Page 1

See discussions, stats, and author profiles for this publication at: https://www.researchgate.net/publication/229800773

Dynamic Development of Action and Thought

Chapter in Journal of Sport & Exercise Psychology � May 2007

Impact Factor: 2.59 � DOI: 10.1002/9780470147658.chpsy0107

CITATIONS

96

READS

59

2 authors, including:

Kurt W. Fischer

Harvard University

66 PUBLICATIONS 1,479 CITATIONS

SEE PROFILE

Available from: Kurt W. Fischer

Retrieved on: 13 May 2016

313

CHAPTER 7

Dynamic Development of Action and Thought

KURT W. FISCHER and THOMAS R. BIDELL

DYNAMIC STRUCTURALISM 314

Variability in the Middle of Things: An Example of

Representing Social Interactions 315

Dynamic Nature of Psychological Structure 317

Dynamic Structure in Living Systems 318

Variation and Order in Development: The

Constructive Web 319

DYNAMIC STRUCTURE IN COGNITIVE AND

EMOTIONAL DEVELOPMENT 320

Psychological Structure as Dynamic Skill 321

Building a Constructive Web for Positive and Negative

Social Interactions 325

How Dynamic Skills Explain Variability

in Development 329

THE CRISIS OF VARIABILITY AND THE

CARTESIAN SYNTHESIS IN

DEVELOPMENTAL SCIENCE 336

The Cartesian Dualist Framework 337

The Tacit Modern Synthesis in Psychology: Nativism and

Empiricism Together 338

The Structure-as-Form Paradigm 340

The Stage Debate and the Discovery of Variability in

Cognitive Development 341

Explaining Variability versus Explaining It Away 342

The Constructivist Alternative 346

METHODOLOGY OF DYNAMIC

STRUCTURAL ANALYSIS 347

Starting in the Middle of Things: Implications

for Design 348

Guidelines for Developmental Research 349

Building and Testing Models of Growth

and Development 356

BUILDING STRUCTURES: TRANSITION

MECHANISMS, MICRODEVELOPMENT, AND

NEW KNOWLEDGE 363

Relations between Micro- and Macrodevelopment 363

Construction Processes: From Micro to Macro 364

EMOTIONS AND THE DYNAMIC ORGANIZATION

OF ACTIVITY AND DEVELOPMENT 370

Emotion and Cognition Together 370

Organizing Effects of Emotions 372

Emotionally Organized Development 376

JOINING NATURE AND NURTURE: GROWTH

CYCLES OF PSYCHOLOGICAL AND

BRAIN ACTIVITY 382

Epigenesis of Action, Feeling, Thought, and Brain 382

Principles for Understanding Growth Patterns of Brain

and Behavior 383

Cycles of Reorganization in Development 385

CONCLUSION: DYNAMICS OF STABILITY AND

VARIABILITY IN DEVELOPMENT 388

REFERENCES 390

Human activity is both organized and variable, dynami-

cally changing in principled ways. Children and adults

are flexible and inventive in their action and thought,

adapting old ideas to new situations and inventing con-

cepts, formulating plans, and constructing hypotheses

while participating in a wide variety of cultural prac-

tices. Few developmentalists today would disagree, as, for

half a century, psychologists have been accumulating a

Preparation of this chapter was supported by grants from Ms.

and Mrs. Frederick P. Rose and the Harvard Graduate School

of Education.

wealth of evidence on the constructive, self-regulating,

and culturally contextualized nature of human psycholog-

ical processes. If psychological function—the way people

act—is constructive, dynamic, and culturally embedded,

then psychological structure—the organization or pattern

of activities—is equally so. Yet remarkably, the most

widely used conceptions of psychological structure and

its development do not reflect this dynamic, constructive,

and contextualized picture of psychological processes.

The opposite is true: The major models of development

describe psychological structure in static, formal terms.

Concepts like universal stages, innate linguistic modules,

and innate cognitive competencies portray psychological

314 Dynamic Development of Action and Thought

organization as fixed and unchanging, insulated from

variation in context and feedback from activity.

The hallmark of the dynamic nature of human behav-

ior is its pervasive variability: People act differently in

different situations, with different people, in different

emotional states. Faced with the large and growing cor-

pus of research evidence for variability in activity and

development, researchers guided by static models have

been continually surprised to find that children’s

performance is nowhere near as stable as the static con-

ceptions predict. A child who can solve an arithmetic

problem (or a social problem) one day or in one situation

frequently cannot solve the same problem the next day or

in a different but apparently similar situation. Different

children of precisely the same age often cannot perform

the same cognitive tasks—sometimes in relation to cul-

tural contexts or family environment, sometimes for rea-

sons that are harder to explain. Even from moment to

moment, a person performs a task differently as she or

he adapts to variations in the situation, social context, or

emotions of self and others. Indeed, when the data of

cognitive developmental research is taken as a whole,

variability in the level of psychological performance is

the norm, not the exception.

The task of developmental science is to detect and de-

scribe patterns in this variability and to propose models

to account for data patterns that reflect both stability

and variability. We show how the concepts and methods

of dynamic structural analysis provide a framework and

tools for analyzing this variability and detecting the

order in it—key findings such as the emergence of qual-

itatively new cognitive abilities or the transitions from

one behavior to another.

In our view, performances vary so greatly because

psychological structure is not static but naturally pro-

duces variability in activity and development arising

from people’s constructive self-organization of their

own psychological structures in relation to situations,

other people, meaning systems, and their own bodies.

Far from being a problem, patterns of developmental

variability are the key to understanding the organiza-

tion of these dynamic systems and the constructive

processes by which human agents create new interrela-

tions and thus new structures. The complexity of these

systems is not something to be controlled for but to be

described and understood. The patterns of variability

that arise from the particular ways in which cognitive

systems are organized are the key to understanding that

organization and thus to understanding psychological

structure. Tools from dynamic systems analysis pro-

vide ways of embracing the variability to find the order

in it.

With this chapter, we present a framework for con-

ceptualizing psychological structure in dynamic sys-

tems constructed by human agents. We show how

this model describes and explains patterns of develop-

mental variability in terms of the structures human be-

ings build. The chapter begins with an introductory

overview of dynamic structuralism as a general ap-

proach to development, elaborating a theoretical model

of psychological structure as the dynamic organization

of self-constructed, socially embedded skills and ac-

tivities (actions and thoughts). We contrast this posi-

tion with traditional static views of psychological

structure, which dominate scientific dialogue in what

amounts to a modern synthesis of traditionally opposed

viewpoints of nativism and empiricism. These static

views derive from reductionist scientific theory inher-

ited from the Cartesian tradition in philosophy, which

leads to systematic misunderstanding of the nature of

psychological structure and blatant failures to explain

the extent of developmental variability.

The dynamic framework and research tools specifi-

cally crafted for analyzing development and learning

provide a research methodology for the study of psycho-

logical structures including both their variability and

the order in the variation. These concepts and tools

apply to both long-term development and short-term mi-

crodevelopmental variations in the building of dynamic

structures, providing powerful methods for testing dy-

namic hypotheses about variation, change, and stability.

Broad in scope and applicability, the dynamic structural

model and methodology elucidate relations between

cognitive, social, emotional, and neurological develop-

ment—which all work together in the activities of

human beings in all their rich complexity.

DYNAMIC STRUCTURALISM

One reason psychological structure has so often been

treated as static is that theorists have confounded struc-

ture with form. Structure refers to the system of relations

(Piaget, 1970) by which complex entities such as biologi-

cal organisms and psychological activities are organized.

There are systematic relations, for instance, between the

nervous system and the cardiovascular system such that

each supports and responds to the other. The relations

between these systems are in a constant balance or equi-

Dynamic Structuralism 315

librium, which can only be maintained by constant activ-

ity on the part of each subsystem. Thus, systems of rela-

tions—structures—are necessarily dynamic.

Form is an abstraction from structure—a fixed pat-

tern that can be detected in a dynamic structure. An or-

ange has cellular and tissue-level structure, which lead

to its cohesion in a spherical shape. The structure of the

orange is dynamic, emerging developmentally, maintain-

ing a dynamic equilibrium for a time, and then decaying.

The concept of sphere, on the other hand, is an abstract

form that we apply to describe one characteristic of the

dynamic structure: the shape it produces. Beyond the or-

ange, the concept of sphere is an ideal form that applies

across myriad realities. The fact that this formal con-

cept is unchanging across many situations is what makes

it useful in describing similarities in many different ob-

jects such as balls, plums, or planets.

A structure/form problem arises when an abstrac-

tion used to describe reality is confounded with the re-

ality described. People commonly expect patterns of

phenomena in the world to conform to their underlying

abstractions, instead of determining which patterns fit

an actual object or experience. In personality and so-

cial relations, people commonly expect others to fit the

stereotype of, for example, a shy, introverted person or

a mother (Greenwald et al., 2002). Similarly, in sci-

ence, researchers who focus on the sphere form may be

surprised that baseballs, basketballs, and soccer balls

are so different from one another, and researchers who

focus on innate knowledge may be surprised to find

that a 3-year-old really does not understand the num-

bers 1, 2, and 3 even though an infant can distinguish

arrays of 1, 2, and 3 dots (Spelke, in press). For the

sphere, the logical fallacy is obvious: The spherical

shape is an abstraction of a common pattern across dif-

ferent objects, not an independently existing form that

somehow dictates what the objects should be like. The

same fallacy applies to the stereotypes and the nativist

explanation of number.

This form fallacy has frequently led to perplexity

among scientists and educators who expect patterns of

thought and action to conform to an independently exist-

ing form such as stage, cognitive competence, or core

knowledge. Scholars have been puzzled when a child

reaches a certain stage or competence for one task or sit-

uation and he or she does not evidence the same ability

in other tasks or situations, as if an underlying abstract

logic could determine an individual’s performance in

the real world (Piaget, 1985). The attempt to preserve

formal conceptions of structure in the face of ever grow-

ing evidence of variability in cognitive performance has

led developmental theorists into pointless arguments

over, for example, which of many varying performances

represent an individual’s “real” logical ability, or at

what age children “really” acquire a concept like object

permanence. We demonstrate later how the confounding

of form with structure has led to an explanatory crisis in

developmental science with ever more tortured attempts

to explain the pervasive evidence of variability in static

conceptions of structure as form. (We also see hopeful

signs that the field is shifting to deal more centrally

with the dynamics of variation.)

Dynamic structuralism offers an alternative to static

conceptions of structure, starting with the recognition

of the complexity inherent in human psychological de-

velopment and the central role of the person in con-

structing dynamic systems of action and thought.

Instead of trying to eliminate or get beyond the com-

plexity of relations among systems, dynamic structural-

ism uses the tools of contemporary developmental

science to analyze patterns in the complexity—how the

constructive activity of human agents leads to new rela-

tions among systems of action and thought. The analysis

of the dynamic structures of human behavior provides a

way of simplifying without discarding complexity, iden-

tifying the essential relations among systems, and ex-

plaining activities and developmental pathways in terms

of those essential system relations. Dynamic structural-

ism thus differs from the classic structuralism of Piaget

(1983), Chomsky (1995), and others, which isolates

structure from the variability of mental dynamics, treats

it as static, and attempts to explain development in

terms of the static forms.

Variability in the Middle of Things: An

Example of Representing Social Interactions

Focusing on the pervasive variability of human activity,

dynamic structuralism analyzes the patterns of stability

and order in diverse patterns of activity in the variation

(Bidell & Fischer, 1992; Fischer, Yan, & Stewart, 2003;

Siegler, Chapter 11, this Handbook, Volume 2; Thelen &

Smith, Chapter 6, this Handbook, this volume; van

Geert, 1998). As in the study of ecology, the analysis

begins in medias res, in the middle of things. Starting in

the middle of things means that people’s activities are

embodied, contextualized, and socially situated—un-

derstood in their ecology (Bronfenbrenner & Morris,

316 Dynamic Development of Action and Thought

Chapter 14, this Handbook, this volume; Cairns, 1979;

Gibson, 1979) as well as their structure. People act and

understand through their bodies acting in the world, not

through a disembodied mind or brain. The brain and

nervous system always function through a person’s body

and through specific contexts composed of particular

people, objects, and events, which afford and support

the actions. People act jointly with other people within

culturally defined social situations, in which activities

are given meaning through cultural frames for interpre-

tation (Rogoff, 1990). Action in context is the center of

who people are and how they develop (Lerner & Busch-

Rossnagel, 1981; Brandtst�dter, Chapter 10, this Hand-

book, this volume).

Starting in the middle of things with embodied, con-

textualized, socially situated individual and joint activ-

ity requires two major steps: (1) to describe basic

structures or organizations of activities in context and

(2) to characterize how those structures vary as a func-

tion of changes in key dimensions of person, body, task,

context, and culture. Whether the focus is on knowl-

edge, action, emotion, social interaction, brain function-

ing, or some combination, the dynamic structural

approach puts the person in the middle of things and

frames the person’s activity in terms of multiple compo-

nents working together. The maturity or complexity of

people’s behavior varies widely and systematically from

moment to moment and across contexts, states, and in-

terpretations or meanings. Each individual shows such

variations, in addition to the wide variations that occur

across ages, cultures, and social groups.

Consider, for example, the wide variation docu-

mented for children’s stories or narratives about posi-

tive and negative social interactions (Fischer & Ayoub,

1994; Hencke, 1996; Rappolt-Schlichtman & Ayoub, in

press; Raya, 1996). The developmental level, content,

and emotional valence of a child’s stories vary dramati-

cally as a function of priming and immediate social sup-

port, emotional state, and cultural experience. For

example, the activities of 5-year-old Susan demonstrate

some of the variations in both developmental complexity

and emotional organization that have been documented

in research. First, she watches her counselor act out a

pretend story with dolls: A child doll named after Susan

makes a drawing of her family and gives it to her father,

who is playing with her. “Daddy, here’s a present for

you. I love you.” Then the daddy doll hugs the girl doll

and says, “I love you too, and thanks for the pretty pic-

ture.” He gives her a toy and says, “Here’s a present for

you too, Susan.” When asked, the girl promptly acts out

a similar story of positive social reciprocity, making

Daddy be nice to Susan because she was nice to him.

Ten minutes later, the counselor asks the girl to show

the best story she can about people being nice to each

other, like the one she did before. Instead of producing

the complex story she did earlier, she acts out a much

simpler story, making the Daddy doll simply give lots of

presents to the child doll, with no reciprocal interaction

between them. There is no social reciprocity in the story

but only a simple social category of nice action.

A few minutes after that, when the girl has sponta-

neously shifted to playing at fighting, the counselor

shows her another nice story about father and child. This

time, when the girl acts out her story, she switches the

content from positive to negative with energetic aggres-

sion. The girl doll hits the Daddy doll, and then he yells

at her, “Don’t you hit me,” slaps her in the face and

pushes her across the room, showing the violence that

often appears in the stories of maltreated children. The

girl doll cries and says she is scared of being hit again.

Note that, despite the shift to negative affect, Susan sus-

tains a story involving social reciprocity: The Daddy

doll hits the Susan doll because she had hit him, and she

becomes afraid because he had hit her.

Then Susan becomes agitated; yelling, she runs

around the room and throws toys. When the counselor

asks her to do another story, she makes the dolls hit and

push each other with no clear reciprocity and no expla-

nation of what is happening. With her distress and disor-

ganization, she no longer acts out a complex aggression

story but is limited to stories of repeated hitting, even

when she is asked to produce the best story she can. She

uses a simple social category of mean action.

What is the “real” story for the child? Does she rep-

resent relationships between fathers and daughters as

positive or negative? Is she capable of representing reci-

procity, or is she not? These are the kinds of questions

that are often asked in child development, but these

questions assume an opposition that makes no sense.

Susan plainly shows four different “competences”—

positive reciprocity, positive social category (without

reciprocity), negative reciprocity, and negative social

category. Depending on the immediate situation, her

emotional state, and the social support from her coun-

selor, she demonstrates each of these four different

“abilities.” Her four skills vary strongly in both emo-

tional valence and developmental level (complexity)

with the different skills linked to the social context, her

emotional state, and her relationship with her father and

her counselor.

Dynamic Structuralism 317

Figure 7.1 Variation in competence for stories as a function

of social-contextual support. In the high-support assessments

the interviewer either modeled a story to a child (Elicited Imi-

tation) or described the gist of a story as well as some content

cues (Prompt), and then the child acted out or told a similar

story. In the low-support assessments the interviewer provided

no such support but either asked for the best story the child

could produce (Best Story) or let the child make up a number

of stories in free play with the most complex story determining

the child’s “competence” for this context (Free Play). Children

had performed similar stories several times before the assess-

ments graphed here. The y-axis indicates steps in the assessed

developmental sequence, as well as skill levels (Rp1 to Rp3),

which will be explained later. Sources: From “The Dynamics

of Competence: How Context Contributes Directly to Skill”

(pp. 93–117), by K. W. Fischer, D. H. Bullock, E. J. Rotenberg,

and P. Raya, in Development in Context: Acting and Thinking

in Specific Environments—The Jean Piaget Symposium Series,

R. H. Wozniak & K. W. Fischer (Eds.), 1993, Hillsdale, NJ:

Erlbaum; and “The Effects of Development, Self-Instruction,

and Environmental Structure on Understanding Social Interac-

tions,” by E. J. Rotenberg, 1988, Dissertation Abstracts Inter-

national, 49(11), p. 5044B.

Low Support

Condition

Elicited Imitation

Free Play

Best Story

Prompt

Free Play

Best Story

1

0

2

3

4

5

6

7

Rp2

Rp1

Rp3

High

Support

Low Support

High

Support

Highest Step/Level

Different contexts for assessment routinely produce

such substantial variations, although most developmen-

tal theories and methods do not deal with this variabil-

ity. Children (and adults) show distinct levels of

competence under different conditions, even for a single

domain such as stories about nice and mean social inter-

actions between peers (A. Brown & Reeve, 1987; Fis-

cher, Bullock, Rotenberg, & Raya, 1993). Figure 7.1

shows the best (most complex) performances of eight 7-

year-old children who were acting in (a) several contexts

in which an interviewer provided high social support for

complex stories, such as prompting the gist of the plot,

and (b) several contexts providing no such support. As

the context shifted, the children’s competence for repre-

senting mean, nice, or nice-and-mean social interactions

shifted dramatically and systematically. Every individ-

ual child showed a similar pattern of shifting across con-

ditions—competence at step 6 or 7 for high-support

conditions, and competence at step 2, 3, or 4 for low-

support conditions. This variation is an example of de-

velopmental range, the spread between competence with

high support and competence with little support. With

both positive and negative stories, Susan demonstrated a

developmental range varying from interactions with so-

cial reciprocity to interactions based on a single, non-

reciprocal category. For example, she showed a higher

competence of social reciprocity when the interviewer

first demonstrated a story of nice reciprocity for her and

a lower competence of nonreciprocal social interaction

when she later made up a story without the interviewer’s

demonstration. Labeling her as having or understanding

social reciprocity misrepresents the range of her compe-

tence, as does labeling her as having only a nonrecipro-

cal social category.

Depending on their emotional state, children also

show different emotional valences in their representa-

tions, just as Susan did in her shift to negative stories.

Maltreated children often shift the content of stories

from positive to negative, and, when they become agi-

tated, the sophistication of their negative stories deteri-

orates and remains low until they become calmer

(Ayoub & Fischer, in press; Buchsbaum, Toth, Clyman,

Cicchetti, & Emde, 1992).

These kinds of variations need to be center stage and

the focus of developmental analysis. Only by including

these variations as a function of context, culture, state,

and other key contributors to behavior can scholars build

an effective framework for explaining the many shapes

of human development. Dynamic structuralism provides

concepts and tools for founding developmental explana-

tion and description of these variations, and it encour-

ages the building of theory and method that capture the

rich complexity that is the legacy of the human species.

