Aaron S. Benjamin, J. Steven de Belle, Bruce Etny

HUMAN LEARNING
Biology, Brain, and Neuroscience

ADVANCES
IN
PSYCHOLOGY
Volume 139
Editor
M. GUADAGNOLI
HUMAN LEARNING
Biology, Brain, and Neuroscience
Edited by
Aaron S. BENJAMIN
University of Illinois, USA
J. Steven DE BELLE
University of Nevada, USA
Bruce ETNYRE
Rice University, USA
Thad A. POLK
University of Michigan, USA
Amsterdam • Boston • Heidelberg • London • New York • Oxford • Paris

San Diego • San Francisco • Singapore • Sydney • Tokyo
North-Holland is an imprint of Elsevier
North-Holland is an imprint of Elsevier
525 B Street, Suite 1900, San Diego, California 92101-4495, USA
The Boulevard, Langford Lane, Kidlington, Oxford OX5 1GB, UK
Radarweg 29, PO Box 211, 1000 AE Amsterdam, The Netherlands
First edition 2008
Copyright 2008 Elsevier Ltd. All rights reserved
No part of this publication may be reproduced, stored in a retrieval system

or transmitted in any form or by any means electronic, mechanical, photocopying,
recording or otherwise without the prior written permission of the publisher
Permissions may be sought directly from Elsevier’s Science & Technology Rights
Department in Oxford, UK: phone (+44) (0) 1865 843830; fax (+44) (0) 1865 853333;
email: permissions@elsevier.com. Alternatively you can submit your request online by
visiting the Elsevier web site at http://elsevier.com/locate/permissions, and selecting
Obtaining permission to use Elsevier material
Notice
No responsibility is assumed by the publisher for any injury and/or damage to persons
or property as a matter of products liability, negligence or otherwise, or from any use
or operation of any methods, products, instructions or ideas contained in the material
herein. Because of rapid advances in the medical sciences, in particular, independent
verification of diagnoses and drug dosages should be made
British Library Cataloguing in Publication Data

A catalogue record for this book is available from the British Library
Library of Congress Cataloging-in-Publication Data

A catalog record for this book is available from the Library of Congress
ISBN-13: 978-0-444-52080-7
ISSN: 0166-4115 (Series)
For information on all North-Holland publications

visit our website at elsevierdirect.com
Printed and bound in Great Britain
08 09 10 11 10 9 8 7 6 5 4 3 2 1
Working together to grow

libraries in developing countries
www.elsevier.com | www.bookaid.org | www.sabre.org
v
Contents
Preface ix
Contributors xi
Part I: Human Learning and Cognition
Introduction: Behavioral Approaches to the Study of Human Learning

and Memory 3
A. S. Benjamin
The Role of Inhibition in Learning 7

J. C. Hulbert & M. C. Anderson
Short- vs. Long-Term Memory 21

I. Neath & A. M. Surprenant
Hemispheric Asymmetries in Verbal Memory 33

K. M. Evans & K. D. Federmeier
Emotional Facilitation and Disruption of Memory 45

S. L. Warren, G. A. Miller, & W. Heller
Scheduling and Learning 61

D. A. Simon
Part II: Cognitive Neuroscience
Introduction: Cognitive Neuroscience of Learning and Memory 75

T. Polk
The Computational Cognitive Neuroscience of Learning

and Memory: Principles and Models 77
L. I. Newman & T. A. Polk
vi Contents
Cognitive Neuroscience of Skill Acquisition 101

J. Bo, J. Langan, & R. D. Seidler
Cognitive Neuroscience of Declarative and Nondeclarative Memory 113

P. J. Reber
Learning and Memory for Emotional Events 125

A. S. Atkins & P. A. Reuter-Lorenz
Age Differences in Memory: Demands on Cognitive Control

and Association Processes 137
C. Lustig & K. Flegal
Part III: Human Motor Learning
Introduction: A Survey of Motor Learning Concepts and Findings 153

B. Etnyre
Two Aspects of Motor Learning: Learning Movements

and Learning Synergies 155
M. L. Latash
Neuroanatomical Correlates of Motor Skill Learning: Inferences

from Neuroimaging to Behavior 167
K. Lindquist & M. A. Guadagnoli
Mechanisms Underlying Short-Term Motor Learning,

Long-Term Motor Learning and Transfer 177
D. M. Corcos, J. Shemmell, & D. E. Vaillancourt
A Dynamical Framework for Human Skill Learning 189

C. Magne & J. A. Scott Kelso
Part IV: Animal Model Systems
Animal Models of Behavioral and Neural Plasticity 207

J. Steven de Belle
A Biological Basis for Animal Model Studies of Learning and Memory 211
B. S. Dunkelberger, C. N. Serway, & J. Steven de Belle
Caenorhabditis elegans as a Model System in Which to Study

the Fundamentals of Learning and Memory 227
T. A. Timbers & C. H. Rankin
Contents vii
The Cell Biology of Learning and Memory in Aplysia 243

D. L. Glanzman
Insect Minds For Human Minds 271

R. Menzel
Patterns of Learning, Memory, and Vocal Production

in the Songbird Brain 287
D. S. Vicario
The cAMP/PKA Pathway and the Modeling of Human

Memory Disorders in Mice 301
F. X. Brennan & T. Abel
Index 317
This page intentionally left blank
ix
Preface
Even in its simplest form, the study of memory and learning is greatly
varied and highly diverse. It would be at best an incomplete list if one
included motor learning, verbal learning, implicit learning, and explicit
learning, as areas of investigation. Additionally, the level at which learning
is investigated varies greatly. For example, some psychologists may inves-
tigate verbal learning from a behavioural perspective, while others may use
brain imagery techniques to investigate both motor and verbal learning at
anatomical and functional level. Yet another approach is that of biologists
who study learning at a cellular or molecular level, and as such have come
to rely on nonhuman species for much of their investigation. Human
Learning: Biology, Brain, and Neuroscience is a collection of chapters
designed to synthesize findings across these levels and types of learning
and memory investigation. More specifically, preeminent authors in the
fields of verbal and motor learning provided chapters that discuss advances
in the areas of cognitive neuroscience, brain chemistry, and brain imaging.
These authors not only brought to light modern advances in science, but
also did it in a fashion that is understandable across disciplines of cognitive
science. To facilitate the flow between the major topics, Section Editors
synthesized and reviewed the most pertinent findings from each section.
The scope and format of Human Learning: Biology, Brain, and Neu-
roscience is designed to contribute to future research and academics. The
book is intended to provide a platform by which findings on learning and
learning theory can be shared across areas and levels of investigation. In
this regard, the book can provide an impetus for future interdisciplinary
investigations: a ground that we believe is truly fertile for scientific dis-
coveries. From an academic perspective, Human Learning: Biology,
Brain, and Neuroscience differs from other books in that it spans multiple
levels and areas of investigation, whereas most texts choose to discuss a
single level or area. Because of our general approach, including the use of
Section Editors as Discussants, our publisher, Elsevier Science, believes
that this book is in a class of its own. In addition to being an excellent
general resource text, they also believe that it can be used for a number of
upper division or graduate course in learning. This would include a number
of cognitive neuroscience courses, motor learning courses, biology of
learning courses, and verbal learning courses. Indeed, Elsevier Science
believes that the book is quite useful to theorists and practitioners alike.
xi
Contributors
Ted Abel (301) Department of Biology, University of Pennsylvania, 204G

Lynch Labs, 433 South University Avenue, Philadelphia, PA 19104
Michael C. Anderson (7) Department of Psychology, University of
Oregon, 345 Straub Hall, 1227, Eugene, OR 97403-1227, USA
Alexandra S. Atkins (125) Department of Psychology, University of
Michigan, 1012 East Hall, 530 Church Street, Ann Arbor, MI 48109-1043,
USA
Aaron S. Benjamin (3) Cognitive Division, University of Illinois, 827
Psychology Building, 2155 Beckman Institute, Champaign, IL 61820, USA
Jin Bo (101) Division of Kinesiology, University of Michigan, 401 Washte-
naw Avenue, Ann Arbor, MI 48109-2214, USA
Francis X. Brennan (301) Redpoint Bio Corporation, Korman Research
Pavilion, Albert Einstein Medical Center, 5501 Old York Rd.,
Philadelphia, PA 19141
Tiffany A. Timbers (227) Department of Psychology & Brain Research
Center, University of British Columbia, 2136 West Mall, Vancouver, B.C.
V6T 1Z4, Canada
Daniel Corcos (177) Kinesiology and Nutrition, College of Applied Health
Sciences, University of Illinois-Chicago, 650 AHSB, MC 517, 1919 W.
Taylor Street, Chicago, IL 60612, USA
J. Steven de Belle (207, 211) School of Life Sciences, University of
Nevada, 4505 Maryland Parkway, Las Vegas, NV 89154, USA
Brian S. Dunkelberger (211) School of Life Sciences, University of
Nevada, 4505 Maryland Parkway, Las Vegas, NV 89154, USA
Bruce Etnyre (153) Kinesiology Department, Rice University, 6100 Main
MS 545, Houston, TX 77005, USA
Karen M. Evans (33) University of Illinois, Urbana-Champaign, 1438 Beck-
man Institute, 405 N. Mathews Avenue, Urbana, IL 61801, USA
Kara D. Federmeier (33) University of Illinois, Urbana-Champaign, 2115
Beckman Institute, 405 N. Mathews Avenue, Urbana, IL 61801, USA
Kristin E. Flegal (137) Department of Psychology, University of Michigan,
1012 East Hall, 530 Church Street, Ann Arbor, MI 48109-1043, USA
xii Contributors
Daivd L. Glanzman (243) and Neurobiology, and the Brain Research

Institute, Depeartment of Physiological Science, University of
California, 621 Charles Young Drive South, Box 951606, Los Angeles
CA 90095-1606
Mark A. Guadagnoli (167) Department of Kinesiology, College of Health
& Human Sciences, University of Nevada, 4505 S. Maryland Parkway, Las
Vegas, NV 89154-3034, USA
Wendy Heller (45) University of Illinois, Urbana-Champaign, 2111 Beck-
man Institute, 405 N. Mathews Avenue, Urbana, IL 61801, USA
Justin C. Hulbert (7) University of St. Andrews School of Psychology, St.
Mary’s College, South Street, St. Andrews, Fife, KY16 9JP, Scotland, UK
J. A. Scott Kelso (189) Human Brain and Behavior Laboratory, Center for
Complex Systems and Brain Sciences, Florida Atlantic University, Boca
Raton, FL 33431, USA
Jeanne M. Langan (101) Division of Kinesiology, University of Michigan,
401 Washtenaw Avenue, Ann Arbor, MI 48109-2214
Mark L. Latash (155) Department of Kinesiology, Pennsylvania State
University, 268N Recreation Building, University Park, PA 16802, USA
K. Lindquist (167) Department of Kinesiology, University of Nevada, 4505
S. Maryland Parkway, Las Vegas, NV 89154-3034, USA
Cindy Lustig (137) Department of Psychology, University of Michigan,
1012 East Hall, 530 Church Street, Ann Arbor, MI 48109-1043, USA
Cyrille Magne (189) Human Brain and Behavior Laboratory, Center for
Complex Systems and Brain Sciences, Florida Atlantic University, Boca
Raton, FL 33431, USA
Randolf Menzel (271) Institut für Biologie – Neurobiologie, Freie Univer-
sität Berlin, Königin-Luise-Strasse 28/30, 14195 Berlin-Dahlem, Germany
Gregory A. Miller (45) University of Illinois, Urbana-Champaign, 2111
Beckman Institute, 405 N. Mathews Ave., Urbana, IL 61801
Ian Neath (21) Department of Psychology, Memorial University of New-
foundland, St. John’s, NL A1C 5S7, P.O. Box 4200, Canada
Lee I. Newman (77) Department of Psychology, University of Michigan,
525 E. University, Ann Arbor, MI 48109-1109, USA
Thad A. Polk (75,77) Department of Psychology, University of Michigan,
525 E. University, Ann Arbor, MI 48109-1109, USA
Catherine H. Rankin (227) Department of Psychology & Brain Research
Center, University of British Columbia, 2136 West Mall, Vancouver, B.C.
V6T 1Z4, Canada
Contributors xiii
Paul J. Reber (113) Psychology Department, Northwestern University, 309

Cresap Laboratories, 2029 Sheridan Road, Evanston, IL 60201, USA
Patricia A Reuter-Lorenz (125) Department of Psychology, University of
Michigan, 1012 East Hall, 530 Church Street, Ann Arbor, MI 48109-1043
Rachael D. Seidler (101) Department of Psychology & Division of Kine-
siology, University of Michigan, 401 Washtenaw Avenue, Ann Arbor, MI
48109-2214, USA
Christine N. Serway (211) School of Life Sciences, University of Nevada,
4505 Maryland Parkway, Las Vegas, NV 89154, USA
Jonathan Shemmell (177) Physical Therapy and Neurology and Rehabi-
litation, University of Illinois at Chicago, Chicago, IL 60612, USA
Dominic A. Simon (61) Department of Psychology, New Mexico State
University, P.O. Box 30001, Las Cruces, NM 88003-8001, USA
Aimée M. Surprenant (21) Department of Psychology, St. John’s, NL A1C
5S7, Canada
David E. Vaillancourt (177) Sensory Motor Performance Program, The
Rehabilitation Institute of Chicago, 345 E. Superior Street, Chicago, IL
60611, USA
David Vicario (287) Department of Psychology, Rutgers University, Busch
Campus, 152 Frelinghuysen Road, Piscataway, NJ 08854-8020
Stacie L. Warren (45) University of Illinois, Urbana-Champaign, Psychol-
ogy Department, 603 East Daniel Street, Champaign, IL 61820, USA
Part I
Human Learning and Cognition

Human Learning 3
Aaron S. Benjamin, J. Steven de Belle, Bruce Etnyre, Thad A. Polk
2008 Elsevier Ltd. All rights reserved
Introduction
Behavioral Approaches to the Study of Human

Learning and Memory
Aaron S. Benjamin
University of Illinois, Urbana-Champaign
The systematic investigation of the ability of humans to learn and

remember information had its beginning near the end of the nineteenth
century. Numerous influential scholars and developments during that per-
iod strongly impacted the nascent field, and it has now matured into
several major research communities that each bring a different focus and
methodology to the study of learning and memory. Arguably, the most
prominent publication of this era was Hermann Ebbinghaus’s Über das
Gedächtnis (Ebbinghaus, 1885), but several important short notes report-
ing experiments similar to Ebbinghaus’s were reported by the physicist
Francis Nipher in the preceding decade (Nipher, 1876, 1878). These three
publications reported systematic treatments of learning, remembering, and
forgetting that employed quite modern techniques, such as the use of
careful stimulus control, quantitative treatment of data, and multiple
measures of retention. In that same period, both Sergei Korsakoff and
Theodore Ribot published monographs that analyzed memory-impaired
populations and provided sophisticated treatments of memory function
(Korsakoff, 1887; Ribot, 1881). Shortly thereafter, Edward Thorndike pub-
lished some of the first work using animal subjects to study learning and
memory (Thorndike, 1898). Earlier in that decade, Thorndike’s mentor,
William James, published the authoritative two-volume set Principles of
Psychology (James, 1890), which provided a comprehensive and scholarly
treatment of many topics in learning and memory.
Despite this common ancestry, current treatments of human learning
and memory are so varied that a volume such as this one is necessary
merely to track parallels and divergences among subject populations and
among methodological approaches. The purpose of this section is to review
4 Part I
some of the current topics that arise in the experimental treatment of

memory in nonpathological adults. This work thus follows most directly
from the tradition established by Ebbinghaus and Nipher, and is character-
ized by three interrelated principles that can be directly traced to those
origins and that are distinct from the traditions established by the other
great scholars of that early time.
1. Experimentation
The principal tool for investigating learning and memory in normal

humans is the experiment. Here the emphasis is on the rigid definition of
the term, which fundamentally implies random assignment as a core prin-
ciple. This may not seem like a particularly distinctive standard for
research, but it is on this principle that this research diverges from the
human neuropsychological research that has lesion studies as its basic
currency. Animal research and human neuroimaging serve as bridges
between lesion studies in humans (which are nonexperimental but are
revealing of brain function) and experimental cognitive psychology
(which is rigorously experimental but mostly uninformative about the
underlying neuroanatomy and neurobiology of cognition).
2. Quantitative standards
Today, behavioral research on learning and memory is one of the most

quantitatively and analytically advanced areas of psychology. At least in
part, this is the heritage of Ebbinghaus and Nipher, who were pioneers in
their quantitative sophistication. Both were acutely aware of measurement
error and variability, and developed techniques to reduce such error (by
using repeated measures) and to evaluate results with an eye toward the
magnitude of such error. Both pioneered a model-based approach to
experimental analysis of psychological data; Nipher, for example, evalu-
ated memory across different serial positions of digit strings by compar-
ison with proportions predicted by a binomial distribution null hypothesis.
3. Abstractness
The final legacy attributable to Ebbinghaus and Nipher was the use of
abstract stimuli. Of all the aspects listed here, this one is the most ques-
tionable, and the one that has undergone the most scrutiny over the history
Introduction 5
of the field. The advantages of such stimuli were well articulated by

Ebbinghaus, who used consonant–vowel–consonant trigrams as stimuli:
The nonsense material . . . offers many advantages, in part because of this

very lack of meaning. First of all, it is simple and relatively homogeneous. In
the case of the material nearest at hand, namely poetry or prose, the content
is now narrative in style, now descriptive, or now reflective; it contains now
a phrase that is pathetic, now one that is humorous; its metaphors are
sometimes beautiful, sometimes harsh; its rhythm is sometimes smooth
and sometimes rough. There is thus brought into play a multiplicity of
influences which change without regularity and are therefore disturbing.
(Ebbinghaus, 1964, p. 23)
By using such stimuli, Ebbinghaus hoped to reduce the problems of

memory to a more tractable set, one that minimized the influence of
stimulus factors. As can be seen in the contributions to this section, this
approach is still the predominant one in the field. However, arguments
have been made that such contrived stimuli impede a full understanding of
the capacities of memory (e.g., Neisser, 1976).
And so this is the stage in which the chapters of this section are players.
The predominately experimental, quantitative, abstract approach will be
evident throughout; remembering the advantages and also the limitations
of such aspects is crucial to gaining a better understanding of what beha-
vioral studies have to offer to the larger picture of learning and memory
that this book provides.
The chapter by Hulbert and Anderson outlines evidence that inhibition
can be actively and strategically used to support memory function and
increase cognitive efficiency. Because memory inhibition is a central
player at the interface of cognitive psychology and clinical psychology—
in which inhibition is presumed to underlie important putative clinical
phenomena such as repression—well-controlled experimental studies
have had a profound influence on both low-level cognitive characteriza-
tions of memory and higher-level clinical and social depictions of memory.
Neath and Surprenant critically consider a distinction popularized by
William James between short-term and long-term memory. The larger
debate that this chapter prominently features is how data are used to
support or reject distinctions between multiple memory systems. This is
a point of contention between lesion studies, for which double dissocia-
tions are the signature datum, and perspectives such as the one embodied
in this chapter, in which quantitative process models are used to interpret
patterns across experiments.
The chapter by Evans and Federmeier provides an excellent example of
how Korsakoff’s legacy can be profitably merged with the tradition of
6 Part I
Ebbinghaus. They review the current state of knowledge on hemispheric

specialization of mnemonic functions, with an eye toward characterizing
the complementary roles the cerebral hemispheres have evolved. Warren,
Miller, and Heller consider the role of lateralized function further by examining
how emotion can promote or disrupt memory. In doing so, they provide linkage
to clinical characterizations by considering how psychopathy—particularly,
depression and anxiety—influences memory function.
The final chapter by Simon reviews how the scheduling of learning
events can have dramatic effects on consequent memory for the involved
materials. Implementing learning regimens that accord with the well-
known effects of spacing and intermixing of materials thus has the poten-
tial to improve learning in education and training settings.
Each chapter in this section provides an example of how modern mem-
ory research has benefited from connections to neuroscience, computer
science, education, and other areas. Fundamentally, this is the lesson of
this book: understanding learning and memory in any useful way requires
investigation and consideration from multiple levels of analysis. To the
degree that we create an environment within which researchers from
different traditions can communicate effectively, we will have done our-
selves, and the field, a good turn.
References
Ebbinghaus, H. (1885/1913/1964). Über das Gedächtnis [(Trans.: On Memory).

Leipzig: Duncker and Humblot (H. A. Ruger & C. E. Bussenius, Trans.); 1913
Translation: New York: Columbia University Press; 1964 Reprint: London: Dover].
James, W. (1890). Principles of psychology (Vols. 1–2). New York: Holt.
Korsakoff, S. S. (1887). Disturbance of psychic function in alcoholic paralysis and
its relation to the disturbance of the psychic sphere in multiple neuritis of
nonalcoholic origin. Vestnik Psychiatrii, 4: fascicle 2.
Neisser, U. (1976). Cognition and reality. San Francisco: Freeman.
Nipher, F. E. (1876). Probability of error in writing a series of numbers. American
Journal of Science and Arts, 12, 79–80.
Nipher, F. E. (1878). On the distribution of errors in numbers written from memory.
Transactions of the Academy of Science of St. Louis, 3, ccx–ccxi.
Ribot, T. (1881). Les Maladies de la Memorie, Paris: Germer Baillere.
Thorndike, E. L. (1898). Animal intelligence: An experimental study of the associa-
tive processes in animals. Psychological Review, Monograph Supplements, 2
(Serial No. 8).
Human Learning 7
The Role of Inhibition in Learning

Justin C. Hulbert and Michael C. Anderson
University of St. Andrews, U.K.
A retentive memory may be a good thing, but the ability to forget is the
true token of greatness.
Elbert Hubbard
None of us wish for a poor memory. Knowledge and skill are precious
commodities we aim to amass, preserve, and disseminate. The pangs of
frustration felt when we forget a colleague’s name, an important appoint-
ment, or a friend’s birthday motivate the desire to have all of our memories
constantly accessible. Yet, one only needs to skim the fictional case of
Funes, the Memorious, by Jorge Luis Borges or the real-life tale of mnemo-
nist Solomon Shereshevskii to appreciate the pain and complications asso-
ciated with being unable to control which memories spring to mind.
Indeed, in everyday life, there are many situations in which remembering
is disadvantageous. For instance, simultaneously recalling all the spots in
which one has ever parked would be more confusing than helpful in
locating the car at the end of the day. In such cases, it is desirable to
selectively retrieve only the most current, contextually relevant informa-
tion (e.g., where one parked today). To the extent that irrelevant details
(e.g., the location of yesterday’s parking spot) intrude, our goals are under-
mined. Similarly, intrusive reminding of unpleasant, upsetting, anxiety-
provoking, or embarrassing events—like memories of trauma or loss—
carry the potential to distract us from our current tasks. When confronted
with such reminding, we may wish to stop retrieval entirely. In both
cases—selective retrieval and stopping retrieval—an inhibitory process
that renders unwanted memories less accessible would prove quite adap-
tive (Anderson, 2003; Bjork, 1989).
Before discussing these two situations in turn, it is worthwhile to
describe the view of forgetting we will advance. Forgetting has long been
relegated to the backseat in discussions of learning. Classically, forgetting
8 Justin C. Hulbert and Michael C. Anderson
was thought to result either from a slow decay caused by disuse or, as
Müller and Pilzecker (1900) proposed, as a passive consequence of learning
new material that interferes with the old. In contrast, we adopt a functional
view of forgetting. As outlined by Anderson (2003), we argue that forgetting
is largely a consequence not of learning, per se, but of the executive control
processes recruited to resolve response competition arising during
memory retrieval. In this chapter, we summarize the ways in which the
forgetting associated with one important facet of executive control—
inhibition—actually facilitates learning.
1. Inhibition in selective retrieval
Memory retrieval is not always as simple as following a straight path

from a reminder to a target. More often than not, a given retrieval cue is
related to many memories, only a subset of which are useful in a given
situation. A rich body of research demonstrates a strong relationship
between the number of memory competitors associated with a cue and
the difficultly in successfully retrieving a particular target memory (e.g.,
Anderson, 1974; Anderson & Neely, 1996; McGeoch, 1942; Postman, 1971).
The problem is compounded when we must overcome a prepotent, yet
inappropriate, memory trace in favor of a weaker, relevant one. Consider
mistakenly dialing a friend’s outdated telephone number right after she
moved or searching for keys in the end table where we normally keep them
rather than in our pants where we left them the night before.
The intrusion of inappropriate memories is distracting and often dele-
terious to our goals. Conceptually, one should be able to resolve the
resultant interference through a combination of boosting the target mem-
ory’s signal and decrementing the strength of competing memories. To
illustrate, repeatedly dialing a friend’s current telephone number would
elevate the accessibility of that number, while attenuating our ability to
access to the older one, even if asked. Decades of empirical research have
highlighted the ways in which we can facilitate target memory retrieval
(Thorndike, 1932); work in our lab, however, has primarily focused on the
latter method of satisfying our mnemonic goals.
Should an unwanted memory impede retrieval of a target trace, we claim
that inhibitory control mechanisms are recruited to override or suppress
the intrusive memory in favor of the more contextually appropriate
response. Such a mechanism would prove especially advantageous if, as
a consequence of retrieving the target memory, it reduced the likelihood
that the competitor would intrude over the long-term. Hence, the very act
of remembering should cause lasting forgetting. If so, when confronted
with a similar situation in the future, we could easily retrieve the relevant
information, unencumbered by intrusions from outmoded or erroneous

knowledge.
Research from our lab and others has shown that selectively retrieving
target traces does suppress competing memories. In the retrieval practice
procedure designed to investigate this phenomenon (Anderson, Bjork, &
Bjork, 1994), participants typically study lists of category–exemplar pairs
(e.g., FRUITS-BANANA, DRINKS-SCOTCH, and FRUITS-ORANGE) and
subsequently practice retrieving half of the exemplars from half of the
categories multiple times, given category and wordstem cues (e.g.,
FRUITS-OR___). Following a 20-min delay, participants are asked to recall
all of the previously studied exemplars. Not surprisingly, retrieving some
items (e.g., ORANGE) during retrieval practice yields improved recall for
those items compared to baseline items from unpracticed categories (e.g.,
SCOTCH). More revealing is the finding that retrieval practice impairs
recall for unpracticed items from practiced categories (e.g., BANANA)
relative to baseline items. This finding, referred to as retrieval-induced
forgetting (RIF), has since been demonstrated using a wide variety
of stimuli (for reviews, see Anderson, 2003; Levy & Anderson, 2002),
including ambiguous words (Shivde & Anderson, 2001), visuospatial
objects (Ciranni & Shimamura, 1999), unusual actions (Koutstaal, Schacter,
Johnson, Angell, & Gross, 1998), personality traits (Koutstaal, Schacter,
Johnson, & Galluccio, 1999; Macrae & MacLeod, 1999b), eyewitness mem-
ories (MacLeod, 2002; Shaw, Bjork, & Handal, 1995), and autobiographical
memories (Barnier, Hung, & Conway, 2004).
Although consistent with the notion that inhibitory control is recruited
to overcome interference during retrieval practice, the basic findings of
RIF could also be explained by several noninhibitory mechanisms (for a
review, see Anderson et al., 1994). For instance, the strengthened, prac-
ticed items may come to intrude so pervasively during later memory tests
that participants’ attempts to retrieve the unpracticed responses are
effectively occluded, a circumstance referred to as associative blocking.
If forgetting is produced solely by strengthening practiced items,
then doing so through extra exposure without actual retrieval practice
should be just as effective in reducing the final recall of unpracticed
competitors. Contrary to a noninhibitory account, such conditions fail
to produce RIF, despite facilitating the practiced items to the same
degree as does retrieval practice (Anderson & Bell, 2001; Anderson,
Bjork, & Bjork, 2000; Bäuml, 1996, 1997, 2002; Ciranni & Shimamura,
1999; Shivde & Anderson, 2001). Thus, RIF appears to be specifically
induced by retrieval and dissociable from the degree to which practiced
items are strengthened.
A second, noninhibitory account called associative unlearning might
frame the basic findings of RIF in terms of damage exacted upon the
associative bond linking the category cue and the unpracticed exemplar.
Accordingly, retrieval practice reduces the viability of the category label to
serve as an effective cue for the unpracticed item—that is, it merely
damages one pathway to the competing memory. Only an inhibitory
account in which the competing memories are actively suppressed, though,
accurately predicts that RIF is independent of the cue used during retrieval
practice and should generalize to novel cues (Anderson & Spellman, 1995).
For example, after retrieving FRUIT-ORANGE from memory, later recall of
BANANA should be impaired regardless of whether it is cued by the
originally studied category (FRUITS-) or by a novel, independent retrieval
cue (e.g., MONKEY-B). Cue-independent forgetting of this sort has since
been demonstrated numerous times (e.g., Anderson & Bell, 2001; Anderson
et al., 2000; Anderson & Spellman, 1995; Aslan, Bäuml, & Pastotter, 2007;
Camp, Pecher, & Schmidt, 2005; Levy, McVeigh, Marful, & Anderson, 2007;
MacLeod & Saunders, 2005; Saunders & MacLeod, 2006), including in a
classic retroactive interference paradigm (Hulbert & Anderson, In Prepara-
tion) and in another, related method referred to as part-set cuing (Aslan,
Bäuml, & Grundgeiger, 2007).
The real-world implications of RIF are brought into focus when one
considers circumstances that demand accurate and complete fact retrieval,
such as eyewitness testimony and academic examination. As Shaw et al.
(1995) and MacLeod (2002) have independently shown, prompts to recall
specific details of mock crime scenes impair the ability to recall related but
previously undiscussed particulars, demonstrating anew that retrieving
some experiential elements impairs others. Clearly, RIF is a double-edged
sword, facilitating the retrievability of practiced items at the expense of
related items that, though not germane at the time of practice, could later
return to relevance. Likewise, students are commonly presented with an
abundance of facts and ideas related to a given topic and tested on specific
items. As their knowledge base for a particular topic grows, retrieval of any
one fact is slowed in response to the need to resolve competition from an
increasing number of memory associations, a finding J.R. Anderson (1974)
termed the fan effect. The necessity to expediently retrieve selected facts
in a testing situation should, therefore, recruit inhibitory control mechan-
isms to resolve the amassed interference. Sure enough, Anderson and Bell
(2001) generalized RIF to fact retrieval by having participants study pro-
positions such as ‘‘The Actor is Looking at the Tulip’’ and practice only a
subset of facts related to a topic. As a result, participants were not only
rendered less able to recall related, unpracticed facts (e.g., ‘‘The Actor is
Looking at the Violin’’) after a delay, but also less likely to recall other
learned facts in which the inhibited object participated (e.g., ‘‘The Teacher
is Lifting the Violin’’) (see also Gomez-Ariza, Lechuga, Pelegrina, & Bajo,
2005; Macrae & MacLeod, 1999a).
Both academic and research settings abound with circumstances that

encourage the selective retrieval of certain facts related to a given topic.
Consider professors and experts, who routinely retrieve and present only a
subset of their wealth of knowledge. Would selective retrieval affect
experts’ untapped competence in the same way it influences their stu-
dents? Though more work is necessary, evidence indicates that specialists
manage to preserve their expertise despite the circumstances. In one
recent example, Carroll, Campbell-Ratcliffe, Murnane, & Perfect (2007)
asked both first-year psychology students (i.e., novices) and psychology
majors (i.e., relative experts) to examine a pair of case studies adapted
from an upper-level abnormal psychology textbook. All participants were
then repetitively quizzed on some of the information for half of the case
studies through a series of questions (with corrective feedback), as though
they were preparing for an upcoming exam with a study partner. Final
testing on the entirety of the studied material after a 15-min delay revealed
that novices were significantly less able to recall unpracticed details from
the practiced case study compared to baseline items. Experts, however,
experienced no measurable RIF, owing, the authors claimed, to experts’
ability to integrate the learned material into existing knowledge schema
and reducing the extant amount of competition and consequent forgetting
(see also Smith, Adams, & Schorr, 1978). In fact, Anderson and McCulloch
(1999) demonstrated that simply instructing naı̈ve participants to integrate
novel information during the initial study phase of a standard RIF paradigm
was enough to significantly reduce forgetting on the final test. In a separate
study, participants who later claimed to have spontaneously engaged in
integrative learning without explicit instruction were also largely shielded
from the forgetting effect experienced by nonintegrating participants
(Anderson & Bell, 2001).
Remarkably, even the material over which we command great expertise
is rendered susceptible to RIF to the extent that it competes with informa-
tion not yet well-integrated into our current knowledge structure. Take
another real-world example of learning: second-language acquisition. Sub-
sequent to immersion in a foreign language environment, such as a seme-
ster abroad, individuals commonly report what is called first-language
attrition, a phenomenon characterized by difficulty in retrieving native-
language words. In order to simulate this experience, Levy et al. (2007)
had native English speakers practice naming objects in a second language
they were studying in school: Spanish. The authors reported significant
forgetting of English phonological labels after naming objects in Spanish a
mere 10 times. In line with prior results regarding expertise, those participants
with relatively greater Spanish fluency were subject to less inhibition than
participants rated less adept in the foreign language. Thus, inhibition appears
to be most heavily recruited during the early stages of second-language
acquisition, thereby facilitating retrieval of the weaker, foreign language

in the face of the otherwise overwhelming dominance of the native
vocabulary. This finding is consistent with interference dependence, the
finding that retrieval only inhibits related traces to the extent that they pose
a considerable threat of interference (Anderson et al., 1994; Shivde &
Anderson, 2001).
In recent years, electrophysiological and functional imaging techniques
have converged to help illuminate the neural correlates of inhibition result-
ing from selective retrieval. Johansson, Aslan, Bäuml, Gabel, and Mecklin-
ger (2006) recorded event-related potentials (ERPs) from participants as
they engaged in retrieval practice or were simply given extra study expo-
sures. Electrical activity over the prefrontal region of the brain was modu-
lated by task, revealing a sustained, positive-going augmentation of ERP
waveforms bilaterally when participants engaged in selective retrieval.
Moreover, activity recorded over this region (specifically, late anterior
frontal amplitudes) during selective retrieval—but not during extra study
exposure—predicted individual differences in the amount of subsequently
observed RIF, accounting for one-third of the variance. These results are in
line with the notion that the prefrontal cortex is heavily involved in cogni-
tive control processes, including selective memory retrieval. Converging
evidence was provided by Kuhl, Dudukovic, Kahn, and Wagner (2007)
who utilized the heightened spatial resolution of functional magnetic reso-
nance imaging (fMRI) to identify brain regions that reflect the level of
demand for cognitive control as competition is reduced over the course
of retrieval practice. The right ventrolateral prefrontal cortex (VLPFC)
exhibited just this property, with the reduction of neural activity from the
first to the third retrieval practice attempt predicting forgetting in the
subsequent final test. This region has often been implicated in a wide
variety of executive control tasks (e.g., Bunge, Ochsner, Desmond, Glover,
& Gabrieli, 2001; Garavan, Ross, Murphy, Roche, & Stein, 2002; Jonides,
Smith, Marshuetz, Koeppe, & Reuter-Lorenz, 1998; Menon, Adleman, White,
Glover, & Reiss, 2001; Nakahara, Hayashi, Konishi, & Miyashita, 2002;
Shimamura, 2000).
The inverse relationship between brain activity and increased forgetting
also held for the anterior cingulate cortex (ACC), a finding that deserves
special discussion. As neuroimaging studies of executive control have tied
this brain region to the detection of conflict between competing responses
(Botvinick, Braver, Barch, Carter, & Cohen, 2001; Botvinick, Cohen, &
Carter, 2004), Kuhl et al. (2007) predicted that as inhibitory control
managed to resolve interference across multiple retrieval practice
attempts, conflict between competing memories measured by the amount
of ACC activation, should be lessened. As hypothesized, the extent of the
reduction in ACC activation across retrieval practices predicted the
amount of RIF observed on the later test. Kuhl et al. further explored
whether high inhibitors (as defined by the magnitude of their later beha-
vioral RIF effect) also showed greater initial ACC activation than did low
inhibitors. In other words, individuals who initially experience a high level
of response conflict (i.e., retrieval competition between the target memory
and other exemplars) should and did exhibit more inhibition for the
competitors later on, due to the heightened amount of executive control
necessary to meet the demand for response resolution. The conflict was
also measurable upstream of the ACC in the medial temporal lobe where
target and competitor memories were actually vying for retrieval. This
ostensibly direct measure of initial conflict in the right hippocampus
correlated with both ACC activity and behavioral RIF.
Thus far, we have outlined how inhibition helps overcome interference.
It should be clear that lingering aftereffects of inhibition—namely reduced
accessibility of competitors—though beneficial to the extent that affected
memories remain contextually inappropriate, become the object of frustra-
tion when our goals change and we want to retrieve an inhibited memory at
some later point. Although a number of recent investigations have demon-
strated forgetting effects after periods as long as a week (Storm, Bjork, &
Bjork, 2007), an inhibited memory is not necessarily a memory lost forever.
Unlike the permanent abolition of a memory trace, inhibition is often
thought to be reversible, so that a memory may regain some portion of its
prior accessibility as retrieval contexts demand. This may explain why
others have found that the effects of RIF diminish after 24-h delays
(MacLeod & Macrae, 2001; Saunders & MacLeod, 2002).
2. Inhibition in memory stopping
Not only is inhibition likely recruited in situations that demand the

resolution of competition, it also plays a significant role in situations that
require the cessation of actions or processes. One can easily appreciate the
usefulness of inhibition on a motor level. When a baseball player is
confronted with an oncoming pitch outside of the strike zone, executive
control kicks in, allowing him to override the prepotent response to swing.
Traditionally, motor stopping of this sort has been studied empirically
using the Go/No-Go task, in which humans (de Zubicaray, Andrew, Zelaya,
Williams, & Dumanoir, 2000; Garavan, Ross, & Stein, 1999) or monkeys
(Sagkagami & Niki, 1994; Sasaki, Gemba, & Tsujimoto, 1989) are asked to
make a physical response on a majority of trials but to withhold response
when so indicated by a less-frequent stimulus. Our lab endeavored to
understand whether memory retrieval can be overridden in a similar man-
ner. Such an occasion might arise when faced with a reminder to an
unpleasant memory or are merely in need of pushing distracting thoughts

out of mind.
Drawing from the motor domain, Anderson and Green (2001) developed
the Think/No-Think (TNT) paradigm that crafts a situation in which parti-
cipants must intentionally exclude certain well-learned memories from
awareness in the face of strong reminders. The procedure occurs in three
parts, commencing with the learning phase during which participants study
a series of cue-target word pairs (e.g., FLAG-SWORD; ORDEAL-ROACH;
LAWN-BEEF). Following this, participants enter the critical TNT phase
when they are presented with the cues of two-thirds of the learned word
pairs (e.g., FLAG-; ORDEAL-). Half of these cues are presented in green,
indicating that participants are to recall the associated memory as quickly
as possible and keep it in mind the entire time that the cue remains on the
screen (i.e., Think trials). If, however, the cue is presented in red as are the
remaining half of the items, participants are instructed to avoid thinking
about the associated memory (i.e., No-Think trials). In the final test phase
that follows, participants are asked to recall the associated targets for all
the learned cues (i.e., Think, No-Think, and Baseline cues, which do not
appear in the TNT phase).
Not surprisingly, recall for those items participants practiced retrieving
during the TNT phase is facilitated as a function of the number of repeti-
tions. Of particular relevance, however, is the deficit in recall for the
avoided memories in comparison to Baseline items that were learned
initially but lacked reminders during the TNT phase. Again, the recallability
of the targets was related to the number of times the cue was presented;
yet, in this case, the relationship is inverted (Anderson & Green, 2001;
Levy & Anderson, 2002). This finding is counterintuitive in that recurrent
reminders, instead of facilitating later recall, actually served to do the
opposite simply by directing subjects to push the unwanted associates
out of mind. In much the same way that RIF is cue-independent, the TNT
effect persists when subjects are tested with an independent probe (e.g.,
INSECT-R___ for ROACH), thus indicating that this impairment is not
simply a result of associative interference, but is reflective of the memory
itself being inhibited. If, however, the TNT paradigm’s typical memory-
suppression instructions are altered only slightly, such that participants
are encouraged to generate an alternative target word for each No-Think
cue as a means of avoiding recollection of the original target during the
second phase (e.g., treating the novel word FIGHT as a target for the cue
ORDEAL-), cue-independent forgetting is lost (Bergström, de Fockert, &
Richardson-Klavehn, Submitted). Given that cue independence is a critical
marker of inhibition, it appears as though, in contrast to thought suppres-
sion, thought substitution is not enough to induce inhibitory forgetting.
Thought substitution does appear, however, to interfere with the retrieval
of the avoided memory when that memory is tested with the original cue
from which the substitute was generated (Bergström et al., Submitted;
Hertel & Calcaterra, 2005).
To summarize, inhibiting memory retrieval bears a strong similarity to
stopping a motor response on a behavioral level, at least. Does the parallel
end there? An fMRI study conducted by Anderson et al. (2004) to identify
the areas of the brain engaged during TNT memory suppression permitted
a further comparison between the neural instantiations of the two types of
stopping. Many of the so-called cognitive control regions associated with
withholding motor responses, including the lateral prefrontal cortex, ante-
rior cingulate cortex, and intraparietal sulcus (Garavan et al., 2002; Menon
et al., 2001), showed evidence of increased activity during Think trials
compared to No-Think trials in Anderson et al.’s (2004) study. These
areas, therefore, appear to be key in inhibiting responses generally; still,
the targets of the inhibitory signals are expected to diverge from the motor
and premotor areas affected in the Go/No-Go task. A likely candidate in the
case of memory control is the hippocampus, the region thought to support
conscious recollection (Eldridge, Knowlton, Furmanski, Bookheimer, &
Engel, 2000; Squire, 1992). Substantiating this claim, Anderson et al.
(2004) first reported bilateral reduction of hippocampal activity for
No-Think relative to Think trials that predicted below-baseline behavioral
suppression, suggesting that subjects can strategically downregulate
the hippocampus to prevent conscious recollection. Depue, Curran, and
Banich (2007), among others, have since replicated the hippocampal deac-
tivation, alternatively contrasting No-Think trials with various baseline
conditions.
Additionally, Depue, Banich, and Curran (2006) established that the TNT
inhibition effect is not limited to verbal stimuli, replicating the below-
baseline performance in both face-word and face-place pairings. In line
with the notion that thought suppression might be especially useful in
inhibiting particularly distressing memories, Depue et al. (2007) found
that forgetting was greatest for negatively valenced items and later showed
that amygdalar activity itself is reduced over the course of No-Think trials.
Forgetting highly salient (and unpleasant) memories is all the more
remarkable in light of research indicating that emotional memories are
more easily retrieved than nonemotional memories (e.g., Bradley, 1994;
Pessoa, Kastner, & Ungerleider, 2002). In fact, negative memories in the
Think condition were facilitated in Depue et al.’s (2006) study, leading to
the conclusion that the effects of executive control are malleable depend-
ing on one’s goals. The desire to avoid unwanted memories is likely
strongest in individuals with more extensive and intrusive thoughts, the
same population that would benefit most greatly from—and have the most
practice exercising—the ability to inhibit those thoughts. In support of this
claim, Anderson and Kuhl (2004) found evidence for enhanced inhibition
effects in college students with more extensive histories of trauma.
Event-related potential recordings obtained during the TNT phase offer a
unique window into how people achieve memory control. Bergström,
Velmans, de Fockert, and Richardson-Klavehn (2007) concluded that strate-
gic control of memories begins at an attentional selection stage preceding
conscious recollection when participants permit Think cues—but not No-
Think cues—to undergo further retrieval processing. They supported this
claim by pointing to an enhanced frontal positivity and posterior negativity
for Think trials that occur well before the late parietal positivity denoting
conscious recollection. As a further indication that participants are able to
intentionally avoid thinking about unwanted memories on the level of indi-
vidual items, Bergström et al. (2007) demonstrated that the ERP signature of
conscious recollection observed during Think trials was practically absent
during No-Think trials. These data converge with the aforementioned neu-
roimaging evidence offered by Anderson et al. (2004) and Depue, Curran, &
Banich (2007) demonstrating that the magnitude of forgetting for suppressed
items is predicted by deactivations of the hippocampus, the brain region tied
to both recollective encoding and retrieval.
Learning effectively requires focused attention, a state difficult to
achieve when we are distracted by intrusive thoughts. Exerting memory
control helps drive out the specter of unwanted memories while simulta-
neously reducing the extent to which those thoughts are likely to intrude in
the future. Again, as we have seen in the case of RIF, inhibition serves to
benefit learning at the expense of retention for other memories. Never-
theless, our understanding of the intricate interface between learning and
forgetting remains ongoing. For example, Kawaguchi, Hotta, and Takei
(2006) presented preliminary evidence that the explicit memory deficit
does not extend to implicit memory tests for the targeted traces. Future
investigations will help clarify the extent and duration of forgetting caused
by voluntary suppression.
3. Concluding remarks
On one hand, learning, by definition, involves the acquisition of knowl-

edge. Inhibition, on the other hand, involves the reduction in accessibility
of a memory trace. Throughout this chapter we have presented evidence
that, though seemingly at odds with each other, learning and inhibition are
tightly integrated. Whether we are studying for an exam, conversing with
our colleagues, or writing up a paper, we depend on executive control
to retrieve the necessary facts (selective retrieval) while excluding goal-
irrelevant or otherwise bothersome memories from consciousness (memory
stopping). When memories compete for retrieval or intrude, the prefrontal

cortex can exert inhibitory control, thereby allowing us to accomplish our
present goals. As we have seen, however, inhibition yields lasting and
potentially frustrating consequences on the targets of memory control.
There is still a great deal to learn about the mechanisms and consequences
of inhibition; yet there is little question that a memory system that serves our
goals relies upon a healthy dose of forgetting.
References
Anderson, J. R. (1974). Retrieval of propositional information from long-term mem-

ory. Cognitive Psychology, 6(4), 451–474.
Anderson, M. C. (2003). Rethinking interference theory: Executive control and the
mechanisms of forgetting. Journal of Memory and Language, 49(4), 415–445.
Anderson, M. C., & Bell, T. (2001). Forgetting our facts: The role of inhibitory
processes in the loss of propositional knowledge. Journal of Experimental
Psychology: General, 130, 544–570.
Anderson, M. C., Bjork, E. L., & Bjork, R. A. (2000). Retrieval-induced forgetting:
Evidence for a recall-specific mechanism. Psychonomic Bulletin and Review, 7,
522–530.
Anderson, M. C., Bjork, R. A., & Bjork, E. L. (1994). Remembering can cause
forgetting: Retrieval dynamics in long-term memory. Journal of Experimental
Psychology: Learning, Memory, and Cognition, 20, 1063–1087.
Anderson, M. C., & Green, C. (2001). Suppressing unwanted memories by executive
control. Nature, 410, 131–134.
Anderson, M. C., & Kuhl, B. (2004). Inhibitory control and the suppression of
unpleasant events. Paper presented at the 45th Annual Meeting of the Psycho-
nomic Society.
Anderson, M. C., & McCulloch, K. C. (1999). Integration as a general boundary
condition on retrieval-induced forgetting. Journal of Experimental Psychology:
Learning, Memory, and Cognition, 25, 608–629.
Anderson, M. C., & Neely, J. H. (1996). Interference And Inhibition In Memory
Retrieval. In E. L. I. Bjork & R. A. Bjork (Eds.), Memory. handbook of perception
and cognition (2nd ed., pp. 237–313). San Diego: Academic Press.
Anderson, M. C., Ochsner, K. N., Cooper, J., Robertson, E., Gabrieli, S. W., Glover, G.
H., et al. (2004). Neural systems underlying the suppression of unwanted mem-
ories. Science, 303, 232–235.
Anderson, M. C., & Spellman, B. A. (1995). On the status of inhibitory mechanisms in
cognition: Memory retrieval as a model case. Psychological Review, 102, 68–100.
Aslan, A., Bäuml, K. H., & Grundgeiger, T. (2007). The role of inhibitory processes in
part-list cuing. Journal of experimental psychology: Learning, Memory, and
Cognition, 33(2), 335–341.
Aslan, A., Bäuml, K. H., & Pastotter, B. (2007). No inhibitory deficit in older adults’
episodic memory. Psychological Science, 18(1), 72–78.
Barnier, A. J., Hung, L., & Conway, M. A. (2004). Retrieval-induced forgetting of

emotional and unemotional autobiographical memories. Cognition and Emotion,
18(4), 457–477.
Bäuml, K. H. (1996). Revisiting an old issue: Retroactive interference as a function
of the degree of original and interpolated learning. Psychonomic Bulletin and
Review, 3(3), 380–384.
Bäuml, K. H. (1997). The list-strength effect: Strength-dependent competition or
suppression? Psychonomic Bulletin and Review, 4(2), 260–264.
Bäuml, K. H. (2002). Semantic generation can cause episodic forgetting. Psycholo-
gical Science, 13(4), 356–360.
Bergström, Z. M., de Fockert, J., & Richardson-Klavehn, A. (Submitted). Direct
Suppression of Unwanted Memory Representations.
Bergström, Z. M., Velmans, M., de Fockert, J., & Richardson-Klavehn, A. (2007). ERP
evidence for successful voluntary avoidance of conscious recollection. Brain
Research, 1151, 119–133.
Bjork, R. A. (1989). Retrieval inhibition as an adaptive mechanism in human mem-
ory. In H. L. I. Roediger and F. I. Craik (eds.), Varieties of memory and con-
sciousness: Essays in honour of Endel Tulving (pp. 309–330). Hillsdale, NJ:
Lawrence Erlbaum Associates.
Botvinick, M. M., Braver, T. S., Barch, D. M., Carter, C. S., & Cohen, J. D. (2001).
Conflict monitoring and cognitive control. Psychological Review, 108(3), 624–652.
Botvinick, M. M., Cohen, J. D., & Carter, C. S. (2004). Conflict monitoring
and anterior cingulate cortex: An update. Trends in Cognitive Sciences,
8(12), 539–546.
Bradley, M. M. (1994). Emotional memory: A dimensional analysis Hillsdale, NJ,
England: Lawrence Erlbaum Associates, Inc.
Bunge, S. A., Ochsner, K. N., Desmond, J. E., Glover, G. H., & Gabrieli, J. D. (2001).
Prefrontal regions involved in keeping information in and out of mind. Brain,
124(10), 2074–2086.
Camp, G., Pecher, D., & Schmidt, H. G. (2005). Retrieval-induced forgetting in
implicit memory tests: The role of test awareness. Psychonomic Bulletin and
Review, 12(3), 490–494.
Carroll, M., Campbell-Ratcliffe, J., Murnane, H., & Perfect, T. J. (2007). Retrieval-
induced forgetting in educational contexts: Monitoring, expertise, text integration
and test format. European Journal of Cognitive Psychology, 19, 580–606.
Ciranni, M. A., & Shimamura, A. P. (1999). Retrieval-induced forgetting in episodic
memory. Journal of Experimental Psychology: Learning, Memory, and Cognition,
25(6), 1403–1414.
Depue, B. E., Banich, M. T., & Curran, T. (2006). Suppression of emotional and
nonemotional content in memory. Effects of repetition on cognitive control.
Psychological Science, 17(5), 441–447.
Depue, B. E., Curran, T., & Banich, M. T. (2007). Prefrontal regions orchestrate
suppression of emotional memories via a two-phase process. Science, 317, 215–219.
de Zubicaray, G. I., Andrew, C., Zelaya, F. O., Williams, S. C. R., & Dumanoir, C.
(2000). Motor response suppression and the prepotent tendency to respond: A
parametric fMRI study. Neuropsychologia, 38, 1280–1291.
Eldridge, L. L., Knowlton, B. J., Furmanski, C. S., Bookheimer, S. Y., & Engel, S. A.
(2000). Remembering episodes: A selective role for the hippocampus during
retrieval. Nature Neuroscience, 3(11), 1149–1152.
Garavan, H., Ross, T. J., Murphy, K., Roche, R. A., & Stein, E. A. (2002). Dissociable
executive functions in the dynamic control of behavior: Inhibition, error detec-
tion, and correction. Neuroimage, 17(4), 1820–1829.
Garavan, H., Ross, T. J., & Stein, E. A. (1999). Right hemispheric dominance of
inhibitory control: An event-related functional MRI study. Proceedings of the
National Academy of Sciences, 96, 8301–8306.
Gomez-Ariza, C. J., Lechuga, M., Pelegrina, S., & Bajo, M. (2005). Retrieval-induced
forgetting in recall and recognition of thematically related and unrelated sen-
tences. Memory and Cognition, 33(8), 1431–1441.
Hertel, P. T., & Calcaterra, G. (2005). Intentional forgetting benefits from thought
substitution. Psychonomic Bulletin and Review, 12(3), 484–489.
Hulbert, J. C., & Anderson, M. C. (In Preparation). Cue-Independent Inhibition in
Retroactive Interference.
Johansson, M., Aslan, A., Bäuml, K. H., Gabel, A., & Mecklinger, A. (2006). When
remembering causes forgetting: Electrophysiological correlates of retrieval-
induced forgetting. Cerebral Cortex, 17(6), 1335–1341.
Jonides, J., Smith, E. E., Marshuetz, C., Koeppe, R. A., & Reuter-Lorenz, P. A. (1998).
Inhibition in verbal working memory revealed by brain activation. Proceedings of
the National Academy of Sciences, 95(14), 8410–8413.
Kawaguchi, J., Hotta, C., & Takei, S. (2006). Implicit memory for intentionally
suppressed information. Poster presented at the association for psychological
science 18th annual convention, New York, NY.
Koutstaal, W., Schacter, D. L., Johnson, M. K., Angell, K. E., & Gross, M. S. (1998).
Post-event review in older and younger adults: Improving memory accessibility of
complex everyday events. Psychology and Aging, 13(2), 277–296.
Koutstaal, W., Schacter, D. L., Johnson, M. K., & Galluccio, L. (1999). Facilitation
and impairment of event memory produced by photograph review. Memory and
Cognition, 27(3), 478–493.
Kuhl, B. A., Dudukovic, N. M., Kahn, I., & Wagner, A. D. (2007). Decreased demands
on cognitive control following memory suppression reveal benefits of forgetting.
Nature Neuroscience, 10, 908–914.
Levy, B. J., & Anderson, M. C. (2002). Inhibitory processes and the control of
memory retrieval. Trends in Cognitive Sciences, 6, 299–305.
Levy, B. J., McVeigh, N. D., Marful, A., & Anderson, M. C. (2007). Inhibiting your
native language: The role of retrieval-induced forgetting during second language
acquisition. Psychological Science, 18(1), 29–34.
MacLeod, M. D. (2002). Retrieval-induced forgetting in eyewitness memory: Forgetting
as a consequence of remembering. Applied Cognitive Psychology, 16(2), 135–149.
MacLeod, M. D., & Macrae, C. N. (2001). Gone but not forgotten: The transient
nature of retrieval-induced forgetting. Psychological Science, 12(2), 148–152.
MacLeod, M. D., & Saunders, J. (2005). The role of inhibitory control in the produc-
tion of misinformation effects. Journal of Experimental Psychology: Learning,
Memory, and Cognition, 31(5), 964–979.
Macrae, C. N., & MacLeod, M. D. (1999a). On recollections lost: When practice

makes imperfect. Journal of Personality and Social Psychology, 77(3), 463–473.
Macrae, C. N., & MacLeod, M. D. (1999b). On recollections lost: When practice
makes imperfect. Journal of Personality and Social Psychology, 77, 463–473.
McGeoch, J. A. (1942). The psychology of human learning: An introduction New
York: Longmans.
Menon, V., Adleman, N., White, C., Glover, G., & Reiss, A. (2001). Error-related brain
activation during a Go/NoGo response inhibition task. Human Brain Mapping,
12(3), 131–143.
Müller, G. E., & Pilzecker, A. (1900). Experimentalle beitrage zur lehre com gedacht-
nis. Zeitschrift Fur Psychologie, 1, 1–288.
Nakahara, K., Hayashi, T., Konishi, S., & Miyashita, Y. (2002). Functional MRI of macaque
monkeys performing a cognitive set-shifting task. Science, 295(5559), 1532–1536.
Pessoa, L., Kastner, S., & Ungerleider, L. G. (2002). Attentional control of the processing
of neutral and emotional stimuli. Cognitive Brain Research, 15(1), 31–45.
Postman, L. (1971). Transfer, interference and forgetting. In J. W. I. Kling & L. A.
Riggs (eds.), Woodworth and schlosberg’s experimental psychology (3rd ed., pp.
1019–1132). New York: Holt, Rinehart and Winston.
Sagkagami, M., & Niki, H. (1994). Spatial selectivity of Go/No-Go neurons in the
monkey prefrontal cortex. Experimental Brain Research, 100, 165–169.
Sasaki, K., Gemba, H., & Tsujimoto, T. (1989). Suppression of visually initiated hand
movement by stimulation of the prefrontal cortex in the monkey. Brain
Research, 495(1), 100–107.
Saunders, J., & MacLeod, M. D. (2002). New evidence on the suggestibility of
memory: The role of retrieval-induced forgetting in misinformation effects. Jour-
nal Of Experimental Psychology: Applied, 8(2), 127–142.
Saunders, J., & MacLeod, M. D. (2006). Can inhibition resolve retrieval competition
through the control of spreading activation? Memory and Cognition, 34(2), 307–322.
Shaw, J. S., Bjork, R. A., & Handal, A. (1995). Retrieval-induced forgetting in an
eyewitness-memory paradigm. Psychonomic Bulletin and Review, 2(2), 249–253.
Shimamura, A. P. (2000). The role of the prefrontal cortex in dynamic filtering.
Psychobiology, 28(2), 207–218.
Shivde, G., & Anderson, M. C. (2001). The role of inhibition in meaning selection:
Insights from retrieval-induced forgetting. In D. I. Gorfein (ed.), On the conse-
quences of meaning selection: Perspectives on resolving lexical ambiguity (pp.
175–190). Washington, D.C: American Psychological Association.
Smith, E. E., Adams, N., & Schorr, D. (1978). Fact retrieval and the paradox of
interference. Cognitive Psychology, 10(4), 438–464.
Squire, L. R. (1992). ‘‘Memory and the hippocampus: A synthesis from findings with
rats, monkeys, and humans’’: Correction. Psychological Review, 99(3), 582.
Storm, B. C., Bjork, E. L., & Bjork, R. A. (2007). When intended remembering
leads to unintended forgetting. Quarterly Journal Of Experimental Psychology,
60(7), 909–915.
Thorndike, E. L. (1932). The physiological basis of the strengthening of connections
by their after-effects. The fundamentals of learning (pp. 314–327). New York:
NY: Teachers College Bureau of Publications.
Human Learning 21
Short- vs. Long-Term Memory

Ian Neath and Aimée M. Surprenant
Memorial University of Newfoundland
Although researchers have been dividing human memory into multiple

systems for almost as long as they have been studying memory scientifi-
cally (Burnham, 1888; James, 1890), short-term memory did not really
become a well-established concept until the late 1950s with the publication
of four highly influential works: Miller (1956), Broadbent (1958), Brown
(1958), and Peterson and Peterson (1959). Since then, the basic idea that
the memory system used to remember over the short term (STM) is
fundamentally different from the system used to remember over the long
term (LTM) has become the dominant view of the field. This view is not
unanimous, however; there have been many researchers who have sug-
gested that the distinction is not needed (e.g., Crowder, 1982; Melton, 1963;
Nairne, 2002; Surprenant & Neath, 2008). The fundamental point of these
critiques is that each finding held up as a reason for postulating a separate
system can be readily explained in terms of ‘‘factors known to operate in
LTM’’ (Melton, 1963, p. 8). In this chapter, we begin by providing a very
brief review of the creation of short-term memory. In the remainder of the
chapter, we address three issues that, in our view, play a major role in the
continued general acceptance of STM: (1) introspection; (2) absence of an
alternative explanation for capacity limits; and (3) lack of an articulated
view of how memory might work without STM.
1. The creation of short-term memory
The modern instantiation of short-term memory came about because

researchers had difficulty explaining a series of experimental results in
terms of the then current theories of long-term memory. Miller (1956)
documented a pervasive and severe limitation on the ability to perceive,
process, and remember information; memory seemed to be limited to
seven (plus or minus two) items in a wide variety of tasks. Brown
22 Ian Neath and Aimée M. Surprenant
(1958) and Peterson and Peterson (1959) introduced a paradigm in which

rehearsal was controlled. The results from this Brown–Peterson task
consistently demonstrated that unless rehearsed, information fades or
decays in as little as 15–20 s. The extant theories of memory were based
on results obtained in multitrial learning experiments and were unsuited
to explain performance in single-trial experiments (for a review, see
Crowder, 1976).
It seemed as though the data required a fast-acting memory system that
could process information experienced only once. Viewed as a buffer, the
new system provided enhanced responsiveness, but at a cost: it could not
store much information and could not store it for very long. Broadbent
(1958) demonstrated how such a theory, borrowing concepts from the
newly developed digital computers, could explain these results by postulat-
ing sensory registers, a short-term buffer, and a limited capacity channel.
Especially, during the 1960s and early 1970s, this theoretical approach
proved highly successful (e.g. Atkinson & Shiffrin, 1968).
Beginning in the mid-1970s, however, results began to appear that chal-
lenged the now-standard dual-store model that divided memory into two
stores. In addition, many of the original findings used to support the
existence of STM have been reinterpreted in light of new data (for a review,
see Surprenant & Neath, 2008). Notwithstanding these empirical and theo-
retical difficulties, dual-store models remain popular.
2. Introspection
One reason for the acceptance of and continued belief in STM is a

compelling phenomenological experience: It certainly does feel as if we
have a limited capacity system in which items fade away once we stop
thinking about them. Moreover, when looking up a phone number, for
example, we certainly do seem to rehearse the item by repeatedly saying
the number over and over. A slight interruption, and the information is
gone. This corresponds nicely with Baddeley’s working memory (1986,
2000), and the operation of the articulatory loop and the phonological
store. In fact, this prototypical situation compares favorably to William
James’ (1890, pp. 608–609) description of primary memory. The phonolo-
gical loop does not act like memory proper, in which an item needs to be
retrieved; rather, an item in the phonological loop, just like an item in
primary memory, ‘‘is not thus brought back; it was never lost; its date
was never cutoff in consciousness from that of the immediately present
moment. In fact, it comes to us as belonging to the rearward portion of the
present space of time, and not to the genuine past.’’
Another current account of STM, the embedded processes model of

Cowan (1999), bears more than a passing similarity to an experience
described by Francis Galton (1883, p. 146):
There seems to be a presence-chamber in my mind where full conscious-

ness holds court, and where two or three ideas are at the same time in
audience, and an ante-chamber full of more or less allied ideas, which is
situated just beyond the full ken of consciousness.
According to the embedded processes view, working memory, the acti-

vated part of long-term memory, corresponds to Galton’s antechamber, and
the focus of attention corresponds to Galton’s presence chamber. Cowan
(2001) reviews much evidence to suggest that the capacity of the focus of
attention is approximately four items.
As compelling as this type of experience is, it is still introspection. As
Crowder (1993, p. 145) noted, ‘‘I share these intuitions . . . but I do not trust
them for a moment.’’ The feelings arise out of our processing, but this does
not make them an accurate description of the nature of the processing. The
reason is straightforward; introspection has had a ‘‘dismal record of failure’’
not only in psychology (Bower & Clapper, 1989, p. 245), but also in science
in general; after all, these intuitions are the same that ‘‘led to a firm belief in
a geocentric universe and a flat earth’’ (Crowder, 1993, p. 145). We do not
deny conscious awareness; we do, however, question whether it is suffi-
cient grounds for postulating a distinct memory system, given the absence
of other supporting data (see Surprenant & Neath, 2008).
3. Capacity limits
Even if one does not accept the notion of STM as a well-supported

concept, one still has to face the repeated observance of some fundamental
limit on the ability to remember certain types of information. Consider the
paradigm known as absolute judgment or absolute identification. In a
typical task, a set of stimuli are constructed that vary on only one dimen-
sion, such as nine tones that vary only in frequency or seven lines that vary
only in length. A label, usually a number, is associated with each item; for
example, tone 1 might have the lowest frequency and tone 9 the highest.
When a tone is played, the subject is asked to provide the label. People
cannot perform this task once the number of items exceeds about 7
(Miller, 1956; Shiffrin & Nosofsky, 1994). However, as soon as a second
dimension is added, performance increases dramatically.
Instead of invoking a box with a limited number of slots or a system in
which items decay rapidly, this limit is better understood in terms of
distinctiveness (Murdock, 1960; Neath & Brown, 2006). In a sequence of

items, the first and last are less likely to be confused with interior items
because they have no neighbors on one side. The closer to the middle of the
sequence, the more items there are on each side to confuse with the
intended item. If the same number of items vary along two dimensions,
then there is less chance of confusing one particular item for any other.1
Now, consider the prototypical task used to measure the capacity of
short-term memory: memory span. In this task, the subject is presented
with a list of items and is asked to recall the items in the same order in
which they were presented. This type of test, immediate serial recall,
remains one of the key paradigms for delineating the properties of short-
term memory. There are various ways to calculate memory span, but all
seem to converge around 7 – 2 items as the limit.
Our claim is that in the typical case, immediate serial recall tasks are
functionally equivalent to identifying items that vary systematically along
only one dimension. We admit that there are some important differences
between immediate serial recall and absolute identification; nonetheless,
because the items in the typical serial recall task do not differ from each
other by much—items are not very distinct from their neighbors—
performance runs into the same sort of limitation as in the absolute
identification tasks.
If this distinctiveness hypothesis is correct, then memory span should
decrease if the items are made more similar and thus relatively less dis-
tinct. This is exactly what happens. Memory span for similar sounding
letters is smaller than memory span for dissimilar sounding letters (e.g.,
Schweickert, Guentert, & Hersberger, 1990) or words (Surprenant, Neath, &
LeCompte, 1999). Similarly, adding another dimension dramatically
improves performance by increasing the difference (and thus the relative
distinctiveness) between items (e.g., Brooks & Watkins, 1990; Klein, 1976).
Another prediction of the distinctiveness hypothesis is that the same
capacity limits should appear regardless of whether the test is one thought
to tap short-term memory or one designed to tap long-term memory. For
two-syllable words, memory span is usually around 5 items (assuming that
systematic variation among the words is controlled). Nairne and Neath
(2001) adapted the traditional span task to a long-term setting by delaying
recall by 5 min. There are no accounts of short-term memory that allow it to
play a role after that length of time. Nonetheless, span was approximately 5
items. The key similarity between this task and the shorter-term version is
the requirement to recall items in order.
1
Neath, Brown, McCormack, Chater and Freeman (2006) provided a detailed description of a
how a model based on distinctiveness can account for absolute identification data.
4. Memory without STM
A third issue that contributes to the persistence of the construct of

short-term memory in the literature is a lack of appreciation for alternate
explanations of the data. Here, we highlight two approaches to memory
that do not invoke short-term memory. One is relatively old, and the other
quite new. Both, however, share the view that postulating a dedicated
memory system to retain information over the short term is not necessary.
4.1. A PROCESSING ACCOUNT
The processing view grows out of the proceduralist tradition (e.g., Bain,
1855; Kolers & Roediger, 1984) and includes both the levels of processing
approach (Craik & Lockhart, 1972) and the subsequent transfer appropriate
processing view (Morris, Bransford, & Franks, 1977). It emphasizes encod-
ing and retrieval processes instead of the system or location in which the
memory might be stored.
STM LTM
Format Phonological Semantic

Capacity Small Infinite?
Duration Seconds Minutes to years.
Consider the above table. If the top row is included, the table shows how
a structuralist—a proponent of dividing memory into different structures—
would explain the data. On the one hand, if information is in STM, it will
have a phonological format, the capacity will be small, and the information
cannot be retained for very long. On the other hand, if information is in
LTM, it will have a semantic format, the capacity will be nearly infinite, and
the information can be retained indefinitely. The processing account
deletes the top row and replaces the word ‘‘format’’ with ‘‘type of proces-
sing.’’ Thus, if you process something phonologically, you will not be able
to recall very much and the information will not be available for very long.
If you process something semantically, you will be able to recall far more
and for far longer intervals. Although the situation is more complex than
this—for example, we ignore interactions between encoding and retrieval
conditions—it is clear that instead of attributing a type of processing to a
particular structure, one can ignore the structure and use the type of
processing as the explanatory concept. By this analysis, the top row is
not needed and adds nothing to the explanation.
4.2. SIMPLE
As another alternative to invoking STM, consider a model of memory
known as SIMPLE: scale-independent memory and perceptual learning
(Brown, Neath, & Chater, 2007; Neath & Brown, 2006, 2007). As the name
suggests, it views memory as being scale-independent, that is, the same
principles of memory apply regardless of the timescale. Memory retrieval,
according to SIMPLE, is really discrimination of items in terms of their
position on one or more dimensions. Items with fewer close neighbors on
the relevant dimensions will be better remembered than items with more
close neighbors. SIMPLE is thus built on the intuitive idea of capacity
limitations described above.
Because space precludes a full description, we focus on how SIMPLE
explains free recall data. According to the model, people represent items in
a free recall task in terms of their time until recall. Importantly, these
temporal cues are on a log-transformed scale. This makes earlier items
generally less distinct than later items, naturally producing a recency effect
(enhanced recall of the last few items in the list). A primacy effect
(enhanced recall of the first one or two items) is produced because the
first item has no neighbors on one side and thus is more relatively distinct
than midlist items.
But how does the model produce scale independence? Consider two
lists of eight items. List A is presented at a rate of 1 item per second, and 1 s
elapses between the presentation of the final item and the test. The final list
item (item 8) has a value of 1 s because it is 1 s delayed from recall. The
seventh item has a value of 2 s (1 s between items 7 and 8, and 1 s for the
retention interval); the sixth item has a value of 3 s (again, adding 1 s for the
presentation rate) and so on. The temporal values for item 8 through item 1
are thus 8, 7, 6, 5, 4, 3, 2, and 1. These values are then log-transformed,
giving 2.079, 1.946, 1.792, 1.609, 1.386, 1.099, 0.693, and 0.000. List B has a
slower presentation rate, 20 s per item, and a longer retention interval, 20 s.
Thus, the final item has a value of 20 s, the seventh item has a value of 40 s,
the sixth item has a value of 60 s, and so on. The temporal values for item
8 through item 1 are thus 160, 140, 120, 100, 80, 60, 40, and 20. The
log-transformed values are 5.075, 4.942, 4.787, 4.605, 4.382, 4.094, 3.689,
and 2.996. The log-transformed values for both lists are shown in Fig. 1,
with the value for the first list item at the top and the last list item at the
bottom.2
2
In this example, we ignore the time taken to recall each item for ease of exposition. For full
details, see Brown, et al. (2007).
6.0
5.0
Represented Temporal Value 4.0

Early list items
3.0
2.0
Late list items

1.0
0.0
List A List B
Fig. 1. Representation, according to SIMPLE, of two eight-item lists presented for

free recall, assuming no rehearsal. The y-axis shows the log-transformed time until
retrieval, so earlier list items have larger values and end of list items have smaller
values. List A has a retention interval (RI) of 1 s and an interitem presentation
interval (IPI) of 1 s; List B has an RI of 20 s and an IPI of 20 s. The figure shows the
relative unimportance of absolute time.
In the model, the next step is to determine the relative distinctiveness of

each item by calculating how different it is from every other item in the list.
The difference between item 1 and item 2 in List A is 2.079 1.946 or 0.133.
The difference between item 1 and item 2 in List B is 5.075 4.942 or 0.133.
Because they have the same value, recall of each item will also be equiva-
lent. Within SIMPLE, the increase in absolute time from 1 to 20 s has no
effect on the relative differences in cases like this where durations are
multiplied by a constant. The figure shows this nicely: the relative spacing
of the dots is the same. The figure also illustrates why typical serial
position functions arise for recall. The first item has no neighbor on one
side, enhancing its relative discriminability and producing a primacy effect.
The final few items are more widely spaced enhancing their relative dis-
criminability, producing a recency effect. The curve will be asymmetrical,
with greater recency than primacy.
The figure also shows how SIMPLE addresses two other issues. First, if
the temporal spacing between items remains constant, but the time
between the final item and the test is increased, the recency effect is
reduced or eliminated whereas the primacy effect is relatively unaffected.
In terms of the figure, increasing only the retention interval will move the
final few items closer to the other list items. The enhanced status of the
first item is unaffected, as it remains the first item. Second, the figure also
illustrates the notion of capacity described earlier. If, for example, some of
the dots were blue instead of black, it would be easier to distinguish them
from their neighbors. Blue would represent adding a second dimension.
This example assumes that people do not rehearse the items other than
at the time of presentation. In the typical setting, however, people usually
do rehearse the items well after it is presented (Tan & Ward, 2000). It turns
out that this does not matter; if we use the time of last rehearsal as the key
value on the temporal dimension rather than the nominal time of presenta-
tion, SIMPLE still produces appropriate results (see Brown, et al., 2007).
More importantly, this acknowledgment that the functional temporal
value can be quite different from the nominal temporal value helps SIMPLE
explain differences in free recall by amnesic and control subjects. With
immediate free recall, the recency portion of the curve (i.e., the last three
positions) does not differ between amnesic patients and healthy control
subjects (e.g., Baddeley & Warrington, 1970). However, all other parts of
the list in immediate free recall are lower for patients than for subjects. The
traditional explanation is that STM is intact but transfer to LTM is impaired.
SIMPLE has no difficulty in simulating these results (Brown, Della Sala,
Foster, & Vousden, 2007; Brown & Lamberts, 2003). The critical assump-
tion has to do with when and how the items are rehearsed. Brown and
colleagues had an amnesic patient perform these tasks using the overt
rehearsal procedure (Tan & Ward, 2000). Brown and colleagues found
that the patient engaged in fixed rehearsal rather than the cumulative
rehearsal, which is characteristic of unimpaired free recall. When the
probability of recalling each item was plotted as a function of the temporal
distance of last rehearsal for each item, both controls and amnesic showed
extended recency and minimal primacy, and the resulting serial position
curves were well fit by SIMPLE.
Finally, we note that there are some intriguing parallels between the
approach taken with SIMPLE and with some theories developed to account
for learning in animals. For example, Gallistel and Gibbon (2002, p. 165)
argue that ‘‘the single most important quantitative fact about conditioning
discovered in a century of experimental work’’ is the timescale invariance
of the acquisition process. Timescale invariance is important in several
different accounts of animal conditioning (e.g., Gallistel & Gibbon, 2000;
Gibbon, Church, & Meck, 1984; Killeen & Taylor, 2000; Wearden, 1994).
5. Summary and conclusion
In our view, the amount of behavioral evidence requiring STM as an expla-

natory construct is fast diminishing: Theories of memory, like SIMPLE, are
being developed that explain STM-related phenomena using ‘‘factors known to
operate in LTM’’ (Melton, 1963, p. 8). We fully expect the range of application of
these models to expand such that the following cycle will be repeated: A case is
made for a set of findings that require a separate STM (Davelaar,
Goshen-Gottstin, & Ashkenazi, 2005) followed by a set of simulations that
show STM is simply not needed (Neath & Brown, 2006).
Acknowledgments
Preparation of this chapter was supported, in part, by grants from

NSERC to each author. The authors are listed alphabetically. Correspon-
dence may be addressed to either Ian Neath at ineath@mun.ca or Aimée M.
Surprenant at asurprenant@mun.ca.
References
Atkinson, R. C., & Shiffrin, R. M. (1968). Human memory: A proposed system and
its control processes. In K. W. Spence & J. T. Spence (Eds.), The Psychology
of Learning and Motivation: Advances in Research and Theory Volume 2
(pp. 89–195). New York: Academic Press.
Baddeley, A. D. (1986). Working Memory. New York: Oxford University Press.
Baddeley, A. D. (2000). The episodic buffer: A new component of working memory?
Trends in Cognitive Sciences, 4, 417–423.
Baddeley, A. D., & Warrington, E. K. (1970). Amnesia and the distinction between long-
and short-term memory. Journal of Verbal Learning and Verbal Behavior, 9, 176–189.
Bain, A. (1855). The Senses and the Intellect. London: John W. Parker and Son.
Bower, G. H., & Clapper, J. P. (1989). Experimental methods in cognitive science. In
M. I. Posner (Ed.), Foundations of Cognitive Science. Cambridge, MA: MIT Press,
pp. 245–300.
Broadbent, D. E. (1958). Perception and Communication. New York: Pergamon.
Brooks, J. O., & Watkins, M. J. (1990). Further evidence of the intricacy of memory
span. Journal of Experimental Psychology: Learning, Memory, and Cognition,
16, 1134–1141.
Brown, J. (1958). Some tests of the decay theory of immediate memory. Quarterly
Journal of Experimental Psychology, 10, 12–21.
Brown, G. D. A., & Lamberts, K. (2003). Double dissociations, models, and serial
position curves. Cortex, 39, 148–152.
Brown, G. D. A., Neath, I., & Chater, N. (2007). A temporal ratio model of memory.
Psychological Review, 114, 539–576.
Brown, G. D. A., Della Sala, S., Foster, J. K., & Vousden, J. I. (2007). Amnesia,
rehearsal, and temporal distinctiveness models of recall. Psychonomic Bulletin
& Review, 14, 256–260.
Burnham, W. H. (1888). Memory, historically and experimentally considered.
American Journal of Psychology, 2, 39–90, 255–270, 431–464, 566–622.
Cowan, N. (1999). An embedded-processes model of working memory. In A. Miyake &
P. Shah (eds.), Models of Working Memory: Mechanisms of Active Maintenance
and Executive Control. New York: Cambridge University Press, pp. 62– 101.
Cowan, N. (2001). The magical number 4 in short-term memory: A reconsideration

of mental storage capacity. Behavioral and Brain Sciences, 24, 87–185.
Craik, F. I. M., & Lockhart, R. S. (1972). Levels of processing: A framework for
memory research. Journal of Verbal Learning and Verbal Behavior, 11, 671–684.
Crowder, R. G. (1976). Principles of Learning and Memory. Hillsdale, NJ: Erlbaum.
Crowder, R. G. (1982). The demise of short-term memory. Acta Psychologica, 5, 291–323.
Crowder, R. G. (1993). Short-term memory: where do we stand? Memory &
Cognition, 21, 142–145.
Davelaar, E. J., Goshen-Gottstin, Y., & Ashkenazi, A. (2005). The demise of
short-term memory revisited: Empirical and computational investigations of
recency effects. Psychological Review, 112, 3–42.
Gallistel, C.R., & Gibbon, J. (2000). Time, rate, and conditioning. Psychological
Review, 107, 289–344.
Gallistel, C.R., & Gibbon, J. (2002). The Symbolic Foundations of Conditioned
Behavior, Mahwah, NJ: Erlbaum.
Galton, F. (1883). Inquiries into Human Faculty and its Development. London: Dent.
Gibbon, J., Church, R. M., & Meck, W. H. (1984). Scalar timing in memory. In
J. Gibbon & L. G. Allan (eds.), Timing and Time Perception (pp. 52–77).
New York: New York Academy of Sciences.
James, W. (1890). The Principles of Psychology. New York: Henry Holt and Company.
Killeen, P.R., & Taylor, T.J. (2000). How the propagation of error through stochastic
counters affects time discrimination and other psychophysical judgments.
Klein, K. (1976). Specifying the mechanisms in a levels-of-processing approach to
memory. Journal of Experimental Psychology: Human Learning and Memory,
2, 671–679.
Kolers, P. A., & Roediger, H. L., III. (1984). Procedures of mind. Journal of Verbal
Learning and Verbal Behavior, 23, 425–449.
Melton, A. W. (1963). Implications of short-term memory for a general theory of
memory. Journal of Verbal Learning and Verbal Behavior, 2, 1–21.
Miller, G. A. (1956). The magical number seven plus or minus two: some limits on
our capacity for processing information. Psychological Review, 63, 81–97.
Morris, C. D., Bransford, J. D., & Franks, J. J. (1977). Levels of processing versus
transfer appropriate processing. Journal of Verbal Learning and Verbal
Behavior, 16, 519–533.
Murdock, B. B. Jr. (1960). The distinctiveness of stimuli. Psychological Review, 67, 16–31.
Nairne, J. S. (2002). Remembering over the short-term: the case against the standard
model. Annual Review of Psychology, 53, 53–81.
Nairne, J. S., & Neath, I. (2001). Long-term memory span. Behavioral and Brain
Sciences, 24, 134–135.
Neath, I., & Brown, G. D. A. (2006). Simple: Further applications of a local distinc-
tiveness model of memory. In B. H. Ross (ed.), The Psychology of Learning and
Motivation (pp. 201–243). San Diego, CA: Academic Press.
Neath, I., & Brown, G. D. A. (2007). Making distinctiveness models of memory
distinct. In J. S. Nairne (ed.), The Foundations of Remembering: Essays in
Honor of Henry L. Roediger III (pp. 125–140). New York: Psychology Press.
Neath, I., Brown, G. D. A., McCormack, T., Chater, N., & Freeman, R. (2006).
Distinctiveness models of memory and absolute identification: Evidence for
local, not global, effects. Quarterly Journal of Experimental Psychology,
59, 121–135.
Peterson, L. R., & Peterson, M. J. (1959). Short-term retention of individual items.
Schweickert, R., Guentert, L., & Hersberger, L. (1990). Phonological similarity,
pronunciation rate, and memory span. Psychological Science, 1, 74–77.
Shiffrin, R. M., & Nosofksy, R. M. (1994). 7 Plus or minus 2: A commentary on
capacity limitations. Psychological Review, 101, 357–361.
Surprenant, A. M., & Neath, I. (2008). The 9 Lives of Short-Term Memory. In
A. Thorn & M. Page (eds.), Interactions between Short-Term and Long-Term
Memory in the Verbal Domain (pp. 16–43). Hove, UK: Psychology Press.
Surprenant, A. M., Neath, I., & LeCompte, D. C. (1999). Irrelevant speech,
phonological similarity, and presentation modality. Memory, 7, 405–420.
Tan, L., & Ward, G. (2000). A recency-based account of primacy effects in free
recall. Journal of Experimental Psychology: Learning, Memory, and Cognition,
26, 1589–1625.
Wearden, J. (1994). Prescriptions for models of biopsychological time. In M. Oaksford &
G.D.A. Brown (eds.), Neurodynamics and Psychology (pp. 215–236). London: Aca-
demic Press.
Human Learning 33
Hemispheric Asymmetries in Verbal

