. seven, eight.
It turned out that computation compared 3 . Counting on from larger. It is more effi-
to retrieval of memorized number facts in cient, and less prone to error, when the
smaller of the two addends is counted.
both Gamm and the controls activated an
The child now selects the larger number
extensive visual processing system bilater-
to start with: Five, and then carries on
ally. According to the authors, this suggests Six, seven, eight. (Butterworth, 2 005 )
that during complex calculation, numbers
are held and manipulated onto a visual type However, many calculators report that
of short term representational medium. their early experiences involved manipula-
This contrasts with the more usual claim that bles. Bidder described how he learned mul-
sub-vocal rehearsal is . . . required for men- tiplication in the following way: I used to
tal arithmetic (Logie, Gilhooly, & Wynn, arrange [peas, marbles, or shot] into squares,
1994), but it would explain how it is possible of 8 on each side, and then on counting them
for brain damage to reduce digit span to two throughout, I found that the whole number
and yet allow a patient to reliably add two amounted to 64 (Quoted by Smith, 1983 ,
orally presented three-digit numbers (But- p. 212). It is probable that Bidder, like oth-
terworth, Cipolotti, & Warrington, 1996). ers who had early experience with manip-
(See this volume for more on LTWM.) ulables, also used a kind of visual coding.
We will see below that Gamms use Salo Finkelstein (born 1896), who seemed
of LTWM is supported by analysis of neu- to have had a standard Polish mathemati-
ral acitivity. cal education, without showing early signs
of exceptionality, calculated by visualizing
Auditory and Visual Working Memory numbers on a freshly washed blackboard.
His calculation ability seemed not to have
Mitchell noticed that there seemed to be
equalled many other prodigies in terms of
two types of working memory used in cal-
time or accuracy, but his ability to memo-
culation, visual and auditory, depending on
rize numbers was. He was able to remem-
how numbers were initially learned. Most
ber numbers up to about 28 digits following
children learn about calculating by count-
a one-second visual exposure; for 3 9 digits
ing aloud using the names of numbers,
he needed a four-second exposure. He was
names often some years before they under-
adept at repeating in either direction with
stand written numerals (Gelman & Gallistel,
equal accuracy, which traditionally suggests
1978). There appear to be three main stages
a visual memory. However, he also used a
in the development of counting as an addi-
wide variety of associations for substrings to
tion strategy:
help him, including numerical facts, such
1. Counting all. For 3 + 5 , children will as the fact that 1.41 is the square root of
count one, two, three and then one, 2, 2,5 92,000 is the number of seconds in a
two, three, four, five countables to month, 10,5 92 is a familiar telephone num-
establish the numerosity of the sets to ber, and 25 95 is the number of paragraphs of
be added, so that two sets will be made Spinozas ethics (Smith 1983 , Chapter 3 3 ).
visible for example, three fingers on Dehaene and colleagues (Dehaene &
one hand and five fingers on the other. Cohen, 1995 ) have proposed that multidigit
The child will then count all the objects.
arithmetic of the sort carried out by calcu-
2 . Counting on from first. Some children
lators depends on visualizing the digits on
come to realise that it is not necessary
to count the first addend. They can start a kind of mental blackboard. There is some
with three, and then count on another evidence that neurological damage can lead
five to get the solution. Using finger to deficits in spatial cognition, which can
counting, the child will no longer count lead to a kind of spatial acalculia where
out the first set, but start with the word the patient has difficulty in maintaining the
560 the cambridge handbook of expertise and expert performance
digits in columns accurately (Hecaen, Angel- approach to the problem, particularly, from
ergues, & Houillier, 1961). increased efficiency in managing memory
In the case of Gamm, it was possible to and accessing the next step in the procedure
identify the brain areas active during calcu- (Charness & Campbell, 1988)
lation, and hence whether verbal or visual Convergent evidence for the composi-
areas were active. It turned out that compu- tionality of arithmetical task comes from
tation compared to retrieval of memorized neurological patients, whose arithmetical
number facts in both Gamm and the con- abilities can be selectively affected in very
trols activated an extensive visual processing specific ways: the memories for facts alone
system bilaterally. can be lost, indeed the memories for facts
from each of the four arithmetical opera-
Domain-specificity in Memory tions can be selectively impaired (Cipolotti
& van Harskamp, 2001; van Harskamp &
Gamm had a forward span of 11 digits (con-
Cipolotti, 2001); arithmetical procedures
trols 7.2, SD = 0.8) and 12 digits backwards
can be lost from memory (Girelli & Delazer,
(controls 5 .8, SD = 0.8), whereas his letter
1996; McCloskey & Caramazza, 1987); and
span was in the normal range (Pesenti, Seron,
the ability to apply arithmetical principles
Samson, & Duroux, 1999).
to problems can be selectively spared or
Mondeux (18261861), a famous nine-
affected (Delazer & Benke, 1997; Hittmair-
teenth-century calculator, was described by
Delazer, Semenza, & Denes, 1994).
a contemporary as never having learned
It is, as has been noted above, that math-
anything besides arithmetic; Facts, dates,
ematical experts and calculating prodigies
places, pass before his brain as before a
build up enormous stores of what Scripture
mirror without leaving a trace (quoted by
called associations numerical facts and
Smith, 1983 , p. 294)
procedures.
Long-Term Working Memory (Ericsson &
Perhaps the greatest of recent calcula-
Kintsch, 1995 ), deployed by experts, is spe-
tors, Wim Klein acquired the multiplication
cific to the domain of expertise; thus, the
tables to 100 100 from experience [he]
musician, the chess master, and the waiter
got by factoring. However, he did set out to
will be normal on for example digit span,
memorize the table of logarithms up to 15 0.
(Ericsson & Kintsch, 1995 ). So, as Ericsson
This training enabled to him to achieve the
and Charness note, exceptional memory is
world record in extracting roots. He could
nearly always restricted to one type of mate-
extract the 13 th root of a 100-digit number
rial (Ericsson & Charness, 1994).
in under two minutes by using a method
that requires taking logarithms of the left-
Management and Strengthening
most group of numbers.
Memories
Aitken, similarly, had an enormous store
Solving even a simple arithmetical prob- of number facts. For him the year 1961
lem can be broken down into separable evoked the thoughts 3 7 5 3 , 44 2 + 5 2 ,
components, which will include retrieving and 40 2 + 19 2 . He could also recite the first
arithmetical facts from memory, retrieving 100 decimal places of (Hunter, 1962).
procedures for calculating (such as borrow- Like other calculating prodigies, Gamm
ing and carrying), understanding the arith- taught himself a vast range of number facts.
metical concepts demanded by the problem, Most of us know our multiplication tables,
and creating a hierarchical set of goals and and perhaps 5 0 simple additions (Ashcraft,
subgoals appropriate for reaching the solu- 1995 ), but Gamm has learned tables of
tion. Charness and Campbell have shown squares, cubes, roots, and so forth. Most
that, in learning a new algorithm for mul- of us know a few procedures for working
tiplying double-digit numbers, the mem- out problems that we cannot retrieve from
ory elements are strengthened by practice, memory, whereas Gamm has an enormous
but there is a larger effect from the overall store of procedures and shortcuts, some of
mathematical expertise 561
which he has learned from books, others he he said, 47 squared is 2209. Well, from that
has worked out for himself. moment, that was the light, and I never went
back (Hunter, 1962).
In the case of Gamm, he said that at
Motivation and Instruction
school he was very bad at arithmetic
because the teachers never explained the
Zeal and Inclination
concepts in ways he could understand. As a
Most exceptional calculators seem to have result he lost interest in mathematics until
been obsessed with numbers from the time about the age of twenty, when he came
they began to count. Jedediah Buxton kept across an algorithm for calendrical calcu-
a record of all the free drinks he received lation. He practiced it for fun, and then
from demonstrating his calculating prowess, entered for a TV competition where he
Thomas Fuller counted the hairs in a cows could win bets by solving various calcu-
tail, and Arthur Griffiths (18801911) kept lations. See Zimmerman, Chapter 3 9, for
track of the grains of corn he fed to the more on motivational factors in the devel-
chickens: 42,173 over three years (Smith, opment of expertise.
195 3 , p. 277). Srinivasa Ramanujan (1887
1920), a prodigious calculator and, accord-
The Role of Practice 10,000 Hours
ing to G. H. Hardy (Hardy, 1969), a natural
mathematical genius in the class of Euler or The highest level of expertise in violinists
Gauss, would work at mathematics in the studied by Ericsson et al. (1993 ) requires
mornings before work, and often stayed up 10,000 hours of practice (by the age of
all night working on problems. 20 years). In general, the level of expert per-
Calculators from an early age develop a formance was related directly (monotoni-
kind of intimacy with numbers. When Bid- cally) to the amount of practice. Similarly,
der was learning to count to 100, the num- expert calculators spend a great deal of
bers became as it were, my friends, and time learning numerical facts and proce-
I knew all their friends and acquaintances dures, though the exact amount has never
(Smith, 1983 , p. 5 ). Klein told Smith that been properly quantified. In preparation for
Numbers are friends for me, more or less. the T.V. program, Gamm started to train
It doesnt mean the same for you, does up to four hours a day, learning number
it, 3 ,844? For you its just a three and an facts and calculation procedures. He now
eight and a four and a four. But I say, Hi, performs professionally. His expertise is rare
62 squared. In a famous story, Hardy vis- enough to be a cause of wonder (the usual
ited Ramanujan in hospital and mentioned definition of prodigy).
that the taxi in which he had come was Some of the best evidence for the pure
number 1729, A rather dull number. No, effects of practice comes from an exper-
Hardy! It is a very interesting number. It iment carried out by Binet in which he
is the smallest number expressible as the compared the performance of two pro-
sum of two cubes in two different ways fessional calculators, Inaudi and Diamandi,
(C. P. Snow in his introduction to [Hardy, with cashiers from the Bon Marche depart-
1969]). ment store in Paris, who had had 14-years
In some cases, there is an incident that experience of calculating (there were no
awakens the interest. For Aitken, a teacher mechanical calculators available in the
chanced to say that you can use the fac- 1890s), but who, presumably, showed no
torization to square a number: a2 + b2 = special early gift for mathematics. He com-
(a + b)(a b) + b2 . Suppose you had 47 pared how long it took them to carry
that was his example he said you could out multidigit multiplications. Although the
take b as 3 . So (a + b) is 5 0 and (a b) is timing was about as accurate as it could have
44, which you can multiply together to give been without voice-activated relays, it is far
2200. Then the square of b is 9, and so, boys, from clear that the conditions were the same
562 the cambridge handbook of expertise and expert performance
simple mathematics. A recent twin study at 17 years the average difference between
of mathematical abilities showed that the boys and girls is still only 1%. The most
concordance rates were 0.73 for monozy- recent cross-national comparisons using the
gotic and 0.5 6 for dizygotic pairs (Alarcon, same tests in all countries, the Third Inter-
Defries, Gillis Light, & Pennington, 1997). national Maths and Science Survey (TIMSS)
Looking at the selective deficit of mathe- (Keys, Harris, & Fernandes, 1996), reinforces
matical ability, dyscalculia, of the dyscalculic the overall picture that in most countries,
probands, 5 8% of monozygotic co-twins and including the USA, there is no statistical dif-
3 9% of dizygotic co-twins were also dyscal- ference in the means, though there are enor-
culic. In a family study, it was found that mous differences among countries, suggest-
approximately half of all siblings of children ing that educational and cultural factors are
with dyscalculia are also dyscalculic, with a vastly more important than gender in the
risk five to ten times greater than for the gen- acquisition of mathematical skills.
eral population (Shalev & Gross-Tsur, 2001).
Another line of research has attempted
to assess whether sex-linked characteris- Brain Systems for Mathematical
tics contribute to mathematical expertise. Expertise
Benbow and colleagues have found in a
host of studies a significant advantage for There is now extensive evidence that routine
talented 12- to 13 -year-old boys over girls numerical tasks involve a fronto-parietal net-
at the upper end of the ability range, as work (Pesenti et al., 2000), where the pari-
measured by SAT-M (Scholastic Apti- etal components, perhaps especially the left
tude Test Mathematics), whereas SAT-V intraparietal sulcus, are relatively special-
(Verbal) showed no comparable difference ized for numbers (Dehaene, Piazza, Pinel,
(see Benbow, 1988, for a review). Ben- & Cohen, 2003 ). It is certainly the case
bow argues that the sex difference cannot that damage to the left parietal lobe can
be explained in terms of environmen- severely affect calculation (Cipolotti & van
tal hypotheses to do with attitudes, con- Harskamp, 2001), though almost nothing is
fidence, or teaching. She argues rather known about its effect on other mathemati-
that a combination of biological differences cal domains.
between the sexes is the cause, in particular More complex calculation in relatively
a more bilateral neural representation of cog- non-expert subjects established that the
nitive functions in the female brain (see next neural basis of simple retrieval (e.g., 3
section). 4 = ?), relative to a reading control, engaged
The differences between boys and girls in a left parieto-precentral circuit representing
SAT-M performance appears to follow from a developmental trace of a finger-counting
the much larger variance in boys scores, representation that mediates, by extension,
which would allow reliable differences at the the numerical knowledge in adult, plus a
top end of the range even if the mean score naming network including the left anterior
for girls were higher than for boys (Becker insula and the right cerebellar cortex (Zago
& Hedges, 1988). When one looks at the et al., 2001). On the other hand, complex
means, girls in England easily outperform computation (e.g., 3 2 24) engaged, addi-
boys in all subjects at all ages. There is one tionally, a left parieto-superior frontal net-
exception to this general rule: mathemat- work for holding multi-digit numbers in
ics. Girls are only just outperforming boys visuospatial working memory along with
(DfES, 2002). bilateral inferior temporal gyri, which is
On the other hand, Geary (Geary, 1996) implicated visual mental imagery. Corre-
reviewed a wide range of industrialized lated activity in the left intraparietal sulcus
countries and showed that boys, on aver- and the precentral gyrus may reflect the
age, still outperform girls in mathematical involvement of a finger movement represen-
problem solving. However, even in the USA tation network in the calculation process.
564 the cambridge handbook of expertise and expert performance
This is not to say that these skilled adults Schneider, Chapter 3 7, concerning brain
are counting on their fingers, but it may be changes with expertise development.
that the childhood use of fingers in learn-
ing to calculate somehow creates the neural
substrate for later acquisition of numerical Conclusions
knowledge (Butterworth, 1999).
There have been very few studies of the Our starting point was Galtons tripartite
brain systems of expert calculators. Ben- theory of eminence: capacity, zeal, and the
bow, OBoyle, and colleagues (e.g., Alexan- ability to do a very great deal of hard
der, OBoyle, & Benbow, 1996; OBoyle, work.
Benbow, & Alexander, 1995 ; OBoyle, Gill, Starting with capacity, it is clear that
Benbow, & Alexander, 1994; Singh & cases of individuals with exceptional math-
OBoyle, 2004) have investigated mathe- ematical, and especially calculating, ability
matically gifted children and adolescents, show enormous variety of cognitive abili-
with special reference to gender and brain ties. Some are highly intelligent, others aver-
organization. In general, they have found agely intelligent, yet others are classed by
more right-hemisphere involvement in a their peers (before standardized IQ test-
range of tasks, though, curiously, mathemat- ing) as stupid. So the kind of general intel-
ical tasks themselves have not been studied. lectual capacity supposed by Galton does
Pesenti and colleagues have published data not seem to apply here. Nor does our sur-
on the brain of an expert calculator carry- vey support Gardners (1983 ) idea of a
ing out mathematical tasks (Pesenti et al., distinct logical-mathematical intelligence,
2001). since many prodigies seem no better than
In a functional neuroimaging study, average, and indeed many are much worse
Pesenti and colleagues found that the than average, in reasoning.
prodigy Gamms calculation processes Zeal seems to be a characteristic common
recruited the same neural network as previ- to all the prodigies described here. They are
ously observed for both simple and complex obsessed with numbers, treat them as famil-
calculation (Zago et al., 2001), plus a sys- iar friends, and actively seek closer acquain-
tem of brain areas implicated in episodic tance with them.
memory, including right medial frontal They also seem to spend a great deal of
and parahippocampal gyri, whereas those time thinking and learning about numbers,
of control subjects did not (Butterworth, presumably for many hours a day: all seem
2001; Pesenti et al., 2001). Functional brain to have the capacity for very hard work.
imaging has established that speech-based Extensive practice has an effect on memory,
working-memory storage, of the kind that as would be expected, and it is quite spe-
supports standard digit-span tasks, involves cific. Exceptional calculators have acquired
the perisylvian language areas (Paulesu, enormous repertories of arithmetical facts
Frith, & Frackowiak, 1993 ). So Gamms and procedures, sometimes deliberately and
activations here are quite different. As noted sometimes by virtue of working with num-
above, it has been suggested that experts bers so much. In some cases, excellent arith-
develop a way of exploiting the unlimited metical memory goes hand in hand with very
storage capacity of long-term memory to poor memory for other materials. Working-
maintain task-relevant information, such memory is frequently cited as a serious lim-
as the sequence of steps and intermediate itation on complex mental calculation, and
results needed for complex calculation, eminent calculators learn or devise tricks to
whereas the rest of us still rely on the very reduce working-memory load.
limited span of working memory (Ericsson Is their exceptional ability confined to
& Kintsch, 1995 ). Gamms activations mathematics? Whereas some seem to excel
are consistent with his having devel- only in calculation, others have shown emi-
oped LTWM for arithmetical calculations nence in fields other than mathematics.
(Butterworth, 2001). See also Hill and Although there appears to be specialized
mathematical expertise 565
brain systems for numerical processing in and hard work, and it may be that we are
the parietal lobes, which have an innate born with dispositions toward them. Charles
basis, this may have little or nothing to do Darwin, in a letter to Galton, wrote I have
with exceptional ability. This is confirmed always maintained that excepting fools, men
by neuroimaging studies: exceptional calcu- did not differ much in intellect, only in zeal
lators such as Gamm seem not to be activat- and hard work; I still think this an eminently
ing the usual brain regions differently, but important difference (quoted by Ericsson &
rather recruiting new regions outside the Charness, 1994). It may also be the case that
parietal lobes to support the current task. some of us are born with a disposition to
There is now ample evidence for activity- enjoy or even be obsessed with an orderly
dependent plasticity: that is, that the func- domain like mathematics. However, there is
tioning, and even the structure, of brain no evidence at the moment for differences in
systems is shaped by practice and experi- innate specific capacities for mathematics.
ence (Amunts et al., 1997; Pascual-Leone &
Torres, 1993 ; Schlaug, Jancke, Huang,
Staiger, & Steinmetz, 1995 ; Schlaug, Jancke, References
Huang, & Steinmetz, 1995 ).
Ericsson and Charness (1994) have
Alarcon, M., Defries, J., Gillis Light, J., &
stressed the role of systematic teaching for Pennington, B. (1997). A twin study of math-
promoting the deliberate practice needed ematics disability. Journal of Learning Disabili-
for the highest levels of expertise. This, at ties, 3 0, 617623 .
least in part, is because deliberate prac- Alexander, J. E., OBoyle, M. W., & Benbow,
tice is not in itself rewarding. There are, in C. P. (1996). Developmentally advanced EEG
the biographies of mathematical prodigies, alpha power in gifted male and female adoles-
many counterexamples to this claim, where cents. International Journal of Psychophysiology,
precocity in mathematics could be nurtured 2 3 , 25 3 1.
in a systematic way, whereas others appear Amunts, K., Schlaug, G., Jancke, L., Steinmetz,
to have acquired exceptional mathematical H., Schleicher, A., Dabringhaus, A., et al.
skills despite very unhelpful early conditions. (1997). Motor cortex and hand motor skills:
It may be that finding solutions to Structural compliance in the human brain.
mathematical problems is, for the zealous, Human Brain Mapping, 5 , 206215 .
intrinsically rewarding. It may also be that Antell, S. E., & Keating, D. P. (1983 ). Percep-
the domain of mathematics is so ordered tion of numerical invariance in neonates. Child
Development, 5 4, 695 701.
that it is propitious for unsupervised learn-
ing since it is easy to check an answer by Ashcraft, M. (1995 ). Cognitive psychology and
simple arithmetic: A review and summary of
using a different method. Many prodigies
new directions. Mathematical Cognition, 1, 3
report external rewards also amazing their
3 4.
friends and family. This may be especially
Barlow, F. (195 2). Mental Prodigies. New York:
relevant in the savant, or near-savant cases,
Greenwood Press.
where there may be few ways to gain the
Becker, B. J., & Hedges, L. V. (1988). The effects
admiration of other people. Perhaps this is
of selection and variability in studies of gender
why parallels between music and mathemat- differences. Commentary on Benbow (1988).
ics are noticed. Both have intrinsic rewards Behavioral and Brain Sciences, 11(2), 183 184.
that are propitious for unsupervised learn- Benbow, C. P. (1988). Sex differences in math-
ing. In music, one can hear whether some- ematical reasoning ability in intellectually tal-
thing sounds right or not there is harmony ented preadolescents: Their nature, effects, and
or there is discord. And there are exter- possible causes. Behavioral and Brain Sciences,
nal rewards that do not require a teacher, 11(2), 169183 .
namely, that other people readily appreciate Binet, A. (1894). Psychologie des grands calcula-
good playing or singing. teurs et joueurs dechecs. Paris: Hachette.
Finally, are exceptional calculators born Butterworth, B. (1999). The Mathematical Brain.
or made? There is ample evidence for zeal London: Macmillan.
566 the cambridge handbook of expertise and expert performance
Butterworth, B. (2001). What makes a prodigy? A. J. Baroody & A. D. Dowker (Eds.), The
Nature Neuroscience, 4(1), 1112. Development of Arithmetic Concepts and Skills:
Butterworth, B. (2005 ). The development of Constructing Adaptive Expertise (pp. 3 3 73 5 8).
arithmetical abilities. Journal of Child Psychol- Mahwah, NJ: Lawrence Erlbaum Associates.
ogy & Psychiatry, 46(1), 3 18. Edwards, C. J., Alder, T. B., & Rose, G. J. (2002).
Butterworth, B., Cipolotti, L., & Warrington, Auditory midbrain neurons that count. Nature
E. K. (1996). Short-term memory impairments Neuroscience, 5 (10), 93 493 6.
and arithmetical ability. Quarterly Journal of Ericsson, K. A., & Charness, N. (1994). Expert
Experimental Psychology, 49A, 25 1262. performance: Its structure and acquisition.
Butterworth, B., Shallice, T., & Watson, F. (1990). American Psychologist, 49(8), 725 747.
Short-term retention of sentences without Ericsson, K. A., & Kintsch, W. (1995 ). Long-
short-term memory. In G. Vallar & T. Shallice term working memory. Psychological Review,
(Eds.), Neuropsychological Impairments of Short- 102 , 211245 .
Term Memory. Cambridge: Cambridge Univer- Ericsson, K. A., Krampe, R. T., & Tesch-Romer,
sity Press. C. (1993 ). The role of deliberate practice in the
Bynner, J., & Parsons, S. (1997). Does Numeracy acquisition of expert performance. Psychologi-
Matter? London: The Basic Skills Agency. cal Review, 100, 3 63 406.
Cappelletti, M., Kopelman, M., & Butterworth, Galton, F. (1979). Hereditary Genius: An Inquiry
B. (2002). Why semantic dementia drives you into its Laws and Consequences (Originally
to the dogs (but not to the horses): A theoret- published in 1869). London: Julian Friedman
ical account. Cognitive Neuropsychology, 19(6), Publishers.
483 5 03 . Gardner, H. (1983 ). Frames of Mind: The The-
Charness, N., & Campbell, J. I. D. (1988). Acquir- ory of Multiple Intelligences. New York: Basic
ing skill at mental calculation in adulthood: Books.
A task decomposition. Journal of Experimental Geary, D. C. (1996). Sexual selection and sex dif-
Psychology: General, 117 (2), 115 129. ferences in mathematical abilities. Behavioral
Cipolotti, L., & van Harskamp, N. (2001). Dis- and Brain Sciences, 19, 229 et seq.
turbances of number processing and calcula- Gelman, R., & Gallistel, C. R. (1978). The Childs
tion. In R. S. Berndt (Ed.), Handbook of Neu- Understanding of Number (1986 ed.). Cam-
ropsychology (2nd. ed., Vol. 3 , pp. 3 05 3 3 4). bridge, MA: Harvard University Press.
Amsterdam: Elsevier Science.
Girelli, L., & Delazer, M. (1996). Subtraction
Cockcroft, W. H. (1982). Mathematics Counts: bugs in an acalculic patient. Cortex, 3 2 , 5 47
Report of the Committee of Inquiry into the 555.
Teaching of Mathematics in Schools under the
Goel, V., & Dolan, R. J. (2004). Differential
Chairmanship of Dr. W. H. Cockcroft. London:
involvement of left prefrontal cortexin induc-
HMSO.
tive and deductive reasoning. Cognition, 93 (3 ),
Dehaene, S. (1997). The Number Sense: How the B109B121.
Mind creates Mathematics. New York: Oxford
Gruber, O., Indefrey, P., Steinmetz, H., & Klein-
University Press.
schmidt, A. (2001). Dissociating neural corre-
Dehaene, S., & Cohen, L. (1995 ). Towards and lates of cognitive components in mental calcu-
anatomical and functional model of number lation. Cerebral Cortex, 11, 3 5 03 5 9.
processing. Mathematical Cognition, 1, 83 120.
Hardy, G. H. (1969). A Mathematicians Apol-
Dehaene, S., Piazza, M., Pinel, P., & Cohen, L. ogy (Originally published in 1940). Cambridge:
(2003 ). Three parietal circuits for number pro- Cambridge University Press.
cessing. Cognitive Neuropsychology, 2 0, 487
Hauser, M., MacNeilage, P., & Ware, M. (1996).
5 06.
Numerical representations in primates. Pro-
Delazer, M., & Benke, T. (1997). Arithmetic facts ceedings of the National Academy of Sciences,
without meaning. Cortex, 3 3 , 697710. USA, 93 , 15 1415 17.
DfES. (2002). GCSE/GNVQ National Summary Hecaen, H., Angelergues, R., & Houillier, S.
Results, from http://www.standards.dfes.gov.uk/ (1961). Les varietes cliniques des acalculies au
performance/2002gcseresults.pdf?version=1. cours des lesions retro-rolandiques: Approche
Donlan, C. (2003 ). The early numeracy of chil- statistique du probleme. Revue Neurologique,
dren with specific language impairments. In 105 , 85 103 .
mathematical expertise 567
Hermelin, B., & OConnor, N. (1990). Factors and ment during linguistic processing. Cortex,
primes: A specific numerical ability. Psycholog- 3 0(3 ), 5 195 26.
ical Medicine, 2 0, 163 169. Pascual-Leone, A., & Torres, F. (1993 ). Plastic-
Hittmair-Delazer, M., Semenza, C., & Denes, ity of the sensorimotor cortex representation
G. (1994). Concepts and facts in calculation. of the reading finger in Braille readers. Brain,
Brain, 117 , 715 728. 116, 3 95 2.
Horwitz, W. A., Deming, W. E., & Winter, R. F. Paulesu, E., Frith, C. D., & Frackowiak, R. S. J.
(1969). A further account of the idiot savants: (1993 ). The neural correlates of the verbal com-
Experts with the calendar. American Journal of ponent of working memory. Nature, 3 62 , 3 42
Psychiatry, 12 6, 160163 . 3 45 .
Hoyles, C., Wolf, A., Molyneux-Hodgson, Pesenti, M. (2005 ). Calculation abilities in expert
S., & Kent, P. (2002). Mathematical Skills calculators. In J. I. D. Campbell (Ed.), Hand-
in the Workplace. London: Institute of book of Mathematical Cognition (pp. 413 43 0).
Education. Hove: Psychology Press.
Hunter, I. M. L. (1962). An exceptional talent for Pesenti, M., Seron, X., Samson, D., & Duroux, B.
calculative thinking. British Journal of Psychol- (1999). Basic and exceptional calculation abil-
ogy, 5 3 , 243 280. ities in a calculating prodigy: A case study.
Jensen, A. R. (1990). Speed of information- Mathematical Cognition, 5 , 97148.
processing in a calculating prodigy. Intelligence, Pesenti, M., Thioux, M., Seron, X., & De
14, 3 . Volder, A. (2000). Neuroanatomical substrates
Keys, W., Harris, S., & Fernandes, C. (1996). Third of Arabic number processing, numerical com-
International Mathematics and Science Study. parison and simple addition: A PET study.
First national Report. Part 1. Slough: National Journal of Cognitive Neuroscience, 12 , 461
Foundation for Educational Research. 479.
Landerl, K., Bevan, A., & Butterworth, B. (2004). Pesenti, M., Zago, L., Crivello, F., Mellet, E.,
Developmental dyscalculia and basic numeri- Samson, D., Duroux, B., et al. (2001). Mental
cal capacities: A Study of 89 Year Old Stu- calculation expertise in a prodigy is sustained
dents. Cognition, 93 , 99125 . by right prefrontal and medial-temporal areas.
Logie, R. H., Gilhooly, K. J., & Wynn, V. (1994). Nature Neuroscience, 4(1), 103 107.
Counting on working memory in arithmetic Rivera-Batiz, F. L. (1992). Quantitative liter-
problem solving. Memory & Cognition, 2 2 , acy and the likelihood of employment among
3 95 410. young adults in the United States. The Journal
McCloskey, M., & Caramazza, A. (1987). Dissoci- of Human Resources, 2 7 (2), 3 13 3 28.
ations of calculation processes. In G. Deloche Schlaug, G., Jancke, L., Huang, Y. X., Staiger,
& X. Seron (Eds.), Mathematical Disabilities: J. F., & Steinmetz, H. (1995 ). Increased corpus
A Cognitive Neuropsychological Perspective. callosum size in musicians. Neuropsychologia,
Hillsdale, NJ: Lowrence Erlbaum Associetes. 3 3 , 1047105 5 .
McComb, K., Packer, C., & Pusey, A. (1994). Schlaug, G., Jancke, L., Huang, Y. X., & Stein-
Roaring and numerical assessment in contests metz, H. (1995 ). In-vivo evidence of structural
between groups of female lions, Panthera leo. brain asymmetry in musicians. Science, 2 67 ,
Animal Behaviour, 47 , 3 793 87. 699701.
Mitchell, F. D. (1907). Mathematical prodi- Scripture, E. W. (1891). Arithmetical prodigies.
gies. Amercian Journal of Psychology, 18(1), 61 Amercian Journal of Psychology, 4(1), 15 9.
143 . Shalev, R. S., & Gross-Tsur, V. (2001). Devel-
OBoyle, M. W., Benbow, C. P., & Alexander, J. E. opmental dyscalculia. Review article. Pediatric
(1995 ). Sex differences, hemispheric laterality, Neurology, 2 4, 3 3 73 42.
and associated brain activity in the intellec- Singh, H., & OBoyle, M. W. (2004). Interhemi-
tually gifted. Developmental Neuropsychology, spheric interaction during global-local pro-
11(4), 415 443 . cessing in mathematically gifted adolescents,
OBoyle, M. W., Gill, H. S., Benbow, C. P., average-ability youth, and college students.
& Alexander, J. E. (1994). Concurrent finger- Neuropsychology, 18(2), 3 713 77.
tapping in mathematically gifted males evi- Smith, S. B. (1983 ). The Great Mental Calcu-
dence for enhanced right-hemisphere involve- lators: The Psychology, Methods, and Lives of
568 the cambridge handbook of expertise and expert performance
Calculating Prodigies. New York: Columbia Wynn, K. (2000). Findings of addition and sub-
University Press. traction in infants are robust and consistent:
Starkey, P., & Cooper, R. G., Jr. (1980). Percep- Reply to Wakeley, Rivera, and Langer. Child
tion of numbers by human infants. Science, 2 10, Development, 71(6), 15 3 5 15 3 6.
103 3 103 5 . Wynn, K. (2002). Do infants have numerical
van Harskamp, N. J., & Cipolotti, L. (2001). Selec- expectations or just perceptual preferences?
tive impairments for addition, subtraction and Commentary. Developmental Science, 5 (2),
multiplication. Implications for the organisa- 207209.
tion of arithmetical facts. Cortex, 3 7 , 3 63 3 88. Wynn, K., Bloom, P., & Chiang, W. C. (2002).
Weinland, J. D., & Schlauch, W. S. (193 7). Enumeration of collective entities by 5 -month-
An examination of the computing ability of old infants. Cognition, 83 (3 ), B5 5 B62.
Mr. Salo Finkelstein. Journal of Experimental Zago, L., Pesenti, M., Mellet, E., Crivello, F.,
Psychology, 2 1, 3 82402. Mazoyer, B., & Tzourio-Mazoyer, N. (2001).
Wynn, K. (1992). Addition and subtraction by Neural correlates of simple and complex men-
human infants. Nature, 3 5 8, 74975 1. tal calculation. Neuroimage, 13 (2), 3 143 27.
C H A P T E R 33
Expertise in History
interpretations. Such problems are termed adding that such study has two components,
ill structured, having more than one pos- continuity and causation, that is, the histo-
sible answer, requiring identification of con- rian documents change and works to deter-
straints, and having no agreed-upon solution. mine its causes.
With respect to the study of expertise Leinhardt, Stainton, and Virji (1994)
in history, although centuries ago histori- developed a definition obtained from inter-
ans as a group were concerned about the views with seven professional historians and
quality and accuracy of their writings, the two history teachers. The composite defini-
study of expertise in history as well as tion that Leinhardt et al. derived is His-
the social sciences emerged in the 1980s, tory is a process of constructing, recon-
the seminal papers being Voss, Greene, Post, structing, and interpreting past events, ideas,
and Penner (1983 ) in political science and and institutions from surviving or inferen-
Wineburg (1991) in history. Two suggestions tial evidence to understand and make mean-
are made regarding why and when these ingful who and what we are today. The
works emerged. One is that the study of process involves dialogues with alternative
expertise began in domains having well- voices from the past itself, with recorders of
structured problems, probably because such the past, and with present interpreters. The
problems are relatively tractable and in many process also involves constructing coherent,
cases computer simulations of performance powerful narratives that describe and inter-
could be obtained. Political science and pret the events, as well as skillful quanti-
history, however, generally are concerned tative and qualitative information from a
with ill-structured problems that have a theoretical perspective (p. 88). The above
large amount of potentially related informa- definitions emphasize change, reconstruc-
tion, and different experts may approach tion, and the importance of historiography,
the same issue differently, depending on that is, the processes whereby the historian
the experts theoretical background, related obtains and uses information.
knowledge, and other factors. Such solutions The beginning of the study of history is
are usually verbal arguments, which typi- usually marked by the writings of Herodotus
cally do not have right or wrong answers, (484425 bc), which contain an account of
but the answers may vary in relative accept- the Persian-Greek Wars and related mat-
ability. Furthermore, evaluation may occur ters (Herodotus, 1987). Although his writ-
by examining the acceptability of the infor- ings include some Homer-like mythological
mation provided per se, the extent to which and religious components, Herodotus pri-
the solution information supports the solu- marily wrote to provide a record for future
tion, and the quality of a counterargument readers. Thucydides (460400 bc), also
or alternative solution that may be offered. explicitly writing for the future, provided
The evaluator may be influenced also by an account (195 4) of the Peloponnesian
the evaluators own beliefs, theoretical ori- War that matched Athens and Sparta. The
entation, or other factors. Thus, because of war began in 43 1 bc, ending in 404 bc with
the nature of the domain, there is relatively the defeat of Athens. Thucydides focused
little opportunity to profit from the ben- on the military and political aspects of the
efits of quantitative analysis, which leads war, considering social, cultural, and eco-
to relatively less certainty in subject-matter nomic aspects when relevant. Moreover, by
knowledge and more heterogeneity in cons- including speeches of leaders he provided
traint usage. examples of political rhetoric, with Pericles
funeral oration probably the most notable.
In a broader sense, a major contribution
History
of these writers was to make history secular,
As expected, a number of definitions of his- that is, to consider human activity as caus-
tory have been advanced. Stearns (1998) re- ing events rather than seeing the forces of
gards history as change over time (p. 281), religion and myth as causes. In addition, the
expertise in history 571
authors used a criterion of evidence for the procedures have changed relatively little. As
historical accounts, namely, eyewitness tes- an example, in the Sixteenth century some
timony, including obtaining corroboration historians were concerned about their meth-
across witnesses. ods and developed rules of writing historical
From the time of Herodotus to the accounts. These included avoidance of ones
present, the field of history expanded sub- own or another persons religious, social, or
stantially, and during this time, history fre- patriotic biases, being detached when writ-
quently has been interpreted to support a ing about recent events, writing in an appro-
particular ideological viewpoint. As exam- priate manner and not for entertainment,
ples, during the Middle Ages history was and being sure to stick to facts rather than
used by the Church to determine Gods inventing them (Lemon, 2003 , p. 119).
Plan, during the Enlightenment it was used To this point we have provided some
as a demonstration of progress, and during background in the fields of expertise and of
the nineteenth century as a means to sup- history, and we turn now to the historian.
port broad socio-cultural ideas of civiliza- The goal of the historian generally is to study
tions progress (Lemon, 2003 ). a particular topic and provide a coherent,
Two historical developments of the twen- interpretive, and persuasive account stating
tieth century are noted. The field of social a position that usually but not necessarily
history developed, addressing more egalitar- is a narrative. The contents of the narrative
ian topics related to the lives of everyday are usually concerned with the changes that
people. Second, the Covering Law, stated took place in the topic being studied and
by Hempel (1942), constituted an attempt what produced them. In the next section we
to subsume history under the theoretical delineate two inter-related tasks of the his-
framework of positivism, the atheoretical torian, obtaining information and writing a
view of science emphasizing experimenta- narrative.
tion and the operation of laws. Under the
Covering Law, history was regarded as event
contingencies mediated by lawful relations,
Expertise in History: The Historians
that is, event A was followed by event B
because of a law operating to produce the Tasks
relation.
The Covering Model raised a number of In this section we are inserting the first
questions. Dray (195 7), for example, argued of ten Characteristics of History Experts
that since each historical set of events is (CHEs). Each CHE is a summary statement
unique, there would need to be many laws of of the findings of expert or expert-related
history. Moreover, historians generally have research in history or, in a few cases, a related
not found such laws, nor actually do they discipline.
seek them (Mink, 1987). In addition, there
is the importance of contextualization, that Obtaining Information
is, events occur in a context and the so-called
The historian may use many resources,
meaning or importance of that event is often
including biographies, autobiographies, jour-
a function of that context. This would sug-
nals, other library sources, museum archives,
gest a need for a law for each occurrence of
letters, paintings, objects, birth and death
A that is in a different context.
records, and photographs, examining them
Finally, comparing the Herodotus period
for their reliability, validity, authenticity, and
to the present, the study of history has
usefulness.
not changed substantially, especially if
compared, unfairly, to the technological
changes in scientific investigation. Specifi- che 1
cally, over time more and better sources have Historians evaluate sources emphasizing
become available, but the basic means and original and authentic information, using
572 the cambridge handbook of expertise and expert performance
criteria different than novices. Historians and contextualization is the act of situat-
also tend to have less intra-group variability ing a document in a concrete temporal and
in source evaluation than novices. spatial context (p. 77). Using both quantita-
Wineburg (1991) had eight high school tive and qualitative indices, experts used the
students and eight professional historians heuristics more frequently and more appro-
evaluate eleven sources pertaining to the priately than novices. Wineburg speculated
Battle of Lexington. Eight written sources about a possible fourth heuristic, identifying
came from diaries (2), an autobiography, a absent evidence, which was also a more fre-
deposition, a newspaper report, and a letter quent strategy among experts. As Wineburg
of protest, each written close to the time of notes, the prior knowledge of the historian,
the battle. The other two written excerpts both in general and in reference to specific
were from a historical novel and a high expertise, is an important factor in the effec-
school textbook, respectively, each written tive use of such heuristics.
in the 1960s. The other three sources were The use of heuristics by historians raises a
paintings done, respectively, in 1775 , 185 9, more general issue. Such heuristics, used as
and 1886. early as the Greek writings, provide the his-
Think-aloud protocols were obtained torian with a means of systematizing their
as well as rankings of each written docu- inquiry, helping to constrain and interpret
ment for its trustworthiness as a source for the material. In other words, the historian is
understanding what happened on Lexing- improving the structure of an ill-structured
ton Green (p. 75 ). The paintings were rated task. These and findings reported later in this
for what most accurately depicts what hap- chapter support a conclusion of Ericsson and
pened on Lexington Green (p. 75 ). Lehmann (1996) that across a wide range of
For the eight written documents, histori- domains, one of the characteristics of exper-
ans provided substantial within-group agree- tise is the ability to adapt and use constraints.
ment, novices having less. There was poor
agreement between the two groups, with
experts, for example, rating one of the diaries che 3
as most trustworthy of the written sources When analyzing sources, historians develop
and novices rating the textbook and novel mental representations of the events and
excerpts as most trustworthy, which histo- activities discussed in the text (situation
rians rated quite low. Rating the paintings models) and also generate subtext.
for their description, analysis, relevance, and Following cognitive text processing the-
qualification, experts provided more infor- ory (e.g., Kintsch, 1988), Wineburg (1994),
mation. History graduate students also were via the use of think-aloud protocols, stud-
shown to evaluate information usefulness in ied how eight historians processed the writ-
a different way than psychology graduate ten sources on the Battle of Lexington
students, with history students also demon- described above. Theoretically, each histo-
strating a more historian-like way of express- rian was presumed to have three mental rep-
ing and supporting their respective positions resentations derived from the text, namely,
(Rouet, Favart, Britt, & Perfetti, 1997). the text contents per se, the event, and
the subtext. The text representation is of
the contents of the text per se. The event
che 2 model is the representation of the event as
Experts use at least three heuristics in their conceived by the historian, and the sub-
analysis of sources, corroboration, sourcing, text, an elaboration of the situation model,
and contextualization (Wineburg, 1991). is constructed by the reader, using what-
Corroboration is the act of comparing ever seems reasonable about the time and
documents with one another, sourcing is the place of the text events, the intention of the
act of looking first to the source of the doc- author, the intended audience, and other fac-
ument before reading the body of the text, tors. To construct the event representation
expertise in history 573
the reader uses the preceding text contents expert, through the application of general
as well as prior knowledge. Using knowl- skills of the historian, was able to perform
edge and inferences based upon the event high-quality analyses but not at the spe-
model, the reader may construct a subtext, cialized experts level. Also, although both
possibly dealing with inferences about the historians spoke of contextual aspects of
authors goals and/or intentions, the moti- the respective documents, such analysis was
vations of the characters being discussed, greater for the specializing expert. Further-
or perhaps hidden assumptions about the more, this result showed a greater devel-
actions. Historians event and especially the opment of subtext by the specialist, that
subtext representations may vary depend- is, this historian used his or her knowl-
ing on the background knowledge of the edge and beliefs to consider matters such
historian. As an example, one historian, in as the intention of the text writer and for
analyzing the diary of a British officer who whom the text was intended. Finally, the-
wrote a description of what happened at the oretically, Leinhardt and Young postulated
Lexington battle, indicated that given what that two schemas operate when historians
the British army did immediately after the read documents, an identify and an inter-
battle, the diary writer did not have time pret schema, which in turn are related to
to write the diary until the next day, and procedures of analysis.
he wrote the diary because he thought he We want to mention in passing that an
may be questioned by his superior about his aspect of the historians task that is virtually
actions. Thus, he wrote the diary to show his never studied in history or any other subject-
actions in a favorable light. Subtext devel- matter domain is the ability of the expert to
opment has the difficulty of varying in the be adroit in selecting and defining the issue
extent to which different historians would to be studied. Problem finding is the crit-
be in agreement. ical first step in problem solving (cf. Get-
zels, 1979), and expert historians must have
skill at posing interesting yet researchable
che 4 questions.
Historians show expert-expert differences in
performance based on differences in areas of Narrative Construction and Analysis
specialization, but they show similarities in
the use of domain-related skills. Much has been written about the narrative,
Both Wineburg (1998) and Leinhardt and by historians, philosophers of history, literary
Young (1996) examined expert-expert dif- critics, and psychologists. We next focus on
ferences. Wineburg presented two history the historical narrative and its purpose and
experts with seven documents concerning use.
statements by or about Abraham Lincoln
and his attitudes concerning race. One che 5
experts specialization was the Civil War Historians have the goal of constructing nar-
and Abraham Lincoln, whereas the other ratives, based on evidence, that provide a
experts field was American History, the lat- reasonable account of particular historical
ter thus having substantial knowledge of the events and actions. As such, narratives are
Civil War but not that of the first expert. rhetorical constructions aimed at building a
Leinhardt and Young (1996) also provided case for a particular position in a manner per-
historians with documents close to or in the suasive to readers.
particular historians field of specialization. In his Rhetoric, Aristotle discussed argu-
Both studies demonstrated that experts ments involving probability or plausibility
specializing on the particular issue in ques- rather than certainty. He described two
tion produced more extensive and detailed types of argument structures employed for
analysis of the documents. In addition, in the this purpose, the enthymeme, that is, a
Wineburg (1998) study, the non-specializing claim and a supporting reason, and the
574 the cambridge handbook of expertise and expert performance
paradeigma or narrative (Aristotle, 195 4; but that this seldom happens. Instead, Mink
McGuire, 1990). In the narrative, the claim maintains that historic information increases
of the argument is the interpretation or con- via new discoveries and interpretations.
clusion of the author and the narrative is
the supporting evidence. Within the narra- che 6
tive there may of course be other arguments, Narrative quality is related to five com-
including extensive development of causal ponents: coherence, chronology, complete-
arguments. Moreover, the narrative may be ness, contextualization, and causation. Al-
written for a specific audience or a more gen- though not studied extensively, evidence
eral audience, as described by Perelman and suggests that narratives deficient in one or
Olbrechts-Tyteca (1969). Similarly, the his- more of these characteristics may produce
torian may develop an argument in expos- lower trustworthiness, the amount depend-
itory or categorical form. The use of these ing on the characteristics and the narrative
two types of argument within a discipline contents.
is current in other fields, such as Schums What constitutes a good narrative has
(1993 ) distinction of temporal and relational been studied empirically by Leinhardt,
arguments in jury decision making. Stainton, Virji, and Odoroff (1994). The
One of the issues raised by critics of the above five characteristics were those stated
historical narrative is the relation of the his- by seven professional historians when asked
torical narrative and the fictional narrative. to indicate the qualities needed for a narra-
One view is that the former is written to tive to be good.
provide an accurate account based on the Coherence, regarded as the most impor-
available evidence, whereas the latter is not tant characteristic, refers to the narratives
based on evidence or seeking truth, except organization and focus on a central theme.
in the most abstract sense (cf. Mink, 1987). Chronology refers to the reasonable and
Although this comparison appears to be rea- accurate discussion of the sequencing of
sonable, historical narratives require further events in time. Completeness, or exhaustiv-
examination. ity, refers to the use of all available evidence
A historical narrative typically consists that supports or opposes the expressed ideas
of organizing and interpreting actions and and arguments. In practice this criterion
events and their consequences. But in doing is quite demanding because a writer must
so, to what extent does the historian fill make judgments regarding what to include
in the dots? Hayden White (1987) has and what to omit. Contextualization refers
argued that a critical factor in historical nar- to placing the narrative subject matter into
rative is emplotment, that is, the events a broader perspective, for example, writ-
of the narrative take on meaning according ing about the causes of the 1991 Gulf War
to the development of the narratives plot. by framing it in the context of the inter-
Chronology is not enough, and the histo- est of the United States to preserve its
rian provides coherence and thematic con- Middle East hegemony. Causation is demon-
tent by generating the plot. To White the strating convincingly that events or actions
plot endows them (real events) with illu- produced particular consequences, thus pro-
sory coherence (White, 1987, p. xi), or sim- viding coherence to the narrative as a
ilarly, he states Reality wears the mask of whole and providing linkages of events and
meaning (White, 1987, p. 21). actions.
Mink (1987) argued that the narrative The importance of some characteristics of
serves as a cognitive instrument, and that narrative quality was studied by Voss, Wiley,
it provides the historian with a valuable and Sandak (1999). Set in a legal context,
tool providing opportunity for interpreta- fictional narratives were presented to college
tion. Mink also noted that the aggregation students, each narrative containing a murder.
of historical information could occur via The texts included a baseline narrative, a nar-
incorporation of narratives with one another, rative in which coherence and chronology
expertise in history 575
together were degraded, and one in which Voss (1996) in which an expert and novices
causation was degraded by using fewer irrel- read two newspaper editorials, making com-
evant (not pertaining to the murder) causal ments as they read. Novices included polit-
connectives than the baseline text. However, ical science or psychology graduate stu-
for all types of text the information per- dents. The expert, a political scientist, in
taining to the respective murder was held reading the first few sentences, categorized
constant. Participants were asked to regard the (anonymous) author of one editorial as
each text as the summary of a prosecuting a neo-realist, indicating problems he felt
attorney and to rate (on a numerical scale) with that position. Although the graduate
whether the defendant was guilty and also students of each discipline provided about
to rate the texts for cohesiveness and for the same number of evaluative comments,
quality. the political science graduate students sup-
The important result is that each of the ported their own positions more than the
degraded texts (coherence and chronology, psychology graduate students, whereas the
and causality) produced not only lower rat- latter provided more comments about pos-
ings of text coherence and quality than the sible bias. Furthermore, only the political sci-
baseline text, but also yielded lower guilty ence graduate students used counterfactual
ratings, that is, the poorer narratives were reasoning. These findings support discipline-
rated as having less convincing content, even related expertise development at the gradu-
though the murder-related contents were ate level.
equivalent.
che 8
wrote an upbeat book concerning its long- ideology, and an unofficial history, held by
term effects while the other author wrote of a number of the citizens, who nevertheless
it as a major disaster. Cronon notes that the are quite knowledgeable about the official
matter does not simply concern the differ- history. Moreover, an official history was
ence in conclusion, but that the stories told organized and coherent, whereas an unoffi-
are also different, an issue leading Cronon cial history was more piecemeal but in con-
into discussion of the historian and the con- flict with the official history. Wertsch and
straints involved. Rozin (1998) suggest three reasons for states
to have an official history: to develop an
che 8b instrument that shows the vision of the state,
to foster community identity, and to produce
Historians may produce alternative narra-
loyalty to the state.
tives because of differences in the time at
A related case is change in official his-
which they write. Such differences may be
tories when one set of government rulers
due to the number and accuracy of sources
replaces another. Ever since at least ancient
as well as the cultural milieu of the time.
Egypt it has been relatively common for
There are many examples of earlier
new regime members to destroy statues,
and subsequent revisionist writings. In this
change names, burn records and pictures,
regard, Lowenthal (2000) commented, We
and even kill family members of the old
are bound to see the Second World War dif-
regime. Even today, in Russia, compared
ferently in 1985 than in 195 0 not merely
to the Soviet days, history books are being
because masses of new evidence have come
rewritten, there are few pictures and posters
to light but because the years have unfolded
of Stalin and Lenin, and Leningrad is once
further consequences the Cold War, the
more St. Petersburg.
United Nations, the revival of the Japanese
Conflict between two official histories
and German economies (p. 78).
was shown in a study by Carretero, Jacott,
Historical events may be interpreted in
and Lopez-Manjon (2002). Comparing his-
terms of present conditions, but when this
tory textbooks in Spain and Mexico, the
is done to an extreme, thereby apparently
Spanish text spoke of Columbuss discovery
distorting events of the past, it is termed
of America, whereas the Mexican text spoke
presentism. On the other hand, Levine
of two cultures colliding. Similarly, whereas
(1989) pointed out that French historian
the Spanish text regarded Columbus as a
Marc Bloch stated that Misunderstanding
leader and hero who endured hardship, the
of the present is the inevitable consequence
Mexican text hardly mentioned Columbus,
of ignorance of the past, . . . But a man may
emphasizing explorers as a group, and indi-
wear himself out just as fruitlessly seeking to
cating that Columbus thought he was in Asia
understand the past, if he is totally ignorant
and at one time was put in chains.
of the present (pp. 671672).
che 8c che 8d
Alternative narratives are found in countries Alternative narratives are produced in class-
in which the government and the citizens rooms by differences in students cultural
are not in agreement regarding historical- backgrounds.
political-social thinking. Epstein (1998) reported that in a U.S. His-
Alternative historical narratives also may tory class having a slightly greater number
coexist because of political-cultural fac- of white than Afro-American students, the
tors. Wertsch and Rozin (1998), studying in teacher included more information about
the former Soviet Union, and Tulviste and the role of Afro-Americans and women
Wertsch (1994), studying in Estonia, dis- than typical U. S. History classes. Whereas
tinguished between an official history, a students generally agreed that the teacher
government-promoted history based on its had done this, questionnaire and interview
expertise in history 577
data indicated that Afro-Americans thought most other disciplines, such inferences usu-
it was not enough, holding that the teacher ally have plausibility or probability, but not
should be Afro-American. The Americans certainty. Finally, these processes typically
identified as most important by Afro- take place in the context of a goal, and
American students were Martin Luther reaching that goal may involve a number of
King, Malcolm X, and Harriet Tubman, such inferences, as well as assumptions and
whereas white students named George knowledge usage.
Washington, John F. Kennedy, and Martin By problem solving, we mean addressing
Luther King Jr. Afro-Americans also rated a domain-related question or problem and
the most important theme in U. S. His- developing a solution or answer to it. As pre-
tory as Afro-American equality (66%) and viously mentioned, the overall structure of
nation building (13 %), as opposed to nation a problem solution is an argument. Specif-
building (5 6%) and equality (22%) for white ically, the problem statement constitutes a
students. Finally, the most important source premise, the solution is viewed as the con-
for learning history was the family for Afro- clusion, and the steps to go from the initial
American students; for white students it was statement to the goal or conclusion com-
the textbook, followed by the teacher. These pletes the argument. The solution process,
findings are of course related to the official by showing the movement from the initial
and unofficial history issue. state to the conclusion, serves to justify the
In another example of cultural differ- conclusion. The form of the solution process
ences, Barton (2001) found that whereas in history or political science may be a nar-
American students had a linear view of his- rative, an expository text, or a series of steps
tory, focusing on the expansion and growth that would likely be in agreement with some
of the United States, students of Northern model of problem solving. Evaluation of the
Ireland considered history in terms of the solution, as stated, takes place by examin-
changes over the centuries of the peoples ing the quality of the evidence, the extent to
that inhabited Northern Ireland, a view of which the evidence supports the solution,
history as a change of context or scene rather and the examination of opposing solutions.
than linear development.
In this section we discussed the work
of obtaining information and writing nar- che 9
ratives. In the performance of such tasks, Because of the conceptual nature of domains
historians engage in a variety of mental activ- such as history, evidence or justification
ities, and we turn now to two such inter- for a claim or conclusion usually is verbal.
related processing skills, reasoning and prob- Moreover, such justification makes use of
lem solving. weak methods of reasoning and problem
solving such as analogy, decomposition, and
hypothesis or scenario generation and test-
Reasoning and Problem Solving
ing (thought experiments), as opposed to
By reasoning we mean the performance of stronger solution methods such as math-
an inferential process by which a person uses ematical proof and inference from well-
his or her knowledge to infer other informa- controlled scientific experiments.
tion related to the initial knowledge (Voss, Weak methods (Newell, 1980), such as
Wiley, & Carretero, 1995 ). The correctness those mentioned above, typically do not
or acceptability of the inference largely rests lead to specific conclusions, as opposed to
upon the proposition or information that strong methods such as mathematics or
justifies and connects the old and new infor- logic that provide certainty. Weak meth-
mation (Toulmin, 195 8). As such, an infer- ods also tend not to be taught in the class-
ence can be an argument, A inferring C, room but are acquired as language struc-
justified by the statement B, connecting A tures from an early age. These methods are
and C. Furthermore, in history, as well as called weak because they do not lead to
578 the cambridge handbook of expertise and expert performance
and decomposition. Second, the expert dition, that is, If X did not happen, would Y
attempted to improve on the solution pro- have happened? when X and Y did happen.
cess by using a strategy called constraint sat- The quality of a counterfactual argument
isfaction. The strategy adds power to the depends on the same three factors as other
solving process by delineating the constraints arguments, namely, the acceptability of the
that are particularly important to satisfy to evidence, the extent to which the evidence is
obtain an acceptable solution. judged to support the claim, and the consid-
A historical example that shows the eration of opposing evidence. However, peo-
importance of constraints in decision and ple may differ in their evaluations, especially
problem-solving processes involves two in relation to their own beliefs and knowl-
examples of the United States going to edge. Tetlock and Belkin (1996), discussing
war. When the Japanese bombed Pearl Har- the difficulties with such evaluation, suggest
bor on December 7, 1941, President Roo- that successful evaluation of counterfactu-
sevelt had virtually no choice in his deci- als is tenuous. Tetlock (1999) presented 5 2
sion to request war be declared on Japan. Soviet Union experts with seven counterfac-
His actions were quite constrained. Indeed, tuals involving hypothetical Soviet events.
probably any president or politician would He presented the entire counterfactual, the
have done the same thing. However, Pres- If-Then statement, or only the antecedent
ident George W. Bushs decision to invade If part of the statement. He also obtained
Iraq was a choice made under much less con- measures of the political positions of the
straint, that is, there were alternative actions participants on a liberal-conservative scale.
possible and many people would likely agree Ideological belief was related to agreement
that not all presidents or politicians would or disagreement with the If-Then state-
have made his decision, even if he himself ments, but not to the If only antecedents.
perhaps felt constrained. Subsequently, participants were told that
new evidence found in the Kremlin made
it more likely that one of the counterfactu-
che 10 als is more likely to have been true. Individ-
Causal reasoning in history faces at least two uals with positions not in agreement with
issues: the general absence of control groups that counterfactual asserted a much more
and the presence of temporally antecedent critical analysis, using three means to attack
events. Attempts to deal with these matters the new information, namely, challenging
include, respectively, counterfactual reason- the authenticity of the documents, chal-
ing and the categorization of prior events. lenging the representativeness of the docu-
In good scientific experimentation, the ments, and questioning the competence and
results frequently provide reliable and verifi- motives of the unknown investigator who
able evidence for the hypothesis under study, found the new information. These results
although such investigation does not pro- are thus in agreement with other findings
duce certainty. Taking the experimental ver- demonstrating belief bias. Tetlock (1999)
sus control design as the fundamental form found a similar effect when experts were
of experimentation, history generally, as life, asked to predict events, that is, when shown
has no control groups; historical events hap- to be wrong, experts felt high confidence in
pen once and control conditions do not exist their prediction and defended themselves by
(although simulations may be made). This use of similar bias procedures. Tetlock fur-
lack of control and the fact that history is ther pointed out that the results suggest that
cumulative, that is, past events and actions one reason why it is so difficult to learn from
are influencing todays events and actions, history is that there is the tendency to defend
constitute two difficulties for causal analysis ones own position even when confronted
in history. by information that brings that position into
Counterfactual reasoning allows the his- question, rather than examining ones posi-
torian to invent a hypothetical control con- tion for alternatives.
580 the cambridge handbook of expertise and expert performance
The above findings are not meant to sug- events are typically not considered as casual
gest that the use of counterfactuals is always antecedents by lay people, as the causal
open to substantial bias. Breslauer (1996), importance of preceding events has been
for example, effectively argues that debates found to be inversely related to their dis-
focused on counterfactuals have occurred tance from the event (Voss et al. 1994). Some
when there is a quite substantial body of writers (Mackie, 1965 ; Ringer, 1989) have
data available concerning the issue in ques- delineated two classes of preceding events,
tion, when the issue suggests a discontinuity namely, enabling conditions and causes. The
that is difficult to explain, and when there former are antecedent conditions enabling
is political partisanship. For example, the events to occur. They are neither necessary
collapse of the Soviet Union would likely nor sufficient to produce that event but may
be an interesting issue to deal with coun- look necessary a posteriori. Ringer (1989)
terfactually since theoretically it was unex- used the analogy of a driver hitting a sheet
pected and the large amount of historical of ice and sliding into a ditch. The fact that
data on pre-Soviet days would suggest the the person made the trip and took the par-
possible operation of a number of important ticular road and that the car is being driven
factors. at the appropriate speed are enabling con-
Pursuing the matter of learning from his- ditions, with the ice being the cause of the
tory, two political scientists (Holsti & Rose- car sliding. Within this general view, histori-
nau, 1977) collected data from over 2,000 cal flow appears to happen such that at any
individuals who worked in ten different point in time a set of conditions exists and
occupations. The study, addressing lessons some action is taken or occurs, as in a nat-
learned from the Vietnam War, had partic- ural disaster, and conditions change. Or a
ipants answer items pertaining to support government official takes an action that not
of or opposition to the war at the begin- only may but quite likely will lead to con-
ning and the end, the individuals politi- sequences that were not intended or antici-
cal belief system, and reasons for the out- pated. The historian attempts to make sense
come of the war. A relationship was found out of this chaos.
between a persons belief system and the
particular reasons for the wars outcome,
with hawkish people holding such reasons Concluding Comments
as not enough force was used early and that
Vietnam was help from Russia and China, In this section we examine expertise
and with dovish people indicating that the research in history and, in so doing, relate it
spirit of the North Vietnamese was under- to expertise research in other subject-matter
estimated and the United States had unre- domains. Before summarizing what factors
alistic goals. Occupation also was related to are important to expertise in history, we first
belief system and outcome selection. These ask whether there likely are any representa-
results also indicate the operation of belief tive tasks in history that enable the inves-
in learning from history. tigator to do some type of measurement of
The second difficulty of causal thinking such expertise. Not considering factual his-
is exemplified by a quotation found in var- torical knowledge as a legitimate indicator,
ious places of history literature, namely, If one is left with using tasks such as provid-
Cleopatras nose would have been one-half ing a substantial amount of information and
inch longer, the history of Western Civiliza- asking that a historical account or analysis be
tion would have been changed. The diffi- written concerning a particular aspect of the
culty is that for any event, there are typically information or perhaps some specific aspect
many preceding related events, thus raising of the information. Scoring would then be in
the question of whether one, a number, or terms of the use of skills of the historian as
all of the preceding events should be con- described above. Or, a number of historical
sidered as a cause (causes). Further, distant accounts or analyses of the same topic could
expertise in history 581
be provided, asking which account is best how are they organizing their knowledge
(or poorest) and why. Such tasks may be of rather than how much time is spent per se.
some value at the undergraduate or graduate The data we have in history suggest that par-
levels, but they probably would not be diag- ticular characteristics of expertise begin to
nostic for experts because skill performance emerge in graduate school, such as defense
likely would be quite high. However, as of ones position, use of counterfactuals, and
reported above, differences among experts the development of techniques of histori-
could be found depending on subject-matter cal analysis. Finally, with respect to history
knowledge. knowledge, we know of no studies that com-
Similarly, especially in political inter- pare the knowledge of the historian to that of
views, historians are sometimes asked to pre- the non-professionally trained history buff.
dict what is likely to happen in a given sit- Such a comparison may provide a better
uation in view of their knowledge of the understanding of the nature of the histo-
past. Although some present situations may rians expertise.
seem to be similar to particular past situ- A second knowledge-related factor is that
ations, these analogous occurrences are as the historian acquires a number of research
weak as other analogies. That is, situations skills. In addition to being excellent and pro-
may seem similar at one level, but the sim- lific readers, historians also have research
ilarity breaks down at some point. Indeed, skills that focus on how to seek and find
situations in history usually have large dif- sources and extract the most authentic and
ferences. Occasionally in the literature you high-quality information from them. How-
read The past is unpredictable or words ever, the research conducted to the present
to that effect, and when past events are not on how historians engage sources is but
predictable given knowledge and perspec- scratching the surface. There is much more
tive of the preceding actions and events, pre- to learn about source seeking and evalua-
dictions of what is going to happen in the tion and especially subtext generation. How
future is even more problematic. Indeed, it is does subtext generation facilitate historical
more accurate to say that historians are more understanding and how consistent is subtext
interested in providing an understanding or generation from one historian to another?
even explanation of the past than in trying to Similarly, how narratives are constructed is
predict it. another issue requiring more study.
Turning now to the major factors of Up to this point we have discussed how
historian expertise, the first factor of the historians and experts in other subject-
expert historian that is noted is the persons matter domains have a high level of domain-
subject-matter knowledge, especially that in specific knowledge and skill. Initially, it
his or her field of specialization, and to a seems that the skills, methods, and pro-
lesser extent in related subject-matter. How- cedures differ considerably across domains,
ever, despite the demonstrated importance and expertise is likely relatively domain-
of subject-matter knowledge across virtu- specific. What we argue, however, is that
ally all domains, our understanding of the across domains there are similarities in the
development of such knowledge is inade- highest level of problem-solving and rea-
quate, in history as well as other domains. soning processes. What is consistent across
Longitudinal studies of knowledge develop- domains is the existence of some problem or
ment are especially needed, although there issue, and the person working in the domain
quite likely are many routes to expertise has the goal of solving that problem. Fur-
in a given domain and it may not be pos- thermore, whatever the domain, the person
sible to identify necessary and/or sufficient needs to provide a solution that includes evi-
conditions to become an expert. Moreover, dence providing support for that solution; in
although experts in a given subject-matter other words, he or she must present a con-
domain no doubt spend much time in study, vincing argument. This general argumenta-
the question is what are they learning and tive structure is what is in common. What
582 the cambridge handbook of expertise and expert performance
differs among domains are the standards This divergence in expert opinion occurs
for what constitutes an acceptable solution because the domain of psychology is one
and the related matter of what in that par- in which answers cannot be demonstrated
ticular domain constitutes appropriate evi- formally. For all the expertise research that
dence and justification, that is, the content has been conducted, little attention has been
of the inquiry and the nature and extent directed to our own field. Expertise in psy-
of proof that is appropriate for that con- chology requires skill in both the collec-
tent. In physics the solution may be jus- tion of empirical evidence and in appropri-
tified by demonstration of the appropriate ate informal reasoning about that evidence.
steps to a solution. In this domain, this evi- As such, psychology would seem to be an
dence for solution is in a sense the proof. excellent candidate for future investigation
In history the same goal exists. The goal of expertise in academic domains.
of the historian is to convince an audience
that a narrative or perhaps an essay is highly
acceptable, and this is done by providing a Author Note
solution or answering a question with con-
The authors thank Melinda Jensen for her
tents that provide evidence, which is done
assistance with this manuscript. During the
by verbal argument. Most often there is not a
preparation of this manuscript, the sec-
deductive argument or mathematical equa-
ond author was supported by grant num-
tion but a position that is being advanced and
ber REC 0126265 from the ROLE Program
defended. As previously stated, a narrative
of the National Science Foundation. Any
often serves as the support and justification
opinions, findings, and conclusions or rec-
for the particular position. Thus, in history,
ommendations expressed in this material are
the solution is not usually correct or incor-
those of the author(s) and do not necessar-
rect but acceptable at some level. However,
ily reflect the views of the National Science
the similarity in goal structure and argu-
Foundation.
ment structure at a high level occurs across
subject-matter domains with the lower-level
specification of the specific domain goals, References
the solution, and the evidence or justifica-
tion of the solution being essentially domain Aristotle (195 4). The Rhetoric and the Poetics of
specific. Aristotle. (W. R. Roberts & I. Bywater, Eds. &
A final case is our own field, psychology. Trans.). New York: Modern Library.
The subject matter of this domain is a combi- Barton, K. (2001). A socio-cultural perspective on
nation of both informal and formal elements. childrens understanding of historical change:
We post constraints to make our problems Comparative findings from Northern Ireland
more tractable. We search for regularities or and the United States. American Educational
Research Journal, 3 8(4), 881913 .
laws. But ultimately, we are trying to explain
human behavior, and that behavior is neces- Breslauer, G. W. (1996). Counterfactual reasoning
in western studies of Soviet Politics and For-
sarily probabilistic. So, in essence, our field
eign Relations. In P. Tetlock & A. Belkin (Eds.),
may be much closer to history that we might
Counterfactual Thought Experiments in World
think initially. We may gather evidence, Politics (pp. 7194). Princeton, NJ: Princeton
but the crux of our argument is the inter- University Press.
pretation of data, and we build a narrative Carretero, M., Jacott, L., & Lopez-Manjon, A.
around the data. We test and operational- (2002). Learning history through textbooks:
ize them via experiments. As a result, one Are Mexican and Spanish students taught the
expert may say that experimental evidence same story? Learning and Instruction, 12 , 65 1
supports a conclusion, whereas another may 665 .
say that the study does not adequately test Chase, W. G., & Simon, H. A. (1973 a). Perception
the hypothesis or lacks a control condition. in chess. Cognitive Psychology, 4, 5 5 81.
expertise in history 583
Chase, W. G., & Simon, H. A. (1973 b). The history. Cognition and Instruction, 14(4), 441
minds eye in chess. In W. G. Chase (Ed.), 486.
Visual Information Processing (pp. 216281). Lemon, M. C. (2003 ). Philosophy of History. New
London: Academic Press. York: Routledge.
Chi, M. T. H., Feltovich, P. J., & Glaser, R (1981). Levine, L. W. (1989). The unpredictable past:
Categorization and representation of physics Reflection on recent American historiography.
problems by experts and novices. Cognitive Sci- The American Historical Review, 94(1), 671
ence, 5 , 12115 2. 679.
Cronon, W. (1992). A place for stories: Nature, Lowenthal, D. (2000). Dilemmas and delights of
history, and narrative. Journal of American His- learning history. In P. N. Stearns, P. Seixas, & S.
tory, 78, 13 4713 76. Wineburg (Eds.), Knowing, Teaching & Learn-
de Groot, A. D. (1965 ). Thought and Choice in ing History (pp. 63 82). New York and London:
Chess. The Hague: Mouton. New York University Press.
Dray, W. (195 7). Laws and Explanation in History. Mackie, J. L. (1965 ). Causes and conditions.
London: Oxford University Press. American Philosophical Quarterly, 2 , 245 264.
Ericsson, K. A., & Lehmann, A. C. (1996). McGuire, M. (1990). The rhetoric of narrative: A
Expert and exceptional performance: evidence hermeneutic critical theory. In B. K. Britton &
of maximal adaptation to task constraints. A. D. Pellegrini (Eds.), Narrative Thought and
Annual Review of Psychology, 47 , 273 3 05 . Narrative Language. Hillsdale, NJ: Erlbaum.
Epstein, T. (1998). Deconstructing differences Mink, L. O. (1987). Historical Understanding.
in African American and European American Ithaca, NY: Cornell University Press.
adolescents perspectives on United States his- Newell, A. (1980). One final word. In D. T. Tuma
tory. Curriculum Inquiry, 2 8, 3 97423 . & F. Reif (Eds.), Problem Solving and Education:
Getzels, J. W. (1979). Problem finding: A theo- Issues in Teaching and Research (pp. 175 189).
retical note. Cognitive Science, 3 , 167172. Hillsdale, NJ: Erlbaum
Hempel, C. (1942). The function of general Perelman, C., & Olbrechts-Tyteca, L. (1969).
laws in history. The Journal of Philosophy, 3 9, The New Rhetoric: A Treatise on Argumen-
3 5 48. tation. (John Wilkinson & Purcell Weaver,
Herodotus (1987). The History. (D. Grene, Trans.) Notre Dame: University of Notre Dame
Trans.). Chicago: University of Chicago Press. Press.
Holsti, O., & Rosenau, J. N. (1977). The meaning Ringer, F. K. (1989). Causal analysis in historical
of Vietnam: Belief systems of American lead- reasoning. History and Theory, 2 8, 15 4172.
ers. International Journal, 3 2 , 45 2474. Rouet, J-F, Favart, M., Britt, A., & Perfetti, C. A.
Kintsch, W. (1988). The role of knowledge (1997). Studying and using multiple docu-
in discourse comprehension: A construction ments in history: Effects of discipline expertise.
integration model. Psychological Review, 95 , Cognition and Instruction, 15 , 85 106.
163 182. Schooler, T. Y. E., Kennet, J., Wiley, J., & Voss,
Larkin, J. H., McDermott, J., Simon, D., & Simon, J. F. (1996). On the processing of political edi-
H. (1980). Expert and novice performance in torials. In R. J. Kreuz & M. S. MacNealy (Eds.),
solving physics problems. Science, 2 08, 140 Empirical Approaches to Literature and Aesthet-
15 6. ics (pp. 445 45 9). Norwood, NJ: Ablex.
Leinhardt, G., Stainton, C., & Virji, S. M. (1994). Schum, D. A. (1993 ). Argument structuring and
A sense of history. Educational Psychologist, 2 9, evidence evaluation. In R. Hastie (Ed.), Inside
7988. the Juror (pp. 175 191). Cambridge: Cambridge
Leinhardt, G., Stainton, C., Virji, S. M., & Odo- University Press.
roff, E. (1994). Learning to reason in history: Stearns, P. N. (1998). Goals in history teaching. In
Mindlessness to mindfulness. In M. Carretero & J. F. Voss & M. Carretero (Eds.), Learning and
J. F. Voss (Eds.), Cognitive and Instructional Pro- Reasoning in History (pp. 281293 ). London:
cesses in History and the Social Sciences (pp. 13 1 Woburn Press.
15 8). Hillsdale, NJ: Erlbaum. Tetlock, P. E., & Belkin, A. (Eds.). (1996). Coun-
Leinhardt, G., & Young, K. (1996). Two texts, terfactual Thought Experiments in World Politics.
three readers: Distance and expertise in reading Princeton: Princeton University Press.
584 the cambridge handbook of expertise and expert performance
Tetlock, P. E. (1999). Theory-driven reasoning Voss, J. F., Wiley, J., & Carretero, M. (1995 ).
about possible pasts and probable futures: Are Acquiring intellectual skills. Annual Review of
we prisoners of our preconceptions? American Psychology, 46, 15 5 181.
Journal of Political Science, 43 , 3 3 5 3 66. Voss, J. F., Wiley, J., & Sandak, R. (1999). On
Thucydides (195 4). History of the Peloponnesian the use of narrative as argument. In S. R.
War. New York: Penguin. Goldman, A. C. Graesser, & P. van den Broek
Toulmin, S. E. (195 8). The Uses of Argument. (Eds.), Narrative Comprehension, Causality,
Cambridge: Cambridge University Press. and Coherence: Essays in Honor of Tom Trabasso
Tulviste, P., & Wertsch, J. V. (1994). Official and (pp. 23 5 25 2). Mahwah, NJ: Erlbaum.
unofficial histories: The case of Estonia. Journal Wertsch, J. V., & Rozin, M. (1998). The Russian
of Narrative and Life History, 4, 3 113 29. Revolution: Official and unofficial accounts.
Tweney, R. D. (1991). Informal reasoning in sci- In J. F. Voss & M. Carretero (Eds.), Learning
ence. In J. F. Voss, D. N. Perkins, & J. A. and Reasoning in History (pp. 3 960). London:
Segel (Eds.), Informal Reasoning and Education Woburn Press.
(pp. 3 16). Hillsdale, NJ: Erlbaum. White, H. (1987). The Content of the Form: Nar-
Voss, J. F., Carretero, M., Kennet, J., & Silfies, rative Discourse and Historical Representation.
L. N. (1994). The collapse of the Soviet Union: Baltimore and London: Johns Hopkins Univer-
A case study in causal reasoning. In M. Car- sity Press.
retero & J. F. Voss (Eds.), Cognitive and Instruc- Wineburg, S. (1991). Historical problem solving:
tional Processes in History and the Social Sciences A study of the cognitive processes used in the
(pp. 403 429). Hillsdale, NJ: Erlbaum. evaluation of documentary and pictorial evi-
Voss, J. F., Greene, T. R., Post, T. A., & Penner, dence. Journal of Educational Psychology, 83 ,
B.C. (1983 ). Problem-solving skill in the social 73 87.
sciences. In G. H. Bower (Ed.), The Psychology Wineburg, S. (1994). The cognitive representa-
of Learning and Motivation, 17 (pp. 165 213 ). tion of historical texts. In G. Leinhardt, I. Beck,
New York: Academic Press. & C. Stainton (Eds.), Teaching and Learning in
Voss, J. F. & Post, T. A. (1988) On the solving History (pp. 85 13 5 ). Hillsdale, NJ: Erlbaum.
of ill-structured problems. In M. T. H. Chi, R. Wineburg, S. (1998). Reading Abraham Lincoln:
Glaser, & M. I. Farr (Eds.), The Nature of Exper- An expert/expert study in the interpretation of
tise (pp. 261285 ). Hillsdale, NJ: Erlbaum. historical texts. Cognitive Science, 2 2 , 3 193 46.
Part VI
GENERALIZABLE
MECHANISMS MEDIATING
EXPERTISE AND GENERAL
ISSUES
C H A P T E R 34
Creations, new insights, and paradigm shifts that of personality a theory that describes
emerge at junctures where different the- what people do and explains why they do
ories different world views meet, come in it. The principal descriptive concept of this
conflict, and are forced to accommodate. So theory is behavioral trait, a characteristic
it can be with a theory of expertise and a the- that persistently distinguishes one individ-
ory of human intelligence. Over the course ual from another despite variation in the cir-
of 20th-century research, the developments cumstances in which individuals are found.
of these two world views have run along The concept of behavioral trait is adopted
separately, rather like Leibnitzs clocks, each by analogy from the concept of biological
addressing much the same question what trait, with which it is easily confused. A
are the major capabilities of the human and behavioral trait is a way of behaving that
how do they come about but neither speak- emerges through learning over a course of
ing to the other. They have arrived at dif- development. It may be shaped partly by
ferent conclusions, neither thoroughly cor- genetic predispositions, but it is shaped also
rect, of course, but neither entirely wrong through societal and cultural influences that
either. Now, we reason, if we put the two involve learning. It becomes characteristic of
theories in newly-met dialog, we can drive an individual as development proceeds.
off the odious irrelevancies of each in a dis- When a particular sample of people is
tillation that captures the truthful essence drawn for study, people are at different levels
of both a new liquor: a theory that is of development and decline of a behavioral
more accurate than any that has gone before. trait. One person is found to reason well rel-
That, immodestly, is what we present in ative to other persons, for example at level
what follows. R8, say, on a ten-point scale and this char-
We deal with the question of how exper- acterizes that person (relative to other per-
tise fits within that part of human person- sons) over a stretch of time. As development
ality we describe with a theory of human proceeds the person may rise to a nearby
intelligence. Thus, the larger perspective is level say, R9 that is then characteristic
587
588 the cambridge handbook of expertise and expert performance
over a stretch of time, or the person may psychology. Good examples are Flanagan
decline to a nearby level (R7) that becomes and Harrison (2005 ), McArdle and Wood-
characteristic. The length of the stretch of cock (1998), McGrew and Flanagan (1998),
time over which the trait is characteristic McGrew and Woodcock (2001), and Wood-
is estimated empirically through analyses of cock (1990). Tests operational definitions
re-measure reliability. of concepts are described in these sources.
There may be function fluctuation We have no particular name for the the-
that is, reliable (non-error) change over the ory of expertise that we accommodate to
stretch of time in a behavioral trait. A per- the theory of intelligence. But the main
son characteristically R8, for example, may ideas of this theory are well described
function at R9 in the morning and R7 in in Ericsson (1996), Ericsson and Charness
the evening. Also, there may be precipitous (1994), Ericsson and Kintsch (1995 ), Eric-
decline in a behavioral trait due to any of sson and Lehmann (1996) and Ericsson,
many possible circumstances, such as neu- Chapter 3 8.
ronal damage in any of several different Let us turn first to the theory of intelli-
parts of the brain. To establish such changes gence.
empirically requires study of period fluctu-
ations in the trait, and the causes of these
fluctuations. Extended Gf-Gc Theory of Intellectual
Behavioral traits thus are at once some- Capabilities
what stable and somewhat dynamic. They
are characteristics that persistently distin- The theory is a generalization of findings
guish one individual from another despite obtained in five kinds of research: (1) struc-
variation in the circumstances in which indi- tural research: studies of the relationships
viduals are found, but they are ever changing and organization of abilities regarded as
as well. indicative of intelligence; (2) developmen-
Also important, behavioral traits are not tal research: studies of age differences and
absolutes; they are probabilistic patterns of changes in cognitive capabilities; (3 ) physio-
behaviors. The patterns indicate constructs logical-function research: studies of the
that is, abstractions. The precise set of behav- physiological mainly neurological cor-
iors that indicate a behavioral trait in one relates of cognitive abilities; (4) education-
individual can be different from the set of occupation research: studies of relationships
behaviors that indicate the same trait in between cognitive abilities and school
other individuals. The behaviors of such dif- and occupational performances; and (5 )
ferent sets are all parts of the same pattern, behavior-genetic research: twin and other
however. That is why they indeed do indi- family-comparison studies of the relation-
cate the same behavioral trait. To put it more ships between cognitive abilities and classifi-
concretely, not all people reasoning at a par- cations representing consanguinity. More of
ticular level get the same set of items cor- the evidence on which the theory is based
rect in a reasoning test that measures that derives from structural and developmental
trait, but the different configurations of right research, than from the other three kinds of
answers that yield the same score neverthe- research.
less indicate the same amount of the trait.1 Expertise theory, also, derives more
In what follows we will consider a trait from structural and developmental evi-
theory of intelligence and seek to integrate dence than from the physiological evi-
it with a trait theory of expertise. The the- dence (but see Hill & Schneider, Chap-
ory of intelligence that we will consider has ter 3 7), education-occupational research,
become known as the extended theory of and behavior-genetic research. In this chap-
fluid (Gf) and crystallized (Gc) intelligence. ter major attention will be given to the struc-
We will refer to this as extended Gf-Gc the- tural and developmental evidence, although
ory. It is described in many publications in a few references will be made to some
a merging theory of expertise and intelligence 589
of the other kinds of evidence (see also sure. The outcome behaviors are a complex
Ackerman & Beier, Chapter 9). expression of many abilities.