Dynamic Nature of Psychological Structure

What is psychological structure? Why is it important in

explanations of development? The answers depend on

assumptions about the nature of the mind and its rela-

tion to other biological, psychological, and social phe-

nomena. Psychological structure is the organizational

property of dynamic systems of activity, and analysis

of dynamic structure starts with assumptions that are

fundamentally different from the traditional view of

318 Dynamic Development of Action and Thought

structure as static form. The concept of structure in

stage theory and related viewpoints equates form with

structure and thus founders on the “discovery” of vari-

ability in development (as do most other traditional

psychological concepts). The continued dominance of

the structure-as-form paradigm has prevented an ade-

quate resolution of the crisis of variability in develop-

mental theory.

To build successful models of dynamic psychological

structure, it is essential to understand how dynamic

structure differs from static form. An essential first

step is to focus simultaneously on variability and stabil-

ity. Indeed, the neglect of variability helps ensure that

models remain static, missing the sources of order in the

variation and treating structures as static forms. Any

adequate account of psychological structure must ex-

plain not only the stability that allows systems to func-

tion and maintain themselves over time and space but

also the wide variability that arises from the dynamics

of self-organizing systems. Models of psychological

structure must specify mechanisms by which activities

are organized dynamically in relation to multiple influ-

ences that are biological, psychological, and social.

In this section, we illustrate how a dynamic struc-

tural framework deals with variability and stability si-

multaneously and thus introduces powerful explanations

of development, including cognition, social interaction,

emotions, and even brain development.

Dynamic Structure in Living Systems

All living systems—whether biological, psychological,

or social—must be organized to function. A living or-

ganism that becomes sufficiently disorganized dies. A

disorganized society collapses. A disorganized mind

leaves a person helpless in the face of everyday prob-

lems. This organizational aspect of living systems is

what we call structure, a dynamic patterning and relat-

ing of components that sustain the organized activities

that define life and living things.

To say that a system is structured or organized im-

plies that specific relations exist among its parts, sub-

systems, or processes. In the human body, for example,

the respiratory, circulatory, digestive, metabolic, and

nervous systems must all function in very specific rela-

tions to maintain the overall functioning and health of

the organism. Similarly in a complex society, the eco-

nomic system, judiciary, political/electoral system, and

government must maintain specific relationships to sus-

tain the society. In this way, dynamic structure exists

only where relationship exists, and relations among the

parts of a system provide its specific organization.

To flourish, living systems must be more than just or-

ganized. They must be dynamic. Systems must con-

stantly move and change if they are to carry out their

functions and maintain their integrity and their interre-

lations with other functioning systems. A system that

becomes static—unable to change and adapt to varying

conditions—will quickly perish. Social, psychological,

or biological systems must be able to stretch the limits

of their current patterns of organization, and even to ac-

tively guide and reorganize the relations that constitute

their structure. An organism or society that becomes in-

flexible and incapable of adaptive response to variations

in its environment will die as surely as one that becomes

disorganized. Thus, structure must be distinguished not

only from disorganization but also from static form,

which really is the antithesis of structure. Structure is

fundamentally dynamic because it is a property of liv-

ing, changing, adapting systems. Susan demonstrated

this dynamic adaptation in her variable representations

of social interactions with her father and counselor. Dy-

namic variation is a fundamental property of human ac-

tion and thought.

The human mind is a specialized living system that

participates in and with other bodily, environmental,

and social systems. The specialized function of the

human mind is to guide and interpret human activity in

relation to the world of people and objects. The activity

takes places in medias res, in the middle of things, not in

the person alone or in the brain. The objects and people

in the physical and social world of the actor are actually

part of the activity.

Moreover, living systems are agentive—self-regulating

and self-organizing, adapting and changing as a conse-

quence of goal-oriented activity, as in Susan’s activities

(Bullock, Grossberg, & Guenther, 1993; G. Gottlieb,

2001; Kauffman, 1996). In seeking its goals, a living sys-

tem is involved in multiple relations with other living

and nonliving systems, and they are part of one another’s

dynamics.

This agency and interaction lead naturally to variabil-

ity in systems. If systems were static, they would be un-

changing; but because they move and change, they give

rise to patterns of variability. The more complex a sys-

tem, the more relations are entailed by its structure and

the greater the variability it is likely to display. Human

beings show more variability in activity than lizards,

Dynamic Structuralism 319

Figure 7.2 Development as a constructive web.

Domains

Counselor Father

Mother

Development

rats, or monkeys. This variability can easily elude overly

simple theoretical models that ignore the dynamic com-

plexity and interrelationships of living systems.

Variation and Order in Development: The

Constructive Web

People unknowingly ground their concepts and activi-

ties in metaphoric frames that give meaning (Lakoff &

Johnson, 1999). Concepts and theories in science derive

from metaphoric frames in the same way as everyday

concepts, except that research systematically tests their

grounding in observation and action. Traditional static

conceptions of development in psychological structure

are closely related to the widespread cultural metaphor

of a ladder. Development is conceived as a simple linear

process of moving from one formal structure to the

next, like climbing the fixed steps of a ladder. It matters

little whether the steps of the ladder are conceived as

cross-domain stages, levels of a domain-specific com-

petence, or points on a psychometrically based scale. In

each case, the beginning point, sequence of steps, and

endpoint of the developmental process are all linear and

relatively fixed, forming a single ladder. With such a

deterministic, reductionist metaphor, it is difficult to

represent the role of constructive activity or contextual

support because there appears to be no choice of where

to go from each step. The richness of children’s devel-

opment, including the variability in their skills across

contexts, is simply lost with the ladder metaphor. Devel-

opment means just moving to the next step—an overly

simple theory that clearly does not capture the variabil-

ity that Susan showed in her stories about nice and

mean interactions.

A more dynamic metaphor for development, which

includes variability as well as stability in development,

is the constructive web (Bidell & Fischer, 1992; Fischer

et al., 2003). The metaphor of a web is useful for dy-

namic models because it supports thinking about active

skill construction in a variety of contexts and for di-

verse variations. Unlike the steps in a ladder, the strands

in a web are not fixed in a determined order but are the

joint product of the web builder’s constructive activity

and the supportive context in which it is built (like

branches, leaves, or the corner of a wall, for a spider

web). The activity of an agent in constructing a web is

particularly clear. For example, a given strand may be

tenuous at first, dependent on surrounding strands for

external support, and like the spider, the person can re-

construct it until it becomes a stable part of the web.

Also, unlike most spider webs, human developmental

webs are constructed jointly by multiple agents, not by

an individual alone, although most psychological re-

search examines individuals isolated from their social

networks. We show how people often join together to

construct parts of their developmental webs.

The separate strands in a web represent the various

pathways along which a person develops. The strands in

a web can start in a number of places, take a variety of

directions, and come out at a range of endpoints, all de-

termined by active construction in specific contexts.

The several strands composing one line may be con-

structed in a different sequential order from the strands

composing another line in a different section of the web.

At the same time, there is order in the web, including

similar orderings of spatial positions for some strands,

separations and junctions of strands, and related starting

and ending points for some strands. Using the construc-

tive web as a metaphor for devising models of develop-

ment facilitates the unpacking of variability relating to

constructive activity and context, which are conflated in

the image of a linear ladder of static structures.

Figure 7.2 depicts an idealized constructive web. The

lines or strands represent potential skill domains. The

connections between strands represent possible rela-

tions among skill domains, and the differing directions

of the strands indicate possible variations in develop-

mental pathways and outcomes as skills are constructed

320 Dynamic Development of Action and Thought

for participation in diverse contexts. Groupings of

strands represent domains of skill, such as mother, fa-

ther, and counselor, for each of the three clusters of

strands. Within each strand, people’s activities also

vary, demonstrating a developmental range (like

Susan’s) varying between high competence with contex-

tual support and lower competence without it (Fischer,

Bullock, et al., 1993; Fischer et al., 2003). In the discus-

sion that follows, the web metaphor is articulated to fa-

cilitate analysis of variability in the development of

dynamic skills.

DYNAMIC STRUCTURE IN COGNITIVE

AND EMOTIONAL DEVELOPMENT

To explain both variability and stability in development

and learning, an alternative framework is needed to re-

place the structure-as-form paradigm as a basis for re-

search and interpretation. Static conceptions of

psychological structure must be replaced with dynamic

ones such as the constructive web. Reified notions of

structures existing separately from human activity must

give way to a new understanding of structure as the dy-

namic organization inherent in the activity itself. Such a

framework is emerging in dynamic systems theory,

which is influencing a variety of fields and a growing

number of researchers. (This volume shows the extent of

the growth of dynamic systems in human development,

with a majority of chapters taking a dynamic systems

perspective.)

Common in many dynamic systems models is a shift

in the treatment of order and variation from being di-

chotomized to being intrinsically related (Hua & Smith,

2004; Kelso, 1995; Port & van Gelder, 1995; van Geert,

1998). Phenomena that were once viewed as random or

chaotic are now seen as organized in complex ways that

lead to specific patterns of variation. Descriptions and

models of the activity and change start with analysis of

relations between organization and variability in spe-

cific phenomena. For instance, the jagged patterns of

seacoasts—seemingly erratic jumbles of random ero-

sions—can be closely modeled with fractal geometry,

revealing an intrinsic organization to a geologic process

of erosion and sedimentation once thought of as disor-

derly. By recognizing that organization is related to

variability, geologists and mathematicians have been

able to create models of the dynamic organization of the

erosion process that can predict and explain the variabil-

ity observed in the changing coastline (Kruhl, Blenkin-

sop, & Kupkova, 2000). Similarly, biologists model the

structures of evolution of living organisms (Kauffman,

1993) and the dynamics of brain functioning and devel-

opment (e.g., Marcus, 2004; Polsky, Mel, & Schiller,

2004; Spruston & Kath, 2004).

Full realization of the potential of dynamic systems

analysis requires not only connecting nonlinear dynamic

concepts to psychological processes but also building

explicit dynamic models of those processes. Global con-

cepts can be powerful and useful, but ultimately they

must be tested out as models with explicitly defined

properties. Only with such models can researchers de-

termine whether the processes they hypothesize in fact

produce the dynamic patterns of development and varia-

tion that they expect (Fischer & Kennedy, 1997; Thelen

& Smith, Chapter 6, this Handbook, this volume; van der

Maas, 1995; van Geert, 1998; van Geert & van Dijk,

2002). Happily, computer-based tools including spread-

sheets such as Excel can be readily used to build explicit

dynamic models and test them against empirical data.

From a dynamic systems viewpoint, psychological

structure is the actual organization of systems of activ-

ity. It is not a separately existing entity, such as a logical

stage dictating behavior, or a preformed linguistic or

cognitive capacity awaiting actualization, but instead is

a property of human activity systems. Because real sys-

tems of activity are dynamic—constantly moving,

adapting, and reorganizing—they must be dynamically

structured. Variability is a natural consequence of sys-

tem dynamics, and because systems are organized, the

variability is not random but patterned, as evident in the

variable stories that Susan told. Just as geologists have

modeled the structures of coastal evolution and biolo-

gists have modeled the structures of evolution of living

species, developmental scientists can build models of

the dynamic structures of development and learning in

human action and thought.

To move beyond a general call to dynamic structural

analysis and model the dynamics of development

successfully, scholars need specific psychological con-

structs that support analyzing structures behind varia-

tion for particular research problems. There is not one

correct construct for a dynamic approach to psychologi-

cal structure. A number of contemporary constructs are

useful for this purpose because they have been devel-

oped specifically to facilitate analysis of variation and

organization of activities in context. The concept of

script, for example, focuses on the organization and

variation in everyday activities for storytelling, narra-

tives, goals, and recall for scripted activities in specific

contexts (Fischer, Shaver, & Carnochan, 1990; Nelson,

Dynamic Structure in Cognitive and Emotional Development 321

1986; Schank & Abelson, 1977). The concept of strat-

egy has a long history of illuminating variations in the

organization of problem-solving activity (Bruner,

Goodnow, & Austin, 1956; Siegler & Jenkins, 1989;

Siegler, Chapter 11, this Handbook, Volume 2). Con-

cepts such as apprenticeship (Rogoff, 1990), environ-

mental niche (Gauvain, 1995), and setting (Whiting &

Edwards, 1988) facilitate analysis of the dynamic social

organization of activities across contexts.

A construct that we find especially useful for facili-

tating a dynamic approach to psychological structure is

dynamic skill; it provides a useful way of integrating

many of the necessary characteristics of dynamic psy-

chological structure into a single, familiar idea (Fischer,

1980b; Fischer & Ayoub, 1994; Fischer, Bullock, et al.,

1993). This construct is based on concepts that were

central to the cognitive revolution of the late 1950s and

1960s (Bruner, 1973; Gardner, 1985), the ecological

revolution of the 1960s and 1970s (Bronfenbrenner &

Morris, Chapter 14, this Handbook, this volume; Gibson,

1979), and the emotive revolution of the 1980s and

1990s (Campos, Barrett, Lamb, Goldsmith, & Stenberg,

1983; Frijda, 1986; Lazarus, 1991). These revolutions

have emphasized, for example, the importance of goals,

self-regulation, organism-environment interaction, bias

or constraint, and the social foundations of activity.

Most importantly, Piaget (1970) and Vygotsky (1978)

insisted on activity as the basis of cognitive structures,

defined as systems of relations among activities.

In the following discussion, we explicate the con-

struct of dynamic skill, using it to articulate essential

characteristics of psychological structures. We show

how the dynamic analysis of structure can both predict

and explain specific patterns of developmental variabil-

ity, focusing on three key types of variability frequently

observed in developmental research: (1) sequence, (2)

synchrony, and (3) range. In subsequent sections, we

show how these dynamic characteristics differ from

those in static views of structure, and we describe key

methodology for studying the dynamics of change, mi-

crodevelopment in learning and problem solving, devel-

opment of emotion, and the role of brain functioning in

development of cognition and emotion.

Psychological Structure as Dynamic Skill

In ordinary English usage, the term skill both de-

notes and connotes essential characteristics of the dy-

namic organization of human activities (Bruner, 1973;

Welford, 1968). Skill is the capacity to act in an organ-

ized way in a specific context. Skills are thus both

action-based and context-specific. People do not have

abstract, general skills, but they have skills for some

specific context: a skill for playing basketball, another

for telling a children’s story, or yet another for interper-

sonal negotiation. Skills do not spring up fully grown

from preformed rules or logical structures. They are

built up gradually through the practice of real activities

in real contexts, and they are gradually extended to

new contexts through this same constructive process

(Fischer & Farrar, 1987; Fischer & Immordino-Yang,

2002; Granott, Fischer, & Parziale, 2002).

The concept of skill also helps to conceptualize the

relations among various psychological, organismic, and

sociocultural processes and to cut through artificial di-

chotomies between mind and action, memory and plan-

ning, or person and context. A skill—such as telling

children stories about emotional interactions with other

children—draws on and unites systems for emotion,

memory, planning, communication, cultural scripts,

speech, gesture, and so forth. Each of these systems

must work in concert with the others for an individual to

tell an organized story to specific children in a particu-

lar context, in a way that it will be understood and ap-

preciated. The concept of dynamic skill facilitates the

study of relations among collaborating systems and the

patterns of variation they produce and inhibits treating

psychological processes as isolated modules that ob-

scure relations among cooperating systems. To see how,

let’s consider some of the characteristics of skills.

Integration and Interparticipation

Skills are not composed atomistically but are necessar-

ily integrated with other skills. The skill of playing bas-

ketball demands that many other skills, such as running,

jumping, and visual-motor coordination, all be inte-

grated to function in a coordinated way. Integrated

skills are not simply interdependent but interparticipa-

tory. True integration means that the systems participate

in one another’s functioning. Atomistic models allow

for simple interdependence: The stones in an arch, the

trusses in a bridge, the modules in a serial computer

comprise atomistic systems in which parts are interde-

pendent but do not obviously participate in each other’s

functioning.

In contrast, the components of living systems not

only depend on one another but participate in one an-

other. Although at first this concept may seem counter-

intuitive, there are many obvious examples in familiar

processes such as human cellular or organ systems. Any

322 Dynamic Development of Action and Thought

system in the human body is composed of multiple sub-

systems whose boundaries defy definition. The cardio-

vascular system, for example, participates in the

functioning of every organ system, because every organ

depends on receiving oxygenated blood. At the same

time, the cardiovascular system includes components

from the nervous system, the muscular system, and so

forth, so that these other systems in turn participate in

the circulatory system. It makes little sense to think of

any of these systems as functioning outside the context

of the other systems: Living systems die when cut off

from the other systems with which they interparticipate.

For living systems, conceptions of structure must reflect

the interparticipation of one system in another.

Systems of activities are central parts of living sys-

tems, especially in complex systems such as human be-

ings. Activities organize into skills, which have many

interparticipating components. When Susan creates a

story of social reciprocity between the positive actions

of the doll Susan and her doll father, the actions of each

character affect each other intimately and recipro-

cally—they participate in each other. Skills normally

involve this interparticipation of components.

Context Specificity and Culture

Skills are context-specific and culturally defined. Real

mental and physical activities are organized to perform

specific functions in particular settings. The precise

way a given skill is organized—its structure—is essen-

tial to its proper functioning, as well as specific to that

skill at any moment. Good basketball players do not au-

tomatically make good baseball players; good story-

tellers in one culture do not automatically have their

stories understood and appreciated in other cultures.

The context specificity of skills is related to the char-

acteristics of integration and interparticipation because

people build skills to participate with other people

directly in specific contexts for particular sociocultural

systems. In turn, people internalize (Cole, 1996;

Wertsch, 1979) or appropriate (Rogoff, 2003) the skills

through the process of building them by participating in

these contexts; and as a result, the skills take on cultural

patterning. Similarly, component systems such as mem-

ory, perception, emotion, and even physiological regula-

tion all participate in the culturally patterned skills.

The context specificity of skills thus implies more than

simply a fit with an environment. Even systems like per-

ception or memory, which are often thought of as being

isolated from sociocultural systems, are linked to them

through the skills in which they participate; research

shows how pervasive and deep the connections are

(Greenwald et al., 2002; Mascolo, Fischer, & Li, 2003).

Self-Organization, Mutual Regulation, and Growth

Skills are self-organizing. Part of the natural function-

ing of skills is that they organize and reorganize them-

selves. These self-organizing properties go beyond

maintenance to include growth of new, more complex

skills. One of the goals of developmental science is to

analyze the processes of organization and change, which

skills undergo with development and learning. Unlike

mechanical systems that must be built and maintained

artificially through an external agency, the agency that

creates and maintains skills (and living systems in gen-

eral) resides in the activities for both individual activity

and social interaction. Construction and maintenance of

skills involves both self-regulation and mutual regula-

tion with other people, because components interpartic-

ipate. In an obvious example from human biology, as

people increase their activity level, their increased use

of energy and oxygen evokes increases in their rates of

breathing and metabolism. No outside agency is involved

in adjusting the controls for this interparticipation of

motor systems with respiratory and metabolic systems.

The living system actively adjusts itself to maintain its

own integrity.

In skills, the components regulate each other in the

same way. Susan’s and her father’s mean actions toward

each other mutually affected the other’s mean actions,

creating adaptations in content, organization, and emo-

tional tone (quality and intensity). Skills are not fixed

abilities but constantly adapting, regulated activity

structures. As Susan, her father, and her counselor act

together, they develop new skills together, coordinating

activities that were previously relatively independent to

form newly integrated wholes. Through coordination

and mutual regulation, they organize their activities into

qualitatively new, integrated systems, with sequences of

coordinations and regulations that build on each other.