Memory
Karen M. Evans and Kara D. Federmeier
University of Illinois, Urbana-Champaign
Just as mapping brain structures to particular functions has helped build

our understanding of how the unified brain operates, studies of cerebral
asymmetries identify the functions of individual hemispheres in order to
better understand how they cooperate during normal processing. Although
the two cerebral hemispheres house similar anatomical structures, each is
specialized for certain types of processing (see Hellige, 1993, for review).
Nowhere is this functional asymmetry more evident than in speech produc-
tion, which for most humans seems to be entirely supported by the left
hemisphere (LH). More general language processes also tend to be left
lateralized, although the right hemisphere (RH) provides critical contribu-
tions (Beeman & Chiarello, 1998; Chiarello, 1988; Joanette, Goulet, &
Hannequin, 1990). Evidence that other cognitive operations—such as mem-
ory—are lateralized in a material-specific manner is less consistent. As
reviewed here, patient data and neuroimaging work provide support for
the idea that memory for verbal materials may be dominated by the LH,
but such studies also routinely uncover contributions from RH regions.
Similarly, studies using visual half-field (VF) presentation methods to exam-
ine processing asymmetries have obtained a mixture of LH and RH perfor-
mance advantages. Verbal memory thus seems to depend on bilateral
processing; accordingly, the focus of recent research has shifted from asking
which hemisphere is better at memory for certain materials to exploiting
asymmetric performance in order to identify the specific processes by which
each hemisphere encodes, retains, and retrieves verbal material.
1. Patient and neuroimaging studies
Initial investigations of cerebral asymmetries in verbal memory exam-

ined individuals with unilateral brain damage, inferring that deficits in
memory performance indicated critical contributions from the damaged
34 Karen M. Evans and Kara D. Federmeier
tissue. Insult to specific LH regions reliably leads to impaired performance

on of verbal memory tasks, such as the recall or recognition of studied
words, particularly when the infarction is to the medial temporal lobe
(Dobbins, Kroll, Tulving, Knight, & Gazzaniga, 1998; Milner, 1971, 1972;
Pillon, Bazin, Deveer, Ehrle, Baulac, & Dubois, 1999), or to posterior cortex
(Vilkki, 1987). Though damage to homologous areas of the RH often pro-
duces some verbal memory deficits relative to control participants (Dobbins
et al., 1998; Pillon et al., 1999), such lesions are more consistently asso-
ciated with impaired memory for nonverbal stimuli such as faces (Milner,
1968; Vilkki, 1987). Patients with frontal and prefrontal infarctions have
also been studied, but damage to these areas does not produce consistent
material-based dissociations; such patients instead manifest more global
memory impairments that are undifferentiated by material type (Milner,
1982; Milner & Petrides, 1984; Vilkki, 1987). Even in cases where temporal
or posterior injury produces strong asymmetries, damage to the LH does
not completely compromise verbal memory, nor does it leave nonverbal
memory abilities intact (Dobbins et al., 1998; Pillon et al., 1999). As such, it
seems that memory for both verbal and nonverbal materials relies on some
degree of contribution from both hemispheres.
The inability to ensure that infarctions are limited to a given anatomical
region warrants caution in interpreting lesion data and suggests the need to
corroborate such findings with other methods. Results from hemodynamic
neuroimaging studies are generally consistent with the neuropsychological
literature in showing that the relative strength of activation of each hemi-
sphere varies with stimulus type. Functional magnetic resonance imaging
(fMRI) studies have found that in the dorsolateral prefrontal cortex and
medial temporal lobe, encoding of words is largely left-lateralized, whereas
encoding of nonverbal materials, such as faces or abstract patterns, pro-
duces right-lateralized activation (Golby et al., 2001; Kelley, et al., 1998;
McDermott, Buckner, Petersen, Kelley, & Sanders, 1999; Wagner, Poldrack,
et al., 1998). When encoding trials are sorted by whether they are later
remembered or forgotten during the test phase, left-lateralized activation in
both prefrontal and temporal regions is predictive of later successful
retrieval of study words (Wagner, Schacter, et al., 1998), and successful
encoding of scenes is associated with right-lateralized prefrontal but bilat-
eral hippocampal activation (Brewer, Zhao, Desmond, Glover, & Gabrieli,
1998). Critically, although neuroimaging studies have consistently found
such material-based asymmetries, encoding activity is not unilateral, as all
stimulus types produce some activation in both hemispheres. Thus, con-
sistent with patient data, neuroimaging studies suggest that regions from
both hemispheres are recruited during memory processing for all material
types, though the LH seems specialized for verbal memory and the right for
spatial memory (e.g., Golby et al., 2001).
2. Visual half-field experiments
Additional evidence of a LH bias for verbal memory has been obtained

using the VF technique to bias the perception and initial processing of
stimuli to one hemisphere. This technique takes advantage of the fact
that information presented peripherally (more than half a visual degree
from fixation) is received exclusively by the contralateral hemisphere.
Although information can eventually be transferred to the other hemi-
sphere across the corpus callosum and other commissures, the temporal
and information-quality advantage afforded to the hemisphere initially
receiving the information biases even later stages of processing. Thus,
measures obtained for stimuli presented in the right visual half-field (rvf )
reflect LH processing biases and measures obtained for stimuli presented
in the left visual half-field (lvf ) reflect RH processing biases (see Banich,
2003; Beaumont, 1982 for review).
Many studies using this method to examine verbal memory briefly pre-
sent a stimulus to one VF and then require either a verbal response
identifying the stimulus or an immediate forced-choice recognition judg-
ment, as participants select the studied stimulus from an array containing
the target along with distractor items. This particular design is somewhat
more perceptual in nature than memory-based, but allows for an assess-
ment of how and how well each hemisphere is initially able to encode
different kinds of material. Such studies have found consistent rvf/LH
advantages for verbal stimuli. When verbal targets must be named, partici-
pants are more accurate to report digits (Geffen, Bradshaw, & Wallace,
1971), letters (Kimura, 1966), and words (Leehey & Cahn, 1979) presented
in the rvf/LH than in the lvf/RH. When targets must be identified in an
immediate forced-choice test, words (Pirozzolo & Rayner, 1977) and pro-
nounceable pseudowords (Fontenot & Benton, 1972) are more accurately
identified if they have been studied in rvf/LH.
By imposing a delay between the perception of the lateralized target
and the presentation of the subsequent test probe, memory demands can be
increased. Under such conditions, Hannay and Malone (1976) reported that
pseudowords are more accurately identified if they have been studied in
the rvf/LH, though this asymmetry was significant only under delayed
conditions. Similar delay-dependent asymmetries have been reported for
lvf/RH advantages in memory for complex line drawings (Bevilacqua,
Capitani, Luzzatti, & Spinnler, 1979; Dee & Fontenot, 1973) and in the
detection of target faces (Moscovitch, Scullion, & Christie, 1976).
The effect of manipulating retention intervals has also been investigated
by testing memory after a series of lateralized items has been presented.
Serial position in this string of stimuli then enables assessment of how time
and interference affect asymmetries. This design has revealed superior recall
for rvf/LH-studied letters (Hines, Satz, & Clementino, 1973) and digits (Hines &
Satz, 1971; Hines, Satz, Schell, & Schmidlin, 1969), which is magnified with
increased retention interval and/or interference. Thus, studies with immediate
testing suggest that there may be standing perceptual biases favoring rvf/LH
presentation of verbal materials and lvf/RH presentation of nonverbal materi-
als, and manipulations of retention interval and number of interfering stimuli
suggest that such asymmetries may increase with memory demands.
Berrini, Della Sala, Spinnler, Sterzi, and Vallar (1982) employed a recogni-
tion test to determine whether lateralizing the study or test phase of a
memory task produced greater asymmetries. When the study phase was
lateralized, verbal materials (meaningless and unpronounceable consonant
pairs) were identified with greater accuracy following rvf/LH encoding, and
nonverbal materials (stars positioned in a matrix) were recognized better
following lvf/RH encoding. When stimuli were studied centrally and tested
laterally, no VF asymmetries emerged for either material type. Thus, this
study suggests that hemisphere-biased encoding produces greater asymme-
tries in verbal (and spatial) memory than biased retrieval, as is consistent
with the neuroimaging literature discussed previously. However, other stu-
dies that lateralized both study and test words have reported critical influ-
ences from the VF of retrieval probes. Although the exact pattern of
recognition performance as a function of study–test VF pairs varies across
studies, this work suggests that test VF effects tend to override study VF
effects, such that rvf/LH test words are recognized faster and more accurately
than lvf/RH test words, regardless of study VF (Berrini, Capitani, Della
Sala, & Spinnler, 1984; Coney & MacDonald, 1988; but see also Leibner, 1982).
As mentioned earlier, material-based VF asymmetries are often observed
even under conditions imposing low memory demands. Although unilater-
ally, presented individual letters and nonpronounceable letter strings
are typically identified with equal facility in the rvf and lvf (Jordan &
Patching, 2004; Jordan, Patching, & Thomas, 2003; Young, Ellis, & Bion,
1984), words and even pseudowords elicit strong rvf/LH advantages (Jordan
et al., 2003; Mishkin & Forgays, 1952; Young et al., 1984). It is possible that
such asymmetries stem from a LH advantage in the use of top-down infor-
mation to aid identification of word and word-like stimuli. Though more
work is needed to elucidate the precise causes of such perceptual asymme-
tries, it is important to note that these effects can obscure memory biases, as
lvf/RH words are more likely to be misread or to require more processing
time than rvf/LH words. When a stimulus is misperceived (i.e., a word is
misread) in a memory task, a memory judgment based on this misperception
will be recorded as a memory error, even though the error may technically
reflect a correct memory judgment based on an incorrect perceptual analy-
sis. Perceptual biases therefore complicate the interpretation of memory
data and license the use of more diverse methods.
3. Manipulations of visual word form
Although LH advantages for word recognition seem fairly ubiquitous,

there are conditions under which RH advantages emerge. Identifying the
verbal memory strengths of the ‘‘linguistically weaker’’ RH has been a goal
of a number of recent studies. In line with evidence that the RH is biased to
encode spatial material, RH advantages for verbal processing have been
found to emerge under conditions in which operations on a verbal stimulus
can be optimized by treating the stimulus as a physical, rather than seman-
tic, entity. Geffen, Bradshaw, and Nettleton (1972) reported that when
participants must match stimuli that share verbal codes but not physical
appearance (e.g., matching a capital to a lowercase exemplar of the same
letter), the typical rvf/LH advantage for verbal materials is obtained. For
physically identical instances (e.g., both capital letters), however, the lvf/
RH letter pairs are matched more quickly (Geffen et al., 1972). Similarly,
Gibson, Dimond, and Gazzaniga (1972) required participants to determine
whether a lateralized word was the same as or different from a word
studied in central vision and found superior match detection with lvf/RH
presentation. As such a task could be performed via a strategy using
linguistic (semantic or phonetic codes) or perceptual (physical stimulus
properties) analysis, results from matching studies (Geffen et al., 1972;
Gibson et al., 1972) suggest that RH spatial abilities may benefit RH
performance in certain verbal tasks. Evidence that the RH is biased to
encode words as pictorial objects was also obtained by Pirozzolo and Rayner
(1977), who found that forced-choice recognition errors to lvf/RH-studied
words were dominated by visually confusable foils. Lvf/RH advantages have
also been reported for verbal stimuli presented in noise (Hellige, 1980;
Hellige & Webster, 1979) or in a novel font (Bryden & Allard, 1976). As
perceptually degraded and unusual letters are novel to participants and
therefore require more processing, such conditions may render RH spatial
abilities even more critical (Bryden & Allard, 1976).
The RH’s fluency with perceptually identical verbal stimuli has also been
suggested as the basis for performance asymmetries observed in stem
completion for both implicit and explicit memory tests. After studying a
list containing both uppercase and lowercase words, participants were
given lateralized word stems that did or did not maintain the case of the
study word from which the stems were derived. Different case stems were
just as likely to be completed by a studied word when they appeared in the
lvf/RH or rvf/LH (Marsolek, Kosslyn, & Squire, 1992; Marsolek, Squire,
Kosslyn, & Lulenski, 1994). However, when the letter case of a study word
was maintained in the test stem, higher completion rates were obtained for
lvf/RH items, both for implicit tests of memory (Marsolek et al., 1992) and
for explicit tests that render form-based encoding strategically sensible
(e.g., when all study words appear in the same letter case: Marsolek et al.,
1994). The RH also yields a higher stem completion rate when the
visual surroundings of the studied word are replicated at test; thus, the RH
benefits from study–test physical similarity of not only the stimulus itself, but
also the context, implicating a more holistic approach to RH encoding
processes (Marsolek, Schacter, & Nicholas, 1996). This body of evidence
has led to the proposal that the RH is biased to encode verbal material in a
veridical manner that maintains form-specific elements of the physical sti-
mulus, whereas the LH is biased to abstract across distinctive physical
properties, forming a more idealized representation that contains shared
physical features of conceptually similar stimuli (Marsolek & Burgund, 1997).
4. False memory experiments
The idea that the RH may hold onto stimulus information that the LH
discards could have important general implications for verbal memory. For
example, although the theory posited by Marsolek and colleagues
(e.g., Marsolek & Burgund, 1997) refers to abstraction across visual proper-
ties rather than at a semantic level, several false memory studies have
reported enhanced discrimination between semantically similar words
when these are presented to the lvf/RH, suggesting that the tendency to
encode verbal stimuli in a more veridical manner may influence even
higher levels of stimulus analysis. Testing a callosotomy patient, Metcalfe,
Funnell, and Gazzaniga (1995) found that unstudied test words (e.g., plum)
that were semantically related to studied words (e.g., apple, peach) were
more accurately rejected with lvf/RH presentation (words were presented
centrally at study and were thus available separately to each hemisphere of
this patient; lateralized test words then sampled the responses of a single
hemisphere). Metcalfe et al. (1995) therefore proposed that only the RH
retains specific information about the visual form of a word, which enables
distinction between studied and unstudied test words that are similar in
meaning; in contrast, the LH is assumed to transform the physical word
into a gist-like semantic representation, even incorporating contextual and
inferential details, so that the resulting representation cannot distinguish
between the actual word studied and semantically related words that fit
with the constructed representation.
Consistent with this interpretation, Fabiani, Stadler, and Wessels (2000)
found that when brain-intact participants do make semantically related mem-
ory errors, only those false memories supported by RH encoding processes
show a neural signature that is distinct from true memories. To promote
semantic errors, Fabiani et al. (2000) employed the DRM method (named for
Deese, 1959; Roediger & McDermott, 1995) in which participants view lists
of related words (e.g., pane, shade, ledge, view, frame, curtain) that all
converge onto one critical target (window).
In Fabiani et al.’s design, the words associated with each target were
always lateralized to the same VF, so that false memories were associated
with encoding initially biased to one hemisphere. The critical comparison
was between targets that had been presented in the study phase (true
memories when endorsed at test) and target lures that were not studied
(false memories when endorsed). Event-related potential (ERP) recordings
to test words showed that only those inappropriately endorsed lures whose
corresponding lists had been studied in the lvf/RH elicited responses that
differed from those to hits; false alarms to lures whose associated lists had
been studied in the rvf/LH elicited ERP responses that were identical to
those of hits. Thus, although a series of highly related words presented to
either VF can elicit many false memories, only those associated with RH
encoding are neurally distinct from true memories.
Behavioral studies of false memory that have used centrally presented
visual (Ito, 2001) or auditory (Westerberg & Marsolek, 2003) study words
and lateralized visual test words have failed to find lvf/RH advantages for
rejecting related words, suggesting that such effects may depend on
hemispheric biases at encoding rather than at test.
5. Long repetition lag experiments
Even when form-based information is not necessary to reject lure items,

retention of specific form information may assist recognition judgments in
standard memory tests, particularly when other memory signals weaken.
Employing a broad range of repetition lags (1, 2, 3, 5, 7, 10, 20, 30, 50),
which extend beyond those used in prior tests of memory asymmetries,
Federmeier and colleagues have identified several advantages for words
studied in the lvf/RH that occur only after long repetition lags. In a
behavioral experiment that stressed the speed of responses, Federmeier
and Benjamin (2005) found that the response time for hits increased more
sharply with lag for rvf/LH-studied words than for lvf/RH-studied words,
such that at the longest lags (10, 20, 30, 50) the typical speed advantage
favoring rvf/LH items reversed, with faster responses to lvf/RH-studied
words. Evans and Federmeier (2007) recorded ERPs in the same design
and found that for long lag hits (20, 30, 50), the memory signal associated
with correct recognition was larger for lvf/RH-studied words than for rvf/
LH-studied words. At short lags, however, memory signals did not differ
based on study VF, bolstering the idea that the rvf/LH response time and
accuracy advantages observed at shorter lags may reflect superior percep-
tion rather than superior memory.
Interestingly, Evans and Federmeier also uncovered asymmetries on an

early ERP effect (P2 repetition effect) associated with visual analysis and
target detection (Luck & Hillyard, 1994). This effect occurred only for hits
studied in the lvf/RH, suggesting that it may reflect perceptual facilitation
afforded by form-specific RH memory representations. The fact that this
apparently form-based memory signal was retained across the lag structure
suggests that it may underlie the RH advantages seen in more explicit
measures at the longest lags.
6. Conclusion
In conclusion, although there seem to be biases in the type of information

each hemisphere processes most adeptly, effects seen in stem completion,
false memory, and repetition lag studies suggest that both hemispheres
contribute to verbal memory, albeit differently. The data thus far suggest
that the RH may tend to encode verbal stimuli more veridically, whereas the
LH may tend to rapidly abstract away from the input, both at perceptual and
at semantic levels of analysis. These encoding differences have conse-
quences for what aspects of verbal stimuli can later be recovered, for the
time course with which information about words will be retained, and for
the types of errors that are likely to be made in memory tasks, among others.
In turn, hemispheric differences in verbal memory have important implica-
tions for language processing (see, e.g., Federmeier, 2007), as comprehen-
ders retain word and message-level information, build and link syntactic
structures, map between referents in a discourse, and reanalyze words and
phrases to appreciate humor and other nonliteral language. A shift away
from the search for global asymmetries and towards an understanding of
how each hemisphere is specialized to process the same material thus
promises to enrich our understanding of how the unified human brain
perceives, remembers, and uses words.
References
Banich, M. T. (2003). Interaction between the hemispheres and its implications for
the processing capacity of the brain. In R. Davidson & K. Hugdahl (eds.), Brain
Asymmetry (2nd ed., pp. 261–302). Cambridge, MA: MIT Press.
Beaumont, J. G. (ed.), (1982). Divided Visual Field Studies of Cerebral
Organisation. New York: Academic Press Inc.
Beeman, M., & Chiarello, C. (eds.), (1998). Right Hemisphere Language
Comprehension: Perspectives from Cognitive Neuroscience. Mahwah, NJ: Erlbaum
Associates.
Berrini, R., Capitani, E., Della Sala, S., & Spinnler, H. (1984). Interaction between
lateralization of memory and probe stimulus in the recognition of verbal and
spatial visual stimuli. Neuropsychologia, 22(4), 517–520.
Berrini, R., Della Sala, S., Spinnler, H. R., Sterzi, R., & Vallar, G. (1982). In eliciting
hemisphere asymmetries which is more important: The stimulus input side or
the recognition side? A tachistoscopic study on normals. Neuropsychologia,
20(1), 91–94.
Bevilacqua, L. Capitani, E. Luzzatti, C., & Spinnler, H. R. (1979). Does the
hemisphere stimulated play a specific role in delayed recognition of complex
abstract patterns? A tachistoscopic study. Neuropsychologia, 17, 93–97.
Brewer, J. B., Zhao, Z., Desmond, J. E., Glover, G. H., & Gabrieli, J. D. E. (1998).
Making memories: Brain activity that predicts how well visual experience will be
remembered. Science, 281, 1185–1187.
Bryden, M. P., & Allard, F. (1976). Visual hemifield differences depend on
typeface. Brain and Language, 3, 191–200.
Chiarello, C. (1988). Right Hemisphere Contributions to Lexical Semantics. New
York: Springer-Verlag.
Coney, J., & MacDonald, S. (1988). The effect of retention interval upon
hemispheric processes in recognition memory. Neuropsychologia, 26, 287–295.
Dee, H. L., & Fontenot, D. J. (1973). Cerebral dominance and lateral differences in
perception and memory. Neuropsychologia, 11, 167–173.
Deese, J. (1959). On the prediction of occurrence of particular verbal intrusions in
immediate recall. Journal of Experimental Psychology: Human Perception and
Performance, 58, 17–22.
Dobbins, I. G., Kroll, N. E. A., Tulving, E., Knight, R. T., & Gazzaniga, M. S. (1998).
Unilateral medial temporal lobe memory impairment: type deficit, function defi-
cit, or both? Neuropsychologia, 36(2), 115–127.
Evans, K. M., & Federmeier, K. D. (2007). The memory that’s right and the memory
that’s left: Event-related potentials reveal hemispheric asymmetries in the encod-
ing and retention of verbal information. Neuropsychologia, 45(8), 1777–1790.
Fabiani, M., Stadler, M. A., & Wessels, P. M. (2000). True but not false memories
produce a sensory signature in human lateralized brain potentials. Journal of
Cognitive Neuroscience, 12, 941–949.
Federmeier, K. D. (2007). Thinking ahead: The role and roots of prediction in
language comprehension. Psychophysiology, 44, 491–505.
Federmeier, K. D., & Benjamin, A. S. (2005). Hemispheric asymmetries in the time
course of recognition memory. Psychonomic Bulletin and Review, 12(6), 993–998.
Fontenot, D. J., & Benton, A. L. (1972). Perception of direction in the right and left
visual fields. Neuropsychologia, 10, 447–452.
Geffen, G., Bradshaw, J. L., & Nettleton, N. C. (1972). Hemispheric asymmetry:
Verbal and spatial encoding of visual stimuli. Journal of Experimental
Psychology, 95(1), 25–31.
Geffen, G., Bradshaw, J. L., & Wallace, G. (1971). Interhemispheric effects on
reaction time to verbal and nonverbal visual stimuli. Journal of Experimental
Psychology, 87(3), 415–422.
Gibson, A. R., Dimond, S. J., & Gazzaniga, M. S. (1972). Left field superiority for
word matching. Neuropsychologia, 10, 463–466.
Golby, A. J., Poldrack, R. A., Brewer, J. B., Spencer, D., Desmond, J. E., Aron, A. P.,
et al. (2001). Material-specific lateralization in the medial temporal lobe and
prefrontal cortex during memory encoding. Brain, 124, 1841–1854.
Hannay, H. J., & Malone, D. R. (1976). Visual field effects and short-term memory for
verbal material. Neuropsychologia, 14, 203–209.
Hellige, J. B. (1980). Effects of perceptual quality and visual field of probe stimulus
presentation on memory search for letters. Journal of Experimental Psychology:
Human Perception and Performance, 6(4), 639–651.
Hellige, J. B. (1993). Hemispheric Asymmetry: What’s Right and What’s Left?
Cambridge, MA: Harvard Press.
Hellige, J. B., & Webster, R. (1979). Right hemisphere superiority for initial stages of
letter processing. Neuropsychologia, 17, 653–660.
Hines, D., & Satz, P. (1971). Superiority of right visual half-fields in right-handers for
recall of digits presented at varying rates. Neuropsychologia, 9, 21–25.
Hines, D., Satz, P., & Clementino, T. (1973). Perceptual and memory components of
the superior recall of letters from the right visual half-fields. Neuropsychologia, 11,
175–180.
Hines, D., Satz, P., Schell, B., & Schmidlin, S. (1969). Differential recall of digits in
the left and right visual half-fields under free and fixed order of recall. Neurop-
sychologia, 7, 13–22.
Ito, Y. (2001). Hemispheric asymmetry in the induction of false memories.
Laterality, 6(4), 337–346.
Joanette, Y., Goulet, P., & Hannequin, D. (eds.). (1990). Right Hemisphere and
Verbal Communication. New York: Springer-Verlag.
Jordan, T. R., & Patching, G. R. (2004). What do lateralized displays tell us about
visual word perception? A cautionary indication from the word-letter effect.
Neuropsychologia, 42, 1504–1514.
Jordan, T. R., Patching, G. R., & Thomas, S. M. (2003). Assessing the role of hemi-
spheric specialization, serial-position processing, and retinal eccentricity in later-
alized word recognition. Cognitive Neuropsychology, 20, 49–71.
Kelley, W. M., Miezin, F. M., McDermott, K. B., Buckner, R. L., Raichle, M. E., Cohen, N. J.,
et al. (1998). Hemispheric specialization in human dorsal frontal cortex and medial
temporal lobe for verbal and nonverbal memory encoding. Neuron, 20, 927–936.
Kimura, D. (1966). Dual functional asymmetry of the brain in visual perception.
Leehey, S. C., & Cahn, A. (1979). Lateral asymmetries in the recognition of words,
familiar faces, and unfamiliar faces. Neuropsychologia, 17, 619–627.
Leibner, L. (1982). Interhemispheric effects in short-term recognition memory for
single words. Cortex, 18, 113–124.
Luck, S. J., & Hillyard, S. A. (1994). Electrophysiological correlates of feature
analysis during visual search. Psychophysiology, 31, 291–308.
Marsolek, C. J., & Burgund, E. D. (1997). Computational analyses and hemispheric
asymmetries in visual-form recognition. In S. Christman, (ed.), Cerebral Asymmetries
in Sensory and Perceptual Processing (pp. 125–158). Amsterdam: Elsevier Science.
Marsolek, C. J., Kosslyn, S. M., & Squire, L. R. (1992). Form-specific visual priming
in the right cerebral hemisphere. Journal of Experimental Psychology:
Learning, Memory, and Cognition, 18(3), 492–508.
Marsolek, C. J., Schacter, D. L., & Nicholas, C. D. (1996). Form-specific visual

priming for new associations in the right cerebral hemisphere. Memory and
Cognition, 24, 539–556.
Marsolek, C. J., Squire, L. R., Kosslyn, S. M., & Lulenski, M. E. (1994).
Form-specific explicit and implicit memory in the right cerebral hemisphere.
Neuropsychology, 8(4), 588–597.
McDermott, K. B., Buckner, R. L., Petersen, S. E., Kelley, W. M., & Sanders, A. L.
(1999). Set- and code-specific activation in the frontal cortex: An fMRI study of
encoding and retrieval of faces and words. Journal of Cognitive Neuroscience,
11(6), 631–640.
Metcalfe, J., Funnell, M., & Gazzaniga, M. S. (1995). Right-hemisphere
memory superiority: Studies of a split-brain patient. Psychological Science,
6, 157–164.
Milner, B. (1968). Visual recognition and recall after right temporal-lobe excision in
man. Neuropsychologia, 6, 191–209.
Milner, B. (1971). Interhemispheric differences in the localization of psychological
processes in man. British Medical Bulletin, 27, 272–277.
Milner, B. (1972). Disorders of learning and memory after temporal lobe lesions in
man. Clinical Neurosurgery, 19, 421–446.
Milner, B. (1982). Some cognitive effects of frontal-lobe lesions in man. Philosophi-
cal transactions of the Royal Society of London. Series B, Biological Sciences,
298, 211–226.
Milner, B., & Petrides, M. (1984). Behavioural effects of frontal-lobe lesions in man.
Trends in Neurosciences, 7, 403–407.
Mishkin, M., & Forgays, D. G. (1952). Word recognition as a function of retinal locus.
Moscovitch, M., Scullion, D., & Christie, D. (1976). Early versus late stages of
processing and their relation to functional hemispheric asymmetries in face
recognition. Journal of Experimental Psychology: Human Perception and
Performance, 2(3), 401–416.
Pillon, B., Bazin, B., Deveer, B., Ehrle, N., Baulac, M., & Dubois, B. (1999). Specifi-
city of memory deficits after right or left temporal lobectomy. Cortex, 35, 561–
571.
Pirozzolo, F. J., & Rayner, K. (1977). Hemispheric specialization in reading and
word recognition. Brain and Language, 4, 248–261.
Roediger, H. L. I., & McDermott, K. B. (1995). Creating false memories: Remember-
ing words not presented in lists. Journal of Experimental Psychology: Learning,
Memory, and Cognition, 21, 803–814.
Vilkki, J. (1987). Incidental and deliberate memory for words and faces after focal
cerebral lesions. Neuropsychologia, 25, 221–230.
Wagner, A. D., Poldrack, R. A., Eldridge, L. L., Desmond, J. E., Glover, G. H., &
Gabrieli, J. D. E. (1998). Material-specific lateralization of prefrontal
activation during episodic encoding and retrieval. NeuroReport, 9(16),
3711–3717.
Wagner, A. D., Schacter, D. L., Rotte, M., Koutstaal, W., Maril, A., Dale, A. M., et al.
(1998). Building memories: Remembering and forgetting of verbal experiences as
predicted by brain activity. Science, 281, 1188–1191.
Westerberg, C. E., & Marsolek, C. J. (2003). Hemispheric asymmetries in memory

processes as measured in a false recognition paradigm. Cortex, 39, 627–642.
Young, A. W., Ellis, A. W., & Bion, P. J. (1984). Left hemisphere superiority for
pronounceable nonwords, but not for unpronounceable letter strings. Brain and
Language, 22, 14–25.
Human Learning 45
Emotional Facilitation
and Disruption of Memory
Stacie L. Warren, Gregory A. Miller, and Wendy Heller
University of Illinois at Urbana-Champaign
Cognitive performance varies as a function of phenomena convention-

ally considered ‘‘emotional’’ in various ways. A large body of research
demonstrates that, to the extent that cognition and emotion can be distin-
guished (see Miller, 1996, for reservations about that), emotion modulates
memory, learning, attention, and executive function (e.g., Bar-Haim, Lamy,
Pergamin, Bakersmans-Kranenburg, & van Ijzendoorn, 2007; Cahill &
McGaugh, 1995; Davidson, 2002; Engels et al., 2007; Gray, 2004; Gray,
Braver, & Raichle, 2002; Heller & Nitschke, 1997; Herrington, Koven, Heller,
Miller, & Nitschke, in press; Herrington, Koven, Miller, & Heller, 2006;
Herrington et al., 2005; Levin, Heller, Mohanty, Herrington, & Miller, 2007;
Mohanty et al., 2005, 2007). Emotion can have an impact on various cogni-
tive mechanisms that influence memory depending on the nature of the
task, the type of emotion, and the circumstances under which the indivi-
dual is engaged in the cognitive work. Furthermore, emotion can affect
memory mechanisms via state conditions (e.g., anxious arousal) as well as
via more chronic or trait conditions (e.g., dysthymia, posttraumatic stress
disorder, avoidance temperament). These can combine to have distinct
influences on memory, such as in the case where a person who is
depressed and prone to anxious apprehension or worry becomes anxiously
aroused (Nitschke, Heller, Palmieri, & Miller, 1999).
1. Dimensional vs. categories approaches to emotion
To understand the diverse relationships between emotion, memory, and

relevant brain function, one strategy has been to explore emotion in terms
of more fundamental components. Major approaches have either placed
emotion in a multidimensional space or cast it as a series of categories,
46 Stacie L. Warren, Gregory A. Miller, and Wendy Heller
either lacking a specific relational structure or employing a simple hier-

archy. Dimensional models have more often been applied to brain data,
including variations on the circumplex model of emotion, in which emotion
is represented as two orthogonal components such as valence and arousal
(Lang, Greenwald, Bradley, & Hamm, 1993; Russell, 1980). The valence
dimension varies from pleasant to unpleasant, with neutral in the middle,
and arousal refers to a continuum that varies from calm to excitement.
Relying on converging evidence from neuropsychological, peripheral phy-
siological, and brain hemodynamic studies as well as developing notions
of brain lateralization and emotion (e.g., Davidson, 1984; Tucker, Stenslie,
Roth, & Shearer, 1981), Heller (1990, 1993) and Heller, Nitschke, Etienne, &
Miller (1997) proposed a neuropsychological model of emotion in which
valence is associated with differential activity in anterior cortical regions
(pleasantness with more left than right activity and unpleasantness with
more right than left), and arousal is associated with activity in more
posterior, right parietotemporal cortex. The right posterior system was
hypothesized to operate distinctly from but in concert with the frontal
lobe. How the interaction of these systems plays out was hypothesized to
bea function of individual differences in affective style (including features
of personality and psychopathology) that contribute to diverse emotional
consequences for cognition (Heller, 1990). Variants of a valence/arousal
approach have been proposed as well. For example, a rotation of the
valence and arousal axes produces positive and negative affect dimensions
(Clark & Watson, 1991), discussed below. Related conceptualizations
emphasize appetitive and defensive (Lang, Bradley, & Cuthbert, 1997) or
approach and avoidance motivation (Davidson & Irwin, 1999) as organiza-
tional principles for brain function in emotion (for comparative reviews,
see Elliott & Thrash, 2002, and Shankman & Klein, 2003).
To date, such dimensional approaches have proven more fruitful than a
categorical approach in which different emotions are presumed to be
instantiated in different regions of the brain (e.g., happiness in one area,
anger in another, fear in a third). Although it is indeed the case that
particular brain areas play a more prominent role in particular emotions,
such approaches have not had much to say about specific brain mechan-
isms likely to operate in such spatially discrete regions and able to differ-
entiate emotions qualitatively. Another trend in the literature is away from
the traditional relegation of neural factors in emotion to subcortical and
phylogenetically old cortical structures. Thus, complementing an early
emphasis on the limbic system and particularly the amygdala for emotional
processing, other newer regions such as prefrontal cortex (PFC) are receiv-
ing more attention (Dolcos, LaBar, & Cabeza, 2004). Further, as tools for
analyzing macrolevel interregional brain connectivity advance, the empha-
sis is moving to multiregional networks. This, in turn, fosters an
appreciation of emotion and cognition as thoroughly interacting or perhaps

as not even distinct types of phenomena (Miller, 1996; Miller, Engels, &
Herrington, 2007; Mohanty et al., 2007).
Arousing events, whether positive or negative, are remembered better
than neutral events. This effect has been attributed to the ‘‘memory-
modulation’’ hypothesis about the amygdala, in which this brain structure
is thought to play a particular role in memory for emotionally arousing
information (Cahill, 2000; McGaugh, Ferry, Vazdarjanova, & Roozendaal,
2000). Converging functional brain imaging studies support the hypothesis
that the amygdala modulates memory storage processes involving other
brain regions (Cahill et al., 1996; Canli, Zhao, Brewer, Gabrieli, & Cahill,
2000; Canli, Desmond, Zhao, & Gabrieli, 2002; Dolcos, Graham, LaBar, &
Cabeza, 2003; Dolcos, Labar, & Cabeza, 2004; Hamann, Ely, Grafton, &
Kilts, 1999; Mohanty et al., 2005, 2007). Since limbic regions such as the
amygdala project to PFC (Barbas, 2000), it is reasonable to assume that
presentation of emotional stimuli can modulate PFC activity during mem-
ory processes. Using an fMRI paradigm, Dolcos et al. (2004) found that
during emotional evaluation, PFC activity showed a hemispheric asymme-
try consistent with the valence hypothesis (left PFC activity greater for
positive than negative pictures, right PFC greater for negative than positive
pictures). In addition, dorsomedial PFC activity was sensitive to arousal,
whereas ventromedial PFC activity was sensitive to positive valence. Suc-
cessful encoding was enhanced by arousal in left ventrolateral and dorso-
lateral PFC regions, hypothesized to reflect an enhancement of strategic,
semantic, and working memory processes. These results suggest that PFC
regions play an important role in the evaluation of emotional stimuli and
are sensitive to both valence and arousal. Using an emotion-word Stroop
fMRI paradigm, Herrington et al. (2005) further supported the prediction of
an association between the valence dimension and frontal laterality in an
unselected sample, and Engels et al. (2007) did so in subjects selected for
high or low anxiety.
2. Depression and memory
Many of the findings in the neuropsychology and cognitive neuroscience

literatures bearing on whether valence affects memory to the extent emo-
tional arousal does come from studies of psychopathology. Emotional
stimulus qualities appear to enhance or impair memory depending on the
nature of the task and the roles of the brain regions involved. For example,
executive functions are enhanced by positive affect associated with activ-
ity in regions of the left PFC (Ashby, Isen, & Turken, 1999; Engels et al.,
2007; Herrington et al., 2005). In contrast, depression has been associated
with impaired memory performance, via prefrontal deactivation with con-

sequent impairment in the use of mnemonic strategies (Heller & Nitschke,
1997; Levin et al., 2007; Mohanty & Heller, 2002; Nitschke, Heller, Etienne, &
Miller, 2004). For example, depression has been associated with impairments
in autobiographical memory (Williams & Broadbent, 1986; Williams &
Dritschel, 1988; Williams & Scott, 1988), episodic memory recall (Cabeza,
Locantore, & Anderson, 2003), and working memory (Elliot et al., 1996). In
addition, depressed subjects perform poorly on explicit memory tasks such
as free recall, cued recall, and recognition. A meta-analysis of recall and
recognition studies by Burt, Zembar, and Niederehe (1995) revealed a stable
association between memory and depression, demonstrating that depression
is linked with particular aspects of memory impairment, specifically deficits
in explicit (and not implicit) memory tasks. These deficits are not explainable
as secondary to reductions in motivation. For example, using a motivation-
enhancing manipulation, Richards and Ruff (1989) demonstrated that
motivated depressed patients performed the same on neuropsychological
measures as did a nonmotivated depressed group.
In other studies, biases have been reported for valenced information,
whereby depression is associated with a tendency to recall negative better
than positive material (for reviews, see Blaney, 1986; Watkins, 2002). Mood
induction studies in clinically depressed and nonpatient individuals have
shown a bias toward recall of negative autobiographical memories (Clark &
Teasdale, 1982; Williams & Scott, 1988). Other research has implicated
withdrawal-related negative emotions and threat perception as factors that
may play a role in negative cognitive biases (Bar-Haim et al., 2007; Heller &
Nitschke, 1997; Nitschke & Heller, 2002). Heller and Nitschke (1997) pro-
posed that executive function impairments account for a large part of the
observed memory deficits in depression. Individuals with depression demon-
strate difficulties in initiating cognitive strategies that enhance their ability to
process and remember information (for review, see Levin et al., 2007).
Prefrontal cortex, especially dorsolateral sectors (DLPFC), is frequently
associated with cognitive control and executive function. A primary function
of DLPFC is the representation of goals and the maintenance of context
information that promotes the means to achieve these goals (Braver &
Barch, 2002; Davidson, 2002; Nitschke et al., 2004). Context information
might include task demands, information regarding the results of previous
behavior, emotional state, or any aspect of the internal or external environment
that would influence the accomplishment of the represented goals (Nitschke
et al., 2004). Recruitment of DLPFC in cognitive control tasks has been found
to be central to performance on various memory tasks, including working
memory and episodic long-term memory (for review, see Nitschke et al.,
2004). In studies examining strategy utilization in depression, Hertel and
colleagues (for reviews, see Hertel, 1994, 1997, 2000) showed that memory
deficits are eliminated when strategies are provided prior to the start of the
task. In line with this research, patients with DLPFC lesions showed impair-
ments when using organizational strategies during episodic memory tasks
but showed improvement when instructed in the use of such strategies
(Gershberg & Shimamura, 1995; Incisa della Rocchetta & Milner, 1993).
Given that depressed individuals can perform cognitively demanding
tasks in the presence of explicit instructions or task constraints, memory
impairments may not be the result of reduced attentional resources but
rather an impairment in the deployment of these resources. Hertel (1994)
proposed that attentional resources are sufficient in patients with depres-
sion but that the initiative to control these resources is missing. This
diminished initiative to attend can be manifested in underrecruitment of
PFC. In an EEG study examining brain mechanisms accompanying the
initiative deficit, Nitschke et al. (2004) found that bilateral activity recorded
over PFC during a preparatory period immediately preceding a sad
narrative was associated with better recall performance in controls but
not in a depressed group. Depressed participants also showed a negative
memory bias. Hyperactivity in right PFC was observed during exposure to
the sad narrative and was associated with improved recognition of words
in that narrative.
In depressed individuals, poor performance on memory tasks may there-
fore be explained in part by a failure to recruit PFC in preparation for
information processing (Nitschke et al., 2004). This impairment may be
associated with a lack of initiative in allocating attentional resources for
performance on cognitive tasks (Hertel & Harden, 1990, Hertel & Rude,
1991) or with problems with sustained attention (Burt et al., 1995).
Nitschke et al.’s (2004) findings support Hertel’s (1994, 2000) model of
memory performance in depression and highlight the importance of distin-
guishing different processes influencing memory performance and cogni-
tive bias. Specifically, the failure to recruit PFC in preparation for
information processing may result in poorer performance on memory
tasks. Recruiting right PFC under conditions of negative emotion or threat
may serve to enhance memory performance (Heller & Nitschke, 1997).
3. Depression/anxiety comorbidity
An important issue to consider in understanding the impact of emotion

on memory and other cognitive processes is the considerable but variable
comorbidity of depression with anxiety. The majority of studies of either
depression or anxiety have not taken this comorbidity into consideration,
nor have they distinguished between types of anxiety such as anxious
apprehension and anxious arousal. Heller, Miller, and colleagues
(1995, 1997, 2003); Engels et al. (2007); Keller et al. (2000); Levin et al.
(2007); Nitschke et al. (1999) have demonstrated that different types of
anxiety and depression are associated with distinct patterns of regional
brain activity using neuropsychological, EEG, and fMRI methods, as well as
having distinct psychometric relationships.
Crucially, unaccounted-for comorbidity may undermine the interpreta-
tion of results in much of the literature on emotion–cognition relationships.
For example, the presence of comorbid anxiety has been shown to cancel
out the effects of depression on neuropsychological performance (Heller
et al., 1995) or to have both additive and nonadditive effects (Keller et al.,
2000). In addition, the literature on brain activity in anxiety and depression
during cognition is inconsistent. Heller and Nitschke (1998) proposed that
these discrepancies can be explained in terms of the differing subtypes of
anxiety and depression represented in different studies. Thus, a failure to
separate anxiety and depression, either experimentally or statistically, may
explain a large portion of the variability in reported executive function
impairments in depression (Levin et al., 2007).
Given the substantial conceptual and epidemiological overlap of anxi-
ety and depression, researchers and clinicians have long desired to under-
stand the relationship between them. An influential line of research has
focused on how they differ in their affective structure. The tripartite model
proposed by Clark and Watson (1991), Mineka, Watson, & Clark (1998),
Watson, Clark, et al. (1995), Watson, Weber, et al. (1995), and Watson,
Weise, Vaidya, & Tellegen (1999) includes a shared general distress factor
characterized by high levels of negative affect that is common to both
anxiety and depression. A separate, positive affect/anhedonia factor is
characterized by low levels of pleasurable engagement with the environ-
ment and is specific to depression. Lastly, arousal characterizes anxiety
and not depression. Heller, Miller, and colleagues have argued that these
components of depression and anxiety are implemented in different brain
regions (e.g., Heller & Nitschke, 1997, 1998; Levin et al., 2007). Further-
more, they have emphasized that the type of anxiety described by Watson
and colleagues (anxious arousal) should be differentiated from anxious
apprehension, or worry, in psychological and neuropsychological terms.
To the degree that one of these types of anxiety is associated with activity
in a particular brain region, it is possible that its presence could disrupt
(either enhance or impair) ongoing cognitive processing typically imple-
mented or influenced by that brain region. For example, reduced brain
activity in PFC associated with depression could account for impaired
performance on various executive function tasks, reviewed above. In
contrast, anxious arousal and anxious apprehension could have other
influences on cognitive processing that would manifest in distinct patterns
of brain activity.
4. Anxiety and memory
Supporting the proposals of Heller, Miller, and colleagues, Shackman et al.

(2006) demonstrated that threat-induced anxiety (essentially anxious
arousal) selectively disrupts spatial but not verbal working memory per-
formance. The tasks used were verbal (letter identity) and spatial (loca-
tion) variants of a three-item N-back task. Threat of shock served as the
affect induction procedure. The authors concluded that threat-induced
anxiety (verified via EMG) disrupted spatial performance (indexed by
accuracy) and not verbal performance because right PFC resources
were engaged in anxiety-related processing and hence were less available
to support working memory performance. Similar results were obtained in
a second experiment testing individuals with high scores on a self-report
measure presumed to index the predisposition to react more strongly to
perceived threat. These findings are consistent with lateralization for
visuospatial working memory in prefrontal and parietal cortices, supporting
the arousal portion of the model of Heller, Miller, and colleagues.
Supporting the model’s proposal of valence lateralization in PFC, with
concomitant effects on cognitive processing, Gray (2001) found a double
dissociation between the effects of inducing positively and negatively
valenced mood on spatial and verbal working memory. Specifically,
performance on tasks relying on verbal working memory, for which there
is considerable evidence of left prefrontal specificity, was enhanced by
the induction of positive mood and impaired by the induction of negative
mood. In contrast, performance on tasks relying on spatial working
memory, for which there is considerable evidence of right prefrontal
specificity, was enhanced by the induction of negative mood and impaired
by the induction of positive mood, supporting the valence portion of
the model of Heller, Miller, and colleagues. Some aspects of Gray’s results
are not consistent with those of Shackman et al. (see Shackman et al., 2006,
for extensive discussion of these discrepancies) but nevertheless serve
to demonstrate differential effects of affect on lateralized cognitive
processes.
In contrast to anxious arousal, anxious apprehension would be expected
to influence tasks associated with the left-hemisphere regions hypothe-
sized to be involved in this type of anxiety (Engels et al., 2007; Nitschke,
Heller, & Miller, 2000). Heller, Miller, and colleagues have suggested that
many of the general decrements in cognitive processing associated with
anxiety (Eysenck & Calvo, 1992; Eysenck, Derakshan, Santos, & Calvo,
2007) can be attributed to the effects of anxious apprehension on the
availability of left-hemisphere processes that support attention, verbal
rehearsal, and verbal working memory (for reviews, see Nitshcke & Heller,
1998; Nitschke et al., 2000).
A particular type of impact on cognition believed to result from

anxiety is cognitive bias. Although memory biases are more common
in depressed individuals, attentional biases are observed in anxious
individuals (Bar-Haim et al., 2007). Anxiety impairs performance on
many tasks, particularly when they are difficult or must be performed
under stressful conditions. Most theorists agree that such deficits in
performance on tasks that require high attentional or short-term memory
demands can be attributed to the interference of worrisome thoughts
with attention to task-relevant information (e.g., McNally, 1998; Sarason,
1988). Reduced recruitment of cognitive control mechanisms may play a
role in anxiety-driven performance impairments. For example, partici-
pants with high anxiety levels showed reduced recruitment of DLPFC
during threat-related distractors in an emotional Stroop task (Mathews &
Mackintosh, 1998). High anxiety is also associated with reduced rostral
anterior cingulate activity, a region of the brain associated with the
assessment of emotionally salient information (Bush, Luu, & Posner,
2000; Engels et al., 2007; Mohanty et al., 2007) and with reduced recruit-
ment of lateral PFC when the expectation of threatening distractors is
established (Bishop, Duncan, Brett, & Lawrence, 2004). Thus, cognitive
control mechanisms that are required to maintain ongoing task proces-
sing are diminished in anxiety during the presence of threat-related
distractors.
Cognitive biases have also been observed to affect the interpretation of
information as well as the ability to remember this information in anxious
individuals (McNally, 1998; Mineka et al., 1998). Anxious individuals have
demonstrated an increased likelihood of interpreting ambiguous informa-
tion in a negative manner across multiple paradigms (Mineka et al., 1998).
For example, anxious individuals are likely to interpret ambiguous homo-
phones (e.g., die/dye, pain/pane; Mathews, Richards, & Eysenck, 1989) and
sentences such as ‘‘The doctor examined Little Emma’s growth’’ (Eysenck,
Mogg, May, Richards, & Mathews, 1991) in a more threatening manner than
do controls.
Evidence also suggests that anxiety disorders are accompanied by
enhanced memory for negative and threatening information (McNally,
1998). Intrusive memories are a common symptom in anxiety disorders.
Some individuals with posttraumatic stress disorder are plagued by horrific
memories reflected in nightmares, intrusive thoughts, and flashbacks. Indi-
viduals with panic disorder frequently experience fear of impending heart
attack, insanity, and death, possibly fueled by memories of their first or
worst episode. Those with obsessive–compulsive disorder may experience
recurrent obsessions about harm, whereas those who suffer from general-
ized anxiety disorder experience uncontrollable worry about looming
threats (Barlow, 2001; Coles & Heimberg, 2002; McNally, 1998). These
phenomenological observations suggest that memory in anxiety disorders

is characterized by enhanced access to threat-related information (Coles &
Heimberg, 2002; Lang, 1979; McNally, 1998).
Such biases in attention and memory may be linked to a right-
hemisphere system differentially involved in responding to threat (Comp-
ton, Heller, Banich, Palmieri, & Miller, 2000; Nitschke et al., 2000). The
right hemisphere is particularly suited to evaluate emotional stimuli (e.g.,
Borod et al., 1998) and has also been identified as important for cognitive
processes such as scanning both sides of space, processing spatial rela-
tionships, and maintaining vigilance. Under normal circumstances, an
integration of functions across prefrontal, parietal, and temporal regions
of the right hemisphere would confer a highly adaptive capacity to moni-
tor the environment for emotional stimuli and modulate responses to
these stimuli (Nitschke & Heller, 2002; Nitschke et al., 2000). In cases of
psychopathology, abnormal emotional responding could be associated
with abnormal patterns of brain activity in these and other regions as
well as with altered (either enhanced or impaired) attention, learning, and
memory.
5. Conclusion
This brief review emphasizes a few aspects of a substantial and grow-

ing literature on relationships between the emotional qualities of stimuli,
the emotion-related processing they prompt, and their effects on mem-
ory. These effects are prominent in, but not confined to, some types of
psychopathology. Indeed, it should be understood that emotion is an
ongoing modulator of memory even in healthy or ‘‘normal’’ individuals,
just as, conversely, memory feeds and alters emotion. Thus, both emo-
tion phenomena and memory phenomena are better understood if each is
better understood. A particularly challenging frontier in this area of
research is determining causal mechanisms, not only between emotion
and memory as psychological processes, or between the brain processes
that implement emotion and memory, but between the psychological
processes and biological processes. Choice of language in this literature
often implies causal relationships, either psychology driving biology or
biology driving psychology, but to date there is no articulated mechanism
by which such causal relationships could arise (Miller, 1996; Miller et al.,
2007). Neither dismissing nor embracing dualism as a type of relationship
between psychological and biological events provides the causal
mechanisms, if any, that link them. The present review provides a sam-
pling of observed associations between these phenomena, awaiting a
mechanistic account.
Acknowledgments
This research was supported by the National Institute of Drug Abuse

(R21 DA14111), the National Institute of Mental Health (R01 MH61358, T32
MH19554), and the University of Illinois Beckman Institute, Department of
Psychology, and Intercampus Research Initiative in Biotechnology.
References
Ashby, F. G., Isen, A. M., & Turken, A. U. (1999). A neuropsychological theory of

positive affect and its influence on cognition. Psychological Review, 106, 529–550.
Barbas, H. (2000). Connections underlying the synthesis of cognition, memory, and
emotion in primate prefrontal cortices. Brain Research Bulletin, 52, 319–330.
Bar-Haim, Y., Lamy, D., Pergamin, L., Bakersmans-Kranenburg, M. J., & van Ijzen-
doorn, M. H. (2007). Threat-related attentional bias in anxious and nonanxious
individuals: A meta-analytic study. Psychological Bulletin, 133, 1–24.
Barlow, D. H. (2001) Clinical handbook of psychological disorders. New York: The
Guilford Press.
Bishop, S., Duncan, J., Brett, M., & Lawrence, A. D. (2004). Prefrontal cortical
function and anxiety: Controlling attention to threat-related stimuli. Nature
Neuroscience, 24, 10364–10368.
Blaney, P. H. (1986). Affect and memory: A review. Psychological Bulletin, 99,
229–246.
Borod, J. C., Cicero, B. A., Obler, L. K., Welkowitz, J., Erhan, H. M., Santschi, C.,
et al. (1998). Right hemisphere emotional perception: Evidence across multiple
channels. Neuropsychology, 12, 446–458.
Braver, T. S., & Barch, D. M. (2002). A theory of cognitive control, aging cognition,
and neuromodulation. Neuroscience and Biobehavioral Reviews, 26, 809–817.
Burt, D. B., Zembar, M. J., & Niederehe, G. (1995). Depression and memory impair-
ment: A metaanalysis of the association, its pattern, and specificity. Psychologi-
cal Bulletin, 117, 285–305.
Bush, G., Luu, P., & Posner, M. I. (2000). Cognitive and emotional influences in
anterior cingulate cortex. Trends in Cognitive Sciences, 4, 215–222.
Cabeza, R., Locantore, J. K., & Anderson, N. D. (2003). Lateralization of prefrontal
activity during episodic memory retrieval: Evidence for the production-monitor-
ing hypothesis. Journal of Cognitive Neuroscience, 2, 249–259.
Cahill, L. (2000). Modulation of Long-Term Memory Storage in Humans by
Emotional Arousal. In J. P. Aggleton (Ed.), The amygdala: A functional analysis
(pp. 425–445). New York: Oxford University Press.
Cahill, L., Haier, R. J., Fallon, J., Alkire, M. T., Tang, C., Keator, D., et al. (1996).
Amygdala activity at encoding correlated with long-term, free recall of emotional
information. Proc Natl Acad Sci U S A, 93, 8016–8021.
Cahill, L., & McGaugh, J. L. (1995). A novel demonstration of enhanced memory
associated with emotional arousal. Consciousness and Cognition, 4, 410–421.
Canli, T., Zhao, Z., Brewer, J., Gabrieli, J. D. E., & Cahill, L. (2000). Event-related
activation in the human amygdala associated with later memory for individual
emotional experience. Journal of Neuroscience, 20 (RC99), 1–5.
Canli, T., Desmond, J. E., Zhao, Z., & Gabrieli, J. D. (2002). Sex differences in the
neural basis of emotional memories.Proceedings of the National Academy of
Sciences of the U S A, 99, 10789–10794.
Clark, D. M., & Teasdale, J. D. (1982). Diurnal variation in clinical depression and
accessibility of memories of positive and negative experiences. Journal of Abnor-
mal Psychology, 91, 87–95.
Clark, L. A., & Watson, D. (1991). Tripartite model of anxiety and depression: Evidence
and taxonomic implications. Journal of Abnormal Psychology, 100, 316–336.
Coles, M. E., & Heimberg, R. G. (2002). Memory biases in the anxiety disorders:
Current status. Clinical Psychology Review, 22, 587–627.
Compton, R. J., Heller, W., Banich, M. T., Palmieri, P. A., & Miller, G. A. (2000).
Responding to threat: Hemispheric asymmetries and interhemispheric division of
input. Neuropsychology, 14, 254–264.
Davidson, R. J. (1984). Affect, cognition and hemispheric specialization. In C. E. Izard &
J. Kagan, & R. Zajonc (Eds.), Emotion, Cognition and Behavior (pp. 320–365).
New York: Cambridge University Press.
Davidson, R. J. (2002). Anxiety and affective style: Role of prefrontal cortex and
amygdala. Biological Psychiatry, 51, 61–80.
Davidson, R. J., & Irwin, W. (1999). The functional neuroanatomy of emotion and
affective style. Trends in Cognitive Sciences, 3, 11–21.
Dolcos, F., Graham, R., LaBar, K., & Cabeza, R. (2003). Coactivation of the amygdala
and hippocampus predicts better recall for emotional than for neutral pictures.
Brain and Cognition, 51, 221–223.
Dolcos, F., Labar, K. S., & Cabeza, R. (2004). Dissociable effects of arousal and
valence on prefrontal activity indexing emotional evaluation and subsequent
memory: An event-related fmri study. Neuroimage, 23, 64–74.
Elliott, R., Sahakian, B. J., McKay, A. P., Herrod, J. J., Robbins, T. W., & Paykel, E. S.
(1996). Neuropsychological impairments in unipolar depression: The influence
of perceived failure on subsequent performance. Psychological Medicine, 26,
975–990.
Elliot, A. J., & Thrash, T. M. (2002). Approach-avoidance motivation in personality:
Approach and avoidance temperaments and goals. Journal of Personality and
Social Psychology, 82, 804–818.
Engels, A. S., Heller, W., Mohanty, A., Herrington, J. D., Banich, M. T., Webb, A. G.,
et al. (2007). Specificity of regional brain activity in anxiety types during emotion
processing. Psychophysiology, 33, 352–363.
Eysenck, M. W., & Calvo, M. G. (1992). Anxiety and performance: The processing
efficiency theory. Cognition and Emotion, 6, 409–434.
Eysenck, M. W., Derakshan, N., Santos, S., & Calvo, M. G. (2007). Anxiety and
cognitive performance: Attentional control theory. Emotion, 7, 336–353.
Eysenck, M. W., Mogg, K., May, J., Richards, A., & Mathews, A. (1991). Bias in
interpretation of ambiguous sentences related to threat in anxiety. Journal
Abnormal Psychology, 100, 144–150.
Gershberg, F. B., & Shimamura, A. P. (1995). Impaired use of organizational strategies

in free recall following frontal lobe damage. Neuropsychologia, 33, 1305–1333.
Gray, J. R. (2001). Emotional modulation of cognitive control: Approach-withdrawal
states double-dissociate spatial from verbal two-back task performance.Journal
of Experimental Psychology: General, 130, 436–452.
Gray, J. R. (2004). Integration of emotion and cognitive control. Current Directions
in Psychological Science, 13, 46–48.
Gray, J. R., Braver, T. S., & Raichle, M. E. (2002). Integration of emotion and
cognition in the lateral prefrontal cortex. Proceedings of the National Academy
of Sciences, 99, 4115–4120.
Hamann, S. B., Ely, T. D., Grafton, S. T., & Kilts, C. D. (1999). Amygdala activity
related to enhanced memory for pleasant and aversive stimuli. Nature Neu-
roscience, 2(3), 289–293.
Heller, W. (1990). The neuropsychology of emotion: Developmental patterns and
implications for psychopathology.. In N. Stein, B. L. Leventhal,& T. Trabasso
(Eds.), Psychological and Biological Approaches to Emotion (pp. 167–211). Hills-
dale, NJ: Erlbaum.
Heller, W. (1993). Neuropsychological mechanisms of individual differences in
emotion, personality, and arousal. Neuropsychology, 7, 476–489.
Heller, W., Etienne, M. A., & Miller, G. A. (1995). Patterns of perceptual asymmetry
in depression and anxiety: Implications for neuropsychological models of emo-
tion and psychopathology. Journal of Abnormal Psychology, 104, 327–333.
Heller, W., Koven, N. S., & Miller, G. A. (2003). Regional brain activity in anxiety
and depression, congnition/emotion interaction, and emotional regulation. In
K. Hugdahl & R. J. Davidson (eds.), The Asymmetrical Brain (pp. 533–563).
Cambridge, MA: MIT Press.
Heller, W., & Nitschke, J. B. (1997). Regional brain activity in emotion: A framework
for understanding cognition in depression. Cognition and Emotion, 11, 637–661.
Heller, W., Nitschke, J. B., Etienne, M. A., & Miller, G. A. (1997). Patterns of regional brain
activity differentiate types of anxiety. Journal of Abnormal Psychology, 106, 376–385.
Heller, W., & Nitschke, J. B. (1998). The puzzle of regional brain activity in depres-
sion and anxiety: The importance of subtypes and comorbidity. Cognition and.
Emotion, 12, 421–447.
Herrington, J. D., Koven, N., Heller, W., Miller, G. A., & Nitschke, J. B. (In press).
Frontal asymmetry in emotion, motivation, personality and psychopathology:
Electrocortical and hemodynamic neuroimaging. In N. Allen & C. Pantelis
(Eds.), Neurobiology of mental disorders.
Herrington, J. D., Koven, N. S., Miller, G. A., & Heller, W. (2006). Mapping the neural
correlates of dimensions of personality, emotion, and motivation. In T. Canli
(Ed.), Biology of Personality and Individual Differences (pp. 133–156). New
York: Guilford Publications.
Herrington, J. D., Mohanty, A., Koven, N. S., Fisher, J. E., Stewart, J. L., Banich, M.
T., et al. (2005). Emotion-modulated performance and activity in left dorsolateral
prefrontal cortex.Emotion, 5, 200–207.
Hertel, P. T. (1994). Depression and memory: Are impairments remediable through
attentional control?. Current Directions in Psychological Science, 3, 190–194.
Hertel, P. T. (1997). On the contributions of deficient cognitive control to memory

impairments in depression. Cognition and Emotion, 11, 569–583.
Hertel, P.T. (2000). The cognitive–initiative account of depression-related impair-
ments in memory. In D. Medlin (ed.), The Psychology of Learing and Motivation:
Advances in Research Theory (pp. 47–71). San Diego, CA: Academic Press.
Hertel, P.T., & Hardin, T. S. (1990). Remembering with and without awareness in a
depressed mood: Evidence for deficits in initiative. Journal of Exprimental
Hertel, P. T., & Rude, S. S. (1991). Depressive deficits in memory: Focusing attention
improves subsequent recall. Journal of Experimental Psychology: General, 120,
301–309.
Incisa della Rocchetta, A., & Milner, B. (1993). Strategic search and retrieval inhibi-
tion: The role of the frontal lobes. Neuropsychologia, 31, 503–524.
Keller, J., Nitschke, J. B., Bhargava, T., Deldin, P. J., Gergen, J. A., Miller, G. A., et al.
(2000). Neuropsychological differentiation of depression and anxiety. Journal of
Lang, P. J. (1979). A bio-informational theory of emotional imagery. Psychophysiol-
ogy, 16, 495–512.
Lang, P. J., Bradley, M. M., & Cuthbert, B. N. (1997). Motivated attention: Affect,
activation, and action. In P. J. Lang, R. F. Simons, & M. T. Balaban (Eds.),
Attention and orienting: Sensory and motivational processes (pp. 97–135).
Mahwah, NJ: Lawrence Erlbaum Associates.
Lang, P. J., Greenwald, M. K., Bradley, M. M., & Hamm, A. O. (1993). Looking at
pictures: Affective, facial, visceral, and behavioral reactions. Psychophysiology,
30, 261–273.
Levin, R. L., Heller, W., Mohanty, A., Herrington, J. D., & Miller, G. A. (2007).
Cognitive deficits in depression and functional specificity of regional brain
activity. In R Atchley & S. Ilardi (Eds.), Cognitive Neuroscience Perspectives on
Depression, Special Issue,Cognitive Therapy and Research, 31, 211–233.
Mathews, A., & Mackintosh, B. (1998). A cognitive model of selective processing in
anxiety. Cognitive Therapy and Research, 22, 539–560.
Mathews, A., Richards, A., & Eysenck, M. (1989). Interpretation of homophones
related to threat in anxiety states. Journal Abnormal Psychology, 98, 31–34.
McGaugh, J. L., Ferry, B., Vazdarjanova, A., & Roozendaal, B. (2000). Amygdala:
Role in modulation of memory storage. In J. P. Aggleton (Ed.), The amygdala: A
Functional Analysis (pp. 391–424). New York: Oxford University Press.
McNally, R. J. (1998). Information-processing abnormalities in anxiety disorders: Impli-
cations for cognitive neuroscience. Cognition and Emotion, 12, 479–495.
Miller, G. A. (1996). Presidential address: How we think about cognition, emotion,
and biology in psychopathology. Psychophysiology, 33, 615–628.
Miller, G. A., Engels, A. S., & Herrington, J. D. (2007). The seduction of clinical
science: Challenges in psychological and biological convergence. In T. Treat , R.
Bootzin. & R. Levenson (Eds.), Psychological clinical science: Papers in honor of
Richard Mcfall (pp. 53–74). Mahwah, NJ: Lawrence Erlbaum.
Mineka, S., Watson, D., & Clark, L. A. (1998). Comorbidity of anxiety and unipolar
mood disorders. Annual Review of Psychology, 49, 377–412.
Mohanty, A., Engels, A. S., Herrington, J. D., Heller, W., Ho, M.-H. R., Banich, M. T.,
et al. (2007). Differential engagement of anterior cingulate cortex subdivisions for
cognitive and emotional function. Psychophysiology, 344, 352–363.
Mohanty, A., & Heller, W. (2002). The neuropsychology of mood disorders: Affect,
cognition, and neural circuitry. In H. D’haenen, J. A. den Boer, H. Westernberg, &
P. Willner (Eds.), Textbook of Biological Psychiatry, (pp.791–802). Chichester:
Wiley.
Mohanty, A., Herrington, J. D., Koven, N. S., Wenzel, E. A., Webb, A. G., Heller, W.,
et al. (2005). Neural mechanisms of affective interference in schizotypy. Journal
of Abnormal Psychology, 114, 16–27.
Nitschke, J. B., & Heller, W. (2002). The neuropsychology of anxiety disorders:
Affect, cognition, and neural circuitry. Biological Psychiatry, (pp. 975–988).
Wiley, Chichester.
Nitschke, J. B., Heller, W., Etienne, M. A., & Miller, G. A. (2004). Prefrontal cortex
activity differentiates processes affecting memory in depression. Biological Psy-
chology, 67, 125–143.
Nitschke, J. B., Heller, W., & Miller, G. A. (2000). Anxiety, stress, and cortical brain
function. In: Borod, J. C. (Ed.), The Neuropsychology of Emotion. (pp. 298–319).
New York: Oxford University Press.
Nitschke, J. B., Heller, W., Palmeiri, P. A., & Miller, G. A. (1999). Contrasting
patterns of brain activity in anxious apprehension and anxious arousal. Psycho-
physiology, 36, 628–637.
Richards, P. M., & Ruff, R. M. (1989). Motivational effects on neuropsychological
functioning: Comparison of depressed versus nondepressed individuals. Journal
of Consulting and Clinical Psychology, 57, 396–402.
Russell, J. A. (1980). A circumplex model of affect. Journal of Personality and
Social Psychology, 39, 1161–1178.
Sarason, I. G. (1988). Anxiety, self-preoccupation and attention. Anxiety Research,
1, 3–7.
Shackman, A. J., Sarinopoulos, I., Maxwell, J. S., Pizzagalli, D. A., Lavric, A., &
Davidson, R. J. (2006). Anxiety selectively disrupts visuospatial working memory.
Emotion, 6, 40–61.
Shankman, S. A., & Klein, D. N. (2003). The relation between depression and
anxiety: an evaluation of the tripartite, approach-withdrawal and valence-arousal
models. Clinical Psychology Review, 23, 605–637.
Tucker, D. M., Stenslie, C. E., Roth, R. S., & Shearer, S. L. (1981). Right frontal lobe
activation and right hemisphere performance: decrement during a depressed
mood. Archives of General Psychiatry, 38, 169–174.
Watkins, P. C. (2002). Implicit memory bias in depression. Cognition and. Emotion,
16, 381–402.
Watson, D., Clark, L. A., Weber, K., Assenheimer, J. S., Strauss, M. E., & McCormick,
R. A. (1995). Testing a tripartite model: II. Exploring the symptom structure of
anxiety and depression in student, adult, and patient samples. Journal of Abnor-
mal Psychology, 104, 15–25.
Watson, D., Weber, K., Assenheimer, J. S., Clark, L. A., Strauss, M. E., & McCormick,
R. A. (1995). Testing a tripartite model. I. Evaluating the convergent and
discriminant validity of anxiety and depression symptom scales. Journal of

Watson, D., Wiese, D., Vaidya, J., & Tellegen, A. (1999). The two general activation
systems of affect: Structural findings, evolutionary considerations, and psycho-
biological evidence. Journal of Personality and Social Psychology, 76, 820–838.
Williams, J. M. G., & Broadbent, K. (1986). Autobiographical memory in attempted
suicide patients. Journal of Abnormal Psychology, 95, 144–149.
Williams, J. M. G., & Dritschel, B. H. (1988). Emotional disturbance and the speci-
ficity of autobiographical memory. Cognition and Emotion, 2, 221–234.
Williams, J. M. G., & Scott, J. (1988). Autobiographical memory in depression.
Psychological Medicine, 18, 689–695.
Human Learning 61
Scheduling and Learning

Dominic A. Simon
New Mexico State University
There is a long history of investigating the impact of different schedules

on learning of new information, skills, and associations. The naı̈ve perspec-
tive on learning is that more study will always be better for learning. Indeed,
there is a ‘‘total time’’ hypothesis which suggests that accumulated benefits
of study grow in proportion to the time spent by the learner on the material
to be learned. There is evidence in support of this hypothesis (Baddeley,
1997), but there is also much evidence to indicate that the benefits of study,
practice, or experience are impacted not only by the amount or frequency of
presentation, but also by how these presentations are arranged: the learning
schedule. This chapter is intended to give a survey and summary of some of
the various ways in which scheduling affects learning.
1. Learning is more than just practice
Closely related to the total time hypothesis is the commonly held belief
that repetition is key to success in learning. However, most people have also
been frustrated by the fact that simple repetition, in and of itself, does not
always lead to success in remembering. A good demonstration of this
phenomenon comes from a study by Craik and Watkins (1973). Participants
were presented with a list of words and told that during list presentation
they should be prepared to report at any time what the most recent word
was that began with a particular letter. The assumption was that the parti-
cipants would covertly repeat the most recent word beginning with that
letter. Craik and Watkins varied the number of items that intervened
between each successive target-letter word and thus indirectly varied the
amount of ‘‘maintenance rehearsal’’ of those items. On a surprise test of all
the target-letter words after the end of the list, there was no relationship
between the likelihood of recalling a word and the amount of maintenance
rehearsal that it had presumably received. Thus, there was no recall
62 Dominic A. Simon
advantage for the items that enjoyed more rehearsal. Similarly, Craik and
Tulving (1975; see also Craik & Lockhart, 1972) showed that the ‘‘level’’ of
processing of materials could account for a great deal of the variability in
later recall or recognition. In various incidental learning protocols, the
manner with which participants processed materials (e.g., by considering
the sound vs. the meaning of studied words) affected later memory drama-
tically. What you do during learning or practice can be just as important as
how much you do in terms of learning and remembering.
It is important to recognize the distinction between factors that impact
immediate performance and those that impact learning as demonstrated by
a lasting change in performance capability. It has long been recognized that
changes in the capability of a learner to display that learning may not be
reflected in immediate performance, but may show up in a different con-
text or at a different time (e.g., Blodgett, 1929). Similarly, immediate
changes in performance may not last and thus are not always a good
guide for anticipating long-term retention. The issue of whether a change
due to experience needs to be long lasting in order to be labeled as
learning—as contrasted with the short-term adaptations seen in such phe-
nomena such as habituation and sensitization—is debatable, but it seems
that for most pragmatic purposes, the relatively long-lasting nature of the
change is a necessary component for learning.
Clearly, good knowledge about the impact of various kinds of scheduling
on the learning processes is crucial to the appropriate design and implemen-
tation of learning interventions of all kinds. A useful principle to keep in
mind when thinking about the effects described in this chapter is that of
‘‘transfer appropriate’’ processing (Lee, 1988; Morris, Bransford & Franks,
1977). The core idea behind this principle is that performance will benefit to
the degree that the cognitive processes induced by the conditions of study
match those under which that learning will be later tested. While trying to
avoid the circularity of the essential truism that whatever conditions of
practice optimize test performance are the most transfer-appropriate, the
idea is to consider conditions of study that will foster the kinds of processes
that will be needed at test. Sometimes that will mean that a close match in
conditions is desirable, but not always—by analogy, track athletes do not
prepare for meets by merely running the distance of their chosen event over
and over. In this chapter, we look at spacing, contextual variety, as well as
the effects of retrieval practice and testing effects, all of which can be used
as part of the experience of learners to enhance learning.1 As we will see,
1
Other aspects of scheduling can have a strong impact on the acquisition and retention of learned
associations, such as the schedule, magnitude, and character of reinforcement—and relatedly of
feedback—but these areas are omitted from the present discussion in the interests of space.
what turns out to be transfer-appropriate is often not what would have been
expected based on intuition.
2. Spacing
There is considerable work in various domains examining the impact of

spacing of learning trials on later memory. Herman Ebbinghaus (1885/
1964), a pioneer of experimental methodology for the study of learning
and memory, investigated the effects of repeated study at various lags
before repetition and found a clear advantage for more widely distributed
study opportunities; these effects have been replicated many times since.
In a now classic analysis, Crowder (1976, p. 274, based on data from
Madigan, 1969) showed that repetition per se of items in a list of words was
not beneficial to recall. However, repetition that occurred with a minimal
interstimulus lag of one or two items did enhance memory. Although the
benefit of further spacing yielded a negatively accelerating relationship
with the lag, there was evidence of continued growth in the probability of
recall as spacing was increased, suggesting that the benefits of spacing
continue to accrue with increasing lags.
The limits of this claim were evaluated in a study by Bahrick, H. P.,
Bahrick, L. E., Bahrick, A. S., and Bahrick, P. E. (1993), who found that
benefits from spacing continued to show positive returns for intervals
between study episodes for foreign–English vocabulary pairs that were
14, 28, or 56 days apart and tested up to 5 years after initial study. Baddeley
and Longman (1978) showed similar benefits for spacing of practice ses-
sions of different lengths occurring daily or twice-daily for postoffice
workers learning to acquire the practical skill of keyboarding to a criterion
level of performance.
It seems then that spacing that occurs both within and across learning
sessions is beneficial to learning and is generally preferable to repetitions
that are massed. Indeed, closely massed presentations may not be more
valuable than a single presentation.
It should be noted that the benefits of spacing are not universal. Glen-
berg (1976) reported data in which the relationship between the probability
of a twice-presented word being recalled and the number of other items
occurring between these presentations was nonmonotonic. However, this
nonmonotonicity was obtained only when the retention delay was brief; a
small number of items between presentation and repetition may be bene-
ficial when testing is scheduled to occur in the very near future (see also
Pashler, Zarow, & Triplett, 2003). Relatedly, presentation duration also
interacts with spacing; Metcalfe and Kornell (2003, Experiment 4) found
that benefits from spacing interacted with presentation time of items.
64 Dominic A. Simon
For an 8-s presentation of Spanish–English word pairs, spacing benefited

retention performance, but for 0.5-s presentation times, recall of massed
items was greater than recall of spaced items.
Another caveat concerning the benefits of spacing comes from work on
the accuracy of source memory. Source memory refers to where some
piece of information came from; Did a particular ‘‘fact’’ that you remember
come from a trustworthy person or from someone prone to confabulation?
Remembering that you read some piece of information can be evaluated
separately from remembering where you read it. Benjamin and Craik (2001)
found that for younger participants who were responding under speeded
conditions, and for older adults without such a speed constraint, items
from a studied but to-be-rejected list were more likely to be wrongly
endorsed as old if they had been spaced during study rather than massed.
These results show that although spacing enhances the basic familiarity of
items, under certain conditions (e.g., speeded responding, or in later life)
this familiarity can be strong enough to override critical information about
its source leading to inaccurate responding.
Many explanations have been offered as to why spacing is beneficial to
learning. Hintzman (1974) suggested that they can be characterized as
being due to either encoding variability or deficient processing. The encod-
ing variability notion is that spaced presentations are more likely to be
processed in somewhat different ways from one another. By virtue of these
different processing experiences, and the greater likelihood of encoding
features overlapping with the test circumstances, a given item is afforded
greater likelihood of later recall. As an example, Gartman and Johnson
(1972) had people study lists that involved two presentations of some
items. For some people, the two presentations of a given item were
designed so that the preceding words encouraged semantically very similar
processing (e.g., arm, leg, foot . . . . chin, knee, foot), while for other people,
the two presentations were designed to encourage semantically distinct
processing (e.g., arm, leg, foot,. . . , inch, meter, foot). On a later recall test,
the key words (e.g., foot) were recalled at a substantially higher rate in the
latter encoding condition. It should be noted that Maki and Hasher (1975)
compared recall for spaced presentations that had occurred in the context
of the same and different other words, as well as for homographs (words
that have multiple meanings, like ‘‘fan’’) where other word cues biased the
repetitions to have the same or different semantic interpretations as the
initial presentation. They did not find evidence for differential recall
between when there had been a change of surrounding words (where the
meaning was presumably not varying), and when the cues at repetition
encouraged different semantic encoding. However, the critical compari-
sons in the Maki and Hasher study did not involve once-presented items. It
is thus hard to determine whether the various conditions benefited in
different but essentially equivalent ways, or whether for some reason the
typical spacing benefits were simply not observed.
In contrast, the deficient processing views center around the idea that
less processing is devoted to repetitions that occur in close proximity to
the initial presentation of an item. As a result of this deficient processing,
massed items fare less well when tested at a delay. It would seem that there
is more theoretical rationale for this account than there is direct empirical
evidence (Dempster, 1996). Indeed, it is hard to reconcile the kinds of
effects found by Bahrick et al. (1993), mentioned above, with the deficient
processing view: the ranges of spacing intervals they used (14, 28 or 56
days) would all seem to allow ample opportunity for initial forgetting in
advance of the subsequent presentations. A definitive theoretical account
of spacing effects has yet to be put forward.
Investigators have also looked at whether learners themselves appreciate
the value of spacing. Baddely and Longman (1978) found that the practice
schedules that emphasized spacing and thus led to superior performance
received less favorable ratings from learners. In contrast, Dail and Christina
(2004) found that distributed practice supported more accurate acquisition
and retention performance than did massed practice of a motor task, and
that learners’ predictions of retention, made throughout practice, echoed
that distributed practice advantage. Son (2004) had learners study word–
synonym pairs, then decide whether they wanted a second presentation, and
if so whether it would occur immediately, after a delay or not at all. Her
findings suggested that choice of whether and when to receive repetitions
was related to the learner’s perceived difficulty ratings of the word pairs;
easier items were spaced, while more difficult items were massed. In a
paradigm with longer study times, Benjamin and Bird (2006) found essen-
tially the opposite effect. This suggests that learners prefer massing for
difficult items when a single presentation is deficient for some rudimentary
level of processing, but prefer spacing for those items when each study
opportunity is sufficient for that basic processing. In combination, these
results suggest that with greater insight into the relevant variables, we may
learn what circumstances tend to yield learner judgments that are or are not
accurate. Given how much of the average person’s learning experiences
occur in what amount to self-paced, self-evaluated study scenarios, under-
standing the factors critical in influencing the accuracy of such metacogni-
tive judgments is of great practical importance (Benjamin, 2007).
3. Contextual variety
Another variation of scheduling that has been shown to differentially

impact on learning concerns contextual variety during study or practice.
66 Dominic A. Simon
Practice that involves intermingling of operations or materials, often called

random practice, has been shown across a wide variety of settings and
materials to yield generally superior learning to situations in which the
same operation or materials are repeated in close succession, often
referred to as blocked practice. As an example, if three operations are to
be practiced, a random schedule would involve repeatedly switching
among those operations on successive trials throughout practice. In con-
trast, blocked practice involves long blocks of trials calling upon a single
operation, and then switching between blocks over the course of practice;
usually there are only as many blocks as there are operations to practice.
Notably, there is often an immediate performance benefit for a blocked
schedule, so this represents a case where the immediate and the longer-
term benefits of a study schedule are conflicting.
The advantages of contextual variation were first noted for learning of
verbal materials (Battig, 1966, 1972), but have been extensively demon-
strated for motor skill acquisition as well (Lee & Magill, 1983; Shea &
Morgan, 1979). Other skills shown to benefit from such manipulations have
included learning of Boolean operations (Carlson & Yaure, 1990), handwrit-
ing in elementary school students (Ste Marie, Clark, Findlay, & Latimer,
2004), foreign language vocabulary acquisition (Schneider, Healy, & Bourne,
1998), and perceptual discriminations (Mitchell, C. Nash, S. & Hall, G., 2008).
In a very practically oriented study, Jamieson and Rogers (2000) found that
both older and younger adults benefited from random practice of different
transaction types on a simulated Automated Teller Machine (ATM) system.
Further, transfer to a different simulation, with novel choice options, was
facilitated by having had random rather than blocked practice.
Though there has not been much work done on metacognitive judg-
ments made under conditions of contextual variety in practice, the findings
so far suggest that learners are sensitive to their immediate performance
rather than to the degree to which their practice schedule supports longer-
term learning. Simon and Bjork (2001, 2002) had participants learn a multi-
segment motor task with a timing goal in blocked and random practice
schedules. Throughout practice, participants made predictions about their
performance on a 24-h retention test. A similar prediction was solicited just
before the actual retention test on the day after acquisition. Both during
practice and immediately prior to the actual retention test, participants’
predictions were in line with acquisition performance; the blocked group
predicted smaller errors than did random. In actual retention performance,
however, the contextual interference effect obtained and the random group
had smaller errors.
Several explanations for the advantage of random practice have been
proposed, including greater retroactive interference for blocked practice
(Shewokis, Del Rey, & Simpson, 1998), an appeal to the general advantage
of spacing (Meeuswen & Magill, 1991), and the development of more

elaborate or distinctive stored representations of the to-be-learned materi-
als (Shea & Zimny, 1983,1988). Perhaps the most compelling explanation is
heavily influenced by the theorizing of Jacoby (1978), Cuddy and Jacoby
(1982), and Lee and Magill (1983, 1985). The crux of this idea is that
repeating the same information or operation on successive learning trials
tends to minimize the effortful stimulus processing that is found when the
information or process is changing from trial to trial. If processes are
changing from trial to trial, then the learner experiences not only greater
difficulty but also benefits from the need to determine and/or execute the
appropriate operation on new trials. By way of analogy, imagine asking a
young child to solve a mental math problem. After some mental effort, they
successfully generate the solution. Blocked practice would be akin to
asking the child exactly the same question right away; little, if any, effort
will be required to provide the correct solution. Random practice, however,
is more akin to asking several other problems first and then returning to the
original problem; if they recall the solution, it will probably be from long-
term memory; if they do not, then it will necessitate their engaging in
another deliberate attempt to find the solution—either way is more likely
to enhance future attempts at the same or similar problems.
It is noteworthy that many aspects of educational practice, such as hand-
writing drills or mathematics homework, tend to engage the learner in close
repetitions of the same operation. It seems likely that learners of all types
might benefit from schedules that engage the students in more intertrial
variation in operations/materials (though benefits may interact with the prior
experience and achievement levels of the learner; Guadagnoli & Lee, 2004).
Thus far, contextual variety has referred to the relationship between
materials or operations across successive study opportunities. However,
psychologists have looked at the impact of variations in physical context
as well. In a famous study, Godden and Baddeley (1975) showed that recall
for words that had been studied in a different context than the one in which
they were tested suffered as compared to words studied in the same context.
This result is one of many suggesting that reinstatement of study context at
test is beneficial. However Psychologists have looked at the impact of
variations in physical context as well (Bjork & Richardson-Klavehn, 1989).
A slightly different issue concerns whether testing in a novel environ-
ment, one not experienced before test, is supported best by having a series
of study sessions in the same place or in various places. Smith and Roth-
kopf (1984) reported findings supporting the hypothesis that variation of
the study setting can be beneficial, probably due to the relatively distinctive
(mental) retrieval cues that may become associated with these different
settings. Thus variation in context can be an important aspect of one’s
learning schedule.
68 Dominic A. Simon
4. Tests as learning events
As a teacher, one of the most surprising aspects of students’ study habits

that I observe is their failure to adequately assess their preparedness for
quizzes and tests. This failure owes perhaps to a lack of appreciation of the
distinction between recall, which is usually relatively difficult (e.g., What is
the name of the current Attorney General?) and recognition, which is
usually somewhat easier (e.g., Was William Rehnquist a Supreme Court
justice?). Students assume that looking at to-be-remembered information
(i.e., looking over their notes or the textbook) will lead that material to be
recallable at the time of test.2 Most of us seem to intuit that given the
choice between an opportunity to study some piece of information and an
opportunity to be tested on it, a further study event is to be preferred.
However, a large and growing body of evidence suggests otherwise.
Landauer and Bjork (1978) had participants either learn new first and
last name pairs or match faces to names, just as one might do upon meeting
a set of new people in a social or professional setting. After initial pre-
sentation of the names or face–name pairs, the spacing of retrieval practice
opportunities was varied: massed, distributed, or expanding. In the expand-
ing case, there was an increasing number of items occurring between
successive attempts to recall either the last name given the first, or the
part of the name given the face. The results favored the expanding retrieval
practice conditions over the others. Landauer and Bjork (1978) did not
compare retrieval practice with repeated study opportunities, but Morris,
Fritz, Jackson, Nichol, and Roberts (2005) did and found a strong effect
favoring retrieval practice over further presentations of the to-be-learned
names, suggesting that cued recall opportunities can be more powerful for
later retention of materials than are additional study opportunities.
These retrieval practice findings can be considered under the more
general phenomenon of the ‘‘testing effect’’ (Roediger & Karpicke, 2006a,
2006b). In various studies using word lists as well as materials that are
closely reminiscent of the things one might encounter in an educational
setting, repeated testing of once-studied materials has yielded superior
retention over relatively long periods (a week) as compared to an equiva-
lent number of repeated study presentations. Two things are quite remark-
able about these findings: first, feedback on the test does not seem to be
necessary to yield the testing advantage (Roediger & Karpicke, 2006a).
2
Of course it is easy to think of tests as the formal examinations that occur in educational
settings, but informal tests of learning occur constantly in our lives: Do you recall how to make
double-sided copies on the office photocopier? When leaving the mall, do you recall where you
parked your car? Do you recall what to do if someone suddenly collapses and needs CPR?
Second, these testing benefits show up not after many study episodes have
occurred, but after a single presentation of the to-be-learned materials.
Clearly then, testing is not simply a way of establishing what has already
been learned, but can be useful in helping to consolidate information and
attenuate its loss. As yet, teachers and learners alike have yet to digest this
important fact.
5. Summary and conclusion
Though the specifics vary, the most important conclusion to be drawn

from the various studies outlined above is that scheduling of study/practice
opportunities matters: simple number of accumulated study opportunities
is often less important than the organizational structure of those opportu-
nities. Learners often appear to be unaware of the kinds of scheduling
phenomena described in this chapter and thus fail to capitalize on the
potential benefits of study to the extent that they might. Part of the
problem clearly stems from a lack of awareness of the distinction between
immediate (performance) and longer-term (learning) consequences of
study events. Alas, this problem seems to persist even among those who
have taken classes that explicitly address such distinctions—perhaps this
stems from the perceived difficulty associated with arranging study to
capitalize on these effects (Benjamin, 2007). Bjork (1994) has suggested
the term ‘‘desirable difficulty’’ to describe aspects of learning scenarios that
incorporate features that make learning tasks more difficult with regard to
immediate acquisition, but that tend to foster enhanced long-term reten-
tion. Educators and researchers need to do a better job of getting the
message across and build such desirable difficulties into the educational
experiences of all learners: variations in scheduling are an excellent place
to begin.
References
Baddeley, A. D. (1997). Human memory: Theory and practice (Rev. ed.). Hove, UK:
Psychology Press.
Baddeley, A. D., & Longman, D. J. A. (1978). The influence of length and frequency
of training session on the rate of learning to type. Ergonomics, 21, 627–635.
Bahrick, H. P., Bahrick, L. E., Bahrick, A. S., & Bahrick, P. E. (1993). Maintenance of
foreign language vocabulary and the spacing effect. Psychological Science, 4,
316–321.
Battig, W. F. (1966). Facilitation and interference. In E. A. Bilodeau (Ed.), Acquisition
of skill New York: Academic Press.
70 Dominic A. Simon
Battig, W. F. (1972). Intratask interference as a source of facilitation on transfer and

retention. In E. F. Thompson & J. F. Voss (Eds.), Topics in learning and
performance, New York: Academic Press.
Benjamin, A. S. (2007). Memory is more than just remembering: Strategic control of
encoding, accessing memory, and making decisions. In A. S. Benjamin & B. H.
Ross (Eds.), The psychology of learning and motivation: Skill and strategy in
memory use (Vol. 48). pp. 175–224. London: Academic Press.
Benjamin, A. S., & Bird, R. D. (2006). Metacognitive control of the spacing of study
repetitions. Journal of Memory and Language, 55, 126–137.
Benjamin, A. S., & Craik, F. I. M. (2001). Parallel effects of aging and time pressure
on memory for source: Evidence from the spacing effect. Memory & Cognition,
29, 691–697.
Bjork, R. A. (1994). Memory and metamemory considerations in the training of
human beings. In J. Metcalfe & A. Shimamura (Eds.), Metacognition: knowing
about knowing (pp. 185–205). Cambridge, MA: MIT Press.
Bjork, R. A., & Richardson-Klavehn, A. (1989). On the puzzling relationship between
environmental context and human memory. In C. Izawa (Ed.), Current issues in
cognitive processes: The tulane flowerree symposium on cognition, (pp. 313–344).
Hillsdale, NJ: Erlbaum.
Blodgett, H. C. (1929). The effect of the introduction of reward upon the maze perfor-
mance of rats. University of California Publications in Psychology, 4, 113–134.
Carlson, R. A., & Yaure, R. G. (1990). Practice schedules and the use of component
skills in problem solving. Journal of Experimental Psychology: Learning, Mem-
ory, and Cognition, 16, 484–496.
Cuddy, L. J., & Jacoby, L. L. (1982). When forgetting helps memory: An analysis of
repetition effects. Journal of Verbal Learning and Verbal Behavior, 21, 451–467.
Craik, F. I. M., & Lockhart, R. S. (1972). Levels of processing: A framework for
memory research. Journal of Verbal Learning and Verbal Behavior, 11, 671–684.
Craik, F. I. M., & Tulving, E. (1975). Depth of processing and the retention of words in
episodic memory. Journal of Experimental Psychology: General, 104, 268–294.
Craik, F. I. M., & Watkins, M. J. (1973). The role of rehearsal in short-term memory.
Journal of Verbal Learning and Verbal Behavior, 12, 599–607.
Crowder, R. G. (1976). Principles of learning and memory. Hillsdale, NJ: Erlbaum.
Dail, T. K., & Christina, R. W. (2004). Distribution of practice and metacognition in
learning and long-term retention of a discrete motor task. Research Quarterly for
Exercise and Sport, 75, 148–155.
Dempster, F. N. (1996). Distributing and managing the conditions of encoding and
practice. In E. L. Bjork & R. A. Bjork (Eds.), Memory (pp. 317–344). San Diego,
CA: Academic Press.
Ebbinghaus, H. (1885). Über das Gedächtnis: Untersuchungen zur experimental-
len Psychologie. Leipzig: Duncker and Humboldt. [Reprinted as H. E. Ebbinghaus.
(1964)]. Memory: A contribution to experimental psychology (H. A. Ruger,
Trans.). New York: Dover..
Gartman, L. M., & Johnson, N. F. (1972). Massed versus distributed repetition of
homographs: A test of the differential-encoding hypothesis. Journal of Verbal
Learning and Verbal Behavior, 11, 801–808.
Glenberg, A. M. (1976). Monotonic and nonmonotonic lag effects in paired-associate

and recognition memory paradigms. Journal of Verbal Learning and Verbal
Behavior, 15, 1–16.
Godden, D. R., & Baddelely, A. D. (1975). Context-dependent memory in tow natural
environments. British Journal of Psychology, 6, 325–331.
Guadagnoli, M. A., & Lee, T. D. (2004). Challenge point: A framework for concep-
tualizing the effects of various practice conditions in motor learning. Journal of
Motor Behavior, 36, 212–224.
Hintzman, D. L. (1974). Theoretical implications of the spacing effect. In R. L. Solso
(Ed.), Theories in cognitive psychology: The Loyola symposium (pp. 77–99)
Potomac, MD: Erlbaum.
Jacoby, L. L. (1978). On interpreting the effects of repetition: Solving a problem
versus remembering a solution. Journal of Verbal Learning and Verbal Beha-
vior, 17, 649–667.
Jamieson, B. A., & Rogers, W. A. (2000). Age-related effects of blocked and random
practice schedules on learning a new technology. Journal of Gerontology:
Psychological Sciences, 55B, 343–353.
Landauer, T. K., & Bjork, R. A. (1978). Optimum rehearsal patterns and name
learning. In M. M. Gruneberg, P. E. Morris, & R. N. Sykes (Eds.), Practical Aspect
of Memory (pp. 625–632). London: Academic Press.
Lee, T. D. (1988). Transfer-appropriate processing: A framework for conceptualizing
practice effects in motor learning. In O. G. Meijer & K. Roth (Eds.), Complex move-
ment behaviour. The motor-action controversy (pp. 201–215). Amsterdam: Elsevier.
Lee, T. D., & Magill, R. A. (1983). The locus of contextual interference in motor-skill
acquisition. Journal of Experimental Psychology: Learning, Memory, and
Cognition, 9, 730–746.
Lee, T. D., & Magill, R. A. (1985). Can forgetting facilitate skill acquisition? In
D. Goodman, R. B. Wilberg, & I. M. Franks (Eds.), Differing perspectives in
motor learning, memory, and control (pp. 3–22). Amsterdam: Elsevier.
Madigan, S. A. (1969). Intraserial repetition and coding processes in free recall.
Journal of Verbal Learning and Verbal Behaviour, 8, 828–835.
Maki, R. H., & Hasher, L. (1975). Encoding variability: A role in immediate and long-
term memory? American Journal of Psychology, 88, 217–231.
Meeuwsen, H. J., & Magill, R. A. (1991). Spacing of repetitions versus contextual
interference effects in motor skill learning. Journal of Human Movement
Studies, 20, 213–228.
Metcalfe, J., & Kornell, N. (2003). The dynamics of learning and allocation of study
time to a region of proximal learning. Journal of Experimental Psychology:
General, 132, 530–542.
Mitchell, C., Nash, S., & Hall, G. (2008). The intermixed-blocked effect in human
perceptual learning is not the consequence of trial spacing. Journal of Experi-
mental Psychology: Learning, Memory, and Cognition, 34, 237–242.
Morris, C. D., Bransford, J. D., & Franks, J. J. (1977). Levels of processing versus
transfer appropriate processing. Journal of Verbal Learning and Verbal Beha-
vior, 16, 519–533.
72 Dominic A. Simon
Morris, P. E., Fritz, C. O., Jackson, L., Nichol, E., & Roberts, E. (2005). Strategies for
learning proper names: Expanding retrieval practice, meaning and imagery.
Applied Cognitive Psychology, 19, 779–798.
Pashler, H., Zarow, G., & Triplett, B. (2003). Is temporal spacing of tests helpful even
when it inflates error rates? Journal of Experimental Psychology: Learning,
Memory, and Cognition, 29, 1051–1057.
Roediger, H. L., & Karpicke, J. D. (2006a). Test-enhanced learning: Taking memory
tests improves long-term retention. Psychological Science, 17, 249–255.
Roediger, H. L., & Karpicke, J. D. (2006b). The power of testing memory: Basic
research and implications for educational practice. Perspectives on Psychologi-
cal Science, 1, 181–210.
Schneider, V. I., Healy, A. F., & Bourne, L. E. Jr. (1998). Contextual interference
effects in foreign language vocabulary acquisition and retention. In A. F. Healy, &
L. E. Bourne Jr. (Eds.), Foreign language learning: Psycholinguistic studies on
training and retention (pp. 77–90). Mahwah, NJ: Erlbaum.
Shea, J. B., & Morgan, R. L. (1979). Contextual interference effects on the acquisi-
tion, retention, and transfer of a motor skill. Journal of Experimental Psychol-
ogy: Human Learning and Memory, 5, 179–187.
Shea, J. B., & Zimny, S. T. (1983). Context Effects in Memory and Learning Move-
ment Information. In R. A. Magill (Ed.), Memory and control of action (pp. 345–
366). Amsterdam: North Holland.
Shea, J. B., & Zimny, S. T. (1988). Knowledge incorporation in motor representation.
In O. G. Meijer & K. Roth (Eds.), Complex movement behaviour: The motor-
action controversy (pp. 289–314). Amsterdam: Elsevier.
Shewokis, P. A., Rey, P. D., & Simpson, K. J. (1998). A test of retroactive inhibition
as an explanation of contextual interference. Research Quarterly For Exercise
and Sport, 69, 70–74.
Simon, D. A., & Bjork, R. A. (2001). Metacogntion in motor learning. Journal of
Experimental Psychology: Learning, Memory and Cognition, 27, 907–912.
Simon, D. A., and Bjork, R. A. (2002). Models of performance in learning multi-
segment movement tasks: Consequences for acquisition, retention and judge-
ments of learning. Journal of Experimental Psychology: Applied, 8, 222–232.
Smith, S. M., & Rothkopf, E. Z. (1984). Contextual enrichment and distribution of
practice in the classroom. Cognition And Instruction, 1, 341–358.
Son, L. K. (2004). Spacing one’s study: Evidence for a metacognitive control strategy.
Journal of Experimental Psychology: Learning, Memory, and Cognition, 30,
601–604.
Ste Marie, D. M., Clark, S. E., Findlay, L. C., & Latimer, A. E. (2004). High levels of
contextual interference enhance handwriting skill acquisition. Journal of Motor
Behavior, 36, 115–126.
Part II
Cognitive Neuroscience
Human Learning 75
Introduction
Cognitive Neuroscience of Learning and Memory

Thad Polk
University of Michigan
The chapters in this section focus on the cognitive neuroscience of

learning and memory. As its name suggests, the field of cognitive neu-
roscience attempts to relate cognition to neuroscience, that is, to under-
stand how thought is implemented in the brain. By attempting to relate
aspects of learning and memory to underlying neural mechanisms, these
chapters therefore serve as a bridge between the chapters discussing
learning and memory at the behavioral level and at the neural level.
Perhaps the single most influential finding from the cognitive neu-
roscience of learning and memory is that there are multiple, relatively inde-
pendent memory systems in the human brain. Long-term memory depends on
different neural substrates than does working memory, and working memory
depends on different substrates than does sensory memory. These systems
themselves can be further subdivided at a neural level. Furthermore, the
executive systems that control these memory systems also depend on differ-
ent neural substrates than do the core memory systems themselves.
Not surprisingly, this fractionation of memory into separate neural com-
ponents is a central theme in most of the chapters in this section. For
example, Reber summarizes current research in the cognitive neuroscience
of long-term memory. Starting from the familiar dissociation between
declarative and nondeclarative memory, he discusses patient studies, animal
studies, and imaging studies that shed light on the neural systems that
underlie each. He also reviews evidence on the neural mechanisms that
subserve meta-memory control processes and how these mechanisms inter-
act with the neural systems involved in declarative memory.
The cognitive control of memory is also a central theme in the chapter
by Lustig and Flegal, although their goal is to review the cognitive neu-
roscience of learning and memory from the perspective of aging. Specifically,
76 Part II
they summarize empirical studies illustrating the effects of aging on different

(and often neurally dissociable) memory systems. Their review demonstrates
that control processes play a particularly important role in many different
types of age-related memory changes.
Atkins and Reuter-Lorenz also discuss effects across various memory
systems, although their focus is on emotion rather than aging. They sum-
marize animal studies, patient studies, and imaging studies, illustrating the
dramatic effects that emotions can have on three major learning and
memory systems: implicit (nondeclarative) memory, declarative memory,
and working memory. Their review demonstrates the central role that the
amygdala plays in emotional learning and memory in all three domains.
Bo, Langan, and Seidler focus on the cognitive neuroscience of skill
acquisition in their chapter. They discuss two different types of motor
learning: (1) sensorimotor adaptation (modifying movements in response
to changes in sensory input or motor output requirements) and (2) sequence
learning (learning to combine separate actions in one coherent skill). They
review many of the major theories for both types of motor learning and point
out similarities and differences among the theoretical approaches.
Finally, the chapter by Newman and Polk represents a very different
branch of the cognitive neuroscience of learning and memory, namely the
computational modeling approach. They show how a few well-established
properties of neural computation can interact to give rise to emergent
functionality and how this emergent functionality has been applied to
explain many aspects of learning and memory. They argue that computa-
tional modeling therefore provides the hope of developing a very explicit
bridge from the brain to the mind.
The study of learning and memory is undoubtedly the largest and the
most active area in all of cognitive neuroscience, and these five chapters
certainly do not provide an exhaustive review of the field. Nevertheless,
they do provide a compelling sampling of some of the hot topics in the area.
Furthermore, when read in the context of parallel work in learning and
memory in the other sections of this book, these chapters illustrate many of
the commonalities (and differences) that have been discovered when dif-
ferent approaches are applied to the study of learning and memory.
Human Learning 77
The Computational Cognitive