We infer ability from correlates. Factor
The Structural Perspective analysis is the method, par excellence, for
summarizing correlates from which to infer
There have been thousands of studies, con- an ability. From the evidence of such a
ducted over more than 100 years, of the summary we infer that a particular abil-
covariations among tests, tasks, paradigms, ity is dominant among all the abilities
and experiments designed to identify fun- involved and give a name to that abil-
damental features of human intelligence. ity. In this manner we describe the orga-
These studies indicate an astonishing num- nization among abilities in what we refer
ber and variety of cognitive abilities. They to as the structure of intellect. Its all a
range from abilities that seem to be elemen- bit circular. Thats the nature of structural
tary and narrow, to abilities that are broad evidence.
and complex abilities that clearly require
development through extensive learning
over long periods of time. principal classes of abilities
These apparent distinctions are some- As just suggested, evidence of structure
what misleading, however, because the among cognitive abilities has come mainly
observed behavior of what we call an abil- from factor-analytic studies. More than a
ity depends on many abilities. There are thousand of these studies have been con-
no clearly discernable demarcations in the ducted over the 100 years since Spearman
behavior that distinguish one ability from (1904) created the first such study. The stud-
another. ies have been directed at determining the
For example, in a measure of backward- number of fundamental capabilities, as well
span memory, a person is asked to repeat, as the nature of the capabilities, that describe
in the reverse of the order in which they measured individual differences in cogni-
were spoken, the names of seven numbers tive abilities thought to indicate intelligence.
spoken at a measured pace. This may seem Summaries of the results of these factor-
to be an elementary, narrow memory abil- analytic studies indicate no fewer than 75
ity, but when we examine what a person of what are called primary-ability factors.
must do to perform the task, it becomes These primary abilities are, for the most
apparent that several abilities are involved. part, positively correlated, and factor analy-
One must comprehend the instructions, dis- ses of selections among them indicate broad,
tinguish between the stimuli, and main- second-order abilities. But these abilities,
tain awareness of those distinctions while also, are usually positively correlated, and
at the same time reciting numbers in the factor analyses of measures of these abili-
asked-for order. Knowledge of number sys- ties indicate even broader third-order abil-
tems is involved. One person might group ities, and these abilities, too, are positively
the numbers as odd or even; another might correlated.
group them in sets of two or three; and Carroll (1993 ) has done a tour-de-force
others might organize them in other ways. summary of over 400 studies (477 data sets),
In structural analyses it turns out that this showing that the reliable covariation among
simple memory task is substantially corre- literally hundreds of ability tests2 designed to
lated with measures of complex reasoning. measure important features of human intel-
The simple and the complex have much in ligence can be described in terms of a system
common. of more than 40 primary abilities and eight
Inability to distinguish precisely the second-order factors of organization of the
abilities producing the scores (measures) primary abilities.
obtained with a particular ability test is The eight second-order factors separate
a problem with all the abilities we mea- classes of abilities are the principal
590 the cambridge handbook of expertise and expert performance
descriptive concepts of extended Gf-Gc the- the span of immediate awareness i.e.,
ory. They provide a sound basis for inte- for a minute or so.
grating a large proportion of the findings r Fluency of retrieval from long-term stor-
of developmental, physiological, education/ age (TSR). This is also labeled long-term
occupation, and behavior-genetic research. memory (Glm). These abilities indicate
Defining features of these abilities can be consolidation in learning that is, abilities
summarized as follows: of reconstruction in associational retrieval
r Acculturation knowledge (Gc). These abil- of things associated hours, weeks, and
years earlier. It can be measured with the
ities indicate the extent to which an indi- same kinds of tasks used to measure SAR,
vidual has incorporated the knowledge except retrieval is called for after a sub-
and language of the dominant culture. stantially longer period of time. It can be
The abilities are explicitly taught in the measured also in various kinds of associa-
curricula of schools, but more generally tion tests for example, associate (recall)
they are the abilities inculcated through words similar in meaning to a given
acculturation. They are the abilities that word.
are most prominently measured in what r Visual processing (Gv). The tasks measur-
are accepted as IQ tests such broad-
ing these abilities involve visual closure
band tests as the Stanford-Binet and
and recognizing the way objects appear in
Wechsler. They are the abilities mea-
space as they are rotated and flip-flopped
sured in school-readiness tests at all levels
in any of many different ways.
of schooling. For example, the scholas-
r Auditory processing (Ga). These are abil-
tic aptitude tests required for admission
to universities and colleges are primarily ities of comprehending patterns among
measures of Gc abilities. sounds, recognizing such patterns under
r Fluid reasoning (Gf). Identified as a com- conditions of distraction or distortion,
and maintaining awareness of order and
mon factor in company with the Gc com-
rhythm among sounds.
mon factor, the Gf abilities are primar-
r Processing speed (Gs). Although involved
ily reasoning abilities and abilities that
support reasoning. They are not as pro- in almost all intellectual tasks, the factor
foundly determined by factors of accul- is measured most purely in rapid scanning
turation and social class as are the Gc and comparisons in tasks in which almost
abilities. They are measured in tasks all people would get the right answer if
requiring reasoning (inductive, conjunc- the task were not highly speeded.
tive, and disjunctive), identifying rela- r Quantitative knowledge (Gq). These are
tionships, comprehending implications, the quantitative thinking and problem-
and drawing inferences in problems the solving abilities of mathematics the abil-
form of which are novel, not dealt with in ities measured in college-readiness tests
schools or other training programs or in such as the SAT and ACT. Although they
the media. Also, the reasoning involved are products of acculturation and indi-
in the factor is required in an immediate cate crystallized knowledge, their predic-
situation, a rather short period of time tors and predictions are different from the
that is, not reasoning that extends over a predictors and predictions of Gc. There
long period of time, such as that involved are good reasons to regard them as a sec-
in writing an article or constructing a ond major class of crystallized abilities,
machine or piece of art. distinct from Gc.
r Short-term apprehension and retrieval
(SAR). This is also referred to as short More detailed and scholarly descriptions
term memory and working memory. The of these abilities are provided in Carroll
abilities of this class are measured in a (1993 ), Flanagan and Harrison (2005 ),
variety of tasks that indicate how many McArdle and Woodcock (1998), McGrew
things one can, without rehearsal, hold in and Flanagan (1998), and Woodcock (1990).
a merging theory of expertise and intelligence 591
Gf does not represent a concept of general unfamiliar (Fagan & McGrath, 1981). Secure
intelligence. It represents only one form of attachment thus is conducive to the devel-
intelligence, not all the essential capabilities opment of Gc abilities.
of intelligence. Secure attachment also supports the
development of the abilities of Gc in other
the developmental perspective ways. This was demonstrated in pioneering
The case made up to this point for dis- studies of Bowlby (195 1) and Spitz (1946).
tinctions among different forms of intelli- They showed that secure attachment in the
gence is based on structural research. This earliest periods of development generalizes
case is made more compellingly by evidence to security in social interactions such as those
derived from the other forms of research, required of a student in learning situations.
particularly developmental and physiologi- This security leads to involvement in learn-
cal research. Let us turn to a brief consider- ing, which in turn promotes development
ation of some of this work. of cognitive abilities such as those of Gc
From about the third year of life onward, (Sroufe, 1977; Shaver & Mikulincer, 2002).
Gf and Gc can be identified as distinctly dif- Children who are securely attached, rela-
ferent abilities: the correlation between the tive to those who are not, are found to be
two is approximately .65 whereas the inter- more enthusiastic, persistent, and involved
nal consistencies of the factors are approx- in classroom learning; and they learn more
imately .90 (see Horn, 1985 , 1989, 1991 for and are able to firmly consolidate their
reviews). The correlation between the two learning.
becomes smaller at successively later stages In the studies of Harlow and his cowork-
of development. In adulthood the correla- ers (e.g., Harlow & Suomi, 1970), rhesus
tion is found to be in a range of about .40 monkeys that were more securely attached
to .5 0 (with factor internal consistencies in to a primary caregiver were better able to
a range of .80 to .90). overcome neophobia, respond to curiosity,
Gc correlates in a range of .40 to .60 with venture into the open field (a strange, some-
the educational or economic level of ones what frightening domain for a rhesus mon-
parents and with ones own educational or key),and deal with novelty.
economic level at later ages. It correlates Secure attachment facilitates neophilia.
also with other indicators of social class. Gf, Even children who are highly predisposed to
in contrast, correlates in a range of .20 to neophobia, but are raised under conditions
.45 with these same indicators of social class that promote security, are able to respond to
(Cattell, 1971; Horn, 1985 , 2002). curiosity, explore novelty, and thus develop
The development of Gc abilities is associ- relatively high levels of Gc. Thus, fear of
ated with individual differences in the qual- the novel need not inhibit the development
ity and amount of formal education, and of Gc.
with child-rearing that promotes the valu- On the other hand, the development of
ing of formal education and the attainment Gf abilities is more dependent on the basic
of the knowledge of the dominant culture. adaptability of neophobia and has little asso-
These educational and child-rearing condi- ciation with factors that promote security in
tions are in turn positively associated with upbringing. Thus, in the development of Gf
the conditions of secure home, neighbor- abilities, the factors that promote security
hood, and school environments. The security have less influence than in the development
of a childs attachment to a primary care- of the Gc abilities, and in virtue of this, an
giver, particularly during the earliest peri- adaptive neophobia will correlate with Gf.
ods of development, sets the stage for devel- It is likely that many factors determine the
opment of Gc (Burgess, Marshall, Rubin, kind of security that enables one to deal with
& Fox, 2003 ; Shaver & Mikulincer, 2002). novelty and consequently develop cognitive
Such attachment enables curiosity about abilities. Early-in-life secure attachment is
what is novel and a drive to investigate the only one such factor. We can not expect that
a merging theory of expertise and intelligence 593
it will account for a great amount of the SAR: Short-term Apprehension and Retri-
covariances. Nevertheless, the evidence sug- eval. The findings of many studies indicate
gests that such attachment has substantial decline with adulthood in a wide variety of
positive relationship with social class, edu- tests measuring aspects of short-term mem-
cation, acculturation, and the development ory (Backman, Small, Wahlin, & Larsson,
of Gc abilities. 2000; Craik, 1977; 2000; Craik & Byrd,
Much of the evidence on the distinction 1982; Craik & Trehub, 1982; Finkel, Reynold,
between the Gf and the Gc classes of abil- McArdle, Gatz, & Pedersen, 2003 ; Lane &
ities has come from studies of people of Zelinski, 2003 ; Schaie, 1996, 2000). The
different ages in adulthood (cross-sectional decline is smaller in tasks that require very
research) and changes in the averages in short periods of retention; there are virtually
repeated measurements of adults (longitu- no age differences for the Sperling (1960)
dinal research). These two different kinds of kind of task, in which retrieval occurs over
research converge in indicating aging decline periods of only a few milliseconds. But for
of Gf abilities, coupled with no decline or measures in which information is presented
even improvement of Gc abilities (see also over periods of a few seconds and retrieval
Krampe & Charness, Chapter 40). is required after short periods of up to a
minute or two, age-related declines gener-
ally have been found. This is true of mem-
ory for information the subject could regard
Declining Capacities during
as meaningful, as well as for nonsense mate-
Adult Development
rial, although age differences appear to be
The evidence of decline comes from smaller for memory for the meaningful kind
both cross-sectional and longitudinal stud- of information (Cavanaugh, 1997; Charness,
ies (Horn, 1991; Horn & Donaldson, 1980; 1991; Craik, 2000; Craik & Trehub, 1982;
McArdle, Ferrer-Caja, Hamagami, & Wood- Ericsson & Delaney, 1998; Gathercole, 1994;
cock, 2002; McArdle, Prescott, Hamagami, Kaufman, 1990; Radvansky & Copeland,
& Horn, 1998; Salthouse, 2001; Schaie, 2000; Radvansky, Copeland, & Zwaan,
2000). These two kinds of studies control 2003 ; Salthouse, 1991a; Salthouse & Ferrer-
for and reveal different kinds of influences, Caja, 2003 ; Schaie, 1996). The more com-
and have different strengths and weaknesses plex the memory task, and the more it
(Horn & Donaldson, 1980). Yet in major requires that material be held in awareness
respects the findings from the two kinds of while doing other things as in definitions
studies are largely in agreement. The lon- of working memory the larger the negative
gitudinal findings suggest that the points in relationship to age.
adulthood at which declines occur are later Tests of working memory also have sub-
than is indicated by the cross-sectional find- stantial correlations with Gf. For example,
ings, but the evidence of which abilities the negative age relationship for backward-
decline, and which decline more and less span memory, in which the subject must
than others, is essentially the same. recall the to-be-remembered elements in the
The principal declines are in Gf, Gs, reverse of the order in which they were
and SAR, which together are referred to presented, is significantly larger (absolute
as age-vulnerable abilities. The abilities that value) than the negative age relationship for
are particularly vulnerable to conditions forward-span memory (Craik, 1977; Craik &
associated with aging are also particularly Trehub, 1982; Masunaga & Horn, 2000, 2001;
vulnerable to conditions associated with Horn et al., 1981; Salthouse, 1991b; Schaie,
brain damage (Horn, 1982, 1985 for early 1996), and the correlation of this test with
reviews). Some features of the neurological Gf can be as large as its correlation with SAR
system decline (on average) with advancing (Masunaga & Horn, 2001).
age (Medina, 1996; Raz, 2000; Tisserand & Gs: Processing Speed. Most cognitive tests
Jolles, 2003 ; Hill & Schneider, Chapter 3 7). are speeded. The results of many studies
594 the cambridge handbook of expertise and expert performance
suggest aging decline in speed of perfor- the participant must provide an answer, not
mance and in speed of thinking (Birren, merely select one from a multiple-choice set.
1974; Kausler, 1990; Christensen, 2001; Moreover, in the instructions the participant
Salthouse, 1992, 1993 , 1994; Salthouse & is taught that some problems do not have a
Ferrer-Caja, 2003 ). Older adults are slower solution, and if what appears to be a prob-
than their younger counterparts in both sim- lem of that kind is encountered, the correct
ple and choice reaction time (RT), although behavior is to mark NC, representing No
the magnitude of the differences, assessed correct solution, and go on to the next prob-
in standard score units, are larger for choice lem. Thus, when one does not work out a
RT. In studies in which young and old sub- solution to a problem, there is little incen-
jects are provided opportunity to practice tive to guess or to estimate a best bet for
a choice RT task, practice does not elimi- a solution. There is no reward for working
nate the age differences, and no notewor- quickly and no reward for guessing. And the
thy age-by-practice interactions are indi- score indicates merely average level of diffi-
cated (Madden & Nebes, 1980; Salthouse & culty of the problems solved (independently
Somberg, 1982). of the number of problems attempted).
Gf: Fluid Reasoning. The research find- Under these conditions of measurement,
ings are consistent in demonstrating steady decline with age from age 25 to age 65 years
decline of Gf over most of the period of is approximately 1.1 standard deviation that
adulthood. The decline is seen with mea- is, expressed in IQ units in which a standard
sures of syllogisms and concept formation deviation is 15 , the decline is about 16.5 IQ
(Fisk & Sharp, 2002; McGrew, Werder, points over 40 years of aging (about 4.1 IQ
& Woodcock, 1991), in reasoning with units for each ten years of aging).
metaphors and analogies (Salthouse, 1987; Even though this lowly-speeded measure
Salthouse, Kausler, & Saults, 1990), with of Gf does not measure speed in providing
measures of comprehending series, as in let- answers, the Gs factor of cognitive speed
ter series, figural series, and number series accounts for about one-half of the decline.
(Horn, 1975 , 1991; Noll & Horn, 1997; Salt- On first consideration, this evidence seems
house, Kausler, & Saults, 1990), and with to indicate that some form of speed of think-
measures of mental rotation, figural rela- ing is involved in fluid reasoning. But we
tions, matrices, and topology (Cattell, 1979; have found reasons to question this first
Horn, 1977; McArdle, Hamagami, Mered- consideration: the amount of decline in Gf
ith, & Broadway, 2000). In each case the that is associated with the Gs speed measure
evidence for Gf decline is cleanest if the is completely accounted for by a measure of
elements of the reasoning test are novel or slow tracing that involves no speed in perfor-
equally familiar to all that is, if the test gives mance and in fact involves only the opposite
no advantage to those with greater knowl- of this, behaving as slowly as one possibly
edge of the culture. can. In the slow-tracing test the requirement
Although many of the tests that have indi- is to trace a line as slowly as possible, and the
cated the aging decline of Gf have a speeded shorter the line one traces in a unit of time,
component, the decline is also indicated very the larger the score.
well by lowly-speeded tests that require res- These, short tracings (large scores) are
olution of high-level (difficult) complexities positively correlated with fast matching and
(Horn, 1991; Horn et al., 1981; Noll & Horn, identifying in the Gs measure and with high
1997). level-of-difficulty scores on the letter-series
In these lowly-speeded tests the score measure of Gf. Moreover, each of these mea-
indicating level of reasoning ability is the sures is negatively related to age over a range
average level of difficulty of the problems of from 25 to 65 years.
correctly solved, not a count of the num- When the Gs measure is parted out of the
ber of problems solved in some unit of time. Gf measure, the 16.5 IQ units of decline of
The items are open-ended letter series, so Gf is reduced to about eight IQ units for the
a merging theory of expertise and intelligence 595
residual Gf (what is left after Gs is removed). Horn & Hofer, 1992; Kaufman, 1990; McAr-
But if the slow-tracing measure is parted out dle et al., 2000; Rabbitt & Abson, 1991;
of Gf, the reduction in decline is approxi- Schaie, 1996, 2000; Stankov & Horn, 1980;
mately the same from 16.5 to about eight Woodcock, 1990).
IQ units when slow-tracing is removed The results are more equivocal in sug-
and the decline in Gf that is associated with gesting whether or not Gc abilities continue
Gs is reduced to near zero: the slow-tracing to increase beyond 60 to 70 of age. Some
measure, which is not at all speeded, com- findings suggest yes, into the 80s (e.g.,
pletely accounts for all the reduction in aging Harwood & Naylor, 1971), some suggest no
decline of Gf that is accounted for by the Gs (Schaie, 1996, 2000). Declines, indicated in
measure. averages, show up in the late 60s and are
The slow-tracing measure requires at first small but accelerating, so that in the
focused concentration and so does the Gs 90s there is notable decline (Schaie, 1996,
measure. That is, although called a measure 2000).
of speed, Gs requires focused concentration If differences in years of formal education
(in order to make many of the speeded are statistically controlled, the increment of
responses). Thus, what seems to be involved Gc through to the 70s is increased (Horn,
in aging decline of Gf reasoning is a capacity 1968, 1972, 1989; Kaufman, 1990). The evi-
for focusing attention and maintaining dence does not make clear whether or not
focused attention. This is measured in slow the declines beyond this age are reduced by
tracing, but also required in identifying rela- control for education.
tionships and working out their implications The standard deviations around the Gc
in solving difficult, highly novel reasoning averages increase as age increases. This
problems of the kind that define Gf. This is means that though generally there is
probably the most important finding of the improvement with age in these abilities,
Horn et al. (1981) series of studies. the improvements for some individuals are
much larger than for others, and for some
people these abilities decline with age. The
Capabilities that Do not Decline during
particular reasons why this occurs have not
Adult Development
been well documented.
In many of the studies in which age-related Generally the findings suggest two pos-
declines were documented, no evidence of sibilities, both of which could be correct:
decline was found (in the same samples of (1) that whereas some people enter profes-
subjects) for reliable measures of Gc and sions and adopt styles of life that require or
TSR abilities abilities indicating breadth otherwise encourage development of crys-
of knowledge, consolidation in learning, and tallized abilities, other people do not, and
fluency of retrieval of information from the (2) that the tests used to assess abilities that
store of knowledge. are regarded as indicating Gc do not measure
Gc: Acculturation Knowledge. As noted well in the areas of knowledge in which some
previously, the abilities of this class are often people are invested. Both of these possibili-
referred to as indicating what is most impor- ties are relevant to the study of the develop-
tant about human intelligence. They are ment of expertise.
indicative of the extent to which an indi- TSR: Tertiary Storage/Retrieval. Two dif-
vidual has incorporated the most valued fea- ferent kinds of variables indicate this class
tures of the intelligence of a culture. of abilities. In one kind of variable sub-
On average, over a period from early jects are required to retrieve information
adulthood into the 60s, there is increase that was associated with other information
with age in these abilities (e.g., Botwinick, several minutes or hours or days prior to
1977, 1978; Cattell, 1971; Finkel et al., 2003 ; the time when retrieval is requested. Mem-
Harwood & Naylor, 1971; Horn, 1968, 1972, ory for Names in the WJ-R (Woodcock,
1982, 1989, 1991; 1997; Horn & Cattell, 1967; McGrew, & Mather, 2001) is a good example
596 the cambridge handbook of expertise and expert performance
of this kind of measure. In the second kind Yet for the TSR class of abilities, as for Gc,
of variable subjects are required to retrieve the evidence indicates either improvement
(or reconstruct), by association, information through age 60 years in adulthood, or if there
from stored knowledge. An example of such is decline, it occurs late and is small, at least
a measure is obtained with the ideas test. until the 80s (Horn, 1968; Horn & Cattell,
In this test subjects provide ideas similar to 1967; Horn & Noll, 1997; McArdle et al.,
a given idea. 2000; Schaie, 1996; Stankov & Horn, 1980;
Tests for both kinds of measurement may Woodcock, 1990).
be given under time limits, but time limits
must be generous, such that subjects have Problems and Limitations of Extended
ample time to provide nearly all the associa- Gf-Gc Theory
tions of which they are capable. Otherwise,
the test will measure Gs and indicate aging Extended Gf-Gc theory is based to a very
decline, rather than aging increase. considerable extent on results from stud-
The retrieval in both of the two kinds ies of age differences and age changes in
of measures of TSR relates to encoding and cognitive abilities. Indeed, it is largely an
consolidation in learning, and to parame- attempt to describe and explain cognitive
ters that characterize individual differences ability development through the adulthood
in initial encoding and consolidation. These years. Regarded in that way, it appears to be
parameters also characterize the retrieval at notably inadequate in several ways.
later times (Bower, 1972, 1975 ; Estes, 1974). For example, the ability declines des-
One such parameter is facility in forming cribed in the theory do not appear to charac-
associations. This is prominent in the first terize the adulthood development of which
kind of measure, where the assessment is we are most aware in ourselves, in those with
over fairly short periods of time although whom we most frequently interact, and in
longer periods than in the measures of SAR. those we see and hear and read about in our
It indicates initial consolidation. various media of communication. Indeed,
The second kind of measure indicates pri- the declines described in the theory present
marily long-retained consolidation of asso- a paradox: they suggest there are serious cog-
ciations consolidation that could have nitive deficits in people otherwise regarded
entered cognition at different times over as exemplars of high intellect in our society
ones history of learning and got stored in a the professors, scholars, researchers of our
system of categories (as described by Broad- major universities and research institutes,
bent, 1966). the CEOs of major businesses and others
Both kinds of variables indicate facility making the major decisions and inventions in
in retrieving the consolidated and stored these institutions, the writers and commen-
information, not the size and variety of tators and pundits of our media of commu-
the storage as such. This latter is indicated nication, the politicians, our representatives
by Gc. in congress, our presidents and their cabi-
TSR thus indicates facility in retrieving nets, and so forth. These people in the most
the knowledge that primarily defines Gc. responsible positions in almost all our major
This facility is correlated with Gc, but the institutions are usually adults well along on
two factors are independent (as previously the age-decline curves described in extended
described) in the sense that the correlation Gf-Gc theory.6
between the two factors is well below their How can it be that these people have
respective internal consistencies, and in the notably declined in reasoning and memory
sense that they have different patterns of cor- abilities that are regarded as indicating essen-
relations with other variables. This means tial features of human intelligence? If true,
that people with a very large store of knowl- how can it be that we dont see this in their
edge may not as readily access their stores as everyday behavior (or if we do, how can it be
people with smaller stores. that we continue to entrust them with the
a merging theory of expertise and intelligence 597
most responsible and difficult jobs requiring hold positions in our society that appear to
intelligence in our society)? There indeed require the highest levels of intellect.
seems to be a paradox here. What is wrong? There are some problems, however, with
Our everyday observations? Are these older the assumption that the measures of Gc and
people really demented? Or are the findings TSR abilities obtained in research really indi-
and the theory deceiving us? Is there some- cate the high levels of cognitive capabil-
thing wrong with the tests? Is there some- ity that are attained in adulthood. We can
thing wrong with the sampling of people see these problems when we look carefully
in the research studies? Are the people we at the measures of Gc that are used in
know and the people we hear about excep- research.
tions to what is going on with the rest of the First, these measures provide only a rather
populace? paltry sample of the abilities that, in theory,
We dont have really good answers to define Gc. This concept represents a much
these questions. There is probably some broader and diverse range of knowledge the
truth in each answer that comes readily to intelligence of the dominant culture than
mind. The older people in high-level posi- is (or probably can be representatively) sam-
tions probably have not lost the most impor- pled with the sets of tests used in research.
tant capabilities of human intelligence; there There is no battery of tests for which there is
is probably something wrong with the tests; compelling evidence that it representatively
the samples of the various studies generally samples the knowledge of the concept of Gc.
did not include the CEOs, professors, and The full set of the achievement tests of the
so forth, alluded to above, and thus the sam- WJ-R (Woodcock et al., 2001) is currently
ples are probably not truly representative of probably the most nearly representative bat-
the smarter segments of our society. But also tery of such tests. It takes three hours to
our friends and the folks at the highest levels administer, more time than can be allocated
in our institutions do sometimes appear to to measure one factor in almost any research.
be fairly obtuse, the tests of the research do Second, even tests such as those of the
appear to measure some important aspects WJ-R battery, and certainly tests of the kind
of intelligence, and samples of subjects used that have typically been used in the research
in the research do represent a segment of our on which the extended Gf-Gc theory is
society. Also, there is the well-replicated evi- based (e.g., vocabulary, similarities, general
dence indicating that normal aging is associ- information, esoteric-word analogies), mea-
ated with loss of neuronal basis for cognitive sure only the elementary knowledge, the
abilities (see Hill & Schneider, Chapter 3 7). beginning knowledge, in the various fields of
There probably is some decline in impor- human culture. The measures sample only
tant cognitive capabilities; however, much surface knowledge in a number of fields, not
of it may not be seen in everyday obser- in-depth knowledge in any field. The Gc fac-
vations. tor is thus more indicative of a dilettante
But there is reason to believe that in the grasp of the intelligence of the culture than
research on which extended Gf-Gc theory it is an indication of what a scholar under-
is based, we have not measured some of the stands. The latter is not only more similar
capabilities that best characterize the intel- to what we mean by intelligence when we
ligence of adult humans. We have seen that opine that middle-aged and elderly CEOs,
the Gc abilities of breadth of knowledge and scientists, writers, and so forth are highly
the TSR abilities of retrieval (or reconstruc- intelligent, it is also more characteristic of
tion) of knowledge on average increase and what is actually meant by the concept of Gc.
for some increase a great amount through Thus, the operational definition of Gc
a major part of adulthood. These measures in the extant research is not quite ade-
thus would seem to be indicative of that quate. As development of cognitive abili-
high level of intelligence we expect to find ties proceeds through childhood into adult-
and think we see in those older people who hood, people develop greater in-depth
598 the cambridge handbook of expertise and expert performance
comprehension, in-depth knowledge, in- tise showing what it is. The developmen-
depth understandings; in consequence of tal research has been directed at describing
this, they reason better and are better how expertise comes about showing how
problem-solvers. This is what characterizes it develops. In what follows we will give
what we refer to as intelligence in adults; first consideration to the structure of cog-
this is what some older adults have more of nitive expertise, then move on to consider
than younger adults. But this is not what is how expertise develops. Ours must be a
well measured in Gc and TSR. rather cursory view of these matters. Other
It is more nearly what is studied in chapters in this volume provide a more com-
research on expertise, including its rela- prehensive look (see many of the chapters in
tionship to aging (see also Krampe & Section III of the handbook).
Charness, Chapter 40). Let us turn to a con-
sideration of what is measured in research on Principal Attributes of Cognitive Expertise
expertise.
We will first identify abilities that consis-
tently characterize expertise and differenti-
ate experts from novices (see also Feltovich,
Theory of Expertise
Prietula, & Ericsson, Chapter 4. Then we
will consider the relations of these to the
The main argument at this point is that kinds of abilities that have been addressed
(1) the abilities that represent the qui- in the theory of intelligence.
ntessential expressions of human intellec- Expert Knowledge. Descriptions of expert
tual capability come to fruition in mature abilities begin with an account of the breadth
adulthood (not in childhood or even early and depth of the knowledge the expert
adulthood), and (2) the measures on which possesses in the domain of the exper-
current theories of intelligence are based do tise (Barnett & Koslowski, 2002; Charness,
not assess these abilities, or at least do not 1981a, 1981b, 1991; Colonia-Willner, 1998;
do so at the high levels that truly character- de Groot, 1978; Ericsson & Kintsch, 1995 ;
ize adult intelligence. Now we advance the Ericsson, 1996; Hershey, Walsh, Read, &
argument that research on expertise more Chulef, 1990; Patel & Arocha, 1999; Walsh
nearly indicates what these abilities are and & Hershey, 1993 ). The expert has greater in-
what the intellectual capacities of humans depth domain knowledge (see Chi, Chap-
become in adulthood. ter 10). As the level of expertise increases,
There are different kinds of expertise, this knowledge increases and the elements of
and some do not at all represent aspects of the knowledge become better organized and
human intelligence, just as there are many integrated. The resulting knowledge system
abilities that were not considered in build- provides the expert with a basis for select-
ing the evidence on which Gf-Gc theory ing, organizing, representing, manipulating,
is based. As we restricted consideration of and interpreting information in the envi-
abilities to cognitive abilities in discussing ronment. Some of the experts reasoning
the theory of human intelligence, so we is knowledge-based (Charness, 1981a, 1981b,
here restrict consideration of expertise to 1991; Colonia-Willner, 1998; Crook, 2002;
what we may call cognitive expertise in de Groot, 1978; Ericsson, 1996; Ericsson &
describing how expertise fits within a sci- Kintsch, 1995 ), and these processes are sup-
entific understanding of human intellectual ported by the expanded working memory
capabilities. that provides flexible access to generated
Just as we found it useful to classify the inferences and intermediate results (Char-
research on the nature of intelligence as ness & Bosman, 1990; Ericsson & Delaney,
structural and developmental, so it is useful 1998; Ericsson & Kintsch, 1995 ; Gobet &
to classify research on expertise as structural Simon, 1996; McGregor & Howes, 2002;
and developmental. The structural research Morrow, Menard, Stine-Morrow, Teller, &
has been directed at identifying and mea- Bryant 2001; Weber, 2003 ).
suring different aspects and levels of exper-
a merging theory of expertise and intelligence 599
In sum, the expert has great knowledge in There has been some, but very limited,
the area of expertise, and this provides a firm study of tests of deductive reasoning within
foundation for the acquisition of expanded the tradition associated with extended Gf-
working memory that supports the experts Gc theory. Perhaps the closest approxima-
reasoning about potential courses of action. tion to the kinds of tasks that have been
Expert Reasoning. The knowledge-based studied in expertise research is tests of
reasoning of expertise has been described reasoning, such as mathematical word-
by a number of researchers (e.g., Barnett problem tests or general reasoning tests.
& Koslowski, 2002; Crook, 2002; Charness, When the covariance of test scores of these
1981a, 1981b, 1991; Colonia-Willner, 1998; reasoning tests is analyzed, factor analyses
de Groot, 1978; Ericsson, 1996; Ericsson & show an involvement of both Gf and Gc
Kintsch, 1995 ; Hershey et al., 1990; Lighten (Carroll, 1993 ; Horn et al., 1981).
& Sternberg, 2002; Patel & Arocha, 1999; These findings do not reflect the high-
Proffitt, Coley, & Medin, 2000; Rikers et al., level reasoning associated with expert
2002; Walsh & Hershey, 1993 ). Expert rea- performance, because these tests require
soning is best described as inferential and limited knowledge of mathematics for suc-
deductive (Charness, 1981a, 1991; Ericsson, cessful performance, and the samples of par-
1996, 1997; Hershey et al., 1990),whereas ticipants are not chosen to represent high
Gf reasoning is best described as induc- levels of expertise. To identify high lev-
tive (Cattell, 1971). In conceptualizing a els of reasoning ability, it would be neces-
problem, the expert is able to comprehend sary to construct tests with complex, novel
its structure and represent the most rele- problems that would challenge expert per-
vant relations before generating alternative formers with extensive experience of the
courses of action, whereas novices generate a associated problems. These kinds of tests
large number of options that often lack high have not yet been constructed and stud-
relevance to the solution (Walsh & Hershey, ied by researchers who developed extended
1993 ). Gf-Gc theory.
For example, chess experts chose the next Expert Memory. Researchers studying ex-
move by first encoding the structure of the pertise have described a form of expanded
current chess position, drawing on their vast working memory (de Groot, 1946, 1978;
knowledge, and then evaluating alternative Ericsson & Kintsch, 1995 ; Ericsson
moves through planning consequences. In & Staszewski, 1989; Holding, 1985 ;
contrast, a person of low expertise, using Koltanowski, 1985 ; McGregor & Howes,
only Gf reasoning, selects the next move by 2002; Morrow et al., 2001; Weber, 2003 ; Fel-
generating and evaluating the various move tovich et al. Chapter 4), which differs from
possibilities that can be seen in the immedi- the short-term working memory (STWM)
ate situation (de Groot, 1978). Even when that is the essence of SAR and is central
experts encounter a problem they have not to extended Gf-Gc theory. This expanded
encountered before, their acquired represen- working memory is illustrated by the ability
tations allow them to see interesting moves of a chess expert in blindfold chess (see also
and evaluate outcomes of these moves by Gobet & Charness, Chapter 3 0).
planning and relying on their acquired chess In playing blindfold chess, the expert is
skills. The novice will be forced to generate never able to literally see the chess board
moves from scratch and will explore moves or the chess pieces. All the outcomes of
suggested by the immediately perceptible sequences of plays must be kept accessible
configuration of the chess pieces. Expert rea- in working memory. The number of alterna-
soning proceeds from the general compre- tive move sequences that the chess expert
hension of essential relations, knowledge of must consider mentally is of the order of
relevant principles to develop specific alter- 3 0 to 40, at least much more than the
native courses of action, whereas novice seven plus or minus two, that is generally
reasoning is stimulated by the salient attr- accepted to be the limit of the capacity
ibutes of a presented situation or problem. of short-term working memory (STWM).
600 the cambridge handbook of expertise and expert performance
Similar superior memory (see also Wilding in the span of immediate apprehension. Sim-
& Valentine, Chapter 3 1) has been docu- ilarly, Masanga and Horn (2001) found that
mented in studies of experts playing mul- experts in playing the game of GO held more
tiple games of chess simultaneously. Cooke, than seven units in the span of immediate
Atlas, Lane, and Berger (1993 ) and Gobet apprehension.
and Simon (1996) demonstrated that highly Experts are able to recall presented ele-
skilled chess players can not only recall infor- ments equally well in orders that are differ-
mation from up to nine chess positions, ent from the presented order. For example,
presented one after the other as rapidly as in a series of chess moves expert chess play-
one every five seconds without pause, but ers can recall the moves nearly as well from
can also back-through numerous sequences the last to the first move as from the order in
(more than seven) to previous positions. which the moves were made. In experiments
There are no such demonstrations of mem- demonstrating the features of STWM, on the
ory feats in the tests used to define the forms other hand, recall of elements in the order in
of memory that indicate SAR. which they were presented is considerably
Ericsson (1996), Ericsson and Kintsch better than recall in the reverse of this order
(1995 ), and Ericsson and Delaney (1998) rea- (e.g. Anderson, 1990). Whereas the limit for
soned from the evidence that the storage forward span STWM is seven plus or minus
in which the work of the working memory two, the limit for backward span is four plus
of experts is carried out must be a different or minus one.
form of storage than the storage of STWM. Experts in chess can play multiple games
Ericsson and Kintsch referred to the mem- of chess at one time, and do very well in all
ory of this different form of storage as long- the games, whereas if one is required to keep
term memory working memory (LTWM). track of the meaning of sentences while also
We note that it operates in the short-term remembering the last word of each sentence,
and refer to it as expertise working memory memory for the last word is likely to be
(ExpWM). very poor.
ExpWM differs in four important ways When experts are unexpectedly asked to
from the STWM that has been most exten- recall information about a complex task in
sively studied in cognitive psychology and in their domain of expertise, their memory is
the research on which extended Gf-Gc the- not only more accurate and more complete
ory is based: (1) the amount of information than that of less-skilled performers, it is also
that is stored is larger, (2) the information considerably more accurate than when one
is less affected by disruption and distrac- is unexpectedly required to recall material of
tion there is more effective resumption STWM, For example, if someone unexpect-
of performance after a disruption, and more edly asked you about a telephone number
is retained under multiple-task demand, you dialed a moment or two before, you are
(3 ) the order of recall is more flexible it not likely to retrieve it, whereas if an expert
can differ from the order of presentation, is unexpectedly asked about the last moves
and (4) information is encoded in long-term in a chess match, hes likely to remember the
memory and thus can be retrieved when moves with considerable accuracy.
recall is requested unexpectedly (Ericsson, In sum, ExpWM is notably different from
1998). The results and analyses of Ericsson STWM (and more generally, SAR).
and Kintsch (1995 ) and Masunaga and Horn
(2001) are consistent in indicating these
The Development of Expertise
differences between ExpWM (LTWM) and
STWM. High-level expertise in such domains as
The span of ExpWM is substantially mathematics, physics, medical diagnosis,
larger than the seven-plus-or-minus-two financial planning, music, sports, and the
span of STWM. As we have mentioned, games of GO and chess is attained through
chess experts hold as many as 40 elements a form of practice that a number of
a merging theory of expertise and intelligence 601
The evidence, although limited, is in line best predictor of performance on all music-
with this prediction. The pinnacle of cogni- related tasks was the amount of practice par-
tive expertise is rarely reached by people in ticipants had maintained during the previ-
their 20s, but more often it is in ones 3 0s ous ten years. Sixty-year-old expert pianists
or 40s or even later. The most outstanding who maintained deliberate practice (over
contributions to science, literature, and the ten hours/week) performed piano-related
arts are made by adults, sometimes youngish tasks at a superior level comparable to the
adults, to be sure, but more often middle- level attained by younger expert pianists
aged adults and even adults in their 5 0s, (Krampe & Ericsson, 1996).
60s, and 70s. The most advanced levels of Charness et al. (1996) and Krampe (2002)
expertise in chess, GO, and financial plan- found that chess-playing skill was most
ning have been attained by middle-aged strongly related to the recent level of delib-
adults (Charness & Bosman, 1990;, Krampe, erate practice, not age. Baltes (1997) found
& Mayr, 1996; Ericsson & Charness, 1994; that in domains of their specialization, older
Kasai, 1986; Walsh & Hershey, 1993 ). adults accessed information more rapidly
Charness (1981a, 1981b, 1991) found that than younger adults. Among architects of
age was correlated near zero with level of rat- different ages and levels of expertise, Salt-
ings of chess skill, and that measured capa- house, Babcock, Skovronek, Mitchell, &
bilities such as depth of search,8 efficiency Palmon (1990) found that at every age, high-
of chunking in memory, and rapidity in eval- level experts consistently scored above low-
uating an end-game position were positively level experts in visualization abilities, and
correlated with level of expertise but not elderly high-level experts scored higher than
with age. youthful-low-level experts.
Rabbitt (1993 ) found that in a sample Deliberate, well-structured practice is
of novices, crossword-puzzle-solving ability required to maintain high levels of expert
was positively correlated with test scores performance (Bahrick, 1984; Bahrick & Hall,
indicating Gf (r = 0.72) and negatively cor- 1991; Kramer & Willis, 2002; Krampe &
related with age (r = 0.25 ), but in a sam- Ericsson, 1996; Walsh & Hershey, 1993 ). If
ple of experts at different levels of expertise, expertise abilities are not used, they decline.
this ability correlated near zero with Gf and To the extent that regular quality practice is
positively with age (r = 0.24). maintained, expertise abilities do not decline
Krampe and Ericsson (1996) studied a with age in adulthood.
sample of classical pianists who ranged from Thus, generally the findings indicate that
amateurs to concert performers with an expertise abilities can be maintained as age
international reputation, and who ranged increases in adulthood that is, if they are
in age from the mid-20s to the mid-60s. used and if there is sustained deliberate
They obtained cognitive speed measures practice. Ericsson (2000) argues that aging
comparable to those that assess the Gs fac- decrease in expert performance is largely
tor of extended Gf-Gc theory, and they attributable to older individuals decisions
obtained cognitive speed measures pertain- to reduce the frequency of engagement
ing to music and piano playing that in challenging activities and decrease the
is, within the domain of expertise. Older intensity of maintained deliberate practice
pianists were notably slower than younger (p. 3 71).
counterparts in the Gs measures just as in
most of the research indicating aging decline
The Relation between Expertise Abilities
of vulnerable abilities but, independently
and Extended Gf-Gc Theory
of age, experts performed better than ama-
teurs on all music-related speeded tasks, that There is thus the suggestion that (1) exper-
is, speed was positively related to exper- tise abilities better indicate the humans
tise but the relationship to age was not sig- intellectual capacities than do the abilities
nificantly different from zero. The single thus far measured in the research on which
a merging theory of expertise and intelligence 603
the current theory of intelligence is based, Grandmasters in chess) are adults, usually
(2) these intellectual capacities are differ- middle aged, but some of them would be
ent from the capacities thus far measured classified as old.
and represented in extended Gf-Gc theory, The game is a competition between two
(3 ) these capacities come to greatest fruition players. One player deploys black stones,
in adulthood, and (4) these capacities there- the other white stones. In turn each player
fore do not decline in adulthood at least places a stone at the intersection of grid
not in the vital first one-half of the adult- lines on a board marked with 19 by 19 such
hood period of development. lines. The object of the game is to place
These arguments are, in effect, the prin- stones in such a way that ones stones sur-
cipal hypotheses evaluated in a recent study round the largest portion of the territory
by Masunaga and Horn (2001). In this on the board, and thus surround the oppo-
study, indicator measures of the ability traits nents stones, but at the same time ensure
of expertise deductive reasoning (ExpDR), that ones own territory and stones are not
expertise working memory (ExpWM), and surrounded by the opponents stones. The
expertise cognitive speed (ExpCS) were rules of the game are simple, but expertise
constructed. The ExpDR indicators were in GO is very complex at least as com-
designed to indicate a form of reason- plex as chess. There are millions of combi-
ing that characterizes adult intelligence. nations of possible stone placements and a
The ExpWM indicators were designed to huge number of possible strategies and con-
be the same as the memory tests that define tingencies to take into account in placing the
STWM, and the ExpCS tests were like those stones. This involves complex and difficult
used to measure Gs, except in each case reasoning, memory, and other indicators of
the elements of the tests were those dealt human intelligence. It is extremely difficult
with in a domain of expertise. The anal- to become expert in the game. It is esti-
yses of the study were directed at deter- mated by those who have become experts
mining, first, whether such variables indi- that it takes at least ten years of concerted
cate factors of intelligence that are distinct effort to reach the apex in GO (Kasai, 1986;
from the age-declining Gf, SAR, and Gs fac- Reitman, 1976).
tors of extended Gf-Gc theory and, second, A factor analysis of the measures of GO-
whether the patterns of relationship to age expertise abilities and cognitive-ability tests
for putative expertise abilities were the same indicated that ExpDR and ExpWM abili-
as, or different from, the patterns seen for ties were distinct from Gf, SAR, and Gs.
the Gf, SAR, and Gs indicators of human The results did not, however, distinguish
intelligence. an ExpCS that differs from Gs. Experts
This research studied expertise in playing were faster than novices in dealing with
the game of GO, utilizing Japanese players. tasks involving stimuli from their domain
In Japan9 this game is very widely played of expertise, but the individual differences
and practiced and thus provides a kind of in the measures that involved these stim-
participant-sampling laboratory for study of uli were largely accounted for by individual
this kind of expertise. Many people of all differences in Gs.
ages strive to become experts in GO, and A further analysis of the separate ExpDR
there are notable differences in the ages at and ExpWM factors examined the effects
which people start, and in the time span of age on these expertise abilities and com-
over which they strive, to gain expertise pared them with aging effects for the Gf and
(somewhat similar to golf in the United SAR indicators of human intelligence. As
States). Although many people learn the outlined previously, the principal hypothesis
rudiments of the game in adolescence, most predicted that the expertise reasoning and
of the development of expertise occurs in working memory would not decline with
adulthood, often extending into old age. All age in adult experts when deliberate, well-
the reigning GO professionals (analogous to structured practice was maintained during
604 the cambridge handbook of expertise and expert performance
adulthood. At lower levels of expertise the particular abilities the human is capable
tests of ExpDR and ExpWM would mea- of developing fall into a relatively small
sure Gf, SAR, or Gs, and thus performance number of different classes. These classes
on these tests would show a similar age represent different ways in which neural
decline as Gf, SAR, or Gs (as reviewed in capacities and learning interact over extend-
the first section of this paper). These find- ed periods of development in both child-
ings would yield a level-of-expertise-by-age hood and adulthood to produce cogni-
interaction for ExpDR and ExpWM with tive ability traits (see also Hill & Schnei-
age-related declines with age at lower levels der, Chapter 3 7). The neural capacities are
of expertise and no declines at high levels laid down genetically, but are altered by
of expertise. environmental influences that directly affect
An ANOVA with expertise and age physiological/neural structure and function.
(13 age groups ranging from an average of The environmental/learning influences are
above 25 to an average of 75 years of age) large in number and diverse, but accul-
revealed a reliable age-by-expertise inter- turation learning influences are particu-
action for both ExpDR and ExpWM. The larly important in shaping cognitive ability
comparable interaction effects for the indi- traits.
cators of Gf, SAR, and Gs were of trivial Four classes of abilities emerge from the
magnitude and not significantly different interactions of neural dispositions and learn-
from zero, thus signaling that the interac- ing influences that operate over the course
tion effect was for expertise abilities of intel- of development Gc, TSR, ExpDR and
ligence, but not for conventional abilities of ExpWM. These abilities reflect individual
extended Gf-Gc theory. These findings were differences in acculturation and learning in
consistent with the hypotheses described different domains in which one can become
above and, in fact, the means for older per- expert. The Gc class of crystallized knowl-
sons at higher levels of expertise were gener- edge abilities indicate dilettante breadth of
ally larger than the means for younger people knowledge of the culture. The TSR abili-
at lower levels of expertise. ties indicate fluency in tertiary retrieval of
The expertise abilities-by-age interac- information from this knowledge store. The
tions were found to be positively related ExpDR abilities indicate expert deductive
to skill rating in playing GO. These results reasoning in particular domains in which
are not notably altered by entering as advanced expert knowledge is built up.
covariate controls the abilities of extended The ExpWM abilities indicate wide-span
Gf-Gc theory, although Gf and STWM were working-memory abilities in the domains in
found to have small negative (suppressor) which one becomes expert. To the extent
relations. Gf was found to be slightly neg- that one works at developing and maintain-
atively related to skill rating in GO, but the ing these abilities, they increase throughout
Gf-by-age interaction was positive, suggest- the lifespan.
ing that those who best retain their Gf with Classes of reasoning abilities (Gf), short-
advancing age perform at a relatively high term apprehension and retention abilities
level, whereas those who show declines in (SAR), and Gs abilities of cognitive process-
Gf perform at a lower level. ing speed are relatively little shaped by influ-
ences focused on acculturation or learning
of particular forms of expertise. The rea-
Summary and Conclusions soning abilities are similar to the abilities
Spearman described as representing appre-
Abilities regarded as indicating important hension, eduction of relations, and eduction
features of human intelligence are numer- of correlates. He argued that these capac-
ous and diverse, but findings from struc- ities are the sine qua non of all the abili-
tural, developmental, physiological, edu- ties of human intelligence. It has not been
cational/occupational and behavior-genetic found, however, that these capacities, as
studies indicate (empirically) that the many measured, represent anything that is general
a merging theory of expertise and intelligence 605
to all the abilities that are accepted as indi- tual capabilities of the human. Reasoning
cating important features of human intel- at the highest levels of which humans seem
ligence. Measured Gf does not represent capable involves use of extensive, integrated
the concept of general intelligence that bodies of knowledge. Gc does not represent
Spearman described under the heading of this knowledge. It indicates only dilettante
g: it represents a more limited form of breadth of knowledge. Extensive, integrated,
intelligence one among other forms of deep knowledge is found in particular kinds
intelligence. of expertise. Gq, representing the abilities
In the earliest period of development, of mathematical/quantitative thinking, is
neophobic and neophilic reaction patterns indicative of such knowledge in one domain
promote cognitive ability development. In of expertise.
conjunction with the extent to which secure The TSR abilities indicate fluency in
attachment is established in infancy and accessing (recreating) information such as
early childhood, neophobia and neophilia that in the Gc store of knowledge. TSR
activate primitive attention-maintaining and abilities also come to fruition in adulthood
reasoning capacities to produce the aware- and thus help characterize adult intelligence.
ness concepts and problem-solving capabil- These are not, however, the abilities for flex-
ities that constitute early intelligence. Neo- ibly maintaining large amounts of informa-
phobia promotes the development of abili- tion in the span of immediate awareness that
ties that influence one to avoid unfamiliar, most characterize intelligence in adulthood.
often dangerous conditions. Secure attach- These abilities the ExpWM abilities are
ment in the earliest years of life enables indicated in displays of expertise.
the child to approach unfamiliar conditions, Also lacking in the Gc and TSR descrip-
which in turn enables neophilia and pro- tions are the reasoning abilities that are most
motes abilities that encourage exploration characteristic of adult intelligence. These
and manipulation of unfamiliar things. Indi- abilities are not well represented by the Gf
vidual differences in secure attachment are abilities, which decline in adulthood. Mea-
associated with many of the factors that pro- sures of Gf abilities are designed to minimize
mote acculturation. These influences, oper- the influences of acculturation and learning
ating throughout development, produce the in any particular domain in which one might
distinction between the Gc and TSR abili- become expert. Yet reasoning at the highest
ties, on the one hand, and the Gf, SAR, and levels involves using best selections of infor-
Gs abilities, on the other hand. mation and problem-solving skills from bod-
The Gf, SAR, and Gs abilities increase ies of knowledge that are difficult to acquire
with age in childhood and adolescence but and require considerable time to acquire
decline in adulthood. One important aspect that is, the bodies of knowledge of domains
of this decline is loss of ability to maintain of expertise. Deductive reasoning with this
focused attention. Measured in tasks that kind of information ExpDR is character-
require one to behave as slowly as possi- istic of the intelligence of adults.10
ble, this ability accounts for a substantial Expertise deductive reasoning (ExpDR)
part of the decline in Gs speed of processing and expertise working memory (ExpWM)
information, SAR retention, and the com- capabilities are distinguishable in adulthood,
plex reasoning of Gf. are indicative of human intelligence, and
The Gc and TSR abilities increase with are different from somewhat comparable
acculturation, and this accumulates through Gf and SAR abilities that heretofore in
childhood and adulthood. The Gc and TSR extended Gf-Gc theory were regarded as
abilities thus indicate aspects of intelligence capturing the essence of the reasoning and
that reach their peaks in adulthood and in memory capabilities of human intelligence.
that sense characterize intelligence of adults, Whereas Gf and SAR decline with age in
but neither well represents the feats of rea- adulthood at all levels of expertise, ExpDR
soning and memory that appear in adult- and ExpWM increase with level of expertise,
hood and are most indicative of intellec- and to the extent that there is deliberate,
606 the cambridge handbook of expertise and expert performance
well-structured practice to develop and three factors were required to fit the intercor-
maintain expertise, these abilities increase relation data at the second order.
with age in adulthood. These abilities exem- 5. Root mean square error of approximation
plify more nearly than the others the full RMSEA, Browne & Cudeck, 1993 ; Steiger,
capacity of human intelligence. 1990.
6. Expressed in IQ units the decline of fluid
reasoning from age 25 to age 65 years is no
Footnotes less than 15 points. Thus the average IQ in
this form of intelligence descends from 100 to
below 85 . Granted that CEOs, Senators, Pro-
1. There is an assumption here that different fessors, etc. would have started at higher levels
items of reasoning are the same. One can con- than other folks say in the 13 0s in IQ units
ceive of specifying behavioral traits without still the evidence of decline suggests that they
this assumption, but in practice it is funda- would be back to fairly ordinary levels of intel-
mental to the definition of all behavioral traits lect by the time they reached the exalted lev-
of personality. This is as true of the trait con- els at which they are found in the real world
structs of expertise theory as it is of the trait of work and responsibility.
constructs of theory of intelligence. The char-
7. While it seems that generally feedback is best
acteristic behaviors that indicate one grand
supplied by a good teacher/coach, Charness
master in chess, for example, are not pre-
(1981a, b; 1991) and Ericsson (1996) have found
cisely the same as the characteristic behav-
instances in developing expertise in chess in
iors that indicate another grand master at the
which self-directed practice, using books and
same level.
studying sequences of moves made by expert
2. Among the tests designed to measure abili- players, could be as effective as teacher/coach-
ties of human intelligence are tests of reason- directed practice.
ing of various forms induction, deduction,
8. Described as the point at which a player can no
conjunctive, disjunctive, eduction of relations,
longer retain accurate information about pro-
eduction of correlates , tests of many forms
jected changes in position.
of problem-solving, tests of abstracting, con-
cept formation, concept attainment, learning, 9. It was comprised of 263 male GO players
knowledge, comprehension, decoding, encod- between 18 to 78 years of age, at 48 levels
ing, communication, creativity, insight, sen- of expertise, drawn with the aid and sponsor-
sitivity to problems, originality, associational ship of the Japanese GO Association, which
fluency, expressional fluency, word naming, conducts examinations that enable individu-
figural fluency, flexibility, associative memory, als to establish (officially) their level of GO
free recall, nonsense memory, visual mem- expertise (which members of the Association
ory, auditory memory, visualization, perceiv- do about once a year on average).
ing spatial relations, visual closure, visual 10. Human intelligence also involves broad classes
integration, spatial scanning, sound localiza- of abilities that are tied to particular organi-
tion, loudness discrimination, pitch discrimi- zations of visual processes (Gv) and auditory
nation, resistance to auditory distraction, judg- processes (Ga). These were not considered in
ing rhythm, temporal integration, perceptual any detail in this chapter.
(visual, auditory, tactile) speed, reaction time
speed, choice reaction time speed, semantic
processing speed, and information processing
References
speed.
3 . Which is often referred to as Spearmans g,
in recognition that Spearman was first to pro- Anderson, J. R. (1990). Cognitive psychology and
pose a testable theory of general intelligence, its implications (3 rd ed.). New York: W. H. Free-
which, however, he designated with the letter man.
g in order to escape from a number of undesir- Backman, L., Small, B. J., Wahlin, A., & Larsson,
able and not-needed connotations of the word M. (2000). Cognitive functioning in very old
as it is used in everyday parlance. age. In F. I. M. Craik & T. A. Salthouse (Eds.),
4. Tests omitted in this selection can be seen to The handbook of aging and cognition (2nd ed,
be indicative of Gc and Gv. Rimoldi found that pp. 4995 5 8). Mahwah, NJ: Erlbaum.
a merging theory of expertise and intelligence 607
Bahrick, H. P. (1984). Semantic memory con- Carroll, J. B. (1993 ). Human cognitive abilities:
tent in permastore: Fifty years of memory for A survey of factor-analytic studies. New York:
Spanish learned in school. Journal of Experimen- Cambridge University Press.
tal Psychology: General, 113 , 129. Cattell, R. B. (1971). Abilities: Their structure,
Bahrick, H. P., & Hall, L. K. (1991). Life- growth and action. Boston: Houghton-Mifflin.
time maintenance of high school mathematics Cattell, R. B. (1979). Are culture-fair intelligence
content. Journal of Experimental Psychology: tests possible and necessary? Journal of Research
General, 12 0, 203 3 . and Development in Education, 12 , 113 .
Baltes, P. B. (1997). On the incomplete archi- Cavanaugh, J. C. (1997). Adult development and
tecture of human ontogeny: Selection, opti- aging, 3 rd ed. New York: ITP.
mization, and compensation as foundation of Charness, N. (1981a). Search in chess: Age
developmental theory. American Psychologist, and skill differences. Journal of Experimental
5 2 , 3 663 80. Psychology: Human Perception and Performance,
Barnett, S. M., & Koslowski, B. (2002). Adaptive 7(2), 467476.
expertise: Effects of type of experience and the Charness, N. (1981b). Visual short-term memory
level of theoretical understanding it generates. and aging in chess players. Journal of Gerontol-
Thinking & Reasoning, 8(4), 23 7267. Journal ogy, 3 6(5 ), 615 619.
of Experimental Psychology: General, 12 0, 20
Charness, N. (1991). Expertise in chess: The bal-
33.
ance between knowledge and search. In K. A.
Birren, J. E. (1974). Translations in gerontology Ericsson & J. Smith (Eds.), Toward a gen-
From lab to life: Psychology and speed of eral theory of expertise (pp. 3 963 ). New York:
response. American psychologist, 2 9, 808815 . Cambridge University Press.
Botwinick, J. (1977). Aging and intelligence. In Charness, N., & Bosman, E. A. (1990). Exper-
J. E. Birren & K. W. Schaie (Eds.), Handbook tise and aging: Life in the lab. In T. M. Hess
of the psychology of aging (pp. 5 80605 ). New (Ed.), Aging and cognition: Knowledge organiza-
York: Van Nostrand Reinhold. tion and utilization (pp. 3 43 3 86). New York:
Botwinick, J. (1978). Aging and behavior: A com- Elsevier.
prehensive integration of research findings. New Charness, N., Krampe, R., & Mayr, U. (1996). The
York: Springer. role of practice and coaching in entrepreneurial
Bower, G. H. (1972). Mental imagery and associa- skill domains: An international comparison of
tive learning. In L. W. Gregg (Ed.), Cognition in life-span chess skill acquisition. In K. A. Eric-
learning and memory (pp. 213 228). New York: sson (Ed.), The road to excellence (pp. 5 180).
Wiley. Mahwah, NJ: Lawrence Erlbaum Associates.
Bower, G. H. (1975 ). Cognitive psychology: An Christensen, H. (2001). What cognitive changes
introduction. In W. K. Estes (Ed.), Handbook can be expected with normal ageing? Aus-
of learning and cognitive processes, vol. 1. New tralian & New Zealand Journal of Psychiatry,
York: Lawrence Erlbaum Associates. 3 5 (6), 768775 .
Bowlby, J. (195 1). Maternal care and mental Colonia-Willner, R. (1998). Practical intelligence
health. Bulletin of the World Health Organiza- at work: Relationship between aging and cogni-
tion, 3 , 3 5 5 5 3 3 . tive efficiency among managers in a bank envi-
Broadbent, D. E. (1966). The well-ordered mind. ronment. Psychology & Aging, 13 (1), 45 5 7.
American Educational Research Journal, 3 , 281 Cooke, N. J., Atlas, R. S., Lane, D. M., & Berger,
295 . R. C. (1993 ). Role of high-level knowledge in
Browne, M. W., & Cudeck, R. (1993 ). Alternative memory for chess positions. American Journal
way of assessing model fit. In K. A. Bollen & J. S. of Psychology, 106, 3 213 5 1.
Long (Eds.), Tesing structural equation modles Craik, F. I. M. (1977). Age differences in human
(pp. 13 6162 ). Thousand Oaks, CA: Sage. memory. In J. E. Birren & K. W. Schaie (Eds.),
Burgess, K. M., Marshall, P. J., Rubin, K. H., & Handbook of the psychology of aging. New York:
Fox, N. A. (2003 ). Infant attachment and tem- Van Nostrand Reinhold.
perament as predictors of subsequent external- Craik, F. I. M. (2000). Age-related changes in
izing problems and cardiac physiology. Jour- human memory. In D. C. Park & N. Schwarz
nal of Child Psychology & Psychiatry, 44, (Eds.), Cognitive aging: A primer (pp. 75 92).
81983 1. Philadelphia: Taylor & Francis.
608 the cambridge handbook of expertise and expert performance
Craik, F. I. M. & Byrd, M. (1982). Aging and cog- ory and thinking. Current issues in think-
nitive deficits: The role of attentional resources. ing & reasoning (pp. 93 114). Hove, England
In F. I. M. Craik & Trehub, S. (Eds.), Aging UK: Psychology Press/Erlbaum (UK) Taylor &
and cognitive processes (pp. 191211). New York: Francis.
Plenum. Ericsson, K. A. & Kintsch, W. (1995 ). Long-term
Craik, F. I. M. & Trehub, S., Eds. (1982). Aging working memory. Psychological Review, 105 ,
and cognitive processes. New York: Plenum. 211245 .
Crook, J. A. (2002). How do expert mental health Ericsson, K. A., Krampe, R. T., & Tesch-Romer,
nurses make on-the-spot clinical decisions? A C. (1993 ). The role of deliberate practice in the
review of the literature. Journal of Psychiatric acquisition of expert performance. Psychologi-
& Mental Health Nursing, 8(1), 16. cal Review, 100(3 ), 3 63 406.
Deakin, J. M., & Cobley, S. (2003 ). A search for Ericsson, K. A., & Lehmann, A. C. (1996).
deliberate practice: an examination of the prac- Expert and exceptional performance: Evidence
tice environments in figure skating and volley- of maximal adaptation to task constraints.
ball. In J. Starkes and K. A. Ericsson (Eds.), Annual Review of Psychology, 47, 273 3 05 .
Expert performance in sport: Recent advances in Ericsson, K. A., & Staszewski, J. (1989). Skilled
research on sport expertise (pp. 115 13 5 ). Cham- memory and expertise: Mechanisms of excep-
paign, IL: Human Kinetics. tional performance. In D. Klahr & K. Kotovsky
de Groot, A. D. (1946). Het denken vun (Eds.), Complex information processing
den schaker [Thought and choice in chess]. (p. 23 5 268). Hillsdale, NJ: Lawrence
Amsterdam: North-Holland. Erlbaum Associates.
de Groot, A. D. (1978). Thought and choice in Estes, W. K. (1974). Learning theory and intelli-
chess. Oxford, England: Mouton. gence. American Psychologist, 2 9, 740749.
Ericsson, K. A. (1996). The acquisition of expert Fagan, J. F., & McGrath, S. K. (1981). Infant recog-
performance. In K. A. Ericsson (Ed.), The road nition memory and later intelligence. Intelli-
to excellence (pp. 15 0). Mahwah, NJ: Lawrence gence, 5 , 12113 0.
Erlbaum Associates. Finkel, D., Reynold, C. A., McArdle, J. J., Gatz,
Ericsson, K. A. (1997). Deliberate practice and M., & Pedersen, N. L. (2003 ). Latent growth
the acquisition of expert performance: An curve analyses of accelerating decline in cogni-
overview. In H. Jorgensen & A. C. Lehmann tive abilities in late adulthood. Developmental
(Eds.), Does practice make perfect?: Current the- Psychology, 3 9(3 ), 5 3 5 5 5 0.
ory and research on instrumental music practice Fisk, J. E. & Sharp, C. (2002). Syllogistic rea-
(pp. 95 1). NMH-publikasjoner. soning and cognitive ageing. Quarterly Journal
Ericsson, K. A. (1998). The scientific study of ofExperimental Psychology: Human Experimen-
expert levels of performance: General implica- tal Psychology, 5 5 A (4), 1273 1293 .
tions for optimal learning and creativity. High Flangan, D. P., & Harrison P. L. (Eds). (2005 ).
Ability Studies, 9, 75 100. Contemporary intellectual Assessment: Theo-
Ericsson, K. A. (2000). How experts attain and ries, tests, and issues(2nd ed). New York:
maintain superior performance: Implications Guilford Press.
for the enhancement of skilled performance in Gathercole, S. E. (1994). The nature and uses of
older individuals. Journal of Aging and Physical working memory. In P. Morris & M. Gruneberg
Activity, 8, 3 463 5 2. (Eds.), Theoretical aspects of memory (pp. 5 0
Ericsson, K. A. (2002). Attaining excellence 78). London: Routledge.
through deliberate practice: Insights from the Gobet, F., & Simon, H. A. (1996). Templates in
study of expert performance. In M. Ferrari chess memory: A mechanism for recalling sev-
(Ed), The pursuit of excellence through educa- eral boards. Cognitive Psychology, 3 1, 140.
tion. The educational psychology series (pp. 21 Harlow, H. F., & Suomi, S. J. (1970). Nature
5 5 ). of love: Simplified. American Psychologist, 2 5 ,
Ericsson, K. A. & Charness, N. (1994). Expert per- 161168.
formance. American Psychologist, 49, 725 747. Harwood, E., & Naylor, G. F. K. (1971). Changes
Ericsson, K. A. & Delaney, P. F. (1998). Work- in the constitution of the WAIS intelligence
ing memory and expert performance. In R. H. pattern with advancing age. Australian Journal
Logie & K. J. Gilhooly (Eds.), Working mem- of Psychology, 2 3 , 2973 03 .
a merging theory of expertise and intelligence 609
Hershey, D. A., Walsh, D. A., Read, S. J., Horn, J. L. (in press). Understanding Human
& Chulef, A. S. (1990). The effects of Intelligence: Where have we come since
expertise on financial problem solving: Evi- Spearman? In R. C. MacCallum & R. Cudeck
dence for goal-directed problem solving scripts. (Eds.), One hundred years after Spearman.
Organizational Behavior and Human Decision Mahwah, NJ: Erlbaum.
Performance, 46, 77101. Horn, J. L. & Cattell, R. B. (1967). Age differences
Holding, D. H. (1985 ). The psychology of chess in fluid and crystallized intelligence. Acta Psy-
skill. Hillsdale, NJ: Lawrence Erlbaum Asso- chologica, 2 6, 107129.
ciates. Horn, J. L. & Donaldson, G. (1980). Cognitive
Horn, J. L. (1968). Organization of abilities and development in adulthood. In O. G. Brim & J.
the development of intelligence. Psychological Kagan (Eds.), Constancy and change in human
Review, 75 , 24225 9. development (pp. 445 5 29). Cambridge, MA:
Horn, J. L. (1972). The structure of intellect: Pri- Harvard University Press.
mary abilities. In R. M. Dreger (Ed.), Multi- Horn, J. L., Donaldson, G., & Engstrom, R. (1981).
variate personality research (pp. 45 15 11). Baton Apprehension, memory and fluid intelligence
Rouge: Claitor Publishing. decline in adulthood. Research on Aging, 3 , 3 3
Horn, J. L. (1975 ). Psychometric studies of aging 84.
and intelligence. In S. Gershon & A. Raskin Horn, J. L. & Hofer, S. M. (1992). Major abili-
(Eds.), Aging, Vol. 2 , Genesis and treatment of ties and development in the adult period. In
psychologic disorders in the elderly (pp. 1943 ). R. J. Sternberg & C. A. Berg (Eds.), Intel-
New York: Raven. lectual development (pp. 4499). New York:
Horn, J. L. (1977). Human abilities: A review of Cambridge University Press.
research and theory in the early 1970s. Annual Horn, J. L. & Noll, J. (1997). Human cogni-
Review of Psychology, 2 7, 43 7485 . tive capabilities: Gf-Gc theory. In Flanagan,
Genshaft, & Harrison (Eds.), Contemporary
Horn, J. L. (1982). The aging of human abilities.
intellectual assessment (pp. 5 3 91). New York:
In B B. Wolman (Ed.), Handbook of developmen-
Guilford Press.
tal psychology (pp. 847870). Englewood Cliffs,
NJ: Prentice Hall. Jensen, A. R. (1998). The g Factor: The Science of
Mental Ability. London: Praeger.
Horn, J. L. (1985 ). Intellectual ability concepts. In
R. J. Sternberg (Ed.), Advances in the psychology Kasai, K. (1986). Tgo de atama ga yoku nara
of human intelligence, vol. 3 (pp. 3 5 77). Hills- hon [Becoming smart with Go]. Tokyo, Japan:
dale, NJ: Lawrence Erlbaum. Shikai.
Horn, J. L. (1989). Cognitive diversity: A frame- Kaufman, A. S. (1990). Assessing adolescent and
work for learning. In P. L. Ackerman, R. J. adult intelligence. Boston: Allyn and Bacon.
Sternberg, & R. Glaser (Eds.), Learning and Kausler, D. H. (1990). Experimental psychol-
individual differences: Advances in theory and ogy, cognition, and human aging. New York:
research (pp. 61114). New York: Freeman. Springer.
Horn, J. L. (1991). Measurement of intellec- Koltanowski, G. (1985 ). In the dark. Coraopolis,
tual capabilities: A review of theory. In K. S. PA: Chess Enterprises.
McGrew, J. K. Werder, & R. W. Wood- Kramer, A. F., & Willis, S. L. (2002). Enhanc-
cock (Eds.), Woodcock-Johnson technical man- ing the cognitive vitality of older adults. Cur-
ual (pp. 197246). Allen, TX: DLM. rent Directions in Psychological Science, 11, 173
Horn, J. L. (1997). A basis for research on age 177.
differences in cognitive capabilities. In J. J. Krampe, R. T. (2002). Aging, expertise and fine
McArdle, & R. Woodcock (Eds.), Human cog- motor development. Neuroscience & Biobehav-
nitive abilities in theory and practice. Chicago, ioral Review, 2 6, 769776.
IL: Riverside. Krampe, R. T., & Ericsson, K. A. (1996). Main-
Horn, J. L. (2002). Selections of evidence, mis- taining excellence: Deliberate practice and
leading assumptions, and oversimplifications: elite performance in young and older pianists.
the political message of The Bell Curve. In J. Journal of Experimental Psychology: General,
Fish (Ed.). Race and Intelligence: Separating Sci- 12 5 (4), 3 3 13 5 9.
ence from Myth (pp. 2973 25 ). Mahwah, NJ: Lane, C. J., & Zelinski, E. M. (2003 ). Longitudi-
Lawrence Erlbaum Associates. nal hierarchical linear models of the Memory
610 the cambridge handbook of expertise and expert performance
The drive to excel has long challenged is whether the development of expertise
humans to push their bodies, minds, and is largely attributable to unusual charac-
technologies in the determined pursuit of teristics of individuals, often thought of in
success. People have demonstrated their terms of largely inherited talents, or of their
devotion to excellence through the years learning histories (see Horn & Masunaga,
of effort and practice they have been will- Chapter 3 4). Because expertise is acquired
ing to invest in accomplishing their goals over many years, perhaps as a result of
(Ericsson, Krampe, & Tesch-Romer, 1993 ). more intense, sustained, and programmatic
For example, Simon and Chase (1973 ) application of the identical mechanisms that
observed that no one had ever attained underlie average levels of attainment (Erics-
the rank of Grandmaster in chess without son et al., 1993 ), the study of expert perfor-
at least a decade of intense preparation. mance is directly relevant to the study of typ-
This observation has since been extended ical human development. Indeed, the study
to many domains, including music, sports, of the development of expertise may rep-
and academia (Bloom, 1985 ; Ericsson et al., resent an opportunity to observe develop-
1993 ; Ericsson, Chapter 3 8). Despite folk mental mechanisms operating at maximum
tales about extraordinary performances by efficiency.
very young individuals, it is clear that the On the flip side, the study of typi-
most eminent individuals in any field do not cal human developmental mechanisms may
exhibit expert levels of performance prior to shed some light on the nature and devel-
an extended period of preparation. opment of expertise. Our research has been
Exploration of the nature of expertise devoted, in part, to creating a better under-
and how it develops has interested schol- standing of the cognitive mechanisms that
ars, professionals, and laypeople alike, and allow people to develop and use the prac-
has involved a wide range of theoretical tical intellectual abilities required for suc-
and methodological approaches. One of the cess in everyday life. As such, our work rep-
enduring debates over many years of study resents a link between the study of typical
613
614 the cambridge handbook of expertise and expert performance
human development and the exploration characteristics (Sternberg, 1994; see also
of expertise. Of particular interest to us is Feltovich, Prietula, & Ericsson, Chapter 4).
the way in which people use their intellec- Among the characteristics explored are
tual capabilities to adapt to and succeed in general information-processing capabil-
particular environments. Expertise, whether ities, such as strategizing and problem
demonstrated in such everyday feats as read- solving, the nature, quantity, and organi-
ing and writing, or in the exceptional accom- zation of expert knowledge (e.g., Chase &
plishments of artists, athletes, and schol- Simon, 1973 ; Larkin, McDermott, Simon,
ars, reflects the outcome of peoples active & Simon, 1980), and, more recently,
engagement in the world around them. cognitive complexity (Day & Lance,
In this chapter, we discuss a psychologi- 2005 ).
cal approach to exploring expertise that is However, definitions of expertise based
based on the theory of practical intelligence on individual differences preclude the recog-
and tacit knowledge (Sternberg, 1988; 1997; nition of expertise demonstrated by the
Wagner & Sternberg, 1985 ). This approach majority of individuals when they succeed
represents an attempt to explain the cogni- in their everyday lives. Another perspec-
tive mechanisms underlying the human abil- tive on expertise, the triarchic perspective
ity to adapt to, select, and shape environ- (Sternberg, 1988, 1997), takes a broader,
ments in the pursuit of personally valued developmental approach, and posits that
goals and success in life. We briefly outline success in life is determined by ones
our conceptualization of expertise; then we analytical, creative, and practical abilities,
describe the fundamental role that practi- which improve with practice and experi-
cal intelligence and tacit knowledge are the- ence. The value of the triarchic perspec-
orized to play in expert performance. We tive to the broader, psychological study of
then describe the scientific exploration of expertise is that it provides an opportu-
practical intelligence and tacit knowledge nity to rethink fundamental issues includ-
their nature, measurement, and relation to ing what is an appropriate definition of
expertise. Finally, we discuss a variety of expertise (see also Evetts, Mieg, & Felt,
approaches used to enhance practical intelli- Chapter 7).
gence and tacit knowledge and suggest some A prototype view of expertise represents
directions for future research. an integration of all these perspectives, and
others as well (Sternberg, 1994; Chapter 7
of this volume). Specifically, the proto-
What Is Expertise? A Prototype View type view of expertise maintains that exper-
tise is relatively domain specific and that
The most common psychological approach the attributes of experts may be specific
to defining expertise is to rely on an to a time and place, Chapter 4. This view
empirical definition based on individual maintains the importance of domain-general
differences. That is, an expert is defined information-processing capabilities, such as
as someone whose level of performance problem solving, while recognizing that
exceeds that of most others. For example, expertise and its requisite knowledge, skills,
Ericsson and Smith (1993 ) observed that and abilities are defined quite differently,
some individuals stand apart from the major- depending on the environment in which
ity. These scholars view the central goal of people develop and express their expertise.
the study of expertise as finding out what Importantly, the prototype view of exper-
distinguishes these outstanding individuals tise recognizes the diversity of skills that
from the less outstanding in some domain can lead to successful performance, and that
of activity. Approximately thirty years of expertise can be thought to exist in degrees
research, explicitly devoted to studying the rather than in an all-or-none fashion.
cognitive basis of expertise, has produced Conceptualizing expertise as a domain-
a rich and varied source of perspectives specific prototype a confluence of
for conceptualizing such distinguishing general information-processing capabilities,
tacit knowledge, practical intelligence, and expertise 615
knowledge depth and organization, and velocity of the striking surface). Such knowl-
environmental opportunity requires an edge remains tacit as people solely attend
understanding of how general psychological instead to the actions of the tool. In effect,
mechanisms interact with acquired knowl- the tool becomes an extension of the per-
edge and situational constraints to produce son, such that the person cannot articulate
expertise in a given domain. The theory of how she uses the tool any more than she
practical intelligence and tacit knowledge can articulate how she uses her own hand.
(Sternberg, 1988, 1997; Wagner & Sternberg, Polanyi (195 8) also emphasized the expe-
1985 ) explicitly addresses the interchange riential nature of tacit knowledge that it
between information-processing capability must be passed on by example and practice,
and experience in particular environmental often implicitly.
contexts, making it an important source of Since Polanyis work, scholars from
insight into expertise. Specifically, the the- domains as diverse as linguistics (Dahl,
ory posits that the development of practi- 2000), cognitive psychology (Reber, 1989;
cally intelligent behavior, a critical aspect of Reber & Lewis, 1977), differential psychol-
success in everyday life, occurs through a ogy (Wagner & Sternberg, 1985 ), and organi-
cycle of inquiry, knowing, and action, that is, zational management (Nonaka & Takeuchi,
a cycle of engaging the environment, acquir- 1995 ) have independently worked toward
ing tacit knowledge, and performing in a understanding the nature and acquisition
practically intelligent manner. of tacit knowledge. These concepts have
proven to be useful, if sometimes controver-
sial (see, e.g., Gottfredson, 2003 ; Stadler &
What Is Tacit Knowledge? Roediger, 1998), for understanding how peo-
ple accomplish much of what they do. This
The word tacit is used to characterize chapter will focus primarily on the work
exchanges that are carried out without the of psychologists Robert J. Sternberg and
use of words or speech, and to describe Richard K. Wagner and their colleagues, who
shared arrangements that have arisen with- have explored extensively the nature, mea-
out explicit agreement or discussion (Oxford surement, and acquisition of tacit knowledge
American Dictionary, 2001). Tacit knowing as it relates specifically to individual differ-
therefore represents a person-environment ences in practically intelligent behavior and
exchange that is not articulated and that expertise.
arises without explicit attempt to link envi- Wagner and Sternberg (1985 ) defined
ronmental stimulation to phenomenological tacit knowledge as knowledge that usually
experience. Although the idea that peoples is not openly expressed or stated . . . is not
actions are subject to unconscious influences directly taught or spoken about, in con-
dates back to Sigmund Freud in the late trast to knowledge directly taught in class-
1800s, scientist and philosopher of science rooms (pp. 43 843 9), with the qualifica-
Michael Polanyi (195 8, 1966) was among the tion that we do not wish to imply that
first to discuss formally the concept of tacit this knowledge is inaccessible to conscious
knowledge, noting its influence on percep- awareness, unspeakable, or unteachable, but
tion and scientific thinking. merely that it is not taught directly to most
Specifically, Polanyi (1966) argued that of us (p. 43 9). Sternberg and his colleagues
we can know more than we can tell (p. 4) have also emphasized that tacit knowledge
and that tacit knowledge underlies a wide is action oriented and procedural in nature,
range of skills, from tool use to application of essentially a complex set of condition-action
the scientific method. Polanyi (195 8, 1966) statements (see, e.g., Sternberg, Forsythe,
claimed that when humans use a tool (e.g., Hedlund et al., 2000). Wagner and Sternberg
a hammer), for example, they are unaware (1985 ) have presented tacit knowledge as
of how the sensations the tool is producing an enabler of practically intelligent behav-
on their palm (i.e., their grip) correspond to ior, which, in turn, is believed to be a critical
the action of the tool (e.g., the direction and aspect of expertise.
616 the cambridge handbook of expertise and expert performance
viewed as developing with effort and experi- performance, and commonly is taught
ence (Sternberg, 1998). Moreover, the devel- explicitly (e.g., Hunter, 1983 ; Ree, Caretta,
opment of practical intelligence through & Teachout, 1995 ). However, occupational
the acquisition of tacit knowledge occurs expertise requires more than implemen-
via an interaction between an individuals tation of facts and concepts learned in
existing level of competency and an environ- formal training, such that employees can
mental context. Ones level of general intel- solve novel problems and think proac-
ligence is believed to exist largely indepen- tively (see, e.g., DuBois & Shalin, 1995 ).
dently of ones knowledge and experience. Tacit knowledge specific to particular job
domains, in contrast, is not explicitly taught,
Crystallized Intelligence but is learned on the job. It facilitates
occupational expertise by bridging the gap
Crystallized intelligence has been defined between formal training and operational
as the outcome of experiential-educative- experience (see, e.g., Sternberg & Horvath,
acculturation influences (Horn & Cattell, 1999). Tacit knowledge can therefore be
1966, p. 25 4) on ones biological capacity. viewed as job-related knowledge, to the
Tests of crystallized intelligence commonly extent that it facilitates job performance
assess vocabulary, reading comprehension, (DuBois & Shalin, 1995 ); but tacit knowl-
and other verbal skills. Practical intelligence edge need not be job related (Sternberg
resembles crystallized intelligence through & the Rainbow Project Collaborators,
their shared dependence on experience and 2005 ).
similar developmental aspects. In contrast
to crystallized intelligence, however, prac- Procedural Knowledge
tical intelligence is applied to identifying
In contrast to declarative knowledge of
and solving problems that often do not have
facts and concepts, procedural knowledge
one correct solution strategy or clearly right
is knowledge of how to execute some task
answer.
(Anderson, 1982). Procedural knowledge
typically is viewed as the end state of a learn-
Personality and Motivation ing process for tasks that can be automa-
Because practical intelligence is theorized tized with practice, such as typing and other
to develop through effortful experience, it psychomotor skills (Ackerman, 1988; Ander-
can be argued that acquiring tacit knowl- son, 1982; Fitts & Posner, 1967). Like proce-
edge and improving practical intelligence are dural knowledge, tacit knowledge is action
in part functions of non-ability constructs oriented, gained from experience, applied
relating to performance, such as personal- unconsciously, and often difficult to verbal-
ity and motivation. With some probing, this ize. However, tacit knowledge is not viewed
argument can be made for the other psy- as an automatic response produced from
chological constructs listed above, including repeated exposures to the same patterns of
general intelligence as it is typically mea- stimuli. Rather, it is viewed as an adap-
sured. Practical intelligence is theorized to tive intellectual resource stemming from the
be distinct from personality and motivation active interaction between individuals and
because it characterizes a persons use of his their dynamic environment.
experiences for learning and performance Of course, making compelling distinc-
and not, in particular, his drive to succeed tions between practical intelligence and tacit
or typical pattern of responding to external knowledge and other psychological con-
events. structs requires empirical evidence, which,
in turn, requires valid measurement. Below
we describe the varied efforts to measure
Job Knowledge
practical intelligence and tacit knowledge
Job knowledge typically is conceptualized that have been made both by Sternberg
as declarative knowledge, such as of facts and his colleagues and other independent
and concepts required for successful job researchers.
618 the cambridge handbook of expertise and expert performance
You and a coworker jointly are responsible for completing a report on a new product by the
end of the week. You are uneasy about this assignment because he has a reputation for not
meeting deadlines. The problem does not appear to be lack of effort. Rather, he seems to lack
certain organizational skills necessary to meet a deadline and also is quite a perfectionist. As
a result, too much time is wasted coming up with the perfect idea, project, or report. Your
goal is to produce the best possible report by the deadline at the end of the week. Rate the
quality of the following strategies for
1 2 3 4 5 6 7
Extremely Neither Bad Extremely
Bad Nor Good Good
meeting your goal on a 1- to 7-point scale.
Divide the work to be done in half and tell him that if he does not complete his part, you
obviously will have to let your immediate superior know it was not our fault.
Politely tell him to be less of a perfectionist.
Set deadlines for completing each part of the report and accept what you have accom-
plished at each deadline as the final part of that report.
Ask your superior to check up on your progress on a daily basis (after explaining why).
Praise your coworker verbally for completion of parts of the assignment.
Get angry with him at the first sign of getting behind schedule.
As soon as he begins to fall behind, take responsibility for doing the report yourself, if need
be, to meet the deadline.
Point out firmly, but politely, how he is holding up the report.
Avoid putting any pressure on him because it will just make him fall even more behind.
Offer to buy him dinner at the end of the week if you both meet the deadline.
Ignore his organizational problem so you dont give attention to maladaptive behavior.
Figure 35.1. Example TKIM Vignette
Hedlund et al., 2000), each of which rep- knowledge draws from approaches com-
resents the degree to which an individuals monly used in managerial assessment and
ratings correspond to (or deviate from) the education, including in-basket tests, which
average ratings of some comparison sample, require prioritizing and responding to job
usually a group of experts. relevant materials (e.g., memos and reports)
in a limited amount of time, and case
studies, which involve critiquing and/or
Case-Study Scenarios
solving a detailed case description. Case-
Case-study scenarios provide detailed, in- study scenarios consist of a brief sum-
depth information about a practical prob- mary of a situation, followed by a detailed
lem situation as it develops over time, fol- description of the particulars as they unfold
lowed by a set of open-ended questions over the near term. Supporting documents
designed to assess knowledge-acquisition such as memos, transcriptions of verbal/
components and practical problem-solving e-mail exchanges, relevant reports, and other
skills, two important aspects of behav- related materials are also provided. Though
ing in a practically intelligent way (Stern- fictitious, case-study scenarios are designed
berg, 1988). This method of assessing tacit to represent realistically practical problems
62 0 the cambridge handbook of expertise and expert performance
Overview
You are 2nd LT Pete Quandry and have recently taken over an infantry platoon with 3 0
soldiers and 4 Bradleys. Because you have just come on board as a PL (Platoon Leader), you
need to learn a lot about weapon systems and procedures. The former PL left nothing on
paper to help you get oriented . . .
The platoon is currently in a state of flux because PSG Joe Forte just left. SSG Ed Newell, a
squad leader, has been promoted from among his peers without a change in rank to replace
him . . . Apparently, your company commander CPT Powers was very dissatisfied with the
previous PL but had a lot of respect for the former PSG because he kept the soldiers in
line . . .
He (CPT Powers) clearly has high expectations of you and the platoon and has already given
you responsibility for a new tactical mission . . .
Background
Apparently, CPT Powers found it so frustrating to work with the former PL that he often com-
municated directly with PSG Forte. PSG Forte had the reputation for being highly demanding
and directive with the platoon. See attachment 1 . . .
The Platoon has a mix of experienced and newly enlisted soldiers. Several were in combat
together. Attachment 2 is an early interaction with PSG Newell about the Platoon.
You have some serious concern about platoon performance because during a recent FTX,
you observed that the soldiers piled out of the vehicles and lit cigarettes rather than setting
up a secure perimeter as their battle drill dictated . . .
One week before the Mission
On Monday morning this week you discover that one of the leader books was not up to date
in the garrison . . .
On Tuesday, there was an accident with one of the Bradleys in a training exercise . . .
This morning (Wednesday) you meet with PSG Newell to discuss details of the upcoming
mission and your concerns about platoon performance . . .
You only have a few days left to motivate your troops and prepare for the mission.
Figure 35.2 . Abridged Military Leadership Case-Study Scenario
students, intermediate scores earned by tors, was administered to all four ranks of
graduate students, and lowest scores earned auditor. The solution-alternative quality rat-
by faculty. Wagner (1987) found this ings of the staff, senior, and manager employ-
same significant linear trend in a sam- ees were then compared to those of the
ple of managers, business students, and partners, who served as the expert crite-
undergraduates. rion. As predicted, Tan and Libby found that
Wagner and Sternberg (1985 , 1990) fur- level of tacit knowledge for auditing distin-
ther have shown that tacit knowledge plays guished top and bottom performers at the
an important role in business management. higher rank of manager (t = 1.72, p = .05 ),
In one study, Wagner and Sternberg (1985 ) but not at the lower ranks of senior (t =
administered an early version of the Tacit- 0.82, p = .21) and staff (t = 0.23 , p = .41).