Dynamic structuralism provides concepts and tools

for taking hold of this adaptive variability to uncover the

order behind the variations. One of the central discover-

ies is a common scale of hierarchical complexity that or-

ders the variations.

A Common Ruler for Skill Development

A key ingredient for advancing developmental science is

common rulers (scales) for measuring change and varia-

Dynamic Structure in Cognitive and Emotional Development 323

Figure 7.3 Developmental cycles of levels and tiers of

skills. Development proceeds through 10 levels of skills

grouped into three tiers between 3 months and adulthood. The

ages of emergence are for optimal levels, the most complex

skill that a person can perform with social-contextual sup-

port, based on research with middle-class American or Euro-

pean children. They may well differ across social groups.

There is some evidence for an additional tier of innate action-

components in the first few months of life. Sources: From “A

Theory of Cognitive Development: The Control and Construc-

tion of Hierarchies of Skills,” by K. W. Fischer, 1980b, Psy-

chological Review, 87, pp. 477–531; and “The Big Picture for

Infant Development: Levels and Variations” (pp. 275–305), by

K. W. Fischer and A. E. Hogan, in Action in Social Context:

Perspectives on Early Development, J. J. Lockman & N. L.

Hazen (Eds.), 1989, New York: Plenum Press.

Abstractions

23–25 years

18–20 years

14–16 years

10–12 years

6–7 years

3�–4� years

2 years

11–13 months

7–8 months

3–4 months

Ab4. Principles

Rp3. Systems

Rp2. Mappings

Sm4/Rp1. Single Representations

Ab3. Systems

Ab2. Mappings

Rp4/Ab1. Single Abstractions

Sm3. Systems

Sm2. Mappings

Sm1. Single Actions

Tiers

Levels

Age of Emergence

Actions

Representations

tion in activity, similar to the Centigrade or Fahrenheit

scale for temperature and the meter or foot for length.

These scales should be grounded in properties of natural

response distributions and applicable across tasks and

domains. However, psychological measurement has pro-

duced mostly arbitrary scales based on one situation

such as those for intelligence, achievement, and person-

ality tests. They do not use naturally occurring response

distributions but statistical models assuming stable

(static) ability and normal distributions (van Geert &

van Dijk, 2002; Wahlsten, 1990), and they assess behav-

ior in one situation, the test. A more useful scale allows

measurement of different skills in various situations and

is not tied to one situation or assessment instrument.

Temperature and length can be measured in many ways

in virtually any situation.

Fortunately, the measurement problem has now

changed with the discovery of a common scale for be-

havioral complexity that captures a central dimension of

both long-term development and short-term change

(Commons, Trudeau, Stein, Richards, & Krause, 1998;

T. L. Dawson & Wilson, 2004; Fischer, 1980b; Fischer

& Immordino-Yang, 2002). Research with various meth-

ods has produced evidence for the same scale, marked

by clusters of discontinuities such as sudden changes in

growth patterns and gaps in Rasch scaling. Analysis of

growth curves has documented these patterns (Fischer

& Rose, 1999; van Geert, 1998), and Rasch (1980) scal-

ing of interview and test data has shown remarkably

consistent evidence of the same patterns of discontinuity

(Dawson, 2003; Dawson, Xie, & Wilson, 2003), forming

a scale of at least 10 levels of hierarchical complexity, as

shown in Figure 7.3. The scale relates to the outline of

developmental stages that Piaget (1983) described, but

the levels on the scale are better grounded empirically,

and performance varies across the scale instead of being

fixed at one point at each age. The scale also has impor-

tant similarities to those suggested by Case (1985),

Biggs and Collis (1982), and others. Interestingly, dis-

continuities in growth of brain activity seem to follow

the same scale, as described later in the chapter (Fischer

& Rose, 1996).

Many developmental scientists have posited stages,

some of which match some of the levels (Biggs & Collis,

1982; Case, 1985; Halford, 1982; McLaughlin, 1963),

but these alternatives have not been based on clear em-

pirical criteria for what constitutes a stage or level—and

what does not (Fischer & Silvern, 1985). Typically,

these investigators have merely described a sequence of

posited cognitive reorganizations without specifying

empirical criteria for stages or levels, except for loosely

defined “qualitative change” and an approximate devel-

opmental sequence.

The skill scale in Figure 7.3 begins with sensorimo-

tor actions, which are coordinated through several com-

plexity levels to eventually form representations, which

are in turn coordinated through several levels to form

abstractions, which continue to develop into adulthood.

The larger growth cycles of actions, representations, and

abstractions are called tiers (left column of the figure),

and the specific changes marked by clusters of disconti-

nuities are called levels (middle column). The ages in the

right column indicate when skills at a level first emerge

under conditions that support optimal performance.

Each level has a characteristic skill structure, as shown

in Figure 7.4, and similar structures recur in each tier,

reflecting a dynamic cyclical growth process. The struc-

tures begin with single sets organized as actions, repre-

sentations, or abstractions. A person coordinates and

differentiates these sets to form mappings, which in

turn are coordinated and differentiated to form systems.

324 Dynamic Development of Action and Thought

Figure 7.4 Cycle of levels of development for a tier: cube

models and skill structures. The fourth level marks the cul-

minating structure for a tier and the formation of a new unit

for the next tier, as shown by the two skill formulas for Level

4/1: Level 4 actions form Level 1 representations, and Level 4

representations form Level 1 abstractions.

In skill formulas, brackets mark a skill structure; and

each letter denotes a skill component, with a large letter des-

ignating a main component (set) and a subscript or super-

script a subset of the main component. A line connecting sets

(—) = A mapping relation, a single-line arrow (↔) = A rela-

tion forming a system, a double-line vertical ( ) arrow = A

relation forming a system of systems, and a greater than sym-

bol (>) = A shift from one skill to another without integra-

tion. Such shifts between skills can occur at every level,

although for simplicity a shift is shown only at the first level.

For skill formulas in later figures and text, bold letters = Sen-

sorimotor actions, italic letters = Representations, and script

letters = Abstractions.

⇔

A

E

B

E

>

A

E

B

E

Level 2: Mappings

A

E

F

B

E

F

Level 4/1: Systems of Systems/Single Sets

B

E

F

A

E

F

C

G

H

D

G

H

T

X

≡

Level 1: Single Sets

Level 3: Systems

Development

At the fourth level of each tier, the person coordinates

and differentiates systems to form systems of systems,

thus constructing a new unit that begins the next tier—a

single set of a new type. At the tenth level, the person

constructs single principles, and there is as yet no evi-

dence for further levels marked by clusters of disconti-

nuities beyond single principles (Fischer et al., 2003).

Contrary to static approaches to development and

learning, the levels on the scale do not indicate the use

of one psychological structure or module across do-

mains, like one of Piaget’s (1985) generalized logical

structures or Chomsky and Fodor’s (1983) modules.

People do not use the same structure across situations,

but they build skills along the same scale. The processes

of growth and variation produce skills that fit a common

scale across tasks and domains, but the skills used dif-

fer, being dynamically adapted to context, emotional

state, and goal. The complexity of separate activities

varies in similar ways for different contexts and states.

Think of temperature, for which physicists discovered a

common scale over the last several hundred years. The

same scale can be used to measure the temperature in

the sun, Antarctica, a refrigerator or furnace in New

York, a person’s mouth, or the bottom of the ocean.

Thermometers measure with a common scale across rad-

ically different situations and methods, even with great

differences in the ways that heat and cold occur.

In this way, skills are organized in multilevel hierar-

chies that follow the scale in Figures 7.3 and 7.4. People

construct skills through a process of coordination, as

when 5-year-old Susan built stories about emotionally

loaded social interactions that coordinated multiple ac-

tions into social categories and then coordinated social

categories into reciprocal activities. Susan used a skill

hierarchy in which individual pretend actions (Sm3 sys-

tems of actions) were embedded in social categories

(Rp1 single representations), which were in turn embed-

ded in socially reciprocal activities (Rp2 representa-

tional mappings). Existing component skills, controlling

activities in specific contexts, were intercoordinated to

create new skills that controlled a more differentiated

and integrated range of activities. In the newly inte-

grated skills, the component skills still functioned as

subsystems in the new skill as a whole. They also could

still be used alone, as when Susan dropped back to sim-

pler actions with less contextual support or with emo-

tional upset. We use representations of positive and

negative social interactions to ground the explanation of

dynamic skills and to illustrate how the skills in the di-

agrams both develop in the long term (macrodevelop-

ment or ontogenesis) and vary from moment to moment

(microdevelopment).

The skill hierarchy in the scale embodies the principles

of self-organization and interparticipation of dynamic

Dynamic Structure in Cognitive and Emotional Development 325

systems. As skills become integrated and differentiated

at later levels, the component skills subordinate them-

selves to new forms of organization and mutual regula-

tion. The very process of creating new skills through

self-organizing coordination leads to a multileveled hier-

archical structuring of living skills. Indeed, “hierarchy”

in this sense has a special meaning. Computer programs,

for example, can be arranged hierarchically in the sense

that lower-level outputs feed higher-level procedures, but

this organization does not typically involve interpartici-

pation and self-organization.

Generalization through Construction

Susan built her skills for representing positive and

negative social interactions in one context, but she nat-

urally tried to generalize those skills across related

contexts—for example, using the skills for represent-

ing interactions with her father to build representa-

tions of interactions with her counselor. The process of

skill construction through coordination is closely re-

lated to skill generalization, and the complexity scale

can illuminate both. Generalization of mental and

physical activity involves specific building of general-

ized skills driven by the goal-oriented activity of an

individual or ensemble (a few people working closely

together), especially for socially constructed domains

such as literacy, mathematics, and science. General-

ization in these domains is not a predetermined, innate

outcome waiting for development to catch up with it,

as some nativists would have it (Baillargeon, 1987;

Fodor, 1983; Spelke, 1988). Several mechanisms of

generalization of dynamic skills through coordination,

differentiation, and bridging from simple to complex

have been specified with some precision (Fischer &

Farrar, 1987; Fischer & Immordino-Yang, 2002;

Siegler & Jenkins, 1989). Studying microdevelopment

is an especially powerful way of analyzing processes

of dynamic generalization, as we describe in a later

section to illuminate how learning general knowledge

takes a long time.

Building a Constructive Web for Positive and

Negative Social Interactions

The complexity scale combines with the constructive

web in Figure 7.2 to support analyzing psychological

structure in dynamic terms. Unlike the traditional

ladder of development, the web highlights integration,

specificity, multiple pathways, active construction, and

other central properties of skill development (Bidell &

Fischer, 1992; Fischer et al., 2003). Building a web is a

self-organizing process in which a person coordinates

and differentiates various activities along the complex-

ity scale. The strands in a web are the joint product of

the person’s constructive activity and the contexts in

which skills are built, including the other people who

coparticipate in building them.

We use stories about nice and mean social interac-

tions to illustrate properties of the constructive web

and its relation to dynamic properties of cognitive and

emotional development. Telling a story or narrative is a

fundamental human activity. To produce a specific

story or narrative, a child needs to organize activities

in a scriptlike way, following specific patterns of se-

quencing of events (Bruner, 1990; Fischer et al., 1990;

Ninio & Snow, 1996; Schank & Abelson, 1977). This

organization helps impart meaning to the narrative, as

with 5-year-old Susan’s stories about interaction be-

tween a girl and her father. Without this script organi-

zation, the story becomes a meaningless jumble; for

example, it becomes unclear who is being nice to whom

and why, or who is hurting whom and why. Yet the or-

ganization of the storytelling activity must also be

flexible, so that a storyteller can create new versions

for changing situations and people, thereby communi-

cating different ideas and feelings, as Susan changed

her stories in relation to her emotional state and to the

contextual cuing and support she received from the

adult interviewer.

Like other skills, the complexity and organization of

story skills varies widely with the dynamics of the con-

structive activity, including story complexity, emotional

state, and social-contextual support from other people.

The skill scale illuminates this variation by providing

a ruler for analyzing and comparing these variations.

When 5-year-old Susan is in a positive mood and has

support from her counselor, she organizes a complex

story about positive interactions. A few minutes later

when she is emotionally stressed, she no longer pro-

duces a complex positive story, even with support from

the counselor, but instead tells an equally complex nega-

tive story. When the counselor does not provide contex-

tual support, Susan can organize only a simpler positive

or negative story. In addition, the form of narrative

organization varies across cultural groupings and

discourse communities because individuals construct

different narrative skills to participate in different

culturally patterned communicative activity. Susan’s

326 Dynamic Development of Action and Thought

Figure 7.5 Developmental web for nice and mean social interactions. The numbers to the left of each set of brackets indicate

the step in complexity ordering of the skill structures. The words inside each set of brackets indicate a skill structure. The left

column designates the first step at each skill level.

NICE

MEAN

NICE and MEAN

Rp1

Level

Rp2

Rp3

1

ME

NICE

1

YOU

MEAN

>

2

YOU

MEAN

ME

NICE

3

YOU

NICE

ME

NICE

3

MEAN

ME

ME

NICE

3

YOU

MEAN

MEAN

ME

>

4

YOU

MEAN

MEAN

ME

YOU

NICE

ME

NICE

5

YOU 1

MEAN

YOU 2

MEAN

MEAN

ME

5

ME

NICE

YOU 1

NICE

YOU 2

NICE

6

YOU 1

NICE

YOU 2

MEAN

MEAN

NICE

ME

7 YOU

NICE 2

NICE 1

ME

NICE 2

NICE 1

7 YOU

MEAN

NICE

ME

MEAN

NICE

7 YOU

MEAN 2

MEAN 1

ME

MEAN 2

MEAN 1

stories fit her cultural community, but would have to be

reorganized to fit others.

Webs and Biases

Figure 7.5 shows a developmental web for stories about

positive and negative social interactions in American

children of diverse ethnicity and social class (Ayoub &

Fischer, in press; Fischer & Ayoub, 1994). When chil-

dren play, they commonly act both nice and mean to

each other, and like 5-year-old Susan, they readily act

out and tell stories about positive and negative interac-

tions between peers. The web has three distinct strands

Dynamic Structure in Cognitive and Emotional Development 327

organized by emotional domains of different valence—

nice on the left, mean on the right, and the combination

of nice and mean in the middle. The tasks are ordered in

steps by skill complexity, marked by the numbers next to

each skill structure. There are normally multiple steps

per level, marking the distinct points in the construction

process that can be discriminated for a particular situa-

tion, which can vary in number. The levels are indicated

in the left-hand column.

In the research, children between 2 and 9 years of age

told stories about two or three boys or girls playing to-

gether, with each story reflecting one of the three emo-

tional domains. One character usually had the name of

the child telling the story, and the others represented his

or her friends or siblings; in some studies, the characters

had the names of unknown children. In a separate as-

sessment, children also told similar stories about parent-

child interactions.

Later steps generally involve more inclusive skills,

constructed by the coordination and differentiation of

lower-level components. For example, in step 3, the

story involves a mapping between two instances of nice-

ness (or meanness), similar to the reciprocity stories of

Susan: One doll acted mean (or nice) to a second doll

who because of the first doll’s action, acted mean (or

nice) in return. In Figure 7.5, each diagram of YOU or

ME acting NICE or MEAN represents a story with a cer-

tain skill structure, varying across the three skill levels

of single representations, mappings, and systems. The

structure

represents a mapping for reciprocity: If you are nice

to me, I will be nice to you. Vertical arrows be-

tween specific story structures in Figure 7.5 indicate

developmental sequencing for those stories, as when

steps 3, 4, and 5 in the left column form part of a path-

way along the strand for nice. The skill formulas focus

on the central elements that children had to control in

the nice/mean stories: roles (you or me), emotional va-

lence (nice or mean), and relations between roles

(shifts without coordination, mappings, and systems).

Like structures in any living system, these elements

subsume many additional components hierarchically

within them such as actions, perceptions, feelings,

goals, and social expectations.

Thus, each step in Figure 7.5 represents a different

level of skill at conceptualizing relations among social

(1)

YOU

NICE

ME

NICE

interactions. Children’s stories develop along strands

for each of the content domains of nice, mean, and

nice-and-mean in combination. When stories are paral-

lel from left to right, they emerge at approximately the

same time in development. Their development also

shows many connections among the strands.

In accord with the general tendency for researchers

to neglect within-person variation and emphasize

between-person variation, people sometimes misunder-

stand this developmental web, interpreting it to mean

that different children are developing along each strand.

To the contrary, each child develops simultaneously

along each of the strands in the web in Figure 7.5. That

is, each child is simultaneously developing understand-

ings about positive valence (how nice interactions

occur), negative valence (how mean interactions occur),

and combined valence (how nice and mean can be com-

bined in an interaction). When the three strands are all

closely parallel, with no clear bias toward one or the

other, then the web looks like Figure 7.5, with complex-

ity as the primary determinant of developmental order-

ing. Steps of the same complexity are parallel in the

web, independent of valence.

One characteristic of emotions, however, is that peo-

ple typically show biases in their actions and thoughts.

Biases toward certain action tendencies are one of the

defining characteristics of emotions, as is discussed in

the later section on Emotional Development. Emotional

biases often have strong effects on a developmental web;

they shift relations between strands, and they change de-

velopmental orderings. For the nice-and-mean web, one

far-reaching emotional bias is a general favoring over

time of one pole of evaluation—toward positive (nice) or

negative (mean). One of the most strongly established

findings in social psychology is that most people show

positive biases in their activities and evaluations, espe-

cially for attributions about themselves (Higgins, 1996;

Osgood, Suci, & Tannenbaum, 1957). Figure 7.6 shows

a global bias toward the positive.

Although positive biases are pervasive, there

are also many instances of negative biases. Powerful

biases toward the negative can be produced by trauma

such as child abuse (Ayoub, Fischer, & O’Connor,

2003; Westen, 1994) and by implicit attitudes (Green-

wald et al., 2002). When children show a strong and

persistent bias toward the negative and against the pos-

itive, their entire developmental web is shifted (biased)

in the opposite direction than in Figure 7.6—toward

the negative pole. That is, mean interactions are under-

stood earlier than nice ones, and the combination of

328 Dynamic Development of Action and Thought

Figure 7.6 Developmental web biased toward nice interactions. This web includes only the first two-thirds of the skills from

the web in Figure 7.5.

NICE

MEAN

NICE and MEAN

1

ME

NICE

3

YOU

NICE

ME

NICE

>

2

YOU

MEAN

ME

NICE

1

YOU

MEAN

5

ME

NICE

YOU 1

NICE

YOU 2

NICE

3

MEAN

ME

ME

NICE

>

4

YOU

MEAN

MEAN

ME

YOU

NICE

ME

NICE

3

YOU

MEAN

MEAN

ME

5

YOU 1

MEAN

YOU 2

MEAN

MEAN

ME

6

YOU 1

NICE

YOU 2

MEAN

MEAN

NICE

ME

nice and mean is delayed as well. A number of abused

children and adolescents show an alternative develop-

mental pathway based on this bias toward the negative

(Fischer et al., 1997; Rappolt-Schlichtman &

Ayoub, in press). Besides the long-term effects of expe-

rience, there are short-term within-person effects as a

function of context, mood, and similar factors, as when

being in a negative mood leads to a bias toward negative

stories. In this way, developmental webs can be use-

ful for representing variations in developmental path-

ways not only between people but also within a person

over time.

Modeling Nonlinear Dynamic Growth in a Web

Besides the representations of weblike relations between

steps and strands like those in Figures 7.2 and 7.5,

various tools can be useful for analyzing different prop-

erties of development. One example that can be particu-

larly powerful is mathematical modeling of growth

functions (Singer & Willett, 2003). Each strand in a web

Dynamic Structure in Cognitive and Emotional Development 329

Figure 7.7 Growth functions showing a bias toward nice in-

teractions. Skill Step refers to the complexity ordering in

Figure 7.5. Level refers to the level of hierarchical complex-

ity in Figure 7.3.

