Professional Documents
Culture Documents
IN
PSYCHOLOGY
50
Editors:
G. E. STELMACH
P. A . VROON
NORTH-HOLLAND
AMSTERDAM. NEW YORK . OXFORD .TOKYO
COMPLEX M O V E ~ N BEHAWOUR
T
'The'motor-actioncontroversy
Edited by
Onno G .MEIJER
Faculty of Movement Sciences
Free University
Amsterdam, The Netherlands
and
Klaus ROTH
Faculty of Psychology and Sports Science
University of Bielefeld
Bielefeld, F: R.G.
<
1988
NORTH-HOLLAND
AMSTERDAM. NEW YORK .OXFORD .TOKYO
@ELSEVIER SCIENCE PUBLISHERS B.V.. 1988
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 permission of the copyright owner.
Pichlish ers:
ELSEVIER SCIENCE PUBLISHERS B.V.
P.O. Box 1991
1000 BZ Amsterdam
The Netherlands
LIBRARY a CONOAESS
Library o f Congress Cataloging-in-Publication Data
This Volume was written in New Century Schoolbookm (text) and New
Helvetica Narrowm (illustrations), on a Macintosh SEW, using
Fontasticm, Fontographerm, Illustratorm, MacDraW, MacPaintm,
PageMakerm, Spellerm, Thunderscanm, and WordTM. The files were
printed on a Laserwriter Plusm before photographic reproduction. The
Editors want to acknowledge Mr. N. Steenhuis, Mr. J. Steenhuis, and Mr.
E. Bindels (Compudress B.V., Kamerik, The Netherlands), and Mr. E.
Grijsen (Free University, Amsterdam) for their advice and help in the
technical production of the text.
vii
Foreword
For the workshop, there were two main precipitating factors. In the first place, a
study year on the topic Perception and Action at the Centre for Interdisciplinary
Research (ZIF) at the University of Bielefeld had been planned to take place
between September 1984 and August 1985. Many interesting publications from
that enterprise have since been committed to print. While the Symposia scheduled
for that event proved attractive to members of the Faculty of Psychology and Sports
Science at the University of Bielefeld - particularly since it was taking place on
their doorstep, so to speak - members of the Sports Science Department (headed by
Klaus Willimczik) had some misgivings, misgivings which were later confirmed
by their attendance at some of the sessions. After all, part of their departmental
brief was to examine the findings of basic research in psychology with respect to
its validity for complex motor behaviour in sports - a topic which was far from
being a direct concern of those contributing to the various symposia on perception
and action.
What did become apparent to members of the Department of Sports Science,
however, from these symposia, was the existence of two markedly different
approaches to the study of skilled behaviour. Approaches which have since come
to be designated the 'motor systems approach' (aligned to an information-
processing perspective) and the 'action systems approach (stimulated on the one
hand by Bernstein's work on the coordination and regulation of movements, and
on the other by Gibson's ecological theory of perception).
At the same time as these events were taking place in the Federal Republic of
Germany, a similar interest in exploring the differences between these two
standpoints and their relevance to an understanding of complex movement
behaviour was already well-established in the Department of Psychology (headed
by John Whiting), and the Department for Theory and History, in the Faculty of
Human Movement Sciences a t the Free University in Amsterdam, The
Netherlands.
...
Vlll Compler Movement Behaviour
While the dialogue between the undersigned that led eventually to the
establishment of the workshop was in some ways fortuitous, in others it was the
more direct consequence of the stimulation provided by the study year on
Perception and Action and the questions to which the content of that programme
gave rise. That this should result in a proposal to explore these dichotomous
approaches to the understanding of skill in a more formal way in the context of
sports and everyday life is, given the circumstances, not surprising. Having
generated the idea it was a relatively simple step to operationalize the proposal.
The presence of Klaus Willimczik in the university of Bielefeld and the concept we
were able to present to the Centre of Interdisciplinary Research provided an
attractive proposition to the Director of the Centre and the rest followed as a
matter of course. This is, perhaps, a suitable place to express our gratitude and
that of all those members who attended the workshop, to the Director of the Centre
for his support and sponsorship. We could hardly have chosen a better
environment or venue for the Workshop on Complex Motor Behaviour - Control
and Learning in Sports and Everyday Life, November 5-8,1985.
Introduction
"the theory of action systems really does allow us to study the skills of everyday life
in a way that is more fruitfbl than traditional motor skills approaches" (p. 82).
Van Wieringen sees the difference as one of kinds of analysis - motor
approaches being explanatory, action approaches descriptive - as well as levels of
analysis - motor approaches being concerned with the level of 'mechanisms'
inside the organism, action approaches with 'molar functional behaviour'. His
paper, 'unmasking' Gibson as a Skinnerian, also serves as an introduction to the
General Discussion which we have tried to present verbatim. As such, the
General Discussion offers interesting material for the study of communication
between scientists: every second page or so there appears to be general agreement
which then explodes into strong opposition and abhorrence. Beek and Meijer,
finally, conclude that no logical demarcation between motor and action
approaches is presently tenable, and present a developmental view: the two sides
have different sources of inspiration.
Reinoud Bootsma and Piet van Wieringen (Chapter 6) analyse aspects of 'visual
control' in table tennis and stress that such an analysis does not a priori preclude
a motor systems approach.
Timothy Lee (Chapter 7) discusses 'transfer-appropriate processing' and
concludes that current interpretations of both Knowledge of Results (KR) and
'variability of practice' data may be wrong.
Zntroduct ion xiii
In Chapter 8 Karl-Heinz Leist presents a framework for the study of 'action' and
investigates how learning to ride the pedalo is represented in the self-perception of
the actor.
Richard Magill describes, in Chapter 9, how 'forgetting' during the post-KR
interval can benefit learning, and arrives at a Bernsteinian interpretation of
learning how to solve 'motor problems'.
Theo Mulder and Wouter Hulstijn (Chapter 10) stress that rehabilitation should
be directed at 'goal-directed actions' rather than 'isolated movements' and claim
that EMG-feedback could be useful in this respect.
In Chapter 11 Klaus Roth claims to have corroborated the 'motor' hypothesis
concerning 'invariance of relative timing', although, at first sight, the majority of
his test persons disagreed.
John B. Shea and Susan T. Zimny state, in Chapter 12, that their model and
data concerning 'knowledge incorporation' in learning, should be of relevance to
both 'motor' and 'action' approaches.
Stephan Swinnen (Chapter 13) remains unambiguously 'motor' in suggesting,
empirically and theoretically, that the until now underestimated KR delay is the
interval which subjects need to actively evaluate their own response.
William Warren remains, in Chapter 14, dedicated to 'action' and needs 13
pages of studying 'insect flight' before it becomes ecologically understandable how
human can climb stairs.
Finally, John Whiting (Chapter 15) emphasizes that 'imitation' may not be
under control of internal representations, although a topological 'image of the act'
plays a role when an external model regulates, rather than produces, movement.
It is, of course, for the reader to decide if we have succeeded in editing a Volume
on Complex Movement Behavwur: 'The' nwtor-actwn controversy which may
help to clarify important issues in the field. On our part, we found it a rewarding
enterprise.
References
Frbse, M. & Sabini, J. (1985). Goal Directed Behavior: The concept of actwn in
psychology. Hillsdale, N.J.: Erlbaum.
Gibson, J.J. (1979). The Ecological A p p m h to Visual Perception. Boston:
Houghton-Mifflin.
Reed, E.S. (1982). An outline of a theory of action systems. Journal of Motor
Behavior, 14, 98-134.
Requin, J., Semjen, A. & Bonnet, M. (1984).Bernstein's purposeful brain. In
H.T.A. Whiting (Ed.), Human Motor Actions: Bernstein reassessed (pp.
467-504).Amsterdam: North-Holland.
Schmidt, R.A. (1982). Motor Control and Learning: A behavioral emphasis.
Champaign, n:Human Kinetics.
Turvey, M.T. (1977).Preliminaries to a theory of action with reference to vision. In
R. Shaw & J. Bransford (Eds.), Perceiving, Acting, and Knowing: Toward
an ecological psychology (pp. 211-265).Hillsdale, N.J.: Erlbaum.
Whiting, H.T.A. (1984, Ed.), Human Motor Actions: Bernstein reassessed.
Amsterdam: North-Holland.
Editing this Volume was something of an adventure right from the start. Each
manuscript was sent to two independent referees - we want to thank them all. To
our surprise, strongly conflicting referee opinion was no exception - we want to
thank the authors for their willingness to enter correspondence in order to ensure
the quality of our book. The Publishers had to bear with our perfectionism and
they did so with grace.
Upon deciding to have English English as the language of our book, we went,
innocently, to Blackwell's in Oxford, and ordered a General Dictionary, a
Dictionary on Word Division, a Guide for Editors, and a Manual for Idiom,
preferably all in the same spelling. It appeared, however, that no such set could
be obtained and so we had to live with dictionaries having different spellings.
John Whiting was willing to provide linguistic advice and read all the Chapters,
and Peter Beek's close reading helped us to remove further inconsistencies. We
thank them both. Fons Blankendaal, Audrey van der Meer and Ruud van der
Weel, in Amsterdam, and Rainer Wollny, in Bielefeld, acted as our assistants;
they did a splendid job which we are unlikely to ever forget.
The Publisher informed us that a Laser Writer print would be best suited. Never
having heard of the contraption, we found out that it had to do with computers
and decided to buy word processing equipment. 'User friendliness' was promised
but it took us twelve days to produce an x with a dot and be proudly able to print i .
That our little Turing machine had a mind of its own became apparent when it
suggested to change the misspelled name 'Schmid into 'Schoolmaster'. We
finally decided to give the Chapters as much of a unified look as possible and to
computerize the Figures. It will be apparent to the reader that we are no artists,
but then ...it is only a book.
List of Referees
Editing this Volume was something of an adventure right from the start. Each
manuscript was sent to two independent referees - we want to thank them all. To
our surprise, strongly conflicting referee opinion was no exception - we want to
thank the authors for their willingness to enter correspondence in order to ensure
the quality of our book. The Publishers had to bear with our perfectionism and
they did so with grace.
Upon deciding to have English English as the language of our book, we went,
innocently, to Blackwell's in Oxford, and ordered a General Dictionary, a
Dictionary on Word Division, a Guide for Editors, and a Manual for Idiom,
preferably all in the same spelling. It appeared, however, that no such set could
be obtained and so we had to live with dictionaries having different spellings.
John Whiting was willing to provide linguistic advice and read all the Chapters,
and Peter Beek's close reading helped us to remove further inconsistencies. We
thank them both. Fons Blankendaal, Audrey van der Meer and Ruud van der
Weel, in Amsterdam, and Rainer Wollny, in Bielefeld, acted as our assistants;
they did a splendid job which we are unlikely to ever forget.
The Publisher informed us that a Laser Writer print would be best suited. Never
having heard of the contraption, we found out that it had to do with computers
and decided to buy word processing equipment. 'User friendliness' was promised
but it took us twelve days to produce an x with a dot and be proudly able to print i .
That our little Turing machine had a mind of its own became apparent when it
suggested to change the misspelled name 'Schmid into 'Schoolmaster'. We
finally decided to give the Chapters as much of a unified look as possible and to
computerize the Figures. It will be apparent to the reader that we are no artists,
but then ...it is only a book.
List of Referees
Chapter 1
Richard A. Schmidt
SUMMARY
reflexes and the control of very simple motor activities, but with almost no
emphasis on the skilled motor behaviour that was presumably the product of the
processes under examination. A second perspective involved scientists in
experimental and applied psychology, physical education, and the
SportwissenschafCen, with a major concern for simple o r complex movement
skills and how they are influenced by factors such as environmental information
and practice. Information processing, particularly as it has been popularized in
the cognitive psychological approach, has been an important feature of this work
recently.
However, there was little or no concern for the underlying physiological mechan-
isms. Recently, we have seen these two approaches come together, to the point
where it is nearly the norm to find measures of neurophysiological activities (e.g.,
EMGs), of kinematics, and of overt skilled movement behaviour being discussed a t
the same conferences and seen in the same journal articles. This new linkage
has been dominant in dealing with problems of movement behaviour during the
late 1970s and early 1980s (e.g., Schmidt, 1982).
But just when we began to feel comfortable with this new synthesis, another
approach emerged which has challenged these ways of thinking. This approach,
originally based on ideas about perception and action from the American psycho-
logist J.J. Gibson (19661,has been continued after Gibson's death by the 'neo-
Gibsonians' (e.g., Turvey, Shaw & Mace, 1978; see Michaels & Carello, 1981, for
an interesting review), and has come to be known as ecological psychology. In a
very influential paper, b e y (1977) extended these ideas to emphasize motor
control, including notions from the Russian physiologist N.A. Bernstein (1967)
and his many successors in the Moscow laboratories. This synthesis is sometimes
referred to as an 'action' view (Fowler et al, 1980;Kugler, Kelso & Turvey, 1980 &
1982;Reed,1982).
Ecological psychology, and the subset which deals with movement control, is
considerably different fmm earlier perspectives, with different assumptions,
methods of analysis, types of experiments, and even different research questions.
It is very broad and comprehensive, attempting to deal in a consistent way with
such diverse issues as perception, motor control, ecological and evolutionary per-
spectives, growth and development, and numerous other (usually distinct) areas,
only a small part of which is immediately relevant to movement control. For these
reasons, it would be presumptuous of me to attempt a critique of all of ecological
psychology, and I do not attempt it here. Rather, this Chapter will be limited to
those aspects of the action view which I consider having the strongest relevance to
motor functioning,and I will leave the remaining issues to others.
After a brief discussion of the main features of the action view, I focus on a
Action and Motor Views 5
-
2. The action perspective a brief description
The action perspective is not really a single view, but a variety of separate
statements with a number of strongly similar features. This, plus the fact that
these viewpoints deal with many separate aspects of biological actions, makes
them particularly difficult to summarize in a brief, yet fair, way. In the Sections
which follow, those aspects of the viewpoints which seem to have the most rele-
vance for the processes underlying the production of human movements are
summarized. For a more complete treatment, the reader should consult the
original sources (e.g., Kugler et al, 1980 & 1982;Kugler & T w e y , 1986;Reed, 1982
& 1988;Turvey, 1977).
2.1. Representatwns
Certainly one of the most important ideas to come from Bernstein's (1967;
Whiting, 1984) writings, and one which has been a cornerstone of the action view
(and often of the motor view, it must be added), is the idea that the motor system is
comprised of huge numbers of degrees of freedom which must somehow be
controlled to produce coordinated and skilled actions. This problem may be viewed
in various ways, such as counting the number of degrees of freedom in the
particular joints (the elbow joint has one, the wrist two, etc.), by counting the
number of muscles which act a t the various joints (the elbow has three muscles
which control flexion alone), or even by considering the massive number of single
motor units which must ultimately be organized in skilled actions (a given
muscle may have several thousand motor units).
Coupled with the fact of having many degrees of freedom to control (however one
chooses to count them), one concern for Bernstein was how the animal can
control the degrees of freedom with minimal intelligence, so that a minimum
amount of computation by the brain is required for coordination. Both the action
and motor views search for ways that the various degrees of freedom are
constrnined, or coupled to one another, so that the control of a number of degrees
of freedom is handled as a single unit, itself with only one degree of freedom. For
the action side, a group of muscles and joints acting together as a unit is termed a
synergy, or a coordinative structure. The motor view has the motor programme
which supposedly links the various degrees of freedom together as a single unit,
the single programmed action. Indeed, I am not even sure that, at a behavioural
level of analysis, one can meaningfully distinguish between coordinative
structures and motor programmes.
However, the action and motor approaches differ considerably with respect to
how they see the various degrees of freedom controlled. For the action proponents
(see Kugler et al, 1980 & 19821,the notion that some movemerit programme stored
in the CNS is responsible for the controlled actions has been particularly
distasteful. They argue that the embodiment of the details of the actions in a
programme or other central representation simply pushes the problem of
coordination deeper into the system, resulting in a logically infinite regress. That
is, if details of, say, the timing of an action are controlled by a programme, then
Action and Motor Views 7
how is the programme structured, 'who' structured it, and what are the rules for
its formation? Thus, to the action view, each additional specification of the action
in the CNS 'takes a loan on intelligence,' in that this represents behaviour (at the
CNS level) which ultimately must be explained. Their strategy, about which I will
have more to say in a subsequent Section, is to account for the order and
regularity in the actions by use of the dynamical properties of the moving systems,
rather than to have them as part of the movement programme structured in the
CNS.
.g
c
1
c
?
0
Muscle Length (cm)
150 120° 90 O 60 30
Angle of Ankle
There are actually two different equilibrium-point hypotheses which differ in terms of what the
motor system is thought to control. In the Fel'dman (1966, 1974a & 1986) view, the motor system
controls the relationship between muscle tension and muscle length, produced via mechanical
characteristics of the contracting muscle and the actions of the muscle spindle and the gamma
system; this relationship is termed the invariant characteristic (IC). In a second view (Polit &
Bizzi, 1978), the motor system controls only the central drive to the muscles, with no involvement of
the gamma system being assumed. At present, Fel'dman's view appears to be the more viable
(Hasan & Enoka, 1985;Vincken, Gielen & Denier van der Gon, 1983).
10 R.A. Schmidt
Houk BE Rymer, 1981)used this idea to explain how the limb could be positioned to
a new location at the end of a movement, with different positions being produced
by specifying different slopes (i.e., stifmss and intercepts of the muscles' length-
tension h c t i o n s (see also Partridge, 1983). With these models, the endpoint can
be reached from a variety of starting positions, under certain external perturba-
tions (see Schmidt, 1980;Schmidt & McGown, 19801,and a t a variety of speeds.
Thus, many of the details of the movements (e.g., the maximum velocity, the
position at which maximum velocity is reached, the time to peak acceleration,
etc.) could in principle be accounted for by dynamical principles of masses and
non-linear springs, and thus one might not have to assume that these features
are controlled directly by the movement programme. This is, in my view, the most
important single contribution of this concept to motor control.
et al, 1980; Kelso C TuUer, in press). With this as a working hypothesis, the Kelso
group has been involved in complex dynamical modeling of (chiefly rhythmic)
behaviours, and they can account for an impressive number of features of the
timing and coordination found without having to assume that the control was
represented centrally. Such a viewpoint is critically opposed to the notion of motor
programmes as I (Schmidt, 1982) and many others have thought of them; I
present my objections to this particular aspect of the action view in a later Section
(see also Schmidt, 1987a).
Next, an important emphasis of the action views has been to remind us that
motor functioning evolved so that the animal could cope with the environment in
moving around, acquiring food, escaping predators, and the like. In this sense,
the studies completed by many of us under an earlier motor tradition, in which
the environments are very unstructured, constant, and controlled, are not
particularly satisfying to those who see the actor's main problem as responding to
changing environmental situations. An important focus in the action view is how
the animal detects the information from the environment relevant to its actions,
what forms this information might take, and how it is used to modify and control
the actor's movements. This emphasis is a pleasant change from the older views
which either did not involve environmental information a t all, or which viewed its
reception by the organism as passive, with information processing activities then
later used to decode it (or make sense of it). Ecological views regard information
pickup as direct (after Gibson, 1966), and not requiring elaborate information
processing to detect its implications for action. Thus, principles learned from
experiments in artificial and constant environments may be at worst
fundamentally different from those involved in real animal-environment
situations, or, at best, may simply be irrelevant to them.
more strongly in the on-going control of movements, arguing that the pickup and
use of external information constrains the nervous system to act in certain ways.
What the muscles receive as activation is modified by the sensory information
being received, which in turn modifies the action, which in turn modifies the
sensory information produced, and so on continuously. This particular point, as
we shall see in a later Section, is a strong issue for disagreement between the two
approaches.
A related issue, brought out well by Reed (1982),is that the motor approach has
tended to separate sensory and motor hnctions arbitrarily. He points out that it is
reasonable to talk of efferent (outgoing) and a e r e n t (incoming) information, but
claiming that the one is strictly motor and the other strictly sensory is mis-
leading. For example, 'motor' processes are supposed to cause movement; yet,
when one stretches a muscle spindle, an afferent ('sensory') pathway is activated,
which ultimately causes the stretched muscle to contract. This leads to the contra-
diction that sensory processes cause movements. Under the action approach,
neither efferent nor afferent pathways cause movements, but it is their mutual
interaction which is ultimately responsbile.
2.7. Summary
The action perspective is, as been pointed out by Reed (1982) and Kugler et al
(1980 & 1982; Kelso et al, 1980), a very different one from the earlier motor
perspective, and it appears to provide solutions to a number of problems which
have faced the motor control field for many decades. But despite the various
authors' elegant arguments, in regard to the implications of this view for
movement contml, there exist a number of concerns, some difliculties, and even
some places where the perspective appears to be incorrect, as we shall see in the
next Sections.
-
3. The action perspective some criticismsand concerns
A first concern is related to the notion that the action and motor perspectives are
in some way competitors - points of view to be pitted against each other in order to
determine which of them turns out to be correct (or more correct). Rather, I have
come to realize lately that, in many if not most ways, these perspectives are really
Action and Motor Views 13
about different things, and thus comparing them may not be particularly
meaningful.
This idea became clear to me in 1984 a t a conference in Bielefeld, Federal
Republic of Germany, during the discussion of a presentation that Mike T w e y
had just made. Larry Stark, whose perspective about movement control employs
an engineering-systems approach, made essentially the following point to T w e y .
In the ecological (or action) approaches, the goal is an understanding of the
mutual relationship between the environmental information and the motor
behaviour that an animal displays; the focus is a t the level of the animal-
environment interface. It is almost like the old 'black box' approach to
understanding behaviour (Stark continued to say), in which input-output
relations are the primary focus, and mechanisms occurring inside the 'box' are
almost never addressed.
Contrast this with Stark's perspective, or any of a number of other motor
viewpoints. Figure 3 contains a flow diagram representing a tracking model from
a paper by Pew (19741, which is typical in many ways of the kinds of engineering-
systems modelling that Stark was discussing. Notice that there is an input signal
from the display indicating shifts in the target pathway, and there is a response
in the form of movements of a control stick which tends to make the hand follow
the track after a delay. But also notice that, between the two there are numerous
intervening hypothetical processes proposed (short-term predictor, motor
command generator), all of which together provide one kind of explanation about
how such hand movements are controlled. In this approach, the focus is on
processes which occur within the model of the human (the box), and experiments
are conducted which presumably bear upon whether or not a particular structure
will account for the behaviour.
My point here is neither to endorse nor criticize engineering-systems models,
but merely to highlight the differences between the action and motor perspectives
in terms of what is the target of explanation. In the action view, the explanation is
in terms of what the motor view would term 'input-output' relations, while in the
motor view it is usually in terms of mechanisms of motor control. More
fundamentally, the difference seems to be in terms of the level of analysis at
which the theorizing occurs. Almost any phenomenon (e.g., Autobahn driving
behaviour) can be analysed at a variety of levels of analysis, ranging h m global
patterns of traffic congestion to biochemical activities in the retina. Typically,
workers a t a given level of analysis tend to regard the immediately 'lower' level as
a 'black box,' where the actual mechanisms are the target of investigation, but are
not actually measured directly. Of course, any level of analysis is appropriate,
depending on what one wants to understand, and it is difficult to say with
14 R.A. Schmidt
confidence that one would ever be 'better' than the other. So, in considering action
versus motor approaches, I am not certain that it even makes sense to draw the
comparison.
Parameter
Signal Comparator
L, Signal
Comparator Predictor -and Hold - Delay
- Short Term Sample
I
Figure 3. Block diagram of a feedback model for a subject peflonning a
compensatory tracking task (adapted from Pew, 1974, p. 19).
However, the argument that the two approaches are operating at completely
different levels of analysis is probably too strong. In fact, some of Kelso's (Kelrw, et
al, 1980;Saltzman & Kelso, 1987)recent work on limb and epeech control (Kelso &
Tuller, 1985)does address the question of mechanisms of motor control. But, as I
will argue in a later Section, the models do not appear to be easily testable,
particularly when contrasted to those proposed by workers in the motor tradition;
and, the evidence for theee mechanisms is equivocal, and can be easily explained
by other, more motor, viewpoints (as Kelso & Tuller, 1985,freely admit). This does
not invalidate their views, of course, but rather weaken8 the impact of the
evidence they cite in favour of their interpretation.
With this distinction (at least in a weak form) in mind, I re-read many of the
papers by the action proponents in preparation for writing this Chapter. My
response is quite positive with respect to a need for a shifting focus in order to
understand the environmental determinants of behaviour, the mutual
interactions of action and perception, and the environmental influences on motor
Action and Motor Views 15
control. But, at the same time, I am compelled to ask, "So, how then do we
actually move ?" I do not get a satisfadory answer from action perspectives,
perhaps because these perspectives do not address this question, at least in the
sense that is implied in Figure 3. Later, I focus on the motor approach to the
problem of how we move, stressing how its proponents have asked questions about
the underlying mechanisms of movement control. But, these questiondanswers
are not very strongly concerned with the animal-environment interface of interest
to the action view; this makes sense if it is mainly two different things that these
approaches are all about.
Recently. in an effort to provide evidence for the claim (Kelso et al, 1980) that
complex motor control processes could be understood by a dynamical systems
approach (and without the assumptions common to a motor programming
approach), Kelso and colleagues have done some elegant modelling of speech
(e.g., Kelso & Tuller, 1985; Kelso et al, 1985) and simple limb movement tasks
(e.g., Saltzman & Kelso, 1987;Kelso t Kay, 1987). In an example from Kelso and
Kay's work on limb control, subjects moved a manipulandum in time to a
metronome using a wrist motion in the horizontal plane. A strong inverse
relationship was found between the subjects' cycling frequency and movement
amplitude, a common finding in movement research. Such relationships are also
frequently found in non-linear dynamical models, called limit-cycle oscillators,
which describe the behaviour of oscillating mass-spring systems that maintain
their amplitude and frequency for relatively long periods of time without 'running
down'.
Kelso and Kay described one such non-linear model that accounts for their
experimental data,seen here as Equation 1:
2 This model may not be so simple after all, as it capitalizes on the large number of ways in which
the various terms can be combined in the equations. And, in the speech modelling, the stiffness
term in the equation has to be changed two times per cycle to account for the data (Kelao et al, 1986).
Action and Motor Views 17
Kelso and Kay (1987) probably chose to characterize their data with this
particular model, in part, because it is a member of a class of models which does
not require central control of timing information to the musculature. Here, the
temporal structure is an emergent property of the dynamics, and does not have to
be controlled directl?. But, as they freely admit, these data can also be described
by a number of other models, one of which is the forced Duffing equation:
nG + b i + k x + l x = fcos (wt)
where 'm', 'b', 'k,'l', and 'w' are constants, and 't' is time. Here, the term at the
right provides an additional input to the system as a function of time, and is more
or less similar to the kinds of ideas specified by the motor programming
viewpoints. The choice of such models is often arbitrary; selecting, say, Equation 1
neither indicates that it is correct, nor provides any assurance that other models
(e.g., Equation 2) may be incorrect.
Kelso and Kay also go on to say that discriminating between these two kinds of
models will be dimcult. Here I disagree strongly. Numerous experiments have
been published which seem to provide a clear separation between their various
predictions. For example, Conrad and Brooks (1974) had monkeys make
oscillating movements of the forearm while grasping a lever, with the
'instruction' to move as rapidly as possible between two stops - a task quite similar
to that used by Kelso and Kay. As one would expect, when the movements were
practiced, the monkeys slammed the lever rapidly back and forth, timing the
bursts of EMG and force such that the movement to the left (right) would be
initiated just as the right (left) stop was struck. Then, without the subject being
able to predict it, the stops were suddenly moved closer together, so that
movements with smaller amplitudes were required. How should this change
alter the monkey's motor control?
The two models have different things to predict here. The dynamical models
without an external timing control (e.g., Equation 1)would predict that, since the
position and velocity of the limb are radically affected a t the new stop position,
there should be marked changes in the timing of the EMGs and forces. On the
other hand, the models with central timing would expect that, since there would
be a considerable interval between the actual change in the stop position until the
motor programming centres are informed of this via feedback, the centrally
Jennings and Nelson (1979) argue similarly, when they say that 'Time'll take care of itself, 80
just leave time alone."
18 R.A. Schmidt
time, per se, cannot be centrally represented, because time is only realizable
through the corresponding changes in other variables, such as space,
temperature, and the like. They say:
claim. Apparently, in order to achieve the control and coordination that they
have in these discrete actions, timing had to be imposed in much the same way as
it is in many other motor programme views which they reject.
Motor commands
A major point made by various proponents of the action view is that it is incorrect
to think that the CNS organizes motor commands to control the musculature
(e.g., Reed, 1982). This thinking comes from Bernstein's (1967; Whiting, 1984)
ideas, mentioned earlier, that (a) various muscular non-linearities (e.g., length-
tension and force-velocity relationships), (b) unknown or unexpected
perturbations, (c) complex reaction forces from gravity or from other attached
limb segments, or (d) modifications provided at the level of the motorneuron pool,
fail to make the motor command (if there were one) univocal with respect to the
movement it is supposed to produce.
While these are important considerations, Reeds rejection of the command
concept does not follow logically from them. Simply because such commands may
be modified by muscle non-linearities, or by the states of the interneuronal pools,
is not sufficient grounds to conclude that they do not exist. In fact, most
proponents of motor programmes would probably agree completely that these non-
linearities are important to consider (e.g., Partridge, 1979), and that research
should be directed at understanding such interactions.
4. Motor programrmn
* g perspectives
In this Section, I discuss assumptions and supporting evidence for the various
motor-programming perspectives. All of these views have in common the notion
that (at least quick) movements are in some way structured in advance, and that
the central representation of action contributes a large part of the ultimate
behaviour that ones sees in the musculature. These viewpoints have been
discussed many times previously (e.g., Brooks, 1979 & 1986; Schmidt, 1982 &
1987b1, and these complete arguments need not be repeated here. Rather, I prefer
to deal with some of the aspects common to almost all of the programming views,
and which seem most strongly at odds with the action view.
There are two problems. First, I don't think that Keele ever meant, as the action
proponents assert, that programmed movements occur without feedback, or that
feedback is irrelevant to motor behaviour. Rather, Keele meant that evidence for
the existence of programmes could be found by examining the movement
capabilities of subjects deprived of feedback in various ways; if the subjects could
still move, then some central representation of the action is implied (e.g., Keele,
1973).
Second, this is a very old definition, and most present-day accounts of motor
programming (e.g., Brooks, 1979; Grillner, 1975; Schmidt, 1982 & 1987b) argue
strongly that sensory information is a critical aspect which modifies the
expression of central representations to meet the various local contingencies.
Such interactions could presumably occur in many ways. There are the well
known effects of local muscle perturbations on the motorneuron pools, which will
then m o m in various ways the nature of a central command on its way to the
muscles. Afferent information can lead to centrally generated spinal tuning
(feedforward), which influences the motor commands in several ways. And
22 R.A. Schmidt
recent work with visual control of action (e.g., Paillard & Amblard, 1984)suggests
that visual information can modify the central command structure as the
movement is unfolding. It seems to me that the action proponents, perhaps in
order to draw most sharply the distinction between their view and earlier ones,
have engaged a very old and feeble straw person in battle.
A big distinction between the views, though, is the action theorists' denial of any
sort of central representation for movements. Kelso and Tuller (1983;Kelso &
Scholz, in press), for example, have argued strongly against viewpoints which
contain central representations of action; their position seems clear when they
say:
the action view as to exactly what Bernstein's position was on this issue, as the
action view must be able to stand on its o w n without relying on his support. But I
do find it puzzling, in view of the action proponents' persistent and strongly
critical statements about motor programmes and central representations, that
Bernstein should be regarded by them as such a strong guiding light.
I find useful one of Bernstein's (1967, chapter n) earlier frameworks for examin-
ing the nature of movement control. Here, he had just pointed out that some
central command must be modified by a host of features 'along the way' to the
actions they produce (muscle properties, afferent information, reaction forces,
etc., as discussed previously), and then he conceptualized the problem as shown
in Figure 4. Curve B shows the hypothetical time-course of the forces produced
during the conduct of some skilled action. These can be decomposed into three
separate sources: (a) the external forces (Curve A), (b) the pattern of forces
actually produced in the action (Curve B), and (c) those forces generated by the
patterning of the central commands to the muscles a t each moment to make up
the difference (the shaded area C). He stressed (as did Reed, 1982) that the control
24 R.A. Schmidt
signal will probably not be a simple transformation of the actions that are
produced, which is a particularly important lesson. But he does not ever say
(contrary to Reed) that the control signal does not exist. My interpretation of
Bernstein's idea (see also Partridge, 1979)is that if we are ever to understand the
nature of the control signals in such tasks, we must be aware of the confounding
factors that can act to disguise them.
Another, older, line of evidence, related strongly to the one just presented, is that
when a motor response is initiated, sensory information indicating that this
response should be changed, or even aborted, requires approximately 200 msec for
the pattern of action to even begin to be modified (e.g., Henry & Harrison, 1961;
Logan & Cowan, 1984;Slater-Hammel, 1960;see Schmidt, 1987b,for a discussion).
Of course, evidence is strong that certain sources of afferent information (e.g.,
muscle stretch) can be responded to (with increased EMG in the stretched
muscle) much faster than this, but these modifications usually take the form of
relatively small adjustments in a given pattern of action, and usually are not
Action and Motor Views 25
responsible for a major alteration4. In some cases (e.g. , Henry & Harrison, 19611,
the signal to stop the movement occurs prior to the movement's initiation (by
approximately 100 maec), and yet the movement is initiated anyway, and is
carried most of the way to completion, before it is altered. Such evidence not only
stresses the point that afferent information processing (especially if it calls for a
major modification1 can be slow, but it also supports the notion that the movement
must have been structured in advance.
Deafferentatwnstudies
Over a half-century ago, Lashley (1917) described the movement capabilities of a
patient with a gunshot wound to the lower back which blocked afferent informa-
tion from the legs to the CNS, yet probably allowed efferent information to be
transmitted essentially normally. This patient could move, and was not obviously
different in simple leg-positioning from another 'normal' subject. These and
many more recent findings on animals (e.g., Taub & Berman, 1968; Wall, 19701,
have shown that movement is indeed possible without sensory information, often
with only minimd impairment. In viewing Taub and Berman's films of monkeys
with deafferented forelimbs, I was impressed that these animals were not
4 There is also evidence that visual information may be involved in movement control with very
short delays (e.g., Bard, Hay & Fleury, 1985; Nashner & Berthoz, 1978; Zelaznik, Hawkins 81
Kisselburgh, 1983). But these modifications =em to ensure that the original pattern of action is
carriedoutfaithfully, rather than initiating an essentially new pattern (Schmidt, 1976 & 1982).
26 R.A. Schmidt
Although discussion of the dynamical view of movement control (Kelso et al, 1980
& 1981) has not focused on this issue, these deafferentation findings raise a
number of serious questions for it. According to their views, as discussed earlier,
movement control is accomplished by the establishment of dynamical 'equations
of constraint' (e.g., Equation 11, which imply various kinds of afferentefferent
coupling. If such coupling is a critical determinant of action, with no central
programming or representation being allowed, then how can movements be
produced without any feedback from the responding limb? These deafferentation
findings continue to present strong challenges to the dynamical perspective to
explain how movement is actually controlled in such cases.
And, Grillner (1975, p. 297) described the interactions between central and
peripheral levels well when he wrote. relative to locomotion control:
The material in the following Sections, most critical for the motor programming
perspective, appears to show that the temporal structure of at least rapid actions
is centrally organized. "his evidence also suggests that the dynamical views, the
extreme forms of which specifically deny such central representations, appear to
be incorrect.
the limb unexpectedly should have resulted in massive disruptions to the pattern
of EMG activities. In particular, these views which hold that the timing of the
antagonist onset is generated as a function of the limbs position during the
movement (Fel'dman, 1974b & 19861, much as an escapement function in a
pendulum clock (e.g., Kelso et al, 19811, cannot account for these findings,
because the limb on the blocked trials never reached the position a t which the
antagonist EMG would be initiated. (See Berkinblit et al, in press, and Schmidt,
1986, for further discussion of this question.)
Proponents of the action view, although aware of these results (and those of
Conrad & Brooks, 1974, discussed earlier) and the implications for their
perspective, have chosen not t o comment on them or to attempt an explanation of
them in terms of their theoretical position. My o w n belief is that these findings
cannot be explained by the action view, and that the action view will need
modificationbecause of them.
At the same time, motor-programme views (see also Adamovich & Fel'dman,
1984) treat these findings as critical empirical support. A temporal structure is
organized centrally prior to the action and, upon response initiation, is 'metred
out' (as I think of it) to the muscles. When the movement is perturbed - even by
such a severe perturbation as a complete block - the timing structure continues
for a short while; considerable time is required to process the afferent information
which informs about the block, so the alteration in movement control is delayed
somewhat. The beginnings of this alteration in patterning are seen after about 120
msec, as a deviation between the 'normal' and 'blocked EMG records. But notice
that, even though the antagonist burst may be affected by the block, it still occurs,
and the second agonist appears as well. The fact that the pattern remains at all
under these blocked conditions seems to deny models which assign this timing to
an emergent property of the dynamics, and seems to support models which have
this organization structured centrally.
and the trajectory from one such movement from a typical subject. Because of the
short MTs and reversal in direction, the EMG onsets and offsets are very abrupt
and strongly reciprocally organized in this task, making their onset times and
durations particularly easy to measure.
Figure 6. Position-time function (upper) and EMGs from the biceps (middle)
and triceps (lower) muscles for a single trial of a rapid reversal response;
reciprocal organization and distinct EMG on- and offsets (particularly for
biceps) are typical of quick actions (jhm Schmidt et al, in press).
trial, and then averaged across trials and subjects for presentation. In all cases,
these temporal durations increased with MT, more or less as the acceleration-
time functions did.
These findings are consistent with a viewpoint in which the temporal structure
of the adion is centrally organized, and particularly with one specific model
termed the generalized motor programme hypothesis (Schmidt, 1982). Here, a
centrally organized temporal structure can be sped up or slowed down to meet the
overall MT requirements of the task. Of course, the amplitude of the EMGs must
increase as the MT is shortened, so that the musculature can develop sufficient
force to complete the distance and time requirements. This view represents a
version of the impulse-timing hypothesis, in which the motor programme is
-
thought to control muscular impulses forces integrated over time - and the
timing of their on- and offsets.
32 R.A. Schmidt
A number of difficulties face this simple model of Ml' control. One is that,
whereas the duration of a number of kinetic and neuromuscular events seems
strongly linearly related to MT, they are not proportional to Ml', as a simple
oscillator model would suggest. The intercepts range from 30 to 80 msec in
various data sets. The various non-linearities in muscle may require a temporal
structure which is not strictly proportional to MT, which may be particularly
critical in the initial portions of quick movements, as Zelaznik, Schmidt and
Gielen (1986;Schmidt & Gielen, 1986)have suggested.
Kelso and Tuller (1985;Tuller & Kelso, 1984)have interpreted similar data from
speech motor control studies somewhat differently. In addition to plots of the
300 L
280 -
260 -
240 -
- -
E
220
200 -
180 -
Y
E
F
160 -
140 -
120 -
100 -
fC
b/ ' I I I I I 1
durations of various events against each other, as in Figure 8, they have shown
that when the position of a segment (e.g., the jaw) is plotted against its velocity in
a phase-plane analysis, there was a constant location on the plot (i.e., the so-
called 'phase angle') at which onset of upper lip motion occurred. In other words,
the upper lip began to move downward when the relationship between the jaw
position and velocity was in a particular phase of its motion. This kind of feature
is typical of many non-linear mass-spring systems, suggesting a dynamical
interpretation. But, these experiments showing temporal expansion of various
kinematic or EMG activities - whether they be analysed as in Figure 8 or with the
phase-plane analysis (e.g., Kelso & Tuller, 1985)- do not discriminate between the
views having or not having central timing control. I tend to favour a central
timing interpretation of even the speech data, however, mainly because the limb-
block findings from Wadman et a1 (1979),Magill et al(1987),and Shapiro & Walter
(1982)implicate such control in quick actions. But it is clearly premature to decide
such a fundamental issue on these few findings, and much more work needs to be
done in this direction.
The previous Sections discuss the rationale for the concept that movement
control, at least in quick actions, is governed by motor programmes with a cen-
trally controlled temporal structure. This evidence seems sufficiently compelling
that it is used as a kind of starting point for many research programmes in motor
control and motor learning.
I --- I
1 180
E 160
140
.t-E im
[ Time to peak triceps
I-"
100
I I I I I I
20 30 40 50 60 70
Movement Distance (deg)
I
distance. and 80 on). I have given one of these examples earlier (see Figures 7 & 8,
and related discussion), where the MT in a reversal task was varied
experimentally, holding other important features of the movements constant.
Other studies in this series have concerned the movement distance and the
inertial loads on the lever as experimental variables (Sherwood, Schmidt &
Walter, in press). For example, Figure 9 shows averaged temporal intervals
measured from individual-trial EMGs (time to peak biceps, duration of biceps
burst, and interval from biceps initiation to peak triceps) for various movement
distance conditions i n the reversal task described earlier, where the MTs are
strictly controlled. Here, there were small but reliable increases in the duration of
all of the temporal intervals in the EMGs a s the movement distances increased
(the amplitudes of the EMGs increased as well, of course). Such findings allow
the evaluation of earlier models of motor control (Schmidt, 1982;Wallace, 1981)
which predict that, when the overall MT for the task is constant, the temporal
durations of the EMG intervals will be uninfluenced by the amplitude of the
actions. Our findings here contradict this position, and suggest that this simple
characterization of the command structure will not be adequate in general (see
also Zelaplik et al, 1986;Martenids et al,1986).
The proponents of the action view have taken strong exception to this style of
inquiry (e.g., Reed, 1982) for reasons I have discussed earlier, but such criticism
does not seem completely warranted. Reed makes a good point that movements in
natural environments seldom begin from stable or stereotyped starting positions5
and that studies of movement control (like ours cited above) should include
changes in postural positions as additional experimental variables. Such work
would probably help to reveal the interactions among motor commands and
atTerent information from joint position and muscle length. But, again,
remembering that if the goal is to understand the nature of movement
commands, complicating the subject's world would seem to make this
understanding more difficult to achieve. Of course, such procedures do make a
great deal of sense if the research goal is the understanding of the animal-
environment 'interface', as in the action view. The difference in research goals of
these groups makes it very difficult to speculate on which of these approaches is
'better', as the answer will surely depend on one's interests.
5 This is debatable. Many activities, such as a cat crouching before a jump, or a batter assuming a
ebreotyped stance before a pitch, suggest that constant initial conditions are quite common
features ofskills.
36 R.A.Schmidt
An important debate in motor control for the past two decades has centred on the
extent to which, and how, afferent information is capable of modifying rapid
actions while in progress. The proponents of the action view (e.g., Kelso et al,
1981; Reed, 1982) argue that afferent information is used continuously in motor
behaviour, at a variety of levels in the CNS, perhaps with very short latencies.
This continuous view may well be correct in some situations, but I doubt that it is
correct in general. Considerable evidence suggests that environmental
Exactly how such environmental information is perceived - 'directly' as in the action view, or
-
'indirectly' 88 in computational views may not be relevant to the question of how movements are
controlled (e.g.,Epstain, 1985).
Action and Motor Views 37
information calling for a change in the movement pattern (e.g., to start action, to
stop it, or to switch from one pattern to another) will require considerable time to
be responded to, with the originally planned movement pattern continuing in the
mean time; and much evidence suggests that this responding is decidedly
discontinuous in nature (see Schmidt, 1976 & 1982, for reviews). The critical
variable determining the latency of such corrections seems to be movement time
(not velocity, distance, or other covarying factors), so that movements with longer
MTs are increasingly available for environmentally based modifications (Schmidt
& Russell, 1972). Of course, lower-level reflex activities can very quickly alter the
patterns of EMGs in response to various perturbations, but I have argued that
these changes are chiefly alterations to ongoing movement patterns rather than
production of new ones (Schmidt,1976 & 1982).
Recently, though, there has been the suggestion (e.g., Turvey, 1977), with some
empirical support (Bard et al, 1985; Carlton, 1981a & b; Nashner & Berthoz, 1978;
Zelaznik et al, 19831, that visual information may act far more quickly than the
usual 200 msec estimates, perhaps even with delays shorter than 100 msec.
Another intriguing possibility from this recent work is that vision might act
continuously as Turvey (1977) suggested, which brings us far closer to the concept
that visual information operates in the form of optical flow variables (Gibson,
1966; Lee & Young, 1984; see also Schmidt, 1987b). If visual information does turn
out to work continuously, and far faster than earlier estimates, then my earlier
assumptions about the similarities in the control of open and closed skills will be
drawn into serious doubt. But, still, such findings would not preclude the
existence of central motor commands; they would only indicate that these
commands can be modified earlier, and/or via different types of information, than
had been understood previously.
The distinctions that have been drawn so sharply between the so-called action and
motor views are probably overdone. One of the contributing factors to this division,
as I have argued here, is that in many ways these views are about different
things. The action view tends to focus on the relationships between the animal
and the environment, and the motor view tends to focus on (internal)
mechanisms of motor control, so that the differences between the views can be
thought of as differences in the level of analysis in which they operate. There are
exceptions to this viewpoint, but to the extent to which this generalization is
38 R.A. Schmidt
correct, it does not come as a surprise that the proponents of the action perspective
have been critical of the motor-control approach's answers about human-
environment relationships, as these problems are not the major concern of the
motor emphasis. As well, many who hold a motor emphasis have been impatient
with the action view of movement control (particularly the dynamical
perspective), arguing that it does not provide sufficiently testable explanations of
movement phenomena. In fact, Epstein (1986) points out that writers representing
these two traditions frequently reference each other's writings (primarily in a
critical vein), but there is little empirical work that can be taken as a basis to
choose between the two viewpointa of a particular problem.
"here is a great deal of mom for compromise with respect to some of the issues
under debate here. For example, the call for the motor proponents to do work with
more consideration for environmental variables, t o examine movements with
somewhat more complexity, to make increasing use of dynamical analyses, and
to consider in detail the ways in which sensory information influences movement
control, are all issues for which I feel considerable sympathy. A t the same time,
the action approach could cast their theoretical ideas in more testable ways, and
conduct research whose outcomes could potentially represent more serious
challenges than they have to the predictions of the motor perspective. But, any
narrowing of the gap between the two views will be fundamentally limited (in my
view) by the action proponents' absolute denial of any central representation for
movements - a position which leaves little middle ground for those of us who see
strong evidence for, and are primarily interested in, such representations. This
division, however, is healthy for our science, as it provides true alternative
paradigms for understanding movement control. It would probably be unwise to
devote much effort at a realistic fusion, even if the consenting parties would be
willing.
ACKNO"LF,DGEMENTS
Thanks to Herbert Heuer, Ron Marteniuk, Onno Meijer, Karl Newell, Ed Reed,
Elliot Saltzrnan, Mike Turvey, and two anonymous reviewers, for their helpful
comments on an earlier draft. Preparation of this manuscript was supported in
part by Contract It MDA903-85-K-0225 from the U.S.Army Research Institute to
Richard A.Schmidt and Diane C.Shapiro.
Action and Motor Views 39
REFEWNCES
Adamovich, S.V. & Fel'dman, A.G. (1984). Model of the central regulation of the
parameters of motor trajectories. Biophysics, 29, 338-342.
Adams, J.A. (1987).Historical review and appraisal of research on the learning,
retention, and transfer of human motor skills. Psychological Bulletin, 41-
74.
Armstrong, T.R. (1970). Training for the production of memorized movement
patterns. Technical Report No. 26, Human Performance Center, University
of Michigan.
Bard, C., Hay, L. & Fleury, M. (1985).Role of peripheral vision in the directional
control of rapid aiming movements. Canadian Journal of Psychology. 39,
151-161.
Berkinblit, M.B., Fel'dman, A.G. & Fukson, 0.1. (1986).Adaptability of innate
motor patterns and motor control mechanisms. The Behavioral and Brain
Sciences, 9, 585-638.
Bernstein, N. (1967).The Co-ordination and Regulation of Movements. Oxford
Pergamon Press.
Bizzi, E. & Abend, W. (1983).Posture control and trajectory formation in single-
and multi-joint arm movements. In J.E. Deamedt (Ed.), Motor Control
Mechanisms in Health and Disease (pp. 31-45).New York Raven Press.
Bizzi, E., Accornero, N., Chapple, N. & Hogan, N. (1982). Arm trajectory
formation in monkeys. Experimental Brain Research, 46, 139-143.
Brooks, V.B. (1979). Motor programs revisited. In R.E. Talbot & D.R. Humphrey
(Eds.), Posture and Mouement (pp. 13-49).New York Raven Press.
Brooks, V.B. (1986). The Neural Basis of Motor Control. New York Oxford
University Press.
Carlton, L.G. (1981a). Processing visual feedback information for movement
control. Journal of Experimental Psychology: Human Perception and
Performance, 7, 1019-1030.
Carlton, L.G. (1981b). Visual information: The control of aiming movements.
Quarterly Journal of Experimental Pqychology, 33A, 87-93.
Conrad, B. & Brooks, V.B. (1974). Effects of dentate cooling on rapid alternating
arm movements. Journal of Neumphyswlogy, 37, 792-804.
Epstein, W. (1986). Contrasting conceptions of perception and action. Acta
Psychologica, 63,103-115.
Fel'dman, A.G. (1966). Functional tuning of the nervous system with control of
movement or maintenance of a steady posture - 111. Mechanographic
analysis of the execution by man of the simplest motor tasks. Biophysics, 11,
766-775.
Fel'dman, A.G. (1974a).Change of muscle length due to shift of the equlibrium
point of the muscle-load system. Biophysics, 19, 544-548.
Fel'dman, A.G. (1974b).Control of muscle length. Biophysics, 19, 766-771.
Fel'dman, A.G. (1986).Once more on the equilibrium point hypothesis (X model)
for motor control. Journal of Motor Behavior, 18, 17-54.
Fowler, C.A., Rubin, P., Remez, R.E. & Turvey, M.T. (1980). Implications for
speech production of a general theory of action. In B. Butterworth (Ed.),
40 R.A. Schmidt
Kelso, J.A.S. & Tuller, B. (1985). Intrinsic time in speech production: Theory,
methodology, and preliminary Observations. Haskins Laboratories Status
Report on Speech Research SR81,New Haven, CT.
Kelso, J.A.S. & Tuller, B. (in press). Intrinsic time in speech production: Theory,
methodology, and preliminary observations. In E. Keller (Ed.), Sensory and
Motor Processes in Language. Hillsdale, NJ: Erlbaum.
Kelso, J.A.S., Tuller, B., Vatikiotis-Bateson, E. & Fowler, C.A. (1984).
Functionally specific articulatory cooperation following jaw perturbations
during speech: Evidence for coordinative structures. Journal of
Experimental Psychology: Human Perception and Performance, 10, 812-
832.
Kelso, J.A.S., Vatikiotis-Bateson, E., Saltzman, E.L. & Kay, B. (1985). A
qualitative dynamic analysis of reiterant speech production: Phase
portraits, kinematics, and dynamic modeling. Journal of the Acoustical
Society ofAmerica, 77, 266-280.
Kugler, P.N., Kelso, J.A.S. & h e y , M.T. (1980).On the concept of coordinative
structures as dissipative structures: I. Theoretical lines of convergence. In
G.E. Stelmach & J. Requin (Eds.), Tutorials in Motor Behavior (pp. 3-47).
Amsterdam: North-Holland.
Kugler, P.N., Kelso, J.A.S. & Turvey, M.T. (1982).On coordination and control in
naturally developing systems. In J.A.S. Kelso & J.E. Clark (Eds.), The
Development of Movement Control and Coordination (pp. 5-78).New York
Wiley.
Kugler, P.N. & Turvey, M.T. (1986). Information, Natural Law and the SelF
Assembly of Rhythmic Movement. Hillsdale, N J Erlbaum.
Lashley, K.S.(1917). The accuracy of movement in the absence of excitation from
the moving organ. American Journal of Physiology, 43, 169-194.
Lee, D.N. & Young, D.S. (1984). Visual timing o f interceptive action. In D.J.
Lngle, M. Jeannerod & D.N. Lee (Eds.), Brain Mechanisms of Spatial
Vision (pp. 1-30).La Haye: Martinus Nijhof.
Logan, G.D. & Cowan, W.B. (1984).On the ability to inhibit thought and action: A
theory o f an act of control. Psychological Review, 91, 295-327.
Magill, R.A., Young, D.E., Schmidt, R.A. & Shapiro, D.C. (1987). Unpublished
data, UCLA.
Marteniuk, R.G., Mackenzie, C.L., Jeannemd, M., Athenes, S. & Dugas, C.
(1986).Unpublished manuscript, University of Waterloo.
McLeod, P. (1983).Unpublished observations, Oxford University.
Michaels, C.F. & Carello, C. (1981). Direct Perception. Englewood CWs, N J
Prentice-Hall.
Nashner, L. & Berthoz, A. (1978).Visual contribution to rapid motor responses
during postural control. Brain Research, 150, 403-407.
Paillard, J. & Amblard, B. (1984). Static versus kinetic visual cues for the
processing of spatial relationships. In D.J. Ingle, M. Jeannerod & D.N. Lee
(Eds.), Brain Mechanisms of Spatial Vision (pp. 299-3301, La Haye:
Martinus Nijhof.
Partridge, L.D. (1979).Muscle properties: A problem for the motor physiologist. In
R.E. Talbot & D.R. Humphrey (Eds.), Posture and Movement (pp. 189-229).
42 R.A. Schmidt
Richard A. Schmidt
Motor Control Laboratory,
Dept. of Kinesiology
University of California,
Los Angeles, CA 90024 USA.
Complex Movement Behaviour: The' motor-action controversy, pp. 45-86
O.G.Meijer & K Roth (editore)
0 Elsevier Science F'ubliehera B.V. (North-Holland),1988
Chapter 2
Edward S. Reed'
SUMMARY
1.Introduction
We commonly speak of motor skills with respect to many diverse action tasks:
sport skill, industrial skill (including tool use), dance, and 'everyday' skills (such
as combing one's hair or opening a door). Students of motor skill have assumed,
by and large, that what makes these various activities all instances of the class
'motor skills', is that they are accomplished by muscular movement. In the 1880s
Marey adapted Muybridge's revolutionary photographic techniques to capture the
displacements of limbs and the body in a static visual rec0rd.h the 1890s
Woodworth (1899 & 1906) showed how experimental procedures could be used to
trace changes in these displacement records, changes that he interpreted in
terms of skill learning. It was not long before Frederick Winslow Taylor applied
such methods to quantifying the movement skills of manual labourers and
industrial workers (see Braverman, 1974)and invented 'time and motion' studies.
The skills analysed by these pioneers of our field were as diverse as dancing,
hauling, writing, and reaching. The assumption these researchers made was
that any skill could be decomposed into patterns of bodily displacement effected by
muscular activity, and that therefore we should understand all skill as being built
up out of elementary neurobehavioural units (see Schmidt, 1982, and Smyth &
Wing, 1984, for brief histories). Psychologists, physiologists, and engineers all
attempted to isolate such basic movement units in order to understand the
mechanisms of motor control and coordination. Wynne Lee (1984,p. 143)recently
expressed this assumption quite succinctly:
And, from coaches to kinesiologists, we have all tended to assume that complex
skills must be made up out of such simple units. A variety of taxonomies of skill
have been proposed (Higgins, 1977),but all of them adhere to this assumption of
neurobehavioural building blocks for all action.
The theory of action systems has a very different origin than does the study of
skill. Action systems theory derives from an attempt to develop a n ecologically
based taxonomy of all actions, not just so-called skills (Reed, 1985). It begins with
the hypothesis that there are many diferent action systems, each evolved by
natural eelection to facilitate a unique behaviourd function, and therefore each
with its own principles of organization. Far from claiming that actions are built
up out of functionally indifferent units of bodily displacement, action systems
theory makes the bold claim that the units of action are themselves functional
actions, and that bodily displacements are a consequence of, not constituents of,
actions. Let me be very clear: I do not deny that neurobehavioural units exist, and
can be demonstrated in the laboratory; what I do deny is the theory that functional
actions must be constructed out of such purely mechanically specific units (see
Reed, 1984). Being functionally based, action systems theory resembles that area
of motor skill research that emphasizes task orientation and offers a taxonomy of
acts based on goals (Gentile et al, 1975;Greene, 19721,but it adds the assertion
Applying the Theory ofAction Systems 47
If action systems theory were nothing more than a fknctional taxonomy of action
it would not be especially relevant to motor skills research. To characterize the
skill of, say, spoon use as an example of manipulative function in the
performatory system may be useful in comparative zoology, but is of dubious value
in human movement science. The problems we students of action face go well
beyond ascertaining the function of particular actions - although it is an
unfortunate fact that the majority of researchers still neglect to ascertain the
functional parameters of the tasks they study. For both theoretical and practical
reasons we need to know not only what function a task fulfills, but also what
activities make up the task. It is these activities that we must rehabilitate or
replace when the action system is damaged, and it is through learning these
subsidiary activities that the development of skill occurs. Action systems theory
3 The older claim that actions are built up out of muscle activities will not be dealt with here, for it
rests on the unproven and almost surely incorrect assertion that there is a unique mapping of
muscle contractionsto bodily displacemanta.
Applying the Theory ofAction Systems 49
water, the diurnal path of the sun through the sky, and the slow procession of the
sun along the ecliptic through the seasons. All of these define absolute features of
the environment. There is plenty of evidence that animals orient to these
environmental features, but little evidence that they orient to objects in space,
despite more than a century of research attempting to specify in what 'coordinate
systems' bodily movements are made. It is a fact that a series of displacements of
our limbs or bodies will have entirely different meanings depending upon our
orientation to the environment. If motor control is a series of limb and body
displacements. then the actions produced by the motor control system (i.e., the
goals or environmental consequences of the actions) w ill necessarily vary. A
cognitive schema is then required to organize the displacements so as to achieve
the desired goal (Arbib,1981;Gallistel, 1980;Newell, 1978;Schmidt, 1975).
4 This statement ought not to be read as implying that the movements cause the actions, or even that
the movements precede the actions in time. It is quite likely that desired changes in an
organism'srelationship to its environmentcause movements.
50 E.S. Reed
and the course of play, orienting one's perceptual systems to the opposing hitters
and the ball's motion t o which they give rise, and changing one's orientation so as
to meet the ball with one's glove. The well prepared player can thus catch a ball
without any displacement of the hands whatsoever, if the batter hits a line drive
A I
I J
Part 1: F i v e
IE F
G
r
to the expected spot - yet surely this non-displacement would count as a highly
skilled action!
My fundamental hypothesis is that an action is made up of whatever
displacements sufKce to realize the agent's goal. What matters is not the
displacements in space-time, but the movements and postures in the
environment. Although it is admittedly confusing, I call the orientation
components of an action 'postures' and the changes in orientation 'movements'
52 E.S. Reed
following J. J. Gibson (see Reed & Jones, 1982, chapter 4.7). In my usage, a
posture is not a static position of the limbs or body. A posture is an action in its
o w n right, a persistent achievement of an organism in the environment. The
fundamental human postures involve our upright pose (whether walking, sitting,
or standing) and our ability to keep our head and eyes on what we are doing,
whether we are locomoting or manipulating objects or engaged in social
interaction. What I call 'movements' - specific changes in posture that are
organized by agents while effecting particular changes in the relation between the
organism and the environment - are also subsidiary actions. To go from standing
to sitting can involve countless different muscle movements and displacements,
but, environmentally considered, is a single kind of act, It is significant that, in
learning to sit and stand, human infants go through a variety of movement
strategies, all of which are attempts at fulfilling this single function of getting the
body erect (Figure 1).
To learn to sit is not to learn to make certain displacements in three
dimensional space, but to learn to solve a very specific environmental problem. In
sitting one is not 'upright' with respect to some abstract axes, but one is upright
with respect to gravity, and with one's trunk pressing down on one's thighs, hips,
and feet, while the head is held up, able to see around.
The inadequacies of the traditional motor control theories are nowhere better
illustrated than in the phenomenon of so-called postural 'compensation' for
movement. The attempt to treat human movements as displacements caused by
mechanisms of reflex, or centrally programmed, responses leads to a whole
series of puzzles concerning how the nervous system balances and coordinates
the body during active movement. I believe these puzzles will only be resolved by
treating movements and postures ecologically, as suggested above. Here I will
first review the literature on this subject, emphasizing the contradictions that
arise from it, and then show how an ecological approach, treating movements
and postures as components in a streamaf action that unites agents with their
environment,helps resolve some of these difliculties.
When a movement of a body member is made there arise simultaneously and
necessarily disbalancing forces on the body. The force required to move a body
member exerta an equal but opposite reactive force against the body's centre of
gravity, and the change in configuration of the body's mass from one position to
another establishes a new equilibrium of forces, a new centre of gravity, and so
Applying the Theory ofActwn Systems 53
a
on. This all seems obvious to us now, but it was only discovered recently by
students of N. Bernstein, who had throughout his career emphasized the role
A
3 I
I
QD
If\,
I HM
A TA
I TS
t
All that motor programmes need worry about, this reasoning went, is the desired
force output andor bodily displacement, "consequently,"Lee continues,
(1980, p. 191) careful study. Thus, although the order of activation of postural
muscles was task specific (first ipsi-, then contralateral), the temporal
relationship between the postural muscles and the prime mover was somewhat
variable. Because of this, Lee argued that the mode of control for postural
precedence is a mixture of centrally driven and sensorimotor feedback control. We
can consider this as evidence against the dichotomy of feedback versus
programmed movement, and we shall see that such,mixed results form the
pattern emerging from this entire field of study.
Bouisset and Zattara (1981) developed the Belen'kii et a1 (1967) paradigm further
through studying more of the muscles involved and using a force plate to monitor
the movements of the relevant parts of their subjects' bodies. They demonstrated
that, prior to raising an arm, there is a forward and upward acceleration of the
body's centre of gravity:
The contralateral knee first moves forward, and the corresponding hip
moves backward, 40-60 msec before the upper limb begins its forward
movement. The ipsilateral knee moves backward a short time later and the
correspondinghip moves forward (268).
Although they claim that "this pattern is modified when the conditions of
voluntary movement are changed (p. 2681, as should be the case for pre-
programmed movements, their study only partially supports this statement. The
pattern did change when subjects were asked to raise both arms (reducing the off-
balancing forces and correspondinglyreducing the postural adjustments) but the:
Again, the distinction between centrally driven and feedback controlled does not
seem to apply to these postural adjustments, which are a bit of both and neither.
When the right arm failed as a support, even though the subject was
unaware that it had done so, it lost postural response (Marsden et al. 1983,
656).
Postural adjustments prior to movement, thus, act as though they are both under
feedback and central control, and as though they are controlled in neither way.
Action systems theory, which rejects the standard sensorimotor and central-
peripheral dichotomy (Reed, 1982) can thus offer a novel and useful
reinterpretation of these phenomena.
Applying the Theory ofAction Systems 59
If movements are not physical displacements of body parts, and if postures are not
lack of displacement (or even tendencies towards a lack of displacement), then we
need not explain how resistance to unintended displacement accompanies
intended displacement. This does not mean that the data discussed above
concerning postural adjustments are unimportant, but simply that they need not
be interpreted as anticipations of the displacement of the body's centre of gravity.
True, such displacements do occur - as the above survey amply illustrates - but
what do these displacements mean ?
The human body is in a state of constant potential disequilibrium with respect to
gravity and the surfaces of support underneath it. If our nervous systems were to
be suddenly shut off in the midst of almost any activity - including sitting and
lying down - our bodies would collapse against the ground. It is a major
achievement of the first year of life to learn t o support our trunk, limb, and head
(Thelen, 1984; Reed, 1984b). Human stance is not, however, the positioning of a
rigid object in space - it is not even like the positioning of a non-rigid object. When
we stand upright our bodies act somewhat like a system of loosely coupled springs
organized into a semi-fluid state. "The standing body does not sway about the
ankles as rigid rod, but as 'a reed in the breeze"' (Thomas &Whitney, 1959, p. 534).
However, we are thinking and feeling reeds, with particularly massive and
mobile heads and limbs. The measurement of the sway of the centre of gravity and
of the head reveals a n enormous range of variation in these postural cycles
(Smith, 1957;Taguchi, 1980; Thomas & Whitney, 1959; see Figure 3).
Neurophysiologists have studied posture as though i t were a reaction to stimuli.
Shenington and Magnus actually treated the force of gravity as 'the stimulus'
creating the upright posture and muscle tone in the so-called 'anti-gravitic'
muscles, and the consequences of this mechanistic view are still being felt
throughout neurology and physical therapy. But posture is not a reaction to a
stimulus, it is an activity in its own right, a function of what I call the 'basic
orienting system'. To use Gibson's (1966) apt terminology, the upright posture is
an achievement of a n action system based on obtained not imposed stimulation.
We must reinterpret experiments on motor reactions in light of what we know
about how the nervous system organizes action in the environment. And we must
begin to study both perceiving and acting in conditions where the agent is
obtaining stimulation, otherwise we will never learn how organisms actually
cope with their surroundings.
Although various parts of the body sway in different ways, and although these
sways are not inconsiderable, the basic orienting system functions to constrain
60 E.S. Reed
and coordinate these oscillations. Perhaps the fundamental ' balancing act' for
humans is to maintain an upright posture while breathing, which changes the
EYES OPEN
Head Movement
3cm L 3 cm
Figure 3. Sway of the centre of gravity of the body & head (Adapted from
Taguchi, 1980, p. 103). Time course in the averaged position of the head and
center of gravity. Each point represents an averaged position for ten
seconds. The initial circle being the second ten second measurement in a
period of a minute, the arrowhead being the last. Subject was engaged in
free standing.
Applying the Theory ofAction Systems 61
shape of the chest and disbalances the trunk above the hips and the head above the
trunk. Gurfmkel et al (1971) have shown that the orienting system produces
compensatory relationships between head and body sway, a kind of antiphonal
voicing in the melody of stance that serves t o keep the head oriented to the
environment. This reciprocal oscillation of the head and trunk is specific to the
task of maintaining the head upright, for supine humans exhibit a totally altered
pattern of respiration, in which the abdomen expands more than the ribcage,
rather than vice versa (Konno & Mead, 1967; Druz & Sharp, 1981). Furthermore,
the amplitude of the heads movement decreases in a standing human when a
perceptual task is added to the task of merely standing still (Fearing, 1926).
Given these results, the action of maintaining an upright posture should be
defined as any set of displacements of the musculo-skeletal system that serves to
keep the head oriented to the environment above the neck and trunk. The
displacements (or lack thereof) of the body and its parts do not constitute the
upright posture, for any such displacements will do, so long as the above
functional requirement is met. I would go further, and argue that the primary
function of the upright posture in normal humans is to keep the visual system
oriented to the environment. Prismatic reversal of t h e visual field - which alters
the eyes' ability to fix on environmental targets - causes what has been called a
disruptive 'swinging' of the visual field and a simultaneous disruption of balance,
to which subjects slowly adapt (Kohler, 1964; Gonshor & Melville-Jones, 1980).
Every clockwise rotation of the gaze line results in optical information that had
previously specified a counter- clockwise gaze movement: the world seems to
swing crazily around, and one's balance and posture are precarious until one
adapts. This is what I mean by saying that postures cannot be defined in terms of
displacements (or lack of motion). What counts is not the displacements of the
body - which are multiple and variable anyway - but the functions served by these
movements. The importance of the basic posture of upright standing and sitting
in humans lies in facilitating the functioning of the perceptual systems which, as
will be discussed below, is crucial for the control of all action.
... a preferred 'set up' maneuvre by which the chest wall is prepared for
phonation .,, . Just before phonatory onset, volume is usually shifted from
the abdomen to the thorax.
This coordinated movement of ribcage and abdomen may place the relevant
articulatory muscles in a position of mechanical advantage (Hixon, Goldman &
Since this was written (September, 1985) several laboratories have extended their studies of
'postural precedence' to include modulations of bodily support during walking. How, for example,
does one support one's weight during the swing phase of the step cycle immediately prior to
reaching out and pulling open a heavy door? Although the results are preliminary as yet, it appears
that the persisting postural tasks constrain the specific muscles involved and the intensity of their
activation (seeForssberga Nashner, in press, and Patla, in press).
Applying the Theory ofActwn Systems 63
Mead, 1973). Thus the lungs are prepared to act as a bellows for vocalization, but
in such a way that neither the respiratory cycle nor the postural cycle are
seriously perturbeds.
The ability to nest movementa into postures requires perceptual information and
control. The oscillations of the head and pelvis could not be in compensatory
counter-phase without the use of information to coordinate the two. Nor could the
head achieve functional orientation to its surroundings without a different, but
equally perceptual, kind of control. The orientation of body members to each other,
as in the antiphonal movements of the hips and head, is a purely propriospecific
function. The hand can be related to the trunk, and the trunk can be related to the
head, regardless of how any of them are oriented to the ground. The vestibular
system and the musculo-skeletal components of the haptic system provide much
of this body-based proprioception (Gibson, 1966,chapters 4 & 6).
But the ability to orient hand to head or head to heels is not functional on the
earth without a coordinate ability to orient one's whole self to the direction of
gravity, the path of locomotion, obstacles on the ground, and so on. The bird flying
desperately into a headwind, and not moving forward at all, should not rely on
haptic information specifying rapid movement of her body. Without information
specifying where one is in the environment (Gibson, 1979, chapter 10) - what
David Lee (1978) calls 'expropriospecific information' - meaningfid posture and
movement is simply impossible. Astronauts in the dark or with their eyes closed
report a total loss of external orientation (Kerwin, 1975). The statocysta in the
vestibular system, and skin and pressure sensitivity in our heels and backsides
provide some expropriospecific information about gravity for non-astronauts. By
and large, however, human beings use vision for exproprioception.
Through vision we perceive the layout of surfaces surrounding us, and especially
the ground. We also see the path of our own movement over the ground and
through the environment. Just as a bird in a headwind can accurately perceive
that she is not going anywhere, so a person walking forward can be made to
falsely perceive that she is going backwards by moving her visible ~ u ~ o u n d i n g ~
It should, however, be noted that the respiratory cycle during speech is highly modified compared
-
to that during quiet breathing; vocalization demands a functional reorganization of respiration,
but not a disruption see Lieberman,1984, for details.
64 E.S.Reed
backwards at a faster rate than her forward movement (Lishman & Lee, 1973).
Similarly, the oscillations of a normal adult's sway can be shown t o depend on
visual control by leaving the mechanical contact of their feet with the ground
stable but 'swinging' their visible environment, which results in efforts to
compensate for the (illusory) sway (Lee & Lishman, 1975; Nashner, 1983; Paulus
et al, 1984).
This perceptual control of posture and movement need not be conscious.
Participants in Lee's experiments found themselves unable to balance as they
compensated for the non-existent sway that was optically specified. Although they
were certainly aware that their balancing was (temporarily) deficient, they were
not aware of the visual basis of the effect (and often blamed it on non-visual
factors). Certainly subjects are unaware of posturally compensating for the effects
of arm or other movements. Marsden et al (1983) report that it is not conscious
awareness of support that creates the locus of their postural adjustment effects,
but perceptual registration of information specifying support. Experiments need
to be done where subjects are made to perceive that they are being supported by,
say, their arms, when in fact they are not, to see whether the locus of postural
anticipation is based on detection of expropriospecific information, or through
some other means. The implicit assumption in the literature that such effects
cannot be perceptual because they are not conscious is unwarranted. For
example, Henry (1953) has shown that when a person is required to stabilize a
moving handle the threshold for consciously reporting the movement is 18 times
greater than that for successfbl action. The term perception should thus be used
as follows:
6. Action cycles
The claim that actions are made up of postures and movements (as opposed to
responses and programmes) does not in itself explain how actions occur. Even if
we knew all the processes underlying postures and movements, which we do not,
and even if we knew all the rules for the perceptual guidance of action - which we
also do not (but see Gibson, 1979, chapter 13) - we would still not know how
movements and postures are organized into functional acts. Any particular
Applying the Theory of Action Systems 65
1. Sitting erect
adjusting seated posture to look at objects, surfaces, etc.
adjusting seated posture to facilitate pouring movements
adjusting seated posture to table / surface of support
2. Breathing
3. Looking at task
4. Positioningobjects for task
5. Hoisting pitcher
6. Holding or hoisting cup (optional)
7. Rotating pitcher
8. Guiding stream of liquid
9. Monitoring splash of liquid / amount poured
10. Righting pitcher
11. Re-positioningobjects (optional).
If action systems theory is correct, at any given time an agent will be engaged in a
variety of functional activities. A healthy adult is always orienting, breathing, and
investigating, and almost always locomoting, exploring perceptual information,
making expressive gestures, and speaking. These separate streams of activity
continually overlap: we do not stop looking around when we go for a walk, nor do
we stop talking during a meal, even though ingestion of food and speech are two
radically different functions of the same effector organs. Each action system has
7 The studies of pouring discussed here were designed in collaboration with Nathaniel Mayer,
MD., Director of the Drucker Head Injury Center, Moss Rehabilitation Hospital, Philadelphia,
PA
66 E.S. Reed
its own characteristic action cycles, all of which are rooted (in humans) in the
diurnal cycle and, to a lesser degree, in the seasonal cycles. The basic orienting
system, the appetitive and investigatory systems are operating continuously in the
waking state and, very likely, also operate (in a much reduced fashion) during
sleep. Locomotion, manipulation, speech, and play break up the day in varying
clusters of activity as, obviously, does eating. In humans, expressive activity only
desists during sleep - and then not completely. Our faces are continually
modulating their expressions as we engage in one activity or another.