Knowledge Inventory for Managers (TKIM, The opposite pattern was found for tech-
Wagner & Sternberg, 1991) to fifty-four busi- nical knowledge. Tacit knowledge became
ness managers and found that scores on the more important for success as the empha-
inventory correlated .46 with salary and .3 4 sis of job demands shifted away from the
with the level of the company where the application of technical skills to the han-
participating manager was employed (i.e., dling of complex practical problems, such
whether the company was among the top as competing goals and career management
companies in the Fortune 5 00 list). Wag- (Tan & Libby, 1997; see also Colonia-Willner,
ner and Sternberg (1985 ) also found that 1998).
in a sample of twenty-nine bank managers, Exploring military applications, Hedlund,
scores on the tacit-knowledge inventory cor- Forsythe, Horvath, Williams, Snook, and
related notably with a variety of indicators Sternberg (2003 ) studied the tacit knowl-
of management success. Specifically, they edge for military leadership of 5 62 com-
found that tacit knowledge correlated .48 missioned Army officers at the platoon,
with the percentage of salary increases, .5 6 company, and battalion levels of command.
with ratings of success in generating new Officers at each level of command filled
business, .29 with ratings of personnel man- out a Tacit Knowledge for Military Lead-
agement capability, and .3 9 with ratings of ership inventory (TKML; Hedlund et al.,
ability to implement company policy. In a 1998) relevant to their particular level of
later study of forty-five business managers command. In addition, supervisors, peers,
taking part in a leadership-development pro- and subordinates rated participating officers
gram, Wagner and Sternberg (1990) exam- on their leadership effectiveness. Hedlund
ined the correlation between scores on the et al. (2003 ) found that tacit knowledge
TKIM and performance ratings on two was independent of subordinate ratings at
management simulations conducted as part both the company and battalion levels of
of the program. The correlation between command (subordinates did not rate pla-
TKIM scores and performance ratings was toon leaders), and of peer ratings at the
.61, with the TKIM outpredicting several platoon and battalion levels of command.
measures of general cognitive ability and However, significant correlations between
personality. tacit knowledge and supervisor ratings were
Tan and Libby (1997) have also explored found at the platoon and battalion levels of
the relationship between tacit knowledge command, with an especially notable cor-
and expertise in a business setting, but with relation at the battalion level (r = .48).
a special focus on financial auditors. They Mirroring Tan and Libbys (1997) findings,
studied 100 auditors at the ranks of staff, lower correlations were found between tacit
senior, manager, and partner employed by knowledge and supervisor ratings at the
the Singapore office of a major account- level of platoon leader (.17) and company
ing firm. A tacit-knowledge inventory for commander (.05 ).
auditing based on the TKIM (Wagner & Legree et al. (2003 ) worked with 5 5 1
Sternberg, 1991), but specialized for audi- enlisted Army personnel, studying the
tacit knowledge, practical intelligence, and expertise 62 3
relationship between tacit knowledge for which has been conducted by Sternberg and
safe driving and accident history. Legree et his colleagues. We summarize below these
al.s Safe Speed Knowledge Test featured diverse research and development efforts.
fourteen scenarios that briefly depicted
various combinations of driving conditions,
including weather (e.g., snow, clear, light Making Tacit Knowledge Explicit:
rain), traffic (heavy, light), or emotional Communities of Practice
states (angry, stressed), among others. Par-
ticipants estimated how much they would Methods to make tacit knowledge explicit
adjust their speed (in one mph increments) are at the heart of a number of accepted
in each scenario, ranging from no adjustment practices recognized for their potential
to slowing down twenty miles per hour. Per- benefit to personal and professional develop-
formance scores on the Safe Speed Knowl- ment. These practices include psychother-
edge Test were derived by calculating the apy, which involves uncovering tacit knowl-
absolute difference between an individuals edge that may be maladaptive, and mentor-
speed adjustments and the average speed ing, which involves the articulation of why
adjustments of the experimental sample (see a particular action should be taken at a par-
Legree, 1995 ), then reflecting this difference ticular time. Although some may reject the
such that larger scores indicated greater tacit notion that tacit knowledge can be made
knowledge. Tacit-knowledge scores showed explicit, its potential to contribute to the
a strong relationship to driver at-fault crash development of expertise has been explored
rates. Specifically, individuals scoring with and advocated by a number of scholars
in one standard deviation of the mean dis- from a variety of theoretical perspectives
tance score were 2.3 times more likely to (Argyris, 1993 ; Brown & Duguid, 1991;
be involved in a crash as individuals scor- Schon, 1983 Sternberg, Forsythe, Hedlund et
ing more than one standard deviation above al., 2000; Wenger, 1998). The central belief
the mean. Individuals scoring more than shared by these scholars, and others, is that
one standard deviation below the mean dis- because most of the relevant know-how that
tance score were five times more likely to be distinguishes different levels of expertise is
involved in a crash. acquired through experience, methods that
stimulate the process of thinking about what
one is doing and why, and talking about
Enhancing Expertise through it with others, will facilitate the develop-
Development of Practical Intelligence ment of expertise.
and Tacit Knowledge Communities of practice, defined as
groups of people who informally come
Given that practical intelligence, through together to exchange knowledge and expe-
tacit-knowledge acquisition and improved rience in a shared domain of interest, have
problem-solving skills, is believed to be been increasingly recognized as an effective
developable (Sternberg, 1998), research has mechanism to develop expertise through
also been devoted to exploring methods for sharing tacit knowledge (Gerardi, Nicolini,
making this development happen. Wagner & Odella, 1998; Lesser & Storck, 2001;
and Sternberg (1990) have noted two ways Nonaka & Takeuchi, 1995 ; Wenger, 2000;
in which ones body of tacit knowledge can Wenger & Snyder, 2000). They are dis-
be enhanced: (1) making tacit knowledge tinguished from workgroups and teams in
explicit and sharing it, and (2) improving that the nature of membership is self-
peoples ability to engage their environ- selection versus assignment by an organiza-
ments and learn from experience. Substan- tional authority, and the purpose of mem-
tial research and program development have bership is to develop capability, and build on
also been devoted to improving peoples and exchange knowledge rather than accom-
practical problem-solving skills, some of plish a more specific task or assignment.
62 4 the cambridge handbook of expertise and expert performance
Examples of professional or trade associ- informally share their knowledge. Now for-
ations akin to communities of practice can mally recognized and endorsed by senior
be found throughout human history, for Army leadership, CompanyCommand.mil is
example, in the artisan guilds in the Mid- defined as an ongoing professional con-
dle Ages (Lave & Wenger, 1991; see also versation about leading soldiers and build-
Amirault and Branson, Chapter 5 ). What ing combat-ready units, where former and
distinguishes them today is that they are experienced company commanders share
not only common among independent prac- leadership-related stories, ideas, and tools
titioners but also are being formed in the with current and future commanders. In
context of large organizations. Enabled by addition, commanders with a question or
advances in information technology, modern problem invite the advice of those who may
communities of practice range in format have knowledge on the topic. When partici-
from regularly scheduled meetings to list- pants receive feedback that challenges their
servs to online discussion forums. Though thinking, they are encouraged to reflect on
typically communities of practice are self- their underlying assumptions and ultimately
organizing, increasingly, organizations are assess their thinking, which promotes the
sponsoring their development to cultivate development of practical problem-solving
needed capabilities and are enjoying sub- ability (Cianciolo, et al. 2004).
stantial return on investment, accomplished
through increased productivity and inno- civilian communities of practice
vation (Crager & Lemons, 2003 ). Below
Sponsorship and support for communities of
we describe some contemporary examples
practice can also be found in a wide range
of communities of practice, with particular
of other private and public sector organi-
attention to how they facilitate making tacit
zations (Crager & Lemons, 2003 ; Lesser &
knowledge explicit and sharing it.
Storck, 2001). Although the type of orga-
army structured professional forums nization and the particular area of exper-
tise may differ, these communities of prac-
Changes in the global political environment
tice share the same objective to create
require that the Army adapt quickly to a
knowledge and stimulate innovation by shar-
broader range of missions, from warfight-
ing experience and expertise across simi-
ing to peacekeeping, while supported by
lar projects or products. For example, at
state-of-the-art information technology. In
Hewlett-Packard, product-delivery consul-
response to this need, the Army is spon-
tants across the country hold monthly tele-
soring the development of a web-enabled
conferences to solve problems they share
knowledge-sharing system called the Bat-
in connection with a particular software
tle Command Knowledge System to pro-
product (Wenger & Snyder, 2000). Pepper-
vide Army leaders and soldiers with spe-
dine University Educational Technology has
cific, relevant knowledge and information to
developed a community of practice to cap-
prepare them to rapidly develop expertise.
ture and transfer knowledge generated by
Structured professional forums communi-
participants engaged in the EdTech Doc-
ties of practice more broadly designed to be
toral program using group memory soft-
resources for professional self-development
ware. Although its design is derived from
and community growth are a critical com-
the perspective of social-learning theory, the
ponent of this knowledge system (Kilner,
importance of capturing and transferring
2002).
tacit knowledge is emphasized (Adams &
One such structured professional forum,
Freeman, 2000).
CompanyCommand.mil (see Dixon, Allen,
Burgess, Kilner, & Schweitzer, 2005 ), was
initially developed as a volunteer effort research on communities of practice
by two Army captains to create a web- Although the financial value of communi-
site where company commanders could ties of practice to organizations has been
tacit knowledge, practical intelligence, and expertise 62 5
demonstrated (Crager & Lemons, 2003 ), standing the problem at hand, and how
there has been relatively little investi- the knowledge acquired from past experi-
gation into how both individual- and ence can be compared to new knowledge to
organizational-level expertise develop from inform decisions and action. It is believed
sharing tacit knowledge (though see Lesser that as students reflect on how they are using
& Storck, 2001). A recent investigation of information from the environment and from
the impact of Army structured professional experience to solve problems, they will come
forums on individual leader competency and to value their experiences as opportunities
professionalism, unit effectiveness, and orga- for learning (Cianciolo et al., 2004).
nizational performance has begun to illu- In one study using this approach,
minate metrics and assessment methods for Sternberg, Wagner, and Okagaki (1993 )
capturing expertise as it relates to the intel- worked with five groups of fifteen college
lectual and social capital developed through students and explored how different instruc-
activity in an online discussion forum (Cian- tional conditions affected the difference
ciolo, Heiden, Prevou, & Psotka, 2005 ). This in tacit-knowledge inventory scores before
study represents only the beginning of the and after intervention. One group of stu-
research that must be conducted to under- dents served as a control in the experi-
stand how the organization and its individ- ment, receiving no intervention, and each
uals leverage tacit knowledge to improve of the three experimental groups received
performance. instruction on how to use one of the three
knowledge-acquisition components while
completing a tacit-knowledge-acquisition
Facilitating Tacit-Knowledge Acquisition
task. The fifth group of students served as a
Facilitating tacit-knowledge acquisition is a second control, completing the knowledge-
more indirect approach to enhancing ones acquisition task, but without an instructional
body of tacit knowledge than is sharing intervention.
explicit tacit knowledge (see also, Hoffman The knowledge-acquisition task required
& Lintern, Chapter 12). However, this indi- participants to play the role of a personnel
rect approach can be expected to have more manager and to evaluate three fictional job
lasting effects on ones practical intelligence. candidates for sales positions, opting to hire
Explicit tacit knowledge, shared in commu- none, one, two, or three of the candidates.
nities of practice, eventually becomes out- On the basis of three fictional interview tran-
dated, sometimes very rapidly, as the mores scripts and a description of the hiring com-
of cultures shift (e.g., by becoming multi- pany, participants were asked to evaluate
ethnic) or as the operational environment the candidates for their ability to manage
changes (e.g., by introducing new techno- themselves, to handle the tasks and problems
logical capability). Learning how to acquire that arise in sales positions, and to handle
tacit knowledge, however, never becomes business relationships with customers, peers,
outdated. and superiors. To facilitate selective encod-
Approaches to facilitating tacit-know- ing, participants in the first experimental
ledge acquisition target the three cognitive group were cued with relevant information
processes thought to underlie knowledge- during the knowledge-acquisition task and
acquisition: selective encoding, selective were provided with relevant rules-of-thumb.
combination, and selective comparison To facilitate selective combination, partici-
(Sternberg, 1988). Specifically, instruction is pants in the second experimental group were
designed to draw students attention to how also cued with relevant information and pro-
the relevant information in the environment vided with rules-of-thumb, but were addi-
or from previous experience can be selected tionally provided with a structured note-
to guide decision making and problem solv- taking sheet to help link the information
ing, how relevant information can be com- to evaluation criteria. Participants in the
bined to form patterns meaningful to under- third experimental group were also cued
62 6 the cambridge handbook of expertise and expert performance
with relevant information and provided with Perkins and his colleagues (Andrade &
rules-of-thumb, but to facilitate selective Perkins, 1998; Grotzer & Perkins, 2000) have
comparison participants in this group also recently explored an approach called Cog-
received an evaluation that had been com- nitive Reorganization, which is designed to
pleted by a predecessor in the company teach school children to think more deeply
such that they could benefit from someone about problems before solving them, to
elses experience. monitor their thinking more closely, to be
Sternberg et al. (1993 ) found that partici- more open, careful, and organized about
pants in the two control groups showed less their thinking, to draw on multiple resources
gain from pre-test to post-test (performance for thinking, and to make better connections
on a tacit-knowledge inventory for sales) between past, present, and future thinking
than the experimental groups (average gain experiences. Because of difficulties in con-
for control groups was 5 .6; for experimen- ducting and evaluating programs designed
tal groups it was 15 .27). Further, the exper- to teach thinking skills (see Grotzer &
imental groups who received instruction on Perkins, 2000), however, a definitive conclu-
selective encoding and selective combina- sion has not yet been reached whether these
tion showed notably greater gain than the programs have lasting effects. In the evalu-
experimental group who received instruc- ation studies that have been conducted, the
tion on selective comparison (16.8, 19.7, and results have been mixed but generally pos-
9.3 , respectively). The results from this brief itive, suggesting that facilitating inquisitive-
intervention indicate that efforts to improve ness and critical thinking in an attempt to
tacit-knowledge acquisition and, by exten- enhance intelligent behavior is not a mis-
sion, practical intelligence, can be successful. guided endeavor (Bransford et al., 1986).
The method of this research is not
altogether isolated from other efforts to
Developing Practical Intelligence
stimulate peoples intellectual curiosity and
critical thinking (see also Zimmerman, Another attempt to improve practical
Chapter 3 9). Its theoretical heritage can be intelligence, Practical Intelligence for
traced to philosopher John Dewey (193 3 ), School (PIFS; Williams, Blythe, White,
who believed that reflective thought the Li, Sternberg, & Gardner, 1996; Williams,
critical analysis of ones ideas and behav- Blythe, White, Li, Gardner, & Sternberg,
iors was an important aspect of good 2002), targeted practical thinking skills
thinking skills and effective problem solv- directly in an effort to boost school achieve-
ing. Deweys work has influenced the think- ment. Based on Sternbergs (1988, 1997)
ing of numerous philosophers (e.g., Ennis, triarchic theory of successful intelligence
1987; Lipman, 1993 ) and psychologists (e.g., and Gardners (1983 , 1999) theory of
Andrade & Perkins, 1998; Bransford, Sher- multiple intelligences, PIFS was a com-
wood, Vye, & Rieser, 1986; Nickerson, prehensive theory-based program designed
1989) who have sought to understand and to enhance middle-school students (fifth
enhance problem-solving and experience- and sixth graders) scholastic achievement.
based learning. These scholars have used a Middle-school students were a focus in this
wide range of methods to facilitate inquis- intervention because differences in practical
itiveness and experience-based learning intelligence for school begin to appear
from using stories to engage children in the during the middle-school years and at that
philosophical analysis of problems (Lipman, time set the stage for future differences in
1993 ) to training the particular cognitive school performance.
processes involved in insight and learning To develop students practical think-
from context (Davidson & Sternberg, 1984; ing skills, the PIFS program was organized
Sternberg, 1987) to exploring complex prob- around five themes: knowing why, know-
lems by discussing multiple points of view ing self, knowing differences, knowing pro-
(Paul, 1987). cess, and revisiting, and targeted four domain
tacit knowledge, practical intelligence, and expertise 62 7
areas: reading, writing, homework, and test- et al., 2002). Long-term benefits of the PIFS
ing. In each of the four domain areas, stu- program have not been assessed.
dents participated in a set of lessons that fea-
tured exercises to engage practical thinking
along the five themes. For example, in the Future Directions for Practical
topic of writing, PIFS students are taught to Intelligence/Tacit Knowledge
discover why writing is important both in Research
and out of school (knowing why), to rec-
ognize their personal strengths and weak- An extensive amount of research has been
nesses in writing (knowing self), to distin- committed to exploring the nature, mea-
guish between different styles and strategies surement, and predictive validity of practi-
of writing (knowing differences), to under- cal intelligence and tacit knowledge. Because
stand the role of planning and organization of the potential practical and theoretical
in the writing process (knowing process), importance of these constructs, they have
and to recognize the importance of revis- been investigated by a wide range of scholars
ing (revisiting). Stand-alone exercises were including Sternberg and his colleagues but
applied to homework and testing skills. Stu- also several independent researchers. There
dents explored, for example, the purposes remains, however, much interesting work
for homework, their personal homework to be done in order to more fully under-
practices, differences in homework require- stand the nature of tacit knowledge and
ments for different classes, and strategies for practical intelligence and their development.
improving the effectiveness of the home- For example, large-scale investigations of the
work process and homework quality. construct validity and predictive efficacy of
The PIFS program took place over a two- practical intelligence and relatively domain-
year period and was implemented in both general tacit-knowledge measures within the
Connecticut and Massachusetts. To main- context of multiple cognitive ability tests
tain an ecologically valid implementation, and non-ability assessments (e.g., personal-
the PIFS researchers encouraged partici- ity inventories) should be conducted.
pating teachers to integrate PIFS instruc- Another area of interest involves the con-
tion into their teaching in a manner that ceptualization and measurement of tacit
best suited their instructional needs. Thus, knowledge as it is acquired and applied
PIFS lessons were either infused into ongo- dynamically during decision making and
ing classroom activity or administered as problem solving. Although practical intel-
stand-alone instruction. At the end of each ligence and tacit knowledge are viewed as
year, teachers and researchers evaluated stu- modifiable (Sternberg, 1998), most mea-
dents practical thinking skills and academic sures of tacit knowledge do not require
achievement via a set of practical and aca- examinees to work through practical prob-
demic assessments. Performance on these lems as they develop over time and as a result
assessments was compared to performance of particular actions. The case-study scenar-
on similar assessments given at the beginning ios recently developed by Sternberg and his
of the year in order to evaluate gains in devel- colleagues represent a major effort to address
opment. At the end of Year 2, students par- this area of interest. However, efforts to cap-
ticipating in the PIFS program showed sig- ture tacit knowledge as it is acquired and
nificantly greater gain in scores on practical applied in real time also can be informed
assessments of reading (F = 19.3 7, p < .001), by theories of social judgment and decision
writing (F = 25 .3 3 , p < .001), homework making (Brehmer & Joyce, 1988), whose pri-
(F = 27.89, p < .001), and test-taking (F = mary focus includes identifying the environ-
10.3 6, p < .01) and on academic assessments mental conditions that people use as the
of reading (F = 13 .63 , p < .001), writing basis for judgment and action.
(F = 16.49, p < .001), and testing (F = 5 .71, Briefly, social judgment theorists attempt
p < .05 ) than non-PIFS students (Williams to mathematically model the basis of human
62 8 the cambridge handbook of expertise and expert performance
judgment and action by identifying the envi- Although the use of various leader-
ronmental cues linked to particular states development methods aimed at facilitating
of the world and how people use these tacit-knowledge acquisition in professional
cues to inform their judgments about these development is theoretically compelling,
states of the world. Conceptualized as the there appears to be limited published
adaptive use of environmental cues in par- empirical research to support claims of their
ticular situations, expert tacit knowledge effectiveness. Moreover, it is still unclear
can be identified and measured using these which of these methods, if any, actually
sophisticated performance-modeling tech- facilitate the acquisition of tacit knowledge
niques, which can be embedded into com- and in turn the development of expertise. In
plex, dynamic tasks, such as command and particular, the distinction between cognitive
control simulations (see, e.g., Rothrock & versus social learning processes involved
Kirlik, 2003 ). in the development of experience-based
Identifying experts tacit knowledge using learning, as suggested by social learning
performance modeling also offers an alter- approaches (Gherardi, Nicolini, & Odella,
native to relying on traditional methods for 1998; Lave & Wenger, 1991), needs to be
eliciting tacit knowledge. To date, the elici- explored and understood (see also Evetts,
tation of tacit knowledge has relied on the Mieg, & Felt, Chapter 7; Mieg, Chapter
verbalized recollections of experts, which 41). Research is also needed to test the
provide unreliable information regarding the effectiveness of methods currently in use
actual conditions (or environmental cues) and the conditions under which they may
that triggered particular actions (Nisbett & be effective. One important question that
Bellows, 1977; Nisbett & Wilson, 1977). warrants examination is whether distinct
Moreover, mathematical modeling of tacit methods are required for the development
knowledge may provide some insight into of novices, as compared to journeymen
what specifically is learned when tacit or experts, who may need to unlearn
knowledge is acquired, thus making it eas- or modify outdated tacit knowledge.
ier to make tacit knowledge explicit or to Another potentially useful area for research
facilitate tacit-knowledge acquisition. is the identification of the individual-
Because much has been discovered differences factors that may influence the
regarding the nature and measurement of effectiveness of one particular method
tacit knowledge, the development of tacit over another.
knowledge and practical intelligence is an
area particularly ripe for research and is of
special interest to leadership-development Conclusion
professionals. In leadership development,
there is movement toward embedding tacit Although it is a delight to be awed by artis-
knowledge acquisition into the work context tic, athletic, or intellectual excellence, or
through methods that stimulate reflection to be struck by the brilliance of everyday
on experience and action, such as executive adaptation, the general psychological mech-
coaching and stretch assignments (Day, anisms underlying the nature and develop-
2000). Methods that facilitate the transfer ment expertise need not, indeed do not,
of tacit knowledge from senior executive to remain a mystery. The theory of practical
junior manager are also commonly featured intelligence and tacit knowledge (Sternberg,
in leadership-development planning, includ- 1988, 1997; Wagner & Sternberg, 1985 )
ing classes taught by senior executives that has provided one means for understand-
focus on sharing stories about pivotal leader- ing expert performance by illuminating the
ship experiences and lessons learned. Other complex, ever-shifting person-environment
methods include mentoring, which provides interaction that allows for success in every-
junior managers an opportunity to observe, day life. This approach to understanding
interact, and reflect on senior executives in expertise complements many of the other
action. approaches discussed in this handbook, by
tacit knowledge, practical intelligence, and expertise 62 9
serving as an integration of multiple per- Andrade, H. G., & Perkins, D. N. (1998). Learn-
spectives and maintaining an inclusive con- able intelligence and intelligent learning. In
ceptualization of what it means to be an R. J. Sternberg & W. M. Williams (Eds.), Intelli-
expert. There is yet much work to be done to gence, instruction, and assessment. (pp. 6794).
fully explore tacit knowledge, practical intel- Mahwah, NJ: Erlbaum.
ligence, and expertise, but this work rep- Argyris, C. (1993 ). Knowledge for action. San
resents exciting movement forward in the Francisco, CA: Jossey-Bass.
quest to understand a phenomenon that has, Bloom, B. S. (Ed.) (1985 ). Developing talent in
in one form or another, inspired us all. young people. New York: Ballentine Books.
Bransford, J., Sherwood, R., Vye, N., & Rieser, J.
(1986). Teaching thinking and problem solving:
Acknowledgments Research foundations. American Psychologist,
41, 10781089.
Preparation of this chapter was supported by Brehmer, B., & Joyce, C. R. B. (1988). Human
the U.S. Army Research Institute. Although judgment: The SJT view. New York: Elsevier.
we are grateful to this agency for its support, Brown, J. S., & Duguid, P. (1991). Organizational
the ideas expressed in this chapter are solely learning and communities-of-practice: Toward
those of the authors and do not represent any a unified view of working, learning, and inno-
official position or policy on the part of ARI. vation. Organization Science, 2 , 405 7.
We wish to thank Dr. Paul Feltovich and Chase, W. G., & Simon, H. A. (1973 ). Perception
the anonymous reviewer whose thought- in chess. Cognitive Psychology, 4, 5 5 81.
ful review and helpful comments signifi- Cianciolo, A. T., Antonakis, J., & Sternberg, R. J.
cantly improved our discussion. Send corre- (2004). Practical intelligence and leadership:
Using experience as a mentor. In D. V. Day, S.
spondence to Anna T. Cianciolo, Command
Zaccaro, & S. Halpin (Eds.), Leader development
Performance Research, Inc., 1201 Waverly for transforming organizations Growing leaders
Dr., Champaign, IL, 61821. for tomorrow (pp. 21123 6). Mahwah, NJ: Erl-
baum.
Cianciolo, A. T., & Grigorenko, E. L., Jarvin,
Footnote L., Gil, G., Drebot, M., & Sternberg, R. J.
(in press). Tacit knowledge and practical
1. For consistency of presentation and ease of intelligence: Advancements in measurement and
interpretation, the signs of the correlation coef- construct validity. Learning & Individual Differ-
ficients from some studies have been reflected, ences.
where appropriate, such that greater lev- Cianciolo, A. T., Heiden, C., Prevou, M. I., &
els of tacit knowledge (i.e., less deviation Psotka, J. (2005 ). Evaluating Army structured
from the criterion comparison group) corre- professional forums: Innovations in under-
spond to greater levels of other variables of standing and assessing effectiveness. Proceed-
interest. ings of the Interservice/Industry Training, Sim-
ulation, and Education Conference (I/ITSEC).
Arlington, VA: National Defense Industrial
References Association.
Colonia-Willner, R. (1998). Practical intelligence
Ackerman, P. L. (1988). Determinants of individ- at work: Relationship between aging and cogni-
ual differences during skill acquisition: Cogni- tive efficiency among managers in a bank envi-
tive abilities and information processing. Jour- ronment. Psychology & Aging, 13 , 45 5 7.
nal of Experimental Psychology: General, 117 , Crager, J., & Lemons, D. (2003 ). Consortium
2883 18. Learning Forum best-practice report: Measuring
Adams, E. C., & Freeman, C. (2000). Com- the impact of knowledge management. Houston,
munities of practice: Bridging technology and TX: American Productivity & Quality Center.
knowledge assessment. Journal of Knowledge Dahl, T. (2000). Text summarisation: From
Management, 4, 3 842. human activity to computer program. The
Anderson, J. R. (1982). Acquisition of cognitive problem of tacit knowledge. Journal of Linguis-
skill. Psychological Review, 89, 3 69406. tics, 2 5 , 113 13 1.
630 the cambridge handbook of expertise and expert performance
Davidson, J. E., & Sternberg, R. J. (1984). The Gherardi, S., Nicolini, D., & Odella, F. (1998).
role of insight in intellectual giftedness. Gifted Toward a social understanding of how people
Child Quarterly, 2 8, 5 864. learn in organizations. Management Learning,
Day, D. V. (2000). Leadership development: A 2 9, 273 297.
review in context. Leadership Quarterly, 11, Gottfredson, L. S. (2003 ). Dissecting practical
5 81613 . intelligence theory: Its claims and evidence.
Day, D. V., & Lance, C. E. (2004). Understand- Intelligence, 3 1, 3 43 3 97.
ing the development of leadership complexity Grotzer, T. A., & Perkins, D. N. (2000). Teaching
through latent growth modeling. In D. V. Day, intelligence. In R. J. Sternberg (Ed.), Handbook
S. Zaccaro, & S. Halpin (Eds.), Leader develop- of intelligence (pp. 4925 15 ). New York: Cam-
ment for transforming organizations Growing bridge University Press.
leaders for tomorrow (pp. 4169). Mahwah, NJ: Hedlund, J., Forsythe, G. B., Horvath, J. A.,
Erlbaum. Williams, W. M., Snook, S., & Stern-
Dewey, J. (193 3 ). How we think: A restatement of berg, R. J. (2003 ). Identifying and assessing
the relation of reflective thinking to the educative tacit knowledge: Understanding the practical
process. Boston, MA: Heath. intelligence of military leaders. Leadership
Dixon, N. M., Allen, N., Burgess, T., Kilner, P., Quarterly. 14, 117140
& Schweitzer, S. (2005 ). CompanyCommand: Hedlund, J., Horvath, J. A., Forsythe, G. B.,
Unleashing the power of the Army profession. Snook, S., Williams, W. M., Bullis, R. C.,
West Point, NY: Center for the Advancement Dennis, M., & Sternberg, R. J. (1998). Tacit
of Leader Development and Organizational knowledge in military leadership: Evidence of
Learning. construct validity, Technical Report 1080, U.S.
DuBois, D., & Shalin, V. L. (1995 ). Adapting Army Research Institute for the Behavioral &
cognitive methods to real-world objectives: An Social Sciences, Alexandria, VA.
application to job knowledge testing. In P. D. Horn, J. L., & Cattell, R. B. (1966). Refinement
Nichols, S. F. Chipman, & R. L. Brennan (Eds.), and test of the theory of fluid and crystal-
Cognitively diagnostic assessment. (pp. 189 lized general intelligences. Journal of Educa-
220). Hillsdale, NJ: Erlbaum. tional Psychology, 5 7 , 25 3 270.
Duncan, J., Seitz, R. J., Kolodny, J., Bor, D., Hunter, J. E. (1983 ). A causal analysis of cognitive
Herzog, H., Ahmed, A., Newell, F. N., & ability, job knowledge, job performance, and
Emslie, H. (2000). A neural basis for general supervisor ratings. In F. J. Landy, S. Zedeck, &
intelligence. Science, 2 89, 45 7460. J. Cleveland (Eds.), Performance measurement
Ennis, R. H. (1987). A taxonomy of critical think- and theory (pp. 25 7275 ). Hillsdale, NJ: Erl-
ing dispositions and abilities. In J. B. Baron & baum.
R. J. Sternberg (Eds.), Teaching thinking skills: Jensen, A. R. (1993 ). Test validity: g versus tacit
Theory and practice (pp. 926). New York: knowledge. Current Directions in Psychological
W. H. Freeman. Science, 2 , 910.
Ericsson, K. A., Krampe, R. T., & Tesch-Romer, Kilner, P. (2002). Transforming army learning
C. (1991). The role of deliberate practice in the through communities of practice. Military
acquisition of expert performance. Psychologi- Review, MayJune, 2127.
cal Review, 100, 3 63 406.
Larkin, J. H., McDermott, J., Simon, D. P., &
Ericsson, K. A., & Smith, J. (Eds.) (1993 ). Toward Simon, H. A. (1980). Expert and novice per-
a general theory of expertise: Prospects and limits. formance in solving physics problems. Science,
Cambridge, UK: Cambridge University Press. 2 08, 13 3 5 13 42.
Fitts, P., & Posner, M. I. (1967). Human perfor- Lave, J., & Wenger, E. (1991). Situated learning:
mance. Belmont, CA: Brooks/Cole. Legitimate peripheral participation. Cambridge,
Gardner, H. (1983 ). Frames of mind: The theory of UK: Cambridge University Press.
multiple intelligences. New York: Basic. Legree, P. J. (1995 ). Evidence for an oblique social
Gardner, H. (1999). Intelligence reframed: Multi- intelligence factor established with a Likert-
ple intelligences for the 2 1st century. New York: based testing procedure. Intelligence, 2 1, 247
Basic. 266.
Garner, W. R., Hake, H. W., & Eriksen, C. W. Legree, P. J., Heffner, T. S., Psotka, J., Martin,
(195 6). Operationism and the concept of per- D. E., & Medsker, G. J. (2003 ). Traffic crash
ception. Psychological Review, 63 , 14915 9. involvement: Experiential driving knowledge
tacit knowledge, practical intelligence, and expertise 631
and stressful contextual antecedents. Journal of edge in complex training performance. Journal
Applied Psychology, 88, 15 26. of Applied Psychology, 80, 72173 0.
Lesser, E. L., & Storck, J. (2001). Communities of Rothrock, L., & Kirlik, A. (2003 ). Inferring rule-
practice and organizational performance. IBM based strategies in dynamic judgment tasks:
Systems Journal, 40, 83 1841. Toward a noncompensatory formulation of the
Lipman, M. (1993 ). Promoting better class- lens model. IEEE Transactions on Systems, Man,
room thinking. Educational Psychology, 13 , 291 and Cybernetics Part A: Systems and Humans,
3 04. 3 3 , 5 872.
McDaniel, M. A., & Nguyen, N. T. (2001). Sit- Schmidt, F. L., & Hunter, J. E. (1993 ).
uational judgment tests: A review of practice Tacit knowledge, practical intelligence, gen-
and constructs assessed. International Journal of eral mental ability, and job knowledge. Cur-
Selection and Assessment, 9, 103 113 . rent Directions in Psychological Science, 2 , 8
Neisser, U. (1976). General, academic, and arti- 9.
ficial intelligence. In L. B. Resnick (Ed.), The Schon, D. A. (1983 ). The reflective practitioner:
nature of intelligence (pp. 13 5 144). Hillsdale, How professionals think in action. New York:
NJ: Erlbaum. Basic Books.
Nickerson, R. S. (1989). On improving thinking Somech, A., & Bogler, R. (1999). Tacit knowl-
through instruction. Review of Research in Edu- edge in academia: Its effects on student learn-
cation, 15 , 3 5 7. ing and achievement. The Journal of Psychology,
Nisbett, R. E., & Bellows, N. (1977). Ver- 13 3 , 605 616.
bal reports about causal influences on social Stadler, M. A., & Roediger, H. L. (1998). The
judgments: Private access versus public theo- question of awareness in research on implicit
ries. Journal of Personality & Social Psychology, learning. In M. A. Stadler & P. A. Frensch
3 5 , 613 624. (Eds.), Handbook of implicit learning (pp. 105
Nisbett, R. E., & Wilson, T. D. (1977). Telling 13 2). Thousand Oaks, CA: Sage.
more than we can know: Verbal reports on Sternberg, R. J. (1987). Most vocabulary is
mental processes. Psychological Review, 84, 23 1 learned from context. In M. G. McKeown
25 9. & M. E. Curtis (Eds.), The nature of vocab-
ulary acquisition (pp. 89105 ). Hillsdale, NJ:
Nonaka, I., & Takeuchi, H. (1995 ). The knowledge-
Erlbaum.
creating company: How Japanese companies cre-
ate the dynamics of innovation. New York: Sternberg, R. J. (1988). Beyond IQ: A triarchic the-
Oxford University Press. ory of human intelligence. New York: Cambridge
University Press.
Oxford American Dictionary. (2001). New York:
Oxford University Press. Sternberg, R. J. (1991). Theory-based testing of
intellectual abilities: Rationale for the Triarchic
Paul, R. W. (1987). Dialogical thinking: Critical
Abilities Test. In H. A. Rowe (Ed.), Intelligence:
thought essential to the acquisition of rational
Reconceptualization and measurement (pp. 183
knowledge and passions. In J. B. Baron & R. J.
202). Hillsdale, NJ: Erlbaum.
Sternberg (Eds.), Teaching thinking skills: The-
ory and Practice (pp. 127148). New York: W. H. Sternberg, R. J. (1994). Cognitive conceptions of
Freeman. expertise. International Journal of Expert Sys-
tems, 7 , 112.
Polanyi, M. (195 8). Personal knowledge: Towards a
post-critical philosophy. Chicago, IL: University Sternberg, R. J. (1997). Successful intelligence.
of Chicago Press. New York: Plume Books.
Polanyi, M. (1966). The tacit dimension. New Sternberg, R. J. (1998). Abilities are forms of
York: Doubleday. developing expertise. Educational Researcher,
2 7 , 1120.
Reber, A. S. (1989). Implicit learning and tacit
Sternberg, R. J., Castejon, J. L., Prieto, M. D.,
knowledge. Journal of Experimental Psychology:
Hautamaki, J., & Grigorenko, E. L. (2001).
General, 118, 21923 5 .
Confirmatory factor analysis of the Sternberg
Reber, A. S., & Lewis, S. (1977). Implicit learning: Triarchic Abilities Test in three international
An analysis of the form and structure of a body samples. European Journal of Psychological
of tacit knowledge. Cognition, 5 , 3 3 3 3 61. Assessment, 17 , 116.
Ree, M. J., Caretta, T. R., & Teachout, M. S. Sternberg, R. J., Forsythe, G. B., Hedlund, J.,
(1995 ). Role of ability and prior job knowl- Horvath, J. A., Wagner, R. K., Williams,
632 the cambridge handbook of expertise and expert performance
W. M., Snook, S. A., & Grigorenko, E. L. Wagner, R. K., & Sternberg, R. J. (1985 ). Practical
(2000). Practical intelligence in everyday intelligence in real-world pursuits: The role of
life. Cambridge, UK: Cambridge University tacit knowledge. Journal of Personality & Social
Press. Psychology, 49, 43 645 8.
Sternberg, R. J., & Horvath, J. A. (Eds.) Wagner, R. K., & Sternberg, R. J. (1989). Tacit
(1999). Tacit knowledge in professional practice: Knowledge Inventory for sales: Written. Unpub-
Researcher and practitioner perspectives. Mah- lished test.
wah, NJ: Erlbaum. Wagner, R. K, & Sternberg, R. J. (1990).
Sternberg, R. J., Nokes, K., Geissler, P. W., Street smarts. In K. E. Clark & M. B.
Prince, R., Okatcha, F., Bundy, D. A., & Clark (Eds.), Measures of leadership (pp. 493
Grigorenko, E. L. (2001). The relationship 5 04). West Orange, NJ: Leadership Library of
between academic and practical intelligence: America.
A case study in Kenya. Intelligence, 2 9, 401 Wagner, R. K., & Sternberg, R. J. (1991). Tacit
418. Knowledge Inventory for managers. San Antonio:
Sternberg, R. J., & Rainbow Project Collabo- The Psychological Corporation.
rators. (2005 ). Augmenting the SAT through Wenger, E. C. (1998). Communities of practice:
assessments of analytical, practical, and cre- Learning, meaning, & identity. Cambridge, UK:
ative skills. In W. J. Camara & E. W. Kimmel Cambridge University Press.
(Eds.), Choosing students: higher education tools Wenger, E. C. (2000). Communities of practice and
for the 2 1st century (pp. 15 9176). Mahwah, NJ: social learning systems. Organization, 7 , 225
Erlbaum. 246.
Sternberg, R. J., Wagner, R. K., & Okagaki, L. Wenger, E. C., & Snyder, W. M. (2000). Commu-
(1993 ). Practical intelligence: The nature and nities of practice: The organizational frontier.
role of tacit knowledge in work and at school. Harvard Business Review, JanuaryFebruary,
In J. M. Puckett, & H. W. Reese, (Eds.), Mech- 13 9145 .
anisms of everyday cognition (pp. 205 223 ). Williams, W., Blythe, T., White, N., Li, J.,
Hillsdale, NJ: Erlbaum. Gardner, H., & Sternberg, R. J. (2002).
Tan, H., & Libby, R. (1997). Tacit managerial Practical intelligence for school: Develop-
versus technical knowledge as determinants of ing metacognitive sources of achievement in
audit expertise in the field. Journal of Account- adolescence. Developmental Review, 2 2 , 162
ing Research, 3 5 , 97113 . 210.
Wagner, R. K. (1987). Tacit knowledge in Williams, W., Blythe, T., White, N., Li, J.,
everyday intelligent behavior. Journal of Per- Sternberg, R. J., & Gardner, H. (1996). Prac-
sonality & Social Psychology, 5 2 (6), 123 6 tical intelligence for school handbook. New York:
1247. Harper Collins.
C H A P T E R 36
Mica R. Endsley
In thinking about expertise, we often focus fans, and sometimes even to the players on
on skilled physical performance (e.g., the the ice, hockey frequently looks like chaos:
world-class tennis player or gymnast) or sticks flailing, bodies falling, the puck ric-
skilled decision making (e.g., the chess ocheting just out of reach. But amid the
grandmaster). In addition to these aspects mayhem, Gretzky can discern the games
underlying pattern and flow, and antici-
of performance, however, situation aware-
pate whats going to happen faster and in
ness (SA), an up-to-date understanding of
more detail than anyone else in the build-
the world around them, forms a critical ing. Several times during a game youll see
cornerstone for expertise in most domains, him making what seem to be aimless cir-
from driving to aviation to military opera- cles on the other side of the rink from the
tions to medical practice. The characteris- traffic, and then, as if answering a signal,
tics that allow people to develop high levels hell dart ahead to a spot where, an instant
of SA often develop silently alongside more later, the puck turns up. (McGrath, 1997)
observable features like skilled physical per-
formance, even in tasks such as sports that Although undoubtedly Wayne Gretzky pos-
are considered primarily physical in nature. sessed the important physical skills associ-
Take for example the following excerpt ated with the sport of hockey, this article
from a magazine story about Wayne points out that the critical attribute that
Gretzky, an all-time leading hockey scorer placed him head and shoulders above his
who set or tied 49 different National Hockey contemporaries was mental his ability to
League records, including most goals, most understand what was happening in the game
points, and most assists. and to anticipate where the puck would be.
This superior situation awareness allowed
Gretzky doesnt look like a hockey him to be ahead of the game and outmatch
player. . . . His shot is only average or, bigger, faster, and better players.
nowadays, below average . . . Gretzkys gift, Similar stories can be found in other
his genius even, is for seeing . . . To most sports (for example, football, basketball, or
633
634 the cambridge handbook of expertise and expert performance
tennis see also Hodges et al., Chapter 27) their meaning and the projection of their sta-
in which anticipating the actions of ones tus in the near future (Endsley, 1988). This
teammates, ones opponents, and where the definition of SA will be further discussed in
ball is going are key to effective individual terms of the three levels of SA embodied
and team performance. The importance of within it.
situation awareness can be found even in a
relatively straightforward sport such as golf, Level 1 SA Perception
which at first glance might appear to be only
a matter of a mechanical match between The perception of relevant information from
the golf swing and the distance and bear- the environment forms the first level of SA
ing to the hole. Even in this sport situation (see also Rosenbaum, Augustyn, Cohen, &
awareness has a role, however. Expert play- Jax, chapter 29). Without basic perception
ers will walk the course ahead of time to of important information (through visual,
take in key situational features that make auditory, tactile, or other means), the odds
one course play differently from another. of forming an incorrect picture of the situa-
In the 2004 Masters Golf Tournament, Phil tion increase dramatically. In highly complex
Mickelson won by sinking a putt on the last and demanding environments, novices may
hole. Credited as critical to that putt was the have significant difficulty in knowing which
fact that the player immediately previous information is most important or in access-
to Mickelson putted from almost the exact ing needed information in a timely manner
same location. This allowed Phil Mickelson to form Level 1 SA. Even with consider-
to read the hole observing the winds and able expertise, this can be quite challeng-
very slight variations of the grass and grades ing. Jones and Endsley (1996), for exam-
of the slope between the ball and the hole. ple, found that 76% of SA errors in pilots
He consciously worked to develop the best could be traced to problems in perception
SA possible before taking his swing. of needed information (due to either failures
Situation awareness plays an even more or shortcomings in the system or problems
important role in other domains, such as mil- with cognitive processes).
itary operations, piloting, or air traffic con-
trol, where there are many factors to keep Level 2 SA Comprehension
track of and these factors can change quickly Situation awareness involves more than sim-
and interact in complex ways. Effective deci- ple perception of information it also
sion making depends on high levels of SA, demands that people understand the mean-
and thus so does effective performance. In ing and significance of what they have
this chapter I will discuss the ways in which perceived (Level 2 SA). Thus it encom-
SA is critical to expert performance and the passes how people combine, interpret, store,
factors that allow it to improve with the and retain information, integrating multi-
development of expertise in a domain. Stud- ple pieces of information and arriving at a
ies from several different arenas will be pre- determination of its relevance to the per-
sented to highlight the many difficulties that sons goals. This is analogous to having a high
novices have in developing good SA and level of reading comprehension, as com-
to show how SA improves as performers pared to just reading words. Twenty percent
develop expertise. of SA errors in pilots have been found to
involve problems with Level 2 SA (Jones &
Endsley, 1996).
Situation Awareness
Level 3 SA Projection
A general definition of SA, which has been
found to be applicable across a wide variety At the highest level of SA, the ability to fore-
of domains, describes SA as the perception of cast future situation events and dynamics
the elements in the environment within a vol- (Level 3 SA) marks individuals who have
ume of time and space, the comprehension of the highest level of understanding of the
expertise and situation awareness 635
System Capability
Interface Design
Stress & Workload
Complexity
Automation
Task/System Factors
Feedback
SITUATION AWARENESS
Individual Factors
Information-Processing
Goals & Objectives Mechanisms
Preconceptions
(Expectations)
Long-Term Automaticity
Memory Stores
Abilities
Experience
Training
Figure 36.1. Model of Situation Awareness in Dynamic Decision Making (Endsley, 1995 )
situation. This ability to project from current (2) The design of the system interface, deter-
events and dynamics to anticipate future mining which information is available to
events (and their implications) allows for the individual, along with the format of
timely decision making. Experts rely heavily the displays for effectively transmitting
on future projections as a hallmark of skilled information.
performance (Yates & Tschirhart, Chapter (3 ) System complexity, including number of
24). components, inter-relatedness of those
components, and rate of change of infor-
mation, affecting the ability of the indi-
SA Model vidual to keep up with needed informa-
The model of SA in Figure 3 6.1 shows how tion and to understand and project future
SA feeds into decision making and perfor- events.
mance in an on-going cycle (Endsley, 1995 ). (4) The level of automation present in the
Key features of the environment affect how system, affecting the ability of the indi-
well people are able to obtain and maintain vidual to stay in-the-loop, aware of
SA, including: what is happening and understanding
what the system is doing.
(1) The capability of the system for provid- (5 ) Stress and workload that occur as a func-
ing the needed information (e.g., rele- tion of the task environment, the system
vant sensors, data transmission capabil- interface, and the operational domain,
ities, networking, etc.). each of which can act to decrease SA.
636 the cambridge handbook of expertise and expert performance
In addition to these external factors, the people will look for information and how
model points out many features of the indi- they interpret what they perceive.
vidual that determine whether a person will (5 ) Pattern-matching to schema & use of
develop good SA, given the same environ- mental models Mental models and
ment and equipment as others. These fea- schema provide cognitive mechanisms
tures include: for interpreting and projecting events
(1) Perceptual processing and limited atten- in complex domains. These long-term
tion As human capacity for attention memory structures can be used to sig-
is limited, the ability of the individual nificantly circumvent the limitations of
to apply conscious focalized attention to working memory.
all relevant information is severely con-
It is worth noting that though the three
strained in most domains. Information
levels of SA represent increasingly better SA,
from different modalities may be more
they do not necessarily indicate fixed, lin-
readily processed in parallel (Wickens,
ear stages. In a linear progression, a person
1992), thereby somewhat reducing this
needs to see or hear a piece of informa-
effect; however, the limits of attention
tion (Level 1 SA), interpret what it means
for taking in needed information forms a
(Level 2 SA), and then project what will
significant problem for SA.
happen next (Level 3 SA). This is from a
(2) Limited working memory Much of SA data-driven perspective. In reality, however,
requires working memory for storing, a simple 123 data driven progression is not
integrating, and processing perceived an efficient processing mechanism in a com-
information and maintaining the current plex and dynamic system, which is where
internalized model of what is happening. expertise and goal-driven processing come
As working memory is constrained, criti- into play.
cal information may be forgotten or may The model recognizes that many times
not be properly integrated to develop people are goal driven. Based on their goals
Level 2 or 3 SA. or their current understanding or projec-
(3 ) Goal-driven processing, alternating with tions (Levels 2 and 3 SA), individuals will
data-driven processing Goals can be look for data to either confirm or deny their
thought of as ideal states of the system or assessments or will search for missing infor-
environment that the individual wishes mation (look for Level 1 data). This is an
to achieve. In a top-down goal-driven iterative process, with understanding driv-
process, the persons goals and plans will ing the search for new data and new data
direct which aspects of the environment combining to build understanding, as repre-
are attended to in the development of sented by the feedback arrow in the model in
SA. This creates significant efficiencies in Figure 3 6.1.
information processing. Conversely, in a Inherent in any discussion of SA is also
bottom-up, data-driven process, the indi- a notion of what is important. For a given
viduals attention is directed across all individual, SA requirements are based on the
relevant information (perhaps in a fixed goals and decision tasks related to their job or
scan pattern) and will be captured by role. The air traffic controller does not need
salient features or by key information to know everything (e.g., the copilots shoe
that can indicate to the individual that size and spouses name), but does need to
different plans will be necessary to meet know a great deal of information related to
goals or that different goals should be the goal of safely separating aircraft within
activated. An ongoing cycling between his or her sector. A doctor has just as great
goal-driven and data-driven processing is a need for situation awareness; however, the
a key feature underlying SA. things she needs to know about will be quite
(4) Expectations Expectations play an different, dependent on a different set of
important role in SA, affecting where goals and decision tasks. Because the things
expertise and situation awareness 637
Novice Expert
Limited attention Schema of prototypical situations
Limited working memory Mental models of domain
Automaticity of processes
Learned skills (e.g., scan patterns,
communications)
that people need to perceive, comprehend, systems. Proper understanding of the signifi-
and project are by nature domain specific cance of what is perceived will likely be error
(based on their role and tasks), high levels prone as well, as a novice would not have
of SA and thus expertise in one domain will the experience base from which to interpret
not necessarily translate into high levels of those cues. So a new driver, pilot, or power
SA or expertise in another domain (see also plant operator simply remains far behind the
Feltovich, Prietula, & Ericsson, Chapter 4). demand curve in taking in and processing the
information that forms the basis for good SA
in these dynamic environments.
Role of Expertise in Situation
The process is further compromised by
Awareness inefficiency. Novices lack the knowledge of
when each piece of information is really
Based on this model of SA, expertise in a par- most important (see also Feltovich et al.,
ticular domain has a significant role in allow- Chapter 4). Scan patterns tend to be spo-
ing people to develop and maintain SA in the radic and non-optimal. They may neglect
face of high volumes of information trans- certain key information or over-sample
fer and system complexity. This can best be other information unnecessarily. This prob-
explained in terms of two divergent ends lem is not just a matter of needing to learn
of a continuous spectrum, as represented in a set way of taking in information. With-
Figure 3 6.2. out knowledge of the underlying relation-
ships among system components, novices do
Novices not realize what information to seek out fol-
lowing receipt of other information. Seeing
In the most extreme case, a person who is
rising temperatures, for example, does not
completely new to the systems and situa-
lead to the immediate shifting of attention
tions in a particular domain (e.g., a person
to system pressure to determine whether an
learning to drive or fly for the first time)
explosion is imminent.
will be considerably overloaded in seek-
Thus, the prototypical novice is quickly
ing to gather information, understand what
overloaded, inefficient, and error prone in
it means, and formulate correct responses.
developing SA. Decision making and perfor-
Novices will be severely hampered in their
mance are highly compromised as a result.
efforts by both limited attention and lim-
Luckily, development of expertise in a par-
ited working memory. Lacking other mech-
ticular domain significantly reduces these
anisms, they will have to think through each
problems through a number of mechanisms.
piece of data and try to process it in work-
ing memory along with other pieces of data.
Expertise
Reading each gauge, or listening to audio
input and interpreting that data, will impose The model also details how with increasing
a significant burden in even mildly complex experience in a particular domain people are
638 the cambridge handbook of expertise and expert performance
patterns consisting of the state of each of the are performed with little if any conscious
relevant elements for that schema or situa- thought at all. The decrease in demand on
tion type). By pattern matching between the mental resources associated with automatic-
current situation and this schema, people ity of physical tasks provides a boon for SA,
can instantly recognize known classes of situ- leaving more attention and working mem-
ations. Oh, this is just like what happened last ory for attending to information and forming
month. These prototypical situations can be SA. Thus freed up from needing to concen-
learned through direct experience or vicar- trate on the demands of steering and shifting,
iously through formal training or the case the driver can give more attention to detect-
studies and storytelling that are endemic ing potential traffic hazards, for example.
in many professions. This is just like what Automaticity may also be considered in
happened to the aircrew in the Azores acci- relation to more cognitive tasks. With a fixed
dent. A critical feature of this schema is pairing of stimulus to response, even more
that new situations need not be exactly like complex cognitive behaviors can develop to
previously encountered situations to achieve a level of automaticity (e.g., braking follow-
a match. Rather, only a few critical cues ing the detection of tail lights or a red traffic
may be required to lead to a match or a light in front of you, and pressing the gas
near match. pedal when the traffic light turns green). A
Pattern matching to learned schema pro- highly experienced driver may begin to pro-
vides a considerable short-cut for SA and cess these cues and convert even cognitive
decision making. Rather than processing portions of driving to a level of automatic-
data to determine Level 2 or 3 SA (requir- ity with little conscious awareness or atten-
ing working memory or exercising the men- tion, freeing up the mind for other important
tal model), that information is already a part matters such as conversation with a passen-
of the schema and must merely be recalled. ger, thinking about what to have for dinner,
Klein has called this recognition-primed or daydreaming. Such a state can be quite
decision making (Klein et al., 1986). Newell common for experienced drivers operating
has called it the big switch, (Newell, 1973 ). in very familiar environments (e.g., the drive
The critical factors for achieving good SA from home to work each day).
through this mechanism are recognizing the As SA requires awareness of informa-
critical cues that are used for pattern match- tion by definition, SA in such a situation may
ing and having a good stock of such schema be fairly low, even though performance may
in memory. In addition, it appears that some be adequate (no wrecks are occurring). A
people are better at pattern matching than reasonable question might be, is SA really
others, and this attribute has been shown to necessary in that situation or in general? I
be correlated with SA (Endsley & Bolstad, believe the answer is yes. The reason that
1994). SA is always required is that good perfor-
A third relevant characteristic of exper- mance involves not just the known normal
tise is the development of automaticity (see situations, but also the many abnormal sit-
also Feltovich et. al, Chapter 4). Automatic- uations that can and do occur. People oper-
ity is normally considered in terms of phys- ating at this level of cognitive automaticity
ical tasks (e.g., riding a bicycle, steering the are not as attentive to cues that are outside
car while shifting gears and operating the the learned routine. Thus, if there is a new
clutch and fuel peddle simultaneously in stop sign on that well-learned route home,
a standard automobile). In these cases, as people often will run right through without
the physical actions are performed more stopping because they are not alert to this
autonomously, less conscious attention and important situational element. If on a partic-
effort are required of the individual (see ular day they intend to stop at the store on
also Hill & Schneider, Chapter 3 7). Tasks their way home from work, they may drive
that initially completely absorbed the atten- right past the store, seeing the store sign but
tion resources of the individual eventually not triggering the significance of its presence.
640 the cambridge handbook of expertise and expert performance
The low SA associated with cognitive tions are encountered, these mechanisms are
automaticity is likely to negatively affect per- of limited value. Relevant schema will not
formance when the situation falls outside be present and the experts mental model
the bounds of the learned routine. For this may be incomplete for the novel situation at
reason experts in various domains take extra hand. The risk for the expert in such a situa-
steps to guard against the deleterious effects tion may be over-pattern matching trying
of automaticity. Pilots, for example, run pro- to use an inappropriate schema or stretch-
cedural checklists to make sure they check ing the mental model beyond its limits to
each item they are supposed to and do not accommodate the new situation.
lapse into automaticity in checking critical Therefore, SA in very novel situations
information. may be hampered by the same factors that
Finally, with experience in a domain, peo- hamper novices, even for people who are
ple learn many specific skills that are rele- generally considered to have a high level of
vant to efficiently and effectively obtaining expertise in that domain. The experienced
critical information. Pilots learn to commu- pilot who has never encountered or been
nicate with ATC and to scan their instru- trained for a particular anomaly will be chal-
ments to make sure their knowledge of the lenged to process information in working
situation does not get out of date. Air traf- memory to determine what is happening,
fic controllers learn how to scan their radar and may be inefficient in searching for rel-
maps efficiently, based on traffic patterns evant information to solve the problem, in
in the sector. Military officers learn how much the same way as when she was a novice
to gather and disseminate key information pilot (although it is likely that she will not
on the radio, how to listen for information be as bad off as a complete novice).
that is relevant to them while ignoring other Most people do not operate at the level of
information, and where to post troops as lis- novice all the time or expert all the time, but
tening and observation posts to insure that rather move around in between, using com-
key information is gathered and passed on in binations of cognitive mechanisms depend-
the first place. The development of SA is not ing on the situation at hand and the availabil-
just passive, waiting for key information to ity of key constructs (e.g., mental models and
be presented, but rather is an active process. schema).
The information that is available to military In order to understand these differences
pilots, for example, is dependant on how better, we will review a few case studies
they set up and operate their radar (search that explore how SA varies as a function of
patterns, where the radar is focused, modes), expertise.
what frequency they tune their radios to, and
when they request information from others.
Thus, their actions determine what informa- SA and Expertise in Aviation Pilots
tion they will obtain. All of these are exam-
ples of learned skills, specific to each domain, Much research on SA has focused on avia-
that improve with expertise and that con- tion, primarily in the realm of the military
tribute to higher SA among experts. or commercial aviation pilot. These pilots
are typically very carefully selected through
a number of screening tests and highly
Running the Gamut
trained. The training is rigorous anyone
It is worth noting that the mechanisms not meeting the bar gets washed out early
underlying good SA in experts are very on and is ongoing, with most pilots being
domain specific (see also, Feltivich et al. required to demonstrate prowess on a vari-
Chapter 4). They may promote good SA in ety of maneuvers and emergency tasks on an
the learned domain (e.g., brain surgery), but annual or semi-annual basis. Consequently,
will not likely transfer to good SA in another most of the SA research has concentrated
domain (e.g., aircraft piloting). And even on a population that is well along the contin-
within a domain, when very novel situa- uum towards the expertise end of the scale.
expertise and situation awareness 641
In addition to benefiting from the addi- cial and military pilots. These much less-
tional cognitive structures discussed pre- experienced and less-current pilots account
viously, anecdotal evidence indicates that for the vast majority of aviation accidents
experienced military pilots who are very and fatalities; 94% of all U.S. civil avia-
good at SA engage in several practices or tion accidents and 92% of U.S. civil aviation
skills that are beneficial (Endsley, 1989, 1995 ; fatalities involve GA pilots (NTSB, 1998).
Ericsson, Chapter 3 8). First, they think This accident rate has remained fairly sta-
ahead of the aircraft during the flight. They ble for nearly two decades (AOPA, 1997).
are continually engaged in projecting what A 1989 NTSB review of 3 61 GA accidents
might happen, allowing them to be ready concluded that 97% of the probable causes
for events if they materialize. Amalberti and were due to pilot error (NTSB, 1989). Trollip
Deblon (1992) found that these pilots report and Jensen (1991) attribute these pilot error
they spend much of their discretionary time accidents to the following factors, in order
engaged in projection and what-if type of frequency:
thinking. They also engage in extensive pre-
mission planning and briefings that form a r loss of directional control
basis for what they expect to encounter dur- r poor judgment
ing the mission. These expectations can lead r airspeed not maintained
to faster, more accurate SA when they are r poor preflight planning and decision
correct, but also can lead to significant SA
making
errors if they are wrong (Jones & Endsley, r
1996). clearance not maintained
r inadvertent stalls
In a study comparing less-experienced
general aviation (GA) pilots (mean experi- r poor crosswind handling
ence level = 720 hours), more-experienced r poor in-flight planning and decision
airline pilots (mean experience level = making
603 6 hours), and line check airmen, consid-
ered to be among the best commercial pilots Most of these issues indicate problems
(mean experience level = 12,3 70 hours), with pilot SA.
Prince and Salas (1998) noted two key differ- Endsley et al. (2002) conducted a more
ences among the groups. First, the amount in-depth analysis of SA problems in low-
of preflight preparation increased as a func- time GA pilots. To do this, they exam-
tion of experience, with more experienced ined 222 incident reports at a popular flight
pilots focusing on planning and preparation school that contained reported problems
specific to the flight and gathering as much with SA. These reports were stratified into
information as possible about the condi- four different experience levels: (1) students
tions and flight elements. Second, there was working on a private pilots license, which
more focus on understanding and projec- typically requires 40 to 60 flight hours,
tion at higher levels of expertise. GA pilots (2) students working on an instrument pilot
described themselves as passive recipients of qualification whose experience levels are
information with an emphasis on informa- typically 75 hours and more, (3 ) students
tion in the immediate environment (Level working on a commercial pilot license whose
1 SA). Line pilots dealt more at the level experience levels are usually 200 hours and
of comprehension (Level 2 SA) and empha- more, (4) students working on a multiengine
sized their active role in seeking out informa- rating whose experience levels are typically
tion. Check airmen were more likely than 200 hours and more.
the other groups to focus on Level 3 SA, Overall, the least experienced group,
seeking to be proactive. They dealt with a those working on their private pilots license,
large number of details and the complex experienced the greatest proportion of inci-
relationships between factors in this process. dents involving SA problems and the more-
GA pilots are typically far less experi- experienced groups encountered the least.
enced and have less training than commer- The total number of incidents/flight hours
642 the cambridge handbook of expertise and expert performance
Private
Instrument
Commercial
Multi-Engine
40
30 Multi-Engine
20 Commercial
10 Instrument
Private
0
Level 1 Level 2 Level 3
SA SA SA
Figure 36.3. SA Error Causal Factors across Pilot Groups.
in the time period was significantly different dents vigilance and monitoring deficien-
between each student group. cies were noted without these overload
As shown in Figure 3 6.3 , these SA errors problems. This may reflect insufficiently
were also distributed differently across the learned scan patterns, attentional nar-
three levels of SA. Significantly more Level 1 rowing, or an inability to prioritize
SA errors were found in the lower expe- information.
rience level groups and fewer in the most (3 ) Insufficiently developed mental mod-
experienced group. None of the four groups els. Many errors in both understand-
was significantly more or less likely to experi- ing perceived information and project-
ence Level 2 SA errors, but there were signif- ing future dynamics could be linked
icantly fewer Level 3 SA errors in the multi- to insufficiently developed mental mod-
engine student group and significantly more els. In particular, the pilots had sig-
in the other three groups. The lower overall nificant difficulties with operations in
SA error rate observed for the most experi- new geographical areas, including rec-
enced group was due to a decrease in Level ognizing landmarks and matching them
1 and Level 3 SA errors, but not Level 2 SA to maps, and understanding new pro-
errors. cedures for flight, landings, and depar-
Overall, a number of problems were tures in unfamiliar airspace. They also
noted as particularly difficult, leading to the had significant difficulties with under-
SA problems found across this group of rel- standing the implications of many envi-
atively inexperienced GA pilots. ronmental factors on aircraft dynam-
(1) Distractions and high workload. Many ics/behaviors. Pilots at these relatively
of the SA errors could be linked to low levels of experience exhibited prob-
problems with managing task distrac- lems with judging relative motion and
tions and task saturation. This may rates of change in other traffic.
reflect the high workload associated (4) Overreliance on mental models. Revert-
with tasks that are not learned to high ing to habitual patterns (learned men-
levels of automaticity, problems with tal models) when new behaviors were
multitasking, or insufficiently developed called for was also a problem for the
task-management strategies. These less low-experience group. They failed to
experienced pilot groups had significant understand the limits of learned models
problems in dealing with distractions and how to properly extend these mod-
and high workload, els to new situations.
(2) Vigilance and monitoring deficiencies.
Though associated with overload in In a further study, Endsley et al.
about half of the cases, in many inci- (2002) conducted challenging simulated
expertise and situation awareness 643
flight scenario studies with both inexperi- scoring experienced pilots clearly outscoring
enced and experienced GA pilots. The inex- the other groups at all levels.
perienced group included ten private pilots The pilots who scored better on SA
with a mean GA flight experience level of within the novice and experienced groups
109 hours (range 41 to 3 00). The experi- were not necessarily those with the most
enced group included ten flight instructors flight hours. In the novice category, two
with a mean GA flight experience level of of the more experienced GA pilots (180
1790 hours (range 3 00 to 9000). Experi- and 3 00 hours) were rated as among the
enced pilot observers completed a rating seven having low SA, and of the three
form during each flight, indicating whether novice pilots with moderate SA, two were
twelve key expected behaviors were never, fairly low-time pilots (5 5 and 80 flight
sometimes, or always performed, and rating hours). In the experienced category, while
each pilots SA and performance on a 1 to all of the very high-time pilots (more than
7 scale. 1000 hours) were rated as having high SA,
In general, the experienced group outper- so were two pilots with only 45 0 hours.
formed the inexperienced group; however, Of the four experienced pilots who were
closer examination showed that the groups rated as having only moderate SA, one had
were not internally consistent. Therefore, 73 0 flight hours of experience. Flight hours
for purposes of analysis, they were redis- were therefore not significantly predictive of
tributed into four groups, based on observer the group.
SA ratings: In examining the key behaviors per-
formed, those pilots who were scored as hav-
(1) Novice pilots with low SA (scores of 1, ing better SA (in both the novice and expe-
2, or 3 on the overall SA scale), rienced categories) all received much higher
(2) Novice pilots with moderate SA (scores ratings for aircraft handling/psychomotor
of 4 or 5 ), skills, cockpit task management, cockpit
(3 ) Experienced pilots with moderate SA task prioritization, and ATC communica-
(scores of 4 or 5 ), and tion/coordination than did the pilots who
(4) Experienced pilots with high SA (scores were rated as having lower SA. A step-wise
of 6 or 7). regression model, accounting for 91.7% of
the variance in SA score across all pilots,
The mean SA scores of these four groups included aircraft handling/psychomotor skill
are shown in Figure 3 6.4 with the high- and ATC communication and coordination.
644 the cambridge handbook of expertise and expert performance
squad leaders, communicates to squads over- With regard to Level 3 SA, lack of con-
all situation and Commanders intent, solic- tingency planning was identified as a prob-
its information from commanders, monitors lem area, as was failure to project the usage
company net, asks for pertinent intelligence rate of ammunition and supplies. Problems
information, communicates key information with SA regarding the opposing force were
to commander). also found at the projection level, as train-
In addition to improvements in how ers noted that new platoon leaders had diffi-
they distributed their attention, these find- culty projecting a likely enemy COA, as well
ings indicate that with experience, the pla- as their disposition around heavy weapons.
toon leaders are significantly improving the Like the inexperienced pilots, inexperi-
skill sets associated with gathering infor- enced platoon leaders appear to have many
mation from the battlefield environment difficulties with forming good SA. Poor skills
and communicating with others. Commu- for gathering information are particularly
nicating information both up and down the problematic in this domain, as very little
echelon is important not only for building information presents itself otherwise. It was
other team members SA, but also because found that the officers often were not com-
such activities prompt the delivery of rele- municating key information because they
vant information back to the platoon leader. assumed they knew what was going on, or
These skills are critical for the development that the information did not need to be
of SA in the infantry environment. passed on.
In a second follow-up study, Strater, The studies also indicate that novice pla-
Jones and Endsley (2001b) surveyed Forty- toon leaders suffer greatly from not having
three individuals (both officers and enlisted) good mental models or schema. They are
who train new platoon leaders. Issues that quickly overwhelmed by information, are
were rated as a frequent problem by more slow to grasp which information is impor-
than 25 % of the respondents are shown tant, and do not know where to look for
in Table 3 6.1. Problems in communication important follow-up information. Without
posed a major problem for new platoon this schema, they also fared poorly in inte-
leaders, ranging from not requesting infor- grating information to understand its signif-
mation to not communicating key informa- icance and in projection and contingency
tion. In addition, there were significant prob- planning.
lems reported in gathering information on
the combat-readiness status of the opposing
force and ones own troops. SA and Expertise in Driving
Comprehension problems were also
noted as frequent problems for SA, includ- Driving provides a fertile ground for SA
ing failing to assemble bits of informa- research because of its ubiquitous nature
tion together into a coherent picture and in modern society. Drivers of various levels
not specifying alternate courses of action of expertise are readily available (although
(COAs), along with not understanding experience is often confounded with age
task priorities, the impact of load and in the general population). A number of
travel on fatigue, and the importance of researchers have performed studies exam-
soldier positioning to minimize fratricide. ining hazard awareness, a form of Level 3
A continuation of Level-1 problems was SA, in drivers at various levels of exper-
noted with detecting information about tise. The researchers determine how far in
the enemy. Instructors rated understanding advance drivers will anticipate or project a
enemy strengths and weaknesses, likely areas potential road hazard. These studies have
of strategic significance to the enemy, and shown a significant negative correlation
enemy expectations of friendly actions as between the lead time for predicting a haz-
major problem areas for SA in new platoon ard in a simulated driving task or traffic
leaders. video and drivers reported accident rates
expertise and situation awareness 647
(Currie, 1969; McKenna & Crick, 1991, 1994; a concurrent memory task (McKenna &
Pelz & Krumpat, 1974). This finding holds Farrand, 1999). Though the physical vehi-
even after age and miles driven are con- cle control portions of the driving task may
trolled (McKenna & Crick, 1991; Quimby become automatized, the cognitive activi-
et al., 1986). ties involved in predicting hazards appear
In using this approach to study exper- to continue to require significant cognitive
tise in drivers, McKenna and Crick (1991) resources. This is consistent with reports
found that experienced drivers (more than from pilots that projection takes up a
ten years of experience) reacted signifi- considerable portion of their discretionary
cantly faster to hazards than novice drivers time. They furthermore conclude that when
(less than three years of experience), and drivers attention is redirected to other tasks
detected significantly more hazards. When (e.g., cell phone usage), they lose much of
they compared drivers with the same num- the advantage that their superior projection
ber of years of experience, half of whom skills provide.
were considered experts (class 1 police Although this body of research focuses
drivers in the UK) who had completed an primarily on Level 3 SA for the driving task,
advanced driving course that included haz- it provides a rich set of data showing that
ard awareness training, they again found an expertise and SA are tightly linked in ways
advantage for those with more expertise as that extend well beyond simple intake of
defined by this training. information. And whereas other measures
Horswill and McKenna (2004) exam- of driving (e.g., vehicle control skills) have
ined possible reasons for the superior haz- failed to effectively discriminate real-world
ard awareness of experienced drivers. While performance (likelihood of an accident), the
they acknowledge that some of the effect hazard awareness aspect of SA has been
may be due to a change in response bias found to have significant predictive ability,
with experience (i.e., a lower response to the extent that it is now incorporated as
threshold for indicating that something is a part of the driving test in Australia and the
a hazard), they present compelling evi- UK (Horswill & McKenna, 2004). Similar
dence that this explanation cannot explain to the research from the aviation and mil-
all of the differences associated with exper- itary domains, this body of work also finds
tise. They also cite evidence that novice significant problems with efficient intake of
drivers tend to focus closer to the front information in less-experienced drivers. In
of their vehicle, use their mirrors less fre- addition, it appears that SA forms a cen-
quently, and fixate in smaller areas compared tral and conscious task for expert drivers,
with more experienced drivers. In addi- who may be freed up to develop SA by
tion, more-experienced drivers were found improved automaticity on physical tasks and
to better adjust their scanning patterns to enabled by effective mental models for orga-
the road type (Mourant & Rockwell, 1972) nizing and directing information search and
and to have better search models regard- interpretation.
ing where to look for hazards (Underwood
et al., 2002). Horswill and McKenna con-
clude experienced drivers are conducting a Conclusions
more efficient and effective search for hazards
rather than simply lowering their criterion for These studies illustrate the significant
what constitutes a hazard. problems that novices have in building
In addition, McKenna and his colleagues SA. They suffer from poor information-
have demonstrated that hazard awareness management strategies, including poorly
is a cognitively demanding task, and not directed information-seeking behaviors and
one prone to high levels of automatic- scan patterns that will allow them to
ity. The superior hazard-detection perfor- detect the most important information from
mance of experts suffers from exposure to amongst the large number of possibilities.
expertise and situation awareness 649
Their ability to process the information they about the picture with an almost identical
perceive and to understand its significance meaning.
to their goals is also quite limited. They The case for SA, as an area of study,
may often fail to appreciate the impor- is that it provides a focal point around
tance or meaning of even the information which experts integrate the information
that they do acquire. In that their cognitive they gather in order to perform their tasks.
resources are often overloaded with carry- This integration, and future projections
ing out the necessary psychomotor and com- made possible by the mechanisms underly-
munications tasks, they remain significantly ing their expertise, is critical to performance
behind the ball in developing needed SA. in these complex and demanding domains.
By contrast, those with increasing lev- By stopping to consider SA, therefore, a
els of expertise exhibit superior strategies richer picture of expert performance can
for gathering information, both proactively be generated and a more focused approach
and as a follow-up to information already to developing systems that support SA has
received. They show superior abilities to been possible.
grasp quickly the significance of informa- Expertise undoubtedly also involves
tion and to project what is going to hap- other skills, including prowess at physically
pen, allowing them to be in the right place carrying out the needed tasks (for example,
in the sports arena or the battlefield, to throwing the ball, driving the car, or fly-
have solutions ready to execute in the cock- ing the plane) and forming effective decision
pit, or to avoid road hazards while driving. strategies. In addition, the ability to develop
Although the act of understanding and pro- and maintain situation awareness is a signif-
jecting can become quick and often effort- icant contributor to the high levels of per-
less, i.e, a clear comprehension of the situa- formance exhibited by experts in the many
tion springing readily to mind, evidence also domains where dynamic situations must be
shows that experts spend considerable effort mastered and understood in order to per-
at the task of situation assessment. Actively form well.
projecting and planning for contingencies forms
the hallmark of expertise in dynamic decision
making.
Acknowledgements
It should be noted that much of what has
been said here about the role of expertise
on SA has much in common with the more This work was conducted with the support
general literature on the effect of expertise of the Advanced Decision Architectures
on cognitive performance (e.g., Ericsson & Collaborative Technology Alliance spon-
Lehmann, 1996; Feltovich et al., Chapter 4). sored by the U.S. Army Research Laboratory
This is because SA is not a new cognitive under Cooperative Agreement DAAD19-01-
construct that has not been a part of the cog- 2-0009.
nition in other research. Rather, SA has been
integral to many of the domains in which
expertise has been historically studied, even References
if it has not been specifically identified as
such.
Amalberti, R., & Deblon, F. (1992). Cognitive
Historically, rather than evolving from
modeling of fighter aircraft process control: A
psychology laboratories, the term situa- step towards an intelligent on-board assistance
tion awareness arose naturally from domain system. International Journal of Man-machine
experts in aviation who speak of getting SA Systems, 3 6, 63 9671.
or trying to keep SA, dating back as far AOPA. (1997). Nall report: General aviation
as World War II (Press, 1986). Similar terms accident trends and factors for 1996. Fredrick,
have been found to exist in other domains as MD: Aircraft Owners and Pilots Association
well. Air traffic controllers, for instance, talk Air Safety Foundation.
650 the cambridge handbook of expertise and expert performance
Currie, L. (1969). The perception of danger in a training. Transport Research Laboratory Report
simulated driving task. Ergonomics, 12 (6), 841 No. 3 13 . Crowthorne, U.K.
849. McKenna, F., & Crick, J. L. (1994). Developments
Endsley, M. R. (1988). Design and evalua- in hazard perception. London, UK: Department
tion for situation awareness enhancement. of Transport.
In Proceedings of the Human Factors Society McKenna, F., & Farrand, P. (1999). The role
3 2 nd Annual Meeting, Human Factors Society of automaticity in driving. In G. B. Grayson
(pp. 97101). Santa Monica, CA: Human Fac- (Ed.), Behavioural research in road safety
tors and Ergonomics Society. ix. Crowthorne, UK: Transport Research
Endsley, M. R. (1989). Pilot situation awareness: Laboratory.
The challenge for the training community. In Mourant, R. R., & Rockwell, T. H. (1972). Strate-
Proceedings of the Interservice/Industry Train- gies of visual search by novice and expe-
ing Systems Conference (I/ITSC) (pp. 111117), rienced drivers. Human Factors, 14(4), 3 25
Ft. Worth, TX: American Defense Prepared- 335.
ness Association. Newell, A. (1973 ). Artificial inteligence and the
Endsley, M. R. (1995 ). Toward a theory of sit- concept of mind. In R. C. Schank & K.
uation awareness in dynamic systems. Human M. Colby (Eds.). Computer models of thought
Factors, 3 7(1), 3 264. and language (pp. 160) San Francisco, W. H.
Endsley, M. R., & Bolstad, C. A. (1994). Indi- Frelman.
vidual differences in pilot situation awareness. NTSB. (1989). General aviation accidents involv-
International Journal of Aviation Psychology, ing visual flight rules flight into instrument
4(3 ), 241264. meteorological conditions (No. NTSB/SR-
Endsley, M. R., Bolstad, C. A., Garland, D., How- 89/01). Washington, DC: Author.
ell, C., Shook, R. W. C., Costello, A., et al. NTSB. (1998). 1997 U.S. Airline fatalities down
(2002). Situation awareness training for gen- substantially from previous year; general avia-
eral aviation pilots: Final report (No. SATECH tion deaths rise. NTSB press release 2/24/98
02-04). Marietta, GA: SA Technologies. (No. SB 9812). Washington, DC: Author.
Ericsson, K. A., & Lehmann, A. C. (1996). OHare, D. (1997). Cognitive ability determi-
Expert and exceptional performance: Evi- nants of elite pilot performance. Human Fac-
dence of maximal adaptation to task con- tors, 3 9(4), 5 405 5 2.
straints. Review Annual of Psychology, 47, 273 Pelz, D. C., & Krumpat, E. (1974). Caution
3 05 . profile and driving record of undergraduate
Horswill, M. S., & McKenna, F. P. (2004). Drivers males. Accident Analysis and Prevention, 6, 45
hazard perception ability: Situation awareness 5 8.
on the road. In S. Banbury & S. Tremblay (Eds.), Press, M. (1986). Situation awareness: Lets
A cognitive approach to situation awareness: The- get serious about the clue-bird. Unpublished
ory, measurement and application. Aldershot, Manuscript.
UK: Ashgate Publishing.
Prince, C., & Salas, E. (1998). Situation assess-
Jones, D. G., & Endsley, M. R. (1996). Sources of ment for routine flight and decision making.
situation awareness errors in aviation. Aviation, International Journal of Cognitive Ergonomics,
Space and Environmental Medicine, 67(6), 5 07 1(4), 3 15 3 24.
5 12.
Quimby, A. R., Maycock, G., Carter, I. D., Dixon,
Klein, G. A., Calderwood, R., & Clinton-Cirocco, R., & Wall, J. G. (1986). Perceptual abilities
A. (1986). Rapid decision making on the fire of accident involved drivers (No. Research
ground. In Proceedings of the Human Fac- Report 27). Transport and Road Research
tors Society 3 0th Annual Meeting (pp 5 76 Laboratory. Crowthorne, UK.
5 80), Santa Monica, CA: Human Factors
Rouse, W. B., & Morris, N. M. (1985 ). On look-
Society.
ing into the black box: Prospects and limits
McGrath, C. (1997, March 13 ). New York Times in the search for mental models (No. DTIC
Magazine. #AD-A15 9080). Atlanta, GA: Center for Man-
McKenna, F., & Crick, J. L. (1991). Hazard per- Machine Systems Research, Georgia Institute
ception in drivers: A methodology for testing and of Technology.
expertise and situation awareness 651
Shattuck, L., Graham, J., Merlo, J., & Hah, Strater, L. D., Jones, D. G., & Endsley, M. R.
S. (2000). Cognitive integration: An inves- (2001b). Analysis of infantry situation aware-
tigation of how expert and novice com- ness training requirements (No. SATech 01
manders process battlefield data. In M. E. 15 ). Marietta, GA: SA Technologies.
Benedict (Ed.), Proceedings of the Fourth Trollip, S. R., & Jensen, R. S. (1991). Human factors
Annual Federated Laboratory Symposium on for general aviation. Englewood, CO: Jeppesen
Displays and Interactive Displays Consor- Sanderson.
tium. Adelphi, MD: U.S. Army Research Underwood, G., Chapman, P., Bowden, K.,
Laboratory. & Crundall, D. (2002). Visual search while
Strater, L. D., Endsley, M. R., Pleban, R. J., driving: Skill and awareness during inspection
& Matthews, M. D. (2001a). Measures of of the scene. Transportation Research, Part F(5 ),
platoon leader situation awareness in vir- 8797.
tual decision making exercises (No. Research Wickens, C. D. (1992). Engineering psychology
Report 1770). Alexandria, VA: Army Research and human performance (2nd ed.). New York:
Institute. Harper Collins.
C H A P T E R 37
There are a variety of patterns of change ror reading, early practice involves mental
in activation seen in skill acquisition experi- rotation of the letters, and late performance
ments, depending on the nature of the task, involves recognition of the rotated word and
practice levels, task difficulty, and nature recall of the meaning without slow algorith-
of control conditions (see Kelly & Garavan, mic rotation).
2005 , for a recent review). The patterns
include: brain activity decreases in the con-
Overview of Brain Anatomy
trol network (see next paragraph), mainte-
and Functional Change
nance of activity in the perceptual and motor
areas, and occasionally increases of brain
activity or shifts in activity to new regions. To understand the changes of the brain with
Throughout the chapter we will review skill acquisition it is useful to take a struc-
examples of changes in brain function (that tural/functional view of the brain. We will
is, changes in the activity patterns), and we provide a short overview of the architecture
will relate these patterns to the task parame- of the brain (i.e., the structure or anatomy),
ters, behavioral responses, and level of expe- that will illustrate the hierarchical nature
rience in performing the task. of perceptual processing by reviewing the
The most common pattern is reduced visual system (specifically the occipitotem-
control network activation with maintained poral pathway), and then comment about
perceptual motor activity (note Kelly & brain misconceptions.
Garavan, 2005 , refer to this a processing
efficiency change). The control network is The Bare Basics of the Cerebrum
a group of brain areas that are believed to The cerebrum of the brain is divided into five
work together to scaffold learning when per- lobes; occipital, parietal, temporal, insular,
forming a new task. This network (see Net- and frontal. Generally speaking,2 the func-
work of Specialized Processing Regions) tion of each lobe is the following:
comprises a set of discrete cortical areas
controlling goal processing, attention, and 1. The occipital lobe is involved in visual
decision making. One would expect to see processing.
those (learning-related) areas active early in 2. The parietal lobe participates in visual
practice and then decrease activity, possi- processing by coding spatial information,
bly even dropping out, as skill is acquired. and it is involved in attentional con-
This is the expected pattern for percep- trol and somatosensory processing (bodily
tual motor tasks, in which stimulus proper- sensation).
ties and (motor) response properties remain 3 . The temporal lobe also contributes to
consistent or unvarying throughout perfor- visual processing at the level of object
mance (see Controlled and Automatic Pro- formation (and face processing) and is
cessing during Learning section). A second involved in coding auditory and verbal
pattern is one of increased cortical tissue information and memory storage.
devoted to the task after very long periods
4. The insular lobe is involved in emotional
of training (e.g., expanded hippocampus for
processing, taste, and learning.
taxi drivers). And a third pattern is func-
tional reorganization. As the term implies, 5 . The frontal lobe is involved in executive
there is a reorganization of active brain areas; function, reasoning, effort and emotional
that is, different brain areas are active in dif- coding, conceptual information and rules,
ferent stages of learning. The engagement motor control, speech, and smell.
and disengagement of brain regions reflects
the fact that unique regions are involved in
Network of Specialized
the various types of processing. As a skill is
Processing Regions
acquired, different strategies are used, and
therefore new areas become active to per- The brain is a network of hierarchically orga-
form the underlying processing (e.g., in mir- nized specialized modules. If one moves a
656 the cambridge handbook of expertise and expert performance
Figure 37.2 . Major parts of the Control Network supporting skill acquisition
and cognitive control. These areas include dorsolateral prefrontal cortex
(DLPFC), anterior cingulate/pre-supplementary motor areas (ACC), posterior
parietal cortex (PPC), anterior insula, and thalamus. These areas appear to be
domain general and decrease in activation as automaticity develops.
centimeter in the brain, typically the tissue body) functions. As information processing
does something distinctively different than flows to higher levels, its spatial localization
the neighboring region. reduces and object specialization increases
The brain has many domain specific rep- (e.g., moving from lines covering 0.1 visual
resentational areas connected in a quasi- area in V1 [the initial visual processing region
hierarchical fashion. There are an estimated located in the occipital lobe] to faces of
5 001000 specialized processing regions individuals anywhere in the visual field in
(Worden & Schneider, 1995 ). A complex the fusiform cortex [a later visual process-
process, such as visual processing, occurs ing region located in the temporal lobe]).
in over thirty distinguishable processing The brain contains a small number of
regions (Felleman & Van Essen, 1991), inclu- domain general control areas that appear to
ding those for detecting lines, colors, shapes, be involved in many tasks and that modu-
structure (e.g., houses, faces), motion, and late cortical activity (see Schneider & Chein,
spatial relationships. These representation 2003 , for details). There is initial evidence
areas appear to be quasi hierarchically con- to suggest some specialization of function
nected. There are reciprocal connections within the different areas. However, these
between these regions, allowing informa- areas work as a tightly coupled unit (e.g.,
tion flow to be both bottom-up and top- typical correlations between activation in
down. Information is coded in the pattern respective areas range from 0.80.95 ). The
of activity, with any one region encoding major areas are listed below, and a subset is
many exemplars and types of stimuli (Ishai, shown in Figure 3 7.2:
Ungerleider, Martin, Schouten, & Haxby, r attentional control posterior parietal
1999). These representation areas include cortex (PPC)
input (visual, audition, somatsensory, gusta- r process monitoring, decision making,
tory) and output motor areas. These areas
conflict management anterior cin-
can be mapped with nearly millimeter pre-
gulate cortex/pre-supplementary motor
cision, showing retinotopic (a retinal [eye]
area (ACC/pre SMA)
based coordinate map in the visual [occip- r goal processing and tasks switching dor-
ital] cortex), tonatopic (a map of acousti-
cal frequencies in the auditory [temporal] solateral prefrontal cortex (DLPFC)
r emotional processing amygdula
cortex), somatotopic (a map of bodily sen-
sation in the associative [parietal] cortex), r episodic coding of association para-
and motor (map of muscle control of the hippocampus and hippocampus
brain changes in the development of expertise 657
and holistic processing, and there is left/right Brain Activation in Working Memory Tasks
laterality of that processing, but there is little and Dual-Tasks, will extend Section one
special status for the global feature of being into working memory and dual-task para-
on the left or right. digms. Themes 14 will be illustrated in
Next we will preview the six patterns of cognitive and motor laboratory tasks that
learning-related brain changes that will be involve a range of training, from minutes to
elaborated further in the chapter. sessions. Finally, Section three, Perceptual-
Motor Learning and Expertise, will review
themes in learning literature, brain visual processing, motor learning, kinemat-
processing, and brain structure ics, navigation, reading, and music training.