Neuroscience of Learning and
Memory: Principles and Models
Lee I. Newman and Thad A. Polk
University of Michigan, Department of Psychology
1. The computational cognitive neuroscience of learning

and memory: computational principles and models
In this chapter, our objective is to provide the reader a brief introduction

to learning and memory as viewed from the perspective of neural computa-
tion. The computational agenda in cognitive neuroscience focuses on two
important tasks, each relevant to the study of learning and memory: to
identify and understand how neural computations are carried out in the
brain and to use computational simulations to better understand how these
neural computations give rise to cognitive function. Our agenda is (i) to
present a set of core principles of neural computation that have been
derived from empirical work in neuroscience and (ii) to demonstrate how
these principles can serve as building blocks for computational models that
have been used to explain many aspects of learning and memory.
2. Computational learning: three approaches
Computational cognitive neuroscientists often divide learning into three

broad classes: supervised, reinforcement, and unsupervised learning. The
three classes share the common objective of forming stored representa-
tions to be later recalled and used to guide behavior; they differ in the
specific mechanisms by which stored representations are derived from
78 Lee I. Newman and Thad A. Polk
experience. We will briefly discuss1 these three types of learning at a

behavioral level using a concrete example: the task faced by a young
child learning the name of an object visually experienced in the environ-
ment.
In a supervised learning task, the learner is provided with explicit feed-
back on a desired outcome. For example, a child might point to a dog and
say ‘‘cat’’ only to be told ‘‘dog’’ by a parent. After a number of such episodes,
this corrective feedback ultimately allows the child to form a stored repre-
sentation associating the visual image of a dog with the utterance ‘‘dog.’’ In
reinforcement learning, the learner is not explicitly corrected, but rather
receives either immediate or delayed feedback in the form of rewards and/
or punishments. Returning to our example, rewarding a child with a snack
every time she says ‘‘dog’’ in the presence of a dog might reinforce the
correct association and facilitate its acquisition. However, this reinforce-
ment-driven task is typically much more difficult than its supervised coun-
terpart because the nature of the feedback used to drive the learning
process is less specific: rather than being told the correct name, the child
is simply given a positive or negative reinforcement. This difficulty may be
further compounded by the fact that reinforcements may not be immediate
and the same reinforcement may be given for different but related beha-
viors (e.g., the child may receive the same snack for correctly saying ‘‘car’’
in the presence of a car).
Although the supervised and reinforcement-learning tasks differ in the
type of feedback received by the learner (i.e., explicit correction vs. less
specific rewards), they have in common the fact that experience (i.e.,
seeing a dog) is always accompanied by some form of behavioral feedback.
In contrast, in an unsupervised learning task, the learner must make sense
of experience in the absence of any feedback. For example, a child may
frequently see a dog and at the same time hear the word ‘‘dog,’’ and
ultimately come to associate the word with the object after many instances
of such experience.
A great deal of work has attempted to understand how these three types
of learning are instantiated in neural computations. One of the earliest
proposals was the seminal work by Hebb (1949) in which he offered a
neural theory of unsupervised learning, which we will discuss in the
following section. Unsupervised learning in the brain has since been widely
studied, for example, in the context of research on the perceptual learning
in sensory cortices (Fahle & Poggio, 2002; Hinton, G.E. & Sejnowski, 1999;
Rolls & Deco, 2002) and the development of topographic cortical maps
1
For more detailed and mathematical treatment of these three classes of learning, see Dayan &
Abbott (2001), Hastie, Tibshirani, & Friedman (2001), and Sutton & Barto (1998).
The Computational Cognitive Neuroscience of Learning and Memory 79
(Obermayer & Sejnowski, 2001). Significant progress has also been made in
understanding the neural basis of reinforcement learning in the mesolimbic
dopamine system (Berridge & Robinson, 1998; Dayan & Balleine, 2002;
O’Doherty, 2004; Schultz, 2006). The neural basis of supervised learning is
less well-understood at present, but supervised learning processes can be
implemented with some of the same computational mechanisms that sup-
port unsupervised learning, and moreover, supervisory feedback can be
provided to one population of neurons by the output of others (Dayan &
Abbott, 2001). Therefore, it may very well be the case that although super-
vised and unsupervised learning differ in their informational content at the
level of behavioral tasks, at a neural level they may be more similar than
different. In this chapter, we have chosen to focus on unsupervised models
of learning and memory, in part because their grounding in neurobiology is
better understood, and in part because these models are amenable to
derivation from established principles of neural computation to which we
now turn our attention.
3. Principles of neural computation
Empirical work in neuroscience has led to models of learning and

memory at multiple levels ranging from molecular models of synaptic
modification (Destexhe, Mainen, & Sejnowski, 1994; Xie, Liaw, Baudry, &
Berger, 1997; Zador, Koch, & Brown, 1990), to models of how a population
of neurons can encode the shape of an object (Logothetis, Pauls, & Poggio,
1995; Poggio & Edelman, 1990), to models of how multiple brain areas
cooperate in support of memory systems (McClelland, McNaughton, &
O’Reilly, 1995). Although this broad range of models differs in the detail
with which they capture neurobiological constraints, they share a common
set of neural building blocks. In this section, we introduce the reader to a
core set of these building blocks. For comprehensive and mathematical
treatments of the core principles of computational neuroscience, we direct
the interested reader to Abbott and Sejnowski (1999), Arbib (2003), Dayan
and Abbott (2001), Gerstner and Kistler (2002), Gluck and Myers (2001),
Koch and Davis (1994), Hinton and Sejnowski (1999), Obermayer and
Sejnowski (2001), and O’Reilly and Munakata (2000).
3.1. PRINCIPLE 1: DISTRIBUTED, OVERLAPPING REPRESENTATIONS
Throughout cortex, information is represented by patterns of

activity across populations of neuronal processing elements as revealed
by both electrophysiological recordings (e.g., Georgopoulos, Schwartz, &
Kettner, 1986) and functional neuroimaging (e.g., Ishai, Ungerleider,
Martin, & Haxby, 2000). These patterns of activity are known as distributed
representations because the burden of representing information is shared
across many neuronal elements. Distributed representations can be con-
trasted with localist representations in which information is represented in
the extreme by a single neuron, often referred to as a grandmother cell
(a single cell capable of representing complex information, such as the face
of one’s grandmother). It is important to note that the degree of distribution
of representations and the degree of localization of representations in the
brain are not identical concepts. A representation can be distributed over a
large number of neurons in a single, small region of cortex (distributed and
localized), or more widely distributed over neurons spanning disparate
cortical regions (distributed and not localized). Closely related to the
distributed nature of neural representations is the fact that distributed
cortical representations also tend to be overlapping: two similar
patterns of information typically share many neuronal processing
elements, while two highly dissimilar patterns of information share few
processing elements.
The distributed and overlapping nature of representations has several
important computational implications for learning and memory. First, dis-
tributed representations are robust to partial damage because patterns are
represented over a large number of neurons. Second, it can be shown
mathematically that distributed representations are efficient because they
can store a large number of mnemonic patterns relative to localist coding
schemes (Rolls & Tovee, 1995; Rolls, Treves, & Tovee, 1997; Rumelhart,
McClelland, & PDP ResearchGroup, 1986, Chapter 3). That the brain makes
use of this efficiency has been shown empirically (e.g., Ishai, Ungerleider,
Martin, Schouten, & Haxby, 1999; Rolls & Tovee, 1995; Rolls et al., 1997).
Third, the tendency of distributed representations to share processing
elements provides a natural computational basis for generalization and
inference. For example, if a neural representation of the concept dog is
associated with a representation of the concept furry, then there is a
neural basis for inferring the association of furriness for other objects
that are similar to dog.
3.2. PRINCIPLE 2: RECURRENT CONNECTIVITY

It is well-established that connectivity in cortex is massively recurrent,
at multiple levels of processing (Felleman & McClendon, 1991; Fuster,
1995; Kandel, Schwartz, & Jessell, 2000). Within a cortical network, two
nearby neurons may be recurrently connected such that they can mutually
excite or inhibit each another. However, neurons within a network need
not be immediate neighbors to be recurrently connected; activity can
spread from one neuron through a chain of many other neurons, ultimately
returning to affect the original neuron. Neurons also need not be in the
same network to be recurrently connected. Neural pathways between two
networks are typically recurrent, thus allowing entire patterns of activity
across networks to mutually affect one another.
There are several important computational implications of recurrent
connectivity. First, recurrence can lead to reverberatory activity within a
network of neurons, allowing the network to sustain a pattern of activity
even when input to the network is no longer present. Second, recurrent
connectivity between networks allows the flow of information processing
in the brain to proceed not only in a bottom-up manner but also in a top-
down manner, thereby enabling higher-order representations (e.g., prior
knowledge and expectations) to affect lower-order representations (e.g.,
visual object identity).
3.3. PRINCIPLE 3: HEBBIAN LEARNING
As mentioned previously, Hebb proposed one of the first neural theories

of learning (Hebb, 1949). Hebb’s idea was that if two connected neurons
are frequently active at the same time, some form of physiological change
in their connectivity (learning) could render them more likely to be coac-
tive in the future, thus providing a physiological basis for memory. This
form of unsupervised learning is frequently captured by the expression
‘‘neurons that fire together, wire together.’’ Hebb’s principle of learning
received empirical support when evidence for synaptic strengthening
(called long-term potentiation or LTP) was discovered in neural circuits
of the mollusk Aplysia (Castellucci & Kandel, 1976) and in hippocampal
neurons of the rabbit (Bliss & Gardner-Medwin, 1973; Bliss & Lomo, 1973).
The principle of Hebbian learning operating at the level of two neurons is
deceptively simple and yet we cannot overstate its importance to the
broader topic of memory formation. By generalizing Hebb’s learning rule
to networks of neurons, we will show in the next section how these
networks are capable of reproducing previously experienced patterns of
activity after repeated exposure to the same or similar inputs. The principle
of Hebbian learning thus provides an explicit account of how a pattern of
activity in a network of neurons can be stored in a pattern of synaptic
connections, thereby serving as a neural substrate of memory.
3.4. PRINCIPLE 4: COOPERATIVE AND COMPETITIVE ACTIVATION DYNAMICS
Hebbian learning is a powerful mechanism, but operating in conjunction

with recurrent connectivity without other constraints would be proble-
matic for the formation of memory. The problem is that because neurons
are highly interconnected, excitatory activity in a few neurons tends to
spread to neighboring neurons, and to their neighbors, and so on. This

problem is compounded by the presence of recurrent connectivity, which
allows activity to reverberate in the network creating mutually reinforcing
activity (cooperation) analogous to the situation when sound from a P.A.
system reverberates back to a microphone resulting in the familiar screech
of feedback. In the presence of runaway cooperative activity, Hebbian
learning would ultimately lead to indiscriminate reinforcement of all
synapses in a network (all neurons firing together would get wired
together), making it impossible to learn the meaningful patterns necessary
for memory formation. Fortunately, competitive processes serve as a bal-
ancing force that prevents such a result from occurring (Freund & Buzsaki,
1996; Gibson, Beierlein, & Connors, 1999). Neural competition is imple-
mented via inhibitory synapses that allow active neurons to inhibit less
active neurons. As activity spreads through a network, the more active
neurons tend to increasingly excite each other (cooperation) and at the
same time increasingly inhibit less active neurons (competition). After a
period of time, it is only the ‘‘winners’’ of these neural competitions that
remain active and that are selectively reinforced by Hebbian learning,
making them more likely to win again when the same pattern is again
presented to the network. In this way, neurons become specialized: they
respond selectively to some patterns of input, but not others.
3.5. PRINCIPLE 5: SPATIAL CONSTRAINTS
Studies of the structure of sensory cortices have revealed a columnar

organization in which columns of neurons are connected via local excita-
tory connections and slightly longer-range inhibitory connections (Fisken,
Garey, & Powell, 1975; Gilbert, 1992; Hubel & Wiesel, 1962; Mountcastle,
1957, 1997; Szentagothai, 1975). As a result of this differing spatial extent
of excitatory and inhibitory connections and the processes of cooperation
and competition discussed in the prior section, activity in sensory
cortices tends to occur in spatially localized bumps of activity. This
fact, combined with the principle of Hebbian learning, implies that learn-
ing via synaptic modifications also tends to be spatially localized. This
spatial constraint on learning has important implications for the structure
of memory, a topic that we address in our discussion of self-organizing
maps.
Taken together, the first four principles – distributed and overlapping
representation, recurrent connectivity, Hebbian learning, and neural coop-
eration and competition – provide a foundation for a computational frame-
work of learning and memory. These principles provide an explicit account
of how information is represented by patterns of activity in networks of
neurons, and how these patterns are transformed into synaptic
connections that serve as a neural substrate of memory. In the next section,

we derive from the first four building blocks an important class of compu-
tational models known as attractor networks. In the following section, we
add the fifth principle and derive a second class of models known as self-
organizing maps.
4. Modeling learning and memory with attractor networks
The attractor network model was originally proposed by Hopfield

(1982), and a range of variants have subsequently been developed, for
example (Ackley, Hinton, & Sejnowski, 1985; Hopfield & Tank, 1986).
This type of models is also commonly referred to as autoassociative net-
work or Hopfield network.
4.1. OVERVIEW OF THE MODEL
4.1.1. Architecture
The architecture of an attractor network is shown in Fig. 1a. The
network consists of set of neural processing units. The activity of each
unit at a given time is specified by a numerical value that serves as an
abstract representation of the unit’s firing rate. In the attractor net formu-
lation presented here, we consider binary-valued units that are either on
(+1) or off (1). Attractor nets can also be built with continuous-valued
units and/or with units that take on only positive values. Units in the
network are linked to every other unit via recurrent connections (Principle 2).
Associated with each connection is a modifiable and persistent synaptic
weight that represents the strength of the connection.
(a) (b) (c)
Input Input Input

Attractor network Attractor network Attractor network
Fig. 1. Attractor network architecture and learning. (a) An attractor net consists of
a set of neural units connected by recurrent synaptic connections (gray lines).
Presentation of an input pattern forces some units to become active (filled
circles) while others remain inactive (open circles). (b) Hebbian learning leads to
a strengthening of connections between coactive units (black lines). (c) When input
activity is no longer present, the pattern is now stored, and can be later recalled, via
the synaptic connections in the network.
4.1.2. Processing
When an external input pattern is presented to the network, each unit in
the network is activated based on the level of activation in the correspond-
ing unit in the input pattern. As shown in Fig. 1a, presentation of an input
pattern leads to a distributed pattern of activity (Principle 1) in which some
units in the network become active (filled circles) while others remain
inactive (open circles). In addition to external input, each unit also
receives activation from other units in the network via recurrent connec-
tions (Principle 2). When an external input is present, it forces a pattern of
activity on the network. However, once the external input is removed, the
activity of each unit becomes entirely determined by the interaction of the
excitatory and inhibitory influence of other units in the network (Principle
4). Units receiving more inhibition than excitation become inactive as
neural competition plays out, and units that receive more excitation than
inhibition ultimately become the neural ‘‘winners’’ and thus are able to
maintain their activity.
4.1.3. Learning
How does an attractor network come to serve as a basis for memory?
When a pattern of input activity (+1, 1 values) is imposed on the network,
the synaptic connection weights between every pair of neurons are
updated based on their coactivity (Principle 3), with the new weight typi-
cally being the product of the activities of the two neurons. The connection
weights between any two active neurons are strengthened (Fig. 1b) and as
a result, in the absence of external input, the input pattern has been stored
in the strengthened connection weights of the network (Fig. 1c). Connec-
tions between pairs of neurons that are not coactive are weakened (not
shown).
4.2. PROPERTIES OF THE MODEL
4.2.1. Pattern completion

Just as in biological neural networks, attractor networks store pre-
viously experienced patterns of activity in the strength of their synaptic
connections. And as in real neural networks, the combination of recurrent
connectivity and cooperative and competitive processes allows these simu-
lated networks to recall stored memories based on partial cues. For exam-
ple, if a noisy or a degraded version of a previously experienced input is
presented to a network, it will activate a subset of units in the previously
stored pattern (Fig. 2a). With the input removed, the activation of the
network will evolve as excitatory and inhibitory activity spreads across
the previously strengthened recurrent connections (Fig. 2b). Although
(a) (b) (c)
Input Input Input

Fig. 2. Pattern completion. (a) A partial version of a previously learned pattern

(thick gray lines) is presented to the network and leads activity of a subset of units
(black circles). (b) With the input removed, activity spreads across previously
strengthened connections (black lines). (c) After activity has settled, the network
is able to recall the previously stored pattern (Fig. 1).
beyond the scope of this chapter, it can be shown mathematically that the
network will settle into a previously stored pattern of activity that is similar
to the novel pattern presented as input (Hopfield, 1982). In this way,
complete patterns can be retrieved from memory based only on partial
input (Fig. 2c).
4.2.2. Pattern generalization

The ability to generalize information is a critical aspect of cognition,
allowing novel objects to be categorized based on their similarity to pre-
viously experienced objects. There is a large body of empirical evidence
from single-unit recording studies (e.g., Tanaka, Saito, Fukada, & Moriya,
1991), and functional imaging studies (e.g., Ishai et al., 2000) showing that
exposure to similar stimuli elicits similar patterns of neural activity in the
cortex. Attractor networks exhibit the same kind of generalization as a
result of their activation dynamics and the overlap in their stored mnemo-
nic representations. In fact, the generalization property of these networks
is directly related to their ability to complete patterns. Consider the net-
work shown in Fig. 3a that receives activation from a novel input. Two
patterns have been previously been stored in the network. The first pattern
is shown by the thick gray connections and the second by the dotted
gray connections. When the input is removed, activation spreads across
the previously strengthened recurrent connections (Fig. 3b). The two pre-
viously stored patterns compete, but because the first input pattern is more
similar to the novel input pattern (they share more units), this first pattern
receives more cooperative excitation than the second pattern (and is also
able to impose greater inhibition on it). The novel pattern is ‘‘attracted’’ to
the most similar pattern that has already been stored (thus the name
attractor network), and in this way, the network is able to generalize on
past experience (Fig. 3c).
(a) (b) (c)
Input Input Input

Fig. 3. Pattern generalization. (a) A novel pattern is presented to a network that

contains two previously stored patterns (thick gray lines, dotted gray lines).
(b) With the input removed, activity spreads across previously strengthened
connections, and the two previously stored patterns compete. (c) After activity
has settled, the network produces the pattern that is most similar to the novel input.
4.2.3. Similarity-based interference

Figure 4a shows a novel pattern (black units and connections) being
learned in a network with a different previously stored pattern (gray con-
nections). The new pattern shares no units with the previously stored
pattern and therefore the connections strengthened in learning the new
pattern (black connections) will not interfere with the previously stored
pattern (gray connections). Now consider (Fig. 4b) what happens when the
new pattern shares units with the previously stored pattern such that the
two representations are overlapping in addition to being distributed (Prin-
ciple 1). This representational overlap has important consequences for the
network. Unit A is part of both patterns, while unit B is a member of the
previously stored pattern and not the input pattern. When unit A becomes
active, as a member of the first pattern it should tend to cooperatively
(a) (b)
B A
Input Input
Attractor network Attractor network
Fig. 4. Similarity-based interference. (a) A novel pattern (black units and

connections) is presented to a network with a previously stored pattern (gray
connections). These two patterns have no units in common, so new learning does
not interfere with previous learning. (b) The novel and previously stored pattern
have unit A in common. Unit A will excite unit B based on the previously
strengthened excitatory connection, but in learning the pattern, the connection
between units A and B should be inhibitory. Learning overlapping patterns thus
forces a compromise on the strength of connections.
excite unit B, but as a member of the new pattern to be learned it should

result in a connection that competitively inhibits unit B. This poses a
problem in that the connection between unit A and B cannot achieve
both goals simultaneously. The learning of new patterns thus can interfere
with previously stored patterns, resulting in connection weights that must
take on values that reflect the relationship between units across all of the
input patterns learned by the network. This interference has both desirable
and undesirable consequences. On the positive side, a network that learns a
set of weights that are a neural compromise based on many experienced
patterns is beneficial in that it supports generalization. On the negative
side, interference can result in a partial input pattern being recalled as a
similar, but incorrect pattern. This type of similarity-based interference is a
common behavioral phenomenon for which the attractor net offers a
neural explanation (e.g., Chappell & Humphreys, 1994; Jones & Polk,
2002). Lastly, the compromise that must be achieved by the network in
learning a set of weights imposes a capacity constraint on the network
limiting the number of patterns that can be accurately stored (Abumostafa &
St. Jacques, 1985; McEliece, Posner, Rodemich, & Venkatesh, 1987; Treves &
Rolls, 1991).
4.2.4. Short-term memory

As mentioned previously, reverberatory activity over recurrent connec-
tions allows an attractor network to maintain a pattern of activity after an
input is no longer present. Thus, in addition to serving as a model of long-
term memory, attractor networks also provide a computational account of
how short-term memory might be instantiated in the brain (Zipser, Kehoe,
Littlewort, & Fuster, 1993).
4.2.5. Resistance to damage

As a result of their distributed representations, recurrent connectivity,
and cooperative and competitive activation dynamics, the recall perfor-
mance of attractor networks is resistant to partial network damage. In a
lesioned network, an external input may not result in a complete pattern of
activity, but often a partial pattern from one network provides a sufficient
input to another network for it to successfully recall the intended pattern
(via pattern completion).
4.3. APPLICATIONS OF THE MODEL
The unique properties of attractor networks have been used to simulate

a wide range of learning and memory phenomena, for example, working
memory (Amit & Brunel, 1997; Compte, Brunel, Goldman-Rakic, & Wang,
2000; Deco & Rolls, 2003; Farrell & Lewandowsky, 2002; Jones & Polk,
2002; Miller, Brody, Romo, & Wang, 2003; Zipser et al., 1993), hippocam-
pally based episodic memory (Byrne, Becker, & Burgess, 2007; Gluck &
Myers, 2001; Hasselmo, Wyble, & Wallenstein, 1996; Stringer, Rolls, &
Trappenberg, 2004), similarity judgments in semantic memory (Cree,
McRae, & McNorgan, 1999; Polk, Behensky, Gonzalez, & Smith, 2002),
word reading (Plaut, McClelland, Seidenberg, & Patterson, 1996),
and motor skill learning (Newell, Liu, & Mayer-Kress, 2001). Because
they can be artificially lesioned, attractor networks have also provided a
valuable means for simulating and understanding the patterns of deficits
exhibited by patients suffering from different types of brain damage and
dysfunction, for example, acquired dyslexia (Hinton, G. E. & Shallice,
1991), schizophrenia (Hoffman & McGlashan, 2001), dysexecutive
syndrome (Polk, Simen, Lewis, & Freedman, 2002), and visual agnosia
(Brunel, 1993).
5. Modeling neural learning and memory with self-organizing maps
Although attractors have proven useful as tools for understanding the

neurocomputational basis of many cognitive phenomena and for generat-
ing novel research hypotheses, they are not fully consistent with what
is currently known about learning and memory in sensory cortices.
Anatomical and physiological studies of sensory cortex have revealed
that topography is an important principle of neural organization (Kandel
et al., 2000). Topography refers to a form of neural organization in which
nearby neurons in the cortex tend to respond to similar inputs. For exam-
ple, most of the visual cortex is organized retinotopically such that neigh-
boring neurons in cortex respond to stimulation of nearby regions of the
retina (e.g., Sereno et al., 1995; Tootell, Switkes, Silverman, & Hamilton,
1988; Tusa, Palmer, & Rosenquist, 1978; Van Essen & Gallant, 1994).
Attractor models do not incorporate the spatial constraints outlined
in Principle 5, and as a result this class of models cannot explain how
topographically organized representations are learned. We, therefore,
turn our focus to a second class of computational models called self
organizing maps (SOMs). The SOM model was originally proposed by
Kohonen (1982a; 1982b) and subsequently many variants have been developed
(e.g., Haese & Goodhill, 2001; Heskes, 2001; Kohonen, Kaski, & Lappalainen,
1997; Koikkalainen & Oja, 1990; Luttrell, 1988, 1989; Tereshko & Allinson, 2002)
and much theoretical work has been conducted to better understand this class
of model analytically (e.g., Cottrell, Fort, & Pages, 1998; Heskes, 2001;
Kohonen, 1993).
5.1. OVERVIEW OF THE MODEL
5.1.1. Architecture
The architecture of an SOM is shown in Fig. 5. The SOM consists of a
grid of neural processing units arranged in a two-dimensional map (a 5 5
map in Fig. 5) that is a computational abstraction of a locally connected
population of neurons in a contiguous region of cortical tissue. Unlike the
attractor network in which units are explicitly connected, in the SOM,
connectivity is implicit in the computational equations that determine
how units become active and how they learn; we will discuss in a later
section how this pattern of connectivity can be derived from a subset of our
five computational principles. Units within the map are indexed based on
their spatial location, and each unit has associated with it a modifiable
weight vector that specifies the input pattern for which the unit is
best tuned (corresponding to the unit’s receptive field). The weight
vector for a given unit is a computational abstraction of a set of synaptic
connections between the unit and all units (in another map) from which it
receives input.
2-D Cortical map
.2
.8
.4
.2
.9 Neighborhood of units
Input .3 updated during learning
vector
.9 Weight vector
.3 of winning unit
.2
.8
Fig. 5. Self-organizing map (SOM) architecture and learning. The SOM is a two-
dimensional map of neural units located on a spatial grid. Each unit is represented
by a weight vector representing its preferred features (here five features), also
known as its receptive field. When an input vector is presented to the network,
similar units become active (black and dark gray units) while dissimilar units
remain relatively inactive (light gray and white units). Learning occurs when the
weight vectors of the most active cell (black) and its neighbors (dark gray units) are
modified so that they become more similar to the input.
5.1.2. Processing
When an input pattern is presented to the SOM, each unit in the map
has access to the input. The activity of each unit is determined by com-
paring the similarity between a unit’s weight vector and the current input
pattern, with similarity typically measured as the inverse of some metric
of vector distance such as Euclidean distance. If the input pattern exactly
matches a unit’s weight vector, this unit will fire at its maximum rate
(typically set to 1); poor matches between the input and a unit’s weight
vector result in little or no activity in that unit. The activity of the map
is thus a distributed representation (Principle 1) in which the input is
represented as the activity pattern induced in the map as shown in Fig. 5.2
Furthermore, because units are represented by weight vectors, nearby
units can (and typically do) have similarly valued elements (e.g., [0.20,
0.40, 0.60, 0.10, 0.70] and [0.20, 0.46, 0.63, 0.11, 0.70]); SOM representa-
tions are therefore also overlapping in addition to being distributed
(Principle 1).
5.1.3. Learning
Units within the map compete to represent an input pattern. This com-
petitive process is carried out computationally by simply searching the
map for the unit with maximum activity; this ‘‘winning’’ unit (black unit in
Fig. 5) will be the one whose weight vector is most similar to the input.
Learning is accomplished by modifying the weight vector of this winning
unit so that its values are more similar to the input pattern, thereby making
this unit more likely to win again if the same input is present. Critically, the
weight vectors of other units in close spatial proximity to the winning unit
(dark gray units in Fig. 5) are also updated (Principle 5), with the magni-
tude of the updates typically proportional to the distance between each
unit and the winning unit. As a result of this spatial constraint on learning,
with experience, nearby neurons in the map come to have similar weight
vectors and thus come to represent similar input patterns. This spatially
constrained learning leads to a process known as self-organization from
which topographic representations develop.
Armed with this knowledge of SOM function, we can now explain how
Principles 2, 3, and 4 are implicitly present in the model. As mentioned
earlier in this chapter, the spatial constraint on learning is based on the
2
In some formulations of the SOM model, the most active unit fully inhibits the activity of all
other units thus resulting in what would be considered a purely localist representation rather
than a distributed representation. However, in determining which unit is most active, the
activity of all units must be compared and therefore in these models, the underlying repre-
sentation is implicitly distributed.
finding that in sensory cortex, the spatial extent of recurrent excitatory

connections is less than the spatial extent of competitive inhibitory connec-
tions. As a result, cooperative activity (Principle 4) across recurrent
connections (Principle 2) leads to spatially localized ‘‘bumps’’ of activity in
cortical tissue. As activity spreads through the network, these bumps of
activity compete (Principle 4) via inhibitory connections. As in attractor
networks, after the competition has played out, only a subset of units
in the network will remain active; however, unlike attractor networks in
which winning neurons are typically scattered throughout the network, in a
SOM, winning units are confined to a spatially localized region of the
cortex—a winning ‘‘bump’’ of activity (black and dark gray units in Fig. 5).
Although these activation dynamics are not directly instantiated in the SOM
model, they are implicitly instantiated in the spatial constraint imposed on
learning. The winning bump of activity corresponds directly to the activities
of the winning unit (identified in the learning process) and its neighbors.
What about Hebbian learning (Principle 3)? As a reminder, Hebbian learning
occurs when the connection between coactive neurons is reinforced. Mod-
ifying the weights of the winning unit and its neighbors so that they are more
similar to the input corresponds directly to Hebbian reinforcement of the
implicit connections between these active units and the input pattern. There-
fore, although not immediately obvious in its formulation, the SOM model
can, in fact, he derived from our five core principles of neural computation.
5.2. PROPERTIES OF THE MODEL
Self-organizing maps share several important properties exhibited by

attractor networks and also possess several unique properties.
5.2.1. Pattern generalization

As in the attractor model, SOMs have an ability to generalize based on
prior learning. If presented with a novel input, the unit most similar to this
input will become partially active as will its neighboring units. Thus,
although a map may never have experienced the input, it is able to produce
a best-guess response based on prior experience.
5.2.2. Similarity-based interference

In the SOM model, learning updates are imposed on a neighborhood of
units. If a new input pattern is similar to a previously learned input pattern,
it will tend to activate a similar neighborhood of units. As a result,
the weight updates made in response to the new pattern will modify the
weights of previously updated units. This same similarity-based interfer-
ence was present in the attractor model, and results in the same set of
benefits and problems, which we will not repeat here.
b
b b
b
b b c
b b c
b e b
e c c
d e b c
e e c c c
d e c c
d e e c
d e e
d e e c
d d f f
d d e e f
d f f
d a a f
d f f
a f f
d a
a a a
a a
Input sequence a
a
{c,a,f,e,d,a,d,c,e...}
Fig. 6. Self-organized topographic organization. Although experiencing a sequence

of different inputs, the SOM identifies statistical clusters based on the features of
these inputs. After learning is complete, the map has self-organized such that there
are spatial neighborhoods of units (labeled a, b, c, d, e, and f) that have similar
weight vectors and therefore that will respond to similar types of inputs.
5.2.3. Resistance to damage

Although an SOM does not have the ability to complete partial input
patterns in the same way as an attractor network, it is nevertheless resis-
tant to damage for a different reason. If a set of units that would normally
respond to a particular input is damaged, they are no longer able to
respond. However, because nearby neurons in an SOM have similar weight
vectors, the winning unit will tend to be an intact unit that is located near
the region where the map response should have occurred, and therefore
this unit is likely to provide reasonable match to the input.
5.2.4. Clustering via emergent topography

As mentioned above, imposing a spatial constraint on learning leads to a
self-organizing process in which neurons in nearby regions of a map learn
to represent similar types of inputs. As a result, spatially organized repre-
sentations, or topographies, naturally emerge from the learning process.
Self-organizing maps are thus able to find and extract statistical clusters of
features from a set of input patterns and to store these clusters in the form
of a spatial topography (Fig. 6).
5.3. APPLICATIONS OF THE MODEL
Self-organizing maps have served an important role in computational neu-

roscience by allowing researchers to explicitly test theories of how topo-
graphic sensory representations are structured and learned in the brain.
Much of this work has been done in the study of visual cortex where simula-
tions using SOMs have been shown to accurately reproduce the same types of
topographies (e.g., orientation, ocular dominance, color blobs, movement
direction) found experimentally in the cortex (Barrow, Bray, & Budd, 1996;
Carreira-Perpinan, Lister, & Goodhill, 2005; Goodhill, 1993; Goodhill & Will-
shaw, 1994; Olson & Grossberg, 1998; Sirosh & Miikkulainen, 1997; Sit &
Miikkulainen, 2006). At a behavioral level, SOMs have been used to model
and help explain a wide range of learning and memory phenomena including
categorical speech perception (Guenther & Gjaja, 1996), lexical development
(Li, Farkas, & MacWhinney, 2004), and category learning and object recogni-
tion (Bradski & Grossberg, 1995; Carpenter, Grossberg, & Rosen, 1991;
Newman & Polk, 2007; Polk & Farah, 1998; Ritter & Kohonen, 1989).
Self-organizing maps have also been used to help understand learning and
memory impairments such as dyslexia (Miikkulainen, 1997; Tuckova &
Zetocha, 2006) and noun–verb naming impairments (Vinson & Vigliocco,
2002).
6. Conclusions
Experimental work in the neurosciences has produced a wealth of

detailed knowledge about neural connectivity, neural processing, and
neural representation. In this chapter, we hope to have demonstrated to
the reader how knowledge of neurobiology can be directly translated
into explicit computational principles; how these principles can serve
as building blocks for constructing computational models; and how
these computational models can help provide neural explanations of
phenomena associated with learning and memory. We have, out of neces-
sity, focused on a small set of computational principles, on only two
classes of models, and only a few representative phenomena that these
models help to explain. Despite this restricted focus, we hope the reader
is left with a compelling sense of how the computational perspective in
cognitive neuroscience can inform our understanding of learning and
memory.
References
Abbott, L., & Sejnowski, T. J. (Eds.). (1999). Neural codes and distributed
representations : foundations of neural computation. Cambridge, Mass.: MIT
Press.
Abumostafa, Y. S., & Stjacques, J. M. (1985). Information capacity of the Hopfield
model. IEEE Transactions on Information Theory, 31(4), 461–464.
Ackley, D. H., Hinton, G. E., & Sejnowski, T. J. (1985). A learning algorithm for
Boltzmann machines. Cognitive Science, 9(1), 147–169.
Amit, D. J., & Brunel, N. (1997). Model of global spontaneous activity and local
structured activity during delay periods in the cerebral cortex. Cerebral Cortex,
7(3), 237–252.
Arbib, M. A. (2003). The Handbook of brain theory and neural networks
(2nd edition). Cambridge, Mass.: MIT Press.
Barrow, H. G., Bray, A. J., & Budd, J. M. L. (1996). A self-organizing model of
‘‘color blob’’ formation. Neural Computation, 8(7), 1427–1448.
Berridge, K. C., & Robinson, T. E. (1998). What is the role of dopamine in reward:
hedonic impact, reward learning, or incentive salience? Brain Research Reviews,
28(3), 309–369.
Bliss, T. V. P., & Gardner-Medwin, A. R. (1973). Long-lasting potentiation of
synaptic transmission in the dentate area of the unanaesthetized rabbit
following stimulation of the perforant path. Journal of Physiology-London,
232(2), 357–374.
Bliss, T. V. P., & Lomo, T. (1973). Long-lasting potentiation of synaptic transmission
in the dentate area of the anaesthetized rabbit following stimulation of the
perforant path. Journal of Physiology-London, 232(2), 331–356.
Bradski, G., & Grossberg, S. (1995). Fast-learning VIEWNET architectures for
recognizing three-dimensional objects from multiple two-dimensional views.
Neural Networks, 8(7–8), 1053–1080.
Brunel, N. (1993). Effect of synapse dilution on the memory retrieval in structured
attractor neural networks. Journal De Physique I, 3(8), 1693–1715.
Byrne, P., Becker, S., & Burgess, N. (2007). Remembering the past and imagining
the future: a neural model of spatial memory and imagery. Psychol Rev, 114(2),
340–375.
Carpenter, G. A., Grossberg, S., & Rosen, D. B. (1991). Fuzzy ART: Fast stable
learning and categorization of analog patterns by an adaptive resonance system.
Neural Networks, 4(6), 759–771.
Carreira-Perpinan, M. A., Lister, R. J., & Goodhill, G. J. (2005). A computational
model for the development of multiple maps in primary visual cortex. Cerebral
Cortex, 15(8), 1222–1233.
Castellucci, V., & Kandel, E. R. (1976). Presynaptic facilitation as a mechanism for
behavioral sensitization in aplysia. Science, 194(4270), 1176–1178.
Chappell, M., & Humphreys, M. S. (1994). An autoassociative neural-network for
sparse representations – analysis and application to models of recognition and
cued-recall. Psychological Review, 101(1), 103–128.
Compte, A., Brunel, N., Goldman-Rakic, P. S., & Wang, X. J. (2000). Synaptic
mechanisms and network dynamics underlying spatial working memory in a
cortical network model. Cerebral Cortex, 10(9), 910-923.
Cottrell, M., Fort, J. C., & Pages, G. (1998). Theoretical aspects of the SOM
algorithm. Neurocomputing, 21(1–3), 119–138.
Cree, G. S., McRae, K., & McNorgan, C. (1999). An attractor model of lexical
conceptual processing: Simulating semantic priming. Cognitive Science, 23(3),
371–414.
Dayan, P., & Abbott, L. F. (2001). Theoretical neuroscience : computational and

mathematical modeling of neural systems. Cambridge, Mass.: Massachusetts
Institute of Technology Press.
Dayan, P., & Balleine, B. W. (2002). Reward, motivation, and reinforcement learn-
ing. Neuron, 36(2), 285–298.
Deco, G., & Rolls, E. T. (2003). Attention and working memory: a dynamical model
of neuronal activity in the prefrontal cortex. European Journal of Neuroscience,
18(8), 2374–2390.
Destexhe, A., Mainen, Z. F., & Sejnowski, T. J. (1994). Synthesis of models for
excitable membranes, synaptic transmission and neuromodulation using a com-
mon kinetic formalism. J Comput Neurosci, 1(3), 195–230.
Fahle, M., & Poggio, T. (2002). Perceptual learning. Cambridge, Mass.: MIT Press.
Farrell, S., & Lewandowsky, S. (2002). An endogenous distributed model of ordering
in serial recall. Psychonomic Bulletin & Review, 9(1), 59–79.
Felleman, D. J., & McClendon, E. (1991). Cortical connections of posterior infer-
otemporal cortex of Macaque monkeys. Investigative ophthalmology & visual
science, 32(4), 1036–1036.
Fisken, R. A., Garey, L. J., & Powell, T. P. S. (1975). Intrinsic, association and
commissural connections of area 17 of visual-cortex. Philosophical Transactions
of the Royal Society of London Series B-Biological Sciences, 272(919), 487–536.
Freund, T. F., & Buzsaki, G. (1996). Interneurons of the hippocampus. Hippocam-
pus, 6(4), 347–470.
Fuster, J. M. (1995). Memory in the cerebral cortex : an empirical approach to
neural networks in the human and nonhuman primate. Cambridge, Mass.:
MIT Press.
Georgopoulos, A. P., Schwartz, A. B., & Kettner, R. E. (1986). Neuronal population
coding of movement direction. Science, 233(4771), 1416–1419.
Gerstner, W., & Kistler, W. M. (2002). Spiking neuron models : single neurons,
populations, plasticity. Cambridge, U.K.; New York: Cambridge University Press.
Gibson, J. R., Beierlein, M., & Connors, B. W. (1999). Two networks of electrically
coupled inhibitory neurons in neocortex. Nature, 402(6757), 75–79.
Gilbert, C. D. (1992). Horizontal integration and cortical dynamics. Neuron, 9(1), 1–13.
Gluck, M. A., & Myers, C. (2001). Gateway to memory : an introduction to neural
network modeling of the hippocampus and learning. Cambridge, Mass.: MIT Press.
Goodhill, G. J. (1993). Topography and ocular dominance – a model exploring
positive correlations. Biological Cybernetics, 69(2), 109–118.
Goodhill, G. J., & Willshaw, D. J. (1994). Elastic net model of ocular dominance - overall
stripe pattern and monocular deprivation. Neural Computation, 6(4), 615–621.
Guenther, F. H., & Gjaja, M. N. (1996). The perceptual magnet effect as an emergent
property of neural map formation. Journal of the Acoustical Society of America,
100(2), 1111–1121.
Haese, K., & Goodhill, G. J. (2001). Auto-SOM: Recursive parameter estimation for
guidance of self-organizing feature maps. Neural Computation, 13(3), 595–619.
Hasselmo, M. E., Wyble, B. P., & Wallenstein, G. V. (1996). Encoding and retrieval of
episodic memories: Role of cholinergic and GABAergic modulation in the hippo-
campus. Hippocampus, 6(6), 693–708.
Hastie, T., Tibshirani, R., & Friedman, J. H. (2001). The elements of statistical
learning : data mining, inference, and prediction. New York: Springer.
Hebb, D. O. (1949). The organization of behavior; a neuropsychological theory.
New York,: Wiley.
Heskes, T. (2001). Self-organizing maps, vector quantization, and mixture modeling.
IEEE Transactions on Neural Networks, 12(6), 1299–1305.
Hinton, G. E., & Sejnowski, T. J. (Eds.). (1999). Unsupervised learning : foundations
of neural computation. Cambridge, Mass.: MIT Press.
Hinton, G. E., & Shallice, T. (1991). Lesioning an attractor network – investigations
in acquired dyslexia. Psychological Review, 98(1), 74–95.
Hoffman, R. E., & McGlashan, T. H. (2001). Neural network models of schizophre-
nia. Neuroscientist, 7(5), 441–454.
Hopfield, J. J. (1982). Neural networks and physical systems with emergent collec-
tive computational abilities. Proceedings of the National Academy of Sciences of
the United States of America-Biological Sciences, 79(8), 2554–2558.
Hopfield, J. J., & Tank, D. W. (1986). Computing with neural circuits – a model.
Science, 233(4764), 625–633.
Hubel, D. H., & Wiesel, T. N. (1962). Receptive fields, binocular interaction and functional
architecture in cats visual cortex. Journal of Physiology-London, 160(1), 106–154.
Ishai, A., Ungerleider, L. G., Martin, A., & Haxby, J. V. (2000). The representation
of objects in the human occipital and temporal cortex. Journal of Cognitive
Ishai, A., Ungerleider, L. G., Martin, A., Schouten, H. L., & Haxby, J. V. (1999).
Distributed representation of objects in the human ventral visual pathway.
Proceedings of the National Academy of Sciences of the United States of
America, 96(16), 9379–9384.
Jones, M., & Polk, T. A. (2002). An attractor network model of serial recall.
Cognitive Systems Research, 3, 45–55.
Kandel, E. R., Schwartz, J. H., & Jessell, T. M. (2000). Principles of neural science.
New York: McGraw-Hill, Health Professions Division.
Koch, C., & Davis, J. L. (Eds.). (1994). Large-scale neuronal theories of the brain.
Cambridge, Mass.: MIT Press.
Kohonen, T. (1982a). Analysis of a simple self-organizing process. Biological Cyber-
netics, 44(2), 135–140.
Kohonen, T. (1982b). Self-organized formation of topologically correct feature
maps. Biological Cybernetics, 43(1), 59–69.
Kohonen, T. (1993). Physiological interpretation of the self-organizing map algo-
rithm. Neural Networks, 6(7), 895–905.
Kohonen, T., Kaski, S., & Lappalainen, H. (1997). Self-organized formation of var-
ious invariant-feature filters in the adaptive-subspace SOM. Neural Computation,
9(6), 1321–1344.
Koikkalainen, P., & Oja, E. (1990). Self-organizing hierarchical feature maps. Paper
presented at the International Joint Conference on Neural Networks (IJCNN),
Washington, D.C.
Li, P., Farkas, I., & MacWhinney, B. (2004). Early lexical development in a self-
organizing neural network. Neural Networks, 17(8–9), 1345–1362.
Logothetis, N. K., Pauls, J., & Poggio, T. (1995). Shape representation in the inferior
temporal cortex of monkeys. Current biology : CB, 5(5), 552–563.
Luttrell, S. P. (1988). Self-organising multilayer topographic mappings. In Proceed-
ings of ICNN’88, IEEE International Conference on Neural Networks, vol. 1,
pp. 93–100. IEEE Service Center, Piscataway, NJ.
Luttrell, S. P. (1989). Hierarchical self-organising networks. In Proceedings of 1st
IEE Conference on Artificial Neural Networks, pp. 2–6. British Neural Network
Society, London, UK.
McClelland, J. L., McNaughton, B. L., & O’Reilly, R. C. (1995). Why there are
complementary learning systems in the hippocampus and neocortex: insights
from the successes and failures of connectionist models of learning and memory.
Psychol Rev, 102(3), 419–457.
McEliece, R. J., Posner, E. C., Rodemich, E. R., & Venkatesh, S. S. (1987). The
capacity of the hopfield associative memory. IEEE Transactions on Information
Theory, 33(4), 461–482.
Miikkulainen, R. (1997). Dyslexic and category-specific aphasic impairments in a self-
organizing feature map model of the lexicon. Brain and Language, 59(2), 334–366.
Miller, P., Brody, C. D., Romo, R., & Wang, X. J. (2003). A recurrent network model
of somatosensory parametric working memory in the prefrontal cortex. Cerebral
Cortex, 13(11), 1208–1218.
Mountcastle, V. B. (1957). Modality and topographic properties of single neurons of
cats somatic sensory cortex. Journal of Neurophysiology, 20(4), 408–434.
Mountcastle, V. B. (1997). The columnar organization of the neocortex. Brain, 120,
701–722.
Newell, K. M., Liu, Y. T., & Mayer-Kress, G. (2001). Time scales in motor learning
and development. Psychological Review, 108(1), 57–82.
Newman, L. I., & Polk, T. A. (2007). The emergence of semantic topography in a
neurally-inspird computational model. Paper presented at the Eighth Interna-
tional Conference on Cognitive Modeling.
O’Doherty, J. P. (2004). Reward representations and reward-related learning in the
human brain: insights from neuroimaging. Current opinion in neurobiology, 14(6),
769-776.
O’Reilly, R. C., & Munakata, Y. (2000). Computational explorations in cognitive
neuroscience
understanding the mind by simulating the brainunderstanding the mind by simulat-
ing the brain. Cambridge, Mass.: MIT Press.
Obermayer, K., & Sejnowski, T. J. (Eds.). (2001). Self-organizing map formation:
foundations of neural computation. Cambridge, Mass.: MIT Press.
Olson, S. J., & Grossberg, S. (1998). A neural network model for the development of
simple and complex cell receptive fields within cortical maps of orientation and
ocular dominance. Neural Networks, 11(2), 189–208.
Plaut, D. C., McClelland, J. L., Seidenberg, M. S., & Patterson, K. (1996). Under-
standing normal and impaired word reading: Computational principles in quasi-
regular domains. Psychological Review, 103(1), 56–115.
Poggio, T., & Edelman, S. (1990). A network that learns to recognize 3-dimensional
objects. Nature, 343(6255), 263–266.
Polk, T. A., Behensky, C., Gonzalez, R., & Smith, E. E. (2002). Rating the similarity of
simple perceptual stimuli: asymmetries induced by manipulating exposure fre-
quency. Cognition, 82(3), B75–B88.
Polk, T. A., & Farah, M. J. (1998). The neural development and organization of letter
recognition: Evidence from functional neuroimaging, computational modeling,
and behavioral studies. Proceedings of the National Academy of Sciences of the
United States of America, 95(3), 847–852.
Polk, T. A., Simen, P., Lewis, R. L., & Freedman, E. (2002). A computational
approach to control in complex cognition. Cognitive Brain Research, 15(1),
71–83.
Ritter, H., & Kohonen, T. (1989). Self-organizing semantic maps. Biological Cyber-
netics, 61(4), 241–254.
Rolls, E. T., & Deco, G. (2002). Computational neuroscience of vision. Oxford
[England] ; New York: Oxford University Press.
Rolls, E. T., & Tovee, M. J. (1995). Sparseness of the neuronal representation of
stimuli in the primate temporal visual cortex. Journal of Neurophysiology, 73(2),
713–726.
Rolls, E. T., Treves, A., & Tovee, M. J. (1997). The representational capacity
of the distributed encoding of information provided by populations of
neurons in primate temporal visual cortex. Experimental Brain Research,
114(1), 149–162.
Rumelhart, D. E., McClelland, J. L., & PDPResearchGroup. (1986). Parallel distrib-
uted processing: explorations in the microstructure of cognition. Cambridge,
Mass.: MIT Press.
Schultz, W. (2006). Behavioral theories and the neurophysiology of reward. Annual
Review of Psychology, 57, 87–115.
Sereno, M. I., Dale, A. M., Reppas, J. B., Kwong, K. K., Belliveau, J. W., Brady, T. J., et
al. (1995). Borders of Multiple Visual Areas in Humans Revealed by Functional
Magnetic-Resonance-Imaging. Science, 268(5212), 889–893.
Sirosh, J., & Miikkulainen, R. (1997). Topographic receptive fields and patterned
lateral interaction in a self-organizing model of the primary visual cortex. Neural
Computation, 9(3), 577–594.
Sit, Y. F., & Miikkulainen, R. (2006). Self-organization of hierarchical visual maps
with feedback connections. Neurocomputing, 69(10–12), 1309–1312.
Stringer, S. M., Rolls, E. T., & Trappenberg, T. P. (2004). Self-organising continuous
attractor networks with multiple activity packets, and the representation of
space. Neural Networks, 17(1), 5–27.
Sutton, R. S., & Barto, A. G. (1998). Reinforcement learning : an introduction.
Cambridge, Mass.: MIT Press.
Szentagothai, J. (1975). Module-concept in cerebral-cortex architecture. Brain
Research, 95(2-3), 475–496.
Tanaka, K., Saito, H., Fukada, Y., & Moriya, M. (1991). Coding visual images of
objects in the inferotemporal cortex of the macaque monkey. Journal of Neuro-
physiology, 66(1), 170–189.
Tereshko, V., & Allinson, N. M. (2002). Combining lateral and elastic interactions:
Topology-preserving elastic nets. Neural Processing Letters, 15(3), 213–223.
Tootell, R. B. H., Switkes, E., Silverman, M. S., & Hamilton, S. L. (1988). Functional-
Anatomy of Macaque Striate Cortex .2. Retinotopic Organization. Journal of
Neuroscience, 8(5), 1531–1568.
Treves, A., & Rolls, E. T. (1991). What determines the capacity of autoassociative
memories in the brain. Network-Computation in Neural Systems, 2(4), 371–397.
Tuckova, J., & Zetocha, P. (2006). Speech analysis of children with developmental
dysphasia by Supervised SOM. Neural Network World, 16(6), 533–545.
Tusa, R. J., Palmer, L. A., & Rosenquist, A. C. (1978). Retinotopic organization of
area-17 (striate cortex) in cat. Journal of Comparative Neurology, 177(2),
213–235.
Van Essen, D. C., & Gallant, J. L. (1994). Review: Neural Mechanisms of Form and
Motion Processing in the Primate Visual-System. Neuron, 13(1), 1–10.
Vinson, D. P., & Vigliocco, G. (2002). A semantic analysis of grammatical class
impairments: semantic representations of object nouns, action nouns and action
verbs. Journal of Neurolinguistics, 15(3–5), 317–351.
Xie, X. P., Liaw, J. S., Baudry, M., & Berger, T. W. (1997). Novel expression
mechanism for synaptic potentiation: Alignment of presynaptic release site and
postsynaptic receptor. Proceedings of the National Academy of Sciences of the
Zador, A., Koch, C., & Brown, T. H. (1990). Biophysical model of a Hebbian synapse.
Proc Natl Acad Sci U S A, 87(17), 6718–6722.
Zipser, D., Kehoe, B., Littlewort, G., & Fuster, J. (1993). A spiking network model of
short-term active memory. Journal of Neuroscience, 13(8), 3406–3420.
Human Learning 101
Cognitive Neuroscience of Skill

Acquisition
J. Bo, J. Langan, and R. D. Seidler
1. Introduction
In this chapter, we provide an overview of contemporary research on the

cognitive neuroscience of human motor skill learning. Other chapters
found in this volume focus on animal models (Part IV) and current beha-
vioral issues (Part III) of skill learning. As such, these topics are not
addressed extensively here. Rather, they are touched upon to draw links
across research paradigms, demonstrate parallels in theoretical questions
addressed with differing techniques, and point out gaps in integration of
the existing literature.
Current cognitive neuroscience techniques amenable to the study of
skill learning include functional magnetic resonance imaging (fMRI), posi-
tron emission tomography (PET), electroencepholography/magnetoence-
pholography (EEG/MEG), and transcranial magnetic stimulation (TMS).
These techniques provide differing relative advantages in terms of spatial
and temporal resolution, and whether they allow for correlative vs. causa-
tive inferences. These issues have been covered extensively elsewhere
(Walsh & Cowey, 2000) and will not be addressed here.
Although the study of behavioral aspects of skill learning has a rich and long
history dating back over 100 years (see James, 1890), determining the under-
lying neural bases of these processes has necessarily lagged behind. Despite
several hundred cognitive neuroscience investigations of skill learning that
have been performed, many of the issues addressed in the behavioral chapters
on skill acquisition in this volume have yet to be investigated with this
approach. This is due in part to the relative infancy of the field (cf. Grafton
et al. 1992; Seitz & Roland 1992), the physical restrictions associated with most
cognitive neuroscience techniques, and the controls required to separate
neural activation associated with performance from that associated with
102 J. Bo, J. Langan, and R. D. Seidler
learning. Parametric manipulation of variables such as movement rate, force,

or motor error has revealed corresponding activation changes in the sensory
and motor cortical and subcortical regions (Ashe, 1997; Dai, Liu, Sahgal,
Brown, & Yue, 2001; Deiber, Honda, Ibanez, Sadato, & Hallett, 1999; Dettmers
et al., 1995, 1996; Jancke et al., 1998; Kitazawa, Kimura, & Yin, 1998; Mattay &
Weinberger, 1999; Muley et al., 2001; Sadato et al., 1997; Schlaug et al., 1996;
Turner, Grafton, Votaw, Delong, &, Hoffman, 1998). Since these same variables
change with skill acquisition, it is difficult to disentangle whether changes in
brain activation are due to differing performance levels that occur with prac-
tice, or rather reflect true contributions to learning.
Skill acquisition (used interchangeably with the term ‘‘motor learning’’ in
this chapter) has been defined as ‘‘. . .a set of processes associated with
practice or experience leading to relatively permanent changes in the
capability for responding’’ (Schmidt, 1988). Researchers studying skill
acquisition have classified learning into at least two broad categories,
including sensorimotor adaptation and sequence learning (Doyon & Benali,
2005; Willingham 1998). In sensorimotor adaptation paradigms, partici-
pants modify movements to adjust to changes in either sensory input or
motor output characteristics. For sequence learning, individuals learn to
combine isolated movements into one smooth, coherent action. The last
decade and a half has seen an explosion of publications using cognitive
neuroscience techniques to study skill acquisition. We have learned much
about the neural underpinnings of the two types of skill acquisition, and
how their contributions are altered for different stages of learning. As
outlined in this chapter, however, there are still some remaining contro-
versies, and much left to be learned regarding how the underlying circuitry
maps onto the many processes of skill learning. For the sake of space
limitations, we have elected to review some of the major current theories
on the cognitive neuroscience of skill learning, and to draw attention to the
overlap and differences among these theories. Although this review is not
meant to be exhaustive, we hope to give the reader a view of the current
status of the field, and the underlying questions that remain to be
addressed.
2. Cognitive neuroscience models of sensorimotor adaptation
Sensorimotor adaptation tasks are used to gain insight into how humans
represent their environment, the mechanics of the body, and interactions
between the two during movement planning and production. These tasks
can be described as either (1) dynamic (or kinetic) paradigms, which alter
anticipated proprioception by having participants move the limb through
an opposing force field (Shadmehr and Holcomb, 1997, 1999; Shadmehr and
Mussa-Ivaldi, 1994), or (2) kinematic paradigms, which disrupt visual feed-

back of movements through the use of displacing wedge prisms (Bossom &
Held, 1959; Redding, Rossetti, & Wallace, 2005; von Helmholtz, 1909/1962;)
or altered visual feedback on a computer screen (Bernotat, 1970; Bock,
1992; Cunningham, 1989; Pine, Krakauer, Gordon, & Ghez, 1996). It was
originally thought that dynamic and kinematic adaptation relied on differ-
ent neural substrates, due to a lack of interference between the two types
of adaptation during learning and consolidation (Krakauer, Ghilardi, &
Ghez, 1999). However, subsequent work demonstrated interference
between the two types of learning when the distortions were dependent
on the same kinematic variable (e.g., position-dependent displacements;
Tong, Wolpert, & Flanagan, 2002), suggesting that there is overlap. Indeed,
error processing of movements made under both dynamic and kinematic
distortions shows extensive neural overlap (Diedrichsen et al., 2005).
One influential model of the neural bases of skill learning is Willingham’s
(1998) control-based learning theory (COBALT). It proposes that learning
can occur via tuning of the processes directly involved in the control of
movement, or through the use of conscious, strategic processes. This model
suggests that a ventral cortical system is engaged for explicit learning, where
the participant is aware of the task and the goal to learn it. In this system,
task goals are transferred from the prefrontal cortex to the posterior tem-
poral lobe. The dorsal cortical learning system involves parietal and premo-
tor areas and operates during implicit learning, where learning occurs
outside of conscious awareness. COBALT proposes that the cerebellum
does not play a direct role in skill learning, regardless of whether sequence
learning or sensorimotor adaptation is taking place. An early PET study of
prism adaptation supports this idea (Clower et al., 1996). These authors
found that only the posterior parietal cortex was associated with adaptation,
after accounting for correction of motor errors, a process that is known to
engage the cerebellar circuitry (Kitazawa et al., 1998).
In contrast, some argue that internal models are updated during sensor-
imotor adaptation, and the cerebellum plays a central role in this process
(Ito, 2002; Miall, Weir, Wolpert, & Stein, 1993; Miall & Wolpert, 1996). An
internal model is thought to use efference copy to predict the sensory
consequences of a motor command (Miall & Wolpert, 1996). These models
are updated via error feedback during sensorimotor adaptation (Shadmehr &
Mussa-Ivaldi, 1994). Work by Imamizu and colleagues (Imamizu, Kuroda,
Miyauchi, Yoshioka, & Kawato, 2003; Imamizu, Kuroda, Yoshioka, &
Kawato, 2004; Imamizu et al., 2000) has consistently shown activation in
the cerebellar regions surrounding the posterior superior fissure during
adaptation of movements to differing visual distortions, even after correc-
tion for performance differences occurring across the time course of
learning.
Doyon and colleagues (Doyon and Benali, 2005; Doyon, Penhune, &
Ungerleider, 2003) have recently proposed a theoretical framework to
describe the dynamic cerebral changes that occur during different phases
of learning, including a fast early learning stage, a slow later stage, con-
solidation, automaticity, and retention. The model suggests that in the fast
learning phase, both the corticostriatal and the corticocerebellar systems
can be recruited depending on the cognitive processes that are required in
the task. Once the skill becomes well learned (slow, later stage of learning),
however, the two systems dissociate; the corticocerebellar system under-
lies sensorimotor adaptation and the corticostriatal system uniquely con-
tributes to the formation of motor sequence representations. This view of
sustained cerebellar contributions to sensorimotor adaptation is consistent
with the internal model view. Whether and how the corticostriatal circuitry
contributes to sensorimotor adaptation remains less clear.
Krakauer et al. (2004) used PET to investigate adaptation to an altered
gain of display of movements or a rotation of the visual display. Consistent
with the notion that direction and extent of movements are planned inde-
pendently (Gordon, Ghilardi, & Ghez, 1994), they found differing patterns
of brain activation associated with the two types of adaptation. The puta-
men was bilaterally activated during gain adaptation, but no basal ganglia
activity was observed during rotation adaptation. In contrast, a recent fMRI
study from our laboratory (Seidler, Noll, & Chintalapati, 2006) found
activation in the right globus pallidus and putamen, and the left globus
pallidus and caudate nucleus, during the early phases of adaptation to a
rotation perturbation. These latter findings are consistent with Doyon’s
model of adaptation (Doyon and Benali, 2005; Doyon et al., 2003) and a
recent computational model of adaptive processes (Grosse-Wentrup &
Contreras-Vidal, 2007).
It is possible that the corticostriatal and corticocerebellar systems inter-
act with each other during the fast, early phase of sensorimotor adaptation
(Doyon & Benali, 2005). Recent anatomical data show connectivity
between the cerebellar dentate nucleus and the striatum (Hoshi, Tremblay,
Feger, Carras, & Strick, 2005). Additionally, the finding that performance
on a sequence learning task is enhanced by prior experience with a motor
adaptation task (Seidler, 2004) suggests sharing of information between the
two systems.
It is possible that early activation of the corticostriatal circuitry during
adaptation reflects the contribution of this system to online error correc-
tions, or movement adjustments that are made within a trial. Such within-
trial corrections do not contribute to the adaptive process; rather,
corrections made from one trial to the next reflect learning or updating
of the motor representation. It has ben shown that patients with Hunting-
ton’s disease, a basal ganglia pathology, are unable to perform within-trial
adjustments for motor errors, but do adapt their performance across trials
(Smith, Brandt, & Shadmehr, 2000; Smith & Shadmehr, 2005). In contrast,
patients with cerebellar damage show the complementary deficit; that is,
they can make online motor adjustments, but they do not show adaptive
performance across trials (Maschke, Gomez, Ebner, & Konczak, 2004;
Morton & Bastian, 2006; Smith & Shadmehr, 2005). Since both within- and
across-trial corrections are performed early in learning, basal ganglia acti-
vation at this point may reflect online adjustments to motor performance.
An alternative hypothesis is that basal ganglia involvement may reflect the
engagement of more cognitive processes such as attention and working
memory (Seidler et al., 2006). This seems plausible, given that the activa-
tion is bilateral (as opposed to contralateral to the moving hand) and
encompasses the caudate nucleus.
In conclusion, the literature shows clear involvement of the cerebellum
and posterior parietal cortex during adaptive sensorimotor processes.
These areas contribute to adaptation through detection and correction of
motor errors, as well as storage of newly acquired internal models. In
contrast, it remains an open question whether engagement of the corticos-
triatal system during adaptation reflects performance (online error correc-
tions) or instead actively contributes to the learning process.
3. Cognitive neuroscience models of sequence learning
Several models have been proposed over the last decade regarding the
neurocognitive bases of sequence learning, taking into account both expli-
cit and implicit processes. For example, COBALT (Willingham, 1998),
described above, suggests that a ventral cortical system is engaged for
explicit learning, in which task goals are transferred from the prefrontal
cortex to the posterior temporal lobe, where target locations are repre-
sented in allocentric space. The dorsal cortical learning system involves
the parietal and premotor areas, operates in the implicit mode, and repre-
sents targets in body-centered space.
Keele, Ivry, Mayr, Hazeltine, and Heuer (2003) have also distinguished
between dorsal and ventral cortical contributions to sequence learning.
These authors suggest that the dorsal system contributes to implicit
sequence learning and is responsible for forming associations among sti-
muli of the same type. In contrast to COBALT, these authors argue that the
ventral system can operate under either an implicit or an explicit learning
mode. An important component of the theory is that the ventral system is
also responsible for forming associations among stimuli across multiple
dimensions, and thus is the source of both cross-task integration and
interference under dual tasking conditions.
Two additional models of sequence learning that have focused on inter-

actions between cortical and subcortical brain regions include those pro-
posed by Doyon and Benali (2005) and Hikosaka, Nakamura, Sakai, and
Nakahara (2002). Doyon’s theoretical framework (Doyon and Benali, 2005;
Doyon et al., 2003) was described above in the section on sensorimotor
adaptation. He suggests that in the early, fast phase of learning, both the
corticostriatal and the corticocerebellar systems can be recruited depend-
ing on the cognitive processes that are required in the task. Once the skill
becomes well learned, however, the two systems dissociate, and the corti-
costriatal system becomes specialized for the formation of motor sequence
representations.
Hikosaka et al. (2002) also focus on corticostriatal and corticocerebellar
circuitry. They propose that motor skill learning occurs independently and
in different coordinates between these two systems. They suggest that
sequences are learned in spatial coordinates in a system involving the
frontoparietal cortices and the associative regions of the basal ganglia
and cerebellum. Simultaneous learning takes place in motor coordinates
in the motor cortical areas, in conjunction with the motor regions of the
basal ganglia and cerebellum. Spatial sequences are acquired explicitly and
quickly, whereas motor sequence representations are usually processed
implicitly and slowly. Thus, for the trial and error sequence learning task
employed by this group (Hikosaka et al., 1999, 2002), early learning activa-
tion is typically seen in the lateral prefrontal cortex and the preSMA. Once
the sequence has been represented in motor coordinates, the spatial learn-
ing mechanism can be engaged to acquire other sequences. In contrast to
Doyon’s theory, this model proposes that both the corticostriatal and the
corticocerebellar circuitry are weighted equally at different stages of learn-
ing. They propose that the distinction between these two systems is not in
the time course of their contributions, but rather in the type of error
information that they encode. Neurons in the basal ganglia encode reward
expectation error and/or novelty, whereas the cerebellar climbing fibers
encode sensorimotor error signals, possibly including timing errors. Both
research groups have shown a shift in sequence representation from the
associative regions of the basal ganglia in early learning to the sensorimo-
tor regions later in learning (Hikosaka et al., 2002; Lehericy et al., 2005).
A newer model of sequence learning (Ashe, Lungu, Basford, & Lu, 2006)
suggests that instead of separable systems for explicit and implicit
sequence learning, there are overlapping neural networks that contribute
to both processes. For instance, when the intention to learn a sequence is
explicit, the processes originate in the prefrontal cortical areas, and then
later in learning, transfer occurs to the premotor and motor cortical
regions. In contrast, when sequences are acquired implicitly, learning
begins in the motor cortical areas, then propagates to premotor regions
and eventually the prefrontal cortex. Further, they suggest that implicit
and explicit processes interact with each other (as has been shown by
Destrebecqz et al., 2005; Willingham, Salidis, & Gabrieli, 2002), with their
relative importance varying depending on the stage of learning.
One remaining controversy regarding our understanding of the neural
bases of sequence learning is the nature of the contribution of prefrontal
cortex to implicit as well as explicit sequence learning. Although COBALT
(Willingham, 1998) proposes that the dorsolateral prefrontal cortex
(DLPFC) only contributes to sequence learning under explicit conditions,
both Keele et al.’s (2003) and Ashe et al.’s (2006) models suggest that
DLPFC can play a role in pure implicit sequence learning. The latter point
is supported by several examples in the literature (Aizenstein et al., 2004;
Pascual-Leone, Wassermann, Grafman, & Hallett, 1996; Robertson, Tormos,
Maeda, & Pascual-Leone, 2001). A TMS study by Robertson et al. (2001)
suggests that the role of the DLPFC is limited to conditions in which stimuli
are spatially presented. The difficulty in measuring explicit awareness
makes interpretation of these DLPFC findings difficult.
An additional contentious issue is whether the cerebellum participates
in such learning. Although studies of sensorimotor adaptation have con-
sistently implicated a role for the cerebellum (with the exception of Clower
et al., 1996), the literature on sequence learning is less consistent. Though
many cognitive neuroscience investigations have implicated a role for the
cerebellum in sequence learning (Aizenstein et al., 2004; Gomez-Beldarrain,
Grafman, Pascual-Leone, & Garcia-Monco, 1999; Pascual-Leone et al., 1993;
Schendan, Searl, Melrose, & Stern, 2003; Wu, Kansaku, & Hallett, 2004), a
study that dissociated performance effects from the learning process found
prominent activation in the cerebellum during the expression of learning
but not during the learning process per se (Seidler et al., 2002). This is
consistent with predictions of the COBALT model (Willingham, 1998).
Recent data from Boyd and Winstein (2004) have also demonstrated that
cerebellar patients have intact learning of the spatial features of a tracking
task, but not its temporal features. Thus, sequence learning tasks with a
temporal component (such as utilized in Penhune & Doyon, 2005) may
elicit cerebellar activation due to the role that this structure plays in motor
timing (Ivry & Spencer, 2004; Spencer, Zelaznik, Diedrichsen, & Ivry, 2003),
whereas spatial sequence learning may occur in the absence of cerebellar
activity.
It has long been known, from both human and nonhuman animal
research, that the striatal system contributes to sequence learning. The
ongoing topics of debate include whether and/or how cerebellar pathways
contribute to this type of learning, and whether engagement of the DLPFC
reflects only explicit processes or temporal and spatial processes under
implicit conditions as well.
4. Conclusions
Advances in technology have opened a new window onto the neural

mechanisms of motor learning. Determining the specific functions of the
engaged circuitry remains an ongoing, and hotly debated, process. Such
progress is important, as it enhances our understanding of brain–behavior
relationships and can provide insight into the mechanisms and treatments
of movement disorders.
References
Aizenstein, H. J., Stenger, V. A., Cochran, J., Clark, K., Johnson, M., & Nebes, R. D.,
et al. (2004). Regional brain activation during concurrent implicit and explicit
sequence learning. Cerebral Cortex, 14, 199–208.
Ashe, J. (1997). Force and the motor cortex. Behavioural Brain Research 87, 255–269.
Ashe, J., Lungu, O. V., Basford, A. T., & Lu, X. (2006). Cortical control of motor
sequences. Current Opinion in Neurobiology, 16, 213–221.
Bernotat, R. K. (1970). Rotation of visual reference systems and its influence on
control quality. IEEE Transactions on Man-Machine Systems, MM11, 129–131.
Bock, O. (1992). Adaptation of aimed arm movements to sensorimotor discor-
dance—evidence for direction-independent gain-control. Behavioural Brain
Research, 51, 41–50.
Bossom, J., & Held, R. (1959). Transfer of error-correction in adaptation to prisms.
American Psychologist, 14, 436.
Boyd, L. A., & Winstein, C. J. (2004). Cerebellar stroke impairs temporal but not
spatial accuracy during implicit motor learning. Neurorehabilitation and Neural
Repair, 18, 134–143.
Clower, D. M., Hoffman, J. M., Votaw, J. R., Faber, T. L., Woods, R. P., & Alexander,
G. E. (1996). Role of posterior parietal cortex in the recalibration of visually
guided reaching. Nature, 383, 618–621.
Cunningham, H. A. (1989). Aiming error under transformed spatial mappings sug-
gests a structure for visual motor maps. Journal of Experimental Psychology:
Human. perception and Performance, 15, 493–506.
Dai, T. H., Liu, J. Z., Sahgal, V., Brown, R. W., & Yue, G. H. (2001). Relationship
between muscle output and functional MRI-measured brain activation. Experi-
mental Brain Research 140, 290–300.
Deiber, M. P., Honda, M., Ibanez, V., Sadato, N., & Hallett, M. (1999). Mesial motor
areas in self-initiated versus externally triggered movements examined with
fMRI: Effect of movement type and rate. Journal of Neurophysiology, 81,
3065–3077.
Destrebecqz, A., Peigneux, P., Laureys, S., Degueldre, C., Del Fiore, G., & Aerts, J.,
et al. (2005). The neural correlates of implicit and explicit sequence learning:
Interacting networks revealed by the process dissociation procedure. Learning &
Memory, 12, 480–490.
Dettmers, C., Connelly, A., Stephan, K. M., Turner, R., Friston, K. J., Frackowiak, R.
S. J., et al. (1996). Quantitative comparison of functional magnetic resonance
imaging with positron emission tomography using a force-related paradigm.
NeuroImage, 4, 201–209.
Dettmers, C., Fink, G. R., Lemon, R. N., Stephan, K. M., Passingham, R. E., Silbers-
weig, D., et al. (1995). Relation between cerebral activity and force in the motor
areas of the human brain. Journal of Neurophysiology, 74, 802–815.
Diedrichsen, J., Hashambhoy, Y., Rane, T., & Shadmehr, R. (2005). Neural correlates
of reach errors. Journal of Neuroscience, 25, 9919–9931.
Doyon, J., & Benali, H. (2005). Reorganization and plasticity in the adult brain
during learning of motor skills. Current Opinion in Neurobiology, 15, 161–167.
Doyon, J., Penhune, V., & Ungerleider, L. G. (2003). Distinct contribution of the
cortico-striatal and cortico-cerebellar systems to motor skill learning. Neuropsy-
chologia, 41, 252–262.
Gomez-Beldarrain, M., Grafman, J., Pascual-Leone, A., & Garcia-Monco, J. C. (1999).
Procedural learning is impaired in patients with prefrontal lesions. Neurology, 52,
1853–1860.
Gordon, J., Ghilardi M. F., & Ghez, C. (1994). Accuracy of planar reaching move-
ments. I. Independence of direction and extent variability. Experimental Brain
Research, 99, 97–111.
Grafton, S. T., Mazziotta, J. C., Presty, S., Friston, K. J., Frackowiak, R. S. & Phelps,
M. E. (1992). Functional anatomy of human procedural learning determined with
regional cerebral blood flow and PET. Journal of Neuroscience, 12, 2542–2548.
Grosse-Wentrup, M., & Contreras-Vidal, J. L. (2007). The role of the striatum in
adaptation learning: a computational model. Biological Cybernetics, 96, 377–388.
Hikosaka, O., Nakahara, H., Rand, M. K., Sakai, K., Lu, X., Nakamura, K. et al. (1999).
Parallel neural networks for learning sequential procedures. Trends in Neuro-
science, 22, 464–471.
Hikosaka, O., Nakamura, K., Sakai, K., & Nakahara, H. (2002). Central mechanisms
of motor skill learning. Current Opinion in Neurobiology, 12, 217–222.
Hoshi, E., Tremblay, L., Feger, J., Carras, P. L., & Strick, P. L. (2005). The cerebellum
communicates with the basal ganglia. Nature Neuroscience, 8, 1491–1493.
Imamizu, H., Miyauchi, S., Tamada, T., Sasaki, Y., Takino, R., Putz, B., et al. (2000).
Human cerebellar activity reflecting an acquired internal model of a new tool.
Nature, 403, 192–195.
Imamizu, H., Kuroda, T., Miyauchi, S., Yoshioka, T., & Kawato, M. (2003). Modular
organization of internal models of tools in the human cerebellum. Proceedings
of the National Academy of Sciences of the United States of America, 100,
5461–5466.
Imamizu, H., Kuroda, T., Yoshioka, T., & Kawato, M. (2004). Functional magnetic
resonance imaging examination of two modular architectures for switching mul-
tiple internal models. Journal of Neuroscience, 24, 1173–1181.
Ito, M. (2002). Historical review of the significance of the cerebellum and the role of
purkinje cells in motor learning. Annals of NY Acad of Science, 978, 273–288.
Ivry, R. B. & Spencer, R. M. C. (2004). The neural representation of time. Current
Opinion in Neurobiology, 14, 225–232.
James, W. (1890). The principles of psychology (vol. 1) NY: Holt.

Jancke, L., Specht, K., Mirzazade, S., Loose, R., Himmelbach, M., Lutz, K., et al.
(1998). A parametric analysis of the rate effect in the sensorimotor cortex: A
functional magnetic resonance imaging analysis in human subjects. Neuroscience
Letters, 252, 37–40.
Keele, S. W., Ivry, R., Mayr, U., Hazeltine, E., & Heuer, H. (2003). The cognitive
and neural architecture of sequence representation. Psychological Review, 110,
316–339.
Kitazawa, S., Kimura, T., & Yin, P. B. (1998). Cerebellar complex spikes encode both
destinations and errors in arm movements. Nature, 392, 494–497.
Krakauer, J. W., Ghilardi, M. F., & Ghez, C. (1999). Independent learning of internal
models for kinematic and dynamic control of reaching. Nature Neuroscience, 2,
1026–1031.
Krakauer, J. W., Ghilardi, M. F., Mentis, M., Barnes, A., Veytsman, M., Eidelberg, D.,
et al. (2004). Differential cortical and subcortical activations in learning rotations
and gains for reaching: A PET study. Journal of. Neurophysiology, 91, 924–933.
Lehericy, S., Benali, H., Van de Moortele, P. F., Pelegrini-Issac, M., Waechter, T.,
Ugurbil, K. et al. (2005). Distinct basal ganglia territories are engaged in early and
advanced motor sequence learning. Proceedings of the National Academy of
Sciences of the United States of America, 102, 12566–12571.
Maschke, M., Gomez, C. M., Ebner, T. J., & Konczak, J. (2004). Hereditary cerebellar
ataxia progressively impairs force adaptation during goal-directed arm move-
ments. J Neurophysiol, 91, 230–238.
Mattay, V.S. & Weinberger, D.R. (1999). Organization of the human motor system as
studied by functional magnetic resonance imaging. European Journal of Radio-
logy, 30, 105–114.
Miall, R. C., Weir, D. J., Wolpert, D. M., & Stein, J. F. (1993). Is the cerebellum a
smith predictor. Journal of Motor Behavior, 25, 203–216.
Miall, R. C. & Wolpert, D. M. (1996). Forward models for physiological motor
control. Neural Networks, 9, 1265–1279.
Morton, S. M., & Bastian, A. J. (2006). Cerebellar contributions to locomotor
adaptations during splitbelt treadmill walking. Journal of Neuroscience, 26,
9107–9116.
Muley, S. A., Strother, S. C., Ashe, J., Frutiger, S. A., anderson, J. R., Sidtis, J. J., et al.
(2001). Effects of changes in experimental design on PET studies of isometric
force. NeuroImage, 13, 185–195.
Pascual-Leone, A., Grafman, J., Clark, K., Stewart, M., Massaquoi, S., Lou, J. S. et al.
(1993). Procedural learning in parkinson’s disease and cerebellar degeneration.
Annals of Neurology, 34, 594–602.
Pascual-Leone, A., Wassermann, E. M., Grafman, J., & Hallett, M. (1996). The role of
the dorsolateral prefrontal cortex in implicit procedural learning. Experimental
Brain Research, 107, 479–485.
Penhune, V. B. & Doyon, J. (2005). Cerebellum and M1 interaction during early
learning of timed motor sequences.Neuroimage, 26, 801–812.
Pine, Z. M., Krakauer, J. W., Gordon, J., & Ghez, C. (1996). Learning of scaling
factors and reference axes for reaching movements. Neuroreport, 7, 2357–2361.
Redding, G. M., Rossetti, Y., & Wallace, B. (2005). Applications of prism adaptation:
A tutorial in theory and method. Neuroscience and Biobehavioral Reviews, 29,
431–444.
Robertson, E. M., Tormos, J. M., Maeda, F., & Pascual-Leone, A. (2001). The role of
the dorsolateral prefrontal cortex during sequence learning is specific for spatial
information. Cerebral Cortex, 11, 628–635.
Sadato, N., Ibanez, V., Campbell, G., Deiber, M.P., Le Bihan, D., & Hallett, M. (1997).
Frequency-dependent changes of regional cerebral blood flow during finger
movements: Functional MRI compared to PET. Journal of Cerebral Blood Flow
and Metabolism, 17, 670–679.
Schendan, H. E., Searl, M. M., Melrose, R. J., & Stern, C. E. (2003). An fMRI study of
the role of the medial temporal lobe in implicit and explicit sequence learning.
Neuron, 37, 1013–1025.
Schlaug, G., Sanes, J. N., Thangaraj, V., Darby, D. G., Jancke, L., Edelman, R. R.,
et al. (1996). Cerebral activation covaries with movement rate. Neuroreport, 7,
879–883.
Schmidt, R. A. (1988). Motor Control and Learning: A Behavioral Emphasis (2nd
ed.) Champaign, IL: Human Kinetics.
Seidler, R. D. (2004). Multiple motor learning experiences enhance motor adapt-
ability. Journal of Cognitive Neuroscience, 16, 65–73.
Seidler, R. D., Noll, D. C., & Chintalapati, P. (2006). Bilateral basal ganglia activation
associated with sensorimotor adaptation. Experimental Brain Research, 175,
544–555.
Seidler, R. D., Purushotham, A., Kim, S. G., Ugurbil, K., Willingham, D., & Ashe, J.
(2002). Cerebellum activation associated with performance change but not motor
learning.Science, 296, 2043–2046.
Seitz, R. J. & Roland P. E. (1992). Learning of sequential finger movements in man:
A combined kinematic and positron emission tomography (PET) study. The
European Journal of Neuroscience, 4, 154–165.
Shadmehr, R. & Holcomb, H. H. (1997). Neural correlates of motor memory
consolidation. Science, 277, 821–825.
Shadmehr, R., & Holcomb, H. H. (1999). Inhibitory control of competing motor
memories. Experimental Brain Research, 126, 235–251.
Shadmehr, R., & Mussa-Ivaldi, F. A. (1994). Adaptive representation of dynamics
during learning of a motor task. Journal of Neuroscience, 14, 3208–3224.
Smith, M. A., Brandt, J., & Shadmehr, R. (2000). Motor disorder in huntington’s
disease begins as a dysfunction in error feedback control. Nature, 403, 544–549.
Smith, M. A. & Shadmehr, R. (2005). Intact ability to learn internal models of arm
dynamics in Huntington’s disease but not cerebellar degeneration. Journal of
Neurophysiology, 93, 2809–2821.
Spencer, R. M. C., Zelaznik, H. N., Diedrichsen, J., & Ivry, R. B.,. (2003). Disrupted
timing of discontinuous but not continuous movements by cerebellar lesions.
Science, 300, 1437–1439.
Tong, C., Wolpert, D. M., & Flanagan, J. R. (2002). Kinematics and dynamics are not
represented independently in motor working memory: Evidence from an inter-
ference study. Journal of Neuroscience, 22(3), 1108–1113.
Turner, R.S., Grafton, S.T., Votaw, J.R., Delong, M.R., & Hoffman, J.M. (1998). Motor
subcircuits mediating the control of movement velocity: A PET study. Journal of.
von Helmholtz, H. (1909/ 1962). A Treatis on Physiological Optics (T. J. P. C
Southall, Trans.). New York: Dover.
Walsh, V. & Cowey A. (2000). Transcranial magnetic stimulation and cognitive
neuroscience. Nature Reviews Neuroscience, 1, 73–79.
Willingham, D. B. (1998). A neuropsychological theory of motor skill learning.
Willingham, D. B., Salidis, J., & Gabrieli, J. D. E. (2002). Direct comparison of neural
systems mediating conscious and unconscious skill learning. Journal of Neuro-
physiology, 88, 1451–1460.
Wu, T., Kansaku, K., & Hallett, M. (2004). How self-initiated memorized movements
become automatic: A functional MRI study. Journal of Neurophysiology, 91,
1690–1698.
Human Learning 113
Cognitive Neuroscience of
Declarative and Nondeclarative
Memory
Paul J. Reber
Northwestern University
The study of memory within cognitive neuroscience is an attempt to

synthesize an account of both the fundamental mnemonic component pro-
cesses and the neural basis of these processes. This approach generally takes
a very broad definition of memory to include any changes within the brain
that reflect long-term storage of new information. This broad definition not
only encompasses traditional definitions of memory, the conscious acquisi-
tion and recollection of facts and events, but also brings a range of additional
phenomena into the memory domain, for example, skill learning, perceptual
learning, eyeblink conditioning, and even plasticity in the aplysia gill with-
drawal reflex (e.g., Milner, Squire, & Kandel 1998). All of these types of
memory depend on changes (plasticity) within identifiable brain regions and
networks. From this perspective, the fundamental questions of memory are to
identify the mechanisms of plasticity, the representations of stored informa-
tion, and the methods by which these changes influence subsequent behavior.
The most well-studied and complex memory system of the brain is based
on the medial temporal lobe (MTL), a neural system that contains the
hippocampus and surrounding cortical areas. It is this system that, when
lesioned in the patient H.M., led to profound selective loss of the ability to
acquire new memories for facts and events (Scoville & Milner, 1957). This
phenomenon of anterograde amnesia resulting from MTL damage has been
extensively studied since the first report of patient H.M. (see Squire, 1992
for a review). In addition, studies of these patients have identified a number
of memory phenomena that are not affected by damage to the MTL. To
provide a taxonomy for organizing the types of memory, the set of related
memory phenomena that depend crucially on the MTL memory system is
termed declarative memory while memory that does not depend on the
MTL is termed nondeclarative memory.
114 Paul J. Reber
1. Declarative memory
Declarative memory refers to the acquisition and retrieval of facts,

events, and episodes. A nearly synonymous term, explicit memory, empha-
sizes the fact that these types of memory are available to awareness, that is,
we consciously recall these memories during retrieval. The bulk of what is
thought of as the ordinary, everyday operation of memory depends on
declarative memory. The conventional use of the term memory generally
refers to the retrieval process (e.g., remembering or failing to remember),
but before memories can be retrieved, they have to be acquired (stored),
and this process depends heavily on the intact function of the MTL.
Patients with damage to the MTL exhibit anterograde amnesia, a loss of
ability to acquire new memories for facts and events. In many of these
patients, retrieval of remote memories is intact. If you converse with a
patient with anterograde amnesia, they will often tell stories of their youth,
which sometimes leads their family members to suggest that their ‘‘memory
seems fine.’’ However, if you meet them again the next day, they will
generally not recognize you and often will tell you again the same story
they told you the previous day. The preservation of older memories is not
absolute, but follows a temporal gradient such that recently acquired
memories (e.g., recently experienced before the event that caused MTL
damage and amnesia) are typically lost or impaired. As memories are
assessed further backwards in time, the number and quality of memories
increases in these patients. This phenomenon of temporally graded retro-
grade amnesia has been carefully studied in animal models of amnesia and
indicates that long-term memories undergo a consolidation process that
depends on the MTL (Milner, Squire, & Kandel, 1998). Damage to the MTL
disrupts this process and in addition to interfering with acquisition of new
memories, recently acquired memories are also lost.
The nature and neurobiology of the consolidation process is an area of
active research and debate. Theories of how the MTL accomplishes con-
solidation range from endogenous processes (e.g., occurring during sleep;
Ellenbogen, Payne, & Stickgold, 2006) to retrieval-triggered strengthening
to the creation of multiple parallel memory traces (Moscovitch et al.,
2005a). In humans, retrograde amnesia can extend years or decades
into the past, indicating that the consolidation process is very gradual
(Rempel-Clower, Zola, Squire, & Amaral, 1996). Because of this, theories
of consolidation become intertwined with questions about the nature of
long-term memories. Episodic memory of prior events is sometimes
described as mental time travel, an internal transportation to something
like a reexperiencing of a prior event. However, much of our long-term
declarative memory is semantic facts that are retrieved and used without
the experience of mental time travel. The multiple-trace theory model of
consolidation suggests that retrieval of episodic memories is always

supported by the MTL and that only retrieval of remote semantic memories
can be fully independent of the MTL (Moscovitch et al., 2005b). The
‘‘standard model’’ of consolidation is that both episodic and semantic
memories initially depend on the MTL and are gradually consolidated by
similar processes to become independent. Although these theories make
important distinctions about the representation of long-term memory, it is
difficult to distinguish between them experimentally. Some severely amne-
sic patients appear to have intact remote episodic memories (Bayley, Gold,
Hopkins, & Squire 2005; Bayley, Hopkins, & Squire, 2003), but it is impos-
sible to determine with certainty if they have the subjective experience of
mental time travel when the episodes are recalled.
This question of the neural basis of long-term memory representation
and retrieval is mirrored in experimental studies that have contrasted
remembering with a strong ‘‘feeling of knowing’’ and retrieval that is
based on familiarity. In this domain, two competing hypotheses about
memory traces have been proposed to describe the neural basis of memory
retrieval. The two-process model (Yonelinas, 2002) proposes that a feeling
of familiarity emerges from cortical regions of the MTL (but not the
hippocampus proper) as a memory trace is retrieved. However, a second
component, specifically dependent on the hippocampus is crucial to having
a strong feeling of knowing (and may also be related to the mental time
travel associated with episodic retrieval). An alternate trace strength model
(Wixted, 2007) suggests that activity increases monotonically with the
strength of the memory across the MTL during retrieval, and there is no
necessarily special role for the hippocampus.
Studies attempting to reconcile these competing hypotheses have often
looked at the contributions to memory function from the prefrontal cortex
(PFC). Although lesion studies show that the MTL is critical for memory
function, even the earliest functional neuroimaging studies found strongly
correlated activity in PFC during memory tasks (e.g., Squire et al., 1992).
The role of the PFC in memory is generally hypothesized to be via meta-
mnemonic processes that guide search, retrieval, and encoding strategies.
Regions within the PFC have been found to show increased activity during
source memory recollection (e.g., Buckner 2002; Rugg, Fletcher, Chua, &
Dolan, 1999), retrieval monitoring (e.g., Dobbins, Foley, Schacter &
Wagner, 2002), and encoding effort (Reber et al., 2002). Although the
specific cognitive functions associated with regions within the PFC are
not yet fully understood, it is generally hypothesized that the PFC mod-
ulates MTL function so that the MTL can support the different functions of
encoding, retrieval, and consolidation within a single neural system.
Although the MTL is frequently the focus of neuroscientific studies of
long-term memory, it is clear that it acts in conjunction with many other
116 Paul J. Reber
brain regions. The lateral temporal cortex is hypothesized to support the

representation of consolidated memories as they become independent of
the MTL. Recent neuroimaging studies have also regularly observed mem-
ory-related changes in activity in posterior parietal areas, suggesting that
these regions may play an important supporting role in memory function
(Wagner, Shannon, Kahn, & Buckner, 2005). A key idea is that memories
are represented as information represented across cortical regions that are
bound together through the MTL (Paller, 2002). This idea is supported by
the analysis of the computational problem posed by rapid acquisition of
complex memory traces, which suggests that this type of neural circuitry
would allow for one-trial learning without interference in prior knowledge
(McClelland, McNaughton, & O’Reilly, 1995). Although other areas support
the MTL, declarative memory is essentially defined by the dependence on
the MTL for memory formation. The memory processes that have been
found not to depend on the MTL and the associated neural circuits are
collectively termed nondeclarative memory.
2. Nondeclarative memory
Although the bulk of what is generally thought of as the everyday

operation of memory involves declarative memory, that is, conscious
recollection of facts and events, a number of tasks indicate the existence
of memory function outside the MTL. This type of memory typically oper-
ates outside awareness and is often referred to as implicit memory. The
terms nondeclarative and implicit memory are nearly (but not completely)
synonymous. Examples of this type of memory will be reviewed here from
a neuroanatomical perspective as nondeclarative memory. In different
paradigms, nondeclarative memory phenomena have been shown to
depend on a number of different brain regions, indicating that this type of
memory is not a single coherent system, but is rather a collection of
phenomena with different neural substrates. Nondeclarative memory
includes perceptual priming, conceptual priming, habit learning, skill learn-
ing, motor sequence learning, certain kinds of category learning, and some
forms of classical conditioning.
An early demonstration of the phenomenon of priming is seen in the
stem completion test. Participants read a list of words including the word
window. At test, no reference is made to the study list but three-letter
stems such as win__ are given and participants are asked to complete the
stem with the first word that comes to mind. The probability of completing
the stem with a word from the study list like window is much higher than if
the word had not been studied. This ‘‘popping to mind’’ phenomenon occurs
when participants are not directed to the study list and even on occasions
where participants are explicitly told not to complete the stem with a word
from the list (Jacoby, 1991). The critical neuroscientific observation is that
patients with memory dysfunction due to damage to the MTL exhibit
normal priming levels even when their declarative memory for the studied
words is remarkably impaired (Hamman & Squire, 1997; Stark & Squire,
2000). This dissociation indicates that the changes in the brain that
occurred during the study phase that leads words to pop to mind later
are not dependent on the normal functioning of the MTL. There must
therefore be neural plasticity elsewhere in the brain supporting changes
that give rise to the priming phenomenon.
Evidence for the location of the neural basis of changes supporting the
phenomenon of priming comes primarily from studies of functional neu-
roimaging. A second encounter with a repeated word (image or face)
typically evokes a smaller response in visual cortical areas associated
with sensory processing of the stimulus than a prior presentation (Schacter &
Buckner, 1998). This ‘‘repetition suppression’’ effect is thought to reflect
a change in the state of sensory cortex that reflects a nondeclarative
memory of the first presentation (Wiggs & Martin, 1998). Similar effects
have also been documented in PFC for semantic priming (e.g., Macotta &
Buckner, 2004). The changes that support the repetition suppression effect
are thought to be local, unavailable to awareness and, to occur independently
of the MTL.
Although nondeclarative memory is a heterogenous collection of phe-
nomena, each type of nondeclarative memory is hypothesized to operate by
similar principles; depending on local changes to a circumscribed brain
region, the representation of these changes is unavailable to awareness,
and the plasticity underlying the changes does not depend on the MTL.
Examples of nondeclarative memory following this form are several forms
of conditioning that have been well studied in experimental animals: delay
eyeblink conditioning (depending on the cerebellum), fear conditioning
(depending on the amygdala), and the gill withdrawal reflex of the aplysia
meets these criteria (Milner, Squire, & Kandel, 1998).
A number of more complex forms of nondeclarative memory have been
studied in humans, which are expressed as skill, habit, and category learn-
ing. Collectively, these studies demonstrate that the plasticity mechanisms
outside the MTL are capable of establishing representations that can be
complex and abstract even though they influence behavior without aware-
ness. Studies of more complex forms of nondeclarative memory have been
defined by a collection of specific tasks that are amenable to demonstra-
tions of learning without awareness. A consistent challenge of this type of
research is that when healthy subjects are able to consciously deduce the
structure behind the experimental paradigm, they may use declarative
memory to support performance. For this reason, tasks that have been
118 Paul J. Reber
effective at showing learning without awareness and preserved learning in