From this point of view, the problem of explaining the organization of action
becomes one of explaining how these various streams of activity are nested into
unified acts. And, where agents are capable of accomplishing more than one
thing at one time, how the streams are kept separately nested, cycling in parallel.
To use pouring again as an example, we have seen a variety of preparatory
actions, all serving the function of enabling such a functionally specific nesting to
occur. Most subjects 'address themselves' to the task by orienting their eyes,
heads, hands, and arms toward the cup, pitcher, tray and table that supports
these objects. They adjust their seated posture if necessary and, almost invariably
(the exceptions are quite rare), adjust the positions of the objects as well. If the
experimenter contrives to make the situation a bit awkward, as placing the
pitcher in position for a left handed pickup with a right handed subject (or vice
versa), then the pourer will always rearrange the objects to make the task easier,
prior to any attempt to actually pour8. Because our pouring was done in a t least a
quasi-experimental situation, our pourers all concentrated intently on the task,
and modulated their facial expressions in a serious manner. In this single,
'simple', everyday task, almost every action system was thus involved - for a list of
the human action systems, see Table 2.
In our work on pouring we are hoping to show that the way to study complex
movements is not to create laboratory experiments that simplify an actor's
situation so that a small range of variation in the action is seen. This
Bimplification procedure - followed by almost all students of motor skill - makes
sense if one accepts the traditional theory of skills as constructed out of
mechanical neurobehavioural units. But I seriously doubt anyone will ever be able
to show how an action like pouring is comprised of hundreds of interacting
reflexes, programmes, and feedback loops (and there would have to be hundreds,
In more than 250 trials, no subject ever reversed the cup and pitcher unless the experimenter had
intentionally placed them in the wrong position.
68 E.S. Reed
look how complicated our models of simple reaching movements have become!).
The only aIternative is to begin the study of action with an attempt to discover the
range of variation naturally found in an action so as to see what various systems
are being nested into a coherent act. Only when one knows what this range of
variation is, can one then attempt t o control it experimentally. We would like to
know, for example, under what circumstances people will stand up to pour (they
never do in the study described, although we have created circumstances where
they will), or where they will always hold the cup, or where they will adjust the
objects before pouring. Note that what our subjects do vary are not responses, but
functionally meaningful units of the pouring action: postures and movements,
ecologically considered. It is to be hoped that certain general principles of the
control of postures and movements will emerge from such studies - for instance,
principles about when actors engage in 'setting up' objects needed for a task - but
we will never discover any such general principles without exploring the
complete range of variations in human action.
7. Basic actions
The action systems of animals have evolved and differentiated because the
resources they use are biologically different. Breathing is one process and
balancing quite another, although neither can go on without the other. Eating is
different yet again, but it is as much a necessity of life as breathing and
balancing, and must also be coordinated with those activities. Although through
the course of evolution, differential selection pressure has resulted in some
anatomical specialization for the action systems - Masseter is a muscle for eating
and not breathing in humans, whereas the reverse is true of the Intercostal
muscles - the systems are most clearly distinguished, behaviourally, by the basic
acts that have evolved in each of them (see Table 2). Given the anatomical and
perceptual adaptations of human beings (or any species of animal), and given
their biological niche, certain characteristic modes of action have been selected.
These basic acts are the most important nesting of postures and movements
under a particular set of evolved environmental constraints, and as such they
constitute the repertory of human daily skills.
It should be stressed that basic acts are not 'innate' in the sense of genetically
inherited patterns of movement; nor are they simply learned due to
environmental constraints. They are ecological facts about the human species,
not genetic or behavioural facts about individuals. Elizabeth Bates (1979,p. 17)
explained this well:
Applying the TheoTy ofAction Systems 69
It is a universal fact of human cultures that people eat with their hands,
sometimes with an intervening fork or chopstick, but certainly never with
the feet, nor with head movements in a trough. Do we want to argue that
hand feeding is innate? Of course we do. But there is no need to invoke a
specific genetic mechanism for eating with the hands. [Given human
anatomy and our ecological niche] ... the solution of hand-feeding is so
probable that no hrther genetic baggage is necessary for the outcome to be
assured. Nor does hand feeding have to be 'learned.
The human infant may have to engage in perceptual learning to gain better
control over feeding herself. And, doubtless, there is some action learning that
goes on as the movements and postures are increasingly coordinated and made
more accurately. But a healthy human infant simply perceives that objects of
certain sizes, shapes and weights are graspable, and she is motivated to reach out
and grasp many things. Whether she brings them to her mouth to taste them or to
explore them haptically is an open question. But there need be no genetic
programme for a 'bringing to the mouth movement nor need there be some kind
of reinforcement of reaching responses for hand-feeding to develop. What is
needed are functional perceptual and action systems in the baby, and enough
interest and motivation to begin to act.
One reason why there is no genetic mechanism for hand feeding, or for any
basic act, is that these are not stereotyped movements, like the ethologists'
mislabelled 'fixed action patterns'. The paths through which even a young infant
gets food into her mouth are many and devious. Sitting on our Glutei maximi is a
basic act of human orientation, but there are many ways to achieve the result of
sitting erect in the environment (see Figure 4). An individual will adopt different
ways to sit depending upon his mood, social situation, or the available sitting
places. What makes a basic act a definite entity is not a stereotype of
configuration, but a definite range of variation which has been established by
natural selection and within which most individuals of a species stay most of the
time. For example, human standing is almost universally bipedal, a definite limit
to the range of variation. Nevertheless, it is possible to stand unipedally, so long as
there is an available horizontal surface of support (Hewes, 1955). Hopping on one
leg is also not a basic mode of locomotion, but it can be used as a mode of
progression under certain circumstances, or for play.
The concept of basic actions and the tentative list given may seem obvious and
even trivial. 'Everyone knows' that people stand and sit, after all. But this is
precisely the problem. We have concentrated so much on ascertaining the facts
about the alleged mechanisms underlying certain simplified movements under
70 E.S. Reed
conditions that are quite subtle and unrepresentative of daily life that we do not
know, in anything like a scientific way, what the skills of everyday living are. We
certainly do not know how to characterize the movements and postures involved
in many of the most fundamental human bodily skills, nor are we well informed
about what constitutes the normal range of variation of those movements and
postures. The list given in Table 2 needs examination and revision until we are
able to specify with some precision not only the functions of the basic acts, but the
typical human movementa and postures that serve to realize those functions.
Our group has been focusing on the skill of eating because of its importance to
the individual (psychosocially as well a8 nutritionally) and its practical and
economic relevance to the management of chronic care in hospitals and the
home.
#B
!i 4
1 L 3
5 6
8. Eating
There are many individual and social preliminaries to eating. The person who
lives in a hunter-gatherer society may simply pick-and-eat a significant portion of
their nutritional requirement, whereas we agriculturalists tend to store food
(Sahlins, 1972). For the sake of simplicity I will here ~ ~ S C U S merely
S the eating of
an already prepared meal. This is, after all, the skill a child learns first, prior to
learning about food preparation.
Humans, as we know, do not eat out of troughs. We usually eat sitting up and,
for the last few thousand years, a goodly number of US, from villager to peasant to
urbanite, have sat at tables. One's body must be adjusted to the table (not all tables
require sitting on chairs, as Japanese style dining shows) in such a way that the
item(s) to be eaten and utensils for use are within reach. Thus, prior to eating, the
self and a number of objects must be positioned for the task. If utensils are used
they must be obtained and made available for the meal. The back and head must
be held sufficiently erect for the passage of food down the throat, with the neck
slightly flexed. This erectness can be interspersed with intervals of bending
forward to meet the food, but the bending cannot be too great or uninterrupted;
some 'swallowing problems' among our patients have turned out to be postural
problems instead (H. Avart, personal communication). Breathing must continue
throughout eating and drinking, and the process of chewing and swallowing
cannot be allowed to interfere with this. Basic orientation to gravity, to local
surfaces of support, and the appetitive action cycle of breathing are the
background cycles into which the movements of eating are nested. In a social
-
eating situation and this 801% of eating in a group, or mealtime, is characteristic
of Hominids (Campbell, 1985) - one must also face others and maintain proper
expressions throughout the meal.
To eat, one must bring the food object to one's mouth, either by grasping it
directly with one's hands or indirectly, through the use of a tool. Liquids, which
cannot be grasped, can be transported to the mouth in vessels or CUPS, one of the
oldest of human inventions. (The use of tools to transport liquids is one of the most
common throughout the whole animal kingdom, see Beck, 1980.) The right
amount of food must be gotten: too much to be chewed off is no use, and can lead
72 E.S. Reed
to choking or gagging; too little may make one bite one's tongue, or at least go
hungry. Chewing and swallowing must be counterphased with inhalation and
exhalation, and it is not very uncommon for small bits of food to be inhaled. The
fact that we can inhale food, unlike most mammals, is connected with the needs
not of our respiratory system, nor of our eating system, but of our speech
apparatus (Lieberman, 1984), illustrating again how nested the functions of
everyday action are. Overly hot items, or unpleasant ones must be avoided (and
desirable items sought out). To bring food to the mouth requires the following of
several of Gibson's (1979,chapter 13) 'rules for the control of locomotion and
manipulation'. One has to approach the object in a controlled way, so as to get
within reaching distance (p. 231) and then reach, grasp and aim it not at another
object, but at the self, specifically at the mouth (pp. 234-235).Reversing this will
withdraw the hand or utensil to the table for repetition or rest. Social norms must
also be followed: one has to look appropriate (head up and with eye contact in
most, but not all, cultures), talk at the correct time, eat what one is allowed, and
share according to custom.
movements are interrelated (Marteniuk, 1978); and breathing and balancing are
coordinately interphased. The standard accounts of motor skills become
incredibly cumbersome when applied to simple everyday acts. In addition to
having to postulate an immensely complex network of motor programmes,
reflexes, and sensorimotor feedback circuits, all somehow adapted to each other
by cognitive and motivational mechanisms, one also has to postulate an inter-
connected series of postural adjustments which are also adaptively coordinated.
For example, how would the programme for postural compensation prior to hand
raising 'know' to change its intensity and direction as an eater looks around the
table or stifles a belch? Traditional motor control research and theory, developed to
explain the displacements of single limbs in space, is simply irrelevant to the
movements and postures of actors in the environment.
The theory of action systems begins with functional acts in the environment. Its
basic concepts are drawn at the level of actions of the organism, not
displacements of parts of the body, or movements of muscle. However, as yet we
know very little about the action cycles of even a single action system, and there
have been no studies whatsoever devoted 'to the rules by which people nest several
action cycles into functional basic acts. We do not even know the parameters of
variation of most of the basic human acts, nor how these are developed in
individuals, or recovered by patients who have lost some motor function. There is
an immense amount of research crying out to be performed. The list of systems
and basic acts in Table 2 can be considered a tentative agenda for developing a
practical theory of action.
The use of tools is fimdamental to human life, and has been for at least two
million years. Humans are by no means the only animals that use or make tools
(Beck, 19801, but we have certainly evolved the largest number and most
sophisticated specializations of tool use in the animal kingdom. Hand tools like
the axe, knife and needle, and tools for transport of material, like the bag and the
jar, are almost indispensable for any human society (Campbell, 1984). It is thus a
r e d indictment of psychology that the analysis of tool use is still in a most
rudimentary stage.
Tools are very special objects in the environment. They are objects that can be
temporarily attached to our bodies, so as to increase our capacity for action
(Gibson, 1979, p. 40). The making of tools involves modifying the substantial
74 E.S. Reed
the role of (indirect) perceptual learning, and the possibility that the use of tools
serving different functions, such as tools for locomoting (tricycles), eating
(spoons), trace-making (crayons) and communicating (telephones), are learned in
different ways.
The traditional theories of motor control must take some of the blame for our
lack of understanding of the psychology of tool use. The concepts and models we
have been using to explain motor skill simply cannot be applied to tool use.
Ironically, it is a commonplace to see experimenters requiring subjects to acquire
a tool use skill in their study (e.g., controlling a joy stick or lever, or pushing a
button to a signal) without even considering the complex perceptuo-motor factors
involved. The much heralded cognitive revolution in psychology began with a
simplistic integration of the concept of feedback into S-R theory in the 'TOTE'
model of action (Miller, Galanter & Pribram, 1960). This model was supposed to
help explain complex cognitive skilled action, such as planning and tool use, but I
have yet to find anyone who was able to use this model to reveal anything about the
psychology of tools. Could it be that such issues as open vs. closed loop, central vs.
peripheral control, are irrelevant to understanding the human use of tools?
Action systems theory suggests that tool use evolved out of ways of enhancing the
basic acts of our human action systems. A tool is not only an object attached to the
body, but it is also an object inserted into an action cycle. Drillis (1963)illustrates
how the manipulative system's basic movements presage a number of important
hand tools (Figure 5). Moreover, the constraints imposed on tools by our postures
and movements have affected tool design. Most hand tools found in peasant
cultures will show some evidence of being the right size, shape, and weight for
relative ease of use. Drillis (1963)argues that this proves that anatomical and
biomechanical factors constrain the building of tools (see Figure 6).
More recently, Warren (1985)has challenged this interpretation, arguing that it
is 'action capacities' - the movements and postures that comprise what I call
-
basic acts that constrain tool design. For example, it is not the shape of the hand,
nor even the biomechanics of the hand-arm system that should be used in
designing handles for cups and pitchers. A pitcher is heavier than a cup, and one
can therefore pick up a cup placed hrther away from one, and with a smaller
grip area than for a pitcher. As Warren rightly emphasizes, there is no such
thing as a 'reaching space' because the size and shape of the region in which one
can manipulate an object depends on the task and the nature of the objects. This is
an important area for study, with practical consequences as well as the
theoretically important value of testing my claims that movements are not
displacementsin space, but changes in an environmental relationship.
76 E.S. Reed
Joining ,
cuttjng
Figure 5 (above). The actwns of the hand as a tool (adapted from Drillis,
1963, p. 428).Although each of these actions are shown in a single pattern,
all of them can in fact be realized by a multiplicity of movement patterns.
For example, one can grasp a tube as easily horizontally as vertically
(shown). I n none of these cases are the actions defined by the movement
patterns.
Figure 6 (right). Tools and body scale (adapted from Drillis, 1963, p. 430).
Note that the systematic use of body-scaled measurement necessarily
implies non-systematicity vis a vis any metric system (see Kula, 1986, for
discussion). Note also that body scale patterns indicate constraints on tool
design, but do not determine the design.
Applying the Theory of Action Systems 77
I \ European Pxe :
I
I 6 Hands :
h e r i c a n Axe I
i:
I
78 E.S. Reed
A knife which has to cut all sorts of things can be of almost any shape;
whilst a tool for some particular purpose must be of some particular shape
(Darwin, 1964, 149; originally 1859).
Once such standardization arose, it became possible to accomplish the same task
with a machine:
If skilled action is made up of displacements of the body and its parts, then
learning a skill as an adult, or developing skills as a child requires the acquisition
of specific neuromuscular units of movement, along with the motivational and
cognitive abilities to control and constrain those movements. On the other hand, if
skilled actions are comprised of ecologically oriented postures and the movements
that change them in the service of particular functions, then skill learning
involves the acquisition and coordination of a number of basic acts for each of the
several human action systems. Learning a skill on this view becomes what
Bernstein (1967;Whiting, 1984) called 'practice without repetition' (see Whiting,
1980): it is not the acquisition of a pattern of movements, but a h c t i o n a l
organization of whatever postures and movements serve to get the job done 9 . The
Robert Woodworth (1906,pp. 374375) knew this long ago: "the great mqiority of purposive
movements are executed for the sake of some effect they produce beyond the mere movement ... . If I
wish to cut a stick, my intention is not that of making certain back and forth movements of my
arm ... my intention is to cut that stick. When I voluntarily start to walk, my intention is not that of
alternately moving my legs in a certain manner; my will is directed toward fcontinws)
80 E.S. Reed
exceptions to this are partial but important: in dance, some sports (e.g.,
gymnastics), and some music skills, part of the definition of the skill is to 'look
right' - i.e., to make movements that are judged to adequately match an ideal. In
this case, learning a skill is learning to make a particular movement pattern.
But this is a pattern as seen by an observer, and it is debatable whether human
observers judge movements in terms of displacements (Gibson, 1979). It is not
necessarily true, even in these skills just mentioned, that the actor learns to
displace his or her body in a certain way. The claim that skill learning involves
learning to move one's body according to a spatial pattern rests on the
unexamined premise that animate movements are displacements in space in the
first place. This is to confound the goal of a foreman on an assembly line with the
goals of purposive agents. Even in the use of a standardized tool, most people do
not make repetitive body movements.
The theory of action systems, if it turns out to be even partially true, will upset
most of this traditional applecart. It denies that animate movements are made up
of displacements of the body, or even of responses to stimuli. Actions are
intrinsically motivated ways of using environmental resources. What counts are
the resources used - not the physical movements, but their biological meaning.
Learning a skill is learning how to use the affordances of the environment, not
how to move one's body. A whole new set of theoretical issues and new areas for
study is thus opened up. Only if these new issues and problems pay off in terms of
research and application will action systems theory prove its mettle. As it stands
now, the theory needs to be tested in most areas. But, in testing it, many
previously unexamined assumptions and hypotheses in the area of motor skills
will be exposed to empirical test, and that should be a positive result, whatever the
conclusions of the testa.
It is a sad fact that the bulk of existing scientific research on motor skills is
simply irrelevant to the tasks of everyday life, to the problems of children learning
skills, or adults learning new skills or specializing in certain skills, or to patients
who, due to trauma, may need to re-learn previously k n o w n skills. The issues we
researchers have been arguing about - central versus peripheral control (Reed,
.
reaching a certain place ... It may conceivably be different with the infant, but this is not probable
... . ...
His movements become organized with reference to situations his motor development is a
process of getting control of the things around him rather than of learning and applying his own
possibilities in the way of bodily movement."I can find no one else who has followed Woodworth's
remarkable insight.
Applying the Theory of Action Systems 81
Some of the issues I have touched upon herein that should be discussed further
are:
Do neuromuscular units underlie basic acts? (Are basic acts mechanically or
functionally specific?)
Can basic acts be characterized empirically in terms of ecologically-defined
postures and movements?
Is the process of action development (the acquisition of basic acts and their
differentiation and integration into action cycles) different in kind from the
learning of skills in later life (the nesting of basic acts to accomplish new tasks)?
Are motor skills whose goal is to achieve a 'movement pattern' (e.g., dance)
intrinsically different from other skills?
What counts as practice of a skill? Are bodily displacements repeated during
practice? Can postures and movements be practiced beneficially independent of
achieving their typical functions?
What is involved in using t o learn tools to extend one's action and perception? Is
the process of 'attaching tools to the self' so they become 'extensions of the body'
one of perceptual or action learning, or both?
Can 'task spaces' be defined objectively and independently of 'movement
spaces'? Is, for example, 'pouring space' really a component of something
meaningfid called 'reaching space'? If task spaces cannot be defined
independently of movement spaces is there any empirical content to my claim that
movements and postures can be studied ecologically, not as physical
displacements?
Are action patterns different when the agent obtains his or her o w n stimulation
ke., adjusts voluntarily to the circumstances) as opposed to using imposed
stimulation (i.e., acting under experimenter-inducedconstraints)?
82 E.S. Reed
ACKNOWLEDGEMENTS
The author of this paper was supported by a fellowship from the National
Endowment for the Humanities #FA-24240-84during its preparation.
REFERENCES
Daly, P., Preston, C.B. & Evans, W.G. (1982).Postural response of the head to bite
opening in adult males. Genetika, 18, 157-160.
Darwin, C.R. (1964). The Origin of Species by Means of Natural Selection.
Cambridge, MA: Harvard University Press, 1964;first edition 1859,London:
Murray.
Diener, H.C., Bootz, F., Dichgans J. & Bruzek, W. (1983).Variability of postural
"reflexes"in humans. Experimental Brain Research, 52, 423-428.
Drillis, RJ.(1963).Folk norms and biomechanics. Human Factors, 5, 427-441.
Druz, W.S. & Sharp, J. (1981). Activity of respiratory muscles in upright and
recumbant humans. Jounuzl ofApplied Physiology, 51, 1552-1561.
El'ner, A. N. (1973). Possibilities of correcting the urgent voluntary movements
and the associated postural activity of human muscles. Biophysics, 18, 907-
911.
Fearing, F. (1971).Reflex Action. Cambridge, MA. MIT Press;first edition 1929.
Nashner, L.N. & Forssberg, H. (in press). The phase dependent organization of
postural adjustments associated with arm movements while walking.
Journal of Neurophysiology.
Friedli, W.G., Hallett, M. & Simon, S R. (1984). Postural adjustments associated
with rapid voluntary arm movements I. Electromyographic data. Journal
of Neurology, Neurosurgery, and Psychiatry, 47, 611-622.
Gallistel, C. R. (1980).The Organization ofAction. Hillsdale, N J Erlbaum.
Gentile, A., Higgins, J., Miller, J. & Rosen, D. (1975). The structure of motor
tasks. In C. Bard, M. Fleury & J. Salmela (Eds.1, Actes du 7e Symposium
en Apprentbage Psych-moteur et Psychologie du Sport, Oct. 1975 (pp. 3-
12).Quebec: L'Association des Professionels de l'Activit4 Physique.
Gibson, J.J. (1966). The Senses Considered as Perceptual Systems. Boston:
Houghton MiBlin.
Gibson, J.J. (1979). The Ecological Approach to Visual Perception. Boston:
Houghton Miillin.
Gibson, J.J. & Y01188,P. (1968). A new theory of scribbling and drawing in
children. In The Analysis of Reading Skill. Final Report, Project No 5-1213,
Cornell University & U. S. Office of Education, Ithaca, NY,pp. 355-370.
Gonshor, A. & Melvill Jones, G. (1980). Postural adaptation to prolonged optical
reversal of vision in man. Brain Research, 192, 239-248.
Greene, P.H. (1972).Problems of organization of motor systems. In R. Rosen & F.
Snell (Eds.), Progress in Theoretical Biology, Vol. 2. New York Academic
h S 8 .
Gurfinkel, V.S. (1973). Muscle afferentation and postural control in man.
Aggressologie, 14, 1-8.
Gurfinkel, V., Kob, Y., Paltsev, Y. & Fel'dman, A.G. (1971). The compensation of
respiratory disturbances of the erect posture of man as an example of the
organization of interarticular interactions. 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. 382-395)
Cambridge: MIT Press.
Henry, F. (1953). Dynamic kinesthetic perception and adjustment. Research
Quarterly, 24, 176-187.
84 E.S. Reed
Edward S. Reed
Complex Movement Behaviour:'The' motor-action controversy,pp. 87-120
O.G. Meijer & K Roth (editors)
Q Elsevier Science Publishers B.V. (North-Holland),1988
Chapter 3
SUMMARY
With respect > motor control and learning two theoretical appmaches are
distinguished. One, the motor systems approach, is characterized by the use of
information processing models and hypothetical constructs framed in cognitive
terms. Such models and constructs are rGected by proponents of the second
approach. The latter, action systems approach, is inspired by GibsonS work. Two
kinds of action systems theories have been developed, a 'Gibsonian' one (e.g.,
Reed), and a 'Neo-Gibsonian' one (e.g., Kugler). The Gibsonians try to explain
motor behaviour as a function of information specifying the environment, the Neo-
Gibsonians resort to physical principles guiding the behaviour of energy-
consuming open biological systems.
In discussing Gibson's ecological psychology attention is drawn to its strong
behavwuristic (Skinnerian) overtones. Analogies between Gibson's and Skinner's
theorizing are pointed out, in particular with regard to their rejection of
*mentalistic' constructs and their scepticism about neurophyswlogical
explanations of behaviour. It is argued that the significance of Gibsonian types of
action theories, in sticking to the ecological level of description and in eschewing
explanations in terms of processes at other (psychological or physiological) levels
of description, is limited to situations exhibiting a perfect fit between organism
and environment. Consequently the Gibsonian approach falls short in accounting
for human motor learning processes, where, at least initially, such a perfect fit
does not exist.
Although the explanatory power of the Neo-Gibsonian theories is judged to be
greater than that of the Gibsonian ones, the former also seem to be incapable of
addressing human learning i n a fruitful way. Therefore, action systems theories
leave many phenomena unexplained, phenomena which are addressable (and
are indeed discussed) in motor systems theories.
88 P.C.W. van Wieringen
1. Introduction
The main topic of this book has to do with the relative merits of what have been
called 'motor systems theories' and 'action systems theories' in explaining motor
control and learning. Although the difference between the two types of theories is
not in all respects clear, and both of them in themselves encompass different
viewpoints, some global indication of the aforementioned difference can be given.
Motor systems theories, then, may be loosely described as information processing
theories making use of constructs like 'programmes', 'schemata', 'memories',
'images', etc. Action systems theories, on the other hand, dismiss the former
approach as an incorrect application of the computer metaphor to biological
organisms. Instead, the complementarity and reciprocity between the organism
and its environment is drawn to the foreground, and attempts t o explain
behaviour do justice to this intimate and non-contingent relationship.
Of course, the differences between the two types of theories, or approaches, may
be evaluated from a variety of viewpoints, philosophical as well as scientific. The
one taken here refers to the level of explanation (the level of description chosen for
explaining the behaviour being studied) as well as to the closely related topic of the
kind of explanation being favoured.
Because the motor systems approach might be considered the 'classical' and
'established one in the motor domain, the accent in this paper will be on the
relatively new action systems approach. However, comparisons between these
approaches will, both implicitly and explicitly,be made throughout the paper.
Where proponents of the (diverse varieties 00 the action systems approach
explicitly pay tribute to Gibson's ecological theory, which found its final
expression in The Ecological Approach to Visual Perception (1979), it seems
apposite to start the inquiry with an evaluation of Gibson's way Qf theorizing. In
Section 2 behaviouristic tendencies in Gibson's work are signalled. In Section 3 it
is argued that the change he made from S-Rpsychology to ecological psychology
did not change his behaviouristic inclinations as evidenced by his enduring
rejection of explanations making use of concepts defined at other levels than the
one at which behaviour is described. In both Section 2 and 3 Gibson's position is
elucidated by comparing central features of his theory with key notions from the
Kinds and Levels of Explanation 89
work of Skinner. This comparison makes clear that Fodor and Pylyshyn's (1981)
intimation that Gibson is the 'Skinner of perception' has a stronger justification
than Turvey et al(1981,footnote pp. 238-239)are willing to admit.
Sections 4 and 5 are devoted to an attempt to show the relevance of different
levels of description for explaining behaviour. Section 4 is concerned with the
neurophysiological level, Section 5 with psychological levels related to hypothetical
constructs, including mental representations. The Neo-Gibsonian type of action
systems theory, accounting for perception and action by extending physical
explanations to the ecological level, is addressed in Section 6. Shortcomings in
both the Gibsonian and Neo-Gibsonian approach are illustrated in Section 7 by
showing that neither of them comes to grips with the phenomenon of human
learning in the domain of perception and action.
In Section 8 the conclusion is drawn that motor systems theories, albeit hybrid
in character, are indispensable if it is wished to explain more than a very
restricted subset of relevant data.
In the fifties and early sixties important developments in Gibson's theorizing took
place. He rejected Gestalt psychology in its attempts to explain perception (both
veridical perception and illusions) by a hypothetical process of sensory
organization (Gibson, 1959, p. 459). However, he also criticized neo-behaviourism
for its making use of hypothetical constructs as intervening variables between
stimulus and response, characterizing his own theory as follows:
The bare logic of the theory is extremely simple: there are dependent
variables of perception and there are independent variables of stimulation.
No intervening variables are presumed (Gibson, 1959,498).
In this regard he takes the same firm stand as Skinner, declaring that his theory
of direct perception and its related 'ecological physics', if successful,
Some years later, Gibson explicitly rejected the S R formula not only because of
its use of intervening variables, but also because of its conception of stimuli and
responses:
90 P.C.W. van Wieringen
The good features of behaviorism are lost in the retreat of psychology into
the old concern with sensations and images. The S-Rformula has failed but
the emphasis on action and adjustment is still valid. The explanation of
perception, accordingly, should be functional. The acceptance of inputs and
images actually goes hand in hand with the belief in discrete stimuli and
countable responses. Both are out of date (Gibson, 198213, 373; originally
1963.)
Those who believe that we see copies of the world may contend that we never
see the world itself, but it is at least equally plausible to say that we never
.
see anything else . .
and he refers to Theophrastus who formulated the basic difficulty of the 'copy
theory' of perception more than two thousand years ago:
It should be noted that the same argument has been used by Gibson against
mental representation:
We certainly do not summon up pictures inside our head for they would
have to be looked at by a little man in the head ... . Moreover, the little man
would have eyes in his head to see with and then a still littler man and so
ad infiniturn (Gibson, 1974,42, italics in original).
Later in this paper it will be argued that both Skinner's and Gibson's ideas of
mental representation are inadequate, and that, moreover, the 'horror
homunculi' is not as justified as it may seem to be. But let us first continue
quoting Skinner:
In an operant analysis, and in the radical behaviorism built upon it, the
environment stays where it is and where it has always been - outside the
body (1974, 82,italics in original).
Kinds and Levels of Explanation 91
... the control exerted by stimuli does not mean that he [a person] infers
what exists in the world around him ...
and even in the case of visual illusions one should not, according to Skinner, say
that the mind infers and predicts reality from incomplete data, but that:
These citations, and chapter 5 of Skinner's 1974 book might be consulted for more,
show that there are at least important similarities between Skinner's and
Gibson's ideas about perception.
It may also be clear by now that Gibson's dismissal of intervening variables was
related very closely to his theory of direct perception. This relationship formed the
topic of a recent discussion between Reed and Katz. Reed stated, and justly so, that
the difference between direct and indirect perception is based on the information
specifying the environment. From this he concluded that the way this information
is processed is no constitutive part of Gibson's theory of direct perception (Reed,
1983,pp. 6-7).Katz, in a reply (1984,pp. 4-51,admits that Gibson indeed defines
direct perception in the way sketched by Reed, but he reiterates that Gibson also
uses it in the sense of perception not mediated by mental representations. A clear
example of this usage, referred to by Katz, may be found in Gibson's last book
(1979,p. 1661,where he describes his theory of direct perception as the opposite of
the theory of mediated perception, the latter theory being characterized by
asserting "that perception is mediated by assumptions, preconceptions,
expectations, mental images, or any of a dozen other hypothetical mediators."
Katz states that for Gibson an information processing approach, far from being
considered compatible with direct perception, is intrinsically rejected, and that,
moreover:
In the light of the discussion in this Section, I think that Katz is quite right in
his appraisal. Of course that should not surprise us too much since, after all,
92 P.C.W. van Wieringen
and I think it is more than accidental that there is talk about "conditioned-
response behaviorism" here, and not about behaviourism per se. Now it must be
admitted, of course, that Gibson's development from S-R psychologist to an
ecological psychologist, even although both may be described as behaviouristic, is
far from trivial. We address this change in the next Section.
theory of perceiving (Reed, 1980, p. 3971, he agrees that he had trouble, himself, in
getting free of the 'baneful influence' of the concept of the stimulus which might
have confused his readers (Gibson, 1982e, p. 399 - originally 1976). However that
may be, in 1966 the stimulus (in which, according to Gibson, lay the severest
weakness of S R psychology) was replaced by 'perceptual information', while later
'postures' and 'movements' were to take the place of responses (Gibson, 1982d, pp.
388-392 - originally 1975).
The ambiguous phrase 'perception as a function of stimulation' (1959) - the
ambiguity leading to possible confusions between energetic and informational
aspects of stimuli - made way, in later works, for perception as a function of
stimulus information. 'Information' was defined not in Shannon and Weaver's
(1949) sense, but in terms of invariants in the optic flow field underlying change
and specifying affordances, i.e., behaviourd possibilities which are a joint
function of the organism and its environment, bridging the gap between the two
(Gibson,1979,pp. 141-142).
The accent in Gibson's work is on the analysis of the 'environmental aspect' of
affordances, and its specification in the optic array, or, in other words, on the
development of ecological optics: a description of the structure of ambient light
which is especially suited to the behaving organism and its ability to extract ('pick-
up') information specifying the affordances. The 'organismic' aspect of
affordances, the way the organism picks up (and learns to pick up) the higher
order invariants in question, is scarcely considered, although it is made clear that
both genetic endowment and learning processes play their role in 'attuning' the
organism to the information. On the one hand
... the basic afTordances of the environment are perceivable and are usually
perceivable directly, without an excessive amount of learning ...(1979, 143)
on the other:
Constraints on the side of the organism are not mentioned. One might just as
well, however, change the focus in proclaiming that perception is wholly
constrained by the organism, taking the environmental constraints for granted.
In this respect I might refer to some remarks of Pribram to the effect that if
Gibson is right in postulating an infinite amount of information in the world:
... there must be constraints determining what an organism can pick up.
But many theorists take the exact opposite view from Gibson and say that
all the constraints are inside the organism. Unlike Gibson and those
extreme constructionists, I think there are constraints both in the
environment and the organism and the brain is special because it
constrains what we can know (Pribmm, in Weimer & Palermo, 1982,233).
It is interesting to note that both Skinner and Gibson seem to deny a special
status to the brain. Skinner notices a current practice of avoiding dualism by
substituting 'brain' for 'mind:
The brain is said to use data, make hypotheses, make choices and so on, as
the mind was once said to have done. In a behavioristic account it is the
person who does these things (1974,86).
The world does not telegraph the brain and the brain does not telegraph the
muscles; only a whole man sends telegrams. The brain is not a receiver nor
a sender, not a homunculus but only an organ (1982e. 399; originally 1976).
In both Skinner's and Gibson's work references to the brain (or central nervous
system) are all but lacking and both have offered pragmatic reasons for this
omission. Skinner states that if a complete description of both observable
behaviour and physiological processes were to be possible:
The organism would be seen to be a unitary system, its behavior clearly part
of its physiology ... [but, he continues:]
A t the moment however, physiological techniques are limited, and we
cannot study everything at once (1969,282).
It will have been noted that the system presented omits reference to the
nervous system. This is because our knowledge of perceptual
neurophysiology, as distinguished from sensory neurophysiology, is very
scant ... . The implication of the present theory is that neural mechanisms
can be discovered to explain how the system "tunes in" on the variables of
stimulation, at first those of simple order and later those of increasingly
higher order (1959,497-498).
In his later works Gibson does not dwell upon these neural mechanisms, but
simply talks in t e r n of alterations in the state of the organism:
... the theory of information pick up does not need memory ... . An image of
the past, if experienced at all, would be only an accidental symptom of the
altered state (1979,254).
And here Gibson's view is very close to Skinner's, the latter noticing that:
The contingencies which affect an organism are not stored by it. They are
never inside it; they simply change it. As a result, the organism behaves in
special ways under special kinds of stimulus control ... [and that:]
-
The computar is a bad model as bad as the clay tablets on which the
metaphor was probably first used (1974, 122).
In conclusion it may be said that neither Skinner nor Gibson have fundamental
objections to neurophysiological theorizing, but both of them deny that there is an
interesting and non-question begging level of description between the motor
behavioural level and neurophysiology. In practice both Skinner and Gibson, and
that is a crucial similarity in the light of the approach we take in evaluating
differences between motor and action systems theories, explain behaviour at the
same level as that of the phenomena to be explained. Both Skinner's (very loosely
defined) molar behavioural level of stimuli and responses, and Gibson's (more
sophisticatedly defined) ecological level are close to the level of everyday
psychological experience, Gibson's even more so than Skinner's. No analysis in
terms of lower level entities or processes is deemed fruitll.
Now, of course, such a way of theorizing has its clear methodological
advantages. Problems of commensurability between concepts being used and
behaviour to be explained will not arise. On the other hand, a price has to be paid,
because the explanations will always be limited. This has to do with the fact that
mechanisms underlying the obserued correlations at the molar level (between S
and R by Skinner, between perceptual information and movements and postures
by Gibson - and I may add, here, Reed) are left out of consideration. All kinds of
disturbances of these lower level mechanisms may came a break down of the
correlations observed a t the molar level, break downs which are not
comprehensible without an insight into the aforementioned mechanisms. They
do, of course, also play a role in improved correlations between behaviour and
environment at the molar level.
Just in passing I'd like to stress that evolutionary accounts of behaviour cannot
act as a substitute for such mechanisms. Evolutionary explanations are very close
to reinforcement interpretations of learning, as several authors have noticed (e.g.,
Dennett, 1979, pp. 75-76;Miller, 1978, p. 77; Pringle, 1956). Such 'ultimate
explanations', as they are called by Lea (1984,p. 12),clearly cannot function as
mechanisms underlying 'proximate explanations' in terms of observed
relationships between behaviour and environment. This is not to deny their value
in providing insight into behaviour but, to point out that, like the perfectly valid
explanations of behaviour favoured by Skinner and Gibson, they are also
explanationsin a limited sense.
In stating explicitly that I see Skinner's and Gibson's accounts of behaviour as
explanatory (and I allude to scientific explanations here) I am more liberal than
is Skinner with respect to his own work. He always emphasized that his analysis
Kinds and Levels of Explanation 97
4. Relevant neurophysiology
If one realizes that the human brain consists of about 1011 neurons with 1015
connections, addressing it as just an organ', although logically impeccable,
seems to foster some inadequate connotations. The mere number of neurons,
which most probably are modularly organized in basic units of about 110 neurons
with similar intrinsic connections (Philips, Zeki & Barlow, 1984; Paillard, 19861,
makes the human CNS an extremely powerful device for information processing.
Paillard (1986) stipulates that in the early stage of ontogenetic development the
human neocortex differs from the homologous structure of more primitive
animals only with respect to the greater number of redundant, modularly
arranged neurons, part of which only specializes during ontogenyl.
Some experimental evidence which might suggest that the mere number of brain cells is of
crucial importance has been reported by Bresler and Bitterman (1969). They transplanted brain
material into the tectal region of the embryo mouth-breeding fish, in this way increasing the size
of this region in the adult fish. The latter showed improved learning performance in a
discrimination task a$ compared to its normal congenitors.
98 P.C.W. van Wieringen
... could account for the astonishing capacity of all nervous systems to
extract regularity and invariant or covariant features of their changing
surroundings. It provides living organisms with the finely tuned adaptivity
that is required for survival in an adverse environment. It has to be
considered as a basic primitive function of neural networks.
We should also not forget that within such networks pattern-specific individual
cells ('feature extractors') are functioning. Well-known instances of such cells are
the 'simple', 'complex' and 'hypercomplex' cells described by Hubel and Wiesel
(1962).These cells respond best, for example, to, respectively, a contour with a
particular orientation, a contour with a particular orientation moving in a
particular direction, and a contour with a particular orientation moving in a
particular direction being a particular number of degrees of visual angle in
length. Hubel and Wiesel (1968) also described colour-specific cells. Other
examples of feature extractors are cells selectively responding to three-
dimensional objects in the colliculus inferior of monkeys (Updyke, 19741, cells
selectively responding to primate's hands in the inferotemporal region of
monkeys (Gross, Rosha-Miranda & Bender, 19721,and cells selectively responding
to particular spatial properties (De Valois & De Valois, 1980).
Neural networks are also involved in regulating movements and postures
comprising actions. We know about such networks being related to locomotion in
several species. In some organisms the organization of such networks has been
shown to be relatively simple. Goldschmeding (19771,for example, studied the
neural control of eating movements in the pool snail G y m m a stagalis). The CNS
of the latter organism comprises 17,000 neurons. The eating movements (mouth
-
opening tongue protrusion - forward movement of tongue over surface - tongue
withdrawal - mouth closing) are realized by 46 muscles in the buccal mass,
control being exerted by two buccal ganglia. Groups of interneurons, because of
their interconnections, were shown to be acting as one activating cell on the motor
neurons innervating all muscles involved in one specific combination (sub-
movement). The motor neurons in question are also interconnected, each one of
them controlling several agonist muscles. Feedback is not necessary, but may be
incorporated.
Kinds and Levels of Explanation 99
... the nerve-circuit of the muscles carries within itself the principle of
motor coordination, when its central or cellular position - a true locomotor
centre - is brought into play by an excitation (Chauveau, 1891,1741.
Delgado (1971) reports experiments with cats and monkeys, showing that
electrical stimulation of the brain (ESB) evokes not only simple responses, but
complex and well-organized behaviour as well, behaviour which may be
indistinguishable from spontaneous activity. Moreover:
... evoked and spontaneous movements influenced each other, and the final
response was a combination of both (Delgado, 1971,104).
There are very few clinical reports of complex movements evoked by ESB
which are comparable to the sequential responses observed in monkeys,
and this may indicate that cerebral organization is less stereotyped in man
than in animals (1971,116).
100 P.C.W. van Wieringen
Massion and his co-workers (1979, referred to by Paillard, 1986) showed that
forelimb liRing in the cat was anticipated by postural adjustments, both in the
case of voluntary movement and in the case of the limb movement being elicited by
electrical stimulation of the corresponding motor area. This is strong evidence for
the existence of feedforward signals being responsible for the anticipatory
postural adjustment.
Some changes in neural control of movement related to the process of learning
will be addressed later. The material being presented in this Section, however,
may be convincing enough to qualify approaches to motor control which totally
neglect CNS activity as being rather incomplete,to say the least.
Realizing the tremendous complexity of neural circuitry in human beings I
have strong reservations about the apparent implications of Reed's contribution to
this book, when he illustrates his method of analysis with an example of the act of
pouring. He stresses the fact that pouring is not a simple artificial laboratory
task, the latter type of task being used, so he says, by people accepting the
traditional theory of skills 'as constructed out of mechanical neurobehavioural
units'. And he continues:
But I seriously doubt anyone will ever be able to show how an action like
pouring is comprised of hundreds of interacting reflexes, programmes and
.
feedback loops ... The only alternative is to begin the study of action with
an attempt to discover the range of uariution naturally found in an action
80 as to see what various systems are being nested into a coherent act ... .
Note that what our subjects do vary are not responsee, but are functionally
meaningful units of the pouring action: postures and movements,
ecologically considered C&ed, 1988, italics in original).
In the first place, I think that the traditional theory of skills as constructed out of
mechanical neurobehavioural units cannot be dismissed as untrue, if at least
'distributed neural networks' may be taken as synonymous to the 'mechanical
neurobehavioural units'. We should not forget that, without the relevant
blueprints, even analysing 'molar behaviour' of modern computers in terms of
elementary activity of its basic units is almost impossible, which, of course, does
not imply that its 'motor behaviour' cannot be subserved by the activity of
(networks OD very simple units (as in fact it is). In the second place I want to
argue that the different ways of pouring may all be realized by specific activity of
the CNS, the different kinds of pouring having to do with differing inputs to the
CNS (see Horridge, 1973).
In sticking to an analysis of behaviour in terms of sub-acts at the molar level
without trying to describe processes underlying that behaviour at the
neurophysiological level one refrains a t least from important explanations of
Kinds and Levels of Explanation 101
motor controls, and sometimes even robs oneself of ways of making relevant
distinctions with regard to disturbances of action. With regard to the latter, one
may think about the analysis of different forms of apraxia as presented, for
example, by Luria (1973,pp. 34-38).The improvement of motor skills, including
locomotion, in Parkinson patients after treatment with GDopa also points to the
relevance of neurophysiologicalknowledge for understanding human actions.
It remains to be seen whether Fodor (1983,pp. 101-119)is right in his suggestion
that, because of the supposed non-modularity of the central systems, the higher
mental processes (in contrast to input processes), are outside the domain of
classical scientific analyses; the reviews of his book (see Fodor, 1985, for
reference), however, demonstrate that this claim is far from being generally
accepted.
Analytic scientific approaches, on the other hand, are like Nagel's non-
teleologicalexplanations:
2Such explanations may be 'explanations in principle', in the sense this term has been used by
Hayek (1955).
102 P.C.W. van Wieringen
This opinion, far from denying the emergence at higher levels of new properties
which are not reducible to properties of lower level subsystems (the former
properties being dependent on the interrelations and interactions between
subsystems), leads Oatley (1978,p. 311to state that
Now most psychologists, and surely the motor systems theorists, draw into their
analysis systems and processes (schematized as boxes and arrows) which are
conceptualized at levels in between the behavioural molar and the atomic
neurophysiological. Systems and processes at these levels often have to be
characterized as 'hypothetical constructs' which have to be connected to higher
and lower level descriptions in order to be useful. In psychological theories the
intermediate levels are 'reached by both so-called top-down and bottom-up
processes. Following either direction, however, one will be forced into jumping
from psychological ('mentalistic') concepts to neurophysiological ones ('top-down
approach), or from neurophysiological concepts t o psychological ('mentalistic')
ones ('bottom-up approach'), the psychological concepts being constructs like
schema's, memories, images, plans, etc.
For that reason, taxonomies of processes a t different levels will be eclectic. As one
of many possible examples, let us take the taxonomy presented by Uttal (1981 &
1983).He distinguishes six levels (Table 1).
Uttal basically takes up a reductionistic viewpoint with regard to his
classification:
B
Figure-Ground Or anization and
Signal Extraction rocessing
(Prequantitative and unidimen-
Organizationof lshihara
patterns
sional)
Integration and Construction Colour constanc and con-
(Quantitative and multidimen- ;I
trast; The Land henom-
sional) ena; simultaneous con-
trast and metacontrast
-
Subsequent Mental Image Pro- Shephard and Cooper's
-
cessing Mental Rotation
-
1cccc-c"1
Immediate, preattentive
Subsequent, manipulative, attentive, active
Physicalistic, deterministic, neural model
Molar, psychological, rationalistic model
104 P.C. W.van Wieringen
It might seem that a 'neural network language' does about the same job he& as
did Carnap's (1932) earlier physicalist language, the latter language being based
on 'protocol sentences' (statements about space-time points), and meant to realize
the ideal of unified science. This interpretation, however, is not compelling. One
might also interpret Uttal's aforementioned statements as indicating that the
more complex psychological functions are both realized by and realized in neural
network activity (in the way that this has been done by Searle, 1983,pp. 262-2721,
without concluding that these psychological functions are 'nothing but' neural
activity. It should be realized that even a reductive explanation of a phenomenon
is something else than explaining such a phenomenon away. As Dennett (1979,p.
89)aptly puts it:
Against the background sketched in this Section, Hinton's (1980,p. 388) claim
that Gibson, in believing that there is no interesting level of description between
Kinds and Levels of Explanation 105
He also states that "such psychological entities map onto physical events in a
very, very abstract way" (p. 681, this statement being made after having drawn
attention to the analogy between such entities and particles in physics (p. 67).
Conceived in this way, nothing is wrong with using mental representations in
psychological theory, or in stating that the brain embodies in symbolic form the
relationships of the real world (Oatley, 1978, p. 15). For illustration: MacKay's
(1981) 'mental nodes', used to explain effects of mental practice on motor
performance, are just as admissable as electrons or quarks in explaining data in
physics, and excluding such concepts from the allowable set of constructs may
easily lead to leaving the explainable unexplained.
In Section 2 I already suggested that homunculi are less dangerous than Gibson
(and Skinner) take them to be: as long as the regressing homunculi need less
intelligence than the ones for which they are 'working', less and less complicated
behaviour has to be explained, eventually ending:
... with homunculi so stupid (all they have to do is remember whether to say
yes or no when asked) that they can be, as one says, "replaced by a
machine". One discharges fancy homunculi from one's scheme by
organizing armies of such idiots to do the work (Dennett, 1979, 124, italics in
or~inal).
Its subject matter is the formulation of principles that are valid for
"systems" in general, whatever the nature of their component elements
and the relations of "forces"between them ... . In elaborate form it would be
a logico-mathematical discipline, in itself purely formal but applicable to
the various empirical sciences Won Bertalanm, 1968,37).
Neo-GibsoNans3 like Kelso, Turvey and Kugler have a related view; however,
they define their generalized theory as 'physics', a physics which has been
extended in order to apply not only to physical phenomena, but to biological and
psychological ones as well (Kugler, 1986). In applying their theory to motor
behaviour they argue that:
The term 'Nec-Gibsonians' is customary by now, nee for example Reed, Kugler and Shaw (1985,
p. 309).
Kinds and Levels of Explanation 107
We will make the eccentric claim that the natural perspective is grounded
in two necessarily complex themes: The proprietary (explanatory)
principles of physical theory, with their underscoring of tendencies in
dynamics, and the proprietary (ontological and epistemological) tenets of
ecological realism (Kugler, Kelso & Turvey, 1982, 15-16).
In a recent paper (Van Wieringen, 1986) I have tried to show that the 'either-or'
dilemma with regard to programmes or constraints is mistaken, and Paillard
(1986,p. 415) argued in the same strain that "the two explanations are clearly not
mutually exclusive but patently complementary". I will not go into the arguments
related to that matter here and confine myself to referring to the latter two
articles, not without mentioning, however, that even for insect locomotion the Neo-
Gibsonian account does not seem the only possible one, witness a survey by
Graham (1985). In the following Section I will address a very important
phenomenon which, in my view, cannot be explained properly by either
Gibsonians or Neo-Gibsonians, the phenomenon in question being human
learning.
108 P.C.W. van Wieringen
7. Human learning
7.1.Introduction
... it is not one of interaction; it's one of, well, reciprocity is not too bad. A
mutual application and a focus point. There are several terms in the
ecological approach to psychology that bridge the gap between animal and
environment. But such a term that bridges the gap points both ways, like
the concept of &ordance. Another one is the ambient optic array. It's
specified at a point where an observer might stand, a station point, where
you have reflected light impinging upon an observer. So with such
concepts, Z don't have to ask a question about the relation between the
animal and its environment. I've defined it out of existence in your own
way (Gibson, in Weirner 6 Palermo, 1982,234, my italics).
I must confess that Gibson's solution looks a bit too easy to me. It might be
adequate for basic dordances to which organisms have been tuned during
evolution (although even here questions about mechanisms concerned in the
reciprocity might be raised). Let's take Von Uexkiill's famous tick Won Uexkiill &
Kriszat, 1934)as an example. The tick is hanging from a branch; when it detects
butyric acid in the air it releases its grasp, drops down and (unless fortune is
against it) into the fur of the animal passing underneath and secreting the acid.
Mechanical stimulation leads to crawling around, till finally heat is detected and
the tick bores itself into its host's skin. The tick indeed is perfectly attuned to the
perceptual information in its ecological niche and this information, specifying
affordances for, respectively, falling, crawling around, and boring its way down,
wholly constrains its actions.
I do not deny that in higher organisms also - and even in man - basic
affordances may be detected and reacted upon on the basis of evolutionary
development having culminated in 'smart mechanisms' (Runeson, 1977).
However, in many cases the fit between organism and environment is far from
perfect and therefore the organism has to learn to adapt to its environment. So the
concept of 'dordance', far from rendering questions about the relationship
between organism and environment superfluous, leads to a very valid question,
viz. how is it that affordances are learned, or, in other words, how do potential
affordances become actual ones. Notwithstanding Gibson's just cited statement
he has himself addressed this topic, albeit in an unsatisfacory way as I will show
later. Subsequently I will go a little deeper into motor learning. Where the
Kinds and Levels of Explanation 109
Gibsonians have not yet addressed this topic, the emphasis will be on the question
if the Neo-Gibsonian approach, as described in Section 6, can, in principle,
provide a satisfactory account of it. I am of course aware of the fact that
perceptual and motor learning are abstractions from one process (often called
perceptual-motor learning), but for reasons of clarity of exposition I think the
division in subsections devoted to perceptual and motor learning is defendable.
7.2Perceptual learning
Further on he continues:
The child becomes aware of the world by looking around and looking at, by
listening, feeling, Smelling, and tasting, but then she begins to be made
aware of the world as well. She is shown things and told things, and given
models and pictures of things, and then instruments and tools and books,
and finally rules and short cuts for finding out more things. Toys, pictures
and words are aids to perceiving, provided by parents and teachers. They
transmit to the next generation the tricks of the human trade. The labors of
the first perceiver are spared their descendants. The extracting and
abstracting of the invariants that specify the environment are made vastly
easier with these aids of comprehension. But they are not in themselves
knowledge, as we are tempted to think. All they can do is facilitate knowing
by the young (258, italics in original).
What is written here is that toys, pictures and words are not in themselves
knowledge, but can only facilitate knowing. Now of course in a literal sense this is
true; toys, pictures and words are not in themselves knowledge, but, contrary to
what Gibson is stating, they not only facilitate, but they also conuey knowledge.
And it is this conveyed knowledge that makes the extracting and abstracting of
invariants easier, and which facilitates M e r knowing. By the way, Gibson, as
we noticed, defines 'knowing' as an extension of perceiving (p. 2581, but some
pages later he states that "Perceiving is the simplest and best kind of knowing" (p.
2631, seemingly implying that the former form of knowing is limited to explicit
knowing in contrast to the latter form which is tacit. It seems to me that Gibson,
in his effort. to eschew mental representations, landed in a muddier muddle then
the one he wishes to avoid. Let me further illustrate this with a simple, almost
trivial, example.
According to Gibson (1979,p. 139) the affordance of a postbox is perceived when
the postbox is identified as such, and it is apprehended whether the postbox is in
sight or out of sight. In Amsterdam there are besides the red postboxes, similar
boxes, which, however, are blue and meant for mailing transfer accounts. Now
Kinds and Levels of Explanation 111
We 6nd it diflicult to accept that Gibson's formulation does away with any
role for memory in understanding perception. We find it unconvincing to
explain a person returning aRer ten years to their grandparents' home and
seeing that a tree has been cut down as having detected directly an event
specified by transformation in the optic array. At this point, we concur with
Fodor and Pylyshyn (1981)that the terms of Gibson's theory are being
extended to the point of becoming empty (Bruce & Green, 1985,332).
In drawing this subsection to a close I may once more emphasize the similarity,
in doing away with memory, between Skinner and Gibson, a similarity which has
already been addressed in Section 3. It is interesting to note that, where Skinner
(e.g., 1974, p. 82) substitutes the environmental history of the organism (in
combination with its genetic endowment) for memory, Shaw and Todd (1980)
make the same move in showing how Gibson can do without memory. They
propose t h a t
7.3.Motor learning
It should be clear, however, that many of the motor skills being learned by
humans are more adequately referred to as 'cultural' than as 'natural' activities,
and that the gradual emergence of such skills cannot be explained by 'free
interplay of forces'. In learning such skills, existing coordinative structures have
to be 'overridden', often at the cost of high energy expenditure, and new ones have
to be built up. Such learning processes have been traditionally described by motor
systems theorists as evolving from a cognitive stage to an autonomous or
automated stage (see, for example, Fitts & Posner, 1967). The early stages of skill
learning are characterized by a non-perfect attunement of the organism and its
environment (which may include a horizontal bar or a pair of parallel bars in
gymnastics, or consist of a ballet studio in dance). Here the affordances of the
Kinds and Levels of Explanation 113
... one in which the learner actively explores a task's potential energy
function in order to discover its topology and identify its extrema. Learning
(from the learner's perspective) is a problem of becoming sensitive to
information carried in the gradients and equilibrium points of potential
surfaces (italics in original).
Without denying the adequacy of this image and the potential value of the
suggested approaoh to the learning process with regard to relatively simple motor
tasks, I think the approach falls short of shedding light on the way complex skills
like ballet movements are mastered. The latter process is better described by
taking into serious account that cognitive guidance plays an important role, and
that movements are, at least partially, based on mental representations of the task
to be performed. In this respect it may be worthwile to remind the reader of
Bandura's (1976,p. 7)claim with regard to delayed imitation:
... the absent modelled events must be internally represented, and factors
such as symbolic transformations and cognitive organisation of modelling
stimuli and event rehearsal, which facilitate retention of acquired contents,
serve as determinants of observational learning.
motor learning. In agreement with this conception of the learning process two
levels of information processing have been suggested by Paillard (1986, p. 432):
One concerns the direct dialogue that the organism entertains with the part
of the physical world to which it is attuned by virtue of its sensori-motor
apparatus. The other is related to cognitive activities that operate on mental
representations of physical reality embodied in memory stores. The general
idea is that progressive evolution of the cognitive apparatus has created
new control systems of action that enlarge the adaptive capacity of the
organism but without, at the same time, undermining the more primitive
and far more economic mode of automatic control that characterises the
sensorimotor level.
8. Conclusion
environment is imperfect Gibson's approach does not bring us very far. Here the
importance of 'internal loops' in regulating behaviour increases relative to the
'external loops' (Shepard, 1984; Shebilske, 19851, and here Gibson's behaviouristic
dislike of internal constructs is in the way of explaining the behaviour in question.
A reconciliation of the ecological approach with an approach using internal
constructs (including mental representations) as favoured for example by
Shepard (1984) and Strelow (1985) seems promising and can account for situations
where the ecological complementarity between organism and environment has no
longer, or has not yet, been realized.
ACKNOWLEDGEMENTS
Thanks are due to P.J. Beek, H.T.A. Whiting, two anonymous reviewers, and the
Editors of this Volume for their comments on a n earlier draR of this paper.
REFERENCES
Bandura, A. (1976).Social learning theory. In J.T. Spence, R.C. Carson & J.W.
Thibaut (Eds.), Behavioral Approaches to Thempy (pp. 1-46). Morristown:
General Learning Press.
Bresler, D.E. & Bitterman, M.E. (1969).Learning in fish with transplanted brain
tissue. Science, 163, 590-592.
Bruce, V. & Green, P. (1985). Visual Perceptwn: Physiology, psychology and
ecology. Hillsdale, N.J.: Erlbaum.
Calis, G. (1984). Zen en kijken. In G.A.M. Kempen & Ch. Sprangers (Eds.),
Kennis, Mens en Computer (pp. 117-124).Lisse: Swetz & Zeitlinger.
Carello, C., Turvey, M.T., Kugler, P.N. & Shaw, R.E. (1984).Inadequacies of the
computer metaphor. In M.S. Gazzaniga (Ed.), Handbook of Cognitive
Neuroscience (pp. 229-248).New York Plenum.
Carnap, R. (1931). Die physikalische Sprache als Universalsprache der
Wissenschaft. Erkenntnis, 2, (423-465).
Carnap, R. (1966).Philosophical Foundutions of Physics. New York Basic Books.
Chauveau, A. (1891).On the sensorimotor nervecircuit of muscles. Bmin, 14, 145-
178.
Delgado, J.M.R. (1969).Physical Control of the Mind. New York Harper & Row.
Dennett, D.C. (1979).Brainstorms: Philosophical essays on mind and psychology.
Hassock, SUSS~X: Harvester Press.
De Valois, R. & De Valois, K. (1980).Spatial vision. Annual Review of Psychology,
31. 309-341.
Fitts, P.M. & Posner, M. I. (1967). Human Performance. Belmont. CA
BrookdCoole.
Fodor, J.A. (1983).The Modularity of the Mind: An essay on faculty psychology.
Cambridge, MA.MIT PresdBadford Books.
Fodor, J.A. (1985).Pr6cis of 'The modularity of mind, followed by 'Open peer
commentary'. The Behavioral and Brain Sciences, 8, 1-33.
Fodor, J.A. & Pylyshyn, Z. (1981). How direct is visual perception? Some
reflections on Gibson's 'Ecological Approach. Cognition, 9, 139-196.
Gibson. J.J. (1959).Perception as a function of stimulation. In S. Koch (Ed.),
Psychology: A study of a science, Vol. I (pp. 456-501).New York McGraw-
Hill.
Gibson, J.J. (1966). The Senses Considered as Perceptual Systems. Boston:
Houghton Mifflin.
Gibson, J.J. (1974).Visualizing conceived as visual apprehending without any
particular point of observation. Leonurdo, 7, 41-42.
Gibson, J.J. (1979). The Ecological Approach to Visual Perception. Boston:
Kinds and Levels of Explanation 117
Houghton Mifltlin.
Gibson, J.J. (1982a). The concept of the stimulus i n psychology. In E. Reed & R.
Jones (Eds.), Reasons for Realism: Selected essays of J.J. Gibson (pp. 333-
349). Hillsdale, N.J.: LEA Publishers; originally 1960.
Gibson, J.J. (198213). The useful dimensions of sensitivity. In E. Reed & R. Jones
(Eds.), Reasons for Realism: Selected essays of J.J. Gibson (pp. 350-373).
Hillsdale, N.J.: LEA Publishers; originally 1963.
Gibson, J.J. (1982~).Autobiography. In E. Reed & R. Jones (Eds.), Reasons for
Realism: Selected essays of J.J. Gibson (pp. 7-22). Hillsdale, N.J.: Lea
Publishers; originally 1967.
Gibson, J.J. (1982d). Notes on action: 11. Note for a tentative redefinition of
behavior. In E. Reed & R. Jones (Eds.), Reasons for Realism: Selected
essays of J.J. Gibson (pp. 388-392). Hillsdale, N.J.: LEA Publishers;
originally 1975.
Gibson, J.J. (1982e). The myth of passive perception: A reply to Richards. In E.
Reed & R. Jones (Eds.), Reasons for Realism: Selected essays of J.J. Gibson
(pp. 397-400). Hillsdale, N.J.: Lea Publishers; originally 1975.
Goldschmeding, J.T. (1977). Neuronale regulatie van de buccale mussa in de
poelslak. PbD. Thesis, Free University, Amsterdam, The Netherlands.
Graham, D. (1985). Pattern and the control of walking insects. Report 48A985.
Research Group of Perception and Action, Centre for Interdisciplinary
Research (ZiF), University of Bielefeld, Federal Republic of Germany.
Gross, C.G., Rosha-Miranda, C.E. & Bender, D.B. (1972). Visual properties of
neurons in inferotemporal cortex of the macaque. Journal of
Neurophysiology, 35, 96-111.
Hayek, F.A. (1955). "Degrees of explanation." British Journal of the Philosophy of
Science, 6, 209-225.
Hinton, G.E. (1980). Inferring the meaning of direct perception. The Behavioral
and Brain Sciences, 3, 387-388.
Holt, E.B. (1914). The Concept of Consciousness. New York MacMillan.
Holt, E.B. (1915). The Fmudian Wish and its Place in Ethics. New York Holt.
Horridge, G.A. (1973). Integration in nervous systems. In E.C. Carlerette & M.P.
Friedman (Eds.), Handbook of Perception, Vol. ZZZ: Biology of Perceptual
Systems (pp. 39-62). New York Academic Press.
Hubel, D.H. & Wiesel, T.N. (1962). Receptive fields, binocular interaction and
functional architecture in the cat's visual cortex. J o u m l of Physiology,
160, 106-154.
Hubel, D.H. & Wiesel, T.N. (1968). Receptive fields and functional architecture of
monkey striate cortex. J o u m l of Physiology, 195, 215-143.
Ito, M. (1984). The Cerebellum and Neural Control. New York Raven.
Johannson, G., Von Hofsten, C. & Jansson, G. (1980). Event perception. Annual
Review O f PSyCholOglY.31, 27-63.
Katz, S. (1984). A reply to Reeds note: 'What is direct perception?" ZSEP
Newsletter, 1(3), 4-5.
Kelso, J.A.S. & Kay, B.A. (in press). Information and control: A macroscopic
analysis of perception-action coupling. In H. Heuer & A.F. Sanders (Eds.),
Perspectives on Perception and Action. Amsterdam: North-Holland.
118 P.C.W. van Wieringen
Koenderink, J.J. (1980).Why argue about direct perception? The Behauioml and
Bmin Sciences, 3, 390-391.
Kottke, F.J. (1976). Facilitation and inhibition as fundamental characteristics of
neuromuscular organization. In A.A. Biirger & J.S. Tobis (Eds.),
Neurophysiologic Aspects of Rehabilitation Medicine (pp. 251-279).
Springfield, IL:Charles C. Thomas.
Kugler, P.N. (1986).A morphological perspective on the origin and evolution of
movement patterns. In M.G. Wade & H.T.A. Whiting (Eds.), Motor
Development in Children: Aspects of coordination and control (pp. 459-525).
The Hague: Martinus Nijhoff.
Kugler, P.N., Kelso, J.A.S. & Turvey, M.T. (1980).On the concept of coordinative
structures aa dissipative structures. I: Theoretical lines of convergence. In
G.E. Stelmach & J. Requin (Eds,), Tutorials in Motor Behavior (pp. 3-47).
Amsterdam: North-Holland.
Kugler, P.N., Kelso, J.A.S. & Turvey, M.T. (1982). On the control and co-
ordination of naturally developing systems. In J.A.S. Kelso & J.E. Clark
(Eds.), The Development of Movement Control and Coordination (pp. 5-78).
New York Wiley.
Lea, S.E.G. (1984).Instinct, Environment and Behaviour. London: Methuen.
Lee, D.N. (1980).The optic flow field The foundation of Vision. Transactions of the
Royal Society, London, Series B, 290, 169-179.
Luria, A.R. (1973). The Working Bmin: An intrwluction to neuropsychology.
Harmondsworth: Penguin Books.
Mace, W.M. (1980).Perceptual activity and direct perception. The Behavioral and
Bmin Sciences, 3, 392-393.
MacKay, D.G. (1981).The problem of rehearsal or mental practice. Journal of
Motor Behavior, 13, 274-285.
Marr, D. (1982). Vision: A computational investigation into the human
representation and processing of visual information. San Francisco:
Freeman.
Massion, J. (1979).Role of motor cortex in postural adjustments associated with
movement. In H. Asanuma & V.V. Wilson (Eds.), Integration in the
Neruous System (pp. 239-258).Tokyo: Igaku-Shoin.
Miller, J.G. (1978).Living Systems. New York McGraw-Hill.
Nagel, E. (1968).The structure of teleological explanations. In P.H. Nidditch (Ed.),
The Philosophy of Science (pp. 80-96).London: Oxford University Press;
originally 1961,
Neisser, U. (1976).Cognitionand Reality. San Francisco: Freeman.
Oatley, K. (1978). Perception and Representations: The theoretical basis of brain
research and psychology. London: Methuen.
Paillard, J. (1986). Development and acquisition of motor skills: A challenging
prospect for neuroscience. In M.G. Wade & H.T.A. Whiting (Eds.), Motor
Development in Children: Aspects of coordination and control (pp. 415-441).
The Hague: Martinus NijhoK
Phillips, C.G., Zeki, S. & Barlow, H.B. (1984).Localization of function in the
cerebral cortex. Past. Present. Future. Bmin, 107, 328-361.
Prechtl, H.F.R. (1986). Prenatal motor development. In M.G. Wade & H.T.A.
Kinds and Levels of Explanation 119
Chapter 4
GENERAL DISCUSSION
I.Neural networks
Whiting: Well, thank you Piet for that presentation. I don't think you could have
been more evocative if you tried, judging from the non-verbal
communication we've been getting (Zaughter) ... people screwed up in their
seats from one place to another. We now have a twenty minutes discussion
on that paper, and I think we should have first the chance to respond to
particular questions which Piet put in regard to the 'pouring' example.
That was a straight question.
Reed: Eh ...
Reed: (Still processing information.) Say it again to make sure that I can ... . That
was the question about whether an act like pouring could be accomplished
via neural networks?
Van Wieringen: Ja. There might be hundreds of these things and you can't get a
grip on them. So you have to work on ... on a high level. Now is that ...
Reed:You can't say that - let me explain - and I didn't write it either ....
Reed: I won't agree - people can read that. But what I am saying is ... first of all
this isn't the context which I gave in my talk ... . We now know that for
almost any movement of an upper limb - let's just talk about upper limbs ...
and the head - that there are postural adjustments prior to the movement in
the lower limbs; not only in the lower limbs but in the trunk, probably in the
opposite limb, depending, however, on where the surface is supported. So,
in a certain situation, for instance if I'm pouring, and my left hand is
gripping a surface, then there will be ... eh ... support movements radiating
outwards from my hand. If I'm standing, they will also be radiating
upwards from the sole of my feet. And it's rather specific. We don't have to
go into all the details. Now, we also know (which contradicts what you said,
I'm afraid) from the literature on motor stimulation of the cortex - I refer
specifically to Asanuma's work, but there are other people - that if you
change the posture of the animal and you stimulate the same regions in the
cortex, you get entirely different movements.
Van Wieringen: That is what Massion does. And then you get a leg lifting, but
before that you see the postural adjustments .. .
Reed: That's right. Now Massion found that. Aaanuma ... did the same - these old
experiments that we were talking about. You stimulate a region of the
cortex, you get a placing response. He did it in cats. Nun Wieringen nods.)
You get something that looks like a placing response. However, if the cat is
in the middle of the air, you get an entirely different response than if the cat
is sitting on a table.
Van Wieringen: That's true, that's true. But what does that prove? Because ...
that's detected in the organism also. That's Heuer's story, I think.
Changing the posture changes the input, so the neural networks are not in
the same condition.
pour with the right hand. Nan Wieringen laughs.) It is possible to pour
while you are looking a t your friend. It is possible to pour while you are
coughing. It is possible to pour while you are sneezing. (General laughter.)
Okay? Now, shall we explain the motor programme before it? It will involve
several hundred muscles, perhaps, won't it, after all those things? There
will be postural precedence effects for the muscles. However, they will not
be structurally specific; they will be support-specific because, if my left
hand is holding, it's different than if my right hand is holding; if my
posture is supported by my belly being against something, it will come from
the trunk. So, now we need another network for this. Fine, if you would like
to build this model, go right ahead. I think, all these things go on. I
wouldn't have mentioned them if I didn't think they go on. I think that the
nervous system is a very important device. (Laughter.)
However, it I want to explain pouring, I had better explain pouring. I
think, at this Conference, we have spent much too much time on the details
of particular movements ... which can be important. I work in
rehabilitation and I need to know this. Sometimes a patient who has, for
example, a problem on the right side ... I need to know these things.
However, when we want to understand pouring, we need to understand
pouring. And we keep avoiding pouring; instead, we want to talk about
everything but pouring.
It is not a model. You say, for instance, - I wrote this down - that neural
networks can produce actions or produce patterns of very similar actions.
The point of the Asanuma experiment is that neural networks, when they
have inputs - I don't believe in the real world they have inputs, but you can
put electrical signals in, those are inputs - ... when you give an input to the
nervous network, you do produce a posture sometimes and a movement
sometimes. But we know that it's not an action because the moment you
change certain configurations of the body, or certain support-
configurations of the animal, you produce a very different effect.
Van Wieringen: Of course, then your network is not in the same condition. But,
maybe, Heuer says something on this, because ...
Reed: So, there is no such thing as a neural network. That's exactly what I'm
saying. There is no such thing as a neural network for an action.
Reed Yeah. Let me just go on a little in order to provoke you. We have at the
beginning different 'levels' of explanation, and different 'kinds' of
explanation. I hope, you see that I don't really believe that these approaches
are differentiated at the 'level'. I think that I have a different 'kind of
approach, which reorganizes how you go to each 'level' because I am not
interested in a particular nervous network. You must tell me how multiple
nervous networks are coordinated, and for each action you must show the
support-network as well as the movement-network. This is not typically
done. Now, I think the difference ... I think it's all in the 'kinds' of
explanations.
I start with functional explanations. I think those provide constraints for
understanding the nervous explanation. It is not my area, and I don't
pretend to be a neurophysiologist, but this theory ... for instance ... would
predict Asanuma's results, and Massion's; it makes predictions about the
kind of systems that might work. They would be multiply interlinked
networks, just as you have said. It is not to avoid that level of explanation;
rather, it gives that level of explanation a different place in the theory.
Whiting:Okay, is there any one else who wants to get in on this ... act?
Meijer: May I strengthen the point just made? ... Well, I may actually weaken it
by wrongly choosing my words. I think this whole 'levels of description
argument' is, firstly, evasive, and, secondly, self destructive. Three people
have used it, I think, in this Conference ... eh ... Dick, Reinoud and
(addresses Van Wieringen) you.
interpreting the same experiments, or the same model and the way we see
that i t works - like Robert and I described ... with the mass-spring
metaphor, which, I think, has a very different place in different models - ,
shows, I think, that the whole 'levels of description argument' does not
work and would, in fact, also hamper discussions, because we just would
have to go home and say: "Well, you do your work and I do my work. And
they concern entirely different things." And that's that. W a n Wieringen
sighs.)