There are six themes evident in the learning Themes 46 will be illustrated primary
literature (Kelly & Garavan, 2005 ; Poldrack, through the study of real-world skills that
2000; Schneider & Chein, 2003 ; see also develop over many years of practice. The
Feltovich et al., Chapter 4). First, learning is themes will be stressed throughout the text;
localized and very specialized, with differ- however, we recommend that the reader
ent portions of the brain showing dynamic briefly review them (see above) before
change and location change depending on proceeding.
the task. Second, in general, learning and
processing occur in the same cortical loca- Controlled and Automatic Processing
tions. Third, learning can produce both during Learning
increases and decreases in the areas of acti-
vation, influencing both the richness of the Skill acquisition involves creating represen-
representation and the efficiency of process- tations and altering control routines. In this
ing. Fourth, in some tasks there is a reor- section, we will focus on the control
ganization of the task that involves differ- routines. Skill acquisition involves priority
ent brain regions when alternate strategies4 learning, learning to code stimuli accor-
are used (e.g., shifting from performing ding to their importance (see Feltovich, Pri-
a math algorithm to using recall to pro- etula, & Ericsson, Chapter 4, on Exper-
duce an answer). Fifth, behaviorally rele- tise Involves Selective Access of Relevant
vant objects and other stimuli are uniquely Information). Through this learning, criti-
processed by experts (i.e., faces for every- cal stimuli (determined by learned priority)
one, birds for bird watchers, ballet move- ultimately transmit information to stimulus
ments for professional ballet dancers). Sixth, processing regions and trigger control area
learning can produce detectable morpho- responses. For example, in the cocktail party
logical changes, such that extended train- effect, hearing ones name in a nearby con-
ing can enlarge the amount of grey matter versation can draw attention, despite com-
dedicated to processing the type of informa- peting background noise and focused atten-
tion that has been trained. The remainder tion on the current conversation. This is
of the chapter will illustrate these themes because ones name has become a high pri-
in three major sections. Section one, Con- ority stimulus that one learns to automati-
trolled and Automatic Processing during cally orient towards over time. (See Proctor
Learning, will introduce the reader to the & Vu, Chapter 15 ; Ericsson, Chapter 3 8, on
dual processing account of skill acquisition, traditional models of skill acquisition, such
by briefly reviewing the longstanding behav- as Fitts & Posner, 1967).
ioral effects and the theorys recent exten- Many of the practice effects of learning
sion into characterizing the underlying neu- can be interpreted within the traditionally
ral substrates. Within this section, themes 1 behavioral framework of automatic process-
and 2 will be illustrated in cognitive labo- ing and controlled processing (Schneider &
ratory tasks that involve several sessions of Shiffrin, 1977; Shiffrin & Schneider, 1977;
training. Section two, Practice Effects on Feltovich et al., Chapter 4, on Expertise
brain changes in the development of expertise 659
Figure 37.3. Activation early and late in a consistent search task. The dark areas
(and dark areas with bright centers) illustrate regions that are active. The ellipses
highlight the frontal and parietal areas. There is a substantial reduction in
activation in these areas as skill is acquired and more automatic processing
develops. The visual areas (lower right parts of the images) show continued
activation early and late in practice with the sensory processing continuing even
while controlled processing decreases.
Involves Automated Basic Strokes), with and other trials the opposite) the reaction
recent extensions to cortical processing times are over 200ms per comparison (Fisk
(Schneider & Chein, 2003 , CAP2 model). & Schneider, 1983 ). In contrast, if the cat-
Novice performance is assumed to involve egories are consistently mapped (CM) (e.g.,
high-effort use of the domain general net- always responding to animals and ignoring
work of control processing through atten- vehicles), reaction time is uninfluenced by
tional routines. The same pieces of brain the number of categories compared (e.g.,
tissue perform key tasks such as attention, slope 1ms) once there is extensive practice.
comparison, and decision across different The novice search task (both CM and VM)
representation modalities (e.g., visual, audi- involves goal processing, memory, attention,
tory or motor). Controlled processing typi- decision making parts of the control net-
cally occurs in novel or varied tasks (see var- work, representation patterns of the visual
ied mapping in next paragraph), when one input, and memory of the targets. The prac-
must compare stimuli and release responses ticed CM search task involves automatic pro-
based on information stored in working cessing in which modules transmit represen-
memory. Automatic processing typically tations to the next stage without involve-
occurs for well-practiced, consistent tasks ment of the control network (Figure 3 7.3 ).
where the responses can occur with lit- Controlled and automatic processing
tle effort. These are representation-specific
love show dramatically different and comple-
operations (e.g., the visual mapping mentary processing benefits and disadvan-
to love will not transfer to the reversed tages (see Schneider & Chein, 2003 ). Con-
phonemes [sounds] of the word love). trolled processing has characteristics that
The classic example of a controlled pro- allow rapid acquisition, easy alteration of
cess is visual search of a varied set of process,5 and modification of memory, but
objects (see Schneider & Shiffrin, 1977). For it is a slow, serial, high-effort mode of pro-
example, in a category search task, subjects cessing that deteriorates under high work-
respond if a word is a member of a category load and stressors (e.g., sleep deprivation,
(e.g., respond to animal or vehicle). If the stress, and alcohol). Automatic process-
memory set alters from trial to trial and there ing has the complementary weakness and
is a varied mapping (VM) (e.g., responding to strengths. Automatic processing shows slow
animals on some trial and ignoring vehicles, acquisition over hundreds of trials, is difficult
660 the cambridge handbook of expertise and expert performance
to alter, does not modify memory, is fast and consistency and practice can modulate cor-
parallel, is low effort, and is robust to high tical activity (Chein & Schneider, in press;
workload and stressors. The CAP2 model Jansma, Ramsey, Slagter, & Kahn, 2001)
(for an in-depth coverage of the model, when acquiring skilled performance. Jansma
we refer the reader to Schneider & Chein, et al. (2001) trained a working-memory con-
2003 ) provides a connectionist implementa- sonant item-recognition task under consis-
tion of controlled and automatic processing tent (i.e., target and distractor items are
and interprets interactions in the control and always from distinct sets) and variable con-
data network and likely brain areas involved ditions (i.e., targets and distractor sets are
in each function. The transition from con- redefined on every trial by selection from a
trolled to automatic processing occurs in this common pool of items) and found that prac-
model as the data modules become capable tice produced reductions in activity only for
of transmitting their output without media- the consistent task.
tion by the control system. The control network appears to be do-
From a neural perspective, the develop- main general. In a study by Chein, McHugo,
ment of automatic processing should result, and Schneider (in prep), early in practice
with practice, in reductions in the same both an auditory task and a visual spatial
regions that support controlled processing task had activated a network of common
(Chein & Schneider, in press; Schneider & areas in frontal, cingulate and parietal cor-
Detweiler, 1988; Kelly & Garavan, 2005 ; tex. In addition there were task specific areas
Jansma et al., 2001; Schneider & Chein, (occipital cortex for the visuospatial task and
2003 ). To determine if a specific group of brain temporal lobe for the auditory task) that
areas are commonly modulated as a func- were unique to the type of task being per-
tion of practice, a meta-analysis of brain- formed. In both tasks, with consistent prac-
imaging studies that utilized practice was tice the control network dropped out. It is
performed (Chein & Schneider, in press; important to note that at the millimeter level
Schneider & Chein, 2003 ). Study selection of cortical tissue, very different tasks initially
was restricted to consistent tasks in which activated the same cortical control areas to
novice performance was contrasted to prac- provide scaffolding to perform and learn the
ticed (i.e., minimum of ten minutes) perfor- task. The same tissue (i.e., the control net-
mance of an explicitly acquired skill. The work brain regions) was controlling different
meta-analysis revealed activity in a reliable representation areas, each of which was later
network (that we refer to as the control net- able to perform its task automatically.
work) of brain regions. This putative con- In a separate study, a paired-associate
trol network includes lateral frontal, medial task, in which all training was kept consis-
frontal, posterior parietal cortices, and the tent, demonstrated engagement of the same
thalamus. According to the theory, such a network and subsequent practice-related
network should be engaged in novel learn- reductions for both verbal and non-verbal
ing situations and should reduce activity or (i.e., novel shapes that are difficult to assign
drop-out as a function of extensive consis- verbal labels) associates (Chein & Schneider,
tent practice. in press).
Automatic processing occurs as the In varied mapping tasks, activation resem-
control-network regions are released, leav- bled initial, untrained levels, both early
ing task-specific processing regions engaged and late in practice (Chein, McHugo, &
to support task performance (Chein & Schneider, in prep). There was no evidence
Schneider, in press; Schneider & Chein, of practice-related increases of activation
2003 ; see also, Feltovich et al., Chapter 4 or reorganization. These studies demon-
on Expertise is Limited in its Scope and strate that consistent practice, but not vari-
Elite Performance does not Transfer). Two able practice, results in processing effi-
fMRI studies have varied consistency of ciency or activity reduction, in the same
learned associations and demonstrated that regions that support untrained performance.
brain changes in the development of expertise 661
Furthermore, some ancilliary task-related (see prior section), parallels can be made
processing regions (such as motor cortex if in terms of behavioral improvement and
the task requires a button press response) neural response. Furthermore, these tasks
continue to activate without regard to con- were trained under a range of practice, both
sistency of practice. briefly (i.e., for 3 0 minutes) and exten-
In summary, the shift from controlled sively (i.e., several weeks), and produced
to automatic processing can occur on a similar results.
relatively short time scale.6 The changes
in behavioral performance (i.e., increases in basic working memory (wm) overview
accuracy and/or decreases in reaction time)
The working memory model is composed
are associated with underlying changes in
of three major parts, an executive control
brain processing (i.e., decreases in brain
system and two slave processing systems,
regions responsible for learning and control).
one involved in visuospatial information and
This shift only occurs through practice of
the other involved in phonological infor-
consistent tasks. Furthermore, brain regions
mation (Baddeley & Hitch, 1974). Distinct
involved in motor and perceptual processing
from both short-term and long-term mem-
remain active even with practice, as ongo-
ory, this system is involved in processing and
ing processing in these areas is necessary for
maintaining information over brief delays
task performance. Finally, automated consis-
and in manipulating information. WM tasks
tent tasks show no evidence of functional
have three phases, encoding of stimuli, delay
reorganization or practice-related activation
(where no stimulus is present and the asso-
increases in the control network regions.
ciated codes must be maintained and/or
The next section will explore prac-
manipulated), and retrieval (of codes or
tice effects in working memory and dual-
transformed codes) to enable responding.
task paradigms. In these studies, outcomes
are mixed; that is, practice-related acti-
vation increases, practice-related activation practice and working memory
decreases, and as well as practice-related Working memory tasks have been frequently
activation increases and decreases are all studied using behavioral measures and in
reported. Although the controlled and auto- brain imaging (Baddeley, 2003 ; Cabeza &
matic processing perspective focuses on Nyberg, 2000), but only recently have stud-
practice-related brain-activation decreases, ies investigated the impact of training in
practice-related activation maintenance, as these tasks. Imaging studies have reported
well as practice-related activation reorgani- practice-related increases and decreases in
zation, are consistent with the theory in neural activity, as well as shifts in activation
cases of variable tasks that require manip- location (for review, see Kelly & Garavan,
ulation of working memory (Jansma et al., 2005 ; Poldrack, 2000). Increases may result
2001; Schneider & Chein, 2003 ). from strengthening of existing activation or
from spatial expansion. Decreases in activity,
conversely, may reflect the reduction in the
Practice Effects on Brain Activation strength of existing activation or a reduc-
in Working Memory Tasks and tion in spatial extent. Whereas increases are
Dual-Tasks believed to reflect the buildup of cortical
representations, decreases are believed to
In this section we will look at practice result from greater neural efficiency, where
effects in cognitive tasks that employ online- only a subset of the initial neurons continue
learning or working memory and tasks that to respond to a particular stimulus or sup-
involve learning to process multiple sub- port the performance of a particular task.
tasks concurrently, that is, dual-task learning. A shift in the location of activity reflects
Although these studies were not designed a reorganization of regions used to support
to investigate automatic processing shifts performance. The reorganization may reflect
662 the cambridge handbook of expertise and expert performance
task, it is also a sequence task (in which typically given response priorities because
participants had to perform both forward one task is designated as the primary task
and backward sequences), and the neural while the other is designated the secondary
response was similar to sequence learning task. The PRP effect is immutable with
paradigms (which typically contrast new [or practice unless performers are allowed to
transformed] and old sequences [Hazeltine respond without regard to task designation
et al., 1997; Honda et al, 1998]) and the (remove the primary and secondary task-
finger opposition paradigm ([Karni et al., response designation). Non-PRP dual-tasks
1995 ]; see the motor learning section). This (ISI > 1 second) do not have immutable con-
suggests a distinction between motor and current performance costs because practice
cognitive tasks in which motor task practice results in the speeding of processing, so that
leads to functional increases of activation, the performer can learn to respond to each
whereas cognitive task practice leads to func- task quickly in isolation, as long as there are
tional decreases reflecting neural efficiency. no structural limitations (such as having only
Cognitive task decreases of activation are not one response finger) (see Meyer & Kieras,
restricted to working memory tasks per se; 1997).
decreases were also found in the studies of There are a variety of patterns of activa-
automatic and controlled processing. tion seen when contrasting single- and dual-
task conditions. This is not surprising when
one considers the range of paradigms that are
dual-task methodology referred to as dual-tasks. A dual-task can be
An important aspect of skilled performance composed of subtasks that involve concurrent
is the ability to perform multiple tasks component tasks (such as PRP tasks) or tem-
at one time. Dual-task performance is an porally separated component tasks. The sub-
area of interest because of its potential tasks are sometimes simple tasks, such as
to inform understanding of attention lim- detection or discrimination, or more com-
itations, learning, and executive process- plex tasks, such as spatial rotation or reading
ing, both through behavioral investigations comprehension.
and more recently in brain-imaging stud- Generally speaking, dual-tasking typi-
ies. Dual-tasks paradigms employ the per- cally involves more effort and time sharing
formance of two or more tasks in close tem- than the single task performance. And moti-
poral proximity (see Meyer & Kieras, 1997; vation and practice history (both single- and
Pashler, 1994; Schneider & Detweiler, 1988). dual-task) must be considered when com-
Simultaneous performance that is untrained paring single- and dual-task performance.
typically results in decreases in accuracy and All of these factors must be interpreted with
increases in reaction time relative to isolated caution when reviewing dual-task effects in
task performance. Although all dual-tasks a behavioral study, and the issues are further
report this concurrent performance cost at complicated in a brain-imaging study. For
some point in training, a distinction is made example, in fMRI, threshold selection can
between dual-tasks that employ short, vari- cause an area to appear active only under
able time lags (i.e., interstimulus interval, dual-task performance, when in fact it is
ISI) between component tasks, known as active, but to a lesser extent, under single-
psychological refractory periods (PRP; short task conditions. In addition, a region that is
ISI < 3 00 msecs), and those dual-tasks that below threshold because of lack of statistical
use relatively longer, fixed time lags (typic- power may in fact be above threshold due to
ally > 1 second). In PRP tasks, reaction time greater demand for the region in a dual-task
increases for the second task response in the setting. Alternatively, time sharing between
short-lag condition, relative to the long-lag two areas while dual-tasking can result in
condition, because short ISIs force the per- an area that was active in a single task
former to use overlapping task processing,8 to drop below threshold during dual-task
which is a limited resource (that is, for short performance. Typically, however, areas
ISI RT2 > long ISI RT2). PRP tasks are also engaged in single tasks are still active in
664 the cambridge handbook of expertise and expert performance
when there was a concurrent performance single-task training outside of the scanner,
cost (i.e., only during the shorter ISI). How- most regions decreased in activity except for
ever, recent work by Erickson et al. (2005 c) dorsolateral prefrontal cortex. An increase
suggests that right IFG activity is not spe- in left DLPFC was associated with mixed
cific to dual-task interference but, alterna- single-task (single-task trials interspersed
tively, is associated with preparing to make with dual-task trials) performance for par-
multiple responses (whether in the context ticipants that received behavioral training.12
of a single or dual-task) and not actual coor- Bi-lateral DLPFC activity was found for the
dinated performance. This area was engaged dual-task condition. These areas were not
by single-task performance when comparing significantly active at session one. Erickson
mixed single-task trials (i.e., interspersed et al. (2005 b) regard the training increase as
with dual-task trials) to pure single-task tri- a shift in processing, where DLPFC begins
als (i.e., exclusively single-task trials) in a to support task coordination as a result of
mixed event related design.11 Furthermore, training.
this area was not engaged by dual-task perfor- Hill and Schneider (2005 ) found
mance, suggesting that the area is sensitive widespread decreases in activity as a result
to preparing to perform multiple tasks as of training an object-word visual search
opposed to the actual performance of mul- dual-task and a pattern-letter visual search
tiple tasks. dual-task. These decreases included pre-
Two of the studies employing psycho- frontal areas, and no areas were found to
logical refractory period paradigm (Herath increase activity with training, suggesting
et al., 2001; Szameitat et al., 2002) found processing efficiency of performance. The
dual-task specific prefrontal activity that training decreases were predicted based on
was spatially distinct from the component- prior work (Chein, McHugo & Schneider,
task activity, which also activated prefrontal in prep; Chein & Schneider, in press;
regions. The inconsistent dual-task specific Schneider & Chein, 2003 ) demonstrating
prefrontal activation may potentially be practice-related reductions when develop-
attributable to whether concurrent perfor- ing automatic processing for consistent tasks
mance results in a performance deficit and (see Controlled and Automatic Processing
to the level of component task complexity. during Learning section).
Although differences in task complexity and Differences in activation dynamics
performance appear to factor into prefrontal between the studies potentially reflect
activity, the extent of the impact is a matter differences in task design and training his-
of speculation. Further research is necessary tory. Hill and Schneider (2005 ) extensively
to elucidate the role of prefrontal cortex in trained all single-tasks prior to any scanning,
task coordination and interference. effectively scanning changes related to nave
versus experienced dual-task performance
(participants were unpracticed on dual-task
dual-task practice effects performance at scan one). Conversely, scan
Few studies have investigated the effects of sessions of Erickson et al. (2005 a) reflect
practice on dual-task related neural activ- untrained task performance (both single
ity (Erickson et al., 2005 a; Erickson et al., and dual-task) compared to trained task
2005 b; Hill & Schneider, 2005 ). Erickson performance. The difference in the direction
et al. (2005 a) found that untrained dual- of DLPFC may reflect different assessment
task performance engaged the same areas as points of learning. In addition, the Erickson
the component tasks (letter and color dis- et al. (2005 a, 2005 b) dual-task involved
crimination), but to a greater extent. This simultaneous concurrent letter and color
study is consistent with those that do not discrimination (ISI = 0), where the Hill
report specific prefrontal dual-task process- and Schneider (2005 ) dual-task involved
ing regions (Adcock et al., 2000; Bunge continuous rapid visual search (nine search
et al., 2000). After extensive dual- and locations changing five times per second);
666 the cambridge handbook of expertise and expert performance
however, simultaneous targets did not occur is developed normally through experience,
in this design (targets could appear at any and music skill, which occurs through inten-
time during the minute search window as tional training.
long as they occurred at least two seconds
after the prior target). Although participants
were instructed to give equal task priority Perceptual-Motor Learning
in both studies, Erickson et al. (2005 c) and Expertise
subjects tended to respond to the color
discrimination task first. PRP interference Much of human skill acquisition and exper-
occurs when one task is instructed to be tise involves perceptual-motor learning (see
given response priority or if this strategy Chapter 29 on perceptual-motor expertise
is employed by the performer (Meyer & by Rosenbaum, Augustyn, Cohen, & Jax).
Kieras, 1997). The DLPFC activity may Learning can occur at many levels within
reflect interference related to strategy the processing hierarchy, depending on the
choice of responding in a fixed order. nature of the task. In the case of vision, train-
In summary, dual-tasks that use a psycho- ing has resulted in improved ability to per-
logical refractory period design elicit infe- form discrimination in various tasks (texture
rior frontal activation under conditions of segregation, motion discrimination, line ori-
high interference, when the ISI is short. The entation, etc.). In some cases, such as line
neural effect of practice on these designs orientation, learning of the trained orienta-
is an unexplored area; however, since prac- tion is specific to the trained location. The
tice does not attenuate behavioral interfer- failure to transfer learning to other loca-
ence, inferior frontal cortex would likely tions argues that the learning occurs early
maintain activation with training. Dual-tasks in the processing stream (that is, V1, the
with longer, fixed ISI (non-PRP tasks) gen- locus of initial visual processing in occipi-
erally do not report dual-task specific pre- tal cortex) where receptive fields are small,
frontal (or otherwise) activity, suggesting tightly tuned to a specific orientation, and
no general locus for task coordination, an topographically organized. This is remark-
executive process. These tasks tend to have able because early visual processing regions
complex subtasks, and therefore it may be were traditionally considered fixed in the
difficult to find particular areas engaged adult brain. The specificity of learning (i.e.,
in task coordination processing. Dual-tasks does training transfer to untrained location,
that contrast the effects of practice have quadrant, or eye) has demonstrated that per-
generally found decreases for inital-task- ceptual learning can occur at different levels
engaged brain regions; however, one study of the processing hierarchy. The specificity
reports a DLPFC increase. Practice must effects, however, have not always been con-
be employed in more studies to determine sistent (i.e., sometimes transfer occurs and
the conditions under which practice-related other times it does not), even when train-
increases would arise. For a discussion ing the same type of perceptual task. Note,
of skilled individuals engaging in high- in a hierarchy of areas, learning can occur
level real-word multiple-task environ- at multiple potential levels that show differ-
ments (such as pilots), see Durso and Dattel ential transfer (e.g., attend to the lower left
(Chapter 20). oriented line or an oriented line anywhere in
Previous sections have looked at perfor- the visual field).
mance and brain changes in laboratory cog- The nature of attention and task diffi-
nitive tasks under conditions of short to culty can influence the specificity of what
moderate amounts of training. The final sec- is learned in a discrimination task. Accord-
tions will look at performance and brain ing to Reverse Hierarchy Theory (Ahissar &
changes that occur over longer amounts Hochstein, 2004), difficult tasks (short vs.
of practice for basic perceptual and motor long ISI and/or fine vs. course line discrim-
skills. Examples are face processing, which ination) are learned with a high degree of
brain changes in the development of expertise 667
specificity. According to this view, learning cessing than in any other domain (Haxby,
is driven by attention, with learning occur- Hoffman, & Gobbini, 2000). Neuroimag-
ring first at the top of the processing hier- ing studies implicate a visual area in the
archy, then proceeding to the lower levels. right mid-fusiform gyrus (though sometimes
Skill on a specific discrimination task also bilateral) that increases its activity when
constrains learning such that the perform- faces are detected (Kanwisher, McDermott,
ers have improved their signal-to-noise ratio & Chun, 1997; McCarthy, Puce, Gore, &
at the lower processing levels and can per- Allison, 1997). This area has been termed
form difficult discriminations. Less-skilled the Fusiform Face Area (FFA) because
performers have poor signal-to-noise ratios although it is responsive to other objects,
at low levels and use high-level representa- it is most responsive to faces (Kanwisher
tion, providing more generalization and per- et al., 1997; Haxby et al., 2000). Imag-
haps faster learning, but without the very ing studies have demonstrated this greater
high levels of performance. Note, in a hier- response to faces, without regard to for-
archical attention network (see Olshausen, mat (photos and line drawing) and without
Anderson, & Van Essen, 1993 ), proficiency regard to familiarity (i.e., not more respon-
allows the performer to determine the opti- sive to famous faces) (Gorno-Tempini &
mal level of processing given the difficulty of Price, 2001; Gorno-Tempini et al., 1998).
the discrimination. FFA activation is greater for faces than for
Learning to perceive phonemes, faces, hands, animals, objects, and scenes (Aguirre,
chess patterns, music, or radiology images Singh, & DEsposito, 1999; Haxby et al.,
all involve multi-level perceptual learning. A 2000; Ishai et al., 1999; Kanwisher et al.,
simple illustration is the inability even to find 1997; Kanwisher, Tong, & Nakayama, 1998;
word boundaries in a spoken language that Yovel & Kanwisher, 2004). Imagined faces
is unfamiliar. With experience, phonemes, also elicit activation in this area (Kanwisher
words, and phrases become units of process- & OCraven, 2000). It is undisputed that
ing (see Feltovich et al., Chapter 4, Exper- right FFA responds greatest to faces; how-
tise Involves Larger Cognitive Units). Imag- ever, it has been suggested that this area
ing data show changes in cortical processing is not specifically modulated for face pro-
at multiple levels of processing as percep- cessing per se, but for processing visual
tual discrimination improves (Karni & Sagi, items for which an individual has devel-
1991). oped high levels of expertise and familiarity
In the following we will see how high- that can be categorized on the individual
level visual areas represent and process level13 (Tarr & Gauthier, 2000; Feltovich
objects in the temporal lobe and other brain et al., Chapter 4, Expertises Is Limited
regions. Visual processing begins in occipital in its Scope). Support for this argument
cortex in the back of the brain. As we move comes from fMRI studies that demonstrate
forward into the temporal cortex, neurons an FFA response to items that are learned
become responsive to larger receptive fields at high levels of expertise, such as cars,
and more complex configurations of stimuli. birds, and greebles, artificial animal-like
Along this pathway, perceptual discrimina- stimuli (Gauthier & Tarr, 1997; Gauthier,
tion develops into object-based representa- Skudlarski, Gore, & Anderson, 2000; Gau-
tion (that is, entities with specific mean- thier & Tarr, 2002). All of these items are
ing) through reciprocal interactions between visual, classifiable at the individual level (like
high-level and low-level processing regions. faces), and only elicit FFA responses in an
individual who has developed expertise for
these items.
Face Processing
The FFA response to non-face objects
Humans are required to process faces on has sparked a debate as to whether
a daily basis, and it has been suggested this area is a module for face detection
they develop greater expertise in this pro- (Kanwisher, 2000; Kanwisher et al., 1997;
668 the cambridge handbook of expertise and expert performance
Kanwisher et al., 1998; Yovel & Kanwisher, objects, one can argue that faces are treated
2004), an area of visual expertise (Gauthier like objects by object-processing regions.
& Tarr, 1997; Gauthier et al., 2000; Tarr & Existence of the specialized area for face
Gauthier, 2000), or one area in a network processing is also supported from studies of
of regions responsible for the distributed- prosopagnosia patients, who have impair-
representation of faces and other learned ment in identifying individuals through
objects. The distributed-representation view facial recognition (Moscovitch, Winocur, &
claims that all objects produce a pattern Behrmann, 1997). This disorder occurs both
of activation across a series of visual areas congenitally and as a consequence of stroke.
that codes the learned category (Haxby These patients do not have a general prob-
et al., 2000; Haxby, Gobbini, Furey, Ishai, lem with (visual) identification, as they
Schouten, & Pietrini, 2001). The intrica- can identify and name individual face parts
cies of this debate are beyond the scope such as noses, and they can identify people
of this chapter, but as a consequence they through other cues such as voices. Prosopag-
have produced evidence that humans pro- nosia patients, however, are not impaired in
cess all faces as members of an expert class face inversion presumably because they pro-
of objects14 in a small localized area of cor- cess inverted faces more like objects (Yin,
tex (e.g., the faces areas in cortex represent 1970). Since object inversion is specific to
less than 1% of the brain). The processing objects for which someone has developed
is not unique to one stimulus class but to a expertise, this may reflect a shift in process-
range of related stimuli (e.g., faces and other ing once a class of objects is extremely well
objects). learned.
FFA appears to activate differently, based
on experience with different types of faces.
Object Processing
Athough FFA is not further activated by
famous faces (Gorno-Tempini & Price, 2001; In addition to face processing, humans spend
Gorno-Tempini et al., 1998), one study a great deal of time processing other types
found greater FFA activation for most sub- of objects. The ventral occiptico-temporal
jects for same-race faces, compared to faces cortex is activated when viewing pictures
from other races, presumably because of of objects, houses, and scenes, compared to
greater experience with same-race faces textures, noises, or scrambled objects (for a
(Golby, Gabrieli, Chiao, & Eberhardt 2001); review see Grill-Spector 2003 ). Many brain-
for a critique see Phelps, 2001). This pro- imaging studies have contrasted faces and
cessing area is also sensitive to inver- objects to differentiate processing between
sion effect, considered a sign of expert- these complex visual items (Aguirre et al.,
level object processing,15 an impairment in 1999; Gauthier et al., 2000; Haxby et al.,
recognition for upside-down objects (Yin, 2001; Ishai et al., 1999; Malach et al., 1995 ;
1969; Yin, 1970). Brain-imaging studies have McCarthy et al. 1997). There are areas that
found that inverted faces elicit the same respond to both parts and whole objects.
or slightly less FFA activation compared Object parts elicit responses from object-
to upright faces (Kanwisher et al. 1998); sensitive regions (Lerner Hendler, Ben-
however, inverted faces further activate Bashat, Harel, & Malach, 2001), in contrast
object-sensitive regions to a greater extent to the fusiform face area, which is not
than upright faces, presumably reflecting responsive to face parts. Temporal lobe areas
that they are processed more like objects perform object processing. These areas are
(Aguirre, Singh, & DEsposito 1999; Haxby more sensitive to greater complexity (i.e.,
et al., 1999). In other words, object sensi- these cells are not simple, single feature
tive regions are highly responsive to inverted detectors; they are most responsive to con-
faces compared to non-inverted faces, and figurations of features; see Network of Spe-
since the same pattern is exhibited for cialized Processing Regions section) and
inverted objects compared to non-inverted exhibit some activity to object scrambling,
brain changes in the development of expertise 669
which has been interpreted as evidence that tendency to respond to fewer or specifi-
object representation is based on component cally one thing). Although there was some
features (Grill-Spector et al., 1998; Lerner enhanced selectivity for individual object
et al, 2001). Recall that objects are typi- parts,18 there was a notable enhancement
cally learned at the basic level (e.g., chairs, for configurations of parts, that is, whole
chairs, and more chairs, though there is some objects. Importantly, this study showed that
amount of individual level categorization the specific enhancement in selectivity for
bar stool versus armchair), whereas faces trained objects is not due to differences in
are identified on a highly individualized the strength of response between trained and
level (e.g., Mary, Albert, Samantha, Jennifer, untrained stimuli. Selectivity allows objects
Sue, Jean, and David). Some objects (birds, to be coded at the individual level (see face-
cars, dogs16 ) can support the development processing section, i.e., specific to a par-
of face-like individual level expertise and ticular category example) instead of at the
elicit responses in face-processing regions in basic level (i.e., at the category level with-
individuals who have acquired this exper- out regard to a specific example). The for-
tise through intentional learning; however, mer is the hallmark of expertise. Typically
most objects are not learned at an expert objects are not coded at this level. For exam-
level, as compared to faces, and are pro- ple, an individual without specific experi-
cessed differently as such (i.e., object sen- ence studying birds should not be able to
sitive regions respond to object scrambling, identify many subtypes relative to an expert.
face sensitive regions do not respond to face The Baker et al. (2002) result (i.e., a neuron
scrambling). that is selective to an individual item) sug-
Experience with objects, however, does gests a mechanism by which bird watchers
impact their processing in other ways. may develop selectivity as they learn to iden-
Object recognition in ventral occipito- tify individual types of birds.
temporal cortex is invariant to size and loca- Although the temporal lobe is clearly
tion (Grill-Spector et al., 1999). Human involved in object learning, training influ-
behavioral work demonstrates that the abil- ences frontal cortical areas as well. A recent
ity to recognize backwards-masked objects monkey study by Rainer and Miller (2000)
improves with specific practice, and ability showed that training enhances specificity in
transfers when trained objects are modulated PF neurons. Though novel objects elicited
in size Furmanski & Engel (2000). Monkey a greater PF response than familiar objects,
single-unit studies (e.g., recording electri- with training neural activity became more
cal activity from individual units (neurons), narrowly tuned for familiar objects in these
to understand neuronal responsiveness)17 of neurons. Training also results in a PF rep-
inferotemporal (IT) neurons have demon- resentation that is robust to the effects of
strated that IT neurons develop view-point stimulus degradation. In addition, Freed-
invariance to objects that prior to training man, Riesenhuber, Poggio, and Miller (2001)
were meaningless and unfamiliar (Logothe- demonstrated that PF neurons are important
sis, Pauls, & Poggio, 1995 ). Other studies con- in learning new object categories.
firm that IT increases its responsiveness to In summary, objects such as faces and
trained objects (Kobatake, Wang, & Tanaka, other highly behaviorally relevant objects
1998) as well as learned patterns (Sakai & (i.e., relevant for the task at hand, such
Miyashita 1991; 1994). as birds are behaviorally relevant to bird
Recent work implicates these neurons in watchers) receive specialized processing in
visual object expertise (i.e., face, acquired the visual-processing stream. In the motor
bird expertise, etc.; Baker, Behrmann, & sections we will present further examples
Olson, 2002; for comments see Connor, of how behaviorally relevant movements
2002). Baker et al. (2002) performed dis- (i.e., finger movements for violinist) and
crimination training on monkeys to deter- stimuli (i.e., words for readers) are rep-
mine if training enhances IT selectivity (i.e., resented uniquely in the cortex. We will
670 the cambridge handbook of expertise and expert performance
also see that practice with these items will What are the areas that support read-
sometimes result in structural expansions of ing change as a function of proficiency? A
cortex. cross-sectional fMRI study (subjects ranged
in age between six and twenty-two) found
a shift in brain regions associated with
Word Reading
an implicit word-processing task as read-
Learning to read is a key skill in our mod- ing ability develops (Turkeltaub, Gareau,
ern society. It involves developing new rep- Flowers, Zeffiro, & Eden, 2003 ). As read-
resentations in a variety of cortical areas. ing began to reflect knowledge of abstract
Of particular importance is the Visual Word word properties (semantics and phonolog-
Form Area (VWFA), an area of left fusiform ical properties) and was less supported by
gyrus that appears sensitive to words that rote memorization of words based on visual
are specifically presented visually. The basic features and context (i.e., stop in a stop
findings with regard to this area have sign), readers demonstrated increased acti-
been reviewed by McCandliss, Cohen, and vation in left middle temporal and inferior
Dehaene (2003 ) and will be summarized frontal gyri and decreases in right inferotem-
below. VWFA is insensitive to visual vari- poral regions.
ation such as changes in case, font, and even Brain plasticity in the reading circuit
location (i.e., no difference in response to can be observed even in adult subjects
left or right hemisphere presented words). after short periods of training. In a novel
It is insensitive to lexical properties of words orthography training experiment, Bolger,
such as word frequency, and it even responds Schneider, and Perfetti (2005 ) trained sub-
to pseudowords as long as these words are jects to learn to read eighty words writ-
well formed according to regularities of the ten in Korean script. Pilot studies conducted
language system. This area is also respon- with training on only sixteen words found
sive to non-word objects for which a per- increases in cortical activation occur rapidly:
son has achieved visual expertise such as a 0.7% increase in BOLD signal from learn-
faces. Therefore VWFA has been suggested ing trials one through four to trials thirteen
as an area specifically implicated in word- through sixteen (Bolger, 2005 ). After four
form processing as a result of developed sessions (twenty words/session) of training,
expertise in processing behaviorally relevant the response in the VWFA increased signif-
stimuli. A recent meta-analysis of imaging icantly and with greater learning in a com-
cross-cultural language processing (Bolger, ponential (i.e., learning letter-sound corre-
Perfetti, & Schneider, 2005 ) provides sup- spondences) compared with a holistic (i.e.,
port by demonstrating that VWFA is con- learning of the whole word) training
sistently activated across word tasks and approach to the material.
writing systems (both eastern and western). How people attend and process stimuli
Furthermore, lesions to the VWFA region alters what cortical areas show plasticity.
have resulted in impairments in recognizing Sandak et al. (2004) explored the effects
and naming words and pronounceable non- of orthographic, phonological, and semantic
words, but are relatively spared in the identi- pseudowords training on overt naming abil-
fication of digits, objects, and, in some cases, ity. Orthographic training involved making
letters themselves. This is a disorder known judgments about consonant and vowel pat-
as pure alexia. Thus, the role of processing terns in pseudowords, phonological training
visually abstract forms of candidate words involved making rhyme judgments in pseu-
has been ascribed to this region (McCarthy dowords, and semantic training involved
& Warrington, 1990; Miozzo & Caramazza, learning novel semantic associations to pseu-
1998). However, owing to the complex vas- dowords. Phonological and semantic train-
culature of the brain, pure alexia stemming ing resulted in equivalent (but superior
from the inferior temporal region rarely ever when compared to orthographic training)
occurs (Price & Devlin, 2003 ). performance on reading ability. Despite
brain changes in the development of expertise 671
comparable behavioral performance, phono- Leone et al, 1993 ; Pascual-Leone and Torres,
logical and semantic training effects were 1993 ). Perhaps most dramatically, structural
driven by different neural processes. Phono- damage, such as a facial nerve lesion, resulted
logical training modulates VWFA process- in the rat primary motor cortex (M1) shift-
ing. The reported reduction in activation was ing representation to a new group of mus-
interpreted as reflecting efficient processing cles (representing the forelimb) within one
in this region. to three hours of the insult (Huntley, 1997).
Studies have shown the structural con- Primary motor cortex learning effects
nectivity of white matter fiber tracts to have been investigated extensively with
be deficient in poor versus skilled read- sequence learning paradigms. An early func-
ers (Klingberg et al., 2000). Similarly, tional magnetic resonance imaging study
functional-connectivity studies of correlated (Karni et al, 1995 ) found that M1 modulates
cortical activity have revealed stronger con- its response to trained finger-thumb opposi-
nectivity between angular gyrus with infe- tion sequences according to the level of prac-
rior frontal and ventral fusiform regions as tice. Early in practice, M1 is sensitive to order,
a function of reading skill (Horwitz et al., initially being more responsive to the first
1998). Pugh et al. (2001) conducted their sequence and later being more responsive to
own functional-connectivity study of the the second sequence, within one session of
angular gyrus comparing normal to impaired training. Karni has referred to the reversal
readers. Their findings reveal that in dyslex- of order effect as the fast learning phase.
ics connectivity in the angular gyrus region However, after four weeks of training M1
is weak for word and pseudoword reading. is more responsive to the trained sequence,
The reading literature illustrates some compared to the untrained sequence, regard-
anatomical mechanisms of learning in the less of practice order termed the slow
brain. Processing is localized and very spe- learning phase. In addition to slow and fast
cialized, with VWFA showing word encod- learning, M1 is believed to be involved in
ing, learning, and processing occurring in the consolidation or in performance improve-
same area. Learning produces both increases ments that occur subsequent to practice
in activation early in practice and decreases (Ungerleider, Doyon, & Karni, 2002). The
as reading becomes more automatic (i.e., if consolidation process is time dependent;
processing rate is controlled). Words are pro- disruption of M1 by repetitive Transcranial
cessed in specialized areas by experts, and Magnetic Stimulation (rTMS) immediately
learning can produce detectable morpholog- after practice diminishes the effects of train-
ical changes. In addition, the training studies ing (Muellbacher et al., 2002). rTMS of
show that the nature of the practice (e.g., control brain19 regions and of M1 six hours
phonological or semantic encoding) impacts after training does not mitigate the effects
where the plastic change takes place. of practice. M1 consolidation effects are evi-
dence that this region is involved in early
learning processing; however, consolidation
Motor Learning
blocking has been found in other regions.
Motor areas can rapidly change as a result This has been regarded as evidence for a dis-
of skilled movement practice and improved tributed network of areas involved in early
performance. Primary motor cortex or M1 is phases of motor learning, particularly the
notable for plastic change with very extensive learning of complex motor skills (Baraduc,
experience and practice (for a review see Sanes Lang, Rothwell, & Wolpert, 2004).
& Donoghue, 2000). M1 motor representa- Although brain-imaging work implicates
tions are experience dependent and highly M1 in sequence learning, (Karni et al.,
modifiable under changing environments. 1995 ), single-unit research suggests that
For example, blind individuals with knowl- M1 is involved in movement execution
edge of Braille have enlarged M1 representa- but is insensitive to temporal order aspects
tion for their (reading) index finger (Pascual- of skilled movement (Tanji & Shima;
672 the cambridge handbook of expertise and expert performance
1994). Single-unit recording in the monkey posterior superior temporal sulcus. These
implicates supplementary motor area areas respond more strongly when an expert
(SMA) and pre-supplementary motor area observer views a movement that was specifi-
(pre-SMA) involvement in sequence learn- cally acquired previously by the observer (e.g.,
ing. Neuron response properties in these seeing a dance move that the observer had
regions are sensitive to particular trained learned to perform). Therefore, the brain is
sequences and rank orders (i.e., always sensitive to complex acquired movement,
respond to the second action); additionally, such that passive viewing of another individ-
they are sensitive to movement interval ual performing behaviorally relevant move-
and movement initiation, both with regard ment results in specialized processing and
to specific movement types and sequence representation. Studies of the macaque mir-
completion (Tanji & Shima, 1994; Shima ror neurons provide a mechanism for this
& Tanji, 2000). The distributions of neural viewer-based processing of relevant move-
responsiveness for the aforementioned ment. Mirror neurons discharge when the
functions vary between SMA and pre-SMA, monkey performs an action or observes
as does the specific selectivity for each func- another monkey or human perform this
tion (for example in the case of rank-order action, hence their name, and have been
neurons, their response may be exclusive proposed to exist in humans (Gallese &
to the second action, or they may respond Goldman, 1998; Rizzolatti et al., 1996).
to both the first and second action but not In the monkey these neurons are known
the third). More pre-SMA neurons (10%) as to exist in premotor and parietal cortex.
compared to SMA neurons (2%) respond According to Calvo-Merino et al. (2005 ),
during the initiation of a new sequence, the human mirror system appears to code
suggesting a role in the early stages of learn- for complete action patterns that are in
ing for pre-SMA. Injecting pre-SMA with an individuals motor repertoire, as opposed
muscimol to produce a reversible lesion to movements that are highly familiar to
resulted in a disruption of performance (in the observer. They scanned professional bal-
terms of button press errors) for novel but let dancers, professional capoeira martial
not learned sequences (Nakamura, Sakai, & artists, and control subjects with no spe-
Hikosaka 1999). Injection of SMA produced cific movement expertise as these individ-
a similar pattern (i.e., disruption occurring uals passively viewed video-taped move-
only for novel sequence performance); ments from both disciplines. They were able
however, this was not a significant effect. to demonstrate the mirror systems exper-
Furthermore, in another study, pre-SMA tise specificity, which even distinguishes bal-
neurons became less active as sequences let and capoeira, despite similar kinematics
become automated (Nakamura, Sakai, & for males. Even though whole movements
Hikosaka 1998). Together these studies were somewhat similar (sub-movements can
provide evidence that pre-SMA is involved be identical) the expert brain is sensitive
in sequence learning. enough to discriminate between acquired
movements in the studied discipline and
similar movements in the non-studied dis-
Body Kinematics
cipline. In other words, if the participants
Complex motor actions, such as those were expert performers, such as a ballet
involved in dance and martial arts (see dancer, their brain had a greater response
Noice and Noice, Chapter 28), are coded to viewing ballet movements when com-
differently by an observer, depending on pared to viewing capoeira, even though
the observers own expertise executing the there is similarity in the types of movements
specific movements (Calvo-Merino, Glaser, being performed in both disciplines (see also,
Grezes, Passingham, & Haggard, 2005 ). Feltovich et al., Chapter 4, on Expertise is
Regions sensitive to motor expertise include Limited . . . ). It is important to stress that
bilateral pre-motor cortex and intraparietal Calvo-Merino et al. (2005 ) argue that exper-
sulcus, right superior parietal lobe, and left tise is operating at the level of being able to
brain changes in the development of expertise 673
perform the movement and not just being London (Maguire et al., 1997). Grey matter
familiar with the movement (i.e., this sug- volume in this region (as well as left poste-
gests that there is a difference between the rior hippocampus; LPH) was subsequently
brain of a professional dancer and an avid shown to be greater in the expert (driver)
dance enthusiast) that appears to drive these population, compared to a non-expert con-
regions. Female and male ballet dancers code trol group (increase relative to non-drivers:
ballet movements differently based on the RPH = 1.93 6%, LPH = 1.5 06%; Maguire
gender of the movement; that is, in clas- et al., 2000) and another control group rang-
sical ballet certain movements are only per- ing in navigational expertise but without
formed by men, and other movements only specific taxi driving experience (Maguire
by women, but many movements are gen- et al., 2003 ).20 Posterior right hippocam-
der neutral. For example, men never learn pal grey matter volume is positively cor-
to dance on point, that is, stand on their related with taxi driving experience (r =
toes. Although no dancer can perform all 0.6; p < 0.05 ) in drivers, but there was
movements, all dancers are highly famil- no associated grey matter relationship in
iar with viewing all movements through non-drivers. Haguire and colleagues suggest
rehearsals, classes, and performances. Left that the structural differences in taxi drivers
parietal cortex is less responsive in female are based on acquired experience and not
ballet dancers when they view male ballet innate ability that might cause high per-
movements. A skilled movement, in this case formers to seek out this profession. Correla-
gender-matched specific movements, modu- tion analysis shows that individuals skilled in
late the level of envoked representation by wayfinding (new route development) acti-
the expert brain. Therefore, the ability and vated anterior hippocampus during novel
personal experience with performing these routes but activated the head of the right
specialized movement patterns appear to be caudate when following well-learned routes
critical to the difference in representation (Hartley, Maguire, Spiers, & Burgess, 2003 ).
of the movements. These results show brain This distinction is not found in individuals
specializations that enable the encoding of who perform poorly at wayfinding.
observed actions into ones own action sys- Expert drivers use these two areas depen-
tems in a way that may potentially enable dent on the task at hand; the hippocampus
replication of the observed actions. is purported to form a modifiable cognitive
map supporting new route development,
whereas the caudate supports fast, auto-
Automotive Spatial Navigation
matic navigation of well-learned routes. Fur-
Brain areas supporting spatial navigation are thermore, these areas have been proposed to
sensitive to expertise with regard to function support learning in a complimentary fashion
and structure (see Durso & Dattel, Chap- (Hartley et al., 2003 ) in other task domains
ter 20). One example of a spatial navigation (classification learning, Poldrack et al.,
expert is a taxi driver. These highly skilled 2001; mirror reading, Poldrack & Gabrieli,
individuals have to know large metropolitan 2001). These studies demonstrate that exper-
areas and how to reach locations in the most tise provides the flexibility to choose the
efficient manner. London taxi drivers rigor- optimal strategy for successful completion
ously train, on average for two years, to pass a of a given task (cf. Feltovich, Spiro, &
series of exams about street names and their Coulson, 1997). These different strategies
locations, which is required for their taxi rely on different brain structures for their
license. Their extensive experience learning execution.
navigation has been suggested to produce
functional and structural changes (Maguire,
Music Training
Frackowiak, & Frith, 1997; Maguire et al.,
2003 ). Functional MRI revealed increased Music expertise is an important topic in
right posterior hippocampus (RPH) activa- skill acquisition (see Lehmann & Gruber,
tion with successful recall of routes around Chapter 26) in which mastery has been
674 the cambridge handbook of expertise and expert performance
et al. (2003 ) comparing the worlds memory general control areas and the domain spe-
experts to control participants brain activa- cific representational areas. The presence
tion during memorization. They found that of a single domain general control network
the differences in brain activation during dif- that supports novice and variable perfor-
ferent memory tasks could be completely mance represents a severe resource limit
accounted for by the superior strategies that for performing novel or varying tasks and
the world experts reported using (similar to working memory dependent tasks (see also
taxi drivers; Hartley et al., 2003 ). The same Feltovich et al., Chapter 4, Expertise is
study did not find any anatomical differences an Adaptation). This network provides the
in the brains of the world experts compared scaffolding to support new learning and
to the control participants,23 which suggests to maintain working memory variables and
that the difference in memory performance operations in order to allow varying the
can be explained in terms of acquired skill nature of the performance and strategy shifts
(Ericsson, 2003 ; Chapter 13 ). in cognitive processing. In consistent tasks,
Of additional interest, Butterworth as processing becomes more automatic,
(Chapter 3 2) describes the evidence on the domain general activity decreases or
brain activation during routine and chal- drops out.
lenging mathematical calculations. He The specific nature of the representa-
reviews evidence that fronto-parietal net- tional areas suggests that both training and
works support the performance of routine performance will be sensitive to the strategy
numerical tasks (Pesenti, Thioux, Seron, and nature of the training. What is learned
& De Volder, 2000) with left intraparietal is based on which representational areas are
sulcus being specialized for numerical pro- active during training. Typically, as prac-
cessing (Dehaene, Piazza, Pinel, & Cohen, tice develops, activity decreases, and there
2003 ). The brain of an expert calculator are rarely new areas that develop in labo-
named Gamm is also discussed. Similar ratory studies of skill acquisition. This sug-
to the taxi driver study of Hartley et al. gests that training causes local changes in the
(2003 ), Pesenti et al. (2001) found that specific representational areas that support
experts use different brain systems to sup- skilled performance. In studies of extensive
port their calculations and also could exhibit training, there is ample evidence for changes
flexibility in strategy choice (supported by in cognitive processing as well as structural
different brain regions) to solve their changes in the nervous system.
problems. Brain training has analogies and differ-
In summary, in addition to process- ences to muscle training. Working a specific
ing efficiency, enriched representations, and brain area can increase the representation
structural expansions, experts can flexibly space and make processing more focused. If
use strategies, by recruiting the associated one wants to strengthen a brain area, one
brain regions, to solve a range of problems, needs to attentively activate those areas to
whereas novice performers can not. alter the neurons in that area. Training of
the domain specific areas typically decreases
activity as processing gets more focused;
Conclusion however, it can cause increases in some
motor tasks as well as some tasks involving
The development and execution of skills exceptional memory. The domain general
has profound effects on the nature of brain areas might be analogous to cardiovascu-
processing. The brain is a plastic structure lar training in muscle training (e.g., train-
that can change the amount of area and ing endurance transfers across many sports).
the activity of areas as a function of train- However, the specific training (e.g., shoot-
ing, effort, and strategy. There are hun- ing in basketball and hitting in baseball) is
dreds of specialized areas of the brain. Train- unlikely to activate the same areas or repre-
ing has differential effects on the domain sentation and do not lead to transfer.
676 the cambridge handbook of expertise and expert performance
Cognitive neuroscience is in a synergis- 9. The paper uses the term SOA for stimulus
tic research development with skill-based onset asynchrony instead of ISI for inter-
research. We know that training dramatically stimulus interval.
effects performance and brain activity. We 10. Marcantoni et al. (2003 ) uses another inter-
are now beginning to relate those changes to ference paradigm, rapid serial visual presen-
better understand both the brain and skilled tation (RSVP).
performance. 11. Dual task and single tasks were scanned in
a mixed event-related design allowing blocks
to contain pure (i.e., exclusively) single-task,
pure dual-task, and mixtures of single- and
Footnotes dual-task trials. This design allowed the activ-
ity associated with individual trial types to
be contrasted against any other trial, regard-
1. FMRI data is not an absolute value and there- less of block, such that differences in dual-
fore the signal is always assessed as a percent task performance when planning to perform
change relative to a baseline or control con- exclusively a dual-task trial versus a mix-
dition. One can determine whether a loca- ture of dual- and single-task trials could be
tion differs in activation over time (e.g., active addressed.
when stimulus present but not during resting
12. A control group was employed to control
periods).
for non-training specific effects in dual-task
2. This is not a comprehensive list of functions. activity.
Often performance relies on dynamic interac-
13 . Most objects, such as tables, are classified
tions among various regions, both within and
at the basic level. Faces are considered rela-
across lobes. Furthermore, there are regions
tively unique because they are processed at
outside the cerebrum, namely, the brain
the individual level, which is with regarded
stem, cerebellum, and spinal cord, that make
to particular examples.
important contributions to the performance
of skills. 14. The authors are not presenting this informa-
tion in support of the Gauthier et al. (2000)
3 . See Controlled and Automatic Processing
visual-expertise interpretation. It should be
during Learning section.
noted that some of these findings have been
4. Evidence will also be presented that experts contested or interpreted as both support for
are more flexible in their strategy utilization. and evidence against this FFA-individual-
5 . That is, performance characteristics can be level-expertise model. The work is men-
flexibly modified on the fly. tioned because it pertains to differential pro-
6. This shift typically involves no more than sev- cessing of expert or experienced-based items.
eral sessions of training for simple cognitive 15 . Dog experts, compared to novices for
tasks such as visual search paradigms (where instance, exhibit dog-inversion decrements in
one searches for targets in a display containing identification (Diamond & Carey, 1986).
distractor items), as opposed to several years 16. To clarify, being familiar with a few specific
of training for learning a musical instrument types of birds or even an individual bird (i.e., a
or high-level chess mastery. pet parrot) does not grant this level of exper-
7. The task involved recognition of faces pre- tise. To develop face-like expertise for a non-
sented at encoding after a delay period. Load face object category such as birds, one must
was varied to compare low-load (one to two become extremely knowledgeable in iden-
faces) with high-load (three to four faces) tifying many bird subtypes. This expertise-
conditions. based processing would develop only in an
8. This is referred to as processing bottlenecks, avid bird watcher and not the typical individ-
which means that a specific process must ual who can identify a few varieties of birds
be performed serially. Serial processing pro- and/or may own birds as pets.
duces interference (increased RT) when in a 17. Much of our understanding of the visual sys-
dual-task environment. The nature of locus of tem traditionally comes from invasive mon-
such bottlenecks is a matter of debate (Meyer key physiology studies. Since the advent of
& Kieras, 1997; Pashler, 1994; Schneider & non-invasive brain imaging techniques, we
Detweiler, 1988; Shiffrin, 1988). have been able to confirm that the human
brain changes in the development of expertise 677
visual system works in a highly similar way. Baddeley, A. (2003 ). Working memory and
Single-unit recording studies allow us to language: An overview. Journal of Communi-
determine the response properties of indi- cations Disorders, 3 6, 189208.
vidual neurons; however, because this tech- Baddeley, A. D., & Hitch, G. J. (1974). Working
nique is invasive, it is typically not per- memory. In G. A. Bower (Ed.), Recent Advances
formed on humans, with the rare exception in Learning and Motivation, Vol. 8 (pp. 4789).
being patients undergoing neurological surgi- New York: Academic Press.
cal procedures. Baker, C. I., Behrmann, M., & Olson, C. R.
18. Selectivity enhancement for object parts has (2002). Impact of learning on representation
been demonstrated also when features are of parts and wholes in monkey inferotemporal
diagnostic (Sigala & Logothetis, 2002). cortex. Nature Neuroscience, 5 (11), 12101216.
19. Control brain regions does not refer to Baraduc, P., Lang, N., Rothwell, J. C., & Wolpert,
the control network, but rather other brain D. M. (2004). Consolidation of dynamic motor
regions serving as experimental controls. learning is not disrupted by rTMS of primary
20. Overall hippocampal size does not differ motor cortex. Current Biology, 14(3 ), 25 225 6.
between drivers and non-drivers, because Bengtsson, S. L., Nagy, Z., Skare, S., Forsman,
non-drivers have an increase in anterior hip- L., Forssberg, H., & Ullen, F. (2005 ). Extensive
pocampi regions relative to drivers. piano practicing has regionally specific effects
21. A split-half correlation analysis was pre- on white matter development. Nature Neuro-
formed, based on a technique developed science, 8(9), 1148115 0.
by Haxby et. al., (2001), on primary motor
Beyerstein, B. L. (1999). Whence cometh the
and somatosensory cortex (respectively M1
myth that we only use ten percent of our
and S1). Digit movement was defined more
brains? In S. D. Sala (Ed.), Mind Myths: Explor-
predictably for a particular region, in this
ing Everyday Mysteries of the Mind and Brain
case M1, if the movement-specific activity
(pp. 124). Chichester, UK: Chichester, UK.
correlated better with itself across the two
halves of the data than with any other digit Bolger, D. J. (2005 ). [Pilot fMRI data of learning
movement. a novel orthography.] Unpublished raw data.
22. In the motor system, the right side of the body Bolger, D. J., Perfetti, C. A., & Schneider, W.
is controlled by the left side of the brain and (2005 ). Cross-cultural effect on the brain
vice versa. revisited: Universal structures plus writing sys-
tem variation, Human Brain Mapping, 2 5 , 92
23 . Taxi-drivers do have morphological expan-
104.
sions based on experience, see Automo-
tive Spatial Navigation section for an Bolger, D. J., Schneider, W., & Perfetti, C. A.
explanation. (2005 ). The Development of Orthographic
Knowledge: A Cognitive Neuroscience Investiga-
tion of Reading. Paper presented at the 12th
Annual Meeting of the Society for the Scien-
References tific Study of Reading, Toronto, Ontario.
Bunge, S. A., Klingberg, T., Jacobsen, R. B., &
Adcock, R. A., Constable, R. T., Gore, J. C., Gabrieli, J. D. E. (2000). A resource model of
& Goldman-Rakic, P. S. (2000). Functional the neural basis of executive working memory.
neuroanatomy of executive processes involved Proceedings of the National Academy of Sciences,
in dual-task performance. Proceedings of the 97 (7), 3 5 73 3 5 78.
National Academy of Sciences, 97 (7), 3 5 67 Buonomano, D. V., & Merzenich, M. M. (1998).
3 5 72. Cortical plasticity: From synapses to maps.
Aguirre, G. K., Singh, R., & DEsposito, M. Annual Review of Neuroscience, 2 1, 149186.
(1999). Stimulus inversion and the responses Cabeza, R., & Nyberg, L. (2000). Imaging cog-
of face and object-sensitive cortical areas. Neu- nition ii: An empirical review of 275 pet and
roreport, 10, 189194. fMRI studies. Journal of Cognitive Neuroscience,
Ahissar, M., & Hochstein, S. (2004). The reverse 12 (1), 147.
hierarchy theory of visual perceptual learn- Calvo-Merino, B., Glaser, D. E., Grezes, J., Pass-
ing. Trends in Cognitive Sciences, 8(10), 45 7 ingham, R. E., & Haggard, P. (2005 ). Action
464. observation and acquired motor skills: An fMRI
678 the cambridge handbook of expertise and expert performance
study with expert dancers. Cerbral Cortex, 15 , characterized by complexity and change. In P. J.
1243 1249. Feltovich, K. M. Ford, & P. R. Hoffman (Eds.),
Carpenter, P. A., Just, M. A., & Reichle, Expertise in Context (pp. 125 146). Menlo PK,
E. D. (2000). Working memory and executive CA: AAAI/MIT Press.
function: Evidence from neuroimaging. Cur- Fisk, A. D., & Schneider, W. (1983 ). Category
rent Opinion in Neurobiology, 10, 195 199. and word search: Generalizing search princi-
Chein, J. M., McHugo, M., & Schneider, W. ples to complex processing. Journal of Experi-
(in preparation). The transition from con- mental Psychology: Learning, Memory, and Cog-
trolled to automatic processing in simple search nition, 9(2), 177195 .
tasks as revealed with fMRI. Manuscript in Fitts, P., & Posner, M. I. (1967). Human Per-
preparation. formance. Monterey, CA: Brooks/Cole. Fitts,
Chein, J. M., & Schneider, W. (in press). Neu- P., & Gibson, E. J. (1969). Principles of Per-
roimaging studies of practice-related change: ceptual Learning and Development. Englewood
fMRI and meta-analytic evidence of a domain- Cliffs, NJ: Prentice Hall. Fitts, P., & Welford,
general control network for learning. Cognitive A. T. (1968). Fundamentals of Skill. London:
Brain Research. Methuen.
Connor, C. E. (2002). Representing whole Freedman, D. J., Riesenhuber, M., Poggio, T., &
objects: Temporal neurons learn to play their Miller, E. K. (2001). Categorical representation
parts. Nature Neuroscience, 5 (11), 1105 1106. of visual stimuli in the primate prefrontal cor-
DEsposito, M., Detre, J. A., Alsop, D. C., & Shin, tex. Science, 2 91, 3 123 16.
R. K. (1995 ). The neural basis of the central Furmanski, C. S., & Engel, S. A. (2000). Per-
executive system of working memory. Nature, ceptual learning in object recognition: Object
3 78(65 5 4), 279281. specificity and size invariance. Vision Research,
Dehaene, S., Piazza, M., Pinel, P., & Cohen, L. 40, 473 484.
(2003 ). Three parietal circuits for number pro- Gallese, V., & Goldman, A. (1998). Mirror
cessing. Cognitive Neuropsychology, 2 0, 487 neurons and the simulation theory of mind-
5 06. reading. Trends in Cognitive Science, 2 (12), 493
Elbert, T., Pantev, C., Wienbruch, C., Rockstroh, 5 01.
B., & Taub, E. (1995 ). Increased cortical repre- Garavan, H., Ross, T. J., Li, S. J., & Stein, E. A.
sentation of the fingers of the left hand in string (2000). A parametric manipulation of cen-
players. Science, 2 70(5 23 4), 3 05 3 07. tral executive functioning. Cerebral Cortex, 10,
Erickson, K. I., Colcombe, S. J., Wadhwa, R., 5 85 5 92.
Bherer, L., Peterson, M. S., Scalf, P. S., Kim, Gaser, C., & Schlaug, G. (2003 ). Gray matter dif-
J. S., Alvarado, M., & Kramer, A. F. (2005 a). ferences between musicians and nonmusicians.
Training-induced plasticity in older adults: Effects Annals of the New York Academy of Sciences,
of training on hemispheric asymmetry. Unpub- 999, 5 145 17.
lished manuscript, Urbana. Gauthier, I., Skudlarski, P., Gore, J. C., & Ander-
Erickson, K. I., Colcombe, S. J., Wadhwa, R., son, A. W. (2000). Expertise for cars and birds
Bherer, L., Peterson, M. S., Scalf, P. S., recruits brain areas involved in face recogni-
Kim, J. S., Alvarado, M., & Kramer, A. F. tion. Nature Neuroscience, 3 (2), 191197.
(2005 b). Training induced changes in dual- Gauthier, I., & Tarr, M. J. (1997). Becoming
task processing: An fMRI study. Unpublished a greeble expert: Exploring mechanisms for
manuscript, Urbana. face recognition. Vision Research, 12 , 1673
Erickson, K. I., Colcombe, S. J., Wadhwa, R., 1682.
Bherer, L., Peterson, M. S., Scalf, P. S., & Gauthier, I., & Tarr, M. J. (2002). Unravel-
Kramer, A. F. (2005 c). Neural correlates of ing mechanisms for expert object recognition:
dual-task performance after minimizing task- Bridging brain activity and behavior. Journal of
preparation. Unpublished manuscript, Urbana. Experimental Psychology: Human Perception and
Felleman, D. J., & Van Essen, D. C., (1991). Performance, 2 8(2), 43 1446.
Distributed hierarchical processing in primate Gazzaniga, M. S., Ivry, R. B., & Mangun, G. R.
cerebral cortex. Cerebral Cortex, 1, 147. (2002). Cognitive Neuroscience: the Biology of
Feltovich, P. J., Spiro, R. J., & Coulson, R. L. the Mind (Second Edition). New York: W. W.
(1997). Issues of expert flexibility in contexts Norton & Company.
brain changes in the development of expertise 679
Golby, A. J., Gabrieli, J. D., Chiao, J. Y., & Eber- during training. American Journal of Psychiatry,
hardt, J. L. (2001). Differential responses in 161(4), 745 747.
the fusiform region to same-race and other-race Herath, P., Klingberg, T., Young, J., Amunts, K.,
faces. Nature Neuroscience, 4(8), 845 85 0. & Roland, P. (2001). Neural correlates of dual
Gorno-Tempini, M. L., & Price, C. J. (2001). task interference can be dissociated from those
Identification of famous faces and buildings: A of divided attention: An fMRI study. Cerebral
functional neuroimaging study of semantically Cortex, 11, 796805 .
unique items. Brain, 12 4, 20872097. Hill, N. M., & Schneider, W. (2005 ). Changes in
Gorno-Tempini, M. L., Price, C. J., Josephs, O., neural activation related to dual-task practice:
Vandenberghe, R., Cappa, S. F., Kapur, N., Evidence for a domain general learning net-
Frackowiak, R. S. J., & Tempini, M. L. (1998). work. Poster presented at the 11th annual meet-
The neural systems sustaining face and proper- ing of Human Brain Mapping. Toronto, ON.
name processing. Brain, 12 1, 2103 2118. Honda, M., Deiber, M. P., Ibanez, V., Pascual-
Grill-Spector, K. (2003 ). The neural basis of Leone, A., Zhuang, P., & Hallett, M. (1998).
object perception. Current Opinion in Neuro- Dynamic cortical involvement in implicit and
biology, 13 , 15 9166. explicit motor sequence learning. A pet study.
Grill-Spector, K., Kushnir, T., Edelman, S., Brain, 12 1, 215 92173 .
Avidan-Carmel, G., Itzchak, Y., & Malach, R. Horwitz, B., Rumsey, J. M., & Donohue, B. C.
(1999). Differential processing of objects under (1998). Functional connectivity of the angular
various viewing conditions in the human lateral gyrus in normal reading and dyslexia. Proceed-
occipital complex. Neuron, 2 4, 187203 . ings of the National Academy of Science of the
Grill-Spector, K., Kushnir, T., Hendler, T., Edel- United States of America, 95 , 893 98944.
man, S., Itzchak, Y., & Malach, R. (1998). A Huntley, G. W. (1997). Correlation between pat-
sequence of object-processing stages revealed terns of horizontal connectivity and the extend
by fMRI in the human occipital lobe. Human of short-term representational plasticity in rat
Brain Mapping, 6, 3 163 28. motor cortex. Cerebral Cortex, 7, 143 15 6.
Hartley, T., Maguire, E. A., Spiers, H. J., & Ishai, A., Ungerleider, L., Martin, A., Schouten,
Burgess, N. (2003 ). The well-worn route and J. L., & Haxby, J. V. (1999). Distributed repre-
the path less traveled: Distinct neural bases sentation of objects in the human ventral visual
of route following and wayfinding in humans. pathway. Proceedings of the National Academy
Neuron, 3 7 , 877888. of Sciences of the United States of America, 96,
Haxby, J. V., Gobbini, M. I., Furey, M. L., Ishai, 93 7993 84.
A., Schouten, J. L., & Pietrini, P. (2001). Dis- Jancke, L., Shah, N. J., & Peters, M. (2000). Cor-
tributed and overlapping representations of tical activation in primary and secondary motor
faces and objects in ventral temporal cortex. areas for complex bimanual movements in pro-
Science, 2 93 , 2425 243 0. fessional pianists. Cognitive Brain Research, 10,
Haxby, J. V., Hoffman, E. A., & Gobbini, M. I. 177183 .
(2000). The distributed human neural system Jaeggi, S. M., Seewer, R., Nirkko, A. C., Eckstein,
for face perception. Trends in Cognitive Sciences, D., Schroth, G., Gronerm R., & Gutbrod, K.
4(6), 223 23 3 . (2003 ). Does excessive memory load attenu-
Haxby, J. V., Ungerleider, L. G., Clark, V. P., ate activation in the prefrontal cortex? Load-
Schouten, J. L., Hoffman, E. A., & Martin, A. dependent processing in single and dual task:
(1999). The effect of face inversion on activity functional magnetic resonance imaging study.
in human neural systems for face and object NeuroImage, 19, 210225 .
perception. Neuron, 2 2 , 189199. Jansma, J. M., Ramsey, N. F., Slagter, H. A., &
Hazeltine, E., Grafton, S. T., & Ivry, R. (1997). Kahn, R. S. (2001). Functional anatomical cor-
Attention and stimulus characteristics deter- relates of controlled and automatic processing.
mine the locus of motor-sequence encoding. Journal of Cognitve Neuroscience, 13 , 73 0743 .
A pet study. Brain, 12 0, 123 140. Jezzard P., Matthews, P. M., & Smith, S. M.
Hempel, A., Giesel, F. L., Garcia Caraballo, (2001). Functional MRI: An introduction to
N. M., Amann, M., Meyer, H., Wustenberg, T., methods. Oxford: Oxford University Press.
Essig, M., & Schroder, J. (2004). Plasticity of Jiang, Y. (2004). Resolving dual-task interference:
cortical activation related to working memory An fMRI study. Neuroimage, 2 2 , 74875 4.
680 the cambridge handbook of expertise and expert performance
Just, M. A., Carpenter, P. A., Keller, T. A., Emery, ability: Evidence from diffusion tensor mag-
L., Zajac, H., & Thulborn, K. R. (2001). Inter- netic resonance imaging. Neuron, 2 5 , 493
dependence of nonoverlapping cortical systems 5 00.
in dual cognitive tasks. Neuroimage, 14, 417 Kobatake, E., Wang, G., & Tanaka, K. (1998).
426. Effects of shape-discrimination training on the
Kanwisher, N. (2000). Domain specificity in face selectivity of inferotemporal cells in adult mon-
perception. Nature Neuroscience, 3 (8), 75 9 keys. Journal of Neurophysiology, 80, 3 243 3 0.
763 . Kolb, B., & Whishaw, I. Q. (2003 ). Fundamen-
Kanwisher, N., McDermott, J., & Chun, M. M. tals of Human Neuropsychology (Fifth Edition).
(1997). The fusiform face area: A module New York: Worth Publishers.
in human extrastriate cortex specialized for Landau, S. M., Schumacher, E. H., Garavan, H.,
face perception. Journal of Neuroscience, 17 (11), Druzgal, T. J., & DEsposito, M. (2004). A
43 0243 11. functional mri study of the influence of practice
Kanwisher, N., & OCraven, K. (2000). Men- on component processes of working memory.
tal imagery of faces and places activates corre- Neuroimage, 2 2 , 211221.
sponding stimulus-specific brain regions. Jour- Lay, B. S., Sparrow, W. A., Hughes, K. M., &
nal of Cognitive Neuroscience, 12 (6) 1013 1023 . ODwyer, N. J. (2002). Practice effects on coor-
Kanwisher, N., Tong, F., & Nakayama, K. (1998). dination and control, metabolic energy expen-
The effect of face inversion on the human diture, and muscle activation. Human Move-
fusiform face area. Cognition, 68, B1B11. ment Science, 2 1, 80783 0.
Karni, A., Meyer, G., Jezzard, P., Adams, M. M., Lerner, Y., Hendler, T., Ben-Bashat, D., Harel,
Turner, R., & Ungerleider, L. G. (1995 ). Func- M., & Malach, R. (2001). A hierarchical axis
tional MRI evidence for adult motor cortex of object processing stages in the human visual
plasticity during motor skill learning. Nature, cortex. Cerebral Cortex, 11(4), 287297.
3 77 , 15 5 15 8. Logothetis, N. K., Pauls, J., & Poggio, T. (1995 ).
Karni, A., & Sagi, D. (1991). Where practice Shape representation in the inferior temporal
makes perfect in texture discrimination: Evi- cortex of monkeys. Current Biology, 5 , 5 5 2
dence for primary visual cortex plasticity. Pro- 5 63 .
ceedings of the National Academy of Science of the Maguire, E. A., Frackowiak, S. J., & Frith, C. D.
United States of America, 88(11), 49664970. (1997). Recalling routes around London: Acti-
Karni, A., Meyer, G., Rey-Hipolito, C., Jezzard, vation of the right hippocampus in taxi drivers.
P., Adams, M. M., Turner, R., & Ungerleider, The Journal of Neuroscience, 17 (18), 7103 7110.
L. G. (1998). The acquisition of skilled motor Maguire, E. A., Gadian, D. G., Johnsrude, I. S.,
performance: Fast and slow experience-driven Good, C. D., Richard, J. A., Frackowiak, S. J.,
changes in primary motor cortex. Proceedings of & Frith, C. D. (2000). Navigation-related struc-
the National Academy of Sciences of the United tural change in the hippocampi of taxi drivers.
States of America, 95 , 861868. Proceedings of the National Academy of Science
Kelly, A. M. C., & Garavan, H. (2005 ). Human United States of America, 97(8), 43 984403 .
functional neuroimaging of brain changes asso- Maguire, E. A., Spiers, H. J., Good, C. D., Hartley,
ciated with practice. Cerebral Cortex, 15 (8), T., Frackowiak, S. J., & Burgess, N. (2003 ). Nav-
10891102. igation expertise and the human hippocampus:
Klingberg, T. (1998). Concurrent performance of A structural brain imaging analysis. Hippocam-
two working memory tasks: Potential mecha- pus, 13 , 25 025 9.
nisms of interference. Cerebral Cortex, 8, 5 93 Malach, R., Reppas, J. B., Benson, R. R., Kwong,
601. K. K., Jiang, H., Kennedy, W. A., Ledden,
Klingberg, T., Forssberg, H., & Westerberg, H. P. J., Brady, T. J., Rosen, B. R., & Tootell,
(2002). Training of working memory in chil- R. B. (1995 ). Object-related activity revealed
dren with adhd. Journal of Clinical and Experi- by functional magnetic resonance imaging in
mental Neuropsychology, 2 4(6), 781791. human occipital cortex. Proceedings of the
Klingberg, T., Hedehus, M., Temple, E., Salz, T., National Academy of Sciences of the United
Gabrieli, J. D., Moseley, M. E., & Poldrack, States of America, 92 (18), 813 5 813 9.
R. A. (2000). Microstructure of temporo- Marcantoni, W. S., Lepage, M., Beaudoin, G.,
parietal white matter as a basis for reading Bourgouin, P., & Richer, F. (2003 ). Neural
brain changes in the development of expertise 681
correlates of dual task interference in rapid tion based on dynamic routing of information.
visual streams: An fMRI study. Brain and Cog- Journal of Neuroscience, 13 (11), 47004719.
nition, 5 3 , 3 183 21. Pashler, H (1994). Dual-task interference in sim-
McCandliss, B. D., Cohen, L. G., & Dehaene, S. ple tasks: Data and theory. Psychological Bul-
(2003 ). The visual word form area: Expertise letin, 116, 220244.
for reading in the fusiform gyrus. Trends in Cog- Pesenti, M., Thioux, M., Seron, X., & De Volder,
nitive Sciences, 7 (7), 293 299. A. (2000). Neuroanatomical substrates of Ara-
McCarthy, G., Puce, A., Gore, J. C., & Allison, T. bic number processing, numerical comparison
(1997). Face-specific processing in the human and simple addition: A PET study. Journal of
fusi-form gyrus. Journal of Cognitive Neuro- Cognitive Neuroscience, 12 , 461479.
science, 9(5 ), 605 610. Pesenti, M., Zago, L., Crivello, F., Mellet, E., Sam-
McCarthy, R., & Warrington, E. K. (1990). The son, D., Duroux, B., Seron, X., Mazoyer, B., &
dissolution of semantics. Nature, 3 43 (625 9), Tzourio-Mazoyer, N. (2001). Mental calcula-
5 99. tion expertise in a prodigy is sustained by right
Meyer, D. E., & Kieras, D. E. (1997). A computat- prefrontal and medial-temporal areas. Nature
ional theory of executive cognitive processes Neuroscience, 4(1), 103 107.
and multiple-task performance: Part I. Basic Phelps, E. A. (2001). Faces and races in the brain.
mechanisms. Psychological Review, 104, 265 . Nature Neuroscience, 4(8), 775 776.
Miozzo, M., & Caramazza, A. (1998). Vari- Poldrack, R. A. (2000). Imaging brain plasticity:
eties of pure alexia: the case of failure to Conceptual and methodological issues. Neu-
access graphemic representations. Cognitive roimage, 12 , 113 .
Neuropsychology, 15 (12), 203 23 8. Poldrack, R. A., Clark, J., Pare-Blagoev, E. J.,
Moscovitch, M., Wincour, G., & Behrmann, M. Shohamy, D., Moyano, J. C., Myers, C., &
(1997). What is special about face recognition? Gluck, M. A. (2001). Interactive memory sys-
Nineteen experiments on a person with visual tems in the human brain. Nature, 414, 5 46
object agnosia and dyslexia but normal face 5 5 0.
recognition. Journal of Cognitive Neuroscience, Poldrack, R. A., & Gabrieli, J. D. E. (2001). Char-
9(5 ), 5 5 5 604. acterizing the neural mechanisms of skill learn-
Muellbacher, W., Ziemann, U., Wissel, J., Dang, ing and repetition priming. Evidence from mir-
N., Kofler, M., Facchini, S., Borrojerdi, B., ror reading. Brain, 12 4, 6782.
Poewe, W., & Hallett, M. (2002). Early consoli- Price, C. J., Devlin, J. T. (2003 ). The myth of the
dation in human primary motor cortex. Nature, visual word form area. Neuroimage, 19, 473
415 , 640644. 481.
Nakamura, K., Sakai, K., & Hikosaka, O. (1998). Pugh, K. R., Mencl, W. E., Jenner, A. R., Katz,
Neuronal activity in medial frontal cortex L., Frost, S. J., Lee, J. R., Shaywitz, S. E., &
during learning of sequential procedures. Jour- Shaywitz, B. A. (2001). Neurobiological stud-
nal of Neurophysiology, 80, 26712687. ies of reading and reading disability. Journal of
Nakamura, K., Sakai, K., & Hikosaka, O. (1999). Communication Disorders, 3 9, 479492.
Effects of local inactivation of monkey medial Rainer, G., & Miller, E. K. (2000). Effects of visual
frontal cortex in learning of sequential proce- experience on the representation of objects
dures. Journal of Neurophysiology, 82 (2), 1063 in the prefrontal cortex. Neuron, 2 7 , 179
1068. 189.
OReilly, R. C., & Munakata, Y. (2000). Compu- Rizzolatti, G., Fadiga, L., Matelli, M., Betti-
tational Explorations in Cognitive Neuroscience: nardi, V., Paulesu, E., Perani, D., & Fazio,
Understanding the Mind by Simulating the F. (1996). Localization of grasp representa-
Brain. Cambridge, MA: MIT Press. tions in humans by pet: 1. Observation ver-
Olesen, P. J., Westerberg, H., & Klingberg, T. sus execution. Experimental Brain Research, 111,
(2004). Increased prefrontal and parietal activ- 103 111.
ity after training of working memory. Nature Sakai, K., & Miyashita, Y. (1991). Neural orga-
Neuroscience, 7 , 75 79. nization for the long-term memory of paired
Olshausen, B. A., Anderson, C. H., & Van associates. Nature, 3 5 4(63 49), 108109.
Essen, D. C. (1993 ). A neurobiological model of Sakai, K., & Miyashita, Y. (1994). Visual imagery:
visual attention and invariant pattern recogni- An interaction between memory retrieval and
682 the cambridge handbook of expertise and expert performance
focal attention. Trends in Neuroscience, 17 (7), the Organization for Human Brain Mapping,
287289. Toronto, Ontario.
Sandak, R., Mencl, W. E., Frost, S. J., Rueckl, Stelzel, C., Schumacher, E. H., Schubert, T.,
J. G., Katz, L., Moore, D. L., Mason, S. A., Ful- & DEsposito (in press). The neural effect of
bright, R. K., Constable, R. T., & Pugh, K. R. stimulus-response modality compatibility on
(2004). The neurobiology of adaptive learning dual-task performance: an fMRI study. Psycho-
in reading: A contrast of different training con- logical Research.
ditions. Cognitive, Affective, & Behavioral Neu- Szameitat, A. J., Lepsien, J., von Cramon,
roscience, 4(1), 6788. D. Y., Sterr, A., & Schubert, T. (in press).
Sanes, J. N., & Donoghue, J. P. (2000). Plasticity Task-order coordination in dual-task perfor-
and primary motor cortex. Annual Review of mance and the lateral pre-frontal cortex:
Neuroscience, 2 3 , 3 93 415 . an event-related fMRI study. Psychological
Schneider, W., & Chein, J. M. (2003 ). Controlled Research.
& automatic processing: Behavior, theory, and Szameitat, A. J., Schubert, T., Muller, K., & von
biological mechanisms. Cognitive Science, 2 7 , Cramon, D. Y. (2002). Localization of execu-
5 25 5 5 9. tive functions in dual-task performance with
Schneider, W., & Detweiler, M. (1988). The role fMRI. Journal of Cognitve Neuroscience, 14(8),
of practice in dual-task performance: Toward 11841199.
workload modeling in a connectionist/control Tanji, J., & Shima, K. (1994). Role for supple-
architecture. Human Factors, 3 0(5 ), 5 3 9 mentary motor area cells in planning several
5 66. movements ahead. Nature, 3 71(6496), 413
Schneider, W., & Shiffrin, R. M. (1977). Con- 416.
trolled and automatic human information pro- Tarr, M. J., & Gauthier, I. (2000). FFA: A flexible
cessing I: Detection, search, and attention. Psy- fusiform area for subordinate-level visual pro-
chological Review, 84(1), 166. cessing automatized by expertise. Nature Neu-
Shiffrin, R. M. (1988). Attention. In R. C. Atkin- roscience, 3 (8), 764769.
son, R. J. Herrnstein, G. Lindzey, & R. D. Luce Turkeltaub, P. E., Gareau, L., Flowers, D. L., Zef-
(Eds.), Stevens Handbook of Experimental Psy- firo, T. A., & Eden, G. F. (2003 ). Development
chology, 2 nd Edition (pp. 73 9811). New York: of neural mechanisms for reading. Nature Neu-
Wiley. roscience, 6, 767777.
Shiffrin, R. M., & Schneider, W. (1977). Con- Ungerleider, L. G., Doyon, J., & Karni, A. (2002).
trolled and automatic human information Imaging brain plasticity during motor skill
processing II: Perceptual learning, automatic learning. Neurobiology of Learning and Memory,
attending, and a general theory. Psychological 78, 5 5 3 5 64.
Review, 84(2), 127190. Worden, M., & Schneider, W. (1995 ). Cogni-
Shima, K., & Tanji, J. (2000). Neuronal activity tive task design for fMRI. International Jour-
in the supplementary and presupplementary nal of Imaging Systems and Technology, 6, 25 3
motor areas for temporal organization of multi- 270.
ple movements. Journal of Neurophysiology, 84, Yin, R. K. (1969). Looking at upside-down faces.
21482160. Journal of Experimental Psychology, 81(1), 141
Sigala, N., & Logothetis, N. K. (2002). Visual cat- 145 .
egorization shapes feature selectivity in the pri- Yin, R. K. (1970). Face recognition by brain-
mate temporal cortex. Nature, 415 , 3 183 20. injured patients. A dissociable ability? Neu-
Small, S., Hlustik, P., Chen, E., Dick, F., Gauthier, ropsychologia, 8, 3 95 402.
J., & Solodkin, A. (2005 ). Distributed popula- Yovel, G., & Kanwisher, N. (2004). Face
tion codes in the primary motor cortex of violinist. perception: Domain specific, not process spe-
Paper presented at the 11th Annual Meeting of cific. Neuron, 44, 889898.
C H A P T E R 38
K. Anders Ericsson
There are several factors that influence the description). According to this view, every
level of professional achievement. First and healthy person will improve initially through
foremost, extensive experience of activi- experience, but these improvements are
ties in a domain is necessary to reach very eventually limited by innate factors that
high levels of performance. Extensive expe- cannot be changed through training; hence
rience in a domain does not, however, invari- attainable performance is constrained by
ably lead to expert levels of achievement. ones basic endowments, such as abili-
When individuals are first introduced to a ties, mental capacities, and innate talents.
professional domain after completing their This general view also explains age-related
basic training and formal education, they declines in professional achievement, owing
often work as apprentices and are supervised to the inevitable degradation of general
by more-experienced professionals as they capacities and processes with age (see also
accomplish their work-related responsibili- Krampe & Charness, Chapter 40). More
ties. After months of experience, they typi- recently, researchers of expert performance
cally attain an acceptable level of proficiency, have found that there are many types of
and with longer experience, often years, experience and that these different types
they are able to work as independent pro- have qualitatively and quantitatively dif-
fessionals. At that time most professionals ferent effects on the continued acquisition
reach a stable, average level of performance, and maintenance of an individuals per-
and then they maintain this pedestrian level formance (Ericsson, 1996, 2002; Ericsson,
for the rest of their careers. In contrast, some Krampe, & Tesch-Romer, 1993 ). This frame-
continue to improve and eventually reach work proposes that some types of experi-
the highest levels of professional mastery. ence, such as merely executing proficiently
Traditionally, individual differences in the during routine work, may not lead to further
performance of professionals have been improvement, and that further improve-
explained by an account given by Galton ments depend on deliberate efforts to change
(1869/1979, see Ericsson, 2003 a, for a particular aspects of performance.