amnesic patients have generally become frequently investigated model
tasks.
A simple task that has been very well studied as a model of nondeclara-
tive memory is the serial reaction time (SRT) task. First reported by
Nissen and Bullemer (1987), the task requires learning a sequence of
motor responses to visual cues. Typically, four cue locations are shown
on a screen with four possible keypress (button) responses aligned under-
neath. Participants wait for the appearance of the cue and simply press the
button beneath the cue as quickly as possible. After cue offset, it reap-
pears in another location, the participant makes the corresponding
response and this continues for several hundred trials. Participants are
not told that the cue order follows a predictable structure, usually a
repeating sequence of 10–12 locations. Response times become much
more rapid with practice and knowledge of the repeating sequence is
assessed by removing the sequence and observing a slight slowdown in
average response time. Knowledge of the repeating sequence occurs even
when participants are unaware of the existence of the sequence, and
learning occurs at a normal rate in patients with memory disorders (Nissen &
Bullemer, 1987; Reber & Squire, 1994). Functional neuroimaging has
implicated the basal ganglia (particularly the putamen) as contributing to
perceptual motor sequence learning in this task (e.g., Destrebecqz et al.,
2005), which is consistent with the finding that patients with basal ganglia
dysfunction due to Parkinson’s disease are impaired at SRT learning (Siegert,
Taylor, Weatherall, & Abernethy, 2006). This form of nondeclarative memory
is hypothesized to depend on changes in corticostriatal circuits involved
in motor response planning.
The phenomenon of artificial grammar learning (AGL) was first reported
by A. Reber (1967). It involves learning sequences of arbitrary symbols.
Participants are shown nonsense strings of letters (e.g., PQXVT) during a
study phase in which they are asked to memorize, copy, or simply observe.
No mention is made of the fact that a complex set of rules is used to create
the letter sequences (the rules are generally represented as a finite state
machine, i.e., an ‘‘artificial grammar’’). After the study phase and a delay,
participants are told that the prior strings were constructed according to
rules and are then shown new strings and asked to judge which of the new
strings correctly follow the same rules. Participants typically report no
knowledge of the rules (but see Dulany, Carson, & Dewey, 1985) but are
able to make ‘‘grammaticality’’ judgments at above chance rates. Although
there has been controversy over the ability to assess a complete lack of rule
knowledge, amnesic patients have been shown to learn the AGL task at a
normal rate (Knowlton, Ramus, & Squire, 1992) as have patients with
Alzheimer’s disease (Reber, Martinez, & Weintraub, 2003). The neural
substrate of AGL is not yet known as different functional neuroimaging

studies have implicated differing regions: the posterior parietal association
cortex (Skosnik, Gitelman, Parrish, Mesulam, & Reber, 2002) or the basal
ganglia (Lieberman, Chang, Chiao, Bookheimer, & Knowlton, 2004; but see
Reber & Squire, 1999).
Various category learning tasks have also been shown to depend on
nondeclarative memory. The dot-pattern classification task described by
Posner and Keele (1968) appears to be learned normally by amnesic
patients (Knowlton & Squire, 1993) and is associated with changes in visual
cortical areas (Reber Gitelman, Parrish & Mesulam, 2003). A task known as
the weather prediction task, sometimes referred to as probabilistic classi-
fication, requires assigning groups of cues to two different categories and
appears to be at least initially learned normally by amnesic patients
(Knowlton, Squire & Gluck, 1994). Learning this task is impaired in patients
with Parkinson’s disease (Knowlton, Mangels & Squire, 1996) and has been
associated with increased activity in the basal ganglia (Poldrack et al.,
2001). Although questions have been raised about the degree to which
this task depends selectively on nondeclarative memory (e.g., after an
initial period of learning, patients do not exhibit learning at the same rate
as controls), a convergent line of evidence supporting the idea of a corti-
costriatal category learning system has emerged (e.g., Nomura et al., 2007).
The second line of evidence uses simple visual stimuli (sine wave gratings)
organized into categories by a decision bound in a two-dimensional stimu-
lus space (Ashby & Ell, 2001). Category structures that do not lend them-
selves to conscious deduction of a verbalizable rule appear to depend on
corticostriatal circuits connecting the posterior caudate to extrastriate
visual areas.
The examples here are not an exhaustive list of the types of memory
function that reflects nondeclarative memory. Skill such as reading mirror-
reversed text (Cohen & Squire 1980) and a number of different forms of
priming (Schacter & Buckner, 1998) have been studied in patients and with
functional neuroimaging. The wide variety of tasks and brain regions
implicated suggests that nondeclarative memory is not a memory system
per se, but rather reflects a general principle of inherent plasticity in neural
circuits that can support certain types of learning.
Understanding the principles behind nondeclarative memory may help
resolve the question of the memory processes that are involved in tasks
where the terms implicit and nondeclarative are not synonymous. For
example, Chun and Phelps (1999) reported a type of priming within visual
search that operates outside of awareness (implicit) but appears to depend
on the MTL (declarative). This type of memory may also be related to the
phenomenon of priming for new associations (Graf & Schacter, 1985), which
is also impaired in patients with MTL damage (Shimamura & Squire, 1989),
120 Paul J. Reber
but appears to operate outside awareness in healthy participants. These

tasks suggest that some principles of nondeclarative memory may operate
even on declarative memory representations. To accommodate these types
of interactions among memory types, cognitive neuroscience theories of
memories will eventually have to move beyond descriptions based solely
on multiple systems.
3. Conclusion
Memory for the ordinary facts, episodes, and events of our lives is
supported by the operation of a complex set of neural circuits depending
critically on the MTL. These circuits are crucial for the acquisition of new
memories, recognition and recall of recently acquired information, and a
consolidation process by which this information gradually becomes inde-
pendent of the MTL. Meta-memory processes are hypothesized to emerge
from interactions between the PFC via modulatory effects on the MTL. This
set of circuitry does not encompass all of the memory abilities of the brain.
A large number of mnemonic processes depend on brain regions operating
independently of the MTL. These memory phenomena typically operate
outside of awareness and are observed as skill, habit, and category learning
in addition to priming and certain kinds of simple conditioning. Some of
these phenomena depend on neural circuits such as corticostriatal loops,
while others depend on local changes within sensory cortex, the amygdale,
or the cerebellum.
References
Ashby, F. G., & Ell, S. W. (2001). The neurobiology of human category learning.
Trends in Cognitive Science, 5, 204–210.
Bayley, P. J., Gold, J. J., Hopkins, R. O., & Squire, L. R. (2005). The neuroanatomy of
remote memory. Neuron, 46, 799–810.
Bayley, P. J., Hopkins, R. O., & Squire, L. R. (2003). Successful recollection of
remote autobiographical memories by amnesic patients with medial temporal
lobe lesions. Neuron, 38, 135–144.
Buckner, R. L. (2002). Frontally mediated control processes contribute to source
memory retrieval. Neuron, 29, 529–535.
Cohen, N. J., & Squire, L. R. (1980). Preserved learning and retention of pattern-
analyzing skill in amnesia: Dissociation of knowing how and knowing that.
Science, 210, 207–210.
Chun, M. M., & Phelps, E. A. (1999). Memory deficits for implicit contextual
information in amnesic subjects with hippocampal damage. Nature
Destrebecqz, A., Peigneux, P., Laureys, S., Degueldre, C., Del Fiore, G., Aerts, J.,
Luxen, A., et al. (2005). The neural correlates of implicit and explicit sequence
learning: Interacting networks revealed by the process dissociation procedure.
Learning & Memory, 12, 480–490.
Dobbins, I. G., Foley, H., Schacter, D. L., & Wagner, A. D. (2002). Executive control
during retrieval: multiple prefrontal processes subserve source memory. Neuron,
35, 989–996.
Dulany, D. E., Carlson, R. A., & Dewey, G. I. (1985). A case of syntactical learning
and judgment: How conscious and how abstract. Journal of Experimental
Ellenbogen, J. M., Payne, J. D., & Stickgold, R. (2006). The role of sleep in declara-
tive memory consolidation: Passive, permissive, active or none? Current Opi-
nion in Neurobiology, 16, 716–722.
Graf, P., & Schacter, D. L. (1985). Implicit and explicit memory for new associations
in normal and amnesic patients. Journal of Experimental Psychology: Learning.
Memory & Cognition, 1l, 501–518.
Hamann, S. B., & Squire, L. R. (1997). Intact perceptual memory in the absence of
conscious memory. Behavioral Neuroscience, 111, 850–854.
Jacoby, L. L. (1991). A process dissociation framework: Separating automatic from
intentional uses of memory. Journal of Memory and Language, 30, 513–541.
Knowlton, B. J., Mangels, J. A., & Squire, L. R. (1996). A neostriatal learning system
in humans. Science, 273, 1399–1402.
Knowlton, B. J., Ramus, S. J., & Squire, L. R. (1992). Intact artificial grammar
learning in amnesia; Dissociation of classification learning and explicit memory
for specific instances. Psychological Science, 3, 172–179.
Knowlton, B. J., & Squire, L. R. (1993). The learning of natural categories: parallel
memory systems for item memory and category-level knowledge. Science, 262,
1747–1749.
Knowlton, B. J., Squire, L. R., & Gluck, M. A. (1994). Probabilistic classification
learning in amnesia. Learning & Memory, 1, 106–120.
Lieberman, M. D., Chang, G. Y., Chiao, J., Bookheimer, S. Y., & Knowlton, B. J.
(2004). An event-related fMRI study of artificial grammar learning in a balanced
chunk strength design. Journal of Cognitive Neuroscience, 16, 427–438.
Maccotta, L., & Buckner, R. L. (2004). Evidence for neural effects of repetition that
directly correlate with behavioral priming. Journal of Cognitive Neuroscience,
16, 1625–1632.
McClelland, J. L., McNaughton, B. L., & O’Reilly, R. C. (1995). Why there are
complementary learning systems in the hippocampus and neocortex: Insights
from the successes and failures of connectionist models of learning and memory.
Milner, B. L., Squire, L. R., & Kandel, E. R. (1998). Cognitive neuroscience and the
study of memory. Neuron, 20, 445–468.
Moscovitch, M., Rosenbaum, R. S., Gilboa, A., Addis, D. R., Westmacott R., Grady, C. L.,
Mcandrews, M. P., et al. (2005a). Functional neuroanatomy of remote episodic,
semantic and spatial memory: A unified account based on multiple trace theory.
Journal of Anatomy, 207, 35–56.
122 Paul J. Reber
Moscovitch, M., Westmacott, R., Gilboa, A., Addis, D. R., Rosenbaum, R. S., Viskontas, I.,
et al. (2005b). Hippocampal complex contribution to retention and retrieval of recent
and remote episodic and semantic memories: Evidence from behavioral and neuro-
imaging studies of healthy and brain-damaged people. In N. Ohta, C. M. MacLeod, &
B. Uttl (Eds.), Dynamic cognitive processes (pp. 333–380). Springer-Verlag.
Nissen, M. J., & Bullemer, P. (1987). Attentional requirements of learning: Evidence
from performance measures. Cognitive Psychology, 19, 1–32.
Nomura, E. M., Maddox, W. T., Filoteo, J. V., Ing, A. D., Gitelman, D. R., Parrish, T.
B., et al. (2007). Neural correlates of rule-based and information-integration
visual category learning. Cerebral Cortex, 17(1), 37–43.
Paller, K. A. (2002). Cross-cortical consolidation as the core defect in amnesia:
Prospects for hypothesis-testing with neuropsychology and neuroimaging. In L.
R. Squire & D. L. Schacter (Eds.), The neuropsychology of memory (3rd ed., pp.
73–87). New York: Guilford Press.
Poldrack, R. A., Clark, J., Pare-Blagoey, E. J., Shohamy, D., Creso Moyano, J.,
Myeres, C., et al. (2001). Interactive memory systems in the human brain. Nature,
414, 546–550.
Posner, M. I., & Keele, S. W. (1968). On the genesis of abstract ideas. Journal of
Experimental Psychology, 77, 353–363.
Reber, A. S. (1967). Implicit learning of artificial grammars. Journal of Verbal
Learning & Verbal Behavior, 6, 855–863.
Reber, P. J., Gitelman, D. R., Parrish, T. B., & Mesulam, M. M. (2003). Dissociating
explicit and implicit category knowledge with fMRI. Journal of Cognitive Neu-
roscience, 15, 574–685.
Reber, P. J., Martinez, L. A., & Weintraub, S. (2003). Artificial grammar learning in
Alzheimer’s disease. Cognitive, Affective and Behavioral Neuroscience, 3, 145–153.
Reber, P. J., Siwiec, R. M., Gitelman, D. R., Parrish, T. B., Mesulam, M.-M., & Paller,
K. A. (2002). Neural correlates of successful encoding identified using fMRI.
Journal of Neuroscience, 22, 9541–9548.
Reber, P. J., & Squire, L. R. (1994). Parallel brain systems for learning with and
without awareness. Learning & Memory, 2, 1–13.
Reber, P. J., & Squire, L. R. (1999). Intact learning of artificial grammars and intact
category learning by patients with Parkinson’s disease. Behavioral Neuroscience,
113, 235–242.
Rempel-Clower, N. L., Zola, S. M., Squire, L. R., & Amaral, D. G. (1996). Three cases
of enduring memory impairment following bilateral damage limited to the hippo-
campal formation. Journal of Neuroscience, 16, 5233–5255.
Rugg, M. D., Fletcher, P. C., Chua, P. M., & Dolan, R. J. (1999). The role of the
prefrontal cortex in recognition memory and memory for source: An fMRI study.
NeuroImage, 10, 520–529.
Schacter, D. L., & Buckner, R. L. (1998). Priming and the brain. Neuron, 20, 185–195.
Scoville, W. B., & Milner, B. (1957). Loss of recent memory after bilateral hippo-
campal lesions. Journal of Neurology, Neurosurgery & Psychiatry, 20, 11–21.
Siegert, R. J., Taylor, K. D., Weatherall, M., & Abernethy, D. A. (2006). Is implicit
sequence learning impaired in Parkinson’s disease? A meta-analysis. Neuropsy-
chology, 20, 490–495.
Shimamura, A. P., & Squire, L. R. (1989). Impaired priming of new associations in

amnesia. Journal of Experimental Psychology: Learning. Memory & Cognition,
15, 72l–728.
Skosnik, P. D., Gitelman, D. R., Parrish, T. B., Mesulam, M.-M., & Reber, P. J. (2002).
Neural correlates of artificial grammar learning. NeuroImage, 17, 1306–1314.
Squire, L. R. (1992). Memory and the hippocampus: A synthesis of findings with rats,
monkeys and humans. Psychological Review, 99, 195–231.
Squire, L. R., Ojemann, J. G., Miezin, F. M., Peterson, S. E., Videen, T. O., & Raichle,
M. E. (1992). Activation of the hippocampus in normal humans: A functional
anatomical study of memory. Proceedings of the National Academy of Science,
USA, 89, 1837–1841.
Stark, C. E., & Squire, L. R. (2000). Recognition memory and familiarity judgments
in severe amnesia: No evidence for a contribution of repetition priming. Beha-
vioral Neuroscience, 114, 459–467.
Wagner, A. D., Shannon, B. J., Kahn, I., & Buckner, R. L. (2005). Parietal lobe
contributions to episodic memory retrieval. Trends in Cognitive Sciences, 9,
445–453.
Wiggs, C. L., & Martin, A. (1998). Properties and mechanisms of perceptual priming.
Current Opinion in Neurobiology, 8, 227–233.
Wixted, J. (2007). Dual-process theory and signal-detection theory of recognition
memory. Psychological Review, 114, 152–176.
Yonelinas, A. P. (2002). The nature of recollection and familiarity: A review of 30
years of research. Journal of Memory and Language, 46 (441–517).
Human Learning 125
Learning and Memory for

Emotional Events
Alexandra S. Atkins and Patricia A. Reuter-Lorenz
Emotions are at the core of human experience. Varying in valence, a term

referring to the positive–negative dimension of emotion, and in intensity, a
dimension related to the arousal or magnitude of the experienced affect,
emotions are considered transient states, with phylogenetic continuity, rela-
tively hard-wired biological components, and characteristic facial and action
predispositions. Here we consider how events that elicit emotional
responses are learned and remembered. We examine three types of memory:
implicit memory, explicit episodic memory, and working memory. Implicit
memory, also referred to as procedural memory in some schemes, is largely
inaccessible to conscious awareness. Pavlovian fear conditioning figures
prominently in research on emotion and implicit memory. Explicit memory,
also referred to as declarative memory, is accessible to awareness and
verbal report and is tested by recall and recognition. Explicit memory
includes long-term memory for specific episodes as well as semantic mem-
ory for facts and information in the public domain. Here we review evidence
pertaining to the recognition and recall of emotionally laden events and
stimuli, as well as personal autobiographical memory (see also LeDoux,
2000, Phelps, 2006). The final section concerns working memory, which
refers to the online maintenance and manipulation of a limited amount of
information for a brief (several seconds) period of time. The relationship
between emotion and working memory has been least well examined; how-
ever, recent work suggests that this form of emotional memory may play an
important role in ‘‘feeling states,’’ emotional regulation, and decision making.
As we discuss, the bilateral amygdalae are central to emotional learn-
ing and memory across all three memory domains. Interactions with
other brain structures, such as the hippocampus and regions of prefron-
tal cortex, mediate different aspects of emotional memory, including the
extinction of emotional responses. We review the importance of the
126 Alexandra S. Atkins and Patricia A. Reuter-Lorenz
amygdalae and related structures by examining animal models as well as

human patient and neuroimaging studies.
1. Implicit emotional learning and memory
How do seemingly neutral places, objects or simple sensory events such

as odors or sounds, come to acquire emotional significance? This question
goes to the heart of the problem of emotional learning. The model task for
investigating emotional learning is fear conditioning, variations of which
have been used to examine the functional properties and neural correlates
in both animals and humans (LeDoux, 1998, 2000; Maren, 2001). In the fear
conditioning paradigm, a neutral stimulus (i.e, conditioned stimulus, CS)
such as a light or a tone is repeatedly paired with an aversive event
(i.e., unconditioned stimulus, US), such as an electric shock to the arm or
foot. As learning proceeds, the CS acquires aversive properties associated
with the shock, so that CS elicits a fear response as if it was the US. Once
the conditioned response has been acquired, animals such as laboratory
rats, will respond to the CS with a stereotypical constellation of measur-
able behaviors including freezing, whereas in humans, conditioned fear can
be measured by sweat gland activity evident in the electrodermal response,
or enhanced eyeblink startle reflex.
Several important properties of this form of emotional learning make
it fundamental to our psychological makeup. First, fear conditioning
falls under the domain of implicit learning or memory, because it can
take place outside of awareness (Critchley, Mathias, & Dolan, 2002;
Morris, Ohman, & Dolan, 1998) and it can be disassociated from the
kind of recollective experience that characterizes declarative, explicit
memory (Becharaet al., 1995). Second, fear conditioning is resilient to
forgetting (LeDoux, 1993, 1998). Even after the conditioned response is
extinguished, by exposure to repeated trials in which the CS does not
predict the US, the conditioned response can be readily reinstated by
contextual cues (Maren, 2001). Thus, the original associations between
the neutral stimulus and fear are not erased, but rather coexist with
new associations, that may or may not control the response depending
on the cues. Third, it has a highly specific, hard-wired circuitry that is
relatively preserved across species, properties that are consistent with
the idea that such forms of emotional learning are adaptive and critical
for survival.
The neural circuitry and cascade of neurobiological processes
that mediate fear conditioning have been specified in exquisite detail
(Maren, 2001). Brain structures such as the amygdala, hippocampus, and
medial prefrontal cortex which figure prominently in fear conditioning,
are also key players in other forms of emotional learning and memory.
Without the amygdala, fear conditioning cannot be acquired, and this is
true in rats and in rare human patients who have bilateral amygdaloid
damage due to surgery or disease (LeDoux, 2000). For example, patients
with Urbach–Wiethe disease, a hereditary disorder often associated with
damage to the bilateral amygdala and adjacent structures, exhibit ‘‘normal’’
fear responses; however, they cannot acquire conditioned associations
between neutral stimuli and aversive events (e.g., Bechara et al., 1995;
Damasio, 1994; LaBar, LeDoux, Spencer, Phelps, 1995). Nevertheless,
they have ‘‘normal’’ unconditioned responses to the shock along with
explicit memory for the learning experience and can discuss their knowl-
edge of repeated pairings of the CS with the US, indicating that their
declarative memory for the contingencies is intact (Phelps, 2006). More-
over, patients with retrograde amnesia due to hippocampal damage can
acquire conditioned fear responses, although they lack declarative memory
for the learning episodes (Bechara et al., 1995). Hippocampal damage also
interferes with contextual reinstatement of conditioned fear responses,
indicating the importance of this structure for encoding and associating
the contextual cues with the learned fear response (Phillips & LeDoux,
1992). Medial prefrontal sites, including regions of the anterior cingulate,
are important for suppressing fear responses during extinction (Morgan, &
LeDoux, 1995; Morgan, Romanski, & LeDoux, 1993; Sotres-Bayon, Cain, &
LeDoux, 2006), and like amygdala and hippocampal regions, are revealed
by functional brain imaging to be activated by the presentation of the CS
once fear learning has taken place (Buchel, Dolan, Armony, & Friston,
1999; LaBar, Gatenby, Gore, LeDoux, & Phelps, 1998).
Indeed, the neuroimaging data are remarkably compatible with the
findings from animals and with neuropsychology data from humans.
For example, the magnitude of the conditioned fear response as reflected
in the magnitude of electrodermal response is correlated with the amount
of amygdala activation obtained using fMRI (e.g., LaBar et al., 1998). Other
components of the emotional learning network do not show this relation-
ship to the acquired fear response. This, coupled with the fact that fear
conditioning cannot be acquired in the absence of an intact amygdala,
reveals that this structure is essential to this form of emotional learning.
2. Declarative emotional memory
Subjectively, we often feel that our memory for emotional events is super-
ior to our memory for neutral, everyday occurrences. Consistent with this
subjective impression, experimental research indicates that episodic memory
is enhanced for emotionally relevant material (Hamann, 2001), presumably
reflecting an evolutionary adaptation that allows us to remember and learn

from emotionally salient events in our lives (Dolcos, LaBar, & Cabeza, 2005;
LaBar & Cabeza, 2006; McGaugh, 2006). Laboratory findings consistently
demonstrate superior recall for emotionally valenced (positive or negative)
versus neutral words, pictures, and films (Anderson, 2006; Phelps, LaBar, &
Spencer, 1997; Hamann, 2001). Furthermore, recall of emotional materials is
positively correlated with individual emotional ratings; stimuli rated high in
emotional content are recalled better than material judged as more neutral
(Anderson, Wais, & Gabrieli, 2006).
What neural mechanisms underlie the selective episodic memory
enhancement for emotional information? As in emotional learning, evi-
dence gleaned from animal models and from lesion and neuroimaging
studies in humans indicates an important role for the bilateral amygdaloid
complex. Patients with bilateral amygdala damage often fail to show
enhanced memory for emotionally charged memoranda (Cahill, Babinsky,
Marksowitsch, & McGaugh 1995; LaBar, & Cabeza, 2006; but see Phelps
et al., 1997). A case study of B.P., a Urbach–Wiethe patient with damage
confined solely to the bilateral amygdalae, showed that the patient retained
normal emotional reaction to a negatively valenced story at encoding, yet
failed to demonstrate the memory benefit for emotional story events, which
was evident in age-matched controls (Cahill et al., 1995).
According to a prominent view, the modulation hypothesis, emotional
arousal at encoding leads to a release of adrenal stress hormones that
activate the amygdala and associated medial temporal lobe structures,
including the hippocampus, via noradrenergic pathways. In intact brains,
release of these hormones may lead to increased neuronal firing in the
memory structures of the medial temporal lobes, leading to enhanced
encoding and consolidation of emotionally relevant stimuli (McGaugh,
McIntyre, & Power, 2002; McGaugh & Roozendaal, 2002; Dolcos, LaBar &
Cabeza, 2004; Paz, Pelletier, Bauer, & Pare, 2006; McGaugh, 2006). The
extent to which release of adrenal hormones modulates and enhances
encoding per se, or operates during postencoding memory consolidation, is
currently unclear. Neverthless, the modulation view is consistent with ani-
mal work, patient studies, and studies involving healthy human subjects that
show increased emotional memory enhancement associated with adminis-
tration of adrenaline following encoding, and reduced or nonexistent emo-
tional memory enhancement when adrenaline is blocked via b-adrenergic
antagonists (McGaugh, 2006).
Data gleaned from functional neuroimaging studies with healthy human
subjects can help to elucidate the role of the amygdala in emotional
memory. Unlike patient or neuropharmocology experiments that lack the
temporal specificity required to distinguish between punctate psychological
stages of processing, functional neuroimaging techniques allow us to
investigate effects at encoding of emotional stimuli separately from

storage/consolidation or retrieval operations.
Several neuroimaging studies show increased event-related amygdala
activation during encoding of emotional memoranda (Hamann, 2001).
Early investigations by Cahill et al. (1996) used PET to examine brain activity
during encoding of emotional and neutral films and correlated this activity
with accuracy on subsequent recall. Results showed that increased activity in
the right (Cahill, et al., 1996) or left (Cahill, et al., 2001) amygdala was
associated with superior recall for emotional, but not neutral, film clips 3
weeks later (Hamann, 2001). Increased amygdala activation at encoding of
negative and positive, but not neutral, picture stimuli has also been observed
(Hamann, Ely, Grafton, & Kilts 1999). Consistent with the modulation hypoth-
esis, this activity predicted increased accuracy on a surprise recognition test
and was positively correlated with increased activity in other medial tem-
poral lobe structures, including the hippocampus and parahippocampal
regions (Hamann et al., 1999).
A recent behavioral study demonstrated that recall of emotional and
neutral pictures was positively correlated with individual emotional rat-
ings given at encoding (Anderson et al., 2006). Intriguingly, recall for
neutral pictures presented prior to highly emotional (negative or positive)
pictures received a retroactive memory benefit, suggesting that encoding
of emotional material may facilitate memory for neutral events that occur
within a specific temporal window. Anderson et al. (2006) found superior
recognition for neutral pictures presented 4 s, but not 9 s prior to an
emotional stimulus, whereas administration of stress hormones 30 s to
1 h after encoding has been shown to lead to memory enhancement.
Although the role of the amygdala in these results is currently unknown,
the authors argue for a fast-acting mechanism at the time of encoding,
which is distinct from the slower neuromodulatory effects of hormones
on memory consolidation processes in the medial temporal lobes (C. F.
McGaugh, 2006). The extent to which increased amygdala activation at
encoding reflects either hormonal modulation or some other neural
mechanism for enhancing encoding of emotional stimuli remains unre-
solved.
Although the majority of affective neuroscience research has focused on
determining the neural mechanisms associated with encoding and conso-
lidation of emotional versus neutral stimuli, there is an emerging literature
examining the role of the amygdala, hippocampus, and related medial
temporal lobe structures in the retrieval of emotional information from
memory (Dolcos et al., 2005; LaBar & Cabeza, 2006). These investigations
are of two types: (1) studies that examine retrieval of emotional memor-
anda studied in the laboratory and (2) studies that examine retrieval of
subjects’ own autobiographical memories.
Data from both types of investigations indicate an important role for

these structures in the retrieval of emotional memories. Studies in the
former category have shown increased activation of visual cortex, the
amygdala, parahippocampal cortex, hippocampus and prefrontal cortex
associated with retrieval of emotional information (Dolan, Lane, Chua, &
Fetcher, 2000; Dolcos et al., 2005; LaBar & Cabeza, 2006). Consideration
relevant to some of these studies, however, is that the close temporal
relationship between encoding and retrieval phases of the tasks might
have caused postencoding consolidation processes to be assessed together
with retrieval processes, thereby conflating two potentially separable
effects.
A study by Dolcos et al. (2005) attempted to isolate retrieval-specific
processes by using event-related fMRI to measure retrieval of emotional
stimuli encoded a full year prior to scanning. At retrieval, accurate recogni-
tion of emotional versus neutral pictures was associated with increased
activity in numerous regions throughout the medial temporal and frontal
lobes, including the amygdala, insula, anterior cingulate, and medial frontal
regions. Increased activity in both the amygdala and hippocampus has been
associated with retrieval of emotional memories for which subjects report
a sense of recollection, as opposed to mere familiarity (LaBar & Cabeza,
2006), suggesting that these medial temporal lobe structures play an impor-
tant role in the successful recollective retrieval, and not just the encoding
or consolidation, of emotional memories (Dolcos et al., 2005; LaBar &
Cabeza, 2006).
The second approach to studying the neural mechanisms underlying
retrieval of emotional memories is to use neuroimaging techniques to
image retrieval of individual autobiographic memories. Although this
approach lacks the experimental control provided by inducing mem-
ories in the laboratory, it has the advantage of allowing us to examine
the neural processes involved in conscious autobiographic recollec-
tion of emotional salient personal experiences. The notion that emo-
tionally arousing personal events hold a special status in episodic
memory is an old one. An early paper by Stratton (1919), for example,
uses the term retroactive hypermnesia to describe the clear, near
photographic subjective memory for highly emotional events such as
natural disasters. Similarly, the term flashbulb memory coined by
Brown and Kulik (1977) refers to highly accurate, detailed, vivid
memories for emotionally arousing events, such as first hearing of
an assassination.
Early work on flashbulb memories generally focused on highly public
events, and in particular on subjects’ memories of learning of the
assassination of President John F. Kennedy (Brown & Kulik, 1977).
Since the terrorist attacks of September 11, 2001, many investigations

have examined autobiographic memories surrounding first learning
of this event. Talarico and Rubin (2003) recorded memories of 9/11
and other neutral memories 1 day after the attacks of 9/11 and subse-
quently, either 1, 6, or 32 weeks later. Results showed the same rate
of forgetting associated with both flashbulb and neutral memories.
9/11 memories were no more consistent than neutral ones, although
subjects rated 9/11 memories higher on scales of recollection, confi-
dence, and vividness.
Talarico and Rubin’s (2003) findings suggest that an increased sense
of subjective recollective confidence, rather than accuracy per se, dis-
tinguishes flashbulb memories from other autobiographical memories.
A recent investigation by Sharot, Martorella, Delgado, and Phelps (2007)
used fMRI to assess whether flashbulb memories are further distin-
guished by a unique neural signature at retrieval. The investigators
asked two groups of subjects to retrieve autobiographical memories
from 9/11 and the preceding summer. All subjects were present in
Manhattan on the day of the attacks but were distinguished by their
proximity to ground zero, the first ‘‘downtown’’ group being those within
approximately two miles of this location and the second ‘‘midtown’’
group an average of four and a half miles away. Behaviorally, only
downtown participants, some of whom had direct personal contact
with falling debris on the day of the attacks, rated 9/11 memories higher
on scales of subjective recollection. For these participants, retrieval-
related activation in the left amygdala was greater during retrieval of
9/11 versus other memories, but no difference was found for the mid-
town group. Within this region of interest, 83% of all downtown subjects
showed increased activation for 9/11 versus other memories, while only
40% of midtown subjects showed this effect. Furthermore, this activa-
tion difference within the left amygdala was positively correlated with
personal ratings of recollection. Relative to midtown participants, down-
town participants showed increased bilateral amygdala activation and
decreased activity in the posterior parahippocampal cortex during
retrieval of 9/11 versus other memories.
Sharot et al.’s (2007) findings indicate that when direct personal
contact with a tragic public event occurs, retrieval of the resulting
flashbulb memories may indeed be associated with a unique neural
signature. Taken together, studies of retrieval of emotional memories
formed in the laboratory or through direct personal experience
suggest that the amygdala and related medial temporal lobe structures
are crucial for both encoding and retrieval of emotionally arousing
memories.
3. Emotion and working memory
In long-term memory, the advantage of emotional information typically

increases as the length of the delay between initial encoding and retrieval
increases. This fact alone would suggest that the brief retention intervals
that characterize working memory tasks would reduce any such advantage.
Although research addressing this question is sparse, there are indications
that maintaining negatively valenced information in working memory, in
the form of fearful faces or disturbing emotional scenes, has a disruptive
effect on performance compared to neutral facial expressions or neutral
scenes (Kensinger & Corkin, 2003; Perlstein, Elbert, & Stenger, 2002).
The disruptive effect of emotional content on working memory is correlated
with a decrease in activity in ‘‘executive’’ brain regions such as dorsolateral
prefrontal cortex (Dolcos & McCarthy, 2006; Perlstein et al., 2002), and with
an increase in the correlation between activity in the amygdala and activity
in regions of prefrontal cortex known to be involved in interference
resolution (Dolcos, Kragel, Wang, & McCarthy, 2006). These results suggest
emotion may have tradeoffs with cognition, which can adversely influence
working memory.
Working memory processes control the online maintenance and
manipulation of memoranda, and therefore may also be important for
affective introspection, emotion regulation, and other emotional pro-
cesses that require an individual to hold a feeling in mind in order to
evaluate it or compare it to other feeling states. These considerations
have led to the investigation of ‘‘affective working memory,’’ which is
hypothesized to be specialized processes dedicated to the online main-
tenance of affective memoranda (Davidson & Irwin, 1999; Mikels, Beyer,
Reuter-Lorenz, & Fredrickson, 2008; Mikels, Larkin, Reuter-Lorenz, & Car-
stensen, 2005). There is supportive evidence that the human working
memory system may include an affective subsystem that involves
regions in orbital frontal cortex and the amygdala (Mikels, & Reuter-
Lorenz, in press; Schaefer, Jackson, Davidson, Kimberg, & Thompson-
Schill, 2002). Within this putative affective working memory subsystem,
there appear to be differences in the efficiency of maintaining different
valances of memoranda. Negative affective states are more readily main-
tained than positive states, consistent with a more general negative emo-
tional bias that characterizes the younger adult. This effect reverses,
however, in older adults where positive emotions become more prominent
and are better retained in working memory (Mikels et al., 2005).
The contents of working memory have been equated with the contents
of consciousness. It follows from this that working memory may be the
system that enables subjective experience, or consciousness of feeling
states (e.g., LeDoux, 2000). If the working memory system includes
mechanisms that are specialized for emotion maintenance, as recent evi-

dence suggests, such mechanisms may reside at the interface between
emotion and cognition, thereby constituting an important agenda item for
future research.
4. Conclusion
Learning the affective significance of otherwise neutral stimuli and

events that predict adverse consequences ensures our survival, while the
emotional richness of recollective experience is a likely inspiration
to survive. As this review documents, the amygdala is vital to emotional
life, and ensures its endurance through memory. Ongoing efforts and
future work will continue to detail the specific interactions of this structure
with other brain mechanisms and the neurobiological processes that
mediate the persistence of emotional memories, their unique subjective
character, and their role in our phylogenetic, ontogenetic, and individual
development.
References
Anderson, A. K., Wais, P. E., & Gabrieli, J. D. E. (2006). Emotion enhances remem-
brance of neutral events past. Proceedings of the National Academy of Sciences
of the United States of America, 103, 1599–1604.
Bechara, A., Tranel, D., Damasio, H., Adolphs, R., Rockland, C., & Damasio, A. R.
(1995). Double dissociation of conditioning and declarative knowledge relative to
the amygdala and hippocampus in humans. Science, 269, 1115–1118.
Brown, R., & Kulik, J. (1977). Flashbulb memories. Cognition, 5, 73–99.
Buchel, C., Dolan, R. J., Armony, J. L., & Friston, K. J. (1999). Amygdala–hippocampal
involvement in human aversive trace conditioning revealed through event-related
functional magnetic resonance imaging. Journal of Neuroscience, 19, 10869–10886.
Cahill, L., Babinsky, R., Marksowitsch, H. J., & McGaugh, J. L. (1995). The amygdala
and emotional memory. Nature, 377, 295–296.
Cahill, L., Haier, R. J., Fallon, J., Alkire, M. T., Tang, C., Keator, D., et al. (1996).
Amygdala activity at encoding correlated with long-term, free recall of emotional
information. Proceedings of the National Academy of Sciences of the United
States of America, 93, 8016–8021.
Cahill, L., Haier, R. J., White, N. S., Fallon, J., Kilpatrick, L., Lawrence, C., et al.
(2001). Sex-related difference in amygdala activity during emotionally influenced
memory storage. Neurobiology of Learning and Memory, 75, 1–9.
Critchley, H. D., Mathias, C. J., & Dolan, R. J. (2002). Fear conditioning in humans:
the influence of awareness and autonomic arousal on functional neuroanatomy.
Neuron, 33, 653–663.
Damasio, A. R. (1994). Descartes’ error: Emotion, reason, and the human brain.
New York: Avon Books.
Davidson, R. J., & Irwin, W. (1999). The functional neuroanatomy of emotion and
affective style. Trends in Cognitive Sciences, 3, 211–221.
Dolan, R. J., Lane, R., Chua, P., & Fletcher, P. (2000). Dissociable temporal lobe
activations during emotional episodic memory retrieval. NeuroImage, 11, 203–209.
Dolcos, F., Kragel, P., Wang, L., & McCarthy, G. (2006). Role of the inferior frontal
cortex in coping with distracting emotions. Neuroreport, 17, 1591–1594.
Dolcos, F., LaBar, K. S., & Cabeza, R. (2004). Interaction between the amygdala and
the medial temporal lobe memory system predicts better memory for emotional
events. Neuron, 42, 855–863.
Dolcos, F., LaBar, K. S., & Cabeza, R. (2005). Remembering one year later: Role of
the amygdala and the medial temporal lobe memory system in retrieving emo-
tional memories. Proceedings of the National Academy of Sciences of the United
Dolcos, F., & McCarthy, G. (2006). Brain systems mediating cognitive interference
by emotional distraction. Journal of Neuroscience, 26, 2072–2079.
Hamann, S. (2001). Cognitive and neural mechanisms of emotional memory. Trends
in Cognitive Sciences, 5, 394–400.
Hamann, S. B., Ely, T. D., Grafton, S. T., & Kilts, C. D. (1999). Amygdala activity
related to enhanced memory for pleasant and aversive stimuli. Nature Neuro-
science, 2, 289–293.
Kensinger, E. A., & Corkin, S. (2003). Effect of negative emotional content on
working memory and long-term memory. Emotion, 3, 378–393.
LaBar, K. S., & Cabeza, R. (2006). Cognitive neuroscience of emotional memory.
Nature Reviews Neuroscience, 7, 54–64.
LaBar, K. S., Gatenby, J. C., Gore, J. C., LeDoux, J. E., & Phelps, E. A. (1998). Human
amygdala activation during conditioned fear acquisition and extinction: A mixed-
trial fMRI study. Neuron, 20, 937–945.
LaBar, K. S., LeDoux, J. E., Spencer, D. D., & Phelps, E. A. (1995). Impaired fear
conditioning following unilateral temporal lobectomy in humans. Journal of
LeDoux, J. (1998). The emotional brain. New York: Touchstone.
LeDoux, J. E. (2000). Emotional circuits in the brain. Annu. Rev. Neurosci., 23,
155–184.
LeDoux, J. E. (1993). Emotional memory system in the brain.Behavioural Brain
Maren, S. (2001). Neurobiology of pavlovian fear conditioning. Annual Review of
McGaugh, J. L. (2006). Make mild moments memorable: add a little arousal. Trends
in Cognitive Sciences, 10, 345–347.
McGaugh, J. L., McIntyre, C. K., & Power, A. E. (2002). Amygdala modulation of
memory consolidation: Interaction with other brain systems. Neurobiology of
Learning and Memory, 78, 539–552.
McGaugh, J. L., & Roozendaal, B. (2002). Role of adrenal stress hormones in forming
lasting memories in the brain. Current Opinion in Neurobiology, 12, 205–210.
Mikels, J. A., Larkin, G. R., Reuter-Lorenz, P. A., & Carstensen, L. L. (2005). Diver-
gent trajectories in the aging mind: changes in working memory for affective
versus visual information with age. Psychology and Aging, 20, 542–553.
Mikels, J. A. & Reuter-Lorenz, P. A. (in press). Affective working memory: Conver-
ging evidence for a new construct. In S. Yoshikawa (Ed.), Emotional mind:
New directions in affective science.
Mikels, J. A., Reuter-Lorenz, P. A., Beyer, J., & Fredrickson, B. L. (2008). Emotion
and working memory: Evidence for domain-specific processes for affect main-
tenance. Emotion, 8 (2), 256–266.
Morgan, M. A., & LeDoux, J. E. (1995). Differential contribution of dorsal and
ventral medial prefrontal cortex to the acquisition and extinction of conditioned
fear in rats Behavioural Neuroscience, 109, 681–688.
Morgan, M. A., Romanski, L. M., & LeDoux, J. E. (1993). Extinction of emotional
learning: contribution of medial prefrontal cortex. Neuroscience Letters, 153,
109–113.
Morris, J. S., Ohman, A., & Dolan, R. J. (1998). Conscious and unconscious emo-
tional learning in the human amygdala. Nature, 393, 467–470.
Paz, R., Pelletier, J. G., Bauer, E. P., & Pare, D. (2006). Emotional enhancement of
memory via amygdala-driven facilitation of rhinal interactions.Nature Neuro-
science, 9, 1321–1329.
Perlstein, W. M., Elbert, T., & Stenger, V. A. (2002). Dissociation in human prefrontal
cortex of affective influences on working memory-related activity. Proceedings of
the National Academy of Sciences of the United States of America, 99, 1736–1741.
Phelps, E. A. (2006). Emotion and cognition: Insights from studies of the human
amygdala.Annual Review of Psychology, 57, 27–53.
Phelps, E. A., LaBar, K. S., & Spencer, D. D. (1997). Memory for emotional words
following unilateral temporal lobectomy. Brain and Cognition, 35, 85–109.
Schaefer, S. M., Jackson, D. C., Davidson, R. J., Kimberg, D. Y., & Thompson-Schill,
S. L. (2002). Modulation of amygdala activity by conscious maintenance of
negative emotion. Journal of Cognitive Neuroscience, 14, 913–921.
Sharot, T., Martorella, E. A., Delgado, M. R., & Phelps, E. A. (2007). How personal
experience modulates the neural circuitry of memories of September 11. Proceed-
ings of the National Academy of Sciences of the United States of America, 104,
389–394.
Sotres-Bayon, F., Cain, C., & LeDoux, J. (2006). Brain mechanisms of fear extinc-
tion: historical perspectives on the contribution of prefrontal cortex. Biological
Psychiatry, 60, 329–336.
Stratton, G. M. (1919). Retroactive hypermnesia and other emotional effects on
memory. Psychological Review, 26, 474–486.
Human Learning 137
Age Differences in Memory
Demands on Cognitive Control and Association

Processes
Cindy Lustig and Kristin Flegal
The most-feared age-related memory declines stem from Alzheimer’s dis-

ease (AD). However, memory function is reduced even in normal aging. There is
growing consensus that the memory declines of AD and normal aging empha-
size different pathways (Buckner, 2004): AD leads to rapid degeneration, begin-
ning in and most strongly affecting medial temporal lobe (MTL) structures such
as the entorhinal cortex and hippocampus. Normal aging effects are compara-
tively slow and are preferentially related to declines in frontal and basal ganglia
volume, white matter volume and integrity, and reduced function of select
neurotransmitter systems, especially the dopamine system (Raz, 2000). The
prominent structural and neurochemical changes in frontal regions associate
with declines in executive function and cognitive control.
As we review below, reducing control demands often reduces and some-
times eliminates age differences in memory performance. However, pre-
dicting age differences in performance is not simply a matter of quantifying
the control demands of a task. Changes in frontal regions are quite promi-
nent, but not the only changes that occur. Besides the changes within brain
regions, white-matter declines also impact communication between
regions. Changes in sensory and motor systems may affect both the quality
of information that older adults receive (e.g., if older adults do not see
something as well, they may not remember it as well) and the quality of
their output. Furthermore, older adults are not passive in the face of these
changes; they often engage different strategies and/or different brain
regions than do young adults. These differences may reflect inefficient
processing, attempts at compensation, or a combination of the two.
Both impairment and attempts at compensation are reflected in brain
activations as well as in behavior. Age differences in brain activation can be
138 Cindy Lustig and Kristin Flegal
roughly grouped into three families. (1) Underactivations—less activity in a

region by older adults than young adults—are typically accompanied by
worse performance. (2) Overactivations—more activity in a region by older
adults—are sometimes associated with preserved performance and are
often interpreted as attempted compensation. Overactivations may be
within the same regions activated by young adults, or in other, often
bilateral regions. Both under- and overactivations are most frequently
found in lateral prefrontal regions, suggesting that they may represent
age-related increases or decreases in the engagement of cognitive control,
but they can also occur elsewhere in the brain. (3) Failures to deactivate
midline regions (posterior cingulate, ventromedial prefrontal cortex) have
only recently become an object of investigation (Lustig et al., 2003). They
appear to reflect failures to disengage from irrelevant processing in order
to focus on the task at hand (Grady, Springer, Hongwanishkul, McIntosh, &
Winocur, 2006; Persson, Lustig, Nelson, & Reuter-Lorenz, 2007), but differ-
ences in spontaneous activity and connectivity between memory-related
brain regions may also play a role (Andrews-Hanna, Snyder,
Vincent, Lustig, Head, Raichle, & Buckner, 2008).
Below, we briefly describe the age differences (or lack thereof) for
several major forms of memory, with particular attention to the age differ-
ences in brain structure or function that may underlie them. As will be
seen, age differences in memory are often (but not always) linked to age
differences in frontal brain regions and in cognitive control.
1. Nonassociative learning
Nonassociative learning (habituation, reduced response to a repeated

stimulus; and sensitization, increased response to a repeated stimulus) and
conditioning are easy to demonstrate even in simple organisms such as
aplysia. Surprisingly, they may not be preserved in aging. Age differences in
habituation are most prominent for low-level physiological responses
(evoked potentials, galvanic skin response) to simple sensory stimuli
(e.g., changes in a string of tones; Eisenstein, Bonheim, & Eisenstein,
1995; Weisz & Czigler, 2006). To some degree, low initial responses by
older adults may make habituation-related reductions difficult to detect
(floor effects). Very salient or complex stimuli that do not have the pro-
blem of low initial responses often do not show age differences in habitua-
tion (Wedig, Rauch, Albert, & Wright, 2005).
However, age differences in habituation often remain even after control-
ling for differences in sensory processes or the magnitude of the initial
response. These remaining differences are often interpreted as reflecting
an age-related deficit in inhibition (Alain & Woods, 1999; McDowd & Filion,
Age Differences in Memory 139
1992). We were unable to find any reports of age deficits in sensitization,

consistent with the theoretical view that aging impairs inhibitory function
while leaving activation processes relatively intact (Hasher & Zacks, 1988).
2. Conditioning
Older adults also show deficits in conditioning. Young adults quickly

begin to produce the unconditioned response (changes in heart rate or skin
response in fear conditioning, the blink in eyeblink conditioning) in
response to the cue, but older adults are usually slower to do so and
often fail entirely (Durkin, Prescott, Furchtgott, Cantor, & Powell, 1993).
Trace conditioning, in which a delay is imposed between the conditioned
(CS) and unconditioned (US) stimulus, is especially sensitive to MTL or
cholinergic dysfunction, and has been proposed as a screening measure for
AD (Woodruff-Pak, 2001).
There is some debate over the degree to which normal age differences in
conditioning are artifacts of age differences in sensory function, the aware-
ness of the CS–US relationship, or strategy use (Bellebaum & Daum, 2004;
Flaten & Friborg, 2005; Labar, Cook, Torpey, & Welsh-Bohmer, 2004).
Physiologically, age-related deterioration of the cerebellum, especially the
Purkinje cells, is thought to contribute to age-related deficits in condition-
ing by disrupting older adults’ ability to judge the time between the CS and
the US. This relatively unstable representation of the CS–US relationship
could make it more difficult for older adults to form the appropriate
association (Woodruff-Pak & Jaeger, 1998). However, age differences in
conditioning extend well beyond the cerebellum. For eyeblink condition-
ing, age differences in activation and functional connectivity have also
been shown in prefrontal and temporal cortex, posterior cingulate, mid-
brain, and caudate structures (Schreurs, Bahro, Molchan, Sunderland, &
McIntosh, 2001).
3. Procedural memory
Tests of procedural or skill memory are usually designed to minimize the

contribution of conscious, controlled processes. Older adults are some-
times slower in initial learning, but subsequent memory and transfer of
skills appear to be preserved (Howard & Howard, 1989; Howard et al., 2004;
Seidler, 2007). When older adults do show deficits, it may be either because
young adults become aware of the task parameters and begin using con-
scious strategies or because older adults have less distinct representations
at a neural level (Dennis, Howard, & Howard, 2006). Young and older adults
activate similar brain regions during procedural memory tasks (Daselaar,

Rombouts, Veltman, Raaijmakers, & Jonker, 2003), but emphasis within
that network may differ (Aizenstein et al., 2006; Fera et al., 2005).
4. Priming
Where procedural memory refers to experience-related improvement of

a skill, priming refers to experience-related improvement on specific items.
Age differences in priming are typically small and nonsignificant within an
individual study, but effect sizes differ across test types (Fleischman,
Gabrieli, Reminger, Vaidya, & Bennett, 1998). Tests that require simple
repetition of the same response typically do not show age differences.
However, those that require response construction often show small
effects favoring young adults, especially if the stimulus does not tightly
constrain response possibilities (e.g., ‘‘WIN-’’ versus ‘‘W_ _D_ W’’ as a cue
for ‘‘WINDOW’’). As with procedural memory, there is always a concern
that young adults may become differentially aware during the task and thus
use deliberate retrieval strategies.
Neuroimaging studies suggest that there may be age differences in
which aspects of a task show repetition benefits; young and older adults
show similar activity reductions in frontal regions related to cognitive
control (Lustig & Buckner, 2004). Young adults may show greater prim-
ing-related reductions in posterior regions related to semantic and percep-
tual representation, although the results differ across studies (Bäckman et al.,
1997; Daselaar, Veltman, Rombouts, Raaijmakers, & Jonker, 2005).
A speculative interpretation is that young and older adults benefit similarly
from repetition-related reductions in control demand, but older adults still
show the effects of weaker initial perceptual representations and greater
word knowledge (semantic memory), resulting in more alternative responses
from which the primed response must be selected.
5. Semantic memory
With age comes experience. Across studies, older adults’ advantage on

semantic knowledge tests is as great as their deficit on episodic memory
tests (Verhaeghen, 2003). Most studies tap verbal semantic knowledge, and
young adults typically show left-lateralized activations (e.g., Broca’s area).
Activation in older adults is often bilateral, and older adults often fail to
deactivate midline regions unrelated to task performance (Logan, Sanders,
Snyder, Morris, & Buckner, 2002; Lustig et al., 2003). These age differences
may be related to control demands; Persson et al. (2004, 2007) found
equivalent patterns of brain activation (and deactivation) in a verb genera-

tion task when interference from competing possibilities was low. How-
ever, when competition was high and placed high demands on controlled
selection processes, older adults showed less activation in left inferior
frontal gyrus, more activation in right inferior frontal gyrus, and less
deactivation of midline default-mode regions. As with priming, older adults’
greater semantic knowledge may paradoxically create a disadvantage if the
task requires fast selection of one specific item from that store.
6. Short-term and working memory
Remembering information for a brief time entails both passive storage

and more active working memory operations such as organization and
manipulation. Behaviorally, older adults often show preserved performance
on passive short-term memory tasks, but increasing deficits with increasing
demands on the controlled, executive processes of working memory. Con-
sistent with a specific deficit in executive processes, older adults often
underactivate prefrontal control-related regions such as left inferior frontal
gyrus, especially in high-conflict situations (e.g., Jonides et al., 2000).
One prominent theory suggests that the most important executive func-
tion differences are in the inhibitory control of irrelevant information
(Hasher & Zacks, 1988). Reducing demands for the inhibition of proactive
interference from now-irrelevant prior trials can eliminate age differences in
working memory span and the ability of span tasks to predict performance in
other areas of cognition (Lustig, May, & Hasher, 2001; but see Hedden &
Park, 2001). Older adults are also more vulnerable to external distractors
that reduce goal-directed attention (Chao & Knight, 1997; Fabiani, Low, Wee,
Sable, & Gratton, 2006) and show more activation in brain regions related to
processing such distractors (Gazzaley, Cooney, Rissman, & D’Esposito,
2005). General declines in goal maintenance, perhaps caused by declines in
resources such as processing speed, are the most popular theoretical alter-
native to inhibitory-based explanations (e.g., Salthouse, 1996).
Reuter-Lorenz and Mikels (2006) proposed a new alternative. Older adults
may engage control processes at lower levels of nominal task difficulty than do
young adults, to compensate for poorer function of perceptual, motor, or main-
tenance processes. As difficulty increases, young adults increasingly engage
executive control, but older adults are already ‘‘maxed out.’’ Consistent with this
idea, Reuter-Lorenz and others have found that older adults have equivalent
performance and overactivation of prefrontal regions at moderate levels of
working memory load. However, only young adults further increased activation
in these regions at higher load levels, so that at high load levels, older adults
underactivated relative to young adults (Reuter-Lorenz & Cappell, in press).
This alternative may provide a principled framework for understanding

why some studies using limited levels of load find underactivations
(Jonides et al., 2000; Rypma & D’Esposito, 2000) and others overactivations
(Cabeza et al., 2004; Reuter-Lorenz et al., 2000, 2001). Consistent with an
increased use of executive processes, overactivations (including bilateral
ones) are usually in frontal regions associated with control and less pro-
minent in posterior regions associated with stimulus representation
(Cabeza, 2002; Reuter-Lorenz & Lustig, 2005). At moderate load levels,
older adults who overactivate often perform better than those who do
not, supporting the idea of compensation (Reuter-Lorenz et al., 2000,
2001; Rypma & D’Esposito, 2000).
Modality differences also point to an interaction between storage and
executive processes. Age effects are especially large for spatial working
memory tasks or if the stimuli are not amenable to verbal codes (Myerson,
Hale, Rhee, & Jenkins, 1999; but see Park et al., 2002), perhaps because
they require greater use of executive processes. In contrast, preserved
semantic memory may provide some protection to verbal working memory
in older adults.
7. Episodic memory
Casual references to ‘‘memory’’ usually imply episodic memory: con-

scious knowledge of specific information or events, tied to a particular
place and time. Older adults typically perform worse than young adults on
tests of episodic memory. As described below, the nature and the size of
age deficits in episodic memory are related strongly—but not solely—to
age differences in cognitive control.
Early neuroimaging studies emphasized age-related underactivations
(Grady et al., 1995). These findings fit well with the behaviorally
derived view that many of older adults’ memory difficulties arise from
a failure to engage the processes needed to support successful remem-
bering (Craik & Lockhart, 1972). These processes, especially at the
encoding stage, are thought to be effortful and demanding. Age changes
in brain function, perhaps especially frontal brain functions related to
attention, might make such processes more ‘‘expensive’’ for older adults
to engage.
7.1. ENCODING
Divided-attention manipulations, often used to simulate hypothesized
age-related deficits in attentional resources, seem to have especially large
effects at encoding. For example, Anderson et al. (2000) found that under
full-attention conditions, older adults had less activity in memory-related

brain regions during encoding than did young adults. Dividing young adults’
attention reduced their left prefrontal cortex activation to the level of older
adults under full attention. Further, dividing attention reduced prefrontal
activation, MTL activation, and subsequent memory for both groups. In
contrast, dividing attention at retrieval had little effect on either perfor-
mance or brain activation for either group.
Other studies support the idea that aging is associated with reduced
attention at encoding, especially in unconstrained situations. Logan, et al.,
(2002) found that both young and older adults had low activity in encoding-
related left prefrontal brain regions in a letter judgment task that is known
to produce poor subsequent memory. Young adults increased activity in
these regions when instructed to intentionally memorize the words, but
older adults did not. Many other studies have found similar age-related
failures to engage needed encoding processes and brain regions. For exam-
ple, if given a prompt to refresh memory for a just-studied word (i.e., think
of that word again), young adults show increases in left prefrontal brain
activity not seen in older adults (Johnson, Mitchell, Raye, & Greene, 2004).
Engaging appropriate encoding processes may be more costly and/or
less likely for older adults, but is apparently not impossible. When Logan
et al. (2002) asked young and older adults to engage in a semantic judgment
task known to engage processes that promote subsequent memory, activity
was equivalent for young and older adults in left prefrontal regions strongly
associated with later subsequent memory. Older adults also had significant
activity in a right prefrontal region not strongly related to young adults’
performance.
Within older adult samples, frontal overactivations at encoding are
often associated with better performance and may reflect compensation
for reduced MTL function (Gutchess et al., 2005; Rosen et al., 2002).
However, overactivation is not restricted to prefrontal regions and is not
necessarily compensatory or useful (see Reuter-Lorenz & Lustig, 2005,
for a recent review). Ventral visual cortex regions that in young adults
are specialized for certain stimulus types (e.g., faces versus scenes)
appear less specialized in older adults (Park et al., 2004). Older adults
may also have more difficulty in inhibiting distraction in order to restrict
encoding processes to the correct stimuli: When instructed to ignore
irrelevant scenes and focus on to-be-remembered faces, young adults
showed significant reductions in the activation of scene-related
regions compared to passive viewing, but older adults did not (Gazzaley
et al., 2005).
Not all age differences in memory are easily linked to control differ-
ences; aging is also associated with reduced memory for new associations
(Mitchell, Johnson, Raye, & D’Esposito, 2000; Naveh-Benjamin, 2000).
These deficits are likely related to medial temporal lobe declines, and
encoding deficits may be especially important.
7.2. RETRIEVAL
Age differences in memory also scale with demands for control and
association at retrieval. Recognition is sometimes described as obligatory
but resource-demanding: It does not decline much under divided attention,
but costs are seen on secondary tasks, especially for older adults (Anderson,
1999). Compared to recognition, age differences are conspicuous on recall
tasks, which require more self-initiation and control, and particularly large on
source memory tasks, which also place heavy demands on association
(McIntyre & Craik, 1987; Spencer & Raz, 1995; Verhaeghen & Marcoen, 1993).
Age differences are small and occasionally reversed on measures of
familiarity, a relatively automatic recognition process, but large on recol-
lection, the controlled retrieval of specific details (Jennings & Jacoby,
1993). Consistent with these behavioral patterns, Daselaar, Fleck, Dobbins,
Madden, & Cabeza (2006) found that older adults had more activation in
familiarity-related MTL regions (rhinal cortex) but less in recollection-
related regions (hippocampus).
Velanova, Lustig, Jacoby, & Buckner, (2007) found that in high-demand
retrieval situations, older adults showed late, extended time courses of
frontal activity relative to young adults. They interpreted this pattern to
mean that young adults used early-stage selection processes to proactively
constrain retrieval to the relevant items, whereas older adults relied on
late-stage evaluation of retrieved items. Age differences in performance
were reduced and age differences in frontal time courses were eliminated
when control demands were reduced by practice. Age differences in epi-
sodic memory may relate not only to how much control different age
groups exert, but also when they exert it.
8. Control processes: targets for training
As reviewed above, cognitive control plays an important role in age

differences in many different memory functions, especially those
(working and episodic memory) that are of particular concern to
older adults. Training programs have begun to focus on improving
cognitive control as a method of improving memory (e.g., Jennings &
Jacoby, 2003). Control-focused programs may have larger effects
and show better transfer of training benefits to other tasks than do
other memory-training programs (cf., Jennings, Webster, Kleykamp, &
Dagenbach, 2005; Willis et al., 2006).
Age and individual differences in control may also influence training

success. Nyberg et al. (2003) trained young and old adults in the method-
of-loci mnemonic. Young adults showed large memory benefits, and
increased activity in both left prefrontal cortex and occipitoparietal cortex.
Many older adults failed to show significant benefits. Even those who did
show smaller benefits than young adults, and only increased activation in
occipitoparietal cortex, not prefrontal cortex. Echoing the results of
Velanova et al. (2007), Bissig and Lustig (2007) found that older adults
who exercised control at early processing stages of a memory-training task
showed greater benefits than did those who focused on later stages.
9. Summary
Cognitive control plays an important—but not exclusive—role in age

differences in memory. Research continues to better understand this role
and to exploit this understanding to improve the memory performance of
older adults.
References
Aizenstein, H. J., Butters, M. A., Clark, K. A., Figurski, J. L., Stenger, V. A., Nebes, R. D.,
et al. (2006). Prefrontal and striatal activation in elderly subjects during
concurrent implicit and explicit sequence learning. Neurobiology of Aging, 27 (5),
741–751.
Alain, C., & Woods, D. L. (1999). Age-related changes in processing auditory stimuli
during visual attention: Evidence for deficits in inhibitory control and sensory
memory. Psychology and Aging, 14 (3), 507–519.
Anderson, N. D. (1999). The attentional demands of encoding and retrieval in
younger and older adults: 2. Evidence from secondary task reaction time distri-
butions. Psychology and Aging, 14 (4), 645–655.
Anderson, N. D., Iidaka, T., Cabeza, R., Kapur, S., McIntosh, A. R., & Craik, F. I. M.
(2000). The effects of divided attention on encoding- and retrieval-related brain
activity: A PET study of younger and older adults. Journal of Cognitive Neu-
roscience, 12 (5), 775–792.
Andrews-Hanna, J. R., Snyder, A. Z., Vincent, J. L., Lustig, C., Head, D., Raichle, M. E., &
Buckner, R. L. (2008). Disruption of large-scale brain systems in advanced aging.
Neuron, 56 (5), 924–935.
Backman, L., Almkvist, O., Andersson, J., Nordberg, A., Winblad, B., Reineck, R.,
et al. (1997). Brain activation in young and older adults during implicit and
explicit retrieval. Journal of Cognitive Neuroscience, 9 (3), 378–391.
Bellebaum, C., & Daum, I. (2004). Effects of age and awareness on eyeblink condi-
tional discrimination learning. Behavioral Neuroscience, 118 (6), 1157–1165.
Bissig, D., & Lustig, C. (2007). Who benefits from memory training? Psychological
Science, 18 (8), 720–726.
Buckner, R. L. (2004). Memory and executive function in aging and AD: Multiple
factors that cause decline and reserve factors that compensate. Neuron, 44 (1),
195–208.
Cabeza, R. (2002). Hemispheric asymmetry reduction in older adults: The HAROLD
model. Psychology & Aging, 17, 85–100.
Cabeza, R., Daselaar, S. M., Dolcos, F., Prince, S. E., Budde, M., & Nyberg, L. (2004).
Task-independent and task-specific age effects on brain activity during working
memory, visual attention and episodic retrieval. Cerebral Cortex, 14 , 364–375.
Chao, L. L., & Knight, R. T. (1997). Prefrontal deficits in attention and inhibitory
control with aging. Cerebral Cortex, 7 , 63-69.
Craik, F. I. M., & Lockhart, R. S. (1972). Levels of Processing - Framework for memory
research.Journal of Verbal Learning and Verbal Behavior, 11 (6), 671–684.
Daselaar, S. M., Fleck, M. S., Dobbins, I. G., Madden, D. J., & Cabeza, R. (2006).
Effects of healthy aging on hippocampal and rhinal memory functions: An event-
related fMRI study. Cerebral Cortex, 16 (12), 1771–1782.
Daselaar, S. M., Rombouts, S., Veltman, D. J., Raaijmakers, J. G. W., & Jonker, C.
(2003). Similar network activated by young and old adults during the acquisition
of a motor sequence. Neurobiology of Aging, 24 (7), 1013–1019.
Daselaar, S. M., Veltman, D. J., Rombouts, S., Raaijmakers, J. G. W., & Jonker, C. (2005).
Aging affects both perceptual and lexical/semantic components of word stem priming:
An event-related MRI study. Neurobiology of Learning and Memory, 83 (3), 251–262.
Dennis, N. A., Howard, J. H., & Howard, D. V. (2006). Implicit sequence learning
without motor sequencing in young and old adults. Experimental Brain
Research, 175 (1), 153–164.
Durkin, M., Prescott, L., Furchtgott, E., Cantor, J., & Powell, D. A. (1993). Conco-
mitant eyeblink and heart-rate classical-conditioning in young, middle-aged, and
elderly human-subjects. Psychology and Aging, 8 (4), 571–581.
Eisenstein, E. M., Bonheim, P., & Eisenstein, D. (1995). Habituation of the galvanic
skin-response to tone as a function of age. Brain Research Bulletin, 37 (4), 343–350.
Fabiani, M., Low, K. A., Wee, E., Sable, J. J., & Gratton, G. (2006). Reduced
suppression or labile memory? Mechanisms of inefficient filtering of irrelevant
information in older adults. Journal of Cognitive Neuroscience, 18, 637–650.
Fera, F., Weickert, T. W., Goldberg, T. E., Tessitore, A., Hariri, A., Das, S., et al.
(2005). Neural mechanisms underlying probabilistic category learning in normal
aging. Journal of Neuroscience, 25 (49), 11340–11348.
Flaten, M. A., & Friborg, O. (2005). Impaired classical eyeblink conditioning in
elderly human subjects: the role of unconditioned response magnitude. Aging
Clinical and Experimental Research, 17 (6), 449–457.
Fleischman, D. A., Gabrieli, J. D. E., Reminger, S. L., Vaidya, C. J., & Bennett, D. A.
(1998). Object decision priming in Alzheimer’s disease. Journal of the Interna-
tional Neuropsychological Society, 4 (5), 435–446.
Gazzaley, A., Cooney, J. W., Rissman, J., & D’Esposito, M. (2005). Top-down sup-
pression deficit underlies working memory impairment in normal aging. Nature
Neuroscience, 8 (10), 1298–1300.
Grady, C. L., McIntosh, A. R., Horwitz, B., Maisog, J. M., Ungerleider, L. G., Mentis,
M. J., et al. (1995). Age-related reductions in human recognition memory due to
impaired encoding. Science, 269 (5221), 218–221.
Grady, C. L., Springer, M. V., Hongwanishkul, D., McIntosh, A. R., & Winocur, G.
(2006). Age-related changes in brain activity across the adult lifespan. Journal of
Cognitive Neuroscience, 18 (2), 227–241.
Gutchess, A. H., Welsh, R. C., Hedden, T., Bangert, A., Minear, M., Liu, L. L., et al.
(2005). Aging and the neural correlates of successful picture encoding: Frontal
activations compensate for decreased medial-temporal activity. Journal of Cog-
nitive Neuroscience, 17 (1), 84–96.
Hasher, L., & Zacks, R. (1988). Working memory, comprehension, and aging: A
review and new view. In G. H. Bower (Ed.), The psychology of learning and
motivation: Advances in research and theory (Vol. 22, pp. 193–225). New York:
Academic Press.
Hedden, T., & Park, D. (2001). Aging and interference in verbal working memory.
Psychology & Aging 16, 666–681.
Howard, D. V., & Howard, J. H. (1989). Age-differences in learning serial patterns –
direct versus indirect measures. Psychology and Aging, 4 (3), 357–364.
Howard, D. V., Howard, J. H., Japikse, K., DiYanni, C., Thompson, A., & Somberg, R.
(2004). Implicit sequence learning: Effects of level of structure, adult age, and
extended practice. Psychology and Aging, 19 (1), 79–92.
Jennings, J. M., & Jacoby, L. L. (1993). Automatic versus intentional uses of memory –
aging, attention, and control. Psychology and Aging, 8 (2), 283–293.
Jennings, J. M., & Jacoby, L. L. (2003). Improving memory in older adults: Training
recollection. Neuropsychological Rehabilitation, 13 (4), 417–440.
Jennings, J. M., Webster, L. M., Kleykamp, B. A., & Dagenbach, D. (2005). Recollec-
tion training and transfer effects in older adults: Successful use of a repetition-lag
procedure. Aging Neuropsychology and Cognition, 12 (3), 278–298.
Johnson, M. K., Mitchell, K. J., Raye, C. L., & Greene, E. J. (2004). An age-related
deficit in prefrontal cortical function associated with refreshing information.
Psychological Science, 15 (2), 127–132.
Jonides, J., Marshuetz, C., Smith, E. E., Reuter-Lorenz, P. A., Koeppe, R. A., &
Hartley, A. (2000). Age differences in behavior and PET activation reveal differ-
ences in interference resolution in verbal working memory. Journal of Cognitive
LaBar, K. S., Cook, C. A., Torpey, D. C., & Welsh-Bohmer, K. A. (2004). Impact of
healthy aging on awareness and fear conditioning. Behavioral Neuroscience, 118
(5), 905–915.
Logan, J. M., Sanders, A. L., Snyder, A. Z., Morris, J. C., & Buckner, R. L. (2002).
Under-recruitment and nonselective recruitment: Dissociable neural mechanisms
associated with aging. Neuron, 33 (5), 827–840.
Lustig, C., & Buckner, R. L. (2004). Preserved neural correlates of priming in old age
and dementia. Neuron, 42 (5), 865–875.
Lustig, C., May, C. P., & Hasher, L. (2001). Working memory span and the role
of proactive interference. Journal of Experimental Psychology: General, 130,
199–207.
Lustig, C., Snyder, A. Z., Bhakta, M., O’Brien, K. C., McAvoy, M., Raichle, M. E., et al.
(2003). Functional deactivations: Change with age and dementia of the Alzheimer
type. Proceedings of the National Academy of Sciences of the United States of
America, 100 (24), 14504–14509.
McDowd, J. M., & Filion, D. L. (1992). Aging, selective attention, and inhibitory
processes – a psychophysiological approach. Psychology and Aging, 7 (1), 65–71.
McIntyre, J. S., & Craik, F. (1987). Age-differences in memory for item and source
information. Canadian Journal of Psychology-Revue Canadienne De Psycholo-
gie, 41 (2), 175–192.
Mitchell, K. J., Johnson, M. K., Raye, C. L., & D’Esposito, M. (2000). fMRI evidence of
age-related hippocampal dysfunction in feature binding in working memory.
Cognitive Brain Research, 10 (1–2), 197–206.
Myerson, J., Hale, S., Rhee, S. H., & Jenkins, L. (1999). Selective interference with
verbal and spatial working memory in young and older adults. Journal of Ger-
ontology Series B Psychological and Social Sciences, 54, P161–P164.
Naveh-Benjamin, M. (2000). Adult age differences in memory performance: Tests of
an associative deficit hypothesis. Journal of Experimental Psychology-Learning
Memory and Cognition, 26 (5), 1170–1187.
Nyberg, L., Sandblom, J., Jones, S., Neely, A. S., Petersson, K. M., Ingvar, M., et al.
(2003). Neural correlates of training-related memory improvement in adulthood
and aging. Proceedings of the National Academy of Sciences of the United States
of America, 100 (23), 13728–13733.
Park, D. C., Lautenschlager, G., Hedden, T., Davidson, N. S., Smith, A. D., & Smith, P.
K. (2002). Models of visuospatial and verbal memory across the adult life span.
Psychology & Aging, 17, 299–320.
Park, D.C., Polk, T.A., Park, R., Minear, M., Savage, A., & Smith, M.R. (2004). Aging
reduces neural specialization in ventral visual cortex. Proceedings of
the National Academy of Sciences of the United States of America,101,
13091–13095.
Persson, J., Lustig, C., Nelson, J.K., & Reuter-Lorenz, P.A. (2007). Age differences in
deactivation: A link to cognitive control? Journal of Cognitive Neuroscience,
19 (6), 1021–1032.
Persson, J., Sylvester, C. Y. C., Nelson, J. K., Welsh, K. M., Jonides, J., & Reuter-
Lorenz, P. A. (2004). Selection requirements during verb generation: differential
recruitment in older and younger adults. Neuroimage, 23 (4), 1382–1390.
Raz, N. (2000). Aging of the brain and its impact on cognitive performance: Integra-
tion of structural and functional findings. In F. I. M. Craik & T. A. Salthouse (Eds.)
The handbook of aging and cognition (2nd ed.). Lawrence Erlbaum, New Jersey.
Reuter-Lorenz, P. A., & Cappell, K. (in press). Neurocognitive aging and the com-
pensation hypothesis. Current Directions in Psychological Science.
Reuter-Lorenz, P. A., Jonides, J., Smith, E. E., Hartley, A., Miller, A., Marshuetz, C.,
et al. (2000). Age differences in the frontal lateralization of verbal and
spatial working memory revealed by PET. Journal of Cognitive Neuroscience,
12 ,174–187.
Reuter-Lorenz, P. A., & Lustig, C. (2005). Brain aging: reorganizing discoveries about
the aging mind. Current Opinion in Neurobiology, 15 (2), 245–251.
Reuter-Lorenz, P. A., Marshuetz, C., Jonides, J., Smith, E. E., Hartley, A., & Koeppe,
R. (2001). Neurocognitive ageing of storage and executive processes.European
Journal of Cognitive Psychology, 13 ,257–278.
Reuter-Lorenz, P. A., & Mikels, J. A. (2006). The aging brain: Implications of endur-
ing plasticity for behavioral and cultural change. In: P. B. Baltes, P. A. Reuter-
Lorenz & F. Rösler (Eds.), Lifespan development and the brain: The perspective
of biocultural co-constructivism (pp. 255–276). New York: Cambridge University
Press.
Rypma, B., & D’Esposito, M. (2000). Isolating the neural mechanisms of age-related
changes in human working memory. Nature Neuroscience, 3, 509–515.
Salthouse, T. A. (1996). The processing-speed theory of adult age differences in
cognition.Psychological Review, 103, 403–428.
Rosen, A. C., Prull, M. W., O’Hara, R., Race, E. A., Desmond, J. E., Glover, G. H., et al.
(2002). Variable effects of aging on frontal lobe contributions to memory. Neu-
roreport, 13 (18), 2425–2428.
Schreurs, B. G., Bahro, M., Molchan, S. E., Sunderland, T., & McIntosh, A. R. (2001).
Interactions of prefrontal cortex during eyeblink conditioning as a function of
age. Neurobiology of Aging, 22 (2), 237–246.
Seidler, R. D. (2007). Aging affects motor learning but not savings at transfer of
learning. Learning & Memory, 14 (1–2), 17–21.
Spencer, W. D., & Raz, N. (1995). Differential effects of aging on memory for content
and context: A meta-analysis. Psychology and Aging, 10 (4), 527–539.
Velanova, K., Lustig, C., Jacoby, L.L., & Buckner, R.L. (2007). Evidence for frontally-
mediated controlled processing differences in older adults. Cerebral Cortex, 17,
1033–1046.
Verhaeghen, P. (2003). Aging and vocabulary scores: A meta-analysis. Psychology
and Aging, 18 (2), 332–339.
Verhaeghen, P., & Marcoen, A. (1993). Memory aging as a general phenomenon –
episodic recall of older adults is a function of episodic recall of young-adults.
Psychology and Aging, 8 (3), 380–388.
Wedig, M. M., Rauch, S. L., Albert, M. S., & Wright, C. I. (2005). Differential amygdala
habituation to neutral faces in young and elderly adults. Neuroscience Letters,
385 (2), 114–119.
Weisz, J., & Czigler, I. (2006). Age and novelty: Event-related brain potentials and
autonomic activity. Psychophysiology, 43 (3), 261–271.
Willis, S. L., Tennstedt, S. L., Marsiske, M., Ball, K., Elias, J., Koepke, K. M., et al.
(2006). Long-term effects of cognitive training on everyday functional outcomes
in older adults. Jama-Journal of the American Medical Association, 296 (23),
2805–2814.
Woodruff-Pak, D. S. (2001). Eyeblink classical conditioning differentiates normal
aging from Alzheimer’s disease. Integrative Physiological and Behavioral
Science, 36 (2), 87–108.
Woodruff-Pak, D. S., & Jaeger, M. E. (1998). Predictors of eyeblink classical con-
ditioning over the adult age span. Psychology and Aging, 13 (2), 193–205.
Part III
Human Motor Learning

Human Learning 153
Introduction
A Survey of Motor Learning Concepts

and Findings
Bruce Etnyre
Rice University
Compared to other areas of scientific study, such as neurophysiology or

experimental psychology, the study of motor control and learning is rela-
tively young. The segregated endeavors in neurophysiology and experimen-
tal psychology began to converge in the late 1960s and early 1970s as
behavioral scientists shifted interest to the intervening processes occurring
between the activities of the central and peripheral nervous systems and
the resultant overt movements. Progress in research of learning skills has
advanced from the process-oriented approach to the identification of
neural mechanisms involved throughout the development of motor learn-
ing and control. The chapters in this section include reviews of issues
related to the merging of neurophysiology and experimental psychology
by motor control and learning scientists.
The complex and elegant interaction of the central nervous system with
how movements are performed and learned is addressed from different
perspectives in the following chapters. Empirical evidence from numerous
studies is summarized in support of learning theories related to central
neural control of simple and complex movements, variability of synergistic
actions for coordinated movements, and neuroanatomical examination of
how brain structures such as the cerebellum, basal ganglia, hippocampus,
and cerebral cortex relate to motor learning and motor memory.
Approaches to our understanding of how the mind controls movements,
both skilled and unskilled, have been addressed from numerous perspec-
tives. These include theoretical descriptions related to performance and
learning of the classic work from Adams, Berstein, Feldman, Fitts, Keele,
Schmidt, and others, as well as the chapter authors. Numerous laboratory
and ‘‘real-life’’ paradigms have been devised from a multitude of movement
154 Part III
tasks from simple to complex for testing motor learning theories from the
combined neurological and psychological perspectives. Technological
advances, such as positron emission tomography (PET scan), functional
magnetic resonance imaging (fMRI), and transcranial magnetic stimulation
(TMS), have also promoted our understanding of how, where, when, and
why motor learning occurs across the phases or stages of learning. We have
also gained valuable information about motor learning from patients with
central nervous system pathologies, such as Parkinson’s disease, cerebral
and cerebellar vascular accidents, as well as from animal studies. All these
concepts and more are discussed in these motor learning chapters.
Human Learning 155
Two Aspects of Motor Learning
Learning Movements and Learning Synergies

Mark L. Latash
Pennsylvania State University
What does a person’s central nervous system (CNS) do when this person
learns a new motor skill? What changes happen in different neural struc-
tures? How are these changes related to the new patterns of control
variables and interaction among the many neuromotor elements? What
aspects of the new movement are memorized? These questions remain
without answers despite the numerous studies in the fields of motor learn-
ing and neural plasticity. The main reason for the lack of progress is the
lack of adequate theoretical frameworks. In other words, before trying to
get answers to these burning questions, one has to formulate them prop-
erly. For the purpose of this paper, I am going to separate (rather artifi-
cially) the main problems of producing a purposeful motor action into
those of control and coordination.
1. Motor control
To perform a voluntary movement, the CNS has to formulate a motor

task in terms of meaningful physical variables such as coordinates (trajec-
tories) of important parts of the body, forces that have to be produced
against external objects. Ultimately, these very variables have to reach
the desired values for the movement to be successful. However, brain
structures can produce movements only by sending neural signals to
intermediate structures such as the interneurons and motoneurons of the
spinal cord. These structures send signals to muscles that convert
the electrical action potentials into mechanical variables such as forces
and displacements.
156 Mark L. Latash
The processes involved in converting brain signals into actions are

complex and poorly predictable. This is due to factors from the following
incomplete list:
(1) Given a level of activation, muscle force depends on muscle length and
velocity.
(2) Muscle activation is defined by signals from descending pathways and
reflex loops converging on the alpha-motoneurons.
(3) Signals along the reflex loops depend on various physical variables
such as muscle length, velocity, force, pressure on the skin, joint
capsule tension, as well as on signals from receptors in other muscles
of the limb.
(4) A change in descending signals to a motoneuronal pool is accompanied
by modulation of sensitivity of muscle receptors to length and velocity.
(5) Transmission time delays may be of the order of several tens of
milliseconds up to 100 ms.
These apparently complicating factors did not escape the attention of a

great Russian physiologist, Nikolai Bernstein, who stated that it was impos-
sible for the brain structures to predict mechanical consequences of signals
sent to spinal structures (Bernstein, 1935, 1967).
Similar conclusions can be reached based on the physics of the system
for movement production. This system uses force generators that are
length- and velocity-dependent (muscles) due to both physical properties
of the peripheral tissues and the action of feedback loops. As such, it may
be described with sets of nonlinear differential equations and viewed as a
dynamic system. For such systems, control can be achieved only by manip-
ulating parameters of the equations but not state variables of the system
(Glansdorf & Prigogine, 1971). For an intact muscle, variables such as
length, velocity, force, stiffness, damping, and activation level are all
state variables. The task of researchers in motor control, therefore, is to
discover parameters that have physiological meaning and can be used as
control variables by the CNS.
Currently, only one theory of motor control is based on accepting all the
mentioned factors and suggesting a hypothetical control variable, that is,
the equilibrium point (EP) hypothesis (Feldman, 1966, 1986; reviewed in
Feldman & Levin, 1995). According to the EP hypothesis, control of a
muscle can be described with a variable related to subthreshold depolar-
ization of the alpha-motoneuronal pool. This variable defines the threshold
(l, hence, it is commonly addressed as the l-model) of the tonic stretch
reflex of the muscle that can be expressed in length units. The EP hypoth-
esis has been developed for the control of multimuscle systems (Feldman &
Levin, 1995; Feldman, Ostry, Levin, Gribble, & Mitnitski, 1998).
Two Aspects of Motor Learning 157
All alternative theories assume that the brain computes either requisite
forces or requisite patterns of muscle activation using stored-in memory
‘‘programs’’ (Schmidt, 1975) or, more recently, cascades of the so-called inter-
nal models (Kawato, 1999; Schweighofer, Arbib, & Kawato, 1998; Wolpert,
Miall, & Kawato, 1998). Note that none of these theories defines physiologi-
cally feasible control variables. Rather the complex, nonlinear transformations
leading to muscle forces and displacements are assumed to be handled by
computational ‘‘crutches.’’ Several recent studies have produced results that
are not well compatible with the idea of internal models (Ostry and Feldman,
2003; Malfait & Ostry, 2004; Malfait et al., 2005; Yang, Scholz, & Latash, 2007).
2. Motor coordination
At any level of analysis, the system for the production of voluntary

movements is redundant. This means that it has more elements than
absolutely necessary to solve typical motor tasks. For example, positioning
a fingertip at a target can be typically achieved with an infinite number of
joint angle combinations. Producing a certain magnitude of joint torque can
be achieved with an infinite number of muscle force combinations. A
required level of muscle activation can be achieved with recruitment of
different subsets of motor units at different frequencies.
Following the original suggestion by Bernstein (1967), the problems of
motor redundancy have been viewed as those of elimination of redundant
degrees of freedom (Newell, 1991; Newell, Broderick, Deutsch, & Slifkin,
2003; Vereijken, van Emmerick, Whiting, & Newell, 1992). An alternative view
was offered by Gelfand and Tsetlin (1966) and further developed as the
principle of abundance (Gelfand & Latash, 1998, 2002), and a computational
apparatus has been developed to analyze coordinated actions of the many
elements that take part in natural voluntary movements. According to the
principle of abundance, the CNS uses all the available degrees of freedom to
ensure stable and flexible production of important performance variables. In
other words, the CNS limits variability in the space of original degrees of
freedom (I am going to address these as elemental variables) along directions
that hurt important characteristics of performance (‘‘bad variability’’) while
allowing more variation along directions that do not hurt these characteris-
tics (‘‘good variability’’). Good variability is indeed good, not irrelevant
because it allows the system to participate in other concurrent tasks or
satisfy constraints without hurting performance in the main task.
A computational method to analyze the good and bad components of
variability has been developed within the framework of the uncontrolled
manifold (UCM) hypothesis (Scholz & Schöner, 1999; reviewed in Latash,
Scholz, & Schöner, 2007). The UCM hypothesis assumes that a controller
158 Mark L. Latash
acts in the space of elemental variables, selects in this space a subspace

(UCM) corresponding to a desired value of an important performance vari-
able, and then organizes covariation among elemental variables that limits
their variability to that subspace. ‘‘Elemental variable’’ is an axiomatic notion
for adequate variables produced by elements of the system at a selected level
of analysis. Since the computational apparatus of the UCM hypothesis is
sensitive to covariation among elemental variables, these variables have to
be chosen in such a way that they do not show ‘‘built-in’’ covariation that
would not be task-specific. For example, analysis of multifinger action and
multimuscle action typically starts with experimen-tal discovery of a set of
elemental variables since both fingers and muscles show considerable inter-
dependence that may not be reflective of a task-specific control strategy
(Latash, Scholz, Danion, & Schöner, 2001; Krishnamoorthy, Latash, Scholz, &
Zatsiorsky, 2003; Zatsiorsky, Li, & Latash, 2000).
The UCM framework has allowed to introduce an operational definition
for a synergy. This approach assumes that synergies are created in the
space of elemental variables with the purpose to ensure certain stability/
flexibility properties of a particular performance variable, that is, a parti-
cular relation between the ‘‘good’’ and ‘‘bad’’ components of variability.
Figure 1 illustrates a simple task of producing a certain total force
F1
UCMM
UCMF
V GOOD
V BAD
F2
Fig. 1. In this mental experiment, a person was asked to press with two effectors (for
example, with the two index fingers) and produce the force of 20 N. Finger forces
were measured in individual trials. The ellipses represent possible data distributions.
The line F1 + F2 = 20 N is the UCM for force stabilization (UCMF). Variability along this
line is ‘‘good’’ – VGOOD, while variability along the orthogonal (dashed line) is ‘‘bad’’ –
VBAD. If VGOOD > VBAD (the solid ellipse), the two elements form a synergy stabilizing
the total force (FTOT). If VGOOD = VBAD (the circle), this is not a synergy with respect to
FTOT. If VGOOD < VBAD (the dashed ellipse), this may be a sign of a synergy stabilizing
another performance variable, in this case, the total moment of force about a pivot in
between the points of force application. Its UCMM corresponds to the dashed line.
magnitude with two effectors pressing on two force sensors. The ellipses
illustrate possible data clouds over several repetitive attempts at the task.
Note that the force–force space of elemental variables has two special
directions, along the line F1 + F2 = FTASK (the solid line) and orthogonal to
that line. Variability along the solid line (which is the UCM for the task of
total force production, UCMF) does not affect total force and, as such, is
‘‘good,’’ while variability in the orthogonal direction is ‘‘bad.’’ To facilitate
comparison between spaces of different dimensionalities, it is handy to use
variance as a quantitative index of variability; so, I will address the two
components as VGOOD and VBAD.
Operationally, VGOOD > VBAD (the ellipse elongated along the UCMF line)
may be viewed as a sign of a synergy stabilizing the total force. The ratio of
VGOOD to VBAD (each quantified per dimension in the corresponding sub-
spaces!) or the normalized difference between the two has been used as
quantitative indices of synergies. A spherical data distribution corresponds
to VGOOD = VBAD and, as such, is an example of a nonsynergy. Comparison of
VGOOD and VBAD can be performed in different spaces of elemental variables
with respect to different performance variables. For example, the dashed
ellipse elongated along the dashed line corresponds to an inequality
VGOOD < VBAD for the total force production task and does not qualify as a
force-stabilizing synergy. However, it suggests the existence of another
synergy stabilizing the total moment of force with respect to a pivot between
the two effectors, for which the dashed line is the UCM (UCMM).
3. Two approaches to motor learning
Motor learning has been traditionally viewed as a staged process involving

steps such as freezing and freeing degrees of freedom (Bernstein, 1967;
Newell, 1991; Vereijken et al., 1992). Experimental studies have frequently
taken this view rather directly and reported practice, leading to a mechanical
degree of freedom being frozen or released based on changes in its range. In
contrast, the principle of abundance directs our attention not to changes in
the variability of elemental variables but in their covariation patterns that can
lead to contrasting changes in the good and bad components of variability.
Imagine that a person learns a new motor skill that requires changing a
performance variable from an initial value to a final value over a required
time interval. Commonly, in laboratory tasks, subjects are instructed to
practice a movement trying to change a selected performance variable as
quickly and as accurately as possible. This variable is produced by the
combined action of many elemental variables (e.g., an accurate pointing
movement is produced by rotation of individual joints, activations of many
muscles), which are in turn produced by changes in control variables.
Following the EP hypothesis, we can associate learning a new movement
160 Mark L. Latash
with a trajectory in the l-space of the participating muscles. However,

there is potentially another component to learning a movement that is
associated with ensuring its desired stability properties. One should not
view stability of a performance variable as a desired feature of any action
(cf. Hasan, 2005). For example, if a quick change in a variable is required, a
mechanism stabilizing its current value is counterproductive. Hence, learn-
ing a movement may be associated with elaborating synergies in the space
of elemental variables that ensure its stability in some phases of the action
and are turned off when a quick change in the variable is needed.
Indeed, a number of studies have shown that preparation to a quick
change in a performance variable is associated with an anticipatory drop in
the index of a synergy stabilizing that performance variable, the so-called
anticipatory synergy adjustments (Kim, Shim, Zatsiorsky, & Latash, 2006;
Olafsdottir, Yoshida, Zatsiorsky, & Latash, 2005; Shim, Olafsdottir,
Zatsiorsky, & Latash, 2005). Further, the index of the synergy drops and
may correspond to a pattern of covariation among elemental variables that is
destabilizing the performance variable (Kim et al., 2006; Shim et al., 2005).
Most studies of motor learning focus on changes in the selected perfor-
mance variable and its indices of variability. Very few of them address an
issue of HOW these changes are brought about. I would like to suggest that
a few other potentially important processes may contribute to improve-
ment in motor performance with practice.
First, one has to organize an optimal set of elemental variables that are used
as the basis for the performance variable. For example, in multimuscle actions,
muscles typically form a few groups with activation levels scaling nearly
linearly within a group. These have been addressed as muscle synergies
(Ivanenko, Poppele, & Lacquaniti, 2004; Ting & Macpherson, 2005; Tresch,
Cheung, & d’Avella, 2006; note that in this context, the word ‘‘synergy’’ has a
different meaning from the one introduced earlier!) or muscle modes (a less
loaded term; Krishnamoorthy, Latash, Scholz, & Zatsiorsky, 2003; Danna-Dos-
Santos, Slomka, Zatsiorsky, & Latash, 2007). The actions are based on selecting
sets of coefficients (gains) at which the modes are recruited in each particular
trial. These gains have been shown to covary to stabilize important physical
variables, that is, evidence for multimode synergies has been demonstrated.
However, can modes themselves be reorganized with practice? Recent
evidence suggests that this indeed may be true (Asaka, et al., 2008). So practice
can lead to learning (creating) a set of variables most adequate for ensuring
required stability properties of the performance variable.
Further, it is necessary to learn mapping between small changes in the
elemental variables and changes in the performance variable (the Jacobian of
the system). Knowledge of the Jacobian is crucial for feedforward control of
multielement systems (Goodman & Latash, 2006). To the best of my knowl-
edge, so far, no studies addressed this important step in motor learning.
The next step is learning (optimizing) synergies within the selected

space of elemental variables. Typical changes in motor patterns with
practice include a drop in all indices of variability of both elemental and
performance variables. However, only recently, changes in synergies sta-
bilizing potentially important performance variables have been addressed.
For simplicity, let us consider the earlier example of two-effector force
production (it is easy to generalize this analysis to multieffector tasks). A
drop in variability of the total force means that VBAD drops with practice.
But what can be expected from VGOOD? By definition, this component of
variability does not affect performance. Does the CNS care about this
seemingly irrelevant component of variability?
Figure 2 illustrates three possible scenarios. It assumes that before
practice, a synergy existed, stabilizing the total force across repetitive
trials (the ellipse representing a data distribution across trials in panel A).
With practice, VBAD drops. VGOOD can stay unchanged (or decrease less,
F1 F1
UCMF Stronger synergy
V GOOD
V BAD
(a) F 2 (b) F2
F1 Unchanged synergy F1 Weaker synergy
(c) F2 (d) F2
Fig. 2. Consider the same task as illustrated in Fig. 1. Practice is expected to lead to
a drop in variability of the performance variable—the total force, reflected in a drop
in ‘‘bad’’ variability, VBAD. ‘‘Good’’ variability, VGOOD, can stay unchanged (panel B),
change proportionally to VBAD (panel C), or change more than VBAD (panel D). This
may be associated with the synergy becoming stronger, staying unchanged, or
becoming weaker, respectively.
162 Mark L. Latash
or even increase), leading to a larger proportion of the total variance

accounted for by VGOOD. This may be interpreted as the synergy becoming
stronger (panel B). VGOOD can decrease in proportion to VBAD. This implies
more accurate performance without a change in the synergy (panel C). And
finally, VGOOD may drop more than VBAD, leading to a more spherical data
distribution illustrated in panel D. Although the last scenario may look
counterintuitive, data compatible with all three scenarios have been
reported.
It is natural to expect synergies to strengthen with practice and indeed
several studies reported results compatible with first scenario illustrated in
Fig. 2. Two of these studies investigated control subjects who practiced
Frisbee throwing (Yang & Scholz, 2005) and the production of an accurate
force pattern with an unusual finger combination (Kang, Shinohara,
Zatsiorsky, & Latash, 2004). One more study investigated the ability of
persons with Down syndrome to produce accurate profiles of the total force
while pressing with the four fingers of the dominant hand on four force sensors
(Latash, Kang, & Patterson, 2002). In this study, prior to practice, the partici-
pants showed predominantly positive covariation of finger forces that appar-
ently destabilized the total force. After three days of practice, negative
covariation among finger forces emerged corresponding to a force-stabilizing
synergy. The participants learned to use their hand not as a fork with four
prongs turned upside down but as a flexible and sophisticated motor
apparatus.
One of the very first studies led to the most unusual outcome, the third
scenario (Fig. 2D) corresponding to a decrease in the synergy strength with
practice (Domkin, Laczko, Jaric, Johansson, & Latash, 2002). In this study,
the subjects practiced a rather simple task of bimanual pointing when one
hand handled the target and the other hand handled the pointer. In a follow-
up (Domkin, Laczko, Djupsjöbacka, Jaric, & Latash, 2005) that involved a
more complex task, the second scenario (Fig. 2C) was observed with
proportional changes in both components of variance, VBAD and VGOOD.
Finally, one more study observed an increase in the synergy index early
in practice followed by its decrease (Latash, Yarrow, & Rothwell,, 2003). In
this study, the subjects reached very high accuracy in the performance of
the task (very low VBAD) about the midpoint over the practice time. With
further practice, they apparently adjusted their control strategies to handle
other components of the task without changing VBAD. This led to shrinking
VGOOD and to an apparent drop in the synergy index. There is more evi-
dence that a purposeful increase in VGOOD may be related to allowing the
system to explore various solutions in unusual conditions or in conditions
of uncertainty (De Freitas, Scholz, & Stehman, 2007; Yang et al., 2007).
To summarize, motor learning does not lead to freezing or freeing of
degrees of freedom. It leads to using all the original degrees of freedom to
form an adequate space in which necessary synergies can be created, and

then optimizing them with respect to important performance variables
defined by the task. Further practice may be directed at optimization of
other factors that may not be specified by the explicit requirements, for
example, postural task components, energy expenditure, avoiding fatigue
and clumsy postures. These processes may be associated with a drop in
quantitative indices of synergies due to the elaboration of more stereoty-
pical solutions.
4. Concluding comments: Where in the brain is motor memory?
The recent development of methods of brain imaging and noninvasive

stimulation of brain structures (such as transcranial magnetic stimulation,
TMS) has resulted in a wealth of information showing plastic changes in
brain structures following both prolonged and short-term practice. Most of
these studies reported higher excitability of brain structures such as the
primary motor cortex (Lemon, Johansson, & Westling, 1995; Pascual-Leone
et al., 1995; Pascual-Leone, 2001), although more subtle and task-specific
changes have also been reported (Classen, Liepert, Wise, Hallett, & Cohen,
1998; Latash et al., 2003). Neural plasticity seems to be everywhere in the
CNS. So it seems that virtually all neural structures have an ability to
change with experience. Do such changes qualify as motor memory? This
seems to be a matter of definition. If we assume that they do, the question
in the subtitle of this section becomes meaningless. The answer is: Every-
where.
Our inability to address the question of localization of motor memory
seems to stem from the lack of answers to a few very basic questions.
Can all neural elements change by experience? Are some neural ele-
ments better at storing such changes over long-time intervals than
others? Is it easier to retrieve memories stored in some neural elements
(structures) than in others? What is the role of diffuse, whole-brain
mechanisms in the processes of creating, storing, and retrieving mem-
ories? Until these questions are addressed, researchers will direct
attention to most conspicuous anatomical structures such as the cere-
bellum, the basal ganglia, the hippocampus, and certainly the cortex of
the large hemispheres in attempts to find where memories are stored.
It is relatively easy to document plastic changes in these large struc-
tures, but do such changes mean that those places are the loci of motor
memories? Do they memorize adequate patterns of control variables,
elemental variables, their relations to performance, and synergies? At
present, I have no answer, but at least it seems that we move closer to
formulating problems more exactly.
164 Mark L. Latash
Acknowledgments
Preparation of this paper has been partially supported by NIH grants