Meijer: I don't deny that there is a level of description phenomenon here but,
since both sides are making rather large claims, there is so much
'common level', that the heated debates always centre around trying to re-
interpret the same phenomena, only in another way.
Schmidt: I just want to reinforce what Herbert was saying ... I think that's the
point, that is, that the phenomena and their explanation don't seem to be
the same. In the ecological perspective, one example is the animal-
environment interface issue, that seems to determine, or seems to provide a
way of asking, about how an animal 'decides', if you will, to jump up
something, or to crawl up it. That's the level of explanation being addressed
in many of the ecological theories. And it's quite a different level of
explanation from asking what the central nervous system tells the muscles
during a limb contraction, for example. Now, I don't see those as the same
at all.
Meijer: Of course, it's true what you said, but it can't be important ... I think. You
present this EMG-evidence and you say: "Well, I agree, but I don't see that
Scott Kelso can get away with that. But also he presents experimental
evidence, and he claims, well .,. we can't get rid of that."
Schmidt:That's just what Herbert said, though. One of the things I tried to say
earlier was that the particular brand of the ecological perspective that Scott
deals with isn't very distinguishable, because they do try to operate at the
126 K. Roth & O.G. Meiier (E&.)
single levels. But if you'd look a t the general ecological perspective, ...
Scott's viewpoint on it is not typical, it represents sort of one end of the
common perspective, I think. So, it's not fair to say, because I'm dealing
with the evidence that tries to argue about Scott's position, that I'm arguing
about the whole ecological position. I don't want to say that.
Reed But what Scott and I would agree on, is that starting at this functional level
reorganizes what kind of explanations you would want. We wouldn't want
-
what you just said with focusing on the muscles. We're interested in what
goes on in the nervous system and the muscles, but that's not what we're
looking at. I think, it's much more a 'kind of explanation issue, here. It's
the 'kinds' of explanation being sought a t a given level that are rather
different.
The way you might approach the functional level is probably different from
the way I would. It's sort of almost a reciprocal relation. At any given level
there's going to be some head batting ... because the attitude towards that
level would be rather different. I'm not sure that the 'levels' distinguish it.
It's much more that there really are 'kinds' of explanation - what you seek
at a given level. And I'm not sure that's too different from what you said, by
the way, right at the beginning. So what you seek a t the 'level' seems to be
different and you can usually find - Onno was right - people talking, a t least
a little bit, at the same 'level', and when you see that, it looks that they're
asking different questions.
Schmidt: J a ...
Van Wieringem J a ... it's not a kind of explanation which brings you very far.
Let's say it like that ... what you said yourself, concerning some patients.
Sometimes, on some level, something breaks down - if I may talk about the
levels. And then you will have to do something ... and if you don't know the
mechanism, often you cannot do very much.
Reed: Oh, no, it's usually the other way around, thank you! N a n Wieringen
laughs.) In my experience, and for most people that I've talked to in
-
rehabilitation and I think that you probably know that too - ... often we're
frustrated by the fact that we can get great deals of structural evidence and
structural data, but we have a very difficult time in achieving function.
This is what Theo's talk is about.
Van Wieringen: But you give your Parkinson patients L-Dopa. Why? Because you
must ...
Reed: No one knows why, or how, that works ...
Van Wieringen: I don't know ...because you have some idea of the mechanism.
Reed:It affects the transmission in the nervous system ...
Van Wieringen: That's what I'm talking about ...
Reed: ... but it afFects a lot of other things too. Not only do you give Parkinson
patients L-Dopa, you also give them visual stimuli.
Whiting: We break for coffee now. Then we will open the general discussion but I
don't believe we can make the 'Skinner - Gibson' question disappear. We
may have to come back on that.
Frischer: Can I say one point about that. It's a slightly related thing; it's about
Parkinson's disease and L-Dopa. Before L-Dopa came along, people
automatically assumed, there was a problem purely with motor
functioning. But when L-Dopa came, a lot of people started to do research
into the mechanism of dopamine and how it works. They began to realize
that it was a problem on a higher level. So, really, just the point I wanted to
make is that the mechanism thing is important in Parkinson's disease,
and it does lead to better understanding of the functional movements in
Parkinson's disease.
Schmidt: Sure! The neurosciences are doing it ... . That's one of the points I
wanted to make, about t h i s blend of experimental psychology and
neurosciences. There appear to be people now - large groups of people - who
have on the one hand a view of the movements, maybe not 'actions', but
'movements' that people are making, or animals are making, and in the
same breath are talking about the neural substrates that are associated
with those. That's a very strong emerging tradition in the neurosciences.
130 K. Roth & O.G. Meijer (Eds.)
Reed: I think, I would accept that. You're right. That's fair, except, I think that
there's a problem in saying 'in the same breath' because, when you look at
it - I'll go back to Asanuma's nice piece of work on this electrical
stimulation effect - the tendency is to say, for instance, that in insects some
-
neurons - in not so large a region form a programme or a centre for a
programme. And then, when we look at it in any context ... it turns out that
it isn't it's a fragment of some larger system. And what we do is to jump to
that level. We all do that, I admit it - I'm not just talking about other people.
We all jump from the level of some sort of meaningful thing, like a
'programme', to a different level. And I think that this is what Piet was
talking about. He called it 'intermediate!. It's pretty much a black box for
everybody. I agree that there is a neuroscience sort of attack on this, which
has emerged in the last couple of years and which is interesting. But, as
you know, it hasn't got us to know how a Parkinson's patient can reach ...
Pickenhain? I think we must distinguish between the 'explanatory' levels and the
'objective' organismic levels, because they are different things. There is an
objective organization within the organism which, up to now, we don't yet
know too much about. But i t is a completely other thing to have different
ezplunatory levels which must be proved to be ... correct, or not correct; ...
they are completely right to go this way without referring obligatory to this
objective, organismic level.
I wish to say that the dopamine system is a very important part of a greater
whole. There are many subsidiary functions, there are many other
connections, and many neurons; ... you can implant new embryonic
material into the substantia nigra - it grows out and Parkinsonism
disappears. This is a completely new method which was used in animals
and now they are trying it also on humans; it will work. This is all in an
organismic 'network.
The second thing I wish to say is that 'network must not be understood as a
fixed-wired connection of neurons; it must be understood as a dynamic
system of very different connections in a degenerate way. This means that
any complex of elements may be used for different purposes and in different
networks, and may answer to different af€erences and give different
efferences, according to different functional aspects, or, finally, different
goals.
Thirdly I propose that with the word 'programme' we must be very
cautious. It means several things and it's better to distinguish them. First,
the genetic 'programme'; I would only talk about genetic 'code', not
'programme', 80 that we can distinguish it from other uses of the word
'programme'. Then, I would propose to call the processes which are going
on in the frontal cortex in preparing the motor programme before the
decision stochastic pre-programming which is 'programming' as a
pmcess. And then, after the decision, one could speak of 'programme
output', which may be changed and adapted accordingto the situation.
I think that many misunderstandings between people have their cause in
the use of the same word for different things they mean. If you ask the
Geneml Discussion 131
Daugs: I think that there is a very simple problem, i.e., the relationship between
explanation and description. If the pre-potentials of Kornhuber are seen as
an 'explanation' of anticipation, or something else, that's wrong. It's a
'description' - I think Kornhuber himself would say just the same. So, the
'explanation' on one level is a 'description' on another one. And so on. It
would be a big mistake to say: "On one level we talk about very global things,
like 'anticipation', and then something arrives which shows us pre-
potentials, and we think them to be an 'explanation' of anticipation." That's
quite wrong.
Bootsma: If you would sort of start from the bottom line and research kicking-a-
ball, you could elicit that behaviour by laying the ball in front of somebody
and hammer on his knee; he then would kick the ball, and that would be an
'explanation' of kicking-the-ball. If you, then, approach from the level of
normally kicking-the-ball, you would say that this is not a valid
explanation; it doesn't work with just reflexes. So I think, what Ed Reed is
saying is that motor control theory has up to now come up with all sorts of
explanations and constructs.
Yet, if we look at the molar level and see what Professor Pickenhain says,
they have to be compatible with experimental findings. I think what action
theorists are claiming is that the results they find show that constructs,
lik4 'schemata' and 'motor programmes', are not compatible with the
findings they come up with, which would mean that you would either have
to fight their results, or say: "Well, then, the schema-script - whatever you
come up with, like 'memory structures' - is not valid for the situations we
are looking at."
Van Wierhgen: I'm not sure that they're incompatible. I go to the question of
Whiting now on applying results. Knowledge of Results: we know that if
people learn motor skills, like we heard here, there are effects, for example,
of analysing your mistakes afterwards; this has an effect on the learning of
motor skills. Also mental practice has an effect on learning them. Now, you
can say: "Okay, I don't look at these data." You can do that, of course, but if
you say: "Well, of course, still these are data and I want to give some
explanation" I am sure you cannot explain it in a way action theorists like
-
to do. The only way and that's the thing I wanted to say - to account for
such data is with these sloppy, intermediate constructs. Then, at least, you
can give some explanation.
132 K.Roth & O.G.Metjer CEdS.)
What you see - Newell has remarked this in his article of 1974 - is that
people have sets of data, and don't look a t any other data. They say: "Well,
that's unpleasant" and they form different camps. But assume that I have
this theory while that is also an interesting finding. How can I explain it?
And then you may say: "Well, okay, maybe he is not so bad aRer all in
taking these sloppy concepts, because I myself cannot say anything about it;
I even don't say that there are any cognitions at all." So, this relates
directly, I think,to the application in teaching situtitions, etc.
Reed: I think that my theory has to explain these data. (By the way, I think that
what you claim are 'explanations', are 'descriptions' - I very much liked
what Professor Daugs said.) I say this in the action systems paper, I list the
semantic system, the expression system; so I'm obligated ... if I can not do
it, then I am wrong.
Whiting: Can we come back to Reinoud, talking about something being valid in
the one situation and not valid in the other? I am not sure in what sense you
have constructed 'validity'.
Beek It's the difference between mechanism underlying function, the phrase
some people use, and ... mechanism subserving function. I would take the
latter to be close to Reeds case. Let's say we have a kind of primacy of
function, and see what happens there. And, alright, Reed wants to stick to
that level which is fine with me. That's good because so little attention has
been paid to just the phenomena of action. Other people have this urge to go
for the mechanism. That's also fine, it fits their information. Okay, but
then, you have to be aware of this distinction, I think. Because, when you
start with mechanisms undedying function, you start to build up from
restrictions which you did not even analyse. You just start with this
mechanism for reaching, and take it as the explanation of the function.
What Reed shows is that this is a very difficult line, a very troublesome way
of proceeding.
Warren: I just want to say something about the notion of 'mechanism', because
it's something the Gibsonians are accused of having nothing to say about.
And now action systems people are getting accused of having nothing to say
about it. The problem is that the notion of 'mechanism' is very muddled.
There are at least three notions of 'mechanism' going around in this room.
A motor programme theorist would say that a motor programme provides
the 'mechanism' which explains a movement. A physiologist would say
that a cell assembly is a 'mechanism' that explains movement. Now,
clearly, the cell assembly is not the same thing as a motor programme,
although there are some people who are trying to identify them. There's yet
a third notion of 'mechanism' which I think is more compatible with the
ecological approach ... I don't want to speak for the late James Gibson.
It questions the role of structures in causing behaviour or in causing the
organization of behaviour. The assumption appears to be that the
General Discussion 133
Schmidt:Well, ...
Warren: Let me finish ... . In physics, for example, the notion of a 'field provided
a 'mechanism' which is at a different level than the causal bumping of one
particle to another. And I think it is those kind of 'mechanisms' that
someone like Kugler, or Turvey, is trying t o identify. Those types of 'field
properties may be realized in ... ja ... one group of cells, or another group of
cells, but the explanation is provided by the laws of 'fields', not the
particular wiring of the neural structure. So, I guess, what I am
suggesting is: Those patterns may get realized in any particular neural
structure; any particular neural structure can be involved in the
production of a whole variety of movements, potentially. And the
explanation is going to reside at the level of the laws, rather than at the
level of particular wiring diagrams.
Beek Well, what I mean by mechanisms subserving function ... comes very close
to what Bill just said about an ecological explanation of these things. But I
just used the phrase in order to communicate. I mean, it's the word all the
people use in this field: 'mechanisms'. So, whatever you take it to be, I
agree: it is a very sloppy term.
Reed Maybe it would help to set an example of what Bill just said that we could
toss around. A very simple example. He gave this tau-principle, and I
won't re-explain it, this time to contact controlling locomotion. There is very
good evidence that applies to bees as well as humans, not to mention other
vertebrates: this works. Well, the mechanisms in a fly's nervous system are
rather different than those in a vertebrate's nervous system. The kind of
explanation that I think Bill and I would seek for this involves the
information, the kind of control systems he was talking about, and, at the
level of 'mechanism', a principle about how any self-organizing system
could be constrained by that information. The detail would be rather
different in the two cases but it might still be a field-theoretical explanation.
I'm very curious if people who tend to go for a motor kind of approach
would see those things as being the same, needing to be explained in the
same way; o r maybe they would see them as entirely different, needing
entirely different explanations, given the fact that they're in a different
nervous network. Anyway, there's an example of how the same thing is
done in two extremely different ways but we might ...
Schmidt: I take your point, Bill, about the different ways that we think of
'mechanism'. But nevertheless, some sort of 'mechanism', however we
might want to conceptualize it, needs to be specified in order that these
viewpoints be testable, at least testable in the domains that we are interested
in working at.
Let me just put Ed on the defence here for a second, and say that my
impression in reading your papers is that I don't find 'mechanisms'
specified, ... however you might like to characterize the 'mechanisms' ...
you mention a tau. But I'm a little concerned about ... I'm a lot concerned
about the idea of calling action theory a 'theory', because in my viewpoint
theories need to have mechanisms or processes, specified implications or
hypotheses derivable, ... things to be tested. And, perhaps, it's too soon to
talk about a 'theory' of these things because I don't see what they predict.
Reed: Well, it was, but (addresses Schmidt) you may not remember it. (General
laughter.)
Schmidt:Thank you!
Reed: You may not remember ... in order to be critical. There is actually a specific
section in that paper on what Bill was calling the 'control problem'; and
that's the level at which I would prefer that theory to be tested. My theory
doesn't make the kind of predictions, let's say, that Scott's doing in his
homeokinetic theory. They are different levels there. The level that I'm
seeking is more ... the functional one. But to test it, you need to find
information specific to functional situations to show that that constrains
postures and movements accordingly. And I do talk about that. That is
where I have to put up or shut up ... to provide in particular cases
descriptions of that information and concrete tests where the information
would vary, and so on.
That would be the procedure. It may not satisfy you; you may very well be
interested in the level that I'm not dealing with. However, I think that what
I'm saying is rather compatible, not completely but rather compatible with
this sort of thing that Kugler, Kelso and Turvey are doing which is close to
your level.
Schmidt: That's the criticism that I most frequently hear about Kelso's
perspective, that it's incredibly difEcult to test and, whenever you provide
testa, you get ... sort of a wriggling response: that you've exceeded the
boundary conditions, or that this is a different kind of task than the ones
they're talking about. There seems to be a strong unwillingness, and not
even a capability of these viewpoints to be subjected to the classical kinds of
critical testa.
General Discussion 135
Reed: Well, I don't think it would be appropriate for me to try to defend Kelso
there. I'm saying that there are specific claims, and I would make specific
predictions. Given a prediction about information that has the control of
behaviour - analogous to the tau - if this would be tested and if it came out
wrong, I couldn't wriggle too far.
Frischer: I'd just like to ask E d If you've got a Parkinson's patient and he's
making rhythmical movements which are very abnormal, how does this
perspective of 'fields' provide any possible help to these patients? It seems to
me that you have to resort to some kind of motor programming. You have to
know what the information is in terms of an internal motor programme, in
order to be able to help that patient.
Reed: Well, that's just not factually true. In some cases with a Parkinson's
patient who's making abnormal movements, if I provide rhythmical
auditory stimulation I can calm him down. I don't know anything about
the programmes, or anything like that, right? So, it's not clear ... I mean ...
that's a fact that doesn't support my theory especially. It doesn't support
anybody'stheory.
Reed: This is a must that I don't understand. I find that purely descriptive, just
the same as we have been hearing. I find the explanation on that level
'descriptive' because every time I make a functional category, like
'pouring', which comprises many, many, many, many, many
mechanisms - and I don't have to go through it again - , people pick out
different mechanisms for different aspects. If you pour standing up, it's
different than if you pour sitting down. Fine. We will describe all those
underlying mechanisms. But they're not mechanisms of pouring. They're
mechanisms that are occasionally involved in pouring. They're
descriptions, they're not explanations of pouring. They're explanations of
fragments of pouring, or things that are occasionally involved in pouring.
But again, they're just descriptive. At my level, I need to understand how
someone goes out and coordinates two hands, uses information about sizes
and shapes, coordinates the stream, understands or uses information.
Frischer: But how can you test one explanation against another in that case? I
mean, in the case studies that Edward just mentioned on rhythmic
stimulation, you can have various hypothetical motor programmes
providing, for example, the timing parameter, and the motor programme
might be involved in several different ways, and you can test those different
motor programmes by providing different rhythmic stimuli, or different
136 K.Roth & O.G. Meuer OMS.)
4. Fitta's law
Lee: I have a very specific question; I think it illustrates a point that Dick was
making earlier. I'd like you to explain to me how the action approach would
explain Fitts's law. What that might do is give us a way of contrasting
empirical ways of testing the motor versus action approach.
Whiting Is the question clear? How the action approach would come up with an
explanation of Fitts's law?
Reed: I don't have it, so I'm not going to explain it. I had some ideas about this
because I happened to be doing something that looks a hell of a lot like a
Fitts's law experiment. I had people reaching for a pitcher and a cup, and
the pitcher had a different amplitude as a target than the cup. For the time
being, I don't get Fitts's law, by the way. That's all right. I might have got
it; I might not have got it. I would, I guess, have to say ... here is a good
place for the 'levels' discussion.
I find myself unable to think that I really could say anything very profound
about Fitts's law, so maybe there's a 'level' problem right there. And then,
I'm interested in reaching, and if Fitts's law doesn't happen to deal with
h c t i o n a l reaching, like pouring o r reaching for a fork, then I'm not
interested. But let's say it did. What I would have to do is, try to find out ...
about the spatial information concerning a target, and so on, constraining
the temporal structure of the movement. And I would have to make a
General Discussion 137
specific hypothesis about that function (it would be a function, Fitts's law is
a logarithmic function) that would be motivated, not ad hoc, across different
functional situations, like picking up a fork first, then picking up my
pitcher, all right? Because I don't have one, I'm not sure what else to say.
This is how I would go about it. Say, I found it. I looked for it, mind you ... .
I was curious if I would get it in picking up, or pouring. And then, I didn't
get it. But say, I got it - and I might switch to studying picking up a fork. I
would first want to quantify the functional situation, which hasn't been
done - and that's peculiar ... . So, I would evaluate the function that really
is, and I would be hard pressed. I would have to start manipulating, I
guess, that spatial information to test how i t rearranged the temporal
structure aoross those functional tasks. I would be very persuaded, by the
way, if in picking up a very light thing, like a fork which is a very small
target, and picking up a very big target, you know, like a gallon container, I
was getting some 801% of the same function. I would be impressed. I would
say that Fitts's law was something that was a law about information use,
and I'd be happy with it. And maybe I'll find it again in forks. I haven't
looked at it in forks yet. Does that help a little bit?
Lee: Yeah.
Reed: How does he measure when they touch the object when they're picking it
up? What does he use as his end point? Just touching?
Schmidt:Anyway, the E t t S ' S function always holds, but the slopes are different ...
Reed: Aha! Maybe if I did more work, I'd be able to say something like that.
Across the hctional tasks I'd be able to see some more patterns holding
up two things,a cup and a pitcher.
Schmidt:Okay.
Warren: I think, Ed's being generous in a sense that there is a stronger response
to that question as well, which is very similar to Runeson's argument in
his 'smart mechanism' paper. You take a system that's evolved, for, you
know, one set of tasks or one set of functions, and you put it in a highly
constrained, somewhat arbitrary, situation, and make observations on it.
And it may, in fact, behave very regularly within the constraints of that
138 K.Roth h O.G.Meiier @&.)
situation. But when you alter the constraints, that regularity may
disappear. In other words, it could be that the regularity that you see in
something like Fitts's law, may be a by-product of the fact that you've taken
a device, put it in a strange situation, a situation for which it was not
evolved to cope with - and tight constraints on the task. And then it does
something very regularly. But that doesn't necessarily reveal to you how
-
that system operates it doesn't generalize necessarily. So then you need
experimentsto see how far it generalizes.
Lee: If indeed, as Dick mentioned, Ron Marteniuk was showing that Fitts's law
seems to hold, albeit with different slopes, across the ecologically valid
tasks, as it were, then I still would like to see an explanation, of why that
occurs, in a fundamentally different explanatory way: from an action
approach.
Lee: Well, perhaps 80, but in the development of the theory, it is the testable
hypothesis that leads to the theory.
Reed: Let me mention a hypothesis I had about why I didn't get Fitts's law. My
hypothesis ... maybe I can work it out with this slope business. The objects
that people are reaching to are of very different weight. They're grasping
the objects and there's this problem of measuring - Fitts's law is always
with pointing, at least the way I know it. I wasn't sure of what was the
endpoint, but anyway, I measured just touching the surface. It's a
preparation to lifting objects with very different dynamic properties, and,
also, for a different function.
I suspected that, whatever Fitts's law was, it was happening to a
relationship between some spatial infomation and a movement that had
only the dynamic properties of the limb without consideration of picking up.
And I'd be very interested to see if the picking up did make a difference. It's
a pure hypothesis, but I didn't test it. It's very testable. I could do what
Marteniuk did with pointing versus picking up. I could also, what I'm
about to do, sample in different ways.
I got the idea from the action approach that the information has to
constrain the dynamics. It may turn out that there is a general form of
Fitts's law of which the simple arm tests are just one example along a more
complex array of functions. I'd be happy with that, but again, a real
explanation of that would have to take into account the tasks of the hand,
the dynamics of it, and - which is very difficult - why that spatial
information does plug into that dynamic task in just that way with just
those parameters.
Reed: I'll go home and do this. It sounds like it would be fun and it might actually
get at whether Fitts's law is a special case or something that might be more
general.
Magill: I would like to come back to a fundamental point that has been raised a
couple of times, in relation to differentiating an action theory approach and
a motor approach. Going back to John's original question about
applicability, it seems in m a n y cases that the more applied you get, the less
clear the differences become. It seems that both sides could handle the
same applied phenomena. But there does seem to be a view that if you hold
to a mechanism, such as a motor programme or schema, therefore you are
in a motor approach. I don't quite follow. Perhaps I'm wrong.
In getting ready to come here, I picked up a book that has just come out, by
Frbse and Sabini who are at the University of Pennsylvania and edited a
book called Gwl Directed Behavior which is based on a Symposium held in
Germany for the exact same purpose that this one was formulated - to get
the two sides of the Atlantic to talk to each other. In their introduction they
talk about what is an action theory and they talk about one section of the gap
between intention and action. So they describe what an action theory
approach holds in that particular way. It is interesting to note that they do
describe that a very important characteristic of action theory, which they
carry over from the cognitive approach, is the notion of 'representation' and
the use of cognitive facilities. They then go on to say that automized plans
are stored schematically with parameters added, depending on the
circumstances under which they are used by the plan in the concrete
circumstance the actor faces. And, ironically, in relation to our
Conference, they reference Schmidt. (General laughter and som
applause.) So, that really puzzled me when I came here. I thought: "Oh,
now I have a better understanding of action theory. It says a lot of things."
Now, I am curious, &er hearing much of this discussion. Is the action
theory that this group addressed, the action theory we are addressing here?
Are they different things ... ?
Shea: First of all, I think that the notion of 'action' in the cognitive area has
probably come to the forefront largely because of the close association of
cognitive science to artificial intelligence. In the artificial intelligence
community 'action' is something that's very important where you have
different types of knowledge systems that have to be operated on through
140
Van Wieringen: I just want to talk about my daughter Rosa. She's on ballet and
takes lessons. I like ballet. At home, she goes to a book, tries to remember,
goes into a room with girlfriends, and they try to do it, and then they go back
and look. And then, next morning, she's having a lesson and she tries ...
and she may say: "It's no good." What's happening there? I think, there
are representations of the state she has to reach. And these representations
guide her behaviour ... . I don't see how you can talk about this without
using a representation. And such processes, of course, they go on
everywhere.
These are really difficult things for action systems theorists to explain. How
do you talk about such things without using cognitive notions? That's my
question.
General Discussion 141
Shea: Yes. One'of the things our research highlights is that the role of movement
is to allow this internal representation the person has constructed to
become anchored and adjusted to the parameters of the real world. In other
words, you have to try it, to find out what those parameters are, to set them.
And then, outside of that, you're okay. In a sense, the movement itself is not
what's really being stamped in, or something like that. When you see your
daughter doing this, she is adjusting and fine-tuning the parameters of
this internal representation to the real world where the action has to occur.
If she didn't do it, she wouldn't know what those parameters are. They
wouldn't be set for the action to occur.
Whiting:We are having two different lines now. I am getting the idea that you are
talking about movements as subservient to cognition, Piet is talking about
cognitions as a subservientto movement.
Reed: I'm even more confused. Firstly, we had a representation in the sense that
many people talk about it. Dr Shea just said that if I were in the act of
pushing a button, or reaching, I have to have a representation or a motor
programme to do that. Now, we have another representation. I don't deny
that a picture is a representation. It's obviously a representation. So, not
only do I have to have pictures in my mind to learn how to reach when I'm
a little kid, but now I have to have pictures of pictures. Because the little girl
looks at the picture, she gets a picture in her mind.
Van Wieringen: Sorry, but that is no explanation. On the other hand ...
you might
accuse me of using explanatory fictions - like Skinner does - with regard to
cognitive constructs. But I think that by talking about representations,
memory, etc., you can ask important questions. Doesn't she memorize
142 K.Roth & O.G.Meiier fEa!s.)
these things very well, is there something wrong with her visualization, or
things like that? I told you, these concepts have something sloppy. I really
cannot say I like them myself. If I could stay on one level, it would be much
more pleasant for me. Also I can see the argument Warren presented just
before, about these mechanisms ... . If you want to address, say, learning
processes like that, I just need these cognitive boxes some way or the other.
Reed I ask: What does the cognitive box get you? If I say that having a picture or a
teacher provides you with information about the goal - which in this case is
to make a pli6 - , ... you not only learn to pick the information up, but you
have to learn to use the information to control your action. We agree on
that, I think. Are we disagreeing or not? I don't even know.
Shea: We don't disagree ... because you could not talk about any form of
representation the way you are talking about it - the representation, if you
want, of some visual image or something - without, a t first, having had
some closely associated procedural knowledge that has to help set up this
thing. I mean, if you are working on a computer, you have to set up
procedures to enter your data into file or into some kind of format. And
then, you also have to have a set of procedures, or the same procedures, that
can be applied to rearrange it, change it the way you want, and then,
finally, you might want to execute it.
But procedural knowledge can't be separated fiom declarative knowledge in
any representational system. So, what we are saying is: "That seems to be
the area that the motor people have concentrated on" whereas the action
people have been able to say: "We want to know really, in a real world sense,
how the hardware carries these actions out. And we don't touch those
issues.''
Pickenhain: Only a short remark. We must have in mind, during this discussion
about representations, features, pictures, and so on, that most of these are
unconscious. And this has a very decisive effect. We must not imagine that
all is conscious. And it looks as if thinking about the mental processes
shows that most of these processes are at an unconscious level, and only a
very small part, which is most exciting and which we include in special
cases with special difficulties, is conscious.
Shea: That's not really the difference. I mean, we would agree that procedural
knowledge is usually unconscious.
Warren:I'll just try to respond to Dr Shea a little bit. Cognitive science has kind of
boiled down to the study of representations.
Shea: I think that you can almost say: "That's the definition of cognitive science."
Warren: Although there are other ones - there are computational procedures as
.
well. But it's true for both perceptual problems and motor problems . . .
Geneml Discussion 143
Shea:The cognitive sciences, I say, probably would not even classify in that.
Warren:Yeah, right. The whole ball game is for representations. And you might
take a debition, or a part of a definition, of the ecological approach, or the
Gibsonian approach t o perception, in that it is trying to explain perceptual
phenomena without representations. For the action approach tries to
explain how you get coordinated action without relying on representations.
Internal representations have a very peculiar ontologicalstatus.
Where do they come from? Are they causally efficient in producing
movement? Are they just a description of something else that's going on in
the system? And I choose to look at something, like a representation or an
algorithm, as a description of some process that is going on. But not
necessarily the best description. And certainly not a lawful one. It's a place
holder, somehow ... .
Shea: I don't think that anyone is going to come up with the best description. Or
the algorithm that solves the problem. Basically, as far as simplicity goes,
in not having to deal with this problem, which is frustrating, the action
approach is certainly one that manages to do that. But I think that,
eventually, you have to come back to this tau, this little box, this special
purpose device, or something. Whether it is a hard wire, or it is a piece of
hardware that attains this goal, it is still represented in the system.
Shea: That's why I don't like this levels-notion that keeps coming up. It is very
confusing. If you take our point of view, ... I have no trouble with much of
what you are doing, because, basically, I feel like saying: "Well, if what
they've found out is true, that's great because that frees us up. Because we
have systems that we can interface with." Eventually, this will happen, I
think. There are af€ordances that allow these things to happen, and right
now I don't think the motor approach comes even close to explaining these.
They don't emphasize movement and how it happens.
Warren: Let me ask you a question ... I've always wanted to ask a cognitive
scientist. (Laughter.) What is the status of a 'representation'? Do
representations exist in organisms? Or do they exist in the brain, or are
they descriptions of something else that exists?
Shea: I find that kind of interesting because it is the action theorists, like Kelso
and Turvey, who have relied so much on metaphors, like pogo stick
metaphors, and other sorts of things, to communicate their ideas.
Basically, you could say: "Well, does the pogo stick exist in the mind
somewhere, or where does this metaphor exist?" In fact, what we are using
144 K.Roth & O.G. MeGer (Eds.)
Warren: Let me just give an example. It disturbs me that, whenever you push
somebody at it, they say: "Oh, no, representation is just a metaphor, it's just
an analogy for something else!" But then, people go ahead, and treat them
as though they were real things that explain something. We were talking
about this a little earlier. But let's take a physical process, like a wave
breaking on a beach. And the form of the wave is a regular organization of
the behaviour of that system that occurs, with some variation, very stably.
And you can explain those pattern on the basis of the laws of
hydrodynamics, the shape of the beach, and the wind forces, and so on, that
are a t work. Just because you have a regular behaviour, there is no reason
to ascribe a representation to the water in order to execute that stably
organized pattern of behaviour. Now we look at stable, organized patterns of
behaviour in animals or people, and we suddenly want to describe
representations that will explain those orderly patterns of behaviour.
Shea: Well, that's an interesting point. Just the fact that you focus on stable,
constant, organized patterns of behaviour, earmarks a focus that is natural
for you, for that particular school of thought. And it's interesting that
people who study representational systems seem to focus on flexible,
modifiable behaviour patterns. And that's natural for those people; that's
probably why the representational systems approach appeals to these
people, because they can look a t it from this perspective. So, it is interesting
that the focus of research in the two areas diverges a t that point because
one type of behaviour is more consistent with what you are looking for, and
the other kind of behaviour is more consistent with our approach.
Warren: But I don't think so, because, as the constraints vary, the patterns of
behaviour vary as well. And you need a theory of the constraints.
.
Shea: That's not incompatible . ,
Reed: Is the way to learn how the beach erodes, and the pattern of the waves
changes, regular?
General Discussion 145
Shea: Well, I think, ... at one level you are right. ... I can see why the physical
sciences are such a nice area for you to get your ideas from. But for those of
us using the representational approach, we don't want to be saddled with
the physical sciences because we don't understand why human behaviour,
which has never been parsimonious in history, should obey parsimonious
physical laws. (General laughter.)
That's a difference again. But these are differences that are trivial if you
ask if an internal model for an action must exist for the action to be
performed. You've got to have systems of affordances, and things like that,
that work. And maybe that's what we will eventually do, what has to be
done.
Warren: Well, I think that if you really do arrive at a theory of those &ordances
... , and the informational constraints, and so on, that the need for
representations would disappear.
Warren: No, we don't need them. And no need for internal models ...
Van Wieringen: Take dance again. Someone has looked, then goes away to a
studio and tries to do that. She, or he, fails. Now, I have a most simple box
model. First, she looks a t a model, then she memorizes something, and
then she tries to imitate what she has seen. Three things: looking,
memorizing, and a kind of output. Now she fails.
I think three things are possible. Maybe she doesn't see it quite well. So, we
can do a test on seeing. Maybe there is something wrong with memory, and
we do a memory test. Maybe she has very stiff legs - like I have - so she will
never succeed. So, I have now the possibility to apply some tests. How do
action theorists see this failure? How do they go into this problem? It's a
very simple question.
Reed: That's very similar. We would look at whether they could pick the
information up, whether they could use the information over Merent
periods of time to control the action, or whether they had problems with
constructing the movements and postures, and things like that ... .
Van Wieringen: I see the importance of information, and also of constraints, for
sure, but you don't come too far with it, as far as I can see.
Reed I just don't know what else you want. I really don't.
Whiting. What do you mean when you say: "He picks up the information from the
book?" Is there, then, no representation? I can see this in a synkinetic
paradigm, but here we are talking about an echokinetic situation.
Reed I am sorry. I think, the question is: How does the picture control the child's
dance? Let us explain that. Let us not just invent another representation
which replicates the problem. It's a very hard problem.
Pickenhain: We can only understand the problem if we see that we also have a
representation of the sequence of behavioural acts, of motor acts. We don't
Geneml Discussion 147
only have pictures, and this is more important - because these pictures only
concern part of behaviour. Someone used the example of motor car driving.
If you have a new motor cdr, you must try all movements which are
necessary to use it. And if you drive it several years, you have a complete
representation of all these features, or other things, especially of all
movements which are necessary, ... and if this doesn't function, you then
consciously must change this. And I said that most representations are
unconscious. During all your driving a car, you have so to speak
unconscious representations of movements which are executed for this
special purpose. And you only use the conscious way if you have to change
your behaviour.
Reed: Dave Young actually said this in his talk. Remember that he was balancing
the stick and he gave the differential equation, if I remember right, for how
the stick would behave - which is a differential equation ...
Reed: Hold on a minute, let me finish! ... And then he said that the stick can be
considered as an information processing system, and obviously the
differential equation at the highest level, as I understand it, describes the
programmes and behaviour of the stick. I have no objection if that's all you
mean. In other words, we have a description of the behaviour. I have an
objection if you think that, somehow, this differential equation has a causal
status in the system.
Warren: Yeah, that's why I have difficulties with the differential analysis
approach. I don't want to try to defend Kelso and Saltzman's analysis
specifically. But, broadening it a little bit, you can use those kinds of
functions too, as a way of describing what a system is doing a t a particular
time. Now the system somehow got organized in that way, and the equation
they describe gives the way that system is functioning. That is not t o say
that the system used a representation of an equation in order t o organize
itself in that particular manner.
Heuer: I think that that is the same with a motor programme. No one, I think,
will have the idea that you have a space-time function in your head. No. It
is an abstract description of some part of the system.
Reed Well, okay, but how is the explanation that a motor programme 'causes' a
behaviour any different, ontologically, from my saying that information,
which is an abstract description in an abstract space, 'controls' behaviour?
It is the same thing, then, isn't it?
Heuer: Well, it depends on whether the information is inside or outside the head.
Reed It's both, right? In both systems it's both. I mean, you have to learn it, at
least, and you have to have got it from some place, right? So, it's both in both
cases. I'm not sure that this is correct, however. I suspect that there is a
disagreementbut ifthat's all you mean - fine.
Whiting: I am getting a bit confused. I think what Bill is saying is that if you
understand how a process works, then you can write an equation for it. But
does that equation add anything extra? Dick says: "Yes."
Warren: I guess that the point that I'm trying to make is that you can have
different types of models of what something like an action system might be
doing. And you can describe that system in different ways. Now, one
description of that system might be something like a set of equations that
describes a physical system. Another type of description might be a motor
programme. I think that the action system is more like a kind of lawfully
behaving physical system than it is like a formal system such as a set of
rules. I am much more predisposed to try to analyse it in terms of a kind of
physical system analogy, than I am in ter1-118 of a formal computer analogy.
General Discussion 149
Whiting: I'd like to shifk the focus of the discussion now, and I have the
impression that Dick had another line.
Schmidt:I d like to try to clarify the distinction between what Ed and I mean ... . I
think I understand what I mean &ugh). Let's say what that viewpoint,
your viewpoint, would predict at a motor control level of analysis. I think I
heard you saying a couple of times that one thing that would provide
support for your viewpoint - and I guess, by implication, d ~ c u l t i e sfor a
motor programme viewpoint - would be that you would find evidence that
information is used to constrain the dynamics. I want to know what you
mean exactly by that. Maybe you could answer that by specifying which of
these two possibilities you mean: By 'constrain', do you mean 'determine'
or do you mean 'affect', 'influence'?
Reed I think, it's pretty hard to demonstrate 'determine'. This is a cute example
of the Fitts's law business. It seems to me that if it is compatible with my
theory - you tell me whether it really is incompatible with another one - that
if I vary the dynamics, so I have the same amplitude object, like a gallon
pitcher, but in one case it's empty and in the next case it's full ... . I'll just
look at the timing, lots of other things about the movement will change but
suppose I can show that timing changes.
Maybe I can show, and purify, the patterns that the timing changes in a
way corresponding to the physical demands, the dynamic demands of
picking that pitcher up. Personally, I would prefer to arrange the
experiments so that they were also fimctional demands, and see if the
timing changed with those. But let's just stay with dynamics a t the
moment. So, I show that the Fitts's amplitude and region relationship
doesn't hold. (hughter.) However, in some kind of higher order, a version
of that holds if you take the dynamics into account. Again, this is just pure
speculation. Doesn't that support the idea that information, related to the
specific dynamic relations of this organism's sitting a t the table and
reaching out, is constrainingthe action?
I wouldn't say 'determining' - it would be much too hard to prove that - but
'constraining' that reaching test. I'll add that if and when I do this
experiment, I will also interact it with a change in function. Because, I
would also predict that on top of the dynamics affecting the Fitts's law
relationship I should get, I hope, a reasonably lawful relationship due to
the functional shiR.
Now, I made those predictions, and those are reasonable predictions, as far
as I understand my own theory. They would be supportive of mine - I'm
unaware of any motivated reason from a Fitts's law kind of paradigm why
those things should change. But I could be wrong. I'm not saying I fully
understand the Fitts's law business.
150 K.Roth & O.G.Meijer (Ea9.)
Schmidt: Well, Fitts's law is not a theory. It's just a description of a relationship.
So it could make no predictions, one way or another.
Reed: It has always been interpreted rather geometrically..,
Schmidt: Oh, it has been interpreted in at least three different ways that I know
of. But, I guess, my response to your answer is that about any
computational - and even 'computational' in the worst case - motor
programme theory would predict exactly what you've just said. Of course,
the goals of the system are going to be important in how the motor system is
pre-programmed. If I know in some sense that I have to pick up a heavy
pitcher, versus a light pitcher, then I'm going to constrain my action
differently. I don't see how that data provides any sort of a test.
Schmidt: That's what Z am trying to do, and I think that many people in this
room wi l l say: "How can we attack these viewpoints and make some
concrete ... "
Reed: Let me point out a couple of things. I agree that you can re-describe almost
any experiment in these different ways. Is it really a principled notion for
any other theory of perception besides Gibson's that you would be perceiving
dynamically relevant information concerning the organism environment
relation? Spatial perception people would say: "I can see the input without a
spatial frequency." They don't know how heavy the thing is. How could a
spatial frequency analyser tell me that this one is heavier than that one
when it has the same contour? So, I think that there are some things being
slipped in here. I, automatically, because of my theory, assume that that
information is available. A Skinnerian will have to talk about some form of
learning: whether that person learned over time that these different
weights were there ...
Van Wierlngen: According to me, Skinner's view is not incompatible with the
idea that action might be geared to perception already during ontogenesis.
Frankly, I would be happy if people did a lot of functional shifting, took their
tasks, and proved my theory wrong.
Schmidt: Would you not just say that I did not change the task enough?
R e e d No, I can't do that. That's why I have the action systems. Go from one
action system to another. That's the best way to do it. Go from a
manipulative to a locomotor test, if you can. Something really big like that
... which goes between the two systems. A good example is the difference
between raising your arms to greet somebody, and raising your arms just
like that. I think Professor Leist said that. That's a nice difference. Those
are very fbnctionally different tasks. I would want to see relevant
parameters change - depending on the level of description. And if you keep
doing that, you get very embarrassed, if nothing else. So, I think that there
is a way of testing it, although it may not do the opposite. It'll test me, it
may not test you. I'm, more interested in testing me anyway.
Bootsma: I want to come back to what Piet said about what happens if it goes
wrong. I think it's from Dick Schmidt's book ... he has this example where
you reach into a refrigerator to take out your pack of milk. And you sort of
forgot how much milk your kids drank. So, there's no information there,
specifying what to do. And you'll make mistakes. So, I can't really see how
somebody would reach out to a pack of milk without him knowing what's in
it.
Reed (alone):In fact, I would obviously make a prediction. Maybe that's the way to
do the experiment. Use an old paint container, a gallon, versus a glass one.
And in one case you can see only the weight of the container, and you might
be able to guess about the weight inside but you cannot see it. You need
learning about the specific situation to pick that up, whereas in the other
case you would not, and I would predict significant differences in the
reaching movement there. Then again, as you say, a lot of people would.
But if they were not found, and if he is doing his functional switches on the
task, and I kept not finding them, then it would be a problem for me. And
that's what I am trying to do. That is my research, to test my own theory; it
is to find functional shifts that do or do not yield functionally relevant
changes in things like task or dynamic parameters.
152 K.Roth & O.G.Meijer (Eds.)
Magilk I just have the feeling that Dick's view would predict the exact same thing
as yours would in the condition that you just described. So, I don't see that
situation you just described as being a way of differentiating the two views.
I think that he would predict the very same things.
Reed But he does need a theory of direct perception, or something close to that, to
explain how the dynamic information is obtained for the system. I mean,
he needs to account for the fact that, instead of perceiving the spatial
parameters of this pitcher, something dynamic, the weight, and so on, is
perceived. If that is convenient, and compatible with your theory, then
that's h e . But it's tricky.
A lot of times, because motor people are not concerned with the perception
side, it is easy to sort of ad hoc say: "Well, sure, that could be perceived." I
want to find out that it is a switch. It is a big shift from ten years ago that
people are talking about this stuff, the pick up of dynamically relevant
information. And that is the contribution of people like Gibson, Runeson,
and others.
Schmidt: Oh, sure, but that's a sleazy way to do it. (Laughter.) We always said,
ten years ago: "Well, he knows in some sense how heavy the glass is." But
that doesn't help us - you are right.
Schmidt: I have a feeling that there are experiments that discriminate between
the two ... . It seems to me that one of the critical things that the action
view, particularly Kelso's viewpoint, says, is that these things are highly
critically dependent upon sensory information, or afferent information. It
is not fashionable any longer to talk about the de-derentation work that
was so often talked about in the fifties and sixties, but that work still exists.
And I wonder how the viewpoint about dynamical systems - that seems to
require afferent information in order to keep the dynamics in its proper
-
state how that viewpoint handles the idea that movements can occur, not
very effectively but they can occur, in some circumstances under highly
deprived, or disturbed, sensory situations. I have a feeling that that's one
line of inquiry which might provide a cutting edge between the two kinds of
styles of motor control explanations.
Shea: I want to comment on Bill's and Dick's point. Both of you are taking the
assumption that eventually, in a long time, enough empirical evidence will
be supplied and then will weigh heavily on one side versus the other. And
then there will be a change. My assumption is that I don't think that
changes like that occur in science very often. Basically, what we're probably
going to see here is a Zeitgeist change, and that will take care of the whole
thing. (Laughter.) I really think that what is going to happen is that, since
the modelling difference is not so much of an empirical nature ... you
know, you cannot generate empirical data that really disproves or proves
one or the other. So, it's really a difference in the model systems - those
things change overnight.
One of the things that's occurring in the AI impact on memory is the use of
highly distributive memory models. And if you couple that with a
procedural notion, or operational view, of human memory, you kind of get
away from the whole problem that seems to be bothering people, which is:
that this motor programme is residing somewhere in the system. The
system is active all the time. You don't go out and cull this programme and
run it.
And, in fact, as you look a t the units and modules that comprise different
types of skills, memory becomes a skill. The more you perform a skill, the
better unite are assembled to do similar things, and in this way you don't
have a repository for motor programmes anywhere present in the system.
And that's a dynamic model which, in fact, kind of moves things a little bit
more toward the position of the action theorists.
This is probably what is going to happen, rather than some critical
experimentbeing conducted that will change things in the future.
Warren: I agree that those models are very interesting &ugh) ... . But still,
those models are going to have to be constrained by information and by the
physical constraints that are operating.
Shea: But they are consistent. That's the whole idea. The flow of information
changes.
154 K. Roth & O.G. Mewer IEdp.)
Warren:I just want to add that I agree. I think that you're saying the same thing
that I was eaying, intended to say. It's as much a sociological phenomenon,
the shift.
Heuer: Let me also comment on Bill and Dick. I think that they are both right in
some sense. In one sense, you cannot test those viewpoints experimentally,
and in the other sense you can, I think that one could draw a distinction
between, say, a 'viewpoint' and a 'theory'. A viewpoint is mainly a
prescription for admissable terms ... whether the resulting description is a
good one or a bad one can't be decided by way of experiments, but just by
criteria of usefulness, for example by seeing how many phenomena it can
be applied at. The second point is that within any framework there can be
formulated, say, statements about specific phenomena. And on that level
where, say, the fiamework is channelled into a theory that makes
predictions, experimental tests are possible. But I think that there will
always be a way out to change predictions within the one system of terms.
Thus,you can never make a critical test of, say, the viewpoint.
Warren: That's absolutely consistent. But there are theories and meta-theories.
And the level at which we distinguish the action system from the motor
.
system is that of meta-theory. ..
Meijer: I'm not sum whether I mean exactly the same or not. My impression is
that there is a difference in experimental style. You just do different things,
which is very obvious to the naive listener. And, there is a difference in
explanatory style. I don't think that the difference in experimental style is
really so crucial, because both sides could use data from the other side,
though having the feeling that it may not be very valid - or something like
that. Bill's research, I think, is a case in point. It is a very clear example of
a new experimental style, which then leads to discussions on the
explanatory level. (Addresses Heuer.) Is that a bit the same as what you
meant?
Heuer: I think not. What I fear is that if we have a certain set of admissable
terms, you will search for experiments where you can apply these terms.
For example, inter-manual interaction is a much more interesting topic in
the one framework than in the other one. The obvious reason is that the
Geneml Discussion 155
terms are particularly suited in the one approach to handle the problem,
while in the other one you need additional terms; you may have t o introduce
some kind of coupling, or whatever else, that is naturally there, essentially
ready in the language of the other approach. Another example would be
stimulus-response language. They have completely different kinds of
experimente, and these experiments are particularly suited for the kinds of
terms they use. That's roughly my impression. It's very basic.
Whiting: Ladies and gentleman (amazingly enough, there were no ladies present;
but that, of course, is not for a chairperson to notice) I am not going to start
a new discussion now. I am impressed by everybody's endurance. Like good
sportsmen, we should get out while the going's good. The discussion is still
going good. Let's stop now before the thing deteriorates.
Hands are shaken and pictures taken to establish the new friendships. After a
quick lunch, the audience return into diaspora and set out to start writing their
papers.
Dramatis personae:
Chapter 5
Martin: Michael, you have got the wrong idea about representation.
MacNamara: That's what I'm trying to tell him. A representation is a
construct of a scientist that lets us know the world. Now what one knows is
the world; one doesn't know the schema nor the representation ... .
Something like representation must form something in my head. I don't
know it directly, but I know of it. Consider meaning: I talk about the world
and there must be some way in which the world out there gets tied up in my
world. It's inour minds.
Member of Audience: What's the relationship between a presentation and a
representation?
Mace: The way that Martin talked about affordance, he was using good
English, but he was not using aEordance the way Gibson would. For
Gibson, there are very severe restrictions on the notion of aEordance, even
though it's certainly not a very precise term as yet.
(From the Gibson-Shaw discussion, in W.B. Weimer & D.S. Palermo - 1982,
Eda. - , Cognition and Symbolic Processes, Vol. 2, 229. Hillsdale, NJ:
Erlbaum. italics in original.)
SUMMARY
It is proposed to simply coin the debate between motor and action theorists : 'The'
motor-action controversy. The aim of the present Chapter is to investigate the
'nature' of this controversy. Aper a short historical introduction action theorists'
characterizations of the controversy are summarized. The idea presented by
motor theorists that action theorists erected a 'straw man' is shown to be
unproductive. It is argued that the difference between the two appnmches cannot
be suffikiently defined as one either in 'domain' or in 'level of analysis'. Although
negative conceptual demarcations are to the point, to date m positive conceptual
demarcation is tevuble. A developmental view on science is presented and it is
concluded that, at present, the differences between the two approaches can only
158 P.J. Beek & O.G. Mever
The field of study today described as 'human movement science' (Brooke &
Whiting, 1973). 'motor control and learning' (Schmidt, 19821, or the study of
'action systems' (Reed, 1982a) - to give but a few of its names - appears, in
retrospect, to have tried to establish its relative independence1 from 1967 onwards.
A t the time, Fitts and Posner (1967,p. 1) coined the term 'human performance'
for that branch of 'experimental psychology' which
Of course, 'human performance' had been studied from at least the time of
Aristotle; nevertheless, the end of the sixties marked the labour to give birth to a
new 'scientific community' (Kuhn's 1962 term). Specialized Journals started to
appear: the Journal of Motor Behavior in 1969, and the Journal of Human
Movement Studies in 1975.
Not everyone necessarily agreed with Fitts and Posner, but their work, as well
as Welfords (1968),marked the emergence of a new psychological field of study to
tackle questions concerning, e.g., the information capacity of discrete motor
responses (Fitts & Peterson, 1964),coding processes in motor short-term memory
(Posner, 19671, attentional demands of simple movements (Posner, 1969;Posner &
Keele, 1969),and the nature of motor programming (Keele, 1968).The new field
appeared promising and information processing was the 'style' of those
approaching it.
'Independence', that is, as a grand domain with ita own 'seiehtific community', i.e., in tams
of the sociology of sfience.
The Nature of the Controversy 159
The Journal Brain, Behaviour and Evolution started to appear in 1968, the Behavioral and Brain
Sciences in 1978.
Examples would be the approaches in which feedback and feedforward mechanisms are seen as
the source of the adaptive flexibility of control functions exerted at different levels of neuronal
organization (Brooks, 1979; Bizzi, 1980), as well as approaches in which the concept of a motor
programme is stretched to include either the elementa of responee elicited by sensory feedback
(Schmidt, 1982; Stemberg et al, 1978) or postural a4ustmenteanticipating and accompanying
movements, taking into account predictable and actual context elements (Gah6ry & Massion,
1981).
160 P.J. Beek & O.G. Mewer
From the late 1970s onwards the 'established approach was seriously challenged
by those who came to defend an 'ecological' approach to action. In several respects
Turvey's (1977)Preliminaries to a theory of action with reference to vision marked
the starting point of this new development. In assimilating the ideas of James J.
Gibson (1950 & 1966)on the intimate relationship between perception and action,
Turvey attempted to convince the scientific community of the need to attribute an
essential role to the continuous stream of perceptual, especially visual,
information. While relying heavily on the work of Nikolai A. Bernstein (cf. 1967)
and the group of Russian investigators who have followed his intuitions, as well
as on the analysis and amplification of the Russian views by Greene (e.g., 19721,
Turvey contributed significantly to the acceptance of Bernstein's formulation of
the degrees of freedom problem as one of the major problems to be solved by any
theory of action. Thus, the thrust of Turvey's 'preliminaries' was to suggest
integrating Gibson's and Bernstein's ideas into a single framework, considered to
be particularly fruitful in explaining action. 'Information processing' turned out
to be challengeable; generally accepted terms such as 'internal' and 'external'
started to lose their meaning; in action control 'peripheral' could take the place of
'central', and vice versa.
After Turvey's lead, a considerable number of manifesto-like papers appeared
setting out to establish the (meta)theoretical assumptions and methodological
principles of the new 'ecological approach to action' (e.g., Fitch & Turvey, 1978;
Fowler & Turvey, 1978; Turvey & Carello, 1981). In broad outline, these papers
precipitated the study of the relationships between perception and action a t an
'ecological scale', that is, as phenomena within the organism-environment
synergy rather than within the organism per se. A strategy was advocated to
search for natural, 'ecological', laws coordinating organism and environment
(e.g., Turvey et al, 1981). The 'action approach aimed a t revealing how organism
and environment mnstmin one another. In order to explain these mutual
constraints, the philosophical u n d e r p k n g s of the approach demanded that one
should
... resist taking out loans of intelligence; and 2. regard with skepticism, and
be prepared to jettison, any assumption, concept, interpretation, fact,
'Resolve', that is, for those who accepted information processing, and both the internal-external
and tho central-peripheraldichotomy.
TheNature of the Controversy 161
Around 1980, Kugler joined the dissidents. His contribution to the development
of the approach is apparent in the emphasis placed on non-equilibrium
thermodynamics. Kugler. Kelso and Turvey (1980) and Kelso et al (1980)
developed, theoretically and empirically, Turvey's (1977F suggestion that
'coordinative structures' may be conceived as self-regulating (oscillating)
systems. They argued that coordinative structures are members of the class of
thermodynamic engines qua dissipative structures, and that this fact should offer
the basis for understanding movement coordination and control with minimal
recourse to intelligence. The gist of these efforts was to seek the solution to
Bernstein's problem of functional non-univocality in principles of modern
dynamics.
Not much later, Reed (1982a) presented his Outline of a theory of action systems,
in sharp contrast to what he called 'motor systems' theories (or theories of
'indirect action'). On the basis of extensive criticism of 'traditional' theories of
motor control - e.g., the idea of 'mechanically specific neuromuscular units of
action triggered by external stimuli and/or controlled by brain commands' - Reed
elaborated and developed the insights and ideas of Gibson with more attention to
'action' than Gibson was ever able to give.
Notwithstanding differences of opinion, 'action theorists' appear to believe
themselves to be working on the theoretical and empirical development of a single
rival approach6 and, gradually, this has also become the belief of other workers in
the field - the sorcalled 'motor theorists' - who only now start to enjoin the invoked
polemics. The present Volume demonstrates that polarization is bilaterally
recognized7, and it appears to be justified to speak of a real 'controversy'. We
choose to dub the debate between motor and action theorists simply: 'The'motor-
action controversy 8.
The aim of the present Chapter is to investigate the 'nature' of this controversy. In
order to do so, we will:
summarize action theorists' characterizations of the controversy (Section 2);
investigate and reject the idea that a 'straw man' argument helps any further
(Section 3);
investigate and reject the idea that a 'domain' argument be sufficient to define
the differences (Section4);
investigate and reject the idea that a 'levels of analysis' argument be sufficient
to define the differences(Section 5);
investigate and reject the idea that a strong 'conceptual' argument be sufficient
to define the differences(Section 6);
present a developmental view on science and argue that, a t present, the
differences between the 'approaches' can not be satisfactorily analysed in logical
terms. It is only with reference to the historical dimension of both apprwches,
while acknowledging 'loose' versions of logical demamations, that a proper
evaluation of the present situation becomes possible (Section 7);
argue that philosophy of science may serve to direct further research in the field
of complex movement behaviour, whereas the field itself offers a case study
enabling one to evaluate theories in philosophy of science (Section 8).
2. Dissidents'contrastingconceptual frameworks
The 'leading', here, is of relevance since as soon as an approach hae established itself' Ph.D.
students will be tempted to elaborate on the views of their Professors 'without questioning':
differential degrees in 'maturity' of the social organization of a ncientific community will tend to
coincide with differential philosophical interest.
TheNature of the Controversy 163
Ecological psychologists (e.g., Gibson, 1979; Michaels & Carello, 1981; Turvey &
Carello, 1981; Shaw, Turvey & Mace, 1982) claim that the most central
assumption of their approach is animal-environment synergy, being the comer-
stone of direct realism. Organism and environment can not be isolated from one
another, since both owe their very identity to the other. In the course of evolution,
the organism's perceptual and action systems have become tailored to the
environment which, according to some ecological theorists, has in its turn been
(relcreated by the organism: a tangled loop which allows both the dualism of
organism and environment and that of mind and body to be rejected.
According to action theorists explanations and explananda ought to be
consistent with the basic ontology (e.g., Michaels & Carello, 1981;Shaw & Turvey,
1981;Turvey et al, 1981;Shaw et al, 1982)and the 'synergy' mentioned above is
presented as the 'primitive' for their characterizations of the differences between
'motor' and 'action' theories. The following may serve as a summary of the action
view on the differences between motor and action approaches:
In action approaches, control is not seen to reside inside the organism but is
localized in the organism-environment system as a whole; organism and
environment constmin one another, and coordination is written over the synergy
per se (Newell, 1984).Here, then, the large number of degrees of freedom of the
motor apparatus is a 'blessing' (Kugler, 19861, allowing for the subtleties of
flexible adaptive behaviour.
In motor approaches 'perception' and 'action' are treated as though they are
logically (but not empirically) independent (e.g., Schmidt, 1982; Hildreth &
Hollerbach, 1985). The goal of perception is to identify the objects present in the
environment whereas the task of the motor apparatus is to select an appropriate
response, set the parameters of the programme, and see over its execution.
Response programming does not constrain stimulus identification.
In an action systems account, perception and action are far more intimately
linked. They are, of necessity, integral (cf. Warren, 1988).Perceptual information
is constrained within the organism-environment synergy and cannot be
considered apart from the potential to act. What we perceive depends on the kind
of act we are engaged in, as well as on our ability to perform that act (Warren,
1984).Vice versa, how we act depends on what we perceive.
This, then, is the gist of the lines of argument adopted by action theorists to
establish their unique identitylo. The motor approach is still Cartesian at heart,
so that adopting an ecological foundation leads to the rejection of its premises.
The easy way out, of course, and not a priori wrong, is to state that nothing in
particular is 'meant' by terms such as 'programme', 'representation', or the like
(cf. the General Discussion in this Volume): action theorists have erected a 'straw
man' in their Quixotic imagination.
It is quite natural, of course, for the establishment to first 'absorb' piecemeal all
ideas presented by the dissidents, denying that any real differences exist. The
dissidents have 'exaggerated and should have known better. In 1984 (p. 477)
-
Requin, Semjen and Bonnet were - as far as we know the first to explicitly attack
ecologists' 'maximizing' conceptual differences.According to them:
loAn 'identity',by the way,to which we are sympathetic;but that is another story.
The Nature of the Controversy 165
Allegedly, 'action theorists' had tried to create a simplistic and outdated image
of 'motor' theories, an image which had more to do with their own idiosyncrasies
than with any real state of affairs. The less well-defined typical 'motor' concepts
are, the easier it is to accuse action theorists of having erected a 'straw man'. In
principle this would lead to claiming that there are no 'real' differences at all: a
'straw man argument'. Harvey (1985,p. 1591)stated:
'As long as we don't ascribe minds to plants just because they have DNA,
Harvey seems to say, we may go on referring to 'internal representations' in
humans without the risk of being accused of dualism. If now indeed Harvey
intended to offer what we have called a 'straw man argument', no real differences
between the approaches should have been allowed to remain. Harvey, however,
goes on to describe what he sees as shortcomings of the ecological approach - as
did Requin et al (1984). Since such shortcomings are relevant insofar as an
essentially different approach is at hand (cf. Lakatos, 19701, Harvey leaves us
where we took off: there are differences between the two approaches - maybe not of
the nature proclaimed by the action theorists, but they are of relevance. This is, of
course, not to say that no straw man was ever erected by action theorists but it just
166 P.J.Beek & O.G.Meiier
happens to be the case that the unmasking of straw men offers no demarcation
criterion whatsoever to distinguish between the two approaches. Hence, for the
present analysis, the straw man argument is unproductive. What, then, is the
nature of the controversy?
Several authors have claimed that the nature of 'the' motor-action controversy
may in fact be characterized as a difference of 'domain' - task and context
selection being responsible for the differences and not the theoretical positions per
se. Harvey appears to adopt such a 'domain argument': Gibsonian theory being
invalid whenever memory plays an important role, or, for example, when we
want to polemically affect the perceptions of our audience.
Shepard (1984, p. 419) comes close to this type of analysis in stating:
Although I agree with Gibson that the brain has evolved to extract
invariants under favorable conditions, I also presume that it has evolved to
serve the organism under less favorable conditions of nighttime,
obstructed, and spatially or temporally limited viewing and, even, of
structural damage of the brain itself.
Van Wieringen (1988, this Volume) gladly accepts this 'rehtation' of ecological
idiosyncrasies:
We agree with Epstein that both approaches have their own preferences with
respect to empirical 'domain'. And indeed, 'domain' characterizations of the
differences abound laboratory vs. real world (cf. Heuer, 1988). ballistic vs.
oscillatory movements (cf. Schmidt, 19881, muscle exercises vs. functional
exercises (cf. Mulder & Hulstijn, 1988), etc. We will, however, demonstrate that
the 'domain argument' - as far as the demarcation between the two approaches is
concerned - evaporates when confronted with its own premises.
for D2:
would emerge, ' Z representing initial conditions and defining the 'domains' in
question. In fields where mathematical statement of 'law' is in its infancy - like
the field of movement behaviour - verbal statements of 'law' obey, in principle, the
same logic.
l1 We happily accept the circular nature of our defmition of 'field: a grand domain in which,
ultimately, the same grand laws are believed to be valid by the scientific community, no matter
whether these laws are known' or expected to be uncovered in the tkture.
168 P.J. Beek & O.G.Meiier
Our analysis of the actual situation leads to leaving the 'domain argument' where
it belongs: it is a speculation about the future. This is not to say that we think such
a future development to be unlikely but it would be a fortune telling anachronism
to treat that intuition as an appropriate demarcation of the actual differences.
There are differential preferences for particular domains - as Epstein (1986)
pointed out - but it happens to be the case that identifying these domain
preferences offers no sufllcient demarcation between the two approaches. Apart
from speculating about possible futures, the question remains: What, a t present,
is the nature of the difference?
Gibson tends to leave the organism, if not empty, stuffed with foam rubber.
Ecologists, allegedly, neglect a level 'inside' the organism, i.e.. that of the
nervous system. Within the realm of a 'levels of analysis' argument, however, it
becomes relatively meaningless to compare, let alone reconcile, different
appronchesl4. In a way, therefore, a critique like Pribram's is vacuous, unless
one would like to dismiss sociology for its stuf€ing individuals with plastic, or
psychology for closing its eyes to astronomyl5. This relative impossibility of
comparison is acknowledged in Schmidt's (1988,this Volume) formulation of the
argument:
In the action view, the explanation is in terms of what the motor view
would term 'input-output' relations, while in the motor view it is usually in
terms of mechanisms of motor control. More fimdamentally, the difference
seems to be in terms of the levels of analysis at which the theorizing occurs.
... So, in considering the action versus motor approaches, I a m not certain
that it even makes sense to draw the comparison.
Apparently, the 'levels of analysis argument' is appealing, but those who invoke a
'levels of analysis argument' tend to over-simplify matters: precisely what levels
of analysis' are distinguished, and how the nature of their relationships is seen,
are, in themselves, dependent on one's scientific stance. Construing 'levels of
l4 Relatively, that is, since levels are not completely independent but do constrain one another:
ultimately, theories concerning different 'levels' have to be mutually consistent and some 'rule of
correspondence'should be agreed upon.
l5 Pribram's complaint, of course, is not vacuous insofar as action theorists do claim to be
describing the 'inside' of an organism. It is true that Gibson creates a host of problems concerning
the physiology of the organism-environment relationship (cf. Epstein, 1986, p. 61, but to attack
action theorists' rejection of the internal/external dichotomy calls for finer phrasing than
Pribram's 1982(cf.Marr, 1982).
170 P.J. Beek & O.G. Mewer
... the argument that the two approaches are operating at completely
different levels of analysis is probably too strong (this Volume).
The conception of a 'hierarchy of levels' is very old16. The central tenet of modern
theorizing on levels is that no phenomenon can be completely reduced to
phenomena a t any other level: levels constrain rather than determine one
another. If two levels - L1 and L2 - are to be distinguished within one field of study,
specific relations have to exist:
"It is indeed true that there is an inherent haziness in level-counting ... ."
(Hofstadter, 1979, p. 13). Behaviourism recognized, in principle, one level of
l6 Twentieth century formulations have been strongly influenced by system theories, e.g., Von
Bertalanfi (19731, Miller (1978).
The Nature of the Controversy 171
Concerning the third requirement, i.e., agreement on the nature of Y 4 V, the rule
of correspondence, as well as agreement on the primary direction of constraint,
uncertainty r e i p s . Linking ecological information to traditional stimuli
l7For example, the apparent spontaneity of behaviour when 'external' eliciting stimuli cannot be
identified,and the diversityofbehavioural (re-)actionsunder 'identical'stimulus conditions.
18 Although Bruce and Green use Mads concept of 'computation', their levelcounting along
disciplinaryboundaries is inconsistentwith Marfs transcendingthesetraditional boundaries.
172 P.J. Beek & O.G. Mever
confronts one with the "sheer difficulty'' (Marr, 1982, p. 30) of the detection of
information. Specifying movements and postures (Reed, 1988)in their relation to
displacements of (parts of) the human body requires a problematic philosophical
reorientation (Tamboer, 1988). Connecting the thermodynamics of physical
machines with cognitive functions, e.g., imagining, remembering, dreaming, is
an enterprise, at the moment hardly conceivable (Van Wieringen, 1988).
The fact that the direction of constraining is not given but chosen, may be
examplified by observing how 'naturally' Weiss (1941)counts his levels 'bottom
up', and Saltzman (1979)'top down', or by comparing the views of Schmidt and
Reed. Schmidt's (1988)reference to Stark's diagram reveals a bottom up approach
whereas in Reeds (e.g., 1982a, 1984, 1985 & 1987) top-down approach an
appropriate description at the level of the animal environment interface
determines the sort of description that is appropriate a t 'lower' levels of
description:
In sum, our analysis of the situation leads to rejecting the 'levels of analysis
argument'. There are, of course, differential preferences for particular levels but
it happens to be the case that to identify these preferences offers no sufficient
demarcation criterion between the two approaches. The present popularity of the
'levels of analysis argument' will, we predict, turn out to be unstable: one may
want to return to the 'domain' argument or decide to agree with the ecologists and
opt for a strong conceptualargument.
... while there are many differences between the views, for me the most
critical one - and for which there does not seem to be any middle ground - is
the action proponents' unbending insistence that movement cannot be
organized centrally (even in part) or, in other words, that there be no
centml representation of action (italics in original).
Neo-Gibsonians, like Shaw and Turvey (19811,have tried to overcome this lack of
clarity by supplementing 'dordances' with 'effectivities' to form a duality (not a
dualism) of structure. Formalized
that '&ordance' is "certainly not a very precise term as yet." We agree, and
expect the concept to undergo further transformations andor ramifications.
7. A developmentalview
Since the concepts of 'affordance' and 'coordinative structure' do not meet the
criteria for an ecological hard core, we fail to see a 'strong conceptual argument'
sufficiently drawing the lines of demarcation between motor and action
approaches. Apart from Schmidt's negative conceptual demarcation, however, it
may still be the case that some form of positive conceptual argument is
appropriate.
In 1982, Cutting (p. 201) formulated his characterization of the nature of the
controversy between information processing and ecology, stating:
19 It is immaterial to the present argument that 'animal-environment mutuality' is, in fact, not a
metaphor at all but a basic assumption. The root in Cutting's 'root-metaphor', we argue,
introduces the historical dimension.
The Nature of the Controversy 177
2o 'Style',here, is not meant in the grand conceptual sense Hacking(l983,p. 12%128) give8 i t
21 Bateson's(1906)coiningtheterm 'genetics'is a case in point (cf. Meijer,in preparation).
22 One of the present authors had the pleasure of entering it as an unbeliever whereupon noticing
that knowing to be 'tricked' did not make any difference whatsoever: it was actually the that
was moving.
178 P.J. Beek & O.G. Meiier
Surely, this is comfort for the discomforted, and it is logically appealing. More
interesting, however, is the historical fact that motor researchers, while they
could have done Warren-like experiments, did not do 80.
Action theorists are using terms like '&ordance', 'higher order invariant', 'optic
array', 'ecological information', 'tuning', 'action', and the like, while avoiding
commands, programmes, and internal representations; they champion
ecologically valid experiments in which the relationships between organismic,
and environmental constraints are manipulated and observed; their preferential
level of analysis is the 'functional' one, i.e., the organism environment interface
where boxes are notoriously lacking.
One may be tempted to describe the difference as one between 'research
traditions' (Laudan, 1977), but such a characterization would require an agelong
history. One may want to opt for a difference of 'style', were it not that 'style' has
been coined to describe grand differences which render communication actually
impossible (cf. Hacking, 1983); this Volume embodies the fact that
communication, although difficult, is possible. Differences in Kuhnian
'paradigm' (cf. Michaels and Carello, 1982) would miss the fact that several
paradigms are involved whereas their precise relation is as yet unclear. A
difference in 'kinds of analysis' has been proposed (Van Wieringen, 1988; cf. the
Discussion in this Volume) which is, of course, to the point but in our opinion not
precise enough: it misses the historical dimension.
According to the present authors, a developmental view is needed, implying
conceptual differences but no strong conceptual argument; allowing for
differential preferences as to domains and levels, without suggesting them to be
demarcation criteria. Motor and action approaches have different sources of
inspiration. Both are trying to develop a unified methodology. Motor approaches
may be more mature in sociological development, but even in them no Lakatosian
'research programme' has evolved as. yet: no sufEcient conceptual unity, no
internationally recognized organization, while papers in specific journals - which
do exist - represent so far almost always the minority in the lists of references of
typical motor publications23. The making of a discipline takes time.
23 One may have noted that in this Volume references to field-specific journal8 are relatively
rare. Instead, referencesto more generaljournals and conference-pmeedingaabound.
The Nature of the Controversy 179
It has been our inclination in this Chapter to offer a demarcation of motor and
action approaches that is consistent with the logic of both. We assumed such a
demarcation to exist. Of course, one is free to reject this assumption. In our
opinion, however, the advantages of emphasizing the historical dimensions of the
differences are much greater than those of the standard accounts. It accepts
differences in inspiration and conceptualization, while avoiding the dangers of
incommensurability and eclecticism. It acknowledges differences in
methodology, while leaving the route open to empirical confrontation. This
recognition of differential methodologies may make scientists in the field more
aware of what is going on and what should be next. In the near future, both
approaches will have to present themselves as powerfully as possible, both
empirically and theoretically. Over time, social development may lead then to full-
fledged 'hard cores'. Or, and this is what we would bet upon, principled changes
in both approaches will occur before the controversy is transcended into a unified
'hard core'.
We have offered the suggestion that the philosophy of science can contribute
si@cantly to an understanding of the nature of 'the' motor-action controversy.
From that perspective we were able to demonstrate that the differences between
motor and action approaches should be described in historical rather than logical
terms, and that many current accounts of these differencesare crippled.
In fact, of come, we have also used 'the' motor-action controversy as a case-
study to contribute to the philosophy of science. It appeared that standard ~~I-XIIS
like 'paradigm', 'research programme', or 'research tradition' are not
appropriate. Those in the field of motor behaviour probably knew this all along:
the general use of the word 'approach tacitly points to a rejection of traditional
philosophy of science. Philosophy of science, then, may learn as much from the
field of complex movement behaviour, as those in the field from the philosophy of
science. Cutting (1982)both ways!
ACKNOWLEDGEMENTS
(New Haven), Ruud van der Wee1 (Amsterdam), John Whiting (Amstsrdam), Piet
van Wieringen (Amsterdam), two anonymous referees, and the participants of
the Conference which led to this book.
