_METACOGNITION, |

MOTIVATION,
AND UNDERSTANDING 3
_ Edited by
2 Franz E. Weinert
Rainer H. Kluwe
DUQUESNE UNIVERSITY LIBRARY
 Digitized by the Internet Archive
in 2022 with funding from
Kahle/Austin Foundation

https://archive.org/details/metacognitionmot0000unse
Metacognition,
Motivation,
and Understanding
THE PSYCHOLOGY OF EDUCATION AND INSTRUCTION
A series of volumes edited by:
Robert Glaser and Lauren Resnick

Bussis, CHITTENDON, AMAREL, KLAUSNER - Inquiry into Meaning: An

Investigation of Learning to Read
CARROLL - Perspectives on School Learning: Selected Writings of John B.
Carroll
CHIPMAN, BRUSH, WILSON - Women and Mathematics: Balancing the
Equation
Dusek - Teacher Expectancies
FOORMAN and SIEGEL - Acquisition of Reading Skills
Paris, OLSON, STEVENSON - Learning and Motivation in the Classroom
Thinking and Learning Skills
SEGAL, CHIPMAN, GLASER - Relating Instruction to Research, Vol. 1
CHIPMAN, SEGAL, GLASER - Research and Open Questions, Vol. 2
BEREITER and SCARDAMALIA - The Psychology of Written Composition
Weinert and KLuwe - Metacognition, Motivation, and Understanding
METACOGNITION,
MOTIVATION,
AND UNDERSTANDING

Edited by

Franz E. Weinert
Max-Planck Institute for Psychological
Research, Munich

Rainer H. Kluwe
University of Bundeswehr, Hamburg

[EA LAWRENCE ERLBAUM ASSOCIATES, PUBLISHERS

1987 Hillsdale, New Jersey London
Copyright © 1987 by Lawrence Erlbaum Associates, Inc.
All rights reserved. No part of this book may be reproduced in
any form, by photostat, microform, retrieval system, or any other
means, without the prior written permission of the publisher.

Lawrence Erlbaum Associates, Inc., Publishers

365 Broadway
Hillsdale, New Jersey 07642

Library of Congress Cataloging-in-Publication Data

Metakognition, Motivation und Lernen.
Metacognition, motivation, and understanding.

Translation of: Metakognition, Motivation, und Lernen.

“Based on the Heidelberg symposium on ‘The Development
of Metacognition, the Formation of Attribution Styles,
and Learning’ at Heidelberg University . . . sponsored by
the German Foundation ‘Volkswagenwerk,’ Hannover’—Pref.
Bibliography: p.
1. Learning, Psychology of—Congresses. 2. Metacogni-
tion—Congresses. 3. Motivation in education—Congresses.
4. Cognition—Congresses. I. Weinert, Franz E., 1930-
II. Kluwe, Rainer. III. Title.

LB1065.M3813 1987 370.15 86-29223

ISBN 0-89859-569-X

Printed in the United States of America

MO Oo 7G F a a FD |
 Contents

CONTRIBUTORS

PREFACE
INTRODUCTION AND OVERVIEW: METACOGNITION
AND MOTIVATION AS DETERMINANTS OF EFFECTIVE
LEARNING AND UNDERSTANDING
Franz E. Weinert

General Laws of Learning

and Individual Learning Efficiency, 2
Competence and Performance as Determinants
of Individual Learning Effectiveness, 4
Metacognition and Motivation as Determinants
of Individual Learning Activity, 7
Metacognition, Motivation, and Learning: Progress
in the Integration of Different Research Paradigms, 13
References, 14

Part | Metacognition

SPECULATIONS ABOUT THE NATURE

AND DEVELOPMENT OF METACOGNITION
John H. Flavell
Metacognitive Knowledge, 22
Metacognitive Experiences, 24
vi CONTENTS

Questions, Problems, and Issues, 25

Acknowledgments, 28
References, 29

EXECUTIVE DECISIONS AND REGULATION

OF PROBLEM SOLVING BEHAVIOR 31
Rainer H. Kluwe

The Role of Executive Decisions, 32

A Set of Executive Decisions, 34
Remarks on the Development of Executive Control
and Regulation and Cognitive Performance, 46
Children’s Regulation of Problem Solving as a
Function of Reversibility Versus Irreversibility
of Actions, 49
Acknowledgments, 61
References, 61

METACOGNITION, EXECUTIVE CONTROL, SELF-REGULATION

AND OTHER MORE MYSTERIOUS MECHANISMS 65
Ann Brown

What is Metacognition? 65
Roots of Metacognition, 69
Status of Metacognition as a Concept, 105
Summary, 109
Acknowledgments, 109
References, 109

METACOGNITIVE COMPONENTS OF INSTRUCTIONAL

RESEARCH WITH PROBLEM LEARNERS 117
Joseph C. Campione

Metacognition and its Various Guises, 119

The Status of Instructional Research, 121
Metacognition and Instructional Attempts, 124
Conclusions, 133
Acknowledgments, 137
References, 137
 CONTENTS vii

Part II Motivation and Attribution Style 141

CAUSAL ATTRIBUTION PATTERNS FOR ACHIEVEMENT

OUTCOMES: INDIVIDUAL DIFFERENCES, POSSIBLE TYPES
AND THEIR ORIGINS 143
Heinz Heckhausen

Intraindividual Asymmetry of Attribution Patterns, 145

Interindividual Contrast in Attribution Patterns:
Convergence of Research Domains, 148
Possible Types of Attribution Patterns
and Their Emergence, 159
Concluding Remarks, 175
Postscript, 178
References, 180

6 HELPLESSNESS AND ATTRIBUTIONAL STYLE

IN DEPRESSION 185
Christopher Peterson and Martin E. P. Seligman
Helplessness and Attributional Style
in Depression, 185
Original Helplessness Theorey, 186
The Reformulation, 189
Attributional Style, 190
The Measurement of Attributional Style, 191
The Attributional Style Questionnaire, 193
The Relationship of Attributional Style
to Depression, 196
Attributional Style as a Cause
of Depression, 203
Conclusion: Summary, Critique, and New Directions, 208
Acknowledgments, 211
References, 211

7 FEELING VERSUS BEING HELPLESS: METACOGNITIVE

MEDIATION OF FAILURE-INDUCED
PERFORMANCE DEFICITS 217
Julius Kuhl

Learned Helplessness, 218

Action-Control, 220
Action-Control and Learned Helplessness, 225
References, 233
 vill CONTENTS

Part III Understanding and Learning 237,

8
REPRESENTING KNOWLEDGE AND METAKNOWLEDGE:
IMPLICATIONS FOR INTERPRETING
METAMEMORY RESEARCH 239
Michelene T. H. Chi
Knowledge Approach to Memory Development, 240
A Framework for Representing Knowledge, 245
Metamemory Research, 252
General Discussion, 262
Acknowledgments, 264
References, 264

APTITUDES FOR LEARNING AND COGNITIVE PROCESSES 267

Robert Glaser and James W. Pellegrino
The Relevance of Rule Induction, 268
Implications, 285
Further Research, 287
Acknowledgments, 287
References, 288

10 PROCESSES AND DEVELOPMENT OF UNDERSTANDING

James Greeno and Mary S. Riley
289

Processes of Understanding, 289

Development of Ability to Understand, 297
Conclusions, 311
Acknowledgments, 312
References, 312

AUTHOR INDEX 317

SUBJECT INDEX 323
 Contributors to This Volume

ANN L. BROWN © Center for the Study of Reading, University of

Illinois at Urbana-Champaign, 51 Gerty Drive, Champaign, IL 61820
JOSEPH C. CAMPIONE + Center for the Study of Reading, University of
Illinois at Urbana-Champaign, 51 Gerty Drive, Champaign, IL 61820
MICHELENE T. H. CHI + Learning Research and Development Center,
University of Pittsburgh, 3939 O'Hara Street, Pittsburgh, PA 15260
JOHN H. FLAVELL + Department of Psychology, Stanford University,
Jordan Hall, Stanford, CA 94305
ROBERT GLASER -» Learning Research and Development Center,
University of Pittsburgh, 3939 O’Hara Street, Pittsburgh, PA 15260
JAMES G. GREENO - University of California, Berkeley, Berkeley, CA
94720
HEINZ HECKHAUSEN + Max-Planck-Institut für Psychologische
Forschung Leopold Strasse 24, 8000 Munich 40, Germany
RAINER H. KLUWE + Universität der Bundeswehr, Abt. Allg.
Psychologie, Holstenhofweg 85, 2000 Hamburg 70, Germany
JULIUS KUHL + Max-Planck Institut für Psychologische Forschung,
Leopold Strasse 24, 8000 Munich 40, Germany
JAMES W. PELLEGRINO + School of Education, University of California,
Santa Barbara, CA 93106
CHRISTOPHER PETERSON + University of Pennsylvania, Department of
Psychology, University of Pennsylvania, 3015 Walnut Street,
Philadelphia, PA 19104
x CONTRIBUTORS

MARY R. RILEY » Learning Research and Development Center,

University of Pittsburgh, 3939 O’Hara Street, Pittsburgh, PA 15260
MARTIN E. P. SELIGMAN +» Department of Psychology, University of
Pennsylvania, 3015 Walnut Street, Philadelphia, PA 19104
FRANZ E. WEINERT + Max-Planck Institut für psychologische
Forschung, Leopold Strasse 24, 8000 Munich, Germany
 Preface

This volume is based on the Heidelberg Symposium on “The Development of

Metacognition, the Formation of Attribution Styles, and Learning” at Heidelberg
University. The Conference was sponsored by the German Foundation “Volk-
swagenwerk,” Hannover.
During the last few years developmental research concerned with metacogni-
tion, attribution styles, and self-instruction extended rapidly. Though there are
common theoretical concepts, as well as research questions, all three fields have
been studied separately.
Metacognition refers to cognition about cognition, and means knowledge
about one’s own thinking, and cognitive activity having as its object own think-
ing processes. Attribution styles are considered to be motivational preferences
with respect to explanations of own success and failure, given different types of
tasks. Self-instruction refers to the way and the degree to which a subject initi-
ates, plans and organizes, monitors and regulates own learning processes.
All three approaches refer to the acquisition and availability of knowledge
about cognition, and of cognitive processes to monitor and regulate the stream of
own thinking. This volume represents a first step to integrate these fields of
developmental research with respect to their theoretical basis, and with respect to
some of the results. The contributors to this volume are considered to be experts
in their field.
The research reported here is important for the elaboration of theories of
cognitive development. Furthermore, this work has consequences for education-
al strategies and related applied problems.

F. E. Weinert
R. H. Kluwe

xi
 Introduction and Overview:
Metacognition and Motivation
as Determinants of Effective
Learning and Understanding

F. E. Weinert
Max-Planck-Institute for Psychological Research
Munich

Also, if somebody knows something, then he knows that he knows

it, and at the same time he knows that he knows that he knows.
—Spinoza, 1632-1677

Do you learn more effectively when you know how to learn? Do you learn more
effectively when you are highly motivated? Although our intuitions suggest
positive answers to each of these questions, everyday experiences provide coun-
ter-examples. On the one hand, we can all think of remarkable accomplishments
made by people who consciously employ special techniques for learning. Exam-
ples are actors or orchestra conductors who learn vast amounts of material for
performances, or mnemonists, the ‘‘*memory experts,’’ who entertain us with the
size and precision of their memories. On the other hand, a group of people
without such techniques—for instance children—also learn and remember large
amounts of material. A general belief is that children are more effective learners
than adults. For example, over half the students in an introductory psychology
class believed that children’s memories are better than adults’ memories
(Vaughan, 1977).
Contradictory examples also come to mind when we ask about the effects of
motivation on learning. We can imagine a student studying for an important
exam whose extreme motivation to do well actually interferes with efficient
learning; however, many school children who seem less gifted than their peers
learn effectively when they are keenly interested and highly motivated.
Everyday experience does not provide definitive answers to our questions
about what contributes to effective learning. Can we find a clarification of the
2 INTRODUCTION AND OVERVIEW

effects of skill and motivation on learning at the level of scientific theory? That
is, does the present psychology of learning and memory provide us with an
adequate, acceptable model for analyzing learning effectiveness and the factors
affecting it? A direct answer to this would have to be negative. This unsatisfacto-
ry situation is due to several reasons, among them the pragmatic fact that learn-
ing in everyday situations is variable and complex, and thus difficult to study,
and the historical fact that such questions have been largely neglected in tradi-
tional learning research. However, this situation is changing. Over the last 15
years, converging ideas arising from different theoretical perspectives are being
systematically developed and integrated, promising that a better understanding of
intra— and inter—individual differences in learning effectiveness may now be
possible.

GENERAL LAWS OF LEARNING AND INDIVIDUAL

LEARNING EFFICIENCY

Regardless of the definition and the theoretical explanation of learning used, the
goal of learning psychology has traditionally been to formulate general laws of
learning, valid for all humans if not all higher organisms. Although Hermann
Ebbinghaus (1885), the founder of modern learning research, recognized the
endless individual differences in human memory, he nonetheless called for and
initiated the study of general, universal laws of learning. Beginning with this
research, learning was the central focus of empirical efforts, and the learner was
considered as a source of error to be ignored or controlled for in the description
of lawful regularities governing the average human. Earlier, Wilhelm Wundt
cautioned that ‘there is not a single psychological law that does not admit to
more exceptions than confirmations’ (1886, p. 473), but this observation was
not sufficient to encourage a parallel development of a psychology of individual
differences in learning and memory. However, even the most strident follower of
a behaviorist perspective had to admit that people differed with respect to learn-
ing effectiveness, even when external conditions were held constant. The dilem-
ma this posed was resolved by positing empirically definable, static individual
difference parameters, such as IQ, which could be inserted into general formulas
to describe learning (Hull, 1945). The theoretical problems arising from such a
solution were expressed clearly by Woodrow (1942):

Now these parameters may refer to anything whatsoever, conscious, physiological,

environmental, psychic, or purely imaginary. Here one is free to follow his pre-
dilections, whether for motives, excitatory or inhibitory substance, field
forces,
states of disequilibrium, inertia of the nervous system, abilities, or what not.
(p. 4)
 METACOGNITION AND MOTIVATION 3

Despite the problems in delineating what the relevant dimensions of indi-

vidual differences might be, there were some interesting early studies that em-
phasized individual differences in learning effectiveness. Noteworthy among
these were: the psychometric classification of memory abilities (Spearman,
1927), the consideration of effects of pre-experimental learning experience on
verbal learning in the lab (Plenderleith & Postman, 1957), studies on moti-
vational and social influences on recall (Bartlett, 1932; Zeigarnik, 1927), the
demonstration of a relationship between personality and learning (Eysenck,
1957; Pawlow, 1953), experiments on the importance of learning habits and
study skills for success at school (Robinson, 1941), and the many studies show-
ing developmental differences in learning and memory performance (Netscha-
jeff, 1900; review of early studies by McGeoch & Irion, 1952). Although there
are other noteworthy examples of an early psychology of individual differences,
none of these findings altered the intensive search for general regularities in
human learning.
The emphasis in the attempt to find regularities in learning began to change in
the 1950s when a ‘‘new look’’ (Bugelski, 1956) in learning research occurred, a
change that mirrored the new look in theories of cognition and memory. Al-
though theories of cognition and memory were also clearly oriented toward
formulating general and universal processes, individual differences assumed a
new role. Differences in the learner’s prior knowledge base, in available cog-
nitive processes and in learning and memory strategies, were used to explain
inter— and intra—individual performance variation in learning and memory tasks.
One lasting result of this new look is that it is, at present, not clear at what level
of generalization regularities in learning, forgetting, and remembering should be
formulated.
Do individual learning histories result in different functional relations be-
tween the external parameters of learning situations and learning outcomes (e.g.,
an aptitude-treatment interaction)? Or should one hypothesize functional invar-
iance in the relation between input and outcome (i.e., a universal learning mech-
anism)? More concretely, one observes learning differences when younger and
older children, or experts and novices, or people differing in intelligence levels
are compared. Do these differences arise because the older, more expert, or more
intelligent people learn faster, more or better (e.g., they have similar, but more
efficient learning mechanisms), or do they arise because of more fundamental
differences in how different people learn, and if so, what are these differences?
Although these are important theoretical questions, contemporary empirical
efforts are reasonably enough not focused on answering such broad issues.
Rather, contemporary work is focused on the clarification of several, more
specific issues concerning learning and memory. The development and effects of
differences in learning competence plays a limited but noteworthy role in current
explanations of differences in learning effectiveness.
 4 INTRODUCTION AND OVERVIEW

COMPETENCE AND PERFORMANCE AS

DETERMINANTS OF INDIVIDUAL LEARNING
EFFECTIVENESS

When one reviews the history of scientific concern with individual differences in
learning effectiveness, it is clear that static approaches have overshadowed dy-
namic ones. This is not surprising if one considers that psychological thinking in
this century is deeply rooted in the ‘‘faculty psychology’? of past eras. From the
faculty psychology perspective, it seems self-evident that individual differences
in learning and memory performance can be explained by stable capacity dif-
ferences in the learner. A classic example of this viewpoint is a formula devel-
oped by Thurstone (1930) to express the amount of time required to learn non-
sense syllable lists. In contrast to many learning theorists of his time, Thurstone
did not consider that the time required to learn a list of items was a simple
function of the number of items. Instead, he added two individual difference
parameters, memory span and learning capacity. His formula is as follows:

T=c/k(nVn-a )

In this formula, T is the time required to recall a row of syllables, c and n are
variables for the task conditions: c refers to the criterion for learning in the
particular task and n refers to the number of items in the list. K and a are
individual difference variables: k is a constant for the individual’s learning capac-
ity and a is a constant for the learner’s memory span. It proved relatively easy to
determine a measure for a, the memory span constant (although by
1925
Guilford and Dallenbach had compiled at least 16 memory span measures).
However, it did not prove possible to find a useful measure for k, the learning
capacity constant in Thurstone’s formula. Despite many attempts to determine
a
valid measure on the basis of general intelligence or specific memory
abilities,
this construct could not be operationalized (McGeogh & Irion,
1952). Later
attempts by developmental psychologists to explain age-dependent
changes in
learning and memory performance in terms of the development of
general intel-
lectual competence of general memory capacity met a similar lack
of success (see
Weinert, 1979). Learning performance, whether in experimental
or school set-
tings, could not be predicted on the basis of correlations with
either global or
specific capacities (McNemar, 1964).
Attempts to predict learning and memory performance met
with more success
when researchers began to move away from the psycho
metric classification of
learning capacities and began to analyze those cognitive structur
es and processes
that were postulated as necessary components for solving
items specific to cog-
nitive abilities. An important impetus for this line of
reasoning was Glaser’s
(1972) conceptualization of the “new aptitudes”’ and
his related empirical work.
Glaser and Pellegrino (this volume) summarize this
work:
 METACOGNITION AND MOTIVATION 2

The research carried out along these lines is predicated on the assumption that
aptitude tests should be viewed as more than primarily predictors of achievement.
Rather, such tests should assist in identifying the processes involved in intellectual
competence, and further indicate how these processes can be influenced and uti-
lized to benefit learning. (p. 267)

The analysis of competence in terms of rule induction allowed Glaser and Pel-
legrino (this volume) to identify:

Three interrelated factors . . . that appear to differentiate high and low skill indi-
viduals. One is the management of memory as this is reflected by speed of perfor-
mance and the handling of demands on working memory. Secondly, individuals
show differences in their knowledge of the constraints of problem-solving pro-
cedures. . . Thirdly, the structure of the declarative-conceptual knowledge base
and the level of representation of this knowledge differ as a function of ability.
High skill individuals employ conceptual forms of knowledge that constrain their
induction of relations, whereas low skill individuals encode their knowledge at
more concrete surface levels; this is manifested by their limited inferential power.
(p. 285)

Greeno and Riley (this volume) studied the comprehension of mathematical

problems, and provide empirical support for the theoretical distinction between
problem schemata (semantic knowledge) and action schemata (procedural
knowledge—plans for solving problems, knowledge about actions at different
levels of generality). M. Chi (this volume) similarly differentiates domain specif-
ic knowledge, general strategies, and metacognitive knowledge as determinants
of learning and memory performance. In this development there has been a
general emphasis on the individual’s accessible, domain-specific knowledge.
Here, command over facts is ascribed less important than the availability of
relevant concepts and rules. Chi (1978), for example, demonstrated that many
typical age differences in memory performance can be explained by differences
in the knowledge base: developmental differences in memory were equalized or
even reversed when younger experts and older non-experts (novices) were com-
pared in a specific content domain (chess). These results have since been repli-
cated many times.
In addition to declarative knowledge, the authors of the chapters pertaining to
learning psychology in this volume postulate another form of knowledge, one
that is important for the discussion of planning and self-direction in problem-
solving on specific tasks. This procedural knowledge allows the learner to use
and manipulate his skills and knowledge through the organization of learning
activities. Such process-related aspects of learning have been emphasized in the
recent developmental literature. There are three frameworks that emphasize this
perspective:
6 INTRODUCTION AND OVERVIEW

(1) The meaning of activity, actions, and related experiences for the ontogeny
of mental abilities is emphasized in the soviet literature. For example, Rubinstein
(1969) defined competence as ‘‘systems of generalized psychic activities, that
are fixed in the individual’’ (p. 110). According to this theoretical viewpoint, the
study of the nature and development of human abilities must take into account
the demands of the socio-cultural environment, and the objectives of the indi-
vidual. Integrating both these demands, this approach begins with a broad,
functional analysis of actions in social context. This perspective, applied to the
study of learning and memory, raises some interesting developmental questions,
although as yet there have been few definitive findings (Reese, 1979; Meacham,
1983).
(2) Development during middle childhood is increasingly interpreted in terms
of an increase in competence and self-regulation. Piaget’s concept of biological
and cognitive autoregulation is an important point of departure for this interpreta-
tion (see Piaget, 1976, 1978; Brown, this volume). According to Piaget (1971):

What knowledge really consists of is not just the acquisition and accumulation of
information, since that alone would remain inert and sightless, as it were. But
knowledge organizes and regulates this information by means of autocontrol sys-
tems directed toward adaptation, in other words, toward the solving of its prob-
lems. (p. 61)

Investigations of knowledge over one’s actions and over the self as actor are one
example of the psychological interest in self-control systems. The assumptions
underlying such studies of metacognitive development have been described by
W. Mischel (1981):

We view children as potentially sophisticated (albeit fallible) intuitive psychol-

ogists who come to know and use psychological principles for understanding social
behavior, for regulating their own conduct, and for achieving increasing mastery
and control over their environments. (p. 240)

In addition to the formation of self-regulatory skills, a number of other pro-

cesses have been studied within the framework of motivation and personality
research. These include the development of control beliefs (Bandura, 1981;
Krampen, 1982), causal attributions (Heckhausen, this volume; Peterson & Se-
ligman, this volume), and other regulatory factors connected with achievement-
related actions (Heckhausen, 1982; Trudewind, 1982).
(3) Research oriented toward a micro-genetic perspective emphasizes that
learning and remembering are not simply determined by available learning and
memory abilities, but that one must also consider the meaning of the learning
activities in the specific task situation. In this orientation, system-theoretic
ap-
proaches (Carver & Scheier, 1981) and action-theoretic approaches (Hacker,
Volpert, & von Cranach, 1982) play an often mixed role. The main interests
are
 METACOGNITION AND MOTIVATION 7

in questions of how task perception, action goals, planning, regulation, self-

evaluation, and self-correction are related to task behavior, as well as how
experience is utilized in the learning process, and how the experience of success
or failure is processed. This orientation focuses on how cognitions and moti-
vations guide learning-related actions (see Kluwe, this volume; Kuhl, this
volume).

METACOGNITION AND MOTIVATION AS

DETERMINANTS OF INDIVIDUAL LEARNING ACTIVITY

To illustrate the phenomena that should be studied in order to analyze how

metacognition and motivation may affect learning activities, imagine a student
attending a lecture just before an important examination. From the beginning of
the lecture, the student concentrates on comparing what is being said with what
he already knows in order to find out what information he lacks. If the student
perceives a gap in his knowledge, he may attempt to take notes on the relevant
information, to understand it, or if possible, to memorize it. If the student has the
impression that these efforts have not been successful, hectic activities or passive
resignation may appear.
After the lecture the student may try to borrow notes from a fellow student,
for further study. If so, we could also observe learning activities from a written
text: the student skims the contents, underlines important passages, looks up or
studies unclear parts, memorizes important points by silently repeating them, and
finally checks whether he has successfully mastered the material. We may then
observe different emotional reactions, depending on whether the student is
pleased or not with the achieved results. If we could then question this imaginary
student about the reasons for his various activities, he might tell us how his
assessment of his own knowledge compares with his expectations about the
demands of the upcoming examination, and about the difficulties he has with the
material. In addition, we may learn about the student’s preferred learning strat-
egies, and his evaluation of his own situation and the possible consequences.
It seems intuitively obvious that behaviors such as those just described must
have an important effect on learning effectiveness. Indeed, asking what such
effects might be is the focus of many lively research efforts subsumed under the
labels of metacognition and achievement motivation.

What Are ‘‘Metacognitions” and What Effects Do They

Have on Learning Behavior and Learning
Performance?
Psychology in general, and developmental psychology in particular, are present-
ly awash in a ‘‘meta’’ flood (metacognition, metacommunication, metamemory,
metalanguage, and so on). Although some believe that these constructs reflect a
8 INTRODUCTION AND OVERVIEW

new and stimulating research perspective (e.g., Paris & Lindauer, 1982), others
find fault with the vague and often arbitrary distinction between cognition and
metacognition. Apart from this fundamental controversy, the construct of meta-
cognition is very attractive, especially within developmental psychology, as
suggested by a number of recent review articles (Brown, this volume; Cavanaugh
& Perlmutter, 1982; Schneider, 1985; Wellman, 1983). At first blush, this
attraction may seem paradoxical. Most developmental authors agree with Cav-
anaugh and Perlmutter’s (1982) evaluation: “‘the present state of metamemory is
not good’’ (p. 22). However, they also would agree with Cavanaugh and Perl-
mutter’s further argument that the concept is useful, and should be retained, and
with their call for an intensification of research in the area. The reasons for
interest in an increase in metacognitive research are presented in several chapters
in this volume: Brown; Campione; Chi; Flavell; Kluwe; Kuhl. In this volume we
present only a brief introduction to the themes, findings, problems and perspec-
tives of metacognition research.
On the surface, it seems easy to distinguish between cognition and metacogni-
tion. Metacognitions are second-order cognitions: thoughts about thoughts,
knowledge about knowledge, or reflections about actions. However, problems
arise when one attempts to apply this general definition to specific instances.
These problems concern whether metacognitive knowledge must be utilized,
whether it must be conscious and verbalizable, and whether it must be gener-
alized across situations. For example, should someone who has learned from a
psychology textbook that distributed practice is more effective than massed
practice, but who does not apply this knowledge in actual learning situations, be
granted metacognitive knowledge of learning processes? Or, consider a child
who takes more time to study difficult than easy items when learning a word list
(a learning strategy), but who is not aware of the easy-difficult distinction, and
who is unable to describe his learning strategy. Does this child have metacog-
nitive knowledge? Finally, consider a driver who strategically finds clever short
cuts when travelling in a familiar part of town, but who becomes disoriented in
unfamiliar districts. Should the driver be denied metacognitive knowledge be-
cause it is not always displayed? The problems in answering such questions
highlight the criticism that the working definition of metacognition is vague and
imprecise.
Similar difficulties arise when one considers a specific type of metacognition,
metamemory. Metamemory is generally defined as: the knowledge one has about
memory in general; knowledge about the peculiarities of one’s own memory
system; sensitivity to past experience with memorizing, storing, and retrieving
different types of information in various situations; and a system of skills for
planning, directing, monitoring, and evaluating one’s behavior during learning
and remembering. Such an omnibus definition does not easily afford a the-
oretically or thematically coherent research program. Indeed, the field is charac-
 METACOGNITION AND MOTIVATION 9

terized by a variety of different lines of research that have little in common apart
from the label ‘‘metamemory.’’ Examples of this variety are studies on: de-
clarative metaknowledge, executive processes, the relationship between domain
specific knowledge and metaknowledge, predictions and evaluations about one’s
own achievements, and so on. The data collection procedures are similarly
variable and include interviews and ‘‘thinking aloud,’’ questionnaires and rating
scales, and observations of behavioral indicators of metacognitive activity. It is
not surprising that the empirical relations among such diverse measures are
variable—a finding that, often noted, need not speak against the utility of the
construct of metamemory.
The metamemory framework is used to provide explanations for some phe-
nomena, and to generate practical suggestions for improving cognitive pro-
cesses. For example, lack of metamemorial knowledge is used to account for the
production deficiency of younger children (available and useful memory strat-
egies were/are not used, Flavell, 1970). On the applied side, training programs
for metacognitive skills have, in many cases, led to stable improvements in
reading, learning, and memory achievements in both average (Pressley, Forrest-
Pressley, & Elliot-Faust, in press) and below average subjects (Campione, this
volume). In addition, metacognitive variables have proved to be the best predic-
tors of memory performance differences for both children and adults (Weinert,
Knopf, & Barann, 1983; Weinert, Knopf, & Körkel, 1983). Given these positive
results, why then is the construct metamemory criticized? Although some of the
problems, such as the diversity in research questions, procedures, and opera-
tionalization have been previously summarized, there are three points that bear
additional comment:

(1) The conceptualization and operationalization of constructs like meta-

memory are hampered by the “‘consciousness’’ problem, a persistently unsolved
issue in psychology. Although this issue is frequently omitted (or ignored) in
studies of cognition, it is not possible to do so when investigating metacognition,
where the measures may include introspection or verbal reports about one’s own
behavior, and where the relation between tacit, implicit, and explicit knowledge
is important. As Brown (this volume) points out, one’s position with respect to
consciousness and implicit knowledge influences the collection, analysis, and
interpretation of metamemory data.
(2) The relationship of memory performance and metamemory remains un-
clear, both theoretically and empirically. Based on a statistical meta-analysis of
approximately 50 studies, Schneider (1985) concluded:

In sum, the meta-analytical findings confirmed the assumption that different pat-
terns of correlations can be found for different classes of memory tasks and strat-
10 INTRODUCTION AND OVERVIEW

egies, and that developmental trends are demonstrable for the memory-metamemo-
ry relationship within each paradigm. On the other hand, most of the numerical
values obtained for the different relationships were surprisingly high. A mean
correlation of .41 between metamemory and memory behavior-performance based
on the data of several hundred subjects clearly contradicts the conclusion conveyed
in most reviews of the field that only a weak link between the two variables had
been found. Rather the quantitative integration of the empirical findings does
indicate that metamemory is substantially related to memory behavior and perfor-
mance. (p. 98)

In fact, the average relationship between memory performance and metamemory

reported by Schneider does seem surprisingly high, and one should not overlook
the considerable variation among the results from different studies. More surpris-
ing, however, is that many developmental psychologists seem to expect or hope
for such strong relationships between memory and metamemory. For one thing,
we know that task difficulty may determine whether we should expect a relation
between memory and metamemory at all. On extremely difficult tasks, the use of
metamemory may not lead to a solution, but may lead only to the realization that
further effort will not be productive. Conversely, on easy tasks, metacognitive
differences do not contribute to performance variance. It is only for tasks of
medium difficulty, where strategic solutions are possible, that learning behavior
and performance can be positively influenced by metamemory skills. Thus, the
functional relation between metamemory and memory performance must be
task— and population—specific. Because the difficulty of any task changes with
general cognitive development, as well as with increasing expertise, in principle
it is not possible to expect a stable and general pattern of correlations between
indicators of metamemory and memory performance.
(3) Metacognition is often regarded as referring to general and flexible knowl-
edge that is not domain specific. It is apparent that in research where relatively
easy and content-general tasks are used (e.g., reading and learning from simple
texts, memorizing and free recall of word lists, and so on), both the effects of
domain-specific knowledge and the role of content-dependent strategies and
metastrategies for learning will be minimized (Chi, this volume: Glaser & Pel-
legrino, this volume; Greeno & Riley, this volume). Although resolving the
problem of the role of domain specific content in the assessment of knowledge is
crucial for planning empirical studies and developing adequate models of meta-
cognitive knowledge, the failure to do so has done little to detract from the
popularity of the metacognitive research domain. However, tackling the problem
is especially important for studies in the field of developmental psychology
because substantial changes in knowledge, abilities, and executive skills occur
that, in comparison to simple autoregulations, allow higher forms of self-reg-
ulatory, goal-adaptive actions and thoughts (Körkel, 1984).
 METACOGNITION AND MOTIVATION 11

Is There a Relationship Between Metacognition and

Motivation Based on Their Effects on Learning
Behavior and Learning Performance?

The distinctions between knowledge about the self and self-related beliefs, be-
tween predictions of one’s future performance and related hopes and fears, and
between observations of one’s own behavior and related, ongoing self-evaluation
are difficult to make, both in everyday experience and in psychological investi-
gations. An attempt to integrate motivation and metacognition means one must
relate theoretical concepts concerned with knowledge about the self, perfor-
mance expectations and monitoring of one’s own actions as perceived in the
metacognition literature with concepts such as self-perception of ability, expecta-
tion of success and fear of failure, causal attributions for success and failure, and
processes of self-evaluation, from the motivation research domain. These two
research traditions, metacognition and motivation, have as yet been largely inde-
pendent, with little common ground. Nonetheless, there is some apparent over-
lap in the variables the two research traditions consider and in how these vari-
ables are operationalized, as suggested in Table 1. The points of similarity in the
phenomena studied, questions asked, and variables considered have not been
completely ignored. To illustrate: Brown has frequently noted that close rela-
tionship between metacognition and motivation; the relationship between meta-
cognitive judgment and causal attributions for success and failure has been
studied (Fabricius & Hagen, 1984; Kurtz & Borkowski, 1984); metacognitive
and motivational determinants of memory performance have been analyzed
(Weinert, Knopf, & Barann, 1983); and it has been suggested that the concept of
metamemory be expanded into a scientific theory of naive models of memory
with motivational components (Wellman, 1983; see also Dixon & Hertzog, in
press).
Although it is noteworthy to point out the similarities, it is also important to
keep in mind the different perspectives of the two research traditions. For exam-
ple, a central point in the analysis of metacognition is the extent to which an
individual’s metacognitive judgments correspond to the actual state of affairs,
whether the judgments are about abilities, learning strategies, or task difficulty.
In contrast, researchers studying motivation are more concerned with individual
biases and subjective distortions in tasks (such as setting aspiration levels, ex-
plaining success or failure, evaluating outcomes, or assessing personal ability).
However, despite the differences in perspective, some very similar and valid
predictions concerning behavior and performance have been proposed by re-
searchers from each area (see Schneider, 1985, for metacognition; see Kuhl,
1983; Peterson & Seligman, this volume; Stipek & Weisz, 1981, for moti-
vation). Indeed, interesting new questions arise from a consideration of the
somewhat different findings from metacognitive and motivation research. For
12 INTRODUCTION AND OVERVIEW

TABLE 1.1
A Summary of Similar Variables and Their Different Functions in
Motivation and Metacognition Research
ee
Motivation Research Metacognition Research

1. Judgments of task difficulty as a function of one’s own experience

—Performance expectations —Performance predictions

—Setting aspiration levels —Action planning
—Effort allocations —Choice of appropriate strategies

2. Causal attributions for success and failure

—Self-concept of own ability —Knowledge of own cognitive

competence
—lIndividual attribution style —Metacognitive judgments about the de-
terminants of performance
—Predictions of future performance —Predictions of future performance

3. Evaluation of action outcomes

— Use of self-relevant evaluation criteria —Metacognitive knowledge about task and

person variables
—Subjective experience of success and —Evaluation of correct and incorrect
failure judgments
—Emotional reactions and their —Improvement of task-related metacogni-
consequences tive knowledge

example, during childhood the ability to process information accurately and to

use metacognitive knowledge improves greatly. Yet, at the same time there is an
increase in the asymmetry with which children make causal attributions for
success and failure, an asymmetry that can be detrimental to self-esteem.
Heckhausen (this volume) asks:

Why do a good many people seem intent on preserving the depressive image they
have of self and own abilities and do not opt for more sanguine expectations, even
when good opportunities are offered? (p. 178)

Perhaps it is possible to identify those people who have both good metacognitive
skills and a characteristically ‘‘helpless’’ attributional style. If so, one could ask
how such a combination affects behavior and achievement in various learning
tasks, and whether attributional biases can be directly influenced by metacog-
nitive knowledge. Another question is whether the frequently observed but unre-
 METACOGNITION AND MOTIVATION 13

alistically pessimistic appraisals older people make about their memories may be
given a motivational explanation (Weinert, sun. & Barann, 1983). One could
also ask whether such an underestimation of one’s abilities affects learning and
remembering.
These questions illustrate the importance of coordinating studies of the cog-
nitive, metacognitive, and motivational determinants of learning behavior and
performance. Additionally, as Kuhl (1982) has noted:

To the extent that emotions are involved in actions, cumulative motivational effects
must be taken into account. . . . This assertion does not exclude the special case of
a purely rational or reflexively driven action, in which the actor chooses among
alternative actions on the basis of specific information in the situation. However, it
does suggest that the generalization of this special case to cover a general theory of
human action, a generalization made in the dominant models of motivation, is not
warranted. (p. 22)

METACOGNITION, MOTIVATION, AND LEARNING:

PROGRESS IN THE INTEGRATION OF DIFFERENT
RESEARCH PARADIGMS?

Books that include contributions from different research traditions often imply
that there is thematic and theoretical convergence, although with closer examina-
tion one sees that variety and diversity predominate. An introduction to such a
volume may strengthen this perception because similarities rather than dif-
ferences are emphasized. Perhaps that is the case in this introductory chapter as
well, although such was not the intention.
There are good reasons to establish connections between research efforts in
metacognition, achievement motivation, and learning, areas that have been rela-
tively isolated from one another. For example, if we attempt to explain indi-
vidual differences in learning and memory as more than the simple consequences
of stable ability differences (which proved impossible to do anyway), the ques-
tions we ask about the relation between learning activities and learning out-
comes, or about the process and conditions of learning, are given a new mean-
ing. Cognition, metacognition, procedural skills, and motivational factors are
important determinants of learning activity, but they must be differentially
weighted depending on task types. An integration of the different research ap-
proaches is also necessary if one wants to study and measure phenomena of
learning by doing, and doing in order to learn.
The difficulties encountered in attempting to combine different research ori-
entations should not, however, be underestimated. For instance, the use of
different prototypical tasks in memory, metamemory, and motivational research
leads inevitably to different conclusions regarding the relative importance of
domain-specific knowledge, procedural skills, and motivational factors in pre-
14 INTRODUCTION AND OVERVIEW

dicting or explaining individual learning and performance differences. This prob-

lem is compounded by the lack of a theoretically grounded taxonomy for learning
and memory tasks. Additionally, the role of motivational processes in learning
has seldom been considered because there have been no systematic efforts to
study how people with specific motivational characteristics behave when given
different types of tasks in situations that the subjects consider to be important.
Keeping these cautions in mind, however, there is sufficient overlap in the
tasks and concerns of the fields of cognition, metacognition, and motivation to
urge the development of an integrated research program.