0

1

2

3

4

5

Age in Years

2

1.5

0

2.5

3

3.5

4

Skill Step and Lev

el

Rp1

Rp2

Nice

Nice and Mean

Mean

can be described in terms of its growth function, which

in this case is represented by a nonlinear dynamic

growth model (Fischer & Kennedy, 1997; van Geert,

1991, 2003). Figure 7.7 shows an example of growth

curves produced by the model for each of the three

strands.

The growth model includes a global positive bias like

that in Figure 7.6, and under certain conditions, it also

produces stagelike jumps in development, which are dis-

cussed in the next section. Complexity scaling provides

the metric for quantifying growth of the strands, with

scaling tools provided by dynamic skill theory. The

graph clearly represents the bias toward positive valence

and away from negative and combined valences, empha-

sizing the quantitative advantage of the nice strand over

the others. The graph also highlights the fits and starts

in growth and the relations between them—something

that is not evident in the web diagram. However, this

quantitative graphing de-emphasizes the ordering rela-

tions among specific story structures, which are clearly

marked in Figures 7.5 and 7.6. Different tools for analy-

sis of developing activity structures capture different

properties of the structures, and no single tool captures

all important aspects.

How Dynamic Skills Explain Variability

in Development

The characteristics of skills, including the weblike pro-

cess of skill construction, can help both explain and

predict patterns of variability that have eluded tradi-

tional static accounts of psychological structure. In this

section, we show how three basic forms of systematic

developmental variability—(1) complexity level, (2) se-

quence, and (3) synchrony—can each be explained by

the characteristics of dynamic skills. In a subsequent

section, we consider issues of methodology and mea-

surement used in the precise description and prediction

of variability in development.

Developmental Range: Optimal and Functional Levels

Children (and adults) routinely perform across a range

of skill levels, like Susan telling stories about nice and

mean at two different levels with her counselor. A fun-

damental error stemming from static conceptions of

psychological structure is that each individual is treated

as “possessing” one fixed level of structure, either

across domains or in a domain, as if cognition were a

sealed bottle with a fixed level of liquid in it. From this

point of view, an individual’s behavior is expected to be

homogeneously consistent with the fixed level of cogni-

tion such as the number of items that a child can sustain

simultaneously in working memory. Deviations from

this fixed level then seem mysterious and appear to call

for complicated explanations. Often the deviations are

ignored, as researchers mistakenly use methods that

sum across individuals, activities, and contexts and treat

true variations in level as errors of measurement (Estes,

1956; Fischer, Knight, & Van Parys, 1993; Skinner,

1938; van Geert & van Dijk, 2002).

A person possesses different competences in differ-

ent contexts and emotional states. The types and com-

plexities of organization found in dynamic skills are

always changing because (a) people constantly vary

their activities as they adjust to varying conditions and

coparticipants, and (b) people commonly reorganize

their skills to deal with new situations, people, and

problems. For instance, a tennis player plays at top level

one day—after a good night’s rest, on an asphalt court,

against a well-known opponent. The same player plays at

a much lower level the next day, with a bad night’s sleep,

on a clay court, against a new adversary. This reduction

in the player’s skill level is a real change in the organi-

zation of activity, not a departure from some underlying

stage or competence that is the “real” thing. The person

unconsciously changes the actual relationships among

the participating systems of perception, motor anticipa-

tion, motor execution, memory (for instance, of the

other player’s strengths), and so on. These relations

330 Dynamic Development of Action and Thought

constitute the dynamic structure of skill. The level of

organization of tennis skills varies because coordina-

tion among the systems is different on the 2 days. To

posit any additional layers of abstract competence or

stage structure to explain this variation is unnecessary,

as it is accounted for by the dynamic properties of real

activity systems.

Comparable variations in skill level occur in most

skills, from playing tennis to interacting socially, plan-

ning a party, and reasoning about scientific or literary

questions. Vygotsky (1978) spoke of the zone of proxi-

mal development (ZPD) and the variation between

performances as a result of presence and absence of

scaffolding by an expert. Our research has documented

an important principle of variation in this zone: the de-

velopmental range introduced earlier, which is the inter-

val between a person’s best performances with and

without social contextual support in some domain.

Susan showed a developmental range in her construction

of stories about nice and mean interactions.

In a study of nice and mean stories, 7-year-old chil-

dren telling stories under conditions of high and low

social-contextual support showed a consistent develop-

mental range, repeatedly changing to a high level with

support and a lower level without it, as shown in Figure

7.1 and Table 7.1 (Fischer, Bullock, et al., 1993). A typ-

ical 7-year-old produced a highest story at step 3 (Level

R2, representational mappings) under low-support con-

ditions but achieved step 6 (Level Rp3, representational

systems) under high-support conditions. The interval

between these two developmental levels (a child’s devel-

opmental range for this domain) is indicated in Table

7.1, which is based on the data in Figure 7.1. The high-

est skill level when functioning independently (under

low support) for a given domain is referred to as the

functional level. The highest level with high-support

conditions is the optimal level.

The interval of variation for a given skill can extend

even farther, as suggested in Table 7.1. Social support

often goes beyond prompting or modeling to actual co-

participation in a task (also called scaffolding), where,

for example, an adult takes on acting out the role of one

of the dolls in a story with a child. With scaffolding, the

level of task performance can be extended several steps

upward because psychological control of the activity is

shared with an expert. In contrast, circumstances such

as emotional stress, fatigue, distraction, or interference

by a coparticipant can lead a person’s skill level to fall

below his or her functional level.

Developmental range seems to characterize perfor-

mance across most tasks, ages, and cultures, and it

grows larger with age, at least through the late twenties

(Fischer, Bullock, et al., 1993; Fischer et al., 2003).

Most people experience the developmental range di-

rectly when they learn something new with a teacher or

mentor. With the prompting of the teacher, they under-

stand a new concept or control a new skill at a relatively

high level. Without the prompting, their level of skill

drops precipitously such as when they leave the class-

room and try to explain the new concept to a friend who

knows nothing about it.

A study of Korean adolescents’ conceptions of

themselves in relationships illustrates the striking gap

that commonly occurs between optimal and functional

levels, as shown in Figure 7.8 (Fischer & Kennedy,

1997; see also Harter & Monsour, 1992; Kennedy,

1991). In this study, adolescents participated in the

Self-in-Relationships Interview (SiR), which assessed

developmental level under two conditions (described

in more detail in the section on Methodology of Dy-

TABLE 7.1 Developmental Range of a 7-Year-Old Telling a Story with

Varying Social Support

Step

Skill Level

Performance Level

Social Support

1

Rp1

2

3

Rp2

Functional level

None

4

5

6

Rp3

Optimal level

Priming through Modeling, etc.

7

8

Ab1

Scaffolded level

Direct participation by adult

9

Note: Functional and optimal levels are upper limits on performance, which show sta-

bility for a task. Scaffolded level involves a range of performance indicated by the

vertical line on the left, with the specific step depending on the nature of the scaf-

folding in combination with the 7-year-old’s skill.

Dynamic Structure in Cognitive and Emotional Development 331

Figure 7.8 Range of developmental levels for Self-

in-Relationships Interview in Korean adolescents.

13

12

11

10

9

8

0

1

2

3

4

5

6

Mean Step and Lev

el

Grade

Abstract Systems

Abstract Mappings

Single Abstractions

Rp3

Ab1

Ab2

Ab3

Optimal Level

Functional Level

0

Figure 7.9 Developmental range of a web for two

relationships.

Mother Best Friend

Automatic

Functional Level

Optimal Level

Scaffolded Level

Degree of Support

Development

namic Structural Analysis). For the optimal-level

condition, high support involved the construction

by each adolescent of a detailed diagram of his or her

own descriptions of self in several different relation-

ships such as with mother, father, sibling, best friend,

and teacher. The diagram as well as the interview

questions supported optimal performance by prompt-

ing key components of skills. For the functional-level

condition, low support involved a relatively open-ended

interview that was similar to most traditional assess-

ments of self for adolescents; they were simply asked

to describe what they were like in each relationship and

how their descriptions related to each other. There was

no prompting of key skill components.

The constructive web provides another useful way of

portraying variability in developmental level. Figure

7.9 represents a developmental web for an individual’s

conceptions of self in two important relationships,

mother and best friend. Along each strand the heavy

solid line indicates a well established, highly automa-

tized skill for a given context. An individual’s perfor-

mance drops to this level in circumstances of high

stress, fatigue, or interference. The thinner solid strand

represents the functional level of independent control

under normal conditions for this context—a level of

skill organization that is well established but less auto-

matic. The optimal-level skills indicated by the dashed

lines are still under construction, occurring when the

person receives modest contextual support such as

modeling or prompting. Finally, the dotted lines indi-

cate a skill level that the individual has recently begun

to construct, in which the person can hold the compo-

nent skills in an integrated structure only if there is

direct scaffolding, coparticipation of a more capable

partner.

From this perspective, it is easy to see why skill

levels vary over a wide range. The variation is a direct

consequence of the active, constructive, and context-

embedded nature of human activity. As Figures 7.8 and

7.9 suggest, adolescents’ conceptions of themselves in

relationships are not fixed capacities but multilevel

structures of dynamic skills under construction. Skills

early in a particular developmental sequence are better

integrated and more stable across time and conditions

than skills more recently constructed or just starting to

be constructed. Variability in the organization of a per-

son’s skill at holding in mind and organizing the events

and characteristics of a social relationship are a natural

consequence of these constructive dynamics. There

is no need to invoke explanations in terms of formal

stage structures or hidden competences hovering over

and guiding activities. Variability is explained by con-

structive dynamics. The task is to build theoretical

models and methods for describing and analyzing these

dynamics.

The Dynamics of Stages and Developmental Synchrony

Besides explaining sources of variability in level, the

concept of dynamic skill also provides a framework for

facilitating analysis of processes of change in construc-

tive dynamics. Specifying the conditions that lead to

332 Dynamic Development of Action and Thought

variability, as in developmental range, allows the control

and use of the conditions to analyze patterns of change.

We have employed this control of conditions to illumi-

nate a classic argument about processes of change, the

stage debate (Bidell & Fischer, 1992). Traditionally the

dialogue about stage has not always been informative,

amounting to assertion without accommodation:

Stage proponent: “There are stages of cognitive

development.”

Nonstage proponent: “No, there are no stages.”

Stage proponent: “Yes, there are.”

Nonstage proponent: “No, there aren’t.”

Instead of arguing about whether stages exist, dynamic

skill analysis provides tools for specifying the conditions

for stagelike change and those for continuous, nonstage-

like change. Stages both do and do not exist, depending

on the dynamics of the conditions of activity!

In the study of Korean adolescents, dynamic skill

theory was used to predict the conditions and age inter-

vals when growth shows discontinuous jumps in level

versus smooth change. High-support conditions were

predicted to produce two discontinuities marking the

emergence of two new levels of coordination of ab-

stractions. Figure 7.8 shows the predicted difference in

growth functions: Optimal-level growth spurted twice,

at grades 11 and 13, which are comparable to the

ages of optimal-level spurts found in research with

American and Chinese samples (Cheng, 1999; Fischer

& Kennedy, 1997; Harter & Monsour, 1992; Wang,

1997). Researchers using the skill theory framework

have observed similar patterns in other types of skills,

in age groups ranging from preschool to adulthood

(e.g., Corrigan, 1983; Fischer & Hogan, 1989;

Fischer et al., 2003; K. Kitchener, Lynch, Fischer, &

Wood, 1993). In each case, the developmental spurt is

associated with a major transition in skill level such as

the transitions to abstract mappings and abstract sys-

tems under optimal conditions in Figure 7.8. When op-

timal and functional levels are lumped together, this

discontinuity is masked because the developmental

function produced is effectively an average of two dif-

ferent developmental functions, a process that in-

evitably masks the true growth functions. In addition,

there is much evidence of other kinds of discontinuities

such as gaps in Rasch scaling and changes in brain-

wave patterns at similar points along the hierarchical

complexity scale (Dawson, 2003; Dawson et al., 2003;

Fischer & Rose, 1996).

As this and many other examples demonstrate, the

developmental level of behavior varies with assessment

context, coparticipant, state of arousal, emotional state,

and goal, just to name a few of the most obvious sources

of variation. Some researchers have argued that these

variations demonstrate an absence of developmental

stages (Brainerd, 1978; Flavell, 1982; Thelen & Smith,

1994), but these arguments overlook the order in the

variability. The organization of behavior develops sys-

tematically, and it also varies from moment to moment.

These facts are contradictory only for overly simple

concepts of stage and variation. Real behaviors—and

real neural networks as well—function not at a single

level but in a range or zone (A. Brown & Reeve, 1987;

Fischer, Bullock, et al., 1993; Grossberg, 1987; Vygot-

sky, 1978). Research to test for stagelike change must

take this range into account and analyze which parts of

the variation show stagelike characteristics and which

do not. Only then will the field move beyond endless ar-

guments in which protagonists focus on only part of the

variation and thus draw half-baked conclusions.

The separation of optimal and functional is one ex-

ample of the way a dynamic skills framework permits

the prediction and explanation of patterns of variability

that have typically been ignored or explained away by

theories relying on static stage or competence models

of psychological structure. Although researchers may

differ with the specific interpretation given to a phe-

nomenon like the discontinuities in optimal level, the

constructive-dynamic framework described here makes

it possible to debate the issues empirically, by providing

concepts and research methodologies to control and ma-

nipulate variations in the developmental process. (These

methodologies are described throughout this chapter;

see also the section on Methodology.)

An important part of “stage” is the expectation of

high developmental synchrony. Stage theories predict

high stability across contexts in the level of performance

an individual will display. The idea of a “hard stage,” an

underlying logical system pervading the mind at a given

stage (Kohlberg, 1984; Piaget, 1985), implies that a given

person should perform logically equivalent tasks at the

same time regardless of state or context—say, Piaget’s

tasks of conservation of liquid and classification of

shapes in matrices. It is as if Piaget touched children’s

heads on their seventh birthday, and instantly they were

transformed into concrete operational thinkers. This

strong “point synchrony” (simultaneous development of

new levels across domains) is seldom empirically sup-

ported (Fischer & Bullock, 1981). Instead, children and

Dynamic Structure in Cognitive and Emotional Development 333

adults show a high degree of variability in levels across

tasks and contexts, even with tasks that are logically sim-

ilar. For example, children who understand tasks for

conservation of number frequently fail tasks for conser-

vation of liquid even when the procedures and questions

are similar.

On the other hand, there is evidence of real develop-

mental synchrony as well when dynamic concepts are

used to analyze how and when synchrony does and does

not occur. Equivalent concepts show what is sometimes

called “interval synchrony,” appearing not at the same

time but within a relatively short time interval of each

other. Moreover, this interval is much smaller for con-

cepts about closely related topics measured in similar

tasks, especially when there is a clearly defined concep-

tual structure that is ecologically valid (T. L. Dawson &

Gabrielian, 2003; K. Kitchener & King, 1990; Pirttil�-

Backman, 1993). The disparity in intervals between

concepts drops as differences in content, context, and

concept are reduced. Case and his colleagues (1996)

have even shown that, with a well-defined central con-

ceptual structure, teaching the structure increases the

degree of synchrony across domains to the point that it

sometimes accounts for approximately 50% of the vari-

ability—a remarkably large effect indicating high inter-

val synchrony. Lamborn, Fischer, and Pipp (1994)

demonstrated that development of understanding of spe-

cific moral concepts such as honesty and kindness re-

lated closely to relevant social problem-solving skills

but not to other problem-solving skills.

The combination of systematic variability and syn-

chrony is hard to explain with static concepts of psycho-

logical structure such as stage or competence. Piaget

and other hard-stage theorists initially waved away evi-

dence by arguing that different tasks posed different

forms of resistance to structures of logic. The resulting

decalages (time gaps) were said to result from different

kinds of resistance, but the processes by which resist-

ance functioned were never explained (Kohlberg, 1969;

Piaget, 1971).

The principles of constructive dynamics explain pat-

terns of variation in stage patterns and synchrony in a

straightforward manner:

• Skills are constructed hierarchically by integrating

earlier skills into a more inclusive whole.

• Skills vary across multiple levels for each individual

depending on context, goal, state, support, and

other factors. An important example is the develop-

mental range.

• Skills are constructed for participation in specific

tasks and contexts and over time can be generalized

to others through specific generalizing activity

(Case et al., 1996; Fischer & Farrar, 1987; Fischer &

Immordino-Yang, 2002).

Even in the simple diagram of two domains in Figure

7.9, it is obvious that among the functional, optimal, and

scaffolded levels, some skills will be the same across do-

mains, and others will be different for the same domain.

Taken together, these principles help explain how inter-

val synchrony occurs as well as how people build gen-

eral skills. This process is elaborated later in the section

titled: Building Structures: Transition Mechanisms and

Microdevelopment.

The scale for hierarchical complexity in Figures 7.3

and 7.4 provides a metric for assessing greater or lesser

synchrony, moving beyond all or none arguments. For

many related skills, levels do not show complete asyn-

chrony but are relatively close even when they differ.

The growth functions for nice and mean in Figure 7.7 il-

lustrate how the same growth curves can simultaneously

show similarities and differences in the ages of change.

Stepping back to look at the broad sweep of change

makes the synchronies evident; stepping close to look at

the details of change highlights the disparities. Each

new skill at a higher level is built from similar lower-

level skills: Each extension of a skill to a new level is a

constructive generalization constrained by the compo-

nent skills available. There is no need to invoke perva-

sive logical structures or innately determined formal

constraints to account for interval synchrony in develop-

ment. The dynamics of the construction of skills in con-

text explain both the variability and the synchrony found

in patterns of variation.

Variability in Sequence of Acquisitions

Another form of variation involves the sequence in

which skills for a given task or context are constructed,

often called developmental sequences or pathways. Al-

though evidence of variation in specific developmental

sequences has been taken as evidence against hierarchi-

cally constructed stages (Brainerd, 1978; Gelman &

Baillargeon, 1983), a dynamic structural analysis illu-

minates when sequences occur and when they do not,

whereas stage and competence theories are hard pressed

to account for observed patterns of variability and sta-

bility in sequences.

An examination of the evidence shows a familiar

pattern: There is high variability in developmental

334 Dynamic Development of Action and Thought

sequences, but this variability is neither random nor ab-

solute. The number and order of steps in developmental

sequences vary as a function of factors like learning

history, cultural background, content domain, context,

coparticipants, and emotional state. In addition, the vari-

ability in steps appears to be contingent on the level of

analysis at which the sequence is examined (Dawson &

Gabrielian, 2003; Fischer, 1980b; Fischer et al., 2003).

Developmental sequences tend to appear mainly at

two levels of analysis: (1) large-scale, broad sequences

covering long times between steps, relatively indepen-

dent of domain, and (2) small-scale, detailed sequences

found within particular domains. Large-scale sequences

appear to be relatively invariant. Children do not, for in-

stance, exhibit concrete operational performances

across a wide range of tasks, and then years later begin

to exhibit preoperational performance on related tasks.

On the other hand, small-scale sequences have often

been found to vary dramatically (Ayoub & Fischer, in

press; Wohlwill & Lowe, 1962).

Typically, variation in small-scale sequences is asso-

ciated with variation in task, context, emotion, copar-

ticipant, or assessment condition. For instance, Kofsky

(1966) constructed an eleven-step developmental se-

quence for classification of objects based on Inhelder

and Piaget’s (1964) concrete-operational thinking and

used scalogram analysis to rigorously test the sequence.