REFERENCES
Chapter6
SUMMARY
An expert table tennis player was required to perform attacking forehand drives
under two ball velocity conditions with normal binocular vision. Under the faster
condition he also performed monocularly. Film analysis results suggest that
under binocular conditions two sources of information were used for timing the
initiation of the drive, viz. ball location and time to frontal eye plane (FEP) - the
time period that would elapse between the moment of movement initiation and the
moment the ball would cross the frontal plane through the eyes. Under the
monocular condition, in which localization of the ball is supposed to be more
d i f f m l t , the subject tended to slow down his movements. Moreover, while under
the binocular conditions the subject was shown to execute a consistent drive,
under the monocular condition he adapted his drive to the very slight variations
in time to FEP. It is tentatively concluded that under normal (binocular)
Conditions the player simply runs off a standard 'motor programme', while under
the more uncertain (monocular) condition 'programme parameters' are
adaptively varied to match the changes in environmental information.
1.Introduction
action have done before him (e.g., Lee, 1976 & 1980; Von Hofsten, 1983). Herein
lies, perhaps, one of the most important contributions of an action systems
approach to the field of motor control and learning. It has enlarged the scope of
traditional motor control research and theorizing from a component-analysis
approach to a holistic approach.
The purpose of this Conference was to critically evaluate the present
controversy in the field of motor control between what have come to be known as
'motor systems approaches' (e.g., Adams, 1971; Schmidt, 1975) and 'action
systems approaches' (e.g., Kugler, Kelso & Turvey, 1980; Lee, Lishman &
Thomson, 1982;Lee et al, 1983;Reed, 1982;Turvey, 1977). Such an evaluation is
beyond the scope of the present study, which explores a much more restricted
topic. It aims at gaining some insight into the constructive role of visual
information sources for the execution and preservation of ongoing action. The
approach which is followed is in line with the work of Lee, referred to earlier.
However, the adoption of such an approach should not be considered to imply an a
priori rejection of motor systems theories; in looking for the mechanisms
subserving the guiding function of visual information such an approach might, at
least in principle, be useful.
The experiment that will be reported here is a small pilot study, conceived to
distinguish between different potential sources of visual information guiding a
real-life action, viz. the attacking forehand drive in table tennis.
David Lee and hie co-workers have provided a host of data, all in support of the
notion that a specific perceptual variable, tau - the inverse of the rate of dilation of
the retinal image - , is used to accommodate action, be it in diving gannets (Lee,
1980), long jumpers running up to the take-off board (Lee, Lishman & Thomson,
1982),car drivers braking to avoid collision (Lee, 19761,or subjects jumping up to
hit a falling ball (Lee et al, 1983). This variable, it is claimed, directly specifies the
time-to-contact between the actor (more specifically, the actor's eye) and the
environmental object of interest, at least in situations in which the velocity of the
object relative to the actor is constant. Von Hofsten (1986)argues in this respect
that, although time-to-contact can indeed be used in the performance of tasks in
which the eye is approached directly by an object, it cannot be the sole determinant
in, for example, catching. If an object is to be contacted at a variable distance
away from the eye, additional information about the object in question must be
available as well.
Visual Control in Table Tennis 191
In the experiment to be reported the subject was an expert table tennis player.
Such expert performers can be distinguished from novices by, amongst other
features, highly consistent movement characteristics (Franks, Weicker &
Robertson, 1985; Spaeth-Arnold, 1977; Tyldesley & Whiting, 1975). Such highly
consistent movement characteristics imply that, if the action is to be successful,
the performer has to be very accurately tuned to the environment.
Three sources of information have, in different studies, been put forward as
determining the temporal initiation point (TIP)of an action like a table tennis
drive:
01. Distance between the player and the object to be hit (Eckert & Hamdorf, 1980);
02. Location in extracorporeal space of the object to be hit (Tyldesley, 1980);
03. Time to contact as specified by tau (Lee et al, 1982 & 1983).
-1.If this kind of information (distance between the player and the object to be hit)
is, in fact, used for initiating the drive, the performer would, necessarily, have to
alter his movement velocity if objects were to approach with different velocities;
-2. The same applies if extracorporeal location of the object to be hit is used, unless
the performer is able to adjust his position in space prior to the execution of the
drive itself in such a way that the time to contact from the extracorporeal location
of the object is kept constant;
-3.If a particular time to contact, as specified by tau, is used to determine the
temporal initiation point of the drive, a constant movement velocity and movement
time would be expected.
Conflicting results with respect to any one of these variables have been reported in
the literature but, it is maintained here, such differences may be more artefactual
than real in the sense that they may be brought about by the differing constraints
imposed upon the player in the different studies. In the Lee et a1 (1983)study, for
example, environmental information was reduced, and in the Tyldesley (1980)
study the use of only one approach velocity meant that all three information
sources previously mentioned were confounded. In the present study a top table
tennis player was required to execute an extensive series of attacking forehand
drives under conditions differing with respect to vision and ball velocity.
2.1. Method
Subject
The subject was a former Dutch national table tennis champion. He was 23 years
of age, right handed, and volunteered for participation in the experiment.
192 R.J. Bootsma & P.C.W. van Wieringen
Task
The experimental set-up was a normal table tennis situation with a Sitco RII-s
ball projection machine standing opposite the player behind the table. Balls were
delivered to the subject at a constant rate of 40 per minute. The subject was
required to smash the balls as hard and as accurately as possible onto a circular
target (55 cm diameter) on the opposite side of the net, utilizing an attacking
forehand drive technique.
Procedure
The subject was tested under three conditions, viz. 1. BILO (binocular vision, low
ball velocity - horizontal ball velocity at TIP 4.1 dsec), 2. BIHI (binocular vision,
high ball velocity - horizontal ball velocity at TIP 4.5 dsec), 3. MOHI (monocular
vision, high ball velocity - horizontal ball velocity at TIP 4.5 dsec). In the latter
condition the non-dominant (left) eye was covered. Under both ball velocity
conditions the flight paths were almost identical; comparable flight times
(preview times about 550 msec) were obtained by moving the robot backwards
under the higher velocity condition. The three conditions were presented
blockwise. After a number of acclimatization trials, which were repeated under
each of the three conditions, film recordings were made of, respectively, 25 trials
in condition BILO, 25 trials in condition BIHI, and 12 trials in condition MOHI.
These recordings were made using a Teledyne camera running at 150 frames per
sec. Following development, the films (Kodak 4 x Reversal 400 ASA) were
projected by means of a NAC (DF-16b) projector onto an opaque screen. Frame by
frame, the coordinates of ball, bat, and right eye were digitized, using a SAC 14"
XY tablet, Connected to an Apple 11microcomputer.
After smoothing, using a second order recursive Butterworth filter (cut-off
frequency 8.0 Hz),'information' parameters, such as location of the ball in space
relative to the table, distance between the ball and the eye, and time to Frontal Eye
Plane (FEP) were calculated for the temporal initiation point of the drive. Time to
FEP refers to the time period which would elapse between TIP and the moment
the ball, when maintaining constant velocity, would cross the frontal plane which
passed through the eye at TIP. This.'virtual' time (virtual, because the ball is
intercepted by the bat) is considered to be very close to 'time to contact' as defined
by Lee (1980). The latter term, however, might lead to confusion with the time
period between TIP and the moment of ballhat contact, and will, therefore, be
avoided. In addition to the aforementioned 'information' parameters, a number of
'execution' parameters, such as spatial location of the bat at the moment of
initiation and at ballhat contact, movement duration, and mean bat velocity, were
derived.
Visual Control in Table Tennis 193
From the trials which were recorded, some, because of imperfect registration,
were unsuitable for analysis. Nineteen trials under the BILO condition, 21 trials
under the BIHI condition, and 10 trials under the MOHI condition, were available
for analysis.
TTF-TIP: estimated time to FEP at the temporal initiation point (TIP) of the
drive; BAL-TIP: horizontal position of the ball in space relative to the
leading edge of the table at TIP; DIST-TIP: distance between the eye and the
ball at TIP; DURAT: duration of the drive; BAT-TIP: horizontal position of
the bat relative to the leading edge of the table at TIP; BAT-HIT: horizontal
position of the bat relative to the leading edge of the table at the moment of
contact between bat and ball; MBV: mean bat velocity during the drive.
VARIABLE
I
LOW VELOCITY iIGH VELOCITY
N=19 N=21
STATISTIC P
2.2. Results
The results from the normal (binocular) vision conditions are summarized in
Table 1. They indicate that mean times to FEP did not differ significantly between
TTF-TIP: estimated time to FEP at the temporal initiation point (TIP) of the
drive; BAL-TIP: horizontal position of the ball in space relative to the
leading edge of the table at TIP; DIST-TIP: distance between the eye and the
ball at TIP; DURAT: duration of the drive; BAT-TIP: horizontal position of
the bat relative to the leading edge of the table at TIP; BAT-HIT: horizontal
position of the bat relative to the leading edge of the table at the moment of
contact between bat and ball; MBV: mean bat velocity during the drive.
the conditions BILO and BIHI, although time to FEP was somewhat more
variable under the former condition than under the latter.
With respect to the extracorporeal location of the ball at TIP (relative to the
leading edge of the table in the direction of the player) no significant difference
between the conditions BILO and BIHI was found. The distance between the
subject and the ball at TIP, however, was significantly different under the BILO
and BIHI conditions. Given the non-significant difference between the times to
FEP under both conditions, the latter result was, of course, to be expected.
The 'execution' parameter, movement duration, showed no significant
difference between the two conditions. The very low variability of the parameter is
remarkable.
An explanation of the simultaneous finding of non-significant differences in
times to FEP and ball locations at TIP between the two velocity conditions is found
in the fact that the subject stepped backwards a little under the high velocity
condition prior to the drive proper, thus transposing his entire body position with
respect to the leading edge of the table. This is reflected in the differences in
position of the bat at TIP and at ballhat contact between both conditions.
In Table 2 the values of the dependent variables under the BIHI condition are
further compared with those obtained under the MOHI condition. Again, the
mean times to FEP did not differ significantly.
An interesting finding concerned the mean location of the ball in space at TIP,
which differed significantly under the two vision conditions, while ball velocity did
not change. Although the length of the drive remained unchanged, mean bat
velocity decreased and movement duration increased under the monocular
situation. Movement time also became more variable.
3. Discussion
The results obtained under the binocular conditions suggest that the subject made
use of two kinds of information to guide his actions, notably time to FEP and ball-
location. By stepping backwards under the higher ball velocity condition, the drive
could be initiated at the moment the ball occupied about the same location in
space, while time to FEP at TIP was also held constant. The very low variability of
movement duration, indicating high consistency in movement execution,
confirms the findings in expert players reported by Tyldesley and Whiting (1975)
and Franks et a1 (1985).
196 R.J. Bootsma & P.C. W.van Wieringen
With respect to the results obtained under the binocular conditions, it was
speculated that the 'triggering' of the motor programme specifying the forehand
drive might be under control of visual information. Post-hoc considerations led to
further explorations. These explorations were motivated by the possibility that,
although the low variability in movement execution may indeed be seen as one
more example of the consistency of top-players, it may, nevertheless, be taken as
signalling the influence of visual control factors. This latter point of view
motivated the assessment of correlation coefficients between time to FEP at TIP on
the one hand, and peak bat velocity and peak bat acceleration on the other. Under
the BILO condition time to FEP correlated positively with peak bat velocity (rp =
.42, p < ,051and peak bat acceleration (rp = .43, p < .05).Under the BIHI condition
these correlations were again positive, although somewhat smaller - the
correlation coefficients being .43 (p < .05)and .36 (p c .lo), respectively. It should
be noted, however, that under the latter condition variations in time to FEP were
extremely small; therefore, significant correlations would be diflcult to obtain
(restriction of range).
Visual Control in Table Tennis 197
Thus, it appears justified to suggest that in the binocular condition the player
operates within a specific bandwidth with regard to time to FEP (in the sense of
starting his drive within the period defined by this bandwidth), and runs off his
drive in the same, ballistic, way from there on. That such a control strategy would
indeed give rise to positive correlations, as have been found, is clear if it is realized
that the bat is continuously accelerating until the ball is hit, implying that a
longer time to FEP at TIP is associated with a higher peak ball velocity and
acceleration. Interestingly, in the monocular, high ball velocity condition negative
correlations between time to F'EP at TIP and both peak bat velocity (rp = -.75,p c
.Ol> and acceleration (rp = -.77,p c .01)were found. Such negative correlations
would be expected if time to FEP at TIP is not used for triggering the same motor
programme time and again, but for 'guiding' the action in eliciting a slower drive
when having a relatively large, and eliciting a faster drive when having a
relatively small value.
Following these lines of argument, it is tentatively concluded that the top player
under the normal (binocular) conditions resorts to highly consistent forehand
drives, supposedly governed by a motor programme of which the parameters are
kept constant. When the situation is less familiar and uncertainty about ball
location and trajectory increases (monocular condition), the player appears to
adapt the execution of his drives to the characteristics of the visual information at
TIP. This might be achieved by specifying different values for the force parameter
of the motor programme from trial to trial. Whether or not a motor programme,
or some other mechanism, is subserving the perception-action coupling can not,
however, be deduced from the results of the present experiment as it is not known
whether the observed trial-to-trial adaptation is accomplished prior to or during
the drive. It should be borne in mind, though, that Lee et a1 (1983)report evidence
against the utilization of a motor programme in subjects jumping up to punch a
falling ball. A n alternative mechanism has, however, not yet been put forward.
It goes without saying that the interpretation of the results obtained in this N=l
study is still speculative. Replication of the experiment on a larger scale is
momentarily under way. At least, the tentative conclusions, drawn about the
modes of control exemplified under the normal (binocular) and more uncertain
(monocular) conditions, serve as hypotheses for these replications. The way in
which they were framed illustrates the point, hinted at in the introduction, that
looking for relations between visual information and actions attuned to it, does
not a priori exclude explanations for such relations in 'motor systems' terms.
198 R.J. Bootsma & P.C.W. van Wieringen
ACKNOWLEDGEMENTS
REFERENCES
Reinoud J. Bootsma
The Netherlands.
Chapter 7
TRANSFERAPPROPRIATE PROCESSING
A FRAMEWORK FOR CONCEFTUALIZINGPRACTICE EFFECTS
INMOTORLEARNING
"imothy D. Lee
SUMMARY
The relationship between practice conditions and transfer conditions is not a new
issue. One of the first theoretical accounts of this relationship was formulated by
Thorndike (1914). His 'theory of identical elements' posited that practice on one
task will facilitate transfer to a second task only to the degree that the two share
common 'elements'. Critics of the theory argued that the comparison of two
different tasks is quite difficult since the performance of one task on two separate
occasions is never entirely the same. Nevertheless, Thorndike's ideas stimulated
considerable theoretical debate and empirical research. McGeoch and Irion's
seminal chapter on 'Transfer of training' provides a comprehensive review of this
work (1952, chapter MI.
A shift in theoretical focus away from the task-oriented approach, however,
precipitated a . decline in research activity regarding the practice / transfer
relationship (Schmidt, 1982). More recent concerns about the relationship between
practice and transfer have adopted a process-oriented approach, focusing on the
compatibility between the feedback conditions available during both practice and
transfer. The results of these studies, in general, supported the position that
transfer performance was facilitated to the extent that the feedback conditions
during transfer matched the feedback conditions available during practice (e.g.,
Lee & Hirota, 1980; Lee & Magill, 1985a; Magill, 1983; Reeve & Mainor, 1983). As
Lee and Magill (1985a) cautioned, however, much of the evidence in support of
this view was derived from studies using simple arm displacement tasks and
these tasks appeared to constrain the learner to pay particular attention to the
movement-produced feedback in order to learn the task. Thus, the generalizability
of this concept to more complex tasks was questionable.
However, a concept that does appear to be generalizable to a wider range of tasks
is the concept of 'transfer-appropriate processing' (Bransford et al, 1979; Morris,
Bransford & Franks, 1977). While maintaining the essential issue of the practice-
transfer relationship espoused in the previously discussed conceptualizations, the
notion of 'transfer-appropriate processing' allocates a greater role to the learner
as an active processor of information.
AS suggested by Morris et al(1977, p. 5281, transfer-appropriate processing:
Note that on a second transfer test, following a 10-day retention interval, the difference between
the blocked and random groups on the blocked transfer test was no longer statistically significant.
While this finding poses a problem for transfer-appropriate processing, a recent study that also
employed a blocked transfer test found significant random group advantages following both 10-
minute and one week retention intervals (Gabriele, Hall Q Buckolz, in press).
The Shea and Morgan experiment is an exception to the typical design in that the pattern
illustrations were still present during transfer. However, it is not known whether subjects in the
random and blocked groups were actually using the illustrations in order to cue retrieval. Indeed,
one would suspect that if the illustrations were actually being depended on, a similar number of
errors in movement production would have occurred. Shea and Morgan found, however, that the
random group had significantly fewer errors than the blocked group, especially during random
retention trials. Random practice may facilitate transfer, then, because these 'internal' sources
are better developed when they must be relied upon or more expedient when external sources are
also available.
Transfer-AppropriuteProcessing 205
A n interesting study reported by Zimny (see Shea & Zimny, 1983 & 1988) revealed
that the verbal reports from subjects in the random group differed considerably
from those in the blocked group. Zimny found that blocked subjects tended to
'automate' their processing during acquisition whereas random subjects tended
to make more comparisons and evaluations of the patterns between practice
trials. These verbal report data supported the argument that the patterns were
learned more distinctly by the random group. During transfer (without the
pattern illustrations available) each of the patterns were subsequently more
readily memorable as a result of this distinctive processing, thereby reducing the
confusion of spatial layout amongst the learned patterns that seemed to be evident
for the blocked group.
A somewhat different view espoused by Lee and Magill (1983 & 1985b) is
consistent with Shea and Zimny's notions when both are considered within a
transfer-appropriate processing framework. Drawing upon research in the
cognitive literature by Jacoby (1978; Cuddy & Jacoby, 1982) we argued that during
random practice a subject engages in more problem-solving analyses because the
results of a previous analysis for a specific action plan (i.e., movement pattern)
are no longer available. The random subject must 're-solve' the problem during
practice by repeated regeneration of action plans. Conversely, the blocked subject
may revise an action plan that currently exists in working memory on each
subsequent practice trial without the need for the more extensive problem-solving
analyses that must be undertaken by random subjects. The critical point is that
the process of learning to generate and implement action plans promoted by
random practice conditions facilitates transfer because these action planning
activities are more similar to the processing requirements of the transfer test
than the learning activities promoted by blocked practice conditions.
... the problem of refining one's general knowledge and skills is not
necessarily equivalent to the problem of learning inputs so that they can
later be retrieved. Instead, growth involves learning fmm experiences in
ways that facilitate subsequent transfer (italics in original).
Indeed, this quote underscores a basic premise upon which Schmidt's schema
theory was conceptualized: the schema representing a knowledge structure that
is developed fmm previous experiences yet utilized for the purpose of generating
and evaluating actions not previously experienced. The process by which the
schema is developed, however, does not appear to be well researched, and, in fact,
may account for why empirical tests of the variability of practice hypothesis have
been equivocal.
A methodological concern amongst previous variability of practice studies may
have resulted in this equivocality. Reviewing the published research that used
adults as subjects, we (Lee, Magill & Weeks, 1985)noted that the primary concern
of these investigations was a comparison of transfer performance following a
period of practice involving either one or multiple parameters of the learning task
(i.e., constant versus vaned practice). However, what was not consistent amongst
these studies (and probably not considered important a t the time) was the manner
by which the variable practice conditions were structured. Some studies had
structured their variability conditions in a blocked order whereas others had
ordered these variability trials randomly (see Lee et al, 1985, Table 1, for a
summary of studies reviewed). In hindsight (based on current knowledge about
contextual interference effects) it was not surprising to us that an advantage for
variability of practice conditions relative to constant practice conditions was
almost exclusively revealed only when these variability trials had been conducted
in a random order. To empirically test our observation we conducted two studies
that each contrasted a constant practice group with two variability groups (one
variable group was blocked, the other random). The results supported our
hypothesis that variable practice conditions only facilitated transfer relative to
constant practice conditions when the variability of practice had been conducted
in a random order. For present purposes however, the focus of discussion
concerns the reason why blocked-variable practice resulted in transfer that was
no better than constant practice.
Transfer-AppropriateProcessing 207
With the exception of the parameter change trials (3 out of 60 total trials in our
experiments), the blocked-variability and constant practice groups performed
similar processing activities during practice - the revision of an action plan
currently available in working memory. From a transfer-appropriateness view,
increased variability of practice in the blocked group (relative to the constant
group) failed to facilitate transfer because the practice conditions did not promote
processing that significantly differentiated the learning activities of the two
groups. As succinctly pointed out in the previous Bransford et al quote, 'growth is
not the learning of but rather the learning from experiences. Given the similarity
of processing activities promoted under both constant and blocked practice
conditions, similar transfer performances would also be expected.
2.3. Frequency of KR
2.4. Bandwidth Mi
-
10 -
trial was 'correct', so I should repeat what I just d i d . This type of processing
activity would be expected to encourage consistency of responding since the
subject would be learning how to repeat a successful action plan: subsequent
attempts, followed by KR or the absence of KR now being interpreted not with
respect to the goal, but rather with respect to the relative success of repeating a
previously successfbl action plan.
During acquisition trials Sherwood found that the 10% bandwidth group
received verbal KR on 31 out of 96 trials. The 5% bandwidth group received verbal
KR on 54 trials. The 0% bandwidth group, of course, received verbal KR on every
210 T.D. Lee
trial. Given the above speculations on the nature of processing activities in the
absence of KR the expectation would be that the 10% bandwidth group learned to
become more consistent than the 5% group which, in turn, learned to become
more consistent than the 0% group. The no-KR transfer trials in Figure 1 (post-
acquisition blocks 1-3)lend support to this speculation since the 10% bandwidth
group was significantly more consistent (i.e., less variable error) than the other
two groups. Further, while no difference appeared between the 5% and 0% groups
on the first no-KR block of trials, the 5% bandwidth subjects were able to maintain
their response consistency while the 0% became more variable.
These findings (and speculations) fit nicely within a transfer-appropriate
processing framework since learning to become consistent during acquisition
trials was followed by consistent behaviour when encouraged by the nature of the
transfer trials3. A major limitation, however, is that the arguments are based on
an assumption that the absence of KR (when inside the bandwidth) promotes the
learning of consistency by encouraging the repetition - as opposed to the
modification - of an action plan. A more direct approach t o this issue is needed,
an attempt at which is discussed below.
While I have argued that the 10%and 5%p u p s were learning to become consistent, the variable
error WE) means amongst the three groups are not greatly different during acquisition trials (see
Figure 1). Thir ir probably a function of the fact that when KR was given (although on less frequent
occasions than the 0% group), the action plan may have been altered more dramatically, thereby
inflatingVE,forthelO%andS%bandwidthgroups.
Tmnsfer-Appropriate Processing 211
following acquisition was a no-KR transfer test at performing the learned task.
Variable error (VE) was calculated within a trial (i.e., over the 5 timing
estimates) and averaged across blocks of 10 trials for the purpose of analysis. The
results are illustrated in Figure 2.
45
40 -
0
E
v 35 -
b
& 30 -
W
3
-5 25 -
>
20 - 0 KR P?
15
0 Paced &a
' /
*
1 2 3 4 5 6 7 8 9 10 1 2
Blocks of 10 Trials
We anticipated that subjecta in the paced group would learn to become more
consistent than the KR subjects because of the invanant guidance over acquisition
trials. The VE results supported this hypothesis (p c 0.001;see Figure 2).Further,
the no-KR transfer test was also performed better by the paced group than the KR
group (pc.05), suggesting that these subjects continued to remain more
consistent. These data are encouraging fmm the view that a test for consistency of
behaviour was performed better by a group that had been guided by an invariant
training method than a group that been trained to encourage consistent behaviour
(i.e.. by means of the 10% bandwidth manipulation) yet in a manner that was
212 T.D. Lee
An interesting aside is also noted regarding this experiment. The role of guidance
in motor learning has received considerable attention although conclusions about
its effectiveness are complex. For example, Schmidt (1982) summarizes that
guidance may be: 1. more effective early, rather than late in practice, and 2. more
effective for slow, feedback-based movements, rather than rapid movements that
may involve the learning of motor programmes. From the concept of transfer-
appropriateness the major concern is the type of processing that is being affected
by the guidance technique and the compatibility of this type of processing with the
processing that will be required without the guidance available. Theoretically, the
transfer-appropriate processing view predicts that guidance will be effective to the
degree that processing activities remain the same following the removal of the
guidance technique. Von Wright's (1957)anticipatory guidance procedure is an
example that would support this prediction.
presumed laws about the way KR 'worked had been based on acquisition
performance during the period that KR was being manipulated. When
performance was measured later, under no-KR transfer trials, some of these
laws changed quite dramatically (the relative frequency manipulation, discussed
earlier, being an example). Salmoni et a1 argued that the no-KR transfer test was
a good test for learning effects for two reasons: 1. the no-KR transfer test is a
procedure that equates all acquisition conditions by allowing the temporary
performance effects of the independent variable to dissipate, thereby allowing the
more permanent influence to be evidenced, and 2. a number of no-KR trials allows
performance to stabilize at some level, thereby providing for a more powerful
estimate of the relative amount learned.
When learning is viewed from the constraints of transfer-appropriate
processing, a serious flaw is noted in the arguments of Salmoni et a1 (1984).Since
the no-KR transfer trials 'equate' all KR groups on a common transfer test, the
processing activities on these transfer trials are also equated. The result is that
the acquisition-transfer processing relationships are not equated. In fact, a KR
procedure that produces an 'optimal' no-KR transfer performance may produce
much poorer performance under a different set of transfer circumstances. The
Salmoni et a1 review provided an answer to the question: "How do KR
manipulations affect transfer performance when KR is no longer available?". One
question, for example, that remains unanswered is: "How do KR manipulations
affect transfer performance when KR is available?". From a transfer-
appropriateness view one suspects that since KR and no-KR transfer tests
promote quite different types of processing activities, the answer to the latter
question might be quite different from the answer supplied by Salmoni et al.
ACKNOWLEDGEMENTS
REFERENCES
95, 355-386.
Schmidt, R.A. (1975). A schema theory of discrete motor skill learning.
Psychological Review, 82, 225-260.
Schmidt, RA. (1982). Motor Control and Learning: A behavioral emphasis.
Champaign, IL:Human Kinetics.
Schmidt, R.A. (1988). Motor and action perspectives on motor behaviour. This
Volume.
Shea, J.B. & Morgan, R.L. (1979). Contextual interference effects on the
acquisition, retention and transfer of a motor skill. Journal of
Experimentbl Psychology:Human Learning and Memory, 5, 179-187.
Shea, J.B. & Zimny, S.T. (1983). Context effects in memory and learning
movement 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. This Volume.
Sherwood, D.E. (1985).A note on the effect of bandwidth knowledge of results on
movement aonsistency. Unpublished manuscript, University of Colorado.
Thorndike, E.L. (1914).The Psychology of Learning. New York: Teachers College.
Von Wright, J.M. (1957). A note on the role of 'guidance' in learning. British
Journal of Psychology, 48,133-137.
c Timothy D. Lee
School of Physical Education,
McMaster University
Hamilton, Ontario,
Canada L8S 4K1.
Complex Movement Behaviour:The'motor-actioncontroversy,pp. 217-230
O.G. Meijer & K Roth (editors)
0 Elsevier Science Publishers B.V. (North-Holland),1988
Chapter 8
Karl-Heinz Leist
SUMMARY
A theoretical framework for the study of action is introduced, distinguishing the
actor's (sub-)goals, segmentation of the action, and intentions concerning
execution and control. Learning to ride the pedalo is used to investigate the
covarying of these variables, their affective colouring, the influence of changing
the frame of reference on the process of learning, and the existence of transfer
effects. It appears that the variables mentioned play specific roles in learning to
solve this particular motor problem. However, their dependence on frames of
reference, the sudden changes in both the structure of mental representations
and the formation of units of execution, and - last but not least - the problem of
verbal articulation of mental representations. lead to considerable methodological
diffiulties in the study of complex situations such as learning to ride the pedalo.
the same continuum, but in terms of perception and action discontinuous jumps
are involved movement of the self is perceived as loss of balance and evokes
balancing reactions, whereas movement of an object leads, depending on the
frame of reference, to following it or trying to avoid it.
I I
I 'I
I
rnovelnent movement
of singular over the
object entire field
In order to assess the value of the above theoretical framework for the study of
action, various experiments have been carried out on learning to ride the pedalo.
The value of this characterization of the hctional units of representation has been established
by Zimmer (1985)in a series ofexperiments.
220 K H.Leist
The motor problem which has to be solved in riding the pedalo (Figure 2), is new
but close to actual sports practice. It involves moving forwards on the pedalo as
smoothly as possible, i.e., with minimal fluctuations in velocity. This implies
keeping balance; the problem is, therefore, one of coordination of the entire body.
The complexity of the movement, however, remains graspable. An added
advantage of this learning task is that the investigator may register various
physical parameters. Pedal position allows for the computation of velocity; EMG-
registration of individual muscles and muscle groups is feasible without
problems.
The driving force - which must be kept constant - can be produced with vertical
momentum on the boards, whereas horizontal momentum is needed to overcome
the extreme upper and lower positions of t h e boards. Given the number of degrees
of freedom of the body, or the legs, various combinations of forces can have the
same propulsive effect on the pedalo. It must be assumed that friction varies with
velocity and board position.
EMG-analysis (Zschorlich, 1986) shows, for all tests and all persons investigated
(n = 2501, that vertical momentum is developed by repositioning the body's centre
of gravity. Horizontal momentum may either be produced by pulling the lower
board backwards, or by pushing the upper one forwards.
In order to further develop and test the theoretical framework mentioned in the
first Section, it was attempted to answer the following 4 questions in learning to
ride the pedalo:
1. Do sub-goals in unguided learning to ride the pedalo covary with represented
segments of the action, identifiable in the actor's self-perception? Do
correspondingintentions exist concerning the execution and control of the action?
2. Are certain situations coloured affectively, and does this influence the process
of learning, as Heckhausen (1964)has it in his distinguishing 'hope to succeed
from 'fear to fail'?
3.How are perception and representation of execution influenced when specific
frames of reference are forced upon the test persons? Is there conscious
perception of task-specificaspects, relevant to solvingthe motor problem?
4.Can transfer-effects be shown in accordance with Zimmer's (1984)notion of
'schema'?
Learning How to Ride the Pedal0 221
At least 30% deueleration during the first l/6 turn of the pedal &r transition of the lower
extreme.
222 K H. Leist
Whenever the diagram format for questions concerning intentions could not be used, these
questions were posed during verbal interviews.
The term 'unit'was introduced to them by showing some examples ofhopping.
Learning How to Ride the Pedal0 223
1 I I'1try to ...
1. keep my balance 16. pull and push
2. stand upright 17. put pressure on the ball of the foot
3. ridesmoothly 18. balance with my arms
4. stand up very straight 19. push the pedals forward
5. ridequickly 20. keep the pedals in swing
6. incline forwards 21. hold on with my feet
7. keep the thighs tense 22. avoid falling over
8. put equal weight on both pedals 23. project my weight forwards
9. relaxtheknees 24. control the upcoming pedal
10. ridequickly 25. use the entire sole of the foot
11. counterpoise with my arms 26. relax one foot while the other goes down
12. avoid falling over 27. use the knees
13. maintain foot contact with board 28. keep the buttocksdown
14. balance with my feet 29. heave pedals over the balancing position
15. project my knees over my feet 30. keep my arms relaxed
I
I
Phase (Sub)goal I Segmentation Execution Control
riding smoothly
22%
uniform motion
29% avoiding to let
riding quickly the board escape
23%
riding round position of the pedal pushing 11% staying upright
1 5% to the next 89% pressing 10% 2170
226 KH. k i s t
Segmentation Valuation
76% manipulation of the upper pedals:
very insecure 40%
insecure 28%
neutral 10%
f
very insecure 15%
insecure 22%
neutral 13%
-
We asked 36 pupils (ages 11 13), who had learned to ride the pedalo, t o do so on
a descending plane; 34% of them indicated before trying that 'braking at the right
time' would be the goal. Perceived changes in own execution concerned often the
upcoming pedal; the execution was differentiated for both boards significantly
more often than in the 'normal' situation; for example: 'push the lower one down,
control the upcoming one.'
The test persons were then asked to go into a reverse run immediately after a
normal one; a significantly higher percentage than in a group which did not have
the descending task in between, knew how to correctly answer questions
concerning the differential effects of upper and lower pedals.
Out of 46 pupils (ages 11 - 13) the mental representations of 'slow learners' (the
slowest 25%) and 'quick learners' (the fastest 25%) were compared in a reversed
riding task, using feature rating as described at the end of Section 3.2. Significant
differences emerged for several items (see Figure 4).
Figure 4. Signifiiantly different ratings on items 2-5 and 7-10 for quick
learners (dashed line), slow learners (thin solid line) and the whole group
(heavy solid line), during Phase 1 (point A), 2 @), 3 (C),and 4 (D).
228 K.H. Leist
Although the total number of data was low, we used (non-rotated) factor
analysis in order to formulate hypotheses concerning the relations between the
variables of our feature rating. The results (Table 5) suggested non-functional
grouping during Phase 1, whereas the grouping during Phase 3 was already
functional. 'Balancing', for example, was grouped in Phase 1 together with
'avoiding to fall' and 'avoiding to be thrown off; in Phase 3 it was coupled with
'regular pushing forward of pedals', 'keep pedals moving', 'avoid being thrown
off and 'use the whole sole of the foot on the pedal.' In the sense of Zimmer's
schema-theory, thus, Phase 3 marks the establishment of adequate rules to group
sub-unitsfor the solution of the motor problem.
Phase 1 Phase 3
remaining in an upright position (-) remaining in an upright position
1 bending forwards bending forwards (-)
standing upright as much as possible (-) standing upright as much as possible
relax arms (-) avoidiig to fall off
grasp with feet
pedalling gently avoid being thrown off
2 putting one's weight in the knees projecting body weight forwards
lowering buttocks
maintaining balance
balancing with arms riding like on a bicycle
3 leaning back pedalling gently
pedalling up and down exerting pressure with balls of feet
maintain the swing of the pedals
maintaining balance standing on complete sole of foot
4 avoiding to fall off
avoiding to be thrown off bending forwards
balancing with arms
5 pedalling gently projecting knees over the feet
projecting body weight forwards projecting body weight forwards
putting one's weight in the knees
6 regulating balance with feet lowering buttocks
applying pressure with balls of feet
stamping feet
leaning backwards
pushing pedals forward
relaxing knees
ride on tips of the toes
Learning How to Ride the Pedal0 229
Factor analysis data were different for quick and slow learners. Grouping for
quick learners appeared to be comparatively univocal, whereas their categories
are more directly available to the transfer situation 'reverse running'. In order to
test this latter suggestion, we instructed a group of 36 pupils (ages 11-14)to use the
categorization of the quick learners. During transfer they appeared not to differ
from quick learners, while their learning time for normal forwards running had
been significantly lower than in all other groups.
5. Concludingcomments
The starting point for the experiments described, was the question whether our
theoretical fiamework for the study of action could lead to research that would be
relevant to actual practice. Our data reveal that goals, segmentation and
intentions concerning execution and control play specific roles in learning t o solve
a particular motor problem. Their dependence on frames of reference, the sudden
changes in both the structure of mental representations and the formation of
units of execution, and - last but not least - the problem of verbal articulation of
mental representations, lead to considerable methodological difficulties in the
study of complex situations such as learning to ride the pedalo.
The development of an adequate methodology should, therefore, be
acknowledged as a first priority issue. Multidimensional scaling, the use of
correlation matrices, and the availability of graphs for the solution of the motor
problem, could be included in this methodology.
ACKNOWLEDGEMENTS
The present author wants to acknowledge two anonymous referees for their
comments on an earlier version of this Chapter, and the Editors for their
linguistic advise.
REFERENCES
Kail-Heinz Leist
Technische Universitat Munchen,
Lehrstuhl fur S rtpadagogik
tp
Conollystrak 2,
8000 Munchen 40,
,Federal Republic of Germany.
Complex MovementBehaviour:'The'motor-action controversy, pp. 231-246
O.G. Meijer & K Roth (editors)
0 Elsevier Science Publishers B.V. (North-Holland),1988
Chapter 9
Richard A. Magill
SUMMARY
1. Introduction
An important type of information for the learner when practising a motor skill is
information about his or her just completed response that is provided to the
learner by an external source, such as a teacher, coach, or therapist. This
information is commonly referred to as 'knowledge of results' (KR) and is best
thought of as information relating to the learner's actions that produced the
232 R.A. Magill
With regard to the influence of engaging in activity during the post-KR interval,
the common prediction has been that if the subject is involved in attention
demanding activity during this interval, then appropriate information processing
activity cannot occur, which should lead to poorer learning than if no activity
occurred. While the logic driving these predictions about post-KR interval length
and activity appears sufficiently valid, especially given the evidence from short-
term memory research showing similar time and activity effects, the empirical
evidence supportirig predictions has been virtually non-existent (see Salmoni et al,
1984,for a review). A possible reason for this lack of empirical support is that the
predictions themselves may not be in the correct direction. This suggestion is a
notable departure from the reasons typically presented for this lack of empirical
evidence (e.g., Magill, 1975;Salmoni et al, 1984;Schendel & Newell, 1976) which
focus on such things as the task being learned or the nature of the interpolated
activity.
While reasons based on the task or the interpolated activity may be valid, it is
also necessary to consider the predictions made relative to the type of influence
post-KR interval length and/or activity has on motor skill learning. The need to
rethink these predictions is based on research evidence rarely related to questions
concerning the post-KR interval. This research, reported almost exclusively using
verbal tasks, concerns the influence on memory and learning of the spacing
between repetitions of a to-be-learned item (see Hintzman, 1974,and Lee & Magill,
1985, for reviews). The well documented evidence here indicates that increasing
the spacing, i.e., length of time and/or other activity, between repetitions of an
item leads to better rather than poorer retention.
It can be argued that it is appropriate to draw parallels between what has been
called the 'spacing of repetitions effect' and the post-KR interval, especially when
KR is provided after each response repetition. This parallel becomes even more
reasonable when it is considered that post-KR interval experimental manipu-
lations involve altering the character of the spacing between repetitions, or
practice trials, of a to-be-learned task. This alteration is brought about either by
manipulating the length of the interval or the type and/or amount of activity in
which the subject engages. When viewed from this perspective, what becomes
apparent is that the predictions about increasing post-KR interval length or
engaging the subject in attention demanding activity during the post-KR interval
should be in the opposite direction from that which traditional motor learning
theories have proposed. Based on a spacing of repetitions effect viewpoint the
prediction should be that these post-KR manipulations should benefit learning.
However, neither previous post-KR interval research in motor learning, as
described earlier, nor spacing of repetitions effect research in motor short-term
234 R.A. Magill
memory (Hall & Buckolz, 1982; Marshall, Jones & Sheehan, 19771, have provided
this support.
The two experiments described in this Chapter provide evidence that increasing
post-KR interval length and engaging subjects in attention demanding activity
during this interval can benefit learning, especially as determined by
performance on a novel response transfer test. With regard t o the activity effect,
two current views must be considered here as they provide direction for how
different types of post-KR interval activity conditions should influence transfer
performance. First, Schmidt (1975) argued in his schema theory that engaging in
response variations allows the subject to develop a schema that leads to a greater
probability of success with a novel variation of the action than if these variations
had not been experienced during practice. This view suggests that if activity
during the post-KR interval is to benefit transfer performance, then that activity
should be a variation of the action being practised. On the other hand, Jacoby
(1978) has argued that the spacing of repetitions benefit occurs because the activity
engaged in between repetitions causes the subject to 'forget' information related to
the previous attempt. As a result, the subject must engage in more effortful
problem solving activity at the next repetition, or trial, than he or she would
otherwise do. This view argues that any attention demanding activity occurring
during the post-KR interval will benefit learning, regardless if that activity is or is
not a variation of the response being practised.
2. Experiment 1
2.1. Introduction
The first experiment to be described here has been reported previously by Magill
and Lee (1984). Of particular interest is the amount of post-KR activity as well as
the type of activity engaged in by the subject during the post-KR interval.
2.2.Method
arranged in two rows 40 cm apart. The three barriers in each row were 19 cm
from centre to centre. The criterion sequence involved a four-segment movement
in which the subject released a starting switch located between himself and the
first pair of barriers, moved to the right to knock down the first barrier on the
right, then moved to knock down the left barrier of the second pair, moved to
knock down the right barrier of the third pair, and then terminated the movement
at a switch just beyond the last pair of barriers.
The forty-eight subjects were randomly assigned to one of four post-KR interval
activity conditions (~12). The 'no activity' condition involved an empty 5-sec post-
KR interval. The 'one-related task' condition required each subject to practise
moving through the same sequence of movements but in a criterion time of 1050
msec. The 'two-related activities' condition required each subject to practice the
same pattern of movement at the 1050 msec criterion and then at a 900 msec
criterion. The 'non-related activity' condition involved the subject in a reciprocal
tapping task in which the subject was required to tap alternately between two 7 cm
wide targets, 21 cm apart, in cadence with a metronome for 20 sec, the
approximate time it took a 'two-related activities' group subject to complete the
two post-KR interval responses. The metronome was set at one of three settings
(100, 132,152 b.p.min) for each trial with the three settings randomly distributed
among the practice trials. All subjects practised the criterion 1200 msec
movement for 18 trials. Following a 10 min retention interval, during which the
subjects performed a card sorting task, all subjects were given 6 no-KR retention
trials at the 1200 msec movement time goal, using the same movement sequence
as had been practised.
2.3.Results
Performance measures were based on deviations from the 1200 msec criterion
time for each trial. For purposes of this report, only the absolute constant error
(ACE) will be reported as it is representative of the primary results of interest for
this experiment. ACE involves calculating the actual criterion time deviation
value (constant error) for each trial for a subject and then using the mean of the
absolute values of those scores for each block of trials for that subject. Trials were
blocked into sets of three trials each. Statistical analysis of ACE for the practice
phase of this experiment involved a 4 (post-KR activity) x 6 (trial blocks) analysis of
variance (ANOVA). For retention test analysis, a separate 4 (post-KR interval
groups) x 3 (trial blocks) ANOVA was computed. The three trial blocks included
the last practice trial block and the two retention test trial blocks to allow for
testing the acquisition by test performance interaction.
236 R.A. Magill
During the practice trials (trial blocks 1 through 6), all groups showed
significant improvement, as evidenced by the trial blocks main effect, F5,220 = 8.20,
p=.OOOl. There was also a significant groups effect, F3,44 = 5.80, p=.002, and a
significant groups by trial blocks interaction, F15,220 = 1.70, pz.05. These effects
were probably due to all groups performing similarly at the end of practice while
early in practice the 'no activity' group performed with less error than the other
groups.
Concerning the difference between the last practice block and the first retention
trial, inspection of the results presented in Table 1 suggests that while the four
groups were performing similarly on the final practice trial block, the 'no-activity'
and the 'one-related activity' groups showed a deterioration in performance on the
first retention trial block while the 'two-related activities' and the 'non-related
activity' groups showed improvements in performance. However, neither the
groups effect nor the groups by trial blocks interaction was significant.
Table 1. Absolute constant error means (msec) for the four post-KR interval
actiuity conditions (n=12)for the 1200 msecgoal pattern.
POST-KR ACTIVITY
I LAST PRACTICE BLOCK
M I S D
FIRST RETENTION BLOCK
M I SD
No activity 51.3 51.3 77.7 47.1
One-related activity 49.4 45.7 64.3 42.4
Two-related activities 55.9 41.8 47.8 47.2
Non-related activity 55.3 46.5 48.6 33.8
I
2.4. Discussion
While the statistical analysis of the retention results of this experiment did not
significantly demonstrate the predicted post-KR activity benefit, the results are
encouraging in that the post-KR activity did not lead to poorer retention
performance than no activity. As such the results argue against the traditional
view that activity d k n g the post-KR interval should lead t o impaired retention
performance. With respect to the prediction conyrning benefit of activity in the
post-KR interval, it is noteworthy that group performances during retention, and
Post ME Activity 237
Based on the data reported in Table 1, there is an obvious need to reduce the
within-subject variability evidenced in the present experiment. Accordingly, the
second experiment reported in this Chapter incorporated a task and procedures
designed to address this concern. Additionally, since the present experiment
confounded length of post-KR interval and activity effects, the second experiment
included a condition designed to separate out these effects.
3. Experiment 2
3.1. Introduction
3.2.Method
All subjects practised the 300-600 msec criterion responses for 30 trials with KFt
provided after every trial for each segment of the pattern. This practice was
followed by a 24-hr no-KR retention test and a transfer test. The transfer test
required subjects to perform a novel response where both the pattern of movement
and the criterion times for the two segments of the pattern were novel.
Four postrKR interval condition groups were formed (n=14) for the practice
trials, Two no-activity conditions were no-activity for a 5-sec post-KR interval and
for a 20-sec post-KR interval. During the post-KR intervals on each trial, these
groups rested quietly until the beginning of the next trial. No specific instructions
were given to subjects in these groups concerning what they should be thinking
about while resting.
Post KR Activity 239
Two post-KR activity conditions were used and were similar to two of the
conditions used in Experiment 1. The 'two-related activities' group was required
to practise two additional patterns during the post-KR interval. The first
interpolated pattern began at the far right corner target, then to the centre target
in 500 msec, and from the centre target to the near right comer target in 800
msec. The second interpolated pattern involved moving from the far left comer
target to the centre target in 700 msec and from the centre target to the far right
corner target in 1000 msec. The rationale for selecting these interpolated task
characteristics was that they provided variability with regard to the goal task in
both spatial and temporal dimensions. The relationship between the transfer task
and the goal task was that the transfer task involved two novel criterion times to
be produced in a new two-segment pattern.
3.3. Results
performed on the ACE data. First, a 4 (post-KR interval groups) x 6 (trial blocks)
ANOVA was performed to test effects during the practice phase of the
experiment, Secondly, a 4 (post-KR interval groups) x 2 (trial blocks) ANOVA was
performed to test retention test results, with the last practice trial block included
in the analysis. Thirdly, a one-way ANOVA was computed for the transfer test to
compare transfer performance for the four post-KR interval groups and the
transfer test only control group.
Table 2. Total absolute constant error means (mec) for the two segments of
the criterion response for the four post-KR interval activity groups and the
transfer test only group (n=14).
I
No activity 5 sec)
No activity 20 sec)
Two-related activities
Non-related activity
Transfer test only
All subjects improved during the practice phase of this experiment as revealed
by a significant trial blocks effect, F 5 , 2 ~= 25.51, p-.0001. However, there was no
group or group by trial block interaction for the practice phase of the experiment.
Retention test analysis revealed no significant group gr blocks effects. Also, there
was no significant group by block interaction, suggesting that the experimental
groups acted similarly from the last practice trial block to the 24 hr no-KR
retention test.
The transfer test analysis revealed a significant group effect, F4,65 = 5.51,
p.0007.The Neumann-Keuls post hoc analysis of t h i s effect indicated that the two
activity groups and the 20-sec postiKR interval non-activity group performed the
novel pattern better than the 5-sec non-activity group and the transfer test only
group. The two activity groups and 20-sec non-activity group did not differ from
each other nor did the 5-see non-activitygroup and the transfer only group.
Post KR Activity 241
3.4. Discussion
TOfurther examine this view that forgetting leads to a greater need for action plan
construction or re-construction on the following trial, which in turn leads t o better
learning, it will be instructive to look more closely at the transfer response
characteristics of the different post-KR conditions. According t o the forgetting
hypothesis, it would be expected that activity during the post-KR interval should
lead to better transfer than an empty 20-sec interval, and that the empty 20-sec
interval should lead to better transfer performance than the empty 5-sec interval.
While there was not a statistical difference between the two activity groups and
the 20-sec non-activity group, there is a noticeable ACE gap between these groups
in exactly the manner predicted by the forgetting hypothesis. These results point
to the need for further research to develop procedures that will allow a statistical
separation of these effects.
242 R.A. Magill
A result from this experiment that did not support the predicted post-KR
conditions benefit for learning was the similarity of performance by all groups on
the no-KR retention test. However, it is important to note here that this result is
counter to the traditional view that these conditions should negatively influence
learning. Clearly, the extended post-KR interval length with no activity and the
activity during this interval did not impede learning of the goal task.
4. General discussion
Investigations of motor learning must address the question of how people solve
the problems involved in learning a new motor skill. Within this question is the
issue of what role is played by various forms of information available to the
learner. In this Chapter, the focus has been on information that is available to the
learner only from an external source and not directly available. This information,
labelled 'knowledge of results' (KR), must be assumed to be important for aiding
learning since manipulations of practice conditions, relative to the presentation
andor use of KR, influence learning a skill.
The manipulation of concern in this Chapter has been the interval of time
following the receipt of XR,known as the post-KR interval. If subjects use this
externally presented information to aid them in error correction and in planning
a more appropriate action strategy for the next trial, then manipulations of the
post-KR interval should influence these processing activities. The unanswered
question, however, has been what type of influence post-KR interval
manipulations will establish. It has been commonly hypothesized that interfering
with KR processing would lead to poorer learning than when this processing had
not been influenced. In this Chapter, evidence has been provided arguing against
this traditional view. For no-KR retention performance, post-KR interval activity
and 20-sec of no activity during practice with KR, did not lead to retention
performance deficits. On the other hand, post-KR activity and 20-sec of no activity
actually led to beneficial transfer to a novel variation of the practised response. In
the case of the novel response transfer, then, it can be argued that interfering
with KR processing can actually benefit motor skill learning if learning is viewed
as enabling the individual to successfully produce previously unpractised
variations of a practised action.
What emerges from such results are two issues that require further exploration.
First, why did this beneficial effect occur? Second, what are the implications of the
results of these experiments for the prediction by schema theory that success on
Post KR Activity 243
post-KR interval. Has KR information been forgotten? Has sensory feedback been
forgotten? Have certain features of the synthesis of these sources of information
been forgotten? A resolve to this problem is critical to establishing the forgetting
hypothesis as the basis for explaining results of the experiments described in this
Chapter (see discussions by Lee & Magill, 1983b & 1985,for further development of
this point.)
In terms of the implications of the results of these experiments for schema theory,
it is obvious that they are especially damaging to the hypothesis that novel
response success is dependent on the amount and variety of practice of response
variations. Variable practice cannot be considered as the only means of
accomplishing this transfer success. The alternative would appear to be, as
Bernstein (1967)suggested, that practice requiring to solve the movement problem
again and again leads to the best motor learning. Taken together, this view
suggests that creating interference in the post-KR interval establishes the
requirement for this type of problem solving activity to repeatedly occur.
ACKNOWLEDGEMENTS
Thanks must be given to Greg Myers, Harry Meeuwsen, Tim Lee, and Bob
Mathews for their assistance in various ways in carrying out the experiments
reported in this Chapter and in preparing this paper.
REFERENCES
Adams, J.A. (1971).A closed loop theory of motor learning. Journal of Motor
Behavior, 3, 111-149.
Bernstein, N. (1967).The Co-ordination and Regulation of Movements. Oxford:
Pergamon.
Bilodeau, E.A. & Bilodeau, I.M. (1958).Variations of temporal intervals among
critical events in five studies of knowledge of results. Journal of
Experimental Psychology, 55, 603-612.
Bilodeau, I.M. (1969). Information feedback. In E.A. Bilodeau & I.M. Bilodeau
(Eds.), Principles of Skill Acquisition (pp. 225-285).New York Academic
Press.
Hall, C.R. & Buckolz, E. (1982). Repetition and lag effects in movement
recognition. Journal of Motor Behavior, 14, 91-94.
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.
Post KR Activity 245
RichardA. Magill
Depts. of Physical Educationand
Louisiana tate University
psycho~o!y~
Baton Rouge,
Louisiana 9893,USA.
Complex Movement Behaviour:The' motor-actioncontroversy, pp. 247-259
O.G. Meijer 8c K Roth (editors)
0Elsevier Science Publishers B.V. (North-Holland),1988
Chapter 10
SUMMARY
There has been a growing interest in the use of artificial sensory feedback
procedures, eg., EMG feedback, for the treatment of patients with motor
disorders. I n spite of two decades of applied research, however, its results are still
disputed, especially regarding the generalization of therapy effects to functional
skills. In the present text the relevance of EMG feedback for rehabilitation
medicine will be discussed. The emphasis will be on the problem of limited
transfer of learned motor control to the performance of functional activities. It is
argued that the efficacy of the treatment procedure could be enhanced if
therapeutical attention was directed at the (re-)acquisition of goal-directed actions
instead of at the learning of isolated movements.
1. Introduction
information, auditory and visual cues etc.) and extrinsic feedback, which refers to
information provided by an. external source such as the therapist or trainer.
Extrinsic feedback, (which is, in fact augmented feedback) plays an important
role during the first stages of the therapy or training process (Holding, 1965) and
in situations where intrinsic information is absent or insufficient (Keele, 19811.
During the first stages of the therapy process a global idea about the correct
movement has to be developed, a cognitive representation or, according to Adams
(1971),a 'perceptual trace' that can function as an internal model or 'reference'
for action. The feedback arising from the action is compared to this internal
model. If such feedback information is not available or is insufficient, as is often
the case in patients suffering from neurological (motor) disorders, the
development of such a model is almost impossible. It is this kind of situation in
which a (physical) therapist could profitably choose to use sources of artificial
sensory feedback.
The term 'artificial sensory' feedback refers to a procedure whereby the patient
is provided with auditory or visual information about a physiological process
normally beyond his or her awareness. For example, the electrical activity of a
muscle or muscle group can be displayed visually on a TV monitor or aurally via
a loudspeaker (EMG feedback), the feedback reflecting the actual muscular events
occurring during the movement. The patient is instructed to use this EMG
information to regain voluntary control over a muscle or muscle group under
training. The rationale for such feedback manipulation is that an increase in the
amount of available information will lead to an increase in learning and
performance.
Since the early 1970's there has been, especially in the American rehabilitation
literature, a growing interest in the use of artificial sensory feedback procedures
(see for a review Mulder, 1985). In numerous articles successful applications of
this technique have been reported. However, in spite of frequent applications and
enthousiastic reports, the question as to the efficacy of artificial sensory feedback
(e.g., EMG feedback) is still difficult to answer. Although several controlled group
studies have been published during the last decade (Basmajian et al, 1975;
Burnside, Tobias & Bursill, 1982; Den Brinker, 1984; Mroczek, Halpern &
McHugh, 1978) the results remain ambiguous. One of the main problems is that,
often, little insight is given in the transfer of therapy effects to more functional
activities. Authors discussing this important topic do not present uniform results.
For example, Wolf, Baker and Kelly (1980) and Wolf (1983) reported results
showing a long-term efficacy of EMG feedback. On the other hand, Seeger and
Caudrey (1983)argue that this long-term effect could not be acquired without
Fmm Movement to Action 249
booster sessions. While Brudny and his co-workers (1976) describe the successful
employment of artificial sensory feedback procedures in the training of upper-
limb control in hemiparetic subjects, the results of Prevo (1979)are in contrast. He
reports no significant differences in the functional use of the arm between
subjects trained by means of EMG feedback and subjects receiving conventional
physical therapy. Hence, even after twenty years of applied research conflicting
results are still being presented, especially with respect to the generalization of
therapy effecta to functional skills.
In the present paper the relevance of artificial sensory feedback for the treatment
o f patients with motor disorders will be discussed. The main focus will be on the
problem of limited transfer of learned motor control to the performance of
functional activities. The discussion will be based on some recent EMG feedback
studies, performed in Nijmegen. These studies will not be presented in detail
since they have been published elsewhere (Mulder, 1985;Mulder, Hulstijn & Van
der Meer, 1986). The studies were carried out with normal subjects as well as
w i t h brain damaged (hemiplegic) patients. With normal adults who had to learn
a novel movement (viz. the abduction of the big toe without moving the other toes
or the total foot), it was indicated that artificial sensory feedback procedures (EMG
feedback and force feedback) lead to better learning than the normal sources of
feedback (proprioceptive, visual) (Mulder & Hulstijn, 1985).However, this finding
could not be replicated in the clinic (Mulder, 1985). In the following Sections this
failure to replicate will be discussed.
In the first clinical study, hemiplegic patients were trained to improve the
dorsiflexion capacity of the foot. The main purpose of the study was to compare the
effects of EMG feedback to the effects of a conventional physical therapy procedure
(NDT).
Twelve hemiparetic subjects (eight male, four female) without cognitive defects
o r aphasia (ranging in age from 34-68 years) volunteered to take part in the study.
All had previously completed other rehabilitation programmes without
reacquiring the dorsiflexion function. They were trained three times a week for 5
weeks.
T h g subjects were randomly assigned to one of the two treatment conditions, i.e.,
Neurodevelopmental Treatment (NDT) and EMG feedback. The subjects in the
NDT condition were trained according to the principles devised by Bobath (1978).
250 Th. Mulder & W.Hulstdn
Prior to the first therapy session each patient's gait was analysed, using
videorecordings (to determine the velocity) and surface electromyographic
registrations of relevant muscle activity during ambulation. The EMG
measurements were taken to determine the 'target muscles' for therapy. The
rationale for this was as follows: dorsiflexion of the foot is mainly the result of the
synergistic activity of the Peroneus longus / brevis and the Tibialis anterior, with
the Gastrocnemius acting as an antagonist.
Hemiparetic subjects very often show an asymmetry between the activities of the
two agonistic muscles (Peroneus longus / brevis and Tibialis anterior), resulting
in inverted or everted foot lifting. The muscle selected for EMG feedback training
was the one which showed a marked electromyographic underactivity during
walking. The training was directed at the restoration of the symmetrical muscle
activity during walking. For each subject, under both conditions, each therapy
session (which lasted about 40 minutes) started with the registration of the
maximum EMG activity of the Peroneus and Tibialis anterior.
This activity was measured during two 5 sec test trials. The changes in the
maximal EMG output across the therapy sessions were considered as one set of
parameters of therapy effect. Range of Motion (ROM), gait velocity and the
phasing defining the EMG signal during walking were used as the other relevant
parameters.
The results indicated that the difference in maximum EMG output across the 15
sessions between the EMG feedback condition and the NDT condition was only
marginally significant (F1,lo = 3.96,p=0.07). The increase in EMG activity across
the 15 therapy sessions was not significant. Also, no significant differences were
obtained for the ROM data, the gait-EMG data and the gait velocity
measurements. Hence, in this study, we were not able to replicate the positive
findings of other authors, although we followed their procedures very closely (for
a review, see Inglis, Campbell & Donald, 1976;Keefe & Surwitt, 1978).
From Movement to Action 251
Each therapy session started, as was the case in the first study, with the
measurement of the maximum EMG output generated during the performance of
a test manoeuvre. The test manoeuvres were patient specific, that is, they differed
across the 16 subjects. For some of the patients the test manoeuvre consisted of
stepping forward and backward, whereas other patients had to perform a
balancing task, or a task directed at the maintenance of hip or knee stability.
The results indicated a significant difference between the pre- and post-test scores
on the h c t i o n a l checklist (F29,290 = 5.949, p<O.OOl). However, the differences
between the two groups were not significant.
Th. Mulder & W.Hulstijn
The ratings of the videorecordings showed an increase in gait quality from pre-
to post-measurement. Again, the differences between the two treatment
conditions were not significant. Also, the EMG values showed no significant
differencesbetween the EMG feedback condition and the NDT condition.
Hence, we have to conclude that EMG feedback, as employed in these two
studies, did not lead to significantly better results than conventional physical
therapy.
3. Discussion
One has to remember that the logic behind the use of EMG feedback is that in
cases where normal sources of information (proprioceptive, visual and auditory
cues etc.) are absent or inadequate, the function of these sources can be replaced
or supplemented by an external feedback loop. In the reported two studies
attention was directed a t a comparison of the effects of EMG feedback as against
conventional physical therapy procedures. The data revealed two interesting
aspects. In the first place, no significant differences were found between the two
treatment conditions (only the difference in EMG output between the two
conditions was marginally significant in the first study), and second, under both
treatment procedures the functional improvements were not impressive. How can
we explain these results?
We have to direct our attention to some crucial aspects of the EMG feedback
treatment as employed in these studies. The following remarks, however, are also
relevant for many physical therapy interventions. EMG feedback is largely
characterized by a strong focus on (isolated) muscle activity. Subjects are
instructed to manipulate (increase or decrease) the displayed EMG signal of one
muscle or muscle group. This EMG signal functions as an important guidance
source containing very specific information concerning the (electrical) activity of
that muscle group.
One could argue that the nature of this feedback forces the subjects into a
'fixation' on the EMG activity instead of on the movement itself. The exercise
effort will, as a consequence, also be focused on the manipulation of the EMG
signal, and more than once this is probably all they learn. Remember that in the
first study the only difference between the groups was found at the EMG level.
Many patients do not use the information to shape their strategy for reacquiring
functional movement control, but the information serves as a goal in itself. They
work hard to manipulate the trace, often not in the least concerned about the
From Movement to Action 253
movement. Indeed, increasing or decreasing the trace represents the final goal of
exercise.
Gonella, Kalish and Hale (1977) criticized such a treatment method, which in
their eyes meant an (implicit) return to classical rehabilitation procedures used
before 1940. Indeed, these procedures were characterized by a strong fixation on
one muscle or muscle group, which was trained in isolation.
If this argument holds, does it mean that the application of artificial sensory
feedback procedures is useless and has to be discarded in favour of conventional
physical therapy? Probably not, but the results do indicate that the EMG feedback
in the present studies has been used in a rather obsolete fashion.
This is a criticism which is valid for all EMG feedback applications where the
attention is directed at a single muscle or muscle group (see also Mulder &
Hulstijn, 1984). It is, of course, not a new critique. Mroczek et a1 (1978)argued
that:
... motor control and skill is not gained simply by contracting or relaxing a
single muecle to full capacity but by a process of acquiring control of
activating and inhibitory mechanisms coordinating the simultaneous
interplay of many muscles (267).
primitive stimuli. Perhaps this mismatch alone accounts for the lack of
relevance (540).
The use of artificial sensory feedback oRen illustrates this lack of relevance; the
treatment is directed at isolated movements, trained in a meaningless context,
instead of the training of goal directed actions, directed to solving a problem
naturally arising in the environment. For example, extension and flexion of the
arm constitutes a 'movement', but reaching for a cup and bringing it to the mouth
(which also involves extension and flexion), is an example of an 'action'. Too often
movements are unduly trained and studied independent of the actions of which
they are elements.
... conviction that the whole is always the s u m of its parts and not more
than this ... and that sufficient knowledge of single bricks would be enough
to comprehend the building constructed of them.
Although everyone agrees that functional control and skills have developed as a
result of a continuous interaction between the 'brain' and the 'environment', or,
in other words, as a result of a multitude of goal-directed actions, it is, in fact,
amazing to observe that following damage to the nervous system, re-development
often takes place in a rigid, laboratory-like environment. The therapeutical
attention focuses on the distorted element instead of on the distorted action. Often
the therapist is concerned to decompose human motor behaviour into segments
which can be manipulated.
However, the restoration of a (motor) skill will be very difficult by concentrating
upon segments to be performed in isolation. At present, the evidence for a close
relationship between recovery of fiulction and environmental factors is
accumulating. The relevant studies (using animal subjects) indicate that optimal
recovery is obtained by providing complex stimulation and exercises in a
meaningful environment. While EMG feedback focusing on single muscles or
isolated movements can result in an improvement, this improvement will
generally be muscle or movement specific with little transfer to the level of actions
performed in daily life.
Fmm Movement to Action 255
With some caution, one could suggest that the effectivity of physical therapy and
EMG feedback could be improved if patients were trained in situations directed at
the acquisition of goal-directed actions instead of the control of single movements.
Of course, this does not mean that all interventions in physical therapy should be
directed at the learning of actions. Indeed, there are situations where an
intervention at the level of isolated muscles seems totally justified. Take, for
256 Th. Mulder & W.Hulstiin
It could be argued that this is all not really new or original since one can observe
an increasing number of attempts to develop a multidisciplinary 24-hours
treatment directed at the re-development of functional behaviour. Furthermore,
one could argue that occupational therapists have always emphasized the
acquisition of goal-directed actions. This cannot be denied but, nevertheless, it can
be observed that still many therapists (implicitly) accept the idea that motor
learning largely consists of linking elementary units (movements) to form larger
patterns (actions). They accept the idea that motor control develops as a result of
repeating correct responses over and over again.
The ideas as described in this Chapter are also not theoretically new. In fact,
they are to a large extent borrowed from European psychology in the first three
decades of the present century. Von Uexkiill, Von Weizsticker (see Frbse & Sabini,
1985) and also the Dutch physiologist Buytendijk (19481, all stressed the concept of
'action'. 'Action' was defined as the performance of goal-directed behaviour in a
direct relationship with the environment. These authors suggested the crucial
link between behaviour and environment, a standpoint which can be found also in
the writings of representatives of the soviet culture-historical school (e.g., Luria,
Leontjew, and Vygotsky - see McLeish, 1975).
Notwithstanding the fact that nothing in this Chapter is provocatively new, it is
still worthwhile to repeat these old arguments in the context of rehabilitation
medicine.
Against this background of action learning, Hubner, Wehner and Stadler (1982)
employed an interesting EMG feedback procedure, which was aimed a t learning
the action of walking instead of a - single dorsiflexion movement. In their
procedure, which they termed 'auto-hetem feedback, the electromyographic
activity of a muscle group (in this case the dorsiflexor group) was fed back in
simplified form to one channel of a headphone worn by the patient during
training. Simultaneously, the electmmyographic activity of the 'same' muscle
group of the walking therapist was fed back to the other channel of the
headphone. The patient had to match these signals during walking. By using this
From Movement to Action 257
procedure it was possible to emphasize the timing and phasing aspects of the task
during the performance of the entire act rather than merely to increase isolated
EMG output sitting in front of a TV monitor.
A similar procedure is now being used in our laboratory. The subject wears a
headphone connected to a portable miniature EMG amplifier, which delivers
auditory information to the subjects about the electromyographic activity of a
muscle (group). The therapist wears a second headphone, also connected to the
patients' EMG amplifier. This provides himher with the possibility t o intervene
during the performance. The device has been used in the training of equilibrium
control, in the training of agonist-antagonist coordination, and in suppressing
spastic antagonistic activity of the Gastrocnemius during walking.
EMG feedback, however, is characterized by the fact that it delivers to the subject
information about a process which forms the condition of a movement but EMG
feedback is not really movement related by itself, that is, it gives no information
concerning (the quality 00 a movement or action. One could ask, therefore,
whether EMG feedback is not too specific to be useful for learning, since too much
precision could be harmful in that it could cause the subject to focus on levels of
errors that are beyond control, ignoring more global aspects of the task. Hence,
beside EMG feedback, force feedback and/or position feedback could be used.
Especially in later stages of the therapy these forms of feedback may be useful.
More research concerning performance-relatedfeedback is necessary.
ACMOWLEDGEMENTS
REFERENCES
Adams, J.A. (1971). A closed loop theory of motor learning. Journal of Motor
Behavior, 3, 111-150.
Basmajian, J.V., Kukulka, C.G., Narayan, M.G. & Tabeke, K. (1975).Biofeedback
treatment of foot-drop after stroke compared with standard rehabilitation
technique: Effects on voluntary control and strength. Archives of Physical
258 Th. Mulder & W. Hulstwn
Mulder, Th. & Hulstijn, W. (1985). Sensory feedback and the learning of a novel
motor task. Journal of Motor Behavior, 17, 110-128.
Mulder, Th., Hulstijn, W. & Van der Meer, J. (1986). EMG feedback and the
restoration of motor control: A controlled group study with 12 hemiparetic
patients. American Journal of Physical Medicine, 65,173-188.
Newell, K.M. (1981). Skill Learning. In D. Holding (Ed.), Human Skills (pp.257-
269). London: Wiley.
Prevo, A.J.H. (1979). Over spastkche parese en revalidatie. Thesis, Free
University, Amsterdam.
Reed, E.S. (1982). An outline of a theory of action systems. Journal of Motor
Behavior, 14, 98-134.
Schmidt, R.A. (1982). Motor Control and Learning: A behavioral emphasis.
Champaign, Ill.: Human Kinetics.
Seeger, B.R. & Caudry, D.J. (1983). Biofeedback therapy to achieve symmetrical
gait in children with hemiplegic cerebral palsy: Long-term efficacy.
Archives of Physical Medicine and Rehabilitation, 64, 160-162.
Stadler, M., Schwab, P.B. & Wehner, Th. (1978). Regulation senso-motorischer
Lernprozesse durch Biosignale. Zeitschrifi fiir Psychologie, 186, 341-381.
Turvey, M.T.(1977). Preliminaries to a theory of action with reference to vision. In
R. Shaw & J. Bransford (Eds.), Perceiving, Acting, and Knowing: Toward
an ecologicalpsychology (pp. 211-265).Hillsdale: Erlbaum.
Whiting, H.T.A. (1980). Dimensions of control in motor learning. In G.E.
Stelmach & J. Requin (Eds.), Tutorials in Motor Behavior (pp.537-551).
Amsterdam: North Holland.
Wolf, S.L. (1983). Electromyographic biofeedback applications t o stroke patients: A
critical review. Physical Therapy, 63, 1448-1455.
Wolf, S.L., Baker, M.P. & Kelley, J.L. (1980). EMG biofeedback in stroke: A 1 year
follow-up on the effect of patient characteristics. Archives of Physical
Medicine and Rehabilitation, 61, 351-355.
Complex Movement Behriviour: "he' motor-action controversy, pp. 261-288
O.G. Meijer & K Roth (editors)
Q Elsevier Science Publishers B.V. (North-Holland), 1988
Chapter 11
Klaus Roth
SUMMARY
1. Introduction
It can not be too strongly emphasized that the theory of movement in sports is far
from fulfilling the expectations from within and outside the discipline just
outlined. At present, research relevant to both theory and practice can only be
done if theoretical assumptions, or model predictions, are tested in series of
investigations, with the emphasis on internal and external validity being
systematically varied.
2. Theoreticalbackground
On the other hand, a certain variability within the movement classes does exist.
The GMP hypothesis implies that programmes can be executed with different
response specifications (parameters). The most important of these mutable or
variable features are duration, speed, absolute force, and movement size.
3.Empirical evaluations
The experiments of group 1 within the research project 'Tactics in Sports Games'
are akin to the investigations of Armstrong (1970),Summers (1977),and Shapiro
et al (1981)with respect to the aims and execution of the tests. The main interest
was directed towards checking timing invariance in the repeated performance of
complex sports techniques. Relative timing was defined as a set of ratios obtained
by dividing the duration of the relevant temporal events in a movement by
movement time (MT).
266 K. Roth
Figure 1 shows the design for the evaluation of the over-arm throw (handball).
The total movement time as well as the following five movement segment times
were measured, using four light barriers:
*T1: lif€ing arm;
*T2: moving arm backwards;
*T3: inverse movement;
*T4:forward swing (part 1);
*T5:forward swing (part 2).
Measurements were made a t the subject's throwing hand. The length of the
trajectory between two light barriers was 0.40 m. A total of 11 persons took part in
these experiments. Each subject completed 10 'normal' throws, 1 0 throws with
increased movement speed, and 10 throws with diminished movement speed.
Results
Regarding the phasing patterns, the most important result of the light barrier
experiment was that there was a difference between two groups of subjects. In
the first group (subjects 1 , 2, 4 & 5 ) a timing invariance for the movement
segments 52 through S5 (the segments of which the times were T2 through T5)
could be detected, albeit not for the S1 phase (see Figure 2). This finding can be
explained by the fact that subjects started the movements in S1 - independent of
MT - always with approximately the same speed, with the consequence that the
Generalized Motor Pmgramme Theory 267
ratio TlMT showed statistically higher values for rapid movements (0.693) than
for normal (0.579)or slow movements (0.513)- F2,117 = 6.84;p c 0.01.
The ensuing differences for the T n T through TS/MT ratios (across different
MTs) disappear ifTi is related to (MT - T1) rather than MT:
52: F2,117 = 1.11 - n.s.;
53:F2,117 = 2.07 - n.8.;
54: F2,117 = 0.24 - n.s.;
55:F2,117 = 2.42 - n.s.
I T i IMT
0.70 1
I----------- 7
0.50 ;o Speeded Up ;
! x Normal !
0.40
Y\ i A Decelerated i
0.30
0.20
0.10
1 2 3 4 5
Segment Number
Figure 2. The quotients TiIMT for the fivemovement segments (group 1).
In the second group of subjects (3, 7, 8, 9 & 10) the phasing remained relatively
constant across all movement segments, i.e., from S1 through 55 (see Figure 3);
nonetheless, the differences between the ratios Tl/MT which were found to be
significant in group 1, could also be observed in group 2, although less
pronounced
S1: F.2147 = 2.91 - n.s.;
S 2 F2,147 = 0.16 - n.8.;
S3: F2,147 = 2.72 - n.s.;
54:F2,~47= 0.32 - n.s.;
S5:F2,147 = 0.14 - n.8.