REFERENCES
Bandura, A. (1981). Self-referent thought: A developmental analysis of self efficacy. In J. H.
Flavell, & C. Ross (Eds.), Social cognitive development (pp. 200—239.) Cambridge University
Press.
Bartlett, F. C. (1932). Remembering: A study in experimental and social psychology. Cambridge:
Cambridge University Press.
Bugelski, B. R. (1956). The psychology of learning. New York: Holt.
Carver, C. S., & Scheier, M. F. (1981). Attention and self-regulation: A control-theory approach
to human behavior. New York: Springer.
Cavanaugh, J. C., & Perlmutter, M. (1982). Metamemory: A critical examination. Child Develop-
ment, 53, 11-28.
Chi, M. T. H. (1978). Knowledge structure and memory development. In R. S. Siegler (Ed.),
Children’s thinking: What develops? (pp. 73-96). Hillsdale, NJ: Lawrence Erlbaum Associates.
Dixon, R. A., & Hertzog, C. (in press). A functional approach to memory and metamemory
development in adulthood. In F. E. Weinert and M. Perlmutter (Eds.), Memory development:
Universal changes and individual differences. Hillsdale, NJ: Lawrence Erlbaum Associates.
Ebbinghaus, H. (1885). Uber das Gedächtnis. Leipzig: Duncker & Humblot.
Eysenck, H. J. (1957). The dynamics of anxiety and hysteria. London: Routledge & Kegan Paul.
Fabricius, W. V., & Hagen, J. W. (1984). The use of causal attributions about recall performance
to assess metamemory and predict memory behavior in young children. Developmental Psychol-
ogy, 20, 975-987.
Flavell, J. H. (1970). Developmental studies of mediated memory. In H. W. Reese, & L. P. Lipsitt
(Eds.), Advances in child development and behavior (Vol. 5, pp. 181-211). New York: Academ-
ic Press.
Glaser, R. (1972). The new aptitudes. Educational Researcher, 1, 5-13.
Guilford, J. P., & Dallenbach, K. M. (1925). The determination of memory span by the method of
constant stimuli. American Journal of Psychology, 36, 621—628.
Hacker, W., Volpert, W., & von Cranach, M. (Eds.). (1982). Kognitive und motivationale Aspekte
der Handlung. Bern: Huber.
Heckhausen, H. (1982): The development of achievement. In W. W. Hartup (Ed.), Review of Child
Development Research (Vol. 6, pp. 600-668). Chicago: The University of Chicago Press.
Hull, C. L. (1945). The place of innate individual and species differences in a natural-science
theory of behavior. Psychological Review, 52, 55-60.
Körkel, J. 1984. Die Entwicklung von Gedächtnis- und Metagedächtnisleistungen in Abhängigkeit
von bereichsspezifischen Vorkenntnissen. Unpublished doctoral dissertation, Heidelberg.
Krampen, G. (1982). Differentialpsychologie der Kontrollüberzeugungen. Göttingen: Hogrefe.
 METACOGNITION AND MOTIVATION 15

Kuhl, J. (1982). Emotion, Kognition und Motivation. I. Auf dem Weg zu einer systemorienti
erten
Betrachtung der Emotionsgenese. Sprache und Kognition /, 1-27.
Kuhl, J. (1983). Leistungsmotivation: Neue Entwicklungen aus modelltheoretischer Sicht. In
H.
Thomae (Ed.), Psychologie der Motive (pp. 229-289). (Enzyklopädie der Psychologie C IV,
2)
Göttingen: Hogrefe.
Kurtz, B. E., & Borkowski, F. G. (1984). Children’s metacognition: Exploring relations among
knowledge, process, and motivational variables. Journal of Experimental Child Psychology, 37,
335-354.
McGeoch, J. A., & Irion, A. L. (1952). The psychology of human learning (2nd ed.). New York:
Longmans, Green.
McNemar, Q. (1964). Lost Our intelligence? Why? American Psychologist, 19, 871-882.
Meacham, J. A. (1983). Memory functions across the life course. Paper presented at the 7th
Meeting of the ISSBD, Munich.
Mischel, W. (1981). Metacognition and the rules of delay. In. J. H. Flavell, & L. Ross (Eds.),
Social Cognitive Development (pp. 240-271). Cambridge: Cambridge University Press.
Netschajeff, H. (1900). Experimentelle Untersuchungen iiber die Gedachtnisentwicklung bei
Schulkindern. Zeitschrift für Psychologie 24, 321-351.
Paris, S. G., & Lindauer, B. K. (1982). The development of cognitive skills during childhood. In
B. B. Wolman (Ed.), Handbook of developmental psychology (pp. 333-349). Englewood Cliffs,
NJ: Prentice Hall.
Pawlow, J. P. (1953). Die Lehre von dem Typen (1935). In Pawlow, J. P., Ausgewählte Werke,
Berlin, 235-256.
Piaget, J. (1971). Biology and knowledge. Chicago: University of Chicago Press.
Piaget, J. (1976). The grasp of consciousness. Cambridge, MA: Harvard University Press.
Piaget, J. (1978). Success and understanding. Cambridge, MA: Harvard University Press.
Plenderleith, M., & Postman, L. (1957). Individual differences in intentional and incidental learn-
ing. British Journal of Psychology, 48, 241-248.
Pressley, M., Forrest-Pressley, D., & Elliot-Faust, D. (in press). How to study strategy instruc-
tional enrichment: Illustrations from research on children’s prose memory and comprehension. In
F. E. Weinert, & M. Perlmutter (Eds.), Memory development: Universal changes and individual
differences. Hillsdale, NJ: Lawrence Erlbaum Associates.
Reese, H. W. (1979). Gedächtnisentwicklung im Verlauf des Lebens: Empirische Befunde und
theoretische Modelle. In L. Montada (Hrsg.), Brennpunkte der Entwicklungspsychologie. Stutt-
gart: Kohlhammer.
Robinson, F. P. (1941). Effective study. New York: Harper & Row.
Rubinstein, S. L. (1969). Prinzipien und Wege der Entwicklung der Psychologie. Berlin: Akademie
Verlag.
Schneider, W. (1985). Developmental trends in the metamemory-memory behavior relationship:
An integrative review. In D. L. Forrest-Pressley, G. E. McKinnon, & T. G. Waller (Eds.),
Metacognition, cognition, and human performance (pp. 57-109). New York: Academic Press.
Spearman, C. (1927). The abilities of man. New York: Macmillan.
Stipek, D. J., & Weisz, J. R. (1981). Perceived personal control and academic achievement.
Review of Educational Research, 51, 101.138.
Thurstone, L. L. (1930). The relation between learning time and length of task. Psychological
Review, 37, 44-53.
Trudewind, C. (1982). The development of achievement motivation and individual differences.
Ecological determinants. In W. W. Hartup (Ed.), Review of Child Development Research (Vol.
6, pp. 669-703). Chicago: The University of Chicago Press.
Vaughan, E. D. (1977). Misconceptions about psychology among introductory psychology students.
Teaching of Psychology, 4, 138-141.
16 INTRODUCTION AND OVERVIEW

s. In L. Mon-
Weinert, F. E. (1979). Entwicklungsabhangigkeit des Lernens und des Gedächtnisse
der Entwicklung spsychologi e (pp. 61-76). Stuttgart: Kohlhammer.
tada (Ed.), Brennpunkte
M., & Barann, G. (1983). Metakognition und Motivation als Determi-
Weinert, F. E., Knopf,
2, 71-87.
nanten der Gedächtnisleistungen im höheren Erwachsenenalter. Sprache & Kognition,
Zusammenhänge zwischen Metawissen, Ver-
Weinert, F. E., Knopf, M., & Körkel, J. (1983).
Er-
halten und Leistung bei der Lösung von Gedächtnisaufgaben durch Kinder und ältere
für
wachsene. In G. Lüer (Ed.), Bericht über den 33. Kongress der Deutschen Gesellschaft
Psychologie in Mainz 1982 (Vol. 2, pp. 262-271). Göttingen: Hogrefe.
Wellman, H. M. (1983). Metamemory revisited. Contribution to Human Development, 9, 31-51.
Woodrow, H. (1942). The problem of general quantitative laws in psychology. Psychological
Bulletin, 39, 1-27.
Wundt, W. (1886). Über den Begriff des Gesetzes mit Rücksicht auf die Frage der Aus-
nahmslosigkeit der Lautgesetze. In Kleine Schriften (Vol. 3). Stuttgart: Enke, 1921.
Zeigarnik, B. (1927). Das Behalten erledigter und unerledigter Handlungen. Psychologische For-
schung, 9, 1-85.
 METACOGNITION

During the last 15 years metacognition has become one

of the major fields of cognitive developmental research.
Research activity in metacognition began with John Flavell.
His work on metamemory provided the first results on the
developmental course of metamemory and raised important
questions for subsequent research. Because of Flavell’s
efforts, a considerable amount of empirical and theoretical
research dealing with metacognition can be registered. Psy-
chological research in metacognition is concerned with two
problems: The first is the availability of factual (declarative)
knowledge about cognition. It is similar to other world
knowledge that individuals store but focuses on general
knowledge about cognitive activities and demands, as well
as the features of one’s own cognitive activities and ca-
pabilities. The second problem is the availability of strat-
egies (procedural knowledge) that may be applied in order
to control (monitor) and regulate cognitive activity (think-
ing about thinking).
Developmental research in metacognition thus far has
emphasized the acquisition and application of knowledge
about cognition. Central questions must be answered in
order to secure the knowledge and the use of the knowledge
necessary to organize and regulate the flow of own think-
ing. The procedural aspects of metacognition have gained
more attention during testing. The chapters by Brown,

17
18 PART I: METACOGNITION

Campione, Flavell, and Kluwe provide an overview of the various research

directions in the field of metacognition. Flavell, the pioneer in the field, dis-
cusses the present state of metacognition research. His analysis is based on a
taxonomy of knowledge components that are important in metacognitive ac-
tivity. He assumes that metacognitive knowledge has its own course of develop-
ment, and introduces the interesting, although not yet precise, concept of meta-
cognitive experiences. If these conditions exist during cognitive endeavors, they
may trigger metacognitive activity. Flavell’s discussion of the problems empha-
sizes the role of instruction in schools as one possible major source for the
acquisition of metacognitive knowledge. His final evaluation of the present state
of developmental research in metacognition, although connected with the hope
for broader models of metacognition and for a more precise conceptualization,
remains skeptical.
The chapter by Kluwe includes an approach for specifying the procedural
aspects of metacognition. Within the framework of Flavell’s metacognitive strat-
egies and Brown’s metacognitive skills, Kluwe focuses essentially on executive
decisions. He refers to production-like rules that, given certain conditions, may
be evoked during problem solving. These rules are directed at the control and
regulation of the solution search. A set of hypothetical executive decisions are
discussed using developmental aspects. It is assumed that during cognitive devel-
opment the knowledge necessary for the formation and application of such exec-
utive rules is acquired. The availability of executive procedural knowledge al-
lows an appropriate regulation of the own cognitive efforts to certain given
demands. In addition, given changing task demands, executive decisions allow
for flexible use of own thinking. Preliminary data, from a developmental study
where children of different ages were required to cope with changing problem
solving conditions, are presented. The significant result of this study is the
awareness that 4-year-old children know when an originally selected approach to
problem solving is no longer adequate, and are able to effectively regulate the
own solution search. However, the relation between such regulatory activity and
level of performance remains an open question.
One of the most comprehensive reviews of metacognition and related fields of
research is found in the chapter by Brown. A discussion of the concept of
metacognition is followed by a review of the four central issues of metacognition
research: reports on own thinking; the role of executive components in informa-
tion processing systems; processes of self-regulation; and processes of an indi-
vidual’s regulation by other persons. Similar to Kluwe, Brown’s discussion
connects cognitive psychological and developmental aspects. Brown emphasizes
the function of success for the acquisition of self-regulatory mechanisms, and the
role of parents and teachers in the formation of cognitive mechanisms that help
control and regulate own thinking.
The chapter by Campione is directly connected to Brown’s contribution con-
cerning the regulation of a child’s thinking by other persons. Campione provides
 PART I: METACOGNITION 19

an Overview of the theoretical and empirical research concerned with the instruc-
tional aspects of metacognition. He focuses on the connection of research on
metacognition with instructional applications and demonstrates that previous
attempts to train and improve children’s cognitive abilities have failed to include
self-regulatory aspects. After considering the developmental results of metacog-
nition research, Campione proposes a modification of training approaches. The
essential feature of this approach is the consideration of the learning and thinking
child in intervention studies. The cognitive resources of the child, the instruction
of strategies connected with insights into the effects of a strategy, and the
instruction of monitoring operations, together result in more successful training
effects. Especially important are improvements in the transfer and persistence of
trained strategies. This type of research demonstrates the importance of metacog-
nitive research to instruction and education.
 u

2 8

“x; u «mr ~ Aue, oS

iu re |hig"
ny OT ars ee voten Aamo $
nah nap eee rs
eu u rn A or a ee
zu al nn ' eee ae ra 5 :
jr ab > Susann tetas
N er ei er pa ok |
are MN <n, Ch any, NT Tag
a eT naar Crees) ise!
040 4 faim line of Cpe =.
tiie Br ona
‘carpet. ==
> 4 ee 8 le

= Oe
é 5 De a u 2
ves >

10%

—~
[5
-y
pat
v
*
fe
nn
~ =

a _ _

— = ©

u „ =
DZ ° ®
ee
Vo
ee
eee
eee
ae
iii oe
US Tu

. m) Ge = omnes

A DAN
Tu

7, En

a ER HRIEEE Gree ‘
a> 7 ¥.> 2, eRe pie
ee he) pat inate a:

(ia Ts
2
Pe | :
25
 Speculations About the Nature
and Development of
Metacognition

John H. Flavell
Stanford University

Metacognition is usually defined as knowledge and cognition about cognitive

objects, that is, about anything cognitive. However, the concept could reasona-
bly be broadened to include anything psychological, rather than just anything
cognitive. For instance, if one has knowledge or cognition about one’s own or
someone else’s emotions or motives, it could be considered metacognitive. Any
kind of monitoring might also be considered a form of metacognition; for exam-
ple, attempts to monitor one’s own motor activity in a motor skill situation. It is
very difficult to rule out the conscious monitoring of motor acts as being some-
thing entirely different from what is ordinarily understood to be metacognition.
Because some metacognitive knowledge and cognitive self-regulatory activity is
not very accessible to consciousness, researchers may eventually feel compelled
to include processes that are not conscious and perhaps not even accessible to
consciousness as forms of metacognition or metacognitive-like phenomena.
This researcher (Flavell, 1979, 1981) tried to classify part of the domain of
metacognition. The taxonomy created is not very satisfactory, but at least it helps
in thinking about the domain. The key concepts in the taxonomy are metacog-
nitive knowledge and metacognitive experience. Metacognitive knowledge refers
to the part of one’s acquired world knowledge that has to do with cognitive (or
perhaps better, psychological) matters. As people grow up, an important part of
what they learn or come to believe concerns the mind and other things
psychological.

21
 22 FLAVELL

METACOGNITIVE KNOWLEDGE

Metacognitive knowledge is conceived as simply that portion of the total knowl-

edge base that pertains to this content area. Metacognitive knowledge can be
subdivided into three categories: knowledge of person variables; task variables;
and strategy variables.

Person Variables

Knowledge of person variables refers to the kind of acquired knowledge and

beliefs that concern what human beings are like as cognitive (affective, moti-
vational, perceptual, etc.) organisms. There are three subcategories of person
variables: intraindividual; interindividual; and universal. An example of an in-
traindividual variable is a person’s belief that he or she is fairly good at dealing
with verbal kinds of material, but poor at spatial tasks; therefore, it is knowledge
or belief about intraindividual variation in one’s own or someone else’s interests,
propensities, aptitudes, and the like. In the case of interindividual variables, the
comparison is between, rather than within, persons. Examples might be the
judgment that one is brighter than one’s parents, but that the parents are more
reflective and thoughtful than certain of their friends. The most interesting vari-
ables are acquired ideas about universal aspects of human cognition or psychol-
ogy. It is hard to imagine a culture in which people grow up without acquiring
any naive psychology; in particular, without developing any intuitions about the
way the human mind works. For example, one cannot imagine anyone growing
to maturity without having some kind of intuition that short term memory is
fallible and of limited capacity. That is something adults know, but did not
always know, about themselves: it must have been acquired in the course of
childhood development. Similarly, can one imagine a culture in which a person
has not acquired the concept of a mistake or error? Something is believed to be
true and it turns out not to be true. You think you understand something and it
turns out that you have misunderstood it or failed to understand it. Surely, adults
the world over have acquired knowledge of such universal mental phenomena
and make use of that knowledge in managing their lives.

Task Variables
The second subcategory is knowledge of task variables. The individual learns
something about how the nature of the information encountered affects
and
constrains how one should deal with it. For instance, experience has
taught that
very difficult, very densely packed, and very low redundancy
information is
troublesome to process. To comprehend and to deal effectively with
such infor-
mation, it is necessary to proceed slowly and carefully and to
process deeply and
self-critically (i.e., with high comprehension monitoring activity).
Other inputs
 1. THE NATURE AND DEVELOPMENT OF METACOGNITION 23

are loosely packed and contain mostly familiar information. People know if they
can comprehend these inputs without paying really close attention, and they are
likely to process them accordingly. Thus, much is learned about the different
kinds of information that are encountered and about the kind of processing that
each kind of information requires or does not require. In addition, given the
information, it is learned that different kinds of tasks place different kinds of
information-processing demands on individuals. An example would be the
knowledge that it is easier to learn the essence or gist of something, such as a
story, than it is to learn it verbatim. All adults recognize that it is easier to recall
the main events of a story than to recite the story word for word. Thus, one learns
about the implications of various task demands for self—processing. One learns
that in some cases the task demands are much more rigorous and difficult than
they are in other cases, and that one must take these demands into account and
act accordingly if the task goal is to be achieved.

Strategy Variables
Much is also learned about cognitive strategies or procedures for getting from
here to there in order to achieve various goals (strategy variables). It has been
suggested (Flavell, 1981) that one can distinguish cognitive strategies from meta-
cognitive strategies. A cognitive strategy is one designed simply to get the
individual to some cognitive goal or subgoal. For instance, a cognitive strategy
for getting the sum of a list of numbers would obviously be to add them up. The
goal is to find the sum, and in order to do so the numbers are added. In the same
situation, a metacognitive strategy might be to add the numbers a second time to
be sure the answer is right. If it is an income tax return or something equally
important, one might even double check by adding them up a third time. The
purpose of the second and third addition is somewhat different from that of the
first. The purpose is no longer to reach the goal (cognitive strategy), but rather to
feel absolutely confident that it has been reached (metacognitive strategy). Simi-
larly, sometimes one reads things slowly simply to learn the content (cognitive
strategy); other times one reads through things quickly to get an idea of how
difficult or easy it is going to be to learn the content (metacognitive strategy).
That is, one skims or scans a text fn order to get some idea of how much work
lies ahead. In the course of development one learns about cognitive strategies for
making cognitive progress and about metacognitive strategies for monitoring the
cognitive progress.
Finally, it should be emphasized that person, task, and strategy variables
always interact, and that intuitions about their interaction are also acquired. For
instance, I may sense that I but not my brother would do better to use strategy A
rather than strategy B, because the task is of this type rather than that. Given
one’s particular cognitive make-up and the particular task, one develops intui-
tions about which strategies are better.
 24 FLAVELL

METACOGNITIVE EXPERIENCES

The other major conceptual entity in the taxonomy is metacognitive experiences.

Metacognitive experiences are conscious experiences that are cognitive and af-
fective. What makes them metacognitive experiences rather than experiences of
another kind is that they have to do with some cognitive endeavor or enterprise,
most frequently a current, ongoing one. For example, if one suddenly has the
anxious feeling that one is not understanding something and wants and needs to
understand it, that feeling would be a metacognitive experience. One is having a
metacognitive experience whenever one has the feeling that something is hard to
perceive, comprehend, remember, or solve; if there is the feeling that one is far
from the cognitive goal; if the feeling exists that one is, in fact, just about to
reach the cognitive goal; or if one has the sense that the material is getting easier
or more difficult than it was a moment ago. Thus, a metacognitive experience
can be any kind of effective or cognitive conscious experience that is pertinent to
the conduct of intellectual life; often, it is pertinent to conduct in an ongoing
Cognitive situation or enterprise. Metacognitive experiences play a very impor-
tant role in everyday cognitive lives. As one grows older one learns how to
interpret and respond appropriately to these experiences. The converse implica-
tion is that young children may have such conscious experiences, but may not
know how to interpret them very well; children simply may not know what these
experiences mean and imply. Recent research supports this implication (Beal &
Flavell, 1982; Flavell, Speer, Green & August, 1981; Singer & Flavell, 1981:
see also various chapters in Dickson, 1981).
In one series of studies (Flavell et al., 1981), the young child subject hears a
brief tape-recorded instruction to build a simple block structure; for example,
“Take the red block and put it on top of the blue block.’’ His task is to make a
structure identical to that of the child speaker, based on her instructions. Some of
the instructions are wholly unambiguous and the child can follow them without
any difficulty. Other instructions are ambiguous, impossible to execute, or other-
wise inadequate. For example: ‘‘Put the big block on the tray. Put the little block
on top of it so you cannot see the big block.’’ Or: ‘‘Put the red block on the
tray,’ when there are two different red blocks to choose between. It has been
found that when kindergarten children (5 or 6 years of age) hear these inadequate
instructions and try to carry them out they often act puzzled or uncertain and
may
even say something like ‘‘Huh?’’ or ‘‘Which red block?’’ However, when
they
are asked if they think that their building looks exactly like that of the speaker’s
they are very likely to say that it does; and when they are next
asked if the
speaker did a good job or a bad job of telling them how to make
their building
exactly like hers, they are even more likely to say that she did a
good job. The
evidence suggests that these curious responses are not artefacts of the
method of
questioning used; rather, it seems that young children often do
not fully under-
stand the meaning and implications of metacognitive experiences
of puzzlement
 1. THE NATURE AND DEVELOPMENT OF METACOGNITION 25

and uncertainty. They may feel puzzled, but they do not know what the implica-
tion of that feeling is for the existing situation. These children cannot be sure
their building matches the speaker’s because the speaker did not describe it
adequately enough. In conclusion, the evidence from these studies and those of
other investigators suggests that young children have more trouble than older
children in properly comprehending their own feelings of incomprehension, and
in properly appreciating the meaning, significance, and implications of such
metacognitive experiences.

QUESTIONS, PROBLEMS, AND ISSUES

The following are some of the questions, problems, and issues concerning meta-
cognition that should be addressed. First, where does metacognition fit in psy-
chological space? That is, what other psychological concepts does it relate to,
and how does it relate to these concepts? The concepts that might be related to
metacognition include: executive processes; formal operations; consciousness;
social cognition; self-efficacy, self-regulation; reflective self-awareness; and the
concept of psychological self or psychological subject. Also related are develop-
ing conceptions of, and about thinking, learning, and other cognitive pro-
cesses—the child’s emerging ‘‘theory of mind.’
Another group of questions concerns the types of metacognitive acquisitions
that develop and the earlier foundations or prerequisites from which they develop
(the taxonomy given will probably not prove to be the best one). Similarly: What
aspects of metacognition are inherent or very early acquired, and what aspects
must be acquired in the course of childhood, adolescence, or even during the
adult years? Some aspects of metacognition, just like some aspects of general
cognition, are probably present almost from the beginning. Also, is the acquisi-
tion, use, and usefulness of some types of metacognition impeded by informa-
tion-processing limitations or biases, by lack of relevant experiences in most
environments, or by other factors? For example, perhaps some types of metacog-
nition are unlikely to occur in a given cognitive domain until some amount of
expertise or knowledge in that domain is acquired.
How might various types of metacognition develop? What might account for
possible individual or cultural differences in what does develop? A rough distinc-
tion can be made between cognitive-developmental changes in the child that
allow for metacognitive acquisitions and experiences the child might have that
could assist or facilitate metacognitive development. In the case of the former,
there might be three closely related but conceptually distinguishable types of
changes. First, there might be cognitive-developmental changes that could lead
directly to metacognitive acquisitions. For example: An increase in the capacity
to plan ahead could lead, more or less directly, to a greater tendency to think
about cognitive means and cognitive goals. Second, there could be changes
 26 FLAVELL

which increase the child’s cognitive readiness to profit from experiences that
promote metacognitive development. Third, and closely related to the second,
one can imagine cognitive-developmental changes in the child that might in-
crease the child’s opportunity to have experiences that could lead to metacog-
nitive acquisitions. For example: Once one has developed sufficiently to start
reading, one can start having the formative experiences that reading brings.
These include the metacognitive experiences of conscious comprehension diffi-
culties, misreadings, and sudden insights; thus, any experiences that can promote
metacognitive growth.
Two changes in the development of the child might possibly contribute to the
acquisition of metacognition. One is the developing sense of the self as an active
cognitive agent and as the causal center of one’s own cognitive activity. The
development of such an internal locus of cognitive control could promote the
monitoring and regulation of one’s own cognitive enterprises. A second kind of
change that should facilitate metacognitive development, an increase in plan-
fulness, has already been mentioned. More generally, an individual that repre-
sents and interrelates past, present, and future actions and events should be in a
good position to acquire metacognitive knowledge. That is, such an individual
could notice and store covariations in person, task, and strategy factors. The
person who can look ahead is also in a position to scan upcoming information or
impending problems, and can plan in advance how processing resources should
be allocated. Thus, an individual who can create conscious and explicit represen-
tations of the past, present, and the future should be in a better position to make
metacognitive progress than one who does not.
There may also be a number of experiences that might assist metacognitive
development; some of them may consist of direct practice in metacognitive
activity. Metacognition, like everything else, undoubtedly improves with prac-
tice. Other kinds of experiences, although not themselves metacognitive ac-
tivities, may simply be heuristic or propaedeutic to metacognitive development.
Therefore, one way to become better at metacognition is to practice it; another
way may be to practice other things which are not metacognitive themselves
but
which indirectly promote metacognitive activity.
One class of these experiences may be supplied by parents (Wertsch
, 1978).
Parents may unintentionally model metacognitive activity for their
young chil-
dren. They may also deliberately demonstrate and teach it, helping
the child to
regulate and monitor his or her actions. Similarly, teachers
in schools may
sometimes model, as well as teach and encourage, metacogn
itive activity.
Schallert and Kleiman (1979) describe some of the things
they have observed
teachers doing to help the child regulate and monitor
own cognition. They
indicate how teachers provide the kind of assistance, not provided
in textbooks,
to help the child wend his or her way through a cognitiv
e endeavor.
There are a variety of other school experiences that may
assist the growth of
metacognitive skills, including reading, which was mentio
ned earlier. A piece of
 1. THE NATURE AND DEVELOPMENT OF METACOGNITION 27

text is very different from an oral communicative interchange between people;

the text has little communicative context (Donaldson, 1978; Olson, 1972). The
child has to treat a written passage as a cognitive object and attempt to figure out
what can be concluded or infered from it alone, without the additional expressive
and situational clues to meaning normally present in an oral communication
situation. In the course of learning to read, the child gets practice in scrutinizing
messages in isolation from context, and thus in evaluating the possible intended
meanings and implications (a form of metacognition). Writing also affords prac-
tice and experience in metacognition. It allows one to critically inspect one’s
own thoughts. It also encourages the individual to imagine the thoughts of
others. For example, to think about whether they will understand or believe what
one is trying to convey in one’s writing. Of course, learning to be a skillful
speaker and critical listener also involves considerable practice and skill in cog-
nitive monitoring. Similarly, learning mathematics provides opportunities for
monitoring all sorts of activities. This was shown earlier, in the example of
checking one’s addition by adding the column a second time.
Good schools should be hotbeds of metacognitive development, for the banal-
sounding reason that so much self-conscious learning goes on in them. In school,
children have repeated opportunities to monitor and regulate their cognition, as
they gradually pass from novice status to semi-expert status in microdomain after
microdomain. They have innumerable metacognitive experiences and innumera-
ble opportunities to acquire person, task, and strategy metacognitive knowledge.
There are still other interesting questions involving metacognition. What is
the particular usefulness and adaptiveness of metacognition? How and why has
metacognition evolved? Metacognition is especially useful for a particular kind
of organism, one that has the following properties. First, the organism should
obviously tend to think a lot; by definition, an abundance of metacognition
presupposes an abundance of cognition. Second, the organism’s thinking should
be fallible and error-prone, and thus in need of careful monitoring and regula-
tion. Third, the organism should want to communicate, explain, and justify its
thinking to other organisms as well as to itself; these activities clearly require
metacognition. Fourth, in order to survive and prosper, the organism should need
to plan ahead and critically evaluate alternative plans. Fifth, if it has to make
weighty, carefully considered decisions, the organism will require metacognitive
skills. Finally, it should have a need or proclivity for inferring and explaining
psychological events in itself and others, a penchant for engaging in those meta-
cognitive acts termed social cognition. Needless to say, human beings are orga-
nisms with just these properties.
An important future endeavor is to try to create detailed process models for
various aspects of metacognition. Process models will have to address questions
such as: How is the information about cognitive processes that is needed to
monitor and regulate these processes obtained? What cues are observed? One cue
that might be used is the speed of processing, especially the processing of sudden
 28 FLAVELL

changes in speed. When you are reading along and suddenly find yourself read-
ing more slowly, the slowdown in processing may function as a cue that the
material is getting difficult, or that something is puzzling, etc. Similarly, indi-
viduals may become aware that they have just read a sentence for the second
time, and that awareness may serve as a metacognitive signal that the material is
difficult, or that attention has wandered, etc. There is also the question of how
monitoring information gets translated into self-regulatory metacognition.
The final question is: When are we most likely to have metacognitive experi-
ences? First, they are obviously apt to occur whenever the situation explicitly
demands or elicits them. For example, someone is asked to justify a conclusion,
or defend a claim. Second, metacognitive experiences may be more apt to occur
when the cognitive situation is something between completely novel and com-
pletely familiar. In this broad range, one knows enough to be puzzled and to
formulate questions, but not enough that the processing is wholly automatic and
effortlessly accurate. Third, metacognitive experiences are likely to occur in
situations where it is important to make correct inferences, judgments, and
decisions. If it really matters whether or not one’s judgments and decisions are
correct, one is apt to monitor them very carefully. Fourth metacognitive antennae
are likely to go up whenever one’s cognitive enterprise seems to be in any sort of
trouble. There is nothing like the sudden awareness of self-contradiction or some
other mental cul-de-sac to cause critical analysis of one’s own thinking. And
finally, one is more likely to have metacognitive experiences (useful ones, at
least) when attentional and mnemonic resources are not wholly preempted by
more urgent subjective experiences, such as pain, anxiety, or depression.
What will the future bring to the area of metacognition? During the next few
years more careful and critical examinations of metacognition and related con-
cepts will probably occur. Undoubtedly, the concept itself will be further re-
fined, clarified, and differentiated. Some methodological advances, better
ways
to measure and assess metacognitive experiences and knowledge than is present-
ly available, should also develop. Finally, deeper insights into the entire
concept
are needed. A number of psychologists have the abiding intutition that metacog-
nition is an extremely important topic, eminently worthy of further theoretical
and experimental investigation. However, none of us has yet come
up with
deeply insightful, detailed proposals about what metacognition is, how
it oper-
ates, and how it develops. Perhaps the future will bring such proposals.

ACKNOWLEDGMENTS
I thank Ellen Markman and the members of a seminar
on metacognitive development she
and I co-taught for their contributions to the questions
and issues raised in the latter half of
this chapter,
 1. THE NATURE AND DEVELOPMENT OF METACOGNITION 29

REFERENCES

Beal, C. R., & Flavell, J. H. (1982). The effects of increasing the salience of message ambiguities
on kindergartners’ evaluations of communicative success. Developmental Psychology, 18, 43—
48.
Dickson, W. P. (Ed.). (1981). Children’s oral communication skills. New York: Academic Press.
Donaldson, M. (1978). Children’s minds. New York: Norton.
Flavell, J. H. (1979). Metacognition and cognitive monitoring: A new area of cognitive-develop-
mental inquiry. American Psychologist, 34, 906-911.
Flavell, J. H. (1981). Cognitive monitoring. In W. P. Dickson (Ed.), Children’s oral communica-
tion skills (pp. 35-60). New York: Academic Press.
Flavell, J. H., Speer, J. R., Green, F. L., & August, D. L. (1981). The development of comprehen-
sion monitoring and knowledge about communication. Monographs of the Society for Research
in Child Development, 46(5, Serial No. 192).
Olson, D. R. (1972). Language use for communicating, instructing, and thinking. In R. O. Freedle
& J. B. Carroll (Eds.), Language comprehension and the acquisition of knowledge (pp. 139-
167). Washington, DC: Winston.
Schallert, D. L., & Kleiman, G. M. (1979). Some reasons why the teacher is easier to understand
than the textbook. Reading Education Report Series, University of Illinois, Center for the Study
of Reading.
Singer, J. B., & Flavell, J. H. (1981). Development of knowledge about communication: Chil-
dren’s evaluations of explicitly ambiguous messages. Child Development, 52, 1211-1215.
Wertsch, J. W. (1978). Adult-child interaction and the roots of metacognition. Quarterly Newslet-
ter of the Institute for Comparative Human Development, 1, 15-18.
nn Aa
.

anecha a: tea
Cae oes fie u
= \ = , Br Yu . ar Ka

RATEN
u
BA
Zu Dam du SR u

= ep owe “<4,
ee Pu
etn u N)
; are ic UNE
n y 1902 oe O19
(hook. eA ER A ED 2
> NR 4 Gat ee A ot eee
I Le

ya.» re cae
meee Toe)
oS eelDS.
 Executive Decisions and
Regulation of Problem Solving
Behavior

Rainer H. Kluwe
University of Bundeswehr, Hamburg

It is well known that Flavell and his co-workers (1970, 1971) initiated research
on metacognition. Since that time ‘‘metas,’’ (e.g., metalistening, metacom-
munication, metapersuasion, metacomponents, and the like) have proliferated in
the literature.
The concept of metacognition relates to a child’s declarative knowledge about
cognition, for example the own cognitive activities and abilities, and to pro-
cedural knowledge, processes directed at the control and regulation of one’s own
thinking, or “‘cognition about cognition’’ (see Kluwe, 1981, 1982). The avail-
able developmental results suggest that younger children lack knowledge about
cognition; in addition, they are less engaged in monitoring the course of own
thinking.
Flavell and Wellman (1977) published a taxonomy of the development of
metacognitive knowledge. This classification of knowledge was the first step
toward a more systematic view of metacognition and provided a starting point for
subsequent research. Metacognition refers mainly to memory functions. The
immense body of research, especially in developmental psychology, published
under the heading of metacognition, relates mainly to ‘‘metacognitive knowl-
edge.’’ That is, it refers to the acquisition of knowledge, the amount of knowl-
edge, and the assumptions and opinions about the states and activities of the
human mind. Such knowledge may refer to the human cognitive system in
general, as well as to an individual’s own cognitive system. There is consider-
able evidence for the development of metacognitive knowledge (Kreutzer,
Leonard & Flavell, 1975), but there are still many open questions. Researchers
do not know enough about how individuals acquire knowledge about their own

31
32 KLUWE

cognitive system or about the usefulness of such knowledge (Thornquist &

Wimmer, 1977; Wimmer & Thornquist, 1979).
Flavell (1981) offers a model of ‘‘cognitive monitoring,’’ which is considered
to be the first approach connecting cognitive and metacognitive activity. Of
central interest is the notion of metacognitive strategies that may be applied in
order to monitor cognitive progress. Another line of research, resulting in equal-
ly important contributions to the field, is the one offered by Brown (1978; Brown
& DeLoache, 1978). The work of Brown and her group has placed increasing
emphasis on the development of “‘metacognitive skills.’’ According to Brown,
metacognitive skills entail the operation of specific mental processes by which
individuals organizes and monitors their own thinking. It is plausible to focus on
the procedural aspects of metacognition, if one examines metacognition less
from the aspect of factual knowledge, but rather as one component of cognitive
activity.
Metacognition has also been receiving increased attention in cognitive psy-
chology (e.g., Sternberg, 1979). There is no doubt that metacognition has a long
history in cognitive psychology, especially in connection with assumptions con-
cerning the function of the executive component in information processing sys-
tems (Neisser, 1967; Reitman, 1965). As a result, it is possible to use the models
proposed by cognitive psychologists to discuss developmental data on
metacognition.
This chapter first outlines metacognitive activity by describing executive pro-
cesses, taking into account current results in developmental research and aspects
of cognitive performance. Second, the preliminary results from an experimental
study, that deals with children’s regulation of their own problem-solving behav-
ior under changing conditions are reported.

THE ROLE OF EXECUTIVE DECISIONS

Brown argues that cognitive activities, such as checking, monitoring, planning,

and prediction (Brown, 1978; Brown & DeLoache, 1978), contribute to cog-
nitive performance of the type investigated by developmental psychologists, and
that they are subject to developmental changes. Brown and DeLoache even claim
that these processes, referred to as metacognitive skills, are the ‘‘basic charac-
teristics of efficient thought, [and] that they are transsituational”’ (p. 15). Similar
cognitive activities are addressed, using the ““imperatives’’ proposed by Flavell
(1976) in his discussion of the improvement of children’s problem-solving
capacities.
Brown (1978) assigns metacognitive skills to the component of an informa-
tion processing system that is designed as an executive; the central processor,
or
the monitoring system. She argues (Campione & Brown, 1978) that the different
 2. EXECUTIVE DECISIONS 3

forms of executive decision making are at the core of intelligent problem solving,
and that “‘variations in the efficiency of the executive’’ may be associated with
differences in intelligence (p. 21). The attribution of metacognitive skills to the
executive component of information processing systems, is important in estab-
lishing the link between cognitive developmental research and cognitive psychol-
ogy. The following discussion concerns executive decisions and processes; the
term metacognition is not used.
Executive functions as described by Neisser (1967) have not been studied
extensively in cognitive psychology (see Ueckert, 1980). One reason for this
might be that research on problem solving was first aimed at the elaboration of a
general model of human problem solving processes. However, there is some
progress, which allows one to ask for procedural variants of problem solving
processes and for the hypothetical mechanisms producing such variants. Simon
(1975) provided the first thorough analysis of the differences between problem-
solving processes.
Studying the procedural attributes of different problem-solving processes is
important because some understanding of the flexibility of thinking processes
with respect to varying situational demands is gained. The central question
concerns the intraindividual variability in problem-solving processes. The execu-
tive requirements for simulation programs, considered to be models of human
problem solving processes, have so far been limited. There are no changing
situational demands similar to those human beings encounter in everyday situa-
tions. Attributes of situational demands that determine the organization and the
course of problem solving processes have been described by Dorner (1974).
These attributes not only indicate the diversity of demands a person has to cope
with; at the same time, make plausible the assumption that there is something
like a central regulatory component in information processing systems, an
executive.
Another aspect of the situation is that of interindividual differences in problem
solving (Simon, 1975). Recent research on differences between successful and
poor problem solvers (Dorner, Kreuzig, & Stäudel, 1978; Lüer, Putz-Osterloh &
Hesse, 1979), research on successful and poor learners (Thorndyke & Stasz,
1980), and analyses of differences between novices and experts in problem
solving (Simon & Simon, 1978) indicate that more research on executive deci-
sions in problem solving and learning is needed. In addition, there is some
evidence that an improvement of executive decision making would presumably
effect an improvement of cognitive performance. Preliminary insight gained
from the research cited and from other investigations (e.g., Sternberg, 1979)
indicates that individual differences in executive functions might contribute to
individual differences in intellectual performance. The work of Reither (1979),
Hesse (1979), and Dorner (1978/1984) on self-reflection and problem solving,
might be considered a move in the direction of the empirical testing of this
34 KLUWE

assumption. The effects on performance of the simple instruction, to reflect on

one’s own problem-solving activity, as it was used in these studies, are indeed
impressive.
In addition to the aspects discussed, there are two other phenomena that
underscore the relevance of the conception of executive decisions and the impor-
tance of the study of their development. First, developmental studies have shown
that the individual’s repertory of cognitive strategies increases during the course
of intellectual development. If this is the case, the developing information pro-
cessing system will increasingly need the ability to decide among alternative
procedures and to evaluate selected approaches; that is, the growing child has to
acquire the ability to make flexible and selective use of its own growing cog-
nitive knowledge. Second, effortful cognitive processes, unlike automated pro-
cesses (see Hasher & Zacks, 1979), may vary in the efficiency of their organiza-
tion and execution. It is assumed that such effortful processes grant flexibility in
problem solving. However, this implies that cognitive activity exists, which not
only can be regulated, but which must be monitored and regulated in order to
ensure its efficiency.
The resulting picture of human thinking, therefore, is not one of an always
predetermined sequence and mode of information processing. Instead, the indi-
vidual may: decide about cognitive processes; interrupt, accelerate, slow down
own thinking directed towards certain goals; increase the intensity of own think-
ing; or deliberately maintain or continue the specific course of thinking. It is,
therefore, assumed that executive decisions contribute to age-related variations
and to intra- and interindividual differences in thinking. Presumably, executive
decisions are necessary in order to efficiently apply the growing repertory of
cognitive operations and to organize and execute effortful processes efficiently.
Finally, in order to maintain cognitive activity, which is subject to emotionally
and motivationally determined variation, the human information processing sys-
tem needs executive decision.

A SET OF EXECUTIVE DECISIONS

Executive decisions deal with the course and the organization of own mental
activity, and therefore, are part of procedural knowledge. Following is a frame-
work for executive decisions. Executive decisions are considered to be pro-
cedural knowledge directed at the flow of own thinking. With regard to the
declarative component, it is assumed that cognitive knowledge is available.
Cognitive knowledge refers to stored facts, assumptions, and beliefs about think-
ing in general and about features of own cognitive endeavors (see Kluwe, 1982;
Kluwe & Friedrichsen, 1985). Executive decisions are based, in part, on the
availability of cognitive knowledge. This discussion is concerned with executive
 2. EXECUTIVE DECISIONS 83

decisions only (for a more complete description see Kluwe & Friedrichsen,
1985),
Executive decisions aim at the acquisition of information about own ongoing
cognitive activity and about the present state of own cognitive endeavor, as well
as the transformation or maintenance of one’s own cognitive activity and states.
Hypothetical executive decisions have three features: (a) they determine how to
solve a problem, but do not actually solve it, rather they guide the selection,
organization, and termination of cognitive operations; (b) they are considered to
be applied, especially to avoid costs, for example, in case of risky states or
failure; and (c) in the course of solution search, they are not always necessary.
Certain conditions that lead to the activation of executive processes, must be
satisfied.
Executive decisions may be conceived of as condition-action-connections (see
aiso Chi, Chapter 1, this volume). Conditions correspond to internal representa-
tions of states during problem solving, such as the amount of solution effort
invested, duration of search, type of problem, distance to goal, etc. Actions
would correspond to cognitive operations directed at the control and regulation of
the solution process. Executive decisions may then be considered as stored rules
for the control and regulation of cognitive activities during problem solving. An
example of a rule could be “‘If very long solution time (condition), then check
efficiency of own approach,’’ or ‘‘then check distance to goal’’ (action). The
action part may correspond to a subprogram that, once called up, achieves the
acquisition of information about features of the previous activity and of the
present state. Another possible action component would be to activate opera-
tions, in order to estimate the effort that still has to be invested; this would be a
control decision.
Regulatory decisions can be described in the same way. For example ‘if no
progress (condition), then increase resources for solution search’’ (action).
Again, this action will correspond to a subprogram that, if activated, would
achieve the goal, for example, by selection of other strategies. It is evident that
the input for the activation of a regulatory decision may be the output of the
action component for a control decision.
Executive decisions are activated when certain states during problem solving
match with stored declarative (cognitive knowledge). A certain state during the
solution search must first be registered; that is, it must be encoded as relevant for
the efficiency, the progress or the success of own approach. Thus, there has to be
declarative knowledge available concerning: the approximate duration of learn-
ing and solution processes; the features of promising approaches; etc. The avail-
ability of cognitive knowledge is necessary in order for a subject to become
aware of searching too long for a solution, searching without a plan, etc. Further-
more, this knowledge component must be connected with subsequent actions,
that guarantee appropriate control or regulation. It also requires knowledge about
36 KLUWE

how to perform, control, or regulate; that is, how to evaluate the own search,
how to speed up, etc. Both knowledge components that constitute rules for
executive decisions are acquired during cognitive development. It is assumed
that experience in problem solving is a major determinant of this development;
and that the connections between conditions, that is, identified states and appro-
priate actions for control and regulation, have to be acquired.
Consequently the concept of metacognition is not needed. The declarative
part, referred to as cognitive knowledge is not different from other factual knowl-
edge; the procedural component corresponds to rules that can be classified as
executive decision, because they are directed at the control and regulation of the
solution effort.
The following two sets of executive decisions are hypothetical; however, they
have some plausibility. Important guidelines have been obtained from the work
of Brown (1978), Flavell (1979), and Sternberg (1979). It has been important to
find that Selz, as early as 1913 and 1922, described processes that are highly
comparable to what in this discussion has been termed executive decisions. Selz
distinguished between processes that prepare for a solution, and those that ac-
company a solution. Among others Selz (1922) mentioned the processes of
pondering and reflection, and emphasized the evaluative selection of strategies
(p. 597, ff.). Following is a discussion of these cognitive activities.

Executive Control

Executive decisions directed at controlling or monitoring the own ongoing cog-

nitive activities generate information about the activity and about the present
cognitive state. Presumably, this may initiate the activation of what Flavell
(1979) has termed metacognitive strategies; he assumes that metacognitive strat-
egies are invoked to monitor cognitive progress (p. 909). The four executive
activities discussed serve monitoring purposes: classification; checking; evalua-
tion; and anticipation.

Classification. Classification of own cognitive activity provides informa-

tion about the status, type, or mode of cognitive activity, and provides an answer
to the question: ‘‘What am I doing here?’’ It is assumed that the accuracy of
classification depends, in part, on the availability of an inventory of concepts that
enable a person to talk about cognition and to look at own cognitive enterprises.
Thus, if someone has knowledge about the problem types available, then he or
she should be better able to identify the perceived cognitive requirements
as a
certain type of problem, that is, to classify it. It is further assumed that in order
to
classify cognitive activity a ‘‘protocol,’’ a may be more or less precise trace
of
its Own Cognitive activity is necessary. To our knowledge, this problem
has not
yet been studied (see Reed & Johnson, 1978: Dörner. 1984; Kluwe,
1980).
 2. EXECUTIVE DECISIONS 37

Kreutzer et al. (1975) report that even young children use such concepts as
forgetting and remembering. Lompscher (1972) instructed children to understand
and use the cognitive concepts of abstraction, comparison, and classification.
However, none of the available results indicate a connection between the avail-
ability of cognitive knowledge, in the sense of a general inventory of cognitive
concepts, and the improvement of monitoring one’s own cognitive activity.

Checking. Checking-steps taken during the process of problem solving pro-

vide information about the state of the cognitive system and the state of cognitive
activity. Classification, as mentioned earlier, provides information about what is
done, in order to find the solution to a problem. Checking is directed at the how
of one’s own cognitive activity, that is, goals, its organization, progress, suc-
cess, and the results. In thinking-aloud protocols checking-steps might take the
form of statements. For example: “‘I can remember most of the text.’’ ‘‘My
planning is pretty detailed and careful.’’ “‘I still have a pretty long way to go
before I get there.’’ ‘‘There’s something I do not understand here.”’
One of the first studies concerned with checking one’s own cognitive activity
was done by Flavell, Friedrichs, and Hoyt (1970). They found that with respect
to certain content, even 8 to 10-year-old children could quite accurately check
their recall readiness. Surprisingly, half of the nursery school children also
responded accurately and appeared able to adequately check their own recall
readiness. On the basis of the analysis of the children’s memorization process,
Flavell et al. concluded that information concerning one’s own memory, in this
case recall readiness, and, over a period time, the strategic deployment of vari-
ous types of memorization activities ‘‘appear to be more closely intertwined than
we had suspected’’ (p. 338). This might be responsible for the higher level of
recall that is especially evident from the memorization procedure of 10-year-old
children.
For adults Perlmutter (1978) reports considerable correlation between check-
ing of own memorization, indicated by the assessment of recall readiness, and
memory performance, indicated by the recall performance. However, the causal
connection between the two variables, memory monitoring and memory perfor-
mance, has not yet been shown.
In a comprehensive study of children between 4 to 10 years of age, Kelly,
Scholnick, Travers, and Johnson (1976) did not obtain age differences with
respect to memory monitoring. Furthermore, they could not detect relations
between checking (as indicated by estimations of recall readiness), study strategy,
and recall performance. This result is in contrast to the findings and assumptions
formulated by Flavell et al. (1970). The study by Kelly and his coworkers also
points to a problem that was first discussed by Masur, McIntyre, and Flavell
(1973). In this study younger children, (6-year-olds) check their own cognitive
state, which can be inferred from the distinction they are able to make between
those items they are actually able to recall and those items they are not yet able
38 KLUWE

recall. However, their strategic procedure, according to the viewpoint of

the experimenter, does not match the actual information available concerning the
cognitive state. The children do not focus their memorization activity on the
difficult items; instead, they memorize all items. It is assumed that younger
children usually do not use the available information about already learned and
about not yet learned items; as a result the memorization strategy they employ will
not be regulated (see Bisanz, Vesonder, & Voss 1978; Brown 1978). However, as
shown by the study of Masur et al., the phenomenon seems to be more complex. It
is possible to provide 6-year-old children with a more appropriate strategy; a
strategy that takes into account the available information about reproduceable and
nonreproduceable items. However, the more appropriate strategy does not im-
prove recall performance. The result of this study, as well as results reported by
Kelly et al., suggest a more complicated relation between the information acquired
through checking, the subsequent use of the information, and the resultant cog-
nitive performance. According to the studies mentioned, one can assume that for
older children, 8—10 years of age, there might be a closer interrelation between
cognitive strategy, checking activity, information acquired from checking, and
the resultant performance. Only the Kelly et al. study, in which no age dependent
variation was obtained, contradicts this assumption. Nevertheless, it is not clear
why younger children, despite having information about the state of their own
memorization activity, maintain a strategy, which from the standpoint of the adult
experimenter, is judged as less efficient, and does not correspond to what they
know.
The principal usefulness of checking activities is shown by another group of
studies. Brown and Barclay (1976) were the first to induce monitoring activities
directed at the children’s own cognitive activities. They trained children of
different age groups to use mnemonic strategies that incorporated self-testing
elements. Training in these strategies (rehearsal, anticipation) improved memory
performance and memory monitoring, indicated by more accurate estimations of
recall readiness. An interesting result is that one year following the training,
older children maintained the strategies. They were even able to transfer the use
of these strategies to new situational demands, and outperform untrained children
(Brown, Campione & Barclay 1979).
The importance of checking is also confirmed in a recent study by Ringel &
Springer (1980). As in Brown and Barclay’s (1976) experiment, the checking
was an independent variable. The authors could show that strategic training alone
was less effective in improving recall performance than was Strategic training
combined with feedback from the effect of training on own performance. Third
graders who received training did transfer the use of the Strategy to new task
demands; this may be attributed to their awareness of the benefits of the strategy,
which were explained by the experimenter. The authors assume that the children
themselves started checking and monitoring the effects of the strategy
on their
own performance. Fifth graders, however, did not gain from this
type of com-
 2. EXECUTIVE DECISIONS 39

bined training; their performance improved even without explicitly provided

monitoring of own performance. However, both studies do show that it is possi-
ble to enhance performance by training and by inducing checking steps directed
at one’s own cognitive endeavour.

