C III)

::::s
--
c=-2
::::s
~
~

---.. r=?
==r C J
,..,...,
c::"C
~
c,r.)
~ cc=?
::::s
- --
c::"C
:::::J
--
c:;.
::x::::J
,

c:::::
C"C
I

c::"C
Co:)
c:::2..
::::i .
c.c:::2
C"C
~ c::::::
C"C
c:::;)
t::J
:3
C"C
::::s
,
------ f
PlJr<~!",:. I
~~ llNIVERSlr':\~J ~:.., .Ji::~': ";'.':' i
~!~
/-"."
.~ I""""
El,ello .. d_'."
!
I
:,;;. CAU I
Copyright Blackwell Publishers Ltd 2000
Editorial matter and organization copyright Kang Lee.
Darwin Muir and Alan Slater 2000
First published 2000

2 4 6 8 10 9 7 5 3 1
Blackwell Publishers Inc.
350 Main Street
Malden. Massachusetts 02148
Contents
USA
Blackwell Publishers Ltd
108 Cowley Road
Oxford OX4 IJF
UK
All rights reserved. Except for the quotation of short passages for the purposes of
criticism and review. 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 publisher.
Preface
Except in the United States of America, this book is sold subject to the condition Acknowledgments
that it shall not. by way of trade or otherwise. be lent, resold. hired out, or
otherwise circulated without the publisher's prior consent in any form of binding Introduction: What Is C<?gnitiveDevelopment Research and
or cover other than that in which it is published and without a similar condition What Is It For?
including this condition being imposed on the subsequent purchaser. Kang Lee
Library oj Congress Cataloging-in-Publication Data !Ias been applied Jar. 1 The History and Future of Cognitive Development
ISBN 0-631-21655-3 (hbk.) Research
ISBN 0-631-21656-1 (pbk.) Cognitive Development: Past, Present, and Future
British Lihrary Cataloguing-in-Publication Data J. H. Flavell
A CIP catalogue record for this book is available from the British Library. 2 Piaget's Theory of Cognitive Development
Piaget's Theory
Typeset in lCYI2 on 13 pt Photina
by Best-set Typesetter Ltd.. Hong Kong
J. Piaget
Printed in Great Britain by MPG Books, Bodmin. Cornwall 3 Developmental Research: Microgenetic Method
This book is printed on acid-free paper
Cognitive Variability: A Key to Understanding Cognitive
Development
R. S. Siegler
4 Information Processing and Connectionism
Development in a Connectionist Framework: Rethinking
the Nature-Nurture Debate
K. Plunkett
 ~-----~---. "

, " . THREE ' I

.....__ .~- ---_ ..... ---....,.-...-----.~~-- ~.--
~"" ..

Developmenta.1 Research:
Microgenetic Method
-
1ntrod uction
One of the goals of cognitive development research is to ~evealhow sys-';
,tematic cognitive change takes place. Cognitive deveiopmentalists have :.
at their disposal two main developmental methods to answer this ques-'
tion: cross-sectional and longitudinal. Cross-sectional designs assess a,
cognitive function of different age groups at the same point in. time, and
age differences obtained are used to suggest a developmental change in
children between these ages. While this design can be efficient for col~ .
lecting sufficient data in a short period of time, it has a major problem: it .
tends to gloss over both inter-individual and intra-individual variations'
and to present only a picture of what the cognitive developmental patterr;
of children as a group may look Wee,which does not necessarily reflect
the developmental pattern of any indivi'dual child.
~ In longitudinal studies the same group of children are t~sted repeat-
ed)y over a period of time.' Although it is'more time-consuming than .
cross-sectional design, because each individual's cognitive functioning is
traqked over time, both inter-individual and intra-individual variations'
can be examined. Traditionally, however, rather long time intervals are,
used in longitudinal studies' (e.g., 'tests ate given monthly or yearly), so
that important developmental changes are often missed. For this reason,
a number of researchers have proposed a new developmental research
approach that is referred to as the IDicrodevelopmental approach
(Karmiloff-Smith, 1979) or the microgenetic method (Kuhn, 1995).
Siegler's article introduces the microgenetic method that he helped
pioneer. It basically is a longitudinal design with much smaller time
intervals (sometimes as short as a few minutes) than the traditional lon-
gitudinal design (normally monthly or yearly). In addition, microgenetic
 Introduction
50 , ," . h'ldr n's cognitive f unctlOnID
"g' , by
method focuses on vanablhtl~s I~ c I ~ ularities in
contrast, the tr~~ition:~~~ngl:~~~a~:s~~~~~~:=s o~ r:asurernent - Cognitive Variability: A Key to
children's co~mtiv~ a I , leser demonstrates that when we pay close,
errors: In thlS artthlc~~r~~~oirregularities in their behavior can be very,
attentIOn to wach , ' .. 1 t
Understanding Cognitive Development
informatlve about regularities in cogmtive deve oprnen . '
Robert S. Siegler