NS035032 and AG018751.
References
Bernstein, N. A. (1935). The problem of interrelation between coordination and

localization. Archivio di scienze biologiche, 38, 1–35 (in Russian).
Bernstein, N. A. (1967). The co-ordination and regulation of movements. Oxford:
Pergamon Press,.
Classen, J., Liepert, J., Wise, S. P., Hallett, M., & Cohen, L. G. (1998). Rapid plasticity
of human cortical movement representation induced by practice. Journal of
Danna-Dos-Santos, A., Slomka, K., Zatsiorsky, V. M., & Latash, M. L. (2007). Muscle
modes and synergies during voluntary body sway. Experimental Brain Research,
179, 533–550.
De Freitas, S. M., Scholz, J. P., & Stehman, A. J. (2007). Effect of motor planning on
use of motor abundance. Neuroscience Letters, 417, 66–71.
Domkin, D., Laczko, J., Jaric, S., Johansson, H., & Latash, M. L. (2002). Structure of
joint variability in bimanual pointing tasks. Experimental Brain Research, 143,
11–23.
Domkin, D., Laczko, J., Djupsjöbacka, M., Jaric, S., & Latash, M. L. (2005). Joint
angle variability in 3D bimanual pointing: Uncontrolled manifold analysis.
Experimental Brain Research, 163, 44–57.
Feldman, A. G. (1966). Functional tuning of the nervous system with control of
movement or maintenance of a steady posture. II. Controllable parameters of the
muscle. Biophysics (Oxf), 11, 565–578.
Feldman, A. G. (1986). Once more on the equilibrium-point hypothesis (l model) for
motor control. Journal of Motor Behavior, 18, 17–54.
Feldman, A. G., & Levin, M. F. (1995). The origin and use of positional
frames of reference in motor control. The Behavioral and Brain Sciences, 18,
723–806.
Feldman, A. G., Ostry, D. J., Levin, M. F., Gribble, P. L., & Mitnitski, A. B. (1998). Recent
tests of the equilibrium-point hypothesis (l model). Motor Control, 2, 189–205.
Gelfand, I. M., & Latash, M. L. (1998). On the problem of adequate language in
movement science. Motor Control, 2, 306–313.
Gelfand, I. M., & Latash, M. L. (2002). On the problem of adequate language in biology.
In: M. L. Latash (Ed.) Progress in motor control: Structure-function relations in
voluntary movement (Vol. 2, pp. 209–228), Urbana, IL: Human Kinetics.
Gelfand, I. M., & Tsetlin, M. L. (1966). On mathematical modeling of the mechanisms
of the central nervous system. In: I. M. Gelfand, V. S. Gurfinkel, S. V. Fomin, &
M. L . Tsetlin (Eds.) Models of the structural–functional organization of certain
biological systems (pp. 9–26), Moscow: Nauka (in Russian, published in English
in 1971 by Cambridge MA: MIT Press.).
Glansdorf, P., & Prigogine, I. (1971). Thermodynamic theory of structures: Stabi-

lity and fluctuations. Wiley: Hoboken, NJ.
Goodman, S. R., & Latash, M. L. (2006). Feedforward control of a redundant motor
system. Biological Cybernetics, 95, 271–280.
Hasan, Z. (2005). The human motor control system’s response to mechanical per-
turbation: should it, can it, and does it ensure stability? Journal of Motor Beha-
vior, 37, 484–493.
Ivanenko, Y. P., Poppele, R. E., & Lacquaniti, F. (2004). Five basic muscle activation
patterns account for muscle activity during human locomotion. Journal of Phy-
siology, 556, 267–282.
Kang, N., Shinohara, M, Zatsiorsky, V. M., & Latash, M. L. (2004). Learning multi-
finger synergies: An uncontrolled manifold analysis. Experimental Brain
Research, 157, 336–350.
Kawato M. (1999). Internal models for motor control and trajectory planning.
Kim, S. W., Shim, J. K., Zatsiorsky, V. M., & Latash, M. L. (2006). Anticipatory
adjustments of multi-finger synergies in preparation for self-triggered perturba-
tions. Experimental Brain Research, 174, 604–612.
Krishnamoorthy, V., Latash, M. L., Scholz, J. P., & Zatsiorsky, V. M. (2003). Muscle
synergies during shifts of the center of pressure by standing persons. Experi-
mental Brain Research, 152, 281–292.
Latash, M. L., Kang, N., & Patterson, D. (2002). Finger coordination in persons with
Down syndrome: A typical patterns of coordination and the effects of practice.
Experimental Brain Research, 146, 345–355.
Latash, M. L., Scholz, J. F., Danion, F., & Schöner, G. (2001). Structure of motor
variability in marginally redundant multi-finger force production tasks. Experi-
Latash, M. L., Scholz, J. P., & Schöner, G. (2007). Toward a new theory of motor
synergies. Motor Control, 11, 275–307.
Latash, M. L., Yarrow, K., & Rothwell, J. C. (2003). Changes in finger coordination
and responses to single pulse TMS of motor cortex during practice of a multi-
finger force production task. Experimental Brain Research, 151, 60–71.
Lemon, R. N., Johansson, R. S., & Westling, G. (1995). Corticospinal control
during reach, grasp and precision lift in man. Journal of Neuroscience, 15,
6145–6156.
Malfait, N., Gribble, P. L., & Ostry, D. J. (2005). Generalization of motor learning
based on multiple field exposures and local adaptation. Journal of Neurophysiol-
ogy, 93, 3327–3338.
Malfait, N., & Ostry, D. J. (2004). Is interlimb transfer of force-field adaptation a
cognitive response to the sudden introduction of load? Journal of Neuroscience,
24, 8084–8089.
Newell, K. M. (1991). Motor skill acquisition. Annual Review of Psychology, 42,
213–237.
Newell, K. M., Broderick, M. P., Deutsch, K. M., & Slifkin, A. B. (2003). Task
goals and change in dynamical degrees of freedom with motor learning.
Journal of Experimental Psychology. Human Perception and Performance,
29, 379–387.
166 Mark L. Latash
Olafsdottir, H., Yoshida, N, Zatsiorsky, V. M., & Latash, M. L. (2005). Anticipatory

covariation of finger forces during self-paced and reaction time force production.
Neuroscience Letters, 381, 92–96.
Ostry, D. J., & Feldman, A. G. (2003). A critical evaluation of the force control
hypothesis in motor control. Experimental Brain Research, 153, 275–288.
Pascual-Leone, A., Dang, N, Cohen, L. G., Brasil-Neto, J. P., Cammarota, A., &
Hallett, M. (1995). Modulation of muscle responses evoked by transcranial mag-
netic stimulation during the acquisition of new fine motor skills. Journal of
Pascual-Leone, A. (2001). The brain that plays music and is changed by it. Annals of
the New York Academy of Sciences, 930, 315–329.
Schmidt, R. A. (1975). A schema theory of discrete motor skill learning. Psycholo-
gical Review, 82, 225–260.
Scholz, J. P., & Schöner, G. (1999). The uncontrolled manifold concept: Identifying
control variables for a functional task.Experimental Brain Research, 126, 289–
306.
Schweighofer, N., Arbib, M. A., & Kawato, M. (1998). Role of the cerebellum in
reaching movements in humans. I. Distributed inverse dynamics control. The
Shim, J. K., Olafsdottir, H, Zatsiorsky, V. M., & Latash, M. L. (2005). The emergence
and disappearance of multi-digit synergies during force production tasks. Experi-
Ting, L. H., & Macpherson, J. M. (2005). A limited set of muscle synergies for force
control during a postural task. Journal of Neurophysiology, 93, 609–613.
Tresch, M. C., Cheung, V. C., & d’Avella, A. (2006). Matrix factorization algorithms
for the identification of muscle synergies: Evaluation on simulated and experi-
mental data sets. Journal of Neurophysiology, 95, 2199–2212.
Vereijken, B., van Emmerick, R. E. A., Whiting, H. T. A., & Newell, K. M. (1992).
Free(z)ing degrees of freedom in skill acquisition. Journal of Motor Behavior, 24,
133–142.
Wolpert, D. M., Miall, R. C., & Kawato, M. (1998). Internal models in the cerebellum.
Trends in Cognitive Sciences, 2, 338–347.
Yang, J.-F., & Scholz, J. P. (2005). Learning a throwing task is associated with
differential changes in the use of motor abundance. Experimental Brain
Yang, J.-F, Scholz, J. P., & Latash, M. L. (2007). The role of kinematic redundancy in
adaptation of reaching. Experimental Brain Research, 176, 54–69.
Zatsiorsky, V. M., Li, Z.- M., & Latash, M. L. (2000). Enslaving effects in multi-finger
force production. Experimental Brain Research, 131, 187–195.
Human Learning 167
Neuroanatomical Correlates of Motor

Skill Learning: Inferences from
Neuroimaging to Behavior
K. Lindquist and M. A. Guadagnoli
University of Nevada, Las Vegas
The learning of motor skills is important from both theoretical and

applied perspectives. As such, it is understandable that motor skill learning
has been a topic of investigation for more than 100 years and, during this
time, behavioral observations have been the building blocks of any model
that attempts to explain motor skill learning. One model of learning that
has stood the test of time is a three-stage motor skill learning model
proposed by Fitts and Posner in 1967. Because of a lack of technology,
much of this work was based on behavioral observations of individuals as
they progressed from beginners to experts. However, since the early work
of Fitts and Posner (1967) and others in the motor learning field, advances
in technology have provided additional insight into the learning process.
Indeed, since 1967, scientists’ understanding of the brain has increased
dramatically, yet the Fitts and Posner motor learning model is as relevant
today as it was when it was first proposed. To this end, this chapter will use
Fitts and Posner’s model as a framework to describe more recent advances
in motor learning. Much of this information results from various modern
research techniques and studies of patient populations.
1. Cognitive-stage skill learning
According to Fitts and Posner, the first stage of learning, known as the
cognitive stage, is marked by an ‘‘intellectualization’’ of the task. This
intellectualization includes learning the importance of cues, responses,
and events that are necessary for task success. The cognitive stage is
essentially a patchwork of old habits put together in a novel fashion and
supplemented by a few new habits (Fitts & Posner, 1967). Since the time of
168 K. Lindquist and M. A. Guadagnoli
Fitts and Posner, neuroimaging studies have revealed two brain structures
that are critical to this stage of learning: the cerebellum and the striatum
(Doyon & Ungerleider, 2002).
The cerebellum acts as the site for integration and coupling of afferent
(sensory) input and efferent (motor) output (Manzoni, 2005) and is
involved in the development and maintenance of generalized motor pro-
grams (Nezafat, Shadmehr, & Holcomb, 2001). Integration of sensory and
motor signals and the development of generalized motor programs are
important for the fast, early stage of learning (Doyon & Ungerleider,
2002). The cerebellum may also contribute to both specific and nonspecific
transfer of motor skills (Obayashi, 2004).
Studies using positron emission tomography scanning (PET scan) and
functional magnetic resonance imaging (fMRI) techniques have shown that
the cerebellum is highly active during the cognitive stage, while this activ-
ity diminishes to nearly undetectable levels as practice continues (Doyon,
Penhune, & Ungerleider, 2003). Additionally, cerebellar activation patterns
vary between sessions of early learning. When activation of the cerebellum
is examined in more detail, the cerebellar cortex ipsilateral to the side
being used is most active over the first two sessions of early learning of a
sequential reaction time task, but less active in the third session of learning
for the same task. The area of activation in the cortex of the cerebellum
decreases greatly over the first three learning sessions. The ipsilateral
dentate nucleus is active in session two of learning only, resulting in the
greatest amount of activation seen in the cerebellar nuclei over three
sessions of early learning (Doyon et al., 2002).
The results of this study suggest that the contributions of the cerebellum
vary depending on the amount of motor sequence practice. In the case of
motor skill transfer and adaptation, the cerebellum acts with the frontal
and parietal association areas to update the stored motor program (or
internal model) to fit the parameters of the transfer task (Obayashi,
2004). Indeed, studies examining conditioned eyeblink responses have
shown that soon after the completion of skill acquisition, adaptation of
behavior is strongly dependent on cerebellar cortex plasticity. However, if
the adaptation situation is presented at a longer time interval from initial
learning, adaptation becomes less dependent on cerebellar cortex plasti-
city (Krakauer & Shadmehr, 2006). As learning progresses through the
cognitive stage, cerebellar cortex plasticity appears to produce plasticity
in the cerebellar nuclei (Krakauer & Shadmehr, 2006). Therefore, as con-
solidation of motor skills occurs, plasticity and activation patterns in the
cerebellum change from cerebellar cortex-dominated pathways to cerebel-
lar nuclei-dominated pathways. Understandably then, investigations of
cerebellar stroke have revealed that patients have difficulty in learning
specific information vital to this stage of learning (Boyd & Winstein, 2004).
Neuroanatomical Correlates of Motor Skill Learning 169
In addition to cerebellar activity, recent neuroimaging studies have

identified two circuits within the basal ganglia (BG) related to the stages
of motor skill learning (Hikosaka, Nakamura, Sakai, & Nakahara, 2002).
The first is the anterior associative/premotor loop, which is active during
the cognitive stage. The second is the posterior sensorimotor/BG loop,
which becomes active in the later stages of motor skill learning. Jueptner,
Firth, Brooks, Frackowiak, and Passingham (1997) hypothesized that the
activation of the sensorimotor/BG loop relates to the use of working
memory and cognitive strategies.
Possible markers for the end of the cognitive stage and progression to
the next stage of learning could therefore be posterior sensorimotor/BG
loop activity and activity of the cerebellar nuclei. Activation of the cere-
bellar nuclei suggests that the learner has progressed from adapting pre-
viously learned motor skills to a novel application to consolidation of the
novel application as a ‘‘new’’ motor skill. Whereas sensorimotor/BG loop
activity suggests that the learner has become familiar enough with the
parameters of the task and is now able to concentrate more on perfor-
mance refinement. Other markers of an individual’s progression from the
cognitive stage to the associative stage of learning are increased activation
in the parietotemporal and occipitotemporal association areas—these
areas are responsible for the association of sensory input and motor
output—and decreased activation of the primary motor and primary soma-
tosensory areas (Mitra, Bhalerao, Summers, & Williams, 2005).
2. Associative-stage skill learning
The second stage of the Fitts and Posner (1967) model is known as the
associative stage because it is the stage in which perceptual and motor aspects
of the task are associated. During this stage, the shift in activation patterns
from the cerebellum to the parietotemporal and occipitotemporal association
areas may support gradual elimination of performance errors. Behavioral
errors in this stage of learning may consist of inappropriate subroutines,
incorrect sequences of acts, or responses to unimportant cues (Fitts & Posner,
1967). Gradual elimination of errors is one of the key characteristics of the
associative stage of learning and serves to refine the skill that is being learned.
Adams (1971) hypothesized that during the performance of a self-paced
skill, the performer creates a memory representation consisting of sensory
information about that skill. Adams referred to this memory representation
as a perceptual trace, which subsequently has evolved into the motor pro-
gram concept. A perceptual trace is stored reference sensory information to
which the performer can compare current sensory information to determine
if the movement is performed without error. As an individual increases in
competence, the perceptual trace becomes more robust (Adams, 1971). In

other words, if an individual performs a skill often, the performer is better
able to anticipate the sensory information generated by that skill.
Logically, Adams’ theory (1971) would suggest that if an individual was
able to perform the skill perfectly over several repetitions, the performer
would develop a strong perceptual trace for the proper performance of
that skill. Therefore, in theory, the most appropriate training technique
to increase the learner’s efficiency in the associative stage of learning
would be one that constrains movement in a way that only allows
for proper performance. A recently proposed motor learning model, the
corticostriatal model, suggests that this theory may indeed be viable
(Doyon & Ungerleider, 2002).
Doyon and colleagues conducted a series of neuroimaging studies to
investigate how activity within the striatum relates to behavior during the
associative stage of learning (Doyon & Ungerleider, 2002; Doyon & Benali,
2005; Doyon et al., 2003; Lehéricy et al., 2005; Monchi, Petrides, Strafella,
Worsley, & Doyon, 2006). These experiments support a model of motor
learning in which motor sequence learning progresses with experience
from a corticocerebellar-dependent pathway to a corticostriatal-dependent
pathway. Motor adaptation (control of already learned movements
that must be adapted to fit environmental parameters) appears to be the
opposite, progressing from a corticostriatal pathway to a corticocerebellar
pathway with experience.
Initial support for this two-circuit model of motor learning arose from an
experiment designed to investigate cerebellar contributions to the motor
learning process (Doyon et al., 2002). This study demonstrated that, as stated
earlier, activation in the cerebellum decreases as learning progresses
through the cognitive stage. Results from this study also suggested that as
participants became more familiar with the motor sequence, activation in
the corticostriatal loop increases. In this experiment, specific activation
increases associated with the corticostriatal model occurred in the right
striatum, sensorimotor area (SMA). The corticostriatal model predictes the
activity of both the striatum and SMA, as striatal output directly influences
the motor-related structures in the frontal lobe (Doyon et al., 2002).
Another study in this series (Lehéricy et al., 2005) used fMRI to examine
which areas of the BG were active during performance and learning of a
novel motor task. The areas found to be active during the first scan (perfor-
mance on the first day of learning) included the dorsal putamen, the asso-
ciative/premotor areas of the BG, and the more rostral striatal areas, such as
the anterodorsal globus pallidus, the output nuclei of the thalamus that
correspond to the motor task, and the subthalamic nucleus (STN). Interest-
ingly, participants’ accuracy and speed increased dramatically over the first
10 min of the first scanning session, and this increase in speed/accuracy
corresponded to increased activations within the sensorimotor territory of

the putamen and globus pallidus (Lehéricy et al., 2005). Most of the observed
changes in activation patterns occurred during the fast, early learning stage
(i.e., as the learner progressed from the cognitive stage to the associative
stage), as activation patterns shifted from associative/premotor areas to the
sensorimotor territories. Less rapid changes in activation occurred over
extended practice intervals (i.e., associative stage to autonomous stage).
These results suggest that the representational changes seen within the
striatum and associated motor cortical regions contribute to the long-lasting
retention of motor skills (Lehéricy et al., 2005).
Therefore, based on Adam’s (1971) perceptual trace theory and Doyon
and colleagues’ (2002) corticostriatal motor learning model, practice con-
ducted in a manner that allows for reactivation of the corticostriatal circuit
should result in a strengthening of the connections between the higher
cortical areas and the striatum. In other words, practice that assists the
individual in achieving the optimal performance parameters should facil-
itate the encoding of proper performance of the task.
Support for the above conclusions is not unanimous. Another point of
view also supported by neuroimaging results is that efficient learning of
novel internal models is dependent on the ability to inhibit previously
learned, and now inappropriate, internal models (Shadmehr & Holcomb,
1999). If efficient performance is composed of the ability to activate the
desired internal model and inhibit previously learned internal models, a
learner should experience many ways to complete the same task after
learning the initial parameters of that task. In this case, the resulting task
performance would be more adaptable and less susceptible to interference
from external perturbations.
Building on the idea of simultaneous activation and inhibition of neural
circuitry, Shadmehr and Holcomb (1999) demonstrated that both activation
of the desired movement plan and inhibition of previously learned similar
movement plans occur during learning in the associative stage. More speci-
fically, Shadmehr and Holcomb found that the contralateral putamen and the
bilateral dorsolateral prefrontal cortex (PFC) were active during initial skill
acquisition with the dominant hand and when participants were exposed to a
force field rotated 180 immediately after original force-field skill acquisi-
tion. However, participants who were exposed to the 180 force-field shift
after a 5-h consolidation period did not reactivate areas used in initial skill
acquisition (Shadmehr & Holcomb, 1999). Additionally, participants in the
delayed exposure group showed lower activation in the bilateral ventrolat-
eral PFC when compared to the immediate exposure group and another
group who performed field A at 5.5 h after initial exposure. These findings
suggest that increases in activation in the ventrolateral PFC relate to suc-
cessful inhibition of competing motor memories.
Increased activation in the ventrolateral PFC may relate to the ability to

switch attention efficiently from stimuli that are no longer appropriate for
performance to novel stimuli that are appropriate for performance in the
new environment (Shadmehr & Holcomb, 1999). Since the caudate nucleus
of the BG receives inhibitory input from the PFC, increased activation in
the ventrolateral PFC would result in decreased activation of the caudate
nucleus. Thus, decreased activation of the caudate nucleus also plays a role
in the successful control of perseveration (the tendency for a memory to
persist in the absence of the appropriate stimuli).
Overall, participants in the delayed exposure condition adapted to
the field changes faster and more accurately than participants who were
in the immediate exposure condition. Participants in the delayed condition
also showed better recall performance in the initial force field condition
than participants in the immediate condition (Shadmehr & Holcomb,
1999). If these task performance measures and neuroimaging data are
considered together, variable practice in the associative stage is beneficial
to learning if memory consolidation occurs before interfering tasks are
performed. Moreover, if interfering tasks are performed too closely
together, learning efficiency for both tasks will decrease and the learner
will need a greater amount of practice in the associative stage. Activation
patterns seen between the two conditions suggest that interpretation of
sensory information and executive decision-making based on sensory
information are important factors that become more efficient after con-
solidation.
To investigate how background sensory information is consolidated and
used to refine movements during the associative stage of learning, Mitra
et al. (2005) used fMRI to examine activity in the association regions of the
cerebrum. This study compared brain activation patterns seen in a learning
condition, where participants were directed to follow sequence cues on a
monitor, to brain activation patterns seen in a test condition, where parti-
cipants were directed to recall a previously learned sequence. Mitra et al.
(2005) found that a greater amount of activation was present in the occipi-
totemporal and parietotemporal regions than in the primary motor or
primary somatosensory regions in the testing, but not in the learning,
condition. This is an interesting finding, as each of these conditions
would have generated a similar amount of motor activation within the
brain. This would suggest that activations seen in the association (occipito-
and parietotemporal) regions are strongly related to the spontaneous gen-
eration of learned motor sequences (Mitra et al., 2005). Findings from this
study support the idea that, with continued practice of a motor skill,
connections form between the brain areas responsible for motor output
and sensory input. In essence, the brain learns the sensory consequences of
the motor skill during the associative stage.
3. Autonomous-stage learning
The progression from associative stage to the autonomous stage of

learning is operationally defined as the ability to perform the skill indepen-
dently, with less cognitive effort than either the cognitive or the associative
stage of learning. The performer is also able to perform the skill without
interference in the presence of environmental distractions or ongoing
activities (Fitts & Posner, 1967).
In the premotor and motor cortices, motor skill learning is defined by
small groups of cells that appear to encode elementary movements and
recruit increasing numbers of neighboring cells with similar output proper-
ties (Monfils, Plautz, & Kleim, 2005). Development of successful muscle
synergies from a larger pool of possible synergies requires selection and
activation of the appropriate neurological circuits and inhibition of inap-
propriate circuits. This process strengthens specific spatiotemporal rela-
tionships, while weakening others simultaneously. Strengthening of
specific spatiotemporal relationships results in a bias of the cortical cir-
cuitry toward practiced movements, such that expression of these move-
ments is most likely when the cortex is stimulated (Hess & Donoghue,
1994; Monfils et al., 2005).
Motor map reorganization occurs as a consolidation effect of training
and may be the reason that some motor skills seem resistant to decay while
others do not (Monfils et al., 2005). Kleim et al. (2004) explored the concept
of consolidation-related motor map reorganization in rats. Kleim et al.
found that rats that were removed from a learning environment
during the cognitive phase of learning did not show motor map reorganiza-
tion when they were mapped 30 days later. These rats were not able
to reproduce the required behavior at the level they were at prior to
removal from the learning environment. Moreover, rats that remained in
the learning environment after the cognitive phase of learning showed
motor map reorganization and maintained motor skill 30 days later
(Kleim et al., 2004).
Humans also demonstrate evidence of motor map reorganization with
skill learning. Not only is evidence for motor map reorganization present in
humans, motor map reorganization occurs at a faster rate in humans than
seen in animal models. For example, participants who practiced a skilled
digit sequence on a piano showed significant motor map reorganization
(detected through transcortical magnetic stimulation or TMS) after 5 days
of practice (Pascual-Leone et al., 1995). These participants showed a sig-
nificant increase in the cortical area of digit representation that corre-
sponded to the trained hand and decreases in activation threshold within
the same area. Similar motor cortex reorganization has been seen in skilled
racquetball players (Pearce, Thickbroom, Byrnes, & Mastaglia, 2000),
skilled braille readers (Pascual-Leone et al., 1993), and in novices with

practice on a skilled tongue protrusion task after 1 week (Svensson, Roma-
niello, Arendt-Nieisen, & Sessle, 2003). These results suggest that the
progression from the associative stage of learning to the autonomous
stage of learning is the result of reorganization in the higher cortical
areas. Therefore, in the case of an overlearned skill, the motor map has
been reorganized to produce the specific sequence of muscle actions
related to that skill. If the skill is performed inappropriately, the motor
map needs to be stimulated to reorganize in order to learn to associate the
proper sequence of muscles with the desired movement.
Throughout this chapter, the Fitts and Posner model of motor skill
learning is used as a framework for more recent advances in the scientific
understanding of the brain as motor skills are learned. Specifically, this
chapter focuses on those areas of the brain in which activity changes the
most during motor learning, not on those that are or are not active during
motor performance. This is an important distinction because while differ-
ent areas of the brain may be active depending on the motor skill being
performed, the areas reviewed here are active depending on the perfor-
mer’s experience level with respect to the motor skill.
Following Fitts and Posner’s model, cognitive stage learning is accom-
panied by activations within the cerebellum. Cerebellar activation is
thought to associate with the observable behavior of adaptation of old
movements to the new movement environment. The associative stage of
learning is typified by the reduction in performance errors and an increase
in activity within the corticostriatal circuit. Whether practice during this
stage should be structured to provide the learner with the opportunity to
practice the motor skill perfectly or whether practice should be structured
to allow the learner to experience many ways to perform the skill is still an
area that requires more experimentation. Finally, when a skill is practiced
to the point of being ‘‘ingrained’’ or ‘‘automatic,’’ the learner is said to be in
the autonomous stage of motor skill learning. When a performer reaches
this stage, motor map reorganization with respect to the individual’s per-
formance of that skill occurs. This is another point at which a distinction
must be made. Simply because the performer reaches autonomous stage,
skill performance does not mean that that performance is necessarily
correct. For example, many individuals allow their knees to come together
(knee valgus position) during performance of the squat movement. These
individuals have reached an autonomous level of performance for the squat
movement, but do not have the biomechanically appropriate performance
encoded for that movement. This raises the question of if it is possible to
regress performance from the autonomous to the associative stage for
correction of performance errors, and if regression is possible, what is
the best method for facilitating these corrections.
The current paper utilized Fitts and Posner’s (1967) behaviorally based
model of motor learning to guide the review of neuroanatomy related to the
stages of learning for motor skills. Fitts and Posner’s model was based on
descriptions of typical behaviors of learners as they progressed from
novices to experts. Interestingly, when this model was proposed, many of
the modern methods used for investigating motor learning had not yet been
invented. Considering this, it is remarkable that the model is such an
appropriate vehicle to discuss recent findings from neuroimaging. By com-
bining the behaviors described by Fitts and Posner’s model and neuroima-
ging results, it becomes clear that as a learner progresses from novice to
expert, there is a predictable succession of reliance on brain structures and
that the observable behaviors associated with learning arise from the use
of these brain structures. As such, in addition to the description provided in
this paper, it is reasonable that the Fitts and Posner model can be used as a
vehicle to develop a model for motor learning based on the areas of the
brain that are involved at each stage of learning.
References
Adams, J. A. (1971). A closed-loop theory of motor learning. Journal of Motor

Behavior, 3(2), 111–150.
Boyd, L. A., & Winstein, C. J. (2004). Cerebellar stroke impairs temporal but not
spatial accuracy during implicit motor learning. Neurorehabilitation and Neural
Repair, 18(3), 134–143.
Doyon, J., & Ungerleider, L. G. (2002). Functional anatomy of motor skill learning.
In L. R. Squire & D. L. Schacter (Eds.), Neuropsychology of memory (3rd ed.,
pp. 225–238). New York, NY: Guilford Press.
Doyon, J., & Benali, H. (2005). Reorganization and plasticity in the adult
brain during learning of motor skills. Current Opinion in Neurobiology, 15(2),
161–167.
Doyon, J., Penhune, V., & Ungerleider, L. G. (2003). Distinct contribution of the
cortico-striatal and cortico-cerebellar systems to motor skill learning. Neuropsy-
chologia, 41(3), 252–262.
Doyon, J., Song, A. W., Karni, A., Lalonde, F., Adams, M. M., & Ungerleider, L. G.
(2002). Experience-dependent changes in cerebellar contributions to motor
sequence learning. Proceedings of the National Academy of Sciences of the
Fitts, P. M., & Posner, M. I. (1967). Basic concepts in psychology series: human
performance. Belmont, CA: Brooks/Cole Publishing Co.
Hess, G., & Donoghue, J. P. (1994). Long-term potentiation of horizontal connec-
tions provides a mechanism to reorganize cortical motor maps. Journal of
Neurophysiology, 71(6), 2543–2547.
Hikosaka, O., Nakamura, K., Sakai, K., & Nakahara, H. (2002). Central mechanisms
of motor skill learning. Current Opinion in Neurobiology, 12(2), 217–222.
Jueptner, M., Frith, C. D., Brooks, D. J., Frackowiak, R. S. J., & Passingham, R. E.
(1997). Anatomy of motor learning. II. Subcortical structures and learning by trial
and error. Journal of Neurophysiology, 77(3), 1325–1337.
Kleim, J. A., Hogg, T. M., VandenBerg, P. M., Cooper, N. R., Bruneau, R., & Remple,
M. (2004). Cortical synaptogenesis and motor map reorganization occur during
late, but not early, phase of motor skill learning. Journal of Neuroscience, 24(3),
628–633.
Krakauer, J. W., & Shadmehr, R. (2006). Consolidation of motor memory. Trends in
Neurosciences, 29(1), 58–64.
Lehéricy, S., Benali, H., Van De Moortele, P.-. Pélégrini-issac, M., Waechter, T.,
Ugurbil, K., et al. (2005). Distinct basal ganglia territories are engaged in early
and advanced motor sequence learning. Proceedings of the National Academy of
Sciences of the United States of America, 102(35), 12566–12571.
Manzoni, D. (2005). The cerebellum May implement the appropriate coupling of
sensory inputs and motor responses: Evidence from vestibular physiology.
Cerebellum, 4(3), 178–188.
Mitra, S., Bhalerao, A., Summers, P., & Williams, S. C. R. (2005). Cortical organiza-
tion of sensory corrections in visuomotor skill acquisition. Neuroscience Letters,
382(1–2), 76–81.
Monchi, O., Petrides, M., Strafella, A. P., Worsley, K. J., & Doyon, J. (2006). Func-
tional role of the basal ganglia in the planning and execution of actions. Annals of
Neurology, 59(2), 257–264.
Monfils, M.-. Plautz, E. J., & Kleim, J. A. (2005). In search of the motor engram:
Motor map plasticity as a mechanism for encoding motor experience. Neuros-
cientist, 11(5), 471–483.
Nezafat, R., Shadmehr, R., & Holcomb, H. H. (2001). Long-term adaptation to
dynamics of reaching movements: A PET study. Experimental Brain Research,
140(1), 66–76.
Obayashi, S. (2004). Possible mechanism for transfer of motor skill learning: Impli-
cation of the cerebellum. Cerebellum, 3(4), 204–211.
Pascual-Leone, A., Cammarota, A., Wassermann, E. M., Brasil-Neto, J. P., Cohen, L.
G., & Hallett, M. (1993). Modulation of motor cortical outputs to the reading hand
of braille readers. Annals of Neurology, 34(1), 33–37.
Pascual-Leone, A., Dang, N., Cohen, L. G., Brasil-Neto, J. P., Cammarota, A., &
Hallett, M. (1995). Modulation of muscle responses evoked by transcranial mag-
netic stimulation during the acquisition of new fine motor skills. Journal of
Neurophysiology, 74(3), 1037–1045.
Pearce, A. J., Thickbroom, G. W., Byrnes, M. L., & Mastaglia, F. L. (2000). Functional
reorganisation of the corticomotor projection to the hand in skilled racquet
players. Experimental Brain Research, 130(2), 238–243.
Shadmehr, R., & Holcomb, H. H. (1999). Inhibitory control of competing motor
memories. Experimental Brain Research, 126(2), 235–251.
Svensson, P., Romaniello, A., Arendt-Nieisen, L., & Sessle, B. J. (2003). Plasticity in
corticomotor control of the human tongue musculature induced by tongue-task
training. Experimental Brain Research, 152(1), 42–51.
Human Learning 177
Mechanisms Underlying Short-Term

Motor Learning, Long-Term Motor
Learning and Transfer
Daniel M. Corcos1, Jonathan Shemmell2, and David E.
Vaillancourt3
1
Departments of Kinesiology and Nutrition, Bioengineering Physical
Therapy and Neurology and Rehabilitation, University of Illinois at
Chicago, Chicago, IL
2
Department of Neurological Sciences, Rush Presbyterian St. Luke’s
Medical Center, Chicago, IL
3
Sensory Motor Performance Program, The Rehabilitation Institute of
Chicago, Chicago, IL
The desire and ability to improve motor performance is an essential

characteristic of human behavior. This desire can manifest itself either in
terms of wanting to learn various new and different movement patterns
such as in a sequence of gymnastic movements, or in terms of improving a
single movement pattern such as a throw. This chapter will focus on
studies in which the goal of the task is to improve a single movement
pattern in terms of speed and accuracy, and then generalize this pattern to
a similar type of movement, such as a longer throw. In the first part of this
chapter, we will focus on selected studies that have investigated how
patterns of muscle activation change to enhance motor performance in
one experimental session. In the second part of this chapter, we will review
studies that have investigated changes in performance and muscle activa-
tion patterns over multiple experimental sessions. Finally, we will review
studies that have investigated the extent to which performance changes
learned when practicing one task transfer to variants of the task. We will
focus on a series of studies that we have performed over the past 20 years
using a single joint model system, and integrate our findings with those
obtained using more complicated movements, and also from animal
neurophysiology.
178 Daniel M. Corcos, Jonathan Shemmell, and David E. Vaillancourt
1. Performance enhancement within an experimental session
One simple type of motor learning entails practice-related performance

changes in simple movements (Sanes, 2003). When neurologically normal
individuals intentionally make elbow movements at increasing speeds over
the same movement distance, both the agonist electromyogram (EMG)
burst and the antagonist EMG burst increase in magnitude. The duration
of the agonist burst is relatively constant. The slope of the agonist burst
rises more steeply for faster movements (Corcos, Gottlieb, & Agarwal,
1989b). Corcos, Jaric, Agarwal, & Gottlieb (1993) wanted to determine
the extent to which practice could alter the myoelectric and kinematic
correlates of such simple movements, and whether the muscle activation
pattern would be altered in the same way as for intentional changes in
movement speed. They had subjects perform 200 movements within one
experimental session. Movement position, velocity, and acceleration were
measured, as were surface EMG recordings from elbow flexor and exten-
sor muscles. They showed that subjects could increase their speed by as
much as 20%, and that the EMG changes were consistent with those
observed when movement speed is changed intentionally [see also Chapter
15 (Corcos, Gottlieb, Jaric, Cromwell, & Agarwal, 1990)]. They also showed
that even a highly skilled baseball pitcher is capable of improving perfor-
mance on this simple task by increasing muscle activation in the same way.
This study meets Sanes’ definition of motor learning in that performance
changes as a result of practice.
One can go beyond this minimal definition of motor learning and ask
whether the time course of learning different parameters associated with a
simple movement remains the same or varies. For example, are the same
numbers of experimental trials needed for peak movement velocity to
reach a plateau in performance as are needed for EMG parameters to
reach a plateau? If the time course is different for different parameters, it
raises the question as to whether there are different neural substrates
underlying the performance changes observed for the different parameters.
Our laboratory used the same single joint model paradigm as previously
discussed, and had neurologically healthy subjects perform 40 blocks of 10
movement trials (Flament, Shapiro, Kempf, & Corcos, 1999). We showed
that reaction time reached a plateau first, and that this was within the first
50 trials. Movement duration, acceleration time, and deceleration time
reached a plateau next. This was followed by EMG parameters reaching a
plateau. Finally, measures of peak velocity and peak acceleration reached a
plateau. In a follow-up study, we showed that the same findings apply when
movements are performed by the wrist joint, and also when instructions
place much greater emphasis on movement accuracy (Kempf, Corcos, &
Flament, 2001). These two studies raise the possibility that the neural
Mechanisms Underlying Motor Learning and Transfer 179
circuits that encode the timing of movements adapt to practice at different

rates to those that are related to increasing the amount of muscle activa-
tion. For example, it is well known that the cerebellum has been implicated
in movement timing (Keele & Ivry, 1990), and cerebellar lesions have been
implicated in abnormal EMG patterns (Hore, Wild, & Diener, 1991; Wild &
Corcos, 1997). As such, the cerebellum may well be instrumental in the
timing of muscle activation, and since short-term practice activates the
cerebellar cortex (Penhune & Doyon, 2005), it may well be involved in
timing changes in muscle activation.
One way to study the role of different neural circuits is to study patients
with movement disorders. Most studies of short-term practice-related
changes in motor performance in patients with Parkinson’s disease (PD)
found that motor learning is either very slightly impaired (Smiley-Oyen,
Worringham, & Cross, 2002) or normal (Laforce & Doyon, 2001). For
example, Jordan and Sager (1994) had patients with PD and healthy con-
trols perform 20 isometric contraction trials. Although there were clear
differences in motor performance confirming that the patients were indeed
bradykinetic, there was no difference in the rate of motor learning.
Flament, Vaillancourt, Kempf, Shannon, & Corcos (2003) used our single
joint paradigm and were interested in whether patients with PD could
change their performance within a practice session. They confirmed the
finding of Jordan and Sagar (1994) and showed that patients can improve
their performance to a similar extent as healthy individuals but their final
performance is still much slower. However, they showed that the EMG
signal was shorter and more fractionated than the EMG patterns observed
in healthy individuals, and it remained this way after 400 practice trials.
This study suggests that learning, as defined by Sanes (2003), is retained by
patients with PD but they do have a motor control deficit that is not
ameliorated by practice.
There are many candidates for brain regions where these short-term
performance changes may occur and may be retained, and these candidate
regions can be probed in detail in animal studies. The principal candidate
brain region is probably the motor cortex because it has the most direct
relationship to changes in patterns of muscle activation, and it can exhibit
rapid changes in its representation patterns. Strong evidence for the role of
the motor cortex in early skill learning comes from an elegant study by
Kargo and Nitz (2003) that involves a more complicated task than our
single joint model system. They simultaneously recorded forelimb muscle
EMG activity and activity from layer 5 of the motor cortex on a reach-to-
grasp task in rodents. They used independent component analysis to con-
struct muscle synergies from the 10–12 muscle data sets used to perform
the task. They showed that learning is associated with both changes in the
selection and tuning of primary motor cortex firing rates. These changes in
firing rates were associated with changes in the synergies or with changes
in the amplitudes of the associated synergies. The change in the amplitude
of the synergies is consistent with our findings of increased muscle activa-
tion with changes in movement speed (Corcos et al., 1993). The fact that
changes may occur in the motor cortex during the task that we use is
bolstered by findings that show the motor cortex has been implicated in
motor learning for ballistic but not ramp movements (it would be helpful to
briefly define rapid- and slow-ramp movements here) (Muellbacher, Ziemann,
Boroojerdi, Cohen, & Hallett, 2001).
2. Performance changes across experimental sessions
The model system that we have studied has placed a premium on

learning to move more quickly. In the experiments discussed in the pre-
vious section, it would be fair to ask whether learning has really taken
place. According to Ito (1976) and Brooks and Watts (1988), learned beha-
viors are those that show the following: (1) movements can be improved
through practice, (2) the improvements can be retained over a ‘‘long’’ time
interval even when the behavior is not performed, and (3) the behavior
becomes less variable. After the initial 200 trial practice session that was
discussed in the previous section, subjects practiced an additional six
sessions of 200 trials per session. The practice effects were very similar
to those observed over one session (Corcos et al., 1993). Performance for
the 54 movements over subsequent sessions continued to improve in
terms of increased peak velocity, decreased peak velocity variability,
increased acceleration and deceleration, a proportionately greater increase
in peak acceleration than deceleration, and greater consistency in terminal
location. Collectively, these findings suggest that the task was indeed
learned. In general, subjects increased the activation of both the agonist
muscle and the antagonist muscle and activated the antagonist earlier as
they did when practicing for 1 day. However, one subject did learn to delay
the activation of the antagonist muscle. This means that the antagonist
muscle was activated proportionately later in the movement and suggests
that there are individual differences in how practice can affect patterns of
muscle activation that are learned. The fact that there are individual differ-
ences in how learning occurs is also seen in the reach to grasp task in rodents
in which some animals adjust activation magnitudes of independent compo-
nents and others do not (Kargo & Nitz, 2003).
Since patients with PD move slowly but display relatively normal short-
term motor learning, it is important to know whether the previously dis-
cussed short-term practice changes continue over multiple experimental
sessions. To answer this question, Agostino, Sanes, & Hallett (1996) had
individuals with PD practice movements daily over a 2-week period. Indi-

viduals with PD improved over the first week to the same extent as healthy
individuals. After 2 weeks, there was less overall improvement. Agostino
and colleagues suggest that this provides support for the idea that long-
term practice benefit might involve basal ganglia circuitry that is compro-
mised in PD.
In addition to their interest in early learning, Kargo and Nitz (2004)
studied practice-related changes in motor performance with the goal of
understanding experience-driven changes in cortical circuitry, and
how this was related to changes in patterns of muscle activation. They
studied the performance changes over 12 days. They identified three
patterns of skill improvement in terms of EMG patterns. First, they identi-
fied improved pattern selection that refers to the fact that some animals
quickly learned successful patterns of forelimb and oral movement.
Second, they managed to adapt the ratios of muscle activation to be
more successful. Finally, they reduced trial-to-trial variability in muscle
recruitment patterns. These changes in muscle activation were linked to
changes in the activity in the motor cortex. These changes consisted of
changes in long-term potention synaptogenesis, and movement representa-
tions in the primary motor cortex, which corresponded to training days 3, 7
and 10, respectively.
One important issue in the study of learning is to separate learning from
performance changes (Cahill, McGaugh, & Weinberger, 2001; Seidler et al.,
2002). The study by Kargo and Nitz (2004) used a very elegant control to
show that the changes they observed in the primary motor cortex were due
to learning. They reported that the same population of cells in the motor
cortex that showed changed firing properties for the reach training period
also showed consistent, nonchanging firing properties during overground
running, which is a well-learned skill. Along similar lines, Plautz, Milliken,
and Nudo (2000) showed that when squirrel monkeys simply repeated
movements that did not require skill acquisition, there was no change in
the representation in the motor cortex.
3. Transfer
The study of performance changes at nonpracticed movement distances

addresses the issue of transfer of training, and is therefore additional
evidence that learning has occurred in our single joint task. If we consider
our simple model task of moving between two stationary targets using
movements that require only elbow flexion, subjects might learn to
improve performance in one of the three ways (Gottlieb, Corcos, Jaric, &
Agarwal, 1988). First, they might learn to activate their muscles in a way
that applies only to the movement distance practiced. Second, they might
be able to transfer the pattern of muscle activation learned at one practiced
distance to other distances. Third, they might partially generalize their
learning and show the most improvement at the practiced distance and
less improvement at other distances. In order to determine which way
performance improves, Corcos and colleagues (1993) had subjects first
perform movements over five distances (18, 36, 54, 72, and 90) as a
pretest. They then had subjects practice 200 movements per day for seven
experimental sessions at 54 as discussed in the two previous sections.
Then subjects performed posttest trials at the same original five distances,
Jaric, Corcos, Agarwal, & Gottlieb (1993). One way to consider the extent
to which there is transfer to nonpracticed distances is in terms of Fitts’ law,
which relates movement time to the distance moved and target size.
The linearized form of this equation, assuming constant target size and a
variable distance (D), is:
MT ¼ a þ bD ð1Þ
Practicing movements at one distance could lead to changes in either the

slope of this equation, the intercept of this equation, or both. As might be
expected, there were individual differences between the subjects in the
amount of transfer demonstrated but the most robust finding was that
there was a reduction in the intercept of this relationship (shorter movement
times), and only a small change in the slope. This suggests a generalized
improvement in performance to nonpracticed distances. One hypothesis to
explain this generalized decrease in movement time across movement dis-
tance is to suggest that some of the subjects learned to increase the intensity
of muscle activation in such a way that it generalizes across all distances.
This can be clearly seen in Fig. 1, which is reprinted from Jaric, et al., (1993).
Studies of single degree-of-freedom movements that we have used in our
model system are often criticized for not being typical of real-world move-
ments (Corcos, Gottlieb, & Agarwal, 1989a). As such, it is informative to
determine the extent to which the findings from single degree-of-freedom
movements in humans that we have focused on apply to movement tasks
that involve more than one degree of freedom. The study of multiple
degree-of-freedom tasks allows two additional questions to be addressed,
which cannot be answered when studying single degree-of-freedom tasks.
The first is whether multiple muscles (synergists) that subserve similar
functions are controlled the same way, and whether they change patterns
of activation in a similar way as a result of practice. The second question
concerns how muscle activation patterns change in muscles that tend to
oppose the intended action. For example, as discussed by Shemmell and
Thick Lines: pre-test Angle

Thin Lines: post-test
75°
Velocity
528 °/s
Acceleration
5600 °/s2
Biceps
1.44
Triceps Lateralis
2.5
0 Time (ms) 500
Fig. 1. Kinematic and myoelectric variables for 36, 54 and 72 movements in the pretest
(thick lines) and the posttest (thin lines) for one subject (5) from the practice group.
The figure depicts angle, velocity, acceleration, and EMGs from biceps (inverted) and
the lateral head of triceps. These variables are the mean of approximately ten trials,
aligned on the graph at t = 50 ms at the onset of the biceps EMG.
colleagues (Shemmell et al., 2005), the biceps brachii is a bifunctional

muscle that generates elbow flexion torque, but when the forearm is
neutral, it will have a supination moment (Ettema, Styles, & Kippers,
1998). As such, there are situations in which this supination moment
must be cancelled out by the activation of another muscle such as the

pronator teres.
To address both these questions, Shemmell and colleagues (2005) had
subjects generate torques in eight directions. The eight torques could be
generated by producing flexion, extension, pronation or supination torques
or by producing a combination of the above torques. They recorded muscle
activation from eight muscle groups, and then performed principal compo-
nent (PC) analysis on the EMG waveforms. They first measured subjects at
20 and 40% of their maximum torque. They then had subjects practice at
30% of their maximal torque for 5 days followed by postpractice measures
at 20 and 40% or their maximal torque. They also had a retention test after
4 weeks in which they measured subjects at 30% Maximum Voluntary
Contraction (MVC). If each of the eight muscles is controlled indepen-
dently for every direction, then there would be one PC for each muscle
and each direction (64). At the other end of the spectrum, if there is a
central neural controller that sends a common pattern to each muscle for
each direction and weighs this pattern in a task-dependent fashion, then
there would be only one PC. Shemmell and colleagues confirmed our
findings that the decreases in the target acquisition time were associated
primarily with increases in the rate of muscle recruitment. These increases
in the rate of agonist recruitment were generalized successfully to non-
practiced levels of force and were recalled after the 4-week period of no
practice. The changes that they observed were mainly observed in bifunc-
tional agonist muscles such as both heads of the biceps brachii. The
contribution of the monofunctional muscles was either maintained or
reduced. In addition, they showed that the CNS does simplify the problem
of controlling muscle groups by using a common set of patterns and can
reduce this number further with practice. Before practice, five PCs were
needed to account for the data. After practice, four PCs accounted for the
data. The results support the theory of a central neural controller.
In a second study, Shemmell et al. (2006) had subjects again practice
generating isometric torques in eight directions using combinations of
flexion and extension, pronation and supination. This study included
two different groups: one group practiced at 20% of maximal voluntary
torque and the other group practice at 40%. Both groups showed posi-
tive transfer to nonpracticed torque levels, suggesting that they had a
stable memory for the torque patterns that they were generating. The
key point of this study is that contraction in directions that required
only pronation or supination was initially performed more quickly than
those that required flexion or extension torques. However, improve-
ment in all directions occurred, suggesting common principles of
motor learning that are independent of the particular muscles perform-
ing the task.
The two studies conducted by Shemmell and colleagues show that

motor control and learning principles learned from our single degree-of-
freedom model system apply to more complex motor tasks. Moreover, they
show the extreme versatility the motor system has in learning CNS com-
mands that can generalize to varied distances, to varied levels of force, and
for contractions performed in varied directions.
4. Summary
In conclusion, the time course of practice that we observe with our

single joint system model in humans is similar to that observed by several
other investigators studying tasks that involve many more degrees of free-
dom at both the kinematic, kinetic and neuromuscular level. Collectively
these studies show the following:
(1) Extensive myoelectric and kinematic changes occur in simple

movements or isometric contractions, and these changes generalize
to simple variants of the task such as moving various distances or
generating various levels of torque.
(2) The time course of the performance changes progresses at variable
rates for varied movement parameters, and this may relate to
varying neural substrates controlling different parameters of the
movement.
(3) Performance can continue improvement and thus learning occurs
when there are disruptions in the circuitry of the basal ganglia as in
the case of PD, but the performance changes may not occur for the
same length of time as in healthy individuals.
Such studies suggest that different neural systems underlie early

learning and late learning. Further studies, similar to the sophisticated
studies that have been conducted in the primary motor cortex, are now
required to show the extent to which short-term practice might increase
activity in cerebellar cortex and how long-term practice might decrease
activation in the cerebellum and/or basal ganglia.
Acknowledgments
This research was supported in part by grants from the National Insti-
tutes of Health (R01-NS52318, R01-NS28127, R01-NS40902, R01 AR44388).
References
Agostino, R., Sanes, J. N., & Hallett, M. (1996). Motor skill learning in Parkinson’s
disease. Journal of Neurological Sciences, 139(2), 218–226.
Brooks, V. B., & Watts, S. L. (1988). Adaptive programming of arm movements.
Journal of Motor Behavior, 20(2), 117–132.
Cahill, L., McGaugh, J. L., & Weinberger, N. M. (2001). The neurobiology of learning and
memory: Some reminders to remember. Trends in Neurosciences, 24(10), 578–581.
Corcos, D. M., Gottlieb, G. L., & Agarwal, G. C. (1989a). Does constraining move-
ments, constrain the development of movement theories? Behavioral and Brain
Sciences, 12(2), 237–246.
Corcos, D. M., Gottlieb, G. L., & Agarwal, G. C. (1989b). Organizing principles for
single-joint movements. II. A speed-sensitive strategy. Journal of Neurophysiol-
ogy, 62(2), 358–368.
Corcos, D. M., Gottlieb, G. L., Jaric, S., Cromwell, R. L., & Agarwal, G. C. (1990).
Organizing principles underlying motor skill acquisition. In J. Winters & S. Woo
(Eds.), Multiple muscle systems: Biomechanics and movement organization
(pp. 251–267). New York, NY: Springer Verlag, New York Inc.
Corcos, D. M., Jaric, S., Agarwal, G. C., & Gottlieb, G. L. (1993). Principles for
learning single-joint movements I. Enhanced performance by practice. Experi-
Ettema, G., Styles, G., & Kippers, V. (1998). The moment arm of 23 muscle segments
of the upper limb with varying elbow and forearm positions: implications for
motor control. Human Movement Science, 17, 201–220.
Flament, D., Shapiro, M. B., Kempf, T., & Corcos, D. M. (1999). Time course
and temporal order of changes in movement kinematics during learning
of fast and accurate elbow flexions.Experimental Brain Research, 129(3),
441–450.
Flament, D., Vaillancourt, D. E., Kempf, T., Shannon, K., & Corcos, D. M.
(2003). EMG remains fractionated in Parkinson’s disease, despite practice-
related improvements in performance. Clinical Neurophysiology, 114(12),
2385–2396.
Gottlieb, G. L., Corcos, D. M., Jaric, S., & Agarwal, G. C. (1988). Practice improves
even the simplest movements. Experimental Brain Research, 73, 436–440.
Hore, J., Wild, B., & Diener, H.-C. (1991). Cerebellar dysmetria at the elbow, wrist
and fingers. Journal of Neurophysiology, 65(3), 563–571.
Ito, M. (1976). Adaptive control of reflexes by the cerebellum. In S. Homma (Ed.),
Progress in Brain research, vol. 44, Understanding the stretch reflex, (pp. 435–443)
Elsevier, Amsterdam.
Jaric, S., Corcos, D. M., Agarwal, G. C., & Gottlieb, G. L. (1993). Principles for
learning single-joint movements II. Generalizing a learned behavior. Experimen-
tal Brain Research, 94, 514–521.
Jordan, N., & Sagar, H. J. (1994). The role of the striatum in motor learning:
Dissociations between isometric motor control processes in Parkinson’s disease.
International Journal of Neuroscience, 77, 153–165.
Kargo, W. J., & Nitz, D. A. (2003). Early skill learning is expressed through selection
and tuning of cortically represented muscle synergies. Journal of Neuroscience,
23(35), 11255–11269.
Kargo, W. J., & Nitz, D. A. (2004). Improvements in the signal-to-noise ratio of motor
cortex cells distinguish early versus late phases of motor skill learning. Journal
ofNeuroscience, 24(24), 5560–5569.
Keele, S. W., & Ivry, R. (1990). Does the cerebellum provide a common computation
for diverse tasks? A timing hypothesis. Annals of the New York Academy of
Sciences, 608, 179–207; discussion 207–111.
Kempf, T., Corcos, D. M., & Flament, D. (2001). Time course and temporal order of
changes in movement kinematics during motor learning: effect of joint and
instruction. Experimental Brain Research, 136(3), 295–302.
Laforce, R., Jr. & Doyon, J. (2001). Distinct contribution of the striatum and
cerebellum to motor learning. Brain and Cognition, 45(2), 189–211.
Muellbacher, W., Ziemann, U., Boroojerdi, B., Cohen, L., & Hallett, M. (2001). Role of
the human motor cortex in rapid motor learning. Experimental Brain Research,
136(4), 431–438.
Penhune, V. B., & Doyon, J. (2005). Cerebellum and M1 interaction during early
learning of timed motor sequences. NeuroImage, 26(3), 801–812.
Plautz, E. J., Milliken, G. W., & Nudo, R. J. (2000). Effects of repetitive motor
training on movement representations in adult squirrel monkeys: role of use
versus learning. Neurobiology of Learning and Memory, 74(1), 27–55.
Sanes, J. N. (2003). Neocortical mechanisms in motor learning. Current Opinion
Neurobiology, 13(2), 225–231.
Seidler, R. D., Purushotham, A., Kim, S. G., Ugurbil, K., Willingham, D., & Ashe, J.
(2002). Cerebellum activation associated with performance change but not motor
learning. Science, 296(5575), 2043–2046.
Shemmell, J., Forner, M., Tathem, B., Tresilian, J. R., Riek, S., Barry, B. K., et al.
(2006). Neuromuscular-skeletal constraints on the acquisition of skill in a dis-
crete torque production task. Experimental Brain Research, 175(3), 400–410.
Shemmell, J., Forner, M., Tresilian, J. R., Riek, S., Barry, B. K., & Carson, R. G.
(2005). Neuromuscular adaptation during skill acquisition on a two degree-of-
freedom target-acquisition task: Isometric torque production. Journal of Neuro-
physiology, 94(5), 3046–3057.
Smiley-Oyen, A. L., Worringham, C. J., & Cross, C. L. (2002). Practice effects in
three-dimensional sequential rapid aiming in Parkinson’s disease. Movement
Disorders, 17(6), 1196–1204.
Wild, B., & Corcos, D. M. (1997). Cerebellar hypermetria: reduction in the
early component of the antagonist electromyogram. Movement Disorders,
12(4), 604–607.
Human Learning 189
A Dynamical Framework for Human

Skill Learning
Cyrille Magne1 and J. A. Scott Kelso2
1
Florida Atlantic University; Middle Tennessee State University
2
Florida Atlantic University
1. Introduction
The acquisition of skills constitutes a key part of our everyday life.

Driving, walking, writing, speaking, riding a bicycle, or playing a musical
instrument are just a few examples of skilled behaviors that have been
acquired through practice and are usually retained over very long periods
of time. Skilled behavior is highly organized in both space and time.
The central problem is how such organization or patterning comes about
(Fitts, 1964).
Human skill learning is a long-standing topic in psychology that is
drawing increasing interest in neuroscience, especially since the advent
of neuroimaging. The traditional approach adopted by most researchers
consists of studying a group of subjects considered as novices for the task
to be learned. Most would certainly agree that individuals are not a blank
slate. Indeed, everyone who enters a new learning environment comes with
his or her own cultural background, history, and experience. However,
determining the initial state of each learner is a huge and (for most
investigators of learning) insurmountable challenge. Thus, to equate for
potential individual differences, arbitrary responses are usually selected
for subjects to learn. Then the performances (error rates, reaction times,
etc.) of the subjects are averaged together and plotted across practice
sessions to produce the well-known learning curve. The learning process
as reflected by the learning curve usually looks like a smooth and gradual
improvement with repeated practice. However, the picture is quite differ-
ent when one examines individual learning curves. Typically, one sees
short periods of increase and decrease in performance, as well as sudden
changes in the slope of the curve, rather than a smooth and continuous
190 Cyrille Magne and J. A. Scott Kelso
improvement as displayed by the average learning curve. In sum, at the

individual level, learning resembles a more dynamic process with rather
more structure than that classically shown by the average learning curve.
In recent years, researchers have examined motor skill learning from the
perspective of informationally coupled self-organizing dynamical systems,
or coordination dynamics (e.g., Amazeen, 2002; Fuchs & Jirsa, 2008; Kelso,
1995; Tschacher & Dauwalder, 2003). In this framework, one of the main
concepts is that a learned skill is the product not only of practice but also
of the dynamics of the preexisting capabilities of the learner. In this
chapter, we present an outline of the theory of learning derived from
coordination dynamics. Then, we review some generic principles derived
from empirical work at the behavioral level and explore how these findings
are supported at the level of brain structure and function.
2. A dynamical theory of learning
The initial empirical impetus for the development of Coordination

Dynamics came mainly from the results of an experiment (Kelso, 1984)
involving a bimanual coordination task in which subjects were instructed
to move their index fingers back and forth rhythmically with the same
frequency, using a pacing metronome. This experiment showed that sub-
jects can stably perform two patterns: in-phase (corresponding to the
simultaneous activation of homologous muscles) and antiphase (involving
the simultaneous activation of nonhomologous muscles). Interestingly, the
two modes did not exhibit the same stability.1 When subjects started by
performing antiphase finger oscillations with respect to the metronome
beat, and the frequency of the oscillation was progressively increased, the
subject spontaneously switched from the antiphase to the in-phase pattern
when the frequency passed a given threshold, usually around 2 Hz. In
contrast, when the subjects started in the in-phase mode, they were
able to perform this pattern throughout the whole session, and no such
switching was observed. Thus, the system that is bistable (in-phase and
antiphase) for certain values of the movement frequency can become
1
Stability is a central theoretical concept in coordination dynamics and can be measured
experimentally in various ways (see Kelso et al., 1987; Schoner & Kelso, 1988). Serious
scientific questions surround the issue of ‘‘stability of what?’’ In coordination dynamics,
stability refers to coordinated states defined by collective variables that capture patterns of
behavior. A chief consequence of the bimanual paradigm is that it allowed such collective
states to be defined and their stability quantified.
monostable for others, and the transition from one tendency to another is
not progressive, but abrupt.
A theoretical model of all these phenomena (multiple patterns, pattern
stability, switching, etc.) was first proposed by Haken, Kelso and Bunz
(1985), was further elaborated to include stochastic (Schöner, Haken, &
Kelso, 1986) and symmetry-breaking effects (Kelso, DelColle & Schöner,
1990) and has been extended in numerous ways (see, e.g. Fuchs & Jirsa,
2000; Kelso, Bressler, Buchanan, DeGuzman, Ding, Fuchs, & Holroyd, 1992).
Quantitative predictions have been tested successfully in many experimental
situations, including synchronization or syncopation of a single limb with a
metronome (Kelso et al., 1990), between two different limbs (Kelso et al.,
1992) or between interacting participants (Schmidt et al., 1990; Schmidt and
Lee 1998). This and other dynamical phenomena have also been demon-
strated in speech (e.g., Tuller, Case, Ding, & Kelso, 1994) and visual percep-
tion (e.g., Ditzinger, Stadler, Struber, & Kelso, 2000). In the last 15 years, the
neural correlates underlying this phenomenon have been explored using
EEG (e.g., Mayville, Bressler, Fuchs, & Kelso, 1999), MEG (e.g., Jantzen,
Fuchs, Mayville, Deecke, & Kelso, 2001; Mayville et al, 2001), and fMRI
methods (e.g., Jantzen, Steinberg, & Kelso, 2004, 2005). One question of
interest within the context of coordination dynamics is the degree to
which the intrinsically less stable patterns may be stabilized through learn-
ing. The hypothesis that learning can be captured in terms of coordination
dynamics is supported by both theoretical (Schöner & Kelso, 1988; Schöner,
Zanone, & Kelso, 1992; Zanone & Kelso, 1994) and empirical research on
bimanual coordination conducted at the behavioral (Kelso & Zanone, 2002;
Zanone & Kelso, 1992, 1997; see also Kostrubiec and Zanone, 2002; Schmidt,
Treffner, Shaw, & Turvey, 1992; Swinnen, Dounskaia, Walter, & Serrien,
1997; Swinnen, Walter, Lee, & Serrien, 1993; Temprado and Swinnen, 2005;
Temprado, Monno, Zanone, & Kelso, 2002) and neural levels (Jantzen et al.,
2001; Jantzen, Steinberg, & Kelso, 2002).
A key concept of coordination dynamics is that each individual enters
the learning environment not as a blank slate but with preferences and
preexisting capabilities. In the language of coordination dynamics, such
predispositions and susceptibilities are referred to as intrinsic dynamics.
Note that this concept does not refer necessarily to innate mechanisms, but
rather to the set of capacities that exist at the time the new task is to be
learned. The constraints imposed by the learning environment, the task to
be learned, the learner’s intention, and so on. constitute a source of beha-
vioral or functional information. Functional information and intrinsic
dynamics are complementary aspects of coordination dynamics (Kelso,
1995; Kelso and Engstrøm, 2006). Functional information is measured by
the same type of coordination variables or order parameters (Haken, 1983)
that are used to characterize the spontaneous coordination patterns that
constitute the intrinsic dynamics. And learning, in the framework of coor-

dination dynamics, is the process by which functional information from the
environment specifying a pattern to be learned becomes memorized.
A coordination pattern is learned to the extent that the intrinsic dynamics
are modified in the direction of the to-be-learned pattern. Once learning is
achieved, the memorized pattern becomes an attractor of the coordination
dynamics (see, e.g. Schöner, Zanone & Kelso, 1992).
Taking into consideration the learner’s intrinsic dynamics prior to learn-
ing is an issue that has long been emphasized by learning theories, but
hardly ever addressed experimentally. The reason is certainly that evaluat-
ing this preexisting repertoire prior to learning and during the evolution of
the learning process remains a very difficult and time-consuming exercise.
Coordination dynamics confronts the problem by providing an operational
means to identify patterns and their dynamics using bimanual phase transi-
tions as an entry point.
3. Behavioral studies
A study by Zanone and Kelso (1992) was the first to directly address the
problem of evaluating the intrinsic dynamics of the learners. To this end,
they carried out a series of experiments using the bimanual coordination
task. The advantage of this task is that it exhibits only two spontaneously
stable patterns of coordination before learning (in-phase and antiphase).
Participants were required to learn a new relative phase, at 90, which
qualifies as a coordination pattern midway between the two stable coordi-
nation states previously mentioned (0 or in-phase and 180 or antiphase).
In addition, the participants’ spontaneous coordination tendencies were
probed before, during, and after learning for the entire 0–180 interval,
thus spanning the 90 relative phase to be learned. Several important points
were highlighted by these experiments. Zanone and Kelso demonstrated
that the stabilization of a new phasing pattern increased the number of
available patterns of coordination, but could also destabilize preexisting
ones, at least temporarily. Moreover, although the stability of the new
phasing pattern could be increased with learning, the degree of stability
of the new pattern to be learned was dependent on whether the novel task
requirement competed or cooperated with the participant’s preexisting
coordination repertoire.
In addition, several findings suggested that learning may take two dif-
ferent forms depending on the level of competition between the intrinsic
dynamics of the subject and the task to be learned (Kostrubiec and Zanone,
2002; Kostrubiec, Tallet, & Zanone, 2006; Zanone and Kostrubiec, 2004;
Zanone & Kelso, 1997). When the competition is relatively weak, learning
leads to a shift of a preexisting attractor to the position required by the task

to be learned. However, when the competition is stronger, because the
distance between the existing attractors and the required value is too far,
this simple shift may not be possible. Consequently, the learning mechan-
ism takes the form of a bifurcation or phase transition in which a new
attractor is created.
Now, notice also that the introduction of a new pattern (90) breaks the
symmetry of the original pattern dynamics (0 and 180). Remarkably,
Zanone and Kelso (1997) showed that not only were the participants able
to perform the 90 pattern after learning, but also that they were able
to perform the 270 pattern (corresponding to 90), although they had
not practiced this relative phase pattern at all. Thus, learning not only
stabilizes the required new phase relationship, but also can spontaneously
stabilize unpracticed phasing patterns. It has been hypothesized that such a
transfer may reflect the preservation of symmetry within the coordination
dynamic pattern. Later, Kelso and Zanone (2002) showed that a new phas-
ing relationship learned with one pair of limbs could be transferred to a
different limb pair without any training. In sum, the results of these experi-
ments tell us that though transfer implies a symmetry breaking of the
original dynamics, it seems that the layout of attractors in the whole system
is modified so that symmetry can be preserved.
Recently, another issue that has been addressed is the attentional
demands associated with learning-related changes in coordination
dynamics. Indeed, it has long been argued that learning a new skill through
practice is accompanied by a change in the amount of information that can
be processed simultaneously at the central level (Fitts, and Posner, 1967;
Schneider and Chein, 2003; Shiffrin and Schneider, 1977). For instance,
Shiffrin and Schneider (1977) argued that some resources at the central
level are dedicated to controlled processing. This controlled processing
plays an important role in the development of expertise, and becomes less
essential as skilled and automatic performances are being acquired.
Temprado et al. (2002) investigated the coevolution of bimanual coordi-
nation dynamics and attentional demand with practice using a dual-task
paradigm. Participants were required to execute a rhythmic antiphase
bimanual coordination pattern while performing a reaction time task in
which they had to depress two buttons simultaneously with their feet as
soon as they heard an auditory cue. A pretest, an intermediate test, and a
posttest determined the individual spontaneous coordination pattern and
critical transition frequency. During the training phase, the participants
were trained to maintain the antiphase coordination pattern at the critical
transition frequency. A retention session was performed 7 days after the
end of the training session. Results showed that not only the number of
transitions decreased following practice, but also that participants were
able to maintain this pattern at a higher frequency in the posttest than in

the pretest. In addition, the analysis of the reaction time during the acquisi-
tion and the retention sessions suggested that participants were able to
maintain a higher degree of stability at a lower attentional cost. However,
the fact that this was observed only in the retention session indicated that
pattern stability and attentional demands do not necessarily change at the
same rate. Although this conclusion may seem trivial for cognitive psychol-
ogists, the nature of the underlying mechanisms remains unclear.
4. Does destabilization of brain activity patterns accompany

changes in learning?
During the last 15 years, an increasing number of studies have been

aimed at studying brain changes following learning and practice using
different neuroimaging methods (TEP, fMRI, EEG, MEG). Aside from a
concentration of work on motor skill learning, the literature spans a wide
range of tasks and paradigms such as verbal learning (e.g., Andreasen et al.,
1995), mirror reading (e.g., Kassubek, Schmidtke, Kimmig, Lücking, &
Greenlee, 2001; Poldrack, Desmond, Glover, & Gabrieli, 1998), artificial
grammar learning (e.g., Fletcher, Buchel, Josephs, Friston, & Dolan,
1999), and abstract design learning (e.g., Petersson, Elfgren, & Ingvar,
1999). In addition, different strategies, mainly two, have been adopted to
track functional brain modifications associated with learning. In the cross-
sectional approach, the brain activity of individuals with varying levels of
expertise is compared while they perform a given task. Though this
approach presents the advantage of allowing the identification of brain
activations associated with long-term practice, it also presents the disad-
vantage of being between-subjects in nature, thus potentially introducing
variability in the observations due to intrinsic differences between indivi-
duals. In contrast, with a longitudinal approach, the same individuals are
examined multiple times over the course of learning. Since its within-
subject nature offers optimal statistical power to identify learning-related
brain changes over the practice sessions, it is not surprising that this latter
approach is the one now mainly used.
What are the main results of neuroimaging studies of learning? Not
surprisingly, behavioral improvement due to practice correlates with
changes in the functional neuroanatomy of the task being learned. This
brain reorganization has been shown to take several forms. In some cases,
when comparing the brain activity pre- and postpractice, an increase in
task-related activation is observed (e.g., Grafton et al 1992; Hazeltine,
Grafton, & Ivry, 1997; Honda et al 1998; Iacoboni, Woods, & Mazziotta,
1996; Rauch et al 1997; Schwartz, Maquet, & Frith, 2002), but in others a
decrease occurs (e.g., Bush et al 1998; Haier et al 1992; Hempel et al 2004;

Kassubek et al 2001; Landau, Schumacher, Garavan, Druzgal, & D’Esposito,
2004; Schiltz et al 1999; Tomasi, Ernst, Caparelli, & Chang, 2004). In a
majority of studies, however, both an increase in certain areas and a
decrease in others are observed (e.g., Raichle et al 1994; Fletcher et al
1999; Frutiger et al 2000; Sakai et al 1998; Debaere, Wenderoth, Sunaert,
Van Hecke, & Swinnen, 2004; Petersson et al 1999; Graydon,
Friston, Thomas, Brooks, & Menon, 2005; Toni, Krams, Turner, & Passing-
ham, 1998; Jenkins, Brooks, Nixon, Frackowiak, & Passingham, 1994;
Parsons, Harrington, & Rao, 2005; Bischoff-Grethe, Goedert, Willingham,
& Grafton, 2004; Willingham, Salidis, & Gabrieli, 2002; Beauchamp, Dagher,
Aston, & Doyon, 2003; Jansma, Ramsey, Slagter, & Kahn, 2001; Shadmehr &
Holcomb, 1997; Andreasen et al 1995; Poldrack & Gabrieli, 2001; Poldrack
et al 1998). Potentially, this mixed pattern could reflect either a redistribu-
tion of the activation levels across a network of brain areas involved in
learning the task or a complete reorganization (Kelly and Garavan, 2005).
New brain areas may be recruited by the end of learning, while some brain
areas that were active at the beginning of learning are no longer involved.
In fact, because functional neuroimaging studies mainly use thresholding
analysis methods, it is rather difficult to determine whether the changes in
activation before and after training reflect a redistribution of the activation
level of brain areas of the same network, or a complete reorganization of
the brain areas involved in the execution of the task. Intriguingly, these two
possible ways of observing a mixed pattern of brain changes make an
interesting parallel in light of the findings from coordination dynamics
showing that learning can take two routes: either a modification of an
existing attractor toward the task to be learned (the ‘‘shift’’ route) or a
modification of the intrinsic dynamics by creating a new attractor (the
‘‘bifurcation’’ or phase transition route). The former path may reflect a
redistribution of activity in an existing network and the latter an entire
reorganization of the brain.
Trying to unify the results of the literature is a very difficult exercise due
to the diversity of research perspectives within which the studies are
conducted. For instance, whether learning is reflected by an increase, a
decrease, or a reorganization of brain activity seems to depend highly on
the task to be learned. Chein and Schneider (2005) recently performed a
meta-analysis over 29 brain imaging studies in order to seek evidence for
the existence of general brain mechanisms involved in learning. Across the
studies, they found a consistent network of distributed brain areas that
display a decrease in activity following practice, regardless of the task to be
learned. This network includes prefrontal, medial frontal, posterior parie-
tal, occipitotemporal, and cerebellar areas that have been associated with
attention, retrieval, and monitoring in several studies (for a review, see
Kelly and Garavan, 2005). Chein and Schneider proposed that the
decreased activity in this distributed network reflects a reduced involve-
ment of a domain-general control system as task performance becomes
more efficient and automatic with practice.
Interestingly, a similar shift in brain activation has been observed in
several studies where subjects switch from syncopation to synchronization
mode. As previously mentioned, when subjects perform syncopated move-
ments either with external stimuli or between two limbs, and the frequency
of the movements is increased, they ultimately switch to a synchronization
mode. Both synchronization and syncopation modes involve a network
including contralateral sensorimotor areas as well as the cerebellum
(Jantzen et al., 2002; Jantzen et al., 2004; Mayville, Jantzen, Fuchs, Steinberg, &
Kelso, 2002). However, syncopating requires the recruitment of additional
regions in the cerebellum and another network including the basal ganglia,
premotor, supplementary motor, and prefrontal areas. It has been hypothe-
sized that this difference in brain activations between the synchronization
and syncopation modes reflects different strategies for performing the
two coordination patterns (Mayville et al., 2002). The synchronization pattern
is performed relatively automatically, with little planning and attention
required, whereas syncopation may involve the planning and execution of
independent movements on each cycle. In general, the results of these
studies may reflect a natural tendency to switch to a less cognitively
demanding and more automatic mode (Sakai, Hikosaka, & Nakamura, 2004).
Are the brain areas involved in the performance of the task to be learned
the only ones to undergo reorganization? Remember that at the behavioral
level, it has been shown that the entire attractor layout changes with learning.
A recent study suggests that this may also be the case in the brain. Jantzen
et al. (2002) investigated how short-term behavioral practice alone affects
intrinsic differences in neural activity between synchronization and syncopa-
tion coordination modes. To this end, in a prepractice session, subjects were
scanned while either synchronizing or syncopating with an external auditory
stimulus at 1.25 Hz. Then subjects practiced the syncopation mode during
four sessions. After practice, they were scanned again while performing the
synchronization and syncopation patterns. Results revealed that practicing
the more difficult syncopation pattern was associated with a reduced activa-
tion in several brain areas that could be linked to a decrease in attentional
demand. But more interestingly, after practicing the syncopation mode, addi-
tional brain activations in both cortical and subcortical areas were observed
for synchronization! Thus practicing the syncopation mode not only had the
effect of modifying the brain areas involved in this task, but also modified the
activation pattern involved in the unpracticed synchronization mode. In
accordance with these results, Rémy et al. (2008) showed that the acquisition
of a new complex coordination pattern (90) was not only correlated with
learning-related changes in brain activations, but also influenced the brain

activations associated with the non-trained, preferred, in-phase coordination
pattern. This study nicely demonstrated that during learning, like the beha-
vioral dynamics, the entire brain dynamics are modified.
5. Conclusion
Trying to find a global trend in the literature dealing with brain changes
associated with learning is a difficult exercise. There is certainly no doubt
that different learning mechanisms exist and different brain areas are
involved in them. Kelly and Garavan (2005) proposed that there are differ-
ences in the neural mechanisms of brain plasticity and cortical representa-
tions associated with practicing a motor/sensory task or a ‘‘higher-level’’
cognitive task. Motor/sensory tasks involve mainly the primary motor and
sensory areas while ‘‘higher-level’’ cognitive tasks involve a larger number of
more distributed brain areas. Although an increase in brain activation reflect-
ing an expansion of the neural representation may be at work in the former, a
decrease in activation reflecting increased efficiency in the neural commu-
nication between different areas of the network may occur in the latter.
Possible methodological issues have also been identified to explain the
discrepancy in the results of the literature, especially related to performance
and time. For instance, a decrease in reaction times is often observed as the
task becomes easier as learning progresses. Several studies have shown that
the level of neural activation is sensitive to the time spent on the task
(D’Esposito et al, 1997). Also, because being in an MRI scanner can be
quite stressful, especially for novice experimental participants, changes in
brain activation could occur across the practice sessions, not because of
learning, but rather due to changes in anxiety or head motion. Another issue
related to time is the delay between the scanning sessions. Most of the
longitudinal studies involve scanning the learners twice or more. However,
in some studies, the sessions are separated by several days, whereas in
others, they are spaced by a couple of weeks. It is thus reasonable to assume
that these different time frames do not necessarily capture the learning-
related brain modifications at the same stages of the learning process.
Coordination dynamics stresses the importance of identifying preexist-
ing tendencies associated with an individual’s history and past experiences
in order to understand learning as a process (as dynamics, we would say).
This research demonstrated that each individual enters the learning situa-
tion with his or her own predispositions and preferred patterns of behavior
(denoted ‘‘intrinsic dynamics’’ due to their (measured) stability properties)
and that the relationship between new information (from the environment,
a task to be learned, etc.) and intrinsic dynamics determines the nature and
rate of learning in the individual. On the one hand, grouping individuals

together without attention to such intrinsic factors hides the basic nature
of the learning process. On the other hand, grouping individuals who share
similar intrinsic dynamics was able to reveal the cooperative and compe-
titive mechanisms that underlie the learning process.
There is now much interest in understanding the intrinsic dynamics of the
brain prior to the introduction of a novel task (e.g., Fox et al., 2005), the
brain’s dark energy (Raichle, 2006). This intrinsic organization consists of
correlated networks in the absence of overt task performance and provides a
critical context within which to understand brain function. Such a change in
methodology that focuses on identifying the intrinsic organization of brain
networks in an individual—analogous to personalized medicine—may usher
in an entirely new approach to understanding the brain dynamics of learning.
Acknowledgments
Much of the research reported herein was supported by grants from the
US Office of Naval Research N00014-05-1-0117, NIMH Grant MH42900 and
NINDS Grant NS48229.
References
Amazeen, P. G. (2002). Is dynamics the content of a generalized motor program for

rhythmic interlimb coordination? Journal of Motor Behavior, 34, 233–251.
Andreasen, N. C., O’Leary, D. S., Arndt, S., Cizadlo, T., Rezai, K., Watkins, G. L., et al.
(1995). PET studies of memory I: Novel and practiced free recall of complex
narratives. NeuroImage, 2, 284–295.
Beauchamp, M. H., Dagher, A., Aston, J. A., & Doyon, J. (2003). Dynamic functional
changes associated with cognitive skill learning of an adapted version of the
Tower of London task. NeuroImage, 20, 1649–1660.
Bischoff-Grethe, A., Goedert, K. M., Willingham, D. T., & Grafton, S. T. (2004).
Neural substrates of response-based sequence learning using fMRI. Journal of
Cognitive Neuroscience, 16, 127–38.
Bush, G., Whalen, P. J., Rosen, B. R., Jenike, M. A., McInerney, S. C., & Rauch, S. L.
(1998). The counting Stroop: An interference task specialized for functional neuroi-
maging—validation study with functional MRI. Human Brain Mapping, 6, 270–282.
Chein, J. M., & Schneider, W. (2005). Neuroimaging studies of practice-related
change: fMRI and meta-analytic evidence of a domain-general control network
for learning. Cognitive Brain Research 25(3), 607–623.
D’Esposito, M., Zarahn, E., Aguirre, G. K., Shin, R. K., Auerbach, P., & Detre, J. A.
(1997). The effect of pacing of experimental stimuli on observed functional MRI
activity. NeuroImage, 6, 113–121.
Debaere, F., Wenderoth, N., Sunaert, S., Van Hecke, P., & Swinnen, S. P. (2004).
Changes in brain activation during the acquisition of a new bimanual coordina-
tion task. Neuropsychologia, 42, 855–867.
Ditzinger, T., Stadler, M., Struber, D., & Kelso, J. A. S. (2000). Noise improves
3D-perception: Stochastic resonance and other impacts of noise on the percep-
tion of autostereograms. Physical Review E, 62, 2566–2575.
Fitts, P. M. (1964). Perceptual-motor skill learning. In A. W. Melton (Ed.),
Categories of Human Learning (pp. 243–285). San Diego, CA: Academic Press.
Fitts, P. M., & Posner, M. I. (1967). Learning and skilled performance in human
performance, Belmont CA: Brock-Cole.
Fox, M. D., Snyder, A. Z., Vincent, J. L., Corbetta, M., Van essen, D. C., & Raichle, M. E.
(2005). The human brain is intrinsically organized into dynamic, anticorrelated
functional networks. Proceedings of the National Academy of Sciences, 102,
9673–9678.
Fletcher, P., Buchel, C., Josephs, O., Friston, K., & Dolan, R. (1999). Learning-related
neuronal responses in prefrontal cortex studied with functional neuroimaging.
Cerebral Cortex, 9, 168–178.
Frutiger, S. A., Strother, S. C., Anderson, JR., Sidtis, J. J., Arnold, J. B., & Rottenberg, D.
A. (2000). Multivariate predictive relationship between kinematic and functional
activation patterns in a PET study of visuomotor learning. NeuroImage, 12, 515–527.
Fuchs, A., & Jirsa, V. K. (2000). The HKB model revisited: How varying the degree of
symmetry controls dynamics. Human Movement Science, 19, 425–449.
Fuchs, A., & Jirsa, V. K. (2008). (eds.). Coordination: Neural, Behavioral and Social
Dynamics, Heidleberg: Springer.
Grafton, S. T., Mazziotta, J. C., Presty, S., Friston, K. J., Frackowiak, R. S. J., &
Phelps, M. E. (1992). Functional anatomy of human procedural learning deter-
mined with regional cerebral blood flow and PET. Journal of Neuroscience, 12,
2542–2548.
Graydon, F. X., Friston, k. J., Thomas, C. G., Brooks, V. B., Menon, R. S. (2005).
Learning related fMRI activation associated with a rotational visuo-motor trans-
formation. Brain Research. Cognitive Brain Research, 22, 373–383.
Haier, R. J., Siegel, B. V., Jr., MacLachlan, A., Soderling, E., Lottenberg, S., Buchs-
baum, M. S. (1992). Regional glucose metabolic changes after learning a complex
visuospatial/motor task: A positron emission tomographic study. Brain Research,
570, 134–143.
Haken, H. (1983). Synergetics: An introduction, Berlin: Springer.
Haken, H., Kelso, J. A. S., & Bunz, H. (1985). A theoretical model of phase transitions
in human hand movements. Biological Cybernetics, 51, 347–356.
Hazeltine, E., Grafton, S. T., & Ivry, R. (1997). Attention and stimulus characteristic
determine the locus of motor-sequence encoding: A PET study. Brain, 120, 123–140.
Hempel, A., Giesel, F. L., Caraballo, N. M. G., Amann, M., Meyer, H., Wustenberg, T.,
et al. (2004). Plasticity of cortical activation related to working memory during
training. The American Journal of Psychiatry, 161, 745–747.
Honda, M., Deiber, M. P., Ibanez, V., Pascual-Leone, A., Zhuang, P., & Hallett, M.
(1998). Dynamic cortical involvement in implicit and explicit motor sequence
learning. A PET study. Brain, 121, 2159–2173.
Iacoboni, M., Woods, R. P., & Mazziotta, J. C. (1996). Brain–behavior relationships:

Evidence from practice effects in spatial stimulus-response compatibility.
Journal of Neurophysiology, 76, 321–331.
Jansma, J. M., Ramsey, N. F., Slagter, H. A., & Kahn, R. S. (2001). Functional
anatomical correlates of controlled and automatic processing. Journal of Cogni-
tive Neuroscience, 13, 730–743.
Jantzen, K. J., Fuchs, A., Mayville, J. M., Deecke, L., & Kelso, J. A. S. (2001).
Neuromagnetic activity in alpha and beta bands reflects learning-induced
increases in coordinative stability. Clinical Neurophysiology, 112, 1685–1697.
Jantzen, K. J., Steinberg, F. L., Kelso, J. A. S. (2004). Brain Networks Underlying
Human Timing Behavior Are Influenced By Prior Context. Proceedings of the
National Academy of Science of the United States, 101, 6815–6820.
Jantzen, K. J., Steinberg, F. L., & Kelso, J. A. S. (2005). Functional MRI reveals the
existence of modality and coordination dependent timing networks. NeuroImage,
25, 1031–1042.
Jantzen, K. J., Steinberg, F. L., & Kelso, J. A. S. (2002). Practice-dependent modulation
of neural activity during human sensorimotor coordination: A functional magnetic
resonance imaging study. Neuroscience Letters, 332, 205–209.
Jenkins, I. H., Brooks, D. J., Nixon, P. D., Frackowiak, R. S., & Passingham, R. E.
(1994). Motor sequence learning: A study with positron emission tomography.
The Journal of Neuroscience, 14, 3775–3790.
Kassubek, J., Schmidtke, K., Kimmig, H., Lücking, C. H., & Greenlee, M. W. (2001).
Changes in cortical activation during mirror reading before and after training: An
FMRI study of procedural learning. Brain Research. Cognitive Brain Research,
10(3), 207–217.
Kelly, C., & Garavan, H. (2005). Human functional neuroimaging of brain changes
associated with practice. Cerebral Cortex, 15, 1089–1102.
Kelso, J. A. S., Schöner, G., Scholz, J. P., & Halken, H. (1987). Phase-locked modes,
phase transitions and component oscillators in coordinated biological motion.
Physica Scripta, 35, 79–87.
Kelso J. A. S., Bressler, S. L., Buchanan, S., DeGuzman, G. C., Ding, M., Fuchs, A., &
Holroyd, T. (1992). A phase transition in human brain and behavior. Physics
Letters A, 169, 134–144.
Kelso, J. A. S., & Engstrom, D. A. (2006). The complementary nature, Cambridge,
MA: MIT Press.
Kelso, J. A. S., & Zanone, P. G. (2002). Coordination dynamics of learning and
transfer across different effector systems. Journal of Experimental Psychology:
Human Perception and Performance, 28, 776–797.
Kelso, J. A. S., DelColle, J. D., & Schöner, G. (1990). Action–perception as a pattern
formation process In M. Jeannerod (Ed.), Attention and Performance XIII
(pp. 139–169). Hillsdale, NJ: Erlbaum.
Kelso, J. A. S., Bressler, S. L., Buchanan, S., DeGuzman, G. C., Ding, M., Fuchs, A.,
et al. (1992). A phase transition in human brain and behavior. Physics Letters. A,
169, 134–144.
Kelso, J. A. S. (1984). Phase transitions and critical behavior in human bimanual
coordination. The American Journal of Physiology, 246, R1000–R1004.
Kelso, J. A. S. (1995). Dynamic Patterns: The Self-Organization of Brain and