Evaluation. Evaluation of one’s own cognitive states and activities provides

information about their quality. Evaluation goes beyond checking, because, in
order to judge the course and the state of own thinking, criteria are applied.
Evaluations made in the course of thinking may be found in thinking-aloud
protocols. For instance, there are statements such as, ‘‘My plan is not good
enough to rule out any risks.’’ or ‘This solution has been great work.’’ Evalua-
tion of own thinking is a complicated activity. As yet, little is known about the
criteria a person uses to evaluate the own cognitive activity. When is a selected
direction rejected because it is inefficient? When does an individual judge a state
to be uncontrollable during problem solving? When is a problem rejected as too
difficult?
Apparently, when an individual decides about the termination of solution
activities, effort and ability play an important role—for instance, when there is
no real progress. Presumably, the decision about progress during problem solv-
ing takes into account past solution efforts, beliefs about one’s own cognitive
abilities and estimations of the effort that has to be invested (see Heckhausen,
1982, for developmental aspects; Kluwe, 1980). This researcher is not aware of
any studies that concern the criteria that may be used to evaluate cognitive
efforts. Such criteria form the basis for judgments, such as systematic, chaotic,
clumsy, or elegant approach. As a result, when monitoring and regulation of
thinking are studied the criteria according to which monitoring and regulation are
performed must be considered. Simon (1975) has provided an important analysis
of different problem-solving strategies. He gives an evaluation of the strategies
that might be useful in solving the “Tower of Hanoi’’ problem; the criteria used
by Simon include memory load and transferability. The question is whether these
criteria may also be regarded as subjective, that is, as criteria that individuals will
take into account when deciding on a particular strategy. What a priori is consid-
ered to be an efficient strategy for a problem solution may not necessarily be the
best strategy for the subject studied; this is especially true for children (see
Richman & Gholson, 1978; Sternberg & Weil, 1980).
Usually, monitoring studies implicitly assume that the subjects do make eval-
uations. This is true for experimental studies referring to recall readiness, as well
as for the developmental experiments concerning memory (Butterfield & Bel-
mont, 1977). These studies deal mainly with evaluation of one’s own cognitive
activity on the basis of internal criteria; for example, the efficiency of one’s own
enterprise.
Presumably, beliefs about one’s own cognitive ability and cognitive effort
constitute, in part, evaluation of one’s own cognitive state. Ruble, Parsons and
40 KLUWE

Ross (1976) report that young children, 4-6 years, estimate the own invested
effort differently from older children 8—9 years. This, in turn, might affect the
evaluation of the efficiency of cognitive activity. Only children older than 8
years take into account social norms when evaluating their own cognitive perfor-
mance (Ruble, Boggiano, Feldman, & Lobe, 1980). However, not much is
known about how beliefs concerning own cognitive abilities, already invested
effort and effort still to be invested, and social norms combine when own cog-
nitive activity is evaluated. Nicholls (1978) elaborated on a developmental se-
quence: Children between the ages of 9 and 10 are able to distinguish between
their own ability and effort and the outcome. The development of this distinction
goes along with the acquisition of a more realistic self-concept, referring to the
cognitive features of an individual (Kluwe, 1980). This means that the beliefs
about one’s own cognitive equipment become more precise and appropriate; it
might also support a more realistic evaluation of own cognitive enterprises dur-
ing cognitive development. However, the question concerning about what chil-
dren younger than 8 or 9 years do is still unanswered; Do they evaluate their own
cognitive activity? If they do, as is assumed, then what criteria do they use?

Prediction. Prediction of cognitive states and cognitive activities provides

information about: the possible alternative options for problem solving; the pos-
sible sequence of solution steps; and the possible outcomes. Zimmer’s (1976)
concepts of “‘anticipation of methods,’’ and ‘“anticipation of goal states’’ refers
to similar phenomena. Prediction will presumably take the form of verbalization,
such as: “If I decide to work on this problem, all those technical details may be
hard to overcome; I’Il have to get someone to help me out with them.”’ or ‘‘In the
next three weeks or so, it should be possible to learn those three chapters on
cognitive psychology for the exam.”’
Prediction, an important component in problem-solving processes, is assumed
to be a function of problem-solving experience in different domains of reality.
Prediction is especially necessary in complex problem situations where the ef-
fects of own solution actions may expand and result in uncertain side effects. In a
similar way, prediction is needed in particularly important or risky situations, in
order to avoid ‘‘costs.’’
The developmental picture is determined essentially by results that concern
the prediction of one’s own memory span. The method used requires children in
different age groups to memorize a set of items; the children are told to memorize
the items for later recall. They are then asked to tell the experimenter how many
of the items they believe they will be able to recall. In such investigations,
(Brown, 1978: Chi, 1978) no age differences are obtained beyond third grade
(age 8-9 years). However, Kelly et al. (1976) did not find any age differences
even for younger children; their study covered kindergarteners through fourth
graders; the same is true of the results reported by Justice and Bray (1979).
According to the findings of Brown (1978), due to the inconsistency in the
 2. EXECUTIVE DECISIONS 41

children’s response, it is necessary to take into account methodological ques-

tions. There is the possibility that because of their predictions of their memory
span, children may erroneously be classified as realistic. Two other factors may
influence children’s predictions with respect to their own cognitive activity: (a)
according to Justice and Bray (1979), one has to assume that the accuracy of
predicting the own memory state depends on task variables, such as context of
the task, and the connection of the task to a desired goal; (b) Chi (1978) shows
that when children predict their own memory span, their declarative knowledge
is an important determinant of accuracy. The conclusions that younger children
in general show a deficit with regard to the prediction of own cognitive states and
activities are not justified by these findings. There are situations in which young-
er children are able to make accurate predictions. Worden and Sladewski-Awig
(1979) reported age differences (Kindergarteners, Second, Fourth, and Sixth
graders) in a study of span prediction. The authors inquired into the reason for
these differences, and their analysis ruled out the possibility of a faulty checking
process for the kindergarteners. According to the authors, overestimation of
memory span in this age group is attributable to the use of a rather liberal
response criterion. A criterion shift towards a more conservative estimate that is
the result of experience with such tasks might provide a more accurate picture; it
would also be in accord with the finding that meaningful contexts enhance the
accuracy of memory span predictions.
For adults, results appear to contradict the assumed importance of the devel-
opment of executive control during childhood. Lachman, Lachman, and
Thronesbery (1979) and Perlmutter (1978) did not find any age differences for
adults and aged subjects with respect to metamemory. Studies concerned with
prediction of cognitive states and activities, other than memory span tasks, are
not available. However, in this context, one study reported by Parsons and Ruble
(1977) can be cited. The authors report that preschoolers required to predict the
possible outcome of own problem solving in terms of success or failure do not
use the information about the outcome of own prior problem solving endeavors.
The four hypothetical processes aiming at control of the own cognitive ac-
tivity are not independent of one another. In order to extrapolate the possible
course of the own thinking process, evaluation requires checking; checking
requires classification; and prediction requires checking. According to certain
criteria, control activities have as a result the decision to regulate the own
cognitive activity. This is discussed in the next section.

Executive Regulation
Executive regulation refers to decisions about the organization, effort, amount,
course, and direction of one’s own cognitive activity. Regulatory decisions do
not necessarily imply a modification of the cognitive activity, they may also
result in the maintenance and continuation of a particular cognitive enterprise. To
42 KLUWE

date, regulatory decisions have been largely ignored in developmental literature.

Here four hypothetical processes of regulation are discussed: regulation of pro-
cessing capacity, regulation of what is processed, regulation of processing inten-
sity, and regulation of the speed of information processing.

Regulation of Processing Capacity. Regulation of processing capacity re-

fers to two related decisions: (a) the decision about the total amount of capacity a
person is willing to invest in order to accomplish certain cognitive demands; and
(b) the decision about the allocation of information processing capacity for
mental operations. Information processing capacity refers to a nonspecific re-
source for cognitive activity. The number of resources may vary and ultimately
may affect the outcome of mental operations (see Hasher & Zacks, 1979, for the
concept of effortful processing).
The first type of decision deals with the total amount of effort a person invests
in a problem-solving situation. The distinction between resource-limited and
data-limited information processing, made by Norman and Bobrow (1975), is
relevant. It relates to the question that might occur to someone involved in the
process of problem solving: Does increasing effort also increase the chances of
success? The second type of decision is closely related to selective attention. It is
assumed that human information processing capacity is limited. Consequently,
for many task demands, the problem solver has a choice as to which elements of
the situation or of the information attended to, should be processed. This, in turn,
means that the capacity limits need selective attention and call for executive
decisions concerning the regulation of information processing capacity. The
question of allocation of capacity has been studied in connection with perfor-
mance differences associated with incidental versus intentional learning. The
variation of capacity for processing affects the efficiency of effortful strategies;
therefore, such tasks should be studied in the context of development. Depending
on the age of the subjects, variations in capacity allocation may be associated
with performance differences.
Crane and Ross (1967) showed that older children are less likely to focus on
task-irrelevant information. Their findings were replicated by Kemler, Shepp,
and Foote (1976). Five- and 7-year-old children process more task-irrelevant
information than do 10-year-old children; however, some of the results are
contradictory. Howard and Goldin (1979) report that even kindergardeners are
able to regulate the allocation of their resources; that is, they are able to focus
attention on only the relevant aspects of the situation. Contrary to earlier studies,
the authors tried to maximize performance by explicitly directing the children’s
attention to the relevant situational features. Without this type of guidance,
efficient allocation of the resources seems to develop rather late in cognitive
growth. Recently, Hale and Alderman (1978) found that extending the exposure
time in a learning paradigm increases the incidental as well as the intentional
scores for 9-year-old children, but not for the 12-year-old children. As children
 2. EXECUTIVE DECISIONS 43

grow older, they regulate the allocation of resources so that task-relevant ele-
ments acquire more attention than irrelevant elements. In all cases concerning the
allocation of resources, the inferred differences go along with performance
differences.
A final, somewhat puzzling, result is reported by Markman (1979). Ina study
on comprehension monitoring, Markman found that even 12-year-old children
did not detect inconsistencies in the text although they claimed that they under-
stood the text. Several interventions to improve children’s comprehension
monitoring were unsuccessful. Finally, Markman warned the children that there
was a problem in each of the texts and that they should read carefully. Now the
12-year-old, but not the 9-year-old children, aware that there was a problem they
would have to detect, were able to apply the necessary control processes. This
relatively simple intervention conceivably could increase the children’s overall
processing capacity. Presumably, the allocation of capacity could also be in-
creased so that the children would focus attention on particular parts of the text.
Given these results, it can be assumed that the ability for deliberate regulation of
one’s Own processing capacity is acquired rather late in cognitive development,
probably not before the age of 10 or 11 years.

Regulation of What is Processed. If one reads through thinking-aloud pro-

tocols, one finds that it is not self-evident at what point or segment the individual
begins to search for a solution to a problem or what is the starting point selected.
In protocols dealing with the problem of solving logical theorems, subjects
decide whether: to work on the left side or the right side; to start working on ‘‘p’’
or “‘q.’’ For example, in cryptarithmetic protocols subjects, decide first, to find
the value for the letter E, etc. Such assumptions are implicit in the type of model
concerning balance scale problems elaborated by Siegler (1976). According to
Siegler’s model, children process weight and distance in a certain sequence.
Developmental studies concerned with this type of regulatory decision, the delib-
erate selection of what is processed, are not available. However, one study with
adult subjects illustrates the problem. Thorndyke and Stasz (1980) report the
differences between good and poor learners. They found that good learners
deliberately regulate, from a complex domain, what they intend to memorize and
what is subsequently to be memorized. Compared to poor learners, good learners
are more involved in deciding what segment of the material should be worked on
and, after successful storage of this segment, how to proceed (they were required
to memorize a map). In their learning processes, poor learners adopt a rather
unsystematic sampling procedure, and in some cases do not adequately focus
attention on the selected segments for memorization. The same holds true for
performance differences in the Raven, ‘Progressive Matrices’’ intelligence test
(Lüer et al., 1979). Apparently, poor performance goes along with insufficient
regulation of the own solution search; as a result, the subjects do not systemat-
ically select and analyze the segments of the task.
 44 KLUWE

Regulation of Processing Intensity. This form of executive decisions is as-

sumed to comprise three different decisions. First, is the regulation of the fre-
quency with which a cognitive operation or process is applied to specific seg-
ments of the problem situation. An example would be the number of
comparisons made between two objects; increasing the frequency of comparisons
between the two objects would also increase the reliability of the final similarity
judgment. Another example would be the repeated reading of an instruction
outlining the important construction steps, in order to avoid irreversible mis-
takes. Second is the regulation of time allotted to the execution of a cognitive
operation. For example, if two objects are compared, the regulation of time
allotted would refer to the time spent comparing objects. This in turn could result
in a longer examination of the attributes of one object, and also in an extension of
the set of attributes which constitute the final judgment; once again, the judgment
might then be more reliable. Third is the regulation of the solution search by
strategy shift and by modification of the perspective. The following two pos-
sibilities are considered: the solution search is not terminated when a strategic
effort is revealed as being unsuccessful, that is, it did not lead to the desired goal
State; rather, the solution search might be continued by shifting to an alternative
strategy (see, for example, the alternative steps in case of failure described by
Dorner, 1976; see also Kluwe, 1979, 1980). Work on a given problem might
also be continued on the basis of some other internal representation of the
problem. In other words, applying a different perspective, which entails different
encoding of the problem, may bring about a new starting point for subsequent
steps in the solution attempt. Therefore, when confronted with failure, a subject
may substitute another strategy for the originally selected approach, for example
by trial and error. Dorner (1976), in his work on problem solving, explicitly
refers to the shift of heuristics as one possibility in overcoming problem solving
barriers.
All three decisions directed at the regulation of processing intensity have one
aspect in common: they affect the degree of persistence of processing. Each one
determines to what degree the own cognitive endeavor will be exhaustive. Sever-
al findings from empirical studies show that younger children process informa-
tion less exhaustively than do older children. Sternberg and Rifkin
(1979), in a
study of the development of analogical reasoning, found that children aged
8—10
years prefer self-terminating operations, operations that need not to be executed
the maximum number of times. They offer one possible explanatio
n for this
result, the memory load associated with exhaustive processing (e.g.,
exhaustive
encoding of an input). It is interesting to note that according
to Sternberg and
Rifkin’s model, the less exhaustive processing characteristic of
younger children
can be connected with their lower performance; that is, with the
higher error rate
in analogical reasoning.
Similar differences may also have been responsible for the results
reported by
Ceci and Howe (1978). They found that in a recall task, 10-year-old
children
 2. EXECUTIVE DECISIONS 45

switch from one mode of memory to another in order to improve recall of the
learned material. This is not the case with 7-year-old children. Perhaps, in order
to reproduce as many items as possible, the younger children do not exhaustively
scan the stored categories. In an earlier study, Kobasigawa (1974) observed that
6- to 8-year-old children, when compared with l-year olds, do not sufficiently
scan the categories of the memorized items for recall. Ceci and Howe report that
10-year-old children, when compared to 7-year-old children do not terminate
their search for items, but instead continue memory search by applying another
organization of the material. They also report that the 10-year-old children
switch from a thematic to a taxonomic organization when instructed to recall
memorized material. The switching from one mode of search to another is related
to higher levels of recall. Examples demonstrating a connection between the
regulation of intensity and preceding executive control of the own approach, are
hard to find. Karmiloff-Smith (1978) reported that children between 7 and 11
years of age deliberately modify a coding system they themselves had worked
out, in order to retrace a longer way of getting around a town. It should be
pointed out that the children modified the coding system even though it had
actually worked. From continuous evaluation of the efficiency of the earlier
representational system it may be inferred that the children did not terminate the
coding activity, but instead, continued the activity.
Rogoff, Newcombe, and Kagan (1974) provide another example. Their re-
sults indicate that when given corresponding task demands, only 8-year-old
children increase the intensity of memorization. Only these children extended the
time for memorizing pictures to be recalled later as a function of the time-interval
between memorization and future recall. Rogoff et al. conclude that the ability to
control and regulate the own mnemonic activity according to the task demands
emerges around 6-7 years of age.
In conclusion, it may be assumed that, in general, younger children tend to
process information less intensively, that is, less exhaustively, than older chil-
dren. The relationship between these findings and Sternberg and Rifkin’s as-
sumption, mentioned earlier, that exhaustive processing is related to a higher
memory load, must still be explored.

The Regulation of Speed of Information Processing. Presumably, intensity

regulation, in those cases where more intensive processing is necessary, results
in more time-consuming activities. Given the same amount of information in a
problem situation, deliberate regulation of the speed of processing might be
reached through executive decisions. These decisions concern the completeness
of the cognitive procedures to be used: Adding certain cognitive operations might
decrease processing speed; skipping some processing steps might increase pro-
cessing speed. The existence of only one small scale study in which subjects
were instructed to regulate the speed of processing, is known; Barstis and Ford
(1977) changed the problem solving conditions for the ‘‘Matching Familiar
46 KLUWE

Figures Test.’’ Instruction emphasized either accuracy or speed of processing.

On the whole, 8-year-old children were better able to increase speed than the
kindergarteners, [who experienced a higher rate of error as speed increased].
There are many situations in which children, especially in school, are required to
deal with a task within a certain time limit; therefore further study of this problem
is necessary. The general hypotheses is that during early childhood a somewhat
stereotyped pace of information processing is typical. It is assumed that in the
process of cognitive development it is easier for children to first slow down
processing, rather than to speed it up. Studies concerning this problem should
investigate not only whether children can regulate the speed of information
processing but also how it can be accomplished by children of different ages.

REMARKS ON THE DEVELOPMENT OF EXECUTIVE

CONTROL AND REGULATION AND COGNITIVE
PERFORMANCE

Executive Control

According to Piaget (1928), children younger than 7 years of age are unable to
classify and check the features of their own thinking. The hypothetically as-
sumed precondition of control, namely the storage of a record of own cognitive
activity, is assumed to be satisfied only in later years, probably around 11 years
of age. Piaget believes that children under 7 years of age are unable to retrace the
steps taken in solving a problem because they are unable to keep a record of their
Own successive attempts. He postulates three stages for the evolution of chil-
dren’s introspection that occur between 7 and 10 years of age. From 7 years of
age onwards we will find ‘‘thinking about thinking’ (p. 146). Piaget considers
‘The shock of our thought coming into contact with that of others”’ (p. 146) to
be an important source for the development of the ability to retrace and evaluate
the own thinking. Until the age of 7, however, the child’s attention is in no way
directed towards thought as a medium interposed between the world and himself.
Yet, the research results discussed show that there are children younger than 7
who engage in control and regulation. It is remarkable that in this context Piaget
suggests that social communication between the child and his parents, relatives,
or other persons in his environment is an essential developmental determinant of
this ability. Recent research by Wertsch (1979) deals with this phenomenon
within the framework of the development of metacognition.
A few aspects of the possible relation between cognitive performance and the
control of one’s own thinking have already been discussed. The following dis-
cussion includes some results from experiments with adult subjects. There
are
individual differences in cognitive performance that apparently are connected
to
differences in executive control. These a posteriori differences may be
examined
 2. EXECUTIVE DECISIONS 47

more thoroughly in follow-up investigations where the influence of executive

control can be studied more systematically.
Owings, Petersen, Bransford, Morris, and Stein (1980) compared the learning
activity of successful and unsuccessful fifth grade learners. They report that
successful learners spontaneously monitored as they read and studied a text.
Thorndyke and Stasz (1980), in their study of learning processes of adult sub-
jects, found that good learners check and evaluate their study progress and
approach more often and more accurately than poor learners. The subjects’ task
was to memorize the contents of a map. Dörner et al. (1978), based on their
investigation of the problem-solving behavior of adult subjects confronted by
complex systems, reported that good problem solvers devote part of their solu-
tion activity to classification, checking, and evaluation of what they did and what
they plan to do. These observations were based mostly on statements made by
the subjects in thinking-aloud protocols. In another experiment subjects were
asked only to reflect on their own approach when solving a problem (Dörner,
1984). The effects of this type of instruction on cognitive performance were not
unequivocal. For instance, Hesse (1979) reported improvements in the perfor-
mance of subjects solving tasks from the Raven Progressive Matrices intelligence
test. However, compared with the control groups, this group’s performance also
was inferior on the easy tasks of this intelligence test. Reither’s (1979) subjects,
after working on switch light problems, were instructed to reflect on their own
solution search. Given a new problem, they needed fewer steps in their search for
a solution, without needing more time for the solution search, than was needed
by the control group.
These results suggest a causal relation between the mode and amount of
executive control and level of cognitive performance. Similar studies of develop-
mental aspects could be undertaken; however, this would require a more careful
examination of the a posteriori differences, such as in the study by Brown and
Barclay (1976). The question is: Is it possible to induce other types of executive
decisions in the same elegant way as was done in the Brown and Barclay’s
(1976) study on self-checking and monitoring of one’s own memory
performance?
In the study of the level of performance, one must consider that the effects of
executive control on performance may not be obvious. Steps directed at control
might affect performance in a way that cannot be captured by the usual success-
failure distinction. One example of the possible effect of executive control,
which requires careful analysis, is the ease with which cognitive strategies are
transferred to new task demands. Other criteria for determining the level of
cognitive performance might be the degree of organization of the subjects’ ap-
proaches, the ease of solution search, memory load, frequency of corrections,
persistence in case of failure, and flexibility of the approach, given changing
situational demands. One problem that exists is that the extension of the criterion
set for evaluating cognitive performance might reflect the effects of executive
48 KLUWE

control. A second problem has been revealed in the results of Kelly et al. (1976);
that is, that the relation between control of own cognitive activity and the subse-
quent use of the acquired information is not clear. Kelly et al. assume that there
may only be a loose relationship between cognitive activity and the information
acquired through control. Regardless of the strength of the hypothetical rela-
tionship, no available knowledge exists concerning the rules that govern the
application of such information. It is not clear when control is activated during
solution search; however, it is usually assumed that unexpected difficulties trig-
ger cognitive operations directed at the checking and evaluation of own cognitive
efforts. Yet, the behavior of good problem solvers and good learners, who are
inclined to control the own approach even in the absence of difficulties, contra-
dicts this assumption. The characteristics of the information acquired through
executive control are unknown. The regulatory decisions described may follow
control steps as a consequence of their results and may be the link between
cognitive activity, control, and performance.

Executive Regulation
It is assumed that regulatory decisions are preceded by control activity; however,
regulatory decisions need not lead to regulatory activity. In light of the issues
discussed previously, it is reasonable to assume that older children (beyond third
grade) are more likely than younger children to use information acquired through
monitoring. However, it is, precipitous to postulate a developmental sequence
from executive control alone to executive control that is connected with appropri-
ate regulation. This sequence might be suggested by the developmental picture
sometimes presented, of the successively emerging use of information acquired
through control. However, it is possible that younger children regulate their
cognitive activity according to criteria other than those being explored here. The
relativity of criteria with regard to the appropriateness of strategies is exemplified
in Sternberg and Weil’s (1980) experiment, which studied the relative efficiency
of strategies.
The following points are of importance to the question of what develops,

I. The ability to regulate one’s own cognitive activity on the basis of infor-
mation about one’s own ongoing effort is developed. For example, being aware,
that one’s own approach is too risky, presumably, requires a decision about
increasing the intensity and decreasing the speed of processing.
2. The ability to transform a regulatory decision into the appropriate cog-
nitive operation is developed. For example, if the solution search is too risky,
an
appropriate implementation of the regulatory decision to decrease speed might
be
to include a planning phase before continuing. A child may become aware
that
proceeding planfully, step by step, is the best way to avoid ‘‘costs’’;
but the child
may not know how to do this.
 2. EXECUTIVE DECISIONS 49

3. It is not assumed that during cognitive development children acquire ex-

plicit knowledge about criteria for efficient thinking. Instead, their flow of think-
ing, organized and directed by executive decisions, indicates implicit knowledge
about efficient thinking (see Greeno and Riley, Chapter 10, this volume, for the
concept of implicit understanding). Greeno and Riley’s cognitive procedures
match principles that might be formulated in a general theory about thinking;
however, the children themselves cannot verbalize these principles. Based on
this viewpoint, it is necessary to focus on the features of children’s and adults’
cognitive procedures.

CHILDREN’S REGULATION OF PROBLEM SOLVING AS

A FUNCTION OF REVERSIBILITY VERSUS
IRREVERSIBILITY OF ACTIONS

Experiment

The following is a preliminary report of the results of a developmental experi-

ment. The goal of the experiment was to create task demands that allow in-
ferences about children’s executive regulation in problem solving. In the
experiment regulatory activity is treated as a dependent variable. The purpose
was to determine if changing the problem solving conditions, that is, changing
from reversibility to irreversibility of actions, would cause children of different
ages to increase the intensity and to decrease the speed of their solution ap-
proach. On the basis of the assumptions previously outlined, one would expect
increments for the time and frequency with which solution operations are applied
under irreversible conditions; one would also expect the application of new
operations, which would not have necessary under reversible conditions. Finally,
it was also to be determined if age differences influence the ways the assumed
regulatory activity is implemented. In this context it was expected that perfor-
mance differences would be associated with regulatory differences.
The subjects of the experiment were 57 children, grouped by age (4, 5, 6, and
7 years). The tasks used were puzzles, which can be applied across a wide age
range; they are the type of problem in which the activities engaged in during
solution search can be easily traced (for criteria for selecting tasks see Brown &
DeLoache, 1978). Each of the children solved five puzzles; puzzles | through 3
under reversible conditions, puzzles 4 and 5 under irreversible conditions. Provi-
sion was made for the irreversibility of action for puzzles 4 and 5. Instead of
using a piece of white cardboard, which had been the working surface for
reversible conditions, the children were given a cardboard that was completely
covered with a strong, colorless adhesive layer. Once a piece of the puzzle was
placed on the working surface, it became fixed and could not be removed.
 50 KLUWE

After the third reversible puzzle was accomplished each child was told that
because all the pieces would fall down, the completed puzzle could not, after all,
be hung on the wall like a picture. This was demonstrated to the children. Then
each child was shown the cardboard with the adhesive and was encouraged to put
his or her hand on the surface of the cardboard. The only information the children
were given, regarding the irreversible condition, was that once the pieces of the
puzzle had been put onto the cardboard, they could not be removed. Although
this information was given twice, the child was not told to be particularly careful
or to pay more attention.
The collected data are based on the registration of eye-fixations and of prob-
lem-solving operations. The recorded data includes:

Code 1: looking at the target picture.

Code 2: looking at the pieces of the puzzle.
Code 3: looking at the cardboard working surface.
Code 22: sorting the pieces of the puzzle.
Code 23: comparing one piece of the puzzle with the partial solution on the
working surface.
Code 44: holding a selected piece close to a fixed piece on the adhesive
surface without attaching it to the surface.
Code 6: model-building, that is, putting two or more pieces together next to
the working surface.

Some of the preliminary results that refer to the regulation of the intensity of
the solution search are reported in what follows. The analysis focusses on the
transition from reversible (puzzle 3) to irreversible (puzzle 4) conditions. Only
the data that refer to the interval between the end of the instruction (providing the
target picture) and putting the first two parts of the solution onto the surface are
considered. Thus, for each child, the segment of the problem-solving process
under reversible conditions was compared with the segment under irreversible
conditions. The preliminary restriction of the analysis with respect to this
seg-
ment of the solution search assumes that during this interval relevant regulatory
activities are especially necessary. The general hypothesis was that
7-year-old
children, when compared to all younger children, would significantly
increase
the frequency and the time of eye-fixations and of certain solution
operations.
The following six regulatory activities were subject to further analysis:

+ h: increasing the relative frequency of solution actions

and/or eye
fixations.
+ t increasing the average time of actions and/or eye-fix
ations;
— h: decreasing the relative frequency of actions and/or
eye-fixations.
— t: decreasing the average time of actions and/or
eye fixations.
add: adding new operations.
sub: skipping operations.
 2. EXECUTIVE DECISIONS 51

According to our discussion of processing intensity, an increase of intensity

should be indicated by +h, +t, and add-regulations.

Results

As a result of changing the task demands from reversibility to irreversibility of

actions, all the children showed a high degree of regulatory activity.
1. Overall changes (+, —, add, sub) were more likely to occur than was the
maintenance of the initial approach.
2. Overall increments of relative frequency and time (+ h, + t), as well as
adding of new operations (add), predominated (Fig. 2. 1a).
3. The regulatory activity extended over the range of encoded solution ac-
tions and eye-fixations. Changes involved primarily eye-fixations (Code 1, 2,
23, Se see rig o.db):

20

(a)
10

+h -h ++ -t add sub (type of regulation)

20
(b)

10

1 2 22 23 3 4h 6 (Code: regulated
operation)
FIG. 2.1a. Percentage of regulatory activities. b: Percentage of actions sub-
ject to regulation.
 (a)

rev irr rev wr rev irr rev irr (condition)

u * 5.5 6 6 7 7 (age)

x (f)

18

16

14

12
(b)
10

[Si
(ee)
fe
>

rev sirr rev irr rev irr rev irr (condition)

dh SS a 8 7 7 (age)
FIG 2.2.
 2. EXECUTIVE DECISIONS 53

4. The analysis of the total time used and the analysis of the total number of
solution actions performed until the second part of the puzzle was placed on the
working surface resulted in the following picture:
The total time of solution activities is increased significantly under irrevers-
ible conditions (F = 14.07; df = 1,53; p = .01; Fig. 2.2a). The total number of
solution actions performed until the second part of the puzzle was placed on the
working surface is also increased significantly under irreversible conditions (F =
212705 aps 1853: p = OL: Bag, 2.26).
5. Some solution actions were systematically related to certain regulatory
activities. First, the relative frequency of applying operation Code 23 (comparing
one piece of the puzzle with the partial solution) increased under irreversible
conditions. The difference between reversible and irreversible conditions with
regard to this variable is statistically significant (F = 16.55; df = 1,53; p = .01;
Fig. 2.3a). However, no comparable effects were found for the other registered
operations. Contrary to what was expected, the subjects did not look more
frequently at the target picture or at the pieces of the puzzles. Second, the
average time for operation Code | (looking at the target picture) and for Code 23
(comparing one piece of the puzzle with the partial solution) also increased
significantly under irreversible conditions. (F (Code 1) = 4.14; df = 1,53; p
2.037 Code 23) = 35-343 df = 1,535p.= .01; see Fig. 2.3b, e).
Third, the adding of new operations (add), which as previously indicated was a
dominant regulatory activity, focused on specific activities (25% for Code 23,
22% for Code 44, 16% for Code 22). The significant deviation from equal
distribution shows that under irreversible conditions, primarily three operations
were included in the solution search: Code 23 (comparing one piece of the puzzle
with the partial solution); Code 22 (sorting the pieces of the puzzle next to the
working surface); and Code 44 (holding a piece ofthe puzzle close to the adhesive
without attaching it to the surface). These operations, added to the repertoire of
already applied operations, provide for a more thorough, detailed analysis of the
puzzle material. Later it is seen that age differences were found for this type of
regulatory activity.
Fourth, the analysis also included the relation between the time spent for a
certain operation and the total time spent on solution search, until the first two
parts of the puzzle were fixed. The analysis of the relative amount of time for
operations yielded statistically significant effects for operation Codes |, 2, and
23. The results indicate that, compared to the total solution time, the proportion
of time for the activities Code 1 and Code 2 decreased; however, the proportion

FIG. 2.2. Effect of the condition reversibility vs. irreversibility on a: the average
of total time t for solution actions (X (t) in min/10): (F = 14.07; df = 1,53; p
<.01). b: the average of total number of solution activities f (F = 21.70; df=
1,53; P = .01). (Averages per age group and condition).
 x(h) 14 7 CODE 23

rev - rev irr rev irr rev irr (condition)

4 & 555 6.6 77 (age)

ZU) CODE 23

b
(b) 3

zZ

rev rev irr rev irr rev irr (condition)

4 Ses Een 7 7 (age)

tip)

rev irr rev irr rev irr rev irr (condihon)

4 & ves 6 6 7 7 (age)
FIG. 2.3. Effect of the condition reversibility vs. irrevers
ibility on a: Relative
frequency h of operation Code 23 P1655
= 1539p =O: b: Average
time (&)r for operation code 23 (F = 35.34:
p= loss p=.0l)ee: Average time
(x)t for operation code I (F = 4.14: df=
1;53sp'= .05). (Averages per age group
and condition; time (&)t in min/ 10)

54
 2. EXECUTIVE DECISIONS 55

of time for Code 23 increased (F [Code 1] = 9.73; df = Loon = 0) )Fi(Code

2 Er O4 raf 015538 pen DIE EI Coded3 |=. 24.616 df =) 1534p: S01).
The solution operation Code | (looking at the target picture) showed an
increment in the average time r, discussed earlier; however, the proportion of
time invested for this activity, compared to the total time, decreased. Presum-
ably, this was due to the extension of the application of operation Code 23, as
well as to the addition of new solution operations (add) under irreversible
conditions.
In sum the results concerning the regulation of solution activities indicate the
following: Under irreversible problem solving conditions new operations are
added to the solution procedure; in general the extension of the solution activity
aims at a more thorough analysis of the pieces of the puzzle (Codes 22, 23, and
44; see Duncker, 1966/1935, and the concept of ‘‘material analysis’’). Howev-
er, regardless of age, under irreversible conditions the average time spent for
“looking at the target picture’’ (Code 1) is increased. In addition, before it is put
down on the surface (Code 23), a selected piece of the puzzle is looked at more
frequently and for a longer period of time. For some solution actions (Code | and
Code 2) the relative amount of time is decreased.
The findings | to 5 indicate that, as a result of changing specific features of
the task demands, it is possible to obtain certain types of regulation. The reg-
ulatory decisions are only directed at a few solution operations. Accordingly, this
deliberate regulatory activity aims at increasing the intensity of the own solution
search by: increasing the relative frequency of operations, increasing the average
time for operations by adding new operations, and, decreasing the speed of
problem solving. Therefore, it appears that regulatory activity is performed to
reduce the risk of failure associated with irreversible problem solving conditions.
This conclusion is referred to when the relation between regulatory decisions and
cognitive performance is discussed.
6. Age differences affect type of regulation. There are certain preferences for
regulatory activities, within age groups, described earlier; these include increas-
ing the average time for solution operations and adding new operations. Howev-
er, this preferential pattern is true only for children 4, 5, and 6 years of age. At
the age of 7 children do not have such clearly focused preferences with regard to
the type of regulation (chi-square (4) = 23.42; chi-square (5) = 16.81; chi-
square (6) 27.28; chi-square (7) = 1.16; df = 5).
It must be pointed out that when the situational demands are changed from
reversible to irreversible conditions, the children of all age groups show quite
similar solving behavior. The average number as well as the average time of
solution operations is increased for all age groups. There were no age differences
in the operation Codes 1, 2, 22, and 23, considering the changes of time and
number. However, important age differences existed in the operation Codes 22,
44, and 6: Code 22 (Sorting of the pieces of the puzzles) is applied under
irreversible conditions significantly more often by older children (6—7 years of
56 KLUWE

% CODE 22
100

80

60
(a)
40

20

% CODE 44
100
80

60
(b)
40

20

% CODE 6
50

40

(c) 30

20

10

7
5
FIG. 2.4a-c.
 2. EXECUTIVE DECISIONS 57

age, chi-square = 9.81; df = 3; p= .025; see Fig. 2.4a). The effect is due to the
fact that only a few 4-year-old children apply this operation for a better analysis
of the pieces under irreversible conditions; however, especially, the 7-year-old
children do use this operation. It can be assumed that the sorting of pieces leads
to decisions about the sequence of fixing pieces on the working surface. There is
no such age effect for the reversible condition. Code 44 (holding one piece of the
puzzle close to the surface without attaching it) is used under irreversible condi-
tions more often by 4- and 5-year-old children (chi-square = 9.85; df = 3; p
= .05). There is no comparable age difference under reversible conditions (see
Fig. 2.4b). Under irreversible conditions Code 6 (model-building, i.e., putting
pieces together next to the working surface), significantly more older children
put two or more pieces of the puzzle together before attaching them to the surface
(chi-square = 8.21; df = 3; p = .05; see Fig. 2.4c).
Both actions, Code 44 and Code 6, are added to the repertoire of solution
activities under irreversible conditions. In the case of Code 6, which may also be
used under reversible conditions, older children add this operation significantly
more often under irreversible conditions (chi-square = 10.97; df = 3; p = .02).
The opposite is true for Code 44: younger children include this operation signifi-
cantly more often in their repertoire under irreversible conditions (chi-square =
12.32; df= 3; p = .01; see Fig. 2.5). In sum, age differences have the following
effect: Older children extend their solution search by adding the operations that
help to get a more thorough analysis of the material (Code 22, 23, 6). In contrast,
younger children under irreversible conditions add Code 44, significantly more
often, to their repertoire.
7. Cognitive performance and regulatory activities differences between suc-
cessful and unsuccessful children apply to all age groups, and do not necessarily
correspond to what is expected. One may assume that successful and unsuc-
cessful problem solvers of different age groups have different regulatory pat-
terns. The following differences emerge: Successful as well as unsuccessful
children increase the average number of solution operations (F [Condition X
success] = .23; df = 1,55; p = .25). The same is true of the total time until
putting the second piece on the surface; there is no interaction (F [condition X
success] = .22; df = 1,55; p = .25). Therefore, according to the analyses, it
must be stated that successful and unsuccessful children are doing about the
same. However, the open question is: Why does one group fail and the other
succeed? There was only one regulatory operation where the difference resulted:

FIG. 2.4a. Percentage of children in different age groups who sort pieces of the
puzzle under irreversible conditions (Code 22). b: Percentage of children in
different age groups that hold one piece close to the pieces on the working surface
without attaching it (Code 44). c: Percentage of children in different age groups
that put two or more pieces together next to the working surface (Code 6; model-
building).
 58 KLUWE

FIG. 2.5: Percentage of children in different age groups that add

model-building
(Code 6) and comparison of pieces with attached pieces (Code 44)
under irrevers-
ible conditions.

Under irreversible conditions unsuccessful children reduce

the proportion of time
spent analyzing the target picture (Code 1: F [unsuccessful]
= 7.48; df = 1,2755
= .05); however, this is not the case with successful
children. The activities of
unsuccessful children are shown in Fig. 2.6a—c. For all
analyzed variables no
remarkable relations with cognitive performance were
found.
The interpretation offered must be kept as hypothetical.
It is not permissable
to say that under irreversible conditions, the mere increm
ent of the total time for
solution search and of the total number of solutio
n operations guarantees a
successful approach. Rather, it is necessary to find
patterns of specific regulatory
activity that correspond to levels of cognitive
performance. Presumably, the
conception of increasing intensity of problem
solving activity under irreversible
conditions, as disussed, is too
simple; beyond the mere increment of
time,
number of operations, and adding of operations,
it is necessary to expect more
comple x regulatory patterns.
 xf)

14

12

fay eee
8

rev irr rev irr (condition)

+= i+ - - (performance)

(b)

rev itr rev irr (condition)

ee - - (performance)

FIG. 2.6. Separated effects of the condition reversibility vs. irreversibility for
successful and unsuccessful children: a: Average number of solution activities
(®)f: (F (condition x success) = .23; df = 1,55; p = .25). b: Average time for
solution activity <(1): (F (condition x success) = Damon= Neyo = 7a Ce Wire
successful children’s average proportion of time for the analysis of the target
picture: (F (Code 1) = 7.48; df = 1,27; p = .05).

59
(c)

rev ‘int rev irr rev irr rev irr (condition)

4 4 caus 6 6 7 7 (age)
FIG. 2.6c.

Discussion

The resulting picture has been surprising. Under changing task demands, 4—5-
year-old children, were not expected to engage in regulatory activity to the extent
they actually did. Remember, the children were not instructed to be particularly
careful or attentive when working with the adhesive surface under irreversible
conditions. Because no age differences were obtained, with respect to the
fre-
quency of puzzle solutions, it may be assumed that each age group, in order
to
cope with the changing task demands, has its own pattern of regulatory activities.
There is still no clear picture of how such regulatory patterns are constituted
and
structured. Further analysis is also needed to account for the factors responsibl
e
for determining performance differences: this also requires a more
accurate de-
scription of performance differences. For instance, with regard to
the features of
the solution outcome, the groups of unsuccessful children could
be divided into
two subgroups; and the features of the process of solution, such
as planfulness,
could be given some consideration.
During the sessions, something that might most aptly be
described as ““per-
sistence’’ of regulatory activity, was observed. The impression
was repeatedly
given that under irreversible conditions the children
tended to fall back on a
careless way of operating. Possibly, this is an additional importa
nt feature of the
regulation of thinking under changing problem solving
conditions.
The study of problem solving under reversible and irrever
sible conditions
clearly revealed the effects of different regulatory
decisions; differences for age
and performance were infrequent. On the whole, under
irreversible conditions
young children appear to apply almost as many activiti
es as older children; the
Same is true for the time variable. The higher freque
ncy in the use of analytical
 2. EXECUTIVE DECISIONS 61

operations (Code 22) and of model-building (Code 6) is most striking for the 7-
year-old children. Only a few of the younger children, especially the 4-year-olds,
use these operations. Instead, younger children in the comparison of fixed pieces
with selected pieces (Code 44), show a higher frequency. However, under irre-
versible conditions, through their regulatory activity, younger and older children
appear to create a condition in which the trial-and-error mode of operation,
practiced under reversible conditions, can be used again. For younger children
this is done by adding the operation Code 44 (comparing a selected piece with a
fixed piece by holding the former close above the surface without attaching it to
the adhesive). For the older children the same effect is achieved by building
models (Code 6), that is, putting pieces together next to the surface; and by more
frequent sorting of the pieces near the working surface (Code 22). Contrary to
expectation, successful problem solving is not necessarily connected with more
frequent visual analysis of the target, of the pieces of the puzzle, or of the
working surface. The increase of time and frequency of solution activities alone
is not a sufficient predictor of performance. It is assumed that one must search
for certain patterns of regulatory activities that are associated with success or
failure. The final question is: How can the solution behavior of children who on
completing the puzzle successfully, proudly say: ‘‘Isn’t that something; I have
done a good job and yet I did not even once look at the picture here’’ (it happened
twice), be analyzed.
The analysis of the available data does not justify Piaget’s (1969/1928) con-
clusion ‘‘childish reasoning before the age of 7-8... resembles a physical
action during which one arm-movement will bring about another arm-move-
Mentone a (Dac 2),

ACKNOWLEDGMENTS

Research reported here has been done as part of a research project supported by a grant
from the Foundation Volkswagenwerk, Hannover, West Germany. Thanks is due to
Kerstin Modrow, Traute Vaihinger, and Heidi Windeit for their contributions to the
completion of this manuscript.