683
684 the cambridge handbook of expertise and expert performance
the execution of those skills, making inten- expertise, Simon and Chase (1973 ) proposed
tional modifications and adjustments diffi- that future experts gradually acquired pat-
cult (see Hill & Schneider, Chapter 3 7). terns and knowledge about how to react
In the automated phase of learning, perfor- in situations by storing memories of their
mance reaches a stable plateau, and no fur- past actions in similar situations. Hence, per-
ther improvements are observed in agree- formance is assumed to improve as a con-
ment with Galtons (1869/1979) assumption sequence of continued experience. Chase
of a performance limit. and Simons (1973 ) research on chess mas-
Similar phases of acquisition and autom- ters extended the pioneering work by de
atization have been shown to account for Groot (1946/1978) and demonstrated that
development in professional domains, such the masters recall for briefly presented reg-
as telegraphy (Bryan & Harter, 1897, 1899) ular game positions was vastly superior to
and typing (Book, 1925 a, 1925 b). Whereas less-skilled players. Simon and Chase (1973 )
initial proficiency in everyday and profes- argued that the masters must have acquired
sional skills may be attained within weeks some 5 0,000 chunks or patterns to enable
and months, development to very high lev- them to retrieve the appropriate moves for
els of achievement appear to require many the current position in a chess game. They
years or even decades of experience. In fact, highlighted the parallels between reaching
Bryan and Harter claimed already in 1899 this highly skilled performance in chess
that over ten years are necessary for becom- and acquiring other cognitive skills, such as
ing an expert. In their seminal theory of speaking a foreign language with its large
686 the cambridge handbook of expertise and expert performance
Reproducibly Superior Performance cians select their own pieces of music for
and Experience their performance. Similarly, the sequence
of moves in a chess game is never the
In many domains of expertise, individuals same and, thus, players never encounter
have been interested in assessing and com- the exact same positions during the mid-
paring levels of performance under fair and dle game. Fortunately, most domains of
controlled circumstances. For thousands of expertise require that experts be able to
years athletes have competed under highly exhibit superior performance for presented
standardized conditions in track and field representative situations. Ericsson and Smith
events, such as running, jumping, and throw- (1991; Ericsson, 1996) proposed a way to find
ing. These competitive conditions approach representative situations that capture the
the controlled conditions generated in mod- essence of expert performance in a domain
ern studies of performance in the labora- and call for immediate action. They also
tory. In a similar manner, musicians, dancers, described general methods for recreating
and chess players have a long history of dis- these situations in the laboratory and then
playing their performance under controlled instructing experts and less skilled individ-
conditions during competitions and tour- uals to reproduce their performance under
naments. Such competitions, together with controlled laboratory conditions, so that
similar tests, such as auditions, serve several investigators can identify the responsible
purposes beyond identifying the best per- mediating mechanisms.
formers and presenting awards. For younger Representative tasks that have been
and developing performers, successful per- found to capture the essence of expertise in
formance at competitions and auditions is three domains are illustrated in Figure 3 8.2
necessary to gain access to the best teachers (see treatments of these and other fields in
and training environments. Sections 5 ). In each example, the measured
Ericsson and Smith (1991) discussed how performance is closely related to the natu-
one could use similar techniques to mea- rally occurring performance. To study chess
sure various types of professional exper- expertise, players at different skill levels are
tise. We argued that a complete under- asked to generate the best move for the same
standing of the structure and acquisition of chess positions that have been taken from
excellence will be possible only in domains actual games between chess masters, which
in which experts exhibit objectively supe- are not publicly available. Different typists
rior performance, in a reproducible manner, are presented the same material and asked
for the representative activities that define to type as much as possible during a fixed
the essence of accomplishment in a given time period. Musicians are asked to play
domain (Ericsson, 1996, 2002). Expert per- familiar or unfamiliar pieces of music while
formers are accustomed to performing in being recorded, and are then asked to repeat
response to external demands, such as dur- their performance exactly. When musicians
ing emergencies in their professional prac- are instructed to repeat their original per-
tice, or at competitions and exhibitions. If formance, experts consecutive renditions
they are able to reproduce their performance show much less variation than renditions by
repeatedly on these types of occasions as less-skilled musicians and, by implication,
well during training, they should be able experts exhibit greater control over their
to reproduce them even under laboratory performance.
conditions, a finding confirmed by recent When a review of evidence is restricted to
research (Ericsson & Lehmann, 1996). only the reproducible superior performance
Unfortunately, expert performance of experts, obtained under directly compa-
occurs naturally in complex and unique rable conditions, it is possible to examine
contexts, where the conditions of perfor- several claims about the relation between
mance differ between performers, making expert performance and experience that
comparison difficult. For example, musi- generalize across domains. First, extensive
688 the cambridge handbook of expertise and expert performance
Figure 38.2 . Three examples of laboratory tasks that capture the consistently superior performance
of domain experts in chess, typing, and music. (From Expertise, by K. A. Ericsson and Andreas C.
Lehmann, 1999, Encyclopedia of Creativity. Copyright by Academic Press.)
performance in many vigorous sports is the fers across domains. For example, elite musi-
mid- to late 20s. For the arts and science, cians (disregarding the biased standards for
it is a decade later, in the 3 0s and 40s (see child prodigies) need closer to 20 to 3 0 years
Schulz & Curnow, 1988, and Simonton, of training and often peak when they are
1997 chapter 18, for reviews). The contin- around 3 0 to 40 years old. Further, outstand-
ued and often extended development of ing scientists and authors normally pub-
expertise past physical maturity shows that lished their first work at around age 25 , and
additional experience is necessary to attain their best work follows around ten years later
ones highest level of performance. (Raskin, 193 6).
Finally, the most compelling evidence Other investigators have pointed to
for the role of vast experience in expertise potential exceptions to the ten-year rule.
comes from investigators who have shown Some of the exceptions are so close to the
that, even for the most talented individ- ten-year rule that they support the neces-
uals, ten years of experience in a domain sity for around ten years to win at the inter-
(ten-year rule) is necessary to become an national level. For example, famous chess
expert (Bryan & Harter, 1899) and to win player Bobby Fischer required nine years
at an international level (Simon & Chase, of intense chess study before being recog-
1973 ). Subsequent reviews have shown that nized as a grand master in chess at age 16
these findings extend to international-level (Ericsson et al, 1993 ). Other examples sug-
success in music composition (Hayes, 1981), gest clearer violations. Very tall basketball
as well as to sports, science, and the arts players (around seven feet) have been able
(Ericsson, Krampe, & Tesch-Romer, 1993 ). to reach the highest professional ranks in
A closer examination of the evidence for less than ten years of training in around
the ten-year rule shows that the number six years. Research on training of memory
ten is not magical. In fact, the number of experts has shown that individuals can reach
years of intense training required to become the highest level in the world after less than
an internationally acclaimed performer dif- a couple of years training (Ericsson, 2003 b;
690 the cambridge handbook of expertise and expert performance
Figure 38.4. Estimated amount of time for solitary practice as a function of age
for the middle-aged professional violinists (triangles), the best expert violinists
(squares), the good expert violinists (empty circles), the least accomplished expert
violinists (filled circles), and amateur pianists (diamonds). (From The role of
deliberate practice in the acquisition of expert performance, by K. A. Ericsson,
R. Th. Krampe, and C. Tesch-Romer, 1993 , Psychological Review, 100(3 ), p. 3 79
and p. 3 84. Copyright 1993 by American Psychological Association. Adapted
with permission.)
musicians had spent over 10,000 hours prac- back. Hence, the requirement for concentra-
ticing, which averages 2,5 00 and 5 ,000 hours tion sets deliberate practice apart from both
more than two less-accomplished groups of mindless, routine performance and playful
musicians at the same academy, respectively engagement, as the latter two types of activ-
(Ericsson et al., 1993 ). In comparison to ama- ities would, if anything, merely strengthen
teur pianists of the same age (Krampe & the current mediating cognitive mecha-
Ericsson, 1996), the best musicians from the nisms, rather than modify them to allow
academy and the professionals had practiced increases in the level of performance. Re-
8,000 more hours. search is currently reevaluating claims that
The core assumption of deliberate prac- some individuals can improve their level
tice (Ericsson, 1996, 2002, 2004; Ericsson of performance without concentration and
et al., 1993 ) is that expert performance deliberate practice. Even the well-known
is acquired gradually and that effective fact that more talented children improve
improvement of performance requires the faster in the beginning of their music devel-
opportunity to find suitable training tasks opment appears to be in large part due to
that the performer can master sequentially the fact that they spend more time in delib-
typically the design of training tasks and erate practice each week (Sloboda, David-
monitoring of the attained performance is son, Howe & Moore, 1996). In a recent
done by a teacher or a coach. Deliberate study of singers Grape, Sandgren, Hans-
practice presents performers with tasks that son, Ericsson, and Theorell (2003 ) revealed
are initially outside their current realm of reliable differences of skill in the level of
reliable performance, yet can be mastered physiological and psychological indicators of
within hours of practice by concentrating concentration and effort during a singing
on critical aspects and by gradually refining lesson. Whereas the amateur singers expe-
performance through repetitions after feed- rienced the lesson as self-actualization and
experience and deliberate practice 693
physiological function will be discussed, fol- induced metabolic demands of the activity
lowed by a description of how practice can by, for example, increases in the number of
improve performance by changes in cogni- capillaries supplying blood to muscles and
tive mechanisms that mediate performance changes in metabolism of the muscle fibers
and further learning. themselves. These adaptations will eventu-
ally allow the individual to execute the given
level of activity without greatly straining
Improving Adaptations by Straining
the physiological systems. To gain further
Physiological Systems
beneficial increases in adaptation, the ath-
Measurable increases in physical fitness do letes need to increase or change their weekly
not simply result from wishful thinking. training activities to induce new and perhaps
Instead people have to engage in intense aer- different types of strain on the key phys-
obic exercise that pushes them well beyond iological systems. For example, improve-
the level of comfortable physical activity ments of strength are attained when indi-
if they are to improve their aerobic fit- viduals lift weights to induce brief maximal
ness (Ericsson, 2003 a; Ericsson et al., 1993 ; efforts of the targeted muscle groups. More
Robergs & Roberts, 1997). Specifically, in generally, athletic training involves push-
order to increase their aerobic fitness young ing the associated physiological systems out-
adults have to exercise at least a couple of side the comfort zone to stimulate phys-
times each week, for at least 3 0 minutes iological growth and adaptation (Ericsson,
per session, with a sustained heart rate that 2001, 2002, 2003 a, 2003 c, 2003 d). Further-
is 70% of their maximal level (around 140 more, recent reviews (Gaser & Schlaug,
beats per minute for a maximal heart rate of 2003 ; Hill & Schneider, Chapter 3 7; Kolb
200). When the human body is put under & Whishaw, 1998) show that the func-
exceptional strain, a range of dormant genes tion and structure of the brain is far more
in the DNA are expressed and extraordi- adaptable than previously thought possi-
nary physiological processes are activated. ble. Especially, early and extended training
Over time the cells of the body, includ- has shown to change the cortical mapping
ing the brain (see Hill & Schneider, Chap- of musicians (Elbert, Pantev, Wienbruch,
ter 3 7) will reorganize in response to the Rockstoh, & Taub, 1995 ), the development
696 the cambridge handbook of expertise and expert performance
performance (and motivation) through a Book, W. F. (1925 a). Learning to typewrite. New
high level of daily practice they can sustain York: The Gregg Publishing Co.
for days, months, and years. The emerging Book, W. F. (1925 b) The psychology of skill. New
insights should be relevant to any motivated York: The Gregg Publishing Co.
individual aspiring to excel in any challeng- Bryan, W. L., & Harter, N. (1897). Studies in the
ing domain (Ericsson, 2004). Although we physiology and psychology of the telegraphic
are already gaining understanding about how language. Psychological Review, 4, 275 3 .
performers improve with deliberate practice Bryan, W. L., & Harter, N. (1899). Studies on the
and reach expert levels, it is unlikely that telegraphic language: The acquisition of a hier-
we will ever be able to fully understand and archy of habits. Psychological Review, 6, 3 45
predict future innovations. We may be able 3 75 .
to reproduce the path of development that Camerer, C. F., & Johnson, E. J. (1991). The
elite performers have taken to reach their process-performance paradox in expert
judgment: How can the experts know so much
highest levels of performance in the past. We
and predict so badly? In K. A. Ericsson and
may also be able to help performers in one J. Smith (Eds.), Towards a general theory of
domain of expertise, such as surgery, learn expertise: Prospects and limits (pp. 195 217).
about the best training methods that have Cambridge: Cambridge University Press.
been developed in domains with a longer Chaffin, R., & Imreh, G. (1997). Pulling teeth
tradition, such as violin performance. We and torture: Musical memory and prob-
may even be able to work in collaboration lem solving. Thinking and Reasoning, 3 , 3 15
with world-class performers who are work- 3 3 6.
ing on improving their performance to new Charness, N., Krampe, R. Th., & Mayr, U.
and undiscovered heights. At the highest (1996). The role of practice and coaching in
levels of expert performance, the drive for entrepreneurial skill domains: An international
improvement will always involve search and comparison of life-span chess skill acquisition.
experimentation at the threshold of under- In K. A. Ericsson (Ed.), The road to excellence:
standing, even for the masters dedicated to The acquisition of expert performance in the arts
redefining the meaning of excellence in their and sciences, sports, and games (pp. 5 180).
Mahwah, NJ: Erlbaum.
fields.
Charness, N., Tuffiash, M. I., Krampe, R., Rein-
gold, E., & Vasyukova E. (2005 ). The role of
deliberate practice in chess expertise. Applied
References Cognitive Psychology, 19, 15 1165 .
Chi, M. T. H., Glaser, R., & Farr, M. J. (Eds.)
Abernethy, B. (1991). Visual search strategies and (1988). The nature of expertise. Hillsdale, NJ:
decision-making in sport. International Journal Erlbaum.
of Sport Psychology, 2 2 , 189210. Cote, J., Ericsson, K. A., & Law, M. (2005 ). Trac-
Anderson, J. R. (1982). Acquisition of cognitive ing the development of athletes using retro-
skill. Psychological Review, 89, 3 69406. spective interview methods: A proposed inter-
Bedard, J., & Chi, M. T. H. (1993 ). Expertise in view and validation procedure for reported
auditing. Auditing, 12 , (Suppl.), 125 . information. Journal of Applied Sport Psychol-
Bengtsson, S. L., Nagy, Z., Skare, S., Forsman, ogy, 17 , 119.
L., Forsberg, H., & Ullen, F. (2005 ). Extensive Cowley, M. (Ed.) (195 9). Writers at work: The
piano practicing has regionally specific effects Paris review interviews. New York: Viking.
on white matter development. Nature Neuro- Dawes, R. M. (1994). House of cards: Psychology
science, 8, 1148115 0. and psychotherapy built on myth. New York:
Bloom, B. S. (1985 a). Generalizations about talent Free Press.
development. In B. S. Bloom (Ed.), Developing Deakin, J. M., & Cobley, S. (2003 ). A search
talent in young people (pp. 5 075 49). New York: for deliberate practice: An examination of the
Ballantine Books. practice environments in figure skating and vol-
Bloom, B. S. (Ed.) (1985 b). Developing talent in leyball. In J. Starkes & K. A. Ericsson (Eds.),
young people. New York: Ballantine Books. Expert performance in sport: Recent advances in
experience and deliberate practice 701
research on sport expertise (pp. 115 13 5 ). Cham- Ericsson, K. A. (2003 b). Exceptional memo-
paign, IL: Human Kinetics. rizers: Made, not born. Trends in Cognitive Sci-
de Groot, A. (1978). Thought and choice in chess ences, 7 , 23 3 23 5 .
(Original work published in 1946). The Hague: Ericsson, K. A. (2003 c). The development of
Mouton. elite performance and deliberate practice: An
Denison, J. (2003 ). Bannister and beyond: The update from the perspective of the expert-
mystique of the four minute mile. Breakaway performance approach. In J. Starkes & K. A.
books. Ericsson (Eds.), Expert performance in sport:
Denney, N. W. (1982). Aging and cognitive Recent advances in research on sport exper-
changes. In B. B. Wolman (Ed.), Handbook tise (pp. 4981). Champaign, IL: Human
of developmental psychology (pp. 807827). Kinetics.
Englewood Cliffs, NJ: Prentice Hall. Ericsson, K. A. (2003 d). How the expert-per-
Doane, S. M., Pellegrino, J. W., & Klatzky, R. L. formance approach differs from traditional
(1990). Expertise in a computer operating approaches to expertise in sports: In search
system: Conceptualization and performance. of a shared theoretical framework for studying
Human-Computer Interaction, 5 , 2673 04. expert performance. In J. Starkes & K. A. Eric-
sson (Eds.), Expert performance in sport: Recent
Duffy, L. J., Baluch, B., & Ericsson, K. A. (2004).
advances in research on sport expertise (pp. 3 71
Dart performance as a function of facets of
401). Champaign, IL: Human Kinetics.
practice amongst professional and amateur
men and women players. International Journal Ericsson, K. A. (2004). Deliberate practice and
of Sport Psychology, 3 5 , 23 2245 . the acquisition and maintenance of expert per-
formance in medicine and related domains.
Dvorak, A., Merrick, N. L., Dealey, W. L., & Ford,
Academic Medicine, 10, s70S81.
G. C. (193 6). Typewriting behavior. New York:
American Book Company. Ericsson, K. A., Delaney, P. F., Weaver, G., &
Mahadevan, R. (2004). Uncovering the struc-
Elbert, T., Pantev, C., Wienbruch, C., Rockstroch,
ture of a memorists superior basic memory
B., & Taub, E. (1995 ). Increased cortical repre-
capacity. Cognitive Psychology, 49, 19123 7.
sentation of the fingers of the left hand in string
players. Science, 2 70, 3 05 3 07. Ericsson, K. A., & Kintsch, W. (1995 ). Long-
term working memory. Psychological Review,
Ericsson, K. A. (1996). The acquisition of expert
102 , 211245 .
performance: An introduction to some of the
issues. In K. A. Ericsson (Ed.), The road to Ericsson, K. A., Krampe, R. Th., & Tesch-Romer,
excellence: The acquisition of expert performance C. (1993 ). The role of deliberate practice in the
in the arts and sciences, sports, and games acquisition of expert performance. Psychologi-
(pp. 15 0). Mahwah, NJ: Erlbaum. cal Review, 100, 3 63 406.
Ericsson, K. A. (2001). The path to expert Ericsson, K. A., & Lehmann, A. C. (1996). Ex-
performance: Insights from the masters on how pert and exceptional performance: Evidence
to improve performance by deliberate practice. on maximal adaptations on task constraints.
In P. Thomas (Ed.), Optimizing performance in Annual Review of Psychology, 47 , 273 3 05 .
golf (pp. 15 7). Brisbane, Australia: Australian Ericsson, K. A., Patel, V. L., & Kintsch, W. (2000).
Academic Press. How experts adaptations to representative
Ericsson, K. A. (2002). Attaining excellence task demands account for the expertise effect
through deliberate practice: Insights from the in memory recall: Comment on Vicente and
study of expert performance. In M. Ferrari Wang (1998). Psychological Review, 107 , 5 78
(Ed.), The pursuit of excellence in education 5 92.
(pp. 215 5 ). Hillsdale, NJ: Erlbaum. Ericsson, K. A., & Smith, J. (1991). Prospects
Ericsson, K. A. (2003 a). The search for general and limits in the empirical study of expertise:
abilities and basic capacities: Theoretical impli- An introduction. In K. A. Ericsson & J. Smith
cations from the modifiability and complexity (Eds.), Toward a general theory of expertise:
of mechanisms mediating expert performance. Prospects and limits (pp. 13 8). Cambridge:
In R. J. Sternberg & E. L. Grigorenko (Eds.), Cambridge University Press.
Perspectives on the psychology of abilities, compe- Feldman, D. H. (1994). Beyond universals in cog-
tencies, and expertise (pp. 93 125 ). Cambridge: nitive development (2nd. Ed.). Norwood, NJ:
Cambridge University Press. Ablex.
702 the cambridge handbook of expertise and expert performance
Fitts, P., & Posner, M. I. (1967). Human perfor- Kolb, B., & Whishaw, I. Q. (1998). Brain plasticity
mance. Belmont, CA: Brooks/Cole. and behavior. Annual Review of Psychology, 49,
Galton, F., Sir (1869/1979). Hereditary genius: 43 64.
An inquiry into its laws and consequences Lehmann, A. C., & Ericsson K. A. (1998).
(Originally published in 1869). London: Julian The historical development of domains of
Friedman Publishers. expertise: Performance standards and innova-
Gaser, C., & Schlaug, G. (2003 ). Brain struc- tions in music. In A. Steptoe (Ed.), Genius
tures differ between musicians and non- and the mind (pp. 6794). Oxford, UK: Oxford
musicians. Journal of Neuroscience, 2 3 , 9240 University Press.
9245 . McDaniel, M. A., Schmidt, F. L., & Hunter, J. E.
Gawel, R. (1997). The use of language by trained (1988). Job experience correlates of job perfor-
and untrained experienced wine tasters. Jour- mance. Journal of Applied Psychology, 73 , 3 27
nal of Sensory Studies, 12 , 267284. 3 3 0.
Glaser, R. (1996). Changing the agency for McKinney, E. H., & Davis, K. J. (2004). Effects
learning: Acquiring expert performance. In of deliberate practice on crisis decision perfor-
K. A. Ericsson (Ed.), The road to excellence: The mance. Human Factors, 45 , 43 6444.
acquisition of expert performance in the arts and Moulaert, V., Verwijnen, M. G. M., Rikers, R.,
sciences, sports, and games (pp. 15 0). Mahwah, & Scherpbier, A. J. J. A. (2004). The effects
NJ: Erlbaum. of deliberate practice in undergraduate medical
Gould, S. J. (1996). Full house: The spread of excel- education. Medical Education, 3 8, 1044105 2.
lence from Plato to Darwin. New York: Harmony Nielsen, S. (1999). Regulation of learning strate-
books. gies during practice: A case study of a single
Grape, C., Sandgren, M., Hansson, L.-O., church organ student preparing a particular
Ericson, M., & Theorell, T. (2003 ). Does work for a concert performance. Psychology of
singing promote well-being?: An empirical Music, 2 7 , 218229.
study of professional and amateur singers dur- Plant, E. A., Ericsson, K. A., Hill, L., & Asberg, K.
ing a singing lesson. Integrative Physiological & (2005 ). Why study time does not predict grade
Behavioral Science, 3 8, 65 71. point average across college students: Implica-
Gruson, L. M. (1988). Rehearsal skill and musi- tions of deliberate practice for academic perfor-
cal competence: Does practice make perfect? mance. Contemporary Educational Psychology,
In J. A. Sloboda (Ed.), Generative processes 3 0, 96116.
in music (pp. 91112). Oxford, UK: Clarenden Plimpton, G. (Ed.) (1977). Writers at work: The
Press. Paris review. Interviews, Second Series. New
Hayes, J. R. (1981). The complete problem solver. York: Penguin.
Philadelphia, PA: Franklin Institute Press. Raskin, E. (193 6). Comparison of scientific and
Helsen, W. F., Starkes, J. L., & Hodges, N. J. literary ability: A biographical study of eminent
(1998). Team sports and the theory of delib- scientists and letters of the nineteenth century.
erate practice. Journal of Sport and Exercise Psy- Journal of Abnormal and Social Psychology, 3 1,
chology, 2 0, 123 4. 203 5 .
Hodges, N. J., & Starkes, J. L. (1996). Wrestling Reif, F., & Allen, S. (1992). Cognition for inter-
with the nature of expertise: A sport spe- preting scientific concepts: A study of acceler-
cific test of Ericsson, Krampe and Tesch- ation. Cognition and Instruction, 9, 144.
Romers (1993 ) theory of Deliberate Prac- Robergs, R. A., & Roberts, S. O. (1997). Exer-
tice. International Journal of Sport Psychology, cise physiology: Exercise, performance, and clin-
27, 400424. ical applications. St. Louis, MO: Mosby-Year
Hoffman, R. R. (Ed.) (1992). The psychology of Book.
expertise: Cognitive research and empirical AI. Rosson, M. B. (1985 ). The role of experi-
New York: Springer-Verlag. ence in editing. Proceedings of INTERACT 84
Krampe, R. Th., & Ericsson, K. A. (1996). Main- IFIP Conference on Human-Computer Interac-
taining excellence: Deliberate practice and tion (pp. 45 5 0). New York: Elsevier.
elite performance in young and older pianists. Salthouse, T. A. (1984). Effects of age and skill
Journal of Experimental Psychology: General, in typing. Journal of Experimental Psychology:
12 5 , 3 3 13 5 9. General, 113 , 3 45 3 71.
experience and deliberate practice 703
Schneider, W. (1993 ). Acquiring expertise: Deter- sports, and games (pp. 81106). Mahwah, NJ:
minants of exceptional performance. In K. A. Erlbaum.
Heller, J. Monks, & H. Passow (Eds.), Interna- Ulijaszek, S. J., Johston, F. E., & Preece, M. A.
tional handbook of research and development of (Eds.) (1998). The Cambridge encyclopedia of
giftedness and talent (pp. 3 113 24). Oxford, UK: human growth and development. Cambridge,
Pergamon Press. UK: Cambridge University Press.
Shanteau, J., & Stewart, T. R. (1992). Why Valentin, D., Pichon, M., de Boishebert, V., &
study expert decision making? Some histori- Abdi, H. (2000). Whats in a wine name? When
cal perspectives and comments. Organizational and why do wine experts perform better than
Behaviour and Human Decision Processes, 5 3 , novices? Abstracts of the Psychonomic Society, 5 ,
95 106. 3 6.
Schulz, R., & Curnow, C. (1988). Peak perfor- Vicente, K. J., & Wang, J. H. (1998). An ecological
mance and age among superathletes: Track and theory of expertise effects in memory recall.
field, swimming, baseball, tennis, and golf. Jour- Psychological Review, 105 , 3 3 5 7.
nal of Gerontology: Psychological Sciences, 43 , Ward. P., Hodges, N. J., Williams, A. M., &
113 120. Starkes, J. L. (2004). Deliberate practice and
Simon, H. A., & Chase, W. G. (1973 ). Skill in expert performance: Defining the path to
chess. American Scientist, 61, 3 94403 . excellence. In A. M. Williams & N. J. Hodges
Simonton, D. K. (1997). Creative productivity: A (Eds.), Skill acquisition in sport: Research, the-
predictive and explanatory model of career tra- ory and practice (pp. 23 125 8). London, UK:
jectories and landmarks. Psychological Review, Routhledge.
104, 6689. Williams, M., & Davids, K. (1995 ). Declarative
Singer, C. (195 8). From magic to science. New knowledge in sport: A by-product of experi-
York: Dover. ence or a characteristic of expertise? Journal of
Sloboda, J. A., Davidson, J. W., Howe, M. J. A., & Sport & Exercise Psychology, 17, 25 9275 .
Moore, D. G. (1996). The role of practice in the Williams, A. M., Ward, P., Knowles, J. M., &
development of performing musicians. British Smeeton, N. J. (2002). Anticipation skill in
Journal of Psychology, 87 , 2873 09. a real-world task: Measurement, training, and
Sonnentag, S. (1998). Expertise in professional transfer in tennis. Journal of Experimental
software design: A process study. Journal of Psychology: Applied, 8, 25 9270.
Applied Psychology, 83 , 703 715 .
Sonnentag, S., & Kleine, B. M. (2000). Deliber-
ate practice at work: A study with insurance Author Notes
agents. Journal of Occupational and Organiza-
tional Psychology, 73 , 87102. This research was supported by the
Starkes, J. L., Deakin, J., Allard, F., Hodges, N. J., FSCW/Conradi Endowment Fund of Florida
& Hayes, A. (1996). Deliberate practice in State University Foundation. The author wants
sports: What is it anyway? In K. A. Ericsson to thank Andreas Lehmann, Len Hill, Robert
(Ed.), The road to excellence: The acquisition Hoffman, and Paul Ward for the helpful
of expert performance in the arts and sciences, comments on earlier drafts of the chapter.
C H A P T E R 39
Barry J. Zimmerman
The attainment of expertise in diverse fields assignments. Expert musicians rated both
requires more than nascent talent, initial lessons with their teacher and their solitary
task interest, and high-quality instruction; practice as two keys to their improvement,
it also involves personal initiative, diligence, but only the latter was solely under their
and especially practice. Both the quality control (Ericsson, Krampe, & Tesch-Romer,
and quantity of an experts practice have 1993 ). Interestingly, the quantity of deliber-
been linked directly to acquisition and ate practice, but not total amount of music-
maintenance of high levels of performance related activity, was predictive of the musi-
(Ericsson, 1996, Ericsson, Chapter 3 8). cians acquisition and maintenance of expert
Regarding its quality, the practice of experts performance. Ericsson (2003 ) has discussed
is characterized by its conscious deliberate a persons attempts to acquire expertise
properties namely, a high level of con- as deliberate problem solving because they
centration and the structuring of specific involve forming a cognitive representation of
training tasks to facilitate setting appropriate the task, choosing appropriate techniques or
personal goals, monitoring informative feed- strategies, and evaluating ones effectiveness.
back, and providing opportunities for repeti- These properties of deliberate practice (e.g.,
tion and error correction (Ericsson, Krampe, task analysis, goal setting, strategy choice,
& Tesch-Romer, 1993 ). Deliberate attention self-monitoring, self-evaluations, and adap-
(i.e., strategic awareness) is believed to be tations) have been studied as key compo-
necessary to overcome prior habits, to self- nents of self-regulation (Boekaerts, Pintrich,
monitor accurately, and to determine neces- & Zeidner, 2000; Schunk & Zimmerman,
sary adjustments. 1998; Winne, 1997; Zimmerman & Schunk,
Although a skilled teacher typically struc- 2001). Self-regulation is defined formally
tures these desirable dimensions of prac- as self-generated thoughts, feelings, and
tice episodes, a student must implement actions that are strategically planned and
them on his or her own before return- adapted to the attainment of personal goals
ing to the teacher for evaluation and new (Zimmerman, 1989). Feedback from ones
705
706 the cambridge handbook of expertise and expert performance
Strategy Use
Feedback Loop
Person
Environment Behavior
Environmental
Self-Regulation
form a relaxing mental image to counter- enhancing cycles of motivation, whereas the
act the pressure. For all three self-regulatory initial task interest of their undistinguished
elements, peoples accuracy and constancy peers fails to sustain dedicated learning and
in self-monitoring of outcomes positively practice? To explain self-enhancing cycles of
influence the effectiveness of their strate- learning, social cognitive researchers (Ban-
gic adjustments and the nature of their dura, 1991; Zimmerman, 2000) have pro-
self-beliefs, such as perceptions of self- posed that self-regulatory processes are
efficacy their self-belief in their capabil- linked to key self-motivational beliefs dur-
ity to perform effectively (Schunk, 1983 b; ing three cyclical phases: forethought, per-
Zimmerman & Kitsantas, 1997). The lat- formance control, and self-reflection (see
ter belief, in turn, is a major source of Figure 3 9.2).
motivation to self-regulate ones functioning The forethought phase involves learn-
(Bandura, 1997; Zimmerman, 1995 ), and ing processes and motivational beliefs that
its cyclical role during self-regulation, along precede and can enhance efforts to learn,
with that of other key self-motivational practice, and perform. The performance
beliefs, is discussed next. phase involves use of processes to improve
the quality and quantity of learning, prac-
tice, and performance, and the self-reflection
A Cyclical Phase View of Self-Regulatory phase involves processes that occur after
Processes and Motivational Beliefs efforts to learn, practice, or perform that
Blooms (1985 ) study revealed that tal- influence a learners cognitive and behav-
ented youth were distinguished by their ioral reactions to that experience. These
initial attraction to their field from first self-reflections, in turn, influence a persons
exposure and by their increasing practice forethought processes and beliefs regard-
time. Their successes led them or their ing subsequent learning, which completes
parents to seek instruction from master the self-regulatory cycle. Although all learn-
teachers. But why does the initial task inter- ers attempt to self-regulate their personal
est of these talented youths lead to self- functioning in some way, developing experts
708 the cambridge handbook of expertise and expert performance
Performance Phase
Self-Control
Task strategies
Imagery
Self-instructions
Time management
Environmental structuring
Help seeking
Self-Observation
Metacognitive self-monitoring
Self-recording
Figure 39.2 . Phases and subprocesses of self-regulation. From Motivating self-regulated problem
solvers, by B. J. Zimmerman & M. Campillo, 2003 , in J. E. Davidson & R. J. Sternberg (Eds.), The
nature of problem solving (p. 23 9). New York: Cambridge University Press. Copyright by Cambridge
University Press. Reprinted with permission.
focus proactively on learning processes (i.e., on women volleyball players (Kitsantas &
as a means to an end) during the forethought Zimmerman, 2002) has shown that experts
and performance control phases, rather than set more specific technique or processes
only reactively on personal outcomes during goals for themselves than non-experts. For
self-reflection (Cleary & Zimmerman, 2001). example, experts reported technique goals
We address these issues next. such as toss the ball properly, whereas
non-experts reported general goals such as
concentrate, and novice learners fail to set
forethought phase goals for themselves at all. In other research,
To prepare to perform at their desired level, learners who set a combination of process
aspiring learners or their instructors analyze and outcome strategies performed better
the learning tasks in order to set appropri- than learners who set singular goals (Filby,
ate practice goals and plan an effective strat- Maynard, & Graydon, 1999; Kingston &
egy for attaining those goals (Ericsson, 1996). Hardy, 1997). Process goals refer to improv-
The self-regulatory process of goal setting ing ones strategy or technique, whereas out-
refers to specifying intended actions or out- come goals refer to enhancing the results of
comes (Locke & Latham, 2002). Research performance, such as points won or applause
development and adaptation of expertise 709
Golen, 1991; Nolen, 1988). The tennis cham- athletes, such as the diver Greg Louganis,
pion Monica Seles described her learning the decathalete Bruce Jenner, the golfer Jack
goal orientation in the following way: I Nicklaus, and the tennis player Chris Evert,
really never enjoyed playing matches, even consider mental imagery to be an asset in
as a youngster. I just love to practice and drill their training (Loehr, 1991). Athletic per-
and that stuff. I just hate the whole thought formers who imagine themselves as success-
that one [player] is better than the other. It ful have reported higher levels of motivation
drives me nuts (Vecsey, 1999, p. D1). and performance than those who do not use
this technique (Munroe, Giacobbi, Hall, &
Weinberg, 2000). Donald Murray (1990), a
performance phase Pulitzer Prize-winning writer, used imagery
Experts advantageous goals, strategic plan- in similar fashion: I see what I write and
ning, and motivational beliefs during the many times the focus of my writing is in my
forethought phase lead to the self-controlled image (p. 97).
and self-observed implementation of these Task strategies refer to advantageous
strategies, methods, or techniques during the methods for learning or performing particu-
performance phase. However, forethought lar tasks. In the domain of academic learning,
phase task analyses that are superficial or an extensive number of task strategies, such
inaccurate, like those of many novices, can as mnemonics, cognitive maps, note-taking,
lead to ineffective or even counterproduc- and outlines, have been found to be effective
tive efforts to control performance phase (e.g., Schneider & Pressley, 1997; Weinstein
processes. Because strategies vary in their & Mayer, 1986). For example, as his task
situational effectiveness, they must be con- strategy, the American author Irving Wallace
stantly self-observed and adjusted, which (1971) prepared extensive notes and outlines
is the second class of performance phase before he began writing. Often task strate-
processes. gies are domain specific in their scope and
The first self-control process to be dis- are context specific in their effectiveness.
cussed is self-instruction. This form of self- For example, the concert pianist Alicia de la
talk refers to vocal or subvocal guidance of Rocha used the practice strategy of playing
ones performance, and there is evidence difficult passages very slowly and very softly
of its effectiveness in enhancing academic to improve her technique (Mach, 1991). As
(Meichenbaum, 1977; Schunk, 1986) and her technique on a passage became profi-
athletic (Hardy, Gammage, & Hall, 2001) cient, she modified her strategy and began
expertise. For example, with athletes who practicing at normal speed. This illustrates
have trouble controlling their negative out- the issue discussed earlier: The utility of
bursts, Loehr (1991), a sports psychologist a particular strategy needs to be carefully
at the elite Nick Bolletierri Tennis academy, monitored to ensure its optimal utilization.
recommended listing all of their negative Time management refers to estimating and
responses and finding a positive alternative budgeting ones use of time (Zimmerman,
for each one, such as saying let it go or Greenberg, & Weinstein, 1994), and experts
come on (p. 47) when they lose a point. often structure their practice and work time
However, self-directive verbalizations must carefully. For example, to improve the quan-
be adapted to task outcomes and gener- tity and quality of his writing, the German
ally should be faded as a skill is mastered poet Goethe recommended, Use the day
(Meichenbaum, 1977), or they can limit before the day. Early morning hours have
further improvement (Zimmerman & Bell, gold in their mouth (Murray, 1990, p. 16).
1972). Although professional writers differ in the
The self-regulatory process of imagery is timing of their optimal states for writing
used to create or recall vivid mental images (such as the morning), those who struc-
to assist learning and performance (Paivio, ture their writing time have reported evi-
1986). Approximately 80 to 85 % of expert dence of its effectiveness (Wallace & Pear,
development and adaptation of expertise 711
1977). Among student instrumental musi- iors were excluded and where a structured
cians, high achievers in annual competitions study method and self-reinforcement were
have reported a greater amount of prac- included (Fox, 1962). All the students in
tice time than low achievers (McPherson & the study reported increases in their grade
Zimmerman, 2002). Thus, time manage- points of at least one letter grade. These
ment can involve regulation of both the favorable results of environmental structur-
quality and quantity of time use. ing were replicated a decade later (Beneke
In research on students academic home- & Harris, 1972). Experts are very sensitive to
work (Zimmerman & Kitsantas, 2005 ), the the impact of their surroundings on qual-
quality of students study time was highly ity and quantity of their functioning. For
correlated with its quantity, and both indices example, the French poet and novelist Cen-
were highly predictive of students grade drars described his need to write in a quiet
point average. undistracted place, such as an enclosed room
It should be noted that implementation with the window shade pulled down. The
of these self-control strategies often involves American bike racer Lance Armstrong pre-
significant others, such as parents and teach- pared himself to win the Tour de France in
ers. The self-regulatory process of adaptive the mountainous sections of the racecourse
help seeking is defined as choosing specific by sleeping in a low oxygen tent to adapt
models, teachers, or books to assist one- himself physiologically to those conditions
self to learn (Newman, 1994). For exam- ahead of time (Lehrer, 2001, July 3 0).
ple, parents often structure practice envi- The key self-observation processes during
ronments for talented youth, and master the performance phase are metacognitive
teachers coach students how to improve monitoring and self-recording, which refer
their practice techniques (Bloom, 1985 ). respectively to mentally tracking or physi-
High academic achievers are not asocial in cally recording ones performance. Experts
their methods of practice, but, rather, selec- observe the implementation and effective-
tively seek instructional assistance in a self- ness of their self-control processes and out-
initiated adaptive manner. By contrast, low comes more systematically than non-experts
achievers are reluctant to seek help because or novices (Kitsantas & Zimmerman, 2002).
of their lack of planning and their resultant Metacognitive self-monitoring is difficult for
fear of adverse reactions from help-givers novices because the amount of information
(Karabenek, 1998). Among expert musi- involved in complex performances can eas-
cians, the concert pianist Janina Fialkowska ily overwhelm and can lead to inconsistent
frequently sought out Arthur Rubinstein as or superficial tracking. Experts are selec-
an exemplary model. He couldnt tell me tive in their cognitive self-monitoring during
how to do something, but he could demon- practice because of the specificity of their
strate how it should sound . . . So when Id learning, practice, and performance goals
play something that wasnt up to par, he (Abrahams, 2001). Experts recall of infor-
became very exasperated, and believe me he mation about a completed task has been
became exasperated very easily. Then hed found to be more accurate and complete
kick me off the bench and play it the way he than that of novices and less accomplished
thought it should be played (Mach, 1991, individuals in the same domain (Ericsson &
pp. 7980). Kintsch, 1995 ). Experts are also more likely
Environmental structuring, which refers to to recall pertinent or substantial information
selecting or creating effective settings for that is pitched at a higher level of abstrac-
learning or performance, is another impor- tion (see also Feltovich et al., Chapter 4,
tant self-control process. For example, stu- Expertise Involves Functional, Abstracted
dents who had difficulty concentrating dur- Representations).
ing studying were taught how to create The legendary golfer Bobby Jones (1966)
an effective study environment where day- described his method of monitoring as fol-
dreaming, eating, or other off-task behav- lows, It has never been possible for me to
712 the cambridge handbook of expertise and expert performance
think of more than two or three details of the dards. I had stopped trying to do a great
swing and still hit the ball correctly . . . The many difficult things perfectly because it had
two or three are not always the same, some- become clear in my mind that this ambi-
time a mans swing will be functioning so tious over-thoroughness was neither possi-
well he need worry about nothing (p. 203 ). ble nor advisable or even necessary (p. 113 ).
Experts can improve the accuracy of Conversely, individuals who fail to set chal-
their self-observations by self-recording lenging standards for themselves have dis-
their progress (see also, Deakin, Cote, & played lower levels of performance than per-
Harvey, Chapter 17). Literary experts, such sons who challenged themselves (Locke &
as Trollope (1905 ) and Hemingway (Wallace Latham, 2002).
& Pear, 1977), were acutely aware of the A second self-judgment that is hypothe-
value of self-recording in enhancing the sized to play a pivotal role in self-reflection
quantity of their literary output and con- involves the causal attribution of errors.
sistently utilized this technique. A persons For example, when errors are attributed to
records are more effective if they track not uncontrollable sources, such as an oppo-
only his or her performance but also the nents luck, learners display negative self-
conditions that surround it, and the results reactions and diminished attainment of skill.
that it produces (Zimmerman & Paulsen, By contrast, when errors are attributed to
1995 ; Ericsson, 1996). Unfortunately, controllable sources, such as ones strategies,
novices often self-record in a cursory and learners experience positive self-reactions
inaccurate way (Hallam, 1997). However, and increased skill (Zimmerman & Kitsantas,
it should be noted that record keeping 1999). Expert golfers have exhibited this
can be time consuming, and as a result, favorable pattern of attributions when dis-
its effectiveness needs to be monitored cussing differences between good and bad
carefully. After a skill has been mastered to rounds. They tend to discount the possibility
a personally acceptable level, people can that chance factors played an important role
often cease record keeping unless problems (Kirschenbaum, OConnor, & Owens, 1999)
arise (Zimmerman & Paulsen, 1995 ). and instead attribute their errors to per-
sonally controllable processes, such as poor
concentration, tenseness, poor imagination
self-reflection phase and feel (McCaffrey & Orlick, 1989). The
Experts increase the accuracy of their feed- swimmer Natalie Coughlin put it this way,
back by generating self-evaluative stan- In general, Im pretty inwardly focused . . . I
dards for themselves (Hamery, 1976). Self- like to concentrate on my stroke and do
evaluation judgments compare self-observed my race, because thats all I can control
information with one of three types of stan- (Grudowski, August, 2003 , p. 73 ). Novices
dards or criteria: (a) a self-improvement are prone to attributing causation for errors
criterion (e.g., comparing current efforts to such uncontrollable sources as a lack
to ones best previous effort), (b) a social of ability, task difficulty, or bad luck
comparison criterion (e.g., comparing ones Cleary & Zimmerman, 2001; Kitsantas &
efforts to those of competitors), or (c) a mas- Zimmerman, 2002). These unfortunate
tery criterion (e.g., comparing ones perfor- attributions occur because of novices poor
mance to a national record). Self-evaluations self-regulatory processes and beliefs during
are not automatic outcomes of performance the forethought and performance phases,
but, rather, depend on an individuals selec- such as vague goal setting, non-strategic
tion and interpretation of an appropriate cri- efforts to learn, and low perceptions of self-
terion (Bandura, 1991). When self-evaluative efficacy (Bandura, 1991).
standards are too high or too low, peo- Self-evaluation and attribution self-
ples learning and performance is diminished judgments are closely linked to two key
(Schunk, 1983 a). For example, the legendary self-reactions: self-satisfaction and adaptive
golfer Ben Hogan (195 7) warned about the inferences. Perceptions of self-satisfaction
dangers of setting unrealistically high stan- or dissatisfaction and associated emotions,
development and adaptation of expertise 713
familiarity with the game that novices typ- We found that goal setting was correlated
ically lack. Our methodology, called micro- with choice of strategy. Athletes who set
analysis, employs specific questions that outcome-specific goals (e.g., to make ten
address well-established psychological pro- out of ten) were more likely to select spe-
cesses at key points during the act of per- cific technique-oriented strategies (e.g., to
forming, such as self-efficacy and attribu- follow through), whereas those athletes set-
tion beliefs. Each participant is separately ting outcome-general goals (e.g., to make
observed, and researchers develop context- them) were more likely to select general
specific information by intensive qualitative technique strategies (e.g., to concentrate on
and quantitative analyses. In this study, the my form). It appears that teaching athletes
boys were questioned regarding their fore- to set specific goals can lead to their selection
thought phase goals, strategy choices, self- of specific strategies to achieve those goals.
efficacy beliefs, and intrinsic interest, as well A key finding about the self-reflection
as their self-reflection phase attributions and phase was that the boys attributions of
feelings of satisfaction as they practiced their errors to strategies were predictive of the
free-throw shooting. boys forethought strategy selections dur-
There were no significant differences ing further efforts to learn. For example,
between experts and non-experts in their boys who attributed their failure to spe-
frequency of practice, playing experience, cific techniques (i.e., I missed the last two
and knowledge of free-throw shooting tech- shots because my elbow was going to the
niques, but there were significant differ- left) were more likely to select a specific
ences in their methods of self-regulation technique-oriented strategy to improve their
during practice. As expected, novices dif- shooting accuracy (e.g., I need to keep
fered from experts and non-experts on all my elbow in). Overall, this study revealed
variables except age. It was found that highly significant differences in the quality
experts set more specific goals, selected more of self-regulation during self-directed prac-
technique-oriented strategies, made more tice efforts by high school basketball play-
attributions to strategy use, and displayed ers of varying ability. Experts were more
higher levels of self-efficacy than either non- focused on specific shooting processes dur-
experts or novices. When asked to self- ing goal setting, strategic planning, and self-
reflect after two consecutive misses, free- reflecting than non-experts or novices, and
throw experts were more mindful of their they were more self-efficacious about their
specific, technique-oriented flaws than boys performance.
in the other two groups. Although 60% of In a study of college womens volleyball
the experts indicated that they needed to practice, Anastasia Kitsantas and I (2002)
focus on their techniques (i.e., to keep my selected a group of experts from the univer-
elbow in, to follow through) in order to sity varsity volleyball team and a group of
make the next shot, only 20% of the non- non-experts from the university volleyball
experts and 7% of the novices mentioned club (i.e., who had been on the club team
this type of strategy. Non-experts preferred for at least three years). The group of
strategies related to the rhythm of shoot- novices had not ever participated in volley-
ing and general focus strategies (e.g., to ball as an organized sport but had played
concentrate or to try harder) for a major- it informally. The three groups of women
ity (i.e., 5 3 %) of their responses. Unfor- were questioned regarding their forethought
tunately, these self-reflections do not cor- phase goals, strategy choices, self-efficacy
rect faulty techniques because they divert beliefs, and intrinsic interest, as well as their
attention away from essential athletic form self-reflection phase attributions and feel-
processes. ings of satisfaction as they practiced their
To see if the results were consistent with a volleyball serves. It was found that experts
cyclical model, we analyzed relations among displayed better goals, planning, strat-
the boys use of self-regulatory processes. egy use, self-monitoring, self-evaluation,
development and adaptation of expertise 715
attributions, and adaptation than either non- group received the same forethought phase
experts or novices. Experts also reported goal setting and performance phase self-
higher self-efficacy beliefs, perceived instru- recording training as the three-phase group,
mentality, intrinsic interest, and self- but they were not instructed how to self-
satisfaction about volleyball serving than reflect. The one-phase self-regulation group
either non-experts or novices. The com- received instruction in only the forethought
bined 12 cyclical measures of self-regulation phase process of goal setting. There was also
explained 90% of the variance in the a practice-only control group and a no-practice
womens volleyball serving skill. Clearly, control group, which did not receive self-
experts differed greatly in self-regulation of regulation training. All of the participants
their practice methods. were randomly assigned to one of the five
conditions and were tested and trained indi-
vidually by an experimenter.
Development of Greater Expertise through
It was expected that one-phase train-
Multi-Phase Self-Regulatory Training
ing would influence subsequent phases
Although there was unambiguous evidence of self-regulation, and two-phase training
of superior self-regulation during athletic would influence self-reflection phase self-
practice by experts, the causal role of these regulation to some degree due to the cyclical
self-regulatory processes and beliefs in the dependence of later phase processes on ear-
development of expertise is another issue. lier phase processes, but we expected that
To develop free-throw expertise of male total phase training, including explicit train-
and female college students, Cleary, Zim- ing in self-reflection phase processes, would
merman, and Keating (in press) trained be optimal. Thus, a positive linear relation-
them to shoot basketball free-throws more ship was predicted between the students
effectively during their physical education free-throw shooting performance and the
classes. The participants assigned to a three- number of self-regulatory phases in which
phase self-regulation group were instructed they were trained.
to set technique goals (a forethought phase The results revealed that there were no
process), self-record (a performance phase gender differences in learning and that there
process), and to make strategic attribu- was in fact a linear relationship between
tions and adjustments following missed free amount of phase training and two key
throws (self-reflection phase processes). Set- measures of learning: free-throw shooting
ting technique goals involved focusing on accuracy and shooting adaptation. A more
properly executing the final four steps of the sensitive measure of shooting accuracy than
shooting process (i.e., grip, elbow position, simple making or missing the basket was
knee bend, follow through) rather than on developed. It involved earning one to five
shooting outcomes. The examiner showed points for each shot according to the follow-
the participants a cue card delineating the ing criteria: (a) five points for swishing the
process goal. This group was then taught shot (not hitting any part of the rim), (b)
how to use a self-recording form in order to four points for making the shot after hitting
monitor the step(s) of the strategy that they the rim, (c) three points for hitting the front
were focusing on while shooting the shots. or back of the rim but not making the shot,
This recording form also allowed the partic- (d) two points for hitting the side of the rim
ipants to monitor whether they missed any and not making the shot, and (e) one point
shots, the reasons for the missed shots, and for completely missing the rim or hitting the
strategies needed to make the next shot. In backboard first. A missed shot hitting the
addition to this self-reflection phase train- front or back of the rim earned more points
ing, the participants were taught how to link (i.e., three points) than a missed shot hit-
poor shots with one or more of the shooting ting the sides of the rim (i.e., two points)
techniques taught in the study. The partici- because the former indicated greater accu-
pants assigned to the two-phase self-regulation racy. Shooting adaptation referred to the
716 the cambridge handbook of expertise and expert performance
frequency of improvements on the next shot sports (Fox, Goudas, Biddle, Duda, & Arm-
following a poor shot. strong, 1994; Williams & Gill, 1995 ) and
The group means ranged in order from academic functioning (Ames, 1992; Pintrich,
lowest to highest as was predicted: no 2000).
practice control group, practice-only con- In terms of causal attributions and adap-
trol group, one-phase training, two-phase tive inferences, significantly more members
training, and three-phase training. This sug- of the multi-phase training group focused
gested that not only did the participants on specific shooting techniques or strate-
who received multiple-phase self-regulation gies following missed free throws, such as
training show greater accuracy when shoot- not keeping my elbow in and not touch-
ing, but they were also able to improve on ing my elbow to my side as I shot the
poor shots with a more successful throw on ball. In contrast, participants from the one-
a more consistent basis than those individu- phase training group or the practice control
als who received only one-phase or no self- group often attributed their misses to gen-
regulation training. Furthermore, the three- eral, non-technique factors, such as a lack
phase group and the two-phase group took of concentration or ability. These technique
significantly fewer practice shots than both attributions and adaptive inferences were
the one-phase and practice-only control associated with more accurate shooting per-
groups, perhaps because they were called formance on the posttest and greater shoot-
on to self-record their shooting techniques ing adaptation during practice. Thus, these
at various points during the practice ses- inexperienced free-throw shooters ability to
sion. Thus, the quality (i.e., defined in terms improve their poor free-throw shots during
of self-regulatory sophistication) of these practice was related to deficiencies in attri-
novices practice methods proved to be more butions and adapting these techniques dur-
important than the quantity of their prac- ing subsequent shot attempts. Focusing on
tice (i.e., number of shots taken) (see also controllable processes is important because
Ericsson, Chapter 3 8). it helps athletes become more aware of
This study focused particular attention what and how they are doing something
on the effects of self-regulation training rather than simply their level of attained
on the participants self-reflective phase outcomes (Cleary & Zimmerman, 2001;
self-judgments (i.e., attributions and self- Clifford, 1986).
evaluations) and self-reactions (i.e., adaptive In another study of multi-phase self-
inferences) to missed free throws because regulatory training, Anastasia Kitsantas and
they reveal how these learners think about I (Zimmerman & Kitsantas, 1997) exam-
their failures as well as their ability to ined the effects of multiple goal setting and
improve future performances. Learners who self-recording on the dart-throwing perfor-
received three-phase training displayed the mance and self-reflections with novice high
most adaptive motivational profile. For school girls. Girls in a process goal group
example, they evaluated their performance focused on practicing strategy steps for
based on personal processes (e.g., use of acquiring high-quality dart-throwing tech-
correct strategy or personal improvement) nique (e.g., the take-back, release, and
more frequently (60%) than all other groups: follow-through positions). By contrast, girls
two-phase group (10%), one-phase group in an outcome goal group focused on improv-
(20%), practice-only control group (20%), ing their scores. The bullseye on the tar-
and the no-practice control group (10%). get had the highest numerical value and
This is consistent with the self-regulation the surrounding concentric circles declined
cyclical phase hypothesis that using a process in value. Previous research had demon-
criterion to evaluate performance is linked to strated that process goals were more effec-
learning or mastery goal orientation, which tive than outcome goals with novice dart
has been found to be related to a variety of throwers (Zimmerman & Kitsantas, 1996).
motivational and achievement variables in From a multiple goal perspective, girls who
development and adaptation of expertise 717
shifted goals from processes to outcomes design of this study closely paralleled the
when automaticity was achieved should dart-throwing study, but in this case, the task
acquire more skill during practice than involved revising a series of writing problems
girls who adhere to only one goal (see drawn from a sentence-combining work-
Zimmerman, 2000). Automaticity was oper- book. These exercises involved transforming
ationally defined as performing the strategy a series of simple and often redundant kernel
steps without error for a specified number sentences into a single non-redundant sen-
of dart-throwing trials. tence. For example, the sentences: It was
Self-recording was taught to half of the a ball. The ball was striped. The ball rolled
girls in each goal group. Girls in the process- across the room could be rewritten as The
monitoring group recorded any strategy striped ball rolled across the room. The
steps they may have missed on each prac- entire group of experimental participants
tice throw, whereas girls in the outcome- was initially taught a three-step writing revi-
monitoring group wrote down their target sion strategy that involved identifying key
scores for each throw. Girls in the shifting- information, deleting duplicate information,
goal group changed their method of self- and combining the remaining words.
monitoring when they shifted goals. Before During a practice session following train-
being asked to practice on their own, all of ing, girls in a process goal group focused on
the girls were taught strategic components implementing the strategy for revising each
of the skill. Thus, the experiment compared writing task, whereas girls in an outcome
the effects of process goals, outcome goals, goal group focused on decreasing the num-
and shifting goals as well as self-recording ber of words in the revised passage, which
during self-directed practice. was the main outcome criterion. Process
The results were supportive of the multi- goals, which focused on strategy steps, had
ple goal hypothesis: Girls who shifted goals been found to be more effective than out-
from processes to outcomes surpassed class- come goals in prior writing research (Schunk
mates who adhered solely to either process & Swartz, 1993 ). As in the dart-throwing
or outcome goals in posttest dart-throwing study (Zimmerman & Kitsantas, 1997), the
skill. Girls who focused on outcome goals most effective goal setting condition was
exclusively were the lowest in dart-throwing expected to involve shifting from process
skill. Self-monitoring assisted learning for goals to outcome goals when automaticity
all goal-setting groups. In addition to their in performance was achieved. Half of the
superior learning outcomes, girls who shifted girls in each goal group were asked to self-
their goals displayed superior forms of record during practice. Girls in the process-
self-reflection than girls who adhered to monitoring group recorded strategy steps
either process or outcome goals exclusively. they missed on each of a series of revi-
The former girls attributed more errors to sion problems, whereas girls in the outcome-
controllable causes (i.e., to strategy use) monitoring group wrote down the number
and reported greater self-satisfaction than of words used on each problem. Girls in the
the latter girls. The girls in the shifting- shifting-goal group changed their method of
goal condition also exhibited superior fore- self-monitoring when they shifted goals.
thought phase motivational beliefs: These The results were supportive of a multi-
girls reported more positive self-efficacy ple goal hypothesis. Girls who shifted fore-
beliefs and greater interest in the dart throw- thought phase goals from processes to out-
ing than girls who adhered exclusively to comes surpassed the writing revision skill
either process or outcome goals. of girls who adhered exclusively to process
The same researchers conducted another goals or to outcome goals. Girls who focused
study of the effects of multiple-goal training on outcomes exclusively displayed the low-
and self-recording on the writing skill of girls est writing skill of the three goal groups.
attending an academically challenging high As in the dart-throwing study, self-recording
school (Zimmerman & Kitsantas, 1999). The enhanced writing skill for all goal-setting
718 the cambridge handbook of expertise and expert performance
groups. Forethought phase goals significantly mance. Second, several coaches rated the
increased the girls performance phase writ- cricketers performance for each match as a
ing skill and also their self-reflection phase subjective measure. Third, the players rated
attributions to strategy use and their self- their strategy use with five scales: imagery
satisfaction reactions. Performance phase ability, mental preparation, self-confidence,
self-recording also enhanced the girls writ- concentration, and activation.
ing skill and self-reflection phase attribu- It was discovered that the experimen-
tions and their self-satisfaction. The lat- tal training group significantly surpassed
ter two self-reflection phase processes were the control group in their level of per-
predictive of increases in the girls task formance, according to both the objective
interest and self-efficacy beliefs regarding and subjective measures. In addition, these
subsequent efforts to learn (i.e., their fore- cricket players displayed significantly greater
thought). These findings provided further consistency in their performance during
evidence of causality in cyclical relations the season according to subjective but not
among self-regulatory processes and self- objective measures of performance. Finally,
motivational beliefs. cricket players in the experimental group
The benefits of training in self-regulatory also reported significantly higher levels of
processes are not limited to novice learn- strategy use for each of the five scales of strat-
ers and regular students. Semi-professional egy use, which included a measure of self-
cricketers (Thelwell & Maynard, 2003 ) were confidence. Clearly, training in the optimal
trained in goal setting, self-talk, mental use of self-regulatory processes improved
imagery, concentration, and activation self- the performance of what many might regard
regulation strategies to improve their bat- as quite expert athletes. It appears that self-
ting and bowling skills. Goal setting involved regulatory training can benefit individuals
both process and outcome (i.e., multiple) across a wide range of expertise.
goals. Self-talk referred to positive self-
statements, task-relevant cues, and personal
Conclusion
goals. Mental imagery dealt with form-
ing mastery images of oneself designed to This chapter dealt with the role of self-
enhance both motivation and execution regulatory processes in the development of
of the skill. Concentration involved ignor- expertise. Although a childs initial inter-
ing distracting cues, especially when ones est in a field of endeavor usually grows
performance suffers, and focusing on task- from and is supported by parents and teach-
relevant cues. This superior concentration ers, his or her ultimate level of exper-
was expected to enhance the cricketers tise depends on self-disciplined practice and
self-confidence, which is similar to self- performance. Experts from diverse disci-
efficacy judgments (Thelwell & Maynard, plines, such as sport, music, and writing,
2002). Activation strategies involved try- rely on well-known self-regulatory processes
ing to create a mental and physical state to practice and perform. Variants of these
of optimal relaxation and alertness when self-regulatory processes can also assist aspir-
performing. ing learners to acquire both knowledge and
These strategies were taught to an exper- skill more effectively. For example, free-
imental group of cricketers between two throw shooters who set specific practice
seasons, during hour-long, weekly training goals, monitored their improvements in per-
sessions for 12 weeks. A control group of formance, and adjusted their shooting strat-
cricket players were trained in team build- egy appropriately learned more quickly than
ing or fielding activities during the training free-throw shooters who practiced without
sessions. Three types of dependent measures employing these self-regulatory processes
were studied. First, the cricketers batting (Cleary & Zimmerman, 2001).
or bowling scores during the matches were However, increases in ones use of self-
analyzed as an objective measure of perfor- regulatory processes will not immediately
development and adaptation of expertise 719
produce expert levels of knowledge and Anderson, D. (2004, September 19). Another
skill. Indeed, learners selection of goals deep hole and not enough rope. New York
and strategies will depend on their lev- Times, Sp, 7.
els of task knowledge and performance Ames, C. (1992). Achievement goals and the
skill, such as when the Olympic swim- classroom motivational climate. In D. H.
mer Natalie Coughlin self-regulated subtle Schunk & J. L. Meece (Eds.), Student percep-
hand positions to improve her performance tions in the classroom (pp. 3 273 48). Hillsdale,
NJ: Lawrence Erlbaum Associates.
(Grudowski, August, 2003 ), whereas a high
school swim team member might focus Bandura, A. (1986). Social foundations of thought
and action: A social cognitive theory. Englewood
on improving a more obvious skill. Many
Cliffs, NJ: Prentice Hall.
training texts, such as for skiing (Tejada-
Bandura, A. (1991). Self-regulation of motivation
Flores, 1986), have organized knowledge and
through anticipatory and self-reactive mecha-
skills into hierarchical levels, such as basic, nisms. In R. A. Dienstbier (Ed.), Perspectives on
intermediate, and advanced, to help learn- motivation: Nebraska symposium on motivation
ers set goals and monitor their performance (Vol. 3 8, pp. 69164). Lincoln: University of
more effectively. Clearly, expertise involves Nebraska Press.
more than self-regulatory competence; it Bandura, A. (1997). Self-efficacy: The exercise of
also involves greater task knowledge and per- control. New York: W. H. Freeman.
formance skill. Beneke, W. M., & Harris, M. B. (1972). Teach-
The use of a cyclical phase model of self- ing self-control of study behavior. Behavior
regulation to investigate differences in the Research and Therapy, 10, 3 5 41.
practice methods of experts, non-experts, Bloom, B. (1985 ). Developing talent in young peo-
and novices has been limited to date, but ple. New York: Ballantine Books.
the initial results appear promising. Recall Boekaerts, M., Pintrich, P. R., & Zeidner,
that the terms expert and novice refer to M. (Eds.) (2000). Handbook of self-regulation
high or low positions respectively on this (pp.13 3 9). San Diego, CA: Academic Press.
continuum of task difficulty in this research. Brody, G. H., & Flor, D. L. (1998). Maternal
Multi-phase self-regulation training that is resources, parenting practices, and youth com-
designed to enhance the quality of ones petence in rural, two-parent African American
practice improved not only skill acquisi- families. Developmental Psychology, 3 2 , 696
tion but also key sources of motivation that 706.
underlie continued striving to learn, such Cleary, T. J., & Zimmerman, B. J. (2001). Self-
as perceptions of self-efficacy or confidence regulation differences during athletic practice
and valuing of the intrinsic properties of by experts, non-experts, and novices. Journal of
Applied Sport Psychology, 13 ,185 206.
the task. The importance of such motiva-
tion to the development of expertise was Cleary, T. J., Zimmerman, B. J., & Keating, T.
(in press). Training physical education students
emphasized by Csikszentmihalyi, Rathunde,
to self-regulate during basketball free throw
and Whalen (1993 ) in their study of the practice. Research Quarterly for Exercise and
roots of success and failure with talented Sport.
teenagers: Unless a person wants to pursue Csikszentmihalyi, M., Rathunde, K., & Whalen,
the difficult path that leads to the develop- S. (1993 ). Talented teenagers: The roots of suc-
ment of talent, neither innate potential nor cess and failure. New York: Cambridge Univer-
all the knowledge in the world will suffice sity Press.
(pp. 3 13 2). Clifford, M. (1986). Comparative effects of strat-
egy and effort attributions. British Journal of
Educational Psychology, 5 6, 75 83 .
Ericsson, K. A. (1996). Acquisition of expert
References performance: An introduction to some of the
issues. In K. A. Ericsson (Ed.), The road to
Abrahams, J. (2001, March). And the winner excellence (pp. 15 0) Mahwah, NJ: Lawrence
is . . . you. Golf Magazine, 43 (3 ), 92100. Erlbaum Associates.
72 0 the cambridge handbook of expertise and expert performance
Ericsson, K. A. (2003 ). The acqusition of expert Hardy, J., Gammage, K., & Hall, C. (2001). A
performance as problem solving. Construction descriptive study of athlete self-talk. The Sport
and modification of mediating mechanisms Psychologist, 15 , 3 063 18.
through deliberate practice. In J. E. Davidson Hogan, B. (with Warren, H.). (195 7). Five lessons:
& R. J. Sternberg (Eds.), The nature of problem The modern fundamentals of golf. New York:
solving (23 3 262). New York: Cambridge Uni- Simon & Schuster.
versity Press. Jones, R. T. (1966). Bobby Jones on golf . New York:
Ericsson, K. A., & Kintsch, W. (1995 ). Long term Doubleday.
memory. Psychological Review, 102 , 211245 . Karniol, R., & Ross, M. (1977). The effects
Ericsson, K. A., Krampe, R. Th., & Tesch-Romer of performance-relevant and performance-
C. (1993 ). The role of deliberate practice in the irrelevant rewards on motivation. Child Devel-
acquisition of expert performance. Psychologi- opment, 48, 482487.
cal Review, 100, 3 63 406. Karabenek, S. A. (1998). Strategic help-seeking:
Filby, W. C. D., Maynard, I. W., & Graydon, J. K. Implications for learning and teaching. Mahwah,
(1999). The effect of multiple-goal strategies on NJ: Lawrence Erlbaum Associates.
performance outcomes in training and compe- Kingston, K. M., & Hardy, L. (1997). The effects
tition. Journal of Applied Sport Psychology, 11, of different types of goals on the processes that
23 0246. support performance. The Sport Psychologist, 11,
Fox, L. (1962). Effecting the use of effi- 277293 .
cient study habits. Journal of Mathematics, 1, Kirschenbaum, D. S., OConnor, E. A., & Owens,
7686. D. (1999). Positive illusions in golf: Empiri-
Fox, K., Goudas, M., Biddle, S., Duda, J., & Arm- cal and conceptual analyses. Journal of Applied
strong, N. (1994). Childrens task and ego goal Sport Psychology, 11, 127.
profiles in sport. British Journal of Educational Kitsantas, A., & Zimmerman, B. J. (2002). Com-
Psychology, 64, 25 3 261. paring self-regulatory processes among novice,
Garcia, T., & Pintrich, P. R. (1994). Regulat- non-expert, and expert volleyball players: A
ing motivation and cognition in the classroom: microanalytic study. Journal of Applied Sport
The role of self-schemas and self-regulatory Psychology, 14, 91105 .
strategies. In D. H. Schunk & B. J. Zim- Lehrer, J. (2001, July 3 0). The News Hour with
merman (Eds.), Self regulation of learning and Jim Lehrer. Televised program on the Public
performance: Issues and educational applications Broadcasting System.
(p. 12715 3 ). Hillsdale, NJ: Lawrence Erlbaum Litsky, F. (1999, August 6). Geena Davis zeros in
Associates. with bow and arrows. New York Times, D4.
Glaser, R. (1996). Changing the agency for learn- Locke, E. A., & Latham, G. P. (2002). Building
ing: Acquiring expert Performance. In K. A. a practically useful theory of goal setting and
Ericsson (Ed.), The road to excellence (pp. 15 0) task motivation: A 3 5 -year odyssey. American
Mahwah, NJ: Lawrence Erlbaum Associates. Psychologist, 5 7 , 705 717.
Graham, S., & Golen, S. (1991). Motivational Loehr, J. E. (1991). The mental game. New York:
influences on cognition: Task involvement, ego Plume.
involvement, and depth of information pro-
Mach, E. (1991). Great contemporary pianists
cessing. Journal of Educational Psychology, 83 ,
speak for themselves (Vols. 12). Toronto,
187194.
Canada: Dover.
Grudowski, M. (August, 2003 ). The girl next McCaffrey, N., & Orlick, T. (1989). Mental fac-
door is hungry. Mens Journal, 12 (7) 7273 . tors related to excellence among top profes-
Hallam, S. (1997). Approaches to instrumental sional golfers. International Journal or Sports
music practice of experts and novices: Implica- Psychology, 2 0, 25 6278.
tions for education. In H. Jorgensen & A. C. McPherson, G. E., & Zimmerman, B. J.
Lehmann (Eds.), Does practice make perfect: (2002). Self-regulation of musical learning:
Current theory and research on instrumental A social cognitive perspective. In R. Col-
music practice (pp. 89108). Oslo, Norway: well & C. Richardson (Eds.), The new hand-
Norges musikkhogskole. book of research on music teaching and learning
Hemory, D. (1976). Another hurdle: The making (pp. 3 273 47). New York: Oxford University
of an Olympic champion. New York: Taplinger. Press.
development and adaptation of expertise 72 1
Meichenbaum, D. (1977). Cognitive behavior son (Ed.), The road to excellence (pp. 81106).
modification: An integrative approach. New Mahwah, NJ: Lawrence Erlbaum Associates.
York: Plenum Press. Stein, J. (195 9). William Faulkner. In M. Cowley
Munroe, K. J., Giacobbi, P. R., Hall, C. R., & (Ed.), Writers at work: The Paris review inter-
Weinberg, R. (2000). The four Ws of imagery views (pp. 119141). New York: Viking Press.
use: Where, when, why, and what. The Sport Tejada-Flores, L. (1986). Breakthrough on skis.
Psychologist, 14, 11913 7. New York: Vintage Books.
Murray, D. M. (1990). Learn to write (3 rd ed.). Thellwell, R. C., & Maynard, I. W. (2002). Anxi-
New York: Holt, Rinehart & Winston. ety performance relationships in cricketers:
Newman, R. S. (1994). Academic help seeking: Testing the zone of optimal functioning
A strategy of self-regulated learning. In D. H. hypothesis. Perceptual and Motor Skills, 87 ,
Schunk & B. J. Zimmerman (Eds.), Self- 675 5 89.
regulation of learning and performance: Issues Thelwell, R. C., & Maynard, I. W. (2003 ). The
and educational applications (pp. 283 3 01). effects of a mental skills package on repeatable
Hillsdale, NJ: Lawrence Erlbaum Associates. good performance in cricketers. Psychology of
Nicklaus, J. (with Bowden, K.). (1974). Golf my Sport and Exercise, 4, 3 773 96.
way. New York: Simon & Schuster. Trollope, A. (1905 ). An autobiography. New York:
Nolan, S. (1988). Reasons for studying: Motiva- Dodd Mead.
tional orientations and study strategies. Cogni- Vecsey, G. (1999, September 3 ). Seles feels windy
tion and Instruction, 5 , 269287. blast from past. New York Times, D1.
Paivio, A. (1986). Mental representations: A dual Wallace, I. (1971). The writing of one novel. Rich-
coding approach. New York: Oxford University mond Hill, Canada: Simon & Schuster of
Press. Canada.
Pintrich, P. R. (2000). Multiple goals, multiple Wallace, I., & Pear, J. J. (1977). Self-control tech-
pathways: The role of goal orientation in learn- niques of famous novelists. Journal of Applied
ing and achievement. Journal of Educational Behavior Analysis, 10, 5 15 5 25 .
Psychology, 92 , 5 445 5 5 . Weinstein, C. E., & Mayer, R. E. (1986). The
Schneider, W., & Pressley, M. (1997). Memory teaching of learning strategies. In M. C. Wit-
development between two and twenty (Chap- trock (Ed.), Handbook of research on teaching
ter 7). Mahwah, NJ: Lawrence Erlbaum (pp. 3 15 3 27). New York: Macmillan.
Associates. Williams, L., & Gill, D. L. (1995 ). The role of per-
Schunk, D. H. (1983 a). Goal difficulty and ceived competence in the motivation of phys-
attainment information: Effects of childrens ical activity. Journal of Sport and Exercise Psy-
achievement behaviors. Human Learning, 1, chology, 1, 3 63 3 78.
107117. Winne, P. H. (1997). Experimenting to bootstrap
Schunk, D. H. (1983 b). Progress self-monitoring: self-regulated learning. Journal of Educational
Effects on childrens self-efficacy and achieve- Psychology, 89, 3 97410.
ment. Journal of Experimental Education, 5 1, Zimmerman, B. J. (1985 ). The development of
88105 . intrinsic motivation: A social learning anal-
Schunk, D. H. (1986). Verbalization and chil- ysis. In G. J. Whitehurst (Ed.), Annals of child
drens self-regulated learning. Contemporary development (pp. 117160). Greenwich, CT: JAI
Educational Psychology, 11, 3 473 69. Press.
Schunk, D. H., & Swartz, C. W. (1993 ). Goals and Zimmerman, B. J. (1989). A social cognitive view
progressive feedback: Effects on self-efficacy of self-regulated academic learning. Journal of
and writing achievement. Contemporary Edu- Educational Psychology, 81, 3 293 3 9.
cational Psychology, 18, 3 3 73 5 4. Zimmerman, B. J. (1995 ). Self-efficacy and educa-
Schunk, D. H., & Zimmerman, B. J. (1998). Self- tional development. In A. Bandura (Ed.), Self-
regulated learning: From teaching to self-reflective efficacy in changing societies (pp. 20223 1). New
practice. New York: Guilford. York: Cambridge University Press.
Starkes, J. L., Deakin, J., Allard, F., Hodges, Zimmerman, B. J. (2000). Attaining self-
N. J., & Hayes, A. (1996). Deliberate pratice regulation: A social cognitive perspective. In
in sports: What is it anyway? In K. A. Erics- M. Boekaerts, P. Pintrich, & M. Zeidner (Eds.),
72 2 the cambridge handbook of expertise and expert performance
Handbook of selfregulation (pp. 13 3 9). San Zimmerman, B. J., & Kitsantas, A. (1997). Devel-
Diego, CA: Acadernic. opmental phases in self-regulation: Shifting
Zimmerman, B. J., & Bandura, A. (1994). Impact from process to outcome goals. Journal of Edu-
of self-regulatory influences on attainment in a cational Psychology, 89, 293 6.
writing course. American Educational Research Zimmerman, B. J., & Kitsantas, A. (1999). Acquir-
Journal, 2 9, 845 862. ing writing revision skill: Shifting from pro-
Zimmerman, B. J., Bandura, A., & Martinez- cess to outcome self-regulatory goals. Journal
Pons, M. (1992). Self-motivation for academic of Educational Psychology, 91, 110.
attainment: The role of self-efficacy beliefs Zimmerman, B. J., & Kitsantas, A. (2005 ). Home-
and personal goal setting. American Educational work practices and academic achievement: The
Research Journal, 2 9, 663 676. mediating role of self-efficacy and perceived
Zimmerman, B. J., & Bell, J. A. (1972). Observer responsibility beliefs. Contemporary Educa-
verbalization and abstraction in vicarious rule tional Psychology, 3 0, 3 97417.
learning, generalization, and retention. Devel- Zimmerman, B. J., & Paulsen, A. S. (1995 ).
opmental Psychology, 7 , 22723 1. Self-monitoring during collegiate studying: An
Zimmerman, B. J., & Campillo, M. (2003 ). Moti- invaluable tool for academic self-regulation. In
vating self-regulated problem solvers. In J. E. P. Pintrich (Ed.), New directions in college teach-
Davidson & R. J. Sternberg (Eds.), The nature ing and learning: Understanding self-regulated
of problem solving (pp. 23 3 262). New York: learning (No. 63 , Fall, pp. 13 27). San Francisco,
Cambridge University Press. CA: Jossey-Bass.
Zimmerman, B. J., Greenberg, D., & Weinstein, Zimmerman, B. J., & Schunk, D. H. (Eds.)
C. E. (1994). Self-regulating academic study (2001). Self-regulated learning and academic
time: A strategy approach. In D. H. Schunk achievement: Theoretical perspectives (2nd ed.),
& B. J. Zimmerman (Eds), Self regulation of Mahwah, NJ: Lawrence Erlbaum Associates.
learning and performance: Issues and educa-
tional applications (pp. 181199). Hillsdale, NJ:
Lawrence Erlbaum Associates. Author Note
Zimmerman, B. J., & Kitsantas, A. (1996). Self-
regulated learning of a motoric skill: The role I would like to thank K. Anders Ericsson and Paul
of goal setting and self-monitoring. Journal of Feltovich for their helpful comments regarding
Applied Sport Psychology, 8, 6984. an earlier draft of this chapter.
C H A P T E R 40
72 3
72 4 the cambridge handbook of expertise and expert performance
development. We then detail the concept with all kinds of professional and everyday
of general, age-related slowing and its rela- challenges.
tions to the different theoretical accounts In their attempts to identify such general
for expert performance in later adulthood. dispositions, researchers in psychometric
Our subsequent review of age-comparative intelligence focused on presumably content-
studies in different domains makes a strong free measures of basic cognitive functioning,
case for the dominance of acquired skills like processing speed, abstract reasoning,
and mechanisms in expert performance. In or spatial abilities in figural transformation
the final sections, we focus on deliberate tasks. These basic capacities were assumed
practice as the prime means to improve- to be the building blocks of complex skills.
ment and its role in maintaining expertise Presumably the most influential theory in
as people age. Particularly, we focus on the this context was Cattell and Horns invest-
trade-offs that may be critical to expertise in ment theory (Cattell, 1971; Horn, 1982),
advancing age: between deliberate practice, which posits that primary mental abilities
and its potential to maintain performance, are invested into the development of more
and aging processes that work to degrade complex abilities (see for related proposals,
performance. Krampe & Baltes, 2003 ). From this perspec-
tive, innate dispositions towards certain pri-
mary abilities have the potential to draw
Historical Background some individuals toward specific domains
and, furthermore, provide them with both
Commonsense notions, which typically a head start and a continued performance
attribute high achievement to innate dis- advantage over less gifted individuals. As
positions, had their scientific origin in the an example, above-average abilities in spatial
19th-century writings of Sir Francis Galton visualization might attract certain individu-
(1979). Galton emphasized three precursors als to professions like architecture or graph-
of exceptional achievements, namely, nat- ics design (e.g., Lindenberger et al., 1992;
ural (innate) capacity, zeal, and the power e.g., Salthouse et al., 1990).
to work hard. It was Galtons first asser- Empirical research in the second part of
tion that received prime attention in later the 20th century was partly successful in
theorizing, and two conceptual trends are supporting the first of Galtons premises.
noteworthy in this context. In certain areas, Tests of intellectual abilities proved to be
most notably in music and the arts, the valid correlates of academic achievement
notion of innate talent became increasingly (specifically, high-school grades), job train-
associated with highly specific dispositions ing, and job performance at the point of
that required only little external stimula- entry (Schmidt & Hunter, 1998). Modern
tion to emerge in those rare, gifted indi- behavioral genetics (Plomin, 1990; Plomin
viduals so endowed (Winner, 1982). Simul- & Rende, 1991) established converging esti-
taneously, Galtons legacy of emphasizing mates of about 5 0% heritability1 for general
innate, interindividual differences as pre- intelligence (the g-factor; see also Horn &
requisites of extraordinary accomplishments Masunaga, Chapter 3 4). The program fared
was also echoed in 20th-century conceptions less well when it came to specific abili-
of intelligence. For the pioneers of intel- ties and their relevance to different occu-
ligence research, notably Binet and Stern, pational specializations. Heritability esti-
the concept of intelligence denoted stable, mates observed for specific capacities are
interindividual differences in general abil- lower than those for general intelligence
ities and capacities that were relevant to (g), even if different reliabilities are con-
acquiring new skills and learning in novel sit- trolled for. Particularly, research in areas that
uations. In the minds of the general public were believed to represent prime exem-
to this day, having a high IQ is synonymous plars of specific talents produced disconcert-
with being smart and having a large poten- ing results. For example, professional musi-
tial for successfully coping with learning and cians showed remarkably poor performances
aging and expertise 72 5
on tests of musical talent (Howe et al., ent stages of skill development (Ackerman,
1999; Sloboda, 1991). Coon and Carey (1989) 1988; Fleishman, 1972; Labouvie et al.,
found reliable estimates of heritability in 1973 ; see also, Proctor & Vu, Chapter 15 ).
their study with identical and fraternal twins, Specifically, general intelligence (g) emerged
who performed tasks similar to those in as a factor at early stages of skill acquisition
standardized musicality tests. However, her- (Fitts, 1964), when understanding the
itability estimates were markedly smaller in nature of the task is a critical requirement
those twins who had undergone systematic (Anderson, 1982). Later stages and per-
musical training. formance after practice tends to be less
Besides the domain-specificity problem, correlated with g, but to show substantial
the Coon and Carey findings also point relations to interindividual differences in
to another problem of ability tests that factors closer to the skill under investi-
had enormous impact on current theorizing gation. For example, in his training study
about abilities, age, and expertise, namely, using an air-traffic control task, Ackerman
the role of continued training. Different (1988) found that g, as well as more specific
from their success in predicting achievement ability factors, indeed correlated with pre-
during academic training, psychometric- test levels of performance. Subsequently,
ability test measures show only weak to these correlations were reduced to non-
moderate correlations with performance in significance with the notable exception
practicing professionals like medical doc- of perceptual speed (measured through
tors (Baird, 1985 ). In their meta-analysis the digit-symbol substitution test), which
Hulin, Henry, and Noon (1990) showed gained in strength and remained the only
that this phenomenon generalizes to other reliable correlation with post-training
domains: Correlations between intellectual performance after ten sessions. At a more
abilities and job performance at the time general level, the contribution of specific
of entry were systematically reduced if lev- over general psychometric abilities also
els of performance at longer time intervals appears to depend on individuals overall
after the end of formal training were consid- level of cognitive functioning, with g having
ered. The increasing relevance of domain- its strongest expression in participants scor-
specific knowledge and skills over general ing within lower values of g (Detterman &
abilities has also been demonstrated for older Daniel, 1989).
professionals. Taken together, these findings demon-
For example, although older bank man- strated that there exists a reliable impact
agers show normal age-related decline on of general psychometric intelligence at early
psychometric ability measures, their degree stages of learning a new skill and in indi-
of professional success depended mainly viduals performing at relatively low levels of
on acquired tacit knowledge about the competence. A sizeable portion of interindi-
bank environment (Colonia-Willner, 1998). vidual differences in this capacity can be
Tacit knowledge (Wagner & Sternberg, 1991; traced to genetic (i.e., inherited and innate)
Cianciolo, Mathew, Wagner, & Sternberg, factors. Levels of experience and other age-
Chapter 3 5 ) refers to practical knowledge correlated factors attenuate or decrease the
about business culture and interpersonal effects of those ability factors that may be
relations that enables managers to work relevant at earlier stages of skill acquisition.
effectively. It is measured with a test that The psychometric study of abilities and their
presents scenarios, followed by different relevance for complex skills provided an
solutions that are to be rated. Degree of invaluable starting point for the systematic
concordance of solution ratings with expert investigation of the processing mechanisms
manager ratings is assessed. that underlie expert performance. To under-
Several laboratory training studies stand the mechanisms and preconditions of
demonstrated changes in the correlational expert performance in later adulthood, we
patterns between psychometric-ability fac- need to consider some general age-related
tors and performance change across differ- changes in processing.