Her predicted sequence followed a logical progression,

but it drew on an assortment of different tasks and mate-

rials to evaluate each step. The results showed weak

scalability with several mini-sequences.

Other sources of variation in small-scale sequences

include cultural background, learning history, learning

style, and emotion. Price-Williams, Gordon, and

Ramirez (1969), for instance, examined the order of ac-

quisition of conservation of number and substance in

two Mexican villages. The villages were comparable in

most ways except that in one village the children partic-

ipated in pottery making from an early age. Children of

the pottery-making families tended to acquire conserva-

tion of substance (tested with clay) before conservation

of number, while nonpottery-making children showed

the opposite tendency.

Affective state can also powerfully affect develop-

mental sequences (Ayoub & Fischer, in press; Fischer &

Ayoub, 1994). For example, inhibited and outgoing chil-

dren show different sequences in representing positive

and negative social interactions, especially those involv-

ing the self. Inhibited children often show the positive

bias in Figure 7.6. Extreme emotional experiences such

as child abuse often lead to highly distinctive develop-

mental sequences for representing self and others in re-

lationships, as we discuss in the section on Emotions.

Furthermore, the failure to consider variation in se-

quences from factors such as learning style, disability,

or cultural difference leads to combining undetected

variations, with the result that task sequences erro-

neously seem to scale poorly (Fischer, Knight, et al.,

1993). As soon as they are resolved into alternative se-

quences, they scale well. For example, a sequence of six

tasks related to reading single words scaled weakly

when tested on a sample of poor readers in first to third

grades (Knight & Fischer, 1992). In each task, a child

dealt with an individual word, reading it directly (Read-

ing Production), reading it through matching it with a

picture (Reading Recognition), producing a word that

rhymes with it (Rhyme Production), recognizing a word

that rhymes with it (Rhyme Recognition), naming the

letters seen in the word (Letter Identification), or de-

scribing what the word means (Word Definition). Use of

a scaling technique for detecting alternative sequences

showed the existence of three different well-ordered se-

quences in the sample. Subsamples of poor readers

showed sequences that reflected their specific reading

difficulties.

The constructive web framework provides a tool for

rethinking these patterns of variation in the construc-

tive dynamics of skill development. Alternative develop-

mental pathways can often be traced for different groups

of children such as the three pathways for good and poor

readers. When the standard metaphor of the develop-

mental ladder is used, children are compared only in rel-

ative progress or delay on a single progression from low

to high performance on a single sequence. As long as

only a single pathway is considered, there seems only

one remedial choice: to work to speed up the apparently

delayed group along the “normal” pathway.

Figure 7.10 shows the three weblike pathways that

the students take through the series of reading tasks. For

each group, the order of acquisition for the six tasks was

tested using partially ordering scaling, a statistical tech-

nique that is based on the logic of Guttman scaling

(Krus, 1977; Tatsuoka, 1986). A line between two tasks

means that the ordering is statistically reliable. A com-

parison of the three developmental pathways shows that

the poor readers are not delayed with respect to a univer-

sal sequence, but actually follow different pathways of

acquiring these skills. Normal readers all showed one

Dynamic Structure in Cognitive and Emotional Development 335

Figure 7.10 Developmental pathways of good and poor readers. The normative pathway for most good readers is shown in (a:

Pathway 1: Normative developmental pathway for reading single words), whereas the two less integrated pathways followed by

poor readers are shown in (b: Pathway 2: Independence of reading and rhyming), and (c: Pathway 3: Independence of reading,

letter identification, and rhyming). From “Growth Cycles of Mind and Brain: Analyzing Developmental Pathways of Learning

Disorders,” by K. W., Fischer, L. T. Rose, and S. P. Rose, in Mind, Brain, and Education in Reading Disorders, K. W. Fischer,

J. H. Bernstein, & M. H. Immordino-Yang (Eds.), in press, Cambridge, England: Cambridge University Press; From “Learning

to Read Words: Individual Differences in Developmental Sequences,” by C. C. Knight and K. W. Fischer, 1992, Journal of Ap-

plied Developmental Psychology, 13, pp. 377–404.

Word Definition

Reading

Production

Rhyme

Production

Reading

Recognition

Rhyme

Recognition

Letter

Identification

Word Definition

Reading

Production

Rhyme

Production

Reading

Recognition

Rhyme

Recognition

Letter

Identification

Word Definition

Reading

Production

Rhyme

Production

Reading

Recognition

Rhyme

Recognition

Letter

Identification

(a)

(b)

(c)

main pathway (a), but poor readers showed two other

pathways different from the normal one (b and c).

This map of alternative pathways suggests a different

remedial educational strategy. Instead of attempting to

speed up development in poor readers, teachers can help

channel children following divergent pathways into al-

ternatives that converge on the goal of skilled reading

(Fink, in press; Wolf & Katzir-Cohen, 2001). By provid-

ing environmental support, teachers can channel devel-

opment, building bridges from the known to the

unknown instead of providing frustrating repetitive en-

counters with the unknown. This approach is being real-

ized most fully in educational efforts for children with

learning disorders and handicaps (Fischer, Bernstein, &

Immordino-Yang, in press; Rose & Meyer, 2002) and

also in some work with maltreated and aggressive chil-

dren (Ayoub & Fischer, in press; Kupersmidt & Dodge,

2004; Watson, Fischer, & Andreas, 2004).

From this perspective, the tool of mapping alternative

developmental pathways is especially important for the

study of development among children of differing so-

cioeconomic groups, cultures, ethnicities, or races, and

children with learning or psychological disorders.

Against the backdrop of a developmental ladder based on

White, middle-class norms, children from different so-

cial groups are frequently seen as exhibiting deficits in

development. Within the web metaphor, many develop-

mental differences become alternative pathways instead

of deficits, and curricula, interventions, or therapies can

be created based on these alternative pathways.

Research methods should allow detection of alterna-

tive sequences instead of forcing all children to either fit

or not fit one sequence. Remarkably, much research on

development has treated sequences not as variable phe-

nomena to be explained but as fixed milestones in a lad-

der. In the early 1970s, Flavell (1971) and Wohlwill

(1971) called for more research on variation in se-

quences, but this call has only recently begun to be taken

seriously. Most neo-Piagetian developmental theories and

domain theories still differentiate only gross stages, ig-

noring completely branches in sequences and variations

among steps, with a resulting overgeneralization of the

336 Dynamic Development of Action and Thought

uniformity and universality of cognitive and emotional

development.

In summary, the organization of human action,

thought, and emotion shows wide, systematic variation

that can be measured, analyzed, and explained in hierar-

chically organized systems of contextually embedded

activity. Patterns of variation in developmental level,

synchrony, and sequence are all consistent with a con-

structivist, dynamic systems interpretation of psycholog-

ical structure. In light of the pervasive evidence of

cognitive variability, it seems surprising that the most

prominent models of psychological structure have been

and continue to be based on static conceptions such as

stage, competence, and innate core knowledge. To under-

stand why these static conceptions of structure continue

to dominate and how dynamic views of psychological

structure move beyond them, we consider the history

and origin of static conceptions of psychological struc-

ture and their shortcomings as explanatory tools in the

next section.

THE CRISIS OF VARIABILITY AND THE

CARTESIAN SYNTHESIS IN

DEVELOPMENTAL SCIENCE

The failure of developmental theory to recognize the

dynamic and constructive nature of psychological

structure has led to an explanatory crisis in develop-

mental science. At the heart of this crisis is the problem

of how to account for the tremendous variability in de-

velopmental phenomena, which during the past 30

years has increasingly moved from the background to

the foreground of developmental research and theory

(Bidell & Fischer, 1992; Damon, 1989; Siegler, 1994,

Chapter 11, this Handbook, Volume 2; Thelen & Smith,

Chapter 6, this Handbook, this volume).

The static stage structure, which dominated theories

of cognitive development from its inception through the

early 1980s, proved incapable of accounting for the mas-

sive evidence of (a) both wide-ranging variation and

sometime consistency within and across individuals in

the age of acquisition of logical concepts across domains

and contexts, (b) systematic sequences in acquisition of

many of these concepts and their components, and (c)

variation from high to low synchrony in development of

concepts under various conditions. By the mid-1980s,

the inability of stage theory to account for this combina-

tion of variability and consistency led many scholars to

virtually abandon stage theory as a framework and to

launch a series of alternative accounts of psychological

structure and its origin.

Many of these alternative accounts attempt to explain

variability without departing from the static structure-

as-form metaphor, but they have consistently come up

short. Traditional Piagetians tried to package up vari-

ability in the concept of “decalage,” which simply

means a gap in ages of acquisition across tasks or indi-

viduals, and then mostly ignored it, thus renaming vari-

ability but not explaining it. Other theorists introduced a

separation between competences or underlying struc-

tures that remain static and performances or surface

manifestations that can vary. However, the separation of

action from competence in competence-performance

and nativist models introduces a major mystery about

what interferes with the expression of competence and

creates an inability to explain how psychological organi-

zation directs action and how structures adapt to a range

of environmental and cultural contexts.

Why do developmental theorists cling to static struc-

tural models? The most important reason, we propose,

is the pervasive influence of the Cartesian epistemolog-

ical tradition in the history of Western psychological

thought (Descartes, 1960; Gottlieb, Wahlsten, & Lick-

liter, Chapter 5; Overton, Chapter 2; Valsiner, Chapter

4, this Handbook, this volume). The Cartesian method

conceptually isolates mental systems from their natural

context of interrelations with the biological and cultural

systems of which they are a part. This intellectual

methodology of isolating an object of study from inter-

relationships with other phenomenon was successful in

the early history of the natural sciences, but it obscures

the complexity and dynamism of mental activity. It

leads to systematic distortions when applied to the ques-

tion of mental organization or psychological structure.

In this section, we review the Cartesian framework

and the empirical debate surrounding the concept of

“stage structure,” showing how this debate led to the

discovery of variability in level, synchrony, and se-

quence, and why the formal view of structure was un-

able to predict or explain this variability. We then argue

that three major theoretical movements since stage

theory—(1) domain specificity theory, (2) nativist com-

petence theory, and (3) competence/performance the-

ory—have also proved inadequate in accounting for

variability in structural development because they too

have failed to move beyond the Cartesian structure-

The Crisis of Variability and the Cartesian Synthesis in Developmental Science 337

as-form paradigm. In the subsequent section, we de-

scribe a set of methods for moving beyond these ap-

proaches to do research that deals with variability more

powerfully within a dynamic structural framework, in-

cluding an outline of how to turn theories about develop-

mental process into specific mathematical models that

can be tested against growth patterns of individual chil-

dren and adults.

The Cartesian Dualist Framework

The debate over nature-versus-nurture explanations of

the origin of knowledge assumes the Cartesian frame-

work, which is accepted by both sides—nature/nativist

and nurture/empiricist. Grounded in the dualism of

mind and world, the two sides necessarily imply one

another. The nativist-rationalist tradition and the

empiricist-learning tradition are two sides of the same

Cartesian coin. The nativist branch of the Cartesian

framework explains the origin of psychological struc-

ture as preformed innate structures such as concepts.

The empiricist branch explains it as experience stamp-

ing its shape on the natural mind. Psychological struc-

ture, conceived as innate form, implies some outside

input to be stored and manipulated. Environmental in-

formation conceived as preexisting packets of knowl-

edge requires some sort of preexisting receptacle or

organizing structure in the mind to receive, contain,

and organize them. In this framework, only two expla-

nations for the origin of psychological structure seem

possible—nature or nurture—and they become the

basis for the two branches of Cartesian epistemology.

The Cartesian tradition in philosophy and science

brings with it the methodologies of reduction and reifi-

cation. These methods, which have been profitably em-

ployed in many areas of science, result in systematic

misconceptions when applied to dynamic processes like

development of action, thought, and emotion. The dy-

namic organization of human mental activity is ab-

stracted from the living systems of which it is a property

and treated as a separately existing “thing,” giving birth

to the conception of static structure. The reification of

psychological structure as a separately existing static

form leads scientists down false paths in trying to un-

derstand the origins and development of psychological

organization. Instead of seeking to understand the con-

structive, self-organizing processes by which children

build new relations among contextually embedded men-

tal activities, theorists have been led into the futile

nature-nurture debate about whether statically con-

ceived psychological structure is somehow insinuated in

the genome or is built up through analysis of perceptual-

motor experience. These reductionist assumptions sup-

port static views of structure and limit the explanatory

power of developmental theories.

The Cartesian method, emerging in the seventeenth

century philosophy of Ren� Descartes (1960) and oth-

ers, gave science a powerful analytical tool to sort out

the complexity of the world and focus on one aspect at a

time for study. This tool, known as Cartesian reduction-

ism, derives simplicity out of complexity by isolating

one aspect of a process from its relations with other as-

pects of the process or from related processes, to be

studied independently. Descartes tried to extract mind

from nature by creating a dualism in which a separately

existing mental structure receives impressions from

the outside world through the sensory apparatus.

Descartes’s famous dissection of the cow’s eye, reveal-

ing the image projected on the retina, supported his

view that innate structures are fed with sensory images

from the environment. Similarly, in his logical empiri-

cism the philosopher John Locke (1794) asserted that

some preexisting logical structure is required to explain

how environmental input leads to higher order knowl-

edge. Locke saw that the simple mechanism of associa-

tion of sensory impressions could not account for

higher-order knowledge involving induction, deduction,

and generalization. Like Descartes, Locke’s account of

knowledge acquisition involved a dualist conception in

which a preexisting psychological structure receives and

processes sensory input from the outside.

Although Cartesian reductionism has been and will

continue to be an indispensable tool of scientific analy-

sis, its strength—the isolation of phenomena from com-

plex relations—is also its weakness (Wilson, 1998).

When Cartesian reductionism is used exclusively as an

analytic method, it eliminates an essential characteris-

tic that needs to be understood—the interrelations of

psychological systems both internally among component

processes and externally with other systems. Under-

standing relations is a requisite for understanding

change and variation in developmental or historical phe-

nomena. In the real world, it is the interrelations among

systems and processes that effect movement and

change. The gravitational relation between the earth

and the moon is key to sustaining the moon’s orbit,

which generates the changing cycles of the moon seen

on earth. To ignore the gravitational relation between

338 Dynamic Development of Action and Thought

earth and moon would preclude understanding the

source of this pattern of variability, and in turn its ex-

planation of an orbital system. The reductionist ap-

proach can be highly efficient for restricted scientific

purposes such as isolating a particular strain of bacteria

that causes a human disease. It is problematic in study-

ing any complex phenomenon involving relations among

elements and systems such as the problem of how some

bacteria evolve more virulent strains in the modern con-

text of changing natural and social ecology, growing

poverty and hopelessness in many locales, and overuse

of antibiotics. The structure grows dynamically out of

the relations among varying systems, neither from a

static innate structure nor a static environmental struc-

ture stamped on the mind.

The exclusive use of reductionism as an analytical

method fosters the related problems of reification and

dualism, both arising from the neglect of relations in

theoretical constructs. Without an account of the rela-

tions among systems that can explain movement and

change, abstractions such as mind, thought, and struc-

ture appear static and isolated from other constructs

such as body, action, or function. These static abstrac-

tions reify the phenomena they refer to, treating dy-

namic processes as frozen objects. The self-organizing,

goal-directed activity of the human agent is ruled out of

the accounts of development.

Moreover, because the relations between such reified

processes are lost, they seem isolated, separate, and

even opposite to one another. This seeming opposition

of reified abstractions is the basis for the classic Carte-

sian dualisms separating mind from body, thought from

action, and structure from function. Since the time of

Descartes, such dualist assumptions have become in-

grained in the mainstream of Western scientific thought

in general and psychological theories in particular. The

result has been static accounts of psychological phenom-

ena and their origins and sterile debates that explain

mental processes by one or another reified abstraction

such as faculties, associations, stimulus-response bonds,

innate concepts, or stages. While such single-construct

explanations have generated intense debate, they have

been notoriously limited in accounting for a broad range

of developmental data.

The Tacit Modern Synthesis in Psychology: Nativism

and Empiricism Together

The result of the debate that has continued for more than

a century between empiricist and nativist theories of de-

velopment is the emergence of a tacitly shared model—a

kind of modern synthesis in psychology—that is neither

strictly empiricist nor strictly nativist but simply Carte-

sian in its assumptions. The emerging model is an amal-

gam of a sort of logical empiricism with a version of

maturationism. According to this view, infants are sup-

plied innately with core knowledge systems that provide

them with predetermined representations of certain as-

pects of the world such as numerosity and object perma-

nence (Carey & Spelke, 1996; Hauser, Chomsky, &

Fitch, 2002; Spelke, 2000). However, these initial repre-

sentations must be extended by learning processes.

Learning processes typically are characterized through

a logical analysis of perceptual-motor input leading to

inductions and generalizations growing from core

knowledge. Debate continues about whether core knowl-

edge systems change qualitatively over time or simply

remain in place into adulthood, which mechanisms lead

from innate representations to new forms of knowledge,

and what roles perceptual analysis and learning mecha-

nisms play in such changes. Yet the framework of the de-

bate remains firmly grounded in Cartesian assumptions.

With respect to the origins and development of knowl-

edge, the debate between empiricism and nativism—and

the emerging modern synthesis—starts with a core set of

shared dualist assumptions: The mind is isolated from its

environmental context, thought is divided from action,

and the way the mind is organized (psychological struc-

ture) is separated from the way it operates in the world

(cognitive function).

Early empiricists tried to explain the origins of

knowledge in sensory impressions of the environment

with little reference to the role of the active person and

mind. In classic empiricist theories, the role of organi-

zation in the mind is minimal (a “blank slate” in the

extreme), and it is shaped by environmental contingen-

cies. Links or associations between ideas are generated

by whatever happens to co-occur: A person sees red

and apple at the same time, so she or he remembers red-

apple. In the behaviorist version of associationism, the

mind is reduced to almost no role at all, and behavior is

organized directly by environmental contingencies

through the stimulus-response bond (Skinner, 1969).

Contemporary empiricist theories tend to rely on an in-

formation processing metaphor in which sensory infor-

mation from the environment is parsed by perceptual

analysis into basic knowledge units that can then be

recombined into higher level knowledge (Newell &

Simon, 1971). However, common to all empiricist theo-

ries of mental development is a dualist separation of

The Crisis of Variability and the Cartesian Synthesis in Developmental Science 339

mind from environmental context, a concomitant reifi-

cation of the mind as a container or mechanistic

processor, and a dualist separation of mental structure

from mental content.

Information processing theories, in the empiricist

tradition, have focused on the input and storage of infor-

mation, building the analysis of cognitive structure on a

model of information flow in a computer. These theo-

ries came late to the problem of where cognitive struc-

tures come from and how they change over time. A few

information processing theories have posited qualitative

hierarchies of cognitive structures (Anderson et al.,

2004; Klahr & Wallace, 1976; Pascual-Leone, 1970),

but they have provided only sketchy accounts of the ori-

gins of these structures and the mechanisms of transi-

tion from one structure to another.

Despite years of vociferous debate with empiricists,

nativists share this set of dualist assumptions but privi-

lege them in different ways (Fischer & Bullock, 1984;

Overton, Chapter 2, this Handbook, this volume). Na-

tivists and the closely associated rationalists also start

from an acceptance of mind-environment, mind-body,

and thought-action dualisms. The difference with em-

piricists is that the structure of the mind is primary in-

stead of the structure of the environment. Nativists

accept the dualism of inner structure and outer sensory

information, but they simply assign them different roles.