268 K.Roth
T i IMT
0.70 1
~~
1 2 3 4 5
Segment Number
It should be mentioned that two of the eleven subjects (6 & 11) could not be
classified within either group 1 nor 2. Their throwing patterns showed a
significant timing variability with respect to all movement segments.
Ti TI T2 T3 T4 T5
0 0 1 1
TI
(.134) (.I47 (.261) (.253)
3 3 3
T2
(.476) (501) (.470)
4 3
T3
(.763) (.790)
T4 4
(.941)
T5 I
Table lb. Correlation matrix of the durations T1 through T5 (subjects 3, 7, 8,
9 & 10).
T5 I
270 K. Roth
T2 3 3 3
(.814)
T3 4 3 3
(.901)
T4 4 3 2
(.Q47)
Table 2b. Comparison of Ti values with MT-Ti) (subjects 3,7,8,9 & 10).
TI 5 3 2
(.644)
T2 5 4 3
(.M3)
T3 4 3 2
(.609)
T4 4 3 2
(.869)
T5 5 3 3
(.845)
Generalized Motor Pmgmmme Theory 271
The results of correlation and regression analyses combined show that the
timing data of T2 through T5 (group 1) and T1 through T5 (group 2) accord well
with the assumptions of the GMP theory. Partial divergences from the model can
be seen only in the last columns of Tables 2a and 2b. The intercepts a of the
regression equations deviate significantly from 0 in 8.3% of the cases in group 1,
and in 25% of the cases in group 2 (positive intercepts). Little is known of potential
causes and interpretations of this result, which has been found as well in most of
the studies quoted at the outset of the Section (see Schmidt, 1984).
Figure 5. The response alternative: Wrist pass / with minor force f to the
left.
In the second experiment (table tennis) the programme dimension was vaned
between a slice and a topspin, and the parameter dimension 'direction' between
right and left. The slice to the right served as a standard movement (Figure 6).
Twenty-ninesports students acted as subjects.
274 K Roth
Results
Table 3 and Figure 7 show the results of the handball experiment. In the last
column of Table 3 are the average CRTs for the standard movements in the
respective block of trials.
The main effect across all precue conditions is statistically significant (F6,174 =
19.46; p c ,011.The differences between the 'programme cued and 'programme
not cued situations are particularly strikingiF1,zs = 56.37;p c ,001).AU conditions
where the programme to be performed was not known (P-conditions) show higher
CRT-values than the situations where the programme could be specified in
advance (prediction 1).
Hypothesis 2.1 and 2.2 were tested, using analysis of variance, as shown in
Figure 7. Checking for interactions (F2.58 = 3.42;p c -05) and simple main effects
Generalized Motor Pmgmmme Theory 275
CRT [msec]
PF
400
PDF
PD
380
360 L
2a22
340
-5
t
a
1 300
320
1 I I
Cued NotCued
Programme
confirm the fixed-order assumption that parameters can not be set before
choosing the movement programme. The advance information on D and F have
effects in the 'programme cued conditions (F2,58 = 6.42; p < .Ol), but do not lead to
any significant changes in the CRT-values (F2,58 = 2.36; not significant) in
'programme not cued situations.
The results of the second experiment (table tennis) are shown in Figure 8.
Statistical comparison of D and P, as well as of D and PD, supports prediction 1,
while contrasting P with PD again shows that in 'programme not cued situations
subjects can not make use of parameter information (Tuckey's HSD procedure:
KD.06; 4.87 = 12.22; CRT (D) < CRT (PD), CRT (PI).
Taken together, the pattern of results obtained from the experiments in handball
and table tennis confirm the hypothesis that the invariants of motor programmes
must be specified before the parameters. The handball experiment, moreover,
shows that with respect to parameters a fixed-order hypothesis might apply in
'programme cued situations (Tuckey's HSD procedure: KD.05; 8,203 = 21.57; CRT
(DF) - CRT (F) = not significant), which assumes that force must be specified
before direction. This latter finding is consistent with the result of a study
conducted by Zelaznik (1981). His investigation concerned the specification
276 K.Roth
of force and direction in isometric contractions of the forearm. He also found that
when force was uncertain, advance infarmation about direction was useless.
CRT [msec]
280 I
260
240
220
200
L
Cued NotCued
Programme
The assumption is that the processing time requirements of the two types of
modifications - one requiring a change in the programme, the other requiring a
change in a parameter value - will not be the same in both cases. Given the
hierarchically structured programming process (see experiments of group 21, the
lower level parameter specifications might be supposed to require shorter
correction times than the higher level programme decisions. This hypothesis has
been confirmed in the majority of the movement correction time experiments thus
far conducted Wince & Welford, 1967; Megaw, 1974; Gielen, Van den Heuvel &
Van der Gon, 1984; Quinn & Sherwood, 1983). These studies investigated
elementary movements, thereby enabling the modification latencies (RT) to be
measured directly and compared with one another.
In the case of complex sports techniques, however, subjects may execute the
programme or parameter corrections in so many different ways that the moment
when the movement correction is initiated is difficult to measure. Therefore,
rather than modification RTs the experiments of group 3 registered
1. the interval from the illumination of the correction signal until the
completion of the movement (SE),and
2. the realization (1) / non realization (0)of the programme or parameter
modification required.
Movement correction data for all subjects were combined and the probability of
(successful) corrections was examined as a function of SE. The expectation was
that S-shaped curves would be observed, similar to that found with the method of
constant stimuli for determining thresholds in perception (Murch & Woodworth,
1978).This would imply that with low SE values the (realization) probability would
remain low but with increasing SE duration it would increase at first slightly,
within the middle range strongly, and finally plateau. Analogous to the
perception experiments, one might define the correction time limit as the SE
value at which the movement corrections were realized in 50%of all cases.
The fit of the collected (observed) data to the theoretically assumed ogive-shaped
curve was statistically tested via z-transformations of the probabilities and a
subsequent test for linearity (Gescheider, 1976).
The investigations of group 3 were conducted using movements and situations
from handball, table tennis, tennis, and volleyball. Each of the experiments
employed a standard movement which had to be executed in at least 50% of the
trials without any modification. Subjects should modify a programme or
parameter only in the case of a signal for movement correction being given. The
illumination of the signal lamp, or the presentation of slides were randomly
selected and the intervals at which these signals occurred were systematically
278 K.Roth
Figure 9 shows the design of the handball experiment employing slides. The
subject jumps off a mat providing the initial stimulus; the behaviour of the goal-
keeper changes (with a timed temporal delay) from the standard (slide-projector
1) to the correction picture shown (slide-projector 2). Before the experiment
started, the players were instructed to respond to this change of picture by
changing the throwing direction from 'bottom right' (standard movement) to 'top
left'. In addition, movement correction slides for parameter modifications t o
'bottom centre', 'bottom left' and 'top right' were presented.
The hrther experiments employing table tennis, tennis and volleyball
contained, aside from direction parameter modification, other parameter
modifications as well as changes in the generalized motor programme. In these
investigations, the signals for movement corrections were presented via
illumination of lamps. In order to determine SE duration, all trials were filmed.
The moment when the ball left the player's hand (handball and volleyball) or the
racket (table tennis and tennis) was defined as the point of movement completion.
Generalized Motor Pmgmmme Theory 279
-Programme: -
backhand lamp backhand slice 556.8
slice to topspin
straight fore- lamp forehand stroke 609.8
hand stroke to lob
smash lamp smash to lob 744.5
-Parameters: -
jump throw lamp throwing 393.9
direction from
right to left
jump throw lamp throwing 382.3
direction from
low to high
jump throw lamp throwing 402.4
direction from
right to left and
from low to high
backhand lamp stroke direction 399.2
slice from right to
left
straight fore- lamp stroke direction 403.5
hand stroke from right to
left
straight fore- lamp length of stroke 41 3.4
hand stroke from long to
short
smash lamp stroke direction 491.4
from right to
left
jump throw slide throwing 327.0
direction from
right to left
jump throw slide throwing 300.6
direction from
low to high
280 K.Roth
Results
The correction time limits (SE values for the standardized realization probabilities
of 50%) for changes in programme and for parameter modifications are
summarized in Table 4. These limits were determined only for those tests in
which values measured exhibited a sufficient fit t o the expected ogive-shaped
curve. In total, data on 3,301 throws I strokes executed by 99 subjects were
collected and analysed. Statistical comparison of the results on changes in
programme and parameters yielded a confirmation of the initial hypothesis. The
programme corrections required significantly higher times in the table tennis
tests (F1.14 = 38.35;p c .Ol), tennis tests (F1,ll = 17.65;p c .Ol),and volleyball tests
(F1.7= 36.35; p < .01)than the considered parameter changes. Comparing the
found correction time limits to the average M!"s of the hitting phases of the
standard movements (table tennis 194 msec; tennis 213 msec; volleyball 296 msec)
shows that the necessity for changing the programme must be recognized at least
two reaction times before the start of the forward swing. Also, parameter changes
are obviously not possible anymore during the hit phases. Here, however, it is
sufficient if the correction signal is presented roughly one reaction time before the
start of the forward swing.
Additional tests indicated that there was no systematic relationship between
correction time limits and sex, level of performance, and other characteristics
(simple reaction time, speed of action, accuracy of movement) of subjects.
Results
The results of the experiment are presented in Table 5. Scree-test (distribution of
Eigenvalues) was used to determine the number of factors which explain 67.1
percent of the total variance. The varimax rotated matrix shows that the three
factors - direction left / right (Factor I), force (Factor 111, and direction high / low
(Factor IV) - can clearly be identified for the movements performed on the tennis
and handball tasks, which are similar in structure. Factor I11 represents the
soccer force parameter tests; the soccer direction tests yield only small factor
loadings and communalities.
The factor pattern found contradicts the hypotheses 4a, 4b and 4d formulated at
the beginning of Section 3. Parameter specification abilities neither appear to be
independent of each other, nor are they correlated in a programme specific way
(across all parameters), or across all parameters and programmes. Opposed to
that, prediction 4c has partially been confirmed. The factors I, 11, and IV indicate
that the abilities for an exact determination of parameters might extend across
different programmes, and thus show a certain degree of generality. Prerequisite
for parameter transfer to appear is obviously a certain structural similarity of the
respective generalized motor programmes (handball - tennis).
Another experiment - not reported here - yielded comparable results.
Researching three different throwing techniques in handball showed that
parameter determination abilities correlate parameter-specifically. Such findings
would especially support those practice oriented approaches that have been tested
284 K.Roth
and reported only in a few domains of sports activities, and are centrally based on
the assumption of parameter transfer occurring across different motor
programmes. Salient examples are the 'inner training' of Gallway (1974), and
Gallway and Kriegel (1977), or the training method for beginners in swimming
devised by Wilke and Madsen (1983).
k TEST
FACTOR
I II 111 IV
4. Concluding remarks
The theory of centrally stored motor programmes can be considered typical for
motor approaches to human movement behaviour. While action theory states
quite explicitly that there is no representation of movement, constructs like
(generalized) programmes, schemes and plans are a main element of motor
theories.
Within the motor perspective the GMP hypothesis is one of the most recent and
most frequently discussed concepts. It is based on the so-called impulse-timing
model and suggests that phasing and relative forces are integral parts of the
movement programmes. Other assumptions about the content and structures of
motor programmes, e.g., of Polit and Bizzi (1978 & 19791, have remained less
influential. These authors have provided evidence leading to the idea that
movement endpoints are stored in the programmes.
The main goal of the experiments described in this Chapter was directed towards
examining specific statements and predictions of GMP theory in new classes of
tasks. Taken together, the findings indicate that the attempt to apply the central
hypothesis of the model to complex movements in sports, and to test them under
sports conditions, appears to be quite fruitful. From the perspective of theory
building mainly the results about the time structure of programming and
reprogramming processes, shown in 3.2 and 3.3, are interesting. They hint at a
hierarchy between programme elements (phasing, relative forces) and
parameters (absolute force, direction), and thereby confirm one of the essential
assumptions of GMP theory.
ACKNOWLEDGEMENTS
REFERENCES
Klaus Roth
University of Bielefeld
De artment of Sports Science,
48fO Bielefeld 1
Federal Republic of Germany.
ComplexMovement Behaviour:'The'motor-actioncontroversy,pp. 289-314
O.G. Meijer & K Roth (editors)
0 Elsevier Science Publishers B.V.Worth-Holland),1988
Chapter 12
SUMMARY
L Introduction
The goal of our current research has been to develop a framework capable of
addressing the foregoing points, and of integrating diverse and often
contradicting findings in motor learning. We have not concerned ourselves with
description at the movement level and our findings are not inconsistent with
either more traditional (e.g., Schmidt, 1975), or action (e.g., Saltzman & Kelso,
1983; Warren, 1987; Reed, 1988) accounts with their emphasis on ecological
validity (Bruce, 1985; Neisser, 1985). We have no problem with the notion that
certain movement taxonomies, hand feeding for example (see Reed, 19881, are
ecologically determined. However, our concern for the learning of the o h n
culturally determined and specific way these skills are accomplished, is treated
in only a tangential way by these theorists.
Relevant findings from three experiments concerning verbal labelling effects
(Shea, Hunt & Zimny, 1985), and one experiment concerning contextual
interference effects on motor retention (Zimny, 1985), will be presented and
' While we feel that the more automatic and lower level components of movement are equally
important to these cognitive mechanisms, and that research illuminating the interdependence of
these control levels should be pursued, explanation at this level has not been our focus.
292 J.B.Shea & S.T.Zinny
Inasmuch as the cognitive model incorporated only one form of knowledge, the
peculiar characteristics of that knowledge form should have been reflected in the
movement. While the task used in these experiments was relatively simple, the
cognitive model that controlled the movement was not.
Table 1 shows the mean accuracy measures for each labelling group for
Experiment 1. These response accuracy findings show that the representational
structure, or cognitive model, that guided performance, was systematically linked
to the information available, and that this information need not be from the motor
domain. These findings conform to what has been found for a representational
structure that is spatial in nature (Levine, Jankovic & Palij, 1982). Such a spatial
representation is thought to maintain configural relations but not necessarily
objective distance (Anderson, 1983). This description uniquely suits the
performance of the LO group, the responses of which were based on the spatial
knowledge provided by the labels, and whose model was presumably highly
configural in nature. The responses of this group were characterized by a large
negative bias but small variability.
294 J.B. Shea & S.T. Zimny
The analysis of verbal reports revealed the use of go", and to a lesser extent 45". as
reference or anchor points in the production of criterion positions. Inspection of
the response curves for trials on which the use of anchors was reported, allowed
LABELLING
GROUP
Clock Degrees Other
VRLM
Group M 20.00 44.16 3.33 26.67
OIo Subjects 50.b0 100.00 50.00 100.00
I
I
VRLO
Group M 16.00 51-34 19.33 63.33
% Subjects 83.33 100.00 50.00 100.00
VRMO
Group M 4.67 30.67 12.67 21.33
% Subjects 50.00 50.00 50.00 83.33
296 J.B. Shea & S. T. Zimny
8
v)
z
2
v)
2.5
2.0 0'"
...4
.........''... :,:.$,.
I
n
-0-
-a-
Label only
Movement only
-m- Label and movement
a: 1.5
W
8
a
w
:
m 1.0
5
z
3
I
0.5
criterion position collapsed over labelling groups. It can be seen that the position
most often used as an anchor in the production of a criterion position was the
position immediately adjacent to and short of it. In fact a general characterization
that can be made is that subjects rarely moved to an anchor point beyond the
criterion position being produced.
Table 3. Mean number of responses at each anchor point for each criterion
position computed over Labelling Groups.
ANCHOR POSITIONS
CRlTERION
POSITION 45" 60" 75" 90" 105" 120" M
-
60 " -20 -00 .05 .05 -00 .OO .21
75 " .62 1.27 .oo -09 .oo .oo .33
90 " .16 .48 .oo .oo .oo .oo -27
105 " .16 .35 .81 2.00 .OO -00 .55
120 " .ll .18 .59 1.14 1.16 .OO .53
-
M .45 .45 .49 .65 .23 .OO
The use of multiple reference systems by all groups and the use of anchor points
in the production of criterion positions documents the constructive nature of
motor retention (Shea, 1977). Given the apparent imprecision and transient
nature of kinesthetic information, the use of well-learned reference systems
would seem necessary for accurate performance on the present task. The finding
that 45" was used as an anchor point was especially interesting since it was one of
the five criterion positions. The use of 45", as well as the increased usage of 90" as
an anchor point by the LO condition underscores the importance of already
existent well-learned knowledge for the storage and retrieval of new information
(Kolodner, 1984; Lawler, 1981).
298 J.B. Shea & S.T. Zimny
The existence of 75" as a highly used anchor point in our experiment can be
accounted for by the fact that some subjects, as evidenced through verbal report
data, in the LO condition thought that 75" was 90".This suggests that they, in fact,
based their movement on a distorted representation of the clock face. This notion
is supported by the large negative bias for the criterion positions in the LO
condition. For example, a subject in the VRLO group (P.L.) said: "Let me try my
reference, go", about 4 inches to the left, 11:30, there." In fact, P.L. stopped his
response at 75" instead of 90°, and his final estimation of 105' was 90".
Interestingly, P.L. consistently reflected this 15" response bias across trials not
only for 90°, but for all other criterion positions. Our view is that if a subject has a
'skewed representation of their most highly used anchor point, then subsequent
production of all criterion positions will reflect the configural distortion of this
representation.
The assumption of many motor skill researchers has been that performance
(output) mirrors input and/or mental representation, but certain findings from
Experiment 3 showed that this is not necessarily the case.
A more complete perspective of the representational structure as well as its
functional significance was obtained by conducting stepwise multiple regressions
on CE scores. These analyses allowed a determination of the amount of variance
in the production of each criterion position accounted for by other criterion
positions. Five separate stepwise multiple regressions were conducted for each
labelling group. Subjects' CE scores for each criterion position were entered as
either a dependent variable (one time) or an independent variable (four times).
Additional information about organizational processes was obtained from the free-
recall test that was given at the end of an experimental session.
The findings of the stepwise multiple regression analyses are summarized in
Table 4. The amount of variance accounted for in the production of criterion
positions by other positions was the greatest for the LO group (M = .78). The MO
group (M = ,691 had an intermediate amount of variance accounted for, and the
LM group (M = .36) had the least amount of variance accounted for in the
production of criterion positions by other positions. Generally, the position
accounting for the greatest amount of variance in the production of criterion
position was the position immediately adjacent to it (either short of it, or beyond
it). Them findings provide evidence that subjects used the configural relation-
ship among positions for the production of individual criterion positions.
Knowledge Incorporation 299
The finding that the variance accounted for was highest for the LO group would
be expected, since this group could only base their movement production on the
configural knowledge present in their representational structure which was
provided by the labels. The finding that the variance accounted for was least for
the LM group was surprising, however. The combination of the well-learned
knowledge structure represented by the labels, and the kinesthetic information
provided by moving to the criterion positions, evidently resulted in a
representation of the individual positions strong enough that their production
could be accomplished with less dependence on other positions. The finding that
the 90" criterion position had the least variance accounted for by other positions is
evidence that highly-learned positions can be produced independently of other
positions.
It is of particular interest that the anchor point responses reported earlier did not
exactly mirror the multiple regression findings. While anchor point responses
were made most often a t the position spatially preceding the criterion position for
a particular trial, they were rarely made at positions beyond the criterion position.
The multiple regression analyses showed, however, that spatially adjacent
300 J.B. Shea & S.T. Zimny
positions either short or beyond the criterion positions were used in response
production. This finding has two implications concerning the use of
representational structures for response production. The first is that a subject can
make a response in reference to other positions within the structure, without
actually having to move to these positions, and the second is that the anchor point
responses have more to do with individual response pmduction per se than with
the exact details of the representation.
In order to investigate the existence of primary recall orders, the differing
orders used by subjects for the free-recall trials were tabulated for each group.
This information, expressed as the percent of total free-recall trials for each
group, is presented in Table 6. The recall orders are categorized by the criterion
position first produced on a free-recall trial. The S-L (short-long) and L-S (long
short) orders are represented as sub-categories of the orders in which the 60" and
120" was produced first, and 24% of the f r e e - r e d trials an S-L order was used
across groups. The 60" and 75" positions were recalled as a unit on 82%, and the
105" and 120° positions were recalled as a unit on 90 of the free-recall trials on
which 90" was recalled first across groups. The ordering of positions within these
units were varied, however. The order in which these units were recalled
following the production of 90" was evenly distributed. On 42% of the trials the 60"-
75" unit was produced first, and on 39% of the trials the 105°-1200 unit was
produced first across groups.
Table 5, Recall orakrs used by each Labelling Group for free-recall test
trials expressed as a percentage of total trials. The order of recall is
designated by the positions recalled first. 'S-L' denotes recall order of short
(609 to long (1209 positions; 'L-S' of long (120" to short (609 positions; these
are sub-categories OL respectively, 60' and 120; their percentages having
been included in these groups. The total number of free-recall trials for
each group was 64.
MO 43 38 41 25 13 1 0
LO 22 13 51 38 27 0 0
LM 27 22 40 20 24 3 1
M 30 24 43 21 21 1 0
Knowledge Incorporation 301
Research concerned with the role of organization in motor retention has in large
part been preoccupied with the order in which movements are presented and later
produced (Diewert & Stelmach, 1978; Stelmach 8z Szendrovitz, 1981).These studies
have shown that order information is important for acquisition, as well as
retention. Thus, the implicit assumption has been that order is coded in the
representational structure for movements (Diewert & Stelmach, 1978). The
present findings demonstrated that subjects based their performance on a spatial
representation and that the nature of this representation facilitates flexibility in
the order of response production.
In studies demonstrating the primacy of order for response production, this
information was emphasized through the provision of substantial practice with
Knowledge of Results and the presentation and recall of all positions in a serial
order. In the present experiments the importance of order information was de-
emphasized in that Knowledge of Results was not given, and performance
entailed the singular presentation and recall of positions.
The important point to be made from this is that depending on the nature of the
task, order information may or may not be coded dong with spatial information
in the representational structure for movement (Anderson, 1983). If order is not
critical to task comprehension, as it is in, for example, language production, then
order of output might be vikwed more profitably as a production strategy and not
necessarily representative of the nature of the encoding.
The present experiment investigated the combined assertions of Shea and Zimny
(1983) that the concurrent presence of tasks in working memory facilitates
multiple processing, and Ericsson and Simon's (1984) model of verbal protocol
analysis that contends that the contents of working memory are available for
report if elicited properly. While earlier work by Zimny (1981) used minimal verbal
reports, a more extensive analysis of verbal reports taken over the entire course of
processing was performed in the present experiment. The tasks used in this
experiment required a number of disks to be tapped in a specified sequence as
quickly as possible in response to an appropriate coloured stimulus light. An
auditory signal (a 'beep') was emitted by the apparatus each time a disk was
tapped. Diagrams depicting the order in which disks should be tapped were
present during acquisition trials. Each diagram was colour coded so that the
subjects would know which stimulus light it was to be paired with. Subjects in
high (random) and low (blocked) contextual interference conditions practised
either five high- or five low-similar tasks (Figure 2).
Knowledge Incorporation 303
5
0 0
0 0 0 0 d o 0 0
Yellow Blue Green Red White
HIGH SIMILARITY ACQUISITION TASKS
'
i
0 . 0 0
were encouraged to 'think aloud at any other time. Verbal reports were recorded
and transcribed.
The dependent measures of total time for the acquisition trials and percent error
for the retention tests were analysed. Briefly, the usual contextual interference
effects of poorer acquisition performance for the random than the blocked
condition was found, while the error data revealed significantly fewer free-recall
errors for the random than for the blocked condition for retention. These findings
held for both the first and second phases of practice.
Rather than provide detailed analyses of the data at this point, we would like to
demonstrate how the subject's verbal reports mapped onto the total time
(acquisition) and error (retention) findings, providing a rich source of information
as to subject's processing activities.
Analysis of total time measures revealed that the random high-similarity group
performed significantly faster on the 'white' task (see Figure 2), while no
difference among tasks was found for the blocked high-similarity group, The
analysis of verbal report data made the source of this facilitation in the
performance of the white task obvious. Sixty percent of the subjects in the random
high-similarity group identified the white task as the letter ' Z and expressed ease
a t recalling and performing this task:
[Subject 1:l That is ah ... Z pattern. I just punched it out as soon as I see
that colour.
.
[Subject 9:l I think this .. that one is the easiest one - it's just like a Z.
3.0 0
fn
h
2.5
s2
- 2.0 Blocked (low contextual interference)
1.5 0 Random (high contextual interference)
11 12
lo TRIALS
Figure 3. Mean total time measures computed over tasks for the Blocked
and Random conditions on the last trial (trial 10) during the first phase of
practice, and on the first (trial 11) and second (trial 12) trials for the second
phase of practice.
[Subject 3:l Well, I just saw the green and it wasn't what I thought it was
on the sheet you just gave me. I moved the whole pattern up about one dot.
Starts out on the second row and ends in the last row. It's the first and the
third.
[Subject 5A:I Um, I'm trying to just incorporate these in my mind. I don't
remember that green one from the last time.
Other direct support for Shea and Zimny's (1983) characterization of the types of
processing, responsible for contextual interference effects, was provided by
categorizing verbal reports according to Shea and Zimny's suggestions. Reports
306 J.B. Shea & S.T. Zimny
Table 6. Percentage of the total number of reports for Blocked and Random
Conditions, accounted for by each statement category.
The finding that subjects in the random condition gave approximately twice as
many intra-item, inter-item, and extra-experimental reports, as subjects in the
blocked condition provided support for Shea and Zimny's suggestions as to
processing differences. In addition, the specific nature of inter-item statements is
exemplified by a statement by a subject in the random high-similarity group:
[Subject l:] Let's see, red, that's ... reds, ah, reds similar to yellow.
Instead of going crisscross, you go straight across from the first move.
The verbal report data also pointed to a differential use of strabgies used in the
high and low task similarity conditions. Subjects in the high similarity condition
made numerous comments about the tasks themselves and comparisons across
tasks. They made extra-experimental comments only concerning the white tasks
Knowledge Incorporation 307
[Subject l A ] Yes, I'm ah, developing a rhythm in working this out. Well, I
have a lot of music background so I think I've been doing stuff like, ta, ta,
tee, tee, ta with the sounds of the dots. Like the yellow one I just did would
be ta,ta,tee, tee, ta,ta, with the finish. That helps me remember them.
3.2. An interpretation
If one combines these findings with the results of the recognition test, a more
thorough interpretation is possible. The choice of items for the recognition test
was principled in that one item represented the correct pattern, one item
represented the correct configural representation of the correct pattern but was
displaced as to orientation on the apparatus, and the third item was highly
dissimilar to the correct pattern. At first glance, the recognition data seemed to
parallel the recall findings with the random condition performing better than the
blocked condition. However, this difference was all but eliminated if the foils that
maintained correct configural information were not considered errors. Subjects
apparently have very good recognition memory for the general shape or the
configural information in patterned movement. This finding agrees with the
earlier work ofMorgan (1981).
Where subject's performance breaks down is in trying precisely to define the
task within the specific constraints of the experimental context. Subjects in the
blocked condition had an accurate general concept of the task but could not
accurately locate it on the apparatus. This notion is in agreement with the total
time performance of subjects in the blocked condition. Here subjects showed a
markedly increased time to accurately specify the task after a time away from it
(see especially Trial 11,following the filled retention and test interval).
308 J.B. Shea & S.T.Zimny
4. Conclusions
A substantial amount of data from experiments dealing with two only remotely
related motor skill topics has been presented. The first set of three experiments
concerned the effects of verbal labels on the short-term retention of self-paced,
lever-positioning responses by blindfolded subjects, and the fourth experiment
concerned contextual interference effects on acquisition of more complex tasks (at
least in comparison to lever-positioning responses). Not only did the tasks used in
these experiments differ, but the experimental procedures, as well as the
dependent variables, differed. We think that findings across these experimental
contexts are worth noting, in that they provide insight into common types of
processing performed during motor skill acquisition, as well as illuminate
divergent types of knowledge incorporated in motor tasks representations.
Findings can be summarized by the following seven points.
01. Subjects easily use a wide range of existent well-known information from a
variety of sources to aid in retention and performance. This was most obvious in
the use of multiple reference systems, as well as the high incidence of using 45"
and 90" as anchor points for the production of criterion positions by subjects in the
verbal labelling experiments. Additional evidence for this use of well-known
knowledge comes from the coding of the 'white' task as the letter 'Z' by the
Knowledge Incorporation 309
among the tasks. Random performance necessitated more detailed and accessible
knowledge of the tasks. The random condition facilitated more active, controlled,
processing, and this processing resulted in a memorial representation adequate
for retrieval.
*6. Response production is not always reflective of the memorial representation
for a task, Because the representation for a task is so closely associated with the
processes used either to operate internally on it or for output purposes, any clear
separation of these is difficult. However, the anchor point multiple-regression
incongruence found in the labelling experiments is good evidence that subjects
have a good deal more knowledge available for response production than that
reflected in the course of the response itself. The flexibility in the order subjects
recalled positions during the free-recall test of the third labelling experiment
provided additional evidence for this view. The spatial nature of the
representation for this task allowed individual positions to be flexibly organized
for production purposes. This processing-representation perspective provides a
framework from which to understand the flexibility typically found in motor
skills. Inflexibility in motor skills can often be attributed to constraints placed on
performance either through instructions or the structure of the task itself (Pew,
1984).
07. The experiments presented have demonstrated the value of verbal reports for
elucidating the underlying processes in motor skill acquisition. This was
especially the case in the reported experiments because care was taken to verify
verbal reports with both no-report control conditions and the behavioural
responses for trials on which reports were given. Complete congruence across
these provided evidence that the verbal reports accurately reflected processing
activities performed during learning. As future research addresses more general
and relevant issues, new flexible paradigms will have to be developed. Verbal
protocol analysis can fruitllly be incorporated into these approaches.
REFERENCES
Adams, J.A. (1983).On the integration of the verbal and motor domains. In R.A.
Magill (Ed.), Memory and Control of Action (pp. 3-15).Amsterdam: North-
Holland.
Allard, F. & Burnett, N. (1985).Skill in sport. Canadian Journal of Psychdogy, 39,
294-312.
Anderson, J.R. (1983). The Architecture of Cognition. Cambridge, MA: Harvard
University Press.
Knowledge Zncorpomtion 311
Zimny, S.T. (1981). The use of verbal protocol analysis to investigate the basis of
contextual interference effects in the learning of a motor task. Unpublished
Master'sThesis. Boulder, CO: University of Colorado.
Zimny, S.T. (1986). Contextual interference and task similarity effects as
determinants of processingactivities. Unpublishedraw data.
Chapter 13
Stephan Swinnen
SUMMARY
In comparison to the post-KR interval, the KR delay interval has been assigned a
minor role in skill learning. It is suggested here that the KR delay is the interval
of response evaluation or error estimation and that its role may have been
underestimated in the past. Several kinds of preliminary evidence point to this
conclusion. First, when interrogated about their activities during the KR delay,
the majority of subjects report that they evaluate their response. They do not
appear to passively wait for the KR to assess the correctness of their responses.
Second, eliminating the KR delay by instantaneous provision of KR is detrimental
to learninglretention. It is hypothesized that this may be due to the blocking of
spontaneous performance evaluation activities. The viewpoint that KR should be
presented as soon as possible is, therefore, debatable. Third, there is limited
evidence that requesting learners to estimate errors is beneficial for
learning Iretention. More research concerning each of these topics will be
necessary before more definite statements can be made.
1.Error-detectioncapabilities
gradually represent the feedback qualities of the correct response. As such, the
perceptual trace becomes a reference of correctness t o which the response-
produced feedback can be related in order to detect errors and eventually correct
them. The postulating of a reference mechanism is at the heart of the 'closed-loop
principle' and it constitutes the basis for a n error-detection mechanism (Adams,
1971 & 1976).
In his schema theory of discrete motor skill learning, Schmidt (1975)has further
elaborated on error-detection mechanisms. When a performer makes a response,
four sources of information are stored: (1)the initial conditions; (2)the response
specifications; (3)the response outcome; and (4) the sensory consequences, Two
schemata are central to the theory. On the one hand, there is a schema for
response production (recall schema). It represents the abstraction of information
concerning the initial conditions, the past response outcomes, and the response
specifications. It enables the performer to choose the required parameter
specifications for the movement.
On the other hand, the recognition schema represents the relation between the
initial conditions and past response outcomes as well as the sensory
consequences. It enables the generation of the expected sensory consequences of
the movement to be produced: expected exteroceptive and proprioceptive feedback.
The expected sensory consequences are compared to the actual feedback produced
by the response.
A third schema, the 'error labelling schema', comes into play here. This
schema has only received minor attention in the research literature. A mismatch
between actual and expected sensory consequences may lead to the detection of a n
error in performance. This raw sensory signal can be labelled and converted into
a reportable form (Newell & Boucher, 1974;Schmidt, 1975).
In a more general vein, Annett (1983)has stated that the action and language
systems are closely linked. There is a bridge between both systems across which
two-way traflic i s possible. "This two-way traffic makes it possible to turn words
into actions, that is to follow instructions and t o describe actions in words"
(Annett, 1983, p. 226). The perspectiye developed in this paper is mainly an
application of the latter direction of traflic albeit in a more limited sense.
According to Schmidt (1975,p. 2391,the labelled error plays a n important role in
learning:
The labeled error signal is termed subjective reinforcement and can serve
as a substitute for knowledge of results (KR), providing outcome infor-
mation that can update the recall schema.
Response Evaluation and Learning 317
This reasoning would imply that the course and magnitude of the within-subject
correlation values through practice would be fundamentally different in fast as
opposed to slow movements. Indeed, in the Schmidt and White (1972) study, a
ballistic movement was used, and the correlation values between the objective and
estimated error scores increased through practice, reaching .90 at the end of
practice. But, such high correlations were not identified in another study in
which subjects learned a slow positioning movement (Schmidt & Russell,
reported in Schmidt, 1975). The within-subject correlations were rather low (.20)
throughout acquisition. This is quite understandable in view of Schmidt's
contention that subjects use error information to correct their errors while
performing slow movements. Subjects move to a position that is judged as being
correct and there is no meaningful error information after the response has been
made.
The afore-mentioned studies appear to suggest that learning a skill does not only
result in better motor performance. Skill acquisition may also be accompanied by
the development of a capability to detect errors. This capability is directly
dependent on the strength of the perceptual trace or recognition schema as well
as on the availability of sensory information. The error-detection mechanism may
prove to be important in skill learning because it may help the learner to maintain
and even improve performance in the absence of KR. The question, then, arises if
techniques can be employed in learning settings to improve this mechanism more
directly.
At least two arguments may preclude the conclusion that error estimation led to
higher learning. First, it is not clear whether estimation facilitated learning or
whether repeating digits hampered learning. Hogan and Yanowitz (1978)asked
subjects of the baseline group to repeat nonsense digits in order to rule out that
just doing something, rather than estimating errors in performance, would
account for the findings. This manipulation may have one disadvantage. Studies
that investigated the effects of activities interpolated in the KR delay interval on
criterion task performance, revealed that some of these activities caused negative
interference as determined by performance on a no-KR test (Boulter, 1964;
Martenidc, 1986; Shea & Upton, 1976). But, these interpolated activities were
generally more demanding than repeating digits (as used by Hogan & Yanowitz)
in that controlled lever movements (Boulter), a to-be-remembered movement
(Shea & Upton), or a to-be-learned movement (Marteniuk), had to be performed.
As a result, the argument that repeating digita hampered performance m a y not
count that strongly although no studies were found in which the interpolations
were continued during a no-KR phase.
320 S. Swinnen
-L
2
&
c
c
c
m
ul
c
0
0
04 I
0 4 8 12 16 20 24
ACQUISITION (blocks of 5 trials)
Figure 1. Mean absolute constant errvr (ICEI) over 5 trial blocks during
acquisition.
J.
0 1 : : : : : : : : I
0 1 2 3 4 5 6 1 2 3 4 5 6
immediate delayed
TRANSFER
Figure 2. Mean absolute constant error (ICE I) over 5 trial blocks in the
immediate and delayed no-KR tmnsfer tests.
in the delayed transfer trials than those who did not (or better: who were not
requested to do so). No significant effect for variable error (VE) was identified
during delayed transfer. The effect of estimating during transfer on transfer
performance was not significant. In sum, estimating errors during acquisition
only affected response bias in delayed transfer.
e
Forc d estimation could in some way have aided in preventing retention losses.
There were some differences between this study and the one by Hogan and
Yanowitz (1978). The findings with respect to the acquisition phase were similar.
However, the immediate transfer data did not support Hogan and Yanowitz's
findings. While the estimators in the latter experiment performed with less
absolute error (AE)than the subjects repeating nonsense letters, no differences in
ICE1 (and in AE3 between estimators and no-estimators were identified in the
Shapiro et a1 (1984) study. The reasons for this discrepancy may be manifold as
the task, the required movement times, the operationalization of the no-KR phase,
etc., were all Werent.
While conducting the error estimation experiments, it was observed that subjects
in the no-estimation conditions displayed signs of performance evaluation
activities during the KR delay. Although not requested to do so, some of them
reported spontaneously that their movements were too fast or too slow before KR
provision. This mainly happened in the initial stage of acquisition. Although not
studied in any systematic way, this observation drew our attention because it left
undecided if a strong contrast was achieved between estimators and no-
estimators in the Shapiro et a1 (1984) experiment. Rather, it appeared that a group
of estimators was compared to another group of subjects who were not explicitly
requested to estimate but who could have done it anyway, albeit in a less detailed
and less systematic way. In the next study, an attempt was made to prevent these
spontaneous activities in order to get a cleaner no-estimation baseline condition
(Swinnen, Schmidt & Shapim, 1986a).
In this study, the timing task of Shapiro et al (1984) was used again. Subjects
(N=78) were instructed to perform the movement in the target time of 1 sec. They
were randomly assigned to three conditions: (1) subjects of the control group
(n=26) disposed of a free 8 sec KR delay interval, (2) subjects of the estimation
group (n=26) were requested to estimate their movement time8 during the 8 sec
KR delay, (3) subjects of the immediate KR group (n=26) had a 0 sec KR delay.
Response Evaluation and Learning 323
Figure 3. Mean absolute constant error ( ICE I) over 5-tricrl blocks during
acquisition
324 S.Swinnen
of the immediate KR condition had a post-KR interval of almost 13 sec while those
of the other two conditions had a post-KR interval of 5 sec.
The acquisition data were analysed using ICE1 and VE as the dependent
variables. The data were collapsed into 5-trial blocks. A 3 x 18 (Groups x Trial
Blocks) ANOVA with repeated measures on the last factor was conducted. No
significant groups effects were identified for I CE I and VE (see Figure 3).
The immediate and delayed transfer data were analysed separately by means of
a 3 x 6 (Groups x Trial Blocks) ANOVA with repeated measures on the last factor.
The analysis of ICE1 and VE did not reveal a significant groups effect with
respect to immediate transfer performance (shown on the left side of Figure 4).
All groups performed with higher error relative to the end of acquisition. A
performance deterioration was also evident from immediate to delayed transfer.
Delayed transfer showed a performance divergence of the groups (shown on the
right of Figure 4). A significant main effect for groups was identified for I CE I.
Subjects who received KR immediately performed with the highest error. Subjects
who estimated, performed with the lowest error. The control group was located in
between the former two groups. Only the difference between the immediate KR
and the estimation group was significant. The immediate KR group showed an
mure 4. Mean absolute constant ermr ( ICEI) over 5-trial blocks in the
immediate and delayed mKR transfer tests.
Response Evaluation and Learning 325
acquisition of sport skills remains unanswered so far. The next two experiments
represent an attempt t o determine the effects of error estimation in gymnastic
skill learning.
Subjects
Subjects (N=38) were 18-year-old students of physical education. They were not
familiar with the skill.
Task
The 'front mill thigh circle' is characterized by a full rotation around the
horizontal bar with one leg lifted over the bar and the other one extended
backwards. The bar is held at the thigh level of the backwards extended leg. From
a high front support position, the subject dives forwards with great impetus. This
is achieved by lifting the body mass in a forwards direction, far from the bar. After
a full turn, the performer tries to get on top of the bar to resume a spread support
position. This is accomplished by a downwards action of the front leg.
Procedure
Subjects individually learned the skill, performing 42 trials over three acquisition
phases, one phase per day (for three consecutive days). One group ('global', n=15)
received only global visual information of the movement, i.e., some live
demonstrations of a model at regular intervals. In addition to this global
information, the second group ('analytic', n=13) received detailed visual and
verbal information about the movement and its constituent parts. In both
conditions, subjects were requested to verbally describe and report movement
errors they thought they had made after each trial. The concern was raised that
328 S. Swinnen
the use of this verbal reporting technique on a trial to trial basis might affect the
acquisition of the gymnastic skill, warranting the addition of a third group. The
latter group received the same information as the global condition but they were
not requested t o report errors ('no-report', n=lO). A comparison of the global and
no-report groups enabled us to investigate the effects of the verbal reporting
technique on gymnastic performance. The visual and verbal information was
provided a t regular intervals: four times in the first phase of acquisition and three
times in the second and third phase. This helped subjects t o get an idea of the
movement 'as it should be performed'. No knowledge of performance was
provided at any time.
The operationalization of the global and analytic instructional conditions was
inspired by the Russian psychologist Gal'perin (in Van Parreren & Carpay, 1980)
who refers in this respect to the 'orientation basis'. The 'orientation basis' is
defined as the whole of the action conditions of which the learner disposes. It
helps to orient the learner toward the action to be performed. The incomplete
orientation basis (= global condition) is characterized by the provision of global
information (a demonstration), leaving the organization and structuring of the
task up to the subject. On the other hand, the complete orientation basis (=
analytic condition) provides more analytic and detailed verbal information besides
global information (Van Parreren & Carpay, 1980).
Before starting practice, initial skill level was determined in the pretest, i.e.,
after seeing one demonstration, subjects attempted to perform the skill.
Movements were recorded on video and analysed later with a protocol constructed
for that purpose.
Scoring
All movements were videotaped. An observation instrument containing 19
different items was constructed on the basis of movement analysis, literature
survey (Blantz & Dickhut, 1959; Geinler et al, 1975; Knirsch, 1976; Van Assche,
1971), observation of slow motion pictures, and directions of specialists in
gymnastics. Items referred t o positions of body parts in relation to each other or to
the environment, actions, and timing or rhythmical aspects of actions.
Movements were analysed and scored per item by observation of slow motion.
Checking to what degree the requirements for items were met, subjects received a
score on each item of the movement. All item scores were added up to arrive at
the overall score on each trial (from 0 to 26). A specialist in gymnastics observed
and scored all the movements by means of the observation protocol. Preceding this
analysis, reliability studies on the instrument were conducted. After a period of
training and independent of each other, two observers (one of whom did the final
Response Evaluation and Learning 329
14
13
8 12
5
g 11
5
a 10
W
.-
c
0
$ 9 0
c3
E 8
0
1 2 3
Pretest Acquisition
Two hypothetical conclusions were drawn from these findings. First, they
provided some support for our concern that the verbal reporting technique, as
operationalized by us, affected gymnastic performance. Stated more carefully, the
request t o report errors after each trial appeared to directly or indirectly enhance
performance. As such, the use of this research technique modified the
phenomenon under investigation, namely skill acquisition. Although not all
verbal reporting techniques might be contaminated with these (undesirable?)
effects, some'authors have warned against the use of this technique in studying
cognitive activities (Cavanaugh & Perlmutter, 1982; Nisbett & Decamp Wilson,
1977).
Second, the study revealed that requesting subjects to report errors was
beneficial to gymnastic performance. The question was raised to what extent
error estimation could be applicable to real-life situations. An additional
advantage of this technique for the instructor would be to learn about the
performers' perception of their errors and their degree of correctness. A big
obstacle in skill acquisition is that learners often fail to make accurate
judgements of performance errors.
The role of error estimation in learning the neck-shoulder spring vault was
investigated using a transfer designl. A group of subjects that was requested to
estimate, was compared to a control group in which the subjects had a free
intertrial interval.
1 It may be argued that the desirability of a transfer design for detecting learning effects be
appraised in relation to the presence during acquisition of those manipulations (like KR) that
induce temporary performance effects. In this respect, the transfer design may be considered more
important in the timing experiments discussed earlier than in the gymnastic studies. This
contention is based upon the idea that the provision of visual and verbal information (at regular
intervals) in the latter experiments does not induce the temporary performance effects that
characterize the acquisition of timing movements (in which KFt is presented on each trial) or, at
least, does so tn a smaller extent.
Response Evaluation and Learning 331
Subjects
Subjects (N=46) were 18-year-old students of physical education. Only those not
familiar with the skill were selected.
Task
Subjects individually learned a neck-shoulder spring vault. The movement
consisted of five phases: the squat position, tumble-in phase, pike phase,
suspension phase, and the landing phase. The subjects started in a squat position
at the end of a spring box, positioned lengthwise. Then they placed their hands on
the box and turned into the pike position. In this phase, the neck and shoulders
were placed between the hands. A simultaneous and total extension of the arms
and hips led to the suspension phase. This extended position was maintained
while the subject landed behind the spring box (Herwanger & Geiger, 1980;
Knirsch, 1976; Weller, 1981;Wiemann, 1968).
Procedure
There were two groups: subjects of Group A were asked to give a complete report
of movement errors during the 1 min intertrial interval (estimation); subjects of
Group B were not requested to report errors (control); they disposed of a free 1 min
intertrial interval. Before the experiment started, initial skill level was
determined as in the previous study. Two performance phases, consisting of 16
trials each, were administered within one experimental session: the acquisition
and the transfer phase (5 min after acquisition). During the acquisition phase,
subjects of both groups received global (demonstration) and detailed visual and
verbal information of the movement to be learned. A videotaped model was shown
on a television screen. The model consisted of a perfect execution of the skill by an
expert gymnast. This information was provided four times during the acquisition
phase at regular intervals. The detailed information was presented twice. All the
verbal information was tape-recorded and provided through headphones. The
provision of movement information was held constant between groups. During
the transfer phase, all subjects had a free intertrial interval (1 min) and no
movement information was given. No knowledge of performance was given at any
time.
Scoring
All movements were recorded on video and analysed later with a protocol
constructed for that purpose. The observation protocol contained 24 different items
of the movement. Preceding movement analysis, reliability studies were
conducted - concerning procedures, we refer to the previous experiment. Inter-
332 S.Swinnen
and intra-scorer reliability was high (rp intra =.96 for observer 1, .97 for observer 2;
rp inter a.90). Reliability studies were conducted on each item separately by means
of Cohen's Kappa (Cohen, 1960; Bishop, Fienberg & Holland, 1977). The item
analysis indicated a satisfactory agreement between two observers for 23 of the 24
item scores.
-0
,,,.o--
/;* --
I
0 ,,'4
6
'
.' ,,*
,
/; _.__.
w'
0' /!
i
i
i
I 7 t
I
o i - 1 : : : : : : : : : : : : : : r
Pre- 1 2 3 4 . 5 , 6 7 8 1 2 4 5 6 7 8
test Acquisition ?ransf er
6. Discussion
A final note concerns the information provided to the subjects during practice in
order for learning to occur. Knowledge of results has been used in the timing
studies. It is information about how the movement was actually performed by the
subject. In the studies on the acquisition of gymnastic skills, subjects were only
informed about how the movement should be performed through the provision of a
perfect model and, in some cases, detailed visual and verbal information about the
movement form. Considerable progress in learning was made in the latter
studies without any provision of knowledge of results (or better, knowledge of
performance - Gentile, 1972). This is not to imply that KR or KP is unimportant for
acquiring these akills but rather that the role of these variables should not be
overemphasized at the expense of other sources of information, such as a
movement demonstration. It is astonishing that, up to now, so little is known
about this important instructional variable to facilitate learning.
336 S. Swinnen
ACKNOWLEDGEMENTS
The author is indebted to the Editors and to two anonymous referees for their
suggestions. Thanks are extended to R.A. Schmidt and D.C. Shapiro for their
help and cooperation. Some studies were performed a t the Motor Control
Laboratory, Dept. of Kinesiology, U.C.L.A. Financial support for these studies
was provided through a grant from the U.S. Army Research Institute (Contract #
MDA 903-85-K-0225)awarded to R.A. Schmidt and D.C. Shapiro.
REFERENCES
Stephan Swinnen
Institute of Physical Education
K.U.L., Tervuurse Vest 101,
3030 Heverlee, Belgium. 1
Complex Movement Behayiour:The' motor-actioneontroversy,pp. 339-380
O.G. Meijer 8 K Roth (edibrs)
43 Elsevier Science publishers B.V. (North-Holland), 1988
Chaphr 14
SUMMARY
1. Introduction
Both the 'motor system' approach (Schmidt, 1975, 1982 & 1987) and the 'action
system' approach (Reed, 1982 & 1987) acknowledge the importance of visual
information in the control of movement. However, on the former account the
perceptual, or 'stimulus identification', component of the system, and the motor,
or 'response programming', component are analysed as though they were
logically independent. Under this view, the task of perception appears to be one of
identifying the objects present in the environment and determining their spatial
and temporal locations, and the task of the motor system is to select an
appropriate response and execute the corresponding motor programme. The
movements produced by the motor system are typically analysed as
'displacements of the limbs', without regard to their functions or the
environmental objects toward which they are directed (Reed, 19881, and' hence,
perceptual information about the environment places few constraints on the
characterization of the motor programme. Informational constraints on
movement are often assumed but seldom analysed in motor system research; as a
result, the phenomena of action become harder, not easier, to explain.
In contrast, on an 'action system' account, perception and action are, of
necessity, integral. The reason for this, I believe, is very basic. By definition, an
action has intentionality (Searle, 1983): it is not simply a movement qua
movement, a displacement of body segments in empty space, but a functionally
specific movement directed at an object or goal, such as the environment or the
relationship between the environment and the ador. Organisms don't act in
vacuo, they act on some object in order to achieve some end. This purposiveness or
goal-directedness of action requires that it be oriented with respect to
environmental objects, and, consequently, informational constraints are central
to the theory of action. The task of a theory of visually controlled action may be
The Visual Guidance ofAction 341
In this paper I will address the problem of visually guided action from an
ecological variant of the 'action system' approach, by developing accounts of its
two component problems: the affordance problem, or how organisms perceive
what actions a given situation affords, and the control problem, or how variables
of optical stimulation directly regulate the free parameters of the action system for
a particular task. I will do this by working through the example of insect flight
control to see how these notions fare when applied to actual behaviour, and will
then present some recent work on the perception of affordances and the visual
control of action in humans. My central argument is that the intentions of the
actor select an affordance to be realized, a corresponding mode of action by which
to realize it, and appropriate laws ofcontrol t o visually regulate the action.
My first claim is that the intentional objects toward which actions are directed
are the afTordances of the environment, that is, the functional utilities of
environmental objects, surfaces, and events for an organism with certain action
capabilities. Consequently, afTordances map one-to-one onto specific modes of
action that are tailored to realize them.
that for each afl'ordance there exists a corresponding mode of action that will
realize it, governed by a unique set of laws of control, such as grasping, throwing,
climbing, sitting down, eating, drinking, and so forth. Each action mode involves
an organization of the musculature into task-specific groupings or coordinative
structures, thereby reducing the dimensionality of the action system to a few free
parameters (Kugler,Kelso & Turvey, 1980 & 1982).
Iberall (1977)has argued that autonomous, self-regulating systems like animals
are characterized by stable periodic modes of activity that act to preserve the
system's integrity in the face of environmental fluctuations. Iberall and
McCulloch (1969)offered a suggestive list of such marginally stable action modes
for human behaviour, which we might generalize to other motile animals to
include foraging, eating, drinking, elimination, mating, defence and flight,
grooming, play, and so on. Such action modes are marginally stable because it
typically requires only a small quantity of energy to switch between them. Each
mode also has a characteristic mean relaxation time or period of one cycle, such
as the average interval between feedings, which places certain constraints on the
animal's activities. The dynamic self-regulation of behaviour thence consists of
switching between marginally stable action modes as required to maintain the
system's integrity, and such switches are observed as phase transitions in the
system's behaviour (Iberall & Soodak, 1978). This treatment of stable action modes
as a consequence of dynamic self-regulatory processes is contrasted with their
conventional reification as the result of drives or instincts.
Iberall's list of activities amounts to an enumeration of global action modes,
which require a whole substructure of more specific actions nested within them.
For example, a higher-order action mode, such as foraging, may be composed of
subsidiary modes, such as walking, grasping, chewing, and swallowing. I want
to suggest that the relationships among action modes may be of two types:
switching and nesting. Switching between action modes involves stopping one
action and starting another governed by different laws of control, as when one
shifts from walking to jumping. Nesting one action mode within another can
involve performing two simultaneous actions governed by non-conflicting laws of
control, such as walking down the street and chewing bubble gum at the same
time, or constraining the laws of control for a higher-order action mode, as when
walking is constrained to approach a particular object. It may be possible to
identifjr basic action modes, akin to Reeds (1988)'basic actions', which are action
modes that can be realized on their own, without embedded actions being
necessary for their performance. I admit that this sort of taxonomy is tricky
because it is potentially bottomless (swallowing is composed of lip pursing, and
jaw clenching, peristalsis, etc.) but I expect some principled lower bound on basic
The Visual Guidance of Action 343
action modes can be agreed to on the basis of their intentionality and functional
specificity.
1.3. Choice
Once an action mode is selected and task-specific constraints arise within the
action system, the remaining free parameters of action are regulated by optical
information to tailor movement to local environmental conditions. I am tempted
to call these relationships between optical variables and action parameters laws of
control, after Gibson's (1979) 'rules for the control of locomotion and
manipulation', to avoid any implication of cognitively represented rules that
prescribe behaviour. They are laws because they are based on lawful relationships
between the structure of light and the movements of an observer through a
stationary environment, according to Gibson's (1960 & 1966) theory of ecological
The Visual Guidance of Action 345
optics. But they are not deterministic laws because they can be selectively
harnessed by the intentions of the actor. Finally, I propose that action modes map
one-to-one onto unique sets of control laws, and hence action modes can be
individuated empirically by discoveringthe operative sets of laws.
The received view of the visual control of movement is that the visual system
determines the objective location of a target object in absolute space and time, and
the motor system programmes a limb movement to arrive at the specified space-
time location. Bernstein (1967)first pointed out the findamental indeterminacy
that exists between any central motor command and the resulting movement, due
to 'context-conditioned variability' in the initial conditions of the body and the
initial state of the interneuron pool (Turvey, Shaw & Mace, 1978). As Bernstein
emphasized, the body is a dynamic system of masses in a gravitational field, and
thus a specific spatio-temporal gesture cannot simply be prescribed without
taking the inertial properties of the system into account. This means that the
motor system must solve the 'inverse dynamics problem', calculating the
dynamic forces needed to produce the desired kinematic trajectory on the basis of
precise information about the complex and variable initial conditions of the
system.
In consideration of Bernstein's problem, current motor programming theory
assumes that afference from joint, tendon, and muscle proprioceptors is utilized
in parametrizing a programme (Schmidt, 1982). However, not only is integrating
this vast amount of sensory information and solving the inverse dynamics
problems a formidable task for real-time computation, such proprioceptive
signals themselves appear to be indeterminate with respect to the initial state of
the system, including its joint angles and muscle tensions (Kelso & Stelmach,
1976;Kugler & Turvey, 1979).
An alternative approach is to seek a description of perceptual information that is
already couched in the 'language' of dynamics (Fitch & Turvey, 19781, such as
optical variables that adjust the relationship among force-related parameters of
an otherwise self-organized action system within a restricted context of
constraint. This addresses part of Bernstein's problem because the requisite
forces are specified by action-scaled visual (or haptic, etc.) information about the
real-world situationl. For example, rather than perceiving the absolute space-
time coordinates of the approaching ball (x, y, z, t), a baseball batter could detect
It does not address the question of how the musculature becomes organized to generate these
forces; but see Kelso (1981) and Kugler & Turvey (in press) for a compatible approach to this
problem.
346 W.H. Warren, Jr.
the asymmetry of its optical expansion pattern to regulate the ratio of vertical to
horizontal force that must be applied to position the bat for an accurate swing. To
regulate the initiation of the swing, the inverse of the ball's rate of optical
expansion could be detected, specifying the time to contact (Fitch 6 T w e y , 1978;
Lee, 1980). The optical variables utilized, and the exact action parameters
regulated (force, impulse, stiffness, etc.), are difficult empirical matters, but the
approach seeks special purpose solutions for domain-specific tasks - that is,
specific laws of control for particular modes of action, A functionally-specific
action cannot be performed by issuing a standard set of muscle commands, nor by
executing a standard sequence of anatomically-specific movements; rather, the
movements must be adapted to local conditions on the basis of adequate perceptual
information.
These relationships provide a basis for the theory of visual perception (Gibson,
1979,p. 184).For the control of action, however, we must go one step further, to
show that not only the s u m of current forces acting on the observer, but also the
The Visual Guiahnce of Action 347
force to be applied by the observer, is specified by optical flow. The difficulty is that
the forces yielding optical flow can be of two types: internal forces applied by the
observer, and external forces due to disturbances such as crosswinds, collisions,
gravity, or passive movements as in riding a train. Thus:
How, then, can the observer determine what component of optical flow is due to
an intentionally applied force and what is due to an external disturbance, so as to
compensate for the latter? This problem leads directly to efference copy theories of
motor control Won Holst, 19541, on the following reasoning: If the actor can
precisely anticipate the self-produced component of flow on the basis of known
efferent commands, then the remaining component can be used to determine the
external forces and required compensation.
The problem with such accounts is that the origin of these expectancies is
inadequately explained. Efference copy theories, and other servomechanisms,
require auxiliary assumptions in order to calculate and predict the often complex
optical consequences of often complex movements from known central motor
commands and adopt the appropriate set-point. Given Bernstein's pervasive
context-conditioned variability, there is no predictable relation between motor
commands and optical results, and hence the origin of accurate expectations is
not, and probably cannot, be accounted for. Even in a simple organism like the fly,
some dirt on a wing can radically alter the relationship between muscle activity
and the resulting movement; this problem is compounded in organisms of many
kinematic links (Kelso, 1979;Reed, 1982;Turvey 19771,& 1979).
Gibson avoided this efference copy regress with a fundamental insight: since
optical information specifies the relationship between the actor and its
environment, the organism can act to achieve such a desired state of affairs,
specified by a corresponding optical pattern. Thus, perception is best
characterized not by a servomechanism that matches current and desired flows,
but by a system that achieves meaningful states of affairs that are specified by
flows. Similarly, changes in the relationship between organism and
environment, whatever their source, are specified by changes in optical flow,
which consequently can be used to compensate for (or produce) such movements.
Throughout evolutionary history, for example, a global transformation of the optic
array has corresponded with a movement of the observer through a stable
environment, regardless of whether the movement resulted from internal or
external forces. An organism intending to maintain its ambient orientation can
simply apply the force required to cancel such a change in flow, given by:
348 W.H.Warren, Jr.
This change in force is precisely the variable (derived from somewhat different
considerations)that Kugler et al(1985) have christened 'control'.
For example, if FeXt changes, the resulting change in optical flow specifies the
change in Fint needed to restore the original flow pattern. Most importantly, such
laws of control specify the necessary action forces regardless of where the
disturbances come from - an external force, an internal force, or a motion of the
surrounding environmental surfaces - based solely on the current optical flow
and a choice to maintain orientation. I think these equations capture what Gibson
meant by moving in order to perceive (0and perceivingin order to move (g).
Let me now try to work this story through for a well-studied example of visually
-
guided action - insect flight control to see whether it holds up in the face of real
behaviour. I would like to show that the concepts of action modes and control laws
apply to the organization of a wide variety of biological systems due to their
common context of constraint, although different species will have exploited these
constraints in particular ways. Even organisms as lowly as fruitflies exhibit
remarkably sophisticated control principles based on specific patterns of optical
flow. After what I hope is an illuminating digression, I will return to the control
of human action.
Insect flight control has been studied extensively in the housefly (Musca
domestics), fruitfly (Dmsophila melanogaster), hoverfly (Syritta pipiens), and
blowfly (Caliphom erythmephula). Although there are some important
differences in flight control between M u m and other species, and in pursuit
behaviour between the sexes, I will emphasize common principles.
Flies have relatively few interests in life, chief among them being foraging for
food and water, maintaining ambient orientation while exploring their
surroundings, finding mates, and resting. These activities define the global
affordances of the fly's niche and their corresponding modes of action. Table 1
represents an inexhaustive list of the &ordances and action modes of the fly,
based on Collett and Lands (1975; Land & Collett, 1974) meticulous observations of
houseflies and hoverflies in free flight. For example, nested within the global
The Visual Guidance of Action 349
action mode of foraging there are several subsidiary action modes: an open
medium affords cruising flight for exploration of the environment and large
stationary objects afford uppmzch for the purpose of closer inspection; if the
objects turn out to be flowers or fruit they afford landing and ultimately nectar
with a particular sucrose concentration affordsfeeding.
Such affordances - what the fly can do in its environment - are determined by the
relationships that hold between properties of the fly's action system and material
properties of that environment. Nectar-bearing flowers afford feeding for an
organism with particular ingestive equipment and metabolism, objects that are
large with respect to the fly afford landing, conspecifics of the opposite sex afford
copulation, and EO on. To the extent that a fly can detect the properties of objects
relative to its own action capabilities, discriminating flowers from other large
objects, conspecifics of the opposite sex from other insects, and so forth, the fly
may be said to perceive the affordances of its environment and can act on them. In
performing such an action, the fly chooses a particular affordance to be realized,
a corresponding mode of action, and the accompanying laws of control.
350 W.H.Warren, Jr.
Let me detail the laws of control for a few of the best understood action modes,
corresponding to Gibson's (1966, p. 59) three general modes of orientation as they
are manifested in the fly: "Basic orientation to the e a r t h or hovering, "oriented
locomotion" or cruising flight, and "temporary orientations to objects" such as
approach t o large stationary objects, pursuit of small moving objects, and
landing. The statements here are interpretations based on my reading of a
literature that is incomplete and occasionally contradictory, and, thus, I will note
my speculations where appropriate.
First a few basics. A flying body possesses six degrees of freedom in a
rectangular coordinate system (Figure 1): pitch (p), the extent of an up-and-down
rotation about the fly's lateral Euds, yuw (w), the extent of a side-&side rotation
about the vertical axis, roll (r), the extent of a rotation about the longitudinal axis,
and tmnslution along each of the three major axes - lateral displacement (XI,
vertical displacement (y), and forward displacement (z). For present purposes, I
will assume a fly-centred coordinate system that translates and rotates with the
fly's body axis (Wagner, 1986a). Thus, "vertical", "lateral", and "forward
movements are defined with reference to the fly's initial position rather than to
gravity. This is appropriate because relations between changes in optical flow and
required force remain invariant under translations and rotations of the fly,
The Visual Guidance of Action 351
X
ivc pitch
The question is, in what form to write control laws? The earlier discussion points
toward relations between kinematic variables of optical flow and dynamic force-
related parameters of action. However, at least two scales of explanation are
required: one at the general scale of flying organisms, relating optical flow a t a
moving point of observation to changes in the application of forces, and one at the
species scale, relating detected variables of flow to motor parameters such as
wingbeat amplitude. Given general laws, different species-specific apparatus
may have evolved for resolving the rotational and translational components of
flow and for generating appropriate forces. Thus, although hoverflies rotate
smoothly by continuously varying torque level and houseflies rotate with a
sequence of 'saccadic' turns by varying the rate of abrupt torque spikes (Wagner,
1986b1, the total amount of required torque may be specified by the same general
law.
In what follows I will suggest both general laws of control and species-specific
relations for a creature such as the hoverfly, which can presumably manipulate
the amplitude (a) of each wing as well as the bilateral stroke angle andor wing
pitch, which adjusts the angle a of the force vector with respect to the body. In
reality, species-specific relations w i l l not be this simple, given the problem of
context-conditionedvariability in the production of forces.
The equation of translatory motion for the fly is:
F = mi + kir + mg
where F is the force applied by the fly, mi is the component that accelerates the
fly, kiis the component absorbed by drag, and mg is the component expended
against gravity,with m equal to body mass and k a drag coefficient. A comparable
equation can be written for torque with rotatory motion. It is important to note that
the force applied by the fly is opposed by a drag force D due to air friction; indeed, it
has been estimated that 99% of the torque produced by a housefly is used to
The Visual Guidance of Action 353
overcome drag (Heisenberg & Wolfe, 1984). I will assume drag increases with
velocity according to D = ki,although there is some controversy on this point2.
Consequently, the generation of a constant force by the fly w i l l result in an initial
acceleration that equilibrates to a constant velocity as the applied force is balanced
by the increasing drag. At equilibrium, m i goes to zero and the fly's terminal
velocity is proportional to the applied force, k = F/k - mgk. It follows that to
change to a new constant translational or angular velocity, the fly must apply a
change in force or torque over some interval of time - our variable of control.
Consequently the constant gravitational term drops out, leaving & = @/k.
What this simple physics means for the control of action is straightforward: since
the fly's velocity (k) is related to rates of global flow (2) by the laws of optics, flow
can conversely modulate the forces applied by the fly which determine its
terminal velocity. Roughly speaking, the laws of flight control follow two general
principles: changes in the rate of flow regulate changes in the net magnitude of
force, and changes in the direction of flow regulate changes in the direction of
force, with respect to initial values. Thus, we have:
This is the form I will adopt for laws of control, adjusted by the proper constants.
The primed optical variables are actually placeholders for an elaborated theory of
optical flow, which I will not attempt here (see Lee, 1974; Longuet-Higgins &
Prazdny, 1980;Koenderink & Van Doorn, 1978;Warren, Morris & Kalish, 1987).
Briefly, rotations of the observer (p, i , i) yield translational (5,
w' ) or rotational
(f') patterns of flow in which the magnitudes of the flow vectors are completely
determined by observer movements independent of the distance of the reflecting
surface. In this case, the optical coefficient c simply scales the fly's rotational
velocity to some measure of the global velocity of flow. On the other hand,
translations of the observer (k, 9, i) produce centrifugal patterns of flow in which
the flow vectors diverge from a stable point in the direction of locomotion and
converge again to a stable point in the direction from whence one came (Gibson,
1950). In this case, the optical variables 3, jr', i' stand for some measure of the
global rate of flow which is not independent of distance, with the point of
expansion in the designated direction. This distance scaling is incorporated in the
optical coefficient c, a sleight-of-hand I will not worry about further here since it
is sufficient to make my point, but which must ultimately be addressed by
ecological optics.
The laws of control for several modes of action, as I currently understand them,
appear in Table 2. They are broken down rather classically with separate laws
governing yaw, roll, and translation, although more appropriate intrinsic
coordinata systems could surely be devised. Corresponding t o each general law is
a provisional species-specific relation for an insect that manipulates the
amplitude of each wing and the force angle independently.
(Vertical)
The result is an upward movement 9 that cancels the upward flow and restores
the fly's orientation. Such modulations of upthrust were found by Srinivasan
(1977) in the house fly and Gotz (1968; Gotz et al, 1979) in the fiuitfly, with insects
tethered to a force transducer and presented with vertically moving striped
patterns (Figure 2). The opposite results obtain for downward flow.
Vertical lift could be adjusted by increasing the amplitude (a) of both wings,
coupled with an appropriate adjustment in stroke angle and wing pitch so as to
yield the force vector at an angle a to the longitudinal axis of the body. Hence, the
change in the vertical component of that vector is given by the species-specific
relation:
The Visuul Guidance of Action 355
+ -9 -w w
As specified in Table 2, similar laws govern the other degrees of freedom for
movement. Briefly, a rotational pattern of flow i' about the longitudinal axis,
specifying that the fly is rolling with respect to the environment, regulates the
differentialupthrust of the two wings by:
-
A (UL U, ) = (WC)Ai' (Roll)
This generates a roll torque in the direction of flow (Srinivasan, 1977; Goodman,
1965; see Figure 2). Analogously, yaw is obtained in the familiar rotating drum
apparatus which creates a lateral flow field with a global rate of i'. This flow
pattern evokes a differential forward thrust
356 W.H. Warren, Jr.
nitial flow rates = 0) Forward : A(FLt FR) = (Wc) = A (aL t aR) cosa
resulting in a yaw movement that tracks the rotating drum (Collett, 1980a & b;
Heisenberg & Wolfe, 1984; Virsik & Reichardt, 1976). Finally, a forward
translation of the surrounding surfaces relative to the fly will produce a
centrifugal flow field with a global rate 2. To maintain stable orientation, this
modulates the total thrust by:
This model of force regulated by flow velocity implies that some constant rate of
flow is required to maintain a constant adjusting force. This is fine for transient
disturbances, such as gusts of wind, but for continuous disturbances it means
that the fly cannot fully cancel the optical flow. For example, if the rotating drum
moves leftward at a constant velocity, the fly's yaw torque will also increase; this
will lead to a decrease in the rate of optical flow as the fly catches up to the drum,
which will in turn decrease the torque, and so on. The fly will equilibrate to a
constant velocity somewhat slower than the velocity of the surround, slipping ever
further behind it.
The data on this point are contradictory: it is exactly what is reported by Collett
(1980b) for yaw in hovedies, where i f l y = (.8) idrum. But Virsik and Reichardt
(1976), based on torque measurements in tethered houseflies, report that the fly's
yaw velocity equals that of the drum, way = id-, as do Heisenberg and Wolfe
(1984) for both tethered and freely rotating fruitflies3. A gain equal to 1 implies
that the fly maintains a constant yaw torque once optical flow goes to zero. This
The latter speculate that Collett's flies may have had stationary as well as moving surfaces in
view.
358 W.H. Warren, Jr.
Am-h
Approach is equivalent to cruising flight toward a large, more or less stationary
object with the added constraint that the fly 'centre' the object, that is, orient its
longitudinal axis toward it, as in Collett and Land's (1975) observations of flight
The Visual Guidance of Action 359
toward flowers. This centering phenomenon often has been called 'hating', or
'slow tracking', of targets because it is studied with tethered flies who can not
actually approach them (Poggio & Reichardt, 1973; Wehrhahn & Reichardt, 19751,
but, as Collett (1980a) points out, it is an aspect of approach under free conditions.
In addition to centering the target object, the fly also translates roughly toward it,
acting, as in Gibson's (1979, p. 233) rule for approach, "to ma& a patch in the
optical array." Thus, the fly yaws until it has centered the target, whereupon
forward thrust will carry it home (Collett & Land, 1975). Due to this added aspect
of centering, the laws of control for approach include the optical position of the
target (w't ) in addition to its optical velocity (w't ) (Virsik & Reichardt, 1976;
Collett, 1980b). For example, the yaw to centre a large stationary target is
governed by:
Pursuit
The hot pursuit of small moving objects, which correspond to conspecifics in the
ecological niche of the fly, is different from approach to large stationary objects.
First, the pursuit mode only occurs in males and often terminates in pouncing;
thus it is embedded in the male mating mode. Second, the fly centres the moving
target by regulating its yaw velocity solely on the basis of the target's optical
position (w't ) (Land & Collett, 1974; Collett & Land, 1975; Collett, 1980 a & b).
360 W.H. Warren, Jr.
The question of 'reflex interaction' arises again here, for laws of ambient
orientation (the 'optomotor response') operating at the same time as laws of
pursuit will act to throw the fly off target. Collett (1980b) found such an effect with
free-flying hoverflies in the rotating drum, forcing him to postulate an efference
copy mechanism to cancel the optical effects of intended pursuit movements
under normal conditions. Virsik and Reichardt (1976) attempted to dispense with
the efference copy by proposing that the effects of the two systems were additive to
explain similar results for female tracking in tethered houseflies. On the other
hand, Wagner (1986~)found no such interaction in free-flying houseflies, male or
female, and suggested that the temporal frequency response of the orientation
system is lower than that of the pursuit system, so that the former simply does not
interfere with the frequent saccadic turns typical of pursuit. Since an ecological
theory is sceptical of resorting to efference copies, it is sympathetic to such
explanations that seek instead to account for the regulating of actions on the basis
of available optical information.
Landing
The landing mode simply adds two constraints to approach. First, when I,
specifying time to contact, reaches a particular margin value (T = d), decelerate by
decreasing forward thrust (or upthrust) to maintain the rate of change of optical
expansion at a value of i = -5, which will guarantee a soft landing (Lee, 1976;
Wagner, 1982):
In the housefly, Wagner (1986b) found that target velocity and other undetermined factors also
contribute to yaw and pitch control during male pursuit.
The Visual Guidance of Action 361
At a second margin value (T = e), extend the landing gear (Coggahall, 1972;
Goodman, 1960;Eckert, 1980;Wehrhahn, Hausen & Zanker, 1981).