References .. . ,
, M' _ and macro-developmental changes mlan- '~
Karmiloff-SIDlth,A. (1979). Jcro, '1 systems Cognitive Science, 3, ,
guage acquisition and other representationa . . ., , .-
91-118,' ., h . What has it told us? !

( 1995). Microgenetic study of c; ange. ,

Kuh 11, D.
Psychological Science, 6, 133-9.,

Among the most remarkable characteristics of human beings is how

further reading, , ,
ic dis la of process: Historical development an ,'
Catan, L, (1986). The dynam ,P y,
contemporary uses of the ffilcrogenetlc me
252-63.
. d much our thinking changes with age. When we compare the thinking
thod H~man DevelOPment, 2~, of an infant, a toddler. an elementary school student. and an adolescent,
'. : . the magnitude of the change is immediately apparent. Accounting for
..
,...
~J. _
how these changes occur is perhaps the central goal of researchers who
study cognitive development.
Alongside this agreement about the importance of the goal of deter-
mining how change occurs. however. is agreement that we traditionally
have not done very well in meeting it. In most models of cognitive devel-
opment. children are depicted as thinking or acting in a certain way for
a prolonged period of time, then undergoing a brief. rather mysterious.
transition. and then thinking or acting in a different way for another
prolonged period, For example, on the classic conservation-of-liquid
quantity problem, children are depicted as believing for several years
that pouring water into a taller, thinner beaker changes the amount of
water; then undergoing a short period of cognitive conflict, in which
they are not sure about the effects of pouring the water; and then real-
izing that pouring does not affect the amount of liquid. How children
get from the earlier to the later understanding is described only
superficially,
Critiques of the inadequacy of such accounts have been leveled most
often at stage models such as Piaget's, The problem, however, is far more
pervasive, Regardless of whether the particular approach describes
52 Developmental Research: Microgenetic Method
Cognitive Variability 53
development in terms of stages. rules. strategies. or theories; regardless
focuses directly on changes in children's thinking. This research has
of whether the focus is on reasoning about the physical or the social
documente~ la~~e-scale variability in children's thinking and suggests
world: regardless of the age group of central interest. most theories
that the vanabllIty contributes directly to cognitive growth.
place static states at center stage and change processes either in the
wings or offstage alLogether. Thus. three-year-olds are said to have non-
representational theories of mind and five-year-olds representational
Pervasive Variability
ones; five-year-olds to have absolute views about justice and ten-year-
aids relativistic ones; ten-year-alds to be incapable and 1S-year-olds
Variability in. c.hildren's thinking exists at every level- not just between
capable of true scientific reasoning. The emphasis in almost all
children o~ dl:feren~ ag~s, or between different children of the same age,
cognitive-developmental theories has been on identifying seq.ue~ces of
but also wlthm an mdlviduaJ solving a set of related problems, within
one-to-one correspondences between ages and ways of thmkmg or
an individual solVing the same problem tWice. and even within an
acting. rather than on specifying how the changes occur. individual on a single trial.
If developmentalists are so interested in change processes, why would
the topic be given such cursory treatment in most contemporary theo-
ries? Part 01' the problem is that studying change is inherently difficult. It Variability within an individual solVing related problems
poses all the conceptual and methodological demands of studying p~r-
Detailed analyses of tasks on which one-to-one correspondences
formance at anyone time. and imposes the added demands of determm-
between age and way of thinking have been postulated indicate that
ing what is changing and how the change is being accomplished.
children's thinking is generally much more variable than past depictions
An additional part of the difficulty. however, may be self-imposed. In
have suggested. To cite an example from language development, rather
our efforts to describe differences among age groups in assimple. dra-
than young children passing through a stage in which they always over-
matic, and memorable terms as possible, we may unwittingly have made
regularize past tense forms (e.g., saying "goed" and "eated" rather than
understanding change more difficult than it needs to be. In particular,
"went" and "ate"), children at all ages between two and a half and five
portraying children's thinking and knowledge as monolithic for several
years produce both substantial numbers of overregularized forms and
a
years at a time creates need to explain the wide gulfs between the suc-
substa.ntial numbers of correct ones. The variability throughout this age
cessive hypothesized understandings - even though such gulfs may not
range IS present for a single child followed throughout the period, as well
exist. The typical depictions make change a rare, dmost exotic, event
as for groups of children sampled at a single age. Adding to the vari-
that demands an exceptional explanation. If children of a given age
ability, children often produce more than one incorrect form of a given
have for several years had a particular understanding. why would they
verb; on different occasions, a given child will say, "I ate it," "I eated it,"
suddenly form a different understanding, and why would they regularly and "I ated it" (Kuczaj,1977).
form it at a particular age? The problem is exacerbated by the fact that
Similar variability has been found in the development of memory