Instead of filling up preexisting mental containers with

experience, the nativist role for sensory information is

to provide inputs, which trigger the emergence or activa-

tion of preexisting psychological structures such as the

syntax of language or the properties of objects. The du-

alist separation of psychological structure from its con-

textual relations with human activity has led to the

reification of psychological structure and the inevitable

conclusion that the structure must be innate. The outside

world provides grist for the cognitive mill, or sometimes

a triggering stimulus to kick off a new level of matura-

tion, but plays a minimal role in the development of the

psychological structures themselves.

When a dynamic system is approached statically, the

complex relations by which it is organized and by which

it develops are lost. The inescapable fact that it is organ-

ized is abstracted and reified as static form. When psy-

chological structure is conceived as static form, with no

activity and no inter-systemic relations to explain its

origin and development, it appears to have an existence

of its own, separate from the reality from which it is ab-

stracted. Therefore, psychological structure must be in-

nate, according to this argument.

The reification of dynamic structure as static form in

the Cartesian tradition has earlier roots in Western cul-

ture (Pepper, 1942) extending back at least to Plato

(1941) 2,000 years ago. His doctrine of ideal, universal

forms provides a particularly clear example of how

concepts and ideas are seen as independent of the mind.

These forms exist independently of the imperfect

material world, which evolves toward them. They are

transferred to each newborn infant, who gradually re-

members them with maturity. In the eighteenth century,

Kant (1958) argued that we inherit preexisting cognitive

structures or categorical imperatives, which determine

how we make sense of our experience. In recent times,

Chomsky (1965) and Fodor (1983) have argued for pre-

determined linguistic structures called modules that im-

pose specific patterns on our learning of languages and

concepts. Following Chomsky’s lead, contemporary

neo-nativists have posited innate structures determining

such developmental achievements as number concept,

object concept, and Euclidian geometry (Baillargeon,

1987; Fodor, 1983; Spelke, 1988).

Because both the empiricist and the nativist versions

of the Cartesian tradition share the same dualist, static

conception of psychological structure, neither has really

challenged the other on the nature of psychological

structure. However, the debate over how much emphasis

to place on innate structures versus learning has forced

each side to examine, rethink, and revise its theories. As

theorists on both sides of the empiricist-nativist debate

have attempted to revise their models to meet these

challenges they have naturally turned to models within

their shared Cartesian framework, and thus have in-

creasingly adopted elements of each other’s theories.

While still emphasizing the importance of perceptual

input, empiricism-based theories have come to rely on

nativist conceptions about the origins of psychological

structure to help explain how that input gets organized

and how its organization changes over time. On the na-

tivist side, theorists have increasingly come to depend

on various functional-learning and perceptual-analysis

mechanisms to explain how innate structures can lead to

knowledge and conceptual change.

At first glance, bringing together two opposing ten-

dencies into a more integrated model may seem like

progress toward a more comprehensive theory. The re-

sulting amalgamated model, however, does not take us

beyond the Cartesian framework of dualism and there-

fore does not offer a way beyond the static conceptions

of psychological structure—a way to explain how struc-

ture emerges from the interrelated activity of people

340 Dynamic Development of Action and Thought

with their world and each other. For this reason, the

Cartesian synthesis is not any more successful in ex-

plaining the broad data of variability in cognitive devel-

opment. Linking static conceptions of psychological

structure to mechanistic information processing models

does not provide us with better explanations for variabil-

ity in cognitive performance than either tradition did on

its own. Understanding why Cartesian models—

whether empiricist, nativist, or a combination of the

two—have trouble explaining variability requires con-

sidering in more depth the static conceptions of psycho-

logical structure inherent in this tradition and the

explanatory limitations they carry with them. This

analysis lays the foundation for understanding how dy-

namic structuralism provides a path to analyzing the dy-

namics of structure in development starting from

activities in context.

The Structure-as-Form Paradigm

Because the Cartesian tradition has been the dominant

framework for scientific theories in general and psycho-

logical theories in particular, reductionism and reifica-

tion have been the rule rather than the exception in

conceptions of psychological structure. The prominence

of these modes of thought in the Western intellectual tra-

dition has encouraged the confounding of dynamic struc-

ture with static form. Accordingly, the structure-as-form

model has tended to serve as an unconscious founda-

tional metaphor (Lakoff & Johnson, 1999; Pepper, 1942)

or paradigm (Kuhn, 1970) for scientific accounts of the

organizational properties of natural and social systems,

especially in psychology (Overton, Chapter 2; Valsiner,

Chapter 4, this Handbook, this volume).

It is no easy matter to move beyond the static

metaphors for structure, which language and cultural

practices strongly support and which people typically

use unaware. A dramatic, pervasive example is the con-

duit metaphor for communication (Lakoff & Johnson,

1980; Reddy, 1979). In ordinary discourse about com-

munication of knowledge, people use this metaphor,

talking as if the mind is a container for knowledge and

as if things that they know are discrete objects. They

treat communication as the transfer of knowledge ob-

jects from one person to another, as if static objects are

being sent through a conduit such as a pipe or telephone

line. This metaphor often leads to the belief that telling

someone an item of information (giving them an object)

is sufficient to communicate it and even to teach it. If a

course or a chapter covers a concept, for example, then

the student or reader is assumed to have been given that

object. If they fail to demonstrate the knowledge speci-

fied by that object, they are taken to be ineffective

learners (stupid, inattentive, or lazy). Research shows

that students do not learn effectively from such presen-

tations, but they require experience with acting on

and manipulating the material to understand it (Crouch,

Fagen, Callan, & Mazur, 2004; Schwartz, 2000;

Schwartz & Fischer, 2005). This static metaphor (and

others as well) omits the constructive nature of learn-

ing, knowing, and understanding from the assumed

structure of communication and education, and their so-

cial nature is minimized too.

The conceptualization of structure as form treats

structure as a static property of knowing that can be

separated from the knowing activities themselves, just

as the conduit metaphor separates objects of knowledge

from activities of knowing. Imagine trying to remove

the structure from the Golden Gate Bridge, gather it up

somehow, and ship it off to someone else, who would

add it to a pile of steel, which would quickly arise to

form a replica of the San Francisco landmark. Even

more absurd would be trying to extract the structure

from the tightly coordinated, self-organizing, physico-

chemical processes of a living cell and then to apply it

to a blob of inert chemical components in hope of gener-

ating a new cell. Structure is an inseparable quality of

real dynamic systems, and it emerges as they develop

(are constructed). In reality, structure cannot be sepa-

rated from its role as the organizational property of dy-

namic systems.

In the study of development, three static conceptions

of psychological structure have predominated, all of

which have used static forms to explain dynamic struc-

tures. In many developmental theories, including Pi-

aget’s (1983, 1985) stage theory, activities take the form

of abstract logical structures. In many linguistic and

cognitive theories, activities take the form of preformed

quasi-logical rules, typified by Chomsky’s (1957, 1995)

theory of innate linguistic competences and its corollary

theories of innate cognitive competences (Baillargeon,

1987; Fodor, 1983; Spelke, 1988). In many traditional

empiricist theories in Anglo-American psychology, ac-

tivities take the form of linear input-output rules, as

typified by linear models in statistics, information pro-

cessing, and behavior genetics (Anderson et al., 2004;

Horn & Hofer, 1992; Plomin, DeFries, McClearn, & Rut-

ter, 1997). This linear form of theory is especially

prominent in approaches that focus on domain speci-

The Crisis of Variability and the Cartesian Synthesis in Developmental Science 341

ficity, the separation of knowledge into distinct parts

tied to domains of experience.

Despite well-publicized disagreements among these

three frameworks, they derive their core assumptions

from the structure-as-form paradigm, portraying psy-

chological structure in abstract forms existing sepa-

rately from real self-organizing human activities. In

stage theory, psychological structure is seen as a univer-

sal abstract logic imposing itself on the developmental

trajectories of every person. Although Piaget believed

that activity is the basis of knowledge and development,

the base metaphor for his stage theory of cognition is

successive stages of logic that determine specific cogni-

tive performances across contexts and domains of

knowledge and are relatively unaffected by the contexts

of those performances. Similarly, nativist competence

theories project a universal preformed code, blueprint,

or set of instructions that somehow exists separately

from the activities that it will someday engender. Like

Platonic forms, these blueprints lurk among the genes,

awaiting the right moment to impose order on behavior.

The experimental/psychometric framework also

bases its core assumptions on structure as form, but

there the structure is hidden behind standard methods

and paradigms for explanation. The assumed linear

combinatorial structures of dichotomies—person and

environment, input and output, heredity and experi-

ence, domain x versus domain y—are embedded in re-

search designs, statistical techniques, and theoretical

concepts, but their implicit assumptions about structure

are seldom acknowledged (Bronfenbrenner, 1979; Fis-

cher & Bullock, 1984; Gottlieb, Wahlsten, & Lickliter,

Chapter 5; Overton, Chapter 2; Thelen, & Smith, Chap-

ter 6; Valsiner, Chapter 4, this Handbook, this volume;

Wahlsten, 1990; Wittgenstein, 1953). Person and envi-

ronment are partitioned into separate groups of factors

instead of being treated as dynamic collaborators in

producing activities. Much of modern biology has

assumed similar reductionist, reifying notions of struc-

ture as form (Goodwin, 1994; Gottlieb, 2001; Kauff-

man, 1996).

The dominance of the structure-as-form paradigm in

cognitive developmental theory has forced scholars to

choose among these three inadequate notions of struc-

ture—stages, innate structures, and linear information

processes. Instead, structure needs to be conceived dy-

namically. Psychological structure exists as a real orga-

nizational property of dynamic systems, just as the

structure of the human skeletal system and the human

circulatory system are real and distinguish humans from

other animals. The structure is a property of the self-

organizing systems that create it—the dynamic organi-

zation exhibited by self-organizing systems of mental

and physical activity, not a free-floating ghost of compe-

tence or logic that dictates behavior to its human ma-

chine. Before we explicate concepts and methods of

dynamic structure, however, we need to ground our ar-

gument with analysis of key problems with the static

conceptions of structure that pervade developmental

and psychological science.

The Stage Debate and the Discovery of

Variability in Cognitive Development

The strength of the stage structure concept, as with all

structure-as-form models, is its account of stability in

development. Skills exhibit patterns of stability both in

the ways they function and the ways they develop. What

would account for such stable patterns in the functioning

and development of cognition? Piaget’s (1983, 1985)

conception of formal logical stages addressed this ques-

tion with what seemed to be a powerful and reasonable

explanation: Individuals construct logical structures

that preserve the organization of their interpretive or be-

havioral activities to be applied again at later times or in

different situations. The existence of these structures

accounts for the ability to apply the same concept or

skill across many situations. Similarly, the emergence of

concepts in specific sequences is accounted for by the

fact that the logical structures underlying the concepts

are constructed gradually, so that a partially complete

logic would give rise to one concept (e.g., one-to-one

correspondence) and the later completion of the logical

structure would give rise to a more extensive and logi-

cally complete concept (e.g., conservation of number).

Piagetian stage theory places all human cognitive activ-

ities into a sequence of abstract logical forms, but it has

proved incapable of explaining the vast array of devia-

tions from stage predictions (Bidell & Fischer, 1992;

Flavell, 1971; Gelman & Baillargeon, 1983).

However, the strength of the stage structure concept

was also its greatest weakness: Whereas universal logical

structures accounted elegantly for stability, they offered

hardly any explanation for variability in the functioning

and development of cognition. Because the stage concept

equated psychological structure (the organization of dy-

namic mental activity) with static form (formal logic), it

provided no model of the real psychological mechanisms

that might lead to variability and change in development.

The idea of a fixed logical structure underlying all of a

342 Dynamic Development of Action and Thought

child’s conceptions at a given stage seems to explain ob-

served consistencies in the form of children’s thinking,

but it predicts much, much more consistency than chil-

dren show, and it has proven incapable of explaining de-

partures from the predicted consistency.

Departures from the consistency predicted by stage

theory proved to be more the norm than the exception as

proliferating replication studies introduced a myriad of

variations on Piaget’s original tasks and procedures. On

the one hand, opponents of Piaget’s theory, doubting the

reality or usefulness of formal stage structures, focused

their research on identifying conditions in which stage

theory predictions failed. In contrast, supporters of

Piaget’s constructivist view tried to validate the pur-

ported products of development—stage sequences, tim-

ing of cognitive achievements, and universality. These

researchers focused a great deal of attention on demon-

strating conditions in which stage predictions were

empirically supported. Today, many researchers still

continue along these independent paths, mostly ignoring

or dismissing findings of people from the other camp.

The outcome of this protracted and often heated em-

pirical debate has been the discovery of remarkable vari-

ability in every aspect of cognitive development studied.

As researchers implemented variations in the nature of

task materials, complexity of tasks, procedures, degree

of modeling, degree of training, and methods of scoring

across a multitude of replication studies, a consistent

pattern of variation emerged (Bidell & Fischer, 1992;

Case, 1991b; Fischer, 1980b; Halford, 1989; Louren�o

& Machado, 1996). To the extent that studies closely ap-

proximated the assessment conditions used by Piaget,

the findings were similar to those he had reported. When

tasks and procedures varied greatly from Piaget’s, the

findings also varied greatly within certain limits.

A classic example of this pattern of variation is found

in research on number conservation. In Piaget’s theory,

number conservation (the ability to conceptually main-

tain the equality of two sets even when one set is trans-

formed to look much larger than the other) was seen as a

product of an underlying stage of concrete operational

logic. In the original number conservation studies, Pi-

aget and Szeminska (1952) had used sets of 8 or 10 ob-

jects each and had identified 6 to 7 years as the typical

age of acquisition for this concept. In one group of repli-

cation studies, Gelman (1972) showed that the age of ac-

quisition for number conservation could be pushed

downward from Piaget’s norms if the task complexity

was simplified by (a) reducing the size of the sets chil-

dren had to compare and (b) eliminating the requirement

for verbal justification of conservation judgments.

Under these conditions, Gelman reported that children

as young as 3 to 4 years of age could answer conserva-

tion questions correctly. Fortunately, the debate about

number eventually produced important new discoveries

spelling out developmental pathways for the early con-

struction of number actions and concepts (Case et al.,

1996; Dehaene, 1997; Spelke, in press).

As replication studies proliferated, this seesaw de-

bate over age of acquisition of logical concepts was ex-

tended to other dimensions of psychological structure

where researchers produced similar patterns of variabil-

ity as a function of assessment conditions. These in-

cluded variability in the three central characteristics we

have described (developmental level, synchrony in level

across domains or contexts, and sequence of develop-

ment in a domain or context).

The growing empirical documentation of variability

in development posed severe problems for the concept of

formal stage structures. If concepts such as conserva-

tion of number are supported by underlying logical

structures, then why wouldn’t the logical structure man-

ifest itself in most if not all situations? Why would a

child show logical thinking one moment and in the next

moment, appear to have lost it? If cognitive development

consists of the emergence of successive forms of under-

lying logic, why wouldn’t developmental sequences re-

main the same across domains, contexts, and cultures?

The formal concept of stage structure could offer no

specific explanation for this pattern of variability, but

only the label of decalage.

In one sense, victory in the stage debate went to the

skeptical. By the mid-1980s, the inability to account for

the dramatic departures from stage theory’s predictions

of cross-domain, cross-individual, and cross-cultural

consistency had resulted in a general flight from stage

theory as an explanatory framework (Beilin, 1983). In a

more important sense, however, there was no winner be-

cause neither side had offered a workable explanation of

the patterns of variation the debate uncovered. What

concept of psychological structure would explain the

fact that cognitive performance varies so greatly with

changing conditions and yet also exhibits great consis-

tency under other conditions?

Explaining Variability versus Explaining It Away

From the perspective of the history of science, one

might think that the discovery of new patterns of vari-

ability would be met with excitement and theoretical ad-

The Crisis of Variability and the Cartesian Synthesis in Developmental Science 343

vance. After all, a central task of science is to discover

and account for variability. Theories are constructed

and reconstructed to interpret the range of variation ob-

served and to search for patterns of order within this

range. Indeed, an essential criterion of sound scientific

theories is that they account for the full range of vari-

ability observed in a phenomenon of interest.

However, change in scientific theories is rarely that

simple. Evidence that threatens a prevailing worldview

or paradigm can lead to attempts to assimilate the dis-

crepant findings into the current paradigm, either by

denying their relevance or by advancing alternative

explanations within the dominant paradigm (Hanson,

1961; Kuhn, 1970). Responses to the discovery of vari-

ability in development have followed this pattern, re-

turning to the prevailing Cartesian framework and

building minor modifications to account for portions of

the observed variability. Instead of attempting to fully

describe the range of variability and explain the reasons

for the observed patterns, responses have tried to ex-

plain away variability through a variety of theoretical

maneuvers that include ignoring variability, accepting

variability without explaining it, and focusing on se-

lected effects of variability to support existing theory

with minor adaptations. Each of these theoretical re-

sponses to variability has served to preserve some

version of the Cartesian framework and the structure-

as-form paradigm in the face of the new evidence and

has led to the modern Cartesian synthesis, despite the

fact that most of the evidence of variability remains

unexplained.

Reasserting Stage Theory

Piaget, Kohlberg (1969), and other stage theorists at

first mostly ignored variability, treating it basically as a

nuisance or as error of measurement. Differences across

domains, tasks, contexts, and coparticipants in phenom-

ena such as age of acquisition, synchrony, and develop-

mental sequence were said to represent varying forms of

resistance to the operation of underlying logical struc-

tures. Although Piaget later acknowledged the inade-

quacy of this position and experimented with alternative

logic frameworks (Piaget, 1985, 1987; Piaget & Garcia,

1991), he never found an alternative concept of struc-

ture that would predict and explain when and how

performance varies. (The discovery of the scale of hier-

archical skill levels, shown in Figure 7.3, came from

analyzing patterns of variation in growth curves,

demonstrating the usefulness of analyzing variation for

understanding stages.)

Several scholars have emphasized Piaget’s belief in

the importance of decalage and other forms of variation

(Beilin, 1983; Chapman, 1988; Louren�o & Machado,

1996), but recognizing that phenomena need to be ex-

plained is not the same as explaining them. Piaget and

other stage theorists have not specified the processes by

which cognitive stage structures and environmental re-

sistance interact to make one kind of task develop later

than another in general. They have dealt even less ade-

quately with variations across individuals in the order

and timing of acquisition of skills and variations within

an individual related to tasks, context, social support,

and experience. In short, stage theory has provided no

explanation for most observed patterns of variation in

developmental level, synchrony, and sequence (Bidell &

Fischer, 1992; Edelstein & Case, 1993).

Domain Specificity Theory

As evidence of variability grew and the inadequacy of

the classic stage concept became clear, the theoretical

crisis deepened. With stage theory losing its potential to

generate interesting and credible research and with no

clear alternative model of psychological structure avail-

able except for Chomskian nativism, some framework

was needed as a basis for the continued empirical study

of development. Domain specificity theory emerged as a

way of freeing the field from its dependence on stage

theory without demanding a new commitment to any

particular model of psychological structure. According

to domain specificity theory, psychological processes

are not organized in universal structures, but within lim-

ited domains such as spatial, linguistic, or mathematical

reasoning, or for groups of similar tasks such as problem

solving, analogical reasoning tasks, and theory of mind

(Demetriou, Christou, Spanoudis, Platsidou, 2002;

Hirschfeld & Gelman, 1994; Turiel & Davidson, 1986;

Wellman, 1990). The structures in these domains are

often referred to as modules, indicating separate, dis-

tinctive structures of brain and behavior (Fodor, 1983).

In education, domain specificity became a major theme

through the influence of Howard Gardner’s (1983) the-

ory of “multiple intelligences,” leading to curricular re-

visions in schools around the world.