REFERENCES

Barstis, S., & Ford, L. H. (1977). Reflection-impulsivity, conversation, and the development of
ability to control cognitive tempo. Child Development, 48, 953-959.
Bisanz, G. L., Vesonder, G. T., & Voss, J. F. (1978). Knowledge of one’s own responding and the
relation of such knowledge to learning. Journal ofExperimental Child Psychology, 25, 116-128.
Brown, A. (1978). Knowing when, where, and how to remember: A problem of metacognition. In
R. Glaser (Ed.), Advances in instructional psychology (Vol. 1, pp. 77-165). Hillsdale, NJ:
Lawrence Erlbaum Associates.
 62 KLUWE

Brown, A., & Barclay, C. R. (1976). The effects of training specific mnemonics on the metamne-
monic efficiency of retarded children. Child Development, 47, 71-80.
Brown, A., Campione, J. C., & Barclay, C. R. (1979). Training self-checking routines for estimat-
ing test readiness: generalization from list learning to prose recall. Child Development, 50, 501-
512)
Brown, A., & DeLoache, J. S. (1978). Skills, plans, and self-regulation. In R. S. Siegler (Ed.),
Children’s thinking: What develops? (pp. 3-35). Hillsdale, NJ: Lawrence Erlbaum Associates.
Butterfield, E. C., & Belmont, J. M. (1977). Assessing and improving the executive cognitive
functions of mentally retarded people. In J. Bailer & M. Sternlicht (Eds.), Psychological issues in
mental retardation. Chicago: Aldine.
Campione, J. C., & Brown, A. (1978). Toward a theory of intelligence: contributions from re-
search with retarded children. Paper presented at the meeting of the American Educational
Research Association, Toronto, Canada.
Ceci, S. J., & Howe, M. (1978). Age-related differences in free recall as a function of retrieval
flexibility. Journal of Experimental Child Psychology, 26, 432-442.
Chi, M. T. H. (1978). Knowledge structures and memory development. In R. S. Siegler (Ed.),
Children’s thinking: What develops?, (pp. 73-96). Hillsdale, NJ: Lawrence Erlbaum Associates.
Crane, N. L., & Ross, L. E. (1967). A developmental study of attention to cue redundancy
follow-
ing discrimination learning. Journal of Experimental Child Psychology, 5, 1-15.
Dörner, D. (1974). Die kognitive Organisation beim Problemlösen. Bern: Huber.
Dörner, D. (1976). Problemlösen als Informationsverarbeitung. Stuttgart: Kohlhammer.
Dörner, D. (1984). Self-reflection and problem-solving. In F. Klix (Ed.) Human
and artificial
intelligence (pp. 101-107). Amsterdam: North-Holland, (Originally published
1978.)
Dormer, D., Kreuzig, H. W., & Stäudel, Th. (1978). Lohhausen. 2. Bericht,
DFG-Projekt ‘‘Sys-
temdenken.’’ Universität Gießen.
Duncker, K. (1966). Zur Psychologie des produktiven Denkens. Berlin:
Springer, (first published
1935).
Flavell, J. H. (1971). First discussant’s comments: What is memory development
the development
of? Human Development, 14, 272-278.
Flavell, J. (1976). Metacognitive aspects of problem solving. In
L. Resnick (Ed.), The nature of
intelligence (pp. 231-235). New York: Wiley.
Flavell, J. (1979). Metacognition and cognitive monitoring:
A new area of cognitive-developmental
inquiry. American Psychologist, 34, 906-911.
Flavell, J. H. (1981). Cognitive monitoring. In W. P. Dickson
(Ed.) Children’s oral communica-
tion skills (pp. 35-60). New York: Academic Press.
Flavell, J., Friedrichs, A. G., & Hoyt, J. D. (1970). Developmental changes in memorization
processes. Cognitive Psychology, 1, 324—340,
Flavell, J., & Wellman, H. (1977). Metamemory.
In R. V. Kail, Jr. & J. W. Hagen (Eds.),
Perspectives on the development of memory and cognitio
n (pp. 3-33). Hillsdale, NJ: Lawrence
Erlbaum Associates.
Hale, G. A., & Alderman, L. B. (1978). Children’s selective attention with variation in
amount of
stimulus exposure. Journal of Experimental Child Psychol
ogy, 26, 320-327.
Hasher, L., & Zacks, R. T. (1979). Automatic
and effortful processes in memory. Journal of
Experimental Psychology: General, 108, 356-388
.
Heckhausen, H. (1982). The development of achieve
ment motivation. In W. W. Hartup (Ed.),
N of Child Development Research (Vol. 6, pp. 600-668
). Chicago: University of Chicago
Tess.
Hesse, F. W. (1979). Trainingsinduzierte
Veränderungen in der heuristischen Struktu
Einfluß auf das Problemlösen. Unveröffentl. r und ihr
Dissertation, Technische Hochschule Aachen
Howard, D., & Goldin, S. (1979). Selecti .
ve processing in encoding and memory: An
resource allocation by Kindergarten children. analysis of
Journal of Experimental Child Psychology,
95. 27, 87—
 2. EXECUTIVE DECISIONS 63

Justice, E., & Bray, N. (1979). The effects of context and feedback on metamemory in young
children. Paper presented at the biennal meeting of the Society for Research in Child Develop-
ment, San Francisco.
Karmiloff-Smith, A. (1978). Micro- and macro-developmental changes in language acquisition
and other representational systems. Paper presented at meeting on Units of Analysis in the Study
of Conceptual Development and Language Development, University of Ghent.
Kelly, M., Scholnick, E., Travers, S., & Johnson, J. W. (1976). Relations among memory, memo-
ty appraisal, and memory strategies. Child Development, 47, 648-659.
Kemler, D. G., Shepp, B., & Foote, K. (1976). The sources of developmental differences in
children’s incidental processing during discrimination trials. Journal of Experimental Child Psy-
chology 21, 226-240.
Kluwe, R. (1979). Wissen und Denken. Stuttgart: Kohlhammer.
Kluwe, R. H. (1980). Metakognition: Komponenten einer theorie zur kontrolle und steuerung
eigenen denkens. Unveröffentl manuskript, Universität München.
Kluwe, R. H. (1981). Metakognition (Positionsreferat). In W. Michaelis (Hrsg.), Bericht tiber den
32. Kongreß der Deutschen Gesellschaft für Psychologie in Zürich 1980 (pp. 246-258). Göt-
tingen: Hogrefe.
Kluwe, R. H. (1982). Cognitive knowledge and executive control: Metacognition. In D. Griffin
(Ed.), Human mind—animal mind (pp. 201-224). New York: Springer.
Kluwe, R. H., & Friedrichsen, G. (1985). Mechanisms of control and regulation in problem solv-
ing. In J. Kuhl & J, Beckmann (Eds.) Action Control (pp. 183-218). Berlin: Springer.
Kobasigawa, A. (1974). Utilization of retrieval cues by children in recall. Child Development, 45,
127-134.
Kreutzer, M. A., Leonard, C., & Flavell, J. (1975). An interview study of children knowledge
about memory. Monographs of the Society for Research in Child Development, 40 (1, Serial No.
159).
Lachman, J., Lachman, R., & Thronesbery, C. (1979). Metamemory through the adult life span.
Developmental Psychology, 15, 543-551.
Lompscher, J. (1972). Theoretische und experimentelle Untersuchungen zur Entwicklung geistiger
Fähigkeiten. Berlin: Volk und Wissen.
Lüer, G., Putz-Osterloh, W., & Hesse, F. W. (1979). Abschlußbericht über das Projekt ‘‘System-
denken’’. Technische Hochschule Aachen.
Markman, E. (1979). Realizing that you don’t understand: Elementary school children’s awareness
of inconsistencies. Child Development, 50, 643-655.
Masur, E. F., McIntyre, C. F., & Flavell, J. H. (1973). Developmental changes in apportionment
of study time among items in a multitrial free recall task. Journal of Experimental Child Psychol-
ogy, 15, 237-246.
Neisser, U. (1967). Cognitive Psychology. New York: Appleton.
Nicholls, J. G. (1978). The development of the concepts of effort, and ability, perception of
academic achievement, and the understanding that difficult tasks require more ability. Child
Development, 49, 800-814.
Norman, D., & Bobrow, D. (1975). On data-limited and resource-limited processes. Cognitive
Psychology, 7, 44-64.
Owings, R., Petersen, G., Bransford, J., Morris, D., & Stein, B. (1980). Spontaneous monitoring
and regulation of learning: a comparison of successful and less successful fifth graders. Journal of
Educational Psychology, 72, 250-256.
Parsons, J., & Ruble, D. N. (1977). The development of achievement-related expectancies. Child
Development, 48, 1075-1079.
Perlmutter, M. (1978). What is memory aging the aging of? Developmental Psychology, 14, 330-
345.
Piaget, J. (1969). Judgment and reasoning in the child. London: Routledge & Kegan Paul. (Origi-
nal work published 1928.)
 64 KLUWE

Reed, S. K., & Johnson, J. A. (1978). Memory for problem solutions. In G. Bower (Ed.) The
psychology of learning and motivation (pp. 161-201). New York: Academic Press.
Reither, F. (1979). Über die Selbstreflektion beim Problemlösen. Unveröffentl. Dissertation, Uni-
versität Gießen.
Reitman, W. (1965). Cognition and thought. New York: Wiley.
Richman, S., & Gholson, B. (1978). Strategy modeling, age, and information processing efficien-
cy. Journal of Experimental Child Psychology, 26, 58-70.
Ringel, B. A., & Springer, C. J. (1980). On knowing how well one is remembering: The per-
sistence of strategy use during transfer. Journal of Experimental Child Psychology, 29, 322-333.
Rogoff, B., Newcombe, N., & Kagan, J. (1974). Planfulness and recognition memory. Child
Development, 45, 972-977.
Ruble, D. N., Parsons, J. E., & Ross, J. (1976). Self-evaluative responses of children in an
achievement setting. Child Development, 47, 990-997.
Ruble, D., Boggiano, A., Feldman, N., & Loebl, J. (1980). Developmental analysis of the role of
social comparison in self-evaluation. Developmental Psychology, 16, 105-115.
Schneider, W., & Shiffrin, R.M. (1977). Controlled and automatic human information processing:
1. Detection, search, and attention. Psychological Review, 84, 1-66.
Selz, O. (1922). Zur Psychologie des produktiven Denkens und des Irrtums. Bonn: Cohen.
Selz, O. (1913). Uber die Gesetze des geordneten Denkverlaufs: 1. Teil. Stuttgart:
Speemann.
Siegler, R. S. (1976). Three aspects of cognitive development. Cognitive Psychology,
8, 481-520.
Simon, H. A. (1975). The functional equivalence of problem solving skills. Cognitive
Psychology,
7, 268-288.
Simon, D. P., & Simon, H. A. (1978). Individual differences in solving physic problems. In R. S.
Siegler (Ed.), Children’s thinking: What develops? (pp. 325-348). Hillsdale,
NJ: Lawrence
Erlbaum Associates.
Sternberg, R. (1979). The nature of mental abilities. American Psychologis
t, 34, 214-230.
Sternberg, R., & Rifkin, B. (1979). The development of analogical reasoning
processes. Journal
Experimental Child Psychology, 27, 195—232.
Sternberg, R., & Weil, E. (1980). An aptitude X treatment interaction
in linear syllogistic reason-
ing. Journal of Educational Psychology, 72, 226-239.
Thorndyke, P., & Stasz, C. (1980). Individual differences in procedures
for knowledge acquisition
from maps. Cognitive Psychology, 12, 137-175.
Thornquist, K., & Wimmer, H. (1977). Meta-gedächtnis
als Bedingung der gedächtnisent-
wicklung. Zeitschrift für Entwicklungspsychologie und Pädagogi
sche Psychologie, 9, 252-264.
Ueckert, H. (1980). The cognitive executive: From artificial
intelligence toward a psychological
theory of consciousness. Paper presented at the International
Congress of Psychology, Leipzig.
Wertsch, J. (1979). The social origins of metacognition.
Paper presented at the meeting of the
Society for Research in Child Development, San Francisco
.
Wimmer, H., & Thornquist, K. (1979). The relation of metamemory to mnemonic performa
nce in
memory development: metamemory is always necessary
but never sufficient. Unpublished manu-
script, University of Salzburg.
Worden, P., & Sladewski-Awig, L. (1979). Children
’s mnemonic judgments and metamemory.
Paper presented at the meeting of the Society for Researc
h in Child Development, San Francisco.
Zimmer, J. (1976). Erscheinungsformen und ausprag
ung der vorausschau im denken von
vorschulkindern. In J. Lompscher (Hrsg.), Verlaufs
qualitäten der geistigen Tätigkeit (pp. 126—
184). Berlin: Volk und Wissen.
 Metacognition, Executive
Control, Self-Regulation, and
Other More Mysterious
Mechanisms

Ann Brown
University of Illinois

WHAT IS METACOGNITION?

In this chapter, the historical roots and current status of the fashionable but
complex, and often poorly understood, concept of metacognition and other top-
ics with which it shares a family resemblance are described. Metacognition refers
to understanding of knowledge, an understanding that can be reflected in either
effective use or overt description of the knowledge in question. Various forms of
metacognition have appeared in the literature, some of these instantiations are
puzzling and mysterious. For example, Marshall and Morton, (1978, p. 226)
refer to the mechanism that permits the detection and correction of errors in
speech production as an EMMA, or “‘even more mysterious apparatus,’’ that
could be an ‘‘optional extra.’’ This researcher argues that far from being an
optional extra, the processes that have recently earned the title metacognitive are
central to learning and development. Although metacognitive concepts may ap-
pear mysterious, what is at issue is the concept of degree of understanding. Thus,
a learner can be said to understand a particular cognitive activity if he or she can
use it appropriately and discuss its use. Understanding admits of degrees, be-
cause learners can often use knowledge effectively without being able to explain
how this can be so. Individuals, with severe learning problems in a domain, can
neither apply nor discuss a piece of knowledge or a rule they have acquired.
An attempt is made to demystify the concept(s) and to illustrate why it is not
an optional extra, or an epiphenomenon (Brown, 1978). The varied forms of
behavior that have been subsumed under the heading, metacognition, are de-
scribed. Then, the historical roots of metacognition are traced through four
somewhat separate strands of inquiry: (a) verbal reports as cognitive processes;

65
 66 BROWN

(b) executive control within an information processing framework; (c) self-reg-

ulation, metaprocedural reorganization, and reflected abstraction from the Piage-
tian school of developmental psychology, and (d) other-regulation, a Vygotskian
notion. Finally, the current status of the concept metacognition is considered and
the importance of the concept for understanding mechanisms of change is
emphasized.
The term ‘‘metamemory’’ was introduced into the literature by John Flavell
(1971), who was also responsible for the first modern study of metamemorial
processes in children (Flavell, Friedrichs, & Hoyt, 1970). The term metacogni-
tion came into vogue sometime around 1975. The term has been problematic
from its inception, denounced as fuzzy and faddish, and even unnecessary
(Marshall & Morton, 1978). It is true that metacognitive-like concepts are
fraught with some of the most difficult and enduring epistomological problems of
psychology. However, this researcher argues that metacognitive-like entities lie
at the very roots of the learning process; and, concomitant with the revival of
interest in things metacognitive, is an exciting revival of interest in mechanisms
of change and development.
Metacognition refers loosely to one’s knowledge and control of own cognitive
system. Two primary problems with the term are: it is often difficult to dis-
tinguish between what is meta and what is cognitive; and there are many different
historical roots from which this area of inquiry developed. The confusion that
follows the use of a single term for a multifaceted problem is the inevitable
outcome of mixing metaphors. Some attempt is made to explicate the confusion
by making the metaphors somewhat more explicit.
Consider first the interchangeability of cognitive and metacognitive functions.
An area where this problem is particularly acute is the currently popular domain
of metacognition and reading, writing, and studying. The following quote
Fla-
vell (1976): ‘‘Asking yourself questions about the chapter might function
either
to improve your knowledge (a cognitive function) or to monitor it
(a metacog-
nitive function),’” demonstrates the interchangeability of cognitive
and metacog-
nitive functions. A particular activity can be seen as the strategy
itself (looking
for main points), its monitoring function (a metacognitive activity),
and a reflec-
tion of the knowledge (also metacognitive) that it is an appropriat
e strategy to
employ in a given situation. Recent reviews of the literature
by Baker and Brown
(1981), have been justly criticized on the grounds that they
have encouraged the
practice of dubbing as metacognitive any strategic action
engaged in while
reading.
Metacognitive reading skills include the following activiti
es, previously dig-
nified with the title of mere Strategies: establishing
the purpose for reading;
modifications in reading due to variations in purpose
; identifying important
ideas; activating prior knowledge; evaluation of
the text for clarity, com-
pleteness, and consistency; compensation for failures
to understand; and assess-
ing one’s level of comprehension (Baker & Brown,
1981). Which of these
 3. METACOGNITION AND OTHER MECHANISMS 67

activities should be deemed metacognitive, or, more subtly, which components

of these complex activities are meta, is not clear.
It is clear that these strategic reading activities were recognized well before
the emergence of the term metacognition. Since the turn of the century educa-
tional psychologists (e.g., Dewey, 1910; Huey, 1908, 1968; Thorndike, ied)
were quite aware that studying and reading involve the type of activities now
called metacognitive. For example, consider Dewey’s system of inducing reflec-
tive reading. The aim was to induce ‘‘active monitoring,’’ ‘‘critical evaluation,”’
and deliberate ‘‘seeking after meanings and relationships.”’ According to Dewey
(1910), learning was *‘learning to think’’ and reading was thinking stimulated by
texts: "Thinking (or reading) is inquiring, investigating, turning over, probing or
delving into, so as to find something new, or to see what is already known in a
new light—in short it is questioning’’ (p. 265).
Another early advocate of ‘‘metacognitive’’ processes of reading was Thorn-
dike (1917):

Understanding a paragraph is like solving a math problem. It consists of selecting

the right elements of the situation and putting them together in the right relations,
and also with the right amount of weight or influence or force for each. The mind is
assailed as it were by every word in the paragraph. It must select, repress, soften,
emphasize, correlate, and organize, all under the influence of the right mental set or
purpose or demand. (p. 329)

Thus, there is considerable historical agreement that reading and learning

from texts involve metacognitive skills. The early paradigms are hauntingly
familiar. For example, Baldwin (1909) introduced the now popular (Myers &
Paris, 1978) reading skills questionnaire, in which he asked 12 to 18-year-olds to
describe their study habits; and Thorndike (1917) invented the error-detection
paradigm now extensively studied (Markman, 1977, 1979; Stein & Trabasso,
1982). Thorndike asked sixth graders to read texts and then answer questions on
what they had read. He found that sixth graders were surprisingly passive readers
who, as they read, did not appear to spontaneously monitor their understanding.
Although they often said they understood, it was obvious they did not. Baldwin
and Thorndike were clearly able to describe these activities without once resort-
ing to the term metacognition.
A second source of confusion concerning the wide spread use of the term
metacognition is that, within the modern psychological literature, it has been
used to refer to two distinct areas of research: knowledge about cognition and
regulation of cognition. The two forms of metacognition are indeed closely
related, each feeding on the other recursively; attempts to separate them lead to
oversimplification. However, they are readily distinguishable, and they do have
different historical roots.
Knowledge about cognition refers to the stable, statable, often fallible, and
often late developing information that human thinkers have about their own
 68 BROWN

cognitive processes; traditionally, this has been referred to as knowing that

(Broudy, 1977). Knowledge about cognition is relatively stable. One would
expect that knowledge of pertinent facts about a domain that it is fallible, se-
verely limited for short-term verbatim retention, etc., for example, memory,
would be a permanent part of one’s naive theory on the topic. This form of
knowledge is often statable; one can reflect on the cognitive processes involved
and discuss them with others. Unfortunately, because the learner is credited with
knowledge only if he or she can state it, this statability criterion is circular in
most empirical studies. Of course, this form of knowledge is often fallible; the
child, or adult for that matter, can perfectly ‘‘know’’ certain facts about cogni-
tion that are not true; naive psychology is not always empirically supportable.
Finally, this type of knowledge is usually assumed to be late developing; it
requires that learners step back and consider their own cognitive processes as
objects of thought and reflection. For Piaget, reflected abstraction requires hy-
pothesis testing and evaluation, and the ability to imagine possible worlds and
their outcomes; therefore, it demands formal operational thought (Piaget, 1976).
For others, earlier signs of emergence are possible; however, reflection is rarely
attributed to the very young child or novice, regardless of how precocious they
might be (Brown & DeLoache, 1978).
The second cluster of activities that is dubbed metacognitive in the develop-
mental literature, consists of the activities used to regulate and oversee learning.
These processes include planning activities (predicting outcomes, scheduling
strategies, and various forms of vicarious trial and error, etc). Prior to undertak-
ing a problem, monitoring activities (monitoring, testing, revising, and re-sched-
uling one’s strategies for learning) during learning; and checking outcomes (eval-
uating the outcome of any strategic actions against criteria of efficiency and
effectiveness). It has been assumed that these activities are relatively unstable,
not necessarily statable, and relatively age independent (i.e., task and situation
dependent). These skills are not necessarily stable; although they are ubiq-
uitously employed by adults, on simple problems. Young children monitor and
regulate their own activities. Indeed, a case can be made that all active learning
involves self-regulation. These activities are often not statable; knowing
how to
do something does not necessarily mean that the activities can be brought
to the
level of conscious awareness and reported on to others.
Although knowledge and regulation of cognition are incestuously
related, the
two forms of activity have different roots and different attendant
problems. The
tension generated by the use of the same term, metacognition,
for the two types
of behavior is illustrated by the fact that the leading proponen
ts in the field tend
to answer questions about the nature of metacognition with:
‘‘It depends.’’ Is
metacognition late developing? It depends on the type of
process referred to. Is
metacognition conscious: It depends. . . .
Tension has bedevilled the area since the original use of
the ‘‘meta’’ terms.
Flavell (1971) in his initial papers on the topic, referre
d to both types of meta-
 3. METACOGNITION AND OTHER MECHANISMS 69

memory. In the landmark paper, where the first experimental studies were intro-
duced (Flavell, Friedrichs, & Hoyt, 1970), two tasks were featured: span-estima-
tion and recall-readiness. Span-estimation is a “‘knowledge-of’’ task where
children are asked to estimate their own memory span; recall-readiness is a
“regulation-of”’ situation where children are observed as they select, revise,
orchestrate, and evaluate strategies for learning.
In order to make explicit the epistomological problems that surround things
metacognitive, and to illustrate the dangers of mixed metaphors, four separate
stands of inquiry, where the current issues of metacognition were introduced and
originally disputed, are considered.

ROOTS OF METACOGNITION

The four historically separate, but obviously interlinked, problems in psychology

that pertain to issues of metacognition are discussed. First, the enduring ques-
tions concerning the status of verbal reports as data (Ericsson & Simon, 1980):
Can people have conscious access to their own cognitive processes? Can they
report on these processes with verisimilitude? How does the act of reporting
influence the processes that are the subject of the report? Second, is the issue of
executive mechanisms within an information processing model of human and
machine intelligence. Who or what is responsible for regulation of cognition?
With what knowledge or form of knowledge must an executive be imbued? How
do processing models deal with the issue of the infinite regress problem of
homunculi within homunculi; how do they deal with the problems of con-
sciousness, intention, and purpose? Third, are the issues of self-regulation and
conceptual reorganization during learning and development that have always
featured in Genevan developmental psychology. Self-regulation and conceptual
reorganization have played a major role in Piaget’s (1976, 1978) modern writ-
ings and those of his co-workers, notably Karmiloff-Smith (1979a, 1979b);
Karmiloff-Smith & Inhelder (1974-75); Inhelder, Sinclair, & Bovet (1974).
Fourth is the topic of other-regulation, which is central to Vygotsky’s (1978)
theory of development.

Verbal Reports as Data

Nature of the Problem. Central to the controversy surrounding the ability to
provide verbal reports of one’s own cognitive processes is the issue of reflective
access (Brown, 1982a; Brown & Campione, 1981; Pylyshyn, 1978), conscious
access (Rozin, 1976; Gardner, 1978), or reflected abstraction (Piaget, 1976). It
refers to the essentially human ability to step back and consider one’s own
cognitive operations as objects of thought; to reflect on one’s own thinking. The
70 BROWN

concept of cognizing about cognition, like most reasonable ideas, has a history
dating back to Plato and Aristotle. Spearman (1923) points out:

Such a cognizing of cognition itself was already announced by Plato. Aristotle

likewise posited a separate power whereby, over and above actually seeing and
hearing, the psyche becomes aware of doing so. Later authors, as Strato, Galen,
Alexander of Aphrodisias, and in particular Plotinus, amplified the doctrine, desig-
nating the processes of cognizing one’s own cognition by several specific names.
Much later, especial stress was laid on this power of ‘reflection’, as it was now
called by Locke. (p. 52-53)

Few people have been able to resist the temptation to play on words. Consider
this example from Spinoza: ‘‘. . . He who knows something knows at the same
time that he knows it and he knows as well that he knows what he knows’’ (cited
in Weinert, this volume).
Special interest in children’s knowledge of their own knowledge was evi-
denced by John Locke, who distinguished two primary sources of ideas: sensa-
tion and reflection. Reflection referred to the ‘‘perception of the state of our own
minds,’’ or ‘‘the notice which the mind takes of its own operations.’’ Neither
children nor the uneducated were thought to be given to bouts of reflection.
According to Locke (1690/1924), primitive minds do not indulge in self-
reflection.

Hence we see the reason why it is late before most children get ideas of the
operations of their own minds; and some have not any very clear or perfect ideas of
the greatest part of them all their lives. Because, though they (ideas) pass there
continually, yet, like floating visions they make not deep impressions enough to
leave in the mind clear, distinct, lasting ideas, till the understanding turns inwards
upon itself, reflects on its own operations, and makes them the object of its own
contemplation. (p. 46)

In the late nineteenth and early twentieth centuries, an interest in self-reflec-

tion was retained as a central topic of debate in the emergent fields of psychol-
ogy. Concern with verbal reports as data distinguished the Wurzburg school of
thought and was vigorously contested by Wundt (for a review see Mandler &
Mandler, 1964; Humphrey, 1963). For example, Buhler’s (1907) protocols of
verbal reports during problem solving have a delightfully contemporary
ring.
Buhler asked his experienced subjects (actually his professional colleagues,
in-
cluding Kulpe and Durr) to talk aloud about their thought processes
as they
attempted to answer complex questions: ‘‘Was the theorem of Pythagoras
known
in the Middle Ages?”’ ‘‘Can you get to Berlin from here in seven hours?”’
‘‘Can
the atomic theory of physics ever be proven untrue by any scientific
discovery?”’
The verbal protocols revealed reasoning on the basis of incomplete
knowledge
that is essentially similar to contemporary work (Collins, 1977;
Norman, 1973).
 3. METACOGNITION AND OTHER MECHANISMS aa

Indeed, many of the thorny problems of awareness, consciousness, and intro-

spective reports, etc. were thoroughly aired by the Wurtzburgers in their battle
with Wundt, who may have been the father of psychology but was certainly not
the father of this form of metacognition! However, differing viewpoints exist
(see Gleitman, in press).
Current interest in knowledge of cognition comes from many sources. A
recurring theme is the distinction between multiple and reflective access that
occurs in several areas of psychology; including cognitive ethology, machine
intelligence, human intelligence, and developmental psychology (for further
treatment see Brown, 1982a; Brown & Campione, 1981). Rozin’s influential
paper on the evolution of intelligence has been especially important in making
the problem of access paramount. Rozin distinguishes between adaptive spe-
cializations related to intelligence, which originated to satisfy specific problems
of survival, and widely accessible cognitive programs. Because the adaptive
specializations, common in the animal world, evolved as solutions to specific
problems, they were originally tightly wired to a narrow set of situations that
called for their evolution; in lower organisms they remained tightly constrained
components of the system. In the course of evolution, cognitive programs be-
come more accessible to other units of the system and therefore, may be used
flexibly in a variety of situations. This flexibility is the hallmark of higher
intelligence, reaching its zenith at the level of conscious access and control,
which affords wide applicability over a broad range of mental functioning.
There are two main points to Rozin’s accessibility theory; first, is the notion
of welding (Brown, 1974, 1978; Shif, 1969). Intelligence components can be
strictly welded to constrained domains (i.e., skills available in one situation are
not readily used in others, even though they are appropriate). According to Rozin
(1976), young children’s programs are “‘not yet usable in all situations, avail-
able to consciousness, or statable.’’ (Development is) *‘the process of gradually
extending and connecting together isolated skills with a possible ultimate exten-
sion into consciousness’’ (p. 262). The second part of each of the preceding
statements refers to Rozin’s second major point, conscious access. Even if skills
are widely applicable, rather than tightly welded, they need not be conscious and
statable. Conscious access to the routines available to the system is the highest
form of mature human intelligence.
A similar distinction between multiple and reflective access is made by
Pylyshyn (1978). Multiple access refers to the ability to use knowledge flexibly;
because it can be systemically varied to fit a wide range of conditions, knowl-
edge is informationally plastic. Reflective access refers to the ability to ““mention
as well as use’’ the components of the system. Similarly, Gardner (1978) cites as
hallmarks of human intelligence: (a) generative, inventive, and experimental use
of knowledge, rather than preprogrammed activities (multiple access); and
(b) the ability to reflect on one’s own activity (reflective access). Piaget, in his
most recent work, notably in The Grasp of Consciousness (1976) and Success
 72 BROWN

and Understanding (1978) also denotes reflected abstractions, *‘Conscious prod-

ucts of reflexive abstraction,’’ as the keystone of formal operations.
The concepts of multiple and reflective access are key issues in the field of
metacognition and in developmental psychology. Original interest in the topic
was spurred by the persistent finding of a production deficiency in young chil-
dren’s learning (Reese, 1962; Flavell, 1970), which is a classic problem of
access. Children who know perfectly well how to use a strategy or have the
relevant prior knowledge, often fail to access it on appropriate occasions. In-
deed, American and Soviet psychologists have suggested that one of the primary
problems with young learners is that they tend to acquire information that is
‘‘welded’’ to the form and context in which it was acquired (Brown, 1974; Shif,
1969). Reflective access was also pinpointed as a problem for the young and, in
particular for the retarded learner. On the basis of the relative absence of these
qualities in slow learners, Brown, Campione, and their colleagues, diagnosed
multiple and reflective access as underlying mechanisms of intelligent behavior
(Brown, 1974, 1978: Brown & Campione, 1978, 1981; Campione & Brown,
1977). The twin concepts of flexibility and reflection are important issues that
have wide implications for theories of learning and development.
In short, several theorists from quite disparate schools agree that the most
stringent criteria of understanding involve the availability of knowledge to con-
sciousness and reflection; thus permitting verbal reports. However, there are
problems associated with verbal reports as data (Ericsson & Simon, 1980): these
problems are well illustrated in developmental psychology (Brown, 1978; Flavell
& Wellman, 1977; Wellman, 1983). First, there is the obvious problem of asking
children to reliably inform on the content of their own conscious processes. As
Piaget and others have pointed out, children are as likely to distort and modify
their observations of their thought processes as they are their observations of
the
world around them. Eyewitness testimony is fallible for the objects and events
of
the internal world, as well as for the external world. This problem
has been
referred to elsewhere as the problem of externalizing mental events (Brown,
1978, 1980). What a child says he or she has done or will do is not
necessarily
related to his or her performance. Reliance on verbal responses and
justification
is risky when the participant is a child, and is not without its problems
when
adults are the informants (Ericsson & Simon, 1980; Nisbett
& Wilson, 1977:
Smith & Miller, 1978; White, 1980).
Second, is the equally thorny problem of what relationship to
expect between
what an informant knows and what he or she does. Studies with
children, which
have been mostly correlational, have yielded only moderat
e correlations: for
example, between performance on a restricted class of memory
tasks and chil-
dren’s statable knowledge of (sometimes different) class
of memory phenomena
(Cavanaugh & Borkowski, 1979: Justice & Bray, 1979;
Kendall, Borkowski, &
Cavanaugh, 1980). However, in some recent studies
(Perlmutter, 1978; Wimmer
 3. METACOGNITION AND OTHER MECHANISMS 73

& Tornquist, 1978) a clearer relationship was found to exist. In most of these
Studies there is a less than compelling rationale as to why one would expect a link
between the particular form of metamemory probed and actual performance
(Wellman, 1983). Similarly, as Flavell (chapter 1, this volume) points out, in
any one particular task, there are many reasons why there should not be a close
link between metamemorial knowledge and memory performance. What is
needed is a theory as to when one may expect to find a relationship between a
subject’s verbal reports and actions, and exactly what that relationship would be.
The beginnings of such a theory have been posited by Ericsson and Simon (1980)
and is readily adaptable to the developmental literature.
However, one must first distinguish between the many forms of contexts that
the child is asked to comment on. Many forms of knowledge about cognitive
things can be assumed to be stable, others are transient and are elicited only in
certain situations. Stable forms of knowledge are the kinds of information learn-
ers may possess about themselves as learners and about the learning context. I
have argued elsewhere that practiced learners come to know a great deal about
the learning situations (Brown, 1982b; Brown, Bransford, Ferrara & Campione,
1983). They know certain stable characteristics about themselves as learners;
they know the demands of certain classes of problems, and they are aware of the
necessity of tailoring their learning activities so that they will be finely in tune
with specific criterial tasks. These are the types of knowledge that Flavell and his
colleagues (Flavell, 1980, 1981; Flavell & Wellman, 1977) have classified as
person, task, and strategy variables. Learners possess naive theories of what it
takes to learn certain types of materials, in order to meet certain criterial task
demands. They also know a lot about their repertoire of the available strategies
needed to accomplish certain ends. That young children are less informed about
stable characteristics of learning is amply documented (Kreutzer, Leonard, &
Flavell, 1975; Myers & Paris, 1978; for reviews see Baker & Brown, 1981, in
press; Cavanaugh & Perlmutter, 1980; Flavell & Wellman, 1977; Wellman,
1983). This lack of stable, statable, knowledge is due to children’s relative lack
of experience in learning situations that occur repeatedly in school; it reflects
their novice status as deliberate learners.
Stable forms of knowledge are the forms of metacognition that are purport-
edly tapped by questionnaire studies or clinical interviews. However, verbal
reports are often taken in situations where the knowledge being assessed may be
transient, is elicted in the face of a particular task or context, or is elicited during
the actual performance of a task. Supposedly, young children are judged to be
incapable of the split mental focus that is required for simultaneously solving
problems and commenting on the process. Therefore, protocol analyses of per-
formers actually solving problems have been restricted to adult or adolescent
subjects. Instead, developmental psychologists have typically asked children to
report retrospectively on what they have just done, with all the attendant prob-
74 BROWN

lems associated with constructing past events (Piaget, 1976). A more common
and more problematic procedure is to ask children to describe how they would
behave in certain hypothetical situations.
Ericsson and Simon (1980) have criticized the practice of asking adult sub-
jects to imagine possible worlds and how they might act in them. For example,
Nisbett and Wilson (1977) asked their subjects how they would react to a story if
some passages had been omitted. Similarly, Reed and Johnson (1977) asked
adults how they would solve a problem if it were presented to them again, a
procedure often employed in the developmental studies of memory-metamemory
linkage. Later that the ability to construct possible worlds (Johnson-Laird, 1980)
and operate in these worlds with facility, is a late developing skill that shares
many of the “‘child-as-scientist’’ features of Piaget’s formal operational period.
It is necessary to distinguish between reports on stable forms of knowledge
and reports on states occurring during problem solving. It is also necessary to
distinguish between: (a) predictive verbalizations about possible performance
before the event; (b) concurrent verbalizations during the actual performance;
and (c) retrospective verbalization after the event has transpired. It is also neces-
sary to be more precise, in order to determine if the information being sought
concerns specific information or very general information. Questions of the
form, “How do you perform these tasks?’’ implicitly request very general infor-
mation and leave open to the informant the creative task of constructing gener-
ality by drawing on a variety of specific prior experiences, including general
knowiedge of what one ought to do in such tasks.
An illustration of the demand characteristics of general inquiries was provided
by Sternberg, Ketron and Powell (1982). They analyzed the systematic strategies
used by adult subjects solving analogy problems and then, for a variety of
theoretical reasons, trained the subjects to adopt an alternative strategy. After
training, performance was again assessed to determine if the subjects had
adopted the new strategy. Sternberg et al. considered the subjects’ latencies and
error patterns, previously found to be highly diagnostic of the particular strategy
employed, and the subjects’ verbal reports of the processes they were using.
Sternberg et al. (p. 160) report: **. . . Subjects’ descriptions were highly con-
sistent with the strategies subjects were trained to use, but not consistent with the
Strategies subjects actually did use . . .”” This researcher does not necessarily
agree with their comment that this discrepancy indicates the subjects’ lack of
awareness of what they were doing. The subjects may very well know what they
are doing, but they may also realize that it is different from what the experiment-
er told them to do. If this situation exists, college-age subjects may well respond
to the demand characteristics of the situation and tell the experimenter that
they
are conforming. Under such circumstances it is clear why verbal reports
may
bear very little relation to the actual cognitive processes used in the
task. Al-
though the example given might be extreme, the tendency to construct
general
 3. METACOGNITION AND OTHER MECHANISMS 15

theories about what one ought to do is encouraged by the probe

for general
procedures.
Ericsson and Simon (1980) point out that in areas of applied psycholo
gy,
where verbal questioning of subjects is important and has a long history,
subjects
are rarely asked for their general theories or impressions. Rather, Flanagan
(1954) and others use what is referred to as the critical incident technique
;
informants are only asked to report about very specific incidents. For example,
combat pilots would be asked to describe a particular, actually experienced
incident, and then to answer questions on how they thought or felt during that
specific event.
An adequate theory of the relation of verbal reports to actual performance
must include some a priori predictions of when verbal reports influence perfor-
mance and when they do not. The current state of the theory concerning the
verbal reports and performance link is summarized in Flavell’s morality argu-
ment. Flavell and Wellman (1977) point out that there are many reasons why
there may not be a direct link between, for example, memory and metamemory.

Suppose a person judges that categorized stimuli are easier to recall than non-
categorized ones. Would he inevitably use categorization as a storage strategy,
given obviously categorizable stimuli? Not at all. He may know about categoriza-
tion but think that something else might be better yet in this situation. He may think
the list easy enough to use simple inspection for storage. He may have enough
knowledge to judge that categorization would be a good strategy, if asked about it,
but not enough to think to utilize such a strategy on his own. Lastly, there are
undoubted gaps between metamemory and memory behavior attributable to Origi-
nal Sin. Moral action does not always accord with moral beliefs, and similarly, we
do not always try to retrieve information or prepare for future retrieval in what we
believe to be the most effective ways.

This is undoubtedly true, there are many reasons to explain a lack of a close
correspondence between what one knows and what one does; however, Original
Sin lacks something as an explanatory construct. Future research could concen-
trate on the specific circumstances which would result in a metacognitive-cog-
nitive link. Under what circumstances would one predict a positive or a negative
relationship between verbal reports and performances? Under what circum-
stances, if any, would no relationship be predicted. (Wellman, 1983)?
Ericsson and Simon (1980), for example predict that asking the learner to
provide concurrent verbal reports as he or she solves a problem, will, depending
on the function of the verbal reports in the ongoing learning process, have a
neutral, positive, or negative effect. Under circumstances where the subject is
asked to describe information that is already available (i.e., in STM), the effect
will be neutral. If the subject is asked to report on information that is available,
but not in verbal or propositional form, the translation process may slow down
76 BROWN

performance but will not otherwise interfere. Several examples of this form of
experiment are quoted by Ericsson and Simon. One example is a discrimination
learning study by Karph (1973). Subjects were divided into think aloud and
control groups. They were asked to learn a discrimination between compound
stimuli varying in form, size, color, shape of border, texture of line under the
form, etc. The correct solution to each problem involved one of the dimensions.
The stimuli were visual; therefore, the translation to a verbal hypothesis should
have been relatively easy; these hypotheses, for example, “‘It’s the black one,”’
were available to the subjects at the time of solution. Subjects were equated for
performance on a pretest. The think aloud subjects verbalized their hypotheses
during the second session and returned to silent problem solving on the third
session, or posttest. No differences in accuracy between the experimental and
control groups were found at any stage of the experiment; however, under
instructions to think aloud, the experimental group was significantly slower.
The relation between ‘‘thinking aloud’’ and problem solving can sometimes
be beneficial. This is particularly true if the type of verbalization that is required
is a statement of a rule or a reason for an action. Ericsson and Simon (1980)
quote many examples where, on a standard laboratory puzzle, such as the Tower
of Hanoi, instructions to state a rule significantly accelerate the learning process.
For example, Gagné and Smith (1962) had subjects state a reason for each move
they made. The simple ‘‘state a reason’’ procedure significantly improved both
learning and transfer performance. Gagné and Smith (p. 17) concluded, rather
vaguely, that instructions to verbalize the reasons for moves ‘‘forced subjects to
think, induced more deliberate planning,’’ etc. The study was criticized on the
grounds that the verbalization group had extra learning time. However, follow-
up studies found that instructions to state the reasons for moves covertly were
equally as effective as overt statements, and that learning time was not a con-
founding factor (Wilder & Harvey, 1971); Davis, Carey, Foxman, and Tarr
(1968) found that the presence or absence of an audience was also insignificant.
The Gagné and Smith findings are robust and suggest that forcing learners to
make a rule explicit helps the learning process, as well as the transfer of the rule.
Verbal reports can often have a negative effect on the learning process. This
situation occurs when the requirement for overt verbalization competes for cen-
tral processing capacity with the processes that must be reported. In on-line
protocols it is characteristic for verbalization to stop when the going gets difficult
and to resume when the cognitive load is lessened. Similarly, verbal reports of
information that is not generally available to consciousness is a disruptive pro-
cedure. For example, one reason why Piaget (1976, 1978) experienced so
much
difficulty getting children to describe their actions may be because the subject of
those descriptions were just that, actions. Perceptual motor activities are
notori-
ously difficult to describe, and it is indeed true that people can do a great
deal
that they cannot describe. Requiring an expert golfer to describe his
or her swing
may, in fact, ruin the experts’ swing, rather than help the novice.
Broadbent
 3. METACOGNITION AND OTHER MECHANISMS 77

(1977) suggests that one is only aware of higher level processes during
the
performance of perceptual-motor tasks; there is some empirical support for this
position. Ruger (1910) found that adult subjects solving mechanical problems
could give only limited explanations of their intermediate solution steps; Klinger
(1974) found a high frequency of general control verbalization, ‘‘Let’s see,
where was I,’’ ‘‘Dammit,’’ and so forth, but very few examples of detailed step
by step verbalization. Many current information processing models claim that
with repeated practice many of the intermediate steps of both thought and action
become automatized, and therefore, they are even less available to conscious
introspection (Norman & Schallice, 1980; Schneider & Shiffrin, 1977; Shiffrin
& Schneider, 1977). Asking subjects to report on internal events that are not
readily available to such inspection may significantly impair the processes on
which they must report.

Developmental Data. Although the majority of developmental work on met-

acognition consists of studies examining children’s verbal reports on their own
mental processes, there are very few empirical observations that could be used to
answer the questions raised by the Ericsson and Simon taxonomy. Systematic
variation of the temporal relation of reporting and the variety of cognitive ac-
tivities considered has not been undertaken.
Questionnaire studies consist primarily of situations where the child must
imagine scenarios and how he or she might act in them. Of the fourteen main
items contained in the Kreutzer, Leonard and Flavell (1975) questionnaire, ten
are completely imaginary (e.g., the child is asked to imagine how two other
children might perform in a task). The remaining four items demand that the
child predict how he or she might perform in an as yet unexperienced activity;
however, the child is given a concrete prop to help the fantasy. For example, if
the child must decide whether colored pictures are easier to remember than black
and white drawings, samples of the materials are provided. Furthermore, al-
though the authors did try to query the young respondents about specific memory
tasks, the kinds of processes that the children were asked to reflect on were very
general. As Ericsson and Simon (1980) point out, asking people to describe the
general processes they might use in imaginary situations is the least favorable
circumstance for producing verbal reports that are closely linked to the cognitive
processes under discussion.
The majority of quasi-experimental studies of metacognition, where children
are asked to predict how they will perform in tasks or describe how they fared,
are also imperfect. Some tentative evidence indicates that children are better able
to identify the items they recall, than to predict in advance how well they will do
(Brown, 1978), however; the data is incidental to the main purpose of the
experiments from which they were gleaned.
Similarly, if one looks hard enough, examples can be found where children
who are forced to state a rule during problem solving are better able to learn and
78 BROWN

transfer (demonstrated with adults in the Gagné and Smith study; however, very
little data is available. One example exists in problem solving situations requir-
ing invention (Resnick & Glaser, 1976), Pellegrino and Schadler (1974). Chil-
dren were required to ‘‘look ahead’’ by verbalizing a sequence of goals, a
procedure that produced a dramatic increase in the number of successful in-
ventions by grade school children. In an on going study, Crisafi and Brown
(work in progress) found greater transfer across problem isomorphs of an in-
ferential reasoning task when, after each problem, three- and four-year-old learn-
ers were required to describe the solution to Kermitt the frog so that he could also
perform the task. Therefore, there is some evidence that verbalizations are help-
ful when they should be; however, there is considerable confusion concerning
when verbalization should or should not be related to cognitive processes (Well-
man, 1983). Questions cannot be addressed, or answered adequately, unless
researchers are precise about the type of verbalization, the type of cognitive
process, and the theoretical rationale for expecting a positive, negative, and
neutral relation between verbalization and the cognitive process.
In summary, there is a desparate need in the developmental literature for
systematic evaluations of children’s verbal reports on their own cognitive pro-
cesses when stringent attention is paid to: (a) the temporal relation between these
reports and the cognition in question; (b) the nature of the cognitions under
evaluation; and (c) the influence of reflection on the operations of thought. It is
simplistic to ask whether or not a certain group of children have reflective access
to their own cognitions without specifying exactly the conditions under which
these observations are made. Ideally, one would like to see programmatic re-
search aimed at uncovering a certain child’s range of understanding within a task
domain. Under what conditions is it reasonable to ask for verbal reports? Can the
child make predictive, concurrent, or retrospective statements about actual or
potential cognitive activity within a problem space? Do the specific restrictions
on adults’ verbalizations, under varying circumstances, apply to children, or do
young learners experience particular difficulties, for example, in imagining pos-
sible actions in situations as yet unexperienced (Brown, 1978)? Do children have
particular problems talking about general rules rather than specific activities?
Can the relation of verbal reports and learning be mapped in terms of predicted
neutral, positive, or negative effects (as Ericsson & Simon have attempted to do
for adults)?
In short, it is necessary to progress from the current piecemeal study of certain
isolated metacognitions, concerning intuitively appealing but seemingly ran-
domly chosen cognitive tasks, to a systematic evaluation of the function of verbal
reports in specific learning situations. Instead of the current tendency to consider
a small range of metacognitive questions in a few, possibly atypical, tasks,
microgenetic case studies are advocated. These studies would involve a small
number of children talking about their cognitive activities in a variety of situa-
tions within a domain. For example, some factorial combinations of predictive,
 3. METACOGNITION AND OTHER MECHANISMS 73

concurrent, retrospective, or imaginative reporting on general or specific rules

could be undertaken. This could be done within a theory that makes prior predic-
tions of neutral, positive, or negative effects of such activities: verbal reports
would not be considered as isolated from the ongoing process of learning. By
examining the conditions and contexts of learning that subjects must reflect on,
the current confusion concerning the relationship of verbal reports and the cog-
nitive operations, which are the subject of these reports, can be eliminated.