Behavior, Cambridge, MA: MIT Press.
Kostrubiec, V., & Zanone, P. G. (2002). Memory dynamics: Distance between the
new task and existing behavioral patterns affects learning and interference in
bimanual coordination. Neuroscience Letters, 331, 193–197.
Kostrubiec, V., Tallet, J., & Zanone, P. -.G. (2006). How a new behavioral pattern is
stabilized with learning determines its persistence and flexibility in memory.
Experimental Brain Research, 170(2), 238–244.
Landau, S. M., Schumacher, E. H., Garavan, H., Druzgal, T. J., & D’Esposito, M.
(2004). A functional MRI study of the influence of practice on component
processes of working memory. NeuroImage, 22, 211–221.
Mayville, J. M., Fuchs, A., Ding, M., Cheyne, D., Deecke, L., & Kelso, J. A. S. (2001).
Event-related changes in neuromagnetic activity associated with syncopation and
synchronization tasks. Human Brain Mapping, 14, 65–80.
Mayville, J. M., Bressler, S. L., Fuchs, A., & Kelso, J. A. S. (1999). Spatiotemporal
reorganization of electrical activity in the human brain associated with a timing
transition in rhythmic auditory–motor coordination. Experimental Brain
Research, 127, 371–381.
Mayville, J. M., Jantzen, K. J., Fuchs, A., Steinberg, F., & Kelso, J. A. S. (2002).
Cortical and subcortical networks underlying syncopated and synchronized coor-
dination revealed using fMRI. Human Brain Mapping, 17, 214–219.
Parsons, M. W., Harrington, D. L., & Rao, S. M. (2005). Distinct neural systems
underlie learning visuomotor and spatial representations of motor skills. Human
Brain Mapping, 24, 229–247.
Petersson, K. M., Elfgren, C., & Ingvar, M. (1999). Dynamic changes in the functional
anatomy of the human brain during recall of abstract designs related to practice.
Poldrack, R. A., Desmond, J. E., Glover, G. H., & Gabrieli, J. D. E. (1998). The neural
basis of visual skill learning: An fMRI study of mirror-reading. Cereb Cortex, 8, 1–10.
Poldrack, R. A., & Gabrieli, J. D. (2001). Characterizing the neural mechanisms of skill
learning and repetition priming: Evidence from mirror reading. Brain, 124, 67–82.
Raichle, M. E. (2006). The brain’s dark energy. Science, 314, 1249–1250.
Raichle, M. E., Fiez, J. A., Videen, T. O., MacLeod, A. M., Pardo, J. V., Fox, P. T., et al.
(1994). Practice-related changes in human brain functional anatomy during non-
motor learning. Cereb Cortex, 4, 8–26.
Rauch, S. L., Whalen, P. J., Savage, C. R., Curran, T., Kendrick, A., Brown, H. D., et al,
(1997). Striatal recruitment during an implicit sequence learning task as measured
by functional magnetic resonance imaging. Human Brain Mapping, 5, 124–132.
Rémy, F., Wenderoth, N., Lipkens, K., & Swinnen, S. P. (2008). Acquisition of a new
bimanual coordination pattern modulates the cerebral activations elicited by an
intrinsic pattern: An fMRI study. Cortex, 44(5), 482–493.
Sakai, K., Hikosaka, O., Miyauchi, S., Takino, R., Sasaki, Y., & Putz, B. (1998).
Transition of brain activation from frontal to parietal areas in visuomotor
sequence learning. The Journal of Neuroscience, 18, 1827–1840.
Sakai, K., Hikosaka, O., & Nakamura, K. (2004). Emergence of rhythm during motor
learning. Trends in Cognitive Sciences, 8, 547–553.
Schiltz, C., Bodart, J. M., Dubois, S., Dejardin, S., Michel, C., Roucoux, A., et al.
(1999). Neuronal mechanisms of perceptual learning: Changes in human brain
activity with training in orientation discrimination. NeuroImage, 9, 46–62.
Schmidt, R. C., Carello, C., & Turvey, M. T. (1990). Phase transitions and critical fluctua-
tions in the visual coordination of rhythmic movements between people. Journal of
Experimental Psychology: Human Perception and Performance, 16, 227–247.
Schmidt, R. C., Treffner, P. J., Shaw, B. K., & Turvey, M. T. (1992) Dynamical aspects
of learning an interlimb rhythmic movement pattern. Journal of Motor Behavior,
21, 122–144.
Schmidt, R. C. Bienvenu, M., Fitzpatrick, P. A., & Amazeen, P. G. (1998). A compar-
ison of within- and between-person coordination: Coordination breakdowns and
coupling strength. Journal of Experimental Psychology: Human Perception and
Performance, 24, 884–900.
Schneider, W., & Chein, J. M. (2003). Controlled & automatic processing: From
mechanisms to biology. Cognitive Science, 27, 525–559.
Schöner, G., Haken, H., & Kelso, J. A. S. (1986). A stochastic theory of phase
transitions in human hand movement. Biological Cybernetics, 53, 247–257.
Schöner, G., & Kelso, J. A. S. (1988). A synergetic theory of environmentally-
specified and learned patterns of movement coordination. Biological Cyber-
netics, 58, 71–80.
Schöner, G., Zanone, P. G., & Kelso, J. A. S. (1992). Learning as change of coordina-
tion dynamics: Theory and experiment. Journal of Motor Behavior, 24, 29–48.
Schwartz, S., Maquet, P., & Frith, C. (2002). Neural correlates of perceptual learning:
A functional MRI study of visual texture discrimination. Proceedings of the
National Academy of Science of the United States, 99, 17137–17142.
Shadmehr, R., & Holcomb, H. H. (1997). Neural correlates of motor memory con-
solidation. Science, 277, 821–825.
Shiffrin, R. M., & Schneider, W. (1977). Controlled and automatic human informa-
tion processing: II. Perceptual learning, automatic attending and a general theory.
Psychological Review, 84(2), 127–190.
Swinnen, S. P., Walter, C. B., Lee, T. D., & Serrien, D. J. (1993). Acquiring bimanual
skills: Contrasting forms of information feedback for interlimb decoupling. Journal
of Experimental Psychology. Learning, Memory, and Cognition, 19, 1328–1344.
Swinnen, S. P., Dounskaia, N., Walter, C. B., & Serrien, D. (1997a). preferred and
induced coordination modes during the acquisition of bimanual movements with
a 2:1 frequency ratio. Journal of Experimental Psychology. Human Perception
and Performance, 1087–1110.
Temprado, J. J., & Swinnen, S. P. (2005). Dynamics of learning and transfer of
muscular and spatial relative phase in bimanual coordination: Evidence for
abstract directional codes. Experimental Brain Research, 160, 180–188.
Temprado, J. J., Monno, A., Zanone, P. G., & Kelso, J. A. S. (2002). Attentional
demands reflect learning-induced alterations of bimanual coordination dynamics.
Tomasi, D., Ernst, T., Caparelli, E. C., & Chang, L. (2004). Practice-induced changes
of brain function during visual attention: A parametric fMRI study at 4 tesla.
NeuroImage, 23(4), 1414–1421.
Toni, I., Krams, M., Turner, R., & Passingham, R. E. (1998). The time course of
changes during motor sequence learning: A whole-brain fMRI study. NeuroImage,
8, 50–61.
Tschacher, W., & Dauwalder, J. P. (2003). The Dynamical systems approach to
cognition: Concepts and empirical paradigms based on self-organization,
embodiment and coordination dynamics, Singapore: World Scientific.
Tuller, B., Case, P., Ding, M., & Kelso, J. A. S. (1994). The nonlinear dynamics of
speech categorization. Journal of Experimental Psychology: Human Perception
and Performance, 20, 1–16.
Willingham, D. B., Salidis, J., & Gabrieli, J. D. (2002). Direct comparison of neural
systems mediating conscious and unconscious skill learning. Journal of Neuro-
physiology, 88, 1451–1460.
Zanone, P. G., & Kelso, J. A. S. (1997). The coordination dynamics of learning and
transfer: Collective and component levels. Journal of Experimental Psychology:
Human Perception and Performance, 23, 1454–1480.
Zanone, P. G., & Kelso, J. A. S. (1994). The coordination dynamics of learning:
Theoretical structure and experimental agenda. In S. Swinnen, H. Heuer,
J. Massion & P. Casaer (Eds.), Interlimb coordination: Neural dynamical and
cognitive constraints (pp. 461–490). San Diego: Academic Press.
Zanone, P. G., & Kelso, J. A. S. (1992). The evolution of behavioral attractors with
learning: Nonequilibrium phase transitions. Journal of Experimental Psychol-
ogy: Human Perception and Performance, 18(2), 403–421.
Zanone, P. G., & Kostrubiec, V. (2004). Searching for (dynamical) principles of
learning. In V. K. Jirsa, J. A. S. Kelso (Eds.), Coordination dynamics: Issues
and trends (pp. 57–89). New-York: Springer.
Part IV
Animal Model Systems

Human Learning 207
Animal Models of Behavioral and

Neural Plasticity
J. Steven de Belle
University of Nevada, Las Vegas
Perhaps no other scientific endeavor generates more curiosity or

absorbs a broader spectrum of research effort than the study of learning
and memory. Philosophers, health practitioners, computer scientists, and
business executives alike have interests in how their particular ‘‘systems’’
respond in the face of repeated exposure to various aspects of the environ-
ment. The preceding sections of this volume provide theoretical, experi-
mental, and applied perspectives on learning and memory, cognitive
neuroscience, and motor learning in ‘‘human’’ systems. As subjects of
these accounts, we humans are drawn to them because they offer explana-
tions for recurrent aspects of our own experiences, our own behavior, and
possibly even clues about who we are as individuals.
For many of the behavioral phenomena discussed in these preceding
chapters, the known neural underpinnings are explained in some detail.
However, several important questions remain unanswered, such as the
following: What is the neural basis of learning? What is the physical
essence of memory? How and where are memories made and stored?
How and under what circumstances are memories retrieved? In spite of
an overwhelming interest in how a human brain solves these questions,
most of the relevant neural mechanisms are inaccessible by noninvasive
experimental techniques currently used to study human subjects. We are
thus forced to look elsewhere for help.
The goal of this final section of the book is to introduce examples of
animal model systems that have enabled significant progress in our under-
standing of the neural mechanisms of learning and memory. Although most
animals cannot communicate with us directly, our bigger challenge is to
ask them appropriate experimental questions when assessing their learning
and memory capacities. Since we are often most interested in plasticity
rather than in complexity of task performance, these questions should be
208 Part IV
unambiguous, perhaps even with binary solutions. Our queries should also
be carefully focused, without extraneous stimuli. For example, a mouse
navigating a radial arm maze may respond to pheromone cues left by a
previous subject, thus masking a capacity to learn a visual task. Likewise, a
well-fed fruit fly in a T-maze may be less motivated to run from a condi-
tioned odor than a hungry fly. These are representative experimental con-
founds, identified through trial and error, that can lead our research
animals to provide confusing answers to our improperly formulated
questions.
Learning and remembering are recognized as conserved (or convergent)
processes across the animal kingdom, at nearly all levels of organization
from molecules to behavior (Barco, Bailey, & Kandel, 2006; Dubnau, 2004;
Hochner, Shomrat, & Fioritohuman, 2006; Sakarya et al., 2007; Walker,
1987). Evidence of behavioral plasticity can even be found in organisms
that lack nervous systems (Sakarya et al., 2007) and consist only of single
cells (Armus, Montgomery, & Gurney, 2006; Saigusa, Tero, Nakagaki, &
Kuramoto, 2008). This remarkable breadth of homology suggests that
plasticity in behavior and in underlying neural mechanisms is indeed
ancient, and strongly justifies the use of a comparative approach for study-
ing learning and memory in animals. Consequently, representative model
research organisms can be selected based on behavioral attributes and
favorable access to the relevant neural functions. Animal researchers
focusing their effort on one particular model system will compare their
findings with those made using other systems, thus testing hypotheses of
homology. For example, the cAMP/PKA second messenger pathway dis-
cussed in this section of the book was independently revealed to be an
essential cellular link for establishing normal memory in neurons of mol-
lusks, flies, and mice. On the strength of this evidence, it was then possible
to recognize the importance of this biochemical pathway in humans and in
other systems not amenable to invasive experimentation.
The following chapters highlight experimental advantages afforded by
prominent model learning and memory systems. We begin with an intro-
duction by Brian Dunkelberger, Christine Serway, and Steven de Belle
(University of Nevada, Las Vegas), who present historical background
and a framework for studies of neurobehavioral plasticity and homology
in the animal kingdom.
Tiffany Timbers and Catherine Rankin (University of British Columbia)
then illustrate some of the experimental advantages of the nematode
Caenorhabditis elegans model system. The countable, well-characterized
neurons in this tiny (1 mm long) worm can be precisely removed or
manipulated to investigate genetic, molecular, and cellular questions
about their roles in learning and memory processes.
Animal Models of Behavioral and Neural Plasticity 209
David Glanzman (University of California, Los Angeles) highlights studies

with the marine snail Aplysia californica. The comparatively simple ner-
vous system of this invertebrate provides a valuable experimental model
for understanding cellular and circuitry mechanisms underlying simple
forms of learning. One exciting advantage of Aplysia is that sensory
and motor neurons mediating specific reflexes can be maintained in cell
culture, where they will reestablish their synaptic connections, providing an
opportunity to examine the cellular and molecular properties of synaptic
plasticity.
One of the most remarkable learners in the animal kingdom is the
honeybee (Apis mellifera). These highly social insects can form life-
long memories after only a few bouts of training. Randolf Menzel (Free
University of Berlin) presents the honeybee as a comparative model
for understanding the neural architecture, physiology, and cognitive func-
tions of the more complex but less experimentally accessible brains of
vertebrates.
Development of vocal communication is most commonly associated
with humans. However, a small number of animal taxa demonstrate similar
abilities and serve as important model systems for investigating the com-
plex neural processes underlying the acquisition, production and compre-
hension of language. David Vicario (Rutgers University) discusses the use
of songbirds in studying basic mechanisms of object recognition, memory
and motor learning in a behavioral context.
In the final chapter, Francis Brennan (RedPoint Bio Corporation) and
Ted Abel (University of Pennsylvania) describe the cAMP/PKA cellular
pathway and its role in memory formation. They then illustrate how phar-
macological and transgenic access to cAMP/PKA pathway elements
in the mouse Mus musculus and pathway homology with humans
may lead to potential therapeutic treatments for three classes of memory
disorders associated with Alzheimer’s disease, mental retardation, and
schizophrenia.
Neurobiologists and cognitive psychologists do not share the same
history, philosophy, or methodology. However, they are converging on
the common goal of determining how brains learn and remember. Learning
is believed to be an ancient property of neurons and brains retained at
multiple levels of organization in all animal lineages. Not surprisingly,
conserved mechanisms are reflected in a common set of learning principles
(Walker, 1987). This homology provides a solid basis for the comparative
model systems approach, which is the common thread running throughout
this section of the book. Effective model systems must (i) offer significant
experimental advantages, (ii) have well-established protocols, resources,
and data bases, and (iii) be representative of other organisms. The exam-
ples presented here satisfy all of these criteria.
210 Part IV
References
Armus, H. L., Montgomery, A. R., & Gurney, R. L. (2006). Discrimination learning

and extinction in paramecia (P. caudatum). Psychological Reports, 98, 705.
Barco, A., Bailey, C. H., & Kandel, E. R. (2006). Common molecular mechanisms in
explicit and implicit memory. Journal of Neurochemistry, 97, 1520.
Dubnau, J. (2004). Neurogenetic dissection of conditioned behavior: Evolution by
analogy or homology. Journal of Neurogenetics, 17, 295–326.
Hochner, B., Shomrat, T., & Fioritohuman, G. (2006). The octopus: a model for a
comparative analysis of the evolution of learning and memory. The Biological
Bulletin, 210, 308.
Saigusa, T., Tero, A., Nakagaki, T., & Kuramoto, Y. (2008). Amoebae anticipate
periodic events. Physical Review Letters, 100, 018101.
Sakarya, O., Armstrong, K. A., Adamska, M., Adamski, M., Wang, I. F., Tidor, B., et al.
(2007). A post-synaptic scaffold at the origin of the animal kingdom. PLoS ONE,
2, e506.
Walker, S. (1987). Animal learning. New York: Routledge & Kegan Paul.
Human Learning 211
A Biological Basis for Animal Model

Studies of Learning and Memory
Brian S. Dunkelberger, Christine N. Serway,
and J. Steven de Belle
University of Nevada
1. Introduction
Plato proposed that humans are born with innate knowledge of all things
and only need the correct keys to unlock the secrets of the universe. Others,
including Aristotle, believed we are blank canvases waiting to be painted by
the events in our lives (Bennett, 2007). Humans are profoundly curious—
particularly about how we are put together, how we are biologically related
to other animals, and how our minds and personalities compare with those of
other people. Much of the scientific endeavor in the biological and social
sciences focuses on these issues. Perhaps one of the oldest and still most
perplexing questions is, ‘‘How do we remember?’’ As humans, we have the
ability not only to form and retain memories of events in which we have
partaken or witnessed, but also to extrapolate, assimilate, and create novel or
abstract ideas from fragments of previous experiences. Although these are
indeed fantastic achievements of neural evolution, they are not uniquely
human qualities and have been demonstrated in several animal taxa.
In this section of the book, we will discuss a classification of behavioral
phenomena and give examples of how behavior is measured in animal
systems. This will be followed by a historical account of animal model
organisms used to investigate the neural mechanisms of learning and
memory. Finally, we discuss homology as the biological basis for the
comparative approach using animal model systems.
2. Learning and Memory Classification
Dozens of different forms of animal learning have been described

(Moore, 2004). It is likely that some reflect unique biological mechanisms
212 Brian S. Dunkelberger, Christine N. Serway, and J. Steven de Belle
while others are based on universal processes found across taxa. The
assays selected for discussion here establish fundamental aspects of
learning, which are likely based on these common learning processes.
We will focus on nonassociative and associative learning, the two main
classes of behavioral plasticity commonly studied in animals.
2.1. NONASSOCIATIVE LEARNING

Nonassociative learning is the response to a change in the salience of an
object or event with interpretations ranging from benign to fearful (Walker,
1987). The two most frequently discussed subsets of nonassociative learn-
ing are habituation and sensitization. Habituation is a decrease in the speed
or severity of a response to a repeated stimulus, whereas sensitization is an
increase. Typically, the difference in behavior is caused by the relative
noxious nature of the stimulus with habituation elicited by a neutral
stimulus and sensitization induced by negative stimuli. For example, habi-
tuation can be elicited in the gill withdrawal reflex of the marine mollusk
Aplysia californica by the repeated application of a light tap to the body
(Carew, Pinsker, & Kandel, 1972). As this is not a harmful stimulus, a
decrease in gill withdrawal speed and duration is seen over time. However,
if the animal receives a tail shock before receiving the repeated light taps,
the animal exhibits sensitization as the gill withdrawal is faster and persists
for longer periods than an animal that has not received a tail shock.
As recently as the early 1900s, it was commonly believed that reflex
behavior was invariant. One of the earliest signs of habituation of a reflex
was observed in spiders (Christoffersen, 1997; Peckham & Peckham, 1887).
When a tuning fork was used to vibrate their web, a spider would drop
away, hanging at some distance on a single thread. After repeated exposure
to this stimulus, they became habituated and discontinued this behavior.
Today habituation and sensitization have been demonstrated as basic
forms of learning throughout the animal kingdom, and include examples
such as defensive withdrawal reflexes in Annelid worms (Boulis & Sahley
1988; Ratner & Gilpin, 1974), the gill withdrawal reflex in marine mollusks
A. californica (Carew, Castellucci, & Kandel, 1971; Pinsker, Kupfermann,
Castellucci, & Kandel, 1970), as well as the umbilical abdominal reflex in
humans Homo sapiens (Hagbarth & Kugelberg, 1958; Lehner, 1941).
2.2. ASSOCIATIVE LEARNING
Associative learning entails the pairing of two or more objects or events

to provide new meaning to the previously novel stimuli (Walker, 1987). For
example, the color green has no inherent meaning. However, with repeated
conditioning to traffic laws, green has acquired the meaning ‘‘go.’’ The green
A Biological Basis for Animal Model 213
light is associated with the task of moving forward, and therefore elicits
that behavior. The two common forms of associative learning are Pavlovian
or classical conditioning and instrumental or operant conditioning. We will
briefly discuss some of the classic studies demonstrating these types of
learning.
The pioneering work of Ivan Pavlov in the early twentieth century gave
rise to classical (Pavlovian) conditioning. Dogs very reliably salivate in
response to the presentation of food. Based on this, Pavlov designed a
simple experiment in which a bell [the conditioned stimulus (CS)] would
ring just before a dog was presented with food [the unconditioned stimulus
(US)] in an attempt to provide a meaningful prediction of the pairing of
food with the bell. Normally the bell on its own does not elicit salivation [an
unconditioned response (UR)]. But after a few training events, the dogs
began to salivate at the sound of the bell in the absence of food [the
conditioned response (CR)]. This response was seen only in conditioned
dogs, as those that were not trained did not salivate at the sound of the bell
(Pavlov, 1927). Classical Pavlovian conditioning has been successfully
adapted to induce learning events in a wide variety of animals including
honeybees Apis mellifera (Hammer & Menzel, 1995; Menzel & Muller,
1996), the common fruit fly Drosophila melanogaster (Quinn, Harris, &
Benzer, 1974; Tully & Quinn, 1985), canaries Serinus canarius (Jarvis,
Mello, & Nottebohm, 1995), and many other model systems.
Instrumental or operant conditioning creates a situation where an ani-
mal’s voluntary behavior operates on the environment, which subsequently
influences future behavioral outcomes. Animals form an association
between their response (behavior) and the stimulus that follows (conse-
quence). At the same time Pavlov was developing his classical conditioning
procedures, ground-breaking work on instrumental conditioning by
Edward Thorndike and B. F. Skinner was being conducted. Thorndike
built puzzle boxes for domestic cats, with a built-in escape mechanism
consisting of a looped string the cat could pull. When placed into the box,
cats showed signs of discomfort and attempted escape until successfully
pulling the string either by accident or trial and error. Interestingly, as the
same animals were repeatedly tested, they rapidly improved their escape
time as they learned the task. The opposite is true for undesirable
responses, which were weakened and occurred less frequently after
repeated testing (Thorndike, 1898).
One drawback to Thorndike’s puzzle box design was that upon solving
the puzzle, the animal was no longer in the box, so he had to artificially
control when a new experiment began. Skinner wanted to look at the rate
at which an animal would perform a learned response on its own. His
‘‘Skinner Box’’ was a small chamber with a lever inside attached to an
electrical monitoring system. It provided a reinforcer when depressed by
the animal, eliminating handling by the experimenter. Instrumental condi-

tioning can involve reinforcement in the form of reward or punishment that
can be either given or withheld, creating many possible experimental
situations. Skinner showed that animals would learn how to maximize a
reward or minimize a punishment (Ferster & Skinner, 1957). He also
developed fixed-interval schedules using a timing device that allowed
only small unit amounts of food reward delivery during a specified period
of time. Interestingly, rats tended to ‘‘pace’’ themselves, with more attempts
immediately before receiving reinforcement and performing fewer
attempts immediately after it. Instrumental conditioning, like the others
discussed so far, has been demonstrated in a wide variety of organisms
including marine mollusks (Baxter & Byrne, 2006), the cockroach Peripla-
neta americana (Sakura, Okada, & Mizunami, 2002), various farm animals
(Arave, 1996), and many others.
2.3. MEMORY CLASSIFIED BY TIME
One defining feature of memory is the amount of time required for its loss.
This memory decay can often be divided into phases having distinct
behavioral, physiological, or cellular properties revealed through experimen-
tation. For example, mechanisms of short-term memory (STM) and long-term
memory (LTM) can be separated through genetic and pharmacological
methods in many model systems. The fly Drosophila has been an important
source of information about learning and memory mechanisms for over
30 years, in both nonassociative and associative learning paradigms (Corfas
& Dudai, 1989; Tully & Quinn, 1985). Dozens of characterized mutations have
facilitated the genetic dissection of memory phases using an associative
assay that pairs a mild electric foot shock with a novel odor. Two of the
first memory mutants isolated in flies, dunce and rutabaga, were examined
in an olfactory conditioning assay. Both exhibited some decrement in initial
learning, but had much sharper decreases in memory retention within the
first hour after training compared with normal wild-type flies (Byers, 1980;
Livingstone, 1985). After this time, their memory decay was relatively normal.
dunce and rutabaga were thus categorized as STM mutants (Margulies,
Tully, & Dubnau, 2005; Tully, Preat, Boynton, & Del Vecchio, 1994).
Genetic and pharmacological studies in Drosophila also established two
distinct longer forms of memory. Early experiments demonstrated an
anesthesia-resistant memory (ARM) phase lasting up to 1 day after a single
training session (Margulies et al., 2005; Quinn, & Dudai, 1976; Tempel,
Bonini, Dawson, & Quinn, 1983). Massed training (10 training sessions
administered one immediately after the other) produces even stronger
memory retention, lasting about 3 days, and this memory is insensitive to
the protein synthesis inhibitor cycloheximide. In contrast, spaced training
(10 training sessions with a 15-min rest interval between each) yields a
protein-synthesis-dependent memory lasting at least 1 week (Tully et al.,
1994). As has been found in many model systems, repetition produces
better memory, and spaced repetition results in the best memory of all.
Along with short and long forms of memory, intermediate memory
processes bridging the gap between them have been described in flies as
well as in several other species. Interestingly, amnesiac mutant flies show
near-normal memory retention immediately after a single training session and
again 7 hours later. In between these time points, memory retention in the
mutants is appreciably lower than in normal (Quinn, Sziber, & Booker, 1979).
Often human and model organism research is conducted independently
with little exchange of information. However, there is much to be
gained from merging ideas between the fields. Figure 1a shows a simplified
(a) Human
Sensory STM LTM

memory 18 s
1s Lifetime
(b) Fly Lifetime

LTM
30 s 1h
Acquisition STM 5h
MTM 3 days
ARM
Fig. 1. A comparison of memory consolidation models for humans and Drosophila.

(a) A three-phase model of human memory proposed by Atkinson and Shiffrin
based on their behavioral observations (Atkinson & Shiffrin, 1971). They suggest
that due to limited processing capacity at higher levels, most information is kept for
1 s in a temporary buffer referred to as sensory memory. Only context-relevant
information is then retained for 18 s in short-term memory (STM), with recurrent
events retained as long-term memory (LTM) that can last a lifetime. (b) One model
of memory in the fruit fly showing genetically and pharmacologically defined
memory phases and their approximate durations. Acquisition (learning) is
achieved within the first 30 s, followed by STM lasting about 1 h, then middle-term
memory (MTM) begin to further consolidate the memory. Middle-term memory is
believed to continue until the 5th hour after acquisition. If a spaced learning
protocol was used (10 rounds of training with 15-min intervals between each
round), this leads to LTM, which is protein-synthesis-dependant and can still be
detected weeks later. However, if a massed training protocol was used (10 rounds
of successive training with no rest intervals), anesthesia-resistant memory (ARM) is
generated, which has a duration of approximately 3 days (Margulies et al., 2005).
Atkinson and Shiffrin model for human learning, which describes three distin-
guishable memory phases based on behavioral observations (Atkinson & Shif-
frin, 1971). Figure 1b illustrates temporal features of memory phases in
Drosophila based on genetic/transgenic dissection and pharmacological
disruptions (Tully et al., 1996, 1994). In the biological sciences, we hope to
describe the neural mechanisms of behavioral phenomena well described in
humans and other systems not amenable to invasive experimentation.
3. A Brief History of Animal Models
Animal models have been useful in demonstrating how neural mechanisms

give rise to behavior and behavioral plasticity, as well as how the nervous
system changes in response to experience and memory consolidation.
3.1. THE IMPORTANCE OF THE BRAIN
In our early written history, it was debated which organ in the human
body was the seat of memory and intelligence. The oldest written record
containing the word ‘‘brain’’ is found in the Edwin Smith surgical papyrus
written in 1700 BC. In this text, brain injuries are noticed to be associated
with changes in the function of other parts of the body, especially the lower
limbs (Gross, 1998). Curiously, the Egyptians did not place such a great
importance on the brain, as they discarded it during the mummification
process while preserving other organs. Aristotle was also convinced that
cognitive processes took place in the heart (Finger, 1994). Alcmaeon used
animal models to address this issue around 500 BC. He dissected the eye of
an animal (of an unnamed species) and noted that the tract leading from
the eye projected into the brain. From this observation he concluded that
the brain was important for the collection of all sensory information
(Lloyd, 1975). The many philosophers and physicians who followed
Alcmaeon began to attribute more behavioral and cognitive functions to
brain activity (Encyclopedia Britannica Online, 2007).
3.2. BRAIN FUNCTIONS
By the beginning of the nineteenth century, almost nothing was known

about how brains work. Marie-Jean-Pierre Flourens performed localized
lesions in the brains of living rabbits and showed that the main divisions of
the brain were responsible for largely different functions (Yildirim &
Sarikcioglu, 2007). Since lesions and other accidental brain damage proved
to be such useful tools to map out functionally relevant regions in human
brains (Marshall & Fink, 2003), people began to look for storage sites of
learning and memory within the brain based on the same principles. Karl
Lashley systematically made various-sized lesions in diverse regions of the
cerebral cortex of rats and examined their behavior in a series of mazes
varying in difficulty (Lashley, 1929). Ultimately, he showed that the locus of
the lesion was less important than the size, particularly for the difficult
mazes. Lashley’s work helped to shape our current view of memory sto-
rage. It is currently believed that different aspects of memory including
color and shape are stored in different locations, potentially accounting for
the difficulty he encountered in finding traces of memory. Animal models
continue to be important for studies of brain function in behavioral plasti-
city. They are especially useful in revealing the neural underpinnings of
diseases that affect learning and memory. Several examples are discussed
in chapter 20 ‘‘The CAMP/PKA Pathway and the Modeling of Human Mem-
ory Disorders in Mice.’’
3.3. NEURONS
In the late nineteenth century, Golgi and Cajal developed staining meth-
ods that for the first time permitted the visualization of detailed fine
structure of the brain (in birds). Cajal argued that the brain was made up
of many small but interconnected cells (Finger, 1994). These elements were
given the name ‘‘neuron’’ in 1891 by Wilhelm Von Waldeyer (Finger, 1994) but
it was not until many years later that people understood anything about how
neurons actually functioned. In 1952, Hodgkin and Huxley published a
computational model describing the flow of electrical current through
neurons (Hodgkin, 1952). They recorded ionic currents in the giant axon
fiber of the Atlantic squid Loligo pealei. The sheer size of this neuron
enabled them to conduct these experiments, which would have been impos-
sible in most other organisms.
4. Homology and the Comparative Approach
Species homology has been the theoretical basis for why researchers
have and continue to ask biologically interesting questions in model organ-
isms. Structural and behavioral similarities among animals can result
through common descent or through convergent evolution (Butler, 2000).
4.1. GENOME HOMOLOGY
Not surprisingly, the genomes of animals with common features and

shared ancestry are homologous to some extent. Looking at this in another
way, if sequence homology in related species contributes to the development
of similar brains, these brains may also drive similar behavior and similar
aspects of behavioral plasticity. Despite relatively long divergence times,
both genome size and genes themselves can be highly conserved even
among distantly related species (Hedges, 2002; Rubin et al., 2000). Most
mammals, for example, have similar genome sizes of 3 billion base pairs
(bp) (Pennisi, 2001). Although humans have an estimated 25,000 genes and
fruit flies have approximately 13,600, it is estimated that 60% of these are
conserved between them (Rubin et al., 2000). Interestingly, many genes
already known to be involved in human neurological diseases have fly
homologues, and mutations in these genes appear to cause similar symp-
toms in both species (Hedges, 2002).
4.2. BRAIN HOMOLOGY
Upon initial observation, the brains of invertebrates (e.g., insects) and

vertebrates appear vastly different. However, there is considerable evidence
that these brains evolved from a common ancestor (Kammermeier &
Reichert, 2001), from which both groups have retained many common
features. All craniate brains develop from three primary rostral–caudal
segments (Fig. 2a) known as the forebrain, midbrain, and hindbrain
(Butler, 2000). Interestingly, higher invertebrate brains also develop in
three primary rostal–caudal segments: protocerebrum, deutocerebrum, and
tritocerebrum (Butler, 2000) (see Fig. 2b).
Further evidence of homology in the brain can be found by examining
the genes known to direct aspects of nervous system development. The
homeotic (Hox) genes produce proteins involved in establishing cellular
identity in early Drosophila embryogenesis and are well conserved in all
bilaterally symmetrical animals. The presence or absence of certain Hox
proteins in very specific patterns drives the development of particular
structures including the central nervous system (CNS) precursor cells.
Mutational inactivation of two specific Drosophila Hox genes as well as
their vertebrate homologs prevents cells from adopting their expected
neuronal cell fate, indicating that these genes have similar neuronal func-
tions in both fruit flies and mice (Hirth & Reichert, 1998).
The Drosophila gene orthodenticle (otd) is a ‘‘gap’’ gene that regulates
the formation of two main regions of the brain: the protocerebrum and the
deutocerebrum (see Fig. 2b). When mutated, the loss of Otd results in the
loss of the rostral brain. Its mammalian ortholog, known as Otx1, produces
a similar effect in mammals as mutations cause the loss of fore- and
midbrain regions (Acampora et al., 1995). Remarkably, full restoration of
the missing brain regions results when normal sequences of these genes
are exchanged between mutants of both species (Acampora et al., 1998;
Leuzinger et al. 1998). This multispecies transgenic physiological rescue of
(a) (b)
Forebrain
(Proencephalon) Protocerebrum
Midbrain
(Mesencephalon) Deuterocerebrum
Hindbrain
(Rhombencephalon) Tritocerebrum
(c) (d)
Hippocampus Mushroom bodies
Fig. 2. A comparison of mammalian and insect brains and functionally equivalent

memory centers. (a) Early development in the mammalian brain establishes three
major divisions: forebrain, midbrain, and hindbrain. (b) Development of the insect
brain gives rise to three divisions: the protocerebrum, the deutocerebrum, and the
tritocerebrum. (c) A human brain. This diagram of a human brain shows the relative
location of the mammalian memory center known as the hippocampus located in
the forebrain. (d) The mushroom bodies are invertebrate memory centers located in
the protocerebrum shown here in a fly brain.
brain defects is strong evidence for similarity in the development of the

CNS in the animal kingdom.
4.3. FUNCTIONAL HOMOLOGY
The greatest differences between vertebrate brains of various species lie

in their environmental adaptations. For example, Radinsky grouped multi-
ple species of otters by how often they used their forepaws to manipulate
food items and then compared this behavior with the somatosensory area
in the cortex (forebrain) where these limbs were represented. The species
with the greatest use of their forepaws had the largest area devoted to
forepaw control (Butler & Hodos, 1996; Radinsky, 1968).
Many brain structures found in vertebrates and invertebrates have simi-

lar connectivity and organization. For example, the sensory systems of
humans map spatial information from the external world onto the brain
in an orderly way, generating topographic maps. In the visual system, cells
in the retina that receive input from adjacent positions in the visual field
have synaptic connections at neighboring positions in the brain (Kandel,
Schwartz, & Jessell, 1999). Topographic organization of neural circuits is
also commonly found in other vertebrates such as the audio and visual
systems in the barn owl (Maczko, Knudsen, & Knudsen, 2006) and the
mechanosensory and olfactory systems in mice (Andermann & Moore, 2006;
Strotmann & Breer, 2006). This type of organization has also been
demonstrated in higher-order invertebrates such as the honeybee and
fruit fly mechanosensory and olfactory systems (Hiroyuki, 2007; Vosshall,
Wong, & Axel, 2000).
Functional homology between vertebrate and invertebrate brains is
supported by a comparison of structures known to mediate aspects of
behavioral plasticity. The vertebrate hippocampus (see Fig. 2c) constitu-
tes part of the limbic system in the forebrain (Ekstrom et al., 2003;
Scoville & Milner, 1957). The functional equivalent of a hippocampus is
the arthropod mushroom body located in the protocerebrum (see Fig. 2d)
(Zars, 2000). Although not obviously similar physically, they are critical
for both establishing memories (Mizunami, Weibrecht, & Strausfeld, 1998;
Muller, 1996) and showing elevated expression of similar learning-related
molecules (Crittenden, Skoulakis, Han, Kalderon, & Davis, 1998; Kandel &
Abel, 1995).
4.4. NEURON HOMOLOGY
Neurons perform essentially the same tasks and utilize similar mechanisms
across species. Sensory neurons relay information to interneurons or perhaps
to motor neurons directly through either electrical gap junctions or chemical
synapses using neurotransmitters. Human embryonic stem cells implanted
into the brains of embryonic mice and chicks (Goldstein, Reubinoff, &
Benvenisty, 2002; Muotri, Nakashima, Toni, Sandler, & Gage, 2005) differenti-
ate into neurons and integrate into the host forebrain. This argues that
neurons are functionally similar and interchangeable among species, lending
further support to the comparative approach using animal model systems.
4.5. BEHAVIOR HOMOLOGY
Homology across species is also seen on a behavioral level. Certainly

most animals perform the same basic behavioral repertoire as humans:
they all feed, sleep, move, and reproduce. Therefore, it should not be
surprising that we share at least some neural mechanisms that drive these
common behaviors. However, we are often astonished when encountering
examples of complex behavior thought to be exclusive to humans, such as
Chimpanzees learning sign language (Fouts & Waters, 2001) and honeybees
that dance to communicate (Menzel & Giurfa, 2006). Even more complex
behavioral interactions have been described in nature. For example, the
white-fronted bee-eaters Merops bullockoides are a type of monogamous
bird species that upon losing its brood, frequently abandon further
breeding attempts and begins to help a closely related pair rear their brood
(Cheney, 1999; Emlen & Demong, 1995). Knowing that animals share some
higher-order cognitive ability with humans makes them ideal candidates for
research into the nervous systems giving rise to these behaviors.
5. Conclusion
There is extensive support for the use of model systems to further our
understanding of learning and memory in all animals, including humans.
This is based on the preponderance of homology at all levels of biological
organization among species, allowing for meaningful comparisons of beha-
vioral plasticity and brain mechanisms from which it is derived. For as long
as there have been paintings on cave walls, tales passed down from gen-
eration to generation, and words written on clay tablets, papyrus, or paper,
we have looked to animals to tell us a little more about ourselves. All
evidence suggests that we are not mistaken in doing so.
Acknowledgments
This work was supported by a grant to JSdB from the National Science
Foundation (0237395). Part of the work was conducted by JSdB while
serving as a Visiting Scientist at the National Science Foundation.
References
Acampora, D., Avantaggiato, V., Tuorto, F., Barone, P., Reichert, H., Finkelstein, R.,
et al. (1998). Murine Otx1 and Drosophila otd genes share conserved genetic
functions required in invertebrate and vertebrate brain development. Develop-
ment, 125, 1691–1702.
Acampora, D., Mazan, S., Lallemand, Y., Avantaggiato, V., Maury, M., Simeone, A.,
et al. (1995). Forebrain and midbrain regions are deleted in Otx2-/- mutants due to
a defective anterior neuroectoderm specification during gastrulation. Develop-
ment, 121, 3279–3290.
Ai, H., Nishino, H., & Itoh, T. (2007). Topographic organization of sensory afferents
of Johnston’s organ in the honeybee brain. The Journal of Comparative Neurol-
ogy, 502, 1030–1046.
Andermann, M. L., & Moore, C. I. (2006). A somatotopic map of vibrissa motion
direction within a barrel column. Nature Neuroscience, 9, 543–551.
Arave, C. W. (1996). Assessing sensory capacity of animals using operant technol-
ogy. Journal of Animal Science, 74, 1996–2009.
Atkinson, R. C., & Shiffrin, R. M. (1971). The control of short-term memory. Scien-
tific American, 225, 82–90.
Baxter, D. A., & Byrne, J. H. (2006). Feeding behavior of Aplysia: A model system for
comparing cellular mechanisms of classical and operant conditioning. Learning
and Memory, 13, 669–680.
Bennett, M. R. (2007). Development of the concept of mind. Australian and
New Zealand Journal of Psychiatry, 41, 943–956.
Boulis, N. M., & Sahley, C. L. (1988). A behavioral analysis of habituation and
sensitization of shortening in the semi-intact leech. Journal of Neuroscience,
8, 4621–4627.
Butler, A. B. (2000). Chordate evolution and the origin of craniates: An old brain in a
new head. The Anatomical Record, 261, 111–125.
Butler, A., & Hodos W. (1996). Comparative vertebrate neuroanatomy: Evolution
and adaptation. New York: Wiley-Liss.
Byers, D. (1980). A review of the behavior and biochemistry of dunce, a mutation of
learning in Drosophila. Basic Life Sciences, 16, 467–474.
Carew, T. J., Castellucci, V. F., & Kandel, E. R. (1971). An analysis of dishabituation
and sensitization of the gill-withdrawal reflex in Aplysia. International Journal
of Neuroscience, 2, 79–98.
Carew, T. J., Pinsker, H. M., & Kandel, E. R. (1972). Long-term habituation of a
defensive withdrawal reflex in Aplysia. Science, 175, 451–454.
Cheney, D. (1999). Animal behavior: Only unthinking intelligence? science.
283.5400.333. Science, 283, 333–334.
Christoffersen, G. R. J. (1997). Habituation: Events in the history of its characteriza-
tion and linkage to synaptic depression. A new proposed kinetic criterion for its
identification. Progress in Neurobiology, 53, 45–66.
Corfas, G., & Dudai, Y. (1989). Habituation and dishabituation of a cleaning reflex in
normal and mutant Drosophila. Journal of Neuroscience, 9, 56–62.
Crittenden, J. R., Skoulakis, E. M., Han, K. A., Kalderon, D., & Davis, R. L. (1998).
Tripartite mushroom body architecture revealed by antigenic markers. Learning
and Memory, 5, 38–51.
Ekstrom, A. D., Kahana, M. J., Caplan, J. B., Fields, T. A., Isham, E. A., Newman, E. L.,
et al. (2003). Cellular networks underlying human spatial navigation. Nature, 425,
184–188.
Emlen, S. T. W., P. H., & Demong, N. J. (1995). Making decisions in the family – an
evolutionary perspective. American Scientist, 83, 148–157.
Encyclopedia Britannica (Encyclopedia Britannica online, 2007).
Ferster, C. B., & Skinner, B. F. (1957). Schedules of reinforcement. New York:
Appleton-Century-Crofts, Inc.
Finger, S. (1994). Origins of neuroscience: A history of explorations into brain

function. New York: Oxford University Press.
Fouts, R. S., & Waters, G. S. (2001). Chimpanzee sign language and Darwinian
continuity: Evidence for a neurological continuity for language. Neurological
Research, 23, 787–794.
Goldstein, R. D., M; Reubinoff, B, & Benvenisty, N. (2002). Integration and differ-
entiation of human embryonic stem cells transplanted to the chick embryo.
Developmental Dynamics, 225, 80–86.
Gross, C. G. (1998). Brain, vision, memory: Tales in the history of neuroscience.
Cambridge, MA: MIT Press.
Hagbarth, K. E., & Kugelberg, E. (1958). Plasticity of the human skin reflex. Brain,
81, 305–318.
Hammer, M., & Menzel, R. (1995). Learning and memory in the honeybee. Journal of
Hedges, S. B. (2002). The origin and evolution of model organisms. Nature Reviews
Genetics, 3, 838–849.
Hirth, F. H., B, & Reichert, H. (1998). Homeotic gene action in embryonic brain
development of Drosophila. Development, 125, 1579–1589.
Hodgkin, A. L., & Hunley A. F. (1952). Propagation of electric signals along ginat
nerve fibers. Proceedings of the Royal Society of London. Series B, 177–183.
Jarvis, E. D., Mello, C. V., & Nottebohm, F. (1995). Associative learning and stimulus
novelty influence the song-induced expression of an immediate early gene in the
canary forebrain. Learning and Memory, 2, 62–80.
Kammermeier, L. R., & Reichert, H. (2001). Common developmental genetic
mechanisms for patterning invertebrate and vertebrate brains. Brain Research
Bulletin, 55, 657–682.
Kandel, E, A., & Abel, T. (1995). Neuropeptides, adenylyl cyclase, and memory
storage. Science, 268, 825–826.
Kandel, E. J., Schwartz, J. H., & Jessell, T. M. (1999). Principles of neural science.
East Norwalk, CT, Hemel Hempstead: Appleton & Lange, Prentice Hall.
Lashley, K. (1929). Brain mechanisms and intelligence: A quantitative study of
injuries to the brain. Chicago: University of Chicago Press.
Lehner, G. F. J. (1941). A study of the extinction of unconditioned reflexes. Journal
of Experimental Psychology, 29, 435–455.
Leuzinger, S., Hirth, F., Gerlich, D., Acampora, D., Simeone, A., Gehring, W. J.,
et al. (1998). Equivalence of the fly orthodenticle gene and the human OTX
genes in embryonic brain development of Drosophila. Development, 125,
1703–1710.
Livingstone, M. S. (1985). Genetic dissection of Drosophila adenylate cyclase.
Proceedings of the National Academy of Sciences of the United States of
America, 82, 5992–5996.
Lloyd, G. (1975). Alcmeon and the early history of dissection New York: Sudhoffs
Archive.
Maczko, K. A., Knudsen, P. F., & Knudsen, E. I. (2006). Auditory and visual space
maps in the cholinergic nucleus isthmi pars parvocellularis of the barn owl.
JNEUROSCI.3946-06.2006. Journal of Neuroscience, 26, 12799–12806.
Margulies, C., Tully, T., & Dubnau, J. (2005). Deconstructing Memory in Drosophila.
Current Biology, 15, R700–R713.
Marshall, J. C., & Fink, G. R. (2003). Cerebral localization, then and now. Neuro-
Image: Convergence and Divergence of Lesion Studies and Functional Imaging
of Cognition, 20, S2–S7.
Menzel, R., & Giurfa, M. (2006). Dimensions of Cognition in an Insect, the Honey-
bee. Behavioral and Cognitive Neuroscience Reviews, 5, 24–40.
Menzel, R., & Muller, U. (1996). Learning and memory in honeybees: from behavior
to neural substrates. Annual Review of Neuroscience, 19, 379–404.
Mizunami, M., Weibrecht, J. M., & Strausfeld, N. J. (1998). Mushroom bodies of the
cockroach: Their participation in place memory. Journal of Comparative
Neurology, 402, 520–537.
Moore, B. R. (2004). The evolution of learning. Biological Reviews, 79, 301–335.
Muller, R. (1996). A quarter of a century of place cells. Neuron, 17, 813–822.
Muotri, A. R., Nakashima, K., Toni, N., Sandler, V. M., & Gage, F. H. (2005). Devel-
opment of functional human embryonic stem cell-derived neurons in mouse
brain. Proceedings of the National Academy of Sciences, 102, 18644–18648.
Pavlov, I. (1927). Conditioned reflexes. Oxford: Oxford University Press.
Peckham, G. W., & Peckham, E. G. (1887). Some observations on the mental powers
of Spiders. Journal of Morphology, 1, 383–419.
Pennisi, E. (2001). The Human Genome. Science, 291, 1177–1180.
Pinsker, H., Kupfermann, I., Castellucci, V., & Kandel, E. (1970). Habituation
and dishabituation of the gill-withdrawal reflex in Aplysia. Science, 167,
1740–1742.
Quinn, W. G., & Dudai, Y. (1976). Memory phases in Drosophila. Nature, 262, 576–577.
Quinn, W. G., Harris, W. A., & Benzer, S. (1974). Conditioned behavior in Drosophila
melanogaster. Proceedings of the National Academy of Sciences of the United
Quinn, W. G., Sziber, P. P., & Booker, R. (1979). The Drosophila memory mutant
amnesiac. Nature, 277, 212–214.
Radinsky, L. (1968). Evolution of somatic sensory specialization in otter brains.
Journal of Comparative Neurology, 134, 495–506.
Ratner, S. C., & Gilpin, A. R. (1974). Habituation and retention of habituation of
responses to air puff of normal and decerebrate earthworms. Journal of Com-
parative and Physiological Psychology, 86, 911–918.
Rubin, G. M., Yandell, M. D., Wortman, J. R., Gabor Miklos, G. L., Nelson, C. R.,
Hariharan, I. K., et al. (2000). Comparative Genomics of the Eukaryotes. Science,
287, 2204–2215.
Sakura, M., Okada, R., & Mizunami, M. (2002). Olfactory discrimination of
structurally similar alcohols by cockroaches. Journal of Comparative Physiology
A: Neuroethology, Sensory, Neural, and Behavioral Physiology, 188, 787–797.
Scoville, W. M., & Milner, B. (1957). Loss of recent memory after bilateral
hippocampal lesions. Journal of Neurology, Neurosurgery and Psychiatry,
20, 11–21.
Strotmann, J. B., & Breer, H. (2006). Formation of glomerular maps in the olfactory
system. Seminars in Cell & Developmental Biology Olfaction; Animal Stem Cell
Types, 17, 402–410.
Tempel, B. L., Bonini, N., Dawson, D. R., & Quinn, W. G. (1983). Reward learning in
normal and mutant Drosophila. Proceedings of the National Academy of Sciences
of the United States of America, 80, 1482–1486.
Thorndike, E. L. (1898). Animal intelligence: An experimental study of the associa-
tive processes of animals. Psychological Review, 2, 1–109.
Tully, T., & Quinn, W. G. (1985). Classical conditioning and retention in normal and
mutant Drosophila melanogaster. Journal of Comparative Physiology. A, Sen-
sory, Neural, and Behavioral Physiology, 157, 263–277.
Tully, T., Bolwig, G., Christensen, J., Connolly, J., DeZazzo, J., Dubnau, J., et al.
(1996). Genetic dissection of memory in Drosophila. Journal de Physiologie
(Paris), 90, 383.
Tully, T., Preat, T., Boynton, S. C., & Del Vecchio, M. (1994). Genetic dissection of
consolidated memory in Drosophila. Cell, 79, 35–47.
Vosshall, L. B. W., Wong, A. M., & Axel, R. (2000). An olfactory sensory map in the
fly brain. Cell, 102, 147–159.
Walker, S. (1987). Animal learning (Coltheart, M., Ed.). New York: Routledge &
Kegan Paul.
Yildirim, F. B., & Sarikcioglu, L. (2007). Marie Jean Pierre Flourens (1794 1867): An
extraordinary scientist of his time. Journal of Neurology, Neurosurgery and
Psychiatry, 78, 852.
Zars, T. (2000). Behavioral functions of the insect mushroom bodies. Current
Opinion in Neurobiology, 10, 790–795.
Human Learning 227
Caenorhabditis elegans as a
Model System in Which to Study
the Fundamentals of Learning
and Memory
Tiffany A. Timbers and Catharine H. Rankin
Department of Psychology & Brain Research Center, University
of British Columbia
1. Introduction
Understanding how humans are capable of learning and remembering

needs to be approached from several levels; the level of behavior, of
organization of the nervous system, of the neuron, and of the gene to
gain a holistic understanding of learning and memory. The well-understood
principles that govern the behavioral changes associated with
human learning and memory have been studied since Ebbinghaus in 1885
(Ebbinghaus, 1964). Technological progress in the last century has
provided researchers with new techniques such as recording of
evoked potentials (EEG electrorecordings) and fMRI that for the first
time allow real-time measurements of neuronal activity in a living human.
Exciting as this is, these measurements are limited in that they only
measure populations of neurons. We have yet to develop techniques
that allow us to ethically observe the changes in neuronal organization, in
individual neurons, and in gene expression in living humans. This places a
huge limitation on our understanding of the processes that govern
human learning and memory. These ethical and technological limitations
are being addressed by neuroscientists through the field of comparative
neuroscience, in which animal models are used for physiological and
cellular analyses. One animal model from this field, Caenorhabditis
elegans, easily overcomes these technological limitations, and it can be
studied at the level of behavior, neuronal organization, single neurons, and
the gene.
228 Tiffany A. Timbers and Catharine H. Rankin
Caenorhabditis elegans is a 1 mm long, transparent, free-living soil nema-

tode that navigates through its environment using taste, smell, touch, and
temperature. In the 1960s, Sydney Brenner developed C. elegans as a novel
experimental model organism to study the genetics of cell differentiation
and development of the nervous system: a task that becomes increasingly
difficult when studying more complex organisms. Brenner thought
C. elegans was an ideal organism to study because of its short life span,
reproductive simplicity, and small economical size. Caenorhabditis elegans
adults are fully developed in 3.5 days, making it easy to study a large number
of animals and possible to perform multigenerational studies. One sex is a
self-fertilizing hermaphrodite that produces clonal offspring. This character-
istic minimizes genetic background differences between individuals. They
eat bacteria, are stored on small Petri plates, and strains do not need to be
constantly maintained as they can be frozen at 80C (Brenner, 1974). Over
the past 40 years, an increasing number of scientists have chosen to study all
aspects of C. elegans biology. This has led to further development of the
worm as a model system, and now studying this organism is like studying an
animal with an instruction manual. Advantages and resources of C. elegans
now include the following: (i) a fully sequenced and annotated genome, (ii) a
cell lineage fate map, (iii) a complete anatomical map (including a neuronal
wiring diagram), (iv) the Caenorhabditis Genetics Centre that distributes
mutant and transgenic strains, (v) fluorescent genetic molecules [such as
Green Fluorescent Protein (GFP) and its variants] that can be manipulated
to show protein localization and function, and (vi) the development of a
large number of simple but robust behavioral assays. These tools and
advantages make C. elegans an ideal model to study the cellular basis of
learning and memory.
Though we stress that C. elegans is a successful model for this purpose
because of its simple behavior, neural organization, and ease of use in the
laboratory, it shares many biological similarities with higher animals. The
genome of C. elegans consists of 96,893,008 base pairs encoding over 19,000
genes and regulatory regions (The C. elegans Sequencing Consortium, 1998).
In comparison, the human genome is 30 times larger, made up of 2.85 billion
base pairs, but it encodes only slightly more genes (the total number being
somewhere between 20,000 and 25,000) (International Human Genome
Sequencing Consortium, 2004). It is estimated that about 35% of C. elegans
genes are closely related to human genes (Genome Sequencing Center,
2008). Like higher vertebrates, C. elegans neurons are organized into centers
of neuronal integration (although the nervous system in general is less
centralized than in higher animals), have both electrical and chemical
synapses, use most of the same neurotransmitter systems, and use many
of the same molecules for cell signaling (reviewed in Hall, Lints, Altan, &
Eric, 2005). With simple forms of learning, such as habituation and classical
Caenorhabditis elegans as a Model System 229
conditioning, the behavioral rules are the same regardless of the organisms
studied (from C. elegans to humans); thus it is likely that these simple forms
of learning appeared very early in evolution and have been maintained
throughout phylogeny. Thus, we believe that many mechanisms of learning
discovered in simple systems will generalize to more complex systems.
2. Types of Learning
Caenorhabditis elegans can learn about a wide variety of types of infor-

mation in its environment. The first studies of learning and memory in
C. elegans looked at the ability of the worm to habituate to nonlocalized
mechanosensory stimuli (Rankin, Beck, & Chiba, 1990). More recently,
several other learning paradigms have been studied; these include thermo-
tactic associative learning, chemosensory habituation and associative learning,
and pathogen avoidance learning. In thermotactic associative learning, the
ambient temperature and the presence of food are associated (Hedgecock &
Russel, 1975, Mohri et al., 2005). In chemosensory habituation, worms that
experience prolonged exposure to low concentrations of diacetyl show a
decrease in approach to the odorant, and this decrement can be dishabituated
(response returns to naive levels) by exposure to a stronger concentration
(Wen et al., 1997). The chemosensory associative learning paradigm shows
that C. elegans can make an association of one soluble ion with food and the
other with lack of food (Wen et al., 1997). Finally, in pathogen avoidance
learning, worms exposed to pathogenic bacteria will avoid them on subsequent
tests (Zang, Lu, & Bargmann, 2005). This is a very impressive learning repertoire
for such a small and compact organism with only 302 neurons. The limitations
of using C. elegans as a model for human learning and memory lie in the fact that
it has been shown to exhibit only these simple forms of learning and memory. It
is not a useful tool to study more complex cognitive forms of learning and
memory. However, the more complex forms of cognition must be built on a
foundation of the simple forms of learning, such as habituation and classical
conditioning; therefore, understanding them will offer insights in more complex
learning. This chapter will focus on short- and long-term mechanosensory
habituation. Although all of the other paradigms are interesting and useful for
studying learning, researchers have only been able to clearly show the presence
of long-term memory in C. elegans using the mechanosensory habituation
paradigm.
2.1. MECHANOSENSORY HABITUATION
The first and most extensively studied learning paradigm in the worm is
habituation to mechanosensory stimuli. In this learning paradigm, the
worm receives a mechanical stimulus, a tap, applied to the side of the Petri
plate within which it resides. The tap results in the worm changing from a
forward-swimming motion to a backward one for a brief period of time,
after which it swims forward again. This response to tap is termed a
reversal, and this behavior is called the tap withdrawal response.
The response magnitude, the distance the worm swims backward,
decreases with repeated stimulation; and this decrease is termed habitua-
tion (Rankin et al., 1990) (Fig. 1).
As in chemotactic olfactory habituation, habituation to tap can be
distinguished from sensory adaptation or fatigue by testing for dishabi-
tuation. After habituation training, an electrical shock can be adminis-
tered to the worm through the agar. This causes dishabituation, such
that worms will respond to the next tap stimulus as if it were novel
(Rankin et al., 1990).
Worms can also form chemosensory context associations with habitua-
tion to tap. When worms plated on Naþ (a taste worms are attracted to)-
treated agar (the substrate on which they live) are habituated to tap, then
transferred onto plain agar for a 1-h rest period, and finally rehabituated on
Naþ-treated agar, they show greater retention of the earlier training
(a) (b)
0.9
0.8
Mean reversal length (mm)
1. 0.7
0.6
0.5
Tap
0.4
0.3
0.2
0.1
1.
0.0
2. 0 5 10 15 20 25 30
Tap stimulus
Fig. 1. Short-term mechanosensory habituation. (a) Reflexive reversal performed by

worms when the Petri plate is tapped (tap withdrawal response). 1, The position of
the tail before the tap; 2, the position of the tail after the reversal. The distance the
worm reverses is measured (grey line). (b) Habituation of mean response amplitude
(reversal length in millimeters) to 30 tap stimuli delivered at a 10-s interstimulus
interval (ISI). Spontaneous recovery is assayed by tapping worms at 30 s, 5 min, and
10 min after the 30th tap (three responses under the grey bar, respectively).
(demonstrated by smaller reversal lengths in the second round of habitua-

tion) than worms that were trained and tested on plates without the Naþ
taste. Similar to associative learning in other organisms, this form of
associative conditioning shows extinction (leaving the worms in the same
context that they are trained for the 1-h rest period, which eliminates the
context effect) and latent inhibition (pre-exposure to the context for 1 h
before training eliminates the context effect) (Rankin, 2000).
3. Neural and molecular mechanisms of mechanosensory

habituation
The neural circuit that mediates the tap withdrawal response was identi-
fied by laser-ablating individual neurons and testing for changes in response
to tap. This method showed that the circuit consists of five mechanosensory
neurons, eight interneurons, and a pool of motor neurons (Fig. 2). Three of
the mechanosensory neurons, ALM left and right (L/R) and AVM, are located
in the anterior body and respond to head touch by inducing a reversal. The
remaining mechanosensory neurons, PLML/R, are located in the posterior
body and respond to tail touch by inducing forward motion. When a tap is
administered to the side of the Petri plate, the worms feel a mechanical
vibration transmitted through the agar at both the anterior and posterior
regions of their body. This causes ‘‘neural competition’’ from the circuit.
Both sets of mechanosensory neurons activate the interneurons (AVAL/R,
AVBL/R, AVDL/R, and PVCL/R), but the posterior neurons are hypothesized
to do so to a lesser extent as there are only two of them. This is integrated by
the interneurons, signaled to the muscles by the motor neurons, and the
resultant behavior is almost always a reversal (Wicks & Rankin, 1995).
Evidence from gene expression studies in the mid-1990s led to a hypoth-
esis about which neurotransmitter system mediated the chemical synapses
between the mechanosensory and interneurons in the tap withdrawal
circuit (Brockie, Madsen, & Maricq, 1997; Dent, Davis, & Avery, 1997;
Hart, Sims, & Kaplan 1995; Maricq, Peckol, Driscoll, & Bargmann, 1995).
Several classes of glutamate receptors were expressed in the many inter-
neurons of the worm, including the four pairs of interneurons that are part
of the tap withdrawal circuit. Confirmation of the hypothesis that gluta-
mate was a critical neurotransmitter came from the study of the first gene
identified to play a role in short-term habituation to tap, EAT-4. This gene
encodes the worm homologue of the mammalian glutamate vesicular
transporter, VGlut1. EAT-4 plays a role in loading the neurotransmitter
glutamate into synaptic vesicles that are then released when the neuron is
depolarized. EAT-4 is expressed in the mechanosensory neurons of the tap
withdrawal circuit (Lee, Sawin, Chalfie, Horvitz, & Avery, 1999), and worms
Touch sensory neurons

Interneurons
Command interneurons
Motor neurons Tail
Head Chemical synapses
Electrical synapses
Nerve ring
Mechanosensation
AVM ALM PLM
PVC AVD
AVB AVA
Forward B A Backward
Fig. 2. Schematic diagram of the components of the neural circuit that mediates the
tap withdrawal response. The circuit consists of five mechanosensory neurons
(triangles), four pairs of interneurons (hexagons), and a pool of motorneurons
(squares). Lines ending with arrows represent chemical synapses, and lines ending
with ovals represent electrical synapses (adapted from Rankin & Dubnau 2007,
p. 316).
with a loss-of-function mutation in this gene have normal initial responses to

tap, but habituate extremely rapidly and cannot be dishabituated by any
stimulus (Wicks & Rankin, 2000). This supported to the hypothesis that the
neurotransmitter between the mechanosensory and interneurons in the tap
withdrawal circuit was glutamatergic and suggested that presynaptic release
of glutamate from the mechanosensory neurons is critical for normal wild-
type short-term habituation. Interestingly, normal glutamate transmission
appears to be critical for normal learning in many species across evolution,
from habituation to sensitization and classical conditioning in the sea snail,
Aplysia (reviewed in Glanzman, 2007), to the intensively studied cellular

models of mammalian learning and memory: long-term potentiation and
long-term depression (reviewed in Citri & Malenka, 2007).
Recently, a role for the dopamine neurotransmitter system in habituation
to tap has also been demonstrated. In the absence of food, wild-type animals
habituate more rapidly than in the presence of food (Kindt et al., 2007); this
contextual information about food availability is provided by dopaminergic
mechanosensory neurons that sense the presence of bacteria. In the worm,
dopamine is synthesized in eight sensory neurons; the two anterior deirid
neurons (ADEs), the two posterior deirid neurons (PDEs), and the four CEP
neurons that are located in head of the worm (Sulston, Dew, & Brenner,
1975). A dopamine receptor, DOP-1, is expressed on the mechanosensory
neurons, ALM and PLM, that receive extrasynaptic connections from the
dopamine neurons, CEPs and PDEs, respectively, (Sanyal et al., 2004; Tsalik
et al., 2003). Worms with a mutation in the DOP-1 receptor habituated faster
than wild-type worms when stimuli were presented at a 10-s interstimulus
interval (ISI) when habituation was measured by the frequency of animals
that respond to tap (Kindt et al., 2007; Sanyal et al., 2004;). This affect of the
dopamine receptor mutation was limited to the rate of habituation, and had
no effect on the asymptotic level and no effect on the spontaneous recovery of
the animals (Kindt et al., 2007; Sanyal et al., 2004; Seamans & Yang, 2004).
A candidate gene approach to look for mutants that showed the same habitua-
tion phenotype as the dop-1 mutants (which would suggest that the genes are
in the same genetic pathway) revealed genes in the phospholipase-C b (PLCb)
pathway (Fig. 3). One of the endpoints of the identified PLCb pathway was the
second messenger IP3. This molecule is a known ligand of IP3-gated calcium
channels on the endoplasmic reticulum (ER) (an intracellular calcium store).
A second PLCb pathway endpoint, PKC-1 (encodes a kinase that is homolo-
gous to mammalian novel PKCs), can phosphorylate voltage-gated calcium
channels. The requirement of these genes suggested a role for calcium in this
habituation mechanism. To follow this up, in vivo Ca2þ imaging of the ALM
mechanosensory neurons revealed that intracellular Ca2þ levels increase
transiently in these neurons after mechanosensory stimulation, and that the
peaks of these transient Ca2þ waves decrease with repeated stimulation.
When Kindt et al. (2007) looked at the effect of mutations in the dopamine-
PLCb pathway, they observed that like the behavior, peaks of the transient
Ca2þ waves in ALML/R decreased more rapidly with repeated stimulation than
those of wild-type animals.
The role of dopamine in glutamate-dependent plasticity and in habituation
has parallels in mammalian studies. In the circuit that sends projections from
the nucleus accumbens to the prefrontal cortex (PFC), dopaminergic input plays
a role in regulating glutamate plasticity (Seamans & Yang, 2004). Also, there
appears to be an interesting convergence of information in the role of dopamine
Fig. 3. Model of the DOP-1 signaling pathway in the mechanosensory neurons identified
by Kindt and colleagues (2007). DOP-1 couples to Gq-a to activate PLCb to produce
the second messengers IP3 and DAG to activate ITR-1 (IP3-gated calcium channel) and
PKC-1, respectively, causing an influx of calcium into the intracellular space.
in C. elegans and in human habituation. Schizophrenia is a disease characterized

by delusions, hallucinations, disturbances in thinking, and withdrawal from
social activity. A disruption in the dopamine neurotransmitter system in the
PFC is thought to be a major contributing factor to the pathogenesis of this
disease, and research has shown that schizophrenic patients exhibit learning
and habituation impairments (Koh, Bergson, Undie, Goldman-Rakic, & Lidow,
2003; Mirnics & Lewis, 2001).
3.1. SITE OF NEURAL PLASTICITY
Laser ablation and behavioral studies by Wicks and Rankin (1997)

showed that the neural circuit of the tap withdrawal response significantly
overlaps with the neural circuits for two other behaviors in C. elegans,
spontaneous reversals and the thermal avoidance response. These three
behaviors share most of the same interneurons and motor neurons that
make up the locomotion pattern generator. If the site of neuroplasticity of
habituation to the tap withdrawal response was located in the interneurons
or motor neurons, Wicks and Rankin (1997) hypothesized that habituation
training would lead to changes in behaviors that shared these components
with the tap withdrawal circuit. Moreover, if the site of neuroplasticity of
the tap withdrawal response was located in the sensory neurons, or their
synapses onto the interneurons, they hypothesized that they would see no
effect on these other behaviors after habituation training. They ran experi-
ments indicating that habituation to tap had no effect on the magnitude or
frequency of either spontaneous reversals or the thermal avoidance
response. This suggests that the locus of plasticity is situated presynaptic
to the interneurons. The genetic analysis supports and extends the beha-
vioral and neural circuit analysis, but gives us more insight into where in
the sensory neurons these changes are happening. The dopamine data
points to cellular excitability of the sensory neurons as a locus of neuro-
plasticity, whereas the EAT-4 data suggests that the synapse is also a site
for plasticity. Thus, it seems that at least two different molecular mechan-
isms mediate the behavioral plasticity of the tap withdrawal response: a
mechanism that modifies the level of sensory cell excitability and a
mechanism that modifies that probability of synaptic release. It is only
through the analysis of this learning at the level of behavior, neural circuit,
individual neurons and the gene that made it possible to identify the sites of
plasticity to this level of specificity.
4. Common elements of memory
As researchers study memory across various species and tasks, it has

become clear that there are a number of different types of memory. In
C. elegans, memory has been studied in terms of its duration, or how long it
lasts. Studies of short-term habituation have shown that habituation is less
rapid and lasts longer if habituating stimuli are separated by 60 s than if
they are separated by 10 s (Rankin & Broster, 1992). This sensitivity to ISI is
a feature of habituation that is present in all species studied (Askew, 1970;
Byrne 1982; Davis, 1970; Groves & Thompson, 1970). These findings
suggested that short-term habituation at different frequencies might be
mediated by different mechanisms. Recent research from our lab has
supported this hypothesis by identifying one gene that affects habituation
at 10-s but not at 60-s ISIs, and a second gene that shows the opposite
pattern (unpublished data).
Despite the fact that worms live only 14–21 days in the lab, they can
express long-term memory for habituation (Rankin et al., 1990). The pro-
tocol to study lasting memory in C. elegans consists of administering 80 tap
stimuli in either a massed or a distributed training protocol. Distributed
training consists of administering four training blocks of 20 taps at a 60-s
ISI, with a 1-h rest period between blocks. In this memory experiment, one
group (the trained group) of worms receives the distributed training and
one group (the control group) of worms receives only a single tap on the
training day. Worms are tested for memory retention 24 h following training
by administering 10 test taps and comparing the response magnitude
between the trained and control groups. Worms that received distributed
training show smaller reversals to tap 24 h after training than control
worms that received only a single tap, suggesting that they have long-
term memory for the tap stimulus (Fig. 4). If the same experiment is
performed, but the worms receive massed training (80 taps at a 60-s ISI)
instead of distributed training, no difference is observed between the
trained and control groups 24 h after training. This indicates that there is
no memory retention at this time point (Rose, Kaun, & Rankin, 2002).
A difference in memory retention for distributed or massed training has
been observed in many different species. As stated, C. elegans can retain
memory for habituation for at least 24 h if training is presented in a
distributed or spaced manner; however, there is no memory for training
at 24 h if worms are given the same number of stimuli in a massed protocol.
This effect of greater memory with spaced rather than massed training has
been observed in rodents (Commins, Cunningham, Harvey, & Walsh, 2003;
Goodrick, 1973; Hasegawa, Shimamura, & Suzuki, 1988;), as well as in
humans. It was first suggested in the late 1800s when Ebbinghaus and
others noted that human subjects benefited from distributed practice
(Ebbinghaus, 1964; Jost, 1897; Thorndike, 1912). Recent meta-analyses
of the literature on massed versus distributed training in humans have shown
that several types of learning (e.g., verbal, and motor learning) benefit from
distributed practice (Cepeda, Pashler, Vul, Wixted, & Rohrer, 2006; Donovan &
Radosevich, 1999; Genovese, 1988; Janiszewski, Noel, & Sawyer, 2003; Lee &
Moss, 1996).
Detailed examination of worm long-term memory for habituation follow-
ing distributed training showed that it shares many characteristics with long-
term memories in other animals, including humans. C. elegans memory can
be long-lasting; initial studies in this organism showed that it can last for 24 h
and more recently it has been demonstrated that it can last for up to 72 h
(Ebrahimi & Rankin, 2006). This is one-fifth of the animal’s life span; a
comparative human memory would be stored for about 16 years. Like
mammals and other invertebrates, long-term memory in the worm requires
new proteins to be synthesized. This was demonstrated by heat-shocking
(a)
Day 1 – Training Day 2 – Testing
Trained
24 h
delay
Control
One hour rest between training blocks
(b) (c)
*
1.2
Control
Mean reversal length
1.0
0.8
Control
0.6 Trained
Trained
0.4
0.2
0.0
Fig. 4. Long-term memory in C. elegans. (a) Training in this paradigm consists of

four blocks of 20 taps given at a 60 s interstimulus interval (ISI) with a 1 h rest
period between each training block. Control worms receive one tap. Twenty-four h
later, 10 taps are administered to both control and trained worms to test for
memory. Each line represents a tap. (b) Memory is seen when trained worms
show significantly smaller responses than control worms. (c) Fluorescent imaging
of GLR-1::GFP fusion protein (expressed in the interneurons of the tap withdrawal
circuit) 24 h after long-term memory training in trained and control worms. GLR-1
expression levels are observed to decrease 24 h after training.
animals after each training block to halt all ongoing protein synthesis.
Worms given this treatment gave reversals to the test taps 24 h after training,
which were no different from control animals (Beck & Rankin, 1995). Long-
term memory in C. elegans is also associated with a decrease in expression
levels of a glutamate receptor subunit, GLR-1 (homologous to the mamma-
lian GluR1 glutamate receptor subunit), 24 h after long-term memory train-
ing. Worms that express a mutant copy of this gene, GLR-1, show short-term
habituation but do not show long-term memory for habituation training
(Rose, Kaun, Chen, & Rankin, 2003). Trafficking of glutamate receptor sub-
units is a common theme in studies of plasticity in various organisms.
Changes in glutamate receptor subunit expression and localization correlate

with long-term plasticity changes in the cellular models of memory, long
term potentiation (LTP) and long term depression (LTD) in rodents
(reviewed in Citri & Malenka, 2007) and in Aplysia (reviewed in Glanzman,
2007). In LTP, an increase in glutamate receptor subunit expression is
observed at the synapses that are stimulated, whereas in LTD, a decrease
in the expression of these subunits is seen. The similarity of mechanism
between LTD and long-term memory of habituation suggests a relationship
between these two processes that should be investigated in greater detail.
C. elegans also shows characteristic reconsolidation of memories after
recall. Reconsolidation is the process of restoring a recalled memory,
allowing memories to be weakened, strengthened, or altered depending
upon the animal’s experiences when the memory is recalled. This process
also occurs across phylogeny from the snail Lymnaea to appetitive and
aversive learning paradigms in rodents to humans (reviewed in Nader,
2003). This was demonstrated in C. elegans by first giving worms long-
term habituation training and 24 h later administering a reminder of the
training (10 taps) followed immediately by heat shock to block protein
synthesis. When these animals were tested for memory for the original
training 24 h after the reminder, the response magnitudes of trained ani-
mals matched response levels of untrained animals; thus, the inhibitory
effects of heat shock on protein synthesis disrupted memory reconsolida-
tion. Further investigation with confocal imaging of GLR-1 tagged with GFP
after reconsolidation blockade showed that not only was the behavioral
decrement eliminated, but so too was the downregulation of GLR-1 that is
correlated with the long-term memory. (Rose & Rankin, 2006). These
results suggest that the conditions that impair memory consolidation simi-
larly disrupt memory reconsolidation, suggesting that similar mechanisms
are involved. These and other studies on reconsolidation suggest that the
notion that memories are susceptible to restructuring after they are
recalled is also a primitive characteristic of memory.
5. Conclusion
This chapter attempts to explain how C. elegans is and is not a model of

human learning and memory. The simple worm is not a model for complex
cognitive forms of learning and memory, such as remembering life events
(e.g., its first birthday) or complex sequences of behavior. It does, though,
show simple forms of associative (e.g., classical conditioning) and nonas-
sociative (e.g., habituation) learning and memory. These simple forms of
learning are present in higher animals, and thus C. elegans is an excellent
model in which to study them. Learning and memory in this species uses
many of the same genes and molecules that mammals use (such as com-
ponents of the glutamate and dopamine neurotransmitter systems and
intracellular signaling pathways), but is a technologically easier animal in
which to discover and study them. It exhibits the following elements and
characteristics of memory that are present in all species:
(1) It is sensitive to temporal spacing during training.

(2) It can be retained for short or very long periods of time depending on
the training protocol.
(3) It relies on the production of new gene products.
(4) It undergoes reconsolidation after it is recalled.
This demonstrates that not only is C. elegans a good model for studying
simple forms of learning and memory that are present in humans, but also that
these forms of plasticity developed early in animal evolution and have been
maintained through natural selection for hundreds of millions of years. Thus,
revealing how critical a process this is to animal survival and reproduction.
References
Askew, H. R. (1970). Effects of stimulus intensity and intertrial interval on habitua-

tion of the head-shake response in the rat. Journal of Comparative Physiological
Psychology, 72, 492.
Beck, C. D. O., Rankin, C. H. (1995). Heat shock disrupts long-term memory con-
solidation in Caenorhabditis elegans. Learning and Memory, 2, 161.
Brenner, S. (1974). The genetics of Caenorhabditis elegans. Genetics, 77, 71.
Brockie, P. J., Madsen, D. M., Maricq, A. V. (1997). Genetic analysis of two
C. elegans putative NMDA receptor subunits nmr-1 and nmr-2. Society for Neu-
roscience Abstract, 23, 936.
Byrne, J. H. (1982). Analysis of synaptic depression contributing to habituation of gill-
withdrawal reflex in Aplysia californica. Journal of Neurophysiology, 48, 431.
Cepeda, N. J., Pashler, H., Vul, E., Wixted, J. T., Rohrer, D. (2006). Distributed
practice in verbal recall tasks: a review and quantitative synthesis. Psychological
Bulletin, 132, 354.
Citri, A., & Malenka, R. C. (2007). Synaptic plasticity: Multiple forms, functions, and
mechanisms. Neuropsychopharmacology Review, 33(1), 1–24.
Commins, S., Cunningham, L., Harvey, D., Walsh, D. (2003). Massed but not spaced
training impairs spatial memory. Behavioural Brain Research, 139, 215.
Davis, M. (1970). Effects of interstimulus interval length and variability on startle-
response habituation in the rat. Journal of Comparative Physiological Psychology,
72, 177.
Dent, J. A., Davis, M. W., Avery, L. (1997). avr-15 encodes a chloride channel
subunit that mediates inhibitory glutamatergic neurotransmission and ivermectin
sensitivity in Caenorhabditis elegans. EMBO Journal, 16: 5867.
Donovan, J. J., Radosevich, D. J. (1999). A meta-analytical review of the distribution

of practice effect. Journal of Applied Psychology, 84, 795.
Ebbinghaus, H. (1964). Memory: A Contribution to Experimental Psychology.
Ruger, H.A., Bussenius, C.E., Hilgard, E.R. (trld.), Dover, New York (Original
work published 1885).
Ebrahimi, C. M., Rankin, C. H. (2006). Early patterned stimulation leads to changes
in adult behavior and gene expression in C. elegans. Genes, Brain, and Beha-
viour, 6, 517.
Genome Sequencing Center (Washington University in St. Louis). (2008). Caenor-
habditis elegans [online]. Available: http://genome.wustl.edu/genome.cgi?GEN-
OME=Caenorhabditis%20elegans [accessed 28 February 2008].
Glanzman, D. L. (2007). Simple minds: the neurobiology of invertebrate learning and
memory. In North, G., Greenspan, R. J. (Eds.), Invertebrate neurobiology
(pp. 347–380). Cold Spring Harbor Laboratory Press.
Goodrick, C. L. (1973). Maze learning of mature–young and aged rats as a function
of distributed practice. Journal of Experimental Psychology, 98, 344.
Groves, P. M., Thompson, R. F. (1970). Habituation: A dual process theory. Psycho-
logical Review, 77, 419.
Hall, D. H., Lints, R., Altan, Z., Eric, A. (2005). Nematode neurons: Anatomy and
anatomical methods in Caenorhabditis elegans. In Aamodt, E. (Ed.), Interna-
tional review of neurobiology, Vol. 69., Elsevier, pp. 1–35. San Diego, California.
Hart, A. C., Sims, S., Kaplan, J. M. (1995). Synaptic code for sensory modalities
revealed by C. elegans GLR-1 glutamate receptor. Nature, 378, 82.
Hasegawa, N., Shimamura, K., Suzuki, K. (1988). Studies on the development of
water maze-learning ability in rats (3). Effects of the learning schedule on learn-
ing acquisition. Jikken Dobutsu, 37, 297.
Hedgecock, E. M., & Russel, R. L. (1975). Normal and mutant thermotaxis in the
nematode Caenorhabditis elegans. Proceedings of the National Academy of
Science, USA 72, 4061.
International Human Genome Sequencing Consortium. (2004). Finishing the
euchromatic sequence of the human genome. Nature, 431, 931.
Janiszewski, C., Noel, H., Sawyer, A. G. (2003). A meta-analysis of the spacing effect
in verbal learning: implications for research on advertisement repetition and
consumer memory. Journal of Consumer Research, 30, 138.
Jost, A. (1897). Die Assoziatiationsfestigkeit in iher Abhängigkeit von der Verteilung de
Wiederholungen [The strength of associations in their dependence on distribution of
repetitions]. Zeitschrift für Psychologie und Physiologie der Sinnesorgane, 14, 436.
Kindt, K. S., Quast, K. B., Giles, A. C., De, S., Hendrey, D., Nicastro, I., Rankin, C. H.,
Schaffer, W. R. (2007). Dopamine mediates context-dependent modulation of
sensory plasticity in C. elegans. Neuron, 55, 662
Koh, P. O., Bergson, C., Undie, A. S., Goldman-Rakic, P. S., Lidow, M. S. (2003). Up-
regulation of D1 dopamine receptor-interacting protein, calcyon, in patients with
schizophrenia. Archives of General Psychiatry, 60, 311.
Lee, R. Y. N., Sawin, E. R., Chalfie, M., Horvitz, H. R., Avery L. (1999). EAT-4, a
homolog of a mammalian sodium-dependent inorganic phosphate cotransporter,
is necessary for glutamatergic neurotransmission in Caenorhabditis elegans.
Journal of Neuroscience, 19, 159.
Lee, T. D., Genovese, E. D. (1988). Distribution of practice in motor skill acquisition:

learning and performance effects reconsidered. Research Quarterly for Exercise
and Sport, 59, 277.
Maricq, A. V., Peckol, E., Driscoll, M., Bargmann, C. I. (1995). Mechanosensory
signaling in C. elegans mediated by the GLR-1 glutamate receptor. Nature, 378, 78.
Mirnics, K., Lewis, D. A. (2001). Genes and subtypes of schizophrenia. Trends in
Molecular Medicine, 7, 281.
Mohri, A., Kodama, E., Kimura, K. D., Koike, M., Mizuno, T., Mori, I. (2005). Genetic
control of temperature preference in the nematode Caenorhabditis elegans.
Genetics, 169, 1437.
Moss, V. D. (1996). The efficacy of massed versus distributed practice as a function
of desired learning outcomes and grade level of the student (Doctoral disserta-
tion, Utah State University). Dissertation Abstracts International, 56, 5204.
Nader, K. (2003). Memory traces unbound. Trends in Neuroscience, 26, 65.
Rankin, C. H. (2000). Context conditioning in habituation in the nematode C.
elegans. Behavioural Neuroscience, 114, 496.
Rankin, C. H., Beck, C. D. O., Chiba, C. M. (1990). Caenorhabditis elegans: a new
model system for learning and memory. Behavioural Brain Research, 37, 89.
Rankin, C. H., Broster, B. S. (1992). Factors affecting habituation and recovery from
habituation in the nematode Caenorhabditis elegans. Behavioural Neuroscience,
106, 239.
Rankin, C. H., Dubnau, J. (2007). Memories of worms and flies. In North, G.,
Greenspan, R. J. (Eds.), Invertebrate neurobiology (pp. 309–346). Cold Spring
Harbor Laboratory Press, Cold Spring Harbor, New York.
Rose, J. K., Kaun, K. R., Chen, S. H., Rankin, C. H. (2003). GLR-1, a non-NMDA
glutamate receptor homolog, is critical for long-term memory in Caenorhabditis
elegans. Journal of Neuroscience, 23, 9595.
Rose, J. K., Kaun, K. R., Rankin, C. H. (2002). A new group training procedure for
habituation demonstrates that presynaptic glutamate release contributes to long-
term memory in C. elegans. Learning and Memory, 9, 130.
Rose, J. K., Rankin, C. H. (2006). Blocking memory reconsolidation reverses memory-
associated changes in glutamate receptor expression. Journal of Neuroscience,
26, 11582.
Sanyal, S., Wintle, R. F., Kindt, K., Nuttley, W. M., Fitzmaurice, P., Bigras, E., Merz, D.,
Herbert, T. E., van der Kooy, D., Schafer, W. R., Culotti, J. G., Van Tol, H. H. M.
(2004). Dopamine modulates the plasticity of mechanosensory responses in
C. elegans. EMBO Journal, 23, 473.
Seamans, J. K., Yang, C. R. (2004). The principal features and mechanisms of
dopamine modulation in the prefrontal cortex. Progress in Neurobiology, 74, 1.
Sulston, J., Dew, M., Brenner, S. (1975). Dopaminergic neurons in the nematode
Caenorhabditis elegans. Journal of Comparative Neurology, 163, 215.
The C. elegans Sequencing Consortium (1998). Genome sequence of the nematode
Caenorhabditis elegans. A platform for investigating biology. Science, 282, 2012.
Thorndike, E. L. (1912) The curve of work. Psychological Review, 19, 165.
Tsalik, E. L., Niacaris, T., Wenick, A. S., Pau, K., Avery, L., Hobert, O. (2003). LIM
homeobox gene-dependent expression of biogenic amine receptors in restricted
regions of the C. elegans nervous system. Developmental Biology, 263, 81.
Wen, J. Y. M., Kumar, N., Morrison, G., Rambaldini, G., Runciman, S., Rousseau, J.,
van der Kooy, D. (1997). Mutations that prevent associative learning in C. elegans.
Behavioural Neuroscience, 111, 354.
Wicks, S. R, Rankin, C. H. (1997). The effects of tap withdrawal response habitua-
tion on other withdrawal behaviors: the localization of habituation in C. elegans.
Behavioural Neuroscience, 111, 1.
Wicks, S. R., Rankin, C. H. (1995). Integration of mechanosensory stimuli in Cae-
norhabditis elegans. Journal of Neuroscience, 15, 2434.
Wicks, S. R., Rankin, C. H. (2000). Mutations of the Caenorhabditis elegans
brain-specific inorganic phosphate transporter eat-4 affect habituation of the
tap-withdrawal response without affecting the response itself. Journal of Neu-
roscience, 20, 4337.
Zang, Y., Lu, H., Bargmann, C. I. (2005). Pathogenic bacteria induce aversive
olfactory learning in Caenorhabditis elegans. Nature, 438, 179.
Human Learning 243
The Cell Biology of Learning

and Memory in Aplysia
David L. Glanzman
Departments of Physiological Science and Neurobiology and the Brain
Research Institute, University of California, Los Angeles, CA, USA
These scenes . . . why do they survive undamaged year after year unless
they are made of something comparatively permanent?
Virginia Woolf, Sketch of the Past (1953)
1. Introduction
How do we remember? This question has preoccupied artists, philo-

sophers, and scientists throughout history. It is fair to say, however,
that the last 40 years has witnessed the most significant progress
toward answering this important question since humans were first
able to articulate it. Indeed, we are now close to being able to actually
identify those ‘‘comparatively permanent’’ changes in our brains that
allow us to retain the scenes of our lives. A major reason for the recent
progress in solving the scientific problem of memory has been the use
of model systems that have permitted us to tackle the problem using
tools of modern biology, including genetic, neurophysiological, and,
more recently, molecular biological, tools. Each of the model systems
that has proved so valuable in research into the biology of memory—
such as Drosophila, Caenorhabditis elegans, Aplysia, rat, and mouse—
offers specific experimental advantages that have been ingeniously
exploited by scientists of memory during the past four decades.
A major insight that has emerged from the study of learning and memory
in these diverse organisms is that the cellular and molecular mechan-
isms of memory have been highly conserved throughout evolution.
Somewhat astonishingly, the biological bases of memory in an organism
244 David L. Glanzman
with only 302 neurons, such as C. elegans, appear to differ little, if at all,
from that in animals, such as a rat or a human, that possess central
nervous systems (CNSs) greater than five orders of magnitude more
complex. As a result, it matters little whether one studies the genetic
basis of learning in the fruit fly, its synaptic basis in the sea snail, or its
molecular basis in a transgenic mouse; current evidence suggests that
the answers from these studies are likely to converge onto a common set
of mechanisms. This is one of the more remarkable conclusions of
modern biology, although it was anticipated, like so many other biologi-
cal discoveries, by Charles Darwin. In his final work, a study of the
habits of earthworms (Darwin, 1881), Darwin wrote that ‘‘worms,
although standing low in the scale of organization possess some degree
of intelligence’’ (p.98). Darwin did not use the word ‘‘intelligence’’ loosely;
rather, he was explicitly comparing the ‘‘mental’’ abilities of worms and
humans. Darwin believed that the distinction between the cognitive
capabilities of these relatively simple invertebrates and those of humans
was quantitative rather than qualitative. Given this—and despite the
almost total lack of knowledge regarding the physical basis of memory
in his day—Darwin would not have been surprised by the striking con-
vergence of results from biological investigations of learning and mem-
ory in such phylogenetically disparate organisms as flies, worms, snails,
mice, and rats.
Studies of the marine snail Aplysia californica have contributed sig-
nificantly to our modern understanding of the biology of learning and
memory (Fig. 1). As appreciated early on by Eric Kandel (Kandel, 2006),
this humble mollusk possesses several major advantages for cell biolo-
gical investigations of memory. Its CNS contains only about 20 000
neurons; although not trivial, this Figure is dwarfed by the 200 million
neurons in the rat brain and the 100 billion in the human brain. Further-
more, many, although by no means all, of the neurons in Aplysia are
quite large; some of the motor neurons range from 100 mM to 1 mm in
diameter (Frazier, Kandel, Kupfermann, Waziri, & Coggeshall, 1967). The
large size of Aplysia neurons was a significant boon to electrophysiolo-
gists in the era before the advent of whole-cell patch-type recording. In
addition, the nervous system of Aplysia, like those of many inverte-
brates, is characterized by possessing identified neurons. These are
typically large neurons that can be found in the same location in the
nervous system in every member of a particular species; moreover, every
instance of a given identified neuron will commonly utilize the same
neurotransmitter(s), as well as have the same electrical properties, phy-
sical characteristics, basic functional synaptic connections, and the same
physiological and/or behavioral roles (Kandel, 1976). On the order of one
(a)
Mantle shelf
Siphon
Gill
(b)
Parapodium
Mantle shelf
Siphon
Gill
Fig. 1. (a) Illustration of the gastropod mollusk Aplysia californica, showing

the siphon, gill, and mantle. (b) The gill- and siphon-withdrawal reflex of
Aplysia, shown from the side. The reflex can be elicited by weak tactile
stimulation (here, a jet of water from a WaterPik). The gill- and siphon-
withdrawal reflex is mediated by neurons in the abdominal ganglion. From
Glanzman (2007).
hundred identified neurons have been described in Aplysia (Kandel,