2.3.Fly choice
The resulta discussed here on insect flight bare a striking resemblance to those
on human visual control of posture and locomotion. For example, the similarity of
postural control to hovering is demonstrated by Lee and Lishman's (1975;Lee &
Aronson, 1974) studies of human balance in the swinging room, in which body
sway is anchored to translations of the visual surround, maintaining ambient
orientation (see also Berthoz et al, 1979). Forward locomotion, akin to cruising
362 W.H. Warren, Jr.
flight, is also detected via centrifugal patterns of optical flow in humans (Warren,
1976; Warren et al, 1987). Studies of treadmill running in the swinging room
reveal that the laws for locomotion are analogously nested within those for
postural control, for adjustments in gait occur to preserve balance when the
visual surround is perturbed during running as well as standing (Young, 1988).
The point here is not simply that human action is like the fly's in its lawfulness,
but that fly action is like the human's in its goal-directedness and self-initiation.
My own recent research has focused on the perception of afTordances and laws of
control in humans.
3. Perceiving affordances
First, consider the ordinary activity of stair climbing (Warren, 1983 & 1984). It is
apparent that some steps in the ground surface afford climbing and others do not,
and that some are easier to climb than others. Many of us have experienced the
frustration of climbing 'monument steps' with short 5" to 6" risers and deep
treads, which are designed more for visual effect than the comfort of climbers. To
analyse the affordance of climbability, the fit between the organism and the
environment that supports the activity of climbing must be determined using
methods of intrinsic measurement.
The Visual Guidance of Action 363
IShort
The same should be true of the optimal point. Optimal riser height was
determined empirically by measuring oxygen consumption during climbing by
short and tall climbers. Riser height varied from 5 ' to lo", with the diagonal
distance between stairs held constant a t 14" and step frequency held constant at 50
stepdmin. The results reveal minimum energy expenditure per vertical metre of
travel at riser heights of 7.7" for the short subjects, and 9.5"for the tall subjects
(Figure 4a). These values are somewhat higher than common indoor risers of 6"
to 7", and help to explain the trouble many people have with monument steps. Far
from making long flights of stairs easier to climb, low risers increase total energy
expenditure by as much as 15%, and deep treads may make an efficient step
length impossible as well.
When the same data are replotted on intrinsic axes in Figure 4b, the curves
become parallel with an optimal xo = .26 for both groups. Thus, the optimal point
also appears to be a constant over scale changes. To see whether the optimal
15
.
Y
-
14
8
13
12
5 6 7 8 9 1 0 1 1 1 2 0.1 0.2 0.3 0.4
Riser Height (inches) R/L
(4
Figure 4. Energy expenditure per vertical metre of travel as a function of (a)
absolute riser height, and (b) the ratio of riser height to leg length, for short
and tall subjects. Arrows indicate perceptually preferred risers (see Figure
5).
366 W.H. Warren, Jr.
I r loo
80
Short
0Tall
(62
c
8 60
0
f
c
40
2
20
i
0 I L
5 6 7 8 9 10 0.1 0.2 0.3 0.4
Riser Height (inches) RIL
(4 (b)
I t loo
-
Small
0
\ a,, 0Large
- ?, ‘s‘I
-
J I
40 60 80 0.6 1.0 1.4 1.8 2.2
Aperture (cm) A/ S
(a) (b)
Relatively little work has been done on the control laws that regulate the forces
applied during human locomotion, in large part because of the technical
problems in recording such gait variables. Recently we tried to determine a law
for the visual control of step length during running over irregular terrain
(Warren, Young & Lee, 1986).
Figure 8. Four hypotheses for how a runner could increase step length (S).
See text for explanation.
running on bare patches of ground. The targets were visible 5 m ahead of the
treadmill.
To determine what gait parameters vaned with step length, we broke the
problem into several parts (Figure 8). First, total step length (S) &om one heel-
strike to the next could be adjusted by changing the horizontal distance ( C )
travelled by the centre of mass between heel-strikes, or by changing the forward
reach (R) of the landing foot relative to the centre of mass:
S=AR+C
372 W.H.Warren, Jr.
However, we found that changes in C accounted for 91% of the total variance in
S, while AR accounted for only 9%. Apparently, runners are acijusting some
parameter that determines the distance travelled by the centre of mass. This
could be done by varying either the total step time TI or the forward flight velocity
v, according to:
C=VT
T = I / mg + A d g
If the runner lands similarly on each step, the last component would be 0. This is
exactly what we found, with Umg accounting for nearly 100% of the variance in
T. It appears a s though runners modulate a single gait parameter of vertical
impulse in order to adjust step length. How, then, might vertical impulse be
regulated by an optical variable - in other words, what is the law of control for step
length?
The required duration of the step between two upcoming targets (T) at the
runner's current velocity is equal to the difference between the time to contact
with the two targets (Figure 9). Since time to contact with a target is specified by
the inverse of its rate of optical expansion, or I, the difference between T for the
two targets (A I = 12 - 11 ) specifies the required duration of the step. Substituting
this into the equation for T, we get the control law:
I=mgAr
I=mgAr
-
-400 -300 -100 0 msec
Figure 9. Vertical impulse (I) on the first target could be directly regulated
by the optical variable Ar, the difference in time to contact between two
upcoming targets. Runners require visual information about the second
target between 100 and 400 msec before landing on it, as they are preparing
to land on, and thrust off from, the first target.
conditions by either covering up the approaching target far ahead of the runners,
forcing them to look down, or covering up the targets right in front of their feet,
forcing them to look ahead. Our error data for landing on target indicate that
for accurate modulation of vertical impulse, the runner must have visual
information about both targets between 100 msec and 400 msec before heel-strike
on target h, corresponding to the stance on target tl and the preceding flight
phase (Figure 9). In short, the data are consistent with a law of control for the
running mode under which the action parameter of vertical impulse is directly
modulated by the optical variable AT , automatically yielding an appropriate step
length.
can be interpreted at two grains of specificity: general laws and their special
cases, the species-specific relations between optical variables and motor
parameters. These latter relations must be interpreted cautiously, for there may
be many movements that can instantiate a given action, and as the context
conditioned variability of movement increases in species with more biokinematic
links, so these relations may become increasingly complex and variable. The
current hypothesis is that perceptual information (visual, haptic, etc.) specifies
the forces that are required to realize an intended action within the constraints of
a given situation. The species-specific relations further constrain the means by
which such forces are applied. Precisely how the musculature becomes organized
so as to generate those forces, is a problem at the level of dynamical self-
organization of the action system (Kugler & Turvey, in press).
ACKNOWLEDGEMENTS
I would like to thank Mike Turvey, Ed Reed, Bob Shaw, Tom Collett, and Herman
Wagner for helpful comments. Thanks to Diane Hargreaves, Stephanie Han, and
Janice Viticonte for their fiuious typing.
REFERENCES
Goodman, L.J. (1960). The landing response of the fly Lucillia sericata and other
calliphorinae. Journal of Experimental Biology, 37,854-878.
Goodman, L.J. (1965). The role of certain optomotor reactions in regulating
stability in the rolling plane during flight in the desert locust, Schistocerca
gregaria. Journal of Experimental Biology, 42,385-407.
Gotz, K.G. (1968). Flight control in Drosophila by visual perception of motion.
Kybernetik, 4,199-208.
Gotz, K.G., Hengstenberg, B. & Biesinger, R. (1979). Optomotor control of wing
beat and body posture in Drosophila. Biological Cybernetics, 35,101-112.
Gotz, K.G. & Wandel, U. (1984). Optomotor control of the force of flight in
Drosophila and Musca. II. Covariance of lift and thrust in still air.
Biological Cybernetics, 51,135-139.
Gotz, R G . & Wenking, H. (1973). Visual control of locomotion in the walking
fruitny Drosophila. Journal of Comparative Physiology, 85,235-266.
Gould, S.J. & Lewontin, R.C. (1979). The spandrels of San Marco and the
Panglossian paradigm: A critique of the adaptationist programme.
Proceedings of the Royal Society of London, B, 205,581-598.
Hallford, E.W. (1984). Sizing up the world: The body as referent in a size
judgement task. Doctoral Dissertation, Ohio State University.
Heisenberg, M. & Wolfe, R. (1984). Vision in Drosophila. Berlin: Springer.
Iberall, A.S. (1977). A field and circuit thermodynamics for integrative
physiology. I. Introduction to general notions. American Journal of
Physiology, 233, R171-Rl80.
Iberall, A. & McCulloch, W. (1969). The organizing principle of complex living
systems.Journa1 of Basic Engineering, 19, 290-294.
Iberall. A. & Soodak, H. (1978). Physical basis for complex systems: Some
propositions relating levels of organization. Collective Phenomena, 3, 9-24.
Kelso, J.A.S. (1979). Motor-sensory feedback formulation: Are we asking the right
questions? Behavioral and Brain Sciences, 2, 72-73.
Kelso, J.A.S. (1981). Contrasting perspectives on order and regulation in
movement. In J. Long & A. Baddeley (Eds.), Attention and Performance Lri
(pp. 437-457). Hillsdale, N J Lawrence Erlbaum.
Kelso, J.A.S. & Stelmach, G.E. (1976). Central and peripheral mechanisms in
motor control. In G.E. Stelmach (Ed.), Motor Control: Issues and trends
(pp. 1-40). New York Academic Press.
Koenderink, J.J. & Van Doom, A.J. (1978). How an ambulant observer can
construct a model of the environment from the geometrical structure of the
visual inflow. In G. Hauske & E. Buttenandt (Eds.), Kybernetik (pp. 224-
247). Miinchen: Oldenburg.
Kugler, P.N., Kelso, J.A.S. & Turvey, M.T. (1980). On the concept of coordinative
structures a s dissipative structures. I. Theoretical lines of convergence. In
G.E. Stelmach & J. Requin (Eds.), Tutorials in Motor Behavior (pp. 3-47).
Amsterdam: North-Holland.
Kugler, P.N., Kelso, J.A.S. & Turvey, M.T. (1982). On the control and co-
ordination of naturally developing systems. In J.A.S. Kelso & J.E. Clark
(Eds.), The Development of Movement Control and Coordination (pp. 5-78).
New York John Wiley.
The Visual Guidance of Action 377
Kugler, P.N. & Turvey, M.T. (1979).Two metaphors for neural afference and
efference. Behavioml and Bmin Sciences, 2,305-307.
Kugler, P.N. & b e y , M.T.(in press). InformatwG Natural Law and the Self-
Assembly of Rhythmic Movement. Hillsdale, N J Lawrence Erlbaum.
Kugler, P.N., Turvey, M.T., Carello, C. & Shaw, R.E. (1985). The physics of
controlled collisions: A reverie about locomotion. In W.H.Warren & R.
Shaw (Eds.), Persistence and Change (pp. 195-229). Hillsdale, N J
Lawrence Erlbaum.
Land, M.F. & Collett, T.S. (1974). Chasing behavior in houseflies. (Fannia
canicularis). Journal of Comparative Physiology, 89,331-357.
Lee, D.N. (1974).Visual information during locomotion. In R.B. McLeod & H.
Pick (Eds.), Perception: Essays in honor of J.J. Gibson (pp. 250-267).Ithaca,
Ny:Cornell University Press.
Lee, D.N. (1976).A theory of visual control of braking based on information about
time to collision. Perception, 5,437-459.
Lee, D.N. (1980).Visuo-motor coordination in space-time. In G.E. Stelmach & J.
Requin (Eds.), Tutorials in Motor Behavior (pp. 281-293). Amsterdam:
North-Holland.
Lee, D.N. & Aronson, E. (1974).Visual proprioceptive control of standing in
human infants. Perception and Psychophysics, 15,529-532.
Lee, D.N. & Lishman, J.R. (1975).Visual proprioceptive control of stance. Journal
of Human Movement Studies, 1, 87-95.
Lee, D.N. & Reddish P.E. (1981).Plummeting gannets: A paradigm of ecological
optics. Nature, 293,293-294.
Longuet-Higgins, H.C. & Prazdny, K. (1980). The interpretation of a moving
retinal image. Proceedings of the Royal Society of London, B, 208,385-397.
Newell, K.M.& S c d y , D.M. (1987).The development of prehension: Constraints
on grip patterns. Unpublished manuscript.
Poggio, T. & Reichardt, W. (1973).A theory of pattern induced flight orientation of
the fly Musca domestica. Kybernetik, 12,185-203.
Reed, E.S. (1982). An outline of a theory of action systems. Journal of Motor
Behavior, 14,98-134.
Reed, E.S. (1988).Applying the theory of action systems to the study of motor
skills. This Volume.
Rosen, R. (1978).Fundamentals of Measurement and Representation of Natural
Systems. New York: North-Holland.
Schmidt, R.A. (1975). A schema theory of discrete motor skill learning.
Psycholosical Review, 82,225-260.
Schmidt, R.A. (1982). Motor Control and Learning: A behavioral emphasis.
Champaign, IL:Human Kinetics Publishers.
Schmidt, R.A (1988).Motor and action perspectives on motor behaviour. This
Volume.
Schuring, D.J. (1977).Scale Models in Engineering. New York Pargamon Press.
Searle, J.R. (1983).Zntentiomlity. Cambridge: University Press.
Srinivasan, M.V. (1977).A visually-evoked roll response in the housefly. Journal
of Comparative Physiology, 119,l-14.
Stahl, W.R. (1961). Dimensional analysis in mathematical biology. I. General
378 W.H. Warren, Jr.
10, 683-803.
Warren, W.H. Morris, M.W. & Kalish, M. (1987). Perception of translational
heading from optical flow. Submitted for publication.
Warren, W.H. & Whang, S. (1987). Visual guidance of walking through
apertures: Body-scaled information for dordances. To appear in Journal of
Experimental Psychology:Human Perception and Performance, 13.
Warren, W.H., Young, D. & Lee, D.N. (1986). Visual control of step length during
running over irregular terrain. Journal of Experimental Psychology:
Human Perception and Performance, 12, 259-266.
Wehrhahn, C.,Hausen, K. & Zanker, J. (1981). Is the landing response of the
housefly (Musea) driven by motion of a flow field? Biological Cybernetics, 41,
91-99.
Wehrhahn, C. & Reichardt, W. (1975). Visually induced height orientation of the
fly Musca domestica. Biological Cybernetics, 41, 91-99.
Young, D.S(1988). Describing the information for action. This Volume.
1
William H.Warren, Jr.,
Depts of Psychology and
Cognitive and Linguistic Sciences
Brown University, Providence,
Rhode Island 02912, USA.
Complex Movement E+haviour: 'The'motor-action controversy, pp. 381-401
O.G. Meijer8t K. Rothi(editors)
Q Elsevier Science Publishers B.V. (North-Holland),1988
Chapter 15
H.TA (John)Whiting
SUMMARY
The concept of 'imitation' is discussed in the context of ideomotor phenomena in
general. A historical sketch confirms that the problems of definition were as
prevalent in the past as they are today - the problems having to do with the extent
reference of the concepts involved. The difficulty of delineating a particular
movement - behaviour - as 'imitative' is illustrated by a comparison of many
recent empirical studies with infants which purport to be studies of imitation. A
distinction is made between two modes of imitation - 'echokinetic' and 'syn-
kinetic' (Prim, in press). The relative absence of studies in which movement
parameters per se are used as dependent variables is taken as the departure
point in the structuring of a research programme on 'observational learning' in
which such parameters are the central concern. The results of an initial
experiment within this programme, utilizing a synkinetic paradigm, are
presented and discussed. Future directions are proposed.
1. Introduction
One needs to know not only that a eunuch is a man who has been castrated,
but also that castration destroys sexual potency, and one must somehow
also know that sexual potency is to be taken as a symbol of efficacy in
general.
The term 'imitation' might be added to Saugstads list. Saugstads point is that
the situation is a natural consequence of insistence - in particular on the part of
positivistic thinkers - on the use of ostensive and operational definitions. He points
out that this kind of procedure leaves theoretical terms vaguely and diffusely
defined allowing the psychologist, who utilizes such approaches, only to be able to
conjecture, loosely, about the nature of the type of events postulated.
In the title of this paper, the reader is faced with two terms, 'imitation' and
'complex cyclical actions', the former of which, in particular, is seldom
accurately defined, and for which a meaningful reference system is often lacking.
It is legitimate, therefore, to ask for more information in order to understand, for
example, what is meant by imitation when applied to human motor actions. What
is the overlap between the extent reference of the concept 'imitation' and related
concepts like 'social facilitation', 'observational learning', 'similarity matching',
etc.? Is there any degree of consistency in the way that different researchers
operationalize the term? Does this lead to dificulties in interpreting or comparing
their findings?
Historically, as Scheerer (in press) in his excellent overview has so succinctly
traced, imitation, as a phenomenon has been repeatedly 'rediscovered. While, as
he suggests, most historical sketches give the impression that 'imitation', as a
concept, only came to light during the period that was dominated by the theory of
Imitation and Learning 383
It is, perhaps, not surprising that imitation should have captured the
imagination of thinkers and experimentalists throughout the ages. Its ubiquity
would, in itself, be a sufficient justification. In everyday life, we are constantly
confronted by dynamic models. It would be most surprising if they were not to
384 H.T.A.Whiting
rational empirical
mechanical reflective
voluntary involuntary
form matter
process product
direct indirect
conscious unconscious
echokinetic synkinetic
Given all the possible combinations of these, and similar dimensions (levels of
analysis), a taxonomy would clearly help to clarify some of the conceptual
problems. A comprehensive taxonomy, however, does not appear to exist.
A useful attempt was made by Aronfreed (1969)in his fourfold categorization of
observationalbehaviour:
Social facilitation: In which there is no necessary correspondence between the
stimulus properties of the evocative and the elicited behaviour. Instead, the
evocative stimulus has a generalized motivational or releasing effect. Social
facilitation might be looked upon as making an observer aware of a new possibility
and his operationalizing that possibility by an action that does not necessarily bear
any direct resemblance to that of the behaviour observed. An example might be
the way in which the mere observation of an experienced player using the side
wall in the game of 'squash induces the beginner to do likewise without there
being any, necessary, correspondencebetween the movements of the two players.
Choice matching: The matching of a person's discrete behavioural choices (e.g.,
an environmental location, the selection of an attribute of objects, or the verbalized
prediction of an impersonal event) to those of another person, on the basis of
Imitation and Learning 385
... thanks to the faculty of instinctive aural and visual imitation, the child,
through frequent repetition of the reflex, develops the activities of different
groups of muscles. This makes its speech articulate; at the same time its
external body movements become intelligent.
Imitation and Learning 389
that the lower leg takes to swing forward and its corresponding return motion,
different gait categories could be simulated and accurately classified by observers.
By plotting the variation in time of the angular displacements of the upper and
lower legs (approximated as a sinusoid and a saw-tooth-like function,
respectively) he was able to establish break points ('bp') varying from
(theoretically) 0-1 (Figure 1).For a bp of 0.6 it was easy to generate - with the
computer - a movement labelled by observers as 'walking' but, with such a bp, it
was hardly possible to arrive at a compelling impression of running. On the other
hand, for a bp of 0.9, the reverse is true. With a breakpoint of 0.4, the movement is
interpreted as skating.
e
tip
In addition to these classes of variable (form and force prediction) there is another
class of (superordinate) movement variable which might be important in both
movement perception and movement production. This is well signalled in
Hopkins's and Prechtl's (1985) qualitative approach to the development of
movements during early infancy. As they point out, an infant's general
movements can be quantified in terms of their individual properties (i.e., their
metrical or form topological characteristics) but this would have nothing to say
about whether these movements were, for example, 'smooth or 'jerky'. Hopkins
and Prechtl consider that for the appraisal of the latter "direct observation based
on Gestalt perception and combined with an element of aesthetic appreciation so
Imitation and Learning 391
that the form quality (Gestaltqualit&) of a movement can be recognized' (p. 180) is
necessary. While such an approach may be necessary for the appraisal of the
movements of babies, alternative methods, as will be demonstrated below, are
available when working with adult subjects.
While these different categories of movement parameter would seem to be
important in delineating potential dependent variables for work on imitation and
related concepts, the literature in this respect has, to date, been very limited. Even
when authors have directed their attention to imitation (in the Katz sense) with
more mature subjects, they have tended to restrict themselves, again, either to
actions or, when movement parameters have been used, t o the sequencing of
movements per se (e.g., Landers & Landers, 1973; Martens, Burwitz &
Zuckermann, 1977; Carroll & Bandura, 1982; Ross et al, 1985).
An exception to this rule is the work of Williams (1985) who, appreciating these
deficiencies in the literature, included in his series of studies a wide range of
movement parameters (timing and displacement of limb traverse, sequencing of
movements, etch In addition, sensitive to the importance of the coupling of the
visual perception of movement and movement production, his work draws heavily
on the theoretical background of those pioneers of 'biological motion perception'
Johansson and Cutting (Johansson, 1950 & 1973; Cutting, 1978, 1981 & 1983;
Cutting & Kozlowski, 1977; Cutting & Profitt, 1981). While his search for links to
the motor learning literature lead him into considerations of motor programming
interpretations, it has t o be appreciated that he sees this more as a heuristic
device. He would be the first to agree that other interpretations of his findings are
possible. Basically, he suggests that:
throwing action which was preceded by between 2 and 8 forward and backward
preparatory movements. Performance under normal and point-light conditions
were compared. Sequence order was shown to be better imitated under normal
display conditions when the number of preparatory movements was 5 or less.
Differences between the conditions were not significant when the number of
movements exceeded 5.
In another experiment, subjects were required to 'learn' a throwing action
consisting of preliminary movements under three viewing conditions, normal,
point-light, and segment-light. They watched the model then produced what was
seen 10 times. Movement response was recorded via goniometry and surface
EMG. There were no significant differences between the groups for each
experimental condition. The order of movements was learned first, during which
time limb displacement was attenuated. When correct order had been achieved,
range of limb movement similar to the model was produced. A measure of
relative timing remained constant across trials in all conditions but was always
slower than that of the model. A further significant finding, in the context of the
research to be reported and of the comments made above, was that individuals
classified - on the basis of an independent appraisal - as 'poor imitators' showed a
higher incidence of muscular co-contraction during movement production than
'good imitators', i.e., they showed less well-coordinatedmovements.
Many of the considerations reported above have been taken into account in
developing a research programme directed towards imitation, and related
phenomena, of movements (in the Katz sense). A primary interest of the author in
learning per se makes the studies on infant 'imitation' reported of less central
importance. They do, however, pose interesting questions and provide useful
guidelines. The studies of Williams (1985) - in that they invoke a range of
movement parameters as dependent variables - provide much more interesting
evidence for the imitation of movement per se although they were construed in the
framework of a performance rather than a learning paradigm. These studies
were also echokinetic in nature and, unlike the studies on infants reported,
subjects were able, of course, to utilize the instruction to imitate the model
provided.
Quite clearly, all the signalled nuances give rise to a whole host of potential
reeearch strategies so that, eventually, the selection made is determined more by
Imitation and Learning 393
the current interests of the researcher and what he considers to be the most
fruitful lines of development than any logical priority that might suggest itself. In
this respect, the present author is attracted by the echokinetic / synkinetic
distinction of Prinz (19851, by the observation that people acquire certain
movement characteristics of significant others - without apparently being
involved in conscious imitation - and Williams's (1985) signalling of those
parameters that might lend themselves to being imitated.
Some of the parameters that might lend themselves to being 'imitated have been
signalled by Williams (1985) namely, sequence (of sub-movements), order,
amplitude, and relative timing. These have been further classified as either
394 H.T.A. Whiting
'topological' or 'metric' by the present author. There is, however, another class of
higher-order parameters, as previously signalled and elaborated by Hopkins and
Prechtl (1985), namely the 'smoothness' or 'jerkiness' of movements.
Incorporation of this class of movement parameters into the experimental
programme in terms of the 'fluency' of movements adds a new dimension.
4.3. Apparatus
The apparatus proposed for the initial experiments in the programme was
determined more by the author's current interest in the learning of cyclical
actions and its availability rather than by any preconceived notions of the most
suitable task. Considerable experience had been built up over the past few years
on the learning of slalom-type ski movements utilizing a so-called 'ski-simulator'
(Den Brinker & Van Hekken, 1982; Den Brinker et al, 1985a & b; Bootsma, Den
Brinker & Whiting, 1986). This apparatus while lending itself to the kind of
paradigm envisaged also reflects the fact that in the natural situation the
learning of skiing movements is often facilitated by 'following' the actions of an
experienced model.
The ski-simulator (Figure 2) consists of a platform, designed to ride over a pair
of bowed metal 'runners', which, because of the spring system beneath, returns
to the central resting position after having been moved in a sideways direction.
When the unloaded platform is disturbed by a horizontal force, it shows a damped
oscillation around 4 Hz. Attached to the centre of the platform is a bank of light-
emitting diodes, the movement characteristics of which are picked up, and
registered, via a SELSPOT system. It is this signal which is used to generate data
about amplitude, frequency and fluency of the movement of the platform
previously shown by Den Brinker and Van Hekken (1982) to be reliable indices of
learning for this taskl.
For methodological and analytical details, the reader is referred to that article.
Imitation and Learning 395
In the proposed research programme the decision was taken to initially explore
the effects of a synkinetic, as against an echokinetic, paradigm. The central
question addressed was what characteristics of the movement of the model were
reflected in the movements of the 'imitator'? Basically, the paradigms proposed
were similar in nature, i.e., a control group left entirely to discovery learning and
an experimental group who either had the benefit of a dynamic video model of an
expert performing the ski-simulator task while they themselves were training
(synkinetic) or, in a second experiment, during the intervals between discovery
learning trials (echokinetic). To date, only the first of these experiments has been
completed (Whiting, Bijlard & Den Brinker, 1987).
The model used was a video presentation of the performance of an expert. The
use of a video model in place of a 'live' model was justified on the basis of a
number of studies which have shown video models to be equally effective (Bandura
396 H.T.A. Whiting
4.6. Results
Somewhat surprisingly (in the sense that an adequate explanation was not
obvious) the experimental group - who had the benefit of the presence of a
dynamic video model during training - showed superior overall mean fluency
scores and showed significantly less overall variability in both the fluency and the
frequency of their movements. The absence of any (conditions x days) interaction
effects for any of the variables prevented any further specification of these
findings.
One possible interpretation of these results is that subjects do not 'pick up' the
fluency information per se but, as suggested in the 'image of the act' theory, they
pay attention to establishing an appropriate movement form - in this case the
movement form of the expert model. If then there proved to be a significant
relationship between this particular movement form and fluency this would
account for the said finding. Unfortunately, information about movement form
was not, a t this stage of the research programme, available. The possible
relationship proposed does, however, point to the importance of having this kind
of information, before adequate interpretation can be made.
Further discussion of the findings of this experiment must await the results of
the similar experiment utilizing an echokinetic paradigm and ultimately an
analysis of movement form. It is, moreover, to be noted that a switch has to be
made from talk about imitation of the model to talk of the mediating effect of the
model. It seems clear that subjects do not imitate the spatio-temporal
Imitation and Learning 397
characteristics of the model per se since, even at the end of the five-day training
period, they moved at a lower frequency than the model and at a lesser amplitude.
In this respect, they showed no significant differences from subjects left entirely t o
discovery learning.
Clearly, then, subjects who had the benefit of the availability of a dynamic model
during training did not copy, exactly, the movements of the model as indexed by
the parameters amplitude and frequency (in the absence of the appropriate
analysis it is not possible t o make a similar statement with respect to movement
form). The effect of the availability of a dynamic model would thus appear to have
a more direct regulatory effect in the manner envisaged by Ricoeur (1966,p. 248):
It still remains to be explained why the regulatory effect should be indexed by the
parameter 'fluency'. In this respect, the position of Prim (1985) - presumably
motivated by the Gestalt school of psychology - is attractive. He postulates the exist-
ence of an immediate (direct) similarity-based match between perceived and
generated event configuration. Such matching may occur on-line (synkinetic) or
off-line (echokinetic). He assumes on a priori grounds that two principles underly
the variables which might most lend themselves to observational learning. In the
first place, event generation tends to rely more on pattern in time than pattern in
space. Secondly, with respect to both spatial and temporal patterns, the
parameters of the to-be generated motor event tend to be based on some abstract,
high order representation of the perceived event. If, he proposes, ideo-motor event
generation relies on the use of parameters of these higher order representations
of spatial and temporal pattern, this should become apparent in closer fits
between imitated and imitating movement - in the higher as compared to the
lower-order fimctions.
It is maintained that the findings from the experiment here reported lend some
credence to Prinz's principle of immediate similarity-based matching. This kind
of issue will be pursued further in subsequent experiments within the research
398 H. T.A. Whiting
programme, If the results are confirmed and if they prove not to be artefactual - in
the sense that movement 'form' is in fact 'picked-up' and this, in turn, has a
positive effect on fluency per se - , support would be provided for the direct
steering of movements by means of a dynamic model, i.e., interpretation of
movement control would be shiRed from some kind of internal representation
(motor programme, schema, etc.) to control via an external agency. This would be
in keeping with Ricoeur's (1966)contention that "the model regulates movement
but does not pmduce it" (p. 249,italics in original).
ACKNOWLEDGEMENTS
REFERENCES
\
H.T.A. John) Whiting
Dept. o Psychology,
8
Facul of Movement Sciences
Free niversity, P.0.Box 7161,
1007 MC Amsterdam,
The Netherlands.
Complex Movement Behaviour:The' motor-actioncontroversy,pp. 405417
O.G. Meijer & K Roth (editors)
Q Elsevier Science Publishers B.V. (North-Holland),1988
Chapter 16
Herbert Heuer
SUMMARY
In view of the discrepancy between the simple movements usuully studied in the
laboratory and complex everyday-life motor skills questions arise about
generalization. It is argued that generalizability depends on different conditions
for different kinds of statements. Generalization of experimental results seems to
require a high similarity of the experimental situation to a reference situation,
while generalization of theoretical statements requires invariance of the system,
that is, invariance of functionally defined components and their interrelations.
Whether or not the conditions required for generalization are fulfilled can never
be established a priori, but rather a research strategy is necessary that permits
this to be decided a posteriori.
1. Introduction
experiments and inferences that are based on these data are related to those skills
that are performed outside the laboratory.
Experiments using highly artificial tasks have a long-standing tradition in
psychology. Every now and then concerns about the utility of such experiments
seem to arise. During recent years criticism of the 'small and simple paradigms',
as Sanders (1984)calls them, has become more vigorous in the field of motor
behaviour. This criticism, based on an action perspective or ecological per-
spective, is directed mainly against the non-naturalness of the tasks studied.
Roughly stated, it is argued that the small and simple experimental paradigms
create something like a new world, and that the behaviour people exhibit in that
world has hardly anything to do with behaviour in normal life.
The thrust of ecological criticism is nicely summarized by Von Hofsten (1987,
pp. 44-45):
This essay grew out of several informal discussions about the ecological critique
and the problem of generalization from the laboratory to the world outside. It is
very much personally biased and it is clearly not intended as representing the
state of the art. It is rather meant as some kind of biased integration of several
ideas that I found lying around, which could be useful for a reader in developing
his or her o w n thinking.
The starting point is the notion that it might be useful not to talk about
generalization in general, but rather to draw distinctions according to what
should be generalized, The most obvious distinction is that between results and
theoretical statements. Theoretical statements may be classified further in
various ways. The major point that I want to make is that different kinds of
generalizations require the fulfilment of different conditions, and that in no case
is there a priori reason to believe that generalizations will be valid. Rather, the
range of generalizabilityis an empirical problem.
The Laboratory and the World Outside 407
2. Generalization of results
There will not be much argument about the statement that the specific results of
experiments - such as error rates, reaction times, movement times, etc. - can
hardly ever be generalized, except when the experimental situation is highly
similar to a reference situation outside the laboratory. Generalization of results,
thus, requires an experimental situation that can be considered as a 'simulation'.
The value of simulation-type experiments for certain purposes is beyond doubt
(see Sanders, 1984).
It is important to note that there appears to be no claim that the results obtained
with small and simple paradigms can be generalized in a straightforward
manner. For many of these paradigms such a claim would be ridiculous. As an
example, consider research on motor programming: outside the laboratory we
hardly ever encounter a situation where we have to choose as quickly as possible,
e.g., between key presses of different durations. For these paradigms it is not
generalization of results that is intended but rather generalization of theoretical
statements. Von Hofsten (19871,for example, seems to miss the point when he
argues that accuracy of children in a coincident timing task is greatly
underestimated if a non-natural experimental task is used.
The limitation of simulation-type experiments is well known and has been
stated several times before (e.g., Sanders, 1984): they contribute little to theory
development. As an illustration, consider some examples from the mental
practice literature where experiments are often designed as simulations of some
reference teaching situation. Clark (1960)studied the effects of mental practice on
performance of the Pacific Coast one-hand foul shot in basket-ball. There was a
certain amount of practice on each day, and in the end i t turned out that mental
practice led to an improvement. Jones (1965)showed that a hock-swing upstart at
the bar could be improved by mental practice. Similar experiments were
performed by Phipps and Morehouse (1969).Of the three tasks they studied only
the soccer hitch kick did not profit from mental practice. Finally, Willimczik et al
(1976)showed the benefits of combined physical and mental practice for two ways
of handling the ball in volleyball.
What can be learned from these results? Exactly this: given particular instruction
conditions, there are some benefits from mental practice for the one-hand foul
shot, the hock-swing upstart, etc., but not for the soccer hitch kick. More
generally, each simulation experiment can be generalized to its reference
situation only. Each prediction of how the system studied will behave in a
particular situation requires to examine the system in exactly this situation or in
408 H. Heuer
a highly similar one. Thus, simulation-type experiments are limited with regard
to the scope of situations to which their results can be generalized. Only if a theory
is available can predictions for situations not encountered so far be derived - as
seems to be widely acknowledged.
3. Generalizationof laws
Why is it that la- relations that hold for a variety of different situations are so
rare? The answer is given in outline by Darner (1983). Consider the system that
the behavioural sciences are dealing with. It consists of a large number of
neurons with a set of interconnections that no one can disentangle. (In fact, no
one would attempt to explain complex behaviour on a neuronal level.) Whatever
behaviour is observed is a product of this huge neuronal network. There are so
many inputs to the system and so many interactions that its behaviour is likely to
appear 'chaotic', and there is no hope of disentangling the complex pattern of
interactions completely.
Given this view, it is quite plausible that laws relating only two, three or a small
number of variables are hard to find. Thefe may exist laws that relate 20, 30 or
even a larger number of variables, but these we are unable to identify. But what is
less plausible is that laws of reasonable generality do exist that relate only two or
three variables, even if these laws are not as general as the physical laws that
served as some kind of model in the search for behavioural laws. In the field of
motor skills we have, for example, Fitts's law (see Keele, 1981) or the power law of
practice (see Newel1 & Rosenbloom, 198l), and in the field of perception there are
various well-known laws like Weber's law and Stevens's law, and less well-
known ones like the Few-Porter law (Brown, 1965). Why is it that such relatively
simple laws can be identified, while in other fields of research identification of
laws seems to be more or less impossible?
Consider Stevens's law which states a power function as the relation between
perceived and physical intensity of a stimulus. The power function not only
410 H. Heuer
From what has been said it is clear that there is no general answer to the question
whether laws that can be generalized to real-life situations can be discovered by
way of using small and simple paradigms. Some laws may show up in the study
of artificial tasks, while other relations will prove to be highly task-specific. It
seems that the former is more often the case if 'simple' perceptual or motor tasks
are studied. If, for example, performance is limited by rather peripheral parts of
the nervous system, critical determination is more likely to occur than, e.g., in
the study of complex mental processes (cf. Sanders, 1984), and it is likely to occur
not only in highly artificial experimental situations but also in everyday-life
situations.
The Labomtory and the World Outside 411
A large part of ongoing research is not intended as a search for laws that could be
generalized to the world outside the laboratory. This is particularly so for
research in the information processing tradition. Theoretical statements this
kind of research is aiming at are concerned with the makeup of the system rather
than with law-like relations between sets of variables. In many cases the variables
studied have no counterpart outside the laboratory, and in other cases the
experimental effects observed are so small that they are utterly irrelevant (except
possibly for some sports, if motor behaviour is considered). Thus, the major
interest here is not in relations between variables but rather in the inferences that
can be drawn from them.
There are many ways to describe the nervous system as a 'system', that is, as a
set of interconnected components. As said already, it is useless to take single
neurons as components if the aim is to predict complex behaviour. Still related to
anatomy, nuclei can be taken as components. More common in the behavioural
sciences, however, is the neglect of anatomy and the use of hnctionally defined
components instead. Theoretical statements mainly concern the existence of such
components or the processes that achieve their functions.
A major approach to the identification of hnctional components has been the
additive-factors logic (for a review, see Sanders, 1980). Here we have some basic
assumptions like seriality of components and independence of the output of each
component from the time it takes to produce the output, and within the general
framework where functional components, called stages, can be identified.
A quite different approach to the identification of functional components is used
in perception. We have reason to hypothesize that stimuli are analysed by a set of
'channels'; each of these channels serves to selectively analyse a particular
stimulus characteristic (see, e.g., Regan, 1982). One of the techniques used to
identify such functional components is selective adaptation. Subjects inspect a
particular stimulus for a while, and if a certain channel becomes adapted during
this time, certain perceptual after-effects are to be expected. Interestingly,
channels identified by this procedure often do correspond to classes of neurons
with particular tuning characteristics. A technique similar to selective
adaptation has also been tried in order to identify functional components involved
in the generation of movement (Rosenbaum,1977;Heuer, 1980 & 1981).
A second kind of theoretical statement about the make-up of the system
presupposes the existence of certain functional components and addresses their
characteristics in more detail. For example, one may presuppose the existence of
something like a motor short-term memory and pose questions about the codes
412 H. Heuer
used (e.g., Posner, 1967). Or one may presuppose something like a motor
programming stage and pose the question which processes run off in this
functional component (e.g., Rosenbaum, 1980 & 1983). A basic assumption in
these kinds of experiments is that the experimental paradigm is in fact suited to
study the component of interest. Thus, the relations observed between various
variables are supposed to be critically determined (cf. Marteniuk, 1976, pp. 6-81.
Although the relations themselves may be of little value for generalization - since
critical determination by the component studied may hardly ever occur outside
the laboratory - the componentwill still exist and contribute to complex behaviour.
These examples may suffice to illustrate what I mean by 'statements about the
system'. To what extent is it possible to generalize statements of this kind to
situations outside the laboratory? The first point to be noted is that knowledge
about the functional components of a system is not in itself suflicient for accurate
predictions of the system's behaviour in novel situations. Of course, detailed
knowledge about the make-up of the system should make predictions possible, but
even if there is complete knowledge, prediction is often hard if the system is of a
sufficient complexity, and as far ae the functional components of the brain are
concerned, there will only be limited knowledge in the foreseeable future. Sanders
(1984),for instance, notes that from knowledge about the psychological refractory
period - which means knowledge about a central decision mechanism - it is not
easy to build a predictive model, e.g., about piano playing. This difficulty may
stimulate concerns about whether theoretical statements about the make-up of the
system are usell, even if their generalizationis possible.
Generalization of statements about the system would be possible if the system
should prove to be invariant, that is, if the same system that performs laboratory
tasks also acts outside the laboratory. Although, of course, it is the same brain
that acts inside and outside the laboratory, this does not imply that it is always the
same system. Even if the brain is the same in performing two tasks, it may
represent two different systems as defined by sets of functional components and
their interconnections. (As an analogy, one and the same computer running with
different programmes can also represent different systems.)
Consider the possibility that the brain represents different systems when it
performs different tasks. For example, functionally or anatomically defined
components may be involved in the performance of some tasks but not in the
performance of others. This is a rather trivial possibility: no one doubts that a
functional component like short-term memory is required for some tasks but not
for others. Things, however, can become worse, in particular when functional
components are considered. It appears plausible that the number of anatomically
defined components is limited when reasonably molar components like nuclei are
The Labomtory and the World Outside 413
considered. It is, however, less plausible that the number of functionally defined
components is limited as well.
It is obvious that functionally defined components have some anatomical
substrate. This, in principle, can be any set of interconnected neurons. The
number of possible neural networks appears rather unlimited, and so may be the
number of possible functional components. While i t is unlikely that in the
laboratory anatomically defined components will be studied that are not involved
in any task outside the laboratory, this possibility is less unlikely in the case of
functionally defined components since the potential number of such components
is larger. The introductory quotation from Von Hofsten (1986) in fact suggests that
highly task-specific systems could be set up, and that the use of non-natural tasks
could result in studying systems or components that do not exist outside the
laboratory. If this were the case, generalization of statements about the system
would no longer be justified.
The existence of some rather general laws indicates that task-specificity is
limited. In studying different tasks we are not dealing with completely different
systems. On the other hand, there is clear evidence for some degree of task-
specificity. As noted by Newel1 (1973),the different paradigms that are studied do
not unite; there seems to be a lack of convergence of conclusions that are based on
different kinds of experiments. This lack of convergence could be due to the fact
that different systems are studied. In the area of sensori-motor coordination, for
example, there is clear evidence for a multiplicity of different systems that
subserve different tasks (see for a brief review Heuer, in press).
The conclusion that can be drawn is that there is no a priori reason to
presuppose that statements about the system that are derived fiom data obtained
with small and simple paradigms have any validity outside the experimental
situation. As far as the experimental task is of no interest per se, such statements
are potentially useless. On the other hand, there is also no a priori reason to
presuppose that theoretical statementa based on small and simple paradigms can
not be generalized. Thus, if we have no a priori reason to decide on
generalizability, we need to find a posteriori reasons by employing an appropriate
research strategy.
5. Methodological implications
There seems to be some agreement that the validity of behavioural laws is limited
and that their range of applicability has to be explored. For example, regarding
Fitta's law, a lot of research has been conducted to establish its range of validity.
414 H.Heuer
With respect to statements about the system such efforts appear to be rare.
Regarding serial stage models, for instance, only recently have suggestions been
made to test their validity outside the experimental paradigm on which their
development is based (Gopher & Sanders, 1984;Sanders, 1984).The possibility of
task-specific systems requires such a 'multi-task research strategy'.
Suppose that the human brain could be described as a system consisting of a well-
defined set of functional components. Each task engages a subset of these
components or even all of them. If this were the case, a suitable research strategy
would be to identify tasks that tap these components in isolation (cf. Marteniuk,
1976, pp. 6-8).Research could focus on a relatively small set of experimental
paradigms. The conclusions from these could be united to form coherent theories
that encompass various paradigms and that allow the prediction of complex
behaviour determined by well-defined sets of components.
If there is some degree of task-specificity, the set of functional components may
no longer be well-defined. There may be components that are unique to certain
tasks, and experimental paradigms that happen to tap such components would
suffer from irrelevance. A strong ecological critic would probably hold that most,
if not all, small and simple paradigms do so. Thus, the problem is no longer only
to identify components and to study them in detail, but also to delineate the set of
tasks which employ the functional components under study. This second problem
can be solved by exploring the implications of a particular theoretical statement
across a variety of tasks.
The possibility of task-specific systems, thus, requires that statements about
functional components are supplied with statements about the set of tasks to
which they are applicable. Ideally, the 'applicability statement' would define a
critical task condition that delineates the range of validity.
Theoretical statements that allow predictions for only one task would be useless
for a multi-task research strategy. For such statements it can never be established
that they are valid for more tasks than just the one that allows to test their
predictions.
6. Concluding remarks
The purpose of this essay was to examine some possibilities to generalize from
small and simple experiments to the world outside the ivory tower. The message
that I wanted to convey is that generalization is possible on various levels, but that
on no level is there an a prwri reason for the belief that generalization w ill be
valid. There is also no simple rule stating that generalization is the more justified
the more the experimental situation mimics some reference situation outside the
laboratory, as ecological criticism sometimes appears to suggest; similarity of
situations seems to be important only if generalization of results is intended, but
not for generalizationof theoretical statements.
The methodological implication of possible task-specific systems which I have
here called a 'multi-task research strategy' appears to be rather trivial and not
particularly new. This makes one wonder why such a strategy is not employed
more frequently. Either researchers believe in the generality of their theoretical
statements - even if they are based on the study of a single task only - or there
must be some special obstacles which prevent the adoption of that strategy. Some
obstacles may, indeed, exist.
First, paradigms tend to develop their own dynamics once they are established.
Even if they are established originally as a tool for the study of a general
theoretical problem, each experiment tends to generate new problems. These
problems then inspire follow-up experiments. While the chain of experiments
develops, the original problems tend to get lost. The experimental paradigm
becomes more and mQre a problem in itself and less and less a tool for solving a
problem that is relevant for more than only that paradigm. A multi-task research
strategy emphasizes that priority is on a general theoretical level and not on
understanding the system engaged in a non-natural experimental task.
A second obstacle against adoption of a multi-task research strategy may be a
practical difficulty. Once a researcher has become familiar with a certain
paradigm he or she has obtained a lot of 'unofficial knowledge'. Setting up a new
task generally requires some pilot studies, for example to make the equipment
robust against the unanticipated responses that subjects tend to produce. The
effort spent in setting up a new experimental task has less returns than the effort
spent in running another experiment with a familiar task, and the law of effect is
among those behavioural laws which have reasonable generality.
416 H. Heuer
ACKNOWLEDGEMENTS
I am grateful for the critical and helpful comments of two anonymous reviewers
and the Editors.
REFERENCES
778.
Posner, M.I. (1967). Characteristics of visual and kinesthetic memory codes.
Journal of Experimental Psychology, 75,103-107.
Regan, D. (1982). Visual information channelling in normal and disordered
vision. Psychological Review, 89,407-444.
Richardson, A. (1967). Mental practice: A review and discussion (Part 1).
Research Quarterly, 38, 95-107.
Rosenbaum, D.A. (1977). Selective adaptation of "command neurons'' in the
human motor system. Neuropsychologia, 1481-91.
Rosenbaum, D.A. (1980). Human movement initiation: Specification of arm,
direction, and extent. Journal of Experimental Psychology: General, 109,
444-474.
Rosenbaum, D.A. (1983). The movement precuing technique: Assumptions,
applications, and extensions. In R.A. Magill (Ed.), Memory and Control of
Action (pp. 231-274).Amsterdam: North-Holland.
Sanders, A.F. (1980). Stage analysis of reaction processes. In G.E. Stelmach & J.
Requin (Eds.), Tutorials in Motor Behavior (pp. 331-354). Amsterdam:
North-Holland.
Sanders, A.F. (1984). Ten symposia on attention and performance: Some issues
and trends. In H. Bouma & D.G. Bouwhuis (Eds.), Attention and
Performance X Control of language pmesses (pp. 3-13). Hillsdale, N.J.:
Erlbaum.
Von Reandt, K. & Moustgaard, I.K. (1977). The Perceptual World. London:
Academic Press.
Von Hofsten, C. (1987). Catching. In H. Heuer & A.F. Sanders (Eds.), Perspectives
on Perception and Action (pp. 33-42).Hillsdale, N.J.: Erlbaum.
Willimczik, R,Boltz, M., Fdhlich, H. & Rother, R. (1976).Zur Effektivitat des
mentalen Trainings im Schulsport. In M. Letzelter & N. Miiller (Eds.),
Sport und Sportwissenschaft (pp. 180-196).Berlin: Bartels.
.
Dept. of Psycholo y,
9
University of Biele eld
P.O.Box 8640,4600 Bielefeld,
Federal Republic of Germany.
Complex Movement Behaviour:The' motor-action controversy, pp. 419-437
O.G. Meijer & K Roth (editors)
Q Elsevier Science Publishers B.V. (North-Holland),1988
Chapter 17
David S. Young
SUMMARY
The control of action can be understood at different levels, where 'level' is meant
in the sense introduced by Marr. It is argued that it is important to maintain the
logical distinctions between the levels, and that action theories must be seen as
descriptions of information transformations at Marr's computational level. The
control of balance during running is taken as an illustrative example.
1. Introduction
In order to control their actions, people and animals require information about
spatio-temporal relationships between themselves and surfaces in their
environments; their actions produce changes in those same relationships. A goal
of action research must therefore be to produce good descriptions of the physical
relationships which drive and are driven by an animal's perceptuo-motor system.
Only on the basis of such descriptions will theories of motor control be able to go on
to specify the processes that link the information driving an action to the effects of
that action. This paper is intended to illustrate a methodology for achieving this
goal of action research, which I shall approach through my view of the motor-
action divide, and its possible resolution.
as a black box, any model drawn to represent its operation must be regarded as
describing an information transformation, and not as describing the nervous
system's actual mode of operation. Some cognitive approaches fail to make this
distinction explicit, and tend to regard such models as operational analogues
rather than as functional descriptions. It is this which lays these approaches
open to charges of unjustified reification (e.g., Turvey & Carello, 1981;Turvey et
al, 1981;Reed, 1981 & 1984). However plausible it may be to propose particular
memory or control structures to explain the resulta of action experiments, such
experiments alone cannot provide evidence for the reality of such structures. An
example to demonstrate this point will be provided shortly, but it has been made
before (e.g., by Anderson, 1978,but see Pylyshyn, 1980).
So, under what circumstances can we talk about the internal mechanisms of
the nervous system? To do so requires information or assumptions about the
properties of neural information processing1 which go beyond the observation of
behaviour. A significant problem is that such assumptions tend to enter
discussion of motor skills through the implicit adoption of an analogy, such as
feedback control from engineering or model-based programming from computer
science, instead of being precisely identified. That a particular model can predict
the behaviour of an animal, however detailed and specific that prediction is, does
not alone mean that the internal mechanisms of the animal correspond to the
mechanism of the model, but rather that they both lie in the class of mechanisms
capable of realizing the particular information transformation that has been
observed. For all but trivial tasks this class is very large.
The set of contending mechanisms may only be narrowed by using information
concerning the limitations or structure of the physical substrate of the
information processing system. Studies based on response times, for instance,
only provide information about mechanisms if there is some underlying
assumption that operations of comparable complexity require comparable
amounts of time (for discussion of this, see Posner, 1976, and blyshyn, 1980).
Reasonable as that aesumption may be, it ie nonetheless dependent on some
model for information processing. The model is rarely stated as precisely as it
should be for a proper evaluation of the hypothetical mechanisms. Similarly,
studies claiming to indicate modes of information storage (for example, whether a
lever movement is stored as an absolute endpoint or as a relative adjustment)
adopt the seductive assumption that if action outcomes can be mapped in a
I UBB the term 'information processing' in ite general sense, without any restrictive con-
notationu
Describing the Information for Action 421
transformation. The problem is that this looks like a model, and so can be
presented as if it were a model of the system rather than a way of specifying the
information transformation. The blurring of this distinction contributes to the
motor-action controversy. It seems better, therefore, to attempt to specify
information transformations in more abstract terms.
To illustrate these ideas, this Chapter will first elaborate the point that even in a
very simple case, many internal mechanisms can correspond to a particular
computational strategy for solving a visuo-motor problem, then will consider how
the results of particular experiments can be described in terms of the information
used by a subject.
Figure lc. All this would amount to a complete description of the information
transformation instantiated by the machine. It says nothing about how the
machine achieves that transformation. As Figure I d shows, it is simple to invent
another mechanism that is equivalent to the first but is fundamentally different in
its operation at the algorithmic and implementational levels. The example is in
fact rather oversimplified, but this does not affect the main point that the
mapping from mechanisms to information transformations is many to one.
A more radical example would be to replace the rod by another mechanism
which sensed how the trolley was moving, and which used a motor to drive a light
pointer to simulate the rods movement. Such a mechanism might use a
microprocessor to process information about trolley acceleration to yield the
correct angle for the pointer. The rod itself is, of course, the simplest way to do
this particular information transformation, and in this sense the rod can be seen
as a simple information processor.
What sort of information about the machine would enable us to suggest which
sort of algorithm was used? Two kinds of knowledge could be relevant. The most
obvious is specific knowledge about the inside of the device. If it could be
dismantled and measurements made of the signals at different points in it, then
in a case as simple as this it would not be difficult to establish how it worked.
Alternatively, if it was known to use, say, electronic components from the
laboratory stores, then the general properties and limitations of these components
would provide strong indications of the class of algorithms that might have been
instantiated using them: for instance, the quality of the capacitors would provide
limits on the time constants available to the system, the signal-to-noise ratio
would limit the degree of resolution of signals and so forth. Together with the
assumption that the device was built in a reasonably economic manner, this
implementation-level information would enable one to distinguish less likely from
more likely algorithmic descriptions.
To sum up, the observations of the device as a black box, interpreted properly,
allow it to be described a t the computational level. This description defines the
class of algorithms that could be implemented by the machine2. Knowledge at the
implementational level, either specific or general, can reduce the size of the class
of admissible algorithms. A description of the information transformation
effected by the machine is complete at the computational level; I do not regard it
as 'globally' complete, but, with Marr, suggest that understanding the 111 story
requires descriptions at the algorithmicand implementational levels too.
Here 'algorithm' is used in a very general sense to mean any defmite method for producing
some behaviour.
424 D.S. Young
la k d
/
Trolley,
control electronics
and drive
Can the control of pitch and the control of roll be investigated separately? The
answer depends on whether or not the information needed for one is independent
of the information needed for the other, and whether the control variables for one
are independent of the control variables for the other. Raibert's machine shows
that a strategy based on this independence can work. His free-hopping machine
has control structures for pitch and roll, each of which is essentially similar to
the one used for a constrained hopping machine which could only fall in a plane.
Intuitively, it is plausible that the same would apply t o human locomotion, in that
changing speed and changing course seem independent goals, and there is no
obvious biomechanical reason why the two should interact. Ultimately the
question must be answered empirically, but for the time being it will be assumed
that regarding the two degrees of freedom as separable is a valid procedure.
In the example of rod-balancing, the angle of the rod to the vertical, and the rate
of change in this angle were the only variables needed for control. In running
balance the line joining the centre of support during stance to the centre of mass
plays the role of the axis of the rod, and there are two angles that describe its
orientation. These are the angles made with the vertical by the projection of the
line into the frontal plane (the roll angle) and by its projection into the sagittal
plane (the pitch angle). These angles and their rates of change specify part of the
state of the body at any moment in time. Many other variables are needed to
specify the full dynamic state of the body, which includes information about the
position and motion of all the body segments. However, the pitch and roll angles
play a key role, because together with the height of the centre of mass they
determine the net gravitational torque, and this alone is what produces changes
3 It should be pointed out here that changing course means changing the direction of travel, and
that this is not the same thing a s changing the direction in which one is facing, which is a problem
not considered here. Bicycles automatically face the direction of travel; people have little
difficulty in doing so;one legged balancingmachines may find it rather hard.
428 D.S.Young
in the runner's total angular momentum. Since changes in the height of the
centre of mass are generally quite small compared with the mean height,
whereas changes in the two angles are much larger than their means (the latter
must be close to zero), changes in the angles are a good deal more important than
changes in the height of the centre of mass, Thus, this simple analysis of the
mechanics indicates that picking up information related to the pitch and roll
angles and their rates of change will be an important part of balancing. In
addition, the system will need much other information: both propriospecific
information about the disposition of the body segments relative to one another, an
nd expropriospecific information about, for example, the timing of the next heel
strike. For the time being, though, I will develop the question of what information
about the pitch and roll angles and their time derivatives is picked up by a runner.
The pitch and roll angles were defined above relative to the vertical. That is, they
were defined in a frame of reference fixed with respect to the direction of gravity.
This will be called the mechanical frame of reference. Alternatively, the
orientation of the centre of mass / centre of support line could be measured
relative to the line where two walls of the laboratory meet. This would amount to
the same thing if, as in a normal room, this intersection was always vertical.
However, in the experiments to be described, the walls of the room surrounding
the subject were moved, and so orientations measured with respect to them were
not the same as orientations in the mechanical frame of reference. The walls of
the room had no mechanical link to the runner, and so their only effects were
through the optic array they establish (Gibson, 1966;Lee, 1980). The frame of
reference fixed relative to the walls of a rigid room will therefore be called the
optical frame of reference. Under normal circumstances, the mechanical and
optical frames are fixed relative to one another, and so orientation information
specified in either is equally useful for balance control. If the room moves,
information in the optical frame of reference will no longer specify the pitch and
roll angles, or their rates of change. By moving the room and so decoupling the
frames, the information used for balance can be identifiedd.
Information about body orientation is available in both frames of reference. For
instance, the torques at a runner's ankles, accessible t o mechanoreceptors,
depend on orientation in the mechanical frame, and the vestibular system can
pick up information about head orientation in the same frame. On the other hand,
the visual system picks up information in the optical frame. Finding out what
Selspot Camera
Treadmill
Method
The arrangement was as shown in Figure 25. Each subject ran on a powered
treadmill at about 2.6 d s e c , inside an open-bottomed box which allowed them
ample space. This 'tilting room' was supported on pivots whose axis passed
roughly through the subject's ankle during support. For one set of experiments
the axis lay in the sagittal plane and so tilting the room produced the optical
change that would normally follow from a change in the subject's roll angle; for
another set the axis lay in the frontal plane and so the optical effect was like a
change in pitch. For the roll experiments the movements of a number of
landmarks on the subject's body were recorded from behind using a Selspot opto-
electronic measuring system; for the pitch experiments the action was recorded
from the sides. Selspot also recorded the room movement.
ARer the subject had been running for a while, movement recordings of 10 sec
were made and, 7 sec into each recording, the room was abruptly tilted through
about 3 degrees in one of the two possible directions (left or right in one set of
experiments, forward or backward in the other), chosen a t random. The subject
knew that the room could tilt, and knew whether the movement would be pitching
or rolling, but he had no warning of the moment at which each lurch of the room
was started, nor in which direction it would be. For safety, the subject always
wore a safety harness which would have prevented injury in the event of his
falling.
sutficient to tell whether the room had been rolled to the right or to the left in the
roll experiments. The subject's body orientation was displayed by playing back the
movement recordings onto a computer display screen, with only the line joining
A
------- c--- -------
Right
4 sec
I rWd
Forward
the coccyx and nape markers shown. A fresh subject watched ten playbacks from
each of two runners and in every case correctly identified whether the room had
rolled left or right.
Thus, one question can be answered immediately: information specified to the
runner in the optical frame of reference was involved in maintaining balance.
The next question to ask is whether this was all the information used, or whether
other frames of reference were involved and if so, how the information was
combined.
To show more exactly how the runners responded, Figure 3 shows example plots
of the roll and pitch angles for the nape-coccyx line, along with the roll and pitch
of the room. These plots have had the cyclical component of the motion removed so
that the transient response to the room movement may be more clearly seen7.
As Figure 3 shows, when the room was rolled, the runner responded by initially
routing with it: the body tilt tracked the room tilt. Subsequently, when the room
stopped moving, the runner recovered, and the body returned to the upright. On
the other hand, when the room pitched, the runner briefly tilted back before
recovering, regardless of whether the room had pitched forwards or backwards.
7 The cyclical component was obtained as follows. The pitch or roll angle curve for a period of
unperturbed running on the treadmill was divided up inta strides, and an average curve for a
stride obtained by summing corresponding point8 in each of these separate stride curves. Now the
curve that the pitch or roll angle would have followed, had the room not moved, was estimated by
repeating this average curve over and over. This cyclical curve was then subtracted from the
observed curve in order to give the plots shown.
8 An example of a quantitative analysis applied to a very closely related task is given by Hubbard
(1980).
Describing the InformationforAction 433
the mechanical frame was used together with the rate of change of roll angle in
the optical frame.
That this scheme produces the observed behaviour can be seen by considering its
predictive properties. Suppose it had in fact been used. When the room was
stationary, either before or after the lurch, the rate of change of body roll obtained
from the optical frame was the same as the rate of change of roll in the
mechanical frame. Thus combining rate of change of angle with angle predicted
the future motion in the mechanical frame. When fed into a corrective control
system, necessarily also operating in the mechanical frame, this prediction led to
correction of any tilt that developed, and hence to maintenance of posture before
the room movement and a return to vertical from the tilted position after the room
movement ceased. While the room was moving, however, the prediction was
changed. Suppose the room lurched to the left and at the start of the lurch the
runner was not rolling in the mechanical frame. Then his rate of roll in the
optical frame became positive t o the right. This predicted a future inclination to
the right: the balance control mechanism inevitably compensated by generating a
leftward torque. Despite the fact that in the mechanical frame of reference this
immediately produced an erroneous leftward tilt, the room motion generated a
prediction, through the optical information, that this tilt was being corrected.
Thus, while the room tilted to the left, the runner tilted with it, producing the sort
of initial tracking behaviour shown. As the room stopped moving, the optical
frame became positive to the left, and hence, combined with the mechanical
frame information, generated a prediction of hrther leftward tilt. From this
prediction followed a corrective torque producing a smooth return to the
mechanical vertical.
This explanation cannot be applied directly to the data for pitching movements, as
in pitching the runner did not track the room when it moved forwards. The
reason for this may be that the backward pitch produced by the runner was a
general purpose emergency response designed to enhance stability by preparing
for a reduction in speed. In any case, the results are still entirely consistent with
the picture that the optical rate of pitch was combined with the mechanical pitch
angle to yield a prediction; the difference from the roll case is how this prediction
was used. It may well be that as roll control is intimately connected with control
over running direction, perturbations to the roll angle occur as a matter of course
in normal running, and so corrections to it can be made in a smooth
compensatory way. On the other hand, the pitch angle may normally only suffer
substantial perturbations when footing is disturbed, for instance by tripping over
434 D.S. Young
9 This explanation for the difference between the pitch and roll responses is supported by the
observation, not shown here, that the timing of footfalls following the lurch was rather more
disruptedby room pitch than by room roll.
Describing the Informationfor Action 435
In the ball-punching study, for instance, the same behaviour would be generated
by using the ball's current distance and speed to predict its future trajectory. It is
not necessary, in fact, to know either the distance o r the speed explicitly: ttcv is
specified directly by the rate of expansion of the ball's image on the retina. The
point I wish to make here is that although the experiment strongly supported the
idea that ttcv prediction was used in controlling the leap and punch, there was no
evidence whatever from the experiment itself that the rate of expansion of the
retinal image was involved. To show that ttcv was the relevant control variable
was to make an important statement about the nature of the spatio-temporal
information used in controlling the action. To infer that ttcv is determined
through the rate of expansion of the ball's retinal image, or through any other
optic variable, requires more information (or more assumptions).
Further experiment could, of course, progressively narrow down the possible
sources of ttcv information without having to consider the algorithmic or
implementational levels. Such an experiment would manipulate the information
available, for instance by removing the view of the ball for part of its approach, or
by removing any texture fiom its background, and so forth. A tighter and tighter
description of the information used by the puncher would ensue: a better and
better understanding of the act at the computational level would follow. Similarly,
tighter and tighter descriptions of the action variables at the output of the system
could be developed, although this is an aspect to which I have given little attention
in this paper.
ACKNOWLEDGEMENTS
The tilting room experiments were carried out at the Department of Psychology,
University of Edinburgh, under a grant from the Medical Research Council. I
thank David Lee for introdacing me to the study of movement control and for
making the experiments possible.
436 D.S. Young
REFERENCES
iDavid S. Young
Cognitive Studies Programme,
Universi of Sussex
Brighton N1 9QN,
United Kingdom.
Complex MovementBehaviour:The' motor-actioncontroversy,pp. 439-461
O.G.Meijer & K Roth (editors)
0 Elsevier Science publishers B.V. (North-Holland),1988
Chapter 18
SUMMARY
1.Introduction
Having made this assertion, some questions naturally arise. On what grounds is
it both meaningful and fruitful to distinguish a 'motor approach from an 'action
' For example, in Best (1978), who speaks, respectively, in terms of '(physical) movement' and
'(intentional)action'.
Images of the Body 441
approach', and what about the tenability of these grounds? Are the two
approaches mutually exclusive? When may a particular approach be classified as
belonging to one or the other cluster, particularly bearing in mind the
multivarious jargon peculiar to this field? Is the mere use of the term 'action' - for
example in a denotation like "muscular action" ( Higgins, 1977, p. 9) - already a
sufficient indication for assigning a certain view to the cluster of 'action
approaches'? And, when should a 'strategy for investigating action' (Kelso, 1982)
be called an 'action' or a 'motor-strategy'?
Answers to questions like these are not exclusively of an empirical-scientific
kind. Whoever tries to answer such questions will quickly encounter a complexity
of epistemological, ontological and anthropological presuppositions in the light of
which different authors take d a e r e n t positions. Sometimes authors show an
explicit awareness of such presuppositions and discuss them accordingly; more
frequently, however, they tend to remain implicit and it becomes dimcult to
uncover them.
The comparison of the human body with a gaol was explicitly made by Plato (see,
e.g., Van Peursen, 1978; Schroten, 1970). In his philosophy the soul was
incarcerated in the gaol (the body), served its time there and could observe the
outer world through the bars (the sense-organs). Plato advocated a dualistic
image of man, in which the soul represented the 'higher' and the body the 'lower'
part. A dualism, therefore, with a negative view of the body. In later times 'gaol'
was replaced by various other images, often with less pronounced devaluative
overtones. Nevertheless, an unmistakable continuous thread runs through Greco-
Western thought in which this negative view of the human body is to be discerned.
In that line of thinking, the human body has, predominantly, in metaphorical
terms been 'silenced' (see Kamper & Rittner, 1976; Kamper & Wulf, 1982;
Tamboer, 1985). Western tradition "has vigorously fought to suppress the life and
truth of the body," states Levin (1985, p. 4).
To date, there has been a revaluation of the body. Being naked - at least in
certain contexts - as well as sexuality are no longer self-evidently associated with
While the steam-engine, for example, was a favourite image in the last century - an image,
which profoundly influenced the metaphorical language of Freudian theory (see Russelman,
19831, such popularity, in modem time, has to be assigned to the computer.
Images of the Body 443
Although less pronounced than in case of a 'gaol', in the alignment of the human
body with an instrument one also makes use of an image in which the being
'silenced', just mentioned, can be recognized (see Klein, 1984; Rumpf, 1983 &
1984). Because of its influential role in many theories about human movement-
behaviour it will be treated at some length.
Human beings often act on their environment in an in-direct way. While a
swimmer is in direct, 'bodily', contact with the water, a rower acts on it
indirectly. His oar can be considered as an extension of his body. Bicycles, cars,
aeroplanes, typewriters and computers are other examples of extensions of the
human body. In using these expedients we place them 'between' our body and the
environment. If we are not yet able t o handle a certain instrument well, we are
certainly aware of that 'between'. The starting hockey-player experiences his stick
as an unmanageable 'barrier' between himself and the ball, not as an extension of
his acting potentiality. In the case of a skilled player that stick has become an
extension to his body to such an extent that a 'between' is no longer experienced.
He literally has 'in-corporated his stick. Similarly, a skilled motorist does not
experience his car as an unmanageable 'inter-mediate' between himself and the
road. And, in the case of a pair of spectacles, a denture, or a pace-maker, their in-
corporation into the body can be so complete, that they are almost considered as
'own' body-parts.
The in-direct way of acting on the environment has inspired many thinkers to
use it - implicitly or explicitly - as a paradigm for coming to an understanding of
the human body itself. The human body can also be understood as an
'intermediate', as a complex 'instrument'. Well-known metaphors in this context
are 'clockwork, 'steam-engine', 'motor-car', 'computer', etc. (see, e.g., Vroon &
Draaisma, 1985). However, from a philosophical-anthropological point of view
such an instrumental conception of the human body is rather problematic. "An
instrument," suggests Zaner (1971,p. 31):
444 J. W.I.Tamboer
The images 'gaol' and 'instrument' are surely not the only possible
interpretations of the human body. That, however, does not alter the fact that
-
these two images, including their various variants, have been and still are - very
influential. Confining ourselves to these two images, then, it is striking that in
both cases 'things' are concerned .to which one can attribute a certain
independence. They can be 'isoZated', and consequently we use 'substantives' in
order to denote them and discriminate them from other 'things'. This reminds
one of what philosophers have traditionally termed 'substance'. That term, then,
stands for: that, which is by itaelf, which can be isolated, which can be understood
as an independent and isolated entity. It is characteristic of a substantial view
that one is able, essentially, to define the 'boundary' between that which does and
Images of the Body 445
that which does not belong to that entity. That boundary, then, divides the 'inside'
('inner') from the 'outside' ('outer').
Attempts to relate the images 'gaol' and 'instrument' to the human body,
generally appear to be understood in such a way as to interpret that body in a
substantial manner. Such images, to put it briefly, are variants of the image of the
substantial body 3. With respect to the image of the substantial body, a n - in Greco-
Western thought predominating - interpretation is invoked in which the body is
understood as an isolated entity, 'the skin and what is 'enveloped by it' (see
Bernard, 19801, which maintains an extrinsic relationship with what is
habitually denoted as a 'psychical andor physical inner-world and a 'physical
and socio-cultural outer-world. A scheme of thinking in terms of 'inner and
outer world - 'internal variables' and 'external variables' - is, as such, directly
related to this image of the body. As 'gaol', 'instrument','motor-apparatus'(see
Rumpf, 1983) or 'medium of expression' (see Benthall & Polhemus, 19751, the body
constitutes an 'inter-mediate' between a stimulating, commanding, triggering or
disturbing inner andor outer-world. The image of the substantial body can be
connected with both a dualistic and a materialistic-monistic image of man.
Likewise it can be understood in a (more) holistic or (more) atomic sense
(Tamboer, 1985). In the course of time, the traditional way of thinking in terms of
isolatable substances has undoubtedly become less rigorous. The old notion of
'substance' has certainly been transformed into a more dynamic and functional
concept. Yet, this by no means alters the fact that a t the present time many of its
traces can still be very well identified, particularly in theories about human
movement-behaviour.
3 In a general w n w 'image of the body' is here understood as a time and place-dependent, more or
less coherent and definable interpretation - which is considered to have a general validity - of
what the human body is and should be (Tamboar, 1985).
446 J.W.I. Tamboer
Dijksterhuis, 1975;Kuhn, 1979). Gibson (1979,p. 93) very properly observed: "The
laws of motion for bodies in space as formulated by Isaac Newton apply only to
idealized detached objects." The image of the substantial body, as just
characterized, lends itself, pre-eminently, to an 'application' of this Newtonian
conception of movement. In that case human movement is appreciated as
'displacement of (parts OD the substantial body'.
3. The 'actionapproach'
The so-called 'action approach is, in reality, also a cluster of approaches, within
which the diversity iseven much more striking than in the case of the 'motor
approach. Contemporary literature on 'action' is not only abundant but also
Images of the Body 449
multiform, as is revealed by the publications of, for example, Van den Beld, 1982;
Brenner et al, 1980;Boesch, 1980;Von Cranach et al, 1982;Hoche, 1973;Hornsby,
1980;Reed, 1982a & 1988;Thomas, 1974;Turvey, 1977;Turvey and Kugler, 1984;
Weinberg, 1978.
With respect to such multiformity, Van den Beld (1982)has shown that, in the
tradition of analytic philosophy alone, at least six theories exist relevant to the
distinction between 'action' and 'mere bodily movement'. This question about
discrimination between an 'action' and a 'mere bodily movement' would appear to
play an important role in most of the literature on 'action'. Van den Beld (1982)
discusses six possible answers to that question:
*the actor-theory; according to this theory the difference between an 'action' and a
'mere bodily movement' is such that the former, in contra-distinction to the latter,
has the actor as its cause. "Someone, who performs an action, e.g., raises his
arm, brings about the result of the action, i.e., the raising of his arm, by himself"
(p. 20).
*the uolition theory: in this case an 'action' is considered "a bodily movement,
resulting from human volition; and, as such, a voluntary bodily movement" (p.
23). (Within the context of this theory two versions can be discerned: volitions as
'mental actions' or as 'mental events'.)
*the contextual theory: unlike the former two theories, the characteristic feature
of an 'action' is not now to be found "in a particular cause or origin, being located
in the actor, so that on him a special authority has to be conferred, when the issue
is whether he did or did not perform an action. No, in answering that question an
observer generally is no less competent than the actor himself. For a bodily
movement only becomes an action in a context of public rules which constitute a
practice" (p. 30).
athe identity-theory: according to this theory a 'basic-action' - about the notion of a
'basic-action' as such, see Van den Beld, p. 17 - is "ontologically identical with a
series of events, starting with an event in the brain and ending with the
movement of ... the body" (p. 34).
*the ascriptiue theory: according to this theory - in Van den Beld's opinion a
variant of the contextual theory - the notion of an 'action' is not a descriptive but
an ascriptive one, "that is, if we confer on a bodily movement the status of an
action, we do not add a new descriptive characteristic to the bodily movement,
e.g., caused by a volition, but we ascribe, with some restriction, responsibility to
the subject of the bodily movement" (p. 35).