Executive Control

Nature of the Problem. The second historical root of things metacognitive is

the notion of executive control taken from the information processing models of
cognition. The majority of information processing models attribute powerful
operations to a central processor, interpretor, supervisor, or executive system
capable of performing an intelligent evaluation of its own operations. Brown
(1978) maintains:
x
' Some form of self-awareness or explicit knowledge of its own workings is critical
for any efficient problem-solving system (Becker, 1975; Bobrow, 1975; Bobrow &
Norman, 1975). The basic requirements of such an executive demonstrate the
complexity of the issue. It must include the ability to (a) predict the system’s
capacity limitations; (b) be aware of its repertoire of heuristic routines and their
appropriate domain of utility; (c) identify and characterize the problem at hand;
(d) plan and schedule appropriate problem-solving strategies; (e) monitor and su-
pervise the effectiveness of those routines it calls into service; and (f) dynamically
evaluate these operations in the face of success or failure so that termination of
activities can be strategically timed. (p. 152)

Thus, very complex operations are attributed to something within the system, a
problem of attribution that is, to say the least, theoretically problematic (Boden,
1977; Dennet, 1978).
Information-processing theories emerged in the mid-1960s along with the
growing interest in computer competence and machine simulation of thought.
The concurrent development of psychological models was greatly influenced by
the theories and jargon of synthetic intelligence; during the past 15 years the
computer metaphor has dominated theories of human cognition. By adopting the
notion of a central processor or executive system imbued with very fancy
powers, developmental psychologists gained a powerful analogy through which
to consider the development of efficient learning. Unfortunately, along with the
notion of a hierarchical cognitive system, complete with a single, god-like super-
visor (Dennett, 1978; Turvey, 1977), they also adopted many of the attendant
problems.
The notion of executive control was in general vogue by the early 1970s
(Greeno & Bjork, 1973) and was introduced into developmental psychology
80 BROWN

(Brown, 1974, 1975). Central within the prototypical information processing

model are the concepts of executive control, implicitly or explicitly stated, and
automated and controlled processes; the notions are interlinked. A description of
the automated-controlled distinction and the executive systems follows.
A two-process approach to thinking, automatic and controlled processing,
predates information-processing models. A notably lucid description of the dis-
tinction between automatic and controlled processes was provided by James
(1890), who stressed the freedom from attention and effort that automatization
provides: ‘‘The more details of our daily life we can hand over to the effortless
custody of automatism, the more our higher powers of mind will be set free for
their own proper work’’ (p. 130). The distinction between automatic and con-
trolled processing, although it masquerades under different titles, is now a com-
mon feature of both the adult and developmental literatures. For example, in the
adult literature are Posner and Snyder’s (1974) ‘‘conscious strategies’’ and ‘‘au-
tomatic activation,’ Shiffrin’s (1975) ‘controlled vs systemic processing,’
Norman and Bobrow’s (1975) “‘resource limited’’ and “‘data limited process-
ing,’ and Laberge’s (1975) ‘‘automatic focusing.’’ Obviously, 1975 was a good
year for this theoretical concept. More recent theories come from Shiffrin and
Schneider (1977), and Logan (1978, 1979). Developmentalists also discuss **de-
liberate and involuntary’’ (Brown, 1975), ‘‘effortful and automatic’? (Hasher &
Zacks, 1979), and ‘‘strategic versus automatic’’ (Naus & Halasz, 1978) proces-
sing.
Automatic processing is a fast, parallel process, not limited by short-term
memory, that requires little subject effort, and demands little direct subject
control. There are two forms of automatic processing: those activities that appear
to be common to all age groups and rarely demand intensive strategic effect, for
example, some forms of recognition (Brown, 1975); and activities that were once
effortful but because of extensive training and experience have become automat-
ic. Controlled processing is a comparatively slow, serial process, limited by
short-term memory constraints, requiring subject effort, and providing a large
degree of subject control (Schneider & Shiffrin, 1977).
Consider a concrete example, skilled reading. The skilled reader can operate
with a lazy, automatic processor. Cognitive activities that once were laborious
and slow have, with practice, become automatized and rapid; they require little
attention. The skilled reader’s top-down and bottom-up (Rumelhart, 1980) pro-
cesses are so fluent that he or she can proceed merrily on automatic pilot until
alerted to a comprehension failure by some triggering event (Brown, 1980:
Collins, Brown, Morgan & Brewer, 1977). If the process is flowing smoothly,
construction of meaning is very rapid, however, when a comprehension failure is
detected the reader must slow down, and allot extra processing capacity to the
problem area. The reader must employ debugging devices and strategies, which
take time and effort. The difference in time and effort between the normal rapid
automatic pilot state and the laborious activity in the debugging state is the
difference between automatic and controlled processing. Whatever the exact
 3. METACOGNITION AND OTHER MECHANISMS 81

nature of the triggering event one reacts by slowing down the rate of processing;
allocating time and effort to the task of clearing up the comprehension failure. In
the process of disambiguation and clarification, the individual enters a con-
trolled, deliberate, planful, strategic state that is quite distinct from the automatic
pilot state, where one is not actively at work on debugging activities. The smooth
flow of reading abruptly stops, and the debugging activities themselves occupy
the lion’s portion of one’s limited capacity processor.
For developmental psychologists, there are many interesting questions con-
cerning automatizations. A major notion is that a great deal of the development
that occurs with increasing expertise (age) is the result of processes that were
originally controlled, effortful, and laborious becoming automated (Brown,
1975; Hasher & Zacks, 1979; Naus & Halasz, 1979). A second well aired notion
is that processes that do not demand strategic control are efficient, even in the
young, and are less sensitive to developmental changes (Brown, 1975; Hasher &
Zacks, 1978). However, in this section, the notion of who or what does the
controlling, and who or what deciphers the output of the automatized system, is
of particular interest.
Heterarchies, hierarchies and demons are involved in attribution. Within the
information processing system, executive power, in large or small degrees, must
be attributed. It is this attribution that causes the models to run into epis-
tomological problems of long standing, problems that are particularly recalcitrant
and uncomfortably metaphysical for a psychology never truly weaned from a
strict radical behaviorist tradition. The major problems are the traditional ones of
consciousness and who has it (1.e., homunculi in their many manifestations). The
problem is nicely stated by Norman (1980) in his inaugural address to the new
Cognitive Science Society:

Consciousness is a peculiar stepchild of our discipline, agreed to be important, but

little explored in research and theory. There are legitimate reasons for this relative
neglect. This is a most difficult topic, one for which it is very difficult to get the
hard, sensible evidence that experimental disciplines require. . . . We cannot un-
derstand (thinking) until we come to a better appreciation of the workings of the
mind, of the several simultaneous trains of thought that can occur, of the dif-
ferences between conscious and subconscious processing, of what it means to focus
upon one train of thought to the exclusion of others. What-who- does the focuss-
ing? . . . And what does it mean to have conscious attention? Can there be atten-
tion that is not conscious? What—who—experiences the results of conscious
attentional processes? (p. 16)

The issue of willed action, and automatic control of behavior is discussed in

depth by Norman (1981), Norman and Shallice (1980), Reason (1979) and
Shallice (1978).
Norman’s self-conscious use of ‘‘who’’ or ‘‘what’’ immediately conjures up
a spectre traditionally feared and derided by psychologists, the ghost in the
machine, the homunculus. Skinner (1971) refers to this entity as the “inner
 82 BROWN

man’’ whose function is ‘‘to provide an explanation which will not be explained
in turn’’ (p. 14). In his article, “‘Behaviorism at Fifty,’’ Skinner (1964) provides
a typical derisive picture of such concepts:

The little man was recently the hero of a television program called ‘Gateway to the
Mind.’ The viewers learned, from animated cartoons, that when a man’s finger is
pricked, electrical impulses resembling flashes of lightning run up the afferent
nerves and appear on a television screen in the brain. The little man wakes up, sees
the flashing screen, reaches out, and pulls the lever. More flashes of lightning go
down the nerves to the muscles, which then contract, as the finger is pulled away
from the threatening stimulus. The behavior of the homunculus was, of course, not
explained. An explanation would presumably require another film. And it, in turn,
another. (p. 80)

Such theories are, indeed, easy to deride, but hard to replace with an alter-
native. As Dennet (1978) points out (see also Boden, 1972, 1977), Skinnerian
outrage at such *’mentalisms’’ can be reduced to the axiom, ‘‘Don’t use inten-
tional idioms in psychology’’ (p. 33). One of the liberations of current theories of
cognitive science is that researchers admit that human beings are intentional, and
that an adequate explanation of human behavior necessitates reference to the
intention, or the meaning of the behavior to the individual who performs it; that
is, the individual’s understanding of what he or she is doing (Boden, 1977;
Brown, 1982a; Dennett, 1978; Flores & Winograd, 1978: Norman, 1980; Shaw
& Bransford, 1977).
How do information processing models deal with the inner man? Most of the
original models were hierarchical, uni-directional systems with a central pro-
cessor initiating and interpreting lower level actions. More recent models tend to
be heterarchical systems that permit lower nodes to feed into upper level nodes so
that control can be distributed throughout the system (Hayes-Roth, 1977; Hayes-
Roth & Hayes-Roth, 1979; Turvey & Shaw, 1979). Heterarchical control is
clearly evident in animal physiological systems (Greene, 1971: Turvey, 1977);
skilled performance, theories of action (Norman, 1981; Norman & Schallice,
1980; Turvey, 1977); and human speech perception (Newell, 1980). This notion
is favored by current artificial intelligence systems (Hayes-Roth & Hayes-Roth,
1979).
Several recent theorists have claimed that heterarchies are the simplest class
of system able to perform processing tasks of the complexity typical of human
behavior (Greene, 1971; Koestler, 1979: Turvey, 1977; Turvey, Shaw & Mace,
1978). These systems do not rid themselves of inner men making decisions; the
demons do not go away. Instead, they are distributed, a democratic solution
that
is much favored in current information processing models (Lindsay & Norman,
1977; Norman & Schallice, 1980), which trace their historical roots to Self-
ridge’s (1959) original Pandemonium model. However, even within the demo-
cratic confederacies or heterarchies, there is a supervisory processor (Lindsay
&
 3. METACOGNITION AND OTHER MECHANISMS 83

Norman, 1977), or a decision demon (Selfridge, 1959) who listens to the pan-
demonium produced by the lower level demons and seiects the most obtrusive.
Conflicts for attentional resources must occur when several subordinate pro-
cessors compete for the same resources. In such systems (McDermott & Forgy,
1978; Norman & Schallice, 1980), some conflict resolution procedure must be
provided.
Central to the issues of metacognition are computer planning models that
attempt to model problem-solving behavior. Only three are discussed here: GPS
(General Problem Solver, Newell & Simon, 1972); NOAH (Nets of Action
Hierarchies, Sacerdoti, 1974); and the recent cognitive model of planning func-
tions of Hayes-Roth and her associates (Goldin & Hayes-Roath, 1980, Hayes-
Roth & Hayes-Roth, 1979, and Hayes-Roth & Thorndyke, 1980). These models
are briefly described.
The concept of planning was first introduced to artificial intelligence by the
programmers of GPS (Ernst & Newell, 1969). The main planning strategy of
GPS was means-end analysis, a hierarchical planning strategy that works back-
wards, from a clear idea of the goal to be achieved. A program using the strategy
employs various comparison procedures to identify the difference(s) between the
goal and the current state. Differences are ordered by priority that is determined
by general purpose heuristics and by domain-specific heuristics. The problem
solver, concentrating on the most important differences first, plans to reduce
them progressively until no differences remain. The simplest situation is where
only one difference exists between the current and goal state, and a single
operator is found that can reduce that type of difference. More typically, no
operator is found that can immediately solve the difference in these cases GPS
establishes subgoals. If, for example, the initial state differs from the goal in
several ways, GPS eliminates the differences, in turn, passing to the next dif-
ference only when the preceding difference has been resolved. GPS is a general
problem solving strategy that can be used in any domain, providing appropriate
information about states and operators is available. Domain-specific heuristics
can also be incorporated if needed.
GPS works well for closed problem systems (Bartlett, 1958) that have well
defined goals that can be reached by fixed means. GPS is a simple state-by-state
planning strategy that does not produce an overall strategic plan of the problem
before the solution is started. This type of decision maker has limited flexibility
in revising and evaluating plans. Sacerdoti (1974) argues, that although it is not
sensible to formulate an epistomologically adequate plan before attempting prob-
lem solution, a broad outline of the plan should be scheduled first, so that the
system can see what adjustments need to be made during execution. Machine
programs, like humans, can not forsee all of the possible contingencies; there-
fore, some form of contingency planning is incorporated in NOAH.
Sacerdoti’s NOAH constructs a preplan that, during execution, can be altered
on a contingency basis; NOAH works by means of a successive refinement
 84 BROWN

approach to planning. However, NOAH is essentially a “‘top-down’’ processor,

the planner making the high level abstract plans that guide and restrict the
subsequent development of low level details. Similarly, the model assumes that
the initial plan is relatively complete and subject to refinement only at lower levels;
therefore, NOAH is essentially a hierarchical planning model. Recently, Hayes-
Roth and Hayes-Roth (1979) developed an opportunistic planning model (OPM)
that departs from the top-down, hierarchical, complete preplan assumptions of
prior planning models. The OPM permits planning at many different levels, and
allows several tentative, incomplete plans to coexist; therefore, it is essentially a
heterarchical system. The OPM has great flexibility; it can opportunistically shift
among several planning levels. At any one point in the planning process, the
planner’s current decisions and observations afford new opportunities for plan
development; these are followed up by the model. Sometimes, opportunistic
decisions result in orderly top-down routes, however, just as in human thinking,
often less coherent sequences are engaged in. The OPM achieves this flexibility by
assuming that the planning process consists of the independent actions of many
distinct cognitive specialists (demons), each able to make tentative decisions for
potential incorporation within a tentative plan. As in the classic Pandemonium
model (Selfridge, 1959), the specialists record these decisions in a common data
structure, known as the blackboard; enabling them to interact with each other, to
influence and be influenced by one another’s decisions.
The blackboard is divided into five conceptual planes that correspond to
different categories of decision. These include: (a) metaplan decisions, dealing
with the general approach to the problem; (b) plan decisions covering what
action to take; (c) plan-abstraction decisions, dealing with the kinds of actions
that are desirable; (d) world-knowledge decisions, noting specific problem en-
vironments; and (e) executive decisions, involved with the organization of the
planning process itself. Each plane on the blackboard is also potentially served at
several levels of abstraction. For example, the levels of the metaplan involve
problem definition, selection of an appropriate problem-solving model, policy
setting, and establishment of appropriate evaluation criteria. Plan-abstraction
levels involve intentions, schemes, strategies, and tactics. The four levels of the
plan plane are outcomes, designs, procedures, and operations. The executive
plane involves decisions of priority, focus, and scheduling. The knowledge-base
plane also has several levels of decisions, but these can only be specificed in
a
context-specific fashion, because knowledge factors are, by definition, problem
specific.
In short, the plan, plan abstraction, and world knowledge decisions determine
features of the developing plan. The executive decisions determine the allocation
of cognitive resources during the planning process (i.e., what kinds of decisions
to generate first, what aspect of the plan to develop next, etc.).
Finally, the
metaplan plane contains general decisions about how to approach
the planning
problem. These high level decisions reflect the planner’s overall understan
ding
of the problem, the methods that he or she will apply, and the criteria
against
 3. METACOGNITION AND OTHER MECHANISMS 85

which plans will be generated, evaluated, scheduled, rescheduled, and given

priority.
The OPM, of Hayes-Roth & Hayes-Roth (1979), is indeed complex; five
planes and three to four levels of decisions within each plane (19 are specified in
the model). Metacognitive—like functions have liberally distributed throughout
the system, where demons with homunculus sounding names also proliferate
(director, referee, top management, middle management, compromisor, archi-
tect, strategist, technician, schemer, inventor, goal-setter, pattern-recognizer,
designer, etc.). The full details of the model are too complex to include here;
however, it is important to note that the OPM is more akin to human planning
than the previous, strictly top-down, single executive, models. OPM is capable
of simulating human planning on realistic problems, such as planning an eco-
nomical route for a series of errands. In order to capture the flexibility and
opportunitistic nature of human planning, the model involves multiple levels of
decisions, of varying degrees of abstraction, taking place on multiple planes. The
very complexity of the model is a reflection of the multifaceted nature of the
problem of executive control within an information processing model. If, as
argued elsewhere (Brown, 1974, 1978), descriptions of executive control func-
tions within systems of machine intelligence are closely parallel to descriptions
of thinking per se, one should not be surprised by the complexity; that the OPM
is so explicit about the planning planes should be seen as an advantage. If one
could describe the types of planning engaged in by adult problem solvers, within
an explicit taxonomy of planning actions, researchers would be in a better posi-
tion to examine the planning procedures of children, in conjunction with devel-
opmental changes in planning functions.
In short, the details of GPS, NOAH, and OPM are not important; instead, the
lesson is that with increasing sophistication, information processing, and ar-
tificial intelligence models have gained more power by paying increasing atten-
tion to the ‘“‘metacognitive’’ aspects of thinking. All models crudely distinguish
between preplanning and planning-in-action (Rogoff & Gardner, in press); plan-
ning and control (Hayes-Roth, 1980); preaction and trouble shooting (Norman &
Schallice, 1980); and planning and monitoring (Brown, 1978). Preplanning in-
volves the formulation of general methods of procedure, prior to the actual onset
of action. During the on-going attempt to solve the problem, there is continual
planning-in-action, trouble shooting, or control processing; involving monitor-
ing, revising, and contingency planning. Both forms of planning, recursively
feed back to each other. Intelligent systems, be they machine or human, are
highly dependent on executive orchestration, resource allocation, and monitoring
functions. Nonintelligent systems, such as the inadequate programs or humans,
are assumed to be deficient in planning functions.

Developmental Data. The central place of executive functions, such as

planning and monitoring, is asserted in most current models of human and
machine information processing, and there is a dearth of clear data, developmen-
86 BROWN

tal or otherwise, that support this assertion. Examples of planning or monitoring

success or failure are often given as incidental reports, following a description of
the main source of interest (i.e., strategy effectiveness). Very few research
programs have concentrated on planning or monitoring processes per se.
In the adult literature, notable exceptions to this rule are the program on
planning in problem-solving by Hayes-Roth and her associates (Goldin & Hayes-
Roth, 1980; Hayes-Roth, 1980); work on expert and novice physics problem
solvers (Chi, Feltovich, & Glaser, 1981; Simon & Simon, 1978); and the recent
work on comprehension-monitoring while reading (Baker & Brown, 1981; in
press). For example, in a recent series of studies, Goldin and Hayes-Roth (1980)
examined the planning strategies of adults. The task was to schedule a series of
errands in a fictious town. The subjects read a scenario that described a series of
desired errands; a starting time and location; an ending time and location; and
sometimes, additional constraints. Invariably, the alloted time was insufficient to
perform all the errands; therefore, the planners were obliged to organize their
schedule in the most economical manner and to set priorities concerning which
errands to perform. The primary data consisted of the protocols of the subject’s
on-line descriptions of their planning processes and the actual route taken.
Good planners made many metaplan and executive decisions and exercised
deliberate control over their planning processes; they also made use of world
knowledge information. Good planners showed greater flexibility than poor
planners; they frequently shifted the focus of their attention among the different
planes of decisions (within the OPM framework) and among the different loci
within the route. More of the decisions of good planners were at a high level of
abstraction; they recognized the importance of global planning, in contrast to the
heavy reliance on local control, or bottom-up plans, shown by the poor planners.
Poor planners tended to switch back and forth between objectives in an idiosyn-
cratic (and often chaotic) fashion. In contrast, effective planners developed a
prototypic procedure for accomplishing the errands, which they maintained over
several instantiations of the task.
Poor problem solvers lack spontaneity and flexibility in both preplanning and
monitoring. More extreme examples of planning deficits in adults come from the
clinical literature on patients with frontal-lobe syndrome. These patients, typ-
ically, omit the initial preaction component (Luria, 1966): they also experience
extraordinary difficulty with error correction (Milner, 1964). Such patients have
been described as simultaneously perservative and distractible, a failure in intel-
ligent focusing attributed to damage to the supervisory attentional mechanism, or
executive system (Norman & Schallice, 1980). Although pathological cases are
extreme, many descriptions of young and retarded childrens’ learning are very
similar. This superficial similarity is intriguing and deserves attention.
The dearth of systematic research specifically concerned with children’s plan-
ning behavior is quite dramatic, although, again, there are many examples in the
literature where failure to plan, monitor or check are described incidentall
y
 3. METACOGNITION AND OTHER MECHANISMS 87

(Brown & DeLoache, 1978). Brown, Bransford, Ferrara and Campione (1983)
discuss some of the incidental cases of very young children planning ahead for
future retrieval attempts.
A good example is provided by Karmiloff-Smith (1979b) who asked children
4-9 years of age, to construct a representation of closed circuit railways, using
cardboard tracks. The children were well aware that they had to (a) use all the
available pieces; (b) fit the tracks together (flush); and (c) make a closed circuit,
so that the train could go around and not leave the tracks. The younger children
started the task immediately, constructing the circuit by simply picking up tracks
as they came to hand and placing them next to each other. No preplanning,
followed by systematic search for a specific shaped piece (a curve or a straight)
was shown. Older children, on the other hand, spent considerable time in the
preplanning phase, piling tracks into stacks, on the basis of identity, and system-
atically selecting from the correct pile the desired shaped (i.e., curved, straight,
curved, straight, curved, straight, etc. or two straight, two curved, two straight,
two curved, etc.) in order to complete the master plan.
A similar example of planning failure in young children was reported by
Kluwe (chapter 2, this volume). Children from 4—7 years of age were asked to
work on a series of reversible puzzles. A reversible puzzle was arranged so that
the children could correct their errors. For example, the task might be to copy a
model of a ship using pieces of colored paper (shapes). If a piece was placed in
the wrong position it could be corrected. After completing some reversible
puzzles, the children were given irreversible puzzles. The base on which the
puzzle pieces had to be assembled was covered with adhesive, so that the
finished construction would be permanent and could be displayed as a picture on
the wall. Even the youngest children, by spending more time looking back and
forth between the model and the pieces before beginning construction, showed
some signs of planning for the new contingency. However, only the older chil-
dren were really clever and adopted strategies, such as constructing the entire
puzzle first on a safe (reversible) surface, before transferring the pieces to the
irreversible surface.
Most of the developmental descriptions of planning in children come from
situations that have not been generated by information processing models of
children’s learning. What is needed in this area is programmatic work on pre-
planning and planning-in-action within a framework that takes into consideration
other factors of the task environment, such as task familiarity and processing
load. An excellent example of such an approach is Shatz’s (1978) reconcep-
tualization of the referential communication literature, within an information
processing framework. Whether or not children will plan ahead and take the
listener’s perspective into account depends on the other processing demands of
the task.
There is surprisingly little data on children’s executive control while per-
forming a task, or the planning-in-action phase; surprising, because of the pro-
88 BROWN

liferation of speculation concerning the centrality of monitoring processes to

effective learning. Notable exceptions are the work on children’s monitoring of
oral and written communications (Flavell, Speer, Green & August, 1981; Mark-
man, 1977, 1979; Stein & Trabasso, 1982). In general, children have difficulty
detecting inconsistencies in messages, unless those inconsistencies are particu-
larly blatant or salient (for a review see Baker & Brown, 1981b).
One example of executive monitoring during learning was provided by Bel-
mont & Butterfield (1977). Adolescent learners were observed as they devised a
systematic rehearsal plan for acquiring an eight item list of words. Over a series of
ten trials the strategy became firmly established. The mature pattern to emerge in
older subjects was the familiar cumulative rehearsal, fast-finish pattern. The
subjects rehearsed the first words in the list cumulatively, pausing at approx-
imately position five to establish the first set in memory, then they viewed the
remaining three items very rapidly before risking a test. The subject’s naive
memory theory agreed with the psychologists’ task analysis. The first part of the
list must be rehearsed repeatedly and cumulatively to ensure memorability; the last
items are viewed rapidly in the hope that they will not have faded before the test
occurs.
After the subjects had established their rehearsal strategy, the experimenters,
without warning, presented the same list of words repeatedly for eight trials. In
the face of the changing task demands, older subjects rapidly abandoned their
rehearsal strategy; younger children took more time to revise their study activity.
In a subsequent test, when a repeated list was presented with two changed words,
the older students were again more efficient at distributing their study time
appropriately (Belmont & Butterfield, 1977). A similar sensitivity to study time
apportionment, during studying, has been shown by children as young as third
grade for simple repetitive lists of pictures (Masur, McIntyre & Flavell, 1973).
In contrast, when the task is to monitor degree of learning from texts, adequate
monitoring is not displayed until well into the high school years (Brown, Smiley
& Lawton, 1978; Brown & Campione, 1978). The influence of task difficulty
and familiarity must be considered in any examination of the child’s propensity
to plan and monitor his or her own activities.

Self Regulation
Nature of the Problem. Any active learning process involves continuous
adjustments and fine-tuning of action via self-regulating processes and ‘‘other
even more mysterious apparatus’’ (Marshall & Morton, 1978 p. 227). Psychol-
ogists interested in mechanisms of growth and change have traditionally
been
concerned with self-regulating processes, because a great deal of learning takes
place in the absence of external agents. However, substantial contributio
ns are
also made by external agents; these are discussed in the next section.
It is
certainly the case that human thinkers ““play’’ with thinking (Gardner,
1978),
that is, they subject their own thought processes to examination and
treat their
 3. METACOGNITION AND OTHER MECHANISMS 89

own thinking as an object of thought. Similarly, learners regulate and refine their
own actions; sometimes this is done in response to feedback concerning errors,
but often it is done in the absence of such feedback. Indeed, even if the system
being experimented with is perfectly adequate, active learners will improve on
their original production (Karmiloff-Smith, 1979a, 1979b).
Recently, the term metacognition has been extended to encompass regulatory
functions, such as error detection and correction (Brown and DeLoache, 1978;
Clark, 1979); the historical roots of these concepts can be found in most of the
major developmental theories. For example, Binet, a pioneer in the empirical
study of cognition and intelligence, was fascinated by individual differences in
his daughters’ cognitive styles of self-regulation (Binet, 1890, 1903); following
intensive study with both normal and retarded children, he selected autocriticism
(Binet, 1909) as a central component of intelligence.
In this section primary consideration is given to relatively recent Genevan
research on self-regulatory mechanisms in children’s thinking and on the grow-
ing emphasis in developmental psycholinguistics on error correction, systemiza-
tion, and metalinguistic awareness.

Piaget's Theory of Regulation. In the latter part of his career, the transfor-
mational period (Brown, 1979; Riegel, 1975), Piaget became more and more
interested in mechanisms of learning and the influence of both conscious and
unconscious regulatory functions in promoting conceptual change. Space re-
strictions, prevent the author from describing Piaget’s complex theory; instead,
the reader is referred to Gelman and Baillargeon (1983). Very briefly (and
probably too simplistically), Piaget (1976) distinguishes between three primary
types of self-regulation: autonomous; active; and conscious regulation.
Autonomous regulation is an inherent part of any ‘‘knowing act.’’ However
small the learner and however simple the action (Bruner, 1973; Koslowski and
Bruner, 1972), learners continually regulate their performance, fine-tuning, and
modulating actions. Active regulation is more akin to trial and error, where the
learner is engaged in constructing and testing “‘theories-in-action’’ (Karmiloff-
Smith & Inhelder, 1974-1975). Under the guidance of a powerful theory-in-
action, the learner tests a current theory via concrete actions that produce tangible
results. Only when a current theory and its range of applicability are confirmed and
consolidated is the learner ready to recognize that there are counterexamples that
have regularity. At this point, some unifying principles for these regularities can
be actively tested. Not until a much later stage can the learner mentally construct
and reflect on the hypothetical situations, which would confirm or refute a current
theory without the need for active regulation. Conscious regulation involves the
mental formulation of hypotheses capable of being tested via imaginary confir-
matory evidence or counterexamples.
Thus, for Piaget, self-regulation, error correction, trial-and-error, theory test-
ing, etc., need not be conscious experiences, but may occur in the plane of
action; however, the highest level of theory building and testing is conscious.
90 BROWN

Consciousness permits mental reasoning that liberates the regulatory functions

from dependence on active testing (Piaget, 1978).
In The Grasp of Consciousness (1976) and the companion volume, Success
and Understanding (1978), Piaget considered degree of consciousness, when he
turned his attention to the gaps that exist between succeeding in action, being
capable of explaining these actions, and being capable of abstract thought that
drives action sequences. According to Piaget (1976):

Action in itself constitutes autonomous and already powerful knowledge. Even if

this knowledge (just knowing how to do something) is not conscious in the sense of
a conceptual understanding, it nevertheless constitutes the latter’s source, since on
almost every point the cognizance (consciousness) lags, and often markedly so,
behind this initial knowledge, which is of remarkable efficacy despite the lack of
understanding. (pp. 346-347)

To encapsulate Piaget’s developmental progression: the initial stage of auton-

omous regulation involves unconscious adjustments and fine-tuning of motor
actions; next the child becomes capable of testing out theories-in-action, via
concrete trial and error. Despite the lack of conscious surveilance on the part of
the learner, this active regulation can lead to successful problem-solving. Even
though the learner cannot describe how they were accomplished actions can be
successfully completed.
Consciousness first emerges as the child becomes capable of reflecting on his
or her own actions in the presence of the actual event. According to Piaget
(1976); ‘‘When the child was asked how he came to discover a specific pro-
cess . . . the younger subjects merely recounted their successive actions (or at
the beginning merely reproduced them with gestures and no verbalizations), the
older children said for instance, I saw that . . . so I thought . . . or so I had the
idea . . . and so on (p. 337). At this initial stage, reportage is tied to concrete
action, but does not direct it. Because consciousness is not directly linked to
conceptualization, the child’s *‘reactions remain elementary, the subject is likely
to distort conceptualizations of what he observes, instead of recording it without
modification.’’ Such distortion can be quite dramatic; for example, having wit-
nessed an event that is contrary to a tenaciously held belief, the "subject contests
the unexpected evidence of his own eyes and thinks that he sees what he pre-
dicted would happen’’ (Piaget, 1976, p. 340).
At the most mature level, which Piaget would prefer restricted to the stage of
formal operations, the entire thinking process can be carried out on the mental
plane. The learner can consciously invent, test, modify, and generalize theories
and discuss these operations with others.

Finally, at the third level (from 11 to 12 years) which is that of reflected abstractio
n
(conscious products of reflexive abstraction) the situation is modified
in that cog-
nizance [consciousness] begins to be extended in a reflexion of
the thought it-
 3. METACOGNITION AND OTHER MECHANISMS 91

self . . . This means that the subject has become capable of theorizing and no
longer only of ‘concrete’, although logically structured, reasoning. The reason for
this is the child’s new power of elaborating operations on operations. . . . He
thereby becomes capable of varying the factors in his experiments, of envisaging
the various models that might explain a phenomenon, and of checking the latter
through actual experimentation. (Piaget, 1976, pp. 352-353)

In brief, the developmental progression is from unconscious autonomous

regulation to active regulation, in the absence of anything more than a ‘‘fleeting
consciousness.’’ The beginning of conscious reflection occurs when the child is
capable of considering his or her actions and describing them to others, albeit
sometimes erroneously. The mature level of reflected abstractions, however, is
characterized by conscious processes, that can be carried out exclusively on the
mental plane coming to direct learning. Mature learners can create imaginary
worlds and theories to explain actions and reactions within them. Such theories
can be confirmed or refuted via the further construction of mental tests, conflict
trials, or thought experiments that extend the limits of generality of the theory.
This is the essence of scientific reasoning and the end state for a Piagetian
development progression of child as scientist.

Metaprocedural Reorganization and Systemization. Piaget’s colleagues, In-

helder and Karmiloff-Smith, have introduced another concept relevant to the
discussion of self-regulation, metaprocedural reorganization (Karmiloff-Smith,
1979a; Karmiloff-Smith and Inhelder, 1974-1975). The basic idea is that learn-
ing within a domain follows a predictable sequence that is characterized by
internal pressure to systematize, consolidate, and generalize knowledge. Repre-
sentational systems are first developed to handle pieces of the problem space
adequately, however, several theories can exist side-by-side in juxtaposition, not
fully integrated into a comprehensive system. Karmiloff-Smith (1979a) main-
tain: ‘‘. . . each time children develop an adequate tool for representing their
knowledge, and once the tool functions well procedurally, then that tool is
considered metaprocedurally as a problem space in its own right’’ (p. 92). Thus,
the theory (tool) that is directing action, itself becomes an object of thought and
experimentation.

—} Stepping-up to a Metaprocedural level appears to take place each time the child has
a handle on his currently functioning system. This enables the child to treat the
system, or parts thereof, as units functioning in their own right. Thus, each time a
representational tool (theory) functions well procedurally, the child somehow takes
it apart to analyze its implicit components and thereby the representational tool
becomes part of the problem space itself. (Karmiloff-Smith, 1979a, p. 115)

The prototypic microgenetic sequence is that the child first works on develop-
ing an adequate theory for a salient aspect of the problem space. This theory is
 92 BROWN

practiced and perfected until it is fully operational. During the initial period the
child cannot attend to other parts of the problem space, only when the initial
theory is consolidated and functioning efficiently can the child step back and
consider the system as a whole. Typically, the child will develop several jux-
taposed theories that are adequate for various parts of the problem space; each
theory operating in isolation from the other. Once the procedures are functioning
well, the next stage of development is possible, and the child ‘‘steps-up’’ and
metaprocedurally reconsiders the problem space. Once children become aware of
the discrepancies or contradictions resulting from the simultaneous existence of
several different functioning units, they begin to attempt to reconcile the dif-
ferences and obviate contradictions resulting from the juxtaposition (Inhelder,
Sinclair and Bovet, 1974).
A concrete example might help to clarify this complicated theoretical notion.
Karmiloff-Smith and Inhelder (1974-1975) asked 4 to 9-year-old children to
balance rectangular wooden blocks on a narrow metal rod fixed to a larger piece
of wood. Length blocks had their weight evenly distributed; the correct solution
was to balance them at the geometric center. Weight blocks had the weight of
each *‘side’’ varied either conspicuously (by gluing a large block to one end of
the base rectangle) or inconspicuously (by inserting a hidden weight into a cavity
on one end of the rectangle).
At first, the children made the blocks balance by brute trial and error, using
proprioceptive information to guide their action. Behavior was purely directed at
the goal of balancing. This ploy was obviously successful, and the child bal-
anced each block in turn. There was no attempt to examine the properties of the
objects that led to balance and no attempt to subject each block to a test of a
unified theory.
This early, but unanalysed, phase, without error, was supplanted by the
emergence of a strong theory-in-action. The theory was directed at uncovering
the rules governing balance in the miniature world of these particular blocks.
Unfortunately, it was an incomplete rule that produced errors. A common theo-
ry, developed early by the children, was to concentrate exclusively on the geo-
metric center and attempt to balance all blocks in this fashion. This worked
for
unweighted blocks; however, when the theory did not result in balance (produced
errors) the blocks were discarded as exceptions (“impossible to balance’’).
After
the theory was well established and working well for length blocks, the
children
became discomforted by the number and regularity of errors. A new
juxtaposed
theory was then developed for the conspicuous weight blocks.
The children
compensated for the weight, obviously added to one end, and adjusted
the point
of balance accordingly. However, for a time, length and weight
were considered
independently. Length blocks were solved by the geometric
center rule and
conspicuous weight blocks were solved by the estimate-weight-first
and then
compensate rule. Inconspicuous weight problems still generated
errors. The
blocks looked identical to the unweighted blocks and were,
therefore, subjected
 3. METACOGNITION AND OTHER MECHANISMS 93

to the dominant geometric center rule; they were discarded as anomalies that
were “‘impossible to balance.’’ The children’s verbal responses reflected these
juxtaposed solutions, exclusively length justifications were given for unweighted
blocks and weight justifications were given for conspicuously weighted blocks.
Gradually and reluctantly, the children entered the period of metaprocedural
reorganization, only possible when both juxtaposed procedures were working
smoothly. Now, the young theorists, made uncomfortable by the remaining
exceptions to their own rules, began to seek a rule for them. In so doing, a
metaprocedural reorganization was induced that resulted in a single rule for all
blocks. The children abandoned the simple theories and reorganized the problem
space so that a single unifying theory predominated. Now, the children paused
before balancing any block, and roughly assessed the point of balance. Verbal
responses reflected their consideration of both length and weight (e.g., ‘‘You
have to be careful, sometimes it’s just as heavy on each side and so the middle is
right, and sometimes it’s heavier on one side.’’) After inferring the probable
point of balance, and only then, did the child place the block on the bar.
There are three main points to note about this example. First, there is the
finding of a developmental lull, or even a seemingly retrogressive stage, when
errors predominate. Initially, the children made no errors, all blocks were bal-
anced. However, during the quest for a comprehensive theory of balance, the
children generated partially adequate procedures that resulted in errors. Only
when the unifying theory was discovered did the children revert to perfect perfor-
mance. If errors alone formed the data base, a curvilinear developmental growth
curve would be apparent (Strauss and Stavey, 1982). Actually, what is happen-
ing is that the children are analyzing the problem space to generate a theory that
would account for behavior. In so doing, they made what looked like errors, but
what are often tests of the existing theory (Bowerman, 1981; Karmiloff-Smith,
1979a).
The second main point is that metaprocedural reorganization, leading to a
“‘stepping-up’’ in theory complexity, is only possible when the partially ade-
quate, juxtaposed systems are well established. It is essential that the child gain
control of simple theories in his or her quest for a more complex adequate theory.
Karmiloff-Smith and Inhelder (1974-1975) refer to this as creative simplifica-
tion.

The construction of false theories or the overgeneralization of limited ones are in

effect productive processes. Overgeneralization, a sometime derogatory term, can
be looked upon as the creative simplification of a problem by ignoring some of the
complicating factors (such as weight in our study). This is implicit in the young
child’s behavior but could be intentional in the scientist’s. Overgeneralization is not
just a means to simplify but also to unify; it is then not surprising that the child and
the scientist often refuse counter-examples since they complicate the unification
process. However, to be capable of unifying positive examples implies that one is
equally capable of attempting to find a unifying principle to cover counterexam-
 94 BROWN

ples . . . [there is] a general tendency to construct a powerful, yet often inappropri-
ate hypothesis which [learners] try to verify rather than refute. This temporarily
blinds the [learner] to counterexamples which should actually suffice to have them
reject their hypothesis immediately. (p. 209)

Progress comes only when the inadequate theory is well established and the
learner is free to attempt to extend the theory to other phenomena. In this way the
theorists, be they children or scientists, are able to discover new properties,
which in turn make it possible for new theories to be constructed.
The third main point is that metaprocedural reorganization is not solely the
response to external pressure or failure, rather it occurs spontaneously when the
child has developed well functioning procedures that are incomplete, but ade-
quate for the task at hand. It is not failure that directs the change, but success,
success that the child wishes to extend throughout the system.
A similar U-shaped developmental pattern has been observed in children’s
language acquisition (Bowerman, 1987; Karmiloff-Smith, 1979a; Nelson &
Nelson, 1979). The phenomena under consideration are ‘‘errors’’ in children’s
spontaneous speech. The particular errors of interest are those that are preceeded
by a period of correct usage; hence, they are referred to as ““late errors’’ (Bower-
man, 1984). For example, consider the child’s use of plural (s) and the past tense
(ed) morphemes. The typical developmental progression is that children produce
correct instances of the plural and past tense of both the regular (dog s, cat s, tree
s, or walk ed, jump ed, climb ed) and irregular (mice, feet, went, broke) forms
(Bowerman, in press). Next, the irregular pattern is replaced by an incorrect,
overgeneralization of the regular form (foots, mouses, goed and breaked).
Eventually the correct forms reappear.
The explanation for this U-shaped development is that the original correct
usage was due to the child having learned the irregular (as well as regular) forms
as individual cases. Repeated experience with the regular pattern causes the child
to recognize the systematicity involved, abstract the general rule, and apply it too
broadly to all plurals (hence, mouses) or all past tense forms (hence, goed,
breaked); errors occur, whereas, previously there had been no errors. When
the
system is fully established, the child is ready to admit exception to the
dominant
rule and the exceptions reappear; however, this time they are part of an integrated
theory, regarded as being exceptions to the rule, not just isolated
forms.
Karmiloff-Smith’s (1979a) description of French children learning
to use
definite and indefinite articles is another case in point. Again, a curvilin
ear trend
was found: the children first used the form correctly, then incorrect
ly, then again
in the appropriate manner. In French the indefinite article (un
or une) has a dual
function; it can either refer to nonspecific reference or act as a numeral. In
English this distinction is marked at a surface level: a man
(nonspecific refer-
ences); one man (numeral). At approximately 3 years of
age, children appear to
be using the adult system correctly, using un in both appropri
ate senses. At a
 3. METACOGNITION AND OTHER MECHANISMS 95

somewhat later stage the child enters a phase when he or she uses two distinct
forms for the two distinct meanings. For example, ‘‘une voiture,’’ for a car, and
“une de voiture,’” for one car. The second is ungrammatical usage; however, the
child finally reverts to the correct use. At a certain point in development, the
child seems to recognize that a single work in his or her repetoire, in different
contexts, functions in different ways. This recognition leads him or her to tem-
porarily create new (ungrammatical) forms that differentiate between the sepa-
rate functions and conserve their separate meanings. The child at the intermedi-
ate stage needs to render tangible the distinctions in meaning by marking them
externally. In the process of analyzing the system the child produces more errors.
Only later will the child allow one external marker to convey many pieces of
information.