1976). (This figure does not include several homogeneous clusters of
smaller neurons, such as the sensory neuron clusters, found in most
central ganglia of Aplysia.) An advantage of identified neurons is that
one can more readily relate neurophysiological changes, such as
alterations in excitability or in synaptic connectivity, in an identified

neuron in Aplysia to a learned behavioral change than in, say, a single
pyramidal neuron in the CA1 region of the mammalian hippocampus.
A final advantage of Aplysia as a model system for learning and memory
is that the neural circuits that mediate some of the animal’s reflexive
behaviors can be dissociated from the CNS and reconstituted in disso-
ciated cell culture (Montarolo et al., 1986; Rayport & Schacher, 1986).
These in vitro preparations permit rigorous cellular and molecular inves-
tigations of learning-related synaptic change. Indeed, as will be described
in this review, many of the most important insights into the basic
mechanisms of learning in Aplysia have come from investigations of
synapses in cell culture.
Of course, none of these advantages would concern students of
memory if sea snails could not learn. However, as shown by Kandel
and his colleagues and others, Aplysia is a robust learner, exhibiting
several forms of nonassociative and associative learning, among them
habituation (Pinsker, Kupfermann, Castellucci, & Kandel, 1970), sensi-
tization (Carew, Castellucci, & Kandel, 1971), classical conditioning
(Carew, Walters, & Kandel, 1981; Lukowiak & Sahley, 1981), and oper-
ant conditioning (Brembs, Lorenzetti, Reyes, Baxter, & Byrne, 2002;
Cook & Carew, 1986; Hawkins, Clark, & Kandel, 2006). Furthermore,
neural correlates of each of these forms of learning have been identi-
fied. More importantly, investigators have succeeded in achieving in
Aplysia something that has eluded students of learning in more com-
plex organisms: the unambiguous demonstration that specific neuronal
changes actually contribute to learned behavioral changes (Antonov,
Kandel, & Hawkins, 1999; Antonov, Antonova, Kandel, & Hawkins,
2003; Li, Roberts, & Glanzman, 2005).
In this chapter, I will focus on what is currently known in Aplysia about
the cellular mechanisms of sensitization and dishabituation, a cognate form
of learning. I have selected sensitization and dishabituation as the subject
of this review because they are perhaps the best understood forms of
learning in Aplysia, and because several exciting discoveries have been
recently made regarding sensitization and dishabituation. These discov-
eries, in addition to significantly advancing our understanding of learning
in Aplysia, dramatically illustrate the extent to which biological mechan-
isms of learning have been conserved during evolution. The interested
reader can refer to Glanzman (2006); Hawkins, Kandel, & Bailey (2006);
and Roberts & Glanzman (2003) for recent reviews of habituation and
classical conditioning in Aplysia. Also, note that in the interest of concise-
ness, I will use the term sensitization to refer to both sensitization and
dishabituation, except where I explicitly wish to refer to dishabituation
alone.
2. Short-Term Sensitization and Dishabituation in Aplysia:

Presynaptic Mechanisms
When disturbed by a sudden tactile stimulus, Aplysia reacts with a

defensive withdrawal reflex. Two organs are involved in this reflex, the
gill and the siphon, a tubular organ that pumps out deoxygenated water
from the branchial cavity during respiration. During the reflex, the gill
contracts beneath the mantle shelf, and the siphon contracts beneath the
parapodia, two large flaps of skin that extend laterally from the animal’s
foot. The gill- and siphon-withdrawal reflex is mediated, in part, by a
monosynaptic pathway within the animal’s abdominal ganglion. This
monosynaptic pathway comprises central mechanoreceptive sensory
neurons and motor neurons. There are 100 sensory neurons within the
abdominal ganglion that innervate the siphon, gill, mantle, and branchial
cavity of Aplysia (Byrne, Castellucci, & Kandel, 1974; Byrne, Castellucci,
& Kandel, 1978; Dubuc & Castellucci, 1991), and 50 motor neurons that
innervate the siphon and gill (Frost & Kandel, 1995; Koester & Kandel,
1977). Most of the research concerning learning-related neuronal plasti-
city in Aplysia has focused on changes in the strength of the monosynap-
tic connection between the central sensory and the motor neurons in
Aplysia (the sensorimotor synapse); consequently, this review will
concern only plasticity of the sensorimotor synapse. Notice that although
facilitation of the sensorimotor synapse has been shown to contribute
significantly to behavioral sensitization in Aplysia (Antonov et al., 1999),
changes within interneuronal pathways also make important
contributions to this form of learning (Cleary, Byrne, & Frost, 1995;
Zecevic et al., 1989).
A single noxious stimulus, such as an electrical shock applied to the
surface of the tail of Aplysia, produces sensitization of the withdrawal
reflex that persists for <30 min. This short-term form of sensitization
does not require the synthesis of new proteins (Sutton, Masters, Bagnall, &
Carew, 2001). Short-term sensitization is accompanied by, and is in
part due to, short-term facilitation (STF) of the sensorimotor synapse
(Antonov et al., 1999; Castellucci, Pinsker, Kupfermann, & Kandel, 1970;
Carew et al., 1971). Early quantitative analyses indicated that sensitiza-
tion-induced synaptic facilitation was due to enhancement of release of
neurotransmitter from the presynaptic sensory neuron (Castellucci &
Kandel, 1976). The first specific intrinsic presynaptic electrophysiologi-
cal change associated with sensitization in Aplysia was broadening of
the sensory neuron action potential (Klein & Kandel, 1978). This phe-
nomenon was observed in experiments in which the investigators
examined the effect of sensitization-related stimuli, such as whole
nerve stimulation, and application of the neurotransmitter serotonin
on the duration of the presynaptic action potential (spike). Serotonin

(5-HT) is an endogenous monoamine released within the CNS of Aply-
sia by sensitizing stimulation (Marinesco & Carew, 2002). Release of
5-HT is required for sensitization in Aplysia (Glanzman et al., 1989).
Furthermore, some of the serotonergic neurons that are activated by
sensitizing stimulation have been identified (Mackey, Kandel, & Hawkins,
1989; Marinesco, Kolkman, & Carew, 2004). How do 5-HT and sensitiz-
ing stimuli broaden the sensory neuron’s action potential? Current
evidence indicates that when 5-HT binds to its receptor it activates a
G-protein, which in turn stimulates adenylyl cyclase within the sensory
neuron. Stimulation of adenylyl cyclase causes synthesis of cyclic ade-
nosine 30 ,50 -monophosphate (cAMP) and activation of protein kinase
A (PKA). PKA activity results in the closure of potassium channels in
the cell membrane of the sensory neuron, which contributes to spike
broadening (see Byrne & Kandel, 1996 for review). (Notice that under
some conditions PKC appears to mediate spike broadening.) According
to a current model of presynaptic facilitation (Byrne & Kandel, 1996),
prolonging the action potential in the sensory neuron results in a
greater influx of calcium into the presynaptic terminal per action
potential, thereby producing enhanced release of the presynaptic trans-
mitter (although see Klein, 1994). Serotonin causes another, as yet
poorly understood, presynaptic change that is independent of spike
broadening and that also contributes to facilitation. This second pro-
cess, sometimes referred to as the (spike duration independent (SDI)
process, may involve enhanced mobilization of presynaptic vesicles
(Byrne & Kandel, 1996). The SDI process appears to play a greater
role in facilitation when the synapse is depressed (see Armitage &
Siegelbaum, 1998; Hochner, Klein, Schacher, & Kandel, 1986). There
may be additional processes that contribute to short-term presynaptic
facilitation of the sensorimotor synapse as well. For example, recent
evidence suggests that 5-HT may cause the switching on of previously
silent release sites in an all-or-none manner (Jiang & Abrams, 1998;
Royer, Coulson, & Klein, 2000).
3. Intermediate-Term Sensitization Memory: Involvement

of Pre- and Postsynaptic Mechanisms
Carew and colleagues have identified a second phase of sensitization

memory in Aplysia (Sutton et al., 2001; Sutton & Carew, 2002). Referred
to as intermediate-term memory (ITM), this phase of memory is both
temporally and mechanistically distinct from short-term memory. ITM
persists for 30 min to 3 h. Furthermore, ITM, unlike STM, depends on
new protein synthesis, but does not require gene transcription. There
are at least two forms of ITM; these differ in the method by which they
can be induced, as well as, to some extent, in their underlying mechan-
isms. Activity-independent ITM can be induced by multiple (e.g., five)
sensitizing shocks; furthermore, its expression depends on persistent
PKA activity (Sutton et al., 2001). By contrast, activity-dependent ITM
can be induced by only a single tail shock; it is revealed when the test
stimulus is applied to same site on the animal’s body as the sensitizing
stimulus (Sutton, Bagnall, Sharma, Shobe, & Carew, 2004). (Activity-
dependent ITM is also referred to as site-specific ITM.) Activity-dependent
ITM, unlike activity-dependent ITM, does not require PKA activity for
either its induction or its expression; rather, both the induction and the
expression of activity-dependent ITM require PKC activity. However, the
induction of activity-dependent ITM and its expression appear to depend
on different isoforms of PKC. This distinction was revealed by experi-
ments that used different PKC inhibitors. Sutton et al. (2004) showed that
the inhibitor calphostin C, which interacts specifically with the regulatory
domain of PKC, blocked the induction of activity-dependent ITM when
applied before sensitization training, but did not disrupt ITM when
applied 30 min after induction. These results imply that the expression
of activity-dependent ITM does not depend on isoforms of PKC posses-
sing a regulatory domain. This conclusion was reinforced by the results
from experiments with two other PKC inhibitors, chelerythrine and bisin-
dolylmaleimide I (Bis I). Chelerythrine inhibits all PKC isoforms
(Laudanna, Mochly-Rosen, Liron, Constantin, & Butcher, 1998), whereas
Bis I most effectively inhibits calcium-dependent (classical) and calcium-
independent (novel) isoforms of PKC, but is a relatively weak inhibitor of
atypical isoforms (Martiny-Baron et al., 1993). One atypical isoform
of PKC that is of particular interest is the PKC catalytic fragment
(PKM). Typically, PKM is generated by cleavage of PKC between the
regulatory and the catalytic domains by the calcium-dependent protease
(calpain), resulting in an autonomously active catalytic PKC fragment
(PKM) (Sossin, 2007). One isoform of PKM in mammals, PKMz, has
been implicated in the maintenance of both hippocampal long-term poten-
tiation (LTP) and memory (Ling et al., 2002; Pastalkova et al., 2006).
Recently, an atypical PKC isoform (Apl III) has been cloned from Aplysia
(Bougie et al., 2007). Furthermore, a PKM form may be generated from
Apl III by proteolytic cleavage. It seems likely that Apl III plays a major
role in the expression of activity-dependent ITM, whereas the induction of
this form of sensitization memory depends on another PKC isoform,
either a classical or a novel form (see below). In support of this idea,
inhibitors of calpain have been shown to block the induction of activity-
dependent ITM, but not its expression (Sutton et al., 2004). This result
suggests that calpain must be present during training in order to generate

the Apl III required to maintain activity-dependent ITM; once Apl III has
been generated, however, calpain is no longer necessary for ITM.
Synaptic parallels for both activity-independent and activity-dependent
ITM have been identified. Five spaced 5-min pulses of 5-HT produce activity-
independent intermediate-term facilitation (ITF) of the sensorimotor synapse
(Ghirardi, Montarolo, & Kandel, 1995; Muller & Carew, 1998). Like activity-
independent ITM, activity-independent ITF requires new protein synthesis and
persistent PKA activation, but not transcription. A single 5-min pulse of 5-HT
produces facilitation of the sensorimotor synapse that decays within 20 min
(Bao, Kandel, & Hawkins, 1998; Montarolo et al., 1986; Sutton & Carew, 2000).
This STF, like STM, does not depend on either protein synthesis or gene
transcription. However, when a single pulse of 5-HT is paired with brief
activation of the sensory neurons, a more persistent facilitation, lasting
20–65 min, is induced. This form of persistent facilitation—known variously
as associative, pairing-specific, or activity-dependent ITF—depends on
translation, but not transcription (Sutton & Carew, 2000). Furthermore,
activity-dependent ITF, like activity-dependent ITM, requires persistent PKC
activation for its expression.
An important question is whether activity-dependent ITF depends on pre-
or postsynaptic PKC or both. This question has been addressed in a recent
study by Zhao and colleagues (Zhao et al., 2006). These investigators
expressed fluorescently tagged Aplysia isoforms of PKC in either sensory
or motor neurons of sensorimotor cocultures. They used two PKC isoforms,
Apl I and Apl II; Apl I is a homolog of the vertebrate classical (calcium-
activated) PKC isoform, whereas Apl II is a homolog of the vertebrate novel
(calcium-independent) PKC isoform (Sossin, 2007). Under basal conditions,
PKC is present in the cytoplasm of neurons. However, stimulation by a
modulatory signal, such as 5-HT, can cause both Apl I and Apl II to translo-
cate to the cell membrane, where they phosphorylate downstream proteins.
Zhao et al. observed that a 5-min application of 5-HT, which produces STF,
caused Apl II, but not Apl I, to translocate to the sensory neuron membrane.
The 5-min 5-HT treatment did not cause translocation of Apl II (or Apl I) in
the motor neuron. Zhao et al. next examined the effect of paired stimulation
on the translocation of the PKC isoforms. Coincident 5-HT application plus
sensory neuron activation, which produces activity-dependent ITF, caused
translocation of the calcium-dependent PKC isoform, Apl I, to the sensory
neuron cell membrane, but did not translocate Apl II. Similarly, overexpres-
sion of a dominant negative form of Apl I in the sensory neuron blocked
activity-dependent ITF in sensorimotor cocultures; overexpression of a domi-
nant negative form of Apl II in the sensory neuron, however, did not disrupt
ITF. Paired stimulation also produced a much more modest translocation of
Apl I in the motor neuron. Zhao et al. did not examine the potential effect on
translocation of Apl I or Apl II in either sensory or motor neurons following

training that induces activity-independent ITF. Nor did they examine the
potential role of pre- or postsynaptic Apl III (Bougie et al., 2006) in ITF.
In our laboratory, we have focused on the role of postsynaptic mechan-
isms in activity-independent ITF. In our experiments ITF is induced by a
single, 10-min application of 5-HT. To isolate the postsynaptic contribu-
tion to ITF, our initial experiments were performed using isolated siphon
motor neurons in dissociated cell culture. The motor neurons were sti-
mulated by brief pulses (puffs) of glutamate, the transmitter used by
Aplysia sensory neurons (Dale & Kandel, 1993; Levenson et al., 2000).
The glutamate was pressure ejected onto the isolated motor neuron using
a puffer pipette, typically at a rate of once per 10 s, and the glutamate
potential evoked in the motor neuron (the Glu-EP) was recorded with a
sharp electrode. When 5-HT was applied for 10 min to the bathing med-
ium, there was a slow onset (5–10 min) enhancement of the Glu-EP
(Chitwood, Li, & Glanzman, 2001). This enhancement persisted for
>40 min after washout of the 5-HT and therefore lies within the range
of intermediate-term facilitation (ITF). (In more recent experiments, we
have found that the enhancement produced by 10 min of 5-HT treatment
lasts for >2 h [G. Villareal and D. L. Glanzman, unpublished].) By com-
parison, a 1-min 5-HT application produces only short-term enhancement
of the Glu-EP (Fig. 2). The 5-HT-induced enhancement of the Glu-EP
depends on both G-protein stimulation and a rise in intracellular calcium
within the motor neuron, because it is blocked by inhibitors of G-protein
activation and by the rapid calcium chelator BAPTA. Furthermore, 5-HT
selectively enhances the AMPA-type receptor-mediated component of the
motor neuron’s response to glutamate. The latter result suggests that
5-HT might be modulating the functional efficacy of AMPA-type receptors
in the motor neuron. Studies in mammals have identified modulation of
AMPA receptor trafficking as a major mechanism underlying LTP of
synapses in the CA1 of the hippocampus (Malinow & Malenka, 2002), as
well at least some forms of learning and memory (Rumpel, LeDoux,
Zador, & Malinow, 2005). One current model of LTP proposes that activa-
tion of NMDA receptors causes additional AMPA receptors to be inserted
into postsynaptic membranes through exocytosis (Passafaro, Piech, &
Sheng, 2001). To test whether 5-HT causes exocytotic insertion of
AMPA receptors into the cell membrane of Aplysia motor neurons, we
injected botulinum toxin, an inhibitor of exocytosis, into motor neurons
before 5-HT treatment. The botulinum toxin blocked 5-HT-dependent
enhancement of the Glu-EP (Chitwood et al., 2001).
Next, we turned to experiments using sensorimotor synapses in disso-
ciated cell culture (Lin & Glanzman, 1994) to test the involvement of
postsynaptic mechanisms in actual activity-independent ITF. A 10-min
(a)
Glutamate
(b1) (b2)
(% Mean – 10–0 min Glu-EPs)
200
Glu-EP amplitude
5-HT/10 min 150
5-HT/1 min
100
Control
1 2
50
1 2
0
–10 –5 0 5 10 15 20 25 30 35 40 45 50
Time (min)
(c1) (c2)
200
Glu-EP amplitude
150
5-HT
Control 100
1 2
50
1 2
0
–10 –5 0 5 10 15 20 25 30 35 40 45 50
Time (min)
Fig. 2. Ten minutes, but not 1 min, of 5-HT stimulation causes prolonged enhancement
of the glutamate-evoked response in motor neurons. (a) Composite micrograph and
cartoon depicting the experimental arrangement. The cell cultures consisted
exclusively of isolated small siphon (LFS) motor neurons (Frost, Clark, & Kandel,
1988). Pulses of glutamate were pressure-ejected from a micropipette (right) onto the
initial segment of the major neurite of the motor neuron once every 10 s (except for the
experiment shown in c). Fast green was used to visualize the glutamate. The evoked
glutamate potentials (Glu-EPs) were recorded from the motor neuron’s cell body using
a sharp microelectrode (left). Scale bar, 50 mm. (b1) sample Glu-EPs. Each pair of
traces shows responses from one experiment. Traces marked ‘‘1’’ represent sample
Glu-EPs evoked at the 5-min time point from the experiments summarized in b2; those
marked ‘‘2’’ represent Glu-EPs evoked at the 40-min time point. Scale bars, 10 mV and
500 ms. (b2) Comparison of the effects of 10 min (black bar) and 1 min (gray bar) of 5-HT
stimulation. Each symbol in the graph represents the mean normalized amplitude of
treatment with 5-HT-induced enhancement of the sensorimotor excitatory

postsynaptic potential (EPSP) that persisted for >40 min. Similar to
5-HT-induced enhancement of the Glu-EP in the isolated motor neuron,
ITF depended on elevated postsynaptic calcium and postsynaptic exocy-
tosis. Moreover, we found that injecting cell membrane-impermeant
inhibitors of intracellular calcium release into the motor neuron before
5-HT application blocked activity-independent ITF. These results
implicated release from intracellular stores in the 5-HT-dependent rise in
postsynaptic calcium. We extended these results from experiments on
cultured synapses to synaptic connections in the abdominal ganglion. In
the experiments on sensorimotor connections in the ganglion, electrical
stimulation of the pedal (tail) nerves was used to mimic the application of a
natural sensitizing stimulus, such as tail shock. Tail nerve stimulation
resulted in persistent facilitation of central sensorimotor synapses that
required elevated postsynaptic calcium and release of calcium
from intracellular stores. Importantly, we also observed that tail nerve
stimulation yielded significantly greater facilitation of the AMPA
receptor-mediated component of the sensorimotor EPSP than of the
NMDA receptor-mediated component. This result is inconsistent with an
exclusively presynaptic model for ITF and provides additional support for
the idea that ITF differs from STF in requiring a contribution from post-
synaptic pathways. Finally, we performed a rigorous test of the involve-
ment of postsynaptic processes in actual learning, rather than in merely
synaptic facilitation. For this test, we used a reduced preparation of Aply-
sia, in which the siphon and tail were preserved, together with the CNS and
the peripheral nerves that connected these organs to the CNS. The siphon
was partially split; one half was pinned to the bottom of the experimental
chamber, while the other half was left free to move. The pinned half of
the siphon was stimulated with weak electrical shocks once every 5 min.
Fig. 2. (Continued) six consecutive Glu-EPs. Motor neurons received 10 min of 5-HT
(n = 9), 1 min of 5-HT (n = 10), or perfusion medium alone (n = 10; control group). The
numbers below the data indicate the times at which the sample Glu-EPs shown in b1
were recorded. Error bars represent –SEM. From Villareal et al. (2007). (c) 5-HT-
dependent enhancement of the glutamate response can be elicited at low rates of test
stimulation. (c1) Sample Glu-EPs from experiments in which glutamate was applied to
the motor neuron once every 5 min. Scale bars, 10 mV and 500 ms. (c2) Effect of 5-HT
when the motor neuron was stimulated with glutamate at a low rate. Each symbol
represents the mean normalized amplitude of six Glu-EPs. Values were normalized to
the Glu-EP recorded immediately before 5-HT application (t = 0 min). Either 5-HT
(n = 7) or normal perfusion medium (n = 6, control group) was applied for 10 min.
The numbers below the data indicate the times at which the sample Glu-EPs shown
in c1 were recorded. Error bars represent –SEM. From Villareal et al. (2007).
A movement transducer was connected to the unpinned half of the

siphon to record the movement of the siphon in response to the weak
shocks. The weak stimulation produced habituation of the siphon
response. After the sixth trial, strong electrical shocks were delivered to
the tail of the preparation; these produced dishabituation of the siphon
response in control preparations. However, in preparations in which botu-
linum toxin was injected into identified siphon motor neurons before the
onset of testing, the dishabituation was blocked (Li et al., 2005). Behavioral
dishabituation therefore requires exocytotic activity upstream from the
neuromuscular junction, because injecting botulinum toxin into
the motor neurons did not disrupt the basic siphon reflex. Together with
the evidence from the experiments on cultured motor neurons and
sensorimotor cocultures, this result represents strong evidence that
modulation of AMPA receptor trafficking mediates sensitization and
dishabituation in Aplysia.
Previous work, described above, has shown that ITF (and ITM) depend
on de novo protein synthesis (Ghirardi et al., 1995; Muller & Carew, 1998;
Sutton & Carew, 2000; Sutton et al., 2001). Accordingly, we asked whether
postsynaptic protein synthesis is necessary for ITF. To address this ques-
tion, we again performed experiments on isolated siphon motor neurons in
culture. We found that the enhancement of the glutamate response in
motor neurons required de novo protein synthesis. Interestingly, the pre-
sence in the bathing medium of inhibitors of protein synthesis, such as
emetine and cyclohexamide, blocked even early enhancement of the Glu-
EP—the enhancement produced during the 10-min 5-HT treatment—as
well as the enhancement that persisted after washout of 5-HT. The rapidity
with which inhibitors of protein synthesis disrupted 5-HT-dependent
enhancement suggested that 5-HT was triggering synthesis of proteins
locally, within the motor neurites, rather than, or in addition to, synthesis
of proteins in the motor neuron’s cell body (Fig. 3). To test this idea, we
used cultured neurites that had been severed from their cell bodies. These
experiments were made possible by the capability of neurites of inverte-
brate neurons, in contrast to mammalian dendrites and axons, to survive
for long periods after being severed from their cell body (Bittner, 1991).
After axotomizing the large gill motor neuron L7, we stimulated the surgi-
cally isolated neurite with glutamate and recorded the resulting Glu-EPs, as
we had done in our experiments on whole motor neurons. A 10-min treat-
ment with 5-HT induced persistent enhancement of the glutamate response
in the neurite, as it had in the intact motor neurons. When 5-HT was added
to the bathing medium in the presence of the protein synthesis inhibitor
emetine, the enhancement was blocked. This result provides an unambig-
uous demonstration that 5-HT produces local synthesis of proteins in
Aplysia motor neurons (Fig. 4). To confirm that ITF also depends on de
(a) (b)

200
Glu-EP amplitude
5-HT 150
5-HT/Emetine 100
Emetine
50 2
1
Control
0
1 2 –10 –5 0 5 10 15 20 25 30 35 40 45 50
Time (min)
(c) (d)
200
5-HT
Glu-EP amplitude
150
5-HT/Emetine
100
Emetine
50 2
Control 1
1 2
0
–10 –5 0 5 10 15 20 25 30 35 40 45 50
Time (min)
Fig. 3. Rapid protein synthesis is required for enhancement of the glutamate

response in motor neurons. (a) sample Glu-EPs. The motor neuron was stimulated
with glutamate once every 10 s. (b) Effect of starting emetine treatment before 5-HT
treatment. Neurons received 5-HT (black bar, n = 6), 5-HT in the presence of emetine
(dark gray bar, n = 6), emetine alone (n = 6), or neither 5-HT nor emetine (n = 6,
control group). The numbers below the data indicate the times at which the sample
Glu-EPs shown in (a) were recorded. Error bars represent – SEM. (c) Sample Glu-EPs
for the experiments shown in (d). (d) effect of simultaneous 5-HT and emetine
treatment. From Villareal et al. (2007).
novo protein synthesis in the motor neuron, we tested the effect of inhibit-
ing postsynaptic protein synthesis on activity-independent ITF using sen-
sorimotor cocultures. To isolate the inhibition of protein synthesis to the
postsynaptic cell, we injected gelonin, a cell membrane-impermeant inhi-
bitor of protein synthesis (Stirpe, Olsnes, & Pihl, 1980), into motor neurons
before treating them with 5-HT. Postsynaptic protein synthesis blocked the
ITF produced by the 5-HT treatment, but not the STF. Together with the
evidence from the experiments on isolated motor neurons/neurites, these
results indicate that ITF depends on local postsynaptic protein synthesis
(Fig. 5), and that inhibition of postsynaptic protein synthesis does not
(a) (a')
0h 24 h
Cut L7
(b) (a")
5-HT
5-HT/Emetine V
Glutamate
Emetine
Control
1 2
(c)
200
Glu-EP amplitude
150
100
50
1 2
0
–10 –5 0 5 10 15 20 25 30 35 40 45 50
Time (min)
Fig. 4. Role of local protein synthesis in enhancement of the glutamate-evoked

response. (a) A gill motor neuron L7 isolated in cell culture before soma
removal. The dashed line shows where the main neurite was severed. Scale
bar, 100 mm. (a0 ) The neurite was allowed to recover in 24–48 h after severing.
Note the swelling of the neurite proximal to the former site of the cell body.
Trial
(a) (b)
0 min 50 min
5-HT
Sensory Motor
Neuron Neuron
V 5-HT/Gelonin
Gelonin
Test alone
(c)
350
EPSP amplitude (%-5-min EPSP)
300
250
200
150
100
50
0
–5 0 5 10 15 20 25 30 35 40 45 50 55 60
Time (min)
Fig. 5. Postsynaptic inhibition of protein synthesis blocks intermediate-term synaptic

facilitation. (a) Experimental arrangement. Scale bar, 50 mm. (b) Representative EPSPs.
Trial times correspond to those shown in the graph in (c). (c) Mean normalized
amplitude of the EPSPs. Error bars represent SEM. From Villareal et al. (2007).
Fig. 4. (Continued) (a00 ) The neurite was impaled with a sharp electrode. The
stimulation and recording methods were like those used in the experiments that
used the LFS motor neurons with their cell bodies (Figs. 2 and 3). Glutamate was
applied to the isolated neurite every 10 s. (b) Sample Glu-EPs from one experiment.
Traces marked ‘‘1’’ represent sample Glu-EPs evoked at the 5-min time point from the
experiments summarized in b2; those marked ‘‘2’’ represent Glu-EPs evoked at the
40-min time point. Scale bars, 10 mV and 500 ms. (c) Effect of inhibiting protein
synthesis on enhancement of the glutamate response in the neurite. The numbers
below the data indicate the times at which the sample Glu-EPs shown in (b) were
recorded. Error bars represent –SEM. From Villareal et al. (2007).
disrupt STF. Recent work has indicated that local postsynaptic protein
synthesis is critical to synaptic plasticity and learning in mammals
(Pfeiffer & Huber, 2006; Sutton & Schuman, 2006). Our results represent
strong evidence that local postsynaptic protein synthesis plays a critical
role in invertebrate learning as well.
What intracellular signal might trigger local protein synthesis within
the motor neuron during ITF? One intriguing candidate for this post-
synaptic signal is PKC. Accordingly, we have examined the potential
involvement of PKC in 5-HT-induced enhancement of the glutamate
response in isolated motor neurons. We found that both chelerythrine
and Bis I block the induction of 5-HT-dependent enhancement of the
Glu-EP. But, whereas chelerythrine also blocks the expression of the
enhancement when it is applied after the enhancement is already estab-
lished, Bis I does not (G. Villareal, Q. Li and D. L. Glanzman, unpub-
lished). These results indicate that the induction of 5-HT-dependent
enhancement depends on a different PKC isoform than the expression.
Induction is probably triggered by the activity of either a classical or a
novel isoform of PKC. The results of Zhao et al. (2006) point to Apl I as
the likely inductive isoform. By contrast, the expression of the
enhancement, because it is resistant to inhibition by Bis I, is likely to
be mediated by the activity of Apl III, the Aplysia homolog of PKM
(Sossin, 2007). In support of this idea, a role for Apl III has been
proposed in the expression of activity-dependent ITF (Sutton et al.,
2004). The idea that a PKM-like molecule mediates the expression of
ITF is particularly attractive in light of our discovery of the role of local
postsynaptic protein synthesis in ITF. Recent evidence indicates that
hippocampal LTP is maintained, in part, by dendritic synthesis of PKMz
(Muslimov et al., 2004). Furthermore, the persistent activity of PKMz
within dendrites appears to play a key role in the insertion of new
AMPA receptors into the postsynaptic membrane during LTP expres-
sion (Hernandez et al., 2003). One possibility is that Apl I, or some
other postsynaptic signal, triggers the rapid, local synthesis of a PKM
form of Apl III, which then drives AMPA receptor insertion during ITF.
But current evidence indicates that the formation of PKM from Apl III
in Aplysia is accomplished by proteolysis rather than by local synthesis
from PKM mRNA (Bougie et al., 2006). Another possibility, therefore, is
that the 5-HT-induced rise in postsynaptic calcium, because of release
from intracellular stores, activates calpain, which leads to proteolysis
of Apl III and formation of PKM; PKM, in turn, may participate in the
local postsynaptic protein synthesis that characterizes ITF. An intri-
guing idea is that PKM causes the synthesis of additional AMPA recep-
tors, which are then inserted into the postsynaptic membrane through
exocytosis (Fig. 6).
(a) Tail
shock
5HT
PLC
G G
AC
NMDAR
NMDAR
DAG
cAMP
PKC Vesicle
PKA mobilization AMPAR
AMPAR
Sensory
Ca2+ K+
neuron
channel channel
(b) Tail
shock
5HT
PLC
PLC G
G G IP3
RyR
AC
AC
NMDAR
Ca2+
PKC CaMKII
Vesicle AMPAR PKC
PKA mobilization
P PKC
syn rotein P
syn rote
the the in
sis sis
Sensory
neuron Ca2+ K+
Retrograde Motor
channel channel signal (s)
neuron
Fig. 6. Cellular models for short-term facilitation (STF) and intermediate-term

facilitation of the sensorimotor synapse in Aplysia. The dashed lines depict pathways
or actions for which experimental evidence is currently lacking. (d) STF. This phase
lasts <30 min. STF is both induced and expressed presynaptically. (b) Intermediate-
term facilitation (ITF). This phase lasts 30 min to 3 h. ITF is induced postsynaptically, by
release of Ca2+ from intracellular stores in the motor neuron, and expressed both pre-
and postsynaptically. The presynaptic expression results from one or more retrograde
signals, activated by postsynaptic Ca2+, which stimulate both PKA and PKC within the
sensory neuron. One likely presynaptic mechanism of expression during ITF is
enhanced presynaptic release (see Wan & Abrams, 2008). ITF appears to involve two
different postsynaptic PKC isoforms; an early, possibly Ca2+-dependent isoform
participates in the induction of ITF, whereas a PKM form (Bougie et al., 2006) appears
to mediate the maintenance of ITF (see the text). Note that ITF depends on both pre- and
postsynaptic protein synthesis, but not gene transcription. From Glanzman (2007).
4. Long-term sensitization memory: requirement for transynaptic

signaling
Sensitization of the withdrawal reflex exhibits a long-term form (Pinsker,

Hening, Carew, & Kandel, 1973; Sutton et al., 2001), as well as short-
and intermediate-term forms. Long-term memory (LTM) for sensitization,
which persists for 24 h, is induced by the spaced application of multiple
sensitizing stimuli, such as shocks delivered to the tail of the animal. LTM
requires both protein synthesis and gene transcription (Castellucci,
Blumenfeld, Goelet, & Kandel, 1989; Sutton et al., 2001). Therefore, LTM
is distinctive from ITM not only temporally but also mechanistically. LTM
is known to involve persistent structural changes, among which are new
presynaptic outgrowth, including the appearance of new sites of presynap-
tic release (Bailey & Chen, 1983, 1988a, 1989; Glanzman, Kandel, & Scha-
cher, 1990; Wainwright, Byrne, & Cleary, 2004), as well as the growth of
new postsynaptic spine-like processes (Bailey & Chen, 1988b). Training
that results in LTM also produces long-term facilitation (LTF) of the sen-
sorimotor synapse (Frost, Castellucci, Hawkins, & Kandel, 1985; Zhang,
Goldsmith, & Byrne,1994). An in vitro analog of LTM-induced LTF can be
demonstrated for sensorimotor synapses in dissociated cell culture, as first
reported by Montarolo et al. (1986). Here, 5-HT is repeatedly applied to
sensorimotor cocultures using a spaced protocol. A commonly used
training protocol for LTF, referred to as the ‘‘5 5-HT’’ protocol, involves
applying 5-HT five times for 5 min per application; between applications
there is a 15-min washout period. This in vitro version of LTF has proved
enormously useful for a mechanistic analysis of LTM. Perhaps most impor-
tantly, LTF of sensorimotor synapses in dissociated cell culture has been
used to demonstrate the requirement for activity of the transcription factor
cyclic AMP response element-binding protein (CREB) in LTM (Bartsch
et al., 1995; Dash, Hochner, & Kandel, 1990; Kaang, Kandel, & Grant,
1993). Other major insights into the cellular and molecular machinery of
LTM gained through studies using in vitro preparations in Aplysia include
the recognition that the induction of LTM is normally inhibited through the
repression of the activational form of CREB (CREB1) by a repressor form
of CREB (CREB2) (Bartsch et al., 1995; Bartsch et al., 2000; Guan et al.,
2002); the insight that the capability for selectively enhancing only some
synapses and not others during learning can be preserved despite the
requirement for gene transcription during LTM through the ‘‘capture’’ by
activated synapses of learning-related gene products (Martin et al., 1997; Si
et al., 2003); the understanding of the importance of protein degradation
through the ubiquitin–proteosome pathway during LTM (Hegde et al.,
1997); and the recognition that LTM depends on epigenetic modifications
(Cohen-Armon et al., 2004; Guan et al., 2002). More recently, evidence from
experiments on LTF of the sensorimotor synapse in culture has contributed

to the idea that prions may function as a ‘‘switch’’ for LTM in Aplysia, and
possibly in other organisms (Si, Lindquist, & Kandel, 2003).
It is widely believed that the major cellular and molecular changes that
underlie LTM in Aplysia are predominantly, albeit perhaps not exclusively,
presynaptic. According to the predominant model of LTM (Hawkins et al.,
2006; Kandel, 2001), repeated, spaced sensitizing stimuli produce
prolonged release of 5-HT within the CNS of Aplysia. 5-HT binds to
G-protein-coupled receptors on the sensory neuron, and produces
enhanced synthesis of cAMP and persistent activation of PKA. Persistent
PKA activity has several downstream actions, particularly stimulation of
CREB-dependent gene transcription. CREB-mediated transcription of
immediate-early genes, in turn, yields other transcription factors, including
the CAAT box enhancer-binding protein (C/EBP). The successive waves of
gene transcription give rise to long-term changes within the sensory neu-
ron, including structural changes, such as the growth of new presynaptic
varicosities (release sites).
Although the experimental evidence for much of the above scheme of
LTM is strong, it is unlikely that LTM in Aplysia should rely solely on
presynaptic changes. First, as I have discussed above, ITM depends criti-
cally on postsynaptic changes. Second, LTM in vertebrates indisputably
involves postsynaptic processes, including NMDA receptor-dependent
LTP, modulation of postsynaptic AMPA receptors, and the growth of new
spines (Engert & Bonhoeffer, 1999; Park et al., 2006). The idea that the
basic mechanistic scheme of LTM in invertebrates would differ qualita-
tively from that of LTM in vertebrates is improbable. Evolution seldom
produces such an outcome.
Recently, we have reexamined the potential role of postsynaptic pro-
cesses in LTF in Aplysia. We have found that LTF, like ITF, requires
elevated postsynaptic calcium and postsynaptic protein synthesis. In addi-
tion, we have found evidence that LTF is maintained by transynaptic
signaling. Therefore, at least some of the persistent presynaptic changes
that have been associated with LTF appear to be triggered postsynaptically,
and involve an as yet unidentified retrograde signal. All of our experiments
were performed on cultured sensorimotor synapses. We used the 5 5-HT
training protocol to induce LTF.
In our first set of experiments, the calcium chelator BAPTA was injected
into the postsynaptic motor neuron in some cocultures 30 min before the
onset of training. We also included cocultures that received the postsynap-
tic BAPTA injections, but not 5-HT stimulation. We found that long-term
(24-h) facilitation was blocked in cocultures following postsynaptic injec-
tion of BAPTA (Cai et al., 2008). Postsynaptic BAPTA treatment by
itself did not appear to have a deleterious effect on the cocultures.
Thus, 5-HT-induced elevation of postsynaptic calcium plays a critical role

in LTF, as it does in ITF. We next examined whether LTF requires post-
synaptic protein synthesis. In a previous report, the cell membrane-imper-
meant protein synthesis inhibitor gelonin was used to show that LTF
involves presynaptic protein synthesis (Martin et al., 1997). To test whether
LTF also involves postsynaptic protein synthesis, we injected gelonin into
motor neurons before treating the cocultures with 5-HT. Postsynaptic
inhibition of protein synthesis impaired LTF. To ensure that these results
were not because of some nonspecific effect of gelonin, we also tested the
effect on LTF of inhibiting postsynaptic protein synthesis with a cell
membrane-impermeant inhibitor of cap-dependent protein synthesis, the
cap analog m7GpppG (Huber, Kayser, & Bear, 2000). Postsynaptic treat-
ment with the cap analog also blocked the induction of LTF (Cai et al.,
2008). The disruptive effect of inhibition of postsynaptic protein synthesis
on LTF was somewhat unexpected: two previous studies had tested the
requirement for postsynaptic protein synthesis in LTF and obtained nega-
tive results (Martin et al., 1997; Trudeau & Castellucci, 1995). The source of
the discrepancy in results is unclear, although the size of the postsynaptic
motor neuron used in the various studies may have played a role. The
postsynaptic neuron in the earlier studies was the large motor neuron L7,
whereas we used the small siphon (LFS-type) motor neuron in our experi-
ments. Possibly, the quantity of the inhibitor introduced into the L7 cell
was insufficient to disrupt protein synthesis; this is particularly likely if, as
our results for ITF imply (Villareal, Li, Cai, & Glanzman, 2007), the post-
synaptic proteins necessary for LTF are synthesized in the synaptic
regions, rather than in the cell body of the motor neuron. (See Sherff &
Carew, 2004, for additional data regarding the requirement for postsynaptic
protein synthesis and LTF.)
Recent work by Schacher and colleagues has shown that a key step
in LTF is the release and increased expression of the sensory neuron
neuropeptide sensorin. Sensorin is expressed in all central sensory neurons
in Aplysia, and appears to be expressed exclusively in sensory neurons
(Brunet, Shapiro, Foster, Kandel, & Iino, 1991). LTF training causes sen-
sorin to be released from the terminals of sensory neurons; disruption of
sensorin release during 5-HT treatment through application of a sensorin
antibody blocks the induction of LTF (Hu, Glickman, Wu, & Schacher,
2004). Thus, sensorin appears to have an autocrine action that is essential
for LTF. In addition to producing enhanced secretion of sensorin, LTF
training causes a rapid increase in the synthesis of sensorin that precedes
the enhanced secretion. The increased sensorin expression occurs in the
axon and terminals of the sensory neurons, and is apparent immediately at
the end of the 5 5-HT protocol (Hu, Wu, & Schacher, 2006). The increase
in sensorin within the sensory neurites is relatively short-lived; sensorin
expression returns to basal levels 2 h after the end of training. (There is a

second, later wave of increased sensorin expression that is preceded by an
increase in sensorin mRNA in the sensory neuron cell body.) The effects of
the 5 5-HT treatment on sensorin appear to result from the activation of
two signaling pathways within the sensory neuron, phosphoinositide
3-kinase (PI3K) and type II PKA (Hu et al., 2006). In addition, the secretion
of sensorin appears to activate mitogen-activated kinase (MAPK), which
then translocates to the nucleus of the sensory neuron (Hu et al., 2004),
where it is believed to play a critical role in stimulating CREB activity
(Martin et al., 1997).
Having determined that elevated postsynaptic calcium is necessary for
LTF, we wished to know whether postsynaptic calcium played a role in the
changes in sensorin release and expression during LTF. Accordingly, we
used quantitative immunohistochemistry to measure sensorin levels in
sensory neurons of sensorimotor cocultures following LTF training. As in
our earlier experiments, BAPTA was injected into some motor neurons
before 5-HT treatment. Again, some cocultures received the postsynaptic
BAPTA injections but not the 5-HT treatment. Immediately after 5-HT
training (or at the equivalent time point in the case of the untreated
cocultures), the cocultures were processed for sensorin immunohisto-
chemistry. In agreement with Hu et al. (Hu et al., 2006), we observed—in
cocultures that did not get the postsynaptic injections of BAPTA—that the
levels of sensorin immunostaining were significantly higher in 5-HT-treated
than in untreated sensory neurons. However, buffering postsynaptic cal-
cium with BAPTA blocked the increase in sensorin immunostaining pro-
duced by 5-HT. Levels of immunostaining in cocultures that received the
postsynaptic BAPTA injection without 5-HT were the same as those in
untreated control cocultures (Cai et al., 2008). These results point to the
conclusion that LTF is regulated postsynaptically. They imply that the
5 5-HT protocol triggers the activation of one or more retrograde signals
through an increase in postsynaptic intracellular calcium and/or postsy-
naptic protein synthesis. The retrograde signals then activate presynaptic
pathways, at least some of which produce the changes in sensorin secre-
tion and expression that are necessary for LTF. We do not yet know the
identity of the retrograde signals. However, it is likely that the presynaptic
targets of the signals include PI3K and/or PKA.
Our LTF results are reminiscent of those from recent studies of learning-
related synaptic plasticity in mammals. For example, it has been reported
that LTP of the mossy fiber!CA3 neuron synapse is triggered by elevated
postsynaptic calcium (Yeckel, Kapur, & Johnston, 1999) (but see Mellor &
Nicoll, 2001). The elevation of postsynaptic calcium stimulates the binding
of EphB receptors in the CA3 neurons with presynaptic ephrin-B ligands.
The EphB receptor–ephrin B ligand interaction activates presynaptic PKA,
leading to enhanced glutamate release from the terminals of the mossy

fibers, the mechanism of LTP expression at this synapse (Zalutsky & Nicoll,
1990). Retrograde signals have been prominently implicated in other forms
of learning-related synaptic plasticity as well (Tao & Poo, 2001). In fact,
transynaptic signaling may eventually come to be regarded as the rule,
rather than the exception, in long-term synaptic plasticity.
5. Conclusion
This chapter has summarized recent insights into a simple form of

learning in a relatively simple organism, the marine snail Aplysia. As I
have tried to show, the cellular and molecular substrates of sensitization in
Aplysia are surprisingly rich and complex. Indeed, there is still much to be
learned about learning in this organism. Importantly, the information
gained by the study of learning and memory in Aplysia is likely to continue
to provide us with valuable insights into how our brains form and hold
memories. The author of To the Lighthouse would probably have been
amazed to learn that the ‘‘comparatively permanent’’ means by which we
retain the poignant scenes of our existence can be glimpsed by inspecting
the nervous system of a mere snail. But Darwin, no doubt would have been
pleased by this outcome.
References
Antonov, I., Antonova, I., Kandel, E. R., & Hawkins, R. D. (2003). Activity-dependent
presynaptic facilitation and Hebbian LTP are both required and interact during
classical conditioning in Aplysia. Neuron, 37, 135–147.
Antonov, I., Kandel, E. R., & Hawkins, R. D. (1999). The contribution of facilitation
of monosynaptic PSPs to dishabituation and sensitization of the Aplysia siphon
withdrawal reflex. J. Neurosci., 19, 10438–10450.
Armitage, B. A., & Siegelbaum, S. A. (1998). Presynaptic induction and expression of
homosynaptic depression at Aplysia sensorimotor neuron synapses. J. Neurosci.,
18, 8770–8779.
Bailey, C. H., & Chen, M. (1983). Morphological basis of long–term habituation and
sensitization in Aplysia. Science, 220, 91–93.
Bailey, C. H., & Chen, M. (1988a). Long-term memory in Aplysia modulates the total
number of varicosities of single identified sensory neurons. Proc. Natl. Acad. Sci.
U.S.A., 85, 2373–2377.
Bailey, C. H., & Chen, M. (1988b). Long-term sensitization in Aplysia increases the
number of presynaptic contacts onto the identified gill motor neuron L7. Proc.
Natl. Acad. Sci. U.S.A., 85, 9356–9359.
Bailey, C. H., & Chen, M. (1989). Time course of structural changes at identified
sensory neuron synapses during long-term sensitization in Aplysia. J. Neurosci.,
9, 1774–1780.
Bao, J. X., Kandel, E. R., & Hawkins, R. D. (1998). Involvement of presynaptic and
postsynaptic mechanisms in a cellular analog of classical conditioning at Aplysia
sensory-motor neuron synapses in isolated cell culture. J. Neurosci., 18, 458–466.
Bartsch, D., Ghirardi, M., Casadio, A., Giustetto, M., Karl, K. A., Zhu, H. et al. (2000).
Enhancement of memory-related long-term facilitation by ApAF, a novel tran-
scription factor that acts downstream from both CREB1 and CREB2. Cell, 103,
595–608.
Bartsch, D., Ghirardi, M., Skehel, P. A., Karl, K. A., Herder, S. P., Chen, M.
et al. (1995). Aplysia CREB2 represses long-term facilitation: relief of repression
converts transient facilitation into long-term functional and structural change.
Cell, 83, 979–992.
Bittner, G. D. (1991). Long-term survival of anucleate neurons and its implications
for nerve regeneration. Trends Neurosci., 14, 188–193.
Bougie, J., Lim, T., Ferraro, G., Manjunath, V., Scott, D., & Sossin, W. S. (2006).
Cloning and characterization of protein kinase C (PKC) Apl III, a homologue of
atypical PKCs in Aplysia. Soc. Neurosci. Abstr., 669.10.
Bougie, J. K., Lim, T., Manjunath, V., Farah-Abi, C., Nagakura, I., & Sossin, W. S. (2007).
The role of atypical protein kinase C (PKC) zeta in synaptic plasticity in Aplysia.
Soc. Neurosci. Abstr., 208.5.
Brembs, B., Lorenzetti, F. D., Reyes, F. D., Baxter, D. A., & Byrne, J. H. (2002).
Operant reward learning in aplysia: neuronal correlates and mechanisms.
Science, 296, 1706–1709.
Brunet, J. F., Shapiro, E., Foster, S. A., Kandel, E. R., & Iino, Y. (1991). Identification
of a peptide specific for Aplysia sensory neurons by PCR-based differential
screening. Science, 252, 856–859.
Byrne, J., Castellucci, V., & Kandel, E. R. (1974). Receptive fields and response
properties of mechanoreceptor neurons innervating siphon skin and mantle shelf
in Aplysia. J. Neurophysiol., 37, 1041–1064.
Byrne, J. H., Castellucci, V. F., & Kandel, E. R. (1978). Contribution of individual
mechanoreceptor sensory neurons to defensive gill-withdrawal reflex in Aplysia.
J. Neurophysiol., 41, 418–431.
Byrne, J. H., & Kandel, E. R. (1996). Presynaptic facilitation revisited: state and time
dependence. J. Neurosci., 16, 425–435.
Cai, D., Chen, S., & Glanzman, D. L . (2008). Postsynaptic resulation of long-term
facilitation in Aplysia. Curr. Biol .
Carew, T. J., Castellucci, V. F., & Kandel, E. R. (1971). An analysis of dishabituation
and sensitization of the gill-withdrawal reflex in Aplysia. Int. J. Neurosci., 2,
79–98.
Carew, T. J., Walters, E. T., & Kandel, E. R. (1981). Classical conditioning in a simple
withdrawal reflex in Aplysia californica. J. Neurosci., 1, 1426–1437.
Castellucci, V. F., Blumenfeld, H., Goelet, P., & Kandel, E. R. (1989). Inhibitor of
protein synthesis blocks long-term behavioral sensitization in the isolated gill-
withdrawal reflex of Aplysia. J. Neurobiol., 20, 1–9.
Castellucci, V. F., & Kandel, E. R. (1976). Presynaptic facilitation as a mechanism

for behavioral sensitization in Aplysia. Science, 194, 1176–1178.
Castellucci, V., Pinsker, H., Kupfermann, I., & Kandel, E. R. (1970). Neuronal
mechanisms of habituation and dishabituation of the gill-withdrawal reflex in
Aplysia. Science, 167, 1745–1748.
Chitwood, R. A., Li, Q., & Glanzman, D. L. (2001). Serotonin facilitates AMPA-type
responses in isolated siphon motor neurons of Aplysia in culture. J. Physiol.,
534, 501–510.
Cleary, L. J., Byrne, J. H., & Frost, W. N. (1995). Role of interneurons in defensive
withdrawal reflexes in Aplysia. Learn. Mem. 2, 133–151.
Cohen-Armon, M., Visochek, L., Katzoff, A., Levitan, D., Susswein, A. J., Klein, R.
et al. (2004). Long-term memory requires polyADP-ribosylation. Science, 304,
1820–1822.
Cook, D. G., & Carew, T. J. (1986). Operant conditioning of head waving in Aplysia.
Proc. Natl. Acad. Sci. USA, 83, 1120–1124.
Dale, N., & Kandel, E. R. (1993). L-glutamate May be the fast excitatory transmitter
of Aplysia sensory neurons. Proc. Natl. Acad. Sci. U.S.A., 90, 7163–7167.
Darwin, C. R. (1881). The formation of vegetable mould, through the actions of
worms, with observations on their habits (London: John Murray).
Dash, P. K., Hochner, B., & Kandel, E. R. (1990). Injection of the cAMP-responsive
element into the nucleus of Aplysia sensory neurons blocks long-term facilita-
tion. Nature, 345, 718–721.
Dubuc, B., & Castellucci, V. F. (1991). Receptive fields and properties of a new cluster
of mechanoreceptor neurons innervating the mantle region and the branchial
cavity of the marine mollusk Aplysia californica. J. Exp. Biol., 156, 315–334.
Engert, F., & Bonhoeffer, T. (1999). Dendritic spine changes associated with hippo-
campal long-term synaptic plasticity. Nature, 399, 66–70.
Frazier, W. T., Kandel, E. R., Kupfermann, I., Waziri, R., & Coggeshall, (1967).
Morphological and functional properties of identified neurons in the abdominal
ganglion of Aplysia californica. J. Neurophysiol., 30, 1288–1351.
Frost, W. N., Castellucci, V. F., Hawkins, R. D., & Kandel, E. R. (1985). Monosynaptic
connections made by the sensory neurons of the gill- and siphon-withdrawal
reflex in Aplysia participate in the storage of long-term memory for sensitization.
Proc. Natl. Acad. Sci. U.S.A., 82, 8266–8269.
Frost, W. N., Clark, G. A., & Kandel, E. R. (1988). Parallel processing of short-term
memory for sensitization in Aplysia. J. Neurobiol., 19, 297–334.
Frost, W. N., & Kandel, E. R. (1995). Structure of the network mediating siphon-
elicited siphon withdrawal in Aplysia. J. Neurophysiol., 73, 2413–2427.
Ghirardi, M., Montarolo, P. G., & Kandel, E. R. (1995). A novel intermediate stage in
the transition between short- and long-term facilitation in the sensory to motor
neuron synapse of Aplysia. Neuron, 14, 413–420.
Glanzman, D. L. (2006). The cellular mechanisms of learning in Aplysia: of blind
men and elephants. Biol. Bull., 210, 271–279.
Glanzman, D. L. (2007). G. North & R. J. Greenspan (Eds.), Simple minds:
the neurobiology of invertebrate learning and memory. In Invertebrate
neurobiology (pp. 347–380) (New York: Cold Spring Harbor Laboratory Press).
Glanzman, D. L., Kandel, E. R., & Schacher, S. (1990). Target-dependent structural

changes accompanying long-term synaptic facilitation in Aplysia neurons.
Science, 249, 799–802.
Glanzman, D. L., Mackey, S. L., Hawkins, R. D., Dyke, A. M., Lloyd, P. E., & Kandel,
E. R. (1989). Depletion of serotonin in the nervous system of Aplysia reduces the
behavioral enhancement of gill withdrawal as well as the heterosynaptic facilita-
tion produced by tail shock. J. Neurosci., 9, 4200–4213.
Guan, Z., Giustetto, M., Lomvardas, S., Kim, J. H., Miniaci, M. C., Schwartz, J. H.
et al. (2002). Integration of long-term-memory-related synaptic plasticity
involves bidirectional regulation of gene expression and chromatin structure.
Cell, 111, 483–493.
Hawkins, R. D., Clark, G. A., & Kandel, E. R. (2006). Operant conditioning of gill
withdrawal in Aplysia. J. Neurosci., 26, 2443–2448.
Hawkins, R. D., Kandel, E. R., & Bailey, C. H. (2006). Molecular mechanisms of
memory storage in Aplysia. Biol. Bull., 210, 174–191.
Hegde, A. N., Inokuchi, K., Pei, W., Casadio, A., Ghirardi, M., Chain, D. G. et al.
(1997). Ubiquitin C-terminal hydrolase is an immediate-early gene essential
for long-term facilitation in Aplysia. Cell, 89, 115–126.
Hernandez, A. I., Blace, N., Crary, J. F., Serrano, P. A., Leitges, M., Libien, J. M.
et al. (2003). Protein kinase M zeta synthesis from a brain mRNA
encoding an independent protein kinase C zeta catalytic domain. Implica-
tions for the molecular mechanism of memory. J. Biol. Chem., 278,
40305–40316.
Hochner, B., Klein, M., Schacher, S., & Kandel, E. R. (1986). Additional component
in the cellular mechanism of presynaptic facilitation contributes to behavioral
dishabituation in Aplysia. Proc. Natl. Acad. Sci. U.S.A., 83, 8794–8798.
Hu, J. Y., Glickman, L., Wu, F., & Schacher, S. (2004). Serotonin regulates the
secretion and autocrine action of a neuropeptide to activate MAPK required for
long-term facilitation in Aplysia. Neuron, 43, 373–385.
Hu, J. Y., Wu, F., & Schacher, S. (2006). Two signaling pathways regulate the
expression and secretion of a neuropeptide required for long-term facilitation in
Aplysia. J. Neurosci., 26, 1026–1035.
Huber, K. M., Kayser, M. S., & Bear, M. F. (2000). Role for rapid dendritic protein
synthesis in hippocampal mGluR-dependent long-term depression. Science, 288,
1254–1257.
Jiang, X. Y., & Abrams, T. W. (1998). Use-dependent decline of paired-pulse facilita-
tion at Aplysia sensory neuron synapses suggests a distinct vesicle pool or
release mechanism. J. Neurosci., 18, 10310–10319.
Kaang, B. K., Kandel, E. R., & Grant, S. G. (1993). Activation of cAMP-responsive
genes by stimuli that produce long-term facilitation in Aplysia sensory neurons.
Neuron, 10, 427–435.
Kandel, E. R. (1976). Cellular basis of behavior: an introduction to behavioral
biology. (San Francisco: W. H. Freeman).
Kandel, E. R. (2001). The molecular biology of memory storage: a dialogue between
genes and synapses. Science, 294, 1030–1038.
Kandel, E. R. (2006). In search of memory. (New York: W. W. Norton & Company).
Klein, M. (1994). Synaptic augmentation by 5-HT at rested Aplysia sensorimotor

synapses: independence of action potential prolongation. Neuron, 13, 159–166.
Klein, M., & Kandel, E. R. (1978). Presynaptic modulation of voltage-dependent Ca2+
current: mechanism for behavioral sensitization in Aplysia californica. Proc.
Natl. Acad. Sci. U.S.A., 75, 3512–3516.
Koester, J., & Kandel, E. R. (1977). Further identification of neurons in the abdom-
inal ganglion of Aplysia using behavioral criteria. Brain Res., 121, 1–20.
Laudanna, C., Mochly-Rosen, D., Liron, T., Constantin, G., & Butcher, E. C. (1998).
Evidence of zeta protein kinase C involvement in polymorphonuclear neutrophil
integrin-dependent adhesion and chemotaxis. J. Biol. Chem., 273, 30306–30315.
Levenson, J., Sherry, D. M., Dryer, L., Chin, J., Byrne, J. H., & Eskin, A. (2000).
Localization of glutamate and glutamate transporters in the sensory neurons of
Aplysia. J. Comp. Neurol., 423, 121–131.
Li, Q., Roberts, A. C., & Glanzman, D. L. (2005). Synaptic facilitation and behavioral
dishabituation in Aplysia: dependence upon release of Ca2+ from postsynaptic
intracellular stores, postsynaptic exocytosis and modulation of postsynaptic
AMPA receptor efficacy. J. Neurosci., 25, 5623–5637.
Lin, X. Y., & Glanzman, D. L. (1994). Long-term potentiation of Aplysia sensorimo-
tor synapses in cell culture: regulation by postsynaptic voltage. Proc. Biol. Sci.,
255, 113–118.
Ling, D. S., Benardo, L. S., Serrano, P. A., Blace, N., Kelly, M. T., Crary, J. F. et al.
(2002). Protein kinase mzeta is necessary and sufficient for LTP maintenance.
Nat. Neurosci., 5, 295–296.
Lukowiak, K., & Sahley, C. (1981). The in vitro classical conditioning of the gill
withdrawal reflex of Aplysia californica. Science, 212, 1516–1518.
Mackey, S. L., Kandel, E. R., & Hawkins, R. D. (1989). Identified serotonergic
neurons LCB1 and RCB1 in the cerebral ganglia of Aplysia produce presynaptic
facilitation of siphon sensory neurons. J. Neurosci., 9, 4227–4235.
Malinow, R., & Malenka, R. C. (2002). AMPA receptor trafficking and synaptic
plasticity. Annu. Rev. Neurosci., 25, 103–126.
Marinesco, S., & Carew, T. J. (2002). Serotonin release evoked by tail nerve stimula-
tion in the CNS of Aplysia: characterization and relationship to heterosynaptic
plasticity. J. Neurosci., 22, 2299–2312.
Marinesco, S., Kolkman, K. E., & Carew, T. J. (2004). Serotonergic modulation in
Aplysia. I. Distributed serotonergic network persistently activated by sensitizing
stimuli. J. Neurophysiol., 92, 2468–2486.
Martin, K. C., Casadio, A., Zhu, H. E. Y., Rose, J. C., Chen, M., Bailey, C. H. et al.
(1997). Synapse-specific, long-term facilitation of Aplysia sensory to motor
synapses: a function for local protein synthesis in memory storage. Cell, 91,
927–938.
Martin, K. C., Michael, D., Rose, J. C., Barad, M., Casadio, A., Zhu, H. et al. (1997).
MAP kinase translocates into the nucleus of the presynaptic cell and is required
for long-term facilitation in Aplysia. Neuron, 18, 899–912.
Martiny-Baron, G., Kazanietz, M. G., Mischak, H., Blumberg, P. M., Kochs, G., Hug, H.
et al. (1993). Selective inhibition of protein kinase C isozymes by the indolocarba-
zole Go 6976. J. Biol. Chem., 268, 9194–9197.
Mellor, J., & Nicoll, R. A. (2001). Hippocampal mossy fiber LTP is independent of
postsynaptic calcium. Nat. Neurosci., 4, 125–126.
Montarolo, P. G., Goelet, P., Castellucci, V. F., Morgan, J., Kandel, E. R., & Schacher,
S. (1986). A critical period for macromolecular synthesis in long-term heterosy-
naptic facilitation in Aplysia. Science, 234, 1249–1254.
Muller, U., & Carew, T. J. (1998). Serotonin induces temporally and mechanistically
distinct phases of persistent PKA activity in Aplysia sensory neurons. Neuron,
21, 1423–1434.
Muslimov, I. A., Nimmrich, V., Hernandez, A. I., Tcherepanov, A., Sacktor,
T. C., & Tiedge, H. (2004). Dendritic transport and localization of protein kinase
mzeta mRNA: implications for molecular memory consolidation. J. Biol. Chem.,
279, 52613–52622.
Park, M., Salgado, J. M., Ostroff, L., Helton, T. D., Robinson, C. G., Harris, K. M. et al.
(2006). Plasticity-induced growth of dendritic spines by exocytic trafficking from
recycling endosomes. Neuron, 52, 817–830.
Passafaro, M., Piech, V., & Sheng, M. (2001). Subunit-specific temporal and spatial
patterns of AMPA receptor exocytosis in hippocampal neurons. Nat. Neurosci., 4,
917–926.
Pastalkova, E., Serrano, P., Pinkhasova, D., Wallace, E., Fenton, A. A., & Sacktor,
T. C. (2006). Storage of spatial information by the maintenance mechanism of
LTP. Science, 313, 1141–1144.
Pfeiffer, B. E., & Huber, K. M. (2006). Current advances in local protein synthesis
and synaptic plasticity. J. Neurosci., 26, 7147–7150.
Pinsker, H. M., Hening, W. A., Carew, T. J., & Kandel, E. R. (1973). Long-term
sensitization of a defensive withdrawal reflex in Aplysia. Science, 182,
1039–1042.
Pinsker, H., Kupfermann, I., Castellucci, V., & Kandel, E. (1970). Habituation
and dishabituation of the gill-withdrawal reflex in Aplysia. Science, 167,
1740–1742.
Rayport, S. G., & Schacher, S. (1986). Synaptic plasticity in vitro: cell culture of
identified Aplysia neurons mediating short-term habituation and sensitization.
J. Neurosci., 6, 759–763.
Roberts, A. C., & Glanzman, D. L. (2003). Learning in Aplysia: looking at synaptic
plasticity from both sides. Trends Neurosci., 26, 662–670.
Royer, S., Coulson, R. L., & Klein, M. (2000). Switching off and on of synaptic sites at
Aplysia sensorimotor synapses. J. Neurosci., 20, 626–638.
Rumpel, S., LeDoux, J., Zador, A., & Malinow, R. (2005). Postsynaptic receptor
trafficking underlying a form of associative learning. Science, 308, 83–88.
Sherff, C. M., & Carew, T. J. (2004). Parallel somatic and synaptic processing in the
induction of intermediate-term and long-term synaptic facilitation in Aplysia.
Proc. Natl. Acad. Sci. U.S.A., 101, 7463–7468.
Si, K., Giustetto, M., Etkin, A., Hsu, R., Janisiewicz, A. M., Miniaci, M. C. et al. (2003).
A neuronal isoform of CPEB regulates local protein synthesis and stabilizes
synapse-specific long-term facilitation in Aplysia. Cell, 115, 893–904.
Si, K., Lindquist, S., & Kandel, E. R. (2003). A neuronal isoform of the Aplysia CPEB
has prion-like properties. Cell, 115, 879–891.
Sossin, W. S. (2007). Isoform specificity of protein kinase Cs in synaptic plasticity.

Learn. Mem., 14, 236–246.
Stirpe, F., Olsnes, S., & Pihl, A. (1980). Gelonin, a new inhibitor of protein synthesis,
nontoxic to intact cells. Isolation, characterization, and preparation of cytotoxic
complexes with concanavalin A. J. Biol. Chem., 255, 6947–6953.
Sutton, M. A., Bagnall, M. W., Sharma, S. K., Shobe, J., & Carew, T. J. (2004).
Intermediate-term memory for site-specific sensitization in Aplysia is maintained
by persistent activation of protein kinase C. J. Neurosci., 24, 3600–3609.
Sutton, M. A., & Carew, T. J. (2000). Parallel molecular pathways mediate expres-
sion of distinct forms of intermediate-term facilitation at tail sensory-motor
synapses in Aplysia. Neuron, 26, 219–231.
Sutton, M. A., & Carew, T. J. (2002). Behavioral, cellular, and molecular analysis of
memory in Aplysia I: intermediate-term memory. Integ. Comp. Biol., 42, 725–735.
Sutton, M. A., Masters, S. E., Bagnall, M. W., & Carew, T. J. (2001). Molecular
mechanisms underlying a unique intermediate phase of memory in Aplysia.
Neuron, 31, 143–154.
Sutton, M. A., & Schuman, E. M. (2006). Dendritic protein synthesis, synaptic
plasticity, and memory. Cell, 127, 49–58.
Tao, H. W., & Poo, M. (2001). Retrograde signaling at central synapses. Proc. Natl.
Acad. Sci. USA, 98, 11009–11015.
Trudeau, L. E., & Castellucci, V. F. (1995). Postsynaptic modifications in long-term
facilitation in Aplysia: upregulation of excitatory amino acid receptors. J. Neu-
rosci., 15, 1275–1284.
Villareal, G., Li, Q., Cai, D., & Glanzman, D. L. (2007). The role of rapid, local
postsynaptic protein synthesis in learning-related synaptic facilitation in Aplysia.
Curr. Biol., 17, 2073–2080.
Wainwright, M. L., Byrne, J. H., & Cleary, L. J. (2004). Dissociation of morphological
and physiological changes associated with long-term memory in Aplysia.
J. Neurophysiol., 92, 2628–2632.
Wan, Q., & Abrams, T. W. (2008). Trans-synaptic plasticity: presynaptic initiation,
postsynaptic memory. Curr. Biol., 18, R220–223.
Yeckel, M. F., Kapur, A., & Johnston, D. (1999). Multiple forms of LTP in hippocam-
pal CA3 neurons use a common postsynaptic mechanism. Nat. Neurosci., 2,
625–633.
Zalutsky, R. A., & Nicoll, R. A. (1990). Comparison of two forms of long-term
potentiation in single hippocampal neurons. Science, 248, 1619–1624.
Zecevic, D., Wu, J.- Y., Cohen, L. B., London, J. A., Hopp, H.- P., & Falk, C. X. (1989).
Hundreds of neurons in the Aplysia abdominal ganglion are active during the gill-
withdrawal reflex. J. Neurosci., 9, 3681–3689.
Zhang, F., Goldsmith, J. R., & Byrne, J. H. (1994). Neural analogue of long-term
sensitization training produces long-term (24 and 48 h) facilitation of the
sensory-to-motor neuron connection in Aplysia. J. Neurophysiol., 72, 778–784.
Zhao, Y., Leal, K., Abi-Farah, C., Martin, K. C., Sossin, W. S., & Klein, M. (2006).
Isoform specificity of PKC translocation in living Aplysia sensory neurons and a
role for Ca2+-dependent PKC APL I in the induction of intermediate-term facilita-
tion. J. Neurosci., 26, 8847–8856.
Human Learning 271
Insect Minds For Human Minds

Randolf Menzel
Freie Universität Berlin, Germany
1. Introduction
Although great progress has been made, nobody will deny that we are
far from understanding even the basic rules governing the machinery of
the human mind. Under these conditions, any approach that promises to
help unravel the mysteries of mind–brain connections is welcome. Mice
brains are cut into thin slices, or neurons are placed in cell culture to
study the molecular and cellular components of neurons and their com-
munication. Brains are recorded with EEG electrodes or imaged with
fMRI methods, although both approaches work only for distant signals
from the neuronal nets. Ideally, one would like to measure the working
of neurons embedded in an animal brain (or, even better: a human brain)
while the brain performs normally, responding to stimuli, recruiting
memory, and learning—in short, behaving as it does in normal life.
There is no such ideal case. The unavoidable compromises, due to the
complexity of the brain machinery and the lack of appropriate methods,
are severe. The limited methods imposed on current neuroscience favor
an additional approach, namely comparative neuroscience. Animals come
with all kinds of brain organizations, and some of them are more suita-
ble for today’s methodology because they provide us with a smaller
number of neurons, larger neurons, more robust behavior under record-
ing conditions, and less complex behavior. Some of these animal species
are even accessible to the most sophisticated armory of molecular–
genetic tools. Since none of these animal species combines all of the
favorable conditions, the best solution is to take advantage of all of
them. Comparative cognitive neuroscience recognizes the evolutionary
links between animal species (including man) and argues that complex
traits like those underlying human cognition have their roots in less-
complex precursors that serve identical or rather similar functions in
animals with smaller brains.
272 Randolf Menzel
2. Comparative architecture of brains: basic features of brains
Thinking about the basic design of a brain that subserves the cognitive
functions of any brain, one recognizes a structure of essential modules and
their interconnectivity (Fig. 1). This architecture of modules appears to be
shared by a large range of animal species and may even apply to the worm-
like creature on the basis of the evolutionary divide between protostomes
and deuterostomes, two largest evolutionary streams of bilateral animals
with a centralized nervous system. Although there are multiples of each of
the modules depicted in Fig. 1 (multiple perceptual systems, multiple
belief-generating systems, multiple desire-generating systems, multiple
action-planning systems, and multiple motor control systems), the basic
idea put forward by Carruthers (2006) in this scheme is that perceptual
systems feed three downstream systems arranged both serially and in
parallel that converge on the action-planning system. Thus perceptual
systems can reach the action-planning systems directly, and, in addition,
the desire- and belief-generating systems receiving the same perceptual
information will act in parallel onto action planning, as well.
When we talk about ‘‘modules’’ and ‘‘systems’’ we mean—in essence—
neurons and neural nets. Therefore, if such a scheme should be of any
heuristic help in understanding the brain, its skeleton needs the flesh of
neurons and their functions. The honeybee brain provides a case for such a
fleshing out. Figure 2 shows a schematic representation of the information
flow from sensory systems (vision, olfaction, and mechanosensory) to a
premotor area of the brain (the lateral horn), and in parallel via the mush-
room body (MB), which through its output neurons also feeds into the
premotor area. The sensory systems are highly developed, allowing the
Belief-
generating
systems
Action Motor
Perceptual
planning control
systems
systems system
Desire-
generating
systems
Body
state
Fig. 1. The cognitive structure of brains composed of modules for perception,

desire, and planning (after Carruthers, 2006, p. 66).
(a) (b)
olfac
calyx vis
mech
mb
MC
Multimodal
vis
output
LC mb
α CB
β
LH
PL
ME LO mech
olfac
AL
SOG
Premotor
descending VUM
fibers
Fig. 2. The honeybee brain. (a) The left hemisphere showing two visual ganglia,
medulla (ME) and lobula (LO), the mushroom body (MB) with the two calyces
(lateral and median calyx, LC, MC) and the two lobes (a and b), the first-order
neurophil of the olfactory pathway, antennal lobe (AL), and half of the
subesophageal ganglion (SOG). (b) Information flow from mechanosensory
(mech), olfactory (olfac), and visual (vis) input to premotor output in the lateral
horn (LH) via descending fibers. Note the parallel pathway via the mushroom body
(MB), a central integrating structure that receives input organized according to the
senses in the calyx and transmits multimodal output via the lobes. Some of the
output neurons feed into the premotor pathway. The desire-generating system is
represented by ventral unpaired median (VUM) neurons of the subesophageal
ganglion that project to the MB calyx, the LH, and the primary neurophil of the
olfactory system, the antennal lobe.
animal to see colors and the polarized light pattern of the sky, recognize
patterns, estimate distance via optic flow, and see depth using motion
parallax. Equally highly developed are the chemosensory and mechanosen-
sory organs, giving the bee the ability to distinguish a large range of odors,
and allow the precise control of body positions and movements both in
walking and flying. The desire-generating system—at least with respect to
one sensory modality (olfaction)—is implemented in the bee brain by a small
set of neurons whose cell bodies are localized in the ventral midline of the
subesophageal ganglion [the ventral unpaired median (VUM) neurons],
and these neurons receive their inputs from desire-related sense organs,
274 Randolf Menzel
e.g., sucrose receptors. Their inputs from body states have not yet been
identified, but since their function depends on the body state (like the levels
of satiation, sleep, arousal, and attention), we must assume that they receive
the respective inputs. Ventral unpaired median neurons feed not only onto
action-planning systems like the premotor output region of the bee brain
(LH), but also onto the belief-generating system (MB) and the perceptual
system (in particular the olfactory antennal lobe). The highlights of the
motor control systems are distributed between various areas in the brain,
including the lateral horn and downstream ganglia of the ventral chord. Their
actions are best characterized by the precise movements during the waggle
dance (see below) and the acrobatic flight in turbulent winds. Thus the
boxicology of general brain functions—as developed for big mammalian
brains (Carruthers, 2006)—applies surprisingly well to the bee brain, and
we may ask whether the boxes and the connections can be filled with
information about the working of the neurons and networks of which they
are composed. We shall focus here on the desire-generating and the belief-
generating systems, the VUM neurons, and the MB.
2.1. DESIRE-GENERATING SYSTEM

The VUM neurons are a major component of the modulatory network in
the bee brain that works with the neurotransmitter octopamine. This
transmitter is involved in arousing the animal, making it hungry and more
sensitive to stimuli. In addition, neural activity in one of the VUM neurons
(that of the maxillary neuromere, VUMmx1) immediately following an odor
stimulation was found to be sufficient to serve as the appetitive-reinforcing
function during olfactory conditioning (Hammer, 1993, 1997). Furthermore,
it was found that this neuron also learns about the stimuli it reinforces.
Taken together, the anatomy and functional properties of VUMmx1 define
it as a neural implementation of the desire-generating system. The proper-
ties of such a system on the level of a single identified neuron become clear
when we more closely examine the significance of the VUMmx1 learning
about the rewarded stimulus (odor). Figure 3a shows that VUMmx1
responds to the forward-paired conditioned stimulus (CSþ) odor with
strong excitation and to the backward-paired CS odor with only very
few spikes. Before training, the neuron responded to both odors equally
strongly with a spike rate significantly lower than the spike rate for CSþ
after learning. This behavior documents the fact that the VUMmx1 neuron
has changed its response properties to both CSþ and CS in the course of
learning. Most importantly, the neuron also develops an ‘‘expectation’’ of
the unconditioned stimulus (US, sucrose reward). This can be seen when
stimulating the animal after learning with CSþ or CS followed by a
sucrose reward (see Fig. 3b). Now the sucrose stimulus—that before
(a)
CS+ US CS–
(b)
CS+ US
20 mV
CS– US
3 sec
Fig. 3. The error-predicting properties of the desire-generating system as

implemented in the VUMmx1 neuron. (a) In the course of learning, VUMmx1
learns to respond to the forward-paired odor CSþ and to reduce its responses to
the backward-paired odor CS. (b) If the animal is stimulated with the rewarding
stimulus (sucrose) following either CSþ or CS, VUMmx1 does not respond to the
predicted US after CSþ but responds vigorously to the unpredicted US after CS.
learning had induced a vigorous response—triggers no spikes in response

to the US after the CSþ, but does cause spikes in response to the US after
the CS stimulation. Thus an expected US (after CSþ) is not responded to,
but an unexpected (after CS) US elicits a strong response. Similar proper-
ties were found in dopamine neurons of the primate ventral tegmentum,
neurons that are known to provide the cortex with appetitive and arousing
signals (Schultz, 2006).
The principle behind this property centers on computing error signals:
whenever the current expectation is violated, error signals trigger pro-
cesses in the brain, which serve to both keep the animal’s expectation
up-to-date and minimize such errors. The most potent error signals are
generated by unexpected situations or events with biological relevance (as
in the case mentioned above, an US). Upon first perceiving such an unex-
pected US, coincidence detectors evaluate any neural activity preceding or
overlapping with the error signal generated by the US. Preceding events
can serve as predictors of the US if their temporal connection with the US
is consistent enough. Depending on the state of the animal and the nature
of the environment, several more encounters with the US and its preceding
events are required for a reliable memory to form. A reliable memory
means that fully expected USs no longer generate any error signals, and
hence the animal does not need to update its expectations any further. In
such a state, the animal can behave adaptively: it expects the consequences
of any behavior or situation. All these properties are implemented in the
response properties of a single identified neuron, the VUMmx1 in the bee
brain (Menzel, Brembs, & Giurfa, 2007).
276 Randolf Menzel
2.2. BELIEF-GENERATING SYSTEM
A major component of the belief-generating systems is the MBs, known

to be essential structures for learning and memory formation. As pointed
out above, the MB is wired in the bee brain very much as shown in Fig. 1
(compare with Fig. 2). Belief-generating systems lie in a parallel pathway
between the perceptual systems and the action-planning systems. This also
applies to the MB, indicating that it adds a value to neural processing that
does not exist in the direct sensory-action connections, and is required for
those neural processes that are based on cross talk between perceptual
systems, on evaluations in reference to experience, and on selection
between behavioral options. Figure 4 gives an impression of the neural
connections of the MB and its intrinsic structure. In many respects, the
design of its neural architecture resembles that of cortical structures in the
mammalian brain, e.g., the hippocampus (Rolls, 1990). All perceptual sys-
tems feed into the MB at its calyx region, where the relatively small number
of input neurons (a few thousand) diverge into a large number (about
160,000) of intrinsic neurons, called Kenyon cells (KC). Each MB has two
Kenyon cells
VUM
Inhibit.
Neuron
Input
neurons
Modulat.
Neurons
Output
neurons
Fig. 4. Intrinsic organization of the mushroom body (see text). VUM: ventral
unpaired median neurons of the subesophageal ganglion representing the desire-
generating system (see text); inhibit. neurons: inhibitory feedback neurons
belonging to the protocerebral–calycal tract, PCT; modulat. neurons: a range of
modulatory neurons of yet unknown functions.
output regions (alpha and beta lobes; only one is shown in Fig. 4) where a
rather small number of extrinsic neurons (a few thousand) leave the MB,
indicating a strong convergence from KC onto extrinsic neurons. In addi-
tion, a considerable number of widely branching neurons invade the MB on
several levels, and these are considered modulatory neurons. These basic
features of the MB architecture can be formalized by an associative neural
net with divergent connectivity at the input site and convergent connectiv-
ity at the output site (Fig. 4). Each KC appears to receive input from several
second- or higher-order sensory neurons and inhibitory neurons, and since
a KC responds only with a short response consisting of a small number of
action potentials (Szyszka, Ditzen, Galkin, Galizia, & Menzel, 2005), it has
been concluded that coincident activation via sensory neurons is required
to overcome inhibition and fire KC. The sparse-coding properties of KC
both at the population level and in the time domain appear to be an ideal
substrate for storing highly selective and cross-modality-specific mem-
ories, properties that KC share with the principal cells in the hippocampus.
However, these conclusions—based on anatomical findings and few
physiological measurements—are still hardly supported by experimental
evidence.
Output neurons of the MB receive input from a large number of KC and
are thus activated by several or many sensory stimuli. This finding is
somewhat surprising because the high selectivity and specificity of coding
and storing (as assumed for the KC) is lost by such properties, and it is not
known yet how the expanded perceptual space and the specific memories
stored in the MB are read out by these neurons. In any case, these neurons
show learning-related plasticity as expected, indicating that they read out
or are even involved in the development of neural substrates of memory.
A telling example is the identified neuron PE1 (Okada, Rybak, Manz, &
Menzel, 2007). This neuron exists only once in each MB and exits it in the
ventral part of the alpha lobe (Fig. 5). PE1 reduces its response to CSþ and
does not change its response to CS after learning. PE1 receives inhibitory
synaptic inputs, and neuroanatomical studies indicate closely attached
GABA-immune-reactive profiles originating at least partially from neurons
of the protocerebral–calycal tract (PCT, see Fig. 4). Since PE1 develops
long-term synaptic plasticity (Menzel & Manz, 2005), it was concluded that
PE1 develops the associative reduction of odor response by intrinsic
mechanism (associative long term potentiation (LTP). However, other
possibilities cannot be ruled out yet and may even exist alongside the
intrinsic plasticity, e.g., associative enhancement of inhibition via PCT
neurons or less input from KC (Okada et al., 2007). PE1 projects to the
premotor output region of the brain (LP, see Fig. 2) where it appears
to terminate on GABA immune neurons that may be presynaptic to
descending neurons. If this connectivity can be verified, it might explain
278 Randolf Menzel
(a) (b)
mPL
LPL
β–L
(c) (d)
% CR
Hz 80
70 CS
60
50
30 40 Ct
10 20
CS
Ctr CS+ CS– 1 2 3 4 5 1 2 3

Rel. response Rel. response Training Test
before training after training
Fig. 5. Learning-related changes in the response properties of the mushroom body

(MB)-extrinsic neuron PE1. (a) Extracellular recording of PE1. The characteristic
double and triple spikes allow reliable assignment of extracellular activity to this
unique and identified neuron. (b) Intracellular marking of PE1. The neuron receives
input across the mushroom body (MB) and projects to the lateral horn (LPL). mPL:
median protocerebral lobe; LPL: lateral protocerebral lobe; b-L: beta lobe of the MB.
(c) Response change of PE1 in the course of odor learning. The ordinate gives the
response to the control odor (Ctr), the CSþ or the CS in spikes per second before
and after training. (d) Acquisition and retention of the animals from which PE1 was
recorded. Percentage of CR (ordinate) gives the response to the respective odors
(Ctr: control odor, CSþ: forward-conditioned odor, CS: backward-conditioned
odor) during the five training trials and during the three retention test trials. The
first test was performed 5 min after the last conditioning trial, the second after
30 min, and the third after 60 min. (after Okada, Rybak, Manz, & Menzel, 2007).
a general property of the parallel pathway via the MB. The MB is known to
provide a general inhibitory effect on behavioral output, for example,
rivalry song in crickets (Huber, 1978), or adaptation of walking behavior
to new stimulus conditions in Drosophila (Heisenberg, 1998). If learning
leads to less inhibition—as was seen in the case of PE1—then the learning-
related signal from the MB would relax this inhibition and the descending
neurons may receive stronger input from the direct sensory-motor
connections.
In such a scenario, the belief system (MB) would not need to code and
store the specifics of the stimuli but rather tag them with the signature
‘‘learned in an appetitive context’’ and provide this signature to the conver-
gence site with the direct sensory-motor connections. This mechanism—if it
exists as it is currently interpreted—is certainly only one of many compo-
nents of sensory coding and memory storage in the belief system. Evidence
for the latter interpretation comes from the finding that other MB extrinsic
neurons show a large range of learning-related plasticity.
3. Comparative cognition: what a little brain can do?
3.1. NONELEMENTAL FORMS OF LEARNING
Elemental forms of associative learning have been described for inverte-

brates, and such associative links between stimuli and responses do not
require neural setups like the ones shown in Fig. 1 with desire and belief
systems. Nonelemental learning, however, which establishes unique links
between specific events and is characterized by the ambiguity of events
under consideration, cannot be explained by simple linear associative con-
nections between the perceptual and motor systems, and these are the usual
forms of learning that animals and humans confront (Rudy & Sutherland,
1995). For example, in a problem called negative-patterning discrimination,
an animal has to learn to discriminate a binary compound AB from its
elements A and B in a protocol in which the compound is not reinforced
while the elements are (Aþ, Bþ vs. AB). This situation is particularly
challenging since each element A and B appears as often reinforced as it
is nonreinforced. Relying on elemental links between A (or B) and reinforce-
ment (or absence of reinforcement) is useless in solving this problem.
Honeybees have no problem with negative-patterning discrimination
(Lachnit, Giurfa, & Menzel, 2004). Other nonelemental forms of learning
are those based on the extraction of rules. A typical example for rule-
based learning is the acquisition of the sameness or difference principle.
These rules are demonstrated through the protocols of delayed matching to
sample (DMTS) and delayed nonmatching to sample (DNMTS), respectively.
In DMTS, animals are presented with a changing set of samples and then
with a set of stimuli, one of which is identical to the sample and is rein-
forced. Honeybees foraging in a Y maze learn both the DMTS and DNMTS
rules. In addition, bees trained with colors and presented in transfer tests
with black-and-white gratings that they have not experienced before solved
the problem and chose the grating identical to the sample at the entrance to
the maze. Transfer was not limited to different kinds of modalities (pattern
vs. color) within the visual domain, but could also operate between
280 Randolf Menzel
drastically different domains such as olfaction and vision (Giurfa, Zhang,

Jenett, Menzel, & Srinivasan, 2001). These results document that bees learn
rules relating stimuli in their environment. The working memory for the
sample underlying the solving of DMTS is around 5 s ( Zhang, Bock, Si,
Tautz, & Srinivasan, 2005) and thus coincides with the duration of other
visual and olfactory short-term memories (STMs) characterized in simpler
forms of associative learning in honeybees (Menzel, 1999).
A whole range of nonelemental forms of learning has been tested in
honeybees; most of the problems posed to the animal could be solved
(Menzel et al., 2007). It is, therefore, more appropriate to consider the
format of memory as it is established by nonelemental learning as repre-
sentation of stimulus–behavior connections that include the context, the
probabilities, the rules, and the internal conditions of the animal itself.
Such memory formats develop over the course of learning events and the
course of time, and this process is called consolidation. I next ask whether
consolidation has been documented in honeybees. Then I shall ask whether
there is evidence for representation as a format of memory.
3.2. MEMORY PROCESSING
Memories exist in multiple forms and functions. They are categorized

according to their physiological substrates along a time scale as short-term,
mid-term, and long-term memory (STM, MTM, and LTM, respectively),
referring to ongoing neural activity as the storage device of STM, inter-
cellular signaling cascades leading to MTM, and gene activation, protein
synthesis, and new structures underlying LTM. The transitions between
these memory stages can be sequential and/or parallel, processes referred
to as physiological correlates of consolidation, a phenomenon originally
known from human psychological studies capturing the fact that over time
early, vulnerable forms of memory are converted into more stable and long-
lasting forms (Müller & Pilzecker, 1900). The similarity of cellular pro-
cesses underlying this temporal structure of memory in invertebrates and
vertebrates argues strongly in favor of homologous processes on the level
of the metabolism of neurons. We may ask whether consolidation at the
network and systems level in an insect like the honeybee resembles fea-
tures known from vertebrate studies and thus may help us trace these
features to functions of identified neurons.
Reward learning in honeybees initiates a sequence of memory phases
that lead to long-lasting memory, passing through at least four forms of
memory (Menzel, 1999; Menzel & Müller, 1996). An associative-learning
trial induces an early form of short-term memory (eSTM) in the seconds
range. This memory is highly dominated by appetitive arousal and sensiti-
zation, is rather unspecific, and is quickly converted into a late STM.
The transition to MTM is a rather slow process and makes the memory
trace unsusceptible to retrograde amnesic treatments. Mid-term memory is
stimulus-specific, indicating that consolidation from STM to MTM includes
processes that incorporate the conditions under which learning has
occurred. Single and multiple learning trials lead to different long-term
forms of memory. Two forms of LTM must be distinguished in honeybees:
early LTM (1–2 days) characterized by translation-dependent retention and
constitutively active PKC, and late LTM (‡3 days) characterized by tran-
scription-dependent retention. The two forms of LTM arise differently,
after massed and spaced multiple learning trials, indicating that the con-
solidation processes between the learning trials structure LTM differently.
Although evidence is sparse on characterizing the contents of the various
memory phases, it is likely that the belief system shaping memory by
consolidation processes does indeed contain intrinsic neural mechanisms
that work on new memory contents in reference to existing memories.
A first hint of such neural processes has been found for MB-extrinsic
neurons that change their properties during consolidation to MTM (Martin
Strube, personal communication).
From an evolutionary point of view, one may expect that memory dynamics
are adapted to choice behavior under natural conditions. Foraging in pollinating
insects is a behavior with a highly regular sequential structure of events, ranging
from actions within seconds to those separated by months and thus may offer
the opportunity to relate memory structure and ecological demands (Menzel,
1999). Different memories are consulted during the sequence of events during
foraging. In the bee, the time courses of successive behaviors during foraging
match the temporal dynamics of memory stages. Choices between flowers
within the same patch quickly succeed each other and are performed during
early STM. Choices between flowers in different patches occur after the transi-
tion to late STM. Successive bouts are interrupted by the return to the hive such
that flower choices in a subsequent bout require retrieving information from
MTM. The separation between the two forms of LTM may be related to the
periods when flower patches are in bloom (Menzel, 2001).
3.3. REPRESENTATIONS: DECIDING BETWEEN OPTIONS
The format of memory is better conceptualized as a representation of all

the conditions that have lead to its establishment, including the consolida-
tion processes. Representations are thought of as neural implementations
of experiences that allow activation into a working memory, processing
within working memory, and the evaluation of predictions of expected
outcomes before and independent of the actual behavior and stimulus
conditions connected to these outcomes. Is there any evidence in honey-
bees that such processing on the level of working memory actually occurs?
282 Randolf Menzel
Honeybees navigate according to a geometrically organized spatial mem-

ory that guides them to goals like the hive and the food source (Menzel et al.,
2005). When bees are released at an unexpected location, several hundred
meters away from the hive and the food source, they can make a decision to
fly either directly back to the hive or first to the food source and then to the
hive (Fig. 6). The decision process underlying this behavior can even be seen
in the sections of the flight path flown during the search period but before
the direct return flights are initiated. Bees seem to ‘‘probe’’ the two potential
return sites by sequentially flying for short periods in one direction or the
other. This probing is performed at locations where no stimuli emanating
(a) (b)
mPL
LPL
β–L
(c) (d)
% CR
Hz 80
70 CS
60
50
30 40 Ct
10 20
CS
Ctr CS+ CS– 1 2 3 4 5 1 2 3

Rel. response Rel. response Training Test
before training after training
Fig. 6. Return flights of bees that had been trained to fly between the hive H and the
feeder F. Each of the 10 bees was caught at the feeder after they had fed to
completion and was transported in the dark to a release site several hundred
meters away from the hive and the feeder either in the north or the south of the
hive. Before they were released, a transponder was attached to their thorax, which
allowed to monitor their flight path using a harmonic radar device. Before bees
started their fast return flights, they first performed a straight flight over 200 m to
the west, which would have brought them back to the hive if they were not
transported to an unexpected release site. They then searched for a while before
they started their fast and rather straight return flights. Only this fast and rather
straight return flights of 10 bees that flew first toward the feeder and then to the
hive are shown here. Only one bee landed at the feeder (after Menzel et al., 2007).
from the two sites (hive and feeder) are available, and both probing and the
final decision are solely based on the location of the two sites in spatial
memory (De Marco, personal communication).
Again, experimental data are sparse on the question of how decisions
are made on the working memory level, but there is no doubt that a brain as
small as the honeybee’s does indeed provide the basic functions of a belief
system. These functions are the following: rich and nonelemental forms of
learning, memory processing that includes incorporation of new memory
into existing memory via consolidation processes, retrieval of remote
memory into working memory, and evaluation of potential outcomes of
behaviors on the basis of the information available in working memory
(Menzel et al., 2006).
4. Conclusion
It has often been said that neuroscience lacks a theory (or theories) of
the brain. Indeed, there appears to be no concept at the level of the
neurons, the networks, or the whole brain that provides enough generality
for developing such a theory. Potentially the ‘‘boxicology’’ of Carruthers
(2006) (see Fig. 1), together with the hard facts obtained from anatomy,
physiology, and behavioral analysis, may provide a path toward developing
a brain theory. As Carruthers (2006) says the following in his book on the
‘‘Architecture of the Mind’’ (p. 68):
To be a believer/desirer . . . means possessing distinct content-bearing

belief-states and desire-states that are discrete, structured, and causally
efficacious in virtue of their structural properties. These are demanding
conditions. But not so demanding that the non-human animals can be
ruled out as candidates immediately. Indeed I propose to argue, on the
contrary, that many invertebrates actually satisfy these requirements.
The honeybee offers a look into the workings of its brain not only using
anatomical and physiological approaches but also via observations of its
communicative process. Here the animals tell us how they process the
information they receive on the basis of their own knowledge independent
of the stimuli and the responses associated with the indicated conditions.
The decisions made based on such evaluation processes reflect ‘‘beliefs’’ in
the sense that expected outcomes must be created as network conditions
in working memory. A model for such neuronal states could be the cor-
ollary discharge phenomenon, the fact that expected outcomes of a motor
behavior are fed to neural instances where they converge with the sensory
input resulting from the motor behavior and where deviations are
284 Randolf Menzel
computed. Single neurons in the insect brain have been reported to provide
such corollary discharge information (Poulet & Hedwig, 2006), and the
VUMmx1’s property of changing its response to the reward whether it is
expected or not (see above) is a telling example for such network proper-
ties. Therefore, in my view, there is hope for both filling the gap between
formal models and boxes as they are provided by science philosophers and
developing step-by-step a theory of the brain based on the hard facts of
neuroscience.
References
Carruthers, P. (2006). The architecture of the mind. Oxford: Clarendon Press.