72 6 the cambridge handbook of expertise and expert performance
with the evidence for superior perfor- deliberate-practice account is that older
mances in older experts and profession- experts actually acquire specific mechanisms
als (Charness & Bosman, 1990; Krampe & to compensate for age-related deterioration
Baltes, 2003 ; Salthouse, 1991b). The first in critical skills.
account maintains that older experts have Studies that tried to disambiguate
always been superior in skill-relevant abil- between the three accounts for expert
ities, such that their advantages at any performance in later adulthood typically
age could be attributed to interindivid- addressed one or more of the following
ual differences with long-term stability that questions: (1) Do older experts, who excel
already existed prior to expertise acquisi- in their domains, also differ from normal,
tion. Such explanations have been termed age-matched individuals in terms of cog-
preserved differentiation (Salthouse et al., nitive abilities, such as general processing
1990) or a priori disposition accounts speed? (2) Does outstanding performance
(Krampe & Baltes, 2003 ) in the literature. in a particular domain also convey an
The second position assumes that the pro- advantage in near transfer domains that
cess of acquiring expertise involves gradual are subject to age-related decline in the
improvements in those abilities that con- normal population? (3 ) Does the level of
strain normal performance (like working- maintained performance in older experts
memory span) such that expertise should depend on individual investment into
transfer to some (but not necessarily all) critical activities, like deliberate practice?
broader cognitive functions. Finally, the
third account posits that outstanding per-
formance rests on specific mechanisms that Cognitive Abilities, Age,
enable experts to circumvent the process and Expertise: Empirical Evidence
limitations constraining performance in nor-
mal individuals (Chase & Ericsson, 1982). The assumption that interindividual (and
According to the deliberate-practice model presumably innate) differences in basic
(Ericsson & Charness, 1994; Ericsson et al., cognitive-motor functions are natural pre-
1993 ; Ericsson & Lehmann, 1996; Ericsson, cursors to, or contribute to interindivid-
Chapter 3 8) these mechanisms must be ual differences in, expert performance can
acquired through individual efforts directed indeed be found in the literature (Keele
at the long-term adaptation to internal (e.g., & Hawkins, 1982). Given that such basic
age-related changes in cognitive functions) components also overlap with measures of
and external (e.g., task-domain, professional performance-IQ as measured by psycho-
environment) constraints. metric tests, an argument can be made
When applied to aging and exper- that interindividual differences in psycho-
tise, the deliberate-practice account implies metric intelligence or relatively broad abil-
that older experts must actively main- ities are causally linked to the ultimate lev-
tain those specific mechanisms that are els of performance. Findings from several
vital to their domain, and we refer to this studies that compared experts and ama-
set of assumptions as the maintenance- teurs appeared to be in line with related
through-deliberate-practice account (Char- assumptions. As examples, maximum finger
ness et al., 1996; Krampe & Ericsson, 1996). tapping rate is correlated with overall typ-
This position maintains that expertise in ing speed (Book, 1924; Salthouse, 1984) or
later adulthood is not merely the out- level of accomplishment in pianists or typ-
come of achievements during younger ages. ists (Keele et al., 1985 ; Krampe & Ericsson,
Rather, older experts must continuously 1996; Telford & Spangler, 193 5 ). Likewise,
invest deliberate effort into the develop- timing capacity (the variability in control-
ment of their skills, while adapting to the ling successive movements) is more efficient
constraints imposed by aging. One spe- in professional musicians than in amateurs
cific variant of the maintenance-through- and controls (Keele et al., 1985 ; Krampe
72 8 the cambridge handbook of expertise and expert performance
et al., 2002). These results, however, could Findings from laboratory research on
also reflect near transfer as hypothesized older experts generally correspond with the
by the expertise-driven specific abilities dominant finding in occupational psychol-
account. Given general age-related changes ogy, namely, that age and skilled perfor-
in performance-IQ and basic cognitive- mance are poorly correlated. There are,
motor speed, age-comparative studies with however, a number of reasons why the
individuals differing in levels of expertise existing literature may be inadequate to
provide a special route to further disentangle detect age-related declines in work produc-
these issues through systematic compar- tivity. These include weaknesses in exist-
isons of interindividual differences in gen- ing studies, such as restricted age ranges,
eral cognitive abilities, near transfer tasks, restricted job types, and uncertain reliabil-
and expertise-specific functions. ity and validity in the productivity mea-
Age-comparative studies with individ- sures (Salthouse & Maurer, 1996). The
uals differing in their levels of exper- demonstration of expertise moderation for
tise have been conducted in such diverse age-effects also faces severe methodolog-
domains as typewriting (Bosman, 1993 ; ical problems. Field-study designs, using
Salthouse, 1984), chess (Charness, 1981a, regression-analytic techniques on stratified
1981b; Charness et al., 1996), bridge age samples, have considerable power to
(Charness, 1983 , 1989), GO (Masunaga & detect age-related changes that generalize
Horn, 2001), piloting (D. Morrow et al., across larger age ranges. However, their sta-
1994; D. G. Morrow et al., 2001), mas- tistical power to detect age-by-expertise
termind (a game requiring identification interactions is limited and requires huge
of hidden patterns of colored pegs on a sample sizes (typically more than a thou-
pegboard) (Maylor, 1994), crossword-puzzle sand, assuming medium-sized effects). To
solving (Hambrick et al., 1999; Rabbitt, this end (McClelland & Judd, 1993 ),
1993 ), management skills (Colonia-Willner, Lindenberger and Potter (1998) demon-
1998; Walsh & Hershey, 1993 ), and music strated that moderator analyses, using hierar-
(Krampe & Ericsson, 1996; Meinz, 2000) chical regression or structural equation mod-
(see also Chapters in Section V). The general eling techniques, suffer from a confirmation
picture emerging from these studies is that bias towards general factor models (like gen-
older experts show normal (i.e., similar to eral age-related performance declines), and
non-expert controls) age-graded declines in that their power to detect age-differential
general measures of processing speed, cog- changes is limited.
nitive abilities as measured through psycho- Although these considerations seem to
metric tests, and performance on unfamiliar favor the extreme-group approach, related
materials. At the same time, older experts studies are susceptible to the criticism
show reduced, if any, age-related declines in of different selection criteria for young
the efficiencies or the speed at which they and older participants. In typical age-by-
perform skill-related tasks. Thus, the evi- expertise designs there is no overlap in bio-
dence from age-comparative expertise stud- graphical ages between young and older
ies speaks largely for the proposition that groups, respectively, and frequently, accom-
expert performance at any age relies more on plished experts are compared with novices
specific rather than general cognitive mech- or amateurs with limited amounts of for-
anisms (see also Feltovich, Prietula & Eric- mal instruction or professional training.
sson, Chapter 4). Consequently, models of Arguably, older experts in these studies
expertise have departed from the assump- could represent the survivors of an age-
tion that the same set of abilities that under- graded winnowing process by which indi-
lie performance in psychometric intelligence viduals with stronger age-related declines
tests can also account for the ultimate level in relevant capacities, or those who have
of expertise attained or the level of expertise been less motivated to continuously invest
maintained in later adulthood. in the development of their skills, have
aging and expertise 72 9
dropped from their fields of expertise or among the public and is also held among
have been promoted to less-challenging some older experts themselves.
positions. Related selection processes occur Krampe and Ericsson (1996) studied
in societal contexts, which are rarely consid- expert and amateur pianists of different
ered in expertise studies. For example, older ages, with a combination of experimental
employees are more likely to be in stable and psychometric measures of ability, along
positions than younger ones who switch jobs with self-report and diary data recording
more frequently (Swaen et al., 2002), and time investment in deliberate practice and
some societies, for example, Germany, take other activities. The expertise-related abil-
radical measures to promote early retire- ities tested comprised virtuoso skills like
ment for those older individuals who feel maximum repetitive tapping and speeded
overwhelmed by occupational challenges. multi-finger sequencing tasks, but also non-
Employers make extensive use of this mech- speeded tasks such as memorization of
anism to rid themselves of older employees sequences and (rated) expressive musical
suspected of declining productivity. interpretation. In line with results for typ-
Despite some limitations of the pertinent ists and chess experts, the authors found that
studies, we argue that the most likely rea- older professional pianists showed normal
son for the age-graded stability of perfor- age-related declines in measures of general
mance in older experts is that increased age processing speed, such as choice reaction
brings with it increased job-specific knowl- time and speed of digit-symbol substitu-
edge and skills. These assets do not come for tion. However, though age-effects within the
free, however. To support the claim that the amateur group, with regard to expertise-
continuation of deliberate practice through- related measures of multiple-finger coordi-
out ones career is a necessary prerequisite nation speed, were similar to those per-
for expert performance in later adulthood, taining to the general speed measures (e.g.,
it is necessary to establish its compensatory choice reaction time), they were reduced
effects over and above age-related changes or fully absent in the expert sample. Taken
in general processing capacities. We now together these findings led to postulation of
turn to studies that provide evidence along an age-by-expertise dissociation of mecha-
these lines. nisms underlying general processing versus
expertise-specific processing.
Krampe and Ericsson (1996) argued that
Deliberate Practice and Expertise this dissociation reflects older experts selec-
Maintenance in Later Adulthood tive maintenance of acquired, expertise-
specific mechanisms. Expertise-specific
The maintenance-through-deliberate-prac- mechanisms in skilled piano performance
tice account is subject to at least one alter- comprise the sequencing of rapid finger
native explanation, which equally acknowl- movements (Rosenbaum et al., 1983 ),
edges the specificity of skilled mechanisms. bimanual coordination and hand inde-
It is feasible that experts acquire the critical pendence (Krampe et al., 2000), and the
skills in their domain at younger ages and efficient executive control of varying motor
that related mechanisms remain available patterns (Krampe, 2002; Krampe et al.,
throughout later adulthood. Thus, deliber- 2005 ), all of which enable experts to opti-
ate practice may well be the key factor mally prepare their movements in advance
in the acquisition phase, whereas compar- and perform in a fluent, seemingly effortless
atively little individual effort and invest- fashion (MacKay, 1982). The selective
ment is necessary to maintain high levels maintenance interpretation in the Krampe
of performance thereafter. The belief that and Ericsson study rests on data pertaining
prolonged experience and usage of once- to older experts investments in deliberate
acquired knowledge and skills suffices to practice at different stages of their develop-
sustain lifelong expertise is widely believed ment. Consistent with this view, the authors
730 the cambridge handbook of expertise and expert performance
showed that the extent to which levels of mance in older age is the idea that
performance in speeded-expertise tasks was older experts compensate for age-related
maintained in old age depended on the declines in certain capacities through the
amounts of deliberate practice invested at development of compensatory, specific,
the later stages of life, namely, in the 5 th and higher-level mechanisms. This idea moti-
6th decade. At the same time measures of vated the molar-equivalence-molecular-
general processing speed did not account for decomposition approach (Charness, 1981a;
the interindividual differences in levels of Over & Thomas, 1995 ; Salthouse, 1984;
maintained expertise in the expert group. In Westerman et al., 1998). This approach
contrast, such measures correlated with per- entails the study of samples of people in
formance in the amateur group, suggesting which correlations between age and over-
that the basis of expertise is decoupled from all levels of performance (e.g., overall typing
general abilities, particularly if later adult- speed, rated chess performance) are essen-
hood is considered (Krampe & Baltes, 2003 ). tially zero (molar equivalence). Based on the
In a similar vein, Charness and colleagues decomposition of a complex expertise into
(Charness et al., 1996; Charness et al., 2005 ) component processes (molecular decom-
found that chess ratings (based on chess position), investigators then use differen-
tournament performance) in a large sample tial patterns of age-related changes among
of rated players, covering ages from 20 to subprocesses to establish evidence for com-
80 years, depended far more on amounts pensatory mechanisms.
of deliberate practice than on chronologi- The first evidence that pointed to com-
cal age (standardized coefficients in a regres- pensatory mechanisms came from a study
sion equation were .3 8 for age and .62 on age and chess expertise conducted by
for deliberate practice). The effects of delib- Charness (1981a, 1981b). He found that the
erate practice were even more pronounced quality of the chess moves subjects selected
in the older players, as seen in a regres- for an unfamiliar chess position was unre-
sion analysis that showed an interaction of lated to age and closely linked to skill level
age and current deliberate practice in pre- (current chess rating on the ELO-scale).
dicting current skill level. It appeared that Detailed analysis of think-aloud protocols
older players needed more current deliber- revealed that older experts engaged in less
ate practice than younger players to reach extensive search (i.e., they generated fewer
equivalent skill levels, again pointing to the potential moves in a move selection task)
need for continued investment in mainte- than their younger counterparts did, but
nance of skills at advanced ages. However, a they nonetheless came up with moves of
second interaction suggested a trend toward comparable quality. One possible interpre-
diminishing returns from deliberate practice tation of these findings is that older play-
later in life. Increasing amounts of cumu- ers compensate for age-related declines in
lative deliberate practice did not reap the search and retrieval speed with more refined
same gains in skill level for older players. knowledge-based processes related to move
Effects such as those just reported regarding selection.
the effects of maintenance practice have also The molar-equivalence-molecular-deco-
been demonstrated in the domain of sports mposition approach was also applied by
(Ericsson, 1990; Starkes et al., 1996; Hodges, Salthouse (1984, 1991b) in his study with
Starkes, & MacMahon, Chapter 27) typists. He found that across age groups
basic components of movement proficiency,
such as the rate of repetitively typing the
Expert Mechanisms as Compensatory same letter, showed a moderate correlation
Means for Age-Related Decline to overall typing speed, accounting for 42%
of the variance. In contrast, measures reflect-
One of the most fascinating theoreti- ing complex expertise-related mechanisms,
cal perspectives on outstanding perfor- like the speed of typing letters with alternate
aging and expertise 731
hands or the eye-hand span (i.e., the num- slower age-related decline or owing to delib-
ber of letters they looked ahead prior to erate activities to maintain these critical
executing the actual keystrokes), accounted capacities.
for more than 70% of the interindividual The concept of compensation also fea-
differences in overall typing speed. Note that tures prominently in extant frameworks of
repetitive tapping rate, like other measures adaptive aging, like the model of selection,
of general processing speed, showed typical optimization, and compensation (Baltes &
age-related decline in this sample, whereas Baltes, 1990). The SOC model depicts com-
the correlation between age and overall typ- pensation as the acquisition and use of novel
ing speed was essentially zero. Interestingly, or alternative means to counter losses in
older expert typists showed larger eye-hand certain functions (like using a wheelchair
spans compared with their younger coun- or a hearing aid). As an example for com-
terparts. Salthouse argued that the success- pensatory strategies in expert performance,
ful maintenance of typing skills in his older Baltes and Baltes cite from the biographi-
expert typists relied on cognitively com- cal self-report of the famous pianist Arthur
plex mechanisms, namely, extensive antic- Rubinstein. He claimed that his compen-
ipation as illustrated by older skilled typ- satory means for age-related decline in
ists longer eye-hand spans (see also Endsley, movement speed was to slow down prior to
Chapter 3 6). More recently, Bosman (1993 ) difficult passages to create a more impres-
argued that older typists might indeed com- sive contrast. In their application of the SOC
pensate for their age-related declines in basic framework to expertise, Krampe and Baltes
motor speed through their extended eye- (2003 ) describe how the selection of critical
hand spans. activities (practice) over alternative engage-
The studies by Charness and Salthouse ments (e.g., leisure or school) shape develop-
broke new ground in that they suggested mental trajectories. At the level of the indi-
that older experts attain the same level of vidual life course, such selective processes
performance as young experts by means of can be viewed to mitigate or compensate for
different mechanisms. Compensation in the age-related changes, a perspective that illus-
narrow sense implies that older experts rely trates how related interpretations depend on
on mechanisms that are not part of young theoretical context and scope.
experts repertoire. A related, but slightly The reported benefits of sustained main-
different interpretation, assumes that aged tenance practice for older experts and
experts rely differentially on different com- the possibility of developing compensatory
ponent processes, preferably with an empha- strategies or mechanisms appear to be good
sis on those that can be easier maintained news for the successful mastery of every-
at advanced ages. The latter view is closer day life and professional competence in the
to the selective skill maintenance interpre- elderly. Some qualifications of this opti-
tation forwarded by Krampe and Ericssson mistic view are in place. Three constraints
(1996), who argued that older expert pianists of successful skill maintenance must be con-
maintain their levels of performance by sidered to evaluate the patchwork findings
selectively training existing skills. In either in the area. The first constraint relates to the
case, deliberate practice is necessary to nature of deliberate-practice activities, that
detect weaknesses and to develop existing is, whether critical abilities remain intact if
or new skills. As a general point, cross- only exercised in everyday life or the nor-
sectional analyses of expert mechanisms face mal course of professional work and what
the challenge to determine whether older these critical abilities are in the first place.
individuals deliberately adopted compen- A second constraint arises from differential
satory mechanisms in response to aging, or sensitivities of different skill components to
whether their performance at younger ages age-related declines. Finally, there is some
was already superior and associated mech- evidence suggesting that the capacity to
anisms were better preserved, owing to a engage in skill-sustaining deliberate-practice
732 the cambridge handbook of expertise and expert performance
practice in professional contexts and its showed age deficits that were mediated by
limited transfer to other skills or abilities age-related differences in working-memory
(see Feltovich et al., Chapter 4, Exper- measures and speed of processing (Taylor
tise is Limited in Scope and Elite Per- et al., 2005 ). Positive expertise effects on
formance does not Transfer). For exam- flying performance were also observed and
ple, the aforementioned studies by Meinz attributed to deliberate-practice differences.
and colleagues assessed sight-reading per- However, expertise did not interact with
formance in musicians. Sight-reading is a age to reduce age-related differences. These
highly specific skill that has high ecolog- findings point to potential limitations of
ical relevance to certain (e.g., accompa- individual adaptation in less-predictable set-
nists and cembalo players) musicians only, tings or to increasing difficulties to counter
whereas most performers are expected to age-related changes through training.
rely on their memories for intensively pre- In sum, professional experience or staying
pared pieces (Lehmann & Ericsson, 1993 , on the job does not guarantee that the rele-
1996). Similarly, Lindenberger et al. (1992) vant capabilities remain intact in older age.
found that negative age effects in imagery- Rather, the available evidence suggests that
based memory performance in a sample of maintaining skills is as effortful as acquiring
older graphic designers were attenuated, but them in the first place, and benefits become
not eliminated. Arguably, the memory task increasingly more specific, that is, limited to
(the Method-of-Loci) presented a case of those skills that are actively practiced and
medium transfer with respect to the occupa- maintained.
tional expertise of older graphic designers.
Another complication is that the rela-
tive importance of component skills (and Differential Sensitivities of Skills
presumably the degree to which they are to Age-Related Decline
exercised) can differ among experts from a
larger age range. Salthouse and colleagues Anecdotal reports cite the famous piano vir-
(1990) found that the occupational rele- tuoso, Wilhelm Backhaus, as explaining that
vance that participants attributed to the he intensified his etude exercises when he
experimental tasks in their study correlated reached his 5 0s, because at that point he
negatively with age, suggesting that differ- felt that his technical skills required system-
ent types of skilled processes are required atic maintenance practice. Indeed, it seems
for young and older architects. In line plausible that certain aspects of skilled per-
with the latter assumption, expert pianists formance are more affected by age than
diary data in the study by Krampe and other aspects, require more intensive main-
Ericsson (1996) showed a larger amount tenance practice, or both. Unfortunately,
of professional activities in older, compared empirical evidence related to these issues
with young experts, but also a pronounced is sparse.
shift in focus (e.g., less practice and more Whereas the speed of repetitive single-
teaching). finger movements (maximum tapping rate)
Domains like piloting or air-traffic con- tends to be reduced in normal older com-
trol maximally tax experts skills to act pared with young adults (for an overview,
on unpredictable events, or their task- Salthouse, 1985 b), Krampe and Ericsson
sharing abilities. Morrow et al. (2003 ) (1996) observed only modest age-related
observed poorer air-traffic control message- declines in two samples of amateur pianists
recall performance of older pilots com- (and no such declines in the experts). Sim-
pared to younger pilots. However, when the ilarly, older amateur musicians performed
task was changed to a realistic one that as well as their young counterparts when
allowed note taking, the age-related dif- performing simple rhythm tasks at a large
ferences in performance disappeared. Sim- range of tempos (Krampe et al., 2001).
ilarly, simulator-based flying accuracy, in These findings suggest that certain basic
response to air-traffic control instructions, motor components can be maintained with
734 the cambridge handbook of expertise and expert performance
relatively small amounts of practice or that ods. Indeed, older expert pianists reported
they can be sustained by merely exer- that they find it easier to develop musical
cising a real-life skill. In contrast, com- interpretations of new pieces and that their
plex cognitive-motor functions, particularly practice is more efficient compared to when
those that involve bimanual coordination, they were younger (Krampe, 1994).
sequencing operations, or executive control, Somewhat different from this positive
show more pronounced declines in both nor- self-perception in older experts, labora-
mal adults and skilled amateurs (Krampe tory training research suggests that cogni-
et al., 2002; Krampe et al., 2005 ). This sug- tive plasticity (i.e., learning rates as well as
gests that mere experience cannot compen- the ultimate outcomes of what individuals
sate for negative age effects, or in turn, that achieve through practice) decrease in later
increased amounts of deliberate practice are adulthood (Kliegl et al., 1989), particularly
required to this end. after the age of 70 (Singer et al., 2003 ;
Another critical factor determining the Yang et al., 2006). Consistent with labora-
need for maintenance practice is the degree tory research, in part because the studies
to which the skill under consideration relies reviewed contained many instances of lab
on specialized, pragmatic knowledge. Age research, meta-analyses of job-related train-
effects in the expert pianist group in the ing, age, and performance show moderate to
Krampe and Ericsson (1996) study were sig- strong negative correlations (Callahan et al.,
nificantly reduced when participants per- 2003 ; Kubeck et al., 1996). That is, older
formed complex sequences from mem- adults seem to benefit less from training
ory, compared with a condition in which than younger ones (Kubeck et al., 1996), or
pianists sight-read the tasks, suggesting that they require specific types of training (self-
older professionals benefitted from their paced training) to approach the degree of
vast knowledge related to harmonic rela- benefit found for young adults (Callahan et
tions or melodic patterns during encoding. In al., 2003 ). Such studies may not general-
addition, older experts and older amateurs ize well to acquiring and maintaining exper-
showed similar levels of performance in a tise because training methods used were not
musical interpretation task, which involved individualized for the most part, and were
a piece that posed little challenge in terms of certainly not geared to promoting high-level
speed or technical virtuosity. This latter find- performance. Nonetheless, they may speak
ing suggests that those skill components that to increasing difficulty expected for older
relate to specialized knowledge can be main- experts, who have to learn new techniques
tained through experience, at least to some that are possibly unrelated to past ones. As
degree. The prominent role of pragmatic mentioned above, Charness et al. (1996) also
knowledge in the development of expertise found a weak interaction between age and
was also evident in a study of chess players deliberate practice, suggesting a diminishing
conducted by Charness et al. (1996). These return for cumulative deliberate practice for
authors found a significant contribution to older chess players.
skill level from the number of chess books It is necessary to draw some distinctions to
owned by participants, which was inde- understand the disconnect that sometimes
pendent of age and amounts of deliberate occurs between the images of aging seen
practice. in laboratory performance and real world
performance. An important distinction is
that between speeded performance and non-
Age-Related Constraints on speeded performance. For instance, except
Improvement through Practice perhaps in assembly-line work, most regu-
lated jobs do not require people to work
Common belief holds that expertise, but as quickly as possible for long periods of
also age, should lead to growing knowledge time. A related distinction is that of usual
about optimal and efficient practice meth- and maximal performance. It seems unlikely
aging and expertise 735
that many jobs in the economy require max- in the Krampe and Ericsson (1996) study
imal output during working hours. However, revealed increased amounts of time spent
most laboratory research stresses human per- on health and body care or medical consul-
formance to the maximum, by requiring that tation compared with young pianists. Such
people respond as quickly and as accurately findings are typical in age-comparative time-
as possible for a long series of repetitive tri- budgeting studies.
als on novel tasks (see also Proctor & Vu, One potential explanation is that the time
Chapter 15 ). Whereas this makes good sense it takes individuals to recuperate from chal-
from the point of view of testing the lim- lenging practice activities increases with age,
its of human adaptability, it may provide a and thus limits the total amounts of time
distorted view of performance likely to be that can be invested into the maintenance
observed in most ordinary work settings. of expertise. The increasing impact of bodily
Nonetheless, when we examine maximal functions and health condition on learning
performance in real settings, the most usual ability and cognitive functioning in older age
pattern observed is a backward inverted is also evident from two recent lines of inves-
J-shaped function (Simonton, 1996). For tigations. Work by Kramer and colleagues
instance, in a highly competitive environ- (Kramer et al., 1999; Kramer & Willis, 2002)
ment, chess playing, within a very elite demonstrated that some intellectual abili-
sample of Grandmasters, Elo (1965 , 1986) ties, most notably those reflecting execu-
observed that there was a rapid rise in ability tive functions, can be improved by aero-
during the teenage and young adult years and bic exercise. There is also growing evidence
then about a one-standard-deviation decline from dual-task research that in older age
in tournament performance from the peak there exists an increasing demand of bodily
years of the mid-3 0s to age 65 . functions for cognitive resources, which in
Another activity in which outstand- turn can no longer be invested into intellec-
ing performance can be fruitfully exam- tual activities (Krampe & Baltes, 2003 ; Rapp
ined is sports, where the clash between et al., 2005 ; Woollacott & Shumway-Cook,
opponents provides strong feedback about 2002). Although this opens the perspective
superiority and inferiority and where the that optimizing ones physical health can
financial incentives to excel are extremely support continued maintenance practice,
high. Multimillion-dollar contracts are not both lines of research also point to growing
unusual for top professionals in soccer, bas- constraints on further improvement, which
ketball, baseball, and football. In those are inescapable eventually.
domains, it is rare to play at top form beyond
the fourth decade (Schulz & Curnow, 1988;
Schulz et al., 1994). Here too we do not Does Expertise Provide General
usually have easy access to training regimens Benefits at Advanced Ages?
for participants, so it is difficult to know to
what extent reduced motivation to maintain There is by now little disagreement in
intense training or biological/physiological the literature that acquired, domain-specific
factors are responsible for decline. mechanisms support expert performance at
One age-related constraint on practice any age. From this perspective, we would
activities that has, until recently, received expect little transfer or benefits to general
too little attention relates to changes in bod- intellectual abilities. In line with this assum-
ily and health conditions. Deliberate practice ption, Hambrick et al. (1999) found similar
is considered to be among the most effortful factor structures for intellectual abilities
activities by experts (Ericsson et al., 1993 ), in individuals who had spent considerable
and there is tentative evidence that aging time on crossword-puzzle solving, as have
professionals in particular must compromise others for non-specialist adults. However,
between skill development and bodily con- in contrast to these studies, there are some
straints. Older expert pianists diary data cross-sectional results that hint at positive
736 the cambridge handbook of expertise and expert performance
and applied researchers to design interven- Relationship to elders intellectual ability and
tions suitable for developing older adults functional status. Psychology and Aging, 17 ,
potentials and supporting opportunities for 101115 .
their lifelong learning (Charness et al., 2001). Anderson, J. R. (1982). Acquisition of cognitive
The good news emerging from research on skill. Psychological Review, 89, 3 69406.
expertise and learning is that older adults can Baird, L. L. (1985 ). Do grades and tests pre-
maintain high levels of skill through their dict adult accomplishment? Research in Higher
own deliberate efforts, at least up to the third Education, 2 3 , 3 85 .
age (i.e., until age 70). To this end, societies Baltes, P. B., & Baltes, M. M. (1990). Psychological
protective mechanisms, that typically guar- perspectives on successful aging: The model of
antee that older employees are more likely selective optimization with compensation. In
P. B. Baltes & M. M. Baltes (Eds.), Successful
to be in stable positions than younger ones
aging: Perspectives from the behavioral sciences
(Swaen et al., 2002), provide a context for (pp. 13 4). Cambridge, NY: Cambridge Uni-
these individuals to selectively maintain rel- versity Press.
evant skills. However, to the extent that job Book, W. F. (1924). Voluntary motor ability of
demands change over time, obsolescence of the worlds champion typists. Journal of Applied
skills becomes a risk, (Sparrow & Davies, Psychology, 8, 283 3 08.
1988) unless, we would argue, people con- Bosman, E. A. (1993 ). Age-related differences
tinue to engage in deliberate practice. There in the motoric aspects of transcription typing
are, however, limits imposed on the contin- skill. Psychology and Aging, 8, 87102.
ued investment of resources into skill devel- Callahan, J. S., Kiker, D. S., & Cross, T. (2003 ).
opment that emerge at even more advanced Does method matter? A meta-analysis of the
ages (the fourth age), that ultimately con- effects of training method on older learner
strain an individuals participation in the training performance. Journal of Management,
long-distance race to achieve and maintain 2 9, 663 680.
high-level expertise. Cattell, R. B. (1971). Abilities: Their structure,
growth, and action. Boston, MA: Houghton
Mifflin.
Footnote Cerella, J. (1985 ). Information processing rates in
the elderly. Psychological Bulletin, 98, 6783 .
Cerella, J. (1990). Aging and information pro-
1. Heritability estimates in behavioral genetics
cessing rates in the elderly. In J. E. Birren &
are based on correlations of criterion measures
K. W. Schaie (Eds.), Handbook of the psychol-
(e.g., IQ-test performances) between samples
ogy of aging (3 rd ed., pp. 201221). San Diego,
that are genetically related. Specifically, heri-
CA: Academic Press.
tability denotes the proportion of interindivid-
ual differences (the variance of the criterion Cerella, J., & Fozard, J. L. (1984). Lexical access
measure) that can be accounted for by kinship. and age. Developmental Psychology, 2 0, 23 5
243 .
Charness, N. (1981a). Aging and skilled prob-
lem solving. Journal of Experimental Psychology:
References General, 110, 213 8.
Charness, N. (1981b). Search in chess: Age
Ackerman, P. L. (1988). Determinants of individ- and skill differences. Journal of Experimental
ual differences during skill acquisition: Cogni- Psychology: Human perception and Performance,
tive abilities and information processing. Jour- 7 , 467476.
nal of Experimental Psychology: General, 117 , Charness, N. (1983 ). Age, skill, and bridge
2883 18. bidding: A chronometric analysis. Journal of
Allaire, J. C., & Marsiske, M. (1999). Everyday Verbal Learning and Verbal Behavior, 2 2 , 406
cognition: Age and intellectual ability corre- 416.
lates. Psychology and Aging, 14, 627644. Charness, N. (1989). Expertise in chess and
Allaire, J. C., & Marsiske, M. (2002). Well- bridge. In D. Klahr & K. Kotovsky (Eds.),
and ill-defined measures of everyday cognition: Complex information processing: The impact of
738 the cambridge handbook of expertise and expert performance
Herbert A. Simon (pp. 183 208). Hillsdale, NJ: Elo, A. E. (1986). The rating of chessplayers, past
Erlbaum. and present (2nd ed.). New York: Arco.
Charness, N., & Bosman, E. A. (1990). Expertise Ericsson, K. A. (1990). Peak performance and age:
and aging: Life in the lab. In T. M. Hess (Ed.), An examination of peak performance in sports.
Aging and cognition: Knowledge organization and In P. B. Baltes & M. M. Baltes (Eds.), Successful
utilization (pp. 3 43 3 85 ). North-Holland: Else- aging: Perspectives from the behavioral sciences
vier Science Publishers B.V. (pp. 164196). New York: Cambridge Univer-
Charness, N., Krampe, R. T., & Mayr, U. sity Press.
(1996). The role of practice and coaching in Ericsson, K. A., & Charness, N. (1994). Expert
entrepreneurial skill domains: An international performance: Its structure and acquisition.
comparison of life-span chess skill acquisition. American Psychologist, 49, 725 747.
In K. A. Ericsson (Ed.), The road to excellence: Ericsson, K. A., & Kintsch, W. (1995 ). Long-
The acquisition of expert performance in the arts term working memory. Psychological Review,
and sciences, sports, and games (pp. 5 180). 102 , 211245 .
Mahwah, NJ: Erlbaum. Ericsson, K. A., Krampe, R. T., & Tesch-Romer,
Charness, N., Park, D. C., & Sabel, B. A. (2001). C. (1993 ). The role of deliberate practice in the
Communication, technology and aging: Oppor- acquisition of expert performance. Psychologi-
tunities and challenges for the future. New York: cal Review, 100, 3 63 406.
Springer. Ericsson, K. A., & Lehmann, A. C. (1996).
Charness, N., Tuffiash, M., Krampe, R., Reingold, Expert and exceptional performance: Evidence
E. M., & Vasyukova, E. (2005 ). The role of on maximal adaptations on task constraints.
deliberate practice in chess expertise. Applied Annual Review of Psychology, 47 , 273 3 05 .
Cognitive Psychology, 19, 15 1165 . Fitts, P. (1964). Perceptual-motor skill learning. In
Chase, W. G., & Ericsson, K. A. (1982). Skill A. W. Melton (Ed.), Categories of human learn-
and working memory. In G. H. Bower (Ed.), ing (pp. 243 285 ). San Diego, CA: Academic
The psychology of learning and motivation Press.
(Vol. 16, pp. 15 8). San Diego, CA: Academic Fleishman, E. A. (1972). On the relation between
Press. abilities, learning, and human performance.
Clarkson-Smith, L., & Hartley, A. A. (1990). The American Psychologist, 2 7 , 1017103 2.
game of bridge as an exercise in working mem- Galton, F., Sir. (1979). Hereditary genius: An
ory and reasoning. Journal of Gerontology, 45 , inquiry into its laws and consequences. London:
P23 3 P23 8. Julian Friedman Publishers. (Originally pub-
Colonia-Willner, R. (1998). Practical intelligence lished in 1869).
at work: Relationship between aging and cogni- Hambrick, D. Z., & Engle, R. W. (2002). Effects of
tive efficiency among managers in a bank envi- domain knowledge, working memory capacity
ronment. Psychology and Aging, 13 , 45 5 7. and age on cognitive performance: An investi-
Coon, H., & Carey, G. (1989). Genetic and envi- gation of the knowledge-is-power hypothesis.
ronmental determinants of musical ability in Cognitive Psychology, 44, 3 3 93 87.
twins. Behavioral Genetics, 19, 183 193 . Hambrick, D. Z., Salthouse, T. A., & Meinz, E. J.
Detterman, D. K., & Daniel, M. H. (1989). Cor- (1999). Predictors of crossword puzzle profi-
relations of mental tests with each other and ciency and moderators of age-cognition rela-
with cognitive variables are highest for low IQ tions. Journal of Experimental Psychology: Gen-
groups. Intelligence, 13 , 3 493 5 9. eral, 12 8, 13 1164.
Dixon, R. A., & Baltes, P. B. (1986). Toward life- Hasher, L., Stoltzfus, E. R., Zacks, R. T., &
span research on the functions and pragmat- Rympa, B. (1991). Age and inhibition. Journal
ics of intelligence. In R. J. Sternberg & R. K. of Experimental Psychology: Learning, Memory,
Wagner (Eds.), Practical intelligence: Nature & Cognition, 17 , 163 169.
and origins of competence in the everyday world Hertzog, C. (1989). Influences of cognitive slow-
(pp. 203 23 5 ). New York: Cambridge Univer- ing on age differences in intelligence. Develop-
sity Press. mental Psychology, 2 5 , 63 665 1.
Elo, A. E. (1965 ). Age changes in master chess Hoffman, R. R., Shadbolt, N. R., Burton, A.
performances. Journal of Gerontology, 2 0, 289 M., & Klein, G. (1995 ). Eliciting knowl-
299. edge from experts: A methodological analysis.
aging and expertise 739
acquisition of expert sight-reading and accom- influence of expertise and task factors on age
panying performance. Psychomusicology, 15 (1 differences in pilot communication. Psychology
2), 129. and Aging, 16, 3 146.
Lima, S. D., Hale, S., & Myerson, J. (1991). How Morrow, D. G., Ridolfo, H. E., Menard, W. E.,
general is general slowing? Evidence from the Sanborn, A., Stine-Morrow, E. A. L., Magnor,
lexical domain. Psychology and Aging, 6, 416 C., et al. (2003 ). Environmental support pro-
425 . motes expertise-based mitigation of age differ-
Lindenberger, U., Kliegl, R., & Baltes, P. B. (1992). ences on pilot communication tasks. Psychology
Professional expertise does not eliminate nega- and Aging, 18, 268284.
tive age differences in imagery-based memory Myerson, J., Hale, S., Wagstaff, D., Poon, L. W.,
performance during adulthood. Psychology and & Smith, G. A. (1990). The information-loss
Aging, 7 , 5 85 5 93 . model: A mathematical theory of age-related
Lindenberger, U., & Potter, U. (1998). The com- cognitive slowing. Psychological Review, 97 ,
plex nature of unique and shared effects in 475 487.
hierarchical linear regression: Implications for Over, R., & Thomas, P. (1995 ). Age and skilled
developmental psychology. Psychological Meth- psychomotor performance: A comparison of
ods, 3 , 21823 0. younger and older golfers. International Jour-
MacKay, D. G. (1982). The problems of flexi- nal of Aging & Human Development, 41(1),
bility, fluency, and speed-accuracy trade-off in 112.
skilled behavior. Psychological Review, 89(5 ), Plomin, R. (1990). The role of inheritance of
483 5 06. behavior. Science, 2 48, 183 188.
Masunaga, H., & Horn, J. (2001). Expertise in Plomin, R., & Rende, R. (1991). Human behav-
relation to aging changes in components of ioral genetics. Annual Review of Psychology, 42 ,
intelligence. Psychology & Aging, 16, 293 3 11. 161190.
Maylor, E. A. (1994). Ageing and the retrieval Rabbitt, P. M. A. (1993 ). Crystal quest. A search
of specialized and general knowledge: Perfor- for the basis of maintenance of practised skills
mance of masterminds. British Journal of Psy- into old age. In A. Baddeley & L. Weiskrantz
chology, 85 , 105 114. (Eds.), Attention: Selection, awareness, and con-
McClelland, G. H., & Judd, C. M. (1993 ). Statis- trol. Oxford, UK: Clarendon Press.
tical difficulties of detecting interactions and Rapp, M., Krampe, R. T., & Baltes, P. B. (2005 ).
moderator effects. Psychological Bulletin, 114, Adaptive task prioritization in aging: Selective
3 763 90. ressource allocation to postural control is pre-
McEvoy, G. M., & Cascio, W. F. (1989). Cumu- served in Alzheimer disease. American Journal
lative evidence of the relationship between of Geriaric Psychiatry 14(1), 5 261.
employee age and job performance. Journal of Rosenbaum, D. A., Kenny, S., & Derr, M.
Applied Psychology, 74, 1117. (1983 ). Hierarchical control of rapid move-
Meinz, E. J. (2000). Experience-based attenua- ment sequences. Journal of Experimental
tion of age-related differences in music cogni- Psychology: Human Perception and Performance,
tion tasks. Psychology and Aging, 15 , 2973 12. 9, 86102.
Meinz, E. J., & Salthouse, T. A. (1998). The effects Salthouse, T. A. (1984). Effects of age and skill
of age and experience on memory for visually in typing. Journal of Experimental Psychology:
presented music. Journals of Gerontology: Psy- General, 113 , 3 45 3 71.
chological Sciences, 5 3 B, P60P69. Salthouse, T. A. (1985 a). Speed of behavior and
Mireles, D. E., & Charness, N. (2002). Com- its implications for cognition. In J. E. Birren &
putational explorations of the influence of K. W. Schaie (Eds.), Handbook of the psychology
structured knowledge on age-related cognitive of aging (pp. 400426). New York: Van Nos-
decline. Psychology & Aging, 17 , 245 25 9. trand Reinhold.
Morrow, D., Leirer, V., Altiteri, P., & Fitzsim- Salthouse, T. A. (1985 b). A theory of cognitive
mons, C. (1994). When expertise reduces age aging. Amsterdam: North Holland Press.
differences in performance. Psychology and Salthouse, T. A. (1991a). Age and experience
Aging, 9, 13 4148. effects on the interpretation of orthographic
Morrow, D. G., Menard, W. E., Stine-Morrow, drawings of three-dimensional objects. Psychol-
E. A. L., Teller, T., & Bryant, D. (2001). The ogy and Aging, 6, 42643 3 .
aging and expertise 741
Salthouse, T. A. (1991b). Expertise as the cir- Ericsson (Ed.), The road to excellence (pp. 227
cumvention of human processing limitations. 25 3 ). Mahwah, NJ: Erlbaum.
In K. A. Ericsson & J. Smith (Eds.), Toward a Singer, T., Lindenberger, U., & Baltes, P. B.
general theory of expertise: Prospects and limits (2003 ). Plasticity of memory for new learning
(pp. 2863 00). New York: Cambridge Univer- in very old age: A story of major loss? Psychol-
sity Press. ogy and Aging, 18, 3 063 17.
Salthouse, T. A. (1991c). Mediation of adult age Sloboda, J. A. (1991). Musical expertise. In K. A.
differences in cognition by reductions in work- Ericsson & J. Smith (Eds.), Toward a general
ing memory and speed of processing. Psycho- theory of expertise. Prospects and limits (pp. 15 3
logical Science, 2 , 179183 . 171). Cambridge, MA: Cambridge University
Salthouse, T. A. (1996). The processing-speed Press.
theory of adult age differences in cognition. Sparrow, P. R., & Davies, D. R. (1988). Effects
Psychological Review, 103 , 403 428. of age, tenure, training, and job complexity on
Salthouse, T. A., Babcock, R. L., Skovronek, technical performance. Psychology and Aging,
E., Mitchell, D. R. D., & Palmon, R. 3 , 3 073 14.
(1990). Age and experience effects in spa- Starkes, J., Deakin, J., Allard, F., Hodges, N. J., &
tial visualization. Developmental Psychology, 2 6, Hayes, A. (1996). Deliberate practice: What is
12813 6. it anyway? In K. A. Ericsson (Ed.), The road to
Salthouse, T. A., & Maurer, T. J. (1996). Aging, excellence: The acquisition of expert performance
job performance, and career development. In in the arts and sciences, sports, and games (pp. 81
J. E. Birren & K. W. Schaie (Eds.), Handbook of 106). Mahwah, NJ: Erlbaum.
the psychology of aging (4th ed., pp. 3 5 3 3 64). Sternberg, R. J. (1999). Intelligence as develop-
New York: Academic Press. ing expertise. Contemporary Educational Psy-
Schaie, K. W. (1989). Perceptual speed in chology, 2 4, 3 5 93 75 .
adulthood: Cross-sectional and longitudinal Sternberg, R. J., & Wagner, R. K. (1986). Practi-
studies. Psychology and Aging, 4, 443 45 3 . cal intelligence: Nature and origins of competence
Schmidt, F. L., & Hunter, J. E. (1998). The valid- in the everyday world. Cambridge, MA: Cam-
ity and utility of selection methods in personnel bridge University Press.
psychology: Practival and theoretical implica- Swaen, G. M. H., Kant, I., van Amelsvoort,
tions of 85 years of research findings. Psycho- L. G. P. M., & Beurskens, A. J. H. M. (2002).
logical Bulletin, 12 4, 262274. Job mobility, its determinants, and its effects:
Schooler, C., & Mulatu, M. S. (2001). The recip- Longitudinal data from the maastricht cohort
rocal effects of leisure time activities and intel- study. Journal of Occupational Health Psychol-
lectual functioning in older people: A longitu- ogy, 7 , 121129.
dinal analysis. Psychology and Aging, 16, 466 Taylor, J. L., OHara, R., Mumenthaler, M. S.,
482. Rosen, A. C., & Yesavage, J. A. (2005 ). Cogni-
Schooler, C., Mulatu, M. S., & Oates, G. (1999). tive ability, expertise, and age differences in fol-
The continuing effects of substantively com- lowing air-traffic control instructions. Psychol-
plex work on the intellectual functioning of ogy and Aging, 2 0, 11713 3 .
older workers. Psychology and Aging, 14, 483 Telford, C. W., & Spangler, H. (193 5 ). Training
5 06. effects in motor skills. Journal of Experimental
Schulz, R., & Curnow, C. (1988). Peak perfor- Psychology, 18, 141147.
mance and age among superathletes: Track and Wagner, R. K., & Sternberg, R. J. (1991). Tacit
field, swimming, baseball, tennis and golf. Jour- knowledge: Its uses in identifying, assessing,
nal of Gerontology: Psychological Sciences, 43 , and developing managerial talent. In J. Jones,
113 120. B. Steffy, & D. Bray (Eds.), Applying psychology
Schulz, R., Musa, D., Staszewski, J., & Siegler, in business: The managers handbook (pp. 3 3 3
R. S. (1994). The relationship between age and 3 44). New York: Human Sciences Press.
major league baseball performance: Implica- Waldman, D. A., & Avolio, B. J. (1986). A meta-
tions for development. Psychology and Aging, analysis of age-differences in job-performance.
9, 274286. Journal of Applied Psychology, 71, 3 3 3 8.
Simonton, D. K. (1996). Creative expertise: A Walsh, D. A., & Hershey, D. A. (1993 ).
life-span developmental perspective. In K. A. Mental models and the maintenance of
742 the cambridge handbook of expertise and expert performance
complex problem-solving skills in old age. In J. and rehabilitation (pp. 183 197). Boston, MA:
Cerella, J. Rybash, W. Hoyer, & M. Commons Kluwer Academic Publishers.
(Eds.), Adult information processing: Limits on Winner, E. (1982). Invented worlds: The psychology
loss (pp. 5 5 3 5 84). San Diego, CA: Academic of the arts. Cambridge, MA: Harvard University
Press. Press.
Westerman, S. J., Davies, D. R., Glendon, A. I., & Woollacott, M., & Shumway-Cook, A. (2002).
Stammers, R. B. (1998). Ageing and word pro- Attention and the control of posture and gait:
cessing competence: Compensation or compi- A review of an emerging area of research. Gait
lation. British Journal of Psychology, 89, 5 79 and Posture, 16, 114.
5 97. Yang, L., Krampe, R. T., & Baltes, P. B. (2006).
Willis, S. L., & Marsiske, M. (1991). Lifespan per- Basic forms of cognitive plasticity extended
spective on practical intelligence. In T. E. Tup- into the oldest-old: Retest learning, age, and
per & K. D. Cicerone (Eds.), The neuropsy- cognitive functioning. Psychology and Aging, in
chology of everyday life: Issues in development press.
C H A P T E R 41
Harald A. Mieg
We have serious difficulties when it comes to well as the effect of music on its audience.
explaining what really defines an expert In short: I understand function as defined
a difficulty that goes beyond the explanatory by what an audience, patient, or customer
range of defining experts by their individual would pay for.1 The function of medical ther-
performance. Take, for example, people who apy is to render a sick person healthy. The
provide political advice or consult multina- function of music is to please the audience
tionals. What would qualify them as experts? (as entertainment) or peer professionals (as
How can we assess their performance? How being excellent). My definition of function
can we disentangle their individual expert as what would be paid for says that there
contribution and the success of the enter- is a potential interest in a particular expert
prise or party they work for? We cannot performance, by the patient, the audience,
understand these cases if we dont consider or other sorts of clients.
what Hoffman, Feltovich, and Ford (1997) In this chapter, I will first introduce an
concluded: the minimum unit of analysis expert role approach that is mainly based on
is the expert-in-context (p. 5 5 3 )(see also attribution theory and will provide an under-
Clancey, Chapter 8). standing of the social functions of experts.
For the purpose of this chapter on social The key to understanding expert roles is to
and sociological factors in the development take into account the layperson or client. In
of expertise, I assume that an expert has other words, to look at expert roles as forms
to be regarded as the connection between of interaction between the expert and his or
a person and a function. The function indi- her client or an audience. In society, expert
cates the social context of the expert perfor- means that you are regarded or addressed as
mance. In the following, I use a broad notion such by someone else. This social conception
of function that includes both the pertinent of expert differs from other ones discussed in
duties and the effects of expert performance, this handbook, such as the expert as an out-
such as the duties and work of a doctor, as standing individual nominated by peers (see
743
744 the cambridge handbook of expertise and expert performance
Chi, Chapter 2) and the expert defined by starting point is the observation that there
his/her superior performance (See Ericsson, are relative experts. In other words: the depth
Chapters 13 , 3 8). of knowledge and skill necessary to provide
In the second part, the expert role an explanation depends on what is required
approach will help us understand the work in a particular context. If I visit a town I have
of experts in various social contexts orga- never been to before and want to know the
nizations, professions, society. We will see direction to the station, I might ask a person
that professions play a decisive role in setting on the street who looks or behaves like a pos-
and controlling quality standards of expert sible resident of that town. In this case I sup-
performance (see also Evetts, Mieg, Felt, pose that a resident knows his or her town
Chapter 7). through personal experience and is able to
The third part of this chapter will be provide me with reasonable instructions on
devoted to mediating mechanisms in the how to get to the station.
development of expertise, such as socializa- There are many open questions regarding
tion. Particularly, we will take a look at the the role of relative experts and beyond. In
assertion that the best context for the social- this first part, I will examine The expert-
ization of experts is the bourgeois middle- interaction by taking five steps to answer
class home. such questions as:
On Relative Experts, Other Experts, The last example in the list is the cor-
and Expert-Performance Criteria porate communications employee. In gen-
eral, this person does not really know the
Almost anyone can under certain circum- plans, strategies, and decision constraints
stances act as an expert. This is based on of the companys board members, nor the
the fact that the level of knowledge and skill industrial-technological processes or scien-
differs in our society, as well as the level tific background a companys production is
of knowledge and skill necessary to serve a based on. However, this persons function is
function in a context. If you are a student of to present the company in the public and
psychology and in trouble with your land- to answer questions by guests or journalists.
lord, you may turn to a friend who studies More specifically, the function of the cor-
law. In this situation your friend acts as an porate communication employee is to act
expert on law a status he or she would as a gatekeeper who prevents the working
never have in the law community. force behind the scene from being involved
In Table 41.1 we see on the left side some in public queries. This person definitely is
examples of experts that are classic to the a relative expert as to everything going on
study of expertise, particularly chess masters in the company and might be an expert
and medical doctors. On the right side we in an absolute sense regarding his or own
see examples of more or less relative experts, job in corporate communications. The mid-
such as the resident or the law student. I have dle column of Table 41.1 contains a list of
also included the example of a Star Wars experts more or less seldomly cited in lit-
film expert whom we might see in a TV erature on expertise, such as entrepreneurs,
game show. Hoffman, Shadbolt, Burton, and master chefs, aborigines, astrologers, or com-
Klein (1995 ) reported that there are studies puter freaks who are expert hackers. Each of
on expertise where psychologists relied on them can act or be addressed as an expert in
the participation of preschool children who certain contexts. I have arranged this list in
were avid fans of Star Wars films (p. 13 1). order to show the importance (or problem)
Other examples include the politician who of expert-performance criteria. In the case of
serves as a minister of foreign affairs and usu- entrepreneurs, a broad and open set of crite-
ally is not a professional diplomat, or as a ria has to be applied (innovativeness, finan-
minister of transport without having mas- cial success, seize of the company, public
tered any university studies on that topic. impact etc.). Astrology lies outside the scope
However, in his or her political party, the of todays accepted sciences, and hacking
politician has become the expert on foreign outside accepted social practices. However,
affairs or transportation by virtue of working there are astrologers, as well as occultists and
on that topic or even by denomination. pendulum specialists, who have been used
746 the cambridge handbook of expertise and expert performance
in productivity is crucial and not limited to I argue that the simplicity and usefulness
engineering the technical basis of industries. of The expert-interaction is based on a sim-
What is the advantage from understand- ple fact: the criteria for finding the expert
ing expertise as human capital? Isnt that result from an extension or generalization of
simply another way of speaking of the divi- ones own experience. By addressing some-
sion of labor? The notion of the division of one as an expert we need to suppose only
labor could invoke ideas of a preestablished that this person has obtained knowledge we
categorization of possible occupations. From could obtain ourselves, supposing we had
this point of view, experts are specialists for the time to do so. This is quite obvious in
specific problems. In this case, it would be the case of asking someone on the street
best to educate everyone at the place he for directions to station. We need to assume
or she will go to work for the rest of their only that a normal resident (we address as
lives. However, this would underestimate expert) has had enough time to develop a
the dynamics of productive knowledge in sufficient picture of the geography of his or
our societies. her town. After being informed about the
Understanding expertise as human capital directions to the station, we can, ourselves,
implies the following: act as experts and instruct others (e.g., our
r Expertise is personalized: Expertise is children) how to get to the station.
Everyone knows a story about how The
embodied in persons. This has the advan- expert-interaction can fail, here is one
tage that we can exchange types of knowl- more: A scientist came to Berlin for the first
edge by exchanging experts. It is much time; he wanted to attend a conference at
easier to exchange a doctor than to the Technical University. In the morning,
change the medical system. his colleague from Berlin picked him up at
r Expertise is priced: To select or exchange
the hotel and walked with him to the Uni-
experts, their expertise has to be val- versity. This was a nice but long walk of
ued. The value of human expertise is more than half an hour passing through a
expressed (measured) by prices paid in spacious urban park (Tiergarten). The next
labor markets or prizes won in profes- day, the visiting scientist used exactly this
sional competitions. way to get from the hotel to the University.
After two or more days, he was accompa-
To fully understand the social function of
nied by other scientists staying at this hotel
experts as human capital (in the broad
and attending the same conference. Some-
sense of function introduced at the outset
times they really had to hurry through the
of this chapter), let us take a short look
park because they left the breakfast table too
at some basics of the psychology of exper-
late. Another day, the colleague from Berlin
tise (see also Feltovich, Prietula, & Ericsson,
picked him up at the hotel again, but this
Chapter 4):
time he went a much shorter way, not cross-
r Expertise seems to be a form of cogni- ing the park. Now the visiting scientist real-
tive/behavioral adaptation to a particu- ized that the first time his colleague from
lar domain of tasks, hence it is domain- Berlin wanted to talk to him and had there-
specific. fore chosen this long deviation through the
r To become an expert requires mas- park. He himself had assumed that this col-
sive domain-specific training and prac- league from Berlin (as a relative expert on
tice deliberate practice (see Ericsson, this area) showed him (the layperson) a rea-
Chapter 3 8). sonable way from the hotel to the Technical
University.
From this point of view, expertise is based on We can say: The core of the experts role con-
focused experience and training. The social sists of providing experience-based knowledge
function of using experts is the time-efficient that we could attain ourselves if we had enough
use of knowledge, based on the experts rou- time to undertake the necessary learning. In
tine through experience. other words: the particular gain from using
social and sociological factors in the development of expertise 749
an expert is the relatively fast utilization of be in a hurry and unwilling to stop; the
the experts compressed experience any rea- doctors main purpose might be to keep
sonable person could make if she or he had his practice running; the appointed experts
enough time to do so (cf. Mieg, 2001). The dominant motivation might be not to say
important remaining question is: How do we anything wrong; the mother instructing her
identify experts of subjects where we lack child might not want to have to lace her
any experience? Why do we trust experts daughters shoes herself; and the scientists
even (or particularly!) in fields where most may want to present themselves in the dis-
of us are complete laypersons, for instance cussion as favorably as possible. Neverthe-
in the diagnosis of brain tumors, asbestos, or less, all of them provide explanations for
international trade regulations? someone else.
To summarize: Expertise can be regarded The expert is a social form, in the same
as human capital. The question of how such way that division of labor and hierarchy
capital is valued led us to recognize the time are social forms. Social forms can be charac-
investment in the development of expertise. terized as
From this we can derive the core social func-
tion of the experts role: It consists of the r extreme generalizations of interactions
relatively fast utilization of the experts com- we find in many societies
pressed experience that any reasonable per- r being independent from the kind of use
son could attain if she or he had enough time
or the motives connected to this use.
to do so.
There is a specific value attached to The assignment of expert roles seems to serve
expert-interactions, truth. In this context as a source of social validation, that is, the
truth means: A sentence such as This or veracity of the information introduced by
that is the shortest way to the station is one group member is confirmed by another
true if (and only if ) this or that is actu- (p. 627).
ally the shortest way to the station. Truth The studies by Simmel and Stasser are
is an option that can potentially be realized backed by the attribution theory. Attribution
in every interaction with experts who have theory reveals much of what is said as a result
sufficient knowledge to share it. When we of the attribution that is inherent to The
say that the value truth is attached to the expert-interaction. It was founded by the
social form The expert, this does not mean psychologist Fritz Heider. His starting point
that every client or person asking for infor- was the question: How does a person inter-
mation expects the expert to tell the truth. pret the actions of another person? Thus he
But, even if someone who is generally sus- started by investigating commonsense psy-
picious of scientific knowledge be it mod- chology. Heider wrote:
ern medicine or political sciences asks a
In everyday life we form ideas about other
scientist about the current trends or infor-
people and about social situations. We
mation on the state of the art, this person
interpret the actions of other people and
would nevertheless expect most scientists we predict what they will do under cer-
to truthfully explain the state of the art in tain circumstances. Though these ideas are
that particular science to the best of their usually not formulated, they often function
knowledge unless they had financial incen- adequately. (195 8, p. 5 )
tives to bias their assessments. The origin of
this truth presupposition in experts, I would Attribution theory basically distinguishes
argue, lies in the extension or generalization two main, dichotomous sources of
of ones own experience: True is what I attributed causality: the person or the
could know myself if I had enough time to situation. A personal attribution is an
undertake the necessary experience (as the internal attribution, a situational attribution
expert did). is an external one. Persons, as well as
The power of the social form The expert situations, have invariant (dispositional) or
can be seen clearly in unstructured con- variable properties (see Weiner, 1986): A
texts, such as in a group of people who are personality trait would be a dispositional
unfamiliar with one another and meet only personal property; pure luck would be
once. This is the situation we encounter in a variable situational factor. Usually, we
experimental groups. Garold Stasser and col- regard expertise as based on experience
leagues have studied the effects of expert and training, thus expertise is a personal
role assignment in groups. A common exper- dispositional characteristic.
imental design is the hidden profile: Some Two implications of the attribution the-
group members have unshared information ory are of particular importance for The
that is necessary to complete the group task expert-interaction:
(see, e.g., Stasser, 1992). In these studies, r There is a tendency to overestimate indi-
unshared information is a basis for the defi- vidual expertise and neglect the context
nition of expertise, expertise signifying that owing to the so-called fundamental attri-
a person has access to more information in bution error (Ross, 1977).
a specific domain than others in the group r This personalized attribution of expertise
(Stewart & Stasser, 1995 , p. 619). It can be
reduces the perceived uncertainty, imply-
shown that the explicit assignment of par-
ing certainty (truth) as well as, to some
ticular expert roles to the group members
extent, trust.
who have unshared information increases
the chance that this piece of information The fundamental attribution error con-
will contribute to the groups work. The sists in the tendency for attributers to
social and sociological factors in the development of expertise 751
with such decision-making experts. Who, the question Can we trust in experts? (soci-
as a professional, wants to see his or her etal context).
expert judgment explained to the pub-
lic by a formal expert? Therefore, experts
Organizational Context: The
desiring to be recognized and consulted
Division of Labor
as formal experts have to professionalize
themselves, for instance as specialized con- In general, organizations are forms of labor
sultants or risk professionals (Dietz & division. Max Weber (1979) regarded ratio-
Rycroft, 1987). nally administered divisions of labor as bure-
An experts role also determines the scope aucracies, impersonal organizations of tasks
of accountability for the experts work. Pro- and specialists. From a more modern per-
fessionals account for the complete profes- spective, companies and similar organiza-
sional task, including treatment. Relative tions are considered resource pools that
experts account only for the information assemble human, financial, material, and
they provide, researchers for the correctness other resources (see, e.g., Engestrom, 1991).
of their analyses. If we take into account In both versions, the organization involves a
what is said about human capital and per- connected distribution of relative expertise.
sonal causal attributions (in previous sec- The success of such an organization depends
tion), we can say that experts represent not on the art of combining the resources in an
only units of expertise (as human capital) effective way. We can reveal the particular
but also units of accountability for the appli- distribution of expertise by assessing who is
cation of expertise in accordance to their used as a source for what sort of informa-
expert role. tion and know-how in an organization (e.g.,
To summarize: We have started our anal- Stein, 1992).
ysis of expert roles in this chapter with the Studies in organizational psychology have
example of asking someone on the street shown that groups and other organizational
for directions to the station. We have seen structures can learn. They develop a kind of
that this simple case of relative expertise memory, transactive memory (Wegner, 1987),
contains all the elements of The expert- that is a coordinated and distributed stor-
interaction, that is, the situation where we age of knowledge. Transactive knowledge is
consult an expert, including expertise as a based on the fact that we can use other peo-
form of human capital. We concluded with ple as external memory, for instance, by
a typology of expert roles. asking a colleague about details on meetings
we have missed. A large part of a secretarys
job consists of reminding his or her superi-
Contexts of Expertise ors of appointments. Transactive memory is
a property of a group (p. 191), which can
I will review expert roles in three of the be a family or a company.
most important social contexts: the organi- A central mechanism in transactive mem-
zational context, the professional context, ory is the attribution of expertise. This has
and the societal context. The general soci- been demonstrated in experiments with co-
ological perspective on experts, highlighting workers or work groups in laboratory set-
professions and the context of science, has tings (see Hollingshead, 2001). Best group
already been addressed in Chapter 7. The performances (in most times recall tasks)
study of the organizational context of expert were found where expertise was clearly at-
work would fill a separate book. Therefore, tributed to particular group members. More-
I can focus only on specific aspects of exper- over, transactive memory seems to be most
tise in context: the division of labor (organi- differentiated when the group members pos-
zational context), the definition of perfor- sess different areas of expertise and incen-
mance criteria (professional context), and tives to remember different information.
754 the cambridge handbook of expertise and expert performance
experts is ubiquitous. We find all administra- damages due to working conditions. At the
tions using expert panels in order to demo- behest of the court, a physician examined
nstrate the severity of problems and the all of the workers and reported findings
necessity of administrative work, thus cre- for each plaintiff. The physicians court-
ating a demand for bureaucratic staff and appointed status was disclosed to the jury,
and the judge reported the jury discounted
funds (as Jasanoff [1994] describes it for the
the experts for each side. In fact, in each
relation between the Environmental Protec-
individual case, the jury followed the find-
tion Agency [EPA] and its Science Advisory ings of the court-appointed expert, finding
Board). And, of course, we find experts who sometimes for the plaintiff and sometimes
directly play into politics, promoting them- for the defendant. (loc. cit., p. 5 4)
selves, as in the case of the IPCC, the Inter-
governmental Panel on Climate Change. The Statutory judicial frameworks, such as in
IPCC has itself become a powerful interna- France and Germany, tend to systematize
tional expert system with an impact on aca- and differentiate legal matters, including
demic course programs and scientific fund actors and functions. In statutory systems,
raising by supporting international climate the role of the expert in the court is more sys-
politics. tematized than in common law. In German
An intriguing case of experts and trust can law, for example, the expert is considered as
be found in courts. In US trials, experts are a judicial clerk or clerk of the judge who,
a type of witness. The expert can testify on under the supervision of the judge, helps the
almost anything that is helpful and relevant court interpret and understand a case.
to the trial (Rossi, 1991). Experts in US tri- To summarize: Expert roles and the attri-
als, as introduced here, are a helpful means bution of expertise serve an important func-
to the parties involved in a trial. The adver- tion in organizations as well as in society.
sary process prevents the jury from undue However, the social form The expert has its
deference to experts. In practice, many of own dynamics that can run into conflict with
the critical aspects of using experts such organizational or public constraints, as we
as, What qualifies a person as an expert? saw in the case of experts in court. This raises
or What data form the basis for an experts awareness for expert-performance criteria.
opinion? are left to cross-examination by
the lawyers. It is in the interest of each
party to examine the experts of the adver- Mediating Processes in the
sary party and demonstrate lack of evidence Development of Expertise
if this proves to be the case. There is also the
possibility of employing a court-appointed Having discussed the contexts of expertise
expert (Federal Rule of Evidence 706). and expert work, we can now turn to the
However, they are employed very infre- question of the contexts that nourish the
quently. Appointment of experts through development of expertise. I will start with
the court transcends the adversary system the process of socialization and then intro-
of common law. Thus the social form The duce some selected more or less psycholog-
expert comes into play again and the con- ical approaches. We will leave the discus-
sequences of The expert-interaction have sion of the social form The expert and
to be taken into account. A survey revealed come to what Simmel had called content:
that juries and judges alike tend to decide expertise. Thereby, we return to the more
cases consistent with the advice and testi- classical examples of expertise (see also
mony of court-appointed experts (Cecil & pertinent Chapters in Section V), such as
Willging, 1993 , p. 5 2): chess, the medical profession, or sports, for
the simple reason that these activities have
The most dramatic illustration of domi- gained a certain social function. However,
nance by a court expert occurred in a case we should not forget that the very chal-
in which a large number of workers claimed lenging cases, the ones where individual
756 the cambridge handbook of expertise and expert performance
expertise and social context are interwoven, ues and rewards. Such a subculture exists,
such as in entrepreneurs or politicians, are for instance, in some families of physicians
not addressed here. that have for generations maintained a cul-
ture of values and routines such as fam-
Socialization ily music as well as an ethics of personal
care for patients and the community. These
The main mediating socio-psychological family subcultures even survived the former
process during the development of expertise German Democratic Republic (19491990)
is socialization. Socialization generally refers that tried to wipe out the traditional health
to the process of social influence through system based on an autonomous medical
which a person acquires the culture or sub- profession (Hoerning, 2003 ).
culture of his or her group, and in the course
of acquiring these cultural elements the indi-
viduals self and personality are shaped the school context
(Gecas, 2001, p. 145 25 ). Classic sociology Schools are an important element in an
considered socialization as a process of inter- education system. They transmit not only
nalizing social roles (e.g., Parsons, 195 5 ). essential skills and knowledge but also cul-
Modern sociology views socialization as the tural norms for excellence. This transmis-
formation of identities (Gecas, 2001). As the sion often happens via narratives, story-like
acquisition of expertise is based on deliber- mental models representing how life may be
ate practice and long-term training, we can (Ferrari, 2002).
expect socialization to exert a strong impact In some states, primary schools are part
on the development of expertise. of an encompassing system of talent screen-
Socialization takes place in different ing and selection. This was, for instance, the
social contexts family, schools, peer groups, basis for the extraordinary success of physi-
work, and so forth. The importance of these cal education in the former German Demo-
contexts varies during the lifetime of a cratic Republic. In the post-1970 German
person. Democratic Republic, every 6th young boy
and every 18th young girl had to start train-
the family context ing in one of the specialized national sports
training centers (Trainingszentren), which
In general, parents are the most effective
formed the basis of a system of sports
agents of socialization when they express
schools, sports research institutes, and com-
a high level of support or nurturance com-
petitions (Teichler & Reinartz, 1999).
bined with the use of inductive control
(Gecas, 2001, p. 145 27). In the development
of extraordinary minds, the orderly life in peer groups
a bourgeois family is a favorable environ- Peer groups are voluntary associations of
ment (Gardner, 1998). status equals and are based on friendship
Families have always played an important bonds (Gecas, 2001, p. 145 28). Among
role in nurturing high levels of expertise. youths, peer groups play an important role
Famous examples are families of musicians in forming and reinforcing the self-identities
such as the Bach family or the Mozart fam- of their members. Peer groups can func-
ily, where the parents trained their children tion as a fertile soil for the development of
from an early age. These German musicians skills in team sports (e.g., basketball, soccer)
families appeared with the Kapellmeister and in the dramatic or performing arts (e.g.,
profession, that is, conductors-composers music bands, acting).We should also not for-
who could make a living by working for one get the ethnic context, such as the socializa-
of many small princedoms that coexisted on tion context of immigrant children in the
German territory after the Thirty Years War USA (Rumbaut & Portes, 2001). Families and
(16181648). peers play an important role in transmitting
Families can create their own particu- impacts within ethnic groups. In some ethnic
lar subculture with its own system of val- groups, children of immigrants display an
social and sociological factors in the development of expertise 757
(p. 3 7). Bandura demonstrated the positive solved is the relationship between socializa-
influence of self-efficacy on performance tion, levels of cognitive development, and
in various domains (see also Zimmerman, transitions between task contexts (school,
Chapter 3 9). positions) that shape and provoke the devel-
opment of expert performance. There are
Individual and Context promising approaches, such as the selection-
optimization-compensation model of aging
The works of Jean Piaget, Lev Vygotsky, and
by Paul Baltes (1997), or the time-span
Sylvia Scribner stand for a series of classical
capacity model of managerial work by
studies on socio-cognitive development. Piaget
Elliott Jaques (1976), both first steps toward
(193 6/195 3 ) described stages of cognitive
a comprehensive theory of human expertise.
child development. Vygotsky (193 4/1962)
introduced the concept of the zone of prox-
imal development, this zone describing the
difference between what a child can do Footnote
unguided, on the one hand, and with
guidance, on the other. Scribner expanded 1. In German, we would use the word Leistung
Vygotskys socio-cultural approach to adult (see Mieg & Pfadenhauer, 2003 ).
cognition. In studies on adult cognition,
such as working intelligence (Scribner,
1984/1997), she showed how the develop- References
ment of particular cognitive capabilities is
linked to work-specific experience (see also
Clancey, Chapter 8). Abbott, A. (1988). The system of professions.
Chicago: University of Chicago Press.
The Context Adler, A. (1912). Uber den nervosen Charak-
ter [The neurotic constitution]. Munchen:
In 1960, Donald T. Campbell wrote a paper Bergmann.
on the blind variation and selective reten- Agnew, N. M., Ford, K., M., & Hayes, P. J.
tion in creative thought as in other knowl- (1997). Expertise in context: Personally con-
edge processes, arguing that the set of cre- structed, socially selected and reality-relevant?
ative personalities is subject to variation and In P. J. Feltovich, K. M. Ford, & R. R., Hoffman
selection. In a similar vein, I argued (Mieg, (Eds.), Expertise in context: Human and machine
2001) that by using an expert, we use the (pp. 219244). Menlo Park, CA: AAAI Press.
specific experience of someone else (the Antons, C. (2004). Folklore protection in
expert) in order to solve a problem or to Australia: Who is expert in aboriginal tradition?
find an explanation. Particularly in domains In E. Kurz-Milcke & G. Gigerenzer (Eds.),
with poor expert decision performance (cf. Experts in science and society (pp. 85 103 ). New
Shanteau, 1992), such as financial markets York: Kluwer Academic.
or business consulting, experts can be used Baltes, P. B. (1997). On the incomplete archi-
like hypotheses (like heuristics) that work tecture of human ontogeny: Selection, opti-
mization, and compensation as foundation of
successfully as long as a certain work envi-
developmental theory. American Psychologist,
ronment does not change. When environ-
5 2 , 3 663 80.
ments change, we can test new approaches
Bandura, A. (1997). Self-efficacy: The exercise of
by exchanging the expert.
control. New York: W. H. Freeman.
Baron, J. N., & Hannan, M. T. (2002). Organi-
to summarize
zational blueprints for success in high-tech start-
The study of mediating processes in the ups: Lessons from the Stanford project on emerging
development of expertise shows the impor- companies. Stanford, CA: Graduate School of
tance of connecting the psychological to Business, Stanford University.
the sociological perspective. One of the Becker, G. S. (1993 ). Human Capital (3 rd ed.).
most challenging scientific puzzles still to be Chicago: The University of Chicago Press.
social and sociological factors in the development of expertise 759
Mieg, H. A. (2001). The social psychology of exper- Shanteau, J. (1992). The psychology of experts.
tise. Mahwah, NJ: Erlbaum. In G. Wright & F. Bolger (Eds.), Expertise and
Mieg, H. A., & Pfadenhauer, M. (Eds.). decision support (pp. 1123 ). New York: Plenum.
(2003 ). Professionelle Leistung Professional Simmel, G. (1908). Soziologie: Untersuchun-
Performance: Positionen der Professionssoziologie gen uber die Formen der Vergesellschaftung
[Professional performance: Approaches to the [Sociology: Studies on the forms of sociation].
sociology of professions]. Konstanz: UVK. Leipzig: Duncker & Humblot.
Kurz-Milcke, E., & Gigerenzer, G. (Eds.). (2004). Simmel, G. (1971). On individuality and social
Experts in science and society. New York: Kluwer forms (trans. by K. H. Wolff, ed. by D. N.
Academic. Levine). Chicago: The University of Chicago
Otway, H., & von Winterfeldt, D. (1992). Expert Press.
judgment in risk analysis and management: Stasser, G. (1992). Pooling of unshared informa-
Process, context, and pitfalls. Risk Analysis, tion during group discussion. In S. Worchel, W.
12 (1), 83 93 . Wood, & J. A. Simpson (Eds.), Group process
Parsons, T. (195 5 ). Family structure and the and productivity Newbury Park, CA: Sage.
socialization of the child. In T. Parsons & R. F. Stein, E. W. (1992). A method to identify can-
Bales (Eds.), Family, socialization, and interac- didates for knowledge acquisition. Journal of
tion Processes. Glencoe, IL: Free Press. Management Information Systems, 9, 161178.
Piaget, J. (195 3 ). Origin of intelligence in the child. Stewart, D. D., & Stasser, G. (1995 ). Expert role
London: Routledge (French original in 193 6.). assignment and information sampling during
Platt, G. M. (2001). Status and role, social psy- collective recall and decision making. Journal
chology of. In N. Smelser (Ed.), International of Personality and Social Psychology, 69(4), 619
encyclopedia of the social & behavioral sciences 628.
(pp. 15 09015 095 ). Amsterdam: Elsevier. Teichler, H. J., & Reinartz, K. (1999). Das Leis-
Putnam, R. D. (1976). The comparative study of tungssportsystem der DDR in den 80er Jahren
political elites. Englewood Cliffs, NJ: Prentice und im Proze der Wende [Training systems
Hall. in professional sports in the GDR in the
Ross, L. (1977). The intuitive psychologist and 1980s and subsequent to German unification].
his shortcomings: Distortions in the attribu- Schorndorf: Verlag Karl Hofmann.
tion process. In L. Berkowitz (Ed.), Advances Vygotsky, L. S. (1962). Thought and language.
in Experimental Social Psychology (pp. 173 220). Cambridge: MIT Press. (Russian original in
New York: Academic Press. 193 4.)
Rossi, F. F. (Ed.). (1991). Expert witnesses. Weber, M. (1979). Economy and society (Vol. I,
Chicago: American Bar Association. G. Roth & C. Wittich, Trans.). Berkeley, CA:
Rumbaut, R. G., & Portes, A. (Eds.). (2001). University of California Press.
Ehnicities: Children of immigrants in America. Wegner, D. M. (1987). Transactive memory: A
Berkeley, CA: University of California Press. contemporary analysis of the group mind. In
Scribner, S. (1997). Studying working intelli- B. Mullen & G. R. Goethals (Eds.), Theories
gence. In E. Tobach et al. (Eds.), Mind and of group behavior (pp. 185 208). New York:
social practice: Selected writings of Sylvia Scribner Springer.
(pp. 3 3 83 66). Cambridge: Cambridge Uni- Weiner, B. (1986). An attributional theory of moti-
versity Press. (Originally & published in 1984.) vation and control. New York: Springer.
C H A P T E R 42
Robert W. Weisberg
that expertise facilitates creative thinking tion, we also use the term expert to refer to
because deliberate practice enables the a person who exhibits a high degree of com-
would-be creator to develop new tech- petence, but we do so irrespective of how
niques or skills, which allow him or her that competence was acquired. In this chap-
to go beyond what had previously been ter I will use expertise in the ordinary sense, to
accomplished. We can derive three testable refer to the capacity to perform consistently
hypotheses from Ericssons proposal: at a superior level, without regard to how
(1) expertise is necessary for creative that capacity was acquired. In a number of
accomplishment; (2) creative advances places I examine the specific role of prac-
develop as the result of new techniques tice in innovation, but when I use the term
and skills; (3 ) creative advances extend the expert or expertise without modification, I
boundaries of the field of endeavor. am including practice and study under one
The present chapter examines the sup- umbrella (see also Weisberg, 1999).
port for those three hypotheses. The first sec- Some clarification is also needed concern-
tion of the chapter provides further elabora- ing my definition of creativity goal-directed
tion on the definitions of relevant concepts, production of novelty and related con-
in order to eliminate several possible points cepts. Most researchers who study creativity
of confusion and to clarify possible roles of usually include the value of an innovation as
expertise in creativity. I then consider sev- a criterion for calling it creative. (See chap-
eral possible objections to the notion that ters in Sternberg, 1999 for numerous exam-
expertise might be relevant to creativity, in ples.) Including positive value in the defini-
order to place the present investigation in tion of creative means, for example, that
a broader context. The next section of the a candidate solution must actually solve the
chapter responds in broad terms to those problem in order to be called creative. Sim-
objections. I then review in detail several ilarly, an invention must carry out the task
case studies of seminal creative advances for which it was designed; a scientific theory
that provide evidence relevant to the three must be useful; and a work of art must find
hypotheses outlined above. The results of an audience. Researchers who include value
the analyses of the case studies indicate that in their definition of creative do so in order
expertise can play several roles in creative to be able to rule out simply bizarre prod-
advances. In the final section of the chapter, ucts from consideration as innovations (e.g.,
possible limitations on the role of expertise the word salad of a schizophrenic, produced
in creative accomplishment are discussed. perhaps in response to a problem). However,
I recently presented evidence that can be if we include intentional production of nov-
taken as support for the hypothesis that elty in the definition, it also precludes the
expertise is necessary for creative thinking schizophrenics word salad, since no one in
(e.g., Weisberg, 1999, 2003 , 2006); the anal- schizophrenic episode would be able to deal
ysis in this chapter examines the limits of intentionally with a problem.
that claim. I believe that including value in the def-
inition also clouds several important issues
(Weisberg, 1993 , Chapter 8; 2003 , 2006).
Further Questions of Definition
Most critically, including value as part of the
One difficulty in examining the role of definition of creative means that if, for exam-
expertise in creative thinking is that the ple, an audience comes to value a previously
terms expert and expertise have meanings in ignored work of art, the attribution of cre-
the research literature that are different from ative to the work, and, ipso facto, to the artist
ordinary language, and this can cause con- who produced it will change: a previously
fusion. As noted, in much of the literature, noncreative artist will become creative (even
an expert is someone who exhibits consis- after death). Conversely, an artist may, if his
tent superior accomplishment as the result or her work falls out of favor, become non-
of deliberate practice. In ordinary conversa- creative. Such changes mean that we could
expertise in creative thinking 763
never develop theories of creative thinking basis for transfer of knowledge to the new sit-
because the data base on which we build our uation, where that knowledge serves as the
theories will be constantly changing. That foundation for innovation. Second, research
is, we would have to add the new people that has studied undergraduates solving lab-
who became creative since we formulated oratory problems has found that creative
our conclusions and subtract those who have products can also come about as the result
become noncreative. We would then have of an individuals analysis of a problem-
to reexamine our data base to determine if atic situation in which he or she is not an
any previously formulated conclusions are expert in the sense discussed earlier (e.g.,
no longer valid. This is obviously an unten- Fleck & Weisberg, 2004; Perkins, 1981; Weis-
able situation for researchers. berg & Suls, 1973 ). In such cases, creative
In my work (e.g., Weisberg, 1993 , 2003 , thinking may not depend on domain-specific
2006), I differentiate between a creative experience, but rather on general expertise,
product (a goal-directed innovation) and what we also call general knowledge, such
one that is valued, influential, significant, or as logical-reasoning ability or mathemati-
important, which I use as near synonyms. cal ability. This distinction between domain-
With a work of art, the audiences reaction specific versus general expertise is similar to
determines whether or not it will be valued. the distinction between strong versus weak
In the case of inventions, scientific theories, methods of problem solving (Newell, 1973
and solutions to problems, the effectiveness Feltovich et al., Chapter 4).
of the innovation is critical in that determi- One example of the role of general exper-
nation. We thus can have valued and nonva- tise in problem solving and, therefore, in
lued innovations; even if an innovation is of creative thinking comes from a study by
no value, however, it is still creative. In the Fleck and Weisberg (2004; see also Weis-
present context, the specifics of the defini- berg, 1980, Chapter 9; Weisberg & Suls,
tion are not of critical concern since all the 1973 ), who examined the processes under-
innovations to be discussed in this chapter lying solution of the Candle Problem. In this
are undoubtedly of the highest value. How- problem, the individual is asked to attach a
ever, it is important to clarify the definition candle to a wall or a similar vertical surface
now so that no one will object that ignoring and is supplied with a box of tacks or fasten-
the value of a product might have affected ers and a book of matches. One solution that
the conclusions drawn concerning creative has been of particular interest to researchers
thinking. is the box solution, that is, the use of the tack
box as a shelf or container for the candle.
This solution is not produced by a majority
of people attempting the problem, and when
Expertise and Creativity
it is produced, it is usually not produced as
the first solution proposed by an individual.
Domain-Specific versus General Modes
Research examining the fine grain of the pro-
of Expertise
cesses involved in attempting to solve the
Recent research studying creative thinking candle problem indicates that the box solu-
has indicated that innovations can develop tion often develops in response to difficulties
in at least two ways. Case studies of cre- that arise when the individual tries to attach
ative thinking at the highest levels (e.g., the candle directly to the wall, as requested
Weisberg, 1999, 2003 , 2004, 2006) have in the instructions. If the individual tries to
indicated that creative ideas can be built use melted wax as an adhesive to attach the
relatively directly on the past, as creative candle, for example, he or she may find that
individuals use what they know about the the candle is too heavy. This failure may lead
domain in question as the basis for creating the individual to search for something to
something new. In such situations, what one use to hold up the candle a shelf or can-
can call domain-specific expertise serves as the dle holder. This search can result in the box
764 the cambridge handbook of expertise and expert performance
being used. In this example, the problem is faced with a new X-ray to interpret. In
solver presumably had no deep expertise that case, the expert can use domain-specific
attaching candles to walls. However, most expertise in solving the problem he or she
people know enough about the properties is facing. Problem-solving exercises used in
of candles, fasteners, and shelves, and pos- laboratory studies are relatively simple and
sess enough general skill working with their bare of information, which may account for
hands, that they can fashion a shelf out of why general expertise can suffice to solve
a tack box as needed. In implementing the them. Creative thinking in the real world,
box solution to the Candle Problem, then, in contrast, occurs in complex, information-
true domain-specific expertise may not be rich environments. It therefore becomes of
available. However, more general expertise interest to determine if there are examples
is used. of creative advances in real-world settings
Similarly, Perkins (1981) examined pro- in which general knowledge, independent of
cesses underlying solution of the Antique domain-specific information, plays the lead-
Coin Problem: ing role.