Levels of Self-Regulation. In this brief and oversimplified synopsis of latter

day Genevan psychology and language acquisition data, in theoretical specula-
tion a central place is afforded to the concept of self-regulation. There is basic
agreement that self-regulatory functions are integral to learning and are central
mechanisms of growth and change. Similarly, the notion of self-regulatory
mechanisms has a central place in the emergent field of metacognition (Brown &
Deloache, 1978).
It is agreed that there are many degrees of self-regulation and that self-
regulation is essential for any ‘‘knowing act.’’ However, it is important to note,
that a sharp distinction is made in both theories of language acquisition and in
Genevan psychology, a distinction that has not been made as clear in the meta-
cognitive literature. The distinction is between conscious awareness and direc-
tion of thought, and self-correction and regulation that can proceed below
the level of consciousness. For example, Karmiloff-Smith (1979a) is quite in-
sistent that ‘‘reference is being made to spontaneous metaprocedural behavior
rather than to explicit awareness’’ (p. 92); and, developmental psycholinguists,
such as Bowerman (1981) and Clark (1979) describe error detection and correc-
tion as an implicit process. “‘Most of the time this is done fluently and com-
pletely unconsciously’? (Bowerman, 1981). In sharp contrast is the notion of
metalinguistic awareness (Gleitman, Gleitman, & Shipley, 1972), where the
child is capable of consciously reflecting on his or her own language and the
language of others.
Piaget is even more emphatic when he distinguishes between success, which
means ‘‘having enough understanding of a situation to attain the requisite ends in
action’’ and understanding, which means ‘‘successful mastery in thought of the
same situation to the point of being able to solve the problem of the ‘how’ and
the ‘why’ of the connections observed and applied in action’’ (Piaget 1978, p.
218). Piaget’s mature stage of reflected abstraction is distinguished by the ““tran-
scendance of action by conceptualization’’; understanding comes “‘out of the
realm of successes into the realm of reason’’ (1978, p. 218). In a normal
 96 BROWN

developmental sequence, success at an action level considerably precedes under-

standing, however, with reflected abstraction the “‘situation is reversed.’’ Con-
ceptual thought drives success in action.
Thus, Piaget distinguishes sharply between active regulation as part of any
knowing act and conscious regulation and direction of thought. The first process
is age independent; even the young learner succeeds in action by regulating,
correcting, and refining his or her current theories. The second process, guided
by reflected abstraction, is late developing, and indeed, for Piaget, this is the
keystone of formal operational thought.
Similarly, error correction during language production is intergral to the pro-
cesses of using language and is present no less in young children (Bowerman,
1981; Clark, 1979) than in adults (Fromkin, 1973; Nooteboom, 1969). In con-
trast, metalinguistic awareness is assumed to be late developing, and in a mature
form, it is also assumed to be a product of adolescent rather than childhood
thinking. The ability to step back and consider one’s own thought (or language)
as an object of thought and, to go further, use the subsequent conceptualization
to direct and redirect one’s cognitive theories, is late developing. Confused in the
metacognitive literature, even lost in some versions of the concept, is the essen-
tial distinction between self-regulation during learning and mental experimenta-
tion with one’s own thoughts. Whatever distinctions must be made, in order to
render metacognition a more malleable concept, this one is a fine candidate for
inclusion in the list.

Developmental Data. The separation of theory and data that seemed neces-
sary in the preceding sections is somewhat artificial in this section. In previous
sections, when discussing verbal reports as data or executive control, the histor-
ical roots and current theories were gleaned from the adult literature. In this
section, on self-regulation, the theories were designed to explain developmental
phenomena and the data base was gathered from children. For symmetry, how-
ever, the separation is maintained; however, for brevity only one example of:
(a) early error correction; (b) metaprocedural reorganization in the absence of
errors; and (c) conscious direction of problem solving are included.
Consider first error correction; the important point is that even very young
children are capable of regulating their activities via a systematic procedure of
error detection and correction. In a recent study, DeLoache, Sugarman
and
Brown (1981) observed young children (24—42 months) as they attempted
to
assemble a set of nesting cups. The number of successful seriators increased
with
age, however, there were no age differences in the number of errors produced
or
the probability of correcting errors. Children in this age range did
not differ in
their likelihood of correcting a set of nonseriated cups; they did however
differ in
their strategies for correction.
The most primitive strategy, used frequently by children below
30 months,
was brute force. When a large cup was placed on a smaller one,
the children
would press down hard on the nonfitting cup. Twisting and banging
were vari-
 3. METACOGNITION AND OTHER MECHANISMS 97

ants of brute pressure, but the same principle held; the selected cup would fit, if
only one could press hard enough. Older children also used the brute force
approach, but only after an unsuccessful series of maneuvers; for them, brute
force appeared to be a last resort.
A second strategy initiated by some of the younger subjects was trying an
alternative. After placing two nonfitting cups together, the child removed the top
cup and did one of two things, he or she either looked for an alternative base for
the nonfitting cup or tried an alternative top for the original base. Both ploys
involve minimal restructuring and necessitate considering the relation between
only two cups at any one time. The third characteristic ploy of children below 30
months was to respond to a cup that would not fit into a partially completed set of
cups by dismantling the entire set and starting again.
Older children (30-42 months) faced with a nonfitting cup engaged in strat-
egies that involve consideration of the entire set of relations in the stack. For
example, one sophisticated strategy was insertion; the children took apart the
stack at a point that enabled them to insert the new cup in its correct position. The
““error’’ is corrected without the need to dismantle work already accomplished.
To do this, however, the child must be cognizant of multiple relations between
the series of cups, not just the single relation between two cups at a time. A
second strategy reversal was also shown by older children. After placing two
nonfitting cups together, the child would immediately reverse the relation be-
tween them (5/4 immediately switched to 4/5).
These rapidly executed strategies of insertion and reversal shown by the older
children were often quite dramatic in their absence in the younger group. Some
young children would repeatedly assemble, for example, cup 4-1, starting with 4
as a base and then inserting 3, 2, and 1. Then they encountered the largest cup, 5,
and attempted to insert it on top of the completed partial stack, pressing and
twisting repeatedly. When brute force failed, they would dismantle the whole
stack. Similarly, having selected 1,2,4, and 5 and then encountering 3, the
younger childrens’ only recourse was to begin again.
The DeLoache, Sugarman and Brown (1981) study of self-correction in
young children is used as one example (see also Koslowski and Bruner, 1972) of
the obvious fact that even very young children correct their errors while solving a
problem. Of more interest is the demonstration that the child’s error correction
strategies provide a window through which the child’s theories-in-action can be
viewed. The very processes used to correct errors reflect the level of understand-
ing the child has of the problem space. Similarly, developmental psycholinguists
have argued that production errors, such as “‘slips of the tongue’’ are very
informative, ‘‘the tongue slips into patterns’? (Nooteboom, 1969). Such errors
reveal a great deal about the organization of the semantic space or cognitive
organization of the speaker (Bowerman, 1981).
Metaprocedural reorganization, in the absence of external pressure, feed-
back, or correction, is the problem solving principle examined extensively by
Karmiloff-Smith. One clear example is the development and refinement of a
 98 BROWN

notational system by children drawing maps of a journey (Karmiloff-Smith,

1979a). Children between 7 and 11 years were shown a roll of paper (12 meters)
on which a long winding road with 20 choice points (bifurcations) was drawn. At
each choice point one road led to a dead-end; the other towards the destination,
which was a hospital. The child’s task was to prepare to drive a patient from his
house to the hospital as quickly as possible. The child was permitted a practice
drive over the route; during the practice drive only one choice point is visible at
any one time (as the experimeter unrolls the paper gradually). The child’s pre-
paratory task is to take notes to help remember the route, which way to turn at
each of the 20 choices.
The details of the notational systems the children developed, although fas-
cinating, are not the major concern. Instead, of interest is the systematic modifi-
cation made to the systems in the course of the journey. All children initially
developed an adequate form of notational systems; therefore, changes in the
system were not made in response to errors, or corrective feedback. During a one
hour session, 70% of the sample made systematic changes to their notational
system, changes that bear a striking resemblance to modifications in language
production, described earlier (Karmiloff-Smith, 1979a). Two typical changes
involved progressive economy and conventionalization of marking. For exam-
ple, the child might begin by writing, ‘‘I will turn to the right, I will turn to the
left,’ and then economize to ‘’left,’’ ‘‘right,’’ ‘‘left.”’ Similarly, a child might
begin by drawing a detailed depiction of a topographical landmark on the map,
such as a house, and then conventionalize these marks by using a line drawing to
symbolize houses or trees, etc.
However, the systematic change of most interest is progressive plurifunc-
tionality. The child would begin by using a notational symbol, which carries all
the necessary information. For example, a topographical marker might be used
both to identify the particular choice point (landmark function) and to indicate
the correct road (directional function), that is, a little house would be drawn on
the correct fork of a bifurcation. Gradually, however, the child moves from using
a form that carries all the necessary information to externally mark each piece of
information with a separate symbol. It is as if once the child becomes aware of
the relevant pieces of information to be conveyed, he or she needs to mark them
externally in a tangible form. Finally, once the system is fully and smoothly
operating, the child reverts to the economical procedure of using one form
to
have multiple meanings. For example, one child began by marking the
choice
point by drawing a topographical sign that simultaneously indicated
the particu-
lar bifuration and the road to take. These functions were separated
when the child
indicated the direction of choice with an arrow on the correct turn of
the bifurca-
tion and, separately drew the topographical marker to indicate
the particular
choice point. Finally, she came to use a topographical marker
which carried a
plurifunctional meaning by drawing it either to the left or right
of the page (e.g.,
to indicate left at tree or right at house).
 3. METACOGNITION AND OTHER MECHANISMS 99

The interesting fact about these changes is that they take place in the absence
of any errors or problems that should make the child dissatisfied with the original
notation, and the types of modification that occur are regular. Karmiloff-Smith
(1979a) argues that the tendency toward

. . . External explication of implicit distinctions has a dynamic psychological func-

tion. It allows the child to get a grip on the newly semanticized distinction, each
piece of information thereby rendered tangible by its own external handle. . . Once
the child becomes aware of the various distinctions to be indicated, he must first
work at each one separately in a juxtaposed fashion . . . hence the need for sepa-
rate, external markers. . . . Gradually, this tangible external marking is replaced
by integration into a more abstract plurifunctional system. (p. 115)

The progress to conscious regulation of problem solving via thought experi-

ments, hypothesis testing, and reflected abstraction is well illustrated in Anzai
and Simon’s (1979) analysis of an adult, systematically refining procedures for
solving a five disc Tower of Hanoi problem. Within a single session the subject
progressed through three stages. First, she was concerned with the goal of
completing the task and therefore looked very like the young block balancers
recorded by Karmiloff-Smiih and Inhelder (1974-1975). In the intermediate
stage she became theory driven, seeking to understand the principles behind the
task, guiding herself explicitly by mentioning intermediate goals, and pausing
after each goal had been reached to plan for the next goal. In this phase theories-
in-action were being created and tested. In the third phase, the subject shows
Piaget’s ‘‘reversed order,’’ the ‘“‘transcendence of action by conceptualization,”’
‘‘reflection directing action.’’ Before undertaking to solve the puzzle again, the
subject tested her understanding by reviewing the moves of the component one,
two, three, and four disc problem. In so doing, she explicitly stated the main
principle of recursivity, and the essential notion of the transfer of pyramids of
discs (Anzai and Simon, 1979).
The microgenetic learning route followed by this adult subject is recapitulated
macrogenetically in Piaget’s protocols of The Hanoi Tower. In the early stage of
solution, children complete a three disc problem by trial-and-error, without being
conscious of the principles. None of the younger subjects (4- to 5-year-olds)
made a plan, or were able to predict how they were going to move the tower.
After the fact, their justifications and explanations were noninformative. In an
intermediate stage correct solutions became stable for three disc problems, there
was evidence of planning ahead and the beginning of the ability to describe the
procedures used during a successful attempt.
The final stage (approximately 11 years of age) was characterized by rapid
and stable success on three disc problems and increasingly inferential anticipa-
tion of the rules for solving five disc problems. Having completed a four disc
problem, one child asked to predict how to solve a five disc problem responded:
100 BROWN

‘‘Theres one more, you have to make more moves, otherwise it’s the same
system — you always take away the smaller one, then the middle one, then you
put the small one on the middle one and you can get at the bigger one; that makes
a small pyramid there, and then the way is clear to do it all again. I can start all
over again; it’s the same story afterward’’ (Piaget, 1976, p. 298).
By Piaget’s Stage III, the child’s understanding of the principles of recursivity
and the pyramid subgoal strategy (Anzai and Simon, 1979) is not only fully
articulated, but it directs the subsequent problem solving attempt. The entire
procedure can be corrected, examined, and revised in thought, before it is at-
tempted in action. In Piaget’s system, this is the essence of conscious control of
action, or reflected abstraction.

Other-Regulation
Nature of the Problem. In this final section the notion of other-regulation, as
it pertains to the general topic of metacognitive development, is discussed.
Important as the processes of self-regulation may be, a great deal of learning
occurs in the presense of, and is fostered by, the activity of others. Supportive-
others, such as parents, teachers, peers, guide a novice to mastery; there seems to
be a systematic regularity in how this guidance works.
A great deal of the work conducted on other-regulation has taken place within
the framework of Vygotsky’s (1978) theory of internalization. Vygotsky argues
that all psychological processes are initially social, shared between people, par-
ticularly between child and adult; and that the basic interpersonal nature of
thought is transformed through experience to an intrapersonal process. Thus, for
Vygotsky, the fundamental process of development is the gradual internalization
and personalization of what was originally a social activity. According to
Vygotsky (1978):

We propose that an essential feature of learning is that it creates the zone of

proximal development; that is, learning awakens a variety of developmental pro-
cesses that are able to operate only when the child is interacting with people in his
environment and in cooperation with his peers. Once these processes are inter-
nalized, they become part of the child’s independent developmental achievement.
(p. 90)
Social settings, where the child interacts with experts in a problem solving
domain, are settings where a great deal of learning occurs. Indeed, some would
argue that the majority of learning is shaped by social processes. (Laboratory for
Comparative Human Cognition, in press). A great deal of this learning involves
the transfer of executive control from the expert to the child; thus, for Vygotsky,
the development of cognitive control is very much a social process. Children first
experience active problem solving activities in the presence of others, then
gradually come to perform these functions for themselves. This process of ‘‘in-
 3. METACOGNITION AND OTHER MECHANISMS 101

ternalization’’ is gradual: First the adult (parent, teacher, etc.) controls and
guides the child’s activity; gradually the adult and the child come to share the
problem solving functions, with the child taking initiative and the adult correct-
ing and guiding when the child falters; finally, the adult cedes control to the child
and functions primarily as a supportive and sympathetic audience (Brown &
French, 1979; Campione, Brown, Ferrara & Bryant, 1984; Laboratory of Com-
parative Human Cognition, in press).
The developmental progression from other-regulation to self-regulation is
nicely illustrated in successful mother-child learning situations. Consider the
following example from Wertsch (1978, p. 17), the mothers and their young
children were given the task of copying a wooden puzzle (a truck). A completed
puzzle was used as the model and the mother and child were asked to complete
an identical puzzle. The mother was told to encourage the child if necessary. The
following is a sample of a video-taped interaction between a mother and her 2%
year-old daughter:

1. C: Oh (glances at model, then looks at pieces pile). Oh now where’s this

one go? (picks up black cargo square, looks at copy, then at pieces pile).
2. M: Where does it go in this other one (the model)? (child puts black cargo
square back down in pieces pile, looks at pieces pile).
3. M: Look at the other truck (model) and then you can tell (child looks at
model, then glances at pieces pile).
4. C: Well (looks at copy then at model).
5. C: I look at it.
6. C: Um, this other puzzle has a black one over there (child points to black
cargo square in model).
7. M: Um-hm.
8. C: A black one (looks at pieces pile).
9. M: So where do you want to put the black one on this (your) puzzle?
(child picks up black cargo square from pieces pile and looks at copy).
10. C: Well, where do you put it in there? Over there? (inserts black cargo
square correctly in copy).
11. M: That looks good.

In this example one can see the mother serving a vital regulatory function,
guiding the problem-solving activity of her child. Good examples of the mother
assuming the regulatory role are statements 2, 3, and 9, where she functions to
keep the child on task and to foster goal relevant search and comparison ac-
tivities. This protocol represents a midpoint between early stages of develop-
ment, where the mother and child speak to each other, but the mother’s utter-
ances do not seem to be interpreted by the child as task relevant; and later stages,
where the child assumes the regulatory functions, the mother functioning as a
sympathetic audience.
102 BROWN

It is argued that supportive ‘‘experts,’’ such as mothers in Wertsch’s exam-

ple, master craftsmen in apprenticeship systems (Childs & Greenfield, 1980),
and more experienced peers in tutoring studies (Allen, 1976), serve a major
function of initially adopting the monitoring and overseeing role; these crucial
regulatory activities are thereby made overt and explicit.
Ideally, teachers function as such mediators in the learning to learn process;
acting as promotors of self-regulation by nurturing the emergence of personal
planning, as they gradually cede their own direction. In schools, effective teach-
ers are those who engage in continual prompts to get children to plan and monitor
their own activities. Schallert and Kleiman (1979) identified four basic strategies
used by successful teachers to promote critical reading. First, they attempted to
tailor the information to the children’s existing level of understanding. They tried
to activate relevant background knowledge by having the students consider new
information in light of what they already knew. Continual attempts were made to
focus the students’ attention on important facts, finally, because of the teachers’
use of clever questioning and such Socratic ploys as invidious generalizations,
counter example, reality testing, etc., students were helped to monitor their own
comprehension. Thus, the expert teacher models many forms of critical thinking
for the students; processes that the students must internalize as part of their own
problem solving activities if they are to develop effective skills of self-
regulation.
Deficiencies in such interactive learning processes have been indicted as
primary sources of developmental retardation. (Brown & French, 1979; Feuer-
stein, 1979). Feuerstein’s theory is essentially similar to Vygotsky’s, for he too
holds that cognitive growth is very heavily dependent on the quality of mediated
learning that the child experiences. According to Feuerstein (1969): ‘‘Mediated
learning is the training given to the human organism by an experienced adult who
frames, selects, focuses, and feeds back an environmental experience in such a
way as to create appropriate learning sets’’ (p. 6). These mediated learning
experiences are an essential aspect of development, beginning when the parent
selects significant objects for the infant to focus on and proceeding throughout
development; the adult systematically shaping the child’s learning experiences. It
is the principal means by which children develop the cognitive operations neces-
sary for learning independently. By interacting with an adult, who guides prob-
lem solving activities and structures learning environments, children gradually
come to adopt structuring and regulatory activities of their own.
Feuerstein believes that the principal reason for the poor academic perfor-
mance of many disadvantaged students is the lack of consistent mediated learn-
ing in their earlier developmental histories, because of parental apathy, igno-
rance, or overcommitment. Quite simply, parents in disadvantaged homes were
themselves disadvantaged children and cannot be expected to teach what they do
not know; large family size and the need for a working mother does not leave a
great deal of time for Socratic dialogue games. In addition, these interactive
 3. METACOGNITION AND OTHER MECHANISMS 103

styles of continually questioning and extending the limits of knowledge are

typical of middle-class social interaction patterns and may be alien to some
cultures (Au, 1980; Bernstein, 1971).
However, mediated learning activities are exactly what occur in schools, and
the middle-class child comes well prepared to take part in these rituals. Not only
does the disadvantaged child lack prior exposure, but there is some evidence that
teachers give less experience in this learning mode to those who, because of their
lack of prior experience, need it most. For example, recent observations of
reading groups (Au, 1980; McDermott 1978) have shown that good and poor
readers are not treated equally. Good readers are questioned about the meaning
behind what they are reading, asked to evaluate and criticize material, etc. By
contrast, poor readers primarily receive drill in pronounciation and decoding.
Rarely are they given practice in qualifying and evaluating their comprehension
(Au, 1980). If, as a result, their reading problems persist, such children may be
singled out for special education; and again, this is likely to decrease rather than
increase their exposure to the ideal mediated learning experience.
Special education classes are more likely to provide step-by-step explicit
instruction for students in basic skills. Heavily programmed learning of this type
may be a practical and efficient means of getting less successful students to
perform better on a particular task; however, such experience is less likely to be
appropriate for promoting insightful learning or the development of self-reg-
ulatory skills. Students may learn something about a particular task, but they are
not likely to learn much about how to learn in general.
The development of cognitive skills proceeds normally via the gradual inter-
nalization of regulatory skills first experienced by the child in social settings
(Vygotsky, 1978). Following repeated experience with experts (mothers, teach-
ers, etc.), who criticize, evaluate, and extend the limits of their experience,
students develop skills of self-regulation. If students are to learn how to learn
independently, the development of a battery of such autocriticial skills (Binet,
1909) is essential. If for some reason the child is deprived of a constant history of
such interaction, the development of a battery of self-regulatory skills is unlikely
to occur.

Developmental Data. Only one example of the basic principle of acquiring

self-regulatory skills, by first observing them in social interaction, is provided.
The most frequently quoted data to support the other-regulation to self-regulation
transition is Wertsch’s (1978) descriptions of mother-child problem solving
dyads, an excerpt was included earlier. To illustrate the generality of the process,
a study from the laboratory, where much older learners are seen acquiring more
complex monitoring skills (Palinscar & Brown, 1981), is used. The basic situa-
tion was an interactive tutoring dyad, where seventh graders were receiving
instruction aimed at improving their reading comprehension skills. The children
were referred by their teachers because, although they were able to decode at
\
104 BROWN

approximately grade level, they had severe comprehension problems; performing

several grade levels behind their peers. Over many sessions the tutor and the
child engaged in an interactive learning game that involved taking turns in
leading a dialogue concerning each segment of text. Both the tutor and the child
would read a text segment; then the dialogue leader would paraphrase the main
idea, question any ambiguities, predict the possible questions that might be asked
about that segment, and hypothesize about the content of the remaining passage
segments. The dialogue leader would then ask the other a question on the seg-
ment; in the next segment the roles were reversed.
Initially, the tutor modelled these activities, however, when his or her turn
came, the child had great difficulty assuming the role of dialogue leader. The
tutor was forced to resort to constructing paraphrases and questions for the tutee
to mimic. In this initial phase, the tutor was modelling effective comprehension
monitoring strategies; the child was a relatively passive observer. In the inter-
mediate phase, the tutee became much more capable of playing her role as
dialogue leader, and by the end of 10 sessions was providing paraphrases and
questions of some sophistication. For example, in the initial sessions, 19% of the
questions produced by the tutees were judged as nonquestions and 36% as
needing clarification. Examples of questions needing clarification are:

1. What was, who, some kings were uh, about the kings? (translated as: Why
is it that kings did not always make the best judges?)
2. What were some of the people? (translated as: What kind of people can
serve on a jury?)
3. What was the Manaus built for? Wait a minute, What was the Manaus
built for? What certain kind of thing? Wait a minute. OK. What was the Manaus
tree built for? (translated as: Why was the city of Manaus built?)

In the final sessions, only 4% of the responses were judged as uninterpretable

or needing clarification.
At the beginning of the sessions, 34% of the questions were aimed at detail
and only 11% at main ideas; however, by the end of the sessions, 54% of all
questions probed comprehension of salient ideas. Examples of main idea ques-
tions are: *‘What are the three main problems with all submarines?’’ ‘‘Plans are
being made to use nuclear power for what?’’ ‘‘What did these people (the
Chinese) invent?’’ Examples of detail questions were: ‘‘What color is the
guard’s uniform?’’ **How far can flying fish leap?”’ ‘‘What are chopsticks made
out of?”
Similar progress was made in producing paraphrases of the main ideas of the
text segment. At the beginning of the sessions, only 11% of summary statements
captured main ideas; at the end of the sessions, 60% of statements were so
classified. The comprehension monitoring activities of the tutees certainly im-
proved, becoming more and more like those modelled by the tutor. With repeat-
 3. METACOGNITION AND OTHER MECHANISMS 105

ed interactive experiences, the tutor and child mutually constructing a cohesive

representation of the text, the tutees were able to employ monitoring functions
for themselves.
Improvement was revealed not just in the interactive sessions but also in
privately read stories, where the students were required to answer comprehension
questions on their own. In the laboratory, tests of comprehension were given
throughout the experiment. On the independent tests performance improved from
10 to 85% correct; in the classroom, compared with all other seventh graders in
the school, the students moved from the 7th to the 40th percentile. Not only did
the students learn to perform comprehension monitoring activities in interaction
with their tutor, they were also able to internalize these procedures as part of
their own cognitive processes for reading.

STATUS OF METACOGNITION AS A CONCEPT

In the preceding review, it is clear that metacognition is not only a monster of

obscure parentage, but a many-headed monster at that. In this final section an
attempt is made to estimate the current status of the offspring and to list some of
the many problems inherent in the current use of the term.
I would like to emphasize my belief that in many ways this status report is
premature. Scientific theorizing, like any other, must pass through stages. Con-
sider as an example, the novice block balancers described earlier. Initially, they
are merely goal oriented; they concentrate on getting the new theory to work.
The next stage is to develop and refine subsystems so that they work fluently.
Only when these subsystems are functioning efficiently can the theorist step back
and consider the entire problem space and systematize or reorganize it into a
cohesive whole. The recent history of theory development in the realm of meta-
cognition can be viewed in this light. In the early 1970s, attracted by the lure of a
new sounding concept, developmental psychologists engaged in demonstration
studies to see how the new idea would work. These early studies were often
ingenious and the wave of enthusiasm they provoked was justified.
The initial stage is now over and the current stage is, and should be, devoted
to the task of developing workable theories and procedures for separate parts of
the problem space. It is for this reason that the strands of inquiry that gave rise to
the step-child, metacognition are examined separately. Currently a great deal of
systematic work is being undertaken, hopefully, it will lead to fluently function-
ing subsystems that at present are merely juxtaposed, existing, and developing
side-by-side; however, this is an essential stage of theory building. Later, per-
haps, when the main subsystems are better understood, metaprocedural re-
organization (Karmiloff-Smith, 1979a) may be possible, and a full understanding
of the domain metacognition will be attained. This chapter is primarily a contri-
bution to the ‘‘juxtaposed procedures’ stage of development.
106 BROWN

However, doubt remains concerning whether the domain(s) covered by meta-

cognition will be tractable enough for a total systematization. If one takes the
wide view, metacognition, as currently used, refers to understanding in a very
broad sense; there is no simple problem space!
In a recent review, Wellman (1981), as others before have done, referred to
the concept of metacognition as a fuzzy concept. Wellman, however, went on to
discuss four features of the fuzzy concept.

First, the concept encompasses an essential, central distinction. However, this

distinction serves to anchor the concept not intentionally define it. Second, pro-
totypic central instances of the concept are easily recognized. However, third, at
the periphery agreement as to whether an activity is legitimately metacognitive
breaks down; the definitional boundaries are truly fuzzy. Related to this, and forth,
different processes all of which partake of the original distinction may be related
only loosely one to another. Thus the term metacognition or metamemory serves
primarily to designate a complex of associated phenomena. (pp. 3—4)

This is nicely put, and illustrates the loose confederation of topics included
under the blanket term, metacognition. Of some concern, however, is whether
the associated phenomena are linked closely enough to warrant the use of a single
family name; that is, does it refer to family resemblances within an ill-defined,
natural, or fuzzy category, or many categories? And, would it not be better, at
this stage, to abandon the global term and work at the level of subordinate
concepts, which are themselves fuzzy?
For clarity and communicative efficiency, a case could be made that the term
metacognition should be pensioned-off, or at least severly restricted in its exten-
sional reference. This is not because the phenomena subsumed under the term are
trivial, but because they are central processes of change and development; the
very theoretical ideas that this author predicts will be the main focus of develop-
mental learning theories in the next decade. Issues of fundamental importance
may be obscured by the current arguments surrounding things metacognitive;
arguments that are obscured because the participants do not make it clear which
head of the beast they are attacking or defending.
Wellman (1981) limits the term metacognition to knowledge, and believes
that this would be an agreed on limitation. However, it is not a limitation at all,
as knowledge can very easily involve all forms of information, both procedural
and declarative. It is clear that may present usages of metacognition do not
exclude process or factual knowledge, or anything else for that matter. One
Suggestion is to limit the term to one of its original uses, knowledge about
cognition, where that knowledge is stable and statable (Gleitman, in press).
Process terms such as planning-ahead, monitoring, resource allocation, self-
questioning, self-directing, etc. would then be used alone, without the adden-
dum, metacognition. This author has no ready solution to offer; but wishes to
 3. METACOGNITION AND OTHER MECHANISMS 107

emphasise that there are nontrivial problems associated with the current blanket
usage of the term.
This brings one back to the problems mentioned at the beginning of the
chapter. At present it is difficult to answer critical questions about metacogni-
tion, such as: “‘Is it late developing?”’ ‘‘Is it general or domain specific? [sit
conscious?”’ without pausing to ascertain which type of knowledge or process is
in question. Although metacognition may turn out to be a fuzzy concept with
indistinct boundaries, this degree of imprecision is not acceptable as a basis of
scientific inquiry. By referring to the process/knowledge under discussion by its
subordinate name (i.e., planning-ahead, error-correction, hypothesis-testing,
etc.) many of the current controversies, but by no means all, would evaporate. At
least one would know where the real problems are, and which problems are the
result of communication failure.
In conclusion, some of the really interesting issues of learning and develop-
ment that the concentration on metacognition has forced one to reconsider as
central concerns are: conscious control over learning; learning without
awareness, transfer of rule learning; relation of age and expertise to various
aspects of planning; monitoring and error correcting; general rules for problem
solving versus domain specific knowledge; and mechanisms of change. From
this seemly random list just one, mechanisms of change, is discussed here (for a
fuller review see Brown, 1979; Brown, 1982a; Brown, Bransford, Ferrara, &
Campione, 1983).
One of the real advances spurred by the interest in metacognition has been a
revived concern for mechanisms of change and development. This has always
been the hidden agenda of developmental psychologists; however, until recently,
there have been surprisingly few attempts to study change directly. Developmen-
tal studies have traditionally consisted of cross-sectional, frozen, one shot looks
at age changes; the performance of groups of children, varying in age or level of
expertise, is compared and contrasted. Even a great deal of longitudinal research
has a surprisingly cross-sectional flavor. Both approaches provide a picture of
cognition in stasis, rather than evolving, as it were, right before one’s eyes.
Many of the studies reviewed in this section depart from this pattern. This is
because many of the studies that have been inspired by the metacognitive boom
have been training studies, or have involved microgenetic analyses of children
learning-by-doing on their own, or learning to develop self-regulatory skills
through the intervention of supportive others.
Reverting to the microgenetic approach advocated by Vygotsky (1978) and
Werner (1961) enables one not only to concentrate on qualitative descriptions of
the stages of expertise, but also to consider transition phenomena and self-
modification techniques underlying the progression from beginning to expert
strategies. Anzai and Simon’s learning-by-doing theory, and developmental mi-
crogenetic studies of young learners (Karmiloff-Smith, 1979a, 1979b; Kar-
miloff-Smith & Inhelder, 1974—1975; Koslowski & Bruner, 1972 for example)
108 BROWN

are some instances of the current interest in within-subject, learning-by-doing

mechanisms of change that can be observed in a single subject within a short
space of time.
Similarly, the current interest in attempts to understand development by en-
gineering change via intervention has clearly been a positive outcome of interest
in metacognition (Brown & Campione, 1981). Training studies in general have
been a traditional tool of the developmental psychologist and are becoming of
increasing importance for both practical and applied reasons (Brown, Campione
& Day, 1981). Although there are many methods of intervention, the essential
element in traditional studies is that the experimenter provides feedback and
direction. It is the experimenter, then, who undertakes the requisite task analysis
and often, it is the experimenter who maintains all of the controlling and decision
making functions.
The current interest in more dynamic learning situations has seen a move
away from direct instruction of the traditional kind towards the interactive pro-
cesses, illustrated by the Palinscar and Brown (1981) study described earlier, and
in Wertsch’s (1978) examination of mother-child teaching dyads. Via the inter-
vention of a supportive, knowledgeable other, the child is led to the limits of his
or her own understanding. The teacher does not, however, tell the child what to
do, he or she enters into an interaction where the child and the teacher are
mutually responsible for getting the task down. As the child adopts more of the
essential skills initially undertaken by the adult, the adult relinquishes control.
Transferrence of power is gradually and mutually agreed upon.
Although the supportive other in the laboratory is usually an experimenter,
these interactive learning experiences are intended to mimic real life learning.
Mothers (Wertsch, 1978), teachers (Schallert & Kleinman, 1979) and master
craftsmen (Childs & Greenfield, 1980) all function as the supportive other: the
agent of change responsible for structuring the child’s environment in such a way
that he or she will experience a judicious mix of compatible and conflicting
experiences. The importance of such interactive learning experiences for general
cognitive development should not be overlooked. Many cognitive activities are
initially experienced in social settings; however, in time, the results of such
experiences become internalized. Initially the supportive other acts as the inter-
rogator, leading the child to more powerful rules and generalizations. The inter-
rogative, regulatory role, however, becomes internalized during the process of
development, and children become able to fulfill some of these functions for
themselves through self-regulation and self-interrogation. Mature thinkers are
those who provide conflict trials for themselves, practice thought experiments,
question their own basic assumptions, provide counterexamples to their own
rules, etc. In short, although a great deal of thinking and learning may remain a
social activity (Brown & French, 1979; Laboratory for Comparative Human
Cognition, in press), through the process of internalization, mature reasoners
become capable of providing the supportive other role for themselves. Under
 3. METACOGNITION AND OTHER MECHANISMS 109

these systems of tutelage, the child learns not only how to get a particular task
done independently, but also how to set about learning new problems. In other
words, the child learns how to learn (Brown, 1982b; Brown, Bransford, Ferrara,
& Campione, 1983).

SUMMARY

Some of the diverse historical roots of the family of concepts that are often
referred to generically as metacognition have been examined. Specifically dis-
cussed are the controversies surrounding such perennial problems for psychology
as: the status of verbal reports as data; executive control; consciousness and who
has it; error correction and systemization in early language production and metal-
inguist awareness; self-regulation, metaprocedural reorganization, and reflected
abstraction from Genevan psychology; and the transferrence from other-regula-
tion to self-regulation, inspired by Vygotsky’s theory of development.
The use of the blanket term metacognition for quite different theories and
various levels of analyses was questioned, as was the utility of the continued use
of such a catholic term. The important redirection of attention to mechanisms of
change inspired by the interest in the metacognitive issues was welcomed.

ACKNOWLEDGMENTS

Preparation of this manuscript was supported in part by U.S.P.H.S. Grant HD-05951

REFERENCES

Allen, V. L. (1976). Children as teachers: Theory and research on tutoring. New York: Academic
Press.
Anzai, Y., & Simon, H.A. (1979). The theory of learning by doing. Psychological Review, 86,
124-140.
Au, K. (1980). A test of the social organizational hypothesis: Relationships between participation
structures and learning to read. Unpublished doctoral dissertation, University of Illinois.
Baker, L., & Brown, A. L. (1981). Metacognition and the reading process. In D. Pearson (Ed.), A
handbook of reading research. New York: Plenum Press, 353-394.
Baker, L., & Brown, A. L. (1984). Cognitive monitoring in reading. In J. Flood (Ed.), Under-
standing reading comprehension. Newark, DE: International Reading Association.
Baldwin, M. J. (1909). How children study. Archives of Psychology, 12, 65-70.
Bartlett, F. C. (1958). Thinking: An experimental and social study. New York: Basic Books.
Belmont, J. M., & Butterfield, E.C. (1977). The instructional approach to developmental cognitive
research. In R. V. Kail Jr. & J. W. Hagen (Eds.), Perspectives on the development of memory
and cognition. Hillsdale, NJ: Lawrence Erlbaum Associates.
Bernstein, B. (1971). Class codes and control (Vol. 1). London: Routledge & Kegan Paul.
Binet, A. (1890). Perceptions d’enfants. Revue philosophique, 30, 582-611.
110 BROWN

Binet, A. (1903). L’ etude experimentale de Iintelligence. Paris: Schleicher Freres.

Binet, A. (1909). Les idees modernes sur les infants. Paris: Ernest Flamarion.
Boden, M. A. (1972). Purposive explanation in psychology. Boston, MA: Harvard University
Press.
Boden, M. A. (1977). Artificial intelligence and natural man. Sussex, England: The Harvester
Press.
Bowerman, M. (1981). Starting to talk worse: Clues to language acquisition from children’s late
speech errors. in S. Strauss (Ed.), U-shaped behavioral growth. New York: Academic Press.
Broadbent, D. E. (1977). Levels, hierarchies, and the locus of control. Quarterly Journal of Experi-
mental Psychology, 29, 181-201.
Broudy, H. S. (1977). Types of knowledge and purposes of education. In R. C. Anderson, R. J.
Spiro & W. E. Montague (Eds.), Schooling and the acquisition of knowledge. Hillsdale, NJ:
Lawrence Erlbaum Associates.
Brown, A. (1974). The role of strategic behavior in retardate memory. In N. R. Ellis (Ed.),
International review of research in mental retardation (Vol. 7). New York: Academic Press.
Brown, A. L. (1975). The development of memory: Knowing, knowing about knowing, and know-
ing how to know. In H. W. Reese (Ed.), Advances in child development and behavior (Vol. 10).
New York: Academic Press.
Brown, A. L. (1978). Knowing when, where, and how to remember: A problem of metacognition.
In R. Glaser (Ed.), Advances in instructional psychology (Vol. 1). Hillsdale, NJ: Lawrence
Erlbaum Associates.
Brown, A. L. (1979). Theories of memory and the problem of development: Activity, growth, and
knowledge. In L. S. Cermak & F. I. M. Craik (Eds.), Levels of processing in human memory.
Hillsdale, NJ: Lawrence Erlbaum Associates.
Brown, A. L. (1980). Metacognitive development and reading. In R. J. Spiro, B. Bruce, & W.
Brewer (Eds.), Theoretical issues in reading comprehension. Hillsdale, NJ: Lawrence Erlbaum
Associates.
Brown, A. L. (1982a). Learning and development: The problem of compatibility, access, and
induction. Human Development, 25, 89-115.
Brown, A. L. (1982b). Learning to learn how to read. In J. Langer & T. Smith-Burke (Eds.),
Reader meets author, bridging the gap: A psycholinguistic and social linguistic perspective.
Newark, NJ: Dell.
Brown, A. L., Bransford, J. D., Ferrara, R. A., & Campione, J. C. (1983). Learning, remember-
ing, and understanding. In J. H. Flavell and E. M. Markman (Eds.), Handbook of child psychol-
ogy (4th ed.). Cognitive development (Vol. 3, pp. 515-529). New York: Wiley.
Brown, A. L., & Campione, J. C. (1978). Permissible inferences from the outcome of training
studies in cognitive development research. Quarterly Newsletter of the Institute for C omparative
Human Development, 2, 46-53.
Brown, A. L., & Campione, J. C. (1981). Inducing flexible thinking: A problem of access. In M.
Friedman, J. P. Das, & N. O’Connor (Eds.), Intelligence and learning (pp. 515-530). New
York: Plenum Press.
Brown, A. L., Campione, J. C., & Day, J. D. (1981a). Learning to learn: On training students to
learn from texts. Educational Researcher, 10 (2), 14-21.
Brown, A. L., & DeLoache, J. S. (1978). Skills, plans, and self-regulation. In R. Siegler (Ed.),
Children’s thinking: What develops? Hillsdale, NJ: Lawrence Erlbaum Associates.
Brown, A. L., & French, L. A. (1979). The zone of potential development: Implications
for
intelligence testing in the year 2000. Intelligence, 3, 255-277.
Brown, A. L., Smiley, S. S., & Lawton, S. C. (1978). The effects of experience on the solution of
suitable retrieval cues for studying texts. Child Development, 49, 829-835.
Bruner, J. S. (1973). Organization of early skilled action. Child Development, 44, 1-11.
Bühler, K. (1907). Remarques sur les problemes de la psychologie de la pensée.
Archives de
Psychologie, 6, 376-386.
 3. METACOGNITION AND OTHER MECHANISMS et

Campione, J. C., & Brown, A. L. (1977). Memory and metamemory developmen

t in educable
retarded children. In R. V. Kail, Jr. & J. W. Hagen (Eds.), Perspectives on
the development of
memory and cognition. Hillsdale, NJ: Lawrence Erlbaum Associates.
Campione, J. C., Brown, A. L., Ferrara, R. A., & Bryant, N. R. (1984). The
zone of proximal
development: Implications for individual differences and learning. In B. Rogoff
& J. V. Wertsch
(Eds.), Children’s Learning in the ‘‘Zone of Proximal Development’. San Francisco:
Jossey-
Bass.
Cavanaugh, J. C., & Borkowski, J. G. (1979). The metamemory-memory ‘‘connection’
’: Effects
of strategy training and transfer. Journal of General Psychology, 101, 161-174.
Cavanaugh, J. C., & Perlmutter, M. (1980). Metamemory: A critical examination.
Unpublished
manuscript, University of Minnesota.
Chi, M. T. H. (1981). Knowledge development and memory performance learning research
and
development center. In M. Friedman, J. P. Das, & N. O’Connor (Eds.), Intelligence
and
Learning. New York: Plenum Press.
Chi, M. T. H., Glaser, R., & Rees, E. (1982). Expertice in problem solving. In R. Sternberg (Ed.),
Advances in the psychology of human intelligence (Vol. 1). Hillsdale, NJ: Lawrence Erlbaum
Associates.
Childs, C. P., & Greenfield, P. M. (1980). Informal modes of learning and teaching. In N. Warren
(Ed.), Advances in cross-cultural psychology (Vol. 2). London: Academic Press.
Clark, E. V. (March, 1979). Children’s error corrections in early language production. Paper
presented at the meeting of the Society for Research in Child Development, San Francisco.
Collins, A. (1977). Processes in acquiring knowledge. In R. C. Anderson, R. J. Spiro & W. E.
Montague (Eds.), Schooling and the acquisition of knowledge. Hillsdale, NJ: Lawrence Erlbaum
Associates.
Collins, A., Brown, A. L., Morgan, J. L., & Brewer, W. F. (April, 1977). The analysis ofreading
tasks and texts (Tech. Rep. 43). Champaign: University of Illinois, Center for the Study of
Reading.
Davis, J. H., Carey, M. H., Foxman, P. N., & Tarr, D. B. (1968). Verbalization in problem
solving. Journal of Personality and Social Psychology, 8, 299-302.
DeLoache, J. S., Sugarman, S., & Brown, A. L. (1981). Self-correction strategies in early cog-
nitive development. Paper presented at SRCD, Boston.
Dennett, D. C. (1978). Brainstorms: Philosophical essays on mind and psychology. Montgomery,
VT: Bradford Books.
Dewey, J. (1910). How we think. Boston: Heath.
Ericsson, K. A., & Simon, H. A. (1980). Verbal reports as data. Psychological Review, 87, 215—
251%
Ernst, G. W., & Newell, A. (1969). GPS: A case study in generality and problem solving. New
York: Academic Press.
Feuerstein, R. (1969). The instrumental enrichment Method: An outline of theory and technique.
Unpublished paper. Hadassah-Wiza- Canada Research Institute.
Feuerstein, R. (1979). Instrumental enrichment. Baltimore, MD: University Park Press.
Flanagan, J. C. (1954). The critical incident technique. Psychological Bulletin, 51, 327-358.
Flavell, J. H. (1970). Developmental studies of mediated memory. In H. W. Reese and L. P. Lipstt
(Eds.), Advances in child development and behavior (Vol. 5). New York: Academic Press.
Flavell, J. H. (1971). First discussant’s comments: What is memory development the development
of? Human Development, 14, 272—278.
Flavell, J. H. (1976). Metacognitive aspects of problem solving. In L. Resnick (Ed.), The nature of
intelligence. Hillsdale, NJ: Lawrence Erlbaum Associates.
Flavell, J. H. (1980). Structures, stages, and sequences in cognitive development. Paper presented
at the Minnesota Symposium on Child Psychology, Minneapolis.
Flavell, J. H. (1981). Cognitive monitoring. In W. P. Dickson (Ed.), Children’s oral communica-
tion skills. New York: Academic Press.
12 BROWN