for many of the competencies of interest, generally relevant experience
strategies. Contrary to the widely cited model that five-year-olds do not
is available at all ages and specifically relevant experience at none.
rehearse and eight-year-olds do, trial-by-trial assessments indicate that
Children see liquids poured into containers of different dimensions at all
the majority of children of both ages sometimes do and sometimes do
acres - and are not ordinarily told at any age that the amount of liquid
not rehearse (McGilly & Siegler, 1990). The percentage of trials on
r:mains the same after pouring as before. Why, then, would they con-
which they rehearse increases with age, but, again, there is variability
sistently have one concept of liquid quantity conservation at age five throughout the age range.
and a different one at age seven?
. Conceptual development evidences the same pattern. Despite claims
Recognition of the unwelcome side effects of the one-to-one depic-
that five-year-olds think of number conservation solely in terms of the
tions of cognitive growth has led to a new generation of research that
lengths of the rows, triaJ-by-trial assessments indicate that most
54 Developmental Research: Mic;rogenetic Method
five-year-olds sometimes rely on the lengths of the rows, sometimes rely
on the type of transformation, and sometimes use other strategies such
as counting or pairing (Siegler, 1993). Again, the frequency of reliance
on these ways of thinking changes with age, but most five-year-olds'
judgments and verbal explanations indicate several different ways of
thinking about the concept.
Development of problem-solVing skills provides yet more evidence for
such within-subject cognitive variability. Contradicting models in which
preschoolers are said to use the sum strategy (counting from 1) to solve
simple addition problems and in which first through third graders are
said to use the min strategy (counting from the larger addend, as when
solving 3 + 6 by counting "6,7,8,9") to solve them, children of all these
ages use a variety of strategies. In one study, most children presented
a set of addition problems used at least three clifferent strategies on
different problems, and most children examined in a more extensive
microlongitudinal study used at least five distinct strategies (Siegler &
Jenkins, 1989).
Variability within an individual solving a single
problem twice
The variability within inclividual children cannot be reduced to children
using different strategies on different problems. Even presented the iden-
tical problem twice within a single session, or on two succe'ssive days,
children use different strategies on roughly one third of the pairs of
trials in addition, time-telling, and block-building tasks (Siegler &
McGilly, 1989; Siegler & Shrager, 1984; Wilkinson, 1982). This vari-
ability within individuals within problems cannot be explained by learn-
ing; in these studies, children used the strategy that appeared more
advanced almost as often for the first presentation of a problem as for
the second (roughly 45% vs. 55%).
Variability within a single trial
In the limiting case, variability has been found even within an individ-
ual solving a particular problem on a single trial. This type of variabil-
ity has been reported by investigators interested in the relation between
children's hand gestures and verbal explanations. In these studies, chil-
dren often express one type of understanding through the gestures and
a quite different understanding through the explanations (Church &
Goldin-Meadow, 1986; Goldin-Meadow & Alibali, in press). For
example, on number conservation problems, children may express a
reliance on relative lengths of the rows in their hand gestures, while at
the same time verbally expressing reliance on the type of transforma-
tion, or vice versa.
These findings suggest that cognitive change is better thought of in
terms of changing distributions of ways of thinking than in terms of
sudden shifts from one way of thinking to another. The types of descrip-
tions of change that emerge from such analyses are illustrated in figures
1 and 2. Figure 1 shows changes in three children's addition strategies
over a three-month period (Siegler & Jenkins, 1989); figure 2 shows
changes in a child's map-drawing strategies over a two-year period
(Feldman, 1980). Similar changes in distributions of strategies have
been found in studies of conceptual understanding, memory strategies,
problem solving, and language. In all these domains, cognitive develop-
ment involves changing clistributions of approaches, rather than
discontinuous movements from one way of thinking to another.
Variability is not just an incidental feature of thinking; it appears to play
a critical.role in promoting cognitive change. Several types of evidence
converge on this conclusion. One comes from observations of children
in the process of cliscovering new strategies. Both the trials immediately
before a discovery and the trial on which the discovery is made fre-
quently involve especially variable behavior - disfluencies, unclear ref-
erences, long pauses, and unusual gestures (Siegler & Jenkins, 1989). A
second type of empirical evidence linking variability to cognitive change
involves analyses of. which children are most likely to make cliscoveries.
Children whose verbal explanations and gestures reflect different initial
misunderstandings of number conservation and of numerical equiva-
lence problems (a + b + c = _ + c) are more likely to make discoveries sub-
sequently than are children whose explanations and gestures reflect the
same initial misunderstanding (Church & Goldin-Meadow, 1986;
Goldin-Meadow & Alibali, in press). Similarly, children whose pretest
explanations reflect varied ways of thinking are more likely to
learn from instruction regarding the meaning of the equal sign in
 ~<#