Macromolecule
Own Ideas
DOMAIN-SPECIFIC
EXPERTISE
Macromolecule Others
Ideas
Other macromolecules
Life Sciences
Physical Sciences
Mathematics
Logic
GENERAL EXPERTISE
of transfer of knowledge or expertise, from both specifically involving love and death
domain-specific to general, when discussing and more general, and so forth, until at the
creative advances, so that findings from case broadest level we might see the incorpo-
studies and laboratory studies can be inte- ration of science, logic, and mathematics.
grated. Again, the specific details of the scheme
Figure 42.1B presents the same sort of in Figure 42.1B are not of concern here;
analysis applied to a hypothetical example the important point is that one can out-
from painting, a domain in artistic creativ- line a movement away from domain-specific
ity. Let us say that an artist has been stimu- expertise to more-general aspects of exper-
lated by the death of a loved one, and he or tise that can be brought to bear on the prob-
she begins to create a painting in response. lem faced by the hypothetical artist. In the
Here too we can outline a set of ever-more- case of a person who possesses general exper-
general domains of expertise that might be tise, we sometimes use the terms knowledge
brought to bear on this project, beginning or general knowledge to describe the state of
with the artists feelings concerning love the person.
and death and his or her earlier works con- The purpose of this chapter is to con-
cerned with those issues. The domain broad- sider various case studies of seminal cre-
ens to include the artists own works on a ative advances, to determine in each case
broader range of topics and works of other if expertise was brought to bear in pro-
artists, both works addressing love and death ducing the innovation and, if so, to deter-
as well as other topics. We then go more mine the domain specificity or generality of
broadly to include influences from other arts, that expertise. It should also be noted that
766 the cambridge handbook of expertise and expert performance
Artists Ideas
on Love and Death
DOMAIN-SPECIFIC
EXPERTISE
Other Arts
Humanities
GENERAL EXPERTISE
Figures 42.1A and 42.1B make clear that, expertise cannot be the basis for creativity.
when we discuss domain-specific versus gen- This view, from my perspective, postulates
eral expertise, we are in actuality talking tension of several types between expertise
about a continuum, ranging from informa- and creativity, so I refer to it as the tension
tion directly relevant to the specifics of the view. The tension view has a long history
problem at hand, through less directly rele- in psychology (e.g., James, 1880) and has
vant but still domain-specific information, to many advocates today (e.g., Csikszentmiha-
information relevant on a more general level lyi, 1996; Simonton, 1999; Sternberg, 1996).
to the problem, and so forth. It is also not Thus, it is important to consider reasons why
necessary for the present discussion that we theorists have rejected expertise as the basis
specify exactly how each example of exper- for creativity.
tise is to be classified. All that is needed is
that we be able to make gross differentia-
tions among examples of expertise of differ- Skepticism about Expertise and
ent degrees of domain specificity. As will be Creativity: The Tension View
seen in the discussion of the case studies in
the chapter, that will be possible. There are a number of different argu-
Before turning to a consideration of the ments that might lead one to believe that
possible role of the various modes of exper- expertise and creativity are unrelated or
tise in creativity, it will be useful to dis- even that expertise might be an impedi-
cuss the opposite view, that is, the idea that ment to creativity (Ericsson, 1996, 1998;
expertise in creative thinking 767
Weisberg, 1999). First, the concept of creative thinking. If an expert musical per-
expertise involves an automatic mode of former (see also Lehmann & Gruber, Chap-
responding, where the individual does not ter 26), for example, is told by a conduc-
think about what he or she is doing. An tor that her playing of a piece should be
example is driving a car, which most of us less emotional, she will adjust her perfor-
carry out expertly and automatically, with- mance to achieve that end. One could argue
out thought. This sort of automatic respond- that this ability to adjust ones behavior to
ing would seem to be incompatible with demands arising in the situation is an exam-
creative thinking, which involves conscious ple of creative thinking. So, by this view, con-
deliberate processing, although automaticity clusions from research on expertise might be
in performance may free up cognitive capac- broadly relevant to creative thinking.
ity for deliberation (see also, Hill & Schnei- In addition, expertise does not include
der, Chapter 3 7; Endsley, Chapter 3 6, this just mindless, automatic processing (al-
volume). Second, many researchers believe though it usually involves some degree
that talent (a constellation of inherited skills of automated processing)(see also Hill &
that makes a person especially suited to excel Schneider, Chapter 3 7, this volume). The
in a specific domain), rather than exper- expert acquires a rich, highly complex con-
tise based on experience and practice, plays ceptual structure that is used consciously to
a critical role in enabling superior levels represent and reason about situations. Evi-
of achievement (Sternberg, 1996; Winner; dence for such a structure can be seen in the
1996, Horn & Masunaga, Chapter 3 4, this ability of chess masters to play several games
volume). If so, then concentrating on expe- at once while blindfolded, that is, solely
rience and practice with regard to exper- from memory (see also Gobert & Char-
tise and its role in creativity is misdirected. ness, Chapter 3 0, this volume). In order to
Finally, it is believed by many researchers, carry out such a task, the expert must have
as well as by many in our society, that cre- available a rich and detailed representation
ative thinking requires that one not rely on of each game so that it can be remembered
the past, as exemplified by knowledge and and effective moves can be made. Experts
habit (e.g., Csikszentmihalyi, 1996; Simon- in other domains show similar abilities.
ton, 1999). This view is captured by the For example, musical performers sometimes
ubiquitous idea that creativity requires that have prodigious memories for the pieces in
we think outside of the box. Expertise, their performance repertoires. The expert
however, involves encapsulation of the past, thus uses a detailed analysis of the situation
so expertise would seem to be in conflict that he or she is facing in order to exercise
with the needs of the creative thinker (e.g., conscious adaptive processing. Only as the
as in the restrictive effects of problem- result of experience and practice will an indi-
solving set and related phenomena; see vidual possess the detailed representations
Luchins & Luchins, 195 9; Scheerer, 1963 ; for of a situation needed to support creative
a different perspective, see Feltovich, Spiro, thinking. Furthermore, there is evidence that
& Coulson, 1997; Weisberg, 1980, ch. 9). raises questions concerning the role of tal-
In contrast to those beliefs concerning ent in superior levels of achievement, at
the possible negative relationship between least in the domain of performance of classi-
expertise and creativity, Ericsson (e.g., 1996, cal music (e.g., Sloboda, 1996), which indi-
1998) has recently proposed that exper- rectly supports the importance of expertise.
tise and creativity are intimately connected. Finally, in contrast to the idea that creativ-
Concerning the notion of automaticity and ity always involves rejecting the past (think
lesser creativity in certain domains, Eric- outside the box), there is, as noted, evi-
sson proposed that domains such as ath- dence that creative thinking can build on the
letic performance, performance of classical past: new ideas can come about as the result
music, and medical diagnosis are more open of an individuals building on old ideas (e.g.,
than many realize and, therefore, can require Weisberg, 1999, 2003 , 2004, 2006). It is an
768 the cambridge handbook of expertise and expert performance
Gardner, 1993 ). These results indirectly sup- reached by an individual. Practice may be
port the notion that expertise is important important in musical performance or swim-
in creativity since it would be expected that ming, but, according to Sternberg, exper-
developing expertise would take time. tise researchers may have ignored domains
One limitation to Hayess (1989) inves- in which talent is more important than
tigation is that, because he presented data practice (e.g., musical composition or paint-
summarized across large groups of indi- ing). According to Sternberg, practice can-
viduals, he presented no information con- not account for the extraordinary early
cerning the specific activities that occurred achievements of Mozart or Picasso.
during the pre-masterwork years. Thus,
although Hayess results are consistent with Why was Mozart so damn good? . . . What
the hypothesis that expertise is necessary made Picasso so good so young? (p. 3 5 0) . . .
[W]hat Mozart did as a child most musi-
for creativity, they do not provide specific
cal experts will never do nor be able to
information concerning the actual develop-
in their lifetimes, even after they have
ment of expertise in any individuals. Accord- passed many times over the amount of time
ingly, I now turn to several case studies of Mozart could possibly have had for delib-
creative achievements of the first rank from erate practice as a child. (p. 3 5 1) . . . We
the arts, science, and invention to investigate fail to see evidence all around usscholarly
in more detail the question of the neces- and common-sensicalthat people differ in
sity of expertise in creativity. We will find their talents, and that no matter how hard
in those studies a range of uses of exper- some people try, they just cannot all become
tise. Some of the cases correspond directly to experts in the mathematical, scientific, lit-
what would be expected on the basis of the erary, musical, or any other domains to
which they may have aspired. (p. 3 5 2 ) . . .
Ten-Year Rule, with innovation dependent
The truth is that practice is only part of
on domain-specific expertise and deliberate
the picture. Most physicists will not become
practice playing a critical role in the devel- Einstein. And most composers will wonder
opment of that expertise. In other cases, why they can never be Mozart. (p. 3 5 3 )
explicit practice may not be seen, but a crit-
ical role is nonetheless played by domain- One piece of evidence that raises ques-
specific expertise. Finally, in two cases, sem- tions for Sternbergs view of Mozart is
inal creative advances may have come about Hayess finding that the Ten-Year Rule holds
by a combination of domain-specific and even for him (and, as we shall see, it holds
general modes of expertise. also for Picasso). As noted, Hayess analy-
sis provides no information about the years
Case Studies of Creative Thinking before the first masterwork. Based on the
hypothesis that expertise is necessary for cre-
Musical Composition ativity, and on the expertise literature, one
might expect to find Mozart developing his
the young mozart skills over those years, as reflected, for exam-
I recently examined in detail the career ple, in increasing production of composi-
development of Mozart (Weisberg, 1999, tions and in their increasing quality. There
2003 ), who, sometimes along with Picasso, should also be evidence for the occurrence
is often cited by researchers as the pro- of deliberate practice during the formative
totype of the creator whose abilities are years.
impossible to understand without invoking In order to test those expectations, I
a concept like talent or giftedness. Stern- looked in detail at Mozarts development, in
berg (1996) discussed Mozarts accomplish- three ways (Weisberg, 2003 ). I examined the
ments in the context of a critique of research number of compositions produced during
on expertise, specifically of the notion that the various years of Mozarts career (Hayes,
practice might be more important than tal- 1989) and found that his output increased
ent in determining the level of achievement over the first ten years or so of his career,
770 the cambridge handbook of expertise and expert performance
supporting the notion that he was master- ductivity over the early years for Beethoven
ing his craft. Second, I measured the qual- and Haydn, for example, mirrored that of
ity of Mozarts early compositions by deter- Mozart. There was an increase in quantity
mining the average number of recordings for and quality of compositions, indicating that
each composition for each year. The qual- those individuals too were developing the
ity of Mozarts compositions increased over skill of writing music. Kozbelt (2004) has
the early years of his career, which also sup- also examined Mozarts career in detail, as
ports the idea that he was honing his skill. well as the careers of other classical com-
Finally, there is evidence that Mozart was posers, and he has also found increases in
carrying out deliberate practice over those quality of work over their careers in a major-
years under the direction of his father, a ity of them. (For further discussion of the
professional musician of some repute. Con- importance of talent versus practice and
sider Mozarts earliest piano concertos, the expertise in music, see Sloboda, 1996.)
first four written at the ripe old age of 11, In conclusion, studies of the development
and the next three written when he was of classical composers support the claim that
16. Those works contain no original music domain-specific expertise was being used as
by Mozart: they are simply arrangements of the basis for composition. If one developed
music of other composers. Mozarts father an analysis similar to that in Figure 42.1B
may have used others music as the basis for classical composition, based on the case
for practice by the young man in writing for studies just discussed, one would place the
groups of instruments. Furthermore, if some influence of expertise relatively close to the
of the published works by the young Mozart domain-specific core of the diagram.
are based completely on the works of oth-
ers, then Mozarts private tutelage from his
father must also have centered on study of the beatles
works of others. So Mozart learned his craft Evidence for the Ten-Year Rule in cre-
over many years, under the watchful eye of a ative thinking also comes from a study of
professional teacher. This training is not dif- the development of the Lennon-McCartney
ferent from that received today in schools of songwriting team, whose songs for The
music by aspiring composers. Beatles broke new ground in popular music
These results call into question Stern- in the 1960s (Weisberg, 1999, 2003 ). The
bergs (1996) claim that most composers Beatles career trajectory corresponds in sev-
will never approach the accomplishments eral ways to the findings from the studies
of Mozarts early years. We have just seen of acquired expertise. When The Beatles
that a number of Mozarts early composi- hit the big time in 1963 , they had already
tions show no originality on his part. Many been working together for several years, and
of his other early works, which do con- they had spent thousands of hours playing
tain his own music, have been more or together. So there was a period of appren-
less ignored by musicians and audiences, ticeship before The Beatles made a signif-
which means that those works are not icant contribution to pop music. Although
so . . . good. They have nothing distinc- this early period did not involve deliberate
tively Mozartian about them. Thus, where- practice in the sense of formal tutelage under
as it is no doubt true that most composers the supervision of a teacher, Lennon and
will not match Mozarts ultimate achieve- McCartney began their careers by immers-
ments, his early achievements are matched ing themselves in the works of others. A
by many composers as they advance through large majority of the songs played by The
music school. Recent analyses of the career Beatles in their early years were cover ver-
development of other seminal classical com- sions of hits recorded by others. This immer-
posers Bach, Beethoven, and Haydn sup- sion in the works of others served as a
ports the findings from Mozart (Weisberg kind of unstructured practice. In addition,
& Sturdivant, 2006). The pattern of pro- there was more explicit practice. Lennon and
expertise in creative thinking 771
McCartney spent much of their early years the environment selectively retains some of
together in active collaboration, in which those blind variations at the expense of oth-
any new information concerning musical ers. Simonton has applied this view to cre-
structure acquired by either (e.g., if either ative thinking, assuming that two stages are
learned new chords on the guitar) was shared involved there also. The first stage, involv-
and served as the basis for explorations of ing production of new ideas, involves ran-
new possibilities of composition (Everett, dom combinations of old ideas. The second
2001). process, selective retention, selects and pre-
It should also be noted that several serves only some of those variations, those
additional years passed before Lennon and which meet a criterion for acceptability.
McCartney made their major contributions If random combination of ideas is the
to popular music. The very earliest songs first step in creative ideation, then expertise
written by The Beatles were not big hits, becomes at least irrelevant and perhaps an
and most of them are forgotten, except by impediment to creativity. Along those lines,
collectors. Most of those songs were only Simonton has presented evidence for what
recorded late in their career, when their he calls the equal-odds rule as support for the
early music became interesting because of role of a random process in the first stage of
what The Beatles had become, not necessar- creative thinking. On the basis of a random
ily because of the quality of the songs them- process as the core of the first stage, Simon-
selves. The most significant music produced tons theory predicts that the probability of
by The Beatles is usually considered to be the a creators producing a masterwork should
middle-period albums Rubber Soul (1965 ), stay constant over a career. The equal-odds
Revolver (1966), and Sergeant Peppers Lonely rule is in conflict with the notion that exper-
Hearts Club Band (1967), created approx- tise is critical in creative thinking. Assum-
imately ten years into their career (Reis- ing that expertise serves as the basis for cre-
ing, 2002). The Beatles development thus ative thinking, leads to the expectation that
supports the notion that domain-specific creative people should develop their skills
expertise was important in their achieve- over time. The results presented in the last
ments, and, comparable to the results few sections, which demonstrated just such
for Mozart and other classical composers, a development in musical composition, thus
deliberate practice seems to have been contradict the equal-odds rule.
involved. Those results leave us with the question of
why Simonton found evidence for the equal-
increasing quality in musical composition odds rule, whereas the results emphasized
versus the equal-odds rule here (e.g., Hayes, 1989; Kozbelt, 2004; Weis-
The finding that the quality of musical com- berg, 2003 ; Weisberg & Sturdivant, 2006) do
positions increased over composers careers not support it. The answer is not clear at this
in classical and popular music is relevant point, but the different conclusions might be
to Simontons (1999) influential Darwinian due to different data bases used by different
theory of creative thinking, in which the cre- investigators. For example, Simonton bases
ative process occurs in two stages, analogous some of his analyses on surveys of classical
to the stages in Darwins theory of organic works compiled 5 0 years ago, whereas Weis-
evolution through natural selection. In evo- berg and Studivant and Kozbelt use more
lution, the first stage is blind variation, as recent (and perhaps more thorough) tabu-
random changes occur in the genetic mate- lations of works. This might contribute to
rial from one generation to the next. Those the different conclusions. At the very least, it
variations result in organisms with differing seems that the equal-odds rule can be called
reproductive capabilities, which therefore into question (see Kozbelt, 2004, for further
will be differentially successful in passing discussion).
their genetic material to the next genera- Let us now turn to a different artistic
tion. Another way to put this is to say that domain the visual arts in order to examine
772 the cambridge handbook of expertise and expert performance
the generality of the conclusion that domain- manner can be seen in at least one other work
specific expertise is at the core of innovation. produced by Picasso in the mid-193 0s. So
Picasso built the structure of Guernica on the
Visual Arts foundation of his own earlier work, that is,
on his domain-specific expertise. That exper-
painting: picassos guernica tise also included knowledge of the work of
In MayJune 193 7, Picasso created what was other artists: a number of the specific char-
to become one of the best-known paintings acters in Guernica can be traced to works
of the 20th century: Guernica. The creation of others, including Goya, a Spaniard whose
of that masterwork was stimulated by the work was particularly important to Picasso.
bombing on April 27, 193 7 of the Basque (See Weisberg, 1999, 2004, 2006, for further
town of Guernica, in northern Spain. The discussion.)
bombing was carried out by the German air One can also find evidence for practice in
force, allies of Francos fascist forces in the Picassos career development, which reveals
Spanish Civil War. The town per se seemed a pattern similar to that seen in Mozart.
to have little strategic value (see Chipp, Picassos father was a painter, as well as a
1988, for discussion), and the Germans teacher of painting, so Picasso, like Mozart,
action was looked on by many as an act of was exposed from an early age to train-
terrorism. When news reports of the bomb- ing from a professional (Weisberg, 1999). In
ing began to reach Paris over the next few addition, Picasso attended art school, and
days, Picasso, a Spaniard who had been liv- some of his early works that have been pre-
ing in Paris for more than 3 0 years, dropped served show him practicing drawing eyes and
his on-going project, which was a painting facial profiles, as well as the human body
of an artist and model in the studio. That in difficult poses. This is concrete evidence
painting was to appear in the Spanish gov- of the young artist carrying out deliberate
ernments pavilion at an international expo- practice. In addition, the Ten-Year Rule also
sition (a worlds fair) to be held in Paris in applies to Picasso: the first works that show
June of 193 7. Over the next six weeks or so, a unique Picasso style did not occur until
he produced a new work, called Guernica, more than ten years into his career (Weis-
which was put on display instead. The Span- berg, 1999). This analysis of Picasso also calls
ish government was losing the Civil War, and into question the claims made by Sternberg
Picassos painting became a great antiwar (1996) concerning the extraordinary level of
and antifascist statement. Picassos early development. Again, it is not
For the student of creativity, Picassos absurd to say that the paintings produced by
working method for Guernica is particularly Picasso over the first ten years of his career
illuminating because he dated and num- are also matched by most painters as they
bered all the preliminary sketches some work their way through art school. Pariser
45 in all that he produced while working (1987), in an analysis of the juvenilia of sev-
out the details of the masterwork. I have eral painters known for precocity, including
analyzed the development of Guernica in Picasso, Klee, and Toulouse-Lautrec, con-
detail elsewhere (Weisberg, 2004), based on cluded that they all went through stages of
the sketches, and several conclusions are rel- development that were the same as those
evant to the present discussion. From the traversed by all painters.
beginning, as can be seen in the first sketch In sum, Picassos overall development
Picasso produced, on May 1, 193 7, he had accords with the Ten-Year Rule, and Guer-
the overall structure of the painting worked nica was based on his domain-specific
out: one can see the main characters in the expertise: he began with information from
same layout as they appear in the final paint- previous works, his own and those of others,
ing. This raises the question of where that and used that as the basis for the creation of
structure and those characters came from, a new work. The outline in Figure 42.1B can
and similar characters organized in a similar be applied to Picassos situation in creating
expertise in creative thinking 773
Guernica, and, to summarize Picassos maneuver. He had also earned money dur-
thought processes in producing that inno- ing the 1920s designing action toys, involv-
vation, we would put a notation close to ing movement, for American manufacturers.
the center of the diagram. As noted ear- Many of those toys can be seen in altered
lier, the adequacy of the precise structure form in the Circus, which became a hit in
of Figure 42.1B is not at issue here. However Parisian art circles.
that structure is depicted, it seems clear that Around 193 0, Calders work took a
Picassos creation of Guernica was another radical turn, becoming abstract or non-
example of domain-specific expertise serv- representational; that is, one could no longer
ing in creative thinking. We now turn to a see people or animals in the pieces he cre-
more seminal advance, from the domain of ated. This relatively sudden shift in style
sculpture. seems to have been triggered by Calders vis-
iting the studio of Piet Mondrian (Calder,
1966), another of the many young artists liv-
calders mobiles ing in Paris at that time, who had met Calder
Abstract wind-driven hanging wire sculp- though a visit to see Calders Circus. Mon-
ture the mobiles with which we all are now drian was a painter whose most well-known
so familiar were created in the early 193 0s work is completely non-representational,
by Alexander Calder (18981976), a young using grids made out of black lines on a
American artist living in Paris (Marter, 1991; white canvas, with some of the spaces in
Weisberg, 1993 , 2006); no one had ever seen the grid filled in with blocks of primary col-
anything like them before. Examination of ors (blue, yellow, red). When Calder saw
this case study allows us to consider the role Mondrians abstract works, he is said to
of expertise in what one could call a radical have remarked to Mondrian that the works
innovation, that is, one that seems to make should move. Soon thereafter, Calder began
a break with the past. to paint in an abstract style, similar to Mon-
When he created the first mobiles, Calder drians, but he quickly turned to wire sculp-
had been a sculptor for several years. He was ture, with which he was more comfort-
born into an artistic family, and he and his able. He produced several abstract works of
sister spent much time during their child- sculpture and soon added movement, usu-
hoods carrying out artistic and construction ally using electric motors. Motorized sculp-
projects of various sorts. In addition, Calder tures were difficult to keep working (the
was trained as a mechanical engineer, which mechanisms kept breaking), and, even when
gave him more formal exposure to mecha- they did work, the possible movements
nisms of various sorts, as well as developing were restricted, and soon became repetitious
further his construction skills. Calders early and boring. Calder then decided to struc-
sculptures, which usually represented peo- ture the sculptures so that they would be
ple or animals, were often constructed out of moved by the wind, a simpler and more
wire and involved movement. In the 1920s, reliable, as well as a less-predictable, source
Calder constructed a circus, with a cast of of movement, and so the first mobiles were
miniature performers made out of wire, bits created.
of wood and cork, and pieces of cloth. There In analyzing Calders creation of mobiles,
were three rings, in which an animal trainer we see further support for the role of dom-
and his wild charges, as well as trapeze ain-specific expertise in innovation. Many
artists, a sword swallower, and acrobats and of his early sculptures, including the Circus
clowns were put through their paces by and the action toys, were made out of wire
the artist. Calder developed ways of hav- and involved movement; also, some of his
ing the miniature people and animals move, early representational works were designed
so the trapeze artists, for example, would to swing in the air. Those aspects of his
swing on the trapeze and then leap from own work his domain-specific expertise
one trapeze to another in a death-defying served as the basis for mobiles. The switch
774 the cambridge handbook of expertise and expert performance
to produce dynamic lines of various textures. that can be traced to Watson and Cricks
He then used those lines to weave highly model of the structure of DNA, ranging
textured compositions of great dynamism, from new drugs to cloned organisms. For-
sometimes on a large scale, which many peo- mulating the double helix was a creative act
ple responded to with emotion. Thus, Pol- of the first order.
locks radical new technique seems to have Watson and Crick collaborated at the
been a direct development out of his expe- Cavendish Laboratory at Cambridge Univer-
rience. As a result of those works, Pollock sity. Watson, an American who had recently
became the leader of the group of artists that earned a Ph.D. in genetics, arrived in the
came to be known as the Abstract Expres- fall of 195 1. Crick, who had been trained
sionists or the New York School and who as a physicist but who had switched to
served to raise American art to the level of biology after World War II, was already at
the equal of Europe. the Cavendish, carrying out graduate-level
work. Soon after Watsons arrival, he and
expertise and creativity in the visual Crick realized that they were both inter-
arts: summary ested in solving the problem of the structure
In conclusion, we have seen that the devel- of DNA, and a close collaboration devel-
opment of innovation in the case studies oped. They decided early on that they would
in visual arts that we have examined is attempt to build a model of the molecule.
parallel to that in the music case studies: The general method of model building was
domain-specific expertise played a critical adopted from Linus Pauling, a world-famous
role in all of them. Thus, there is consistency chemist who had had great success with
across different domains in the arts. We now model building in his recent research.
turn to case studies in science and technol- Watson and Crick also adopted a more
ogy in order to examine further the finding specific strategy from Pauling, who had
that domain-specific expertise is of central recently published a structural model of the
importance in creative thinking. We have protein alpha-keratin, which makes up hair,
also not found any unequivocal examples horn, and fingernails, among other things.
of general expertise playing a role in real- Paulings model of alpha-keratin was in the
world innovation, comparable to that found form of a helix (the alpha-helix), and Wat-
in the laboratory problem-solving results son and Crick assumed, based on Paulings
(e.g., Fleck & Weisberg, 2004; Perkins, 1981; work, that DNA was also helical. This was
Weisberg & Suls, 1973 ). not an unreasonable assumption to make,
since DNA and alpha-keratin are analo-
Science and Technology gous in several ways: both are large organic
molecules, constructed out of smaller ele-
the double helix ments that repeat again and again, in differ-
Early in 195 3 , Watson and Crick published ent combinations. Proteins are constructed
the double-helix model of the structure of out of peptide units, and DNA is made out
DNA, the genetic material (the discussion of of nucleotides. Thus, Watson and Crick used
the double helix is based on Olby, 1994; Wat- information from a closely related area
son, 1968; Weisberg, 1993 , 2006). A number from their domain-specific expertise as
of research teams were at that time trying the foundation on which they constructed
to determine that structure, because it was their model.
believed correctly, as it turned out that Those two strategic assumptions made
understanding the structure of the genetic by Watson and Crick model building and
material would enable scientists to under- starting with helical structures led to sev-
stand how it replicated. It was assumed that eral advantages on their part. First, they
this knowledge would ultimately allow sci- began to examine all the available informa-
entists to control developmental processes, tion from the perspective of what each piece
and we have all seen the astounding advances could tell them about the helical structure
776 the cambridge handbook of expertise and expert performance
of DNA. That meant that they did not have the wright brothers invention
to spend time examining information that of the airplane
might have taken them offtrack. Since DNA The Wright brothers first successful pow-
turned out to be helical in shape, Watson and ered flights, on December 17, 1903 , at Kitty
Crick moved far along the correct path to Hawk, NC, came after several years of
the answer without having expended undue intense work (Weisberg, 2006). Wilbur and
amounts of time and effort. In addition, Orville Wrights interest in flying was kin-
Watson and Crick were in contact with Mau- dled (or rekindled, since they had had some
rice Wilkins, who was also working on the interest in flying machines earlier in their
structure of DNA. Wilkins told them that lives) by news accounts of the death of
he believed that DNA was helical in struc- Otto Lilienthal in August 1896, in a glid-
ture and that it was thicker than a single ing accident (Heppenheimer, 2003 ). Lilien-
strand. He also provided Watson and Crick thal, a German engineer, had for several
with some experimental results that sup- years been experimenting with gliders of
ported that view. Wilkins did not at that his own design as part of a project to pro-
time build models of possible structures of duce a powered flying machine. Lilienthals
DNA, as he was less committed to that gliders had wings shaped like those of bats,
strategy than were Watson and Crick. This and he flew by hanging suspended from the
lack of commitment to building models may wing. The gliders were controlled by Lilien-
have resulted in Wilkins being left behind by thal moving his body, thereby shifting the
Watson and Crick. center of gravity of the apparatus, to coun-
The creation of the double-helix model teract the lifting force of the wing. During
of DNA was obviously a much more com- one flight, a gust of wind brought up the
plex process than has been outlined here (for front of the wing of the glider, and the craft
further discussion, see Olby, 1994, Watson, stalled (it stopped moving, thereby losing
1968, and Weisberg, 2003 , 2006). Many lift, the capacity to stay aloft). Lilienthal was
other specific pieces of information had to unable to bring the glider under control by
be determined before a specific model could shifting his weight, and it crashed, break-
be constructed, such as how many strands ing his back. He died the next day. Lilien-
were in the molecule; how the strands were thals death was reported in newspapers and
structured; how far apart the strands were; magazines, and the Wrights read about it.
the angle, or pitch of the helical spiral; how It was not until 1899, however, that Wilbur
the strands of the helix were held together, Wright wrote to the Smithsonian Institution
and so forth. However, the answers to those to inquire about any available information
questions do not introduce any issues that recounting research on flight. He received a
will change the present conclusions, espe- list of the materials, including several books,
cially the principal one, that the double and he also received several pamphlets pub-
helix, a creative product of the first rank, was lished by the Smithsonian.
firmly built on the domain-specific expertise There were several research projects on
of Watson and Crick. flight beyond that of Lilienthal that were
In conclusion, we see here evidence for described in the materials the Wrights
the critical role of domain-specific exper- received (Weisberg, 2006). Octave Chanute,
tise in a seminal example of creative think- a retired engineer, was heading a team car-
ing in science. The example in Figure 42.1A rying out research using gliders, and several
can be used to outline the development of investigators had worked on powered flying
DNA just discussed. We see that all of what machines, including Samuel P. Langley, the
has just been summarized would fall near Secretary of the Smithsonian. The Wrights
the center of the diagram at the domain- thus acquired information as the result of
specific area. I now turn to case studies in their study of other inventors work, and this
invention to examine further the generality domain-specific expertise played a role in
of this finding. their own work. As an example, the biplane
expertise in creative thinking 777
(two-wing) configuration of the gliders with one steers the front wheel in the direction
which they began their work, and also of one wishes to go by moving the handlebars,
their powered Flyer, were similar to that of but one also leans to the side that one is turn-
Chanutes gliders. However, perhaps more ing to, so that the bicycle tilts (banks) to the
important for the Wrights, was what they inside of the turn. That is, one begins to fall
perceived of as missing from the work of when one is making a turn. The experienced
other would-be inventors of the airplane. rider keeps the bicycles speed high enough
On reading the accounts of those projects, so that it leans into the turn but does not
the Wrights were most struck by the fact fall; a novice rider when making a turn is
that none of those would-be inventors had likely to go too slowly and will have to put
attempted to tackle what to the Wrights was his or her inside foot on the ground to pre-
the most pressing problem in building a fly- vent a fall. When the turn is completed, the
ing machine: development of a system that rider reestablishes equilibrium by straight-
would enable the pilot to control the aircraft ening the front wheel and sitting straight on
in the air. the bicycle.
As an example of the lack of focus on The Wrights surmised that control of a
a control system, the steam-powered air- plane in flight might be like control of a mov-
planes called aerodromes that Langley ing bicycle. One might say that the Wrights
had under development had wings and a thought of the airplane as a bicycle with
tail designed to automatically keep them sta- wings (Heppenheimer, 2003 , p. 89). Thus,
ble in response to changes in wind velocity the Wrights belief in the need for a system
and direction (Heppenheimer, 2003 , p. 88). to enable a pilot to control an aircraft in flight
There were no controls to enable the pilot was the outgrowth of their expertise with
to actively control the craft. Chanutes glid- bicycles. Other researchers conceived of an
ers were constructed similarly. There was airplane as a boat in the air, which is con-
concern on the part of many of the early trolled very differently. Langley, for exam-
researchers that a pilot would not be able ple, designed his aerodromes with a rudder
to respond quickly to changes in wind direc- at the rear, like that of a boat, to control
tion and speed and thus would be useless in turns. It should be noted, however, that some
an emergency. The Wrights, in contrast, felt individuals who preceded the Wrights in
that the issue of control was so important speculating about the possibility of human-
that a method had to be devised so that a powered flight had also considered riding
human would be able to pilot the craft. bicycles as analogous to piloting an aircraft.
The Wrights belief in the necessity for James Means, a commentator on the flight
control, and in the ability of a human to carry scene, predicted (in a book that was on the
out that task, may have arisen from their Smithsonian list sent to Wilbur Wright and
experiences with bicycles (Heppenheimer, probably read by the brothers) that the air-
2003 , p. 88). The Wrights had built and sold plane would be perfected by bicycle men,
bicycles of their own design, so they were because to fly is like wheeling: To learn to
well versed in the specifics of bicycling. Bicy- wheel one must learn to balance. To learn
cles as vehicles are analogous to airplanes in to fly one must learn to balance (quoted
important ways, because both require rel- in Heppenheimer, 2003 , p. 88). Lilienthal
atively complex control on the part of the had written to Means in praise of Meanss
pilot. A person riding a bicycle makes con- analysis of the relationship between riding a
stant adjustments to speed, body position, bicycle and flying. If we analyze this aspect
and orientation of the front wheel (through of the Wrights thinking based on the out-
the handlebars) in order to maintain equilib- lines presented in Figure 42.1A and 42.1B,
rium and to proceed in the chosen direction. their use of the bicycle as the basis for con-
However, and this is most important, the ceiving of control in flight would be clas-
rider also at times deliberately upsets equi- sified as being based on relatively domain-
librium, most specifically, in order to turn: specific expertise since both are modes of
778 the cambridge handbook of expertise and expert performance
transportation, although one is land based able surfaces, analogous to the birds wing
and the other is not (Weisberg, 2006). tips, through metal rods and gears (Heppen-
The Wrights then relatively quickly heimer, 2003 ). The system was designed to
developed an idea for a control system allow the pilot to move the wing tips up
based on observations they had made of and down in opposite directions and to tilt
birds in flight, another example of domain- the machine when necessary, either to main-
specific expertise (Weisberg, 2006). Here tain equilibrium in the face of wind gusts,
too, there were precedents in the commu- or to disturb equilibrium intentionally in
nity of researchers with which the Wrights order to bank into a turn. This is an example
were familiar. L.-P. Mouillard had written a where they used their expertise as mechan-
book in which he discussed bird flight and ics general expertise to implement the
urged others to observe birds gliding effort- birds system in human materials. However,
lessly on air currents (Heppenheimer, 2003 ), the rod-and-gear system was too heavy to
and Lilienthal carried out observations of be practical. They then, again, used their
birds. In a magazine article on Lilienthal mechanical expertise to develop a system
that had appeared in the United States, the wherein the pilot, lying prone on the lower
author noted that Lilienthals observations wing of a two-winged glider (a biplane),
of birds in flight had led him to conclusions controlled the orientation of the wing tips,
about the optimal shape of the wings for his through wires that he could pull in one direc-
gliders. The Wrights had read this article and tion or another by swinging his hips in a cra-
had also read elsewhere about bird flight. dle to which the wires were attached. The
It is also possible that the Wrights and pilots movement caused the wires to pull
many others interested in the possibility one set of wing tips up and the other down,
of human flight were led to study birds which was called wing warping. An early ver-
because, if one wants to learn how to fly, one sion of a wing-warping system was devel-
should study the behavior of an organism oped by the Wrights at their home in Day-
that already knows how. That is, use of infor- ton, Ohio, during the summer of 1899. They
mation gleaned from birds would be another tested it on a five-foot wingspan biplane kite
example of relatively specific transfer (Weis- model that they built. The person flying the
berg, 2006), and indeed one could argue that kite was able to warp the wings by pulling
birds are closer to flying machines than on two sets of strings, one of which con-
bicycles are. The Wrights reported that they trolled each set of wing tips. They found that
had observed birds gliding on wind currents, the system worked as they hoped. When the
with their wings essentially motionless, in a pilots hip cradle was incorporated in their
dihedral or V shape. The animals could be first glider in 1900, it added little weight to
seen sometimes being tilted to one side or the machine and worked well enough that it
the other by changing winds and air currents was used to control all their gliders (1900
(the V would no longer be vertical), and 1902) and the first powered flying machine
somehow making adjustments that allowed (1903 ).
them to return to level flight. Close obser- We have seen what we can designate as
vation indicated that the birds responded to three stages in the Wrights development of
changes from level flight by altering the ori- a control system for their aircraft (1) decid-
entations of their wing tips. By moving the ing that there was a need for a control sys-
tips of their wings in opposite directions, tem, (2) using birds control of their wing
the birds essentially turned themselves into tips as an example of a control system, (3 )
windmills, and were turned by the wind back implementing the system were outgrowths
toward level flight. of different aspects of their expertise. The
The Wrights discovery of birds use of first two stages, which were dependent on
their wing tips to control their orienta- the bicycle and bird flight, respectively, were
tion led them to develop a mechanical sys- the results of domain-specific expertise. The
tem whereby the pilot could control mov- final stage, implementation of a method for
expertise in creative thinking 779
controlling the wing tips of their aircraft, pump. The vacuum pumps then available
seems to have been independent of domain- could not produce anything near a complete
specific expertise since they did not have vacuum, so the burner could not be pro-
any experience constructing a flight-control tected, and the bulbs quickly failed. Plat-
system analogous to that of birds. That is, inum burners presented a different problem:
the wing-warping system was based on the their temperature had to be controlled very
Wrights general expertise as mechanics and carefully because if the burner got too hot,
carpenters, which in turn was based on years it would melt and crack, thereby rendering
of construction projects, as well as their the bulb useless. As with the Wrights, Edi-
experiences as manufacturers of bicycles. son began his work relatively knowledgeable
In conclusion, the accomplishments of about what had been done before, that is,
the Wright brothers provide evidence that other researchers failures.
the two modes at the ends of the contin- Edison started his electric-light work in
uum of expertise as outlined earlier may play 1877 with a carbon burner in a vacuum.
roles in major creative accomplishments. We This work, built directly on the past, was
will see further evidence of creative thinking not successful: the burner oxidized. Since
based on both modes of expertise in the next he knew of no way to improve the vac-
case study, which examines a seminal inven- uum, Edison abandoned work on the car-
tion produced by one of the most prolific bon burner. About a year later, he carried
inventors who ever lived. out a second phase of work on the light bulb,
this time with platinum burners, again build-
edison and the light bulb ing directly on what had been done in the
On New Years Eve of 1879, Thomas Edi- past. In order to try to stop the platinum
son opened his Menlo Park, NJ, laboratory from melting, Edisons bulbs contained reg-
to the public so that they could see and mar- ulators, devices like thermostats in modern
vel at the electric lighting system that had heating systems, to regulate the temperature
been installed there. This demonstration cul- of the platinum and keep it from melting
minated several years work in Edisons labo- (Friedel & Israel, 1986). Edison had seen reg-
ratory. In Edisons light bulb, electric current ulators in electric-lighting circuits designed
was passed through a thin filament of carbon by others. However, it proved impossible
(the burner), which was enclosed inside a to control the temperature of the platinum
glass bulb, in a vacuum. The current flowing burner. Thus, one could summarize Edisons
through the carbon caused it to heat to the early work on the light bulb by saying that
point of glowing or incandescence. it was based on domain-specific expertise,
Edison is usually referred to as the inven- that is, relatively direct transfer of informa-
tor of the light bulb, but there had been tion from the same domain. Unfortunately,
numerous earlier attempts to produce an Edisons work also suffered the same fate as
incandescent electric light bulb, and he the earlier attempts on which it was based.
was aware of that work (Friedel & Israel, In response to the failure with platinum
1986; Weisberg, Buonanno, & Israel, 2006). burners, Edison tried to determine exactly
Almost all of those earlier attempts used why they failed. He observed the broken
either carbon or platinum as the burner, but burners under a microscope, and he and his
there were difficulties with each of those ele- staff concluded that the melting and crack-
ments. When carbon was heated to a temper- ing was caused by escaping hydrogen gas,
ature sufficient to produce light, it would which platinum under normal conditions
quickly oxidize (burn up), rendering the had absorbed from the atmosphere. The
bulb useless. In order to eliminate oxidation, hydrogen escaped when the platinum was
it was necessary to remove the carbon burner heated, causing holes to form, which facil-
from the presence of oxygen, and many of itated melting and cracking of the burner.
the earlier workers had placed the carbon Edison reasoned that the platinum might be
burner in a vacuum produced by a vacuum stopped from cracking if the hydrogen could
780 the cambridge handbook of expertise and expert performance
previous use of wire and motion in sculpture the double helix may indicate that domain-
were critical in his development of mobiles; specific expertise may not be sufficient for
(2) the nonrepresentational content of the creativity. Similarly, and perhaps potentially
new sculpture came from Mondrian. That is, more important in this context, Pauling,
a critical component of Calders innovation who was a world-renowned expert in anal-
came from others, as did Pollocks. Watson ysis of the structure of complex organic
and Cricks domain-specific expertise their macromolecules through modeling, also was
knowledge of Paulings work served as the not successful in determining the structure
foundation for their own work in two ways: of DNA. In a parallel with the discussion
construction of models and attempts to of Wilkins, Paulings failure to discover the
model helical structures. The Wrights used double helix can be interpreted as indicat-
information from several related domains ing that domain-specific expertise may not
bicycles; birds flight as the basis for their be sufficient for creative achievement.
work, perhaps because at that time there was However, there is a critical assumption
little in the way of useful information avail- underlying such a conclusion: one must
able in aviation. Finally, Edison built on ear- assume that the expertise of Wilkins and
lier work, but in a negative way: he began Pauling concerning DNA was equivalent to
within the framework of earlier attempts that of Watson and Crick. I have argued else-
(domain-specific expertise) and devised a where (Weisberg, 1993 ; 2006) that the rea-
way to overcome the problem of oxidation son that Watson and Crick were successful
of the carbon burners (based on general and when others, including Wilkins and Pauling,
domain-specific expertise). were not was because only Watson and Crick
Turning to the more specific question of possessed all the information necessary for
whether domain-specific or general exper- the construction of the double helix. So the
tise is necessary for creativity, it seems that reason that Wilkins and Pauling each failed
domain-specific expertise may be necessary was not because each of them lacked some
because in all the case studies considered sort of ingredient a creative spark per-
here the innovations depended at least in haps that played a critical role in their
part on domain-specific expertise. Based as creative thinking. Rather, the reason Wilkins
well on the present case studies, general and Pauling each failed was because each of
expertise is not always necessary for cre- them did not know what Watson and Crick
ativity, because in a majority of the cases did.
the innovations did not seem to depend on In conclusion, the present results paint
general expertise (see Table 42.1). It might a complicated answer to the question of
be the case that if an individual possesses whether domain-specific expertise is suffi-
detailed-enough domain-specific expertise, cient for creativity. Analysis of more case
general expertise may not be called on. studies may help to provide more illumina-
An important related question concerns tion on this question.
whether either domain-specific or general
expertise is sufficient for creativity. We saw
Creative Advances and the Extension
that the Wrights and Edison had to go
of Technique
beyond domain-specific expertise in order
to successfully complete their work, so The second hypothesis was that creative
domain-specific expertise was not sufficient advances come about through extensions of
for them. Two additional pieces of evi- technique, in an analogy to the achievement
dence that might be relevant here come of new performance standards by elite ath-
from the case of DNA. First, we saw that letes. This hypothesis is not strongly sup-
Wilkins provided Watson and Crick with ported by the case studies since most of the
some pieces of information that played a creative breakthroughs examined here, even
role in their creating their successful model. those that depended on domain-specific
The fact that Wilkins himself did not create expertise and deliberate practice, did not
expertise in creative thinking 783
require that the innovator go beyond exist- poured paintings and Calders nonrepresen-
ing technique. Pollocks development of his tational sculptures, for example, did not go
dripping or pouring technique was a creative beyond anything in a quantitative sense, they
advance that went beyond existing tech- were simply new and different than exist-
niques (Weisberg, 1993 , 2003 ), but is was ing forms. Creators artists, for example
the only example of such an advance in are usually not competing in a quantitative
the case studies that we examined. Calders sense, as athletes are (even though, as noted
nonrepresentational sculpture, for example, earlier, works of art sometimes receive pri-
was simpler than his representational work. zes). A better analogy might be to think of
Essentially the same technique was used artists as explorers, each of whom takes a
twisted pieces of wire but the content was different path through heretofore unknown
changed (and simplified). Similarly, Guer- territory. One innovation is thus not better
nica was not based on new technical devel- than another; it does not go further along
opments on Picassos part. the same path (as an athlete does when he or
Also, some radical artistic advances do not she jumps higher than anyone else, for exam-
involve any technique at all. As an exam- ple). Rather, the innovation takes a different
ple, consider the use of found objects in path, so a direct quantitative comparison is
painting and sculpture, which involves incor- not appropriate.
porating objects found in the environment It is also interesting to note that old
into a work of art or, even more radically, styles of art and music, and old ideas in sci-
presenting a found object by itself as a work ence and invention, sometimes come back
of art. The most striking example of such into favor, although perhaps not in iden-
a practice was Marcel Duchamps Bicycle tical form, which means that the old has
Wheel, which was a sculpture consisting not been surpassed, it has just been put
of a bicycle wheel and the fork in which it aside for a while until it becomes relevant
spins, mounted on a painted kitchen stool. again. Examples of such recyclings include
Similarly, Duchamps Fountain consisted of the development of the neoclassical style in
a urinal, which he had signed R. Mutt, sit- classical music of the 1980s, as well as the
ting on a base (not mounted on the wall). realistic-based styles of painting that devel-
The purpose behind Duchamps use of those oped after the ascension of the modern non-
objects is not of relevance here. The critical representational art of the 195 0s and 1960s.
point for us is that those works of art (and Also, Edisons return to carbon burners in
they are displayed in museums and are dis- his lightbulb is an example of a previously
cussed in texts and histories of art see, e.g., rejected idea coming back, although the
Arnason, 1986, p. 229 so they are works of time frame was considerably shorter than
art) require nothing in the way of technique that which usually occurs when old ideas are
acquired through practice. recycled. In conclusion, the analogy made
The important conclusion to arise from by Ericsson between expertise in perfor-
those examples is that creative advances are mance domains, such as athletics, and cre-
not analogous to a new performance stan- ative domains might not be that useful,
dard set by an elite athlete. When we say musical and artistic competitions notwith-
that an innovation goes beyond the borders of standing.
some domain, one may conclude that one is
talking about the innovation surpassing the
Do Creative Advances Based on
old, in a quantitative manner, along some
Expertise Redefine Domains?
dimension. That is not the way to analyze the
relation between the old and the new, how- The final hypothesis is that acquiring exper-
ever. An innovation is different than the past tise enables creative thinkers to redefine
but not necessarily better; this is in contrast their domains. It should first be noted in
to an ice-skaters five-revolution jump, say, response that not all creative advances rede-
which is better than one with four. Pollocks fine their domains. Most importantly, some
784 the cambridge handbook of expertise and expert performance
Table 42 .2 . Case studies: Relationship between expertise and redefinition of the domain
Did the achievement How was the achievement brought about?
redefine the domain? Domain-specific expertise (Including outside
information from same domain) General expertise
No Mozart Edisons analysis of
Picassos Guernica failures of platinum
DNA (Paulings modeling and alpha-helix)
Edisons Light Bulb (began with others
unsuccessful bulbs; combined vacuum pumps)
Yes Beatles Middle-Period Works Wright Brothers
Calders Mobiles (non-representational style from implementation of
Mondrian) specific wing-
Pollocks Poured Paintings (pouring, spilling, and warping control
throwing paint as mode of application; system
Siqueiros workshop)
Wright Brothers (bicycles for need for control
system; bird flight for example of a control
system)
implications for optimal learning and creativ- Olby, R. (1994). The path to the double helix. The
ity. High Ability Studies, 9, 75 100. discovery of DNA. New York: Dover.
Ericsson, K. A. (1999). Creative expertise as supe- Pariser, D. (1987). The juvenile drawings of
rior reproducible performance: Innovative and Klee, Toulouse-Lautrec, and Picasso. Visual
flexible aspects of expert performance. Psycho- Arts Research, 13 , 5 3 67.
logical Inquiry, 10, 3 293 3 3 . Perkins, D. N. (1981). The minds best work. Cam-
Everett, W. (2001). The Beatles as musicians. The bridge, MA: Havard.
Quarry Men through Rubber Soul. New York: Reeves, L. M., & Weisberg, R. W. (1994). Models
Oxford. of analogical transfer in problem solving. Psy-
Feltovich, P. J., Spiro, R. R., & Coulson, R. L. chological Bulletin, 116, 3 81400.
(1997). Issues of expert flexibility in contexts Reising, R. (2002). Every sound there is. The Beat-
characterized by complexity and change. In P. J. les Revolver and the transformation of rock and
Feltovich, K. M. Ford, & R. R. Hoffman (Eds.), roll. Burlington, VT: Ashgate.
Expertise in context (pp 125 146). Menlo Park, Scheerer, M. (1963 ). On problem-solving. Scien-
CA: AAAI/MIT Press. tific American, 2 08, 118128.
Fleck, J. I., & Weisberg, R. W. (2004). The use of Shiffrin, R. M. (1996). Laboratory experimenta-
verbal protocols as data: An analysis of insight tion on the genesis of expertise. In K. A. Eric-
in the candle problem. Memory & Cognition, sson (Ed.), The road to excellence. The acquisi-
3 2 , 9901006. tion of expert performance in the arts and sciences,
Friedel, R., & Israel, P. (1986). Edisons elec- sports, and games (pp. 3 3 73 46). Mahwah, NJ:
tric light. Biography of an invention. New Erlbaum.
Brunswick, NJ: Rutgers University Press. Simonton, D. K. (1999). Origins of genius. Dar-
Gardner, H. (1993 ). Creating minds. An anatomy winian perspectives on creativity. New York:
of creativity seen through the lives of Freud, Ein- Oxford.
stein, Picasso, Stravinsky, Eliot, Graham, and Sloboda, J. (1996). The acquisition of musi-
Gandhi. New York: Basic. cal performance expertise: Deconstructing the
Hayes, J. R. (1989). Cognitive processes in cre- talent account of individual differences in
ativity. In J. A. Glover, R. R. Ronning, & musical expressivity. In K. A. Ericsson (Ed.),
C. R. Reynolds (Eds.), Handbook of creativity The road to excellence. The acquisition of
(pp. 13 5 145 ). New York: Plenum. expert performance in the arts and sciences,
Heppenheimer, T. A. (2003 ). First flight. The sports, and games (pp. 107126). Mahwah, NJ:
Wright brothers and the invention of the airplane. Erlbaum.
Hoboken, NJ: Wiley. Solomon, M. (1977). Beethoven. New York:
James, W. (1880). Great men, great thoughts, and Schirmer.
the environment. Atlantic Monthly, 46, 441 Sternberg, R. J. (1996). Costs of expertise. In K. A.
45 9. Ericsson (Ed.), The road to excellence. The acqui-
Kozbelt, A. (2004). Reexamining the equal odds sition of expert performance in the arts and sci-
rule in classical composers. In J. P. Frois, P. ences, sports, and games (pp. 3 473 5 4). Mah-
Andrade, & J. F. Marques (Eds.), Art and sci- wah, NJ: Erlbaum.
ence. Proceedings of the XVIII Congress of the Sternberg, R. (Ed.) (1999). Handbook of creativity.
International Association of Empirical Esthetics, New York: Cambridge.
5 405 43 . Lisbon, Portugal: IAEA. Watson, J. D. (1968). The double helix: A personal
Landau, E. (1989). Jackson Pollock. New York: account of the discovery of the structure of DNA.
Abrams. New York: New American Library.
Luchins, A. S., & Luchins, E. H. (195 9). Rigidity Weisberg, R. W. (1980). Memory, thought, and
of behavior. Eugene, OR: University of Oregon behavior. New York: Oxford.
Press. Weisberg, R. W. (1993 ). Creativity: Beyond the
Marter, J. (1991). Alexander Calder. Cambridge, myth of genius. New York: Freeman.
MA: MIT Press. Weisberg, R. W. (1999). Creativity and
Newell, A. (1973 ). Artificial intelligence and the knowledge: A challenge to theories. In
concept of mind. In R. C. Shank & K. M. R. J. Sternberg (Ed.), Handbook of creativ-
Colby (Eds.), Computer models of language and ity (pp. 22625 0). New York: Cambridge
thought. San Francisco: W. H. Freeman. University Press.
expertise in creative thinking 787
789
790 author index
Bedard, J., 4, 17, 23 , 27, 686, 700 Bilodeau, E. A., 265 , 283
Bee, B., 495 , 5 01 Bilodeau, I. M., 265 , 283
Beek, P. J., 472, 476, 477, 480, 484, 485 , 486, 5 14, Binder, C., 80, 84
5 16, 5 20 Binet, A., 163 , 223 , 225 , 23 6, 23 8, 5 23 , 5 26, 5 3 0, 5 3 1,
Behrmann, M., 668, 669, 677, 681 5 3 4, 5 40, 5 5 0, 5 5 4, 5 65
Beier, M. B., 160, 161, 163 Binier, B. L., 5 08, 5 17
Beier, M. E., 12, 3 4, 3 7, 49 Binks, M. B., 5 3 , 64
Beilock, S. L., 15 , 3 61, 3 69, 475 , 479, 484, 5 13 , 5 16 Birdwhistell, R., 13 0, 142
Beinlich, I. A., 89, 102 Birnbaum, L., 222
Belkin, A., 5 79, 5 83 Birren, J. E., 5 94, 607
Bell, B. S., 440, 446, 45 0, 45 1 Bisseret, A., 3 69
Bell, J. A., 710, 722 Bizzi, E., 5 07, 5 17
Bell, J. F., III, 13 4, 144 Bjork, R. A., 5 06, 5 18, 5 19
Bellenkes, A. H., 249, 25 9, 3 62, 3 69 Blair, V., 444, 45 3
Bellows, N., 628, 63 1 Blaiwes, A. S., 441, 45 2
Belohoubek, P., 448, 45 0 Blakemore, S. J., 5 11, 5 16
Ben-Bashat, D., 668, 669, 680 Blehar, M. C., 606
Benbow, C. P., 3 4, 3 6, 3 7, 5 63 , 5 64, 5 65 , 5 67 Blendell, C., 411, 415 , 417
Bender, W. W., 3 45 , 3 5 3 Bleske-Rechek, A., 3 6, 3 7
Bendix, R., 120 Blickensderfer, E. L., 443 , 45 3
Beneke, W. M., 711, 719 Bliese, P. D., 448, 449
Bengtsson, S. L., 674, 677, 696, 700 Bloch, S., 495 , 5 02
Benke, T., 5 60, 5 66 Block, R. A., 3 05 , 3 16
Benner, P. E., 12, 17 Blomberg, J., 142
Bennett, J. S., 97, 100 Bloom, B. S., 3 , 12, 13 , 15 , 17, 46, 62, 79, 84, 287, 288,
Bennett, S. J., 246, 247, 262, 476, 477, 487 289, 3 00, 3 05 , 3 16, 462, 467, 613 , 629, 691, 700,
Ben-Shoham, I., 3 5 6, 3 5 7, 3 60, 3 71 706, 707, 709, 711, 719
Benson, R. R., 668, 680 Bloom, P., 5 5 5 , 5 68
Berardi-Coletta, B., 23 0, 23 8 Bloomfield, J., 9, 17
Bereiter, C., 82, 86, 297, 3 00, 3 91, 402 Blum, B. I., 13 1, 142
Bereiter, S., 400 Blythe, T., 626627, 63 2
Berg, C. A., 88, 101 Bock, J. K., 400
Bergen, P. C., 3 47, 3 5 3 Boecker, H.-D., 3 84, 3 86
Berger, R. C., 5 27, 5 3 0, 5 3 5 , 600, 607 Boekaerts, M., 705 , 713 , 719
Beringer, D. B., 3 5 6, 3 5 7, 3 67, 3 68, 3 70 Bogler, R., 618, 63 1
Berliner, D. C., 173 , 183 Bogot, Y., 3 5 0
Berliner, H., 5 25 , 5 3 4 Bohmer, R. M., 444, 446, 448, 45 0
Berliner, P., 45 8, 467 Boice, R., 400
Berlyne, D. E., 44, 62 Boldrini, M., 444, 45 0
Bernard, H. R., 129, 142 Bolger, D. J., 670, 677
Bernasconi, P., 480, 484 Bolger, F., 13 , 17
Berners-Lee, 99, 101 Bolstad, C. A., 63 9, 641, 642, 644, 65 0
Berninger, V. W., 3 99 Bonaceto, C., 213 , 218
Bernoulli, D., 441, 449 Bond, N. A., Jr., 15 8, 164
Bernstein, M., 3 23 , 3 3 3 Book, W. F., 685 , 700, 727, 73 7
Bernstein, N. A., 479, 484, 5 16 Boose, J. H., 97, 101, 204, 219, 405 , 415
Berry, C., 3 23 , 3 3 2 Bootsma, R. J., 480, 484
Berryman, R. G., 23 5 , 240 Bor, D., 616, 63 0
Berstein, L. M., 26, 27 Bordage, G., 3 46, 3 5 0, 3 5 2
Berti, S., 465 , 470 Borgeaud, P., 245 , 25 9, 478, 484
Bertini, G., 269, 284 Boring, E. B., 223 , 23 8
Bertrand, L., 247, 248, 262 Boring, E. G., 76, 84
Besson, M., 463 , 467 Borman, W. C., 45 1
Bettinardi, V., 672, 681 Bornschier, V., 120
Bettman, J. R., 425 , 43 7 Borrojerdi, B., 671, 681
Beurskens, A. J. H. M., 729, 73 7, 741 Borron, J., 99, 101
Bevan, A., 5 5 5 , 5 5 6, 5 67 Borstein, B. H., 405 , 415
Bevans, G. E., 3 04, 3 16 Boshuizen, H. P. A., 25 , 26, 28, 29, 23 5 , 23 8, 241, 3 43 ,
Beyer, H., 129, 142 3 49, 3 5 0, 3 5 1, 3 5 2, 3 5 3 , 463 , 467, 494, 5 03 , 5 99,
Beyerlein, M. M., 45 1 610
Beyerstein, B. L., 65 7, 677 Bosman, E. A., 5 98, 602, 607, 727, 728, 73 1, 73 7, 73 8
Bhalla, M., 5 14, 5 16, 5 19 Botwinick, J., 5 95 , 607
Bherer, L., 65 7, 664, 665 , 666, 678 Bouffard, V., 499, 5 01
Biddle, S., 716, 720 Bourdieu, P., 118, 120, 75 7, 75 9
Biederman, I., 268, 269, 283 , 5 45 , 5 5 0 Bourdin, B., 400
Bieman, J. M., 3 74, 3 83 , 3 87 Bourgouin, P., 664, 676, 680
Bienias, J. L., 496, 5 03 Bourne, L. E., Jr., 276, 279, 281, 283 , 284, 285
Billroth, T., 45 7, 467 Boutilier, C., 498, 5 03
792 author index
Bowden, K., 648, 65 1 Buchanan, B. G., 12, 14, 43 , 48, 61, 62, 87, 88, 90, 91,
Bowen, K. R., 160, 161, 163 92, 95 , 96, 97, 98, 99, 100, 101, 102, 103 , 13 0, 13 5 ,
Bower, G. H., 265 , 283 , 5 49, 5 5 2, 5 96, 607 142, 204, 219
Bower, J. M., 5 08, 5 17 Buchner-Jeziorska, A., 107, 120, 121
Bowerman, W. G., 3 26, 3 3 2 Buhler, K., 225 , 228, 23 8
Bowers, C. A., 215 , 219, 244, 248, 25 3 , 25 8, 261 Bukstel, L., 5 0, 63
Bowlby, J., 5 92, 607 Bullemer, P., 274, 275 , 286, 3 5 1, 5 12, 5 19
Boyatzis, R. E., 15 7, 164 Bullis, R. C., 618, 622, 63 0
Boyes-Braem, P., 176, 179, 183 Bunderson, J. S., 446, 45 0
Boyle, J. D., 45 7, 467 Bundy, D. A., 621, 63 2
Bracke-Tolkmitt, R., 5 08, 5 17 Bunge, S. A., 664, 665 , 677
Bradburn, N. M., 23 7, 241 Buonomano, D. V., 65 7, 677
Bradshaw, G. L., 25 3 , 25 8, 262 Burchell, G., 111, 121
Brady, T. J., 668, 680 Burgess, N., 5 48, 5 5 1, 5 92, 673 , 674, 675 , 679, 680
Brainin, E., 448, 45 3 Burgess, T., 624, 63 0
Brainthwaite, A., 401 Burgess-Limerick, R., 483
Bramer, M., 218, 405 , 415 Burian, K. V., 497, 5 01
Bramwell, B. S., 3 21, 3 27, 3 3 2 Burke, C. S., 15 , 43 9, 440, 441, 442, 443 , 444, 448,
Brand, A. G., 400 45 0, 45 1, 45 2, 45 3
Brand, C. R., 3 2, 3 7 Burns, B. D., 3 5 0
Brandfonbrener, A., 465 , 467 Burns, C. M., 209, 210, 211, 218
Bransford, J., 626, 629 Burns, K., 213 , 218
Branson, R. K., 5 , 14, 41, 46, 69, 76, 77, 84 Burton, A. M., 97, 102, 170, 176, 180, 182, 183 , 198,
Brashers-Krug, T., 5 07, 5 17 200, 206, 215 , 220, 222, 73 6, 745 , 75 9
Brasil-Neto, J. P., 671, 674, 681 Burwitz, L., 245 , 246, 25 6, 262, 475 , 477, 478, 487
Brauer, J., 3 49, 3 5 3 Butterworth, B., 16, 5 9, 60, 23 5 , 5 5 3 , 5 5 5 , 5 5 6, 5 5 7,
Bray, D. W., 3 3 , 3 7 5 5 8, 5 5 9, 5 64, 5 65 5 66, 5 67, 675 , 693
Bray, S. R., 448, 449 Button, G., 128, 129, 13 1, 13 3 , 13 4, 13 5 , 13 8, 142, 143
Breen, T. J., 180, 184, 3 5 6, 3 5 7, 3 65 , 3 70 Buyer, L. S., 23 0, 23 8
Brehe, S., 400 Byard, L., 429, 43 7
Brehmer, B., 243 , 25 9, 45 0, 45 1, 627, 629 Bynner, J., 5 5 3 , 5 66
Breslauer, G. W., 5 80, 5 82 Byrd, M., 5 93 , 608
Brezovic, C. P., 408, 415 , 417
Brialovsky, C., 3 5 3 Cabeza, R., 661, 662, 664, 677
Bridwell-Bowles, L., 400 Cacciabue, P. C., 188, 200, 205 , 220
Briggs, G. E., 278, 283 Cadopi, M., 499, 5 02
Briggs, L. J., 78, 85 Caicco, M., 498, 5 03
Britt, A., 5 72, 5 83 Calder, A., 773 , 785
Britton, B. K., 400, 401 Calderwood, R., 171, 182, 192, 200, 206, 209, 221, 403 ,
Broadbent, D. E., 5 17, 5 96, 607 404, 406, 407, 408, 410, 415 , 416, 417, 43 6, 43 7,
Broadway, K. P., 5 94, 5 95 , 5 96, 610 45 1, 5 29, 5 3 5 , 63 9, 65 0
Brochet, F., 268, 283 Calkins, V., 3 48, 3 5 2
Brockett, O. G., 489, 5 01 Callahan, J. S., 73 4, 73 7
Brody, G. H., 706, 719 Calvin, S., 5 16, 5 17
Brooks, D., 5 08, 5 18 Calvo-Merino, B., 672, 677
Brooks, L. R., 15 , 47, 5 5 , 23 5 , 240, 25 0, 3 42, 3 46, Camerer, C. F., 13 , 17, 43 3 , 43 6, 686, 700
3 5 0, 3 5 1, 3 5 2, 3 5 3 Caminiti, M. F., 3 48, 3 5 3
Brou, R. J., 249, 260 Cammarota, A., 671, 674, 681
Brown, A., 5 8, 62 Camp, C. J., 5 49, 5 5 0
Brown, E. S., 43 3 , 43 8 Campbell, D. J., 3 74, 3 75 , 3 77, 3 81, 3 84
Brown, I. D., 3 5 6, 3 5 7, 3 60, 3 69 Campbell, D. T., 760
Brown, J. S., 46, 48, 67, 623 , 629 Campbell, J. P., 443 , 45 0
Brown, M., 3 5 0 Campbell, J. I .D., 280, 283 , 5 60, 5 66
Brown, N. R., 3 74, 3 75 , 3 77, 3 84 Campillo, M., 708, 724
Brown, R., 5 8, 62 Campion, M. A., 187, 200, 3 84, 3 87, 448, 45 0
Brown, S., 267, 284 Campitelli, G., 174, 182, 5 3 1, 5 3 2, 5 3 3 , 5 3 5 , 5 3 6
Bruer, J., 117, 123 Canas, A. J., 212, 213 , 218, 219
Bruhn, E., 498, 5 01 Canavan, A. G. M., 5 08, 5 17
Brun, W., 445 , 448, 449 Candia, V., 466, 468
Bruner, J. S., 44, 62, 191 Candolle, A. de, 3 21, 3 26, 3 27, 3 28, 3 3 2
Brunk, C. A., 97, 102 Cannon, J. R., 3 5 6, 3 5 7, 3 61, 3 67, 3 68, 3 70
Brunswik, E., 15 7, 163 Cannon, M. D., 446, 448, 45 0
Bryan, W. L., 11, 12, 17, 225 , 23 8, 266, 267, 282, 283 , Cannon-Bowers, J. A., 43 9, 440, 441, 443 , 45 0, 45 3
474, 484, 5 09, 5 10, 5 17, 685 , 689, 700 Cantor, N. F., 73 , 84
Bryant, D., 172, 183 , 211, 218, 5 98, 5 99, 610, 728, 740 Caplan, J., 3 80, 3 85
Bryant, W. K., 3 04, 3 16 Caplan, R. A., 425 , 43 6
Bryman, A., 176, 182, 205 , 219 Cappa, S. F., 667, 668, 679
author index 793
Cappelletti, M., 5 5 7, 5 66 607, 608, 65 7, 685 , 693 , 696, 697, 699, 700, 723 ,
Caramazza, A., 400, 5 60, 5 67, 670, 681 726, 727, 728, 73 0, 73 4, 73 7, 73 8, 740
Carbotte, R., 3 49, 3 5 3 Chase, W. G., 3 , 11, 12, 17, 19, 27, 44, 46, 49, 5 0, 5 2,
Card, S. K., 188, 191, 199 5 7, 5 8, 60, 61, 63 , 67, 96, 100, 103 , 169, 171, 172,
Carello, C., 5 14, 5 17 173 , 178, 182, 207, 218, 23 5 , 23 6, 23 7, 23 8, 23 9,
Caretta, T. R., 617, 63 1 244, 245 , 25 9, 292, 297, 3 01, 3 05 , 3 16, 3 18, 3 5 3 ,
Carey, G., 725 , 73 8 3 69, 402, 43 1, 43 6, 474, 478, 484, 493 , 5 01, 5 10,
Carey, L., 401 5 17, 5 23 , 5 27, 5 3 1, 5 3 5 , 5 41, 5 42, 5 47, 5 5 0, 5 69,
Carey, S., 676, 678 5 82, 5 83 , 601, 611, 613 , 614, 629, 685 , 689, 696,
Carff, R., 212, 218 700, 703 , 727, 73 8, 768, 785
Carlsen, J. C., 463 , 469 Chassin, M. R., 3 49, 3 5 1
Carlson, B., 289, 3 00 Chein, J. M., 269, 285 , 65 3 , 65 6, 65 8, 65 9, 660, 661,
Carlson, R. A., 281, 286, 5 06, 5 19 665 , 678, 682
Carlton, E., 117, 121 Chen, 476
Carnahan, H., 471, 484 Chen, C. H., 496, 5 01
Carnot, M. J., 217, 220 Chen, D., 25 6, 261
Carpenter, P. A., 662, 664, 678, 680 Chen, E., 674, 682
Carr, T. H., 3 61, 3 69, 475 , 479, 484, 5 08, 5 13 , 5 16, Chen, Z., 5 0, 63
5 17 Cheney, F. W., 425 , 43 6
Carraher, D. W., 26, 29 Cheney, G., 498, 5 01
Carreras, A., 464, 468 Cheng, P. C. H., 64, 5 27, 5 3 6
Carretero, M., 5 75 , 5 76, 5 77, 5 80, 5 82, 5 84 Chenowith, N. A., 400
Carroll, J. B., 3 2, 3 7, 78, 79, 84, 5 44, 5 5 0, 5 89, 5 90, Chevalier, A., 3 76, 3 84
5 91, 5 99, 607 Chi, M. T. H., 3 , 5 , 12, 14, 15 , 17, 18, 21, 22, 23 , 24, 25 ,
Carroll, J. M., 3 76, 3 86 27, 28, 3 1, 3 7, 44, 46, 47, 48, 49, 5 0, 5 1, 5 2, 5 4, 5 5 ,
Carroll, J. S., 444, 45 0 63 , 64, 83 , 84, 95 , 100, 101, 13 0, 13 1, 142, 163 , 167,
Carron, A. V., 448, 449 169, 170, 172, 174, 175 , 176, 177, 178, 179, 180, 181,
Carr-Saunders, A. M., 107, 121 182, 204, 205 , 219, 228, 23 0, 23 8, 244, 25 9, 287,
Carter, F. J., 3 48, 3 5 1 3 01, 3 05 , 3 16, 3 49, 3 5 1, 3 69, 3 76, 3 84, 406, 412,
Carter, I. D., 648, 65 0 415 , 416, 43 6, 440, 45 0, 5 3 2, 5 3 5 , 5 69, 5 83 , 5 98,
Carter, M., 400 65 3 , 686, 700, 744
Carvajal, R., 212, 218 Chiang, W. C., 5 5 5 , 5 68
Cascio, W. F., 726, 740 Chiao, J. Y., 668, 679
Caspi, R., 3 5 0 Chidester, T. R., 446, 448, 45 0
Cass, J., 497, 5 01 Chien, J. M., 5 12, 5 19
Cassandro, V. J., 3 23 , 3 3 2 Chiesi, H. L., 48, 5 1, 5 5 , 63 , 67, 179, 182, 471, 484
Castejon, J. L., 618, 63 1 Chignell, M. H., 25 3 , 260
Castellan, N. J. J., 45 0 Chipman, S. F., 185 , 192, 199, 200, 201
Castka, P., 448, 45 0 Chipp, H. B., 772, 785
Castro, C. A., 448, 449 Chiu, M.-H., 23 0, 23 8
Catchpole, L. J., 411, 415 , 417 Chivers, P., 25 6, 260
Cattell, J. M., 3 05 , 3 16, 3 21, 3 23 , 3 3 2 Cho, K., 26, 27
Cattell, R. B., 3 2, 3 7, 5 92, 5 94, 5 95 , 5 96, 5 99, 607, Chomsky, N., 43 , 63
609, 617, 63 0, 724, 73 7 Choudhry, N. K., 3 49, 3 5 0
Caulford, P. G., 3 49, 3 5 0, 3 5 2 Chow, R., 209, 219
Cauraugh, J. H., 25 6, 261, 476, 477, 484 Christal, R. E., 3 2, 3 7
Cauzinille-Marmeche, E., 5 3 2, 5 3 5 Christensen, C., 26, 27
Cavanaugh, J. C., 5 93 , 607 Christensen, H., 5 94, 607
Cecil, J. S., 75 5 , 75 9 Christensen, P. R., 15 8, 164
Cellier, J. M., 3 69 Christiaen, J., 5 3 3 , 5 3 5
Cerella, J., 726, 73 7 Christoffersen, K., 211, 222
Chabris, C. F., 23 3 , 23 8, 5 29, 5 3 1, 5 3 3 , 5 3 5 Chulef, A. S., 5 98, 5 99, 609
Chaffin, R., 23 7, 23 8, 461, 463 , 467, 698, 700 Chun, M. M., 667, 680
Chalmers, B., 211, 218 Cianciolo, A. T., 12, 16, 3 2, 3 7, 91, 15 1, 163 , 613 , 621,
Chang, A., 3 5 2 624, 625 , 626, 629, 727
Chang, R. W., 3 5 0 Cipolotti, L., 5 5 5 , 5 5 9, 5 60, 5 63 , 5 66, 5 68
Chapin, R. S., 3 05 , 3 16 Clancey, W. J., 12, 15 , 45 , 46, 63 , 95 , 98, 99, 101, 103 ,
Chapman, C., 25 7, 262 116, 127, 13 5 , 142, 143 , 144, 206, 208, 219, 243 , 745 ,
Chapman, G. B., 405 , 415 760
Chapman, P. R., 3 5 6, 3 5 7, 3 62, 3 63 , 3 64, 3 69, 3 71, Clark, J., 673 , 681
648, 65 1 Clark, R. D., 3 23 , 3 3 2
Charness, N., 11, 16, 25 , 3 4, 3 7, 44, 49, 5 0, 5 2, 5 4, 60, Clark, V. P., 668, 679
63 , 101, 168, 23 3 , 23 4, 23 5 , 23 8, 244, 25 9, 297, 3 00, Clarkson, G. P., 23 6, 23 8, 5 26, 5 3 6
3 06, 3 16, 3 27, 3 28, 3 3 2, 3 48, 412, 416, 462, 463 , Clarkson-Smith, L., 73 6, 73 8
467, 478, 484, 5 23 , 5 24, 5 25 , 5 26, 5 27, 5 28, 5 29, Clawson, D. M., 279, 283
5 3 0, 5 3 2, 5 3 3 , 5 3 4, 5 3 5 , 5 3 7, 5 3 8, 5 5 3 , 5 60, 5 62, Clayton, J. E., 97, 102
5 645 65 , 5 66, 5 88, 5 93 , 5 98, 5 99, 601, 602, 606, Cleary, T. J., 708, 709, 712, 713 , 715 , 716, 718, 719
794 author index
Cleeremans, A., 274, 283 Coulson, R. L., 46, 5 6, 64, 83 , 86, 249, 260, 3 5 1, 3 85 ,
Clegg, B. A., 273 , 283 675 , 767, 786
Cleveland, A. A., 5 23 , 5 3 5 Courchesne, E., 5 08, 5 17
Clifford, M., 716, 719 Couture, B., 3 90, 400
Clifton, J., 463 , 467 Cowan, N., 5 0, 5 9, 63
Cline, J., 3 5 3 Cowan, T., 5 40, 5 42, 5 46, 5 5 2
Clinton-Cirocco, A., 406, 407, 410, 417, 63 9, 65 0 Cowley, M., 400, 699, 700
Cobley, S., 23 7, 23 8, 3 05 , 3 07, 3 16, 601, 608, 698, 701 Cox, C., 3 21, 3 22, 3 23 , 3 26, 3 27, 3 3 2
Cockcroft, W. H., 5 5 3 , 5 66 Coyle, T., 5 16, 5 17
Coderre, S., 3 5 0 Craft, J. L., 272, 286
Coffey, J. W., 13 1, 143 , 178, 183 , 207, 208, 211, 212, 213 , Crager, J., 624, 625 , 629
215 , 216, 217, 218, 219, 220 Craig, J. E., 75 , 84