Flavell, J. H., Friedrichs, A. G. & Hoyt, J. D. (1970). Developmental changes in memorization

processes. Cognitive Psychology, 1, 324-340.
Flavell, J. H., Speer, J. R., Green, F. L., & August, D. L. (1981). The development of comprehen-
sion monitoring and knowledge about communication. Monographs of the Society for Research
in Child Development, (46, Whole No. 192).
Flavell, J. H., & Wellman, H. M. (1977). InR. V. Kail, Jr., & W. Hagen (Eds.), Perspectives on
the development of memory and cognition. Hillsdale, NJ: Lawrence Erlbaum Associates.
Flores, C. F., & Winograd, T. (1978). Understanding cognition as understanding. Unpublished
manuscript, Stanford University.
Fromkin, V. A. (Ed.). (1973). Speech errors as linguistic evidence. The Hague: Mouton.
Gagné, R. H., & Smith, E. C. (1962). A study of the effects of verbalization and problem solving.
Journal of Experimental Psychology, 63, 12-18.
Gardner, H. (1978). Commentary on animal awareness papers. Behavioral and Brain Sciences, 4,
SDR:
Gelman, R., & Baillargeon, R. (1983). A review of some Piagetian concepts. In J. H. Flavell & E.
M. Markman (Eds.), Handbook of Child Development: Vol. 1. Cognitive Development. New
York: John Wiley.
Gleitman, H. (in press). Some trends in the study of cognition. In S. Koch & D. E. Leary (Eds.), A
century of psychology as science: Retrospectives and assessments. New York: McGraw-Hill.
Gleitman, L. R., Gleitman, H., & Shipley, E. F. (1972). The emergence of the child as gram-
marian. Cognition, 1, 137-164.
Goldin, S. E., & Hayes-Roth, B. (1980). Individual differences in planning processes (Tech. Rep.
N-1488-ONR). Santa Monica, CA: The Rand Corporation.
Greene, P. H. (1971). Introduction in I. M. Gelford, U. S. Garjinkel, S. V. Fourer & M. L. Tsetlin
(Eds.), Model of the structural functional organization of certain biological systems (pp. xi—
xxv). Cambridge, MA: M.I.T. Press.
Greeno, J. G., & Bjork, R. A. (1973). Mathematical learning theory and the new ‘‘mental for-
estry.’’ In P. H. Mussen & M. R. Rozenzweig (Eds.), Annual review of psychology, 24.
Hasher, L., & Zacks, R. T. (1979). Automatic and effortful processes in memory. Journal of
Experimental Psychology, 108, 356-388.
Hayes-Roth, B. (1977). Evolution of cognitive structure and process. Psychological Review, 84,
260-278.
Hayes-Roth, B., & Thorndyke, P. W. (1980). Decision making during the planning process (Tech.
Rep. N-1213-ONR). Santa Monica, CA: The Rand Corporation.
Hayes-Roth, B., & Hayes-Roth, F. (1979). A cognitive model of planning. Cognitive Science, 3,
275-310.
Hayes-Roth, B. (1980). Flexibility in executive strategies. (Tech. Rep. N-1170-ONR). Santa
Monica, CA: The Rand Corporation.
Huey, E. B. (1968). The psychology and pedagogy of reading. Cambridge, MA: M.I.T. Press.
(Originally published, 1908).
Humphrey, G. (1963). Thinking: An introduction to its experimental psychology. New York: John
Wiley.
Inhelder, B., Sinclair, H., & Bovet, M. (1974). Learning and the development of cognition.
Cambridge, MA: Harvard University Press.
James, W. (1890). Principle ofpsychology (Vol. 1). New York: Holt.
Johnson-Laird, P. N. (1980). Mental models in cognitive science. Cognitive Science, 4, 71-115.
Justice, E. M., & Bray, N. W. (1979). The effects of context and feedback on metamemory in young
children. Paper presented at the meeting of the society for Research in Child Development, San
Francisco.
Karmiloff-Smith, A. (1979a). Micro- and macro- developmental changes in language acquisition
and other representational systems. Cognitive Science, 3, 91-118.
Karmiloff-Smith, A. (1979b). Problem solving construction and representations of closed railway
circuits. Archives of Psychology, 47, 37-59.
 3. METACOGNITION AND OTHER MECHANISMS 113

Karmiloff-Smith, A., & Inhelder, B. (1974-75). If you want to get ahead, get a theory. Cognition,
3, 195-212.
Karph, D. A. (1973). Thinking aloud in human discrimination learning. Doctoral
dissertation, State
University of New York at Stony Brook.
Kendall, C. R., Borkowski, J. G., & Cavanaugh, J. C. (1980). Maintenance and
generalization of
an interrogative strategy by EMR children. Intelligence, 4, 255-270.
Klinger, E. (1974). Utterances to evaluate steps and control attention distinguish
operant from
respondent thought while thinking out loud. Bulletin of the Psychonomic Society,
4, 44-45.
Koestler, A. (1979). Janis: A summing up. New York: Random House.
Koslowski, B., & Bruner, J. S. (1972). Learning to use a lever. Child Development, 43, 790-799.
Kreutzer, M. A., Leonard, S. C., & Flavell, J. H. (1975). An interview study: Children’s
knowl-
edge about memory. Monographs of the Society for Research in Child Development, 40 (1, Serial
No. 159).
LaBerge, D. (1975). Acquisition of automatic processing in perceptual and associative learning. In
P. M. A. Rabbitt & S. Dornic (Eds.), Attention and performance. New York: Academic Press.
Laboratory of Comparative Human Cognition. (in press). The zone of proximal development:
Where culture and cognitive create one author. In J. Wertsch (Ed.), Culture, Communication and
Cognition: Vygotskian Perspectives. New York: Cambridge University Press.
Lindsay, P. H., & Norman, D. A. (1977). Human information processing (2nd ed.). New York:
Academic Press.
Locke, J. (1924). An essay concerning human understanding. A. L. Pringle-Pattison (Ed.). Oxford:
Clarendon Press.
Logan, G. D. (1978). Attention in character classification tasks: Evidence for automaticity of
component stages. Journal of Experimental Psychology: General, 107, 32-63.
Logan, G. D. (1979). On the use of a concurrent memory load of measure attention and automat-
icity. Journal of Experimental Psychology: Human Perception and Performance, 5, 189-207.
Luria, A. R. (1966). Higher certical function in man. London: Tavistock Press.
Mandler, J. M., & Mandler, G. (1964). Thinking: From association to Gestalt. New York: John
Wiley.
Markman, E. M. (1977). Realizing that you don’t understand: A preliminary investigation. Child
Development, 46, 986-992.
Markman, E. M. (1979). Realizing that you don’t understand: Elementary school children’s
awareness of inconsistencies. Child Development, 50, 643-655.
Marshall, J. C., & Morton, J. (1978). On the mechanics of EMMA. In A. Sinclair, R. J. Jarvella,
& W. J. M. Levelt (Eds.), The child’s conception of language. Berlin: Springer.
Masur, E. F., McIntyre, C. W., & Flavell, J. H. (1973). Developmental changes in apportionment
of study time among items in a multitrial free recall task. Journal of Experimental Child Psychol-
ogy, 15, 237-246.
McDermott, J., & Forgy, C. (1978). Production system conflict resolution strategies. In D. A.
Waterman & F. Hayes-Roth (Eds.), Pattern-directed influence systems. New York: Academic
Press.
McDermott, R. D. (1978). Some reasons for focusing on classrooms in reading research. In P. D.
Pearson & J. Hansen (Eds.), Reading: Disciplined inquiry in process and practice (27th Year-
book of the National Reading Conference). Clemson, SC: National Reading Conference.
Milner, B. (1964). Some effects of frontal lobectomy in man. In J. M. Warren & K. Avert (Eds.),
The frontal granular certex and behavior. New York: McGraw-Hill.
Myers, M., & Paris, S. (1978). Children’s metacognitive knowledge about reading. Journal of
Educational Psychology, 70, 680-690.
Naus, M. J. & Halasz, F. G. (1978). Developmental perspectives on cognitive processing. In L. S.
Cermak & F. I. Craik (Eds.), Levels of processing in human memory. Hillsdale, NJ: Lawrence
Erlbaum Associates.
Nelson, K. E. & Nelson, K. (1979). Cognitive pendulums and their linguistic realization. In K.
Nelson (Ed.), Children’s Language, New York: Gardner.
114 BROWN

Newell, A. (1980). Production systems and human cognition. In R. A. Cole (Ed.), Perceptive and
production of fluent speech. Hillsdale, NJ: Lawrence Erlbaum Associates.
Newell, A., & Simon, H. A. (1972). Human problem solving. Englewood Cliffs, NJ: Prentice-
Hall.
Nisbett, R. E., & Wilson, D. (1977). Telling more than we know: Verbal reports on mental
processes. Psychological Review, 84, 231-279.
Nooteboom, S. C. (1969). The tongue slips into patterns. In A. G. Sciarone, A. J. Van Essen & A.
A. Van Raad (Eds.), Nomen: Leyden studies in linguistics and pragmatics. The Hague: Mouton.
Norman, D. A. (1973). Memory, knowledge, and the answering of questions. In R. L. Solso (Ed.),
Contemporary issues in cognitive psychology: The Loyola Symposium. Washington, DC: V. H.
Winston.
Norman, D. A. (1980). Twelve issues for cognitive science. Cognitive Science, 4, 1-32.
Norman, D. A. (1981). Categorization of action slips. Psychological Review, 88, 1-15.
Norman, D. A., & Bobrow, G. A. (1975). On data-limited and resource-limited processes. Cog-
nitive Psychology, 7, 44—64.
Norman, D. A., & Shallice, T. (December, 1980). Attention to action: Willed and automatic
control of behavior (CHIP Technical Report 99). LaJolla, CA: University of California.
Palinscar, A. S., & Brown, A. L. (1981). Training comprehension-monitoring skills in an interac-
tive learning game. Unpublished manuscript, University of Illinois.
Pellegrino, J. W., & Schadler, M. (1974). Maximizing performance in a problem-solving task.
Unpublished manuscript, University of Pittsburgh.
Perlmutter, M. (1978). What is memory aging the aging of? Developmental Psychology, 14, 330-
345.
Piaget, J. (1976). The grasp of consciousness: Action and concept in the young child. Cambridge,
MA: Harvard University Press.
Piaget, J. (1978). Success and understanding. Cambridge, MA: Harvard University Press.
Posner, M. I., & Snyder, C. R. R. (1974). Attention and cognitive control. In R. L. Solso (Ed.),
Information processing and cognition: The Loyola Symposium. Potomac, MD; Hillsdale, NJ:
Lawrence Erlbaum Associates.
Pylyshyn, Z. (1978). Computational models and empirical constraints. The Behavioral and Brain
Sciences, 1, 93-99.
Reason, J. T. (1979). Actions not as planned. In G. Underwood and R. Stevens (Eds.), Aspects of
consciousness. London: Academic Press.
Reed, S. K., & Johnson, J. A. (1977). Memory for problem solutions. In G. Bower (Ed.), The
psychology of learning and motivation (Vol. 11). New York: Academic Press.
Reese, W. (1962). Verbal mediation as a function of age level. Psychological Bulletin, 59, 502—
509.
Resnick, L. B. & Glaser, R. (1976). Problem solving and intelligence. In L. B. Resnick (Ed.), The
Nature of Intelligence. Hillsdale, NJ: Lawrence Erlbaum Associates.
Riegel, K. F. (1975). Structure and transformation in modern intellectual history. In K. F. Riegel &
G. C. Rosenwald (Eds.), Structure and transformation: Developmental and historical aspects.
New York: Wiley.
Rogoff, B., & Gardner, W. P. (in press). Developing cognitive skills in social interaction. In B.
Rogoff & J. Lave (Eds.), Everyday cognition: Its development in social context. Hillsdale, NJ:
Lawrence Erlbaum Associates.
Rozin, P. (1976). The evolution of intelligence and access to the cognitive unconscious. Progres-
sion in Psychobiology and Physiological Psychology, 6, 245-280.
Ruger, H. A. (1910). The psychology of efficiency. Archive of Psychology, 15.
Rumelhart, D. E. (1980). Schemata: The building blocks of cognition. In R. J. Spiro, B. C. Bruce
& W. F. Brewer (Eds.), Theoretical issues in reading comprehension. Hillsdale, NJ: Lawrence
Erlbaum Associates.
 3. METACOGNITION AND OTHER MECHANISMS 15

Sacerdoti, E. D. (1974). Planning in a hierarchy of abstraction spaces.

Artificial Intelligence, 5,
115-136.
Schallert, D. L., & Kleiman, G. M. (June, 1979). Some reasons
why the teacher is easier to
understand than the text book (Reading Education Report No. 9). Urbana:
University of Illinois,
Center for the Study of Reading (ERIC Document Reproduction Service No.
ED 172 189)
Schallice, T. (1978). The dominant action system: An information-processing
approach to con-
sciousness. In K. Pope & J. E. Springer (Eds.), The flow of conscious experience.
New York:
Plenum Press.
Schneider, W., & Shiffrin, R. M. (1977). Controlled and automatic human information
processing:
|. Direction, search, and attention. Psychological Review, 84, 1-66.
Selfridge, O. (1959). Pandemonium: A paradigm for learning. Symposium on
the mechanization of
thought. London: H. M. Stationery Office.
Shatz, M. (1978). On the development of communication understandings: An early
strategy for
interpreting and responding to messages. Cognitive Psychology, 10, 271-301.
Shaw, R., & Bransford, J. (1977). Approaches to the problem of knowledge. In R. Shaw
& J.
Bransford (Eds.), Perceiving, acting, and knowing: Toward an ecological psychology. Hillsdale,
NJ: Lawrence Erlbaum Associates.
Shif, Z. I. (1969). Development of children in schools for mentally retarded. In M. Cole & I.
Maltzman (Eds.), A handbook of contemporary Soviet psychology. New York: Basic Books.
Shiffrin, R. M. (1975). The locus and role of attention in memory systems. In P. M. A. Rabbitt, &
S. Dornic (Eds.), Attention and performance V. NY: Academic Press.
Shiffrin, R. M., & Schneider, W. (1977). Controlled and automatic human information processing:
2. Perceptual learning, automatic attending, and a general theory. Psychological Review, 84,
127-190.
Simon, D. P., & Simon, H. A. (1978). Individual differences in solving physics problems. In
R. S. Siegler (Ed.), Children’s thinking: What develops? Hillsdale, NJ: Lawrence Erlbaum
Associates.
Skinner, B. F. (1964). Behaviorism at fifty. In T. W. Wann (Ed.), Behaviorism and phe-
nomenology. Chicago: University of Chicago Press.
Skinner, B. F. (1971). Beyond freedom and dignity. New York: Knopf.
Smith, E. R., & Miller, F. S. (1978). Limits on perception of cognitive processes. A reply to
Nisbett and Wilson. Psychological Review, 85, 355-362.
Spearman, C. (1923). The nature of intelligence and principle of cognition. London: MacMillan.
Stein, N. L. & Trabasso, T. (1982). What’s in a story: An approach to comprehension and instruc-
tion. In R. Glaser (Ed.), Advances in the psychology of instruction (Vol. 2). Hillsdale, NJ:
Lawrence Erlbaum Associates.
Sternberg, R. J., Ketron, J. L., & Powell, J. S. (1982). Confidential approaches to the training of
intelligence performance. In D. K. Detterman & R. J. Sternberg (Eds.), How and how much can
Intelligence be Increased. Norwood, NJ: Ablex, (155-172).
Strauss, S., & Stavey, R. (1982). U-shaped behavioral growth: Implications for theories of devel-
opment. In W. W. Hartup (Ed.), Review of child development research (Vol. 6). Chicago:
University of Chicago Press.
Thorndike, E. L. (1917). Reading as reasoning: A study of mistakes in paragraph reading. Journal
of Educational Psychology, 8, 323-332.
Turvey, M. T. (1977). Preliminaries to a theory of action with reference to vision. InR. Shaw & J.
Bransford (Eds.), Perceiving, acting, and knowing: Toward an ecological psychology. Hillsdale,
NJ: Lawrence Erlbaum Associates.
Turvey, M. T., & Shaw, R. (1979). The primacy of perceiving: An ecological reformulation of
perception for understanding memory. In L. G. Nillson (Ed.), Perspective of Uppsola Univer-
sity’s 500 Anniversary. Hillsdale, NJ: Lawrence Erlbaum Associates.
Turvey, M. T., Shaw, R. E., & Mace, W. (1978). Issues in the theory of action: Degrees of
116 BROWN

freedom, coordinative structures and coalitions. In J. Requin (Ed.), Attention and performance
VII. Hillside, NJ: Lawrence Erlbaum Associates.
Vygotsky, L. S. (1978). Mind in Society: The Development of Higher Psychological Processes (M.
Cole, V. John-Steiner, S. Scribner, & E. Souberman, Eds. and Trans.). Cambridge, MA:
Harvard University Press.
Wellman, H. M. (1983). Metamemory revisited. In M. Chi (Ed.), What is memory development the
development of? A look after a decade (pp. 31—51). Basel: Karger.
Werner, H. (1961). Comparative psychology of mental development. New York: Science Editions.
Wertsch, J. V. (1978). Adult-child interaction and the roots of metacognition. Quarterly Newsletter
of the Institute for Comparative Human Development, 1, 15-18.
White, P. (1980). Limitations on verbal reports of internal events. A refutation of Nisbett and
Wilson, and Bem. Psychological Review, 87, 105-112.
Wilder, L., & Harvey, D. J. (1971). Overt and covert verbalization in problem solving. Speech
Monographs, 38, 171-176.
Wimmer, H., & Thornquist, K. (1978). The relation of metamemory to mnemonic performance in
memory development: Metamemory is always necessary but never sufficient. Unpublished manu-
script University of Salzburg, Austria.
 Metacognitive Components of
Instructional Research with
Problem Learners

Joseph C. Campione
University of Illinois at Urbana-Champaign

The aim of this paper is to describe the influence that theory and research in the
broad area of metacognition have had on psychoeducational research; particu-
larly on work with slow-learning children. Any comparison of the literature 10
years ago, with the literature emerging today, makes it clear that metacognition
has had a pronounced impact. It can be seen directly in a number of research
efforts, and more broadly in the recent overall orientation guiding this research.
On one level, specific findings, concerning the nature of individual and com-
parative differences in metacognitive functioning, have been borrowed directly
and incorporated into instructional research; the content of the research changed.
A related, but more general, effect can also be seen: Metacognitive work has
reawakened an interest in the role of consciousness, or awareness, or understand-
ing, in thinking and problem-solving. These topics bring with them considerable
problems, nevertheless, they are intimately involved in important aspects of
thinking. At first, it was difficult to see how complex forms of thinking could be
inculcated without making learners aware of the operations, facts, and pro-
cedures that are the goals of instruction. However, as this fact became more
apparent, the form of training studies changed.
In order to trace the changes in training studies, this paper is divided into three
main sections: the first section deals with a description of training studies, prior
to the interest in metacognition; the second section involves a brief discussion of
the types of metacognitive research that emerged in the early 1970s, and influ-
enced the instructional work; and the third section attempts a classification of a
number of different types of metacognitive-instructional experiments that have
been conducted, and an analysis of their effectiveness.

7
 118 _CAMPIONE

During the past decade, there has been an explosion of interest in “‘metacog-
nitive’’ aspects of academic and social performance. Viewed in one perspective,
this interest has stemmed from a concentration on the memory development
research pioneered by Flavell and his colleagues (e.g., Flavell, 1971; Flavell,
Friedrichs, & Hoyt, 1970; Masur, McIntyre, & Flavell, 1973). As a result, terms
that sound suspiciously like metacognition and its various referents have become
prominent in such diverse areas as reading and comprehension (Baker & Brown,
1981; Brown, 1980; Markman, 1981); cognitive ethology (Rozin, 1976); general
cognitive development (Brown, 1982; Flavell, 1979); and intelligence, machine
intelligence (e.g., Bobrow & Norman, 1975; Pylyshyn, 1978) and human intel-
ligence (Brown, 1978; Campione & Brown, 1978; Campione, Brown, & Fer-
rara, 1982; Sternberg, 1980). In each case, the more advanced or successful
performers are characterized as possessing more fully developed metacognitive
skills than the less successful performers. Metacognitive skills are not merely
factors underlying performance; frequently, they have been regarded as being of
particular importance, by a number of theorists (cf. Campione, Brown & Fer-
rara, 1982, for a discussion of the role of metacognitive factors in theories of
intelligence). Thus, despite pronounced variations in the subject populations
investigated, the specific topics being addressed, and the general theoretical
orientations guiding the different investigators, some agreement on the impor-
tance and centrality of metacognitive factors seems to have been reached.
The frequency with which metacognitive deficiencies have been cited as a
factor in poor academic performance has led, not surprisingly many psychol-
ogists engaged in instructional research, to include metacognitive skills as part of
their overall training packages. Similar instructional routines resulted from more
clinical approaches to the design of curricula developed for use with slow-
learning children; both at the turn of the century (Binet, 1909) and more currently
(e.g., Feuerstein, 1980). The aim of this chapter is to review some of the
relevant
theories and resultant instructional research, summarizing the current status
of
these approaches. The more specific aim is to classify and illustrate some of
the
different types of experiments that have been done and to use the review
as a
means of describing the various aspects of metacognition that seem
to be of
particular relevance to the topic of instructional practice.
In order to do this, it is necessary to briefly review a set of
experiments
primarily, though not exclusively, from our laboratory. The studies
reviewed
typically involve students with learning problems, most frequent
ly mildly re-
tarded (IQ around 70) adolescents. This is the area in which
the author has
worked most closely, and one in which much of the metacogn
itively-oriented
instructional work has been conducted. Some of the specific
topics covered are,
no doubt, a reflection of the emphasis placed on problem
learners. The conclu-
sions reached might be different if more capable students
were the objects of
instruction; however, in the limited space available, such
a concentration seems
Justified.
 4. METACOGNITION AND PROBLEM LEARNERS’ 119

To motivate the discussion, before proceeding to the actual research, it is

necessary to address two preliminary topics. The first topic deals with the various
referents of the term ‘‘metacognition’’; the second deals with the status of
instructional work with slow-learning children, prior to the advent of the recent
interest in metacognition.

METACOGNITION AND ITS VARIOUS GUISES

The emphasis on metacognitive-like processes in recent theories is ubiquitious;

however, this ‘‘agreement’’ should not be interpreted too strongly. Although, on
a general level, there is agreement that metacognition is an important component
of successful performance, there are many different kinds of knowledge and
processes subsumed under that term; metacognition has been used by different
authors to mean quite different things. Brown (Chapter 3, this volume) provided
a detailed analysis of these variations; in this discussion, the author is much less
thorough. On the most general level, the topic of metacognition includes knowl-
edge concerning, understanding of, and access to cognition and cognitive re-
sources. For example, Flavell (1979) has defined metacognition as ‘knowledge
that takes as its object or regulates any aspect of any cognitive endeavor.’’ There
are two distinct referents involved in this definition: statable knowledge about the
cognitive system and its contents; and the effective regulation and control of that
system.
The first topic is concerned with knowledge about one’s own cognitive re-
sources. For example, mature thinkers know: that their immediate memory span
is limited, that some kinds of information (e.g., organized lists of items) are
easier to remember than others (disorganized lists); that some memory tasks
(e.g., recognition) are easier than others (recall); and what they might do to
remember a telephone number. That is, they have knowledge about ‘person,
task, and strategy variables’’ (Flavell & Wellman, 1977). Some features of this
knowledge, as reflected in the developmental literature, are that it is stable and
statable. It is stable in that once an individual acquires some pertinent fact about
cognition, he or she should continue to have that fact available; it is statable, by
definition, in that the knowledge is assessed through verbal reports.
The second set of activities includes those concerned with the self-regulatory
mechanisms, engaged during an ongoing attempt to learn or solve a problem.
These include checking, planning, monitoring, selecting, revising, etc. (Brown,
1978, Chapter 3, this volume). These are not stable, but depend on the subject’s
familiarity with the task, motivation, and the like. Individuals may be capable of
regulating their thought processes in one area, but not in another area. Also,
these activities are not necessarily statable, because much of self-regulation must
go on at a level below conscious awareness.
 120 CAMPIONE

The distinction between the awareness of resources and the operations applied
to those resources appears in various forms elsewhere in the literature (cf. Brown
& Campione, 1981). For example, it is essentially the same as one embodied in
Pylyshyn’s (1978) discussion of reflective and multiple access. Reflective access
refers to the ability to ‘‘mention as well as use’’ the components of the cognitive
system. It implies that the organism not only has the relevant information repre-
sented in memory, but also the ability to ‘‘represent the representing relation
itself”’; that is, the individual possesses not only fact X but also the notion of fact
X. Multiple access refers to the ability to use information flexibly. It implies that
information is not delimited to a constrained set of circumstances, but rather is
informationally plastic and can be ‘‘systematically varied to fit a wide range of
conditions which have nothing in common other than that they allow the valid
inference that, say, a certain state of affairs holds’? (Pylyshyn, 1978, p593):
A number of writers have proposed awareness of and control of specific
resources as criteria for the understanding of those resources. For example,
Moore and Newell (1974) argue that learners understand some routine when they
use it appropriately. Although not denying this component, Gardner (1978) and
Rozin (1976) suggest that a criterion of understanding is that knowledge be
available to consciousness and, perhaps, be statable. Similarly, Piaget (1978)
discussed the importance to human intelligence of the concept of reflected ab-
straction, with the result that cognitions be made statable and available to con-
sciousness, at which point they can be worked on and extended.
In the instructional work conducted, both types of metacognition have been
incorporated. Interventions have taken the form of either providing information
(statable knowledge) about, for example, the memory system or teaching the use
of a number of executive functions, such as planning, overseeing, and monitor-
ing. We are interested in, among other things, questions about the possible
differential effects of the two emphases. To provide some additional back-
ground, it is worthwhile to illustrate the way in which the distinction manifested
itself in the early memory development literature; work that served as
the impe-
tus for the instructional research. In fact, this can be done quite well
by reference
to one of the earliest papers in the area.
In a two-part experiment, Flavell et al. (1970) investigated the develop
ment
of preschool through fourth grade children’s ability to predict their
own immedi-
ate memory span and to monitor the current state of their learning
during an on
going study attempt. In the span estimation procedure, subjects
were given a
series of trials on which they were shown a number of picture
s—one on Trial 1,
two on Trial 2, three on Trial 3, etc., to a maximum of
10 trials. On each trial,
the experimenter showed Subjects the pictures, labeled them,
and then covered
the pictures up. Then the subjects were asked to predict
whether or not they
would be able to repeat the names (n); no recall was attempte
d. The procedure
continued until the subjects reported an inability to
repeat the set, or to 10 items,
the maximum tested. Each Subject’s span was set at
the longest series of names
 4. METACOGNITION AND PROBLEM LEARNERS 121

that could be recalled. After this process was terminated, each subject’s actual
span for those materials was assessed.
In the second part of the experiment, subjects were given a series of three
trials. On each trial they were to study a set of pictures (the number of pictures
for each subject was set equal to the actual span as determined in the first phase
of the experiment) for as long as they liked, until they were sure they could recall
all of the pictures in order. They were told to study the items for as long as they
wanted, and to ring a bell when they were sure they could repeat all the names in
the correct order. At this point, recall was attempted.
The data from each phase indicated clear developmental differences. In the
span estimation experiment, accuracy increased dramatically, with the mean
predicted; actual span differences were 3.7, 4.4, 1.6, and .6 items for nursery
school, kindergarten, second, and fourth graders, respectively. The consistent
tendency was for subjects who made errors to overestimate how much they
would be able to recall. For example, 60% of the nursery school and kinder-
garten children and 24% of the older children predicted that they would be able to
recall all 10 items, the maximum tested. Turning to the recall readiness data, a
clear developmental trend again emerged. Even though the task was not an
exceedingly demanding one (the number of items each subject had to study and
recall was equal to his or her span on this task and thus, demonstrably within the
subject’s capabilities), only 40% of the nursery school and kindergarten children
were able to recall without error on more than one of the three trials, whereas,
95% of the older subjects were able to do so. On the other hand, 12% of the
younger and 83% of the older children were consistently correct (no errors on
any of the three trials). Thus, although all subjects are able to recall the total set,
the younger children have a strong tendency to terminate study before they are
ready to recall, that is, they do not appear adept at monitoring their current state
of learning. This failure is also reflected in the study times for the various groups;
older subjects spent significantly more time studying the sets than did the young-
er subjects.
Although there are currently many more examples in each category, the data
from the initial study are sufficient to indicate that with development, children
come to know more about their memory system and show an increasing tendency
to oversee their cognitive activities. In addition to these developmental dif-
ferences, there are strong comparative differences; children with learning prob-
lems seem to perform poorly on assessments of cognitive abilities (see Brown,
1978, for some relevant data and a review).

THE STATUS OF INSTRUCTIONAL RESEARCH

was an
Around 1970, when metacognition work was beginning to appear, there
nal research being conducted . In fact, the
increase in the amount of instructio
as one of the favorite vehicles for
training study was beginning to emerge
 122 CAMPIONE

conducting basic research in the area of mental retardation. Why this occurred
has been spelled out in considerable detail elsewhere (Belmont & Butterfield,
1977; Brown & Campione, 1978; Butterfield, 1979) and is not of concern to this
discussion. Of relevance, is that it became clear that retarded children did poorly
on a variety of memory and problem-solving tasks, in part because they con-
sistently failed to produce the appropriate and necessary strategies (Brown,
1974).
In a host of studies, it was also shown that retarded children could readily be
taught to employ task-specific strategies, such as rehearsal and elaboration, with
the result that their performance would improve, often quite dramatically. It is
also fair to say that the results tended to be of limited practical significance. This
is not surprising inasmuch as the studies were done for purely theoretical reasons;
they were designed to answer basic questions about the nature of intelligence-
related sources of individual differences in memory performance. A prototypical
situation in which a training study would be conducted is as follows: There is a
specific task. Indications are that different groups of subjects perform differently
on that task, and experimenters would like to know why. To deal with this
question, one needs an analysis of what individuals must do to perform well on
that task and some hypothesis about the specific source(s) of individual differ-
ences. As an example, consider a memory task where it is assumed that effective
performance requires, among other things, the use of a rehearsal strategy.
It is
assumed also that retarded children, because of a failure on their part to
rehearse,
perform more poorly than nonretarded children. By training the retarded
children
to rehearse, both assumptions can be tested simultaneously. If their
performance
does improve significantly, it can be concluded that the original analysis
of the
task was correct (if rehearsal were not an important component
of performance
on that task, instructing individuals in its use would not lead to
improved perfor-
mance), and that retarded children perform poorly, at least in
part, because they
fail to rehearse without prompting (if they did rehearse spontaneously
, training
would not have been necessary). Note also that all that is required
to evaluate the
theoretical question is that the subjects execute the target
activity. To “‘teach”’
rehearsal, the experimenter may simply model the strategy,
tell the subject
exactly what to do, and then continue instructing and prompting
for the duration
of the study (e.g., Brown, Campione, Bray, & Wilcox,
1973).
These theoretical questions can be evaluated on the
basis of the subjects’
immediate response to training; in the early 1970s it
was typical for the research
to stop at that point. The contribution of this work is obvious
: it demonstrated the
potential for instruction-based improvement in perfor
mance: the improvements
were not trivial. In fact, given extant theories,
their magnitude was frequently
surprisingly large; it was this feature of the researc
h that encouraged those
interested in remediation, However, to stop the
research at that point requires
additional theoretical and practical questions.
Strategy training, even on the
training task, did not eliminate differences
between retarded and nonretarded
 4. METACOGNITION AND PROBLEM LEARNERS 123

children; the implication is that there are other sources of comparative dif-
ferences that remain ‘‘untreated.’’ To advance any theory of individual dif-
ferences in this domain, it is necessary to specify what those sources are; similar-
ly, the sources must be known in order to improve on instruction. Therefore, one
line of research, can address more detailed within-task analyses.
Considering the problem from a practical standpoint, demonstrations that
performance on a specific task can be improved when an instructor leads a
subject through a learning activity is not particularly overwhelming. Therefore,
the immediate and standard questions concern the maintenance of and gener-
alizability (Campione & Brown, 1977) of the instructed routines. Will the sub-
ject continue to execute the activity when the experimenter/instructor is absent or
ceases prompting; that is, will the activity be maintained? Will the subject apply
what he or she has learned to novel but related problems; that is, will the effects
of training be generalized effectively? If the answers to these questions are
negative, the instructional effects, although theoretically interesting, remain of
dubious practical significance. A second line of inquiry would then concern
between-tasks effects of training. (It may come as no surprise that both within-
and between-tasks analyses led to an emphasis on metacognitive factors.)
Reports of relevant research began to appear in the mid 1970s; Campione and
Brown (1977) provided an early review. At that time a brief summary of the
existing situation was that retarded children: (a) did not produce the kinds of
strategies necessary for efficient performance on a variety of tasks; (b) could be
taught to carry out strategies, which resulted in improvements in performance,
but not to the level of nonretarded comparison groups; (c) frequently abandoned
the strategy when the experimenter ceased prompting its use, however, extensive
training might overcome that problem; and (d) failed to apply the strategies to
new problems where they would be appropriate (i.e., failed to recognize an array
of problem isomorphs). In some sense, points (a) and (d) can be regarded as
similar; a novel problem is presented, and the subject does not bring to bear the
resources he or she has available to deal with the problem. The question, of both
theoretical and practical significance, then became, Why do failures in applica-
tion arise?
It was at this point that people involved in mental retardation training work
became attracted to the work on metamemory. Failures to produce strategies, to
carry them out most efficiently, or to transfer them widely might be due to a
failure on the part of the learner to understand the significance of the instructed
activities and procedures. Retarded children might not know enough about the
memory system to appreciate why the strategies were necessary. Also, they
might fail to monitor the effects of instructed strategy use and thus, fail to carry it
out as well as might be expected. Therefore, they would derive less than the
maximum benefit and/or not realize that the strategy was helpful. As a specific
example, consider a child faced with the task of remembering 10 items. If he or
she believes that the items can be readily remembered (overestimates his or her
 124 _CAMPIONE

span), there is no reason for engaging any specific learning activity, even if one
were available. Further, if the learner did not monitor performance on recall
trials, he or she would not become disabused of the original notion. In addition,
if instruction is provided, and if performance improves, it is reasonable to expect
that the learner will continue to employ that strategy only if he or she monitors its
use and notes that it has actually helped; if no such monitoring takes place, it is
not surprising that transfer is limited.

METACOGNITION AND INSTRUCTIONAL ATTEMPTS

Failure to produce strategies or to transfer them might be seen as reflecting a

general metacognitive deficit. The first step in evaluating this possibility was to
collect some metacognitive data on retarded children to assess their competence
in this domain. In a series of laboratory studies (Brown, 1978: Campione &
Brown, 1977), it became clear that retarded children did experience particular
problems in this area. They are not as insightful about their memories as non-
retarded children of comparable age. For example, they do not appear to be
aware of the severity of working memory limitations, and drastically overesti-
mate their memory span (Brown, Campione, & Murphy, 1977). Also they are
not as cognizant of the contents of their memory. In a study on the feeling
of
knowing phenomenon, Brown and Lawton (1977) found mildly retarded subjects
less able to predict their ability to recognize a name they had previously
been
unable to recall, than nonretarded children. As a final example, retarded
children
do not appear to know as much as nonretarded children, about
the kinds of
material that are easy or hard to remember. Retarded children were
asked to
generate lists of items that would be easy to remember. They were
given a series
of key words and asked to generate items which would be easy to
remember in a
list, including the key items and additional items. In response
to this request, the
retarded children were less likely than nonretarded children
(Tenney, 1975) to
construct organized lists; and when they did include items from
which a (tax-
onomic or thematic) category, their choices were more likely
to be ‘“‘broad’’ (1 4
2 7) than ‘‘narrow’’ (1 2 3 4) (Brown, 1978).
Executive decision making also presented difficulty. One
early demonstration
of the severity of the problem for retarded children was
made in the context of an
attempt to teach retarded adolescents a rehearsal Strateg
y (Butterfield, Wambold,
and Belmont, 1973). In addition to providing all
of the components of the
Strategy, the subjects were forced to provide detailed
instruction in the coordina-
tion and sequencing of those components. Even
when each component was
taught in excruciating detail, direct instruction in
their management was also
required. In our own work, it was found that
retarded children do not tend to
monitor their state of learning accurately (Brown
& Barclay, 1976; Brown,
 4. METACOGNITION AND PROBLEM LEARNERS 125

Campione & Barclay, 1979), nor do they seem to allocate study time differen-
tially to items that have been hard or easy for them to remember (Brown &
Campione, 1977).
Given the fairly impressive evidence that retarded children do experience a
variety of metacognitive problems and the hypothesized relation between meta-
cognition and transfer, it seemed obvious that training attempts should include a
metacognitive component. It was also easy to rationalize the transfer failures
existing in the literature; they could be regarded as attacking the symptom, but
not the cause, of the problem. The typical training study, termed blind training
by Brown, Campione, and Day (1981), did not in any way involve the students
as active conspirators in the training process. They were induced to use a strat-
egy, or even tricked into carrying out the desired activities; no attempt was made
to impart to them any reason why those activities might be helpful. For example,
the child is taught to use a cumulative rehearsal strategy by initially copying an
adult but is not explicitly told why acting that way helps performance or that it is
an activity appropriate to a certain class of memory situations. If the original
problem was, due in part, to metacognitive limitations, there is nothing in the
instructional interaction that would alter that state of affairs; hence the need for
modified training approaches.
_ There was one additional reason for believing that training attempts involving
general metacognitive skills would have widespread effects. One problem with
the skills that were the object of training originally, was that they were specific to
a very small class of situations. For learners to generalize the effects of instruc-
tion in the use of specific routines, they would have to be able to discriminate the
situations in which the routine would be appropriate from those in which it would
not. Adequate generalization requires both extended use in novel situations and
decisions not to use the trained activity in situations where it would not be
beneficial. In the case of truly general skills, this discrimination would not be
necessary; the skill or routine could simply be used in a whole battery of problem
solving situations, eliminating the need for a detailed analysis of the task being
attempted (Brown & Campione, 1978).
This leads to one of the oldest and most discussed topics in the problem-
solving area, the question of the generality/specificity of problem-solving skills.
Are there truly general skills? No attempt is made to enter into that discussion;
later it becomes clear that our preference (see Campione, Brown, & Ferrara,
1982) in this domain is to borrow the expanding cone metaphor proposed by
Newell (1980). According to this view, there are general skills; however, their
generality comes at the expense of power. General skills cannot solve problems
by themselves but may provide access to the more specific and powerful skills
that are necessary for problem solution. In this context, the questions are: Are
(some of) the self-regulatory mechanisms general ones? If so, can they be taught
alone or only in combination with the more specific skills to which they allow
access?
 126 _CAMPIONE

Given these considerations, researchers have sought to incorporate theory and

data from the metacognitive domain into their instructional packages. The blind
training studies have been replaced or supplemented in a number of ways. One
method of classifying the different approaches is in two dimensions: the kind of
metacognitive supplement involved (knowledge or regulation); and the presence
or absence of additional specific skills training. Reviewing some of the instruc-
tional work, Brown, Campione and Day (1981) identified a three leveled classifi-
cation of training studies, the aforementioned blind studies; informed training,
where the learner is given some information about him or herself, a task, or a
strategy; and self-control training, in which the learner is given explicit instruc-
tions about the monitoring, checking, or evaluating of some of his or her cog-
nitive resources. The latter two correspond to the two ‘types’ of metacognition
discussed earlier. In either case, the metacognitive component can be taught
either independently (as an object in itself) or in conjunction with a specific skill
or strategy. Examples of each type of study and a statement of the benefits of
each approach, is discussed.

Knowledge Alone. Retarded children do not know much about their memo-
ries. One hypothesis that might be entertained is that teaching them some facts
about their memory might have a somewhat general effect on performance. One
reason that retarded children might fail to produce strategies spontaneously
or
maintain them after (blind) training is that they do not know that the strategies
are
necessary; they overestimate their ability and believe they can remember
the
information without any undue effort. Brown, Campione, and Murphy (1977)
showed retarded children with mental ages (MAs) of 6 and 8 a set of 10
pictures
and asked them to predict how many of them they would be able to recall.
Only
about 25% of the children could be classed as “‘realistic’’ (the difference
be-
tween their actual and predicted spans was no more than two
items) in their
predictions. The unrealistic estimators always overestimated
their ability; most
predicted that they could recall all 10 pictures. Given this result,
training seemed
appropriate.
The training consisted of a series of 10 predict-recall pairs. The
subject would
be shown an array of ten items and asked how many of them
he or she would be
able to recall (exactly as on the initial assessment). The
items were then hidden
and recall assessed. For half the Subjects, explicit feedbac
k was given after each
pair: “You predicted that you would recall , and you actually recalled
. For the remaining half, no such feedback was provide
d. After training,
the subjects were given a series of maintenance and
generalization tests.
The major findings were as follows. These subject
s could be induced to
become realistic, although it was not easy. For
example, none of the younger
(MA 6) children became realistic unless provided explici
t feedback about their
performance. They would predict 10 and recall 4
consistently; the fact that over
the 10 trials they could never recall more than
half the items did not affect their
 4. METACOGNITION AND PROBLEM LEARNERS 127

predictions. With feedback, 62% became realistic, but the effect was not a
durable one; two weeks later, they reverted to overestimation. For the older
children, 65% became realistic following training; however only those given
explicit feedback remained realistic one year later. Thus, explicit feedback was
necessary to bring about any change in the younger children and to bring about a
durable change for the older ones.
One other aspect of the data from the training phase of the experiment is
important to this discussion. A possibility is that the subjects who became real-
istic after training did not learn anything about their memory, instead, they
learned a simple verbal response (e.g., to say “‘four,’’ whenever they were asked
a prediction question); this is very unlikely to be true. The prediction-recall series
included two types of sets of items. On half the trials, the sets were categorized,
and on the remaining half, they were not. All subjects actually recalled more on
the categorized sets. Of the originally realistic subjects, about 75% predicted that
they would recall more on the categorized, as opposed to uncategorized, sets; of
the subjects who became realistic through training, almost 90% made this accu-
rate prediction; finally, of those who remained unrealistic even following train-
ing, only 50% predicted better recall on the categorized sets, that is, their
predictions were at the chance level. The result indicates that the subjects who
were trained to be realistic were doing more than simply learning a single
number; they came to realize that categorized lists would be easier to recall than
uncategorized ones, even though that manipulation was never explicitly
mentioned.
This was an encouraging beginning. The results of a number of generalization
probes were quite dramatic; however, only one of the experiments is discussed
here. During the initial phase of the experiment, subjects’ predictions and spans
were assessed by showing them a 10 item array, asking them how many they
would be able to remember, and later having them attempt recall. As a transfer
test, we used Flavell, Friedrichs & Hoyt’s (1970) method, in which the subject
was asked if he or she could remember one item, then two items, and so on. The
subjects who were realistic estimators prior to training in the initial phase of the
study (25% of the sample) were also realistic on this task; evidence that the same
skills are being assessed. However, none of the subjects who became realistic
only after training, continued to be realistic on this variant of the training task.
Teaching retarded subjects facts (at least this fact) about their memory is not
very profitable. As strategies taught in isolation tend to be welded to the learning
context, facts taught in isolation can suffer the same fate, a finding that is not
exactly new (see Bereiter & Scardamalia’s, 1980, discussion of inert
knowledge).

Strategies Plus Knowledge. In a number of studies, strategy training has

been supplemente d by information about the effects of that strategy. The argu-
ment is that trained subjects may abandon an instructed routine when prompting
 128 CAMPIONE

is withdrawn, because they do not realize that they performed better when using
the strategy, or they do not realize that it may be helpful on more than one task.
One way to deal with these problems is to simply provide subjects with that
additional information; either directly or indirectly. For example, in a study with
nonretarded children, Kennedy and Miller (1976) found that an instructed re-
hearsal strategy was more likely to be maintained in the absence of experimenter
prompts if it had been made clear that the use of the strategy did result in
improved performance. A similar effect with retarded children has been obtained
by Kendall, Borkowski, and Cavanaugh (1980). In a similar vein, Burger,
Blackman, Holmes, and Zetlin (1978) taught retarded adolescents to use an
organizational strategy to facilitate recall of a list of pictures. Their students were
given practice in sorting items into categories and were informed that this would
help them remember. Again, the result was improved performance on the train-
ing task and increased maintenance of the activities by the students when faced
with subsequent similar problems.
In these studies, the subjects were told about the effectiveness of the in-
structed routines. It is, of course, also possible to use less direct methods of
providing information to subjects. For example, if transfer is desired, it is impor-
tant that subjects be aware that a class of strategies may be applicable in a variety
of situations. One approach would be to simply tell them. A more indirect
approach was employed by Belmont, Butterfield, and Borkowski (1978): they
resorted to training in multiple contexts. Their retarded subjects were trained to
rehearse in either one or two situations. This indirect manipulation had the
desired effect; the twice-trained group showed more evidence of transfer to a
third (similar) rehearsal task.
To summarize, supplementing strategy training by providing additional infor-
mation about their effectiveness or range of utility does lead to increased mainte-
nance. It does seem to allow the experimenter to withdraw prompting without
having the subjects abandon the trained routines, at least over short intervals. At
this point, whether it is likely to lead to much generalization is unknown, for
a
number of reasons, it seems unlikely that it would. Recall the span estimation
experiment described in the previous section. In that experiment the provision
of
explicit feedback was also necessary to effect either learning (for the younger
subjects) or maintenance (for the older subjects); even when long-term mainte-
nance (one year) was achieved, there was no evidence for generalizati
on.
Why would generalization not be expected? To answer that question
, another
question can be asked: Why is it necessary to provide explicit feedback
in the
first place? A possibility is that the Subjects do not monitor their
own perfor-
mance levels spontaneously; thus, they would not know that the
strategy resulted
in enhanced recall. (It is not that they cannot monitor when told
to. For example,
in the span estimation study, subjects who were not given
explicit feedback
could, when asked, report how many items they had just
recalled.) Provision of
feedback would be helpful; however, it would still not
address the underlying
 4. METACOGNITION AND PROBLEM LEARNERS 129

cause of the problem. Again, the symptom (lack of awareness of performance

levels) would be treated, but the underlying cause (failure to monitor perfor-
mance) would not be. Alternatively stated, the necessary product would be
provided, but not the processes leading to that product. As the product is, in fact,
task specific and the processes are more general, it must be the latter that is of
more importance; this leads to the next group of experiments.