,,'
..
'~,
.. ....
% use of .: __
__
sum
short-cut sum 60 ... .'........... Level of map

..,.
"--0--' fingers ,,'
strategy -<r-II
-- .. - min
I/l -e--III
"5'" 40
.IV
'0
o ?f
1-5 6-1011-1516-2021-2526-3031-33
Sessions 20

70

60
......... " .
50 d/ .....
c.... .-- sum
40 ". __ short-cut sum
% use of ". ..0
"'-0-" retrieval Figure 2 Changes in distributions of map-drawing approaches across five
strategy
30 cr ---r" min
sessions, conducted over a two-year period. Higher numbers indicate more
20 advanced levels of map drawing; thus, LevelIV maps are more advanced than
10
Level III ones.

mathematics than are children whose pretest explanations reflect crisp,

100
specific misunderstandings (Graham & Perry, 1993).
A different type of evidence for the contribution of variability to cog-
80 nitive change comes from formal models of development. Theorists who
differ in many particular assumptions have found that modeling change
60 _sum
% use of __ short-cut sum requires both mechanisms that produce variability and mechanisms
strategy "'-0-" retrieval that produce adaptive choices, among the variants. Connectionist
__.. _ min
40
models of development are based on connection strengths among pro-
20 cessing units varying at all points in learning, from initial. randomly
varying strengths to final, asymptotic levels; change occurs through
redistributions of the varying connection strengths. Dynamic systems
models also treat variability as a fundamental property of development;
they aim to explain how local variability gives rise to global regularities.
FIgur e 1 Changes in distributions of addition strategies
,
of three children over
'h 'b'l Similarly, recent symbolic-processing models of development focus on
roughly 30 sessions conducted over a three-month penod: N~tlce t e van a 1- how varying strategies, analogies, and other higher-order units come to
ity that is present within each child's performance wlthm each block of
be used increasingly in the situations in which they are most effective.
sessions, as well as the changes in distributions of strategy used over the
course of the study.
Source: Data from Siegler & Jenkins, 1989.
58 Developmental Research: Microgenetic Method
At a less formal level. operant conditioning models. evolutionarily
based models. and generate-and-test models are all based on the
. assumption that change occurs through selection processes operating
on omnipresent. spontaneously produced variability in behavior (e.g ..
see CampbelL 1974; McClelland & Jenkins. 199]; Smith & Thelen. in
press; Skinner. 1981).
A striking empirical finding about the variability in children's think-
ing. and one that is important for its ability to contribute to cognitive
development. is the constrained quality of the variations that children
generate. Far from conforming to a trial-and-error model, in which all
types of variations might be expected, the new approaches that children
attempt consistently conform to the principles that define legal strategies
in the domain (except when children are forced to solve problems for
which they do not possess any adequate strategy). For example, in a 30-
session study of preschoolers' discovery of new addition strategies. none
of the children ever attempted strategies that violated the principles
underlying addition (Siegler & Jenkins, 1989). They invented legitimate
new strategies. such as the min strategy. but never illegitimate ones. such
as adding the smaller addend to itself or counting from the larger addend
the number of times indicated by the first addend. The question is how
they limit their newly generated strategies to legal forms.
One possibility is that even before discovering new strategies. children
often understand the goals that legitimate strategies in the domain must
satisfy. Such understanding would allow them, without trial and error,
to discriminate between legitimate new strategies that meet the essen-
tial goals and illegitimate strategies that do not. A very recent study
revealed that children possessed such knowledge in both of the domains
that were examined - simple addition and tic-tac-toe (Siegler & Crowley,
in press). In simple addition, children who had not yet discovered the
min strategy nonetheless judged that strategy (demonstrated by the
experimenter) to be as smart as the strategy they themselves most often
used - counting from 1 - and significantly smarter than an equally
novel but illegitimate strategy that the experimenter demonstrated. In
tic-tac-toe, children rated a novel strategy that they did not yet use -