Cohen, A., 275 , 283 , 5 13 , 5 17 Craik, F. I. M., 3 85 , 5 93 , 607, 608
Cohen, L., 5 5 9, 5 63 , 5 66, 675 , 678 Craine, D., 498, 5 01
Cohen, L. G., 670, 671, 674, 681 Cranberg, L., 5 3 3 , 5 3 5
Cohen, M. S., 404, 405 , 406, 416, 445 , 45 0 Crandall, B. W., 170, 171, 182, 183 , 192, 200, 209, 212,
Cohen, N. J., 73 5 , 73 9 219, 220, 406, 407, 408, 415 , 416, 417, 5 29, 5 3 5
Cohen, R., 5 01 Cratty, 473 , 484
Cohen, R. G., 16, 47, 5 05 , 5 07, 5 09, 5 17, 666 Crawford, M., 461, 467
Cohen, R. L., 489, 496, 5 01 Crawford, S. W., 25 4, 25 9
Colcombe, S. J., 65 7, 664, 665 , 666, 678 Cresswell, A. B., 3 48, 3 5 1
Cole, J., 117, 123 Crick, J. L., 3 63 , 3 70, 648, 65 0
Cole, R., 74, 84 Crisp, F., 498, 499, 5 00, 5 03
Coley, J. D., 175 , 180, 183 , 184, 5 99, 610 Crivello, F., 5 5 4, 5 63 , 5 64, 5 67, 5 68, 675 , 681
Collani, 3 77, 3 78, 3 84 Croghan, J. W., 204, 222
Collard, R., 5 10, 5 19 Croker, S., 64, 5 27, 5 3 6
Colley, A., 462, 467 Crompton, R., 106, 108, 111, 121
Collins, B. P., 46, 67 Cronbach, L. J., 163
Colonia-Willner, R., 5 98, 5 99, 607, 621, 622, 629, Cronon, W., 5 75 , 5 83
725 , 728, 73 8 Crook, J. A., 5 98, 5 99, 608
Colt, H. G., 25 4, 25 9 Cross, F. L., 72, 73 , 74, 84
Combs, D. M., 102 Cross, T., 73 4, 73 7
Compton, P., 97, 102 Crossman, E. R. F. W., 5 17
Conditt, M. A., 5 12, 5 17 Crovitz, H. F., 224, 23 8
Connally, T., 422, 43 7 Crowder, R. G., 205 , 218, 265 , 283
Connell, K. J., 3 5 0 Crowley, R. S., 23 4, 23 8
Connolly, T., 403 , 417 Crozier, M., 75 4, 75 9
Connor, C. E., 669, 678 Crundall, D., 3 62, 3 64, 3 69, 3 71, 648, 65 1
Consolini, P. M., 448, 45 1 Crutcher, R. J., 224, 23 8, 23 9, 462, 468
Constable, R. T., 664, 665 , 670, 677, 682 Csikszentmihalyi, M., 291, 299, 3 00, 400, 45 8, 468,
Contreni, J. J., 72, 73 , 74, 75 , 84 719, 766, 767, 785
Converse, S., 440, 441, 443 , 45 0, 45 3 Cullen, J., 176, 182, 205 , 219, 473 , 475 , 486
Cook, C. R., 3 79, 3 86 Cunningham, A. E., 402
Cook, E. F., 43 4, 43 7 Curatola, L., 5 3 3 , 5 3 7
Cook, S., 5 44, 5 5 2 Curley, S. P., 425 , 43 8
Cooke, J., 72, 86 Curnow, C., 3 19, 3 22, 3 23 , 3 29, 3 3 0, 3 3 3 , 689, 690,
Cooke, N. J., 176, 182, 191, 192, 200, 443 , 446, 45 0, 703 , 73 5 , 741
5 27, 5 3 0, 5 3 5 , 600, 607 Curran, T., 400
Cooke, N. M., 180, 184, 215 , 219, 3 5 6, 3 5 7, 3 65 , 3 70 Currie, L., 648, 65 0
Coombs, C. H., 405 , 416 Curtis, B., 3 74, 3 80, 3 82, 3 85
Coon, H., 725 , 73 8 Cuschieri, A., 3 48, 3 5 1
Cooper, D., 106, 109, 121 Cushing, K. S., 173 , 183
Cooper, D. E., 71, 84 Cusimano, M. D., 3 48, 3 5 0, 3 5 2, 3 5 3
Cooper, E. E., 5 45 , 5 5 0 Custers, E. J., 3 5 1
Cooper, R. G., Jr., 5 5 5 , 5 68
Cooper, W. E., 5 09, 5 10, 5 17 Dabringhaus, A., 5 65
Copeland, D. E., 5 93 , 610 Daffertshofer, A., 472, 477, 480, 484, 485 , 486, 5 16
Corbet, J. M., 13 1, 143 Dagenbach, D., 5 08, 5 17
Corcos, D. M., 465 , 469, 727, 73 9 Dahl, T., 615 , 629
Cork, C., 462, 468 DAlembert, J. L. R., 6, 7, 8, 18
Corlett, E. N., 187, 200 Dallop, P., 499, 5 03
Corrigan, J. M., 25 5 , 260, 43 3 , 43 6 Dang, N., 671, 674, 681
Corrigan, S., 215 , 221 Daniel, M. H., 3 2, 3 7, 725 , 73 8
Costello, A., 641, 642, 65 0 Darst, P. W., 3 14, 3 15 , 3 16
Cote, J., 14, 15 , 60, 3 03 , 3 14, 3 16, 474, 481, 483 , 484, Darzi, A., 25 0, 25 4, 261, 262, 3 47, 3 5 2
693 , 700, 714 Das, T. L., 3 05 , 3 16
Coughlin, L. D., 3 5 1 Daston, L. J., 115 , 121
author index 795
Dattel, A. R., 15 , 5 2, 94, 3 5 5 , 3 65 , 3 69, 666, 673 Detre, J. A., 664, 678
Dauphinee, W. D., 248, 3 5 3 Detterman, D. K., 3 2, 3 7, 725 , 73 8
Davenport, T. H., 217, 219 Detweiler, M., 660, 663 , 676, 682
David, J.-M., 204, 219 Devlin, J. T., 670, 681
Davids, K., 245 , 246, 25 6, 262, 3 83 , 3 85 , 471, 474, De Volder, A., 5 5 4, 5 63 , 5 67, 675 , 681
475 , 476, 477, 478, 487, 691, 703 de Voogt, A., 5 24, 5 3 6
Davidson, J. E., 626, 63 0, 710 Dewalt, B. R., 129, 143
Davidson, J. W., 10, 18, 45 9, 461, 468, 469, 470, 692, Dewalt, K. M., 129, 143
703 , 725 , 73 9 Dewey, J., 626, 63 0
Davies, C., 109, 121 Dholakia, U. M., 43 5 , 43 6
Davies, D. R., 73 0, 73 7, 741, 742 dHondt, W., 5 3 3 , 5 3 6
Davies, I. R. L., 174, 184, 268, 286 Diamond, R., 676, 678
Davies, N., 75 , 85 Diaper, D., 185 , 192, 199, 200
Davies, S. P., 3 77, 3 78, 3 85 DiBello, L. A., 3 74, 3 75 , 3 77, 3 84
Davis, D. A., 3 49, 3 5 0, 3 5 2 Dick, F., 674, 682
Davis, J. G., 3 76, 3 84 Dickinson, T., 441, 45 3
Davis, K. J., 3 5 8, 3 70, 693 , 702 Dickson, G. W., 3 81, 3 82, 3 87
Davis, R., 12, 14, 43 , 48, 87, 91, 95 , 96, 97, 99, 101 Dickson, M. W., 441, 45 1
Davison, A., 25 2, 25 8, 261 Diderot, D., 6, 7, 8, 18, 203 , 219, 494, 5 01
Dawes, R. M., 26, 28, 405 , 416, 43 3 , 43 6, 686, 700 Didierjean, A., 5 3 2, 5 3 5
Dawis, R. V., 15 8, 164 Diedrich, F. J., 206, 215 , 221, 480, 484
Day, D. V., 448, 45 0, 614, 628, 63 0 Diener, E., 43 7
Day, L. J., 25 6, 25 9 Diener, H. C., 5 08, 5 17
Deakin, J. M., 14, 15 , 60, 23 7, 23 8, 3 03 , 3 05 , 3 06, Dietz, T. M., 120, 121, 75 3 , 75 9
3 07, 3 09, 3 11, 3 16, 3 18, 474, 478, 481, 483 , 485 , DiGirolamo, G. J., 273 , 283
486, 498, 499, 5 00, 5 03 , 601, 608, 693 , 698, 701, Dingwall, R., 107, 109, 110, 121
703 , 709, 714, 721, 73 0, 741 Dinse, R., 465 , 470
Dealey, W. L., 697, 701 Dippel, K., 746, 75 9
Deary, I. J., 3 2, 3 7 Disanto-Rose, M., 499, 5 01
Deblon, F., 641, 649 Dise, M. L., 25 0, 260
de Boishebert, 686, 703 Dissanayake, 476
Decortis, F., 208, 219 Dixon, N. M., 624, 63 0
Deffenbacher, K. A., 186, 191, 200, 203 , 205 , 220, 244, Dixon, R. A., 648, 65 0, 73 6, 73 8
245 , 260 Dizio, P., 5 12, 5 17
Defries, J., 5 63 , 5 65 Djakow, I. N., 10, 18, 226, 23 8, 5 23 , 5 3 3 , 5 3 5
Degner, S., 45 8, 460, 468 Djerassii, C., 91, 101
de Groot, A. D., 11, 13 , 18, 23 , 28, 41, 44, 49, 63 , 169, Doane, S. M., 248, 249, 25 9, 260, 261, 279, 280, 283 ,
171, 182, 226, 23 1, 23 2, 23 8, 244, 260, 3 05 , 3 16, 3 5 1, 3 5 6, 3 5 7, 3 65 , 3 66, 3 68, 3 69, 3 71, 686, 701
478, 484, 5 23 , 5 25 , 5 27, 5 28, 5 29, 5 3 0, 5 3 5 , 5 69, Dobrin, D., 402
5 83 , 5 98, 5 99, 608, 685 , 696, 701, 761, Dogan, M., 117, 121
785 Dogan, N., 401
Dehaene, S., 5 5 6, 5 5 9, 5 63 , 5 66, 670, 675 , 678, 681 Dolan, R. J., 5 5 5 , 5 66
Deiber, M. P., 662, 663 , 679 Doll, J., 10, 18, 5 3 3 , 5 3 6
De Keyser, V., 208, 219, 3 69 Domeshek, E. A., 405 , 411, 418
Dekker, S. W. A., 143 , 199, 201, 208, 219 Dominowski, R. L., 23 0, 23 8
Delaney, P. F., 83 , 85 , 181, 182, 23 7, 23 8, 23 9, 268, Donaldson, G., 5 91, 5 93 , 5 94, 5 95 , 5 99, 609
283 , 5 43 , 5 45 , 5 5 0, 5 93 , 5 98, 600, 608, 690, 701 Donaldson, M., 25 5 , 260
Delazer, M., 5 60, 5 66 Donlan, C., 5 5 45 5 9, 5 66
de Leeuw, N., 23 0, 23 8 Dougals, A., 45 1
Dell, G. S., 5 09, 5 17 Donoghue, J. P., 671, 682
Delp, N. D., 73 4, 73 9 Donoghue, L., 107, 121
DeMaio, J. C., 180, 184, 3 5 6, 3 5 7, 3 65 , 3 70 Donohue, B. C., 671, 679
Demarco, G., 3 12, 3 16 Dorner, D., 243 , 25 9
Dember, W. N., 429, 43 6 Dougherty, M. R. P., 43 1, 43 6
De Mille, A., 498, 5 01 Dourish, P., 128, 13 1, 13 4, 13 8, 143
Deming, W. E., 5 5 7, 5 67 Douthitt, R. A., 3 04, 3 05 , 3 16
Denes, G., 5 60, 5 67 Doverspike, D., 163
Denison, J., 690, 701 Dow, R. S., 5 08, 5 18
Denney, N. W., 684, 701 Dowdy, D., 400
Dennis, M., 618, 622, 63 0 Down, J., 462, 467
Dennis, W., 3 24, 3 26, 3 3 2 Doyle-Wilch, B., 5 01
Derr, M. A., 5 10, 5 19, 729, 740 Doyon, J., 671, 682
DeShon, R. P., 442, 45 0 Dray, W., 5 71, 5 83
Desmond, J. E., 5 08, 5 17 Drebot, M., 621, 625 , 629
DEsposito, M., 63 , 662, 664, 667, 668, 677, 678, 680 Drevdahl, J. E., 3 05 , 3 16
Destrebecqz, A., 274, 283 Dreyfus, H. L., 12, 18
Detienne, F., 3 74, 3 77, 3 85 Dreyfus, S. E., 12, 18
796 author index
Driskell, J. E., 410, 416, 443 , 45 0 Elo, A. E., 21, 28, 3 19, 3 22, 3 23 , 3 29, 3 3 0, 3 3 2, 5 24,
Drury, C. G., 187, 200 5 3 6, 73 5 , 73 8
Druzgal, T. J., 662, 680 Elstein, A. S., 26, 27, 44, 46, 47, 63 , 88, 101, 3 5 1
Duarte, M., 5 14, 5 20 Emanuel, T. W., 25 3 , 25 8, 262
Dubar, C., 106, 121 Emery, F. E., 129, 143
DuBois, D., 617, 63 0 Emery, L., 664, 680
Dubourdieu, D., 268, 283 Emler, A. C., 405 , 415
Duda, J., 204, 219, 716, 720 Emslie, H., 616, 63 0
Duffy, L. J., 481, 484, 693 , 701 Endsley, M. R., 16, 47, 5 2, 5 5 , 213 , 219, 248, 25 9, 3 64,
Duffy, T., 400 3 66, 3 69, 406, 416, 63 3 , 63 4, 63 5 , 63 8, 63 9, 641,
Duguid, P., 623 , 629 642, 644, 645 , 646, 65 0, 65 1, 699, 73 3 , 769
Dumville, B. C., 443 , 45 2 Engbert, R., 727, 729, 73 3 , 73 4, 73 9
Duncan, J., 3 5 6, 3 5 7, 3 60, 3 69, 5 18, 616, Engel, S. A., 669, 678
63 0 Engelien, A., 465 , 470
Duncan, K. D., 189, 199 Engelkamp, J., 496, 5 00, 5 01
Duncker, K., 41, 63 , 168, 182, 224, 23 8 Engelmore, R., 92, 101
Dunn, J. C., 26, 29, 476 Engestrom, Y., 129, 13 5 , 143 , 144, 75 3 , 75 9
Dunning, D., 5 7, 65 Engle, R. W., 3 2, 3 7, 5 0, 5 1, 63 , 64, 73 2, 73 8
Dunsmore, H. E., 3 76, 3 85 Engstrom, R., 5 91, 5 93 , 5 94, 5 95 , 5 99, 609
Dupui, P., 5 00, 5 02 Ennis, R. H., 626, 63 0
Duran, A. S., 5 1, 64 Ensley, M. D., 443 , 446, 448, 45 1, 45 2
Durant, W., 73 , 74, 85 Entin, E. B., 244, 25 9, 406, 418
Durkheim, E., 107, 110, 121 Entin, E. E., 406, 418, 443 , 45 1
Durlach, N. I., 465 , 468 Epel, N., 400
Duroux, B., 5 5 4, 5 60, 5 64, 5 67, 675 , 681 Epstein, S. A., 5 , 18
Durso, F. T., 15 , 5 2, 99, 180, 184, 248, 3 5 5 , 3 5 6, 3 5 7, Epstein, T., 5 76, 5 83
3 64, 3 65 , 3 66, 3 69, 3 70, 668, 673 Epstein, W., 25 , 28
Dutta, A., 19, 272, 284, 285 Erickson, K. I., 65 7, 664, 665 , 666, 678
Dvorak, A., 697, 701 Ericson, M., 692, 702
Dyer, J. L., 45 0 Ericsson, K. A., 3 , 4, 10, 11, 12, 13 , 14, 16, 18, 19, 23 , 24,
25 , 28, 3 1, 3 7, 41, 44, 45 , 46, 47, 49, 5 0, 5 2, 5 4, 5 5 ,
Eastman, R., 205 , 220 5 6, 5 7, 5 9, 60, 63 , 64, 67, 70, 71, 75 , 83 , 85 , 87,
Easton, C., 45 8, 468 96, 101, 105 , 168, 176, 177, 181, 182, 183 , 191, 200,
Ebbinghaus, H., 49, 63 205 , 219, 223 , 224, 226, 227, 228, 229, 23 0, 23 1,
Ebeling, C., 5 25 , 5 3 4 23 2, 23 3 , 23 4, 23 5 , 23 6, 23 7, 23 8, 23 9, 241, 244,
Eberhardt, J. L., 668, 679 248, 249, 25 1, 25 9, 261, 266, 279, 283 , 284, 292,
Eccles, D. W., 203 , 206, 208, 221, 473 , 484 297, 3 00, 3 05 , 3 06, 3 07, 3 08, 3 11, 3 14, 3 16, 3 17,
Edelman, G. M., 5 17 3 21, 3 27, 3 3 2, 3 49, 3 5 1, 3 60, 3 65 , 3 69, 3 70, 3 74,
Edelman, S., 269, 284, 669, 679 3 75 , 3 82, 3 83 , 3 85 , 3 86, 400, 405 , 412, 416, 427,
Eden, G. F., 670, 682 43 1, 43 6, 45 8, 45 9, 460, 462, 463 , 464, 466, 468,
Edmondson, A. C., 444, 446, 448, 45 0 469, 471, 472, 475 , 479, 480, 481, 484, 485 , 491,
Edwards, C. J., 5 5 5 , 5 66 496, 5 01, 5 03 , 5 05 , 5 10, 5 17, 5 20, 5 26, 5 29, 5 3 1,
Edwards, P., 476, 487 5 3 2, 5 3 6, 5 40, 5 41, 5 42, 5 43 , 5 45 , 5 47, 5 5 0, 5 5 3 ,
Edwards, W., 424, 43 6 5 5 8, 5 60, 5 61, 5 62, 5 645 65 , 5 66, 5 72, 5 83 , 5 88,
Egan, D. E., 5 0, 5 1, 63 , 172, 179, 182 5 93 , 5 98, 5 99, 600, 601, 602, 606, 608, 609, 613 ,
Egan, V., 3 2, 3 7 614, 63 0, 63 9, 643 , 646, 649, 65 0, 65 3 , 65 8, 65 9,
Eggleston, R. G., 193 , 200 667, 675 , 683 , 685 , 686, 687, 688, 689, 690, 691,
Ehn, P., 129, 13 0, 143 692, 693 , 694, 695 , 696, 697, 698, 699, 700, 701,
Ehrlich, K., 3 78, 3 86 702, 705 , 706, 708, 711, 712, 718, 719, 720, 726,
Eid, J., 445 , 448, 449 727, 728, 729, 73 0, 73 1, 73 2, 73 3 , 73 4, 73 5 , 73 8,
Eilers, A. T., 176, 183 73 9, 744, 748, 761, 766, 767, 768, 781, 785 ,
Eisenstadt, J. M., 3 27, 3 3 2 786
Eisenstadt, M., 47, 63 , 171, 182 Eriksen, 618
Ekman, P., 493 , 5 02 Erman, L., 92, 101
Ekornas, B., 445 , 448, 449 Ernst, G., 42, 64
Elander, J., 3 63 , 3 69 Ernst, G. W., 11, 18
Elbert, T., 465 , 466, 468, 5 08, 5 17, 5 3 3 , 5 3 4, 674, Eskridge, T., 212, 218
678, 695 , 701 Essig, M., 662, 679
Elchardus, M., 3 05 , 3 16 Estes, W. K., 5 91, 5 96, 608
El Guindi, F., 13 0, 143 Etelapelto, A., 3 79, 3 85
Elias, J. L., 71, 85 Etzioni, A., 108, 121
Elias, N., 118, 121 Etzioni-Halevy, E., 118, 119, 121
Elliott, D. H., 3 04, 3 12, 3 17, 476, 487 Euhus, D. M., 3 47, 3 5 3
Elliott, L. R., 244, 25 9 Eva, K. W., 15 , 47, 5 5 , 23 5 , 25 0, 3 46, 3 49, 3 5 1
Ellis, H., 3 26, 3 3 2 Evans, A. W., 215 , 219
Elm, W. C., 193 , 201, 208, 222 Evans, D. A., 3 5 2, 496, 5 03
Elms, A. C., 3 20, 3 3 2 Everett, W., 771, 778
author index 797
Evetts, J., 9, 15 , 105 , 107, 110, 111, 112, 121, 614, 628, Fiore, S. M., 15 , 43 9, 444, 45 0
746, 75 3 , 75 4, 75 9 Fischer, U., 445 , 45 2
Eyferth, K., 3 78, 3 87 Fischhoff, B., 244, 260
Eyrolle, H., 3 69 Fisher, J. A., 3 62, 3 70
Eys, M. A., 448, 449 Fishwick, R. J., 25 4, 261
Fisk, A. D., 5 3 , 66, 5 26, 5 3 2, 5 3 6, 65 9, 678
Facchini, S., 671, 681 Fisk, J. E., 5 94, 608
Fagan, J. F., 5 92, 608 Fitts, P. M., 18, 47, 5 9, 60, 64, 83 , 85 , 267, 271, 284,
Fagan, L. M., 89, 102 462, 468, 475 , 485 , 5 12, 5 13 , 5 17, 617, 63 0, 65 8,
Fagerhaugh, S., 144 678, 684, 694, 702, 725 , 73 8
Fahey, J. L., 495 , 5 01 Fitzgerald, J., 400
Fahle, M., 268, 269, 283 , 284 Fitzsimmons, C., 728, 740
Faidiga, L., 672, 681 Fitzsimmons, M. P., 6, 18
Fajen, B. R., 5 15 , 5 17 Fix, J. L., 43 5 , 43 6
Falk, G., 3 22, 3 23 , 3 3 3 Fix, V., 3 78, 3 87
Falkenhainer, B., 180, 183 Fiz, J. A., 464, 468
Fallesen, J. J., 410, 416 Flanagan, D. P., 5 88, 5 90, 608, 610
Faloon, S., 23 6, 23 9, 5 42, 5 5 0 Flanagan, J. C., 188, 189, 192, 200
Farnsworth, P. R., 45 8, 468 Flavell, J., 5 5 , 64
Farquhar, A., 100, 103 Fleck, 763 , 775
Farr, B. R., 89, 102 Fleishman, E. A., 3 2, 3 7, 164, 443 , 45 1, 725 , 73 8
Farr, M. J., 3 , 12, 18, 23 , 27, 3 1, 3 7, 46, 63 , 95 , 101, 13 0, Fletcher, R. H., 3 49, 3 5 0
13 1, 142, 244, 25 9, 3 69, 412, 416, 43 6, 686, 700 Fleury, C., 471, 475 , 476, 485
Farr, M. L., 205 , 219 Fleury, M., 475 , 485
Farrand, P., 3 63 , 3 70, 648, 65 0 Flin, R., 409, 416, 45 1
Farrell, J. M., 275 , 276, 286 Flor, D. L., 706, 719
Farrow, D., 25 6, 260 Flores, F., 405 , 419
Fassina, N. E., 3 83 , 3 86 Flower, L. S., 3 90, 400, 401
Favart, M., 5 72, 5 83 Flowers, D. L., 670, 682
Fayol, M., 400 Foley, M. A., 499, 5 01
Fazio, F., 672, 681 Forbus, K. D., 46, 180, 183
Fehr, T., 5 3 3 , 5 3 4 Ford, G. C., 697, 701
Feigenbaum, E. A., 12, 14, 43 , 48, 62, 87, 90, 91, 99, Ford, J. K., 440, 45 3
101, 102, 204, 219 Ford, K. M., 3 , 18, 23 , 24, 28, 46, 64, 95 , 101, 13 1, 13 4,
Feightner, J. W., 46, 47, 62, 3 5 0, 3 5 2 13 5 , 143 , 178, 183 , 206, 207, 208, 211, 212, 213 , 215 ,
Feldman, D. H., 291, 292, 299, 3 00, 690, 702 216, 217, 218, 219, 220, 3 85 , 748, 75 9, 760
Feldman, J., 87, 101 Ford, P., 485
Fellander-Tsai, L., 25 0, 261 Forsberg, H., 696, 700
Felleman, D. J., 65 6, 678 Forsman, L., 674, 677, 696, 700
Felt, U., 9, 15 , 105 , 744, 746, 75 3 Forssberg, H., 662, 674, 677, 680
Feltovich, P. J., 3 , 12, 14, 17, 18, 23 , 25 , 27, 28, 29, 41, Forsyth, D. E., 204, 219
44, 46, 5 1, 5 2, 5 3 , 5 4, 5 5 , 5 6, 63 , 64, 65 , 67, 76, 77, Forsyth, E., 3 70
81, 83 , 86, 87, 89, 95 , 96, 100, 101, 105 , 13 1, 13 4, Forsythe, 615 , 616, 622
13 5 , 143 , 169, 172, 174, 175 , 177, 179, 180, 181, 182, Forsythe, D. E., 95 , 97, 101, 128, 13 1, 13 3 , 143
183 , 204, 205 , 219, 249, 260, 3 5 0, 3 5 1, 3 76, 3 84, Forsythe, G. B., 3 2, 3 8, 615 , 616, 618, 622, 623 , 63 0,
3 85 , 406, 415 , 440, 45 0, 5 69, 5 83 , 614, 63 9, 641, 63 1
647, 65 3 , 65 8, 65 9, 667, 674, 675 , 677, 708, 713 , Foster, S. L., 3 13 , 3 14, 3 17
73 0, 73 5 , 743 , 75 0, 761, 763 , 765 , 75 9, 767, 786 Fournier, V., 111, 112, 113 , 121
Feltz, D. L., 15 9, 164 Foushee, H. C., 445 , 45 1
Fencsik, D. E., 5 9, 67, 277, 285 Fox, K., 716, 720
Fendrich, D. W., 276, 284 Fox, L., 711, 720
Ferguson, L. W., 163 , 164 Fox, P. T., 5 08, 5 19
Fernandes, C., 5 63 , 5 67 Fox, P. W., 5 45 , 5 5 0
Ferrah-Caja, E., 5 93 , 5 94, 611 Fozard, J. L., 726, 73 7
Ferrari, M., 75 6, 75 9 Frackowiak, R.S.J., 5 08, 5 18, 5 48, 5 5 1, 5 64, 5 67, 667,
Ferrer-Caja, E., 5 93 , 610 668, 673 , 674, 679, 680
Ferris, G. R., 3 81, 3 85 Francis, N. K., 3 48, 3 5 1
Ferster, C. B., 45 , 64 Frank, A., 5 3 3 , 5 3 6
Fetterman, D. M., 128, 143 Frankenberg, R., 3 05 , 3 17
Feyerabend, P., 119, 121 Franks, A., 478, 486
Fick, G. H., 3 5 0 Frederick, J. A., 3 04, 3 17
Fiedler, K., 27, 3 0 Frederickson, B. L., 43 1, 43 6
Fielder, C., 4, 18 Frederiksen, J. R., 278, 279, 284
Fiez, J. A., 5 08, 5 17 Fredrick, J. A., 3 04, 3 17
Filby, W. C. D., 708, 720 Freeark, R. J., 3 48, 3 5 3
Fincher-Kiefer, R. H., 3 75 , 3 87 Freedman, D. J., 669, 678
Finkel, D., 5 93 , 5 95 , 608 Freeman, C., 186, 200, 624, 629
798 author index
Freeman, J. T., 295 , 3 01, 3 74, 3 80, 3 86, 406, 416, 445 , Gaser, C., 465 , 468, 674, 678, 695 , 702
45 0 Gathercole, S. E., 5 93 , 608
Frehlich, S. G., 25 6, 261, 476 Gatz, M., 5 93 , 5 95 , 608
Freidson, E., 106, 109, 110, 111, 112, 114, 121, 75 4, 75 9 Gauthier, I., 5 08, 5 17, 667, 668, 676, 678, 682
French, D., 3 63 , 3 69 Gauthier, J., 674, 682
French, K. E., 23 4, 23 9, 473 , 474, 479, 485 , 486 Gawel, R., 686, 702
Frenkel, S., 107, 121 Gazzaniga, M. S., 65 3 , 678
Frensch, P. A., 24, 26, 29, 205 , 222, 267, 284, 3 49, 3 5 3 Geary, D. C., 5 63 , 5 66
Frenzel, L. E., 103 Gecas, V., 75 6, 75 9
Frey, P. W., 5 27, 5 3 6 Gecht, M. R., 3 5 0
Freyhoff, H., 5 27, 5 3 6 Geis, C. E., 446, 448, 45 0
Friedel, 780, 781, 788 Geisler, C., 401
Friedland, R. P., 496, 5 01 Geiss, A., 3 48, 3 5 3
Friedman, C. P., 27, 28, 23 4, 23 8, 3 49, 3 5 1 Geissler, P. W., 621, 63 2
Frieman, J., 5 40, 5 42, 5 46, 5 5 2 Gellert, E., 178, 183
Friesen, W. V., 493 , 5 02 Gellman, L., 3 48, 3 5 3
Frith, C. D., 5 11, 5 16, 5 48, 5 5 1, 5 64, 5 67, 673 , 680 Gelman, R., 5 5 9, 5 66
Fritsch, T., 496, 5 01 Gembris, H., 45 8, 468
Frohna, A. Z., 3 42, 3 5 1 Gentile, C. A., 3 99, 401
Frohring, W. R., 725 , 73 9 Gentner, D., 5 2, 64, 180, 183 , 205 , 219, 3 66, 3 70
Fromkin, V. A., 5 09, 5 17 Gerchak, Y., 3 27, 3 28, 3 3 2, 5 3 2, 5 3 5
Frost, S. J., 670, 671, 681, 682 Gernsbacher, M. A., 402
Frydman, M., 5 3 3 , 5 3 6 Gersick, C. J., 441, 45 1
Fujigaki, Y., 3 82, 3 85 Gesi, A. T., 276, 284
Fukuyama, F., 75 4, 75 9 Getzels, J. W., 5 73 , 5 83
Fulbright, R. K., 670, 682 Gherardi, S., 623 , 628, 63 0
Fulcomer, M., 3 05 , 3 17 Giachino, A. A., 3 48, 3 5 3
Fulgente, T., 5 3 3 , 5 3 7 Giacobbi, P. R., 710, 721
Furby, L., 163 Giacomi, J., 13 8, 142
Furey, M. L., 668, 677, 679 Gibson, E. J., 268, 284, 5 14, 5 17
Furmanski, C. S., 669, 678 Gibson, E. L., 97, 102
Futterman, A. D., 495 , 5 01 Gibson, G. J., 3 2, 3 7
Gibson, J. J., 268, 284, 480, 485 , 5 13 , 5 17
Gabrieli, J. D. E., 5 08, 5 17, 664, 665 , 668, 671, 673 , Giddens, A., 75 4, 75 9
677, 679, 680, 681 Gieryn, T., 114, 122
Gadea, C., 106, 121 Giesel, F. L., 662, 679
Gadian, D. G., 5 48, 5 5 1, 673 , 680 Gil, G., 621, 625 , 629
Gagne, R. M., 77, 78, 80, 82, 85 , 201, 204, 219, 226, Gilbreth, F., 187
228, 240 Gilhooly, K. J.,24, 28, 174, 178, 181, 23 4, 241, 5 5 9, 5 67
Gaines, B. R., 101, 102, 204, 219 Gill, D. L., 716, 721
Galanter, E., 41, 44, 65 , 226, 240 Gill, H. S., 5 64, 5 67
Galbraith, O., 25 4, 25 9 Gill, R. T., 205 , 220
Gale, T., 247, 248, 262 Gillis Light, J., 5 63 , 5 65
Galer, I. A. R., 3 62, 3 70 Gilmartin, K. J., 5 27, 5 3 8
Gallagher, A. G., 25 5 , 260, 261 Gilmore, R. O., 5 06, 5 19
Gallagher, J., 472, 482, 483 , 487 Gingrich, K. F., 3 81, 3 84
Gallese, V., 672, 678 Girard, N., 3 5 3
Gallistel, C. R., 5 5 9, 5 66 Girelli, L., 5 60, 5 66
Galton, F., 10, 18, 71, 224, 225 , 240, 3 05 , 3 17, 3 21, 3 23 , Givon, T., 401
3 26, 3 27, 3 3 2, 45 8, 468, 5 5 3 , 5 66, 684, 685 , 702, Gizzo, D. P., 3 13 , 3 14, 3 17
724, 73 8 Glaser, D. E., 672, 677
Gammage, K., 710, 720 Glaser, R., 3 , 12, 17, 18, 23 , 24, 27, 29, 3 1, 3 7, 44, 45 ,
Gandolfo, F., 5 12, 5 17 46, 47, 49, 5 0, 5 1, 5 2, 5 3 , 5 4, 5 5 , 63 , 64, 65 , 66, 79,
Gao, J. H., 5 08, 5 17 80, 82, 83 , 84, 85 , 95 , 101, 13 0, 13 1, 142, 15 7, 163 ,
Garavan, H., 65 5 , 65 8, 660, 661, 662, 678, 680 169, 172, 174, 175 , 177, 179, 180, 181, 182, 183 , 204,
Garcia, T., 713 , 720 205 , 219, 220, 244, 25 9, 287, 3 01, 3 05 , 3 16, 3 65 ,
Garcia Caraballo, N. M., 662, 679 3 69, 3 70, 3 76, 3 84, 406, 412, 415 , 416, 43 6, 440,
Gardner, H., 3 4, 3 7, 71, 77, 85 , 191, 200, 5 5 4, 5 64, 45 0, 5 69, 5 83 , 686, 694, 700, 702, 706, 720
5 66, 626627, 63 0, 63 2, 75 6, 75 9, 769, 786 Glass, J. M., 5 9, 67, 277, 285
Gareau, L., 670, 682 Glass, R. L., 3 82, 3 85
Garhammer, M., 3 05 , 3 17 Glenberg, A. M., 25 , 28, 497, 5 02, 5 06, 5 17
Garland, D., 641, 642, 65 0 Glencross, D., 475 , 486
Garland, D. J., 478, 485 Glendon, A. I., 73 0, 742
Garland, D. L., 213 , 219 Glickman, A. S., 441, 45 2
Garling, T., 43 6, 43 7 Glorieux, I., 3 05 , 3 16
Garner, W. L., 77, 85 , 618 Glover, G. H., 5 08, 5 17
Garnier, H., 3 23 , 3 3 3 Gluck, M. A., 673 , 681
Gaschnig, J., 204, 219 Glynn, S. M., 401
author index 799
Gobbini, M. I., 667, 668, 677, 679 Grafton, S. T., 662, 663 , 679
Gobet, F., 11, 16, 18, 19, 24, 25 , 28, 3 1, 3 7, 44, 49, 5 0, Graham, D. J., 280, 283
5 2, 5 4, 5 8, 60, 64, 66, 168, 174, 182, 23 3 , 23 5 , 3 5 1, Graham, J., 645 , 65 1
5 23 , 5 24, 5 25 , 5 26, 5 27, 5 28, 5 29, 5 3 0, 5 3 1, 5 3 2, Graham, K. C., 23 4, 23 9, 474, 485
5 3 3 , 5 3 5 , 5 3 6, 5 3 7, 5 3 8, 5 98, 600, 608, 693 , 696, Graham, S., 245 , 25 9, 478, 483 , 709, 720
769 Grande, G. E., 3 70
Godbey, G., 3 04, 3 18 GrandMaison, P., 3 5 3
Goel, A. K., 178, 183 Granovetter, M. S., 118, 122, 75 7, 75 9
Goel, V., 5 5 5 , 5 66 Grape, C., 692, 702
Goertzel, M. G., 3 27, 3 3 2 Gray, W. D., 176, 179, 183 , 244, 25 3 , 260
Goertzel, T. G., 3 27, 3 3 2 Graydon, J. K., 708, 720
Goertzel, V., 3 27, 3 3 2 Grazioli, S., 23 5 , 240
Goetz, E. T., 402 Green, C., 48, 64
Goff, G. N., 3 2, 3 8 Green, A. J. F., 23 7, 240
Goff, M., 163 Green, B. F. Jr., 3 3 , 3 8
Goffman, E., 13 5 , 143 , 749, 75 9 Green, C., 24, 28
Goh, J., 3 64, 3 71 Green, L., 45 9, 468
Golby, A. J., 668, 679 Green, M. L., 43 4, 43 7
Gold, A., 5 3 2, 5 3 3 , 5 3 8 Greenbaum, J., 129, 13 0, 143
Goldfield, E. C., 5 14, 5 17 Greenberg, D., 710, 722
Goldiez, B., 243 , 260 Greenberg, L., 205 , 220
Goldin, S. E., 5 28, 5 3 7 Greene, T. R., 23 , 29, 47, 67, 3 75 , 3 87, 5 70, 5 78, 5 84
Goldman, A., 672, 678 Greenwood, R., 106, 108, 122
Goldman, L., 43 4, 43 7 Gregorich, S. E., 446, 448, 45 0
Goldman-Rakic, P. S., 664, 665 , 677 Grelon, A., 110, 122
Goldschmidt-Clearmont, L., 3 05 , 3 17 Grey, C., 113 , 122
Goldsmith, L. T., 292, 293 , 3 01 Grey, S. M., 187, 200
Goldsmith, T. E., 180, 184, 3 5 6, 3 5 7, 3 65 , 3 70 Grezes, J., 672, 677
Goldstein, I., 91, 101 Grigorenko, E. L., 10, 19, 3 1, 3 2, 3 8, 615 , 616, 618, 621,
Goldstone, R. L., 268, 284 623 , 625 , 629, 63 1, 63 2
Golen, S., 709, 712, 720 Grill-Spector, K., 668, 669, 679
Golomer, E., 5 00, 5 02 Gritter, R. J., 91, 101
Gomez, G., 212, 218 Grobe, C., 401
Gomez de Silva Garza, A., 178, 183 Grober, E. D., 3 48, 3 5 1, 3 5 3
Gonzales, P., 746, 75 9 Grocki, M. J., 5 13 , 5 19
Gonzalez, A., 217 Groeger, J. A., 3 70
Good, C. D., 5 48, 5 5 1, 673 , 674, 680 Groen, G. J., 5 5 , 5 6, 64, 66, 102, 23 5 , 240, 25 1, 261,
Good, R., 177, 184 352
Goodbody, S. J., 5 12, 5 18 Gronlund, S. D., 3 64, 3 66, 3 69
Goodeve, P. J., 5 17 Gronn, P., 448, 45 0
Goodnow, J. J., 44, 62 Grossin, W., 3 05 , 3 17
Goodstein, L. P., 144, 208, 222 Gross-Tsur, V., 5 63 , 5 67
Goodwin, C., 5 2, 65 Grotzer, T. A., 626, 63 0
Goodwin, G. F., 15 , 43 9, 441, 45 3 Gruber, H. E., 16, 47, 5 9, 60, 287, 3 01, 45 7, 45 8, 460,
Goolsby, T. W., 465 , 468 464, 468, 5 27, 5 3 2, 5 3 3 , 5 3 6, 5 3 8, 673 , 693 , 769
Goossens, L., 5 49, 5 5 2 Gruber, O., 5 5 4, 5 66
Gordon, C., 111, 121 Grudin, J. G., 5 09, 5 18
Gordon, J., 96, 101 Grudowski, M., 709, 714, 719, 720
Gordon, P., 5 40, 5 5 1 Grue, N., 178, 183
Gordon, S. E., 205 , 211, 220 Gruhn, W., 463 , 468
Gore, J. C., 5 08, 5 17, 664, 665 , 667, 668, 676, 677, Gruppen, L. D., 3 42, 3 5 1
678, 681 Gruson, L. M., 461, 468, 698, 702
Gorno-Tempini, M. L., 667, 668, 679 Gudagunti, R., 107, 122
Gorry, G. A., 43 , 66 Guerin, B., 3 78, 3 79, 3 85
Gorter, S., 3 49, 3 5 3 Guha, R. V., 99, 102
Gossweiler, R., 5 14, 5 19 Guilford, J. P., 15 8, 164
Gott, S. P., 204, 221 Gully, S. M., 441, 45 1
Gottfredson, L. S., 615 , 63 0 Gurfinkel, V. S., 5 14, 5 16
Gottlieb, R., 497, 5 02 Guskey, T. R., 78, 79, 85
Gottsdanker, R., 277, 284 Guskin, H., 490, 5 02
Goudas, M., 716, 720 Gutenschwager, G. A., 3 05 , 3 17
Gould, S. J., 691, 702 Guttman, L., 164
Goulet, C., 471, 475 , 476, 485 Guzzo, R. A., 440, 441, 45 1
Govinderaj, T., 178, 183 Guzzon, R., 45 0
Gowin, D. B., 222
Grace, D. M., 3 48, 3 5 1 Haber, L., 3 5 9, 3 60, 3 63 , 3 70
Graf, P., 497, 5 03 Haber, R. N., 3 5 9, 3 60, 3 63 , 3 70
Grafman, J., 5 3 3 , 5 3 7 Hackman, J. R., 43 9, 441, 443 , 45 1
800 author index
Kaufman, A. S., 5 93 , 5 95 , 609, 726, 73 9 Kitsantas, A., 707, 708, 709, 711, 712, 713 , 714, 716,
Kaufman, D. R., 24, 29 717, 720, 722
Kaufmann, D. R., 5 2, 66, 23 5 , 240 Kivlighan, D. M., 174, 175 , 183
Kausler, D. H., 5 94, 609, 611 Kjellin, A., 25 0, 261
Kay, B. A., 5 14, 5 17 Klahr, D., 205 , 208, 221
Kayes, C. D., 446, 45 1 Klahr, P., 99, 101
Kazanas, H. C., 81, 86 Klapp, S. T., 5 09, 5 18
Keating, D. P., 5 5 5 , 5 65 Klatzky, R. L., 43 2, 43 8, 686, 701
Keating, T., 715 , 719 Klaus, D. J., 80, 85
Keck, J. W., 3 48, 3 5 2 Klausmeier, H. J., 79, 85
Keele, S. W., 273 , 275 , 283 , 465 , 469, 5 08, 5 13 , 5 17, Klein, D. E., 200
5 18, 727, 73 9 Klein, G. A., 15 , 23 , 28, 3 3 , 3 7, 46, 5 2, 5 4, 5 6, 64, 65 ,
Keeney, R. L., 424, 43 4, 43 6 97, 102, 13 8, 170, 171, 173 , 176, 182, 183 , 187, 192,
Kellaghan, T., 3 2, 3 7 198, 199, 200, 201, 205 , 206, 209, 212, 216, 217, 219,
Keller, E. F., 117, 122 220, 221, 243 , 244, 261, 3 63 , 3 67, 3 70, 3 71, 3 74,
Keller, F. S., 225 , 240 3 80, 3 86, 403 , 404, 405 , 406, 407, 408, 410, 411,
Keller, R., 405 , 416 412, 413 , 414, 415 , 416, 417, 418, 419, 422, 426, 43 0,
Keller, T. A., 664, 680 43 3 , 43 6, 43 7, 43 8, 441, 442, 444, 445 , 45 0, 45 1,
Kelley, R., 3 80, 3 85 45 2, 45 3 , 5 29, 5 3 5 , 63 7, 63 9, 640, 65 0, 73 6, 745 ,
Kellogg, R. T., 15 , 44, 60, 23 5 , 3 89, 401, 693 75 9
Kelly, A. M. C., 65 5 , 65 8, 660, 661, 680 Kleine, B. M., 3 83 , 3 86
Kelly, B. C., 443 , 45 1 Kleinman, D. L., 443 , 45 1
Kelso, J. A. S., 5 14, 5 20 Kleinschmidt, A., 5 5 4, 5 66
Kelso, M. T., 495 , 5 02 Kliegl, R., 5 47, 5 49, 5 5 0, 5 5 1, 724, 727, 729, 73 3 ,
Kemeny, M. E., 495 , 5 01 73 4, 73 9, 740
Kemper, K., 3 5 6, 3 5 7, 3 64, 3 66, 3 67, 3 71, 445 , 45 3 Klimoski, R., 443 , 45 1
Kendall, D., 442, 45 0 Kline, D. A., 444, 45 1
Kennedy, C., 496, 5 02 Kling, R., 13 1, 144
Kennedy, W. A., 668, 680 Klingberg, T., 662, 664, 665 , 671, 677, 680, 681
Kennet, J., 5 75 , 5 80, 5 83 , 5 84 Klinger, D. W., 413 , 417, 419
Kenny, S., 5 10, 5 19, 729, 740 Klopfer, D., 23 , 29, 44, 5 3 , 65 , 172, 181, 183
Kent, P., 5 5 3 , 5 67 Kneebone, R., 25 5 , 260
Keren, G., 405 , 416 Kneeland, H., 3 04, 3 17
Kerlirzin, Y., 476, 486 Knights, D., 3 05 , 3 17
Kernodle, M., 479, 486 Knopf, M., 73 6, 73 9
Kerr, T., 481, 485 , 487, 5 18 Knorr-Cetina, K. D., 116, 122, 205 , 206, 208, 221
Keys, W., 5 63 , 5 67 Knowles, J. M., 246, 247, 25 2, 25 6, 25 7, 25 8, 262,
Khatwa, R., 25 3 , 260 477, 488, 697, 703
Kida, T., 405 , 418, 43 3 , 43 7 Kobatake, E., 669, 680
Kiekel, P. A., 446, 45 0 Kobbeltvedt, T., 445 , 448, 449
Kieras, D. E., 5 9, 65 , 67, 191, 200, 277, 285 , 663 , 666, Kobus, D. A., 406, 417
676, 681 Koehler, D. J., 43 7
Kiker, D. S., 73 4, 73 7 Koeske, R., 175 , 176, 182
Killackey, H. P., 5 08, 5 18 Kofler, M., 671, 681
Kilner, P., 624, 63 0 Koh, K., 5 12, 5 18
Kim, J. S., 665 , 678 Kohl, J., 480, 484
Kim, N. S., 3 42, 3 5 2 Kohn, L. T., 25 5 , 260
Kimball, D. R., 3 78, 3 85 Kolabinska, M., 119, 122
Kimball, H. R., 3 5 2 Kolb, B., 65 7, 680, 695 , 702
King, J., 91, 101 Kolodner, J., 92, 102
Kingberg, T., 664, 665 , 679 Kolodny, J., 616, 63 0
Kingston, K. M., 708, 720 Koltanowski, G., 23 3 , 240, 5 99, 609
Kinsley, B., 3 04, 3 17 Komarovsky, M., 3 04, 3 17
Kintsch, W., 11, 18, 25 , 28, 5 0, 5 2, 5 4, 5 6, 63 , 64, 23 2, Konijn, E. A., 495 , 5 02
23 3 , 23 5 , 23 9, 248, 249, 25 9, 260, 262, 3 5 2, 401, Koning, P., 23 4, 23 8, 478, 483
43 1, 43 6, 463 , 468, 496, 5 01, 5 26, 5 29, 5 3 6, 5 47, Kooman, J. P., 3 5 3
5 5 0, 5 5 8, 5 60, 5 64, 5 66, 5 72, 5 83 , 5 88, 5 98, 5 99, Koonce, J. M., 25 3 , 261
600, 608, 686, 696, 701, 711, 720, 726, 73 8 Koopman, P., 215 , 221
Kirby, I. K., 24, 28 Kopelman, M., 5 5 7, 5 66
Kirk, E. P., 5 9, 63 Kopiez, R., 45 8, 460, 464, 469
Kirkpatrick, A. E., 25 1, 260 Korczynski, M., 107, 121
Kirlik, A., 628, 63 1 Kording, K. P., 5 12, 5 18
Kirschenbaum, D. S., 712, 720 Koriat, A., 497, 5 02
Kirsner, K., 26, 29, 266, 286 Kornblum, S., 271, 284, 5 12, 5 19
Kirwan, B., 185 , 200 Korotkin, A. L., 443 , 45 1
Kiss, I., 424, 43 6 Koschmann, T. D., 5 2, 65
Kite, K., 3 5 6, 3 5 7, 3 64, 3 66, 3 67, 3 71, 445 , 45 3 Koslowski, B., 5 98, 5 99, 607
804 author index
Shanteau, J., 4, 19, 26, 29, 5 2, 66, 88, 102, 205 , 222, Simonson, I., 43 4, 43 7
3 70, 405 , 418, 426, 43 2, 43 7, 43 8, 686, 703 , 75 8, Simonton, D. K., 12, 15 , 19, 21, 22, 29, 60, 164, 3 19,
760 3 20, 3 21, 3 22, 3 23 , 3 24, 3 25 , 3 26, 3 27, 3 28, 3 29,
Shapin, S., 115 , 123 3 3 0, 3 3 1, 3 3 3 , 3 3 4, 3 3 5 , 45 8, 689, 703 , 73 5 , 741,
Shapira, Z., 43 4, 43 7 766, 767, 771, 786
Shapiro, D., 495 , 5 01 Simpson, S. A., 23 4, 241
Sharp, C., 5 94, 608 Sims, H. P., 444, 45 2
Sharp, J., 448, 45 0 Sinacore, J. M., 3 5 0
Shattuck, L., 645 , 65 1 Sinangil, H. K., 45 0
Shaver, P. R., 5 92, 611 Singer, 476
Shaw, J. C., 90, 102 Singer, C., 6, 19, 690, 703
Shaw, M. L. G., 102 Singer, R. N., 25 6, 261, 476, 486
Shaywitz, B. A., 671, 681 Singer, T., 73 4, 741
Shaywitz, S. E., 671, 681 Singh, H., 5 64, 5 67
Shea, J. B., 5 06, 5 19 Singh, R., 667, 668, 677
Shefy, E., 43 0, 43 7 Singleton, S., 3 05 , 3 17
Sheiner, L. B., 89, 102 Sinha, A. P., 3 76, 3 77, 3 78, 3 84
Shepherd, A., 185 , 190, 201 Skare, S., 674, 677, 696, 700
Sheppard, S., 3 74, 3 86 Skinner, B. F., 45 , 64, 67
Shertz, J., 175 , 184, 3 79, 3 87 Skovronek, E., 602, 611, 724, 727, 73 2, 73 3 , 741
Sherwood, R., 626, 629 Skudlarski, P., 5 08, 5 17, 667, 668, 676, 678
Shiffrar, M. M., 268, 269, 283 Slagter, H. A., 5 3 , 64, 660, 661, 679
Shiffrin, R. M., 3 1, 3 8, 5 3 , 60, 66, 67, 269, 270, 281, Slamecka, N. Y., 497, 5 03
286, 475 , 486, 5 12, 5 13 , 5 19, 65 8, 65 9, 676, 682 Slaughter, J. E., 3 23 , 3 24, 3 25 , 3 3 0, 3 3 5
Shima, K., 671, 672, 682 Slaven, G., 409, 416
Shin, R. K., 664, 678 Sleeman, D., 46, 67
Shinar, D., 3 5 6, 3 5 7, 3 60, 3 71 Sloboda, J. A., 10, 18, 3 1, 293 , 299, 3 01, 45 9, 461, 463 ,
Shipp, S., 5 7, 67 468, 470, 692, 703 , 725 , 73 9, 741, 767, 770, 786
Shire, K., 107, 121 Small, B. J., 5 93 , 606
Shohamy, D., 673 , 681 Small, S., 674, 682
Shook, R. W. C., 641, 642, 65 0 Smeeton, N. J., 245 , 246, 247, 25 2, 25 6, 25 7, 25 8, 261,
Shortliffe, E. H., 43 , 46, 48, 63 , 67, 91, 92, 94, 96, 97, 262, 476, 477, 478, 486, 487, 488, 697, 703
98, 101, 102, 13 0, 13 5 , 142, 204, 222 Smith, D. H., 91, 101
Shriffin, R. M., 267, 269, 285 , 286, 761 Smith, E. C., 204, 219, 226, 228, 240
Shrobe, H. E., 95 , 101 Smith, E. M., 23 5 , 240, 3 46, 3 5 2, 440, 45 3
Shulman, L. S., 44, 46, 47, 63 , 88, 101, 3 40, 3 41, 3 5 1 Smith, G. A., 726, 740
Shumway-Cook, A., 73 5 , 742 Smith, J., 3 , 11, 13 , 18, 23 , 28, 3 1, 3 7, 46, 5 9, 64, 75 ,
Shunn, C., 5 7, 66 85 , 23 1, 23 2, 23 9, 244, 25 9, 266, 284, 3 60, 3 70,
Shuter-Dyson, R., 45 7, 470 3 74, 3 75 , 3 85 , 400, 43 6, 471, 5 47, 5 49, 5 5 0, 5 5 1,
Sidhu, R. S., 3 47, 3 5 1 614, 63 0, 686, 687, 702, 73 4, 73 9
Sieck, W. R., 405 , 417, 422, 43 0, 43 7 Smith, J. F., 405 , 418, 43 3 , 43 7
Siegler, R. S., 24, 29, 73 5 , 741 Smith, J. E. K., 5 12, 5 19
Sierhuis, M., 143 , 144 Smith, M. C., 23 7, 240
Sigala, N., 677, 682 Smith, M. U., 177, 184
Sigma-Mugan, C., 43 3 , 43 7 Smith, P. L., 78, 79, 86
Siler, S. A., 177, 182 Smith, R., 100, 103
Silfies, L. N., 5 75 , 5 80, 5 84 Smith, R. G., 91, 101
Silverman, S. M., 3 27, 3 3 3 Smith, S. B., 5 5 4, 5 5 5 , 5 5 7, 5 5 9, 5 60, 5 61, 5 67
Simmel, G., 749, 760 Smith, S. M., 65 3 , 679
Simmons, R., 204, 219 Smith, S. G. T., 25 0, 25 4, 262
Simon, D. P., 23 , 24, 29, 44, 65 , 88, 102, 169, 177, 179, Smith-Jentsch, K., 445 , 448, 45 3
184, 5 69, 5 83 , 614, 63 0 Smode, A. F., 25 3 , 261
Simon, H. A., 11, 12, 17, 18, 19, 23 , 24, 27, 28, 29, 3 1, Smyth, K. A., 496, 5 01
3 7, 41, 42, 44, 47, 49, 5 0, 5 2, 5 5 , 5 7, 5 8, 60, 61, 63 , Smyth, M. M., 491, 5 00, 5 02, 5 03
64, 65 , 66, 67, 87, 88, 90, 96, 100, 102, 103 , 13 4, Snoddy, G. S., 267, 286
144, 168, 169, 171, 173 , 176, 177, 178, 179, 182, 183 , Snook, S. A., 3 2, 3 8, 615 , 616, 618, 622, 623 , 63 0, 63 1
184, 191, 200, 201, 205 , 207, 218, 219, 222, 224, 226, Snow, A. J., 3 71
227, 228, 229, 23 0, 23 1, 23 5 , 23 6, 23 7, 23 9, 240, Snow, R. E., 15 9, 164
241, 244, 245 , 25 9, 292, 297, 3 01, 3 05 , 3 16, 3 18, Snowden, P. T., 174, 184
3 5 1, 3 5 3 , 3 69, 3 74, 3 86, 3 89, 402, 405 , 406, 416, Snyder, A. J., 25 0, 260
418, 43 1, 43 6, 474, 478, 479, 484, 485 , 493 , 5 01, Snyder, C., 5 3 , 66
5 10, 5 19, 5 23 , 5 25 , 5 26, 5 27, 5 28, 5 29, 5 3 0, 5 3 1, Snyder, W. M., 623 , 624, 63 2
5 3 4, 5 3 5 , 5 3 6, 5 3 7, 5 3 8, 5 69, 5 82, 5 83 , 5 98, 600, Sohn, M.-H., 281, 286
601, 608, 611, 613 , 614, 629, 63 0, 685 , 689, 696, Sohn, Y. W., 248, 249, 25 9, 261, 279, 280, 283 , 3 5 6,
700, 703 , 768, 785 3 5 7, 3 65 , 3 66, 3 68, 3 69, 3 71
Simon, J. R., 272, 286 Soloboda, J. A., 3 1, 3 8
Simon, Th., 163 Solodkin, A., 674, 682
814 author index
819
82 0 subject index
cognitive deficits in exemplars of high intellect, 5 96 Cognitive Psychology and Its Implications, 287
cognitive demands Cognitive Psychology by Neissen, 191
of operations, 5 3 cognitive reorganization, 626
of writing, 3 903 91 cognitive representations
cognitive development of experts, 5 0
of children, 75 8 mediating performance and continued learning by
importance compared to physical skill, 478 experts, 5 9
cognitive differences between experts and novices, 44 mediating skilled anticipation, 697
cognitive effort, retrieving domain knowledge and of musical structure, 463
strategies, 24 cognitive research on sport, 472
cognitive elements in naturalistic decision making, 414 cognitive resources
cognitive engineering. See also knowledge engineering driving hazard detection requirements, 648
emergence of, 186 increasing demand of bodily functions in older age,
foundational methods of, 208 73 5
Cognitive Engineering Based upon Expert Skill overloading of novice, 649
(CEBES), 25 2 pooling by teams, 442
cognitive expertise, 5 98 released by practice, 5 3
principal attributes of, 5 98600 cognitive science, 42
reaching the pinnacle of, 602 computational problem solving models and, 5 3 0
requiring experience, 3 6 perceptual-motor expertise and, 5 05
cognitive functions cognitive skill and expertise, study of, 14
content-free measures of, 724 cognitive skills
perceptual-motor expertise and, 5 08 combining with movement skills, 472
transferring expertise to some broader, 727 relationship between fundamental and higher order,
cognitive information processing, language of, 87 53
cognitive instruments, 5 74 for team sport experts, 482
cognitive involvement, 480 cognitive stage of perceptual-motor skill acquisition,
cognitive load, writers managing, 3 923 93 5 12
cognitive mechanisms cognitive strategies
adaptive abilities and, 614 in a learning outcome taxonomy, 78
case-study scenario and, 620 of writers, 3 93
expertise level combinations and, 640 Cognitive Strategies learning outcome, 80
musical knowledge and, 464 cognitive structures, 266
situation projections and, 63 6 cognitive systems, designing joint, 192
cognitive operations, 5 3 cognitive systems engineering, 193
cognitive performance, 5 49, 649 cognitive task analysis (CTA), 13 0, 177, 192193 , 204,
cognitive phase 229
of improvement in performance, 685 of air-traffic controllers, 3 67
of skill acquisition, 267 analyzing transcriptions of air traffic controllers, 3 61
cognitive plasticity, decreasing in later adulthood, 73 4 denoting a large number of different techniques, 192
cognitive probes in CDM, 192 era of, 206208
cognitive processes major issues remaining to be resolved, 192
acquired knowledge in a domain associated with novel systems a major challenge for, 192
changes in, 48 in reaction to behavioral task analysis, 208
associated with changes in performance, 23 0 review of methods, 213
chess players selecting superior moves, 23 2 from the study of instructional design and enhanced
creativity and, 761 human learning, 208
of designers and programmers, 3 74 of troubleshooting, 196
knowledge acquisition and executive, 616 understanding expert decision making in field
in knowledge acquisitions, 616, 625 settings, 192
in musical practice, 460 usability of the products of, 192
in self-regulation, 706 cognitive tasks
situation awareness information transformation by, automaticity and, 63 9
645 646 not directly addressed by Taylor and Gilbreth, 187
verbal-reporting procedures changing, 228 practice leading to functional decreases, 663
cognitive psychologists cognitive text process theory, 5 72
describing mechanisms responsible for superior cognitive training, increasing plasticity, 65 7
human performance, 83 cognitive traits, 148, 15 5 15 7
differences with software engineers, 192 cognitive units, larger and more integrated, 495 0
suggesting the information processing perspective, Cognitive work analysis, 13 8
82 cognitive work, independent of particular
cognitive psychology technologies, 215
collaboration with Computer Science, 42 cognitive/behavior adaptation, expertise as, 748
information processing language and computer cognitive/intellectual correlations with initial task
metaphor, 44 performance, 15 6
perceptual-motor expertise and, 5 06 cognitive-motor performance, systematic age-related
taking a turn toward applications, 205 declines, 726
832 subject index
cognitive-motor tasks, reduced speed or accuracy, 723 as psychological mechanism for superiority, 75 7
cognitivism, rise of, 78 in the SOC-model, 73 1
coherence compensatory behaviors of drivers, 3 5 8
as decision making perspective, 424425 compensatory mechanisms, 73 0
expected utility and, 425 compensatory strategies in expert performance,
historians providing, 5 74 73 1
as a narrative quality criterion, 5 74 competence
as process decomposition, 427 efficacy and, 444
using to appraise expertise, 425 expertise and, 762
cohort effects networks and individual, 75 7
as challenges to retrospective interviews, 296 overlap with expertise, 81
on the expected performance of an individual, 3 26 in sports, music, and chess, 687
co-incidence or co-construction of expertise, 299 stated goal often for ISD, 81
Colburn, Zerah, 5 62 strong positive correlation with years of experience,
collaborative process, knowledge elicitation as, 216 3 49
collectives transition to expertise, 297
decision making proficiency and, 43 6 competition
excluding female actors to a large degree, 117 of creative domains, 768
as a unit of analysis, 13 7 to enter medical school, 3 3 9
college students, excessive restrained by professionalism, 110
college-educated adults, words known by, 178 between professions, 75 4
color discrimination task, 666 compilation phase of skill acquisition, 267
colour vision, 5 5 5 Compiled level of expertise, 3 44
Columbia mission, 13 6 completeness
Comenius, Jan, 74 of knowledge, 178
commercial flying, expertise in a function of the as a narrative quality criterion, 5 74
aircraft, 3 5 8 complex abilities, developing, 724
commitment, 423 complex acquired movement, 672
common sense complex computation, brain system for, 5 63
in AI programs, 99 complex human activity, 43
large body of good enough, 99 complex systems, high fidelity simulations of, 243
commonalities, among abilities, personality, and complex tasks. See also tasks
interests, 15 9 decomposing into distinct subtasks, 278
commonly-held knowledge, 99 skill at, 276
communality, among predictors and trait complexes, subtasks as, 663
15 9162 complex units. See chunks
communication complexity
assessments of, 3 83 of environmental information and situation
aviation pilot situation awareness and cockpit, 643 awareness, 63 4
development of a formalized system for science, 115 as a situation awareness model feature, 63 5
errors by new platoon leaders, 646 component skills among experts, 73 3
of exceptional software designers, 3 80 componential training approach, 670
in expert teams, 443 , 446, 448, 449 composers
overload and psychomotor skills in novice, 644 of classical music, differential eminence of, 3 28
skills and experience, 640, 646 expertise acquisition in classical, 3 24
in the software design and programming domain, faster start for outstanding, 3 29
3 803 81 composite eminence measure for classical composers,
training in, 3 84 3 28
communities of practice, 128, 403 composite evaluation, 3 3 0
Army structured professional forums as, 624 composition instructors, 3 97
civilian organization sponsored, 624 compositional fallacy, 3 26
domain of interest tacit knowledge sharing by, 623 compositional preparation for classical composers,
expert standard definition by, 746 3 28
ordered world of, 13 4 compositions, 3 29
professional cultures as, 75 7 comprehension
research on, 624 coding for readers during, 3 92
for respective talent fields, 290 of conjunctions, 5 91
tacit knowledge and, 623 625 relationship to reading skills, 5 3
companies, power and organization, 75 4 situation awareness and, 646
CompanyComand.mil, 624 as situation awareness level, 63 4
comparison groups of a text, 3 91
absence of, 294 computation
possible created by key findings, 295 perceptual-motor skill acquisition and, 5 07
compatible mapping, 271 supporting intelligent behavior, 42
compensation as visual processing, 5 5 9
in extant frameworks of adaptive aging, 73 1 computational methods, describing human
by older experts, 73 1 performance with, 41
subject index 833
in the software design and programming domain, situation assessment and, 409
3 803 81 situation awareness comprehension, 646
training in, 3 84 court-appointed experts, 75 5
cooperative activity, technology mediating, 208 as expert witnesses, 75 5
cooperative work settings, 3 80 as relative experts, 746
coordination roles of, 75 5
expert teams and, 442, 449 status authority of, 75 5
expert team shared mental models and, 446 use of, 75 5
improving as skills are refined over time, 25 1 Covering Law, 5 71
of medical knowledge, 3 463 47 Covering Model, 5 71
perceptual-motor expertise and, 5 16 covert self-regulation, 706
corporate knowledge management, 217 Cox, Catherine, 3 21
corrective actions for malfunctioning devices or CPCs (certified performance controllers), 3 61
processes, 94 CPD (Continuing Professional Development), 111
correlates, inferring ability from, 5 89 craft guilds, 7475 , 203
correlation analyses for classical composers, 3 28 crafts, skilled, 6
correlational data analyses, 3 22 craftsmen, 5 , 74
correlational data, statistical techniques suitable for creative accomplishment, expertise and, 762
the analysis of, 3 3 2 creative achievement, 785
correlational method, lacking causal inference, 3 3 1 creative activities, role of deliberate practice, 693
correlations creative advances
attenuating between measurements, 15 5 domain specificity-general mode transfer, 765
maximizing between predictors and criteria, 15 7 as expert redefined domains and, 783 785
corroboration as a historical source heuristic, 5 72 expertise in real world, 764
cortex, faces areas in, 668 performance standards and, 783
cortical activity in real world settings, 764
consistency and practice modulating, 660 technique and, 782783
functional connectivity studies of correlated, 671 techniques and skills in, 762
modulating, 65 6 creative development, 3 28
cortical areas creative domains, curvilinear function seen in, 3 3 0
changing and adapting function, 283 creative expertise, 3 20
very different tasks activating the same, 660 creative intelligence, 616
cortical plasticity creative output, quantity and quality of, 3 20
of normal elderly, 65 7 creative performance, 3 29
on a slower time scale through extensive training, creative process, 761
662 creative productivity, 3 20
cortical reorganization in musical experts and novices, creative products
465 double helix domain specificity expertise and, 776
cortical representation, increased, 674 valued and, 763
cortical tissue, increasing for a task, 65 5 creative solutions, 27
cost of failure (c), 190 creative thinking. See also thinking
cost savings of expert systems, 94 Calders domain specific expertise in, 774
Coughlin, Natalie, 709, 712 case studies, 769780
counselors as cognitive, 761
categorizing based on abstract information, 175 critical vs. random, 771
categorizing client statements, 175 Darwinian theory of, 771
listening to a counseling session, 174 deliberation and, 767
novice, 175 double-helix model and, 775 776
counter-elites Edison light bulb development, 779780
emergence of, 119 evolution of, 771
role in the generation of cultural change, 119 expertise and, 762
counterfactual reasoning, historians use of, 5 79 expertise facilitation of, 768
counterfactuals, 5 80 expertise in, 761787
counting expertise modes and Wright Bothers, 779
prodigious abilities growing out of, 5 5 4 information and, 782
stages in the development of, 5 5 9 knowledge and habit in, 767
course of action (COA) in scion and technology, 775 780
experience and, 409 ten year rule and, 768769
expert recognition and, 410 of Wright Brothers, 776779
generation of, 410 creative thought, 75 8
mental simulation of, 406 creative writers, 3 95 , 3 99
mental wargaming and, 410 creativity
in military decision making, 410 arising from chance and unique innate talent, 22
natural production of, 410 decision expertise scholarship and, 429
preferred, 411 as decision option expertise, 43 1
prototype linked to, 406 definition, 761, 762
quality of, 410 deliberate practice and, 768
836 subject index
expertise working memory. See ExpWM recording techniques and occlusion studies, 476
expertise-driver specific abilities account, 727 search patterns of skilled performers, 476
expertise-related mechanisms, 73 0 simulated by CHREST, 5 27
expert-lay dichotomy of surgeons laparoscopic simulation, 25 1
expert-interaction and, 747 vs. external environment, 5 11
knowledge gradient and, eye-hand spans of older skilled typists, 73 1
as relational notion, 746
expert-novice differences face inversion, prosopagnosia patients not impaired in,
in dance, 499 668
features of expert-interaction in, 746 face-like expertise, developing for a non-face object
leading directly to new methods of instruction, 46 category, 676
musician tonality recall and, 463 faces
experts. See also memory experts; older experts inverted activating object-sensitive regions, 668
explanations processing, 667668
coding methods for, 177 same race, 668
compared to thinking aloud, 228 tests for, 5 45
expert experience use in, 75 8 treated like objects by object processing regions, 668
by expert systems, 9798 working-memory task, 662
insuficiency, 204 facial expression, emotional experience and, 493
of the line of reasoning, 93 facilitative trait complexes, 15 9
verbal reporting as, 176 factor analysis, 5 89
explicit awareness, sequential learning not dependent factor analytic studies, 5 89
on, 274 failures
explicit concrete entities, novices solving a problem on attention in perceptual-motor expertise and, 5 13
the basis of, 181 essential to the development of expert levels of skill,
explicit-instruction training groups, contrasted, 25 7 45
exponential law of practice, 267 of experts, 23 , 5 6
expository writing, 5 74, 5 75 likely to arise in deliberate practice, 698
expression in dance, 5 00 viewing as opportunities to improve, 601
ExpWM (expertise working memory), 600, 604 false associates, activating by way of other problems,
abilities as distinct from Gf, SAR, and Gs, 603 280
abilities indicated in displays of expertise, 605 families
as different from STWM and memories of SAR, 600 of Calder as artistic, 774
indicative of intelligence, 605 expertise socialization and, 75 6
level-of-expertise-by-age interaction for, 604 of German musicians, 75 6
measures of the ability traits of, 603 mental development and bourgeois, 75 6
reliable age-by-expertise interaction for, 604 music societal factors and, 466
extended Gf-Gc theory. See Gf-Gc theory musical abilities and, 45 745 8
extended practice. See also deliberate practice as subcultures for expertise, 75 6
leading to improvements in performance, 3 1 support by, 13
PRP effect persisting across, 277 family background, world-class expertise emerging
refining and improving rules as a function of, 479 from, 3 27
extended training, 61 family circumstances, influencing the acquisition of
extensive experience extraordinary expertise, 3 27
of activities in a domain, 683 family influences, providing early experiences and
necessary to attain superior expert performance,687 motivating learning, 298
extensive watching, not the same as extensive playing, fast learning phase of M1, 671
691 Faulkner, William, 713
external demands, performing in response to, 687 faults. See also errors
external supports, elimination of, 706 considering possible, 193
external variables in Carrolls system, 79 detecting in writing, 3 90
extreme base rates, problem of, 15 4 diagnosing, 94
extrinsic rewards, writers creativity and, 3 95 feather, sharpening for writing with ink, 6
extroversion, decision vigilance and, 429 features, identifying, 268
Extroversion personality trait in the Social Trait feedback
complex, 15 9 in adaptive expert teams, 442
eye fixations in decision skills training, 412
measured by Chase and Simon, 5 26 in deliberate practice for writers, 3 96
recording and analyzing, 23 3 in expert teams, 446448
sequences of, 229 expertise development and, 705
eye movements in the ISD process, 81
of chess experts, 5 25 5 26 for motor control, 273
data indicating expert search strategies, 246 perceptual-motor skill learning and, 5 06, 5 08
developments in the recording and analyses of, 471 required for deliberate practice, 601
of experienced vs. inexperienced drivers, 3 62 responding to, 5 115 12
by flight instructors versus student pilots, 25 0 situation analysis information development and, 63 6
of music instrumentalists, 465 vital role of, 45
recording in an action component, 246 to writers from composition instructors, 3 97
848 subject index
forward-span memory, negative age relationship for, scenarios and recall, 478
5 93 time constraints in, 473
forward-working search strategy Gamm, Rudiger
experts representation characterized as, 169 active brain areas, 5 60
used by the physics expert, 177 brain activity of, 5 64
frame. See structured object brain of, 675
frame theory, 178 brain regions used by, 5 65
frameworks, 13 413 6 calculating prodigy, 5 5 7
Franklin, Benjamin, 3 97 learning to use LTWM facility, 5 5 9
Frasca 142 flight simulator, 3 64 memory specificity, 5 60
fraud detection, 23 5 neural network for calculation processes, 5 64
free recall, 171 on practice, 5 61
Freud, Sigmund, 615 practice and, 5 61
fronto-parietal networks as self-taught, 5 60
supporting performance of routine numerical tasks, study of, 5 5 4
675 visual processing computation and, 5 5 9
use in numerical tasks, 5 63 gatekeepers, 745
frustration, skill demands and, 3 95 Gates, Bill, 14
Fuller, Thomas, 5 5 7, 5 61 Gauss, Carl Fredrich, 5 5 4
functional fixedness, 27 gaze-contingent paradigm for chess player perception,
functional hierarchical representation, 195 5 25
functional magnetic resonance imaging. See fMRI Gc abilities
functional organization in brain activation, 65 3 in the Intellectual/Cultural trait complex, 15 9
functional reorganization no decline or improvement of with aging, 5 93
of brain areas, 65 5 Gc (acculturation knowledge), 5 90
versus process efficiency, 662 abilities increasing with acculturation, 605
functional validity of behaviors, 3 13 correlating with the educational or economic level,
functions 5 92
behavioral trait fluctuations in, 5 88 development of associated abilities, 5 92
behaviorally valid, 3 13 as dilettant breadth of knowledge of the culture,
expert performance as, 743 604
expertise as knowledge of, 747 improvements for some individuals with age much
named in a production rule, 92 larger than for others, 5 95
fusiform cortex, 65 6 increasing in adulthood, 5 95
Fusiform Face Area. See FFA indicating dilettante breadth of knowledge, 605
future events measures of, 5 97
experts ability to anticipate, 246 operational definition not adequate, 5 97
projection as situation awareness level, 63 4 security conducive to the development of, 5 92
Gc (crystallized intelligence), 3 2, 161
g (general intelligence), 3 2, 5 91 characterization of, 617
characterization, 616 components within a person, 3 2
discovery and measurement of, 5 91 correlating with Gf, 3 2
as a factor at early stages of skill acquisition, 725 domain general tacit knowledge inventories and,
heritability for, 724 621
as missing, 5 91 instruments for measuring, 3 2
practical intelligence and, 616, 620 practical intelligence and, 616, 621
tacit knowledge and, 621 tacit knowledge and, 621
Ga (auditory processing), 5 90 gender
Gagne, 80 accounting for performance differences in sprint
Galton, Sir Francis events, 481
attempts to measure a generalized, inheritable scientific expertise perspective of, 117
intelligent quotient, 71 in self-regulatory training, 715 716, 717, 718
criteria of eminence, 5 5 3 general ability, importance of, 616
first behavioral scientist to publish a truly influential general expertise
historiometric study, 3 20 creativity in Edison and Wright Brothers, 780
hypothesis of a general superiority for experts, 10 as mechanical in Wright Brothers, 770
on inherited abilities, 5 5 6 in Wright Brothers flight control development,
innate biological capacities limiting an individuals 779
potential, 684 general intelligence. See g
innovation setting the groundwork for empirical General Problem Solver, 11, 42, 90
studies of thinking, 224 general systems theory in military problems, 77
on natural ability and mathematical expertise, generalists, 46
555 generalized reasoning ability. See Gf
precursors of exceptional achievements, 724 generate and test weak method, 43
games genetic endowment, relevance of, 3 27
chunking of arrays in, 171 genetic inheritance, as a relevant component for
presenting situations to chess players, 23 2 expertise in music and sports, 22
850 subject index
Institute for Human and Machine Cognition, 212 intelligences; naturalistic intelligence; practical
Institute of Social Research (ISR), 3 04 intelligence; psychometric intelligence; working
institutional structures, scientific knowledge intelligence
demonstration and, 115 ability in mathematics and, 5 5 65 5 7
institutionalization as-process, 161
of expertise, 105 , 114, 73 6 as-reasoning, 3 3
of scientific expertise, 115 calculator ability and, 5 5 7
institutions, professions as, 108109 chess skill and, 5 3 3
instruction cognitive skills covered by, 87
dance formal, 498 computational device approaches to, 43
early availability of, 13 creative, 616
individualized, 79 denoting stable, interindividual differences, 724
instruction classes and teachers in, 75 extended theory of fluid (Gf) and crystallized (Gc),
of ISD movement, 81 5 88
in knowledge acquisition, 625 hierarchical model of, 3 2
learning fit and, 83 inheritance of, 3 21
in mathematical expertise, 5 61 integrating trait theory of with theory of expertise,
programmed, 77 5 88
in Socratic context, 71 as lacking in self-taught calculators, 5 62
systematic design of, 79 memory and, 5 475 48
teachers and trainers, 79 as a reasonably good predictor of performance early
theatrical forms and, 491 in learning, 3 2
time spent on, 289 role of, 3 2
instruction design by domain experts, 81 successful use of, 3 4
instructional design tests designed to measure abilities of, 606
pioneers in the development of, 14 theory of, 5 87
research projects on, 204 working, 75 8
instructional sequence, student self-assessment as, 77 intelligence research, pioneers of, 724
instructional systems, 81 intelligence tests, tacit knowledge and, 621
of Bloom, 79 intelligent behavior
development, 77 artificial methods for producing, 42
experts and, 81 child thinking skills instruction and, 626
by intelligent tutoring, 46 intelligent systems
Instructional Systems Development movement. See creation, 217
ISD movement interfaces for, 213
instructional techniques for tutoring, 46
medieval, 74 intensity. See quality
systematic nature of sophist, 72 intentions
instructivist perspectives versus constructivist, 83 actor identification of character, 492
instructors, changing the role of, 70 actor long-term memory and, 496
insurance companies, 3 83 actor performance and, 492
integration actor script segmentation and expert chunks, 493
of experts representations, 180181 as key decision feature, 423
as interaction of features, 180 interaction analysis, 13 0, 141
intellect, adaptive, 617 interaction patterns between people, 207
intellectual ability. See cognitive abilities interests
intellectual and cultural activity interests, 3 4 clusters of, 3 4
intellectual capabilities, 5 885 98 expertise and, 3 4
intellectual capacity, mathematical expertise and, matching with job characteristics, 15 8
5 64 talents channeled by, 3 4
intellectual development, investment theory of, interference
15 9 attributing to different stages of processing, 664
intellectual endeavor, tasks captured in an expert in a dual-task environment, 676
system, 88 dual-task specific processing and, 664
intellectual skills related to strategy choice, 666
acquisition of, 5 06 interindividual differences, 147, 727
in a learning outcome taxonomy, 78 factors leading to changes in, 15 1
vs. perceptual-motor expertise, 5 065 08 individual kinship differences, 73 7
Intellectual Skills learning outcome, 80 during learning or skill acquisition, 15 1
intellectual stage of perceptual-motor skill acquisition, practice reducing the range of, 3 1
5 12 intermediate levels
Intellectual/Cultural trait complex, 15 9, 160 acquisition by future experts, 62
intelligence. See also academic intelligence; AI; non-experts and, 179
analytical intelligence; creative intelligence; g intermediates
(general intelligence); Gc (crystallized medical student recall as, 3 41
intelligence); Gf (fluid intelligence); performance assessment by, 408
logical-mathematical intelligence; multiple situational assessment by, 409
856 subject index
mental operations, including as part of the description method of tough cases, 205
of learning, 78 methodical orderliness of human activity, 13 4
mental realm, researchers progressively encroaching, methodological artifacts, 3 25
44 methodological issues of jurors, 13 3
mental rehearsal methodologies
dancer movement encoding, 499 benefiting from opportunism, 217
dancer pattern use of, 499 formal experts and, 75 2
mental representations. See also representation(s) importance of convergence of findings across, 296
functions of, 5 6 relationship to research questions proposed, 296
of historians, 5 725 73 studied in a workplace, 13 3
instrument implementation plan and, 464 metrics, used by social scientists, 141
musical performance and, 463 Mickelson, Phil, 63 4
for performance and continued learning, 696698 micro level for time spent in an activity, 3 03
of prototypical movements, 499 microanalysis, 714
of readers, 3 91 microcognition in naturalistic decision making, 414
triangular model for musicians, 464 microcomputer chess programs, drawing matches with
mental resources. See also resources the best human players, 5 25
automaticity and situation awareness, 63 9 micro-game simulations of team sports, 25 7
decision chores and, 43 1 microscopic pathology, experts encoding essential
mental set fixedness, 27 information more accurately, 23 4
mental simulation Microsoft Flight Simulator, 249
as even sequence envisioning, 406 middle ages, expertise in, 75
in juror decision making, 43 3 middle-school students, 626
in naturalistic decision making, 406 military
mental walk along a well-known route, 5 40 commanders experience with recognition-primed
mental wargaming in military decision making, 410 decision models, 411
Mentice Procedicus, 25 4 company commanders CompanyComand.mil as
mentoring, 628 Army forum for, 624
mentors, influence of domain-specific, 3 24 enlisted men and women performance predictions,
MER (Mars Exploration Rover) 33
application of the rock abrasion tool, 13 4 intelligence and information techniques, 645
improving mission work processes, 13 2 jobs, 77
mission study limited by the number of observers, officers information skills and experience, 640
142 social background of, 75 7
rover operations, 13 9 tacit knowledge for leadership, 620, 622
science and operations support teams, 13 213 3 training, 78
merit, 118 military decision making. See also army command and
merit principle, 119 control; decision making; platoon leaders
Merton, Robert K., 115 cognition in, 410, 411
meta-analysis of sports expertise findings, 482483 naturalistic, 409412
metacognition, 5 5 rationale of, 410
as automatic, 5 7 situation awareness and, 644
important to test understanding and partial Military Decision Making Process (MDMP), 409, 410,
solutions, 5 6 411, 412
within the information processing model, 5 5 Military Leadership Case-Study Scenario, 620
in naturalistic decision making, 406 Military Operations in Urban Terrain. See MOUT
self-observation processes and, 711 facilities
metacognitive activity, 5 7 military pilots, situation projection by, 641
metacognitive knowledge, 5 7, 3 79 Miller, George, 191
metacognitive self-monitoring, 711 Miller, Robert B., 188, 189
metacognitive skills mind
in decision skills training, 412 computer metaphor of, 5 09, 5 10
of music learners, 464 multiple representations in, 3 89
musician self-regulation and, 461 mine detection clearance operations, 25 2
metacognitive strategies, 5 7 minimal access training, 25 4
Meta-DENDRAL learning program, 91 minimal invasive simulation trainers, 25 4
meta-level knowledge in an expert system, 96 Minimal Invasive Surgery Trainer in Virtual Reality
metaphorical imagery, dancers and, 5 00 (MIST-VR), 25 4
metaphors, reasoning with, 5 94 Minimal Invasive Surgical Trainer (MIST), 25 4
method acting, 490 mirror neurons, studies of macaque, 672
method of loci mirror system
brain activity during training in, 5 48 coding for complete action patterns, 672
effect of training in on delayed serial recall in the expertise specificity of, 672
elderly, 5 49 misconceptions
as a memory retrieval structure, 5 47 about simulation and training, 25 8
use by memory experts, 5 48 about the brain and expertise, 65 7
used by Shereshevskii, 5 41 factors contributing to for medical practitioners,3 43
subject index 865
motor task practice, leading to functional increases of compositional training for classical composers, 3 28
activation, 663 cues in dance, 5 00
motor tracking task, brain activation as a function of deliberate practice and, 693
practice in learning, 65 4 deliberate practice and proficiency, 45 9
motorization of transport, civil economy, and war, 186 deliberate practice improving, 23 7
Mouillar, L. P., 778 deliberate practice related to high performance,
MOUT facilities, 243 3 83
move sequences, memory for in blindfold chess, 5 3 1 development of technique, 466
movement expert performance in, 15
actor recall and, 496 expertise, 45 7470
central role in sports, 473 genres, 45 8
cerebellar control of, 5 08 as grammar-based non-semantic temporal
combining with cognitive skill, 472 phenomenon, 467
encoding by dancers, 499 as highly effortful, 460
execution, 671 home environment and excellence in, 45 8
memory enhancement and physical, 497 increased cortical representation associated with,
mental representation of prototypical, 499 674
production, age-related declines, 726 Indian and Mid Eastern improvisation and problem
sequences, 498, 5 09 solving, 466
skill inherent in world class sport performances, 472 individual achievement differences in, 45 745 8
time, 473 innate abilities vs. environmental factors, 45 8
between two or move athletes, 473 knowledge, 463
moves. See chess moves laboratory task capturing superior performance in,
movie directors 688
hierarchical linear modeling assessing, 3 25 metacognitive and self regulation skills of learners,
recent historiometric study of top, 3 3 0 464
movies. See films Mozarts training in, 770
Mozart, Wolfgang Amadeus non-European genres, 466
case study of, 769770 perceptual processing and knowledge of, 463
expertise and creativity in, 781 performers mastering music considered unplayable
expertise investigation, 45 7 in the 19th century, 690
expertise research on, 45 7 phenomenological learning account, 462
music expertise domain redefinition and, 784 practice and performance in, 45 845 9, 462
in a musical household, 5 62 practice and styles, 460
Picasso similarity, 772 practice skills of experts, 461
surpassing the technical virtuosity of, 690 practices hours and instrumental, 460
ten year rule and, 462, 768 proficiency of experts, 467
Mozartians, 3 93 psychological research and, 467
MRI, 5 48. See also fMRI skill acquisition model, 462
multi level perceptual learning, 667 societal factors in performance of, 466
multidimensional scaling algorithms. See MDS style recycling in, 783
Multinational Time Use Study, 3 04 music composition
multiple cognitive ability tests, 627 case studies of, 769772
multiple intelligences equal-odds rule and, 771
Gardners popular theory of, 5 5 4 practice vs. talent in Mozart, 769
in school performance enhancement program, quality in, 771772
626 ten-year rule and, 689
multiple perspectives, principle of, 13 613 7 music instrumentalist, perceptual-motor adaptation,
multiple players, naturalistic decision making and, 465
403 music learning, practice and performance level of
Munsterberg, Hugo, 186 instrumental, 45 9
Murray, Donald, 710 music practice. See also deliberate practice
muscles ability difference compensation by, 45 9
compared to the brain, 65 7 instrumental music learning and, 45 9
fibers, 695 as investment of effort, 45 8460, 462
training, 675 medical problems of musicians and, 465
muscular-skeletal problems, musicians and, 465 methods improvement, 466
music musical performance and, 45 8462
the Beatles early, 771 musical performance role of, 45 8
age-comparative studies, 728 musical styles and, 460
aptitude tests, 45 7 as necessary for high-level achievement, 45 8
attainment and accumulated practice, 45 9 perceptual-motor skill adaptation, 465
autistic savant knowledge and, 463 as predictive of success, 460, 5 11
characteristics of experts in, 3 05 quality enhancement of, 460
cognitive adaptations of experts, 463 464 skill maintenance through continuous, 462
cognitive representation and, 463 stages of, 461462
compared to chess, 697 as systematic activity, 461
composition of classical, 3 28 as time investment, 45 8460
subject index 867
political fragmentation, exceptional creators likely to practice. See also deliberate practice
develop, 3 28 the Beatles and, 770
political interviews of historians, 5 81 in academic learning, 711
political science activation increases and decreases, 661
as an expertise domain with ill-structured problems, actual active less than reported, 3 08
5 70 adaptive inferences during, 713
as an ill-structured domain, 5 69 age and efficiency, 45 9
problem representation in, 5 78 age leading to optimal and efficient methods,
solving of ill-structured problems in, 5 78 73 4
time use literature on, 3 05 age-based interactions with, 481
politicians as relative experts, 745 by Calder, 774
politics, public policy experts and, 75 5 changing mediating mechanisms, 14
Pollock, Jackson, 774775 , 784 in chess mastery, 5 3 2
polygons consistent, 660
determining whether identical or not, 279 dance technique as skill derived from, 497
illustrations of, 280 disciplined, 709, 718
polymath, 72, 76 domain-specificity of in professional contexts,
Ponomariov, Ruslan, 5 24 73 3
positions, jobs consisting of, 187 effects on dual-task performance on experts, 5 3
positivism, Covering Law and, 5 71 expertise development and, 705
possibilities exponential law of, 267
as cardinal decision issue, 43 2 exposure to particular exemplars and, 3 45
stress and neglect in decision making, 43 2 extreme amounts on a circumscribed set of tasks,
posterior parietal cortex (PPC), 65 6 53
posterior right hippocampal grey matter volume, hours required, 207
correlated with taxi driving, 673 importance of, 3 1, 480482, 706
post-industrial educational model, 75 massed over space, 5 06
power in mathematical expertise, 5 615 62
of expert systems as knowledge, 100 mathematical expertise and, 5 64
of knowledge, 90 as means to automaticity, 5 3
scientific expertise intertwined with, 117 memory elements strengthened by, 5 60
Power Law of Learning, 5 10 memory superiority and, 5 45
power law of practice, 267 need for repeated experiences, 45
PPIK theory, 15 9, 161 overestimation of, 3 07, 3 08
practical abilities, expertise as, 72 perceptual-motor expertise and spaced, 5 06
practical approaches in work settings, 3 83 3 84 in Picasso creative thinking case study, 772
practical intelligence power law of, 267
case-study scenarios assessments, 619620 practice vs. talent in Mozart, 769
characterization, 616 process distinction, 13 5
as critical in everyday life, 615 profound effects of extended focus, 5 9
crystallized intelligence and, 617 quality and quantity of, 716
distinctiveness, 621 research on, 5 3
domain general tacit knowledge inventories and, 621 schedules for motor control, 273
expertise and, 613 63 2 self-directed, 714
expertise enhancement and, 623 627 self-enhancing cycles of, 707
expertise research and, 614 as self-regulation, 705
future research on tacit knowledge and, 627 self-regulatory methods during, 714
general intelligence and, 616 shift from attention-demanding controlled
improvement, 626627 processing to more automatic mode, 282
measurement and, 618620 solitary, 705
middle school student degeneration in, 626 structuring of, 705
personality and motivation and, 617 tasks and mappings, 271
psychological constructs and, 616617, 621 technique focus of, 709
reflection techniques in tacit knowledge acquisition understanding as a skill acquisition variable, 3 05
and, variable as ineffective, 660
research findings, 620623 working memory and, 661663
research on, 3 2 by writers, 3 97
tacit knowledge as enabler in, 615 practice activities
tacit knowledge currency and, 625 age-related constraint on, 73 5
tacit knowledge enhancement by, assessment of, 3 14
tests of, 618 changing states into complex states, 694
triarchic theory and, 616 isolating to meet all the criteria for deliberate
practical problems, tacit knowledge importance and, practice, 693
622 mediating improved physiological function,
practical thinking 695 696
academic achievement, 626, 627 necessary to improve performance, 60
skills development program, 626 rating for wrestlers and figure skaters, 3 07
876 subject index
quality readers
in decision making, 423 427 awareness of, 3 94
expert certification of, 75 4 poor versus skilled, 671
measures of, 3 14 text comprehension of, 3 91
in music composition, 771772 reading
musical practice, 460 brain areas supporting, 670
of service value, 111 inferior frontal and ventral fusiform regions as a
quantitative analyses, 3 19 function of, 671
quantitative changes, 266 as a knowledge predictor, 3 973 98
quantitative knowledge. See Gq relationship to comprehension skills, 5 3
quantitative measurement, 147 Realistic interest personality trait, 15 9
quantitative measures, 3 13 3 14 realization problem, 42
quantitative methods, 187 real-life decisions
quantitative scale, 3 24 as cardinal decision issue, 427
Quenault, 3 5 9 relying on analogical reasoning and schematic
questions techniques, 3 3
abstract, 25 real-world demands
asking to elicit expert knowledge, 213 capturing, 246
concrete, 25 reproducing in a standardized setting, 25 0
direct, 177 real-world domains
interview, 176 creative thinking in, 764
in interviews, 177 studying expertise in, 170
open-ended, 177 real-world perceptual characteristics, 245
private, 73 real-world performance
probe, 209 improving via simulation, 25 7
research, 292 usefulness of training under simulated conditions in
why, 23 0 improving, 25 8
Quetelet, Adolphe, 3 203 21 real-world tasks, future studies using more complex,
quick diary log. See stylized activity list 3 82
quiet eye periods, 476 reappraisal of professionalism, 110111
reasoning. See also medical reasoning
racecar drivers, 3 5 9 by analogy in chess, 5 3 2
racial differences, 45 7 blackboard model of, 92
radiologists, 172, 174 causal, 5 795 80
railway motormen, 186 chains used by radiologists, 181
Rajan Mahadevan, 5 425 43 , 5 45 , 5 46 by a computer, 87
Ramanujan, Srinivasa, 5 61 dependent on knowledge, 48
random chess moves, recall in blindfold chess and, development of new methods for different kinds, 96
531 domain-general or global strategy, 167
random chess positions, recall of, 5 27 in early learning, 15 6
random processes engine, 91
creative thinking evolution and, 771 expertise residing in the power of methods, 90
music composition quality and, 771772 experts graceful in, 5 5
range, restriction of, 15 3 15 4 by historians, 5 775 80
rank order neurons, 672 IF-THEN rules and, 92
rapid chess games, grandmasters rapid play quality,5 29 mathematical, 618
Rasmussen, Jens, 188, 208 methods of knowledge engineering, 91
rating system models creation, 209
of chess, 5 24 novice performance limits and, 5 7
for chess, 5 24 separation from knowledge, 48
rational behavior, normative model of, 404 skill as predictor, 73 2
rational-analytic theories in military decision making, strategies of experts, 215
409 tests of, 606
rationalist paradigm types of events occuring for effective, 5 8
as cognitive, 404 with uncertainty, 93
of decision making, 404 under uncertainty, 96
rationality as applying knowledge, 13 6 weak methods of, 5 77
reaction time (RT) reasoning abilities, 23 , 5 90
interval, 475 adult intelligence and, 605
older adults slower, 5 94 high levels of, 5 99
reaction times, 174, 473 recall
reactive consequences of extensive verbal descriptions, actual performance insight and, 245
228 concrete versus abstract language and, 3 92
reactive effects of instructing students to explain as a contrived task, 171172
performance, 23 0 dancer music cues use by, 5 00
reactivity delayed, 5 43
avoiding the problem of, 224 expert-novice differences paradigm, 181
of verbal reporting, 227 of experts, 600
subject index 881
changing during an experimental session, 23 1 subject matter expertise, decision making expertise
developed to satisfy task goals, 282 and, 426
differences in experts and novices, 3 67 subject matter experts
discrepancies between observations and reported, instructor as, 70
223 judgment accuracy of, 43 2
employed by experts across divergent scenarios, substantive decision making procedures and, 43 3
25 7 value issue proficiency and, 43 4
experts selecting fewer, 3 68 subject matter knowledge of historians, 5 81
flexibly used by experts, 675 subjective ratings, using during practices to evaluate
managing the cognitive load on working memory, quality, 3 14
3 99 subjectivity
metacognitive, 5 7 of activity, 3 12
more appropriate chosen by experts, 24 decision making expertise research and, 423
shifts and skill acquisition discontinuity, 268 subject-performed tasks (SPT)
validity of general descriptions of, 23 1 actor recall and, 496
street experts, inflexibility in the use of strategies, 26 dance movements and, 499
stress sub-optimal moves, diagnosing the source of, 697
attention in decision making and, 43 2 subordinate category, 176
impact on experts vs. non-experts, 3 82 subordinate level
musical performance from memory and, 463 experts categorizing at, 176
as situation awareness model factor, 63 5 objects, 179
situation diagnoses and decision making subsequent learning, 80
performance, 443 subsidiary study by Roe, 294
strong methods, 43 substantive variables, historiometrics study of
in AI research, 90 composers and, 3 28
providing certainty, 5 77 subtasks of a dual-task, 663
Stroop-like interference task, 5 26 subtext, 5 72
structural changes generation by historians, 5 81
in brain tissue size, 65 3 by historian specialists, 5 73
music training inducing in the brain, 674 success
structural equation modeling, 728 ethnic group and social, 75 7
structural game sequences, 478 as a poor predictor, 3 41
structural research summary statistics, 141
of abilities indicative of intelligence, 5 88 superior memory
on expertise, 5 98 evidence of, 5 46
Gf-Gc theory and, 5 895 92 main methods used in the study of, 5 40
structured interviews most striking examples as strategy-dependent,
methods of, 205 5 46
verbal reporting as, 176 neurological basis of, 5 485 49
yield of, 206 organization and, 244
structured objects, 92. See also schema scientific study of, 5 40
Strumilin, S.G., 3 04 superior performance. See also reproducibly superior
students performance
achievement variation in, 79 as domain specific, 10
aptitude, 78 mechanisms identification, 49
expertise in, 79 objective reproducibility of, 687
intelligent tutoring systems use by, 46 psychometric factors, 49
knowledge, 211 social and experience-based indicators and, 686
medical, 25 superiority as psychological mechanism, 75 7
peer feedback incorporation by, 26 superiority of expertise, limited to a specific domain,
practical intelligence development program, 626 25
practice implementation by, 706 superiority of experts, found to be specific to specific
preparation, 298 aspects, 10
self-views of, 289 superordinate categories, 175
study environment for, 711 supervision of beginning musicians, 461
teaching to work like experts, 297 supervisor ratings, tacit knowledge and, 622
study environment, 711 supervisory control, 188
study methods, consistent with deliberate practice characterized by monitoring displays, 186
predicting achievement, 699 resource management and, 3 62
STWM. See short-term working memory unobservable cognitive activities of, 189
styles supervisory role, knowledge and cognition importance
of acting, 489 in, 188
expressing prewriting strategies as, 3 93 supplementary motor area. See SMA
recycling in art, 783 supply side theory of professionalism, 109
stylized activity list, 3 09 support, required for expertise, 3 5
subassemblies, 94 surface features, undergraduate problems sorting,
subcomponents, 282 175
892 subject index