*the control-theory: the crux of this theory is that "the difference between a mere
bodily movement and a basic action" lies in the contention, "that in the latter case
450 J.W.I. Tamboer
It is the merit of an author like Merleau-Ponty (1945& 1964)to have shown that
in Western thought - particularly from the Renaissance onwards - the
'simultaneity' of man's grandeur and mishre has been predominantly interpreted
in terms of contra-distinctions. 'Freedom and determinism', 'subject and object',
'inner and outer world, 'sensory and rational', 'body and mind, are all contra-
distinctions which he shows to be closely related to one another and to which he is
antagonistic. Distinctions such as these are not, as was explicitly discerned by
Merleau-Ponty, independent of a particular conception of the human body. "The
human body ... has suffered centuries of systematic misrepresentation" (Levin,
1985,p. 51, a mis-representation which was characterized by Merleau- Ponty as
'body-object'(see also h e r , 1971).
For the sake of an adequate articulation of the 'oneness of human existence'
(Van Olst, 1981)it is necessary that the human body be revalued in such a way
that its suppressed potentialities be reaflirmed. An important impulse to such a
revaluation is to be found in Merleau-Ponty's 'basic discovery' (Kwant, 1968)of the
'body-subject', with its emphasis on man's primordial being-in-the-world, as pre-
eminently experienced in pre-reflective ways of 'knowing-the-world-in-action',
even though he was not yet f d y able to break away from the traditional subject-
object dichotomy. Nevertheless, in his Le Visible et l'lnvisible (1964)he contrives,
by means of his basic notion 'chair du monde', to give an approximate indication
of the primordial connection between 'man and world in general
('connaturalit47 and between 'man and fellow man' in particular
('intercorporQt.4'). In his connaturalit4 and intercorpor4iM man discovers that
other things and animate beings are fbdamentally akin to him. Man,
essentially, is embodied; he is "a mode of bodily being-in-the-world (Levin, 1985,
p. 44).
In the light of a fundamental division between the 'subjective' and the 'objective'
point of view, existential-phenomologicaloriented authors such as Van den Berg
(19501,Grupe (1969),Gordijn et a1 (1975;see also Tamboer, 1979;Pluge, 1967),
following in the footsteps of Marcel, Sartre and Merleau-Ponty - have added to the
protest against the traditional 'silencing' of the human body. In that sense, the
significance of their plea for the primordial role of 'the body as subject' can hardly
be overestimated. However, the distinctions made by these authors - 'to have one's
body' I 'to be one's body', 'the lived body' I 'the experienced body', 'the body as
subject' / 'the body as object' - hardly lend themselves to an adequate classification
of 'images of the body'. Such polarities are, in general, so conceived that they both
support and affirm the traditional subjectobject dichotomy since neither the
452 J. W.I. Tamboer
'subjective' nor the 'objective' point of view can clearly and consistently be
maintained (Tamboer, 1985). A typology of images of the body indifferent to these
points of view is needed.
Within the framework of the image of the relational body, 'actions' have to be
understood aa ways of 'knowing-the-world-in-action'.Actions provide evidence of
an intentional discernment and realization of particular 'relations of meaning'4.
This conception of 'actions' is quite similar to characterizations which can also be
found in other publications. Best (1974, p. 193), for example, writes: "...the
character of any intentional action can be recognized only by considering its
relational significance" and Reed (1982a) states, among other things: "Whereas
meanings are imposed on motor systems, they are intrinsic to action systems" (p.
112), ... actions are realizations of afTordances by agents" (p. 1241, and "because
"
affordances are potential relations between an animal and its environment, they
cannot be made actual unless an animal alters its relations with the
environment" (p. 124). In Reeds opinion "movements and postures form the
components of action" (p. 118), components which are considered "units of
relations" (p. 117) themselves. Beek (19851, however, has demonstrated that Reed
is not always consistent in employing his "new definition of movement and
posture" (Reed, 1985). It appears that this inconsistency can, for the greater part,
be explained by the absence of due reflection on the peculiarity of the human body,
which leaves him committed to the old-established image of the substantial body.
As a matter of fact, this also holds for Best, as was shown by Tamboer (1985). This
in itself, however, does not alter the fact that - notably in the 'ecological theory of
action' (e.g., Turvey, 1977; Turvey et al, 1981; Turvey & Kugler, 1984; Reed, 1982a
& 1988; Leist, 1983) important impulses to a theory of action have been given,
which manage quite nicely with an image of the relational body.
In what follows an attempt will be made at characterizing the peculiarity of
'motor-actions' ('Bewegungshandlungen'), in the light of an image of the
relational body.
3.2. 'Motor-actions'
'cycling', 'swimming', 'playing soccer', 'dancing', etc. Similar verbs are in daily
life operative as 'natural interpretations' (Tamboer, 1985) of human movement.
In everyday life such verbs are 'self-evidently' recognized as paradigms of human
movement. The doctor's advice surely did not contain a n incitement - and also
will not have been apprehended in that way - to watch more TV, participate in
more discussions, read more novels or solve more mathematical problems than
previously. In everyday life, the term 'movement' is directly associated with
activities like walking, cycling, swimming, etc., and not with the last-mentioned
ones. Even though it is, surely, possible to perceive some 'movement' in the case of
someone sitting in a chair and watching TV,such a perception takes place in the
light of a conception of movement which is different from that underlying the
hypothetical doctor's advice. (See, for examples of - often - unrecognized
confusions in the use of the term 'movement', Best, 1978.) It could also be
maintained &at verbs such as 'watching TV','reading', 'speaking', etc., belong
to a different group from those, experienced and recognized as 'natural
interpretations' of human movement. What, however, is the defining peculiarity
of this latter group?
In attempting to answer this question, within the framework of the conception
'displacement of (parts 00 the substantial body', the inclination will present itself
to remark that in the case of, for example, 'cycling', different parts of the body are
active, different effects are aimed at, and different causes play a role from those
which are involved in the case of, for example, 'speaking'. Answers of this kind
can hardly be contested. Yet, as soon as the statements 'different parts of the
body', 'different effects' and 'different causes', used in this example, are used in
an attempt to formulate an answer, it is clear that the proper question is virtually
disregarded. This can be briefly illustrated with respect to the verb 'speaking'. In
the light of the answer, just given, 'speaking' can be operationalized in terms of
the activity of certain parts of the body. That activity is 'caused in a certain way
and brings about certain 'effects'. As to 'activity of parts of the body', processes in
and around, in particular, the pharynx come to mind. Nobody would deny that
such processes are, for speaking, a conditio sine qua non. Even so, good reasons
can be produced by virtue of which it can be maintained that analyses at that level
cannot bring about an adequate comprehension of the multiform nature of
human speaking. Speaking can be understood, for example, as 'asserting
something', 'giving an order', 'asking something', 'promising something',
'whooping with delight', 'giving a reprimand, etc. If attention were to be focused
on the peculiarity of such 'speech acts', or '-actions' (see, e.g., Van Eemeren &
Koning, 19Sl), a level of analysis is chosen, in respect of which the previously
mentioned 'activity of parts of the body' does, indeed, constitute a necessary
Images of the Body 455
condition. From that level, however, such acts surely cannot be so apprehended.
Similarly, the question can be posed as to whether the very essentials of, for
example, 'cycling' can be identified if, first of all, that activity is construed in
terms of displacements of (parts of) the substantial body. That question has,
again, to be answered in the negative. "The specificity of the action lies in the
meaningful changes wrought, not in the pattern of the movements made" (Reed,
1984, p. 163), and this applies as much to 'speech-actions' as to 'motor-actions'
(the last group containing 'cycling' as a paradigm case). If the guiding principle
be 'displacement of (parts 00 the substantial body' it is certain that 'speaking'
differs considerably from 'cycling'. In that case, such differences are conditioned
by operationalizations of 'speaking' and 'cycling', which are very restrictive in
nature. For the sake of understanding the essential peculiarity of such activities
an 'action approach is necessary. In view of the tenability of such an approach it
has, itself, to be rooted in an image of the relational body.
Since the specificity of an action lies in 'the meaningful changes wrought' (Reed),
the question should be posed as to the 'meaningful changes' which are intended
in motor-actions. Within the framework of the image of the relational body a
motor-action has to be understood as a specifl way of 'knowing-the-world-in-
action'. Actions provide evidence of a discernment and realization of particular
'realizations of meaning'. The relevant question can, then, be reformulated as: in
what respect is a motor-action 'specific' and what 'relations of meaning' must be
held in mind in the case of such an action? In developing the answer to this
question it has to be emphasized that the category 'change of position' is, in that
context, indispensable. However, that category surely needs t o be re-interpreted,
expressed terminologically by substituting the verb 'to displace' ('to change its
position') for the substantive 'displacement' ('change of position'). In the light of
the image of the relational body the term 'to displace' has to be understood in such
a way, that something is said about the typicality of the network of 'relations of
meaning', which becomes manifest in motor-actions.
A motor-action is to be perceived as an actualization of the concordance between
human intentionality and the intended 'world while that concordance is
primarily realized in a displacing way. This can be illustrated by means of the
following example borrowed from Nijk (1984,p. 90). Nijk writes: "Cycling can be
an action, but has not necessarily to be so. A coach who, cycling along the water-
side, follows and instructs his rowers, undoubtedly performs an action, but his
acting surely will not for choice be defined as cycling." Why not? Within the
particular context the 'acting' of the cycling man should primarily be
characterized as 'coaching' and not as 'cycling' (or 'balancing', or 'being
456 J.W.I. Tambwr
discontented, etc.). In his coaching he - as contrasted with his rowers (!) - gives
evidence of a differently (from primarily displacing) directed intentionality.
'Cycling', in this example, is as atypical of his action as 'sitting in a chair'
characterizes the action of someone who is watching TV or reading a novel. It is
suggested that an adequate description of a motor-action, to be qualified as such,
has to contain references to at least the following three components:
the person - 'actor' - with respect to his, primarily, displacing-directed
intentionality;
the "bewegungsrelevante Umwelt" (Scherler, 19791, being the 'world', as
understood ('known') in a displacing way (as to this, Gibson's notion of
'affordances' could be firther elaborated);
the manner of displacing, in terms of spatio-temporal relations (in verbs such
as, e.g., 'high-jumping','long-jumping',or 'synchronizedjumping').
In their interdependency, these three components constitute the necessary
conditions for the characterization of a motor-action. Considered in the light of the
image of the relational body they are the central nodal points in the network of
'relations of meaning' that can be discerned and realized in motor-actions. The
specificity of motor-action lies in the fact that the network of 'relations of
meaning', becoming manifest in such an action, is primarily discerned and
realized in a displacing way. In contrast to the case of 'displacement of (parts OD
the substantial body' this 'displacing' can only be adequately described in terms in
which the three aforementioned components are, collectively, recognizable.
Conversely, the notion 'displacing' constitutes the connecting thread between
these components, without which no central nodal points in the network of
'relations of meaning' that are to be discerned and realized in motor-actions,
could exist. With respect to the demarcation of motor-actions from actions of a
different kind it can be said that the notion 'displacing', considered as the
connecting thread amongst the components just mentioned, constitutes the
necessary condition for the characterization of an action as a motor action
(differentfrom, for example, a 'speech action').
The necessary conditions just mentioned are surely not sufficient if the concern
is with a characterization of the 'relations of meaning' that can be discerned and
realized in motor-actions in the context of practical fields like movement-
education and sport. In that context the nodal points in question are embedded in
a wider network of 'relations of meaning'. Motor-actions in the context of
movemenbeducation and sport always appear in a particular intervention
perspective, specified by distinct patterns of constitutive rules (e.g., 'citius, altius,
fortius'). However, the brief space afforded for this Chapter constrains an
interpretation of the necessary conditions to those just mentioned.
Images of the Body 457
4. Epilogue
The fact that certain 'displacements of (parts of) the substantial body' play a role
in motor actions, as in all other types of action, is beyond question. Yet, the
peculiarity of 'actions' cannot be adequately characterized if the displacements in
question are taken as the point of departure for further reasoning. This can be
more comprehensively understood by calling attention to the interrelation of the
two images of the body, introduced above. With respect to the aforementioned
'unitary vision' of Merleau-Ponty: the 'unity', which he indicates by 'chair du
monde', should not be conceived in such a way as to be incompatible with
'distinctions' and 'differences'. From the outset, the concern is with a 'unity in
diversity', a unity that dynamically manifests and unfolds itself in differences.
The images of the substantial and of the relational body (including their variants)
have to be understood in the sense of unfolding manifestations of that unity. As
soon as the images of the substantial and the relational body are made absolute,
they become, logically speaking, mutually exclusive and function only under a
pretence of being comprehensive. Yet, even then, they still appear to evoke each
other, thus exposing the incompleteness of either. The two images of the body are
both complementary and connected in the sense that they refer, in the last resort,
to an underlying 'unity', which unfolds itself in both, albeit in a different way.
Both images of the body are concerned with time and place-dependent
transformations of the same dynamic basic-structure, which makes their
existence possible. As 'reductionistic' and 'relational transformations' (Tamboer,
1985) of an ontological reIationality, they, respectively, map out an empirical
reality, within which impervious dividing lines, ultimately, do not arise. Even so,
the image of the relational body, in being a 'relational transformation', has a
certain ontological priority. If the image of the substantial body is taken as the
point of departure for further reasoning, 'relationships' can only become
apparent in an 'extrinsic' form. On the other hand, by discerning the ontological
primacy of a relational point of view, the reductionistic nature of extrinsic
relations can not only be understood as such, but also truly valued.
Consequently the so called 'motor-approach, by taking its point of departure in
the image of the substantial body, can only be really understood, and truly valued,
in the light of the primordial perspective of an 'action approach - with the
restriction that the latter be considered a cluster of approaches rooted in an image
of the relational body. Such an approach can not, then, be considered a variant
within the framework of a 'motor approach. Consequently, the relevant question
should not be: what turns a 'mere bodily movement' into an 'action' but
conversely: what turns an 'action' into a 'mere bodily movement'? Reed (1982a, p.
458 J.W.Z.Tamboer
ACKNOWLEDGEMENTS
The present author is indebted to two anonymous referees for helpful suggestions,
and to Peter J. Beek and the Editors of this Volume for their help in the final
shaping of this Chapter.
REFERENCES
Gorcum.
Van den Berg, J.H. (1950). Menselijk lichaam, menselijke beweging. Nederlands
T&ischrift vwr a!e Psychlogie, 4 & 5 , 267-299& 400-429.
Van Eemeren, F.H. & Koning, W.K.B. (1981,Eds.). Studies over Taulhundelingen.
Meppe1:Boom.
Van Olst, E.H. (1981). Psychologie en wijsgerige anthropologie. In F. Orlebeke,
P.J. Drenth, RH.C. Janssen & C. Sanders (Eds.), Compendium van uk
Psychlogie. I. Achtergmnden en Stromingen (pp.73-104). Muiderberg:
coutiliho.
Van Peursen, C.A. (1978).Lichuum-Ziel-Geest,Utrecht: Erven J. Bijleveld.
Von C m c h , M. & H a d , R. (1982). The Analysis of Action. Cambridge:
University Press.
Vroon, P. & Draaisma, D. (1985).De Mens als Metafoor.Baarn: Arnbo.
Weinberg, P. (1978). Handlungstheorie und Sprtwissenschfl. K6h: Pahl-
Rugenstein.
Whiting, H.T.A. (1984, Ed.). Human Motor Actions: Bernstein reassessed.
Amsterdam: North-Holland.
Willimczik, K. & Roth, K. (1983). Bewegungslehre. Grundlagen, Methoden,
Anulysen. Reinbek Rowohlt.
h e r , R.M. (1971).The Problem of Embodiment. Den Haag: Martinus Nijhoff.
Zaner, R.M. (1981). The Context of SeZfi Athens, OH University Press.
Complex MovementBehaviour:The' motor-actioncontroversy,pp. 463-487
O.G. Meijer & K Roth (editors)
0 Elsevier Science Publishers B.V. (North-Holland),1988
Chapter 19
Lothar Pickenhain
SUMMARY
It is argued that theories of complex movement behaviour can only be valid if they
are embedded in a holistic view on human behaviour. Such a view has to
comprise biotic (motor, autonomic) as well as psychic components, unified by self-
sustaining integrative systemic processing. The leading notion of this view is the
active effort of the organism to reach special need- induced andlor cognitive aims.
Goal-directed behuvwur is determined by genetic and epigenetic factors, and
leads, while using former experience and information concerning the actual
situation, to the fulfilment of the existent biotic andlor psychic needs through
internal probabilistic anticipation of future events. In Man, the highest-order
steering and controlling mechanisms of goal-directed behavwur have acquired
such complexity that the qualitatively new functions of verbal and non-verbal
cognitive (mental) phenomena have evolved on the basis of social relations,
steering his subjective voluntary behaviour in all its complexity.
1. Introduction
'Complex movement behaviour' is a very important topic which can only be dealt
with by interdisciplinary research. I am especially interested in this topic because
it touches on the problem of the highest level of organization and control in the
brain and the relations between the organism and its ecological environment
without neglecting the lower input and output levels, and because it considers
mental and cognitive, conscious and unconscious, processes as well. In most
cases, research on complex movement behaviour is phenomenological 'and
regards the phenomena in the context of biomechanical, physiological,
psychological, clinical, and other types of experiments. But now, the tendency is
prevailing to seek for a common theoretical frame, and with regard to this
framework I wish to make some personal remarks.
They concern the following issues:
the unity of perception and action in the theory of functional systems;
the goal as the deciding link in the architecture of behaviour;
the gap between the study of the elementary and the highest level of morpho-
functional organization;
the relevance of neurophysiology regarding programming and controlling
processes in the brain,
advantages and risks of modelling brain function;
some words on the genetic question;
some new ideas about selection versus induction;
mental (psychological) processes as the highest steering and controlling level in
the brain.
Perception and action must be taken together as an obligatory entity. This was
already emphasized by Viktor von Weizsilcker in 1939. Neither function may be
understood if we regard them as distinct. Of course, there must be separate study
of the elementary structures and functions of perception and action but this can
only be the initial step for understanding their joint functioning within the
organism as a whole. Perception and action are joint constituents of the behaviour
of the organism, and we cannot understand the elementary events if they are not
related to the behavioural function of the organism as a whole. Each perception is
influenced by efferent impulses (e.g., the level of arousal, attention, preselected
evaluation), and every motor action is embedded in the whole field of internal and
external (environmental)relations.
A Neuroscientist's View 465
Pavlov (see 1954) used to talk about the 'analytic' and 'synthetic' approaches in
physiology; he combined these two in his experimental studies on circulatory and
gastro-intestinal systems, and, later on, in his studies on 'higher nervous activity'
(behaviour of the organism). Both approaches are necessary and vital but the
holistic view - regarding the function of the entire organism in the determination
of behaviour to which all elementary processes are contributing - must be the
leading one. What we investigate as 'elementary processes' or 'elementary
functional events' comprise the immense bulk of all morpho-functional
possibilities. But in any actual situation only a small part of these possibilities is
used to reach the intended goal of the organism. Reaching that goal is the
deciding issue.
Anochin (see 1978) distinguished and studied various so-called 'functional
systems' - being organized and developing according to special patterns to fulfil
vital needs of the organism in relation to the changing ecological environment. He
showed that during ontogenesis different elementary structures and functions
develop primarily as part of such functional systems and not, as many believe,
according to a proximal-distal principle. In new-born babies one of the first
coordinated goal-directed complex movements is sucking, soon complemented by
thumb-sucking. For this purpose some distal motor units, moving the thumb,
must have developed very early. In a similar way, during the early development of
reaching, the grasping movements of the hand show an elementary coordination
before the proximal parts of the arm are able to execute well-coordinated smooth
movements. In this idea of 'functional systems', specialized motor programmes,
linking perception and action, emerge as functional entities.
Summarizing his ample investigations and suggestions, Anochin concludes in
his 1968 paper (see 1978)The systemogenesis as a general rule of development that
'systemogenesis', i.e., "the adequate and timely consolidation of vital hctional
systems of the organism" (p. 98),acts as the "regulator of the development of all
brain structures and brain functions." (p. 124). The cause for the development of
the organism's functional systems - well coordinated in space and time and
adapted to the environmental conditions and the needs of the organism - is given
by natural selection during evolution. The genetic code contains the principal
rules, the realization of which is influenced by ecological and interindividual
factors subtly adapting the functional systems to the actual situation.
Anochin's concept of 'systemogenesis' is an attempt to understand the dynamic
development of relations between organism and environment originating during
the history of life. According to this view the genetic code is the result of
phylogenetic selection by adaptation to the respective ecological niche; its
realization occurs during ontogenesis of the individual organism - in the
466 L.Pickenhain
A B
Afferent Synthesis
zn
K
A
EF F
(1958) proposed to call this type of biotic event 'teleonomic'. 'Teleonomy' is one of
the main principles of all biotic beings. All self-sustaining, self-regulating living
systems must display active behaviour with regard t o their environment as an
essential condition for their survival. This means that all behaviour must be goal-
directed t o assure internal and external homeostasis. In this connection the
terms 'teleonomy' and 'teleonomic' have only a descriptive, not a n explanatory
value.
A Neuroscientist's View 469
According to Edelman (1975) we can assume that the cerebral cortex of Man
consists of 2-6 million self-regulating neuronal circuits or 'modules', each
comprising 6-10 thousand neurons. These neuronal groups are formed during
embryogenesis and ontogenetic development. Intrinsic connections exist within
each neuronal group, involving a variety of synaptic as well as non-synaptic
modes of interaction within a local circuit. These intrinsic connections vary
greatly from group to group. Extrinsic connections exist among the groups,
specified by genetic codes and synaptic selectivity.
It is supposed that neuronal groups of different structure and connectivity form
groups of a higher order which can respond to, or recognize, a particular signal
pattern or initiate a special motor action. Through the existence of a many-one
principle with degeneracy, there must be more than one way of satisfactorily
recognizing a given input signal. This implies the presence of multiple neuronal
groups with different structures, capable of carrying out the same function more
or less efficiently. Therefore, degenerate groups are isofunctional, but not
isomorphic. The polling of such degenerate groups of higher order by signals
leads to associative recognition. Led by these assumptions, Edelman considers a
response hierarchy that in a later stage will be non-linear because of the presence
of feedback and feedforward loops with their associated alterations of temporal
patterns and response timing. These higher-order neuronal groups are
represented in the association cortex and the temporal, frontal and prefrontal
lobes, forming changing subgroups that act as 'recognizer of recognizers'. They
alter their properties as a result of experienced input, changing the probability of
their future selection in preference to their neighbours; in this way a collection of
different higher-order neuronal groups is produced.
Thus, in these selective subgroups of modules and their mutual connections,
genetically coded and epigenetically coined by experience, we may have those
'elementary self-organizing entities' which, variably and hierarchically
connected into higher-order self-regulating systems, continuously amalgamate
the activity of the nervous system into the steering unity of the whole organism.
In such self-organizing and self-regulating circuits of neuronal modules, the
external world is being represented, including past experience as well as current
influences, and the internal image of environmental events is produced. By their
activity, the probabilistic image of future situations, and of the possible effects of
actively changing them, determines goal-directed / future-oriented behaviour (cf.
Anochin, 1978; Bernstein, 1975). Included in this activity is the genetically and
epigenetically determined preprogramming of future actions which, after the
decision, are performed through subprogrammes of different levels that include
motor as well as autonomous components.
A Neuroscientist's View 471
Up to 15 years ago the single method to obtain more detailed knowledge on the
highest fimctional level of the human brain appeared to be the performance of
psychological tests. Since the time of Wilhelm Wundt purely subjective statements
were supplemented by experimental tests with objective measurements, and a
special psychological terminology was developed. But it became evident very soon
that, despite the accumulation of a large body of knowledge - especially in the
fields of perception, learning, language, and other psychic functions - , a real
elucidation of the underlying dynamic and functional mechanisms in the brain
could not be reached from this side. Exact study of the psychic consequences of
brain damage also did not have the expected results and could only reveal
relatively coarse correlations between biotic and psychic phenomena. This state of
472 L Pickenhain
At this point, we shall not analyse further preconditions 1 and 3 but we shall
stress some new pathways of research which, especially if they are applied within
a complex programme, promise to open new insights into the intricate dynamic
functioning of the brain. In this connection, we do not take into consideration all
the new fascinating morpho-biochemical discoveries made in the human brain
A Neuroscientist‘s View 473
APV
apparent movement occurs. But already 800 msec before, a negative shift starts,
indicating the activation of broad areas in the brain. This negative shift is to be
observed on both sides of the frontal and middle parts of the human scalp (area 5
in particular). In psychological terms this is the period in which the subject
'intends' to perform the voluntary movement. It corresponds to the period of
probabilistic programming assumed by Bernstein (see 1975) in which broad areas,
including the frontal, association, and sensorimotor cortex, are participating.
This so-called 'readiness-potential' ('Bereitschaftspotential',BP) is followed by a
'premotor positivity' (PMP) which is differently expressed at the various
derivation points of both hemispheres. Only shortly before the execution of the
movement, i.e., after the decision what voluntarily to do, do we see a negative
'motor potential' (MP) which is clearly restricted to the representation area in the
motor cortex contralateral to the finger that shall be moved. At the same time we
can record tonic pre-innervation impulses in the periphery preparing the optimal
performance of the intended movement.
As we see, such electrophysiological recordings from the human scalp can give
us some very interesting hints concerning the participation of various brain
structures. Further refinements of experimental design and technical
performance of the experiments promise far-reaching conclusions, especially if
combined with other research methods and compared with analogous exper-
iments in animals. How far we already can go may be shown by an experiment of
Keidel (1983).In this experiment the subject was instructed to perform every 3
seconds 'mental', i.e., without any real movement, dorsal flexions of the large toe
of one foot. The EEG was derived from the contralateral motor area and computed
by autocorrelation. As can be seen in Figure 4, the 'mental' dorsal flexion of the
toe produced a clear change of the autocorrelogram in the range of about 3 sec.
This reflects the mental process of 'imagined movements', In other experiments,
changes in autonomous functions during mental training could be shown of
which the mentally training subjects were not aware.
Changes in the slow wave electrical activity of the brain by mental influences
are also shown in Figure 5. The subject is performing a tracking experiment in
which a moving line on the screen must be followed by a second line triggered by
elbow movements (MG). The task consists in following a jump function. During
the experiment electrical brain activity is derived from the vertex. T w o different
instructions are given: either to follow 'exactly' or t o follow 'quickly' (the type of
motor reaction can be seen in MG).Dependent on the instruction, the slow
readiness potential (BPI changes; it looks completely different after each of the two
instructions, reflecting the different preparatory mental processes in the brain
(my own experiments).
A Neuroscientist's View 475
1 2 3 4 5 sec
h B
1 2 3 4 5sec
I
I
I
1 sec
+--------------
frontal cortex
(t memory, t emotion, ,
4- - _ _ _ _ _ _ _ _ _ I i imagination
of aim
t situation) I
I
I
I
I
I
’1
i
I
probabilistic
basal
J lateral
I
I
I
I
programrnin
z
ganglia cerebellum
(---.
intermediate
cerebellum
+---
+
-
I
I
I
I
I
I
h
1
I U
-
I Q ,
: a
I
I
I
motor action I
Other types of approaches have been developed during the last years that we can
only mention briefly. There is especially the evoked bio-electrical activity (evoked
potentials, EP) elicited by various external stimuli or internal events. We must
also point to the patterns of brain activity revealed by measuring the regional
cerebral blood flow (e.g., Ingvar, 19751, the distribution of radioactive glucose in
different brain regions during mental activity, and the application of positron
emission tomography which opens the way for detecting various local biochemical
changes in the living human brain (Phelps Q Mazziotta, 1985). By this method
also deeper brain structures may be investigated and the consequences for the
development of a subtle functional neuroanatomy, by using these and other
methods, are not foreseeable.
Of course, we cannot yet demonstrate (in biological terms) the fluent stream of
behavioural actions as they usually occur in everyday life or during special
training procedures. But serious work is going on in this direction and multiple
recordings of different functions are especially promising.
of biotic processes. Arshavsky, Gelfand and Orlovsky (1984). coming from the
Bernstein school, underline their critical position regarding cybernetic modelling
in their book on the Cerebellum and the Regulation of Rhythmic Movements.
They write:
I think it is completely clear that there must be a genetic predetermination for all
principal rules according to which the internal structures and functions, and the
relations of the organism to its ecological environment, develop and function. This
genetic code evolved by adaptation and selection within the ecological niche in
which the respective species was developing during evolution. The question
remains in which way this genetic code is realized and how many degrees of
freedom it leaves open for subtle dynamic adaptation during ontogeny and later
life.
I think that in this question too Bernstein (see 1975)gave us some very important
suggestions. He pointed to the fact that everyone can distinguish different types of
trees. We know how the bark, the leaves and the branches of the oak, the lime, or
other trees look, and one leafis enough to recognize what type of tree it is. Without
any doubt the principal structural plan of each type of trees is genetically
determined. But not determined are the exact number of leaves, the number of
branches and ramifications, the special direction in which the tree is growing,
etc. This means that the genetic code cannot enumerate those details which are
not essential or vital for this special type of biological being. Bernstein (see 1975)
supposes that the limit number of elements the genetic code can specify lies in the
range of about 26. The realization of the genetic code lets open a wide range of
variations which gives the possibility for necessary adaptations during the
individual development of the single specimen.
The same holds true for the genetic predetermination of the organization and
development of the peripheral and central nervous system. Nowadays we know of
the enormous surplus on motor neurons and synapses growing out during early
480 L.Pickenhain
ontogeny. The large msjority of them are deleted during development, because
they don't reach their destination point and are not used. The same applies to
connections inside higher levels of the central nervous system. The genetic code
for building up the central nervous system comprises all necessary and sufKcient
rules according to which the distinct hierarchical and heterarchical levels of the
central steering and controlling organ are built-up. The principles of organization
are determined, but not the details, Genetic rules do not determine the number of
cells in the different subsystems, but do determine their principal functional plan.
They determine the construction and functional activity of locomotor generators at
different levels of the central nervous system and give the sequence in which the
motor generators of the different levels start working. But the genetic code does
not give the exact timing of this sequence, i.e., when they will actually start to
function. This depends on many additional internal and external factors. The
general rules of the genetic code do not define the semantic content and
grammatical rules of our language but they prepare the morpho-functional
subsystems in the brain on the basis of which the influence of our social
communication can create the amazing world of human language and mental
processes.
Therefore, these general rules may be relatively few and straightforward. We
know that in the developing brain a multiple surplus of neurons and synapses is
produced, which during early ontogeny is reduced to the necessary amount to
fulfil the behavioural demands and the demands of internal organization. But we
also know that during early and later development - within the boundaries of the
given genetic rules - many plastic changes occur and give rise to the plasticity
needed for adaptation and learning. This occurs especially in higher-order
subsystems which are the basis of the highly integrated function of sign and
programme storage for the probabilistic prediction of future events, the decision
for action according to the actual demands, and the top-down control of all
internal functions.
We already mentioned the immense number of neumnal groups and assemblies
which are organized according to degenerate principles to fulfil different
functional tasks (Edelman, 1975). The existence - not the number - and the
principle type of development of these neuronal groups is determined by the
genetic code, but their degenerate functioning, i.e., the degenerate selection to
fulfil special tasks, is mostly determined by internal laws of the self-sustaining
developing systems and by the actual afferent input, especially during so-called
'critical periods' of development. Similarly, the extrinsic connections between the
various groups and group configurations are specified by both the genetic code
and acquired selectivity.
A Neuroscientist's View 481
'Selection' implies that after ontogeny and early development, the brain
contains cellular configurations that can already respond discriminatingly
to outside signals because of their genetically determined structures or
because of epigenetic alterations that have occurred independently of the
structure of outside signals. These signals serve merely to select among pre-
existing configurations of cells or cell groups in order to create an
appropriate response.
2. The total behaviour repertoire of animals and Man may be classified in four
types:
Goal-seeking behaviour which is triggered by internal drives and/or external
stimuli and will change over into goal-directed action, or vanish if the
requirementa disappear.
Goal-directed action which may be initiated by internal needs or external cues
on the basis of inborn and/or learned, (partly) automatized, and variable, adaptive
mechanisms.
*The rest state of the organism during which recovery processes occur;
484 L.Pickenhain
ACKNOWLEDGEMENTS
I very much acknowledge the invitation of Prof K. Willimczik and Prof H.T.A.
Whiting to participate in the Bielefeld Conference on 'Complex Motor Behaviour',
and I express my warmest thanks to Dr Onno G. Meijer for improving my
imperfect English.
REFERENCES
Lothar Pickenhain
e
HolzhliuserstraOe8
DDR-7027, Leipzi ,
German Democra c Republic.
Complex Movement Behaviour:'The'motor-action controversy, pp. 489-561
O.G. Meijer 8 K Roth (editors)
0 Elsevier Science Publishers B.V. (North-Holland),1988
Chapter 20
SUMMARY
It is argued that 'the' motor-action controversy revives the agelong debate on the
existence and power of 'higher I central 1 commanding 1 controlling' entities,
such as the soul, the brain, a motor programme, or an internal representation:
'the hierarchy debate'. Action theories are anti-hierarchical at heart. The history
of decapitated frogs shows how difficult it has been to positiuely specify an anti-
hierarchical position from the seventeenth century onwards. Disentangling
'analysis' from 'control', and acknowledging the importance of 'low level
autonomy' (self-organization), may bring the solution to this problem. Modern
dynamical study of self-organization Cteleodynamics3 is the most attractive
starting point for the analysis of motor control. Turing machines have
'secondary' self-organization and 'secondary' self-reference. The advantages of
models based upon natural Turing machines on top of low level autonomy
CTuringuesque madelling') are discussed from an evolutionary standpoint. It is
concluded that Turinguesque modelling allows for the understanding of
'learning'. Hence, motor theories are incomplete and will have to acknowledge
the partnership of teleodynamical modelling, action theories are incomplete and
will have to acknowledge the partnership of Turinguesque modelling.
490 O.G.Mewer,R.C. Wagenaar & A.C.M. Blankendaal
Also Reed (1982a), on grounds very different from those of Schmidt, claimed a solution for the
dichotomy: by 'transcending' it.
This is not to say that Schmidt allows for no feedback a t all during ballistic movements. The
execution of any programme may be modified by feedback after a certain amount of time:
"Immediately following the initiation of the motor program, the impulses begin to flow out to the
muscular system with all of the details of at least the first 200 msec of the movement specified. As
the movement is carried out, sensory receptors in the body provide information about the
movements occurring. ... The raw error is fed back to the schema so that in positioning
movements, for example, additwnnl at(luntments can be made to reduce the error to zero."
(Schmidt, 1976,p. 239, our italics).
Their position (see also Shaw & McIntyre, 1974). in fact, goes beyond classical heterarchy: rather
than analysing relationships between 'levels' inside the organism, they emphasize the dual
complementarity between organism and environment. Such 'self-organization' is anti-
hierarchical par excellence.
The Hierarchy Debate 491
peripheral, about serial versus parallel, and so on. ... These issues, I
claim, about whether one or the other characterizes some phenomenon,
serve to drive a large part of the experimental endeavour.
It was probably Michael Turvey (1977,pp. 223-224) who first used the term
'heterarchy' in an ecological context.'Heterarchy' was, in 1977, rather en vogue
(cf. Broadbent, 1977, pp. 189-190),and it was, at the time, a bit hard to decide
whom in the field of movement behaviour Turvey was attacking. It has been
speculated (cf. Requin, Semjen & Bonnet, 1984)that Keele (1968)offered a major
example of the hierarchical position Turvey was referring to, 'although4 feedback
plays an important role in Keele's 'programme' theory (6. Keele, 1968, p. 397).
Turvey, however, did not want just to reject open-loop positions; he wanted to go
further. If one takes Bernstein and Gibson seriously5, he seemed to argue, one
has to accept the fact that in action control 'the executive' is deemed to be
'ignorant' of what exactly the 'lower' levels will do:
This form of "ignorance" has been duly recognized by students of action. ...
The executive ... is to modify the mode of interaction among elements a t a
lower level ... . As a general rule, however, ... the executive does not have
advance knowledge of which particular state, out of a set of possible states,
a lower level will arrive at after a mode of interaction has been specified ...
(Turvey, 1977,226, our italics).
Turvey's 1977 predicament reminds one somewhat of Darwin's vow never to use
the terms 'higher' and 'lower' while consistently refusing to adhere to that very
rule (see Ospovat, 1981, p. 227). Already in 1978, however, Turvey et al's
'coalitions' allowed for the fading away of 'higher' and 'lower', and in 1980
This 'although' (d.Schmidt, 1988) misses the point that ecological anti-hierarchism includes
rejectingopen-looppositions which do allow for feedback, as did Keele, 1968, p. 397.
5 For Bernstein, one is referred to, e.g., Bemstein, 1967 & 1975, Pickenhain and Schnabel, 1975,
and Whiting, 1984; for Gibson we mainly used Gibson, 1950, 1966 & 1979, and Reed and Jones,
1982.
492 O.G.Mewer,R.C. Wagenaar & A.C.M.
Blankendaal
'coordinative structures' appeared on the scene (Kugler, Kelso & Turvey, 1980;
Kelso et al, 1980).
No vortex in a river - to use our own paradigm example - can ever be understood
by focusing on the static architecture of its building blocks. Inasmuch as the
behaviour of the water molecules in the vortex is coherent, the ,'structure' of this
coherence is not a property of 'things' but one of 'processes' (Jantsch, 1985). In
this sense 'coordinative structures' refer to the dynamic coherence of ongoing
actions and not to the cooperation of the building blocka of the locomotor
apparatus6. The locus of 'coordinative structures' is not the organism per se but
the 'naturally evolved system of the organism and the environment' (Kugler et al,
1980, p. 9), a system which is both autonomous and task-specifi. Just so, the
coherence in a vortex does not result fram that vortex in isolation, but from the
whole system of the inflowing water, the stones, the plants, the sand, etc., a
system which is autonomous because it has no higher I central I' commanding I
controlling entities, whereas its specificity rests in its being-a-vortex.
At first, Turvey and his co-workens wanted 'control' within the system to take
place with "a minimum recourse to an 'intelligent regulator"' (Kugler et al, 1980,
p. 40): as if this entity still mildly existed; it took some years before the 'field could
appear to "specify its own parameters" (Carello et al, 1984, p. 242): as if the
controller and the controlled could be seen to coincide. In any case, Turvey's
heterarchism went beyond rejecting Keele: new concepts of 'control' were on their
way.
If our understanding of 'coordinative structure' is correct, then its description as "a groyp of
muscles often spanning a number of joints that is constrained to act aa a single functional unit"
(Kugler, Kelso & Turvey, 1982, p. 60) is a bit unfortunate, suggesting implicitly that actual
coordination may result from morphology. Coordinative structwes qua dissipative structures
make use of the underlying morphology which, in its turn, allows the dynamics (compare the
vortex). Early morphological statements about 'coordinative structures' should not be seen as
definitive descriptions but as historical transitions ptween Bernsteinian 'synergies' and
dynamical 'fields'.
The Hierarchy Debate 493
... The theory of indirect action ... holds that we act on our environment
only indirectly, by causing our bodies to move. This theory, which goes back
at least to Descartes' Passions of the Soul ... underlies most of our thinking
about action. Bernstein's arguments showing the impossibility of univocal
peripheral results of cerltral "commands" should have led him to rdject this
entire approach, but even he did not see this implication in 1935, and he
only saw it dimly as late as in the 1960s (Reed, 1984,163, italics in original).
Bernstein qua source of inspiration, may have been confusing indeed in his early
period, Bernstein opted for 'cybernetical' modelling in the regulation of motor
acts: although central does not know much about peripheral, it still can compare
the actual situation with the ideal one, and issue 'commands' to correct for errors
(cf. Bernstein 1957, in Whiting, 1984, pp. 343-371).Later, Bernstein started to
wonder about the relevance of non-equilibrium thermodynamics, questioning the
"honeymoon"between cybernetics and the biological sciences:
Bernstein is not the only one to have shifted his position: in 1977 Turvey's
'executive' retained some power, in 1984 the 'field apparently disposed of 'him'
(Carello et all; in 1982 'indicational commands' still appeared acceptable to Reed,
nowadays all control by 'commands' is rejected; in 1975 Schmidt's 'programme'
tells the lower levels how to operate, in 1988 it is uncertain what aspects of the
movement are controlled by the 'command structure'.
494 O.G.Mewer,R.C. Wagennur & A.C.M. Blankendaal
As such, the hierarchy debate is as messy as can be. The term 'heterarchism'
was coined by McCulloch (19451, one of the intellectual fathers of artificial
intelligence (cf. McCulloch & Pitts, 1943; Minsky & Papert, 19721, whereas the
term has also been used by those who want to oppose artificial intelligence;
'hierarchy theory' has been developed by those who are sympathetic towards
modern dynamics (e.g., Pattee, 1973a & b) but modern dynamics is, in a sense,
anti-hierarchical. Positions have repeatedly shifted and even reversed. It is our
contention that an analysis of the historical background is due before setting out to
critically evaluate the bewildering multitude of positions in today's version(s1 of
the hierarchy debate.
A commanding key-role has been played by animals whose 'command keys'
were removed, as Turvey (1977), Reed (1982),and Schmidt (1982, but not 1975) duly
recognized. The history of, in particular, decapitated frogs illustrates both the
variances and invariances of the positions taken. It is, therefore, the beheading of
frogs which we will describe as a major feat of arms in this agelong struggle,
dedicating our Chapter to all those frogs who rendered their Heads to the Queen of
Hearts!
2. The Queen'sEvidence
"Give your evidence," said the King; "and don't be nervous, or I'll have you
executed on the spot." (147) 7.
Take a frog and strip it. You may see the circulation of blood if you hold him
up against the sun.
Take out the heart and the frog will leap about a great while after.
We went to the Bodleian Library to study the context of these remarks. There
was none8.
7' All quotes from 'Alice in Wonderland are from Martin Gardner's1960 edition.
In the margin of the first sentence Locke wrote 'circulatio sanguinis', of the second 'cor'.
Kenneth Dewhurst (1963a, p. 867) stated the following : "Another note shows that Locke confirmed
Harvey's work on the circulation of the blood, and demonstrated in a simple animal experiment
the reflex activity of the nervous system." We wanted to establish in how far, as hontinues)
The Hierarchy Debate 495
Lockeg, at the time, was at Christ Church where later he would attend Willis's
(1 620-1675) lectures (1664), before switching allegiances to Sydenham ( 1 624-16 8 9 )
and moving to London (1666)lO. But neither Willis - author of Cerebri Anatome
- nor Sydenham - the 'empirick - refer, as far as we know, in their works t o
heartless frogs". We do know, however, that Locke was acquainted with Robert
Boyle (1627-1691) at least from 1660 onwards12 and we want to suggest that Boyle's
staggering experiments may have induced Locke to scribble down the above notes
as soon as he heard about them around, say, 165413.
(continued) Dewhurst has it, both the circulation of the blood and the nervous system were meant by
Locke. The text itself did not give any answer to that question: the note is placed between one on the
'ovum' and one on 'pallor'. We want to emphasize that nothing in Locke's note suggests that he
himself performed the experiments mentioned.
lo It may be interesting to note that, coincident with his switch to Sydenham, Locke's authoritarian
political views changed radically into, allegedly, egalitarian ones concerned with public liberty
(d.Gough, 1973).
l1 For Willis, see, eg., Dewhurst, 1964,1980 & 1981; Feindel, 1965; Hierons & Meyer, 1962 & 1964;
Meyer & Hierons, 1965; Willis, 1664 & 1670. For Sydenham, see, e.g., Cunningham, 1986;
Dewhurst, 1966; Payne, 1900.
l2 For Boyle, see, e.g., Birch, 1772; Debus, 1977, Vol. 2, pp. 473484. In 1660 Ivie wrote to Locke: "I
hope Sir; you will let slippe noe occasion whereby you may better yourself, and soe me, by your
acquaintance with Mr. Boyle" (Dewhurst, 1963b, p. 7).
l 3 In 1958 Dewhurst suggested that Locke, during his early years a t Oxford, was greatly
influenced by his friendship with Boyle. At the beginning, Lower may have been their go-between.
According to Boyle himself, the relevant experiments had been completed a s early a s 1649 (cf.
Debus, 1977,Vol. 2, p. 474).
496 O.G.Mewer,R.C. Wagenoor & A.C.M.Blankendaal
... The first of these ... shall be an experiment that we remember ourselves
formerly to have made upon frogs. For having opened one of them alive,
and carefully cut out his heart ... the frog notwithstanding leaped up and
down the room as before, dragging his entrails (that hung out) after him;
and, when he rested, would upon a puncture leap again, and being put into
the water would swim, whilst I felt his heart beating betwixt my fingers.
On the same page, Boyle's fascination with heartless frogs becomes embedded in
a context of philosophical questioning:
... How this experiment will be reconciled to the doctrine ascribed to Mr.
Hobbes o r that of the Aristotelians, who tell us, that their master taught the
heart to be the seat of sense ... let those, that are pleased to concern
themselves to maintain all his opinions, consider.
Boyle, evidently, was concerned with the localization of the 'seat of sense': he was
trying to localize the soul.
In 1664 Thomas Willie had connected 'psychologia', the study of mental
functioning, with the substance of the brain, rather than the fluids in its
ventricles - as had been the common view during the Middle Ages (Clarke &
Dewhurst, 1972). Substance, as opposed to fluid, has a neat localization: it can be
isolated and hrther divided. The animu mtwnulis, the 'rational soul', was for
Willis as much "beyond the realm of the scalpel" (Clarke & Dewhurst, 1972,p. 70)
as had been the res cogitans for Rent5 Descartes; but Willis speculated that an
Anima brutorum - a 'brute' or 'sensitive soul' common to man and animals - had
its 'fire' in the blood and was turned into 'light and air' whenever it travelled
through the brain, and then into the nerves.
For those, indeed, who wanted to take their scalpel and try to localize the
sensitive soul, the target organs were to be the heart and the brain. It was Robert
Boyle who indulged in cutting off heads from, among others, chickens, frogs,
tortoises and:
... vipers: ... not only their hearts clearly severed from their bodies may be
observed to beat for some hours .,. but the body itself may be, sometimes, two
or three days after the skin, heart, head, and all the entrails are separated
from it, seen to move in a twining or friggling manner: they ... may appear
to be manifestly sensible of punctures, being put into a fresh and vivid
motion, when it lay still before, upon being pricked, especially on the spine
or m m w , with a pin or needle (1663; cited from Birch, 1772, Vol. IZ, 70).
The Hierarchy Debate 497
Now although I will not say, that these experiments prove, that ... the bmin
may not be confined to the head, but may reach into the rest of the body,
after another manner than is wont to be taught; yet it may safely &rmed,
that such experiments as these may be of great concernment in reference to
the common doctrine of the necessity of unceasing influence from the
brain, being so requisite to sense and motion; especially if to the lately
mentioned particulars we add ... what we have observed of the butterflies ...
that they may, not only, like common flies and divers other winged insects,
survive a pretty while the loss of their heads, but may sometimes be capable
of procreation aRer having lost them ... (1663; cited f i m BimhJ772, Vol. ZI,
71; our italics ).
'There's more evidence to come yet, please your Majesty," said the White
Rabbit,jumping up in a great hurry: "this paper has just been picked up."
498 0.G.MeGert R.C. Wagenaar & A.C.M. Blankendaal
For as the best musician cannot make a flute the sound of a violin, nor a
harpsichord that of a French horn ... in like manner the soul can only ...
exercise its rational powers in the brain; it can only taste in the tongue,
smell in the nose, see in the eyes, hear in the ears, and feel hunger in the
stomach (Whytt. 1755, quoted from 1768,288).
l4 French (1972) translates "l'empire qui a l'8me sur le coeuf' as "the power of the soul in the
heart".
l5 "It i s true Dr. Stahl, by extending the influence of the soul, as a original agent, over the body a
great deal too far, and thus carrying this doctrine beyond all reasonable bounds, has been the
occasion, why it has rather for many years been looked upon as a subject of ridicule, than
deservinga serious and rational answer''(Whytt, 1751,p. 267).
l 6 For Whytt, see, e.g., French, 1969& 197%Carmichael, 1927;Whytt,1751& 1768.
The Hierarchy Debate 499
If then the soul in pigeons, h g s , vipers, and tortoises, is not confined to the
brain, but can continue for a long time to actuate their bodies independent
of that organ; ... why should we deny, that, in man and such animals as
resemble him most, the parts may continue to be actuated by the soul or
sentient principle for some few minutes after their communication with the
brain has been cut off? ...
If then life in animals be owing to the energy of a principle distinct from
matter, and of powers superior to it, we have reason to conclude, that, as
long as any signs of life remain in the bodies of animals, or in any of their
parts, t h i s principle still continues to actuate them (1768,291, our italics).
Nothing is more normal than to observe the frog's heart remain beating,
and the muscles irritable, even when both the spinal cord and the head
have been severed. It is true that Mr Whytt renders the time of death rather
l 7Concerning the "lateReverend and learned Dr Hales",see Whytt, 1768, p. 290; originally 1755.
l8 For Von Haller, see, e.g., Foster, 1901; kench, 1969 & 1972; Siidhoff, 1922; Von Haller, 1777,
1778,1779&1966.
500 O.G.Meder,R.C. Wagenaar & A.C.M. Blankendaal
shrewdly uncertain, and he believes that the animal still has life even when
it has appeared dead for a long time ... . Since, however, i t is certain that
the seat of the soul is in the head, and since the soul, when the head has
been cut off, loses its empire over the rest of the body, ... since, furthermore,
irritability remains perfect notwithstanding these circumstances, ... it
becomes clear from all this ... that irritability does not depend on the soul
(52-53;our italics; our translation).
Later, an Edinburgh student was to write: "Dr. Whytte had a great ...
controversy with the celebrated German Physician Baron Haller on anatomical
subjects ... , and he beat the baron hollow" (French, 1969, p. 11). Neither of the
gentlemen, however, appeared to have been "deficient in irritability" (French,
1969, p. 111,and Von Haller, in a 1772 postscript to his lectures, boasted that even
in Edinburgh most of his views were finally adopted (Siidhoff,1922, p. 57).
The executioner's argument was, that you couldn't cut off a head unless
there was a body to cut it off from; .that he had never had to do such a thing
l9 DeSauvages' rejoicing the reestablishment of the 'empire' of the soul over the heart could also
refer to the relation between reason and emotion but, as yet, we found no explicit mention of this
topic in the documents studied.
20 We are aware of the fact that this is a rather unorthodox characterization of Von Haller's views.
Piliigw (1853). Huxley (1870), and Reed (1986), all remain silent as to the possibility of
interpretingVon Haller as one who defended the 'liberty'of the 'lower'parts.
The Hierarchy Debate 501
The separated head and trunk may be sent a hundred miles in opposite
directions, and at the end of the journey each will be as purposive in its
actions as before. ... I am [therefore] unable t o see in what respect the soul
of the frog differs from matter (7).
from frogs, in that they have a will and a moral consciousness." (p. 1). This,
according to Manning (p. l), pointed to a faculty "distinct from the thinking
brain":
... A chest of carpenter's tools is inactive, and has neither invention nor
product without the mind and will of the carpenter. What have the brain
and the hand more than the lathe and the chisel, without the Agent from
whom they derive guidance and activity?(7).
Suppose that ... the anterior division of the brain ... is removed. ... The frog
may be kept in a state of full bodily vigour for months, or it may be for years;
21 For Huxley, we mainly used: Burrow, 1968; Chalmers, 1900; Di Gregorio, 1984; Huxley, 1900;
Huxley, 1893 & 1898;Irvine, 1955.
502 O.G.MeUer,R.C. Wagenaar & A.C.M. Blankendaal
but it will sit unmoved. It sees nothing: it hears nothing. It will starve
sooner than feed itself, although food put into its mouth is swallowed. On
irritation, it jumps or walks; if thrown into the water it swims. If it be put
on the hand, it sits there, crouched, perfectly quiet, and would sit there for
ever. If the hand be inclined very gently and slowly, so that the frog would
naturally tend to slip off, the creatures fore paws are shifted on the edge of
the hand, until he can just prevent himself from falling. If the turning of
the hand be slowly continued, he mounts up with great care and
deliberation, putting first one leg forwards and then another, until he
balances himself with perfect precision upon the edge; and if the turning of
the hand is continued, he goes through the needful set of muscular
operations, until he comes to be seated in security, upon the back of the
hand. The doing of all this requires a delicacy of coordination, and a
precision of acliustment of the muscular apparatus of the body, which are
only comparable to those of a rope dancer (quoted from the 1893 edition, 224-
225, our italics).
~~ ~~~ ~
22 For the rather infamous Huxley-Wilberforce discussion during the Oxford-Meeting &r the
publication of Darwin's theory, see, e.g., Ruse,1979.
23 For Pniiger, see: Du Bois-Reyrnond, 1910; Nussbaum, 1909; Robertson, 1968, pp. 264-268;
Miiger, 1853,1877,1889 & 1906; Pfliiger, not dated.
24 Pfeiffer's(1891, p. xviii) phrasing.
The Hierarchy Debate 503
movements like a writer over the letters of the alphabet (Lotze, 1844, p. 195).
According to Lotze, these most simple reflex movements can be observed par
excellence in decapitated animals because in them the soul is unable to compose
the letters into sentences (ibidem). Pfluger, regularly touring in his carriage
around the Dottendorfer Loch where his daughter would catch frogs for him
(Pfluger, not dated, p.3), is of a different opinion and presents some astonishing
experiments:
When a pair [of copulating frogs] is captured, and one only grips the male,
the female will be pulled out of the water as well because the male strongly
holds [her] enclosed... . When the spinal cord of the male is cut between the
atlas and the second vertebra ... he does not let go ... . If ... she attempts to
withdraw, he just holds her stronger ... , If one now drops some acetic acid
on one of his arms, he lets go with that arm, whereas the other one holds
the female, and he rubs off the biting substance with the ipsilateral leg.
Then, however, he again holds her enclosed with both arms ... . Be the
female quickly removed while he opens his arms ... or, in case this is
impossible, the female be excised from his arms, one may put different
objects on top of him ... without eliciting any further response. ... Well now,
I observed repeatedly that the decapitated animal, when one puts an active
frog on top of him, all of a sudden ... straightened himself, opening his ...
arms, reaching out for the withdrawing frog, grasping it, embracing it,
and again held it enclosed so tightly with both arms, that ... it could not
escape, or, if it did, had to carry the other one (Pfliiger, 1853, 17-18; our
tmnslation).
Stronger experimental evidence for the 'intelligence' of the spinal cord is hardly
conceivable. Pfluger, however, goes on and on and on. The fact that masturbation,
known to be an irritation of the spinal cord, causes loss of sense (p. 60); that the
tail of a decapitated eel tries to escape the heat of a flame (p. 114); or that the
decapitated frog used its other leg to rub off acid whenever the ipsilateral leg has
been cut (p. 124-126; Figure 2), proves beyond doubt, that every part of the
organism has consciousness (p. 331, and that the spinal cord is a seat for sensory
perception and will (p. 115).
25 For Lotze, see: Wentscher, 1903; Wentscher,1913 & 1925; Lotze. 1842,1844,1846,1851,1852,1856-
1864 & 1891.
504 0.G.Meijer. R.C.Wagenaar & A.C.M. Blankenduul
abhorring plebeian misuse of power and admiring the splendour of the old
nobility - he revolutionized German psychology by emphasizing the ability of the
organism to behave in teleomechunical fashion26. Lotze's acid review (1853, we
used 1891) of Pfliiger's experiments, however, shows that he was in trouble.
Unless one grants 'direct revelation' to the soul in a n eel's tail, one has to refer to
former experience engraved in its spinal cord. "The eel, however, before being
fried, tends to have experience with water but not with fire" (p. 172). Is one, then,
really to believe that the soul in the tail calculates that it is in the air, and not in
water (p. 174)? But the whole fabric of Pfliiger's reasoning collapses upon closer
26 The soul is, at birth, a tabula m a , the body a tabula inscripta (1844, p. 202) and so, the soul
neither has to 'invent' new movements, nor haa to 'wait' until goal-directed movements perchance
obtain (1863,p. 149). For the early history ofteleomechanism,see Lenoir, 1982.
The Hierarchy Debate 505
analysis since we cannot accept the existence of more than one sod (p. 153). It is
mechanically preformed teleology in the spinal cord (p. 1571, resulting from the
association (pp. 161-162)of physical traces (p. 168) of former experience, which
allows the decapitated animal to behave in a fashion accommodated to present
Circumstances (p. 152).
The Pfluger-Lotze controversy precipitated heated discussions in the literature27
and stimulated quite a number of frogs to loose their heads. Political metaphor
became the fashion in the second half of the nineteenth century and one may well
try to explain non-empirical differences between Pfluger and Lotze as correlates of
political views. At the surface, such a hypothesis appears t o work quite well: both
in politics and in science Pfluger gave more power to the lower strata than Lotze
did; similar 'explanations' might, anachronistically, be construed for Von Haller
/ Whytt and Boyle / Willis. In fact, one of the present authors' favourite pastimes
is to map our colleagues' neurophysiological hierarchy / anti-hierarchy onto their
political stance. Were an eventual correlation to have full explanatory power,
however, one would expect Pfluger to have been at least ambivalent towards Von
Haller28. He was not: up until the middle of the nineteenth century 'the hierarchy
debate' was explicitly concerned with the localization of the soul.
"If there's no meaning in it," said the King, "that saves a world of trouble,
you know, as we needn't try to find any. And yet I don't know," he went on,
spreading out the verses on his knee, and looking at them with one eye; "I
seem to see some meaning in them, after all ... ."(159).
27 Historically, this is somewhat of an anachronism: it is certainly the case that Huger's views
drew much attention (see, e.g., Robertson, 1968, pp. 264-268; originally 1878), but it was only
Fearing, 1930 (we used: 1970, pp. 161-186) who turned Pniiger's differences with Lotze into a
'controversy'.
28 Whereas Von Haller's theory of muscle contraction was rather heterarchical, his views on the
empire of the soul-in-the-headwere of a hierarchical nature.
29 Edward Reed in a note to one of the present authors (O.G.M.), accompanying a xerox of the
manuscript for his "Motor variability but functional specificity: Demise of the concept of motor
commands."(1986).
506 O.G.Meijer, R.C. Wagenaar & A.C.M. Blankendaal
Fukeon and her co-workers (1980) explicitly refer back to Pfluger'e experiments
and his spinal soul. The experiments could be confirmed, the interpretation,
however, was different: "The spinal frog can make complex leg movements
according to the scheme of the body", while "central commands ... define the final
hip angle", and "the motor program provides sufficient stiffness of the hindlimb
that the influence of gravity on the movement is minimized (p. 1261). In 1980,
thus, the spinal cord of the frog was able to issue 'commands'.Why?
At least one of them, Anatoly Fel'dman, has been greatly influenced by
Bernstein (cf. Fel'dman, 1986). Well now, there are at least two principally
diferent ways to solve Bernstein's problem of how to cope with the tremendous
number of degrees of freedom in the control of action. One can focus on dynamical
self-organization (as Turvey and his co-workers are doing), or one can postulate
low-level mechanisms which take care of the degrees of freedom (as Asatryan and
-
Fel'dman - 1965 did)30, thereby allowing for univocal 'commands'. It was this
latter 'solution' which led to Fukson et al's h c t i o n a l analysis of the spinal cord
in decapitated frogs.
Edward S. Reed could take it no longer. He wanted to regard the spinal frog's
wiping reflex as "fhctionally specific to the problem caused by irritation"
(manuscript for his 1986 publication, pp. 3-4).Berkinblit, Fel'dman, and Fukson
(1986) may present important experimental evidence, but:
What the authors do not say ... is that the central signals are not 'motor
commands'. These signals create the conditions under which muscle
stretches occur, but they do not establish either muscle lengths or forces
and they do not specify movement parameters. ... A signal that influences
but does not determine an act should not be called a 'command (4).
And then, Reed goes on to present a short history of the spinal frog, relating it to
the present situation. Full circle:
Berkinblit et al's hypothesis that the spinal cord of the frog contains a body
image ... is a descendant of Whytt's "sentient principle". The only way to
save the hypothesis of mechanical specificity ... is to postulate souls which
adapt stimuli to the body's needs and fashion responses accordingly. ... The
30 The mass-spring metaphor may have it both ways, its equifinality dramatically reducing the
number of degrees of freedom: for Kugler et al (1980) the mass-spring is an example of a self-
organizing dynamical phenomenon, for Schmidt (1982) it allows central commands to define
endpoints of mechanical muscle-movement(cf.Wagenaar, 1986).
The Hierarchy Debate 507
FINE
We may describe the office of the brain as that of avemging the interests of
life, physical, intellectual, moral, social; and a good brain is one in which
the desires answering to these respective interests are so balanced, that the
conduct they jointly dictate sacrifices none of them. Similarly, we may
describe the office of a parliament as that of averaging the interests of
various classes in a community; and a good parliament is one in which the
parties answering to these respective interests are so balanced, that their
united legislation concedes to each class as much as consists with the
claims of the rest. Besides being comparable in their duties, these great
directive centres, social and individual, are comparable in the processes by
which their duties are discharged (Spencer, 1972, p. 53; originally 1860;
italics in original).
508 O.G.Mewer,R.C. Wagenaar & A.C.M. Blankendaal
In the present stage of our trial concerning the existence and power of 'higher /
central / commanding / controlling' entities, the Jury is bound to be codbed by
the striking fuzziness of central notions in the debate, such as 'hierarchy' and
'control' - a fortiori 'the hierarchy debate' itself. In order to reach a Verdict, the
members of the Jury will have to reach at least some form of agreement as to the
meaning of these notions.
God
Angels
Man
Animals
Plants
Minerals
B. Hierarchy
H m
Heterarchy Self-organization
Figure 3.Chartingpositwns.
Both Spencer's physiological and political views may be wen as heterarchical' (cf. Peel, 1972);
it was, however, not his heterarchy per ee. which inspired others, but his politically inspired
confoundingmeaning, analysis,and control.
32 See Bekkenutte et al, 1985; Weindling, 1981; cf. Haeckel, 1923, p. 45 (originally 1878); Verwom,
1917, p. 13.
510 O.G.Meijer, R.C. Wagenaar & A.C.M. Blankendaal
For similar attempts, compare Pew, 1974, with Pew, 1984; and see, e.g., Razran, 1971, Pattee, 1973a,
Houk 81 Rymer, 1981.
The Hierarchy Debate 511
The Jury, then, gladly accepts analytical hierarchies, but is of the opinion that
control hierarchies call for empirical research, rather than theoretical a priori
decisions, Of course, in order for such empirical research to be feasible, at least
some theoretical agreement has to be reached as to the meaning of 'control'.
34 Houk and Rymer's 1982 definition of the nature of cybernetic control is more specific than, but
not inconsistent with, our own understanding. Our quarrel with Houk and Rymer is not that their
expos4 be unclear - in fact it appears to be more sophisticated than our own one; the problem we have
is that cybernetic control theory reveals one position within a multitude - this bias, of course,
allowing Houk and Rymer to be both more specific and more sophisticated. They define the
function of the controller as "to generate appropriate forcing functions, based on its inputs, called
references, set points, or command signals that designate the desired behavior of the overall
system" (pp. 259-260). To us, the term 'forcing' is a concession to low level autonomy while
'commands', when used within an information processing context, appear to be inappropriate for
the specification offorcing functions (cf.Section 5).
35 Since this language draws heavily fmm corporate management, it offers another example of the
difficultyto disentangle analysis from control.
a6 Since self-organization 'specifies its own variables', such 'external' control cannot refer to
purely 'external' variables. External changes in variables which are both internal to the system
and part of its environment, are the candidates for control on top of self-organization.
37 In fact, one may be tempted to say that Von Haller was the outstanding exception in the proto-
history of the hierarchy debate: in a way, his 'irritability' was a form of self-organization, on top
of which external forcing could come from the 'soul-in-the-head. We certainly want to further
investigate Von Haller's theories and their reception.
512 O.G.Mewer,R.C. Wagenaar & A.C.M. Blankendaal
possibility, it is quite possible that our quest would have been superfluous! But
then, self-organization / autonomy38 appears to have been unpalatable for quite
some time.
The dominant feature is that one process alters the nature of another
process, rather then merely supplying it with input. Notice ... that this type
of theory embodies many of the points from traditional observations, which
are hard to handle in open chain or closed loop theories. For example, the
lower level can function in the absence of the upper, so that the obsel-vations
of Hughlings Jackson on epilepsy or alcohol ... become reasonable. The
upper level is concerned with nwdifmbility and response to novel situations,
mther than transmission of information according to an unchanging code
(199,our italics).
39 Arbib and Szentegothai do emphasize,however, the existence of hetermhies (cf. Section 6).
The Hierarchy Debate 513
modes of control: supplying a process with input without altering its nature - we
rather arbitrarily opted for the term external pammetrization40 - and changing
the way a process is organized - external modif~ation4l.Obviously, the clarity of
the Broadbents' analysis may lead us close to the solution of our problems.
But why should one have to Notice ... that this type of theory embodies many of
"
the points from traditional observations, which are hard to handle in open chain
or closed loop theories"? Evidently, the Broadbents were convinced that open chain
o r closed loop theories have failed to acknowledge the possibility that the lower
level can function in the absence of the upper. The importance of this observation
may call for an example.
It has been a well-known fact from, at least, Boyle onwards that the heart will
continue to beat for some time even if it is removed from the body. Modem
understanding has it that the heart in principle specifies its own rhythm: a case
of low-level autonomy. The locus of this 'autonomy' in the 'well-hearted
organism is not just the morphological structure we call 'heart' but the heart in
its normal 'environment', given the inflow of blood, etc. Moreover, a continuing
supply of energy is needed to allow the heart to go on beating, but this energy does
not specifj, the ensuing rhythm: the heart 'taps' its own energy (cf. P.J. Beek,
1987). All this, however, does not preclude the heart from being under the
influence of 'higher' levels, the activity of the n e m u vagus being a case in point:
external forcing on top of existing self-organization.
This being known, there is no need to claim, or investigate, a high-level
'programme' for the specification of heart beats, the maybe trivial, but not
unimportant, argument being that if the heart can take care of its own rhythm,
we are freed from the need to research how that rhythm is externally specified.
In this sense, description and explanation appear to coincide (cf. the General
Discussion in this Volume), but that is not to say that all further research
questions disappear: it will both be needed to investigate how it is that the heart
can specify its own rhythm, given the influx of blood, etc., and to study how
external forcing succeeds in adequately adapting the heart rhythms to the actual
demands.
40 We take Pattee's definition of control an 'changing the rate of a process without a corresponding
freezing out of its configurational degrees of freedom' (1973a. p. 84) to be closely related to our
definition of' external parametrization'.
41 It may be interesting to note that also Turvey (1977) uses the term 'modification' in this sense.
Reed (1986)writes of 'signals'which create 'conditions'.
514 O.G.MeGer,R.C. Wagenaar & A.C.M. Blankendaal
Because our human body consists of many parts, and needs to be controlled
by reason, it may be seen as a clockwork with different cogwheels and
measures. And just as a clockwork is worth nothing unless regulated, so
our human body does not function unless governed by Moderation
(Christine de Pisan, around 1400, quoted from V m n & Draaisma, 1985, 91;
our tmnslation).
42 In hindsight, it is quite amusing to see how classical mechanicism missed the fact that the
escapement brought the clock back into the realm of dynamics. We are presently (continues)
The Hierarchy Debate 515
(continued) unaware of the actual etymology of the term kchoppement; it is tempting, although
probably wrong, to speculate that it was conceived as an escape from the messiness of a non-
mechanical universe.
43 In refemng to 'mechanicism', or 'mechanical modelling', it is the historical tradition of The
Majestic Clockwork which we have in mind. Modem mechanics, or, for that matter,
biomechanics, has little to do with now that it has been recognized that matter is, in fact, much
more sophisticated than seventeenthcentury science wanted to believe.
44 Claiming that mechanicism was the only important creed of seventeenth century scientists is
clearly counterfactual (cf. Webster, 1975; Meijer, 1983). Boyle was inspired by alchemy, but even
Newton, the Great Alchemist, never completely succeeded in unifying alchemy and mechanics
mates, 1964, Dobbs, 1975, Debus, 1977, Westfall, 1983). It was, we argue, the dominance of
mechanicism, rather than its lack of competitors, which was responsible for the failure of early
anti-hierarchical attempts.
516 O.G.Meijer, R.C. Wagenaar & A.C.M. Blankendaal
45 It may be unfair to 'credit' Schelling (cf. Kirchhoff, 1982) with the rise of teleomechanics. Its
root is, in fact, Kantian (cf Lenoir, 1982); we are, however, of the opinion that it was mainly
Schelling's dynamical rejection of mechanics that led to the general popularity of
teleomechanics.
46 It may be interesting to note that both Darwin and Sechenov published their theories after Lotze-
Pfliiger (1863),but before Huxley's disposing of the soul (1870).
The Hieramhy Debate 517
... My major sin against our time is that I do not conceive Nature in a
mechanical, but in a dynamical fashion. If one only could convince me that
nature is nothing but mechanism, I would be converted on the spot; in that
~~
47 In this sense, one may be tempted to agree with, e.g., Leibniz's contention that even Newton's
'pravity' was conceptually empty, although empirically useful (cf. Westfall, 1983). (It was only
Enlightment philosophy which succeeded in hiding the cracks in the foundations of Newton's
temple.) Of course, conceptual 'emptiness' is somewhat tricky; if it would only imply that no
practical models exist to ostensively define the meaning of such concepts, the square root of minus
one were to remain conceptually 'empty' for ever; a more important characteristic may be that
'empty' concepts have no place in a coherent conceptual framework which is generally regarded
as being acceptable.
518 O.G.Meier, R.C. Wugenaar & A.C.M. Blankendaal
case, there would be no doubt that Nature is dead, and every other
philosopher may be right, but not I. Ever since Descartes, all dominant
philosophy has been modelled on this mechanical viewpoint ... . In essence,
its warfare against Naturphilosophie is mechanics fighting dynamics, and
she, therefore, enters an unequal and, from her viewpoint, highly reckless
combat (Friedrich Wilhelm Joseph von Schelling, 1806, quoted from
Kirchhofi 1982,59;our tmnslation)48.
There exist gardens where noticing which flowers are open may tell one the time,
a mode of timing which is definitely less 'straightforward than that of a clock. Let
us assume 'temperature' to be the relevant parameter. At first, the bud swells a
little, to then, rather suddenly, open into a flower, which, eventually, will open up
somewhat more. But a sudden spell of cold weather will cause the flower to close
again - to be exact: when a temperature is reached which is lower than a t the
earlier moment of opening. It is convenient to say that the inflorescence may
switch between bud and flower, two different steady states to which it is attracted
periodically, each switch as such being a topological catastrophe (cf., e.g., Thom,
1973; Zeeman, 1977; Prigogine & Stengers, 1986; Warren, 1988). Switches take
place at specifids values of the relevant variables - values which are different,
depending on the immediate pagt of the system: hysteresis (Figure 5); it is only
outside the domain of hyeteresis that mechanical modelling may be appropriate,
being subsumed under, rather than in control over, dynamics.
What now is crucial in the way an inflorescence is organized when compared
with traditional mechanical modelling? A pendulum, when left alone and
without escapement, will cease to oscillate: the oscillation is said to be positively
damped. The above flowering cycle, however, will - within the limits of the
flowering season - continue to present itself: if one wants to model it as an
oscillation, the oscillation has no damping, or, better still, negative damping. If I
put a ruler on my desk-top so that it protrudes into the open air, a single tap on its
exposed extremity will cause it to oscillate for some time; and then, it stops. In
fact, if it did not stop, I would have to conclude that the oscillating ruler by itself
'taps' energy from the surroundings to counter the to-be-expected damping. The
48 We have often wondered in how far Shelling was influenced by the Newton-who-really-was,
rather than the popularized Enlightment version of Sir Isaac. There is this fascinating possibility
that Newton conceived the world embraced by the love of God the Father, gravity being His
instrument, while the world was redeemed by the sacrifice of God the Son, manifesting Himself in
alchemy (see Dobbe, 1975; but Westfall, 1983).
49 Better still: narrow zones around specific values. During catastrophes 'fluctuations' take place
in these zones.
The Hierarchy Debate 519
It was possibly Watt's 1785 invention of the governor which allowed the steam
engine to metaphorically replace the clock (Arbib, 1982). The machine itself was to
send the steam periodically to one or the other side of the piston: it appeared to be
self-organizing, akin, in that respect, to the human body (cf. Bernard, 1878;
Cannon, 1939). 'Energy' could now appear (Kuhn, 1959), giving birth to
thermodynamics, and Maxwell, after having theorized about governors (18681,
50 For example, in phase transitions in human bimanual oscillations (cf., e.g., Kelso, 1984;
Haken, Kelso & Bunz, 1985).