forking - as even smarter than the strategy they themselves usually
employed - trying to complete a single row or column. Ability to anti-
cipate the value of untried strategies may promote cognitive growth by
filtering out unpromising possibilities and thus channeling innovations
in potentially useful directions.
Thinking is far more variable than usually depicted. In the past,
researchers have usually ignored such variability or viewed it as a
bother. This stance has led to subjects being given practice periods, not
so that especially variable behavior in those periods can be studied, but
so that it can be discarded, in order that it not obscure the more orderly
patterns in later performance. When such variability has been explicitly
noted at all. it has usually been viewed as an unfortunate limitation of
human beings, a kind of design defect, something to be overcome
through practice. Computers, robots, and other machines not subject
to this flaw can perform many tasks more accurately than people can.
Presumably, people's performance would also be enhanced if it were
less variable.
This view of variability as detracting from efficient performance
misses at least half the story, though. The variability of cognition and
action allows us to discover a great deal about the environments toward
which the thinking and action are directed. Our difficulty in reproduc-
ing the way we pronounced a word in an unfamiliar foreign language
may lead to some even less adequate pronunciations in the short run,
but in the longer run may lead us to generate and then learn better pro-
nunciations. Likewise, our inability to give a colloquium in the same
words twice, even when we want to. may lead to some parts being less
clear than in the best of our previous presentations. but it also allows
us to observe audience reaction to new lines of argument and to learn
which ones are best received. In general, cognitive variability may lead
to performance never incorporating on anyone occasion all the best fea-
tures of previous performance, but also may be critical to our becoming
increasingly proficient over time.
If cognitive variability does indeed facilitate learning, it would be
adaptive if such variability were most pronounced when learning,
rather than efficient performance, is most important - that is, in infancy
and early childhood. This appears to be the case. Across many domains,
expertise brings with it decreasingly variable performance. To the extent
that young children are "universal novices," their lack of expertise
alone would lead to their performance being more variable than that of
older children and adults. A number of cognitive neuroscientists have
hypothesized that above and beyond such effects of practice, the process
of synaptogenesis, which results in children from roughly birth to age
seven having far more synaptic connections than older children and
adults, may contribute both to the high variability of early behavior and
to young children's special ability to acquire language, perceptual skills,
and other competencies under abnormal organismic and environ-
mental conditions (Goldman-Rakic, ] 987; Greenough, Black, &
Wallace, 1987). That is, young children's greater variability at the
neural level seems to allow them to learn useful behaviors under a
greater range of circumstances. The general lesson seems to be that
explicitly recognizing the great variability of infants' and young chil-
dren's thinking, and attempting to explain how it is generated and con-
strained, will advance our understanding of the central mystery about
cognitive development - how change occurs.
Siegler, R. S., and Crowley, K. Goal sketches constrain children's strategy
discoveries. Cognitive Psychology (in press).
Siegler, R. S., and Jenkins, E. How children discover new strategies. Erlbaum,
Hillsdale, NJ (1989).
Siegler. R. S., and McGUly,K. Strategy choices in children's time-telling. In Time
and human cognition: A liJe span perspective, Levin, 1.. and Zakay, D. Eds.
Elsevier Science, Amsterdam (1989).
Siegler, R. S., and Shrager, J. Strategy choices in addition and subtraction: How
do children know what to do? In The origins oj cognitive skills, Sophian, C. Ed.
Erlbaum, Hillsdale, NJ (1984).
Smith, L. B.. and Thelen, E., Eds. Dynamical systems in development: Applications.
Bradford Books, Cambridge, MA (in press).
Wilkinson, A. C. Partial knowledge and self-correction: Developmental studies
of a quantitative concept. Developmental Psychology, 28, 876-893 (1982).