Description of development and learning within im-

portant domains has great value for both developmental

science and education, but many scholars have stopped

with the domain description. They thus avoid having to

explain patterns of variability—for example, the differ-

ences and similarities in age of acquisition across dif-

ferent logical concepts such as number and theory of

344 Dynamic Development of Action and Thought

mind. Instead, they simply assert that cognition is or-

ganized locally and so cross-domain relations do not

have to be explained. This theoretical stance simply ac-

knowledges the fact of variability and sidesteps a sys-

tematic account of its origins.

In some ways, this acknowledgment has represented

an advance for a field once dominated by stage theory

with its assumption of a single logic that catalyzes

change across all aspects of the mind. However, to the

extent that domain specificity creates the illusion of

having solved the problem of variation, it is an unfortu-

nate theoretical detour. Developmental scientists need

to explain why clusters of many (structurally equiva-

lent) concepts emerge in different domains around the

same time, showing interval synchrony (Case, 1991b;

Fischer & Silvern, 1985). They need to explain how an

individual who is working within a single domain and

task exhibits one skill level when working alone, but a

distinctly higher level when working with the support of

a helpful adult (Fischer, Bullock, et al., 1993; Rogoff,

1990). Although domain specificity theory provides im-

portant recognition of developmental variability, it of-

fers no explanation of variability across domains and

within individuals.

Neo-Nativism

An important response to the evidence of variability has

been the neo-nativist movement (Carey & Gelman,

1991; Fodor, 1983; Spelke, 1988), which represents a

major theoretical alternative to stage theory within the

structure-as-form paradigm. Researchers taking this

perspective have used ingenious experiments to uncover

surprising capacities of infants and young children and

have led to the creation of the modern Cartesian synthe-

sis. With the rejection of the concept of structure as

stages of formal logic, the other predominant concept of

structure—innate formal rules—seems to be the only

remaining alternative within the structure-as-form par-

adigm. Unfortunately, the concept of innate formal rules

has the same fundamental limitation as its sister concept

of formal logic: As a static conception of structure, it

cannot adequately account for the variability that arises

from dynamic human activity (Fischer & Bidell, 1991).

Neo-nativist researchers have focused on selected ef-

fects of cognitive variability that seem to support the

existence of innate competences within prominent do-

mains such as number, space, language, object proper-

ties, and theory of mind (Carey & Spelke, 1994). For the

most part, they have not attempted to deal with the ex-

tensive variability found in performance. Indeed, the

modern father of this movement, Noam Chomsky (1965,

1995), specifically rejects the evidence of variability in

language, asserting that it is illusory and that all people

“really” speak the same fundamental language. The

Chomskian theory of linguistic competence accounts for

human linguistic behavior on the basis of a set of innate

rules, only a few of which have been specified. Despite

almost 50 years of effort, nativism has been notoriously

unable to account for either the variations of human lan-

guages (Chinese is different from English!) or the highly

variable everyday communication skills that individuals

develop in a language within and across diverse settings

(Lakoff, 1987; Ninio & Snow, 1996; Slobin, 1997). Nev-

ertheless, the nativist approach has had great appeal to

many developmental scientists because of its important

discoveries about children’s early abilities.

The basic paradigm for neo-nativist research is to de-

sign tasks that drive ages of acquisition much lower than

traditional Piagetian norms (Baillargeon, 1987; Spelke,

1988, in press). Nativist researchers introduced tech-

niques for simplifying Piagetian task materials and pro-

cedures, requiring only minimal activity from a child or

providing modeling, training, and other forms of sup-

port for more complex activity. They have shown great

ingenuity in discovering capacities of young infants and

children, demonstrating strong violations of Piaget’s

age norms for various logical concepts. Their neo-

nativist argument is that cognitive structure must be in-

nate because acquisition of certain concepts can be

demonstrated at very young ages. However, this argu-

ment from precocity takes into account only half the evi-

dence for variability—the downward half (Fischer &

Bidell, 1991; Halford, 1989). It treats the earliest age as

the “real” age for a concept’s emergence, ignoring evi-

dence of wide variations in age of acquisition both up-

ward and downward.

A good example of the focus on early age instead of

variation is the extensive research on infants’ acquisition

of knowledge of objects, especially object permanence

(objects continue to exist even when they have been dis-

placed and are not perceived) and object tracking. Re-

searchers have used the procedure of dishabituation,

which is designed to assess preferences for stimuli with-

out requiring much behavior. Infants are shown a stimu-

lus until they are used to it (habituated), and then they

are shown an altered stimulus. If they show increased at-

tention to the new stimulus (dishabituation), the conclu-

sion drawn is that they have noticed the difference.

The Crisis of Variability and the Cartesian Synthesis in Developmental Science 345

A well-known case is Baillargeon’s research on ob-

ject permanence in young infants (Baillargeon, 1987,

1999). To appreciate the problems with focusing on

only selective aspects of variability, it is useful to place

this study in the context of Piaget’s (1954) original

findings and interpretations regarding infant object per-

manence. Piaget described a six-stage sequence in in-

fants’ construction of object permanence, which

subsequent research confirmed with some revision and

clarification (McCall, Eichorn, & Hogarty, 1977; Uz-

giris & Hunt, 1987).

Piaget offered a constructivist interpretation of his

observations: a simple activity-based mechanism to

explain transitions from one stage to another. By coor-

dinating early sensorimotor activities on objects to

form new, more comprehensive action systems, infants

gradually construct more inclusive understandings of

what they can do with objects and therefore how ob-

jects can behave. For instance, by coordinating the

sensorimotor actions for looking at and grasping ob-

jects at Stage 2, infants of about 5 to 6 months of age

move to a new Stage 3 structure for dealing with ob-

jects—visually guided reaching, in which they simul-

taneously hold and observe an object. Piaget described

an especially important transition at stage 4, when in-

fants of about 8 months coordinate different visually

guided reaching skills into a system for searching out

objects that have been displaced or hidden. For in-

stance, infants coordinate two skills (what Piaget

called “schemes”): reaching for a rattle to grasp it, and

reaching for a cloth that is covering the rattle to re-

move it. With this stage 4 coordination, they can begin

to understand how objects come to be hidden by other

objects and why hidden objects remain available to be

retrieved. Later stages in this understanding extend to

late in the second year of life, when infants become

able to search exhaustively for hidden objects in many

possible hiding places.

In contrast to Piaget’s model of gradual construc-

tion of object permanence, Baillargeon focused on the

lower end of the age range and a simple looking task.

Infants from 3 to 5 months of age were habituated to

the sight of a small door that rotated upward from a flat

position in front of them, tracing a 180% arc away from

them to lie flat again on a solid surface. They were then

shown two scenes with objects inserted behind the ro-

tating door. In the possible event, the door swung up

but stopped at the object. In the impossible event, the

object was surreptitiously removed and the door was

seen to swing right through the space the object had

occupied, as if it moved through the object. Infants as

young as 31⁄2 to 41⁄2 months dishabituated to the impos-

sible event significantly more than they did to the pos-

sible, and Baillargeon took this behavior as evidence of

object permanence. She concluded that infants acquire

object permanence 4 to 5 months earlier than the age of

8 months that Piaget had reported.

This argument from precocity is straightforward: If

behaviors associated with a conceptlike object perma-

nence can be found much earlier than in prior research,

then the concept in question must be present innately.

Similar evidence has led to claims of innate determina-

tion for a growing list of concepts, including object

properties, space, number, and theory of mind (Carey &

Gelman, 1991; Saxe, Carey, & Kanwisher, 2004; Spelke,

2000). Based on the static Cartesian model, these

claims have important limitations centered on the fail-

ure to consider the full range of variability involved in

developmental phenomena.

The crux of the problem is a simplification that ig-

nores the gradual epigenetic construction of activities

that vary in complexity. Baillargeon’s task and proce-

dure were dramatically different from the more complex

method of assessment used by Piaget. In place of inde-

pendent problem solving in which the infant must ac-

tively search for an object hidden in several successive

places, Baillargeon substituted a simple look toward one

of two displays. This procedure simplifies the task so

greatly that it shifts from a conceptual task to one of

perceptual anticipation. Indeed, on a computer a neural-

network model of the situation can solve a similar task

with a simple visual strategy and no coordination of ob-

ject characteristics with spatial location (Mareschal,

Plunkett, & Harris, 1999).

Baillargeon and other nativists claim that the object

concept appears very early, even though the more com-

plex behaviors described by Piaget still develop at the

usual later ages, as shown by overwhelming evidence.

The selective focus on one early age for one behavior

obscures the constructive mechanisms of development

and makes it seem that the concept of object perma-

nence has suddenly leaped up, fully formed, at 31⁄2

months of age. Within this framework, innate concepts

emerge abruptly in the first few months of life, and de-

velopment disappears. How could such early develop-

ment arise except through innate concepts? The answer

to this question is another question: How do skills de-

velop through a long sequence of increasingly complex

346 Dynamic Development of Action and Thought

object-related activities of which the looking behavior is

only the beginning?

Competence/Performance Models

Nativists and many other cognitive scientists answer by

distinguishing between competence and performance.

The modern version of the competence/performance

distinction was proposed by Chomsky (1965) in an ef-

fort to explain why his theory of innate linguistic rules

could not predict the wide range of variability observed

in actual language usage. Chomsky argued that innate

language rules existed separately from the performance

of specific acts of communication. The rules governed

which communication practices are possible but not

which ones will actually take place in a given situation.

Many developmental scientists, faced with the similar

problem of explaining why formal Piagetian conceptions

of logic do not predict observed patterns of variability in

cognitive performance, adopted this distinction (Flavell

& Wohlwill, 1969; Gelman, 1978; Klahr & Wallace,

1976; Overton & Newman, 1982).

Competence/performance theories based on the Pi-

agetian and Chomskian models portray cognitive struc-

tures as fixed rule sets in the mind/brain that specify

behaviors but are somehow impervious to or indepen-

dent of the contexts of the behaviors. The structures

exist somewhere in the background and serve a limiting

function: They determine the upper limit on the range of

actions possible at a given time, but they leave open the

specific action that will take place. For example, in

arithmetic, the counting behavior of a preschool child

arises from a mathematical competence such as being

able to directly perceive numbers of objects of 1, 2, or 3.

When a child fails to count, say, three pretzels accu-

rately, the failure is explained by some interference such

as memory failure or distraction (Freeno, Riley, & Gel-

man, 1984; Spelke, in press). A skilled person can in-

deed mess up a performance here and there because of

memory failure or distraction, but when the 3-year-old

fails almost all tasks for counting three objects, what

sense does such an explanation make?

These models dismiss variability in cognitive and lan-

guage performance by asserting that fixed competence is

differentially expressed because of intervening cognitive

processes (vaguely specified) or as a result of unanalyzed

environmental resistance to the competence, as Piaget

suggested for decalage. Although most nativist theories

assume such a framework, some competence/perfor-

mance theories do not require that psychological struc-

tures exist innately, but only that they are firmly sepa-

rated from the actions that instantiate them. The

dynamics of construction of activities leading to wide

variation are lost in the muddy mediators that somehow

prevent competence from being realized in activity. Such

conceptions of disembodied structure seem not too dis-

tant from the humorous idea of bottling up the structure

of the Golden Gate Bridge. Why is it necessary to posit

separate levels of structure, existing somewhere (it is un-

clear where) outside the real activity in question? Why

not model the organization of the actual mental and phys-

ical activity as it exists in its everyday contexts?

In short, domain specificity, nativist, and compe-

tence/performance models share the same fatal limita-

tions as the logical stage models they were meant to

replace. Although the newer models do not make the

cross-domain claims that stage models did, they retain a

conception of psychological structure as static form

existing separately from the behavior it organizes.

Whether such static forms are seen as universal logics

or domain-specific modules, they offer accounts only of

stability in the organization of behavior while ignoring

or marginalizing variability. The challenge for contem-

porary developmental science is not to explain away evi-

dence of variability in performance. Instead, scholars

need to build dynamic models of psychological struc-

ture, using concepts such as skill, hierarchical complex-

ity, contextual support, and developmental web to build

methods for analyzing and explaining both the variabil-

ity and the stability in the organization of dynamic

human activity.

The Constructivist Alternative

The constructivist alternative takes as its starting point

what the Cartesian framework rules out: the construc-

tive agency of a human being acting in the context of re-

lationships among systems—biological, psychological,

and sociocultural. As we have shown in the opening sec-

tions of this chapter, the dynamic structural framework

provides a straightforward, comprehensive alternative

to the conundrums created by the Cartesian synthesis

and the related structure-as-form paradigm. Human

knowledge is neither passively received from the envi-

ronment nor passively received from the genome. In-

stead, people construct knowledge through the active

coordination of action systems beginning with the earli-

Methodology of Dynamic Structural Analysis 347

est sensorimotor activities of newborns, influenced by

environmental and genetic systems. By coordinating the

systems of activity (including perceptual activities)

through which they participate in the social and physical

worlds, infants create new relations among these sys-

tems and thus new potentials for acting in and under-

standing the world. These new relations among action

systems constitute psychological structures—the orga-

nizational aspect of human knowledge, which we refer

to as skills. They exhibit both wide variations and pat-

terns of order within the variations.

Dynamic systems research provides the framework

for this alternative account, drawing on traditions that

have developed outside of or as an alternative to the

Cartesian tradition. Important concepts and methods

come from epistemological constructivism and related

sociocultural/sociohistorical theory (Cole, 1992; Ro-

goff, 2003), traditional systems theory (Dixon &

Lerner, 1992; von Bertalanffy, 1976), dynamic systems

theory (Thelen & Smith, 1994, Chapter 6, this Hand-

book, this volume; van der Maas & Molenaar, 1992; van

Geert, 1991), and the developmental science group

(Cairns, Elder, & Costello, 1996; Cairns, Chapter 3;

Valsiner, Chapter 4, this Handbook, this volume). These

traditions, while differing in many ways, share a con-

structivist focus on action, interrelatedness, and com-

plexity of psychological, biological, and sociocultural

systems. From this perspective, the person is the pri-

mary agent of cognitive change, constructing new kinds

of relations among psychological systems with biologi-

cal and cultural systems (Bidell & Fischer, 1996; R.

Kitchener, 1986). These relations are organized in par-

ticular ways that give rise to specific patterns of perfor-

mance, and they are complex and variable because they

are living systems.

People construct the skills of human understanding

and action through their diverse bodies, the variable

physical world, different sociocultural relations, and

distinct developmental histories, thus producing highly

variable activities. If this variability is ignored, it

acts as noise disguising the nature of developmental

processes and thus misleading researchers and educa-

tors. However, if the tools of developmental analysis are

used to control and manipulate conditions contributing

to variability, then the systematicity of the variability

can be uncovered and it becomes a key to understanding

the nature of psychological structure. In the next sec-

tion, we discuss some of the methodological tools by

which developmental variability can be used to under-

stand and describe the development of dynamic psycho-

logical structure.

METHODOLOGY OF DYNAMIC

STRUCTURAL ANALYSIS

To overcome the limitations of structure as static form,

we need to articulate a framework for dynamic develop-

ment, which includes a set of methods that embody dy-

namic concepts. Classical research methods use static

notions, indicating the age when a competence emerges

(really, the mean or modal age for one context and one

group), forcing growth into linear models, and partition-

ing analysis of activities into dichotomies such as hered-

ity and environment or input and output (Anderson

et al., 2004; Horn & Hofer, 1992; Plomin et al., 1997;

Wahlsten, 1990). Most importantly, effective research

needs to be designed so that it can detect variability and,

in turn, use the variability to uncover sources of order or

regularity in development.

Effective research should be built with designs, mea-

sures, analytic methods, and models that can detect

variations in growth patterns. Research must be de-

signed to deal with variability, or it is doomed to fail to

provide an adequate analysis of development. This chap-

ter focuses on activities in which people coordinate and

differentiate lower-order components to form higher-

order control systems, which encompasses most activi-

ties of interest to developmental and educational

researchers. The components of these control systems

range from neural networks to parts of the body, imme-

diate contexts (including objects and other people), and

sociocultural frameworks for action. Moment by mo-

ment, people construct and modify control systems, and

the context and goal of the moment have dramatic ef-

fects on the nature and complexity of the systems. Fre-

quently, people do the construction jointly with others.

To go beyond static stereotyping of development and

learning, research must deal directly with these facts of

variation. Research must be designed to deal with the

wide range of shapes of development that occur for dif-

ferent characteristics of action and thought in diverse

contexts and conditions.

Developmental regularities can be found at several

levels of analysis, from brain activities to simple actions,

complex activities, and collaborations in dyads or larger

groups. In analyzing these developmental regularities, it

348 Dynamic Development of Action and Thought

is important to avoid a common mistake. No one regular-

ity applies to all characteristics of developing activity or

all levels of analysis. The same developmental regulari-

ties will not be found everywhere. That is an essential

principle of the variability of human activity.

In one major realization of this principle, develop-

ment has many different shapes! Some behaviors and

brain characteristics show continuous growth, others

show clusters of spurts and drops, still others show

oscillation, and some show growth followed by decay

(Fischer & Kennedy, 1997; Siegler, Chapter 11, this

Handbook, Volume 2; Tabor & Kendler, 1981; Thatcher,

1994; Thelen & Smith, Chapter 6, this Handbook, this

volume; van Geert, 1998). Ages of development likewise

vary dynamically, even for the same child measured in

the same domain: Assessment condition, task, emotional

state, and many other factors cause ages to vary dramat-

ically. There are no legitimate developmental milestones,

stones fixed in the developmental roadway in one posi-

tion. Instead, there are developmental buoys, moving

within a range of locations affected dynamically by var-

ious supports and currents.

It is remarkable how pervasively researchers ignore

or even deny variations in shape and age of development.

Scholars committed to a continuous view of development

typically ignore the spurts and drops in many develop-

mental functions, insisting that development is smooth

and continuous despite major evidence to the contrary.

Physical and psychological development are both rou-

tinely graphed with smooth curves, as in the charts in a

pediatrician’s office, even though research on individ-

ual growth consistently shows patterns of fits and starts

in virtually all aspects of physical growth (Lampl &

Johnson, 1998). The distortion is just as pervasive in

psychological development. For example, Diamond’s

(1985) findings of linear growth of memory for hidden

objects in infancy are frequently cited, even though

replications by others with the same tasks and measures

show nonlinear, S-shaped growth (Bell & Fox, 1992,

1994). Many data sets show powerfully nonlinear indi-

vidual growth as the norm in infant cognitive and emo-

tional development as well as development at later ages

(Fischer & Hogan, 1989; McCall, Eichorn, & Hogarty,

1977; Reznick & Goldfield, 1992; Ruhland & van

Geert, 1998; Shultz, 2003).

In a similar manner, at the other pole of argument,

scholars committed to stage theory often ignore the evi-

dence for continuous growth, even in their own data.

For example, Colby, Kohlberg, Gibbs, and Lieberman

(1983) asserted that their longitudinal data on moral de-

velopment showed stages in growth even in the face of

clear evidence that growth was gradual and continuous

(Fischer, 1983). In the same way for age, scholars rou-

tinely talk as if there are developmental milestones at

specific ages, despite the massive evidence of variability

in age of development with variations in conditions of as-

sessment (Baron-Cohen, 1995; Case, 1985; Spelke, in

press). Common claims, for example, are that object per-

manence develops at 8 months in Piagetian assessments,

conservation at 7 years, and combinatorial reasoning at

12 years, although no such statement is tenable without

more specification because the ages vary greatly with

task, support, and so on. Classic research on reflexes in

very young infants even demonstrates variability in the

ages at which they emerge and disappear (Touwen, 1976).

Starting in the Middle of Things:

Implications for Design

To study development in medias res—in the middle of

things—research designs need to be broadened so that

they capture the range of variation and diversity of

human activities in real-life settings. If development is

assessed with an instrument that places all behavior on a

single linear scale, for example, then nothing but that lin-

ear change can be detected. The limitations of most clas-

sical research arise from assumptions that restrict

observation and theory to one-dimensional analysis.