520 O.G.Meoer,R.C. Wagenaar h A.C.M. Blankendaal
conceived the self-organizing 'vortex atom' (1875). Later, Freud used the steam
engine to foster his model of the mind (Russelman, 19831,while Bateson wanted to
build his science of genetics on the hope that: "If we could make a vortex which
would continue to divide spontaneously, we should - consciousness apart - have a
rude but not uneffective ... model of life" (1924, quoted from Coleman, 1970, p.
310151.
In hindsight, one may wonder why dynamical self-organization did not take
firmer roots in the life sciences before the second half of the twentieth century.
Apparently, it had to wait the further development of a mathematics of non-linear
equations (cf., e.g., PoincarB, 1893;"ham, 1973;May, 1976;Crutchfield et al, 19861,
and the experimental demonstration of physico-chemical phenomena escaping
the second law of thermodynamics (cf. Bonner, 1973; Winfree, 1974;Prigogine &
Stengers, 1985, we used 1986). Even today, however, the New Science has no
generally accepted name52. For obvious historical reasons, we propose to christen
it teleodynamics63.
51 It is amusing to note that, for Coleman, Bateson's long standing rejection of mechanical models
of the chromosome made him a 'romantic conservative', while Cock (1985) defended Bateson by
emphasizing that Bateson had also good reaeons for not accepting the Morgan tradition. To us, it
seems that some rehabilitation of Bateson's dynamical views is due (cf. Weiss, 1971; Pattee,
1973a).
52 E.g., 'catastrophe theory', 'dissipative structure theory', 'modem dynamics', 'natural physics',
'non-equilibrium thermodynamics', 'synergetics' (cf. Haken, 1977 & 1983; Iberall, 1972;
Prigogine& Stengers, 1980 & 1986;Thom, 1973;Zeeman, 1977).
63 Our christening the New Science is a christening for-the-time-being: we will have to leave it to
the physicists themselves to reach agreement on the name of their own field.
The Hierarchy Debate 521
thus, the 'coefficients' in the formulae are terms which contain the changing
variables themselves: we have to do with non-linear phenonenab4.
The non-linear formulae specify the states to which the process is attracted, as
well as the road to reach such a state - given, within limits, any set of initial
conditions. In, for example, the case of a limit cycle oscillator, both the state
which the process will reach - the limit cycle - and its way to reach it from the
state specified by the initial conditions, are captured in the formulae, the
empirical counterpart being that one may actually observe the process attain its
limit cycle, or return to it after perturbations (cf. Beek & Beek, 1987).
The boundaries of any system may be blurred, or it may be nested into a larger
system. But in teleodynamical phenomena, even forcing by 'external' changes in
the variables does not necessarily invalidate the existing self-organization of the
system. Environmental temperature may be said to 'control' the inflorescence; in
fact, changes in environmental temperature become changes in temperature
within the system, able to then modify its present steady state. In this important,
'ultimate', sense, then, it is left to self-organization how the inflorescence
behaves, unless, that is, changes are so extreme that all organization is 'frozen
out' and the flower dies.
54 Being neither mathematicians nor physicists ourselves, we had, of course, 8ome trouble in even
reaching a crude understanding of the mathematics involved. We want to thank Arne
Wunderlin, Peter Beek and Wiero Beek for their help. It was through them that we came to
understand that the intuitive way to grasp non-linearity is seeing then when Xi is a solution to the
-
equation involved, nxi - n being a natural number is not, or, at least, not necessarily. For the
mathematics involved, see, e.g., Haken, 1977 & 1983; Kugler, 1986; Warren, 1988; P.J. Beek,
1987; Haken & Wunderlin, 1987; W.J.Beek, 1987; Beek and Beek, 1987; Kugler and Turvey, in
press.
522 O.G.MeCer,R.C. Wagenaar & A.C.M. Blankendaal
found in nature. We are led to consider in all seriousness the traditional image of
the snake eating its own tail as the guiding image for autonomy as self-law and
self-regulation"(Varela, 1979,p. xii).
One may want an external observer to recognize the 'intrinsic' meaning of a
system. This is, in fact, what Francisco Varela tries to do, stressing that "The
way a system is identified and specified through our interactions with it is not
separable from the way its cognitive performance is understood (1979, p. xii,
italics in original). Upon analysing the 'Eigenbehaviour' of systems, however, he
states:
... Every natural process decomposes into structurally stable islands, the
chdodes. The totality of chdodes and the multidimensional syntax which
directs their respective positions, forms a senantic model.
... If one regards a chr6ode C as a word in that multidimensional
language, its meaning @a signification de ce mot] is no other than the
global topology of the associated attractor(s) and that of the catastrophes
they undergo (1973,321; italics in original, our translation).
The present authors' usual way of establishing the time is to pull down the Apple
menu and select 'Alarm clock, after which the time will appear on screen
(Figure 6). In this mode of timing, both self-organization and self-reference are
. i ? d I.CX %. ~ ~ ~ i h i r f i ~ y ] l
ent authors' usual way of establishing the time is to pull down the Apple me
'Alarm clock', after which the time will appear on screen (Figure 7). I n t
ng, both self-organization and self-reference are tricky concepts, Imagine
with 0's and l's, fed into a machine which can be in different 'configuratio:
tricky concepts. Imagine an endless tape with 0's and 1'8, fed into a machine
which can be in different 'configurations' - 'mental states', as Turing used to call
them (cf. Hodges, 1985). It is the 0's and 1's on the tape which modify the
configuration; it is the machine-in-that-configuration which, in turn, changes
the 0's and 1's: in this sense, the controller and the controlled appear to be one
and the ~ame55.
The actual behaviour of artificial Turing machines comes close to the old
dreams of dualism: the organization of the machine and the meaning of its
operations derive, ultimately, from the creativity of the engineer I programmer
who built / instructed the machine in the first place. If now we want to conceive
natural Turing machines, they must, we argue, have evolved on top of self-
organizing systems.
Hence, our argument reverses the role of classical mechanics: rather than its
representing the necessary bottom-level of our understanding, it represents a
possible top-level.
The 0's and 1's of an artificial Turing machine do not intrinsically stand for
anything at all; they do, however, in a formal sense 'represent' the entities I
values of variables for which they stand. As long as the system is formal, this
'representation' is arbitrary (cf. Pattee, 1979).
Once given the fact, however, that the 0's and 1's represent anything at all, a
strange phenomenon occurs: the system of 0's and 1's is closed upon itself, and a
tangled form of (secondary) self-reference emerges:
... No one had the slightest idea how to do this until A1 and the notion of
data-structures came along. Data-structures may or may not be biologically
or psychologically realistic representations, but they are, if not living,
breathing examples, at least clanking, functioning examples of
representations that can be said in the requisite sense t o understand
themselves (Dennett, 1979,p. 123; italics in original).
56 -
Cf. Turing, 1936 - we used 1965a; Turing, 1950 we used 1963; Von Neumann. 1958 - we used
1976;Dennett, 1979.
The Hierarchy Debate 525
Computer modelling of the mind (cf, e.g., Gardner, 198556), or of motor control
and learning (cf, e.g., Arbib, 1972; Schmidt, 1982), has been criticized from many
sides57. The main issue in the literature has been that computers are formal
devices (Dennett, 1979; Johnson, 1986), but a8 early as 1939 Turing conceived an
informal 'oracle' (we used 1965b, pp. 166-167) to form a non-mechanical part of his
universal machine.
For a natuml machine to model workings of the mind, there is, we argue, first
and foremost this problem: the origin of 'natural' meaning as it came to be
represented by the machine. Unless the problem of the origin of meaning in
biological Turing machines is solved, Turinguesque modelling of human
behaviour, if not question begging, is deemed to rely on superior authority. It is
our contention that the heritage of superior authority, has, in fact, lured quite a
number of theorists into saving meaning by putting a little programmer in the
head.
56 We have also relied on: Boden, 1977; Bolter, 1984; Davis, 1965; Dennett, 1979; Feigenbum &
Feldman, 1963; Gardner, 1983; Hodges, 1985; Hofstadter, 1979 & 1986; Johnson, 1986; McCorduck,
1979; Turing, 1963, 1965a & b; Von Neumann, 1976. Since the general history of this type of
modelling has been excellently described in these books, we will largely omit historiography From
this Section.
57 E.g., Bolter, 1984; Carello et al, 1984; Dreyfus, 1972; Gauld & Shotter, 1977; Searle, 1984, Shotter,
1975.
526 O.G.Meuer, R.C.Wagenaar & A.C.M. Blankendaal
Possibly the first sign of this was the rechristening of Turing's 1950 'note pad as
a 'memory', as if today's machines can 'remember', very much a high level
phenomenon. But artificial computers, as we know them today, just change their
'note pad without having any overview over the meaning of these changes. In
this sense, we disagree with Van Wieringen's (1988) critique of Skinner's and
Gibson's dismissing 'control' by memory. Skinner and Gibson state that
organisms just change over history, which is exactly what today's computers are
able to do. One cannot take out a 'loan of intelligence' (Dennett's 1979 term), to
then 'explain' it away as a high level phenomenon - loans of intelligence are to be
repaid as low level phenomena.
A more serious mishap occurred in the field of movement science. At least
since Bernstein, 'commands' are given by high authority in the brain, and today,
the use of 'commands' in cognitive modelling is widespread (e.g., Kilmer,
McCulloch & Blum, 1969;Arbib, 1972;Greene, 1972;Dennett, 1979;Schmidt, 1988).
Relying on 'commands', however, may be misleading. In a 'specificational' (cf.
Turvey & Kugler, 1984) sense, it supposes the ability to adequately predict the
formal results of their execution. But the gist of Turing's 1936 argument was to
prove that there can be no mechanical process for determining whether the same
process "ever prints 0" (cited from 1966a, p. 148). In an 'indicational' sense,
coming closer to the normal human use of the term, issuing 'commands'
presupposes, we argue, knowledge of 'what it is like' (Nagel's 1974 phrasing) to
receive a command - another high level phenomenon. Hence, not only
specificational use of the term 'commands' relies on superior authority but also
indicational use begs the question.
The hard way to rescue meaning is to conceive the evolution of dynamically
contingent biological meaning in natural machines. In cellular automata theory
we have already been informed that Turing machines may spontaneously
originate and start to reduplicate (cf. Gardner, 1983,pp. 214-257),but, as yet, that
does not sufliciently highlight the origin of represented biological meaning.
Bateson's problem
Ironically, Bateson's conception of 'life' (apart from consciousness) as a vortex,
capable of reduplication, was ridiculed from the start. 'Reduplication' came to be
explained by the static architecture of DNA, and indeed, the possibility to
reduplicate is no intrinsic property of teleodynamical systems - although logic
compels us to believe that at least some teleodynamical systems must have had
the potential to reduplicate, allowing the emergence of 'life'. A dynamical variant
of Crick's 'wobble hypothesis' (196658)may help to conceive how this happened.
Footnote on next page.
The Hierarchy Debate 527
Suppose (cf., e.g., Eigen & Winkler, 1981) there once co-existed an autocatalytic
cycle of proteins, and a similar cycle of nucleic acids; as soon as any dynamically
contingent code between them would stimulate the production of proteins which
would then stimulate the production of nucleic acids, the code would 'stick', and
quite soon rule out all other physical possibilities.
The point of interest would be the following: given the underlying dynamics, the
number of possible codes is limited but it is larger than one; there is, then, an
element of stochasticity in the actual choice for one code. We will coin such
'wobbling' as dynamical contingence59. The number of possible 'wobbles' is
constrained by the underlying dynamics, but only dynamically contingent actual
'wobbles' will determine the future of the system. In this way, we argue, 'life' has
emerged within the constraints of the underlying dynamics, contingent
nevertheless, and open to natural selection to sieve the 'survivors' out of the varied
pool that actually presented itself.
DNA must have derived its informational content - its 'meaning' - from the
place in the original system of the proteins for which it contained
'representations'. The role of mutations in the phylogeny of the organism shows
the possibility of DNA to modify, indeed, the organization of the whole system;
once a representation exists, nature can start 'working' on that representation,
thereby changing the fate of the whole substrate. Again, the results of these
changes are 'dynamically contingent'. 'Novelty', then, is not an intrinsic aspect of
teleodynamics.
As soon as 'life' had emerged, the original 'persistence of pattern' acquired the
properties of 'survival', a certain type of error presenting itself. No one can
reasonably blame a vortex for failing to remain in existence. Variation and
selection, however, blind as they may be, allow for a certain 'blameability':
mutations may give rise to some organisms being 'better adapted to
environmental circumstances than other ones. It is for this very reason - the
58 To 'wobble' (of a person or animal): 'Tomove from side to side unsteadily or with uncertain
direction" ("he Shorter Oxford English Dictionary, Vol 2, 1986, p. 2563). We have also been
inspired by the style of reasoning in, e.g., Jacob, 1970 & 1982; Monod, 1970; and Eigen & Winkler,
1981.
59 It is important to emphasize that our 'dynamical contingency' goes beyond classical mutation
theory in its relying upon underlying self-organization. It is, however, in accordance with
modem developments in genetics. Of course, our 'dynamical wobble hypothesis' used Cricks
reasoning as a stepping stone only. 'Dynamically contingent' changes are constrained by the
underlying dynamics, appear to leave open at least some wobble, and are subjected to some form of
selection. As far as self-organization is concerned, the locus of 'dynamical contingency' reveals
a relatively independent degree of freedom; in the study of 'change', therefore, it should be
consideredto be 'external' to the self-organizingsystem.
528 G.Meoer, R.C. Wagenaar & A.C.M. Blankendaal
0.
to the mating cry, which is centrally represented in order to allow the organism to
actively 'pick up' the meaningful information from among the meaningless noise
in its environment.
Some sort of 'wobble hypothesis' of the nervous system must be invoked in order to
understand how this new situation came about. For the purpose of the present
argument, we will delegate the evolution of animals in general to general
evolutionary theory, pausing, after that Great Leap Forward, at the point where
the next watershed has been passed. Suppose an antelope is grazing in the
savanna. All of a sudden 'lion-ness' is signalled in, say, its mesencephalon. A
decent mesencephalon would switch the general state of the animal from
'feeding' to 'flight'. Although the animal cannot know exactly what will happen if
it runs away, or stays, with the help of a mesencephalic 'representation' the lion
is (instinctively)'anticipated' as 'danger' before it even starts to attack.
Since all sorts of representations of 'lion-ness' (or 'danger') are conceivable, this
mesencephalic representation must be dynamically contingent. Contingent
representations, calling for Turinguesque modelling, turn the deer's
mesencephalon into an anvil for the environment to hammer upon, a process the
results of which we used to call 'conditioning'. Its emergence must have been
allowed by the lower levels; once it has emerged, however, these levels may, and
will, be changed by it.
Perkel and Bullocks 'code', we propose, must be seen as a nervous
'representation' of the underlying organization of normal ongoing movement61.
First of all, we have to conclude that the organism is able to adequately fly without
this representation; this ability has, at least in part, to be understood as self-
organization (Figure 7a). Moreover, environmental information, such as
concentrations of chemical substances in the air, is already able to force the
organization of movement (Figure 7b). Since Perkel and Bullocks 'code' appears
to be meaningless, we believe that it resultad from a mutation doubling the lower
'circuitry' (Figure 74. The nervous system has the capacity to 'rewire' itself
through laws of (contingent) effect (cf. Arbib C SzentAgothai, 1974; Paillard, 1986).
Well now, for the 'code' under discussion, all this niakes sense only, we argue, if
the code allows novel perceptual input to meaninfilly force the underlying
dynamics (Figure 7d). As soon as this is the case, the evolutionary advantages of
61 Although clearly different, our understanding comes close to important aspects of views as
divergent as those of Calis, 19&; Fodor & Pylyshyn, 1981; Neisser, 1976; Reed, Kugler & Shaw,
1985;Shepard, 1984;Turvey,1977: and Whiting&Den Brinker, 1982.
530 O.G.Meijer,R.C. Wagenaar & A.C.M. Blankendaal
environment
1 forcing
the 'code' are tremendous (Figure 7d): it allows the organism to adequately cope
with novel environmental information.
The hovering locust, upon meeting a mate, is now finally able to fly along with
the mate, rather than continue hovering. The locust is able to 'learn'.
Our Turinguesque representation of the underlying organization of movement
~ organism-environment systbm to let the animal do what it is doing
a i 1 0 ~the
while secondary self-organizatwn allows the system to prepare for new
meaningful variables that will add novel adequate forcing functions. This
capacity, we argue, is presently best understood in Turinguesque modelling. In
fact, it is our contention that Turinguesque modelling is essential for the
understanding of being pre-adapted to novelty.
Werner's highllevel 'representations' of behavioural modes. and Johannesma's
mesencephalic 'representation' of the mating cry, apparently serve similar
functions: their existence leads to claiming that actions may be understood in the
context of global action modes which consist, in themselves, of meaningful sub-
actions (cf. Tinbergen, 1950; Reed, 1988), and that the animal-environment system
may be prepared for relevant perceptual control of action through the construction
of representations of meaningful perceptual occurrences. Of course, these
representations are not identical to what they represent; nor are they capable of
'control' by themselves; they just are the instruments allowing for environmental
forcing which allows for novelty.
A new level has, again, emerged: the level of 'action / perception' implying the
appearance of terms such as 'satisfaction', 'effect', 'failure', 'mistakes' (cf. Ben-
Zeev, 1984, etc., based now on Laws of Effect (Dennett, 1979) which are somewhat
less blind than those of phylogeny, but certainly not as directly purposive as
human planning, programming, or commanding. Of course, 'satisfaction' is not
incompatible in principle with 'persistence of pattern' (dynamics), or 'survival'
(life); nor can it preclude the emergence of cultural 'rules'. It is just the case that
'satisfaction', 'effect', 'failure', 'mistakes', etc., mark the new level, the dual
unfolding of action and perception calling for new laws to be expected.
Given the contingencies of the environment-organism relationship, we, for
ourselves, expect 'heterarchical control' to be the rule in ongoing motor control,
whereas quick environmental modification of ongoing action, may very well call
for 'hierarchical control', as is the case in jumping back from a hot stove
('reflexes'). Of course, well-learned perceptual control of action could be
completely in the realm of teleodynamics. None of these issues, however, are
decidable on theoretical grounds only. The presence of forcing in the organization
of action, then, is an empirical problem rather than theoretical.
532 0.G.Mewer. R.C. Wagenaar h A.C.M. Blankendaal
The real problem, when asking a computer the time, comes when the answer
appears dubitable. We could, of course, type: "Are you sure?" but even if the
machine were programmed to understand that question, it would probably repeat
the procedure and come up with the same answer. Artificial comput8r8, as far as
technology has produced them, are unable to identi@2, unable, that is, to think in
terms such as: 'What on earth may have led them to that question? Is it dark
already and, therefore, bound to be later than Z thought it to be?" The system
would run berserk.
Shotter (1975) described three modes of analysing human existence: the
mechanical, the organismic, and the personal ones. So far, we have only
modelled the organismic mode, and left a vacancy for understanding 'persons'.
Inspired by Russian 'physiology of activity' and Sartre's (1943) 'phenomenological
ontology'63, we like to believe that the hall-mark of understanding persons is,
recognizing the importance of their social context. Consciousness, as an
instrument of control, is the ability to identify with the view of the Other - Sartre's
L'autrui. It is 'through the eyes of the relevant other' that we come to see
ourselves, constructing a narmtiue about our person a8 if experienced from the
62 Searle's (1984) well-known 'Chinese room' misses this aspect with breathtaking
marksmanship. There is this person, surrounded by Chinese characters, unable to understand
their meaning. A manual instructs how to respond when similar characters are brought in: by
selecting certain ones and sending them out in a particular order. The bystanders are astonished
by the room's 'ability' to correctly answer Chinese questions while inside the room only the
-
instructions of a manual are followed. The Turing test in passed, but and that is the point Searle
might have wanted to make - we see two apparently irreconcilable systems of meaning: the room-
worker is unable to idenlib with the bystanders, unable to 'share' the meaning they perceive.
63 It was only through discussions with Huib Laoren de Jong, and reading Kozulin (1977; 1984 &
19@8), that we came to underatand the historical connections of works we had read 80 far in
isolation. Apparently, Janet - the founding father of French psychiatry - can be taken as the source
(cf. Riimke, 1969, pp. 170-176). Janet's 'hierarchical psychiatry' was translated into Russian,
- - -
leading to a tradition from Vygotsky (cf. 1926 we used 1979; 1931 we used 1978; 1934 we used
-
1962) to Luria (cf. 1976 & 19791, Leontiew & Zaporozhets (1960) d Wertech, 1981; 1985a & b;
-
Zinchenko, 1939 we used 1984. Probably also Bematein's biology of activity' (1966; we used
Pickenhain & Schnabel, 1976, pp. 218-233 & 242-243) must be understood in this tradition. Its
-
French root is revealed in, e.g., Sartre's work (19431, while Meads views (1934 we used 1974)
reveal how much of a world wide tradition this really is.
The Hierarchy Debate 533
outside, the crucial moment probably being that first time we say 'I' in order to
create the narrative with which we are identi-fying64. We do identify, of course,
with several relevant others, and their identities become blurred, the remaining
continuous narrative of myself about myself turning into a well-formed 'I'. The
m i n d s I (Hofstadter & Dennett, 1981).
64 For the differenceabetweenthisview and Piapt's ideas, see Kozulin, 1977 C 1986.
66 It appears that even DNA's role in development can be 'shared by organisms': there exist fish
where, upon the death of a dominant male, a dominant female taken over his role, while actually
changingber'gender.
66 We don't think that 'neuronese' Nan de Grind, 1987) is the relevant level of understanding
this language; it would rather have to be eome higher level form of 'meta-neuronese'.
534 O.G.Meijer, R.C. Wagenaar & A.C.M. Blankendaal
ourselves in the past - penetrable for consciousness - that we can imagine what
may happen in the future.
In a playing and manipulating species, where language was developing, the
socially looking at our own actions must have caused the miracle to occur: all of a
sudden the reasons could emerge for doing this or that, allowing the reversal of
past and future (Figure 8; cf. Bernstein 1966, we used Pickenhain & Schnabel,
1975,pp. 218-233& 242-243).Early Man in the savanna, running away for a lion,
and seeing himself doing so, claimed 'reasons' for running away, rather than
'causes'. Dennett (1979)has shown how much we can only 'understand a good
chess-computer by referring to its 'reasons' for moves, rather than the underlying
machine-language: Turinguesque modelling of language-upon-language allows
us,at least in principle, to understand how the mind could emerge ...
-
My running away for the I ion: I
I My moving Knight b8-c6:
I
a. was caused by a mesen- I a. was caused by 00101
cephalic switch due to auto- I
1000101101 arriving in my
lo1
I
matized danger signalling I
I central processing unit
I
I
I
I
1
I
I b. was chosen because I
I
I
wanted to engage in at-
danger of a possible I
I
tacking his Queen
attack I
I
Meta antelope I
I
Meta chess computer
In order to study motor control and learning, the cooperation of Two New Sciences
is needed - we called them teleodynamics and Turinguesque modelling, but the
terms themselves are irrelevant to the present argument. The two sciences have
been described as being 'complementary' (Pattee, 1979), but we argue that our
evolutionary narrative allows a sharper understanding of their relative
importance. No meaningful model can build itself on top of mechanics, unless
one wants to restore superior authority (e.g., Popper & Eccles, 1977); all natural
modelling, thus, must have an autonomous dynamical system as its foundation.
Representations may, and do, come on top of this; failing to recognize them as
being relatively independent, may lead to the unability to understand how the
representations themselves can, and do, change, thereby changing the fate of the
whole system.
The general image may, for the time being, be sdiiciently clear, but at least two
serious problems must be recognized before reaching a Verdict. The first problem
is that, within the framework of our model, i t is always possible to telescopically
enlarge the dynamical box, so that it includes one, or several, levels of
representation. The legitimacy of this procedures depends on the control question
asked. There is no crucial experiment conceivable to distinguish definitively
between the two sciences: observing any movement sequence XYZ is per definition
dynamically interpretable, since it is impossible to observe impossible dynamics,
And per definition interpretable as the manipulation of representations - since it is
inconceivable to observe a movement sequence which cannot be formally
536 O.G.MeOer,R.C. W a g e m r & A.C.M. Blankendaal
described. Experiments, however, may reveal the relative efficiency of one or the
other type of analysis (cf. Beek & Beek, 1987). These experiments, we argue, will
serve to pinpoint the plausible limits of teleodynamica and the locus of forcing.
We have tried to localize the major watersheds where the division of labour
between the Two New Sciences takes place. The second problem of our model is
that it is desparately in the need of refinement. I t is in principle possible to build
box upon box upon box (6. Fentress, 1986) - in fact, there is no conceivable end to
constructing Russian dolls. Our narrative m a y have shed light upon the major
watersheds; in fact, it also contained descriptions of minor ones. The further
definition, then, of the natural boundaries of boxes still is a n enterprise of
paradigm importance, all the more so, since the appearance of higher levels may
change the lower ones - a possible explanation why headless frogs do so much
better than headless persons. The fact remains, however, that low level autonomy
and higher level representations form the evolutionary sine qua non of our model.
For the time being, we think that this testimony is suflicient for the Jury to do
their job.
5. The verdict
Our conceptual hard core will be that of low level autonomy on top of which
representations allow for the emergence of novel meaningful forcing functions.
Empirical decidability will not be reached before some conceptual hard core is
agreed upon.
The difficulty of just that - revealed in the history of the spinal frog - is part of our
metaphorical heritage: one is at risk of either uncritically invoking superior
authority, or of being so afraid for such authority that representational control
devices become unacceptable.
Iron control
In 1978, Kupfermann and Weiss, following Wiersma and Ikeda's (1964) lead,
defended the concept of 'command neuron': cells whose activity would result in
always the same neuromuscular behaviour. According to Kupfermann and
Weiss a neuron should prove to be both necessary and sufficient in the production
of a certain behaviour, in order to be designated as the 'command neuron' for that
behaviour.
'Necessity' is operationalized as the disappearance of the behaviour upon
(reversible) incapacitation of the neuron; 'sufliciency' is supposed to arise if
artificial imitation of its firing pattern leads to the reproduction of that behaviour.
Kupfermann and Weiss, referring to, e.g., the giant 'escape fibres' in crayfish
(Larimer et al, 19711,and the Mauthner escape neurons in fish (Diamond, 19711,
express themselves to be optimistic about the possibility of localizing different
'command neurons in different species.
The existence of specific electrical results in both neurons and muscle fibres
upon the stimulation of certain neurons cannot be disputed. In fact, they are the
hard core of open loop theories. Bernsteinian reasoning, however, leads to the
conclusion that neuromuscular specificity has very little to do with meaningful
movements or behaviours, actions or postures: given different initial conditions
the same firing of a 'command neuron may lead to different movements,
whereas, in turn, different movements may be 'caused by the same 'command
neurons. Deterministic chains of mechanical control are akin to 'letters' of
Lotze's 'alphabet', allowing the organism, at best, to quickly react with
meaningless jerks. Unless they are built into meaningful loops, on top of
meaningful self-organization,they cry out for superior authority.
Similar reasoning, we argue, is valid for the demise of strict open loop motor
control (e.g., Lashley, 1917).
538 O.G.Meger, R.C. Wagenaar & A.C.M. Blankendaal
Take Hill 71
One could, of course, try to save the 'enchanted loom' (Sherrington's 1906 term) by
invoking indicational, rather than specificational commands. This is what
Greene (1972,p. 308)tried to do in his analysis of Bernstein's work:
When a cat turns its head to look at a mouse, the angles of tilting of its head
and flexion and torsion of its neck will tune spinal motor centres in such a
way that its brain has only to command 'jump!' and the jump will be in the
right direction. ... The cat's brain is like a general, who commands 'Take
Hill 7!'without having to specify as many individual movements as there
are soldiers, because low-levelorganization can furnish the details.
Those who were inspired by the digital computer have been particularly sensitive
to the attractiveness of Greene's 'general'. Maybe Gelfand et al (1971) played a
crucial role in this defining the establishment position: peripherally, 'context
conditioned variability' is solved, while centrally, 'indicational commands' are
issued. Schmidt's (1975,more so 1982, cf. also 1988) theory of 'mixed control',
'mixed referring to both open loop and closed loop phenomena, is a case in point.
It is, however, important to note that, within the context of the level of action /
perception, it makes no sense to even speak of indicational 'commands', although
-
modifying signals may be given switching the antelope from grazing to running
away. It is only within the high level language of consciousness, or between
conscious organisms, that indicational commands are meaningful - "Today we
really must finish this paper, or we will never make it!". No doubt, such high
level switches are able to sequentially modify low level states, but there simply is
no way to meaningfully conceive these control signals as 'commands'.
'crucial' evidence for the existence of 'programmes', a fairy tale, again, may be
due:
Always when the air above the village became colder, rain would turn into
snow; snow would fall upon the mountain, the sheer weight of it, at a certain
moment, causing an avalanche; the avalanche, year after year, formed a dam in
the river, broken only in the spring to cause destruction and distress. Is their a
'programming' entity responsible for the distress? The villagers may believe so.
We don't.
Distributed control
In fact, as soon as Greene's (1972) 'commands' are replaced by 'modifying
signals' (6. Broadbent, 1977; Turvey, 1977; Reed, 1986) we have no quarrel
anymore. We have tried to show, however, that the issue is more than just
quibbling over words. Speaking of Greene, it is only fair to point to the fact that the
hierarchical nature of his general is no essential aspect of his theory: according to
Greene the general may become 'soldier', and vice versa (cf. also Turvey, 1977).
Evidently, hierarchical and heterarchical control become, in a way,
interchangeable concepts. The clearest manifestation of such a position, we found
in Arbib and Szentagothai (1974,p. 326):
Two levels, A and B (such as the cortex and the superior colliculus), may be
so intimately interlinked that one cannot even make statements such as
'Level A commands Level B and Level B informs Level A , since it might be
equally valid to say 'Level B commands Level A and Level A informs Level
B'.
Within each box, smaller boxes may always be construed, having either
hierarchical or heterarchical relations amongst each other. In 1947 Pitts and
McCulloch devised a 'distributed model of the superior colliculus'; in 1969 Kilmer,
McCulloch and Blum conceived a distributed "model of the vertebrate command
system" (our italics). McCulloch (see also McCulloch, 1945; McCulloch & Pitts,
1943) evidently saw fit to combine 'heterarchism' with 'commands'. So does Arbib
(cf. 1972 & 1982;Arbib, Iberall & Lyons, 1983;Arbib and SzentAgothai, 1974).Being
a man of all seasons, however, may insufficiently protect one against the muddles
of computer metaphors, rather than apposite Turinguesque modelling.
In 1966 Bernstein (we used Pickenhain & Schnabel, pp. 218-233 & 242-243)
argued contra Pitts and McCulloch (1947)that the study of the nervous system in
isolation is of little relevance to the biology of activity. It is, we argue, through the
interaction between teleodynamical systems and 'environmental' as well as
-
'societal' systems of representations embodied in the nervous system and in the
-
higher languages of the brain that we may come to start to understand motor
control and learning. Ignoring the importance of dynamics, andlor neglecting the
distinction between levels of meaning, may lead back to 'superior authority'. We
fail to understand how Arbib could, with SzentAgothai, be an important
spokesman for heterarchy - in the form of distributed control -, while at the same
time being one of the most generous donator of commands (1972), and even
organizing slide shows in the brain (1972 & 19821, as if the conception of nervous
representations as images, rather than control devices, would not necessarily
involve superior authority.
The Hierarchy Debate 541
The Verdict, of course, is far from just negative. Time and again we are
impressed by the clarity of Schmidt's arguments; we agree that knowledge of
Results is a n important domain for research; time and again we are impressed by
Arbib's emphasizing neurophysiology; we even agree that 'command neurons',
as Kupfermann and Weiss see them, exist, although we would use a different
term. Neglecting the importance of low level autonomy, however, renders motor
approaches at best incomplete, at worst inconsistent. The general 'hierarchy' of
their models leads easily - although not necessarily - to invoking superior
authority. Although we fail to see a sharp conceptual definition of motor
approaches, they may be understood through their historical inspiration (cf. Beek
& Meijer, 1988). Historically, they are in line with Willis, Whytt, and Lotze. A
meaningful, intergrated, approach, then, calls out for a re-orientation:
acknowledgingthe partnership of teleodynamic modelling.
5.2.Action positions
Fixed points and perodic attractors ... generate some ... behavioral
characteristics observed in discrete and rhythmical movements ... . The
topology of an attractor may change abruptly when a key parameter crosses
a bifurcation point - a distinct change to a new form may occur. ... Similar
phenomena abound in nature, including biological motion, from the
transitions in phase observed in simple material (e.g., from solid to liquid to
gas) to the transitions in gait patterns observed in horses (walk to trot to
gallop ...) to transitions in human posture ... (Kelso & Kay, 1984-1985,21-
22).
The fact, then, that living organisms are qua dynamical systems reasonably
adequate, may call for at least some genetics in the study of motor control and
learning (cf. Pickenhain, 1988).
Our contingent representations in the nervous system appear as 'constraints' in
the views of Kugler and Turvey (cf. Kugler & Turvey, 1984;Kugler et al, 1982).In a
ceiling perspective, this is perfectly adequate. Stating that our Two New Sciences
are complementary but inconsistent (Kugler, 1986, p., 471), however, fails to
acknowledge the way the environment may use the anvil, fails to acknowledge
how antelopes in the savanna actually 'learn' - fails to acknowledge 'secondary
self-organization' and 'dynamical contingency'. Sometimes, an animal may
'decide' to attack, rather than running away for a potential adversary. Maybe this
is what a courting male would 'prefer' to do in order to positively condition the
female: the organism-environment interface has come to contain mateability,
attackability, as well as better-to-running-away-for-ability. In order to realize the
first, the animal must have come to learn the distinction between the latter two,
'learning' being the environment's hammering the nervous anvil. Learning as
such, is a floor-phenomenon, and not a ceiling one. Only the effect of learning -
constraining lower levels - may be seen as belonging to the ceiling.
The demise of Pylyshyn's (1984;cf. Fodor and Pylyshyn, 1981) defence of the
intentional stance whenever physics is insufficient, as "a rich man's game of
science" (Kugler, 1986, p. 475; italics in original), is adequate for the workers in
the shaft, but incomplete for the construction of 'a poor man's game of science'.
The Great Proletarian Revolution, however, has provided help, particularly in the
Vygotsky tradition. It is all well to conceive high level language 'symbols'
constmining underlying nervous mechanisms and dynamical self-organization
mates & Kugler, 1984),but, again, one has to wonder how these symbols came
about. During a mountain walk near Our0 Pretro, Brazil, one of the present
authors all of a sudden heard a definite rattling, and found himself running
away. The author had never seen a rattle snake, not even in the movies. He had,
however, read about them in his (very contingent) mother tongue, and had stored
an imaginary narrative in which his preferred relationship with rattle snakes
was specified.
Evidently, contingent societal symbols, as in books, have worked on high level
representations, no matter, in fact, the underlying hard ware (lots of other brains
would have done the same), or the wrapping of the symbols (a German book might
have had the same effect). A consistent incompleteness is emerging. It is our
contention that ew'oying the splendour of Turinguesque modelling, on top of
teleodynamics, opens the way for understanding learning. Failing to do so leaves
us ignorant about the ignorant executive. The executive, however, does not have to
The Hierarchy Debate 545
be fired. He just has to move. His new Suburbia Address, by the way, is
'environment', his Summer Palace 'society'.
Control by information
In fact, as soon as Kugler's (1986) 'constraints' are supplemented with
'representations' we have no quarrel anymore. We have tried to show, however,
that the issue is more than just quibbling over words. And 'supplementing'
brings us to our second major source of inspiration: the Gibsonians. The much
more biological style of reasoning of this group has not only led to the paradigm
experiments of action approaches (cf., e.g., Lee & Lishman, 1975; Warren, 19881,
but also to practical recommendations on how to proceed (cf., e.g., Reed, 1982 &
1988):
Reeds theory of 'action systems' would allow to define the boundaries of the
discrete behavioural 'modes' into which the organism may be switched, thereby
filling one of the major gaps in our model. Some caution, however, is needed.
The organism's being controlled by ecological information snugly fits into our
own theoretical fiamework. But it does not specify how this information interferes
with self-organization, nor - and this is the point we want to make - how the
organism-society interface may, within limits, give rise to apparently mechanical
movements. Humans are not only born with remarkably little cognitive actualities
(rather than potentials), but also with movement skills which have relatively little
to do with the repertoire of the adult. Why would that be the case? Rather than
argue that lack of adult skills in the neonate is the price one has to pay for
cognitive learning, our image is that of evolution carefully nursing the human
potential to actively force dynamical autonomy into mechanical movement. We do,
after all, manipulate objects, work in factories, play the piano, handle
typewriters, dance according to geometrical prescriptions (cf. Van Wieringen,
19881, move like robots in Prague's Laterna Magica, or even engage in one of
Monty Pythons 'Silly Walks'. Quite literally, 'mechanical' movement is 'top of the
world.
546 O.G.Mewer,R.C. Wagenaar & A.C.M. Blankendaal
The human ability to engage in Silly Walks requires, we argue, the presence of
high level (meta-)representations which allow us to imagine ourselves-when-
walking, and to construct a sequential narrative of idiotic inefficiencies. Not that
we ever will,or can, have any exact a prwri knowledge of how our Silly Walk is
going to look like - a really Silly Walk may well take ten years to master - but we do
switch our behavioural modes, seeking Knowledge of Results from the faces of the
audience. Resenting the superior-authority flavour of 'representations', as most
Gibsonians do, may lead to the unability, in principle, to engage in a Silly Walk.
Saving the Silly Walk, then, may have been o u r most important objective in
criticizing action approaches.
The verdict, of course, is far from negative. Time and again we are impressed by
the clarity of Turvey's arguments; we agree that teleodynamics is a prerequisite
for adequately theorizing on motor control and learning; time and again we are
inspired by Reeds tenacity; we even agree that 'kinematic transformations' - as
Saltzman and Kelso see them - are useful, although we would give them a
different place. Neglecting the importance of representations, as well as the
distinctions between levels, however, renders action approaches at least
incomplete, a t worst inconsistent. The general 'anti-hierarchy' of their models,
leads easily,- although not necessarily - to overlooking the importance of
contingent change. Although we fail to see a sharp conceptual defenition of action
approaches, they may be understood through their historical inspiration (cf. Beek
& Meijer, 1988). Historically they are in line with Boyle, Von Haller, and Pfluger.
A meaningful, integrated, approach, then, calls out for a re-orientation:
acknowledgingthe partnership of Turinguesque modelling.
But enough, for some time now I have been forgetting myself, going further
than the goal I had in mind. By the way, if my words may have sounded too
challenging or tattling, remember that Foolishness was speaking, and,
furthermore, a woman! One is reminded of that saying of the Greeks: "A
fool may oRen speak a bit of truth." Or do you think that that has nothing to
do with women?
I see that you are waiting for a final conclusion, but you are way too daft if
you expect that I remember anything at all from everything I said; I have
blurted out so much! The old proverb goes: "I hate the man who does not
forget what has been said in drinking." The new one says: "I hate the man
who does not forget what has been heard while listening." Therefore:
farewell, express your approval, enjoy life, and drink, celebrated servants of
The Hierarchy Debate 547
DA CAPO AL FZNE
ACKNOWLEDGEMENTS
REFERENCES
Boden, M. (1977). Aritificial Intelligence and Natural Man. New York: Basic
Books.
Bolter, J.D. (1984). Turing's Man: Western culture in the computer age. Chapel
Hill, NC: The University of North Carolina Press.
Bonner, J. (1973). Hierarchical control programs in biological development. In
H.H. Pattee (Ed.), Hierarchy Theory: The challenge of complex systems (pp.
49-70).New York George Braziller.
Broadbent, D.E.(1977). Levels, hierarchies, and the locus of control. Quarterly
Journal of Experimental Psychology, 29, 181-201.
Bronowski, J. (1974). The Ascent of Man. London: British Broadcasting
Corporation; originally 1973.
Burrow, J.W. (1968).Evolution and Society. Cambridge: At the University Press.
Calis, G. (1984).Concerning Gibson's 'on the face of it': Immediate perception
and single-glanceface recognition. Acta Psychologica. 55,195-214.
Cannon, W.B. (1939).The Wisdom of the Body. New York: Norton.
Carello, C., Turvey, M.T., Kugler, P.N. & Shaw, R.E. (1984). Inadequacies of the
computer metaphor. In M. Gazzaniga (Ed.), Handbook of Cognitive
Neuroscience (pp. 229-248).New York: Plenum.
Carmichael, L. (1927). Robert Whytt: A contribution to the history of physiological
psychology. Psychological Review, 34, 287-304.
Chalmers, M.P. (1900). Thomas Henry Hwcley: A sketch of his life and work.
London: G.P. Putman's Sons.
Clarke, E. & Dewhurst, K. (1972). An Illustrated History of Brain Function.
Berkeley: University of California Press.
Cock, A.G. (1983). William Bateson's rejection and eventual acceptance of
chromosome theory. Annals of Science, 40, 19-53.
Coleman, W. (1970).Bateson and chromosomes: Conservative thought in science.
Centaurus, 15,228-314.
Cranston, M. (1966). John Locke: A biography. London: Longmans, Green and
Co; originally 1957.
Crick, F.H. (1966). Codon-anticodon paring: The wobble hypothesis. J o u m l of
Molecular Biology, 19,548-555.
Crutchiield, J.P., Farmer, J.D., Packard, N.H. & Shaw, R.S. (1986). Chaos.
ScientifK American, 255, 38-49.
Cunningham, A. (1986). Thomas Sydenham: Epidemics and the 'good old cause'.
Paper presented a t the Conference on Seventeenth Century Medicine,
Corpus Christi, Cambridge, September 16-19.
Darwin, Ch. (1859). The Origin of Species by Means of Natural Selection, or the
Preservation of Favoured Races in the Struggle for Life. London: Murray.
Davis, M. (Ed, 1965). The Undecidable: Basic papers on undecidable propositions,
unsolvable problems and computable functions. New York Raven Press.
Dawkins, R. (1976). Hierarchical organisation: A candidate principle for ethology.
In P.P.G. Bateson & R.A. Hinde (Eds.), Growing Points in Ethology (pp. 7-
54).Cambridge: University Press.
Debus, A.G. (1977). The Chemical Philosophy: Paracelsian science and medicine
in the sixteenth and seventeenth centuries. h u o Volumes. New York:
Neale Watson Academic Publications.
550 O.G.Meoer, R.C. Wagemar Q A.C.M. Blankendaal
Deliagina, T.G., Fel'dman, A.G., Gelfand, I.M. & Orlovsky, G.N. (1975). On the
role of central program and afferent inflow in the control of scratching
movements in the cat. Brain Research. 100,297-313.
Dennett, D.C. (1979). Bminstorms: Philosophical essays on mind and psychology.
Hassocks: Harvester Press.
Dewhurst, K. (1958).Locke and Sydenham on the teaching of anatomy. Medical
History, 2, 1-12.
Dewhurst, K. (1963a).An Oxford Medical Quartet: Sydenham, Willis, Locke, and
Lower. British Medical Journal, ii, 857-860.
Dewhurst, K. (1963b). John Locke, Physician and Philosopher: A medical
biography. London: The Wellcome Historical Medical Library.
Dewhurst, K. (1964). T h o r n s Willis as a Physician. Los Angeles: University of
California.
Dewhurst, K. (1966). Dr. Thomas Sydenham. His Life and Original Writings.
Berkeley, CA: University of California Press.
Dewhuret, I& (1980). Thomas Willis's Oxford Lectures. Oxford: Sanford
Publications.
Dewhurst, K. (1981). Willis's Oxford C a s e b k (1650-52). Oxford Sanford
Publications.
Diamond, J. (1971).The Mauthner cell. In W.S. Hoar & D.J. Randall (Eds), Fish
Physiology. Vol 5: Sensory systems and electric organs (265-346).New York
Academic Press.
DiGregorio, M.A. (1984). T.H. Huzley's Place in Natuml Science. New Haven,
C N Yale University Press.
Dijksterhuis, E.J. (1980). De Mechanisering van het Wereldbeeld. Amsterdam;
Meulenhof; originally 1950.
Dobbs, B.J.T. (1975). The Foundations of Newton's Alchemy, or: "The hunting of
the greene lyon". Cambridge: University Press.
Dreyfus, H.L. (1972).What Computers Can't Do. New York: Harper & Row.
Du Bois-Reymond, E. (1910). Nachruf a d ' Pfluger. Berliner klinische
Wochenschrift, 47, 658-659.
Duintjer, O.D. (1977).Rondom Regels: Wosgerige gedachten omtrent regel-geleid
gedmg. Meppel: Boom.
Dullemeijer, P. (1974). Concepts and Apprwches in Animal Morphology. Assen:
Van Gorcum.
Eccles, J.C. (1953). The Neurophysiological Basis of Mind. Oxford: Clarendon
Press.
Eigen, M. & Winkler, R. (1981).Laws of the Game. London: Allen Lane; originally
1975.
Erasmus, D. (1969). Lof der Zotheid. Utrecht; Het Spectrum; originally Laus
Stultitiae, 1515.
Fearing, F. (1970). Refla Action: A study in the history of physiological
psychology. Cambridge, MA:MI" Press; originally 1930.
Feigenbaum, E.A. & J. Feldman (Eds, 1963). Computers and Thought. New York:
McGmw-Hill.
Feindel, W. (Ed., 1965). Thomas Willis. The Anatomy of the Brain and Nerves.
Montreal: McGill University Press; originally 1664.
The Hierarchy Debate 551
John Wiley.
Kuhn, T.S. (1959). Conservation of energy as an example o f simultaneous
discovery. In M. Clagett (Ed.), Critical Problems in te History of Science
(pp. 321 -356). Madison, WI:University of Wisconsin Press.
Kupfermann, I. & Weiss, K.R. (1978). The command neuron concept. The
Behavioral and Brain Sciences, 1. 3-10; for peer commentary, see 10-39.
LaMettrie, J.O. (1748).L’Homme Machine. Leyde: E. Luzac Fils.
Larimer, J.L., Eggleston, A.G., Masukawa, L.M. & Kennedy, D. (1971). The
different connections and motor outputs of lateral and medial giant fibres
in the crayfish. Journal of Experimental Biology, 54, 391-402.
Lashley, K.S.(1917). The accuracy of movement in the absence of excitation from
the moving organ. The American Journal of Physiology, 43,169-194.
Lee, D.N. & Lishman, J.R. (1975). The optic flow field The foundation of vision.
Philosophical fiansactwm of the Royal Society of London. B, 290,169-179.
Lee, D.N. & Reddish, P.E. (1981). Plummeting gannets: A paradigm of ecological
optics. Nature, 293,293-294.
Lee, D.N. & Young, D.S. (1985). Visual timing of interceptive action. In D.J.
Ingle, M. Jeannerod & D.N. Lee (Eds.), Brain Mechanisms and Spatial
Vision (pp. 1-30).Dordrecht: Martinus Nijhoff.
Lenoir, T. (1982). The Strategy of Life: Teleology and mechanics in nineteenth
century German biology. Dordrecht: D. Reidel.
Leontiew, A.N. & Zaporozhets, A.V. (1960). Rehabilitation of Hand Function.
London: Pergamon Press.
Locke, J. (not dated). Medical notebooks. Bodleian Library, Oxford, MSS Locke
e.4.
Long, P. (1959). A Summary Catalogue of the Lovelace Collection of the Papers of
John Locke. Oxford:University Press.
Lotze, R.H. (1842). Leben und LebenskraR. In Wagner, R. (Ed.), Handwdrterbuch
der Physiologie, Band 1 (pp. 9-58).Braunschweig: F. Bieweg und Sohn.
Lotze, R.H. (1844). Instinct. In Wagner, R. (Ed.), Handwdrterbuch der
Physiologie, Band 2 (pp. 191-209).Braunschweig: F. Bieweg und S o b .
Lotze, R.H. (1846). Seele und Seeleleben. In Wagner, R. (Ed.), Handwdrterbuch
der Physwlogie, Band 3 (pp. 142-269).Braunschweig: F. Bieweg und Soh.
Lotze, R.H. (1851). Allgemeine Physwlogie des karperlichen Lebens. Leipzig:
Weidman’sBuchhandlung.
Lotze, R.H. (1852). Meduinische Psychologie oder Physiologie der Seele. Leipzig:
Weidman’sBuchhandlung.
Lotze, R.H. (1856-1864).Mikmkosmus, 3 Blinder. Leipzig: Hinzel.
Lotze, R.H. (1891). Recension von Eduard Pfliiger, die sensorischen Functionen
des Ruckenmarks der Wirbelthiere nebst einer neuen Lehre tiber die
Leitungsgesetze der Reflexionen. In Pfeiffer, D. (Ed.), Kleine Schriften von
Hermann Lotze, Band 3 (191-209).Leipzig S.Hirzel; originally 1853.
Lovejoy, A.O. (1973). The Great Chain of Being. cambridge, MA: Harvard
University Press.; first edition 1936.
Luria, A. (1976). Cognitive Development. Cambridge, MA: Harvard University
Press.
Luria, A. (1979). The Making of Mind. Cambridge, MA: Harvard University
The Hierarchy Debate 555
Macey, S.L. (1980). Clocks and the Cosmos. Hamden, CN: Archon Books.
Magill, R.A. (1988). Activity during the post-knowledge of results interval can
benefit motor skill learning. This Volume.
Manning, H.E. (1871). What is the relation of the will to thought? Papers of the
Metaphysical Society, 1 (for 1869-1874), jan. 11,1871,l-13;Manchester
College Library, Oxford.
Maxwell, J.C. (1868).On governors. Proceedings of the Royal Society of London, 16,
270-283.
Maxwell, J.C. (1875).Atom. In Encyclopedia Brittanica, 9th Edition, V o l 3 (36-49).
Edinburgh.
May, R.M. (1976).Simple mathematical models with very complicated dynamics.
Nature, 263, 459-467.
McCorduck, P., (1979).Machines Who Think. San Francisco: W.H. Freeman.
McCulloch, W.S. (1945). A heterarchy of values determined by the topology of
nervous nets. Bulletin of Mathematical Biophysics, 7, 89-93.
McCulloch, W.S. & Pitts, W. (1943). A logical calculus of the ideas immanent in
nervous activity. Bulletin of Mathematical Biophysics, 5, 115-133.
Mead, G.H. (1974). Mind, Self and Society. Chicago: University of Chicago Press;
originally 1934.
Mecacci, L. (1979). Brain and History: The relationship between neurophysiology
and psychology in Soviet research. New York: BrunnerDfazel; original Italian
edition 1977.
Meijer, O.G. (1983). Het Verboden Denkraam. Two Volumes. Amsterdam: Free
University.
Meyer, A. & Hierons, R. (1965).On Thomas Willis's concepts of neurophysiology.
Medical History, 9 1-15 & 142-155.
Minsky, M. & Papert, S. (1972).Artificial Intelligence. Arificial Intelligence Memo,
252, Cambridge, MA: Artificial Intelligence Laboratory, MIT.
Monod, J. (1970). Le Hazard et la Nkcessitk. Paris: Editions du Seuil.
Musee dHistoire de NeuchAtel (not dated). Les Automates des Jacquets Droz.
Nagel, T. (1974). What is it like to be a bet? Philosophical Review, 83,435-451.
Neisser, U. (1976). Cognition and Reality. San Francisco: Freeman.
Newell, A. (1973). You can't play 20 questions with nature and win: Projective
comments on the papers of this Symposium. In W.G. Chase (Ed.), Visual
Information Processing (pp. 283-308). New York: Academic Press.
Nussbaum, M. (1909).E.F. W. Pflilger als Naturforscher. Bonn: Martin Hayer.
Ospovat, D. (1981). The Development of Darwin's Theory: Natzrral history, natural
theology & natural selection, 1838-1859. Cambridge: University Press.
Paillard, J. (1977). La machine organisbe et la machine organisante. Revue de
1'Education Physique, 17, 19-48.
Paillard, J. (1986). Development and acquisition of motor skills: A challenging
prospect for neuroscience. In M.G. Wade & H.T.A. Whiting (Eds.), Motor
Development in Children: Aspects of coordination and control (p. 41 5-441).
Dordrecht: Martinus Nijhoff.
Pattee, H.H. (1973a). The physical basis and origin of hierarchical control. In H.H.
Pattee (Ed.), Hierarchy Theory: The challenge of complex systems (pp. 71-108).
556 0.G.MeGer. R.C. Wagenaar & A.C.M. Blankendaal
originally 1863.
Shaw, R.E. & McIntyre, M. (1974).Shaw-McIntyre discussion. In W.B. Weimer &
D.S. Palermo (Eds.), Cognition and the Symbolic Processes (pp. 363-366).
Hillsdale, N J Erlbaum.
Shepard, R.N. (1984). Ecological constraints on internal representation: Resonant
kinematics of perceiving, imagining, thinking, and dreaming.
Psychological Review, 91,417-447.
Sherrington, C.S. (1906). Zntegmtiue Action of the Nervous System. New Haven,
C N Yale University Press.
Shotter, J.(1975).Images of Man in Psychological Research. London: Methuen.
Spencer, H. (1972).The social organism. In J.D.Y. Peel (Ed.), Herbert Spencer on
Social Evolution (pp. 53-72). Chicago: University of Chicago Press;
originally 1860.
Siidhoff,K. (1922).Albrecht Huller. Von den empfrndlichen und reizbaren Teilen
&s menschlichen K&irpers. Leipzig: Johann Ambrosius Barth; presented
in 1752,originally published in 1753.
Tamboer, J.W.I. (1988). Images of the body underlying concepts of action. This
Volume.
The Shorter Oxford English Dictionary, 2 Vols (1986).Oxford Clarendon Press.
Thom, R. (1973). Stabilitk Strwturelle et MorphogdnAse: Essai d'une thbrie
gkndrale des mcdhles. Reading, MA: W.A. Benjamin; originally 1972.
Thomson, D.W. (1942). On Growth and Form. London: Cambridge University
Press; originally 1917.
Thorndike, L. (1923-1958).A History of Magic and Experimental Science, 8 Vols.
New York Columbia University Press.
Tinbergen, N. (1950). The hierarchical organization of nervous mechanisms
underlying instinctive behaviour. Symposia of the Society for Experimental
Bwlogy, 4, Physiological Mechanisms in Animal Behauwur, 305-312.
Turing, A. (1963). Computing machinery and intelligence. In E.A. Feigenbaum &
J. Feldman (Eds.), Computers and Thought (pp. 11-35). New York
McGraw-Hill; originally 1950.
Turing, A. (1965a). On computable numbers, with an appplication to the
Entscheidungsproblem. In M. Davis (Ed.), The Undecidable: Basic papers
on undecidable propositions, unsolvable problems and computable
functions (pp. 115-154).New York Raven Press; originally 1936.
Turing, A. (196513). Systems of logic based on ordinals. In M. Davis (Ed.), The
Undecidable: Basic papers on undecidable propositions, unsolvable
problems and computable functions (pp. 154-222).New York Raven Press;
originally 1939.
Turvey, M.T. (1977).Preliminaries to a theory of action with reference to vision. In
R. Shaw & G. Bransford (Eds.), Perceiving, Acting, and Knowing: Toward
an ecological psychoZogy (211-265).Hillsdale, NJ: Erlbaum.
Turvey, M.T., Shaw, R.E. & Mace, W. (1978). Issues in the theory of action:
Degrees of freedom, coordinative structures and coalitions. In J. Requin
(Ed.), Attention and Performance VZZ (pp. 557-595). Hillsdale, N J
Erlbaum.
Turvey, M.T. & Kugler, P.N. (1984).A n ecological approach to perception and
The Hierarchy Debate 559
Fischer.
Verworn, M. (1917). Bwlogische Richtlinien der stautliche Organisation.
Naturwissemchaflliche Anregungen fiir die politische neuorientierung
Deutschlancls. Jena: Gustav Fischer.
Vesalius, A. (1975). De Humani Corporis Fabrica. Nieuwendijk: De Forel;
originally 1543.
Von Haller, A. (1777).Bibliotheca Anatomica, Band 2. Hildesheim: Georg Olms.
Von Haller, A. (1778). Usong, Carlsruhe: Christian Gottlieb Schmieder;
originally 1771.
Von Haller, A. (1779).Fabius und Cuto. Carlsruhe: Christian Gottlieb Schmieder;
originally 1774.
Von Haller, A. (1966).First Lines of Physiology. New York Johnson Reprint
Corporation; first translated in 1786.
Von Neumann, J. (1976).The Computer and the Brain. New Haven, C N Yale
University Press; originally 1958.
Vroon, P. & Draaisma, D. (1985).De Mens als Metafmr. Baarn: Ambo.
Vygotsky, L. (1962). Thought and Language. Cambridge, IMA: MIT-Press;
originally 1934.
Vygotsky, L. (1978). Mind in Society: The development of higher psychological
processes. Cambridge, MA: Harvard University Press; originally 1931.
Vygotsky, L. (1979).Consciousness as a problem of psychology of behavior. Soviet
r n , originally 1925.
P ~ ~ h l ~17,535;
Wadman, W.J., Denier van der a n , J.J., Geuze, R.H. & Mol, C.R. (1979).Control
of fast goal-directed arm movements. Journal of Human Movement
Studies, 5, 3-17.
Wagenaar, R.C (1985).De massa-veer-metafoor: Een kritische beschouwing van
dynamische bewegingssturingstheorieh. Typescript, Interfaculty of
Human Movement Science and Education, Free University, Amsterdam.
Warren, W.H. (1988).Action modes and laws of control for the visual guidance of
action. This Volume.
Webster, C. (1975).The Great Instauration: Science, medicine and reform, 1626-
1660.London; Duckworth.
Weinding, P. (1981).Theories of the cell state in Imperial Germany. In Ch.
560 O.G.Mecer, R.C. Wagenaar & A.C.M. Blankendaal
P.0.Box 7161,
Robert C. Wagenaar
Department for Ph sical Therapy,
University Hospitar
Free University, P.0.Box 7161,
1007 MC Amsterdam,
4, 6, 7, 20, 22, 23,
Author Index 212
Baken, R.J. 24, 33, 53, 79,160,
62 177, 243, 254, 345,
Baker, K.E. 347,466,470,474,
292 479,491,492,493,
Ballreich, R. 506, 516, 526, 528,
263 532, 534, 537, 538,
Abranavel, E. Bandura, A. 540,542,543
385,387 113,395 Best, D.
Adamovich, S.V. Bard, C. 440, 447, 453, 454,
29 25,37 458
Adams, J.A. Barret, W. Bilodeau, E.A.
3,159,190,232,248, 445 232
290, 315, 316,317, Basmdian, J.V. Bilodeau, I. McD.
325,333 248 333
Melt, Z. Bates, E. Birch, T.
505 68 495,496,497
Allard, F. Bateson, B. Bischof. N.
291 508 408,409
Anderson, J.R Bateson, W. Bishop,Y.
289,291,293,301, 177, 508, 520, 526, 332
420 543 Bizzi, E.
Annett, J. Battig, W.F. 9,159
316,485 292,301,302 Blantz, H.
Anochin, P.K Baum, E. 328
466,467,470 221 Bobath, B.
Arbib, M A Beck, B.B. 249
48,49,81,510,512, 71,73 Boden, M.
519, 525, 526, 529, Beek, P.J. 510,525,528
540,541 453, 511,513,521, Boesch, E.E.
Aristotle 536,541,542,546 449
445,467,496 Bekkenutte, M.M. Bolter, J.D.
Armstrong, D. 509 515,525
452 Belen’lcii,V.Y. Bonner, J.
Armstrong. T.R. 53,54,55,56 520
29,265 Bennet-Clark, H.C. Boon, L.
Aronfreed, J. 341 162
384,385,386 Benthall, J . Bootama, RJ.
Arshavsky, J.I. 445 394
479 Ben-Zeev,A. Bouisset, S.
Asanuma, H. 531 55,56,58
122,123,124,130 Bergson, H. Boulter, L.R.
Asatryan, D.G. 458 319
506 Berkinblit, M.B. Boyle, R.
Avart, H. 9,28,506 495, 496,497,500,
71 Berard, C. 505, 507, 511, 513,
519 514, 515, 517, 523,
Baart, 1. Bernard, M. 546
443 445,452 Bransford. J.D.
Baddeley, A. Bernstein, N.A 202,206,207
564 Complex Movement Behaviour
12-15,37,88,124-128,136,168-172,
Lambda model 421,452,454
27 -computational, 421,423,434,435
Learning - implementational, 421, 423, 429,
- Adams's theory of, 159, 248, 315- 434,435
317 Low-level autonomy
- and action systems theory, 88, 510,512-514,516,520,528,535,537,
108-114,217-229.253-257, 544-546 538,541,543
-and affordances,80-108,109
-and EMG feedback,247-257 Mass-spring hypothesis
- and contextual interference, 302- 9,10,15,33,125,285,506
308 Mechanisms
- and error detection 126,128-136,139,142,146
mechanisms, 315-335 -'smart', 137
- and guidance, 210-212,335 - vs. functions, 128-136
- and imitation, 381-398 Memory
- and KR, 131, 204, 207-213. 231- 95,102,109,111,131,141,145,146,
244,247-257,315-335 153,166,171,233,302,326,334,411,
- and motor disorders, 247-257 412,420,469,478,526
- and motor systems theory, 88, - codingprocessesin, 158,243
140,253-257,263,264 - theory of skilled, 308
- and postKR activities, 231-244 Mentalism
- and post-performance activities, 92,102,104,444
315-335 Mental practice
- a n d practice, 79,80, 81, 131,140, 105,131,407,409,474,484
201-213,229,301-310,315-318,335 Merleau-Ponty
- and transfer-appropriate 439,451-453,457,458
processing, 201-213 Model - demonstration of a, 327,328,331,
-autonomous stage of, 112-115 335,392,395
-cognitive control of, 289-310 -dynamic, 383-386,393-398
- cognitive stage of, 112-115, 232, - mediating effect of, 396
393 - video-taped, 331,385,395,396
- evolutionary explanation of, 96, Motor-action controversy
528-534 3-38,52-58,75,87-115,124-128,134-
- genetic predetermination of, 479- 155,160-179,253-257,285,340,341,
482 422,439-441,444,457,458,490-494,
- goal-oriented,289,291,309,469 514,538-547
- molecular I neuronal changes -domain argument, 166-168
during, 469 - historical dimension, 158, 162,
-observational, 381-398 176-179
- of complex cyclical actions, 381- -implications, 88-115
398 - levels of analysis argument, 12-
- of gymnastic skills, 326-333 15,37,88,124-128,136,168-172,419,
-perceptual, 109-112 421,423,429,434,435
- phases of, 112-115, 221-229, 232, -nature of, 158-179
393 - strawman argument, 139, 164-
-reinforcement, 96,385 166
Length-tension relationship - strong conceptual argument, 5,
7,9,10,20,27 6,172-176
Levels of analysis Motor commands
578 Complex Movement Behaviour
-langUage,104 323-335
Neurobehavioural units, 100 -type ofactivities during, 234-240
Neurodevelopmental treatment Posture
249-252 - adjustments prior to
Neuron movements, 52-58,100,122,159
97, 98, 121-124, 130, 409, - as function of basic orienting
411,413,470,479,480 system, 59,60
-command, 537,541 -definition, 52
Non-equilibrium thermodynamics - ecological vs. mechanistic
161,172,175,493 definition,47-52
- perceptual control of, 63,64,361,
Observational learning 362
381-398 -upright, 50.51,60,61
Optical flow Practice
346-348,352-357,543 - and motor learning, 79-81, 201-
Organism-environment synergy 213,229,302-310,315-318,335
163,164,172,340,341,362,374,490, -and post-KRinterval, 231-246
492,528,531 - and error detection mechanism,
Oscillators 316-318
-limit cycle, 15,175,521 - and transfer conditions, 201-213,
234-240,280,281
Parkinsonism -blocked, 203-207,302-310
127,129,130,135 -constant, 206-207,243
Perception and action - feedback conditions during, 202,
11,12,72,81,96,160,163,164,189, 231-246,315
190,217,218,317,340,341,344,346- -mental, 105,131,407,409,474,484
348,350-374,393,426-435,444,464, -random, 203-207,302-310
466,531,543 - variability of, 205-207, 234, 241,
- andlearning, 109-112,141 242,244
- and mental acts, 91 Precuing method
- and w Of tools, 73-79,81 264,271-276
- copy theory of, 90 - fixed S-Rpairs, 271
-definition, 64
- direct, 11,64,89-91,97,109,143, Reaction time analysis
152,163,164 25,54-56,58,177,271-277,280
- direct vs. indirect, 91, 163, 164, Reafference principle
444 466,468
- indirect, 74,163,164,444 Readiness potential
- affordances, 341, 343, 349, 362- 473,474,476
374 Recognition
- Skinner's conceptualization of, 290,303-310,470
90 -schema, 316,318
Perceptual trace Reflex 5468,67,131,175,361,388,531
248,315-318 -arc, 466
Post-KRinterval -chain, 58
231,244,316,320,323-335 - stretch, 56
-activities during, 231-244,316 -Wiping, 504-506
- amount of activities during, 234- Rehabilitation
240 123,127,426
- length of, 232-234, 237-244, 320, - EMGfeedhack, 247-257
680 Complex Movement Behaviour
- neurodevelopmental,249-252 Sequencing
-physical, 252,255 55,385,391-393,533,539
Relational body Simulation-type experiments
- image of the, 439,448-458 407,408
Relative timing Skill learning, see learning
261,262,265-271,285,392,393 Ski-simulator
Representations 141,394,395
5,6,10,22-37,89-92,113,115,139- Skills 45-47, 128, 201-203, 231, 232, 243,
148,157,163,165,171,177,178,203, 247,253,290,291,405,406
221-229,248,262-285,289-310,344, - complex,261,262,405
383,393,398,421,444,490-494.528- -cultural, 112,115,291
531,535,537,542,546 -elementary, 262,277,405
-and action approach, 3,5,6,22,89- -functional, 247,249
92,105,110,139-148,163,165,178, - gymnastic, 326-333
219-221,223,227-229,285,344,444, -natural, 112
490494 -restoration of, 254
- and reference systems, 294-301, -taxonomies of, 46
308 Skinner
-evidence for, 2433 89-97,105,111, 112,128,141,150,
-investigationsof, 221-229 526
- oftiming, 27-37,261,262,265-271, Social facilitation
286 384
- verbal articulation of, 229, 294- Soal - and brain, 496-507,515
310 - and heart, 494-500
Response modifications -localization of, 496-507,515
24,25,37,140,261,264,276-280,316- - spinal, 503-507
318 Speech control
-time requirementsof, 276-280 9,10,14,15,22,62,388,454-456
Response selection Speed-accuracytrade-off
276,280,281,285,316 289,292
-errors in, 276-280 Spinal cord
Running balance 496,499,502-507
426434 - intelligence of, 503407,517
- and speed I course control, 426, -programmes,506,507
427 S-R psycho lo^
- information used for, 426434 89-97,104,131,292
- and Gibson, 89-92,128,444
Schema theory Stevens's law
-Schmidt, 205-207, 231, 234, 241, 409,410
242,244,265,290,316-318,490,539 Stimulus
- Zimmer, 219,220,228 89-97,99,125,163,164,254,194,343,
Selective adaptation 361,409-411,444
411,465,479,483 -and response, 89,92-97,350
Self-organization - discrete, 90
606,509-525,528,530,635,537,539, -energeticaspects, 93,409,410
642,544 -informational aspecta, 93
- and Turing machines, 524, 525, -rhythmical auditory, 135
642 Subjectivereinforcement, 316,317
-primary, 525 Substantialbody
- secondary,525,528,530,535,542 -
image of the, 439, 442-448, 450,
Subject Index 581
452458 524,525,542
- as gaol ofthe soul, 439,442-445 - and the problem of meaning, 521-
- 88 an instrument, 439,442445 523,526
- displacements of, 439, 445448, - artificial, 524,525
450,454-457 - natural, 524.525
Systemogenesis Turinguesque modelling
465 525,529,531,533-535,542,544,546
- and teleodynamics,535,542,544
Teleonomy
468 Validity
Teleodynamics -ecological, 138,291,458
504,505,516,520-523,525,526,531, - external, 262,40541 5
535,636,541-543 -face, 128,132
Theory offunctional systems - internal, 262
464466,469,472,478,479,481-485 Variability ofpractice hypothesis
Theory ofidentical elements 205-207,234,241,242,244
202 Verbal labelling method
Tiltingroom 289-310,327-335
429434 Visual information
Time-to-contact(tau) 127,146,160,190-197,248,250,252.
133-136,140,143,177,190-1 92,346, 327,328,331,335,339-374,426-434
434,435,543 - and balance control, 422434
Timing - and control of action, 339-374,
6,11,15-20.27-37,115,135-137,149, 419434
257,261,262,265-271,328,385,391, -binocular, 192-197
428,434,470,. 538 -monocular, 192-197
- central control of, 27-33,261,262,
265-271,285 Wadman et a1
- objections to central control of, 27-29,33,538
1520,638
- relative, 261, 262, 265-271, 285,
392,393
-research strategy, 33-37
Tools 73-79
- and hand actions, 75-77
- and measuring systems, 78
- developmentof the use of, 73,74
- standardization of, 78,79
TOTE-model
75
Transfer
227-229,234-240,247-249,254
- of movement parameters, 280-
2a4
Transfer-appropriate processing
201-213
-empirical evidence,203-212
-limitations, 212-213
Turing machines
- and self-organizing systems,
Permissions
p. 8, Fig. 1: Rack., P.M.H. & Westbury, D.R. (1969).The effects of length and
stimulus rate on tension in the isometric cat soleus muscle. Journal of
Physiology, 204, 443-460; p. 455.
p. 10, Fig. 2: Schmidt, R.A. (1982).Motor Control and Learning: A behavioral
emphasis. Champaign, IL:Human Kinetics Publishers; p. 268.
p. 14, Fig. 3: Pew, R.W. (1974).Human perceptual-motor performance. In B.H.
Kantowitz (Ed.), Human Information Processing: Tutorials in Performance
and cognition (pp. 1-39).Hillsdale, N J Erlbaum; p. 19.
p. 23, Fig. 4: Bernstein, N. (1967). The Co-ordination and Regulation of
Movements. Oxford Pergamon Press.
p. 28, Fig. 5: Wadman, W.J., Denier van der Gon, J.J., Geuze, R.H. & Mol, C.R.
(1979). Control of fast goal-directed arm movements. Journal of Human
Movement Studies, 5, 3-17;p. 10.
p. 34, Fig. 9 Sherwood, D.E., Schmidt, R.A. & Walter, C.B. (in press). Rapid
movements with reversals in direction: 11. Control of movement amplitude
and inertial load. Expermental Brain Research.
p. 53, Fig. 2: Lee, W.A. (1984).Neuromotor synergies as a basis for coordinated
intentional action. Journal of Motor Behavior, 16,135-170.
p. 60, Fig. 3: Taguchi, K. (1980).Relationship between the heads and the body's
center of gravity during normal standing. Acta Otolaryngologica, 90, 100-
105;p. 103.
p. 70, Fig. 4: Hewes, G. (1955).World distribution of certain postural habits.
American Anthropologist, 57, 397-405;p. 400.
pp. 76-77,Figs. 5 & 6: Drillis, R.J. (1963).Folk norms and biomechanics. Human
Factors, 5,427-441;pp. 428 & 430.
p. 103,Table 1: Uttal, W.R. (1981).A Taxonomy of Visual Processes. Hillsdale, N J
LEA Publishers.
p. 209,Fig. 1: Sherwood, D.E. (1985).A note on the effect of bandwidth knowledge
of results on movement consistency. Unpublished manuscript, University
of Colorado.
p. 218, Fig. 1: Bischof, N. (1981).Aristoteles, Galilei, Lewin - und die Folgen. In
W. Michaelis (Ed.), Bericht iiber den 32. KongreD der Deutschen
Gesellschaft far Psychologie in Zurich, 1980, Vol. I (pp. 21-78).Gottingen:
Hochrefe.
p. 355, Fig. 2: Srinivasan, M.V. (1977).A visually-evoked roll response in the
housefly. Journal of Comparative Physiology, 119,l-14.
p. 467, Fig. 1: Anochin, P.K. (1978). Beitriige zur allgemeinen Theorie des
funktionellen Systems. Jena: Fischer.
p. 468, Fig. 2: Von Holst, E. & Mittelstaedt, H. (1950). Das Reafferenzprinzip.
Naturwissenschaften, 37,464-476.
p. 473, Fig. 3: Deecke, L., Scheid, P. & Kornhuber, H.H. (1969).Distribution of
readiness potential, pre-motion positivity and motor potential of the human
584