Giurfa, M., Zhang, S. W., Jenett, A., Menzel, R., & Srinivasan, M. V. (2001). The
concepts of ‘sameness’ and ‘difference’ in an insect. Nature, 410, 930.
Hammer, M. (1993). An identified neuron mediates the unconditioned stimulus in
associative olfactory learning in honeybees. Nature, 366, 59.
Hammer, M. (1997). The neural basis of associative reward learning in honeybees.
Trends in Neuroscience, 20, 245.
Heisenberg, M. (1998). What do the mushroom bodies do for the insect brain? An
introduction. Learning and Memory, 5, 1.
Huber, F. (1978). The insect nervous system and insect behaviour. Animal Beha-
viour, 26, 969.
Lachnit, H., Giurfa, M., & Menzel, R. (2004). Odor processing in honeybees: is the
whole equal to, more than, or different from the sum of its parts. In P. J. G. Slater,
J. S. Rosenblatt, T. J. Roper, C. T. Snowdon, H. J. Brockmann & M. Naguib (Eds.),
Advances in the Study of Behavior, Vol. 34 (pp. 241–264). Amsterdam: Elsevier.
Menzel, R. (1999). Memory dynamics in the honeybee. Journal Of comparative
Physiology [A], 185, 323.
Menzel, R. (2001). Behavioral and neural mechanisms of learning and memory as
determinants of flower constancy. In L. Chittka & J. D. Thomson (Eds.), Cogni-
tive ecology of pollination. Animal behavior and floral evolution (pp. 21–40).
Cambridge.
Menzel, R., Brembs, B., & Giurfa, M. (2007). Cognition in Invertebrates. In J. H. Kaas
(Ed.), Evolution of nervous systems, Vol. II: evolution of nervous systems in
invertebrates (pp. 403–422). Oxford: Academic Press.
Menzel, R., Greggers, U., Smith, A., Berger, S., Brandt, R., Brunke, S., et al. (2005).
Honeybees navigate According To a map-like spatial memory. Proceedings of the
National Academy of Science, USA 102, 3040.
Menzel, R., Leboulle, G., & Eisenhardt, D. (2006). Small brains, bright minds. Cell,
124, 237.
Menzel, R., & Manz, G. (2005). Neural plasticity of mushroom body-extrinsic neu-
rons in the honeybee brain. Journal of Experimental Biology, 208, 4317.
Menzel, R., & Müller, U. (1996). Learning and memory in honeybees: from behavior
to neural substrates. Annual Review of Neuroscience, 19, 379.
Müller, G. E., & Pilzecker, A. (1900). Experimentelle Beiträge Zur Lehre Vom
Gedächtnis. Z Psychol, 1, 1.
Okada, R., Rybak, J., Manz, G., & Menzel, R. (2007). Learning-related plasticity in
PE1 and other mushroom body-extrinsic neurons in the honeybee brain. Journal
of Neuroscience, 27, 11736.
Poulet, J. F., & Hedwig, B. (2006). The cellular basis of a corollary discharge.
Science, 311, 518.
Rolls, E. T. (1990). Theoretical and neurophysiological analysis of the functions of
the primate hippocampus in memory. Cold spring harbor symposium. Quantum
Biology, LV, 995.
Rudy, J. W., & Sutherland, R. J. (1995). Configural association theory and the
hippocampal formation: an appraisal and reconfiguration. Hippocampus, 5, 375.
Schultz, W. (2006). Behavioral theories and the neurophysiology of reward. Annual
Review of Psychology, 57, 87.
Szyszka, P., Ditzen, M., Galkin, A., Galizia, C. G., & Menzel, R. (2005). Sparsening
and temporal sharpening of olfactory representations in the honeybee mushroom
bodies. Journal of Neurophysiology, 94, 3303.
Zhang, S. W., Bock, F., Si, A., Tautz, J., & Srinivasan, M. V. (2005). Visual working
memory in decision making by honeybees. Proceedings of the National Academy
of Science, USA 102, 5250.
Human Learning 287
Patterns of Learning, Memory,

and Vocal Production in the
Songbird Brain
David S. Vicario
Department of Psychology, Rutgers University
1. Introduction
The complex system of verbal signals used for communication is a

defining adaptation of the human species, and it depends on each infant
learning to perceive and produce the rapid speech signals used in its
particular language environment. Of the very few animal taxa that learn
the vocal signals they use for social communication as humans do, song-
birds are the most easily studied at the neurobiological level (Doupe &
Kuhl, 1999). In both humans and songbirds, infants learn vocalizations by
imitating the sounds of conspecific tutors heard during an early sensitive
period. An acoustic pattern is stored in the brain and used as a reference
that guides vocal motor development to eventually mimic the same acous-
tic pattern. This process combines many forms of perceptual and motor
learning whose brain mechanisms we seek to understand. The songbird
brain contains a specialized auditory area where neurons respond more
strongly to conspecific than to heterospecific vocal sounds and form long-
lasting memories for the unique vocalizations of individual conspecifics
(Chew et al., 1995; Mello, Vicario, & Clayton, 1992). The way song patterns
are processed and remembered in this area is under active investigation.
The songbird brain also contains specialized vocal production areas whose
size and function are correlated with vocal learning (Nottebohm, Kasparian,
& Pandazis, 1981; Nottebohm, Stokes, & Leonard, 1976), and that show
intriguing properties, e.g., sexual dimorphism and lateralization. Many labora-
tories are studying the contribution of activity in these areas to the detailed
progression of sensorimotor events during development that ultimately pro-
duce the imitated pattern. Thus, the birdsong system enables us to study
basic mechanisms of perceptual memory and motor learning in a behavioral
288 David S. Vicario
context that may shed light on the process of speech acquisition. Learning
at this level of complexity is a system phenomenon that may involve changes
in the mapping relationships between central representations of sensory
and motor variables. At this point, we cannot directly describe it in terms
of specific processes of synaptic modification, although these processes
undoubtedly are involved.
2. The vocal learning process
The songbird vocal system has emerged as the best-developed model for
the neural basis of speech acquisition because songbirds are among the
very few animals that learn their vocalizations through a process of vocal
imitation, as humans do (Doupe & Kuhl, 1999). Among the known vocal
learners (humans, cetaceans, hummingbirds, parrots, and songbirds), the
songbird is the most tractable for experimental study with invasive tech-
niques. It should be noted that auditory discrimination learning is wide-
spread in the animal kingdom; many types of animals discriminate complex
sounds and recognize other individuals by their idiosyncratic voices, yet do
not display vocal imitation. Vocal learners have the special ability to
remember sounds and then use these sound memories to shape their
vocalizations. Furthermore, many of the phenomena of songbird vocal
learning have striking parallels with human speech acquisition, e.g., learn-
ing during a critical period, strongest imitation from a social tutor, and
lateralization of vocal control (Doupe & Kuhl, 1999). In addition, to our
knowledge, among all nonhuman animals, only songbirds have the capacity
to learn recursive, phrase-structured syntax, which was once considered to
be a unique characteristic of human language (Gentner, Fenn, Margoliash, &
Nusbaum, 2006). In an intriguing recent parallel, a gene that, when mutated,
produces a failure of normal speech development in humans, FOX-P2, is
present and developmentally regulated in the basal ganglia of songbirds
(Haesler et al., 2007); when it is blocked in the songbird, vocal learning is
degraded (Scharff & Haesler, 2005).
Although the songbird forebrain differs in structure from the mammalian
forebrain, neurobiological research into song learning mechanisms has been
enabled because many important parts of the brain circuit for vocal produc-
tion have been identified (Fig. 1a ). Song learning and production are served
by a specialized set of brain nuclei that exist in songbirds but not in other
birds (Nottebohm et al., 1976). These areas are larger in males than in
females (in most songbirds studied, the male sings more than the female),
and their development occurs postnatally, in parallel with vocal develop-
ment (Bottjer, Glaessner, & Arnold, 1985; Konishi & Akutagawa, 1985). The
telencephalic nucleus HVC (formerly known as Higher Vocal Center)
Patterns of Learning, Memory, and Vocal Production in the Songbird Brain 289
(a)
HVc
cHV
NCM L LMAN
NH
RA Area X
DLM
Nxllis
Syrinx Respiratory muscles
(b)
8
Amplitude
(Volts)
8
7
Frequency (kHz)
6
5
4
3
2
1
0
0 0.2 0.4 0.6 0.8 1.0 1.2 1.4
Time (s)
Fig. 1. (a) Neural substrates for learning: the song system. The motor pathway (dark
gray) is necessary for normal song production throughout life and includes HVC and the
robust nucleus of the arcopallium (RA). The robust nucleus of the arcopallium projects
to the tracheosyringeal portion of the hypoglossal nucleus (nXIIts), which controls the
bird’s vocal organ or syrinx, and to nuclei involved in the control of respiration during
song. HVC sends a second projection to the anterior forebrain pathway (AFP, medium
gray). The AFP includes area X, which is homologous to mammalian basal ganglia and
receives dopaminergic inputs, the medial nucleus of the dorsolateral thalamus (DLM),
and the lateral magnocellular nucleus of the anterior neostriatum (LMAN; a frontal
cortex-like nucleus). Lateral magnocellular nucleus of the anterior neostriatum sends a
projection to the motor pathway at the level of RA. The Field L complex is the avian
primary forebrain auditory area and projects to a complex network of higher auditory
areas (light gray), including the caudomedial nidopallium (NCM) (modified from
Brainard and Doupe, 2002). (b) Example of a zebra finch song. Songs typically consist
of a series of differentiated wide-band syllables, often frequency-modulated, separated
by silences in a fixed order that is repeated to form a motif (black bar).
projects to the robust nucleus of the arcopallium (RA). The projection from
RA to areas of the brainstem that control respiratory and syringeal vocal
effectors is known to be the final common output pathway from the tele-
ncephalon for song production. HVC also projects to area X in the basal
forebrain; area X has been shown to be part of the avian basal ganglia.
Damage to area X and associated structures has been shown to disrupt
song learning in development and prevent song plasticity in the adult
(Brainard & Doupe, 2002; Williams & Mehta, 1999). Figure 1 also shows
the auditory pathways of the telencephalon, described below.
Learned vocal signals consist of modulated respiratory pulses that are
shaped by the vocal tract. In songbirds, specialized forebrain areas serve the
production of these vocalizations through separable projection pathways
from motor nucleus RA to respiratory control areas and to motor neurons
innervating the bird’s vocal organ, the syrinx (Vicario, 1991). Unlike the
human larynx, the two halves of the syrinx can function as separate sound
sources that are independently controlled by the two brain hemispheres.
Learned song in the zebra finch (the most-studied songbird species) consists
of a stereotyped sequence of distinct sounds and silences (see Fig. 1b) pro-
duced by the modulatory action of the two syringeal halves on patterned
respiratory pressure pulses, further modified by the upper vocal tract (Goller &
Cooper, 2004; Suthers, 1997). Birdsongs consist of rapid motor sequences that
are naturally learned through a process of vocal imitation that has much in
common with human speech acquisition. Song is learned from an external
model, often the male parent. The young bird forms an auditory memory of the
tutor’s song and then uses that memory to guide sensorimotor practice until its
vocalizations match the tutor’s; this depends on auditory feedback (Konishi,
2004; Marler, 1970; Nottebohm, 1968). Song learning in the zebra finch pro-
ceeds rapidly, during a short developmental period. The bird transforms its
limited, unstable vocal repertoire into a pattern of distinct syllable types
defined by temporal and acoustic parameters—a process of differentiation
and stabilization (Derégnaucourt, Mitra, Fahér, Maul, Lints & Tchernichovski,
2004; Tchernichovski, Mitra, Lints, & Nottebohm, 2001).
This progressive but nonmonotonic process is a form of motor skill learn-
ing, like learning to play a musical instrument or perfecting a tennis stroke.
The extraordinary thing is that the ultimate result is a reasonable copy of the
original sound that was heard. Several learning processes overlap in time, at
least for the zebra finch and other songbirds that learn their songs during an
early period of development. At the start of the receptive period for acquiring
the model song, vocal production mechanisms are immature, and the bird
forms an auditory memory of the tutor song that will serve later imitation.
Copying does not need to take place in the presence of the song model; in fact,
in some species, the auditory and motor phases of song learning are separated
in time (Brainard & Doupe, 2002; Marler, 1970, 1991). In effect, the young
songbird first memorizes a tune and then learns to play it on an instrument that
it is simultaneously learning how to play for the first time (cf. Nottebohm,
1968). Memory processes are also involved in human speech acquisition;
infants bias their early vocal practice—‘‘babbling’’—toward recently heard
sounds (Kuhl & Meltzoff, 1996). Implicitly, the young bird or infant is learning
a set of correspondences between neural control, motor gestures, and their
auditory consequences, while favoring patterns that match the auditory
memory. The degree to which imitation results from instructive processes
that direct vocal changes vs. processes that select from a larger repertoire, or
some combination of the two, remains an active area for theory, experiment
and modeling efforts (Adret, 2004; Aronov et al., 2008; Marler, 1997).
3. Central and peripheral motor control and constraints
When an animal ‘‘learns’’ a behavior, it produces a given pattern of motor

behavior in the new associative context. The motor pattern itself might
appear new, but at a fine-grained level, it is assembled out of preexisting
patterns, themselves previously ‘‘learned’’ and/or unlearned. In effect,
motor skill learning emerges from the modification and recombination of
preexisting patterns. In the case of vocal learning, the gestural pattern must
be adjusted to fit a somewhat arbitrary external model acquired through
the acoustic modality. There is evidence that learned vocalizations are
assembled by modifying preexisting unlearned vocal gestures. This can
be seen in the maps of song development (based on vocalizations recorded
over the entire ontogeny) that show how distinct syllables in the mature
song can be traced back over ontogeny to less-differentiated early vocaliza-
tions present before exposure to tutor song (Derégnaucourt et al., 2004;
Tchernichovski et al., 2001). Figure 2 shows two-dimensional cluster plots
based on acoustic features of sounds produced by a young zebra finch at
different developmental stages. When a song is mature, the bird produces
differentiated sounds (the song syllables) that can be seen as tight clusters
in Fig. 2a. Early in the development, the bird produces highly variable
sounds (see Fig. 2c) that, under the influence of exposure to the tutor
song, begin to form clusters (see Fig. 2b).
The way in which a learned behavior is formed from unlearned compo-
nents can also be seen at the mechanistic level in the study of another
learned vocalization in the zebra finch, the long call (LC) (Price, 1979).
Male and female birds use the LC in the same behavioral context (Zann,
1984), typically when a zebra finch hears the vocalizations of another bird,
but cannot see it. However, the LCs of the two sexes differ in form as a
result of learning in males, but not in females (Zann, 1984, 1985). Three
acoustic features that distinguish the LCs of males from the simpler LCs of
Zebra Finch
A: 90 days old B: 47 days old (4th training day) C: 43 days old (1st training day)
90 90 90
Introductory
Mean FM
notes
Calls
0
400 0 400 0 400
Syllable duration (ms) Syllable duration (ms) Syllable duration (ms)
Time
Fig. 2. Feature maps plotting syllable duration against frequency modulation for
vocalizations recorded at different times during song development. The feature space
of song before tutor exposure is continuous with only one or two vague clusters of
simple calls (c). Within 4 days after the onset of training (b), the production space
takes the form of clusters. Even these very early clusters correspond to those of the
mature song (a) (modified from Derégnaucourt et al., 2004).
females are the following: (1) shorter, much more stable duration, (2) higher
fundamental frequency, and (3) presence of fast frequency modulations
(see also Simpson & Vicario, 1990). The LC of every male contains at least
one, and often two or three, of these features; each feature can vary para-
metrically, with the result that the LC of each male is virtually unique.
Although they are unlearned, the LCs of individual females contain subtler
idiosyncratic features that also differentiate them acoustically. Male LCs
resemble song syllables, and many males include a version of their LC as a
song syllable. These learned LCs also depend on the telencephalic song
control pathway (Nottebohm et al., 1976), as shown by lesion studies. The
learned frequency modulations depend on active control of the syrinx
because they are abolished when the nerves that innervate that organ are
cut; however, the short, stable duration remains (Fig. 3a). Lesion of HVC or
RA dramatically affected not only the acoustic structure but also the dura-
tion of the learned LC, but interestingly, did not abolish it (Simpson &
Vicario, 1990). Males with bilateral lesions of RA, for example, continued
to produce LCs in the correct behavioral context, but their LCs became
completely feminized: they no longer possessed male-typical features. In
females, bilateral RA lesions had no significant effect on the LC. Behavioral
responses to acoustic features of LCs also differ between the sexes; males
discriminate the male-typical spectral features, but females do not (Vicario,
2004). An intriguing additional effect of these RA lesions was that males lost
their ability to discriminate the male-typical features of LCs that they heard,
and thus were feminized in their discrimination of these features as well as
being unable to produce them.
(a) (b)
Female Male Female Male
8
7
6
5
4
3
2
Frequency (kHz)
Frequency (kHz)
1
8
7
6
5
4
3
2
1
150 ms
Fig. 3. Contributions of the syringeal nerves and robust nucleus of the arcopallium
(RA) to acoustic features of the long call (LC) in males in females. (a) Left: LCs of
one female before (top) and after (bottom) bilateral syringeal nerve section,
showing no change in the call. Right: LCs of one male before (top) and after
(bottom) bilateral syringeal nerve section, showing loss of fast frequency
modulation (open arrow) and lowered fundamental frequency. (b) Long calls of
one female before (top) and after (bottom) bilateral RA lesion, showing no change
in the call. Right: LCs of one male before (top) and after (bottom) bilateral RA
lesion, showing loss of fast frequency modulation (open arrow), lowered
fundamental frequency, and increase in duration. Time bar: 150 ms (modified
from Simpson & Vicario, 1990).
These results suggest that the smaller vocal control structures in the
female telencephalon make no essential contribution to the structure of
female calls. In males, these structures are necessary for the production of
learned male-typical features, but not for call production per se. This leads to
the interpretation that (1) brainstem vocal pathways are responsible for the
production of female-type LC in females and in RA-lesioned males, and
(2) the learned LCs of males are produced through modification of the
unlearned female-type call by the action of the telencephalic vocal pathway.
This picture also fits with recent data showing that HVC lesions do not affect
the early vocalizations produced by young males (Aronov, Andalman, &
Fee, 2008), suggesting that these are produced by another vocal system.
4. A specialized area for processing and remembering

conspecific vocal sounds
Auditory information reaches the avian auditory telencephalon (Field L

complex) from the thalamic homologue of the mammalian medial geniculate.
The Field L complex has a tonotopic organization and neurons there respond
briskly to simple stimuli, such as tones (Lewicki et al., 1995, 1996). Field L
neurons project directly and indirectly to the caudomedial nidopallium (NCM;
Vates, Mello, & Nottebohm, 1996). Caudomedial nidopallium was originally
identified as an important processing station for communication signals
because playback of novel conspecific songs is much more effective than
familiar conspecific song, heterospecific song, or nonsong sounds in inducing
the expression of an immediate early gene, zenk (also known as zif-268, egr-1,
ngfi-a, or krox-24) in the NCM of songbirds (Mello, Vicario, & Clayton, 1992).
Stronger responses to conspecific song and call stimuli were also documented
in electrophysiological studies of NCM (Chew et al., 1995, 1996a, 1996b).
Caudomedial nidopallium neurons have a further property: auditory
responses to novel stimuli are initially vigorous, but decrease (adapt) rapidly
to repeated stimulation, declining by about 50% over the first 50 iterations.
This adaptation is stimulus-specific and occurs independently and specifi-
cally for each song presented. The individually unique songs of different
birds provide a large array of stimuli. When a stimulus is novel, the rate of
adaptation (measured as the percent change in response amplitude for each
iteration of the stimulus) is fast; when a stimulus is recently familiar, the rate
of adaptation is slow. This difference in rate can be used to determine
whether any given stimulus is familiar to neurons at a site in NCM at any
given point in time. This method was used to show that the adaptation is
long-lasting, but that its duration depends on the type of auditory stimulus.
For conspecific song and LCs, adaptation lasts for up to 44 h after repeated
stimulus playback, but adaptation to heterospecific stimuli lasts less than 6 h
(Chew et al., 1995). Thus, in the adult, stimulus-specific adaptation in NCM
seems to be part of a specialized system of recognition memories that could
potentially store information about the vocalizations of recently heard indi-
vidual conspecifics for a period of hours or days.
A striking feature of vocal learning in songbirds is that young birds imitate
an adult tutor heard early in development. This depends on an auditory
memory of the tutor song, called a ‘‘template’’ (Konishi, 2004), that functions
to guide vocal development. We speculated that the ability to form auditory
memories in NCM could play a role in this process. To test this, young male
birds were raised in isolation and either tutored with an adult song during
the critical period for forming the template memory or untutored (left in
silence). Birds were then kept in silence for 40 days until adulthood. Their
songs were recorded and most tutored birds were found to have copied the
tutor song, but with varying degrees of fidelity measured by a similarity

index (SI) (Derégnaucourt et al., 2004). They were then tested for an auditory
memory of the tutor song experience in NCM (Phan et al., 2006). The birds
heard playbacks of the tutor song and of completely novel songs. The
adaptation rates of neural responses in NCM were then used to calculate a
familiarity index (FI) that had a value of 1 if a test song was not distin-
guished from novel songs and a value >1 if the song had a lower adaptation
rate than novel songs, indicative of familiarity. Tutored birds had a
significantly higher FI than untutored birds. When the FI was compared to
the SI for individual birds, they were highly correlated (Fig. 4a and 4b). This
(a) (b) (c)

2.0 2.0 2.0
Familiarity index for novel/BOS

Familiarity index for novel/tutor
Familiarity index for novel/tutor
1.5 1.5 1.5
1.0 1.0 1.0
0.5 0.5 0.5
2
R = 0.5148 2
R = 0.9036 R 2 = 0.084
0.0 0.0 0.0
0 20 40 60 80 100 0 20 40 60 80 100 0 20 40 60 80 100
% Similarity % Similarity % Similarity
Tutor song
Pupil-Tutored
87% similarity
50 msec
Fig. 4. The familiarity index (FI) for tutor song observed in NCM is a strong predictor
of imitative success. (a) The habituation rate in the caudomedial nidopallium (NCM)
is significantly correlated with the degree of similarity between bird’s own song
(BOS) and the tutor song (p = 0.0039) for all of the tutored intact and untutored
intact male birds. The correlation for the group of untutored intact birds was not
significant, as expected (p = 0.247). Further analysis showed that the measurements
of the percent similarity for two of the nine tutored intact birds did not differ
significantly from the untutored intact birds, indicating that they did not learn the
tutor song. (b) If the regression is limited to the birds that showed significant
learning, the correlation is significant (p = 0.001). (c) In these same birds, the FI for
the BOS was not significantly related to the similarity measurements of the percent
(p = 0.528), indicating that the strong correlation seen for the FI of tutor song is not
simply mediated by the similarity between tutor song and BOS. Lower panel shows
an example of the tutor song and the song produced by a pupil who achieved 87%
similarity score (modified from Phan et al., 2006).
suggested that birds that formed a stronger recognition memory for the tutor
song went on to make better copies. This effect was not simply due to
similarities between the tutor song and the bird’s own song (BOS), because
the SI was not correlated with the FI for BOS (see Fig. 4c). This result
established not only that NCM could form an auditory memory that lasted
for >40 days under these conditions, but also that the quality of that memory
predicted the fidelity of vocal imitation. Although this is only a correlation, it
focuses attention on the role of NCM in sensory and mnemonic aspects of the
vocal imitation process.
5. Conclusions
The songbird provides a potential laboratory for studying sensorimotor

learning on the systems level, which is not well-understood, despite great
progress in understanding the role of synaptic plasticity in associative learn-
ing. In the songbird, work focuses on a set of interconnected nuclei that are
specialized for vocal control—the song control system. This system devel-
ops during the critical period when song learning occurs in many avian
species, so that the overlap between the learning process and the develop-
mental process is an important feature of this preparation. However, intrigu-
ingly, ‘‘learning’’ is an operational and slippery concept, so we need a better
definition of kinds of processes involved in songbird vocal learning in order
to ask the right questions of the relevant brain structures at critical times.
We must try to ask the right questions by refining our concepts, meta-
phors, and hypotheses in light of new data and methodological opportu-
nities. In that spirit, this paper has attempted to distinguish components of
the learning process along several dimensions. First, a distinction must be
made between new behaviors and the modification/recombination of pre-
existing behavior patterns. The developmental maps of song syllables in
zebra finches strongly suggest that each sound evolves from earlier ver-
sions; there are few, if any, completely de novo events. However, recent
analysis of the temporal, rhythmic structure of emerging song (Saar &
Mitra, 2008) does suggest abrupt rhythmic transitions. The origin and
mechanism of these events remains to be elucidated. Nonetheless, the
effects of RA lesions on LC production—stripping the learned features
from the male LC to reveal the unlearned female form—strongly suggest
that the learning process sculpts preexisting material. Second, song pro-
duction involves the coordination of many effectors; notably, different
control signals must drive respiration, multiple syringeal muscles, and
other structures. Signals from the two hemispheres must be coordinated.
To some degree, each component may be separately learned and then
assembled into the form we hear. For example, it is often observed that
individual syllables are produced before their sequence and rhythm

become fixed. Third, vocal learning in the juvenile period occurs at a
time when many physiological systems may still be undergoing develop-
mental plasticity and critical period phenomena, especially in altricial
songbirds. In particular, sensory processing may still be maturing in the
auditory system, especially in auditory association areas like NCM, known
to process and remember conspecific songs selectively (Chew et al., 1995;
Mello, Velho, & Pinaud, 2004; Mello, Vicario, & Clayton, 1992), and in CMM,
which encodes learned auditory objects (Gentner, 2004). Fourth, although
auditory responses to the BOS in the vocal control pathway are undoubt-
edly a significant clue to sensorimotor mechanisms, they have tended to
limit our thinking about a larger range of perceptual and memory processes
that may be essential to vocal learning and vocal communication, as
cogently argued by Bolhuis and Gahr (2006). Recent evidence from both
gene expression and electrophysiological studies suggests a likely role for
NCM and connected areas in forming auditory memories for songs heard in
development (Phan et al., 2006; Terpstra, Bolhuis, Riebel, van der Burg, &
Boer-Visser, 2006).
The question of how sensory information, for example, the auditory
memory of the tutor song, influences motor structures to alter behavior
remains a fundamental one in neuroscience; the answer is likely to be
elusive. A more modest hope is that the song system affords us an oppor-
tunity to study a simplified form of vocal learning that has significant
mechanistic as well phenomenological parallels with human speech acqui-
sition. This approach has the potential to provide a descriptive model of
how auditory stimuli are processed and stored during development, how
the development of perceptual systems interacts with vocal development,
and how multiple motor patterns, controlling different vocal effectors,
combine to produce the song we hear.
Acknowledgments
I apologize to the many authors who have contributed to the field and to
the ideas presented here but could not be cited due to space restrictions.
References
Adret, P. (2004). In search of the song template. Ann N Y Acad Sci. 1016: 303–24.
Aronov, D., Andalman, A. S., & Fee, M. S. (2008). A specialized forebrain circuit for
vocal babbling in the juvenile songbird. Science, 320, 630–634.
Bolhuis, J. J., & Gahr, M. (2006). Neural mechanisms of birdsong memory. Nature
Reviews Neuroscience, 7(5), 347–357.
Bottjer, S. W., Glaessner, S. L., & Arnold, A. P. (1985). Ontogeny of brain nuclei controlling
song learning and behavior in zebra finches. Journal of Neuroscience, 5, 1556–1562.
Brainard, M. S., & Doupe, A. J. (2002). What songbirds teach us about learning.
Nature, 417, 351–358.
Chew, S. J., Mello, C., Nottebohm, F., Jarvis, E., & Vicario, D. S. (1995). Decrements in
auditory responses to a repeated conspecific song are long-lasting and require two
periods of protein synthesis in the songbird forebrain. Proceedings of the National
Academy of Sciences of the United States of America, 92, 3406–3410.
Chew, S. J., Vicario, D. S., & Nottebohm, F. (1996a). A large-capacity memory
system that recognizes the calls and songs of individual birds. Proceedings of
the National Academy of Sciences of the United States of America, 93, 1950–1955.
Chew, S. J., Vicario, D. S., & Nottebohm, F. (1996b). Quantal duration of auditory
memories. Science, 274, 1909–1914.
Derégnaucourt, S., Mitra, P. P., Fehér, O., Maul, K. K., Lints, T. J., & Tchernichovski, O.
(2004). Song development: in search of the error-signal. Annals of the New York
Academy of Sciences, 1016, 364–376.
Doupe, A. J., & Kuhl, P. K. (1999). Birdsong and human speech: common themes and
mechanisms. Annual Review of Neuroscience, 22, 567–631.
Gentner, T. Q. (2004). Neural systems for individual song recognition in adult birds.
Annals of the New York Academy of Sciences, 1016, 282–302.
Gentner, T. Q., Fenn, K. M., Margoliash, D., & Nusbaum, H. C. (2006). Recursive
syntactic pattern learning by songbirds. Nature, 440, 1204–1207.
Goller, F., & Cooper, B. G. (2004). Peripheral motor dynamics of song production in
the zebra finch. Annals of the New York Academy of Sciences, 1016, 130–152.
Haesler, S., Rochefort, C., Georgi, B., Licznerski, P., Osten, P., & Scharff, C. (2007).
Incomplete and inaccurate vocal imitation after knockdown of FoxP2 in songbird
basal ganglia nucleus area X. PLoS Biology, 5, e321.
Konishi, M., & Akutagawa, E. (1985). Neuronal growth, atrophy and death in a
sexually dimorphic song nucleus in the zebra finch brain. Nature, 315, 145–147.
Konishi, M. (2004). The role of auditory feedback in birdsong. Annals of the
New York Academy of Sciences, 1016, 463–475.
Kuhl, P. K., & Meltzoff, A. N. (1996). Infant vocalizations in response to speech:
vocal imitation and developmental change. Journal of the Acoustical Society of
America, 100(4), 2425–2438.
Lewicki, M. S., & Konishi, M. (1995). Mechanisms underlying the sensitivity of
songbird forebrain neurons to temporal order. Proceedings of the National
Academy of Sciences of the United States of America, 92, 5582–5586.
Lewicki, M. S., & Arthur, B. J. (1996). Hierarchical organization of auditory temporal
context sensitivity. Journal of Neuroscience, 16, 6987–6998.
Marler, P. (1970). A comparative approach to vocal learning: song development in
white-crowned sparrows. Journal of Comparative and Physiological Psychology
Monographs, 71, 1–25.
Marler, P. (1991). Song-learning behavior: the interface with neuroethology. Trends
in Neurosciences, 14, 199–206.
Marler, P. (1997). Three models of song learning: evidence from behavior. Journal
of Neurobiology, 33, 501–516.
Mello, C. V., Velho, T. A., & Pinaud, R. (2004). Song-induced gene expression:
a window on song auditory processing and perception. Annals of the New York
Academy of Sciences, 1016, 263–281.
Mello, C. V., Vicario, D. S., & Clayton, D. F. (1992). Song presentation induces gene
expression in the songbird forebrain. Proceedings of the National Academy of
Sciences of the United States of America, 89, 6818–6822.
Nottebohm, F. (1968). Auditory experience and song development in the Chaffinch,
Fringilla coelebs. The Ibis, 110, 549–568.
Nottebohm, F., Kasparian, S., & Pandazis, C. (1981). Brain space for a learned task.
Brain Research, 213, 99–109.
Nottebohm, F., Stokes, T. M., & Leonard, C. M. (1976). Central control of song in the
canary, Serinus canarius. Journal of Comparative Neurology, 165, 457–486.
Phan, M. L., Pytte, C. L., & Vicario, D. S. (2006). Early auditory experience generates
long-lasting memories that May subserve vocal learning in songbirds. Proceedings
of the National Academy of Sciences of the United States of America, Jan, 24;103.
Price, P. H. (1979). Developmental determinants of structure in zebra finch song.
Journal of Comparative and Physiological Psychology, 93, 260–277.
Saar, S., & Mitra, P. P. (2008). A technique for characterizing the development of
rhythms in bird song. PLoS ONE, 3, e1461.
Scharff, C., & Haesler, S. (2005). An evolutionary perspective on FoxP2: strictly for
the birds. Current Opinion in Neurobiology, 15, 694–703.
Simpson, H. B, & Vicario, D. S. (1990). Brain pathways for learned and unlearned
vocalizations differ in zebra finches. Journal of Neuroscience, 10, 1541–1556.
Suthers, R. A. (1997). Peripheral control and lateralization of birdsong. Journal of
Neurobiology, 33, 632–652.
Tchernichovski, O., Mitra, P. P., Lints, T., & Nottebohm, F. (2001). Dynamics of the
vocal imitation process: how a zebra finch learns its song. Science, 291, 2564–2569.
Terpstra, N. J., Bolhuis, J. J., Riebel, K., van der Burg, J. M., & Boer-Visser, A. M. D.
(2006). Localized brain activation specific to auditory memory in a female song-
bird. Journal of Comparative Neurology, 494, 784–791.
Vates, G. E., Mello, C. V., & Nottebohm, F. (1996). Auditory pathways of caudal
telencephalon and their relation to the song system of adult male zebra finches.
Vicario, D. S. (1991). Organization of the zebra finch song control system:
II. Functional organization of outputs from nucleus robustus archistriatalis.
Vicario, D. S. (2004). Using learned calls to study sensory–motor integration in
songbirds. Annals of the New York Academy of Sciences, Jun, 1016, 246–262.
Williams, H., & Mehta, N. (1999). Changes in adult zebra finch song require a
forebrain nucleus that is not necessary for song production. Journal of Neuro-
biology, 39, 14–28.
Zann, R. (1984). Structural variation in the zebra finch distance call. Zeitschrift fur
Tierpsychologie, 66, 328–345.
Zann, R. (1985). Ontogeny of the zebra finch distance call: I. Effects of cross-
fostering to Bengalese finches. Zeitschrift fur Tierpsychologie, 68, 1–23.
Human Learning 301
The cAMP/PKA Pathway and the

Modeling of Human Memory
Disorders in Mice
Francis X. Brennan1 and Ted Abel2
1
RedPoint Bio Corporation
2
Department of Biology, University of Pennsylvania
Memory is a man’s real possession . . . In nothing else is he rich, in

nothing else is he poor.
Alexander Smith
Scottish essayist & poet (1830–1867)
1. Introduction: Memory and the cyclic AMP (cAMP)/ protein kinase

A (PKA) pathway
Memory is the retention of information over time and is important on a

multitude of levels, ranging from basic survival to quality of life in humans.
A number of human disease states, including Alzheimer’s disease (AD),
mental retardation (MR), and schizophrenia, are characterized by signifi-
cant memory impairments. Although a great deal of progress has been
made in recent years in understanding the neurobiology of memory
(Barco, Bailey, & Kandel, 2006), many of the molecular details remain
elusive. However, it is apparent that the formation of long-term memories
in all animals thus far investigated involves gene expression and protein
synthesis via the activation of a number of intracellular pathways (Sutton &
Schuman, 2006). The cAMP/PKA second messenger signal transduction
pathway is the most extensively characterized of the various second mes-
senger systems (see Korzus, 2003), and is illustrated in Fig. 1.
G protein-coupled receptors comprise a large superfamily associated
with a number of different transmitters including dopamine, serotonin,
GABA, and glutamate (Luttrell, 2006). The cAMP/PKA system was initially
shown to be important for the formation of long-term memories in elegant
studies using fruit flies (Drosophila melanogaster). Flies with genetic
(a) (c)
302
450 WT, Spaced
fEPSP slope (% baseline)

400 R(AB), Spaced CaMKIV
350 MAP Kinase
P Ac
N
Ac
Ac
300 CBP
Ac
CaMKII P P N
Ac Ac
N
N
Ac
250 CREB CREB
Ac
CRE CRE TARGET GENES
200
150
I-1 Nucleus
100 Calcineurin
50 5 min, spaced PDE

Tyrosine Kinase PPI
0 Protein Kinase C PKA
Francis X. Brennan and Ted Abel

0 20 40 60 80 100 120 140 160 CaMKII
cAMP
Time (min)
2–
Ca Calmodulin
Cytoplasm
(b) Adenylyl
Postsynaptic Cyclase
450 WT, Massed
fEPSP slope (% baseline)
Neuron 9 AK
400 R(AB), Massed AP AP
5
AK Ca2+
350
NMDAR AMPAR Modulatory Inputs
300
(Dopamine)
250 Ca2+
200
150
100
Presynaptic
50 3 s, massed Neuron
0
0 20 40 60 80 100 120 140 160
Time (min)
Fig. 1. Schematic of G protein signaling. The cAMP/PKA pathway becomes activated when a ligand binds to a G protein-coupled
receptor (e.g., D1/D5), or by calcium influx. After ligand binding, the stimulatory a subunit (Gas) dissociates from the receptor
complex and stimulates adenylyl cyclase (AC), a membrane-bound enzyme that causes cAMP formation. Cyclic AMP then activates
a number of targets, including cAMP-dependent protein kinase (PKA). Activated PKA moves to the cell nucleus and phosphorylates
the cAMP response element-binding (CREB) protein.
The cAMP/PKA Pathway and the Modeling 303
mutations of various components of the cAMP/PKA pathway all showed

deficits in classical conditioning (Dubnau, 2004). Flies with alterations in
phosphodiesterase (PDE) (dunce) or adenylyl cyclase (AC) (rutabaga)
show significant deficits in memory formation (reviewed in Margulies,
Tully, & Dubnau, 2005). Subsequently, cAMP response element-binding
(CREB) protein has been shown to play a similar role in Aplysia neurons
(Dash et al., 1990). Although it was research in Drosophila that illustrated
the behavioral role of PKA, it was research in Aplysia that demonstrated
that PKA has a role in synaptic plasticity (Hawkins, Kandel, & Bailey, 2006).
Further, transgenic mice with reduced PKA activity also showed reduc-
tions in both the late phase of long-term potentiation (L-LTP) and memory
deficits (Abel, Nguyen, Barad, et al., 1997). The results from the animal
models thus strongly implicate the cAMP/PKA pathway as a critical path-
way involved in memory storage.
Since the basic cAMP/PKA signaling pathway has been retained in
humans, an interesting question is whether impairments in the functioning
of this pathway may be related to the memory impairments in various
human disorders. There are preliminary but exciting data to suggest that
this is in fact the case. For example, patients with schizophrenia have been
shown to have abnormalities of the cAMP system (Muly, 2002), including
overactivity of G protein signaling (Avissar, Barki-Harrington, Nechamkin,
Roitman, & Schreiber, 2001). This link, although exciting, is of course
extremely preliminary.
Cyclic AMP levels in vertebrates are regulated partly by a negative
feedback system that involves the PDEs, in particular PDE4 (Ghavami,
Hirst, & Novak, 2006). Because cAMP activity has been associated with
memory formation, one therapeutic strategy has been to increase its activ-
ity by reducing the negative feedback of the PDEs (Blokland, Schreiber, &
Prickaerts, 2006). Thus, drugs such as rolipram, a PDE4 inhibitor, have
been used to increase activity in the cAMP pathway and improve memory.
As we discuss in detail below, rolipram and related compounds may have
significant potential as novel pharmacotherapeutics (Kanes et al., 2007).
Animal models are critical to understand and evaluate potential treat-
ments of human psychiatric disorders (e.g., see Arguello & Gogos, 2006;
Bartke, 2006). In recent years, mice have become the vertebrate species of
choice in neurobiology because of the ability to perform genetic manipula-
tions (Bucan & Abel, 2002; Tecott, 2003). It is of course critical to under-
stand that a genetically altered mouse used to model a human psychiatric
condition does not in any meaningful way have the human disorder. How-
ever, the animals may exhibit endophenotypes, relevant phenotypic com-
ponents of the disorder (Bucan & Abel, 2002). An implicit assumption is
that studying the relevant endophenotypes in animal models will lead to an
improved understanding of the disorder in humans. A final issue is
304 Francis X. Brennan and Ted Abel
modeling cognition or memory in a species with demonstrably less-

advanced cognitive ability than humans. It is clear that mice do not possess
the same level of processing ability as humans. However, with the selection
of appropriate behavioral tasks, it is thought that a broad construct such as
‘‘working memory’’ is retained across phylogeny. This retention will then be
translated into an endophenotype that can be used to study the disorder.
In this chapter, we focus on three classes of memory disorders: AD, MR,
and schizophrenia. There is strong evidence that perturbations in the
cAMP/PKA signaling system may be related to the memory problems in
each disorder. We briefly review each disorder, and then discuss the
evidence that disruptions in the cAMP/PKA pathway are at least partially
responsible for the memory impairments. Finally, we discuss novel treat-
ment options that build on our increasing knowledge of intracellular sig-
naling pathways.
2. Alzheimer’s disease
2.1. CLINICAL SYMPTOMS
There are a number of human disease processes associated with or

caused by degeneration of neural tissues. The most intensively studied of
these is AD, a disorder initially of memory that is becoming more prevalent
as the population ages (Libow, 2007). Alzheimer’s disease is the most
frequent type of dementia affecting over 20 million people worldwide,
and is related to a build-up of b-amyloid protein in the brain, in particular
limbic structures such as hippocampus, as well as prefrontal cortex.
Impairments in short-term memory are the first clinically relevant sign
and correlate with hippocampal damage (Blennow, de Leon, & Zetterberg,
2006). Eventually, neuronal death results with attendant multiple cognitive
deficits in language, spatial orientation, executive functioning, and signifi-
cant impairments in function (Newman, Musgrave, & Lardelli, 2007).
Considerable progress has been made recently in understanding the
pathophysiology of AD. The b-amyloid protein plaques characteristic of
AD are derived from proteolytic cleaving of the amyloid precursor protein
(APP; Shen & Kelleher, 2007). Presenilins are a class of transmembrane
proteins that are associated with the protease complex. There are two
presenilin genes (PS1 and PS2; Newman et al., 2007), mutations of which
have been associated with early onset AD. The much more common late-
onset AD is also hypothesized to be related to presenilin dysfunction.
Further, a presenilin knockout (KO) in mice mimics many of the cognitive
and neurodegenerative phenotypes of AD (Shen & Kelleher, 2007).
A buildup of b-amyloid protein at levels too low to induce neurotoxicity
may inhibit both PKA and LTP and produce memory disturbances early
in the disorder prior to actual loss of neurons (Arancio & Chao, 2007).
A very recent report described a reduction in PKA in brains of patients with
AD (Liang, Liu, Grundke-Iqbal, Iqbal, & Gong, 2007). This may subsequently
lead to downregulation of CREB proteins and problems in cognition and
memory.
2.2. AD AND THE cAMP/PKA PATHWAY
Although there are a number of putative mouse models of AD, the most
common are transgenic mice over- or underexpressing the genes thought to
be important in AD pathology (Ryman & Lamb, 2006). There are a number
of reports that implicate the cAMP/PKA pathway in the symptom complex
of AD. Arancio and colleagues have published a number of illuminating
reports on this topic (Arancio & Chao, 2007; Gong et al., 2004). Interest-
ingly, although late in the course of AD, b-amyloid protein accumulation is
clearly related to neuronal loss and cognitive function; these relationships
are less clear earlier in the disease process when memory impairments first
become evident (Arancio & Chao, 2007). Early in the course of the disease
is where the cAMP/PKA system may be impacted and also may represent a
novel therapeutic target.
Vitolo et al. (2002) reported a study conducted in rat hippocampal slices
that illustrated the potential efficacy of the PDE4 inhibitor rolipram. Slices
treated with b-amyloid protein resulted in a dose-dependent decrease in
PKA activity. Interestingly, the b-amyloid doses used were well below the
level necessary to induce cell death in vivo. Further, the inhibition was
eliminated by concurrently treating the slices with rolipram. Thus, early in
the course of AD, drugs such as rolipram may be a novel treatment strategy
to improve cognitive function. Arancio and colleagues have further
advanced this area with a study of double transgenic mice (Gong et al.,
2004). These animals overexpress both APP and presenilin-1. Hippocampal
slices from the double transgenics were deficient in LTP. This deficit was
also eliminated by rolipram treatment (Gong et al., 2004). A deficit
in contextual fear conditioning in the double transgenics was also reported.
Contextual fear conditioning in mice involves returning the animals to a
context where they received shock and measuring freezing, a behavioral
index of fear. The basic paradigm is depicted in Fig. 2. The deficit in fear
conditioning, manifested as reduced freezing upon reexposure to the con-
text where shock was received, was also eliminated by rolipram treatment.
Finally, a 3-week treatment regimen of rolipram eliminated the electrophy-
siological and behavioral deficits 2 months after treatment cessation (Gong
et al., 2004). This indicates that rolipram may function to alter synaptic
circuitry in a long-term beneficial manner. This finding has important
ramifications for the potential treatment of human patients.
Cued conditioning
Test at 24 h
Onset of sound (CS)
Onset of shock (US)
Onset of sound (CS)

Contextual conditioning Different context
Test at 24 h
Exposure to context
Onset of shock (US)
Same context
Fig. 2. Schematic of cued and contextual fear conditioning procedure. Animals are
first placed in a specific context and briefly allowed to acclimate. In the cued
conditioning procedure (top), the animal receives one or more tone–footshock
pairings. In contextual conditioning (bottom), the animal receives shock alone. At
some time later, typically 24 h, the animal is either exposed to the tone in a different
context or returned to the context where shock was received. Memory of the shock
presumably induces fear, which is behaviorally measured by the amount of time
that the mouse freezes or is completely immobile.
Other studies have focused on transgenic animals that attempt to

produce relevant endophenotypes (Comery et al., 2005). Thus, mice that
constitutively overexpress APP display a number of biochemical and
behavioral characteristics relevant to AD. These mice show impair-
ments in contextual fear conditioning that develop as early as 16
weeks of age. Further, the animals display normal fear at retention
intervals of 1- or 6-h post-training, but are significantly impaired at
20 h (Comery et al., 2005). This appears to indicate impairment in
long-term memory storage, analogous to AD patients. Interestingly,
the PDE4 inhibitor rolipram significantly improved retention in the
transgenics. Rolipram also improved memory in the controls, indicating

that the cognitive-enhancing effects are independent of APP and
b-amyloid. Finally, treatment of the transgenics with an inhibitor of the
APP proteolytic enzymes both reduced b-amyloid levels and improved
retention. Thus, this preliminary evidence indicates that inhibiting the
proteolytic enzymes that cleave APP into b-amyloid may be a novel treat-
ment for both the cognitive and the neurological features of AD.
3. Mental retardation
3.1. A MULTITUDE OF CAUSES
Mental retardation is a general term that describes cognitive impair-

ments in humans (DSM-IV, 1994). It displays a highly diverse phenotype
in terms of the scope of cognitive impairment, as well as the presence or
absence of additional symptoms. It is typically initially recognized in chil-
dren by delays in meeting developmental milestones. There are a tremen-
dous number of causes of MR, ranging from environmental perturbations
such as anoxia, infection, or fetal alcohol syndrome to a number of genetic
conditions including Down’s syndrome, phenylketonuria, and Fragile X
syndrome (Dierssen & Ramakers, 2005). It has been hypothesized recently
that although a number of different mutations are associated with MR, the
common feature may be a dysfunction in synaptic connectivity, with resul-
tant cognitive deficits (Dierssen & Ramakers, 2006). If this is in fact the
case, then as neuroscience knowledge of the factors that influence synaptic
connections expands, the potential for the development of pharmacothera-
pies becomes greater.
3.2. MR AND THE cAMP/PKA PATHWAY
There are several types of MR that are characterized by alterations in

the cAMP/PKA pathway. One initially described over 60 years ago is
Albright hereditary osteodystrophy (AHO), a disorder characterized by
a mutation in the gene coding for Gas (Albright, Burnett, Smith, &
Parson, 1942). Albright hereditary osteodystrophy is associated with a
number of endocrine deficiencies, as well as MR in most cases (Farfel &
Friedman, 1986). Another genetic disorder characterized by MR as well
as other physical symptoms is Rubinstein–Taybi syndrome (RTS),
a rare disorder linked to chromosome 16 (Hallam & Bourtchouladze,
2006). The specific mutation involves CREB-binding protein (CBP),
although RTS can also be produced by mutations in the EP300 gene,
which codes for the homologous p300 protein (Roelfsema et al., 2005).
CBP is a transcriptional coactivator that binds to phosporylated CREB
proteins and results in gene transcription (Kwok et al., 1994). cAMP
response element-binding protein-induced gene transcription is a critical
step in the formation of long-term memory (Tully et al., 2003). Recently, a
mouse model of RTS has been developed by producing an animal that
develops a truncated form of CBP (Oike et al., 1999). These mice showed a
number of physical endophenotypes characteristic of RTS and also dis-
played some apparent deficits in long-term memory. The CBP-deficient
mice were less active and showed less rearing responses in an open-field
environment. Interestingly, Y-maze alternation learning was normal in the
CBP-deficient animals. The CBP-deficient animals were impaired at
retaining a passive avoidance response, indicated by shorter latencies to
enter a chamber where shock had been received 1 day after training. In a
cued fear conditioning procedure, the CBP-deficient animals showed less
freezing to the CS only 24 h after training, but not before, indicating a
deficit specific to long-term memory formation.
Another investigation of the cognitive abilities of the CBP-deficient
animals was described by Bourtchouladze et al. (2003). The CBP-deficient
animals were significantly impaired in an object recognition task, which
requires the animal to discriminate between a novel object and one that
they have seen before. Again, the impairment was seen only in a version
that tested long-term memory. As in the report from Oike et al (1999),
short-term memory appeared unaffected. Most strikingly, the authors
tested two different PDE4 inhibitors, rolipram and HT0712. Both com-
pounds dose-dependently reversed the memory impairments in the CBP-
deficient animals. Collectively, these data indicate that PDE inhibitors may
represent a novel pharmacotherapeutic strategy for RTS patients.
Both the Oike et al. (1999) and the Bourtchouladze et al. (2003) reports
utilized animals that had the impaired allele throughout development. It is
thus impossible to differentiate developmental from acute effects in these
studies. Accordingly, Wood and colleagues developed a transgenic model
of truncated CBP protein that was driven by the CaMKIIa promoter, which
is restricted to forebrain neurons. Since the CaMKIIa promoter does not
become active until several weeks of postpartum, developmental effects
are reduced. The mice were physically similar to wild types, without the
abnormalities in the KOs described by Oike et al. (1999). Importantly, the
transgenic animals were also impaired in two hippocampal-based tasks,
the hidden platform version of the Morris water maze and contextual fear
conditioning (Wood et al., 2005). These data indicate that CBP is involved
in synaptic plasticity in the hippocampus and hippocampal-dependent
long-term memory formation.
4. Schizophrenia
4.1. A COMPLEX DISORDER WITH MANY SYMPTOMS
Schizophrenia is a psychiatric disorder that describes a group of highly

debilitating mental disorders characterized by abnormalities in emotion,
sensory processing, as well as memory and cognition (DSM-IV, 1994).
Although a number of antipsychotic medications are highly effective in redu-
cing the ‘‘positive’’ symptoms such as auditory hallucinations and delusions,
cognitive, and memory disturbances often persist (Gray & Roth, 2007). Even
with the elimination of the positive symptoms, these attentional and cognitive
endophenotypes typically prevent a patient from functioning effectively in
everyday life (Bowie & Harvey, 2005). This illustrates the need for novel
medications for schizophrenics to resume normal lives. The memory impair-
ments appear to be due in part to an underlying hippocampal dysfunction, and
a number of studies have reported neuroanatomical abnormalities in schizo-
phrenic patients (reviewed in Boyer, Phillips, Rousseau, & Ilivitsky, 2007).
The ‘‘dopamine theory of schizophrenia’’ is commonly accepted and
overactivity of dopaminergic systems is thought to be causally linked to
hallucinations and delusions; accordingly, all effective antipsychotics are
antagonists to the D2 receptor subtype (Guillin, Abi-Dargham, & Laruelle,
2007). The neurobiology of the attentional and cognitive symptoms is not
as well understood. Alterations in glutamatergic, as well as dopamine,
cholinergic, and serotonergic function, have been postulated (Sawa &
Snyder, 2002). The myriad of transmitter systems postulated to be involved
have led some investigators to search for a common ‘‘downstream’’ intra-
cellular mechanism. The cAMP/PKA pathway appears to be an excellent
candidate to be the unifying intracellular mechanism behind the cognitive,
memory, and attentional deficits in schizophrenia. This is further sup-
ported by a number of genes such as DISC-1 that have been linked to
schizophrenia and synaptic plasticity (Roberts, 2007).
4.2. SCHIZOPHRENIA AND THE cAMP/PKA PATHWAY
A number of converging lines of evidence point to dysregulation in the

cAMP/PKA pathway in schizophrenia. In humans, Gas is encoded by
the GNAS1 gene. A specific polymorphism in the gene has been linked
to the incidence of deficit (negative) schizophrenia in an Italian cohort
(Minoretti et al., 2006). Also, measures of Gas activity in leukocytes have
been shown to be increased in untreated schizophrenics compared to
healthy controls (Avissar et al., 2001). Finally, Monteleone et al. (2002)
reported a significant inverse correlation between symptom severity in a
population of deficit schizophrenics and Gas activity in leukocytes.
The use of genetically modified animals is an invaluable tool for both

modeling human disorders and evaluating the effects of potential novel ther-
apeutics. Accordingly, a transgenic strain of mice that overexpress Gas (Gas*)
has been developed (Abel et al., 1997). As it is dependent on the CaMKIIa
promoter, the overexpression is limited to the cortex, amygdala, hippocampus,
striatum, and olfactory bulb. The overactivity of Gas should theoretically lead
to an excess of cAMP. However, similar to schizophrenic patients (Muly, 2002),
these animals actually have decreased cAMP levels in cortical neurons. The
decrease in cAMP appears to be caused by PKA-dependent upregulation of
PDEs (Kelly et al., 2007a), enzymes that degrade cAMP.
The Gas* transgenic mouse has been behaviorally characterized in a
number of tasks relevant to schizophrenia. These animals show deficits in
prepulse inhibition (PPI) similar to schizophrenic patients (Gould et al.,
2004; Kelly et al., 2007a). They also show electrophysiological impairments
in stimulus encoding (Maxwell et al., 2006). Interestingly, the PPI deficits
are reversible after treatment with either haloperidol or rolipram (Kelly
et al., 2007a). The fact that rolipram reversed the deficits in PPI indicates
that PDE inhibitors have potential as a novel class of antipsychotics. The
transgenics have also been studied in several long-term memory tasks and
show impairments in both contextual and cued fear conditioning para-
digms, and perform poorly in the hidden platform version of the Morris
water maze (Bourtchouladze et al., 2006). Both contextual fear condition-
ing and the hidden version of the Morris maze are tasks that require an
intact hippocampus (Phillips & LeDoux, 1992), and thus the deficits in the
transgenics mimic the memory impairments of schizophrenic patients.
5. Treatment/future directions
5.1. HISTONE DEACETYLASE INHIBITORS
We have previously discussed the therapeutic potential of PDE inhibi-

tors such as rolipram (e.g., Kanes et al., 2007). Although rolipram has been
shown to be effective in animal models for quite some time (e.g., Weishaar,
Cain, & Bristol, 1985), translation to humans has proven difficult. Another
potential avenue for treatment of disorders of memory and cognition is
histone deacetylase (HDAC) inhibitors. Both PDE and HDAC inhibitors
target memory suppressor genes (e.g., Abel et al., 1998). Histones are
proteins surrounded by DNA. Adding an acetyl group to a histone appears
to improve gene transcription and thus memory. Removing an acetyl group,
a process called deacetylation, has the opposite effect. Compounds that
deacetylate histones are referred to as histone deactylators. HDAC inhibi-
tors would inhibit this process and presumably improve gene transcription
and memory. Therefore, HDAC inhibitors such as trichostatin A (TSA) have

been used as cognitive enhancers (Vecsey et al., 2007). Infusion of TSA into
the hippocampus improved both memory in a contextual fear conditioning
procedure and LTP (Vecsey et al., 2007). Interestingly, TSA did not enhance
cued fear memory, a nonhippocampal task, indicating that HDAC inhibitors
may work on hippocampal-dependent tasks and have utility for a number
of human disorders characterized by memory problems. Further, gene
transcription was not globally increased by TSA, but rather only specific
genes related to CBP. A recent report also indicates that TSA improved
extinction in a contextual fear conditioning paradigm (Lattal, Barret, &
Wood, 2007).
5.2. VARIATIONS IN NORMAL HUMAN MEMORY
We have argued in the current chapter that malfunction in a given

system such as cAMP/PKA may be responsible for phenotypic impairments
in memory associated with conditions such as AD, MR, and schizophrenia.
This may lead to novel treatment options for the memory symptoms of
these disorders that are refractory to current treatments.
An interesting corollary to the description of disease states is the idea
that genetic variations in these molecules, outside the realm of disease,
may be responsible for individual differences in human memory. Concei-
vably, the development of novel compounds for cognition and memory in
different disorders could also be used to improve the long-term memory
ability of people with less than ideal memory due to aging, drug or alcohol
abuse, or simply innate ability. Thus, we describe an extremely interest-
ing report by De Quervain and Papassotiropoulos (2006). They reported
that certain genes in ‘‘normal humans’’ correlated with performance in an
episodic memory task. Among a total of seven genes in a cluster that
correlated with performance on the memory task were genes associated
with adenylyl cyclase, PKA, and CaMKII (De Quervain & Papassotiropou-
los, 2006). These data support and extend the numerous animal studies
that implicate the cAMP/PKA system in memory. A complete understand-
ing of this system may lead to novel medications for human memory
disorders. If this comes to fruition, the basic research in mice will be
indispensable.
Acknowledgments
Due to space restrictions, we were able to describe in a number of

instances only review articles. We apologize to many scientists whose
contributions were not covered in detail that the work warranted.
References
Abel, T., Nguyen, P. V., Barad, M., Deuel, T. A., Kandel, E. R., & Bourtchouladze, R.
(1997). Genetic demonstration of a role for PKA in the late phase of LTP and in
hippocampus-based long-term memory. Cell, 88, 615–626.
Abel, T., Martin, K. C., Bartsch, D., & Kandel, E. R. (1998). Memory suppressor genes:
inhibitory constraints on the storage of long-term memory. Science, 279, 338–341.
Albright, F., Burnett, C. H., Smith, C. H., & Parson, W. (1942). Pseudohypoparathyr-
oidism: an example of Seabright-bantam syndrome. Endocrinology, 30, 922–932.
Arancio, O., & Chao, M. (2007). Neurotrophins, synaptic plasticity and dementia.
Arguello, P. A., & Gogos, J. A. (2006). Modeling madness in mice: one piece at a
time. Neuron, 52, 179–196.
Avissar, S., Barki-Harrington, L., Nechamkin, Y., Roitman, G., & Schreiber, G. (2001).
Elevated dopamine receptor-coupled G(s) protein measures in mononuclear leuko-
cytes of patients with schizophrenia. Schizophrenia Research, 47, 37–47.
Barco, A., Bailey, C. H., & Kandel, E. R. (2006). Common molecular mechanisms in
explicit and implicit memory. Journal of Neurochemistry, 97, 1520–1533.
Bartke, A. (2006). New findings in transgenic, gene knockout, and mutant mice.
Experimental Gerontology, 41, 1217–1219.
Blennow, K., de Leon, M. J., & Zetterberg, H. (2006). Alzheimer’s disease. The
Lancet, 368, 387–403.
Blokland, A., Schreiber, R., & Prickaerts, J. (2006). Improving memory: a role for
phosphodiesterases. Current Pharmaceutical Design, 12, 2511–2523.
Bourtchouladze, R., Patterson, S. L., Kelly, M. P., Kreibich, A., Kandel, E. R., & Abel, T.
(2006). Chronically increased gsalpha signaling disrupts associative and spatial
learning. Learning Memory, 13, 745–752.
Bowie, C. R., & Harvey, P. D. (2005). Cognition in schizophrenia: impairments,
determinants, and functional importance. Psychiatric Clinics of North America,
28, 613–633.
Boyer, P., Phillips, J. L., Rousseau, F. L., & Ilivitsky, S. (2007). Hippocampal
abnormalities and memory deficits: new evidence of a strong pathophysiological
link in schizophrenia. Brain Research Reviews, 54, 92–112.
Bućan, M., & Abel, T. (2002). The mouse: genetics meets behaviour. Nature Reviews
Genetics, 3, 114–123.
Comery, T. A., Martone, R. L., Aschmies, S., Atchison, K. P., Diamantidis, G., Gong, X.,
et al. (2005). Acute gamma-secretase inhibition improves contextual fear condition-
ing in the Tg2576 mouse model of Alzheimer’s disease. Journal of Neuroscience,
25, 8898–8902.
Dash, P. K., Houchner, B., & Kandel, E. R. (1990). Injection of the cAMP-responsive
element into the nucleus of Aplysia sensory neurons blocks long-term facilitation.
Nature, 345, 718–721.
Diagnostic and Statistical Manual of Mental Disorders (DSM-IV) (1994). Washing-
ton, D. C: American Psychiatric Association.
Dierssen, M., & Ramakers, G. J. (2006). Dendritic pathology in mental retardation:
from molecular genetics to neurobiology. Genes, Brain Behavior, 5(S2), 48–60.
Dubnau, J. (2004). Neurogenetic dissection of conditioned behavior: evolution by

analogy or homology? Journal of Neurogenetics, 17, 295–326.
Dubnau, J., & Tully, T. (1998). Gene discovery in Drosophila: new insights for
learning and memory. Annual Review of Neuroscience, 21, 407–444.
Farfel, Z., & Friedman, E. (1986). Mental deficiency in pseudohypoparathyroidism
type I is associated with ns-protein deficiency. Annals of Internal Medicine, 105,
197–199.
Ghavami, A., Hirst, W. D., & Novak, T. J. (2006). Selective phosphodiesterase (PDE)-4
inhibitors: a novel approach to treating memory deficit. Drugs R D, 7, 63–71.
Gong, B., Vitolo, O. V., Trinchese, F., Liu, S., Shelanski, M., & Arancio, O. (2004).
Persistent improvement in synaptic and cognitive functions in an Alzheimer mouse
model after rolipram treatment. Journal of Clinical Investigation, 114, 1624–1634.
Gould, T. J., Bizily, S. P., Tokarczyk, J., Kelly, M. P., Siegel, S. J., Kanes, S. J., et al.
(2004). Sensorimotor gating deficits in transgenic mice expressing a constitu-
tively active form of GS. Neuropsychopharmacology, 29, 494–501.
Gray, J. A., & Roth, B. L. (2007). Molecular targets for treating cognitive dysfunction
in schizophrenia. Schizophrenia Bulletin, 33, 100–119.
Guillin, O., Abi-Dargham, A., & Laruelle, M. (2007). Neurobiology of dopamine in
schizophrenia. International Review of Neurobiology, 78, 1–39.
Hallam, T. M., & Bourtchouladze, R. (2006). Rubinstein–Taybi syndrome: molecular
findings and therapeutic approaches to improve cognitive dysfunction. Cellular
and Molecular Life Sciences, 63, 1725–1735.
Hawkins, R. D., Kandel, E. R., & Bailey, C. H. (2006). Molecular mechanisms of
memory storage in Aplysia Biological Bulletin, 210, 174–191.
Kanes, S. J., Tokarczyk, J., Siegel, S. J., Bilker, W., Abel, T., & Kelly, M. P. (2007).
Rolipram: a specific phosphodiesterase 4 inhibitor with potential antipsychotic
activity. Neuroscience, 144, 239–246.
Kelly, M. P., Isiegas, C., Cheung, Y. F., Tokarczyk, J., Yang, X., Esposito, M. F., et al.
(2007a). Constitutive activation of galphas within forebrain neurons causes
deficits in sensorimotor gating because of PKA-dependent decreases in cAMP.
Neuropsychopharmacology, 32, 577–588.
Kelly, M. P., Vecsey, C., Stein, J., Yang, X., Tokarczyk, J., Esposito, M. F., et al. (2007b).
Regulated overexpression of a single G-protein subunit is sufficient to alter
biochemistry, neuroanatomy, and behavior. Manuscript submitted to. Journal of
Neuroscience.
Korzus, E. (2003). The relation of transcription to memory formation. Acta Biochi-
mica Polonica, 50, 775–782.
Kwok, R. P., Lundblad, J. R., Chrivia, J. C., Richards, J. P., Bachinger, H. P., Brennan,
R. G., et al. (1994). Nuclear protein CBP is a coactivator for the transcription
factor CREB. Nature, 370, 223–226.
Lattal, K. M., Barret, R. M., & Wood, M. A. (2007). Systemic or intrahippocampal
delivery of histone deacetylase inhibitors facilitates fear extinction. Behavioral
Liang, Z., Liu, F., Grundke-Iqbal, I., Iqbal, K., & Gong, C. X. (2007). Down-regulation
of cAMP-dependent protein kinase by over-activated calpain in alzheimer disease
brain. Journal of Neurochemistry, 103, 2462–2470.
Libow, L. S. (2007). Alzheimer’s disease: overview of research and clinical advances.

Managed Care Interface, 20, 20–23.
Luttrell, L. M. (2006). Transmembrane signaling by G protein-coupled receptors.
Methods in Molecular Biology, 332, 3–49.
Margulies, C., Tully, T., & Dubnau, J. (2005). Deconstructing memory in Drosophila
Current Biology, 15, R700–R713.
Maxwell, C. R., Liang, Y., Kelly, M. P., Kanes, S. J., Abel, T., & Siegel, S. J. (2006).
Mice expressing constitutively active gsa exhibit stimulus encoding deficits
Similar To those in schizophrenia patients. Neuroscience, 141, 1257–1264.
Minoretti, P., Politi, P., Coen, E., Di Vito, C., Bertona, M., Bianchi, M., et al. (2006). The
T393C polymorphism of the GNAS1 gene is associated with deficit schizophrenia in
an Italian population sample. Neuroscience Letters, 397, 159–163.
Monteleone, P., Di Lieto, A., Martiadis, V. Bartoli, L., & Maj, M. (2002). Correlations
between negative symptoms and peripheral G protein levels inmonouclear leu-
kocytes of deficit and nondeficit schizophrenics. Preliminary results. European
Archives of Psychiatry and Clinical Neuroscience, 252, 214–218.
Muly, E. C. (2002). Signal transduction abnormalities in schizophrenia: the cAMP
system. General Psychopharmacology, 36, 92–105.
Newman, M., Musgrave, F. I., & Lardelli, M. (2007). Alzheimer disease: amyloido-
genesis, the presenilins and animal models. Biochimica Et Biophysica Acta,
1772, 285–297.
Oike, Y., Hata, A., Mamiya, T., Kaname, T., Noda, Y., Suzuki, M., et al. (1999).
Truncated CBP protein leads to classical Rubinstein–Taybi syndrome phenotypes
in mice: implications for a dominant-negative mechanism. Human and Molecular
Genetics, 8, 387–396.
Phillips, R. G., & LeDoux, J. E. (1992). Differential contribution of amygdala and
hippocampus to cued and contextual fear conditioning. Behavioral Neu-
roscience, 106, 274–285.
Roberts, R. C. (2007). Disrupted in schizophrenia (DISC1): integrating clinical and
basic findings. Schizophrenia Bulletin, 33, 11–15.
Roelfsema, J. H., White, S. J., Ariyurek, Y., Bartholdi, D., Niedrist, D., Papadia, F.,
et al. (2005). Genetic heterogeneity in Rubinstein–Taybi syndrome: mutations in
both the CBP and EP300 genes cause disease. American Journal of Human
Genetics, 76, 572–580.
Rose, G., Hopper, M., De Vivo, A., Tehim, M., & A. (2005). Phosphodiesterase
inhibitors for cognitive enhancement. Current Pharmaceutical Design, 11,
3329–3334.
Ryman, D., & Lamb, B. T. (2006). Genetic and environmental modifiers of Alzheimer’s
disease phenotypes in the mouse. Current Alzheimer’s Research, 3, 465–473.
Sawa, A., & Snyder, S. H. (2002). Schizophrenia: diverse approaches to a complex
disease. Science, 296, 692–695.
Shen, J., & Kelleher, R. J. (2007). The presenilin hypothesis of Alzheimer’s disease:
evidence for a loss-of-function pathogenic mechanism. Proceedings of the
National Academy of Science USA, 104, 403–409.
Sutton, M. A., & Schuman, E. M. (2006). Dendritic protein synthesis, synaptic
plasticity, and memory. Cell, 127, 49–58.
Tecott, L. H. (2003). The genes and brains of mice and men. American Journal of
Psychiatry, 160, 646–656.
Tully, T., Bourtchouladze, R., Scott, R., & Tallman, J. (2003). Targeting the CREB
pathway for memory enhancers. Nature Reviews Drug Discovery, 2, 267–277.
Vecsey, C. G., Hawk, J. D., Lattal, K. M., Stein, J. M., Fabian, S. A., Attner, M. A., et al.
(2007). Histone deacetylase inhibitiors enhance memory and synaptic plasticity
via CREB: CBP-dependent transcriptional activation. Journal of Neuroscience,
27, 6128–6140.
Vitolo, O. V., Sant’Angelo, A., Costnazo, V., Battaglia, F., Arancio, O., & Shalenski, M.
(2002). Amyloid, 99, 13217–13221.
Weishaar, R. E., Cain, M. H., & Bristol, J. A. (1985). A new generation of phospho-
diesterase inhibitors: multiple molecular forms of phosphodiesterase and the
potential for drug selectivity. Journal of Medicinal Chemistry, 28, 537–545.
Wood, M. A., Kaplan, M. P., Park, A., Blanchard, E. J., Oliveria, A. M., Lombardi, T. L.,
et al. (2005). Transgenic mice expressing a truncated form of CREB-binding protein
(CBP) exhibit deficits in hippocampal synaptic plasticity and memory storage.
Learning Memory, 12, 111–119.
317
Index
Abstractness, 4–5 Anxiety, memory and, 51–3
Age differences in memory, 137–45 depression/anxiety comorbidity,
conditioning, 139 49–50
control processes, 144–5 Aplysia californica, 212, 244–64, 303
episodic memory, 142–4 learning, 246–64
encoding, 142–4 intermediate-term sensitization,
retrieval, 144 248–59
nonassociative learning, 138–9 long-term sensitization, 260–4
priming, 140 short-term sensitization, 247–8
procedural memory, 139–40 neuronal organization, 244–6
semantic memory, 140–1 Area X, songbirds, 290
short-term memory, 141–2 Artificial grammar learning (AGL),
working memory, 141–2 118–19
Albright hereditary osteodystrophy Associative learning, 212–14
(AHO), 307 Caenorhabditis elegans, 229, 230–1
Alzheimer’s disease (AD), 137, 301, elemental learning, 279
304–307 nonelemental learning, 279–80
cAMP/PKA pathway and, 305–307 Associative unlearning, 9–10
clinical symptoms, 304–305 Attractor network model, 83–8
AMPA receptors, 251, 253, 258 applications, 87–8
Amygdalae, 125–31 architecture, 83
Amyloid precursor protein (APP), 304, learning, 84
305, 306–307 pattern completion, 84–5
ß-amyloid protein, 304, 307 pattern generalization, 85–6
Anesthesia-resistant memory processing, 84
(ARM), 214 resistance to damage, 87
Animal model systems, 207–209 short-term memory, 87
comparative approach, 217–21, 227, similarity-based interference, 86–7
243–4, 271 Auditory memories, songbirds, 294–6
behavior homology, 220–1
brain architecture, 272–9 BAPTA calcium chelator, 251, 261–2, 263
brain homology, 218–19 Basal ganglia (BG), 169, 170
cognition, 279–83 Bee see Honeybee
functional homology, 219–20 Behavior homology, 220–1
genome homology, 217–18 Belief-generating system, 276–9
neuron homology, 220 Birds see Songbirds
history, 216–17 Brain, 216
brain functions, 216–17 comparative architecture, 272–9
importance of the brain, 216 belief-generating system, 276–89
neurons, 217 desire-generating system, 274–5
see also Specific animal models functions, 216–17
Anterior cingulate cortex homology, 218–19
(ACC), 12–13 theory of, 283–4
318 Index
Caenorhabditis elegans, 227–39 cooperative and competitive

biology, 228 activation dynamics, 81–2
genome, 228 distributed, overlapping
mechanosensory habituation, 229–35 representations, 79–80
mechanisms, 231–5 Hebbian learning, 81
memory for, 235–8 recurrent connectivity, 80–1
site of neural plasticity, 234–5 spatial constraints, 82–3
neuronal organization, 228–9 Conditioning:
types of learning, 229 age differences, 139
Category learning tasks, 119 classical (Pavlovian), 213
Caudomedial nidopallium (NCM), fear conditioning, 125, 126–7,
songbirds, 294–6, 297 305–306, 310
Cerebellum, 168 conditioned stimulus (CS), 274–5
Cerebral asymmetries, 33 defects, 303
false memory experiments, 38–9 instrumental/operant, 213–14
long repetition lag experiments, Consolidation process, 114–15
39–40 Consonant–vowel–consonant
manipulations of visual work form, trigrams, 5
37–8 Contextual variety, 65–7
patient and neuroimaging studies, Control-based learning theory
33–4 (COBALT), 103, 105, 107
visual half-field experiments, 35–6 Coordination dynamics, 190–2, 197–8
Chemosensory habituation, behavioral studies, 192–4
Caenorhabditis elegans, 229 brain activity changes, 194–7
Classical (Pavlovian) conditioning, 213 Cortex:
defects, 303 anterior cingulate cortex (ACC),
fear conditioning, 125, 126–7, 12–13
305–306, 310 distributed representations, 79–80
Cognitive control, 144–5 motor cortex, 179–80
Cognitive neuroscience, 75 overlapping representations, 80
computational, 77–93 recurrent connectivity, 80–1
attractor network model, 83–8 see also Prefrontal cortex (PFC)
principles of neural computation, CREB see Cyclic AMP response
79–83 element-binding protein (CREB)
self-organizing maps (SOMs) Cyclic AMP (cAMP), 303
model, 88–93 Cyclic AMP/protein kinase A pathway
skill acquisition, 101–108 (cAMP/PKA), 301–304
sensorimotor adaptation models, memory impairments and,
102–105 303–304, 311
sequence learning models, Alzheimer’s disease, 305–307
105–107 mental retardation, 307–308
Communication, 287 schizophrenia, 309–310
see also Songbirds Cyclic AMP response element-binding
Computational learning, 77–9 protein (CREB), 260–1, 303
principles of neural computation, CREB-binding protein (CBP)
79–83 deficiency, 307–308
Index 319
Declarative memory, 113, 114–16 Episodic memory, 114–15, 125

emotional, 127–31 age differences, 142–4
Delayed matching to sample (DMTS), encoding, 142–4
279–80 retrieval, 144
Delayed nonmatching to sample emotional facilitation, 127–8
(DNMTS), 279–80 Equilibrium point (EP) hypothesis, 156,
Dementia see Alzheimer’s 159–60
disease (AD) Event-related potentials (ERPs):
Depression, memory and, 47–9 cerebral asymmetry studies, 39–40
depression/anxiety comorbidity, selective retrieval studies, 12
49–50 Experimentation, 4
Desire-generating system, 274–5 Explicit memory, 125
Dishabituation, Aplysia, 247–8
Distinctiveness hypothesis, 24 False memory experiments, 38–9
Distributed representations, 79–80 Familiarity index (FI), songbirds, 295–6
Dopamine: Fear conditioning, 125, 126–7,
Caenorhabditis elegans habituation, 305–306, 310
233–4 Field L complex, songbirds, 294
schizophrenia and, 309 First-language attrition, 11–13
Dorsolateral prefrontal cortex Flashbulb memories, 130–1
(DLPFC), 48–9, 52 Forgetting, 7–8
motor learning and, 107, 171 retrieval-induced (RIF), 9–13
Dot-pattern classification Functional homology, 219–20
task, 119 Functional magnetic resonance
Drosophila, 214–16, 218, 301–303 imaging (fMRI):
cerebral asymmetries, 34
Earthworms, 244 emotional learning and memory, 47,
Ebbinghaus, Hermann, 3, 4, 5 127, 130, 131
Electromyogram (EMG) changes, motor learning, 104, 108, 170, 172
motor learning, 178–9, 184 selective retrieval, 12, 15
Elemental learning, 279 see also Neuroimaging studies
Emotion, 45, 125
memory and, 45–53, 125–33 Galton, Francis, 23
anxiety, 51–3 Genome homology, 217–18
declarative emotional memory, Glutamate:
127–31 Aplysia sensitization, 251, 254
depression, 47–9 Caenorhabditis elegans habituation,
depression/anxiety comorbidity, 231–3
49–50 memory and, 237–8
dimensional vs. categorical Go/No-Go tasks, 13–14
approaches, 45–7
implicit emotional memory, 126–7 Habituation, 212
working memory, 132–3 age differences, 138–9
Emotional learning, 125–7 Caenorhabditis elegans, 229–35
implicit, 126–7 chemosensory, 229
Endophenotypes, 303–304, 306 long-term memory, 236–8
320 Index
Habituation (Continued) comparative cognition, 279–83

mechanisms, 231–5 memory processing, 280–1
mechanosensory, 229–35 nonelemental forms of learning,
short-term memory, 235 279–80
site of neural plasticity, 234–5 representations, 281–3
dishabituation, Aplysia, 247–8 Instrumental conditioning, 213–14
Hebbian learning, 81, 91 Intermediate memory, 215
Hemispheric asymmetries see Cerebral Aplysia, 248–59
asymmetries processing, 281
Hippocampus, 15 Introspection, 22–3
in emotional learning and memory,
127, 129–30 James, William, 3
homology, 220
Histone deacetylase inhibitors, 310–11 Kenyon cells (KC), 276–7
Homology, 217–21 Korsakoff, Sergei, 3
behavior, 220–1
brain, 218–19 Learning, 61–3
functional, 219–20 Aplysia, 246–64
genome, 217–18 intermediate-term sensitization,
neuron, 220 248–59
Honeybee, 272–4, 283–4 long-term sensitization, 260–4
belief-generating system, 276–9 short-term sensitization and
desire-generating system, 274–5 dishabituation, 247–8
memory processing, 280–1 associative, 212–14
nonelemental forms of learning, attractor network model, 83–8
279–80 Caenorhabditis elegans, 229–35
representations, 281–3 mechanosensory habituation,
5-HT see Serotonin (5-HT), Aplysia 229–35
sensitization classification, 211–14
Huntington’s disease, 104–105 computational, 77–9
HVC telencephalic nucleus, songbirds, neural computation principles,
288–9, 293 79–83
emotional, 125–7
Implicit memory, 116, 125 implicit, 126–7
emotional, 126–7 nonassociative, 212
Inhibition: age differences, 138–9
age differences and, 141 nonelemental forms of, 279–80
in memory stopping, 13–16, 141 scheduling, 62–3
prepulse inhibition (PPI) deficits, 310 contextual variety advantages, 65–7
in selective retrieval, 8–13 presentation duration, 63–4
neural correlates, 12–13 spacing benefits, 63–5
Insect models, 271–84 self-organizing maps (SOMs) model,
comparative brain architecture, 88–93
272–9 songbird vocal learning, 288–91
belief-generating system, 276–9 motor control and constraints,
desire-generating system, 274–5 291–3
Index 321
tests as learning events, 68–9 implicit emotional memory, 126–7

total time hypothesis, 61 modulation hypothesis, 128
see also Animal model systems; impairments, 301
Motor learning cAMP/PKA pathway and,
Left hemisphere (LH), 33 303–304, 311
see also Cerebral asymmetries see also Alzheimer’s disease (AD);
Localist representations, 80 Mental retardation (MR);
Long-term depression (LTD), 238 Schizophrenia
Long-term memory (LTM), 214–15 nondeclarative, 113, 116–20
anesthesia-resistant memory see also Implicit memory
(ARM), 214 processing account, 25
Aplysia, 260–4 processing methods, 280–1
Caenorhabditis elegans, 236–8 representations, 281–3
cAMP/PKA pathway importance, selective retrieval, 7
301–303 inhibition in, 8–13
consolidation process, 114–15 self-organizing maps (SOMs) model,
emotion and, 132 88–93
processing, 281 SIMPLE model, 26–8
retrieval, 115 source memory, 64
see also Memory training programs, 144–5
Long-term potentiation variations in, 311
(LTP), 238 see also Age differences in memory;
Animal model systems; Long-
Mechanosensory habituation, term memory (LTM);
Caenorhabditis elegans, 229–35 Short-term memory (STM);
mechanisms, 231–5 Working memory
memory for, 235–8 Memory-modulation hypothesis, 47
site of neural plasticity, 234–5 Memory span, 24
Medial temporal lobe (MTL), 113, Mental retardation (MR), 301, 307–308
114–17, 119–20 cAMP/PKA pathway and, 307–308
Alzheimer’s disease and, 137 multitude of causes, 307
Memory, 113 Mid-term memory see Intermediate
attractor network model, 83–8 memory
auditory memories, songbirds, Motor control, songbirds, 291–3
294–6 Motor cortex, 179–80
Caenorhabditis elegans, 235–8 Motor learning, 101–108, 153–4, 155,
classification, 214–16 167, 177
declarative, 113, 114–16 approaches to, 159–63
emotional, 127–31 associative-stage skill learning, 169–72
emotional facilitation, 45–53, 127–8 autonomous-stage learning, 173–5
anxiety and, 51–3 behavioral studies, 192–4
declarative emotional memory, brain activity changes, 194–8
127–31 cognitive-stage skill learning, 167–9
depression and, 47–9 dynamical theory, 190–2
depression/anxiety comorbidity motor control, 155–7
and, 49–50 motor coordination, 157–9
322 Index
Motor learning (Continued) Parkinson’s disease (PD), 179, 180–1

performance changes across experi- Pavlovian (classical) conditioning, 213
mental sessions, 180–1 fear conditioning, 125, 126–7,
performance enhancement within an 305–306, 310
experimental session, 178–80 PE1 neuron, honeybee, 277–8
sensorimotor adaptation models, Phosphodiesterase (PDE), 303, 310
102–105 Positron emission tomography (PET):
sequence learning models, 105–107, 159 motor learning, 104, 168
transfer, 181–5 see also Neuroimaging studies
Motor map reorganization, 173–4 Prefrontal cortex (PFC), 46–7, 51–2, 115
Motor memory localization, 163 dorsolateral (DLPFC), 48–9, 52
Motor redundancy, 157 motor learning and, 107, 171
Mouse model, 303–304 ventrolateral (VLPFC), 12
Alzheimer’s disease, 304, 305–307 motor learning and, 171–2
mental retardation, 308 see also Cortex
schizophrenia, 310 Prepulse inhibition (PPI) deficits, 310
Mushroom body (MB), 272, 274, 276–9 Presenilins, 304, 305
Priming, 116–17, 119–20
NCM (caudomedial nidopallium), age differences, 140
songbirds, 294–6, 297 Principal component (PC) analysis, 184
Negative-patterning discrimination, 279 Principle of abundance, 157
Neural cooperation and competition, Procedural memory, age differences,
81–2 139–40
Neural plasticity: Protein kinase C (PKC), 249, 250, 258
Aplysia sensitization, 247 Protein kinase M (PKM), 249, 258
Caenorhabditis elegans habituation, Protein kinase A (PKA), 248, 249, 303
234–5 Alzheimer’s disease and, 304–305
Neuroimaging studies: see also Cyclic AMP/protein kinase A
cerebral asymmetries, 33–4 pathway (cAMP/PKA)
emotional learning and memory, 127, Protein synthesis, Aplysia sensitization,
128–30 254–8, 262
motor learning, 168–9, 170–2, 194–7
selective retrieval, 12, 15 Quantitative standards, 4
see also Specific imaging techniques
Neurons, 217, 272 Recognition, 144
homology, 220 word recognition, 37–8
Nipher, Francis, 3, 4 Reconsolidation, 238
NMDA receptors, 251, 253 Recurrent connectivity, 80–1
Nonassociative learning, 212 Reflex behavior, 212
age differences, 138–9 Aplysia withdrawal reflex, 247
Nondeclarative memory, 113, 116–20 Reinforcement learning, 78, 79
Nonelemental learning, 279–80 Repetition suppression effect, 117
Representations, 281–3
Octopamine, 274 distributed, 79–80
Operant conditioning, 213–14 localist, 80
Overlapping representations, 80 overlapping, 80
Index 323
Retrieval-induced forgetting (RIF), Short-term memory (STM), 214

9–13 age differences, 141–2
Retrograde amnesia, 114 Caenorhabditis elegans, 235
Ribot, Theodore, 3 capacity limits, 23–4
Right hemisphere (RH), 33 creation of, 21–2
see also Cerebral asymmetries distinctiveness hypothesis, 24
Robust nucleus of the arcopallium processing, 280
(RA), songbirds, 290, 296 see also Memory
Rolipram, 303, 305, 306–307, 310 Similarity index (SI), songbirds, 295
Rubinstein–Taybi syndrome (RTS), SIMPLE model, 26–8
307–308 Skill acquisition, 101–108, 189–90
Rule-based learning, 279–80 sensorimotor adaptation models,
102–105
Scheduling, 62–3 sequence learning models, 105–107
Schizophrenia, 301, 309–310 see also Motor learning
cAMP/PKA pathway ‘‘Skinner Box’’, 213–14
and, 309–310 Songbirds, 287–97
symptoms, 309 specialized area for auditory
Selective retrieval, 7 processing and memory, 294–6
inhibition in, 8–13 vocal learning process, 288–91
neural correlates, 12–13 motor control and constraints,
Self-organizing maps (SOMs) model, 291–3
88–93 Source memory, 64
applications, 92–3 Spacing, 63–5, 214–15
architecture, 89 Spatial constraints, 82
clustering via emergent Speech acquisition, 288
topography, 92 see also Songbirds
learning, 90 Spike duration independent (SDI)
pattern generalization, 91 process, 248
processing, 90 Supervised learning, 78, 79
resistance to damage, 92 Synchronization, 196
similarity-based interference, 91–2 Syncopation, 196
Semantic memory, age differences, Synergies in motor learning, 160–3
140–1
Sensitization, 212 Tap withdrawal response, Caenorhab-
Aplysia, 247–64 ditis elegans:
intermediate-term, 248–59 habituation, 230–5
long-term, 260–4 memory, 235–8
short-term, 247–8 Telencephalon, songbirds, 288–9,
Sensorimotor adaptation models, 293, 294
102–105 Testing effect, 68–9
Sensorin, 262–3 Think/No-Think (TNT) paradigm, 14–16
Sequence learning models, 105–107 Thorndike, Edward, 3
Serial reaction time (SRT) task, 118 Thorndike’s puzzle boxes, 213
Serotonin (5-HT), Aplysia sensitiza- Topography, 88
tion, 247–8, 250–3, 254–8, 260, 262 Total time hypothesis, 61
324 Index
Transfer of training, 181–5 Voluntary movements see Motor

Trichostatin A (TSA), 311 learning
Unconditioned stimulus (US), 274–5 Withdrawal reflex, Aplysia,

Uncontrolled manifold (ICM) 247–64
hypothesis, 157–9 intermediate-term sensitization,
Unsupervised learning, 78–9 248–59
Urbach–Wiethe disease, 127, 128 long-term sensitization, 260–4
short-term sensitization and
Ventral unpaired median (VUM) dishabituation, 247–8
neurons, 273–5 Word recognition, 37–8
Ventrolateral prefrontal cortex Working memory, 22, 125
(VLPFC), 12 age differences, 141–2
motor learning and, 171–2 emotion and, 132–3
Visual field (VF) asymmetries, 35–6
Vocal communication, 287 Zebra finch, 290, 291
see also Songbirds see also Songbirds

Aaron S. Benjamin, J. Steven de Belle, Bruce Etny

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Aaron S. Benjamin, J. Steven de Belle, Bruce Etny

Uploaded by

Copyright:

Available Formats

HUMAN LEARNING

Biology, Brain, and Neuroscience

Biology, Brain, and Neuroscience

Amsterdam • Boston • Heidelberg • London • New York • Oxford • Paris

First edition 2008

Copyright 2008 Elsevier Ltd. All rights reserved

No part of this publication may be reproduced, stored in a retrieval system

British Library Cataloguing in Publication Data

Library of Congress Cataloging-in-Publication Data

For information on all North-Holland publications

Printed and bound in Great Britain

Working together to grow

Part I: Human Learning and Cognition

Introduction: Behavioral Approaches to the Study of Human Learning

The Role of Inhibition in Learning 7

Short- vs. Long-Term Memory 21

Hemispheric Asymmetries in Verbal Memory 33

Emotional Facilitation and Disruption of Memory 45

Scheduling and Learning 61

Part II: Cognitive Neuroscience

Introduction: Cognitive Neuroscience of Learning and Memory 75

The Computational Cognitive Neuroscience of Learning

Cognitive Neuroscience of Skill Acquisition 101

Cognitive Neuroscience of Declarative and Nondeclarative Memory 113

Learning and Memory for Emotional Events 125

Age Differences in Memory: Demands on Cognitive Control

Part III: Human Motor Learning

Introduction: A Survey of Motor Learning Concepts and Findings 153

Two Aspects of Motor Learning: Learning Movements

Neuroanatomical Correlates of Motor Skill Learning: Inferences

Mechanisms Underlying Short-Term Motor Learning,

A Dynamical Framework for Human Skill Learning 189

Part IV: Animal Model Systems

Animal Models of Behavioral and Neural Plasticity 207

Caenorhabditis elegans as a Model System in Which to Study

The Cell Biology of Learning and Memory in Aplysia 243

Insect Minds For Human Minds 271

Patterns of Learning, Memory, and Vocal Production

The cAMP/PKA Pathway and the Modeling of Human

Ted Abel (301) Department of Biology, University of Pennsylvania, 204G

Daivd L. Glanzman (243) and Neurobiology, and the Brain Research

Paul J. Reber (113) Psychology Department, Northwestern University, 309

Human Learning and Cognition

Behavioral Approaches to the Study of Human

The systematic investigation of the ability of humans to learn and

some of the current topics that arise in the experimental treatment of

The principal tool for investigating learning and memory in normal

Today, behavioral research on learning and memory is one of the most

of the field. The advantages of such stimuli were well articulated by

The nonsense material . . . offers many advantages, in part because of this

By using such stimuli, Ebbinghaus hoped to reduce the problems of

Ebbinghaus. They review the current state of knowledge on hemispheric

Ebbinghaus, H. (1885/1913/1964). Über das Gedächtnis [(Trans.: On Memory).

The Role of Inhibition in Learning

1. Inhibition in selective retrieval

Memory retrieval is not always as simple as following a straight path

information, unencumbered by intrusions from outmoded or erroneous

Both academic and research settings abound with circumstances that

acquisition, thereby facilitating retrieval of the weaker, foreign language

2. Inhibition in memory stopping

Not only is inhibition likely recruited in situations that demand the

unpleasant memory or are merely in need of pushing distracting thoughts

On one hand, learning, by definition, involves the acquisition of knowl-