Campbell, D. T, Evolutionary epistemology. In The Philosophy oj Karl Popper, Vol.

14, Schilpp, P. A. Ed. Open Court, La Salle, 11(1974). Skinner, B. F.. Selection
by consequences. Science, 213, 501-504 (1981).
Church, R. B., and Goldin-Meadow, S. The mismatch between gesture and
speech as an index of transitional knowledge. Cognition, 23, 43-71 (1986).
Feldman, D. H. Beyond universals in cognitive development. Ablex, Norwood, NJ,
1980.
Goldin-Meadow, S., and Alibali, M. W. Transitions in concept acquisition: Using
the hand to read the mind. Psychological Review (in press).
Goldman-Rakic, P. S. Development of cortical circuitry and cognitive function.
Child Development, 58, 601-622 (1987).
Graham, T, and Perry, M. Indexing transitional knowledge. Developmental
Psychology, 29, 779-788 (1993).
Greenough, W. T, Black, J. E.. and Wallace, C. Experience and brain develop-
ment. Child Development, 58, 539-559 (1987).
Kuczaj, S. A. The acquisition of regular and irregular past tense forms. Journal
oj Verbal Learning and Verbal Bellavior, 26, 589-600 (1977).
McClelland, J. 1., and Jenkins, E. Nature, nurture, and connections: Implications
of connectionist models for cognitive development. In Architectures Jor intelli-
gence, Van Lehn, K. Ed. Erlbaum, Hillsdale, NJ (1991).
McGilly, K.. and Siegler, R. S. The influence of encoding and strategic knowl-
edge on children's choices among serial recall strategies. Developmental
Psychology, 26, 931-941 (1990).
Siegler. R. S. A microgenetic study of number conservation. Manuscript in
preparation, Carnegie Mellon University, Pittsburgh (1993).