When those assumptions are changed, research opens up

to encompass the full range of human activity. By limit-

ing developmental observation and explanation to one-

dimensional processes, the static assumptions have

stymied investigation of the richly textured dynamic

variations of development. To do research that facilitates

multidimensional-process explanation requires building

research designs that go beyond one-dimensional as-

sumptions to provide for detection of the dynamics of

variability (Edelstein & Case, 1993; G. Gottlieb,

Wahlsten, & Lickliter, Chapter 5, this Handbook, this

volume; Lerner, 2002; Thelen & Smith, 1994, Chapter 6;

Valsiner, Chapter 4, this Handbook, this volume; van

Geert & van Dijk, 2002).

Here are four important one-dimensional assump-

tions that are typically incorrect and that are embodied

in research designs that implicitly assume static struc-

ture. These all need to be avoided in designs for assess-

ing the dynamics of change by addressing variability

and diversity.

Methodology of Dynamic Structural Analysis 349

1. Single-level, single-competence assumption—not. At

any one moment, a person functions at a single

cognitive stage or a single level of complexity and

possesses a single competence. Contrary to this one-

level, one-pathway assumption, people function at

multiple developmental levels concurrently, even

within the same situation (Fischer & Ayoub, 1994;

Goldin-Meadow & Alibali, 2002; Siegler, Chapter

11, this Handbook, Volume 2). In development, a per-

son moves through a web of connected pathways

composed of multiple strands (domains or tasks),

each involving variation within a range or zone of de-

velopmental levels, as illustrated in the webs in Fig-

ures 7.2 and 7.9. Assessments must include multiple

pathways and multiple conditions so that the full

range of levels and competences can be detected.

2. Single-shape assumption—not. Each developmental

pathway shows essentially similar linear or monoto-

nic shapes. Contrary to this linearity assumption,

developmental pathways or strands take many dif-

ferent shapes, which frequently include reversals in

direction—not only increases but also decreases, as

illustrated in Figure 7.11. Individual people nor-

mally grow in fits and starts both physically and

psychologically, as we described in the introduction

to this section. In development, these fits and starts

seem to be especially prevalent and systematic

when people are functioning at optimum or when

they are building a new skill in microdevelopment.

Developmental pathways or strands for individual

activities move through nonlinear dynamic patterns

of change, seldom showing straight lines. In long-

term development, there are periodic movements to

a lower level (regressions), especially after develop-

mental spurts (Fischer & Kennedy, 1997). In mi-

crodevelopment, backward movement to a low-level

skill is common before construction of a new skill

(Granott & Parziale, 2002), as we discuss in the

section on Microdevelopment.

3. Single-person assumption—not. People develop and

learn individually, and they sometimes interact and

affect each other. Contrary to this individualist as-

sumption, people do not usually function solo, but in-

stead from birth they act in a fundamentally social

way, working together in ensembles that distribute a

task across several collaborating partners (Bronfen-

brenner & Morris, Chapter 14, this Handbook, this

volume; A. Brown & Palincsar, 1989; Scardamalia &

Bereiter, 1999; Vygotsky, 1978). Studying develop-

ment socially is not only more realistic, but it can also

make the processes of development more transparent.

When people work together, communicating about

what they are doing, the internal processes of learn-

ing and thinking become externalized, and the

processes of social collaboration and interference be-

come evident (Fischer & Granott, 1995).

4. Single context assumption—not. The most effective

research typically focuses on one task and variations

on it or one context for assessment. Contrary to this

uniformity assumption, research needs to combine

multiple tasks and assessment contexts so that it can

capture the range of levels and competences, path-

ways, and social interactions that characterize devel-

opment (Bronfenbrenner, 1993; Campbell & Stanley,

1963; Fischer, Knight, et al., 1993). To accurately de-

scribe people’s developing activities, research must

be designed with an array of assessment conditions

and an array of tasks within conditions.

Guidelines for Developmental Research

To analyze and understand natural variations in de-

velopment as well as consistencies across variations,

research needs to move beyond these limiting as-

sumptions. Analyzing the dynamics of change requires

Figure 7.11 Three different growth curves based on the

same growth model. The growth curves are all generated by

the same nonlinear hierarchical model of development of self-

in-relationships used in this chapter, but variations in the val-

ues of the parameters in the equations produce vastly different

shapes. The same growth processes produce essentially monot-

onic growth (Grower 1), growth with stagelike spurts and

drops (Grower 2), and fluctuating change (Grower 3).

800

700

600

500

400

300

200

100

0

1

2

3

4

5

6

7

8

Events or Age

Level

0

2

1

3

350 Dynamic Development of Action and Thought

methods that allow detection of variations in develop-

ment and learning:

• People develop along multiple concurrent path-

ways in a web.

• From moment to moment people function across a

range of different levels and competences.

• People develop in the long run and learn in short time

periods according to diverse shapes of growth, in-

cluding the complex nonlinear fits and starts in many

growth curves.

• People learn and develop in social ensembles, and re-

search should reflect this fundamentally social na-

ture of development.

• People act differently in different tasks and condi-

tions, and so research needs to include a range of

tasks and conditions to detect the full range of vari-

ability in action and thought.

Only through analyzing the natural variability in devel-

opment and learning can researchers come to understand

the consistencies inside the variation.

Putting together all these contributions to variation

can seem daunting, but it need not be. A few straight-

forward guidelines in designing research and analyzing

observations facilitate uncovering the variation and di-

versity of development. Investigators should focus on

(a) using well-designed clocks and rulers to measure

change and variation, (b) studying several tasks and do-

mains to determine the generality and variation in

pathways, (c) varying assessment conditions to uncover

the range of variability in level and content, and (d) in-

vestigating diverse sociocultural contexts to discover

the effects of different cultural groups on development.

No one study can investigate all sources of variation at

once, but investigators can make sure that several

sources are evaluated in each study. Also researchers

need to situate their findings within a conceptual map

of the multiple sources of dynamic development, avoid-

ing the pitfall of reductionist description, which as-

sumes that one study captures the important sources

of variation.

Clocks, Rulers, and Repeated Measures

Detection of variation in developmental shapes requires

both good clocks and good rulers to measure change. To

capture either smooth growth or fits and starts requires

a clock that can detect the speed of change. Ages or

events need to be sampled frequently enough to provide

several assessments for each period of increase and

decrease. Otherwise, the shape of growth cannot be de-

tected. Also, the distribution of ages or times of assess-

ment must be chosen carefully so that estimates of

changes in item or response distributions are not dis-

torted by biases in time sampling. Much developmental

research uses clustered ages such as groups of 2- and 4-

year-olds clustered tightly around the mean ages of 2

and 4. This design assumes the importance of mean dif-

ferences and provides a bad clock for development, be-

cause it represents only a few of the many points along

the time scale from 2 to 4 years. If major reorganiza-

tions of activity are hypothesized to occur, for example,

every 6 months in the early preschool years as Case

(1985) predicted, then assessments must take place at

least every 2 or 3 months to reliably detect the periods

of reorganization, and the distribution of ages across 2-

or 3-month intervals should be uniform, not clustered at

the mean age.

Capturing the shapes of development requires a good

ruler as well, one that provides a scale sensitive enough

to detect the ups and downs of growth. The best assess-

ments provide a relatively continuous developmental

scale of increasing complexity, such as the Uzgiris-Hunt

(1987) scales to assess infant development and the

scales for nice and mean social interaction (Ayoub &

Fischer, in press; Fischer, Hencke, & Hand, 1994). It is

crucial to avoid scales that combine items in a way that

forces growth into a particular function, as when intelli-

gence tests force test data into scales that show linear

increase with age.

A single task seldom makes a good ruler because it

provides such a limited sample of behavior. Better is a

series of tasks or a grouping of tasks that forms coher-

ent developmental scales. A series of tasks can be used

to assess either (a) a Guttman-type developmental scale

measuring one linear pathway in a developmental web

(Guttman, 1944), like the Uzgiris-Hunt scales, or (b)

branching pathways like the tasks for nice and mean in-

teractions and those for reading single words in Figures

7.5 and 7.10. Through analysis of profiles across tasks,

a good ruler can be created for either pathway. A partic-

ularly useful method is Rasch scaling, which is based

on a sensible, nonlinear (logistic) developmental model

and allows detection of Guttman scales as well as

branches (Bond & Fox, 2001; Dawson, 2003; Rasch,

1980). The discovery of the general ruler for hierarchi-

cal skill development came from research assessing per-

formance profiles with these and related methods, as

Methodology of Dynamic Structural Analysis 351

TABLE 7.2 Task Profiles for Normative Developmental Sequence for Reading Words

Word

Letter

Rhyme

Reading

Rhyme

Reading

Step

Definition

Identification

Recognition

Recognition

Production

Production

0

−

−

−

−

−

−

1

+

−

−

−

−

−

2a

+

+

−

−

−

−

2b

+

−

+

−

−

−

3

+

+

+

−

−

−

4

+

+

+

+

−

−

5

+

+

+

+

+

−

6

+

+

+

+

+

+

Note: Pass = +; Fail = −.

Adapted from “Learning to Read Words: Individual Differences in Developmental Sequences,” by

C. C. Knight and K. W. Fischer, 1992, Journal of Applied Developmental Psychology, 13, pp. 377–404.

discussed in the section titled A Common Ruler for

Skill Development.

Table 7.2 shows a set of profiles for defining the sim-

plest developmental pathway in the development of read-

ing words—Figure 7.10 (a), the pathway for normal

readers, which includes only one simple branch (Knight

& Fischer, 1992). The sequence is determined by the or-

dering patterns for every pair of tasks. For most profiles

in this simple sequence, every task is passed up to a cer-

tain point in the table from left to right, and then all tasks

are failed thereafter, which is characteristic of a Guttman

scale. Branching is indicated by profiles that show varia-

tions in this simple pattern, such as Step 2b in Table 7.2,

where there is a failed task in the middle of a string of

passes. Based on analysis of performance across tasks for

each word, a child is assigned a profile in Table 7.2, and

therefore a step in the pathway, even when assessment is

at a single time rather than longitudinal. The table shows

pass/fail tasks for simplicity, but multistep scales can be

used, with scaling tested by tasks earlier in a sequence

having higher scores than those later.

Profile analysis can detect webs as simple as the one

for normal readers in Figure 7.10 (a), or as complex as

the one for nice and mean social interactions in Figure

7.5. The logic of analysis is the same for branched webs

as for linear Guttman scales, and sequencing is deter-

mined by the ordering patterns of all pairs of tasks. In-

deed, the same set of tasks can define different webs for

different children. For example, different sets of profiles

for the tasks in Table 7.2 define the unintegrated webs

for poor readers in Figure 7.10 (b) and (c), such as the

web in which the three domains of identifying letters,

reading words, and rhyming words are all independent.

An important characteristic to keep in mind when de-

vising tasks to build rulers for change is the similarities

and differences among tasks. Simple ordering like that

in Table 7.2 is typically eliminated by differences in

content or procedure between tasks. When researchers

have attempted to build scales using distinctive tasks to

assess different steps, the task differences have wiped

out scaling of steps (Kofsky, 1966; Wohlwill & Lowe,

1962). A good, simple Guttman-type ruler uses tasks

that include only variations in complexity or difficulty,

with minimal differences in content and procedure. Dif-

ferences between distinctive tasks are captured by hav-

ing separate Guttman rulers for each set of tasks (each

domain). In a similar way, measuring the temperature of

a refrigerator in New York requires a different ther-

mometer from measuring the temperature of the surface

of Mars. Rasch analysis can also facilitate using a com-

mon scale across tasks and domains (Bond & Fox, 2001;

Dawson et al., 2003), as it has helped test the generality

of the ruler for skill complexity, showing simultaneously

the same scale across domains and large domain effects.

Another method for devising a ruler uses groupings of

similar tasks to assess a scale. For example, in early lan-

guage development, Ruhland and van Geert (1998)

grouped words into syntactical classes based on Dutch

children’s spontaneous speech to form a sensitive devel-

opmental scale. With pronouns, for example, they found

a large growth spurt late in the second year, as shown in

Figure 7.12. Other groupings that have proved useful in

studies of development have included arithmetic prob-

lems of similar complexity (Fischer, Pipp, & Bullock,

1984) and explications of dilemmas about the bases of

knowledge called reflective judgment (K. Kitchener

et al., 1993). Scales based on such groupings of similar

tasks can be used to specify the shapes of development

in various domains and to compare relations among de-

velopment across domains or levels in individual sub-

jects or groups. Like scalogram analysis, they also

provide a way of testing developmental functions with

352 Dynamic Development of Action and Thought

Figure 7.12 Development of pronoun use in the Dutch boy

Tomas. Source: From “Jumping into Syntax: Transitions in

the Development of Closed Class Words,” by R. Ruhland and

P. van Geert, 1998, British Journal of Developmental Psychol-

ogy, 16(Pt. 1), pp. 65–95.

160

140

120

100

80

0

20

40

60

80

100

120

Age in Weeks

Number of Pronouns/Session

cross-sectional designs. For example, this method can

test for both spurts and bimodal distributions on emer-

gence of developmental levels or growth of separate

strategies for approaching a task (Siegler, 2002). The de-

sign must included separate groups of tasks for each

level or strategy. The grouping method, however, does

not provide a sensitive index of the intervals in a scale

between points of discontinuity, or levels.

Rasch analysis fills this need, providing powerful

tools for assessing the steps and intervals along a scale

as well as discontinuities (Bond & Fox, 2001; Rasch,

1980). Only recently have researchers begun to realize

its potential for assessing developmental scales and de-

termining the distances between items along a scale.

Most investigators have used it to determine whether

items in a domain fit a single Guttman scale and what

the distances are between items along that scale, and it

can also be used to assess for several independent scales

or branches in a web. Rasch scaling provides one of the

most convincing sources of evidence for the scale for

hierarchical complexity of skills shown in Figure 7.3

(Dawson, 2003; Dawson et al., 2003).

The three techniques for combining tasks to form

developmental scales (Guttman scaling, groupings of

similar tasks, and Rasch analysis) provide a repeated-

measures assessment that has many of the desirable

characteristics of longitudinal assessment, even when

there is only a single session. Through analysis of task

profiles and distributions, each person can be tested to

determine whether he or she follows a particular devel-

opmental pathway or growth function. Contrary to

the conventional wisdom that development can only

be effectively assessed longitudinally over months

and years, these repeated-measures assessments can

provide powerful tools for describing and testing devel-

opmental pathways and growth functions. They can

also be combined with longitudinal designs, where

they provide even more powerful tools for assessing

development.

General Structure across Tasks in a Domain

Task differences are typically controlled for and

systematically manipulated in developmental scales.

However, task differences are important in their own

right. Task is one of the most powerful sources of vari-

ability in behavior, as documented by thousands of

psychometric and experimental studies across many

decades (Fleishman, 1975; Mischel, 1968). An accurate

portrait of development requires assessment of different

tasks and domains to capture patterns of variation in de-

velopmental pathways and growth functions.

One of the most common hypotheses in cognitive and

developmental science is that behavior divides into do-

mains, which are built on general psychological

structures. That is the core of the domain specificity

framework and of neo-nativist explanations. However,

evidence for generality in a conceptual structure is rela-

tively rare in the research literature, where careful tests

of generalization are infrequent (Fischer & Immordino-

Yang, 2002). Many abilities that have been described as

general competences seem not to be coherent abilities at

all but instead summary variables, with at best weak cor-

relations among items. Examples include the hypothe-

sized domains of theory of mind, metamemory, visual

thinking, and ego resiliency, for each of which there is

no clear evidence of a central generalized structure that

generates common activity across a wide array of tasks.

For example, ego resiliency has been posited as a broad

characteristic of effective people and it has been sub-

jected to extensive longitudinal study by Jack and Jeanne

Block (Block, 1993). Research on this general compe-

tence in Dutch and American children indicates that ego

resiliency does not affect relevant specific competences

such as school achievement and social preference (Har-

ter, 1999; van Aken, 1992). That is, it does not show a

generalizing relation with specific skills, which would

indicate a common structure applied across tasks. Ego

resiliency may be a useful social construct, but it does

not seem to be a central psychological structure that or-

ganizes various activities together in development.

Methodology of Dynamic Structural Analysis 353

One convincing case of a general structure in a do-

main is the central conceptual structure documented by

Robbie Case and his colleagues (1996). It provides a

model for defining a general structure and testing its

generality. Assessment of the development of a general

concept of number requires an array of tasks that all re-

quire the use of that concept. Case and his colleagues

have constructed such a task array for the elementary

number line, which represents number as quantitative

variation along a line. This representation constitutes

what they call a central conceptual structure for number

in young children, a framework for thinking about num-

ber that facilitates numerical understanding across many

situations. Tasks like reading the time on a clock, count-

ing gifts at a birthday party, and doing simple arithmetic

problems in school all make use of this same structure.

Discovery of general conceptual structures like the

number line would be a strong boon for educators,

greatly streamlining their efforts to teach children the

basic concepts and skills required by modern society.

From approximately 4 to 8 years of age, children

build the central conceptual structure for number.

When instructors and curriculum explicitly teach the

structure, children evidence a major improvement in

performance across a wide array of number tasks but

not for tasks in other domains such as understanding so-

cial interactions. The change amounts to as much as

50% of the variance in test scores, which is a remark-

ably large effect. The use of many tasks allowed Case

and his colleagues to determine how general the struc-

ture is—where children apply it and where they do not.

Note also that along with the general change across

number tasks, the researchers still found large task ef-

fects and considerable developmental variation in level.

The generality of the structure operates within this

substantial variability.

In the behavioral sciences, researchers commonly

wish to generalize from their data to the development

of a domain, but the two standard methods preclude le-

gitimate generalization by artificially reducing varia-

tion instead of analyzing it. First, in the “psychometric

method,” commonly used in intelligence, education, and

personality testing, many tasks are summed and only

the summary scores are considered. A boy’s IQ score is

116, or the college entrance test score for a young

woman is 575. Most of the variation in each person’s

performances on the tasks is ignored. Second, in the

“experimental method,” commonly used in experimental

psychology and neuroscience, a researcher analyzes one

task by varying a parameter and calculating mean per-

formance differences for specific values of the parame-

ter. Variations in performance in the task other than the

means are treated as error variance and not analyzed

further. Also, variations among diverse tasks are ig-

nored because only one task is examined.

The psychometric strategy is evident in ability theo-

ries, where researchers study some hypothesized gen-

eral ability such as spatial intelligence or verbal

intelligence (Demetriou et al., 2002; Sternberg, Lautrey,

& Lubart, 2003). The evidence for the coherence of

these supposedly modular abilities is modest in compar-

ison to Case’s evidence for a central conceptual struc-

ture for number. Most tasks or items that measure each

ability or intelligence have only minimal variance in

common, with correlations between pairs of items typi-

cally accounting for approximately 4% of the variance

(an average correlation of .2 between individual items).

Educational researchers have regularly thrown up

their hands in dismay that they have found so little gen-

eralization or transfer of concepts to tasks that are dis-

tinct from those taught (Salomon & Perkins, 1989). For

example, when instructors teach a concept such as grav-

ity, evolution, or working memory, they commonly find

that even intelligent students have difficulty using the

concept in tasks different from those explicitly taught

in class. The reason for the difficulty of this far gener-

alization (use of knowledge in tasks far from the origi-

nal object of learning) is that the construction of

generalized skills requires time and effort (Fischer &

Immordino-Yang, 2002). Furthermore, even with a

strong conceptual structure like Case’s number line,

generalization is not perfect. For a weak structure such

as spatial intelligence or eg