Self-Regulation. The initial failures to obtain transfer of specific skills pro-

duced an increased interest in dealing with more general, transituational skills. In
some cases, there appeared to be an either-or approach to the specific-general
issue. For example, Butterfield and Belmont (1977) argued that attempts should
be made to provide direct training in executive functions rather than in specific
strategies. Candidates for general executive routines include the skills discussed
throughout this paper: planning, checking, monitoring, etc. In the cognitive
domain, this movement received further impetus from the cognitive behavior
modification programs stimulated, in part, by Meichenbaum’s (1977) research.
In that work, teaching general self-regulatory routines, such as continually
monitoring progress through some task, questioning whether they were remain-
ing on task has helped a variety of clinical populations. These routines are
successful, and some (e.g., Belmont, Butterfield, & Ferretti, 1982) have argued
that they also result in generalized effects of training.
Some problems exist with the cognitive behavior modification-cognitive skills
training alliance (see Brown & Campione, 1982); the main problem is that the
situations being analyzed in the two areas are quite different. From the cognitive
behavior modification research, there is often good reason to believe that many
of the specific skills required for efficient performance are relatively intact and
that for some patients (e.g., some impulsive children), the major problem may be
a failure to use those resources effectively. In that case, self-regulation instruc-
tion by itself, could be very powerful. In the cognitive domain, slow-learning
children are often lacking specific skills as well as regulatory mechanisms. The
result is that concentrating exclusively on regulatory mechanisms is unlikely to
be of much help; whether or not self-regulation training will help depends on
other factors.
As an example, consider an experiment by Brown and Campione (1977); the
object was to train retarded children to allocate their study time effectively. The
situation, developed by Masur, McIntyre & Flavell (1973), is an analogue of
many study situations. A test is upcoming, and the student does not have suffi-
cient time to study all the material. He or she must then decide which portions of
the material have been well learned versus those which have not been learned.
The strategic decision would be to spend more time studying the less known
material. In the experimental situation, subjects are given a multitrial, free recall
task involving N items. They study the N items and then attempt recall. Then,
they are given time for further study; however, they are allowed to have only half
 130 CAMPIONE

the set available to them during that period. The question of interest concerns the
items they choose for study. According to the data from Masur et al., college
students and third grade children select previously unrecalled items for study;
that is, they allocate more attention to the difficult (for them), unrecalled items.
In contrast, first graders seem to choose randomly.
The study by Brown and Campione (1977) included two groups of retarded
children, who were tested in the same situation. They saw and studied a set of 12
pictures, and then attempted recall. They were then allowed to select six pictures
for further study; were told that they would have to recall all 12 pictures. During
this test phase, their selections were essentially random; they showed no prefer-
ences for selecting study items they had missed earlier (subsequent tests demon-
strated that they were very accurate in indicating the items they had recalled on
the previous trial). During the intervention phase, the ‘‘correct executive deci-
sions’’ were made for some of the subjects. The experimenter selected the six
items the students would have available for further study. In one condition, she
gave the students items they had missed previously; in another, she gave them a
random set of six items. Thus, in one condition, the mature selection was
simulated; in the other, the more primitive one.
The effect of this manipulation was different for the two groups. For the
older, more academically advanced group, the subjects given previously missed
items performed significantly better on the recall task than those given a random
set. However, for the younger group, the treatments did not differ. Simulating
the “‘correct’” executive decision making was effective for one group, but not for
the other. The proposed explanation for the difference was that selecting missed
items for study was only one part of an overall plan for dealing with the task. It
was presumed that subjects would have to identify missed items, select those
items for study, and maintain in memory, possibly by means of rehearsal,
the
previously recalled items. For the older subjects, this latter step was apparently
within their capabilities, and the intervention was successful: for the younger
subjects, it was not, and the intervention failed. This experiment
demonstrates
some of the risks associated with training general skills in isolation. If
some of
the more specific routines are not available (or inferrable) to the learner.
the
training is bound to be ineffective.

Strategies Plus Regulation. In a number of earlier studies, it became clear

that even when strategies and/or their various components were
taught directly to
retarded or young children, they did not regulate them effectiv
ely. The Butter-
field et al. (1973) experiment, cited earlier, provides one
type of example. They
taught their subjects a rehearsal strategy and a corresponding
retrieval strategy
(corresponding in the sense that the rehearsal strategy would
be effective only if
the appropriate retrieval plan were also used). Teachin
g those components sepa-
rately was not sufficient to lead to optimal performance;
Butterfield et al. also
had to teach their subjects how to coordinate the two
plans. In a number of other
 4. METACOGNITION AND PROBLEM LEARNERS 131

situations where subjects of varying cognitive maturity are taught specific strat-
egies, there is evidence that they ‘‘execute’’ the strategies in accordance with
instructions, but do not monitor them spontaneously. For example, Belmont and
Butterfield (1971) trained retarded subjects to rehearse in a short-term memory
task (the subjects saw a series of six consonants exposed sequentially and were
then asked to indicate the position in which a randomly selected one had oc-
curred). Their procedure allowed subjects to pace themselves through the list in
preparation for the recall test, and thus to spend as much time as they liked
studying. Rehearsal training involved having the subjects expose the first four
items quickly and then pause to rehearse that set of four items; they could
rehearse as long as they wanted before going on to study the final two items. In
fact, they were told to pause long enough to repeat the set three times.
The retarded subjects did pause appropriately, and performance did increase,
but not as much as could be expected or hoped. The presumption is that they
executed the strategy in a rote fashion and did not monitor how well learning was
proceeding during the pause interval. That is, they did rehearse but ceased that
activity before the items were learned well enough to guarantee excellent recall.
In a slightly different paradigm, a similar effect with young children has been
obtained by Flavell et al. (1970). As described earlier, younger children in their
recall readiness study were less likely to assess their readiness for recall than
older children. Given unlimited study time, they studied for shorter periods and
did not recall as well.
The recall readiness paradigm was borrowed by Brown and Barclay (1976); it
seemed an excellent vehicle for investigating the combined effects of strategy
training and instruction in self-regulatory activities. These authors investigated
the performance of groups of mildly retarded children differing in MA (6 vs. 8);
the recall readiness paradigm was somewhat modified. In an initial phase of the
experiment, each subject’s memory span for pictures was determined. In the
subsequent phases of the research, subjects were given a series of trials in which
they were asked to study and recall 1/2 times their span—rather than just the
span, as in Flavell et al. (1970). This was done to make the task more demanding
for the students.
On each recall readiness trial, the subjects were told that they could spend as
much time studying as they wished and that they could look at the pictures in any
order and as often as they wanted. They were also told to ring a bell when they
were sure they could recall all the items in left-to-right order. On a pretraining
series of trials involving these supraspan lists, their performance was extremely
poor. Even though they had indicated that they were ready for recall, they were
able to recall only about % of the items. Only one of 27 younger children and
five of 39 older children were able to recall even one list (of three) perfectly.
During the training portion of the study, children were taught strategies that
could be used to facilitate their learning of the lists, along with the overseeing or
monitoring of those strategies. The latter aspect of training was accomplished by
 132 CAMPIONE

employing strategies that included a self-testing component and by telling the

children to monitor their own state of learning. For example, in a rehearsal
condition, the subjects were told to break the list down into manageable subsets
(three items) and to rehearse those subsets separately. They were also instructed
to continue rehearsing the subsets until they were sure they could recall all of the
items. Note that one can continue to rehearse all the items (including prior
subsets) only if one can remember them well enough to produce them for re-
hearsal (Flavell, 1979). Thus, in this situation, rehearsal both facilitates learning
and provides a check on the state of learning. Anticipation was another trained
strategy that included self-testing features. The children were instructed to try to
remember the name of a picture before they pressed the window exposing it. Ina
final condition, labeling, children served as a control group; they were told to go
through the list repeatedly, labeling each item exposed. In all conditions, it was
emphasized that children should engage in the instructed activity until they were
sure they were ready to recall all the picture names.
The immediate effects of training were impressive. For the rehearsal and
anticipation groups, recall improved significantly; one index, 13 of 18 younger
subjects in these conditions had at least one trial of perfect recall, compared with
0 of 18 before instruction. Of the 26 older children given self-control training,
the figures were 2 before instruction and 24 after instruction. In addition, for the
older (MA 8) subjects, these effects were extremely durable; differences between
the trained and control subjects were still significant on a retest given one year
after training.
Also of considerable import, on a test of generalization, these trained students
outperformed a number of control groups (Brown, Campione, & Barclay, 1979).
On that test, the subjects were given a series of prose passages to study. As in the
original experiment, they were told that they could study each passage for as long
as they liked, until they were sure they would be able to recall all the important
information; however, no mention was made of the earlier study. The rehearsal-
and anticipation-trained students recalled more of texts and also showed some
evidence of better comprehension than the did the controlled subjects.
Another relevant study (Day, 1980) compared, within one experiment, a
number of the approaches outlined in this discussion. The subject population and
the type of learning activity were different from the others described. The experi-
menter was interested in the ability of junior college students to summarize prose
passages. The students were divided into two groups: ‘‘normal’’ students with no
identified reading or writing problems (who were, however, reading at only
the
seventh grade level); and remedial students, who although purportedly of normal
reading ability, had writing problems.
Within each of the two ability groups, there were four instructional condi-
tions: (a) self-management—the students were given general encourag
ement to
write a good summary, to capture the main ideas, to dispense with
trivia and all
unnecessary words, but they were not given explict rules for
achieving these
ends; (b) rules—the students were given explicit instructions
and modeling of
 4. METACOGNITION AND PROBLEM LEARNERS 133

the use of a predetermined set of rules (Brown & Day, 1980) (e.g., they were
shown how to delete trivia, delete redundant information, write in superordinates
for lists of items, underline topic sentences for inclusion in their summary, and
invent topic sentences for paragraphs not containing them [see Day, 1980, for a
more detailed discussion of the rules]); (c) rules plus self-management—the
students were given the general self-management instructions of the first group
and the rules instructions of the second group, but they were left to integrate the
two sets of information for themselves; (d) control of the rules (strategies plus
regulation in the current terminology)—involved training of the rules, as in the
second group, and additional training in the control and overseeing of those rules
(i.e., the students were shown how to check that they had a topic sentence for
each paragraph, to check that all redundancies were deleted, etc). For the stu-
dents in this condition, the integration of the rules and appropriate self-regulatory
routines were explicitly modeled.
In this discussion attention is restricted to a small subset of the data, and
concentration is given to the use of the most difficult rules—select topic sentence
and invent topic sentence. The invent topic sentence rule is much more difficult
than the selct topic sentence rule. Consider first the select data. All training
conditions had some effect; however, for the less sophisticated learners, the most
effective condition was the most explicit training, control of the rules. Training
in rule use alone did help performance, but adding the general self-managment
instruction did not provide any additional help. The poorer students did not
integrate the rules and self-managment instructions for themselves; they needed
explicit instruction in rule control to maximize their performance. The more
sophisticated junior college students responded better to all forms of training and
were able to integrate the general self-management and rule training for them-
selves; there was no difference between conditions 3 and 4.
A similar pattern was obtained with the very difficult invention rule. Accord-
ing to Brown and Day (1980), even regular college students used the rule on only
50% of the occasions where it would be appropriate. The lower ability junior
college students in Day’s research, only improved with the most explicit train-
ing, condition 4, and then only slightly. The higher ability junior college students
showed more improvement, however, in order to achieve maximum performance
levels, explicit instructions in control of the rules were required. For the easier
select rule, conditions 3 and 4 were equal; with the more difficult rule, condition
4 is superior to condition 3. As the complexity of the rule being taught increases,
the need for additional instruction in the monitoring of that rule also increases.

CONCLUSIONS

Although it would be useful to have more data from studies representing the
different categories, some clear patterns are emerging. One pattern, which has
not been emphasized in this discussion, is that the type of instruction necessary to
134 CAMPIONE

effect meaningful change in performance varies with the maturity of the learner
and the complexity of the skills being taught. This was mentioned directly in the
discussion of Day’s (1980) work, but only in passing elsewhere. In each of the
studies from our laboratory (e.g., Brown & Barclay, 1976; Brown & Campione,
1977; Brown et al., 1977; Brown et al., 1979), instructions that produced effects
with more mature (typically MA 8) retarded children were insufficient to bring
about comparable results with younger (typically MA 6) subjects. In fact, in all
of those studies, as in Day’s, when the same intervention was applied to groups
of subjects differing in ability, differences between the two groups increased.
Typically, the different groups were approximately equal prior to training; fol-
lowing reasonable training, the more capable group was performing significantly
better than their less capable counterparts. The implications of these findings for
the design of instructional packages, notions of individual differences, and theo-
ries of intelligence have been discussed in some detail by Campione et al.
(1982).
As illustrated by Day’s data, these ‘‘between-groups’’ differences in response
to instruction can also be captured within subjects. Although less cognitively
mature learners require more complete and explicit instruction than more mature
learners within a homogeneous group of learners the need for such complete
instruction varies with the difficulty of the task: more complete instruction was
needed to bring about use of the invent topic sentence rule than the select rule.
Ignoring these complexities, consider for example, low ability subjects and
fairly complex tasks: What can one conclude from this review of intervention
studies? Several features seem clear. One is that the potential effects of the
varying approaches differ. On the basis of the one relevant study reviewed, it
seems improbable that simply providing facts about the cognitive system is likely
to be very productive. Given that low ability subjects do not have that informa-
tion available to them and that a case can be made for the importance of that
information, teaching such knowledge seems reasonable. Within some domains
experts differ from novices, in terms of domain-specific knowledge; providing
that information might be one way to narrow the expertise gap. However, the
clear problem for slow learning children remains; it is the same problem that led
to the metacognitive work in the first place. Availability of knowledge, either
declarative or procedural, does not in itself guarantee flexible access to and use
of those resources.
Providing knowledge to supplement skills training does produce some desir-
able effects. The most obvious one is that it can lead to maintenance of instructed
routines. As we have seen, this effect can be obtained fairly easily and should not
be taken too lightly. There are many situations where flexibility and creativity
are not particularly necessary; where there are fairly specific routines that
do
need to be applied repeatedly. The contrast is between domains where there
are
many powerful and specific routines and ones where there may be a much more
limited number, possibly only one or two. In the former case, access
to and use
 4. METACOGNITION AND PROBLEM LEARNERS 135

of the particular procedures needed in a given problem is important; in the second

case, the recognition conditions are much simpler. In the limiting case, where
there is one routine to be used, the goal would be to guarantee that the routine is
engaged; maintenance is all that is required, and the provision of explicit infor-
mation about the recognition conditions and effectiveness of the instructed ap-
proaches may be all that is necessary.
It is known that low ability performers are deficient in both knowledge and the
resources for controlling that knowledge; however, providing them with relevant
information is not very effective. Why? One answer is that deficient knowledge
is simply a symptom of the more important and pervasive problem; knowledge
differences reflect processing differences. Poor learners are deficient in knowl-
edge because of poor learning skills, in particular, the self-regulatory, executive
mechanisms illustrated earlier. Providing the products of that processing (e.g.,
specific knowledge about the memory system) does not affect the underlying
problems. It is worth noting that an asymmetry is clearly possible. For example,
teaching subjects to monitor performance may well lead to the development of
factual knowledge about the memory systems (e.g., span limitations and strategy
effectiveness).
If this is true, more impressive differences can be expected only if the general
executive processes are the subject of training. The preliminary data indicate that
this is the case. Teaching (older) retarded children self-checking routines in
conjunction with memory strategies leads to impressive immediate training ef-
fects, extremely durable outcomes, and evidence for generalized application of
those training experiences. The differential effects of knowledge provision and
regulation instruction can be seen in another way. Many of the subjects who
transferred so well in the recall readiness study were the same ones who failed so
dramatically in the span estimation experiment.
The general view is the same as that reached by Feuerstein (1980), among
others. His Instrumental Enrichment program, which has been insightfully ana-
lyzed by Bransford, Stein, Arbitman-Smith, and Vye (in press), is essentially a
program aimed at inculcating metacognitive skills. Feuerstein begins by attempt-
ing to deal with the same problem that motivated much of the metacogni-
tion/instructional work; poor academic achievers experience particular problems
with transfer. His goal is to provide students with the skills of learning that allow
them to deal with unfamiliar problem spaces; the general skills he focuses on are
the self-regulatory and executive ones emphasized by those in the metacognitive
arena (see Bransford et al., in press, and Campione et al., 1982, for further
discussions).
The mention of Feuerstein’s program leads into the final issue to be discussed.
Presuming self-regulatory skills are to be taught, should they be taught separately
or in conjunction with the more specific skills they are to control? Feuerstein’s
position is that those skills should be imparted as content-free activities in a
setting of their own, rather than primarily in the context of specific content areas.
136 CAMPIONE

He also advocates the use of special materials, designed specifically as vehicles

for teaching the skills. Feuerstein is certainly aware of the potential problems
with this approach, namely, that the goal is to have these skills transferred to and
used in traditional academic areas. His main mechanism for bringing this about is
via ‘‘bridging’’; the instructor points out various situations where the skills are
applicable.
Feuerstein’s overall argument, which includes a number of important prag-
matic/social/motivational aspects, is an important one and is supported by some
of the data reviewed. The first comment is straightforward; teaching general
components of performance in the absence of the more powerful domain-specific
skills, cannot work. Although, as argued earlier, there are certainly cases where
such instruction can bring about impressive effects, it is necessary to have a
detailed analysis of the domain and of the resources a learner can bring to bear on
that domain, before an intelligent decision about “‘pure’’ self-regulation instruc-
tion could be made. The specific example cited was that inducing some children
to apportion their study time ‘‘appropriately’’ did not result in any improvements
in performance; instruction would have to be supplemented with specific skill
information to be effective.
Feuerstein’s assumption, of course, is that the teaching of specific content-
dependent procedures is taking place in other classroom settings. Even if this
were happening, the data suggest that it might still be necessary to include
explicit instruction regarding regulatory mechanisms in conjunction with some
specified skills. This fact is a reflection of one of the features of this class of
metacognitive skills; they are not stable. Even if the learning of regulatory
mechanisms is proceeding smoothly, their use by any individual is likely to vary
with task requirements. As pointed out elsewhere (e.g., Brown, Chapter 3, this
volume), there are conditions where executive activities are relatively less likely
to be employed. For example, the routines may compete for attentional resources
with the operations they are overseeing; it might be expected that they would be
most likely to be used effectively and spontaneously with simple or well-learned
routines. When new, and particularly difficult, skills are being taught, it may be
necessary to include self-regulation training, even for subjects who, on other
occasions, have been known to engage those skills themselves (see Day, 1980).
The general conclusion is that effective performance within some domain
requires knowledge about that domain, specific procedures for operating within
the domain, and more general regulatory processes that are task independent;
slow learners experience problems in all three areas. Presumably, the most
effective teaching plan would attend to all three areas. As the data indicate,
attention to only one category is unlikely to be sufficient to bring about major
effects. The extent to which all three need to be treated explicitly, within any
instructional package, varies with the ability and starting resources of the learner,
the complexity of the target activities, and the goals of the instruction. Although
no single factor is sufficient, each leads to worthwhile outcomes, and a single
 4. METACOGNITION AND PROBLEM LEARNERS 137

package incorporating all three areas would be expected to yield truly impressive
results. An evaluation of that statement, however, would require an experimental
approach different from that represented in the research reviewed earlier; in that
work, relatively few training sessions were included. An attempt to borrow as
much as is now known about metacognition and apply it seriously to the topic of
instruction would require a much more extended intervention. Researchers have
begun to do such experiments, and the results of these endeavors should be
available in the near future.
In summary, the nature of instructional efforts with slow-learning children has
changed dramatically since the emergence of interest in metacognitive skills.
Rather than trying simply to impart specific skills, the programs now attempt to
foster understanding of the routines by providing knowledge about the domain
and the place of the skill(s) within the domain. Close attention is also paid to
imparting the regulatory schemes necessary to make maximal use of resources.
In these endeavors, an overriding emphasis is on making learners aware of
themselves and of the availability and potential use of cognitive resources. These
changes have been positive ones; the outcomes of current training studies are
more dramatic than prior ones, and their relevance to educational issues is much
clearer.
The instructional data relates to more general issues regarding the topic of
metacognition. The knowledge/control distinction is powerful, enough to make
it preferable to refrain from grouping the two categories together under one
category; the data also indicates primacy of the control component. One sug-
gestion is that the knowledge of the cognitive system, which develops with age
and expertise, is the product of the operation of the general executive
mechanisms.

ACKNOWLEDGMENTS

Preparation of this manuscript was supported in part by U.S. P.H.S. Grant HD-05951.

REFERENCES

Baker, L., & Brown, A. L. (1981). Metacognition and the reading process. In D. Pearson (Ed.), A
Handbook of reading research. New York: Plenum Press.
Belmont, J. M., & Butterfield, E. C. (1971). Learning strategies as determinants of memory
deficiencies. Cognitive Psychology, 2, 411-420.
Belmont, J. M., & Butterfield, E.C. (1977). The instructional approach to developmental cognitive
research. InR. V. Kail & J. W. Hagen (Eds.), Perspectives on the development of memory and
cognition. Hillsdale, NJ: Lawrence Erlbaum Associates.
Belmont, J. M., Butterfield, E. C., & Borkowski, J. G. (1978). Training retarded people to gener-
alize memorization methods across memory tasks. In M. M. Gruneberg, P. E. Morris, & R. N.
Sykes (Eds.), Practical aspects of memory. London: Academic Press.
138 CAMPIONE

Belmont, J. M., Butterfield, E. C., & Ferretti, R. P. (1982). To secure transfer of training, instruct
self-management skills. In D. K. Detterman & R. J. Sternberg (Eds.), How and how much can
intelligence be increased. Norwood, NJ: Ablex.
Bereiter, C., & Scardamalia, M. (1980). From conversation to composition: The role of instruction
in the developmental process. In R. Glaser (Ed.), Advances in Instructional Psychology (Vol. 2).
Hillsdale, NJ: Lawrence Erlbaum Associates.
Binet, A. (1909). Les idees modernes sur les infants. Paris: Ernest Flamarion.
Bobrow, D. G., & Norman, D. A. (1975). Some principles of memory schemata. In D. G. Bobrow
& A. Collins (Eds.), Representation and understanding: Studies in cognitive science. New York:
Academic Press.
Bransford, J. D., Stein, B. S., Arbitman-Smith, R., & Vye, N. J. (in press). Three approaches to
improving thinking and learning skills. In S. Chipman, J. Segal, & R. Glaser (Eds.), Cognitive
skills and instructions. Hillsdale, NJ: Lawrence Erlbaum Associates.
Brown, A. (1974). The role of strategic behavior in retardate memory. In N. R. Ellis (Ed.),
International review of research in mental retardation (Vol. 7). New York: Academic Press.
Brown, A. L. (1978). Knowing when, where, and how to remember: A problem of metacognition.
In R. Glaser (Ed.), Advances in instructional psychology (Vol. 1). Hillsdale, NJ: Lawrence
Erlbaum Associates.
Brown, A. L. (1980). Metacognitive development and reading. In R. J. Sprio, B. Bruce, & W.
Brewer (Eds.), Theoretical issues in reading comprehenson. Hillsdale, NJ: Lawrence Erlbaum
Associates.
Brown, A. L. (1982). Learning and development: The problem of compatibility, access, and induc-
tion. Human Development, 25, 89-115.
Brown, A. L., & Barclay, C. R. (1976). The effects of training specific mnemonics on the meta-
mnemonic efficiency of retarded children. Child Development, 47, 71-80.
Brown, A. L., & Campione, J. C. (1977). Training strategic study time apportionment in educable
retarded children. Intelligence, 1, 94-107.
Brown, A. L., & Campione, J. C. (1978). Permissible inferences from the outcome of training
for Comparative
studies in cognitive development research. Quarterly Newsletter of the Institute
Human Development, 2, 46-53.
Brown, A. L., & Campione, J. C. (1981). Inducing flexible thinking: A problem of access. In M.
Friedman, J. D. Das & N. O’Connor (Eds.), Intelligence and learning. New York: Plenum Press.
Brown, A. L., & Campione, J. C. (1982). Modifying intelligence versus modifying cognitive
skills: More than a semantic quibble. In D. K. Detterman, & R. J. Sternberg (Eds.), How and
how much can intelligence be increased. Norwood, NJ: Albex.
Brown, A. L., Campione, J. C., & Barclay, C. R. (1979). Training self-checking routines for
estimating test readiness: Generalization from list learning to prose recall. Child Development,
50, 501-512.
Brown, A. L., Campione, J. C., Bray, N. W., & Wilcox, B. L. (1973). Keeping track of changing
variables: Effects of rehearsal training and rehearsal prevention in normal and retarded adoles-
cents. Journal of Experimental Psychology, 101, 123-131.
Brown, A. L., Campione, J. C., & Day, J. D. (1981). Learning to learn: On training students
to
learn from texts. Educational Researcher, 10, 14—21.
Brown, A. L., Campione, J. C., & Murphy, M. D. (1977). Maintenance and
generalization of
trained metamnemonic awareness by educable retarded children: Span estimation. Journal
Ex-
perimental Child Psychology, 24, 191-211.
Brown, A. L., & Day, J. D. (1980). Strategies and knowledgfor e summarizing texts: The develop-
ment and facilitation of expertise. Unpublished manuscript, University of
Illinois.
Brown, A. L., & Lawton, S. C. (1977). The feeling of knowing experience
in educable retarded
children. Developmental Psychology, 13, 364—370.
Burger, A. L., Blackman, L. S., Holmes, M., & Zetlin, A. (1978). Use of active sorting and
 4. METACOGNITION AND PROBLEM LEARNERS 139

retrieval strategies as a facilitator of recall, clustering, and sorting by EMR and nonretarded
children. American Journal of Mental Deficiency, 83, 253-261.
Butterfield, E. C. (1979). On testing process theories of intelligence. Paper presented at NATO
Conference on Intelligence and Learning, York, England, July.
Butterfield, E. C., & Belmont, J. M. (1977). Assessing and improving the cognitive functions of
mentally retarded people. In I. Bailer & M. Steinlicht (Eds.), Psychological issues in mental
retardation. Chicago: Aldine Press.
Butterfield, E. C., Wambold, C., & Belmont, J. M. (1973). On the theory and practice of improv-
ing short-term memory. American Journal of Mental Deficiency, 77, 654-669.
Campione, J. C., & Brown, A. L. (1977). Memory and metamemory development in educable
retarded children. In R. V. Kail, Jr. & J. W. Hagen (Eds.), Perspectives on the development of
memory and cognition. Hillsdale, NJ: Lawrence Erlbaum Associates.
Campione, J. C., & Brown, A. L. (1978). Toward a theory of intelligence: Contributions from
research with retarded children. /ntelligence, 2, 279-304.
Campione, J. C., Brown, A. L., & Ferrara, R. A. (1982). Experimental and clinical interventions
of retarded individuals: Intelligence, learning, and transfer. In R. Sternberg (Ed.), Handbook of
human intelligence. New York: Cambridge University Press.
Day, J. D. (1980). Training summarization skills: A comparison of teaching methods. Unpublished
doctoral dissertation, University of Illinois.
Feuerstein, R. (1980). Instrumental enrichment: An intervention program for cognitive modi-
fiability. Baltimore: University Park Press.
Flavell, J. H. (1971). First discussant’s comments: What is memory development the development
of? Human Development, 14, 272-278.
Flavell, J. H. (1979). Monitoring social-cognitive enterprises: Something else that may develop in
the area of social cognition. Paper prepared for the Social Sciences Research Council Committe
on Social and Affective Development During Childhood, January.
Flavell, J. H., Friedrichs, A. G., & Hoyt, J. D. (1970). Developmental changes in memorization
processes. Cognitive Psychology, 1, 324-340.
Flavell, J. E., & Wellman, H. M. (1977). Metamemory. In R. V. Kail & W. Hagen (Eds.),
Perspectives on the development of memory and cognition. Hillsdale, NJ: Lawrence Erlbaum
Associates.
Gardner, E. (1978). Commentary on animal awareness papers. Behavioral and Brain Sciences, 4,
572.
Kendall, C. R., Borkowski, J. G., & Cavanaugh, J. C. (1980). Maintenance and generalization of
an interrogative strategy by EMR children. /ntelligence, 4, 255-270.
Kennedy, B. A., & Miller, D. J. (1976). Persistent use of verbal rehearsal as a function of informa-
tion about its value. Child Development, 47, 566-569.
Markman, E. M. (1981). Comprehension monitoring. In W. P. Dickson (Ed.), Children’s or com-
munication skills. New York: Academic Press.
Masur, E. F., McIntyre, C. W., & Flavell, J. H. (1973). Developmental changes in apportionment
of study time among items in a multitrial free recall task. Journal of Experimental Child Psychol-
ogy, 15, 237-246.
Meichenbaum, D. (1977). Cognitive behavior modification: An integrative approach. New York:
Plenum Press.
Moore, J., & Newell, A. (1974). How can Merlin understand? In L. W. Gregg (Ed.), Knowledge
and cognition. Hillsdale, NJ: Lawrence Erlbaum Associates.
Newell, A. (1980). Production systems and human cognition. In R. A. Cole (Ed.), Perception and
production of fluent speech. Hillsdale, NJ: Lawrence Erlbaum Associates.
Piaget, J. (1978). Success and understanding. Cambridge, MA: Harvard University Press.
Pylyshyn, Z. (1978). Computational models and empirical constraints. The Behavioral and Brain
Sciences, 1, 93-99.
140 CAMPIONE

Rozin, P. (1976). The evolution of intelligence and access to the cognitive unconscious. Progres-
sion in Psychobiology and Physiological Psychology, 6, 245—280.
Sternberg, R. J. (1980). Sketch of a componential subtheory of human intelligence. The Behavioral
and Brain Sciences, 3, 573-584.
Tenney, Y. J. (1975). The child’s conception of organization and recall. Journal of Experimental
Child Psychology, 19, 100-114.
 MOTIVATION AND
ATTRIBUTION STYLE

For many years, the analysis of achievement-related behav-

iors has been an important part of the psychology of moti-
vation. Through investigations of the influence of
dispositional factors such as hope-for-success and fear-of-
failure on subjective experience, activities and performance
in different task situations, the prevalence of inter-indi-
vidual differences in causal attributions for success and
failure was first noted. Such causal attributions have figured
prominently in many of the new, cognitive models of moti-
vation, and are also central to the learned helplessness
model developed by Abramson, Seligman, and others. In
this model, cumulative experiences of being unable to con-
trol situations or to influence outcomes by goal-directed
actions are postulated to result in behavioral deficits and
performance decrements. The impact of fear-of-failure and
learned helplessness on learning activities and performance
has been demonstrated in both experimental and school set-
tings. The phenomena of learned helplessness and issues
related to causal attributions for success and failure are
central to the following three chapters.
Heinz Heckhausen examines individual differences in
causal attributions for achievement outcomes. He describes
stable intra-individual asymmetries in attribution patterns
that are different for individuals who are motivated to
achieve success or to avoid failure. Heckhausen dis-

141
142 PART Il: MOTIVATION AND ATTRIBUTION STYLE

tinguishes a ‘‘positive attribution pattern (PAP) from a “‘depressive attribution

pattern’? (DAP), and discusses how these patterns affect behavior and how they
emerge.
Christopher Peterson and Martin Seligman focus on the relationship between
a helpless attribution style and depression. Following a review of the original
learned helplessness model, they analyze the role played by relatively stable
attribution styles, and discuss how such attribution styles are assessed in children
and adults. Peterson and Seligman then present and discuss empirical findings
that speak to the relation between attribution style and depressive symptoms.
They give particular attention to the question of how a depressive attribution
style might develop.
Julius Kuhl proposes an alternative explanation for the phenomena of help-
lessness, drawing on achievement motivation theory and data. The central idea
behind Kuhl’s model concerns action control. Kuhl postulates that action control
is determined both by dispositional differences that bias an individual toward an
action or state orientation, and by situational variables such as the extent to
which an outcome can be controlled by one’s own actions. Kuhl presents em-
pirical findings consistent with the model of action control, and concludes with a
discussion of how this model may be applied to learning and performance prob-
lems encountered in school settings.
 Causal Attribution Patterns for
Achievement Outcomes:
Individual Differences,
Possible Types and Their
Origins

Heinz Heckhausen
Max-Planck-Institute for Psychological Research
Munich

Given the same achievement outcomes, why will one person remain confident of
success and another fearful of failure— why, in the face of the same failure will
one person increase or renew efforts, and another soon become helpless and give
up on the task? Typical answers are that each has a different ‘‘achievement
motive.’’ The former, to use the concepts of some research traditions, is
““achievement-oriented,’’ ‘‘nondepressive,’’ and ‘‘internally controlled,’’ the
PIE EG
latter, “‘helpless, depressive,’ and ‘“‘externally controlled.’’ Within each
tradition, measures have been developed to assess some dispositional difference
that covaries with some aspects of achievement behavior.
To better understand differences in achievement behavior in seemingly identi-
cal situations, one should be better able to answer two questions. The first relates
to motivation: In given situations, what specific determinants of personality,
other than the summary constructs, will give rise to different achievement behav-
iors? The second question relates to development: What are the antecedents of
diverging determinants of personality in the earlier stages of the life cycle?
To the developmentalist, the question is a difficult one; the further one reach-
es back into the past, the more difficult is it to isolate the ontogenetic factors.
Each developmental stage has a specific environmental impact; therefore, it is
difficult for the developmentalist to know what specific stage to search for
personality determinants. The search for a developmental explanation cannot be
more specific than the determinants of personality being sought allow. The
genesis of a dispositional factor as summary as the achievement motive can, with
regard to the ontogenetically relevant age spans and their environmental determi-
nants, at best be summarily accounted for only. The explanation is even more
difficult because the time gap between the present and past skips long periods of

143
144 HECKHAUSEN

development (and their accompanying psychological implications), and also be-

cause it entails attempts to reconstruct the past on the basis of current reports. For
these reasons, the inconclusiveness of the findings seems to be an inevitable
predicament of the usual socialization research, it also seems to be the plight of
attempts to clarify the emergence of individual differences in achievement
motivation.
To be sure, relationships have been found between achievement motivation
and such diverging variables, such as ‘‘a democratic family climate,’’ *‘love-
oriented child-rearing techniques,’’ and “‘child-centered independence training’’
(see Heckhausen, 1972). However, from the developmentalist’s point of view,
such relationships are trivial for two reasons: First, it is still not known at which
specific age spans the variables discussed actually generate individual dif-
ferences, until what age they do not as yet have this effect, and from what age on
they no longer have it. Second, these determinants are too unspecific to account
for the motivational differences to be accounted for because they covary with
differences in other dispositional factors, such as development of conscien-
tiousness, I.Q., and social competence.
In what follows, I attempt to avoid the aforementioned drawbacks of develop-
mental accountability. Hence, summary constructs are not used to explain indi-
vidual differences in achievement behavior. Instead, a more specific personality
variable is singled out. In light of current results of achievement motivation
research, three specific personality variables are suggested the valence of incen-
tives for success or failure, personal standards of excellence (level of aspiration),
and attribution patterns (see Heckhausen, 1980a).
This discussion is concerned only with attribution patterns, primarily because
they are personality variables uncovered (in some studies under the term of
“attribution styles’’) by investigators working within diverse research traditions.
Attribution theory has achieved a remarkable overlap between explanatory con-
structs and research results.
The first section examines the intraindividual asymmetry of attribution pat-
terns for success or failure and the assumed underlying motivations. This is
followed by an overview of various lines of research that have brought to light an
interindividual disparity in attribution patterns, which may produce high or low
self-evaluation. Finally, a number of possible attribution patterns are specified in
terms of their potential to explain success and failure according to a personal
bias.
Depending on the weighting given to the causal factors, the effects of any of
the attribution patterns are apt to heighten or lower self-evaluation. Accordingly
each of them is analyzed to determine the level of cognitive development that
must be presupposed in each case to specify the earliest level of development
from which a biased attribution can be expected. The earliest level of develop-
ment serves as a referent for generating assumptions concerning environmental
 5. CAUSAL ATTRIBUTION PATTERNS 145

factors that might have produced individual differences in the various attribution
patterns.
My procedure is entirely hypothetical, even though it does comprehend the
few research findings in this field. The intention is not to extrapolate prematurely
across the few results; rather, it is designed to stimulate a systematic search for
missing data.

INTRAINDIVIDUAL ASYMMETRY OF ATTRIBUTION

PATTERNS

Among the individual differences in attribution patterns, a general bias is per-

haps, above all, the most salient one. People tend to explain their own success or
failure by asymmetric attribution patterns, even for tasks of identical difficulty.
One tends to attribute success to oneself, and failure to external factors for which
one is not responsible (e.g. Luginbuhl, Crowe, & Kahan, 1975). On the other
hand, in explaining the achievements of others, subjects tend to be more impar-
tial (Snyder, Stephen, & Rosenfeld, 1976). In relation to own success and
failure, asymmetric attribution patterns evidently have a motivational basis; they
serve to enhance one’s sense of self-esteem. Miller and Ross (1975) have ques-
tioned this assumption and attribute the asymmetry to nonmotivational factors in
information processing—for instance, that success, rather than failure, is
intended and expected, and that accordingly, one tends to ascribe the achieve-
ment outcome to oneself. Recent empirical evidence, however, has confirmed
the motivational nature of a general, self-serving bias in the attribution asymme-
try (see Bradley, 1978; Miller, 1976; Stevens & Jones, 1976).
If one keeps in mind the many causes that may facilitate or inhibit a successful
achievement outcome and that none of these causes can be objectified, the
latitude for subjective arrangements of an attribution pattern is enormous. Fur-
ther, considering the extent to which various skills and abilities (intellectual
abilities in particular) are valued in our culture, and taking into account that most
people want to see themselves and want to be regarded by others as competent in
some respect, it is not surprising that in most cases the scope for subjective
assessment of abilities is used in ways that attribute success to one’s own abili-
ties, rather than to other facilitating factors, such as ease of task, the effort
expended, or good luck. By the same token, failure is ascribed less to one’s own
lack of abilities than to other adverse circumstances, such as insufficient efforts,
too difficult a task, or bad luck. To be sure, such self-esteem enhancing propen-
sities may be unrealistic; they may even be generated by self-deception. Nev-
ertheless, those self-concepts will foster the motivation to expend more effort
instead of giving up after experiencing failure, and in the final analysis will
facilitate the further development of one’s abilities. In this sense, a somewhat
146 HECKHAUSEN

overoptimistic assessment of oneself, or even a kind of self-conceit, may be a

characteristic that has survival value—possibly human evolution would not have
taken place at all in its absence.
Against the background of the general aspiration to make one’s own abilities
appear in the most favorable light, attribution patterns involving self-criticism or
that detract from self-esteem appear to be exceptional. Yet this type of attribution
pattern has attracted attention in a number of research studies, not only in
individual cases, but for whole groups of subjects selected according to very
different diagnostic procedures. The selection criteria included such divergent
personality traits as a perception of external control (Rotter, 1966), ‘‘help-
lessness’’ (Diener & Dweck, 1978), ‘‘depression’’ (Beck, Ward, Mendelsohn,
Mock & Erbaugh, 1961), ‘‘failure motivation’’ (Heckhausen, 1963; Schmalt,
1976), “‘low self-concept’’ (various questionaires), and, yes, even in the female
gender of the subjects (e.g., Deaux & Farris, 1977).
One may ask why whole groups of persons should disparge their own effec-
tiveness, abilities, and competence in their own eyes and in the eyes of others,
even when there is no compelling reason—when the same achievement out-
comes could have been attributed so that subjects can retain a much higher sense
of self-esteem without diminishing attributional plausibility. The question, then
is, whether such persons have acquired a negative self-image to which they
intend to adhere, even after having had more positive experiences (see Fitch,
1970). Recent findings are surprising, they show the degree to which individuals
will hold to a preconceived notion even though it has been disproved by the facts
(Ross & Lepper, 1980). Can this phenomenon be explained by the assumption
that an attribution pattern acquired in an earlier developmental phase is subse-
quently transformed into a mode of information processing that already distorts
perception at a pre-attentive level? Does this become an attribution pattern inca-
pable of self-correction? Even though the individual has had positive experiences
that contradict the negative self image, does this attribution pattern become
“‘immune’’ to modification (see Heckhausen, 1975)? Or, on the contrary, does
attribution that reduces self-esteem constitute a cognitive process motivated by
the intention to avoid longer exposure to achievement situations and the self-
evaluation entailed? So far, no conclusive statement can be made as to which of
these explanations might be correct. One cannot dispose of investigation tech-
niques that would make it possible to balance the effect of automatized sub-
routines of information processing against currently motivated distortions of
judgment.
Nonetheless, there are two reasons that favor the assumption that even this
type of self-esteem-reducing attribution pattern should be held to be a motivated
process. The first reason is simply that its counterpart, self-esteem-enhanc
ing
attribution, is undoubtedly a motivated attribution pattern. Therefore
, without
cogent reasons, self-esteem-reducing attribution should not be regarded
as being
exceptional. This argument has gained validity with investigations
showing that
 5. CAUSAL ATTRIBUTION PATTERNS 147

when a particular situational context provides grounds for a modest self-image

(e.g., when performance is being assessed by a team of experts), the same
Subjects will promptly switch from a self-esteem-enhancing attribution to its
opposite. In such situations, a ‘‘counterdefensive’’ attribution may be observed:
Subjects do not feel primarily responsible for their successes, but for their
failures (Arkin, Appelman & Burger, 1980; Schlenker, 1975; Wortman, Con-
stanzo & Witt, 1973). Even more convincing is the motivational nature of what
has been termed the ‘“‘self handicap.’’ Self-handicap is a strategy of self-repre-
sentation by which individuals set up in advance difficulties in their solution
attempts (sometimes even at the risk of preprogramming their own failure), so
that subsequent to the event they need not call their own capacity into question
(Berglas & Jones, 1978). Obviously, if one is successful, with this strategy one’s
own abilities will appear even more impressive. The self-handicap, a precaution-
ary attribution strategy, can be observed when it takes the form of the academic
ritual, often used in the captatio benevolentiae turns of phrase in introductions to
lectures.
Also in support of the argument that attribution pattern asymmetry is moti-
vationally based is that the ‘‘self-esteem reducing’’ label itself appears to be an
overstatement of what is actually at issue. In most cases, it is not possible to
determine the extent to which either of these attribution patterns involves an
attributional error—there are no objective criteria for “‘realistic’’ attribution.
Thus, one can contrast these two attribution patterns only in their complementary
relationship. It is possible that the problem group—the depressive, the failure
motivated, and the like—are merely less prone to adopt a self-esteem-enhancing
attribution pattern than are the ‘‘normal’’ groups that for this reason they make
attributions that are realistic and self-critical rather than “‘self-esteem’’ reducing.
Indicators of this possibility have been found.
One way to assess propensity to attributional errors is to make an intrain-
dividual comparison of success and failure attributions, that is, to determine the
asymmetry of attribution. For instance, Meyer (1973) found that failure-moti-
vated subjects, after experiencing unexpectedly low or unexpectedly high
achievement outcomes, modified their attributions concerning ability, effort, and
chance only very little, success-motivated subjects were prone to tap all the
possibilities provided by self-esteem-enhancing settings. Success-motivated sub-
jects ascribed unexpectedly high performance outcomes to their own ability and
unexpectedly low outcomes to (low) effort and to chance. Alloy and Abramson
(1979), on the basis of stringent objectivity criteria, demonstrated that de-
pressives were able to draw a surprisingly accurate distinction between the
efficacy (contingency) of their own undertakings and chance events; non-
depressive subjects appeared to overestimate their ability to exert control when
the outcome aimed at was either frequently obtained as a result of chance or had
been made valuable. Correspondingly, they underrated their control when the
outcome was not valuable. Alloy’s and Abramson’s conclusion was that de-
 148 HECKHAUSEN

pressives are ‘‘sadder but wiser.’’ Perhaps the same is true for other ‘problem
groups’’—the failure-motivated, the externally-controlled, the helpless, subjects
with a low self-concept.
It may appear paradoxical that of all people it should be these problem groups
who attain a more realistic self-evaluation. But is a realistic self-evaluation
‘“‘normal’’ and helpful? Is not a certain measure of self-esteem-enhancing self-
deception ‘‘normal,’’ or at least necessary, for one to cope with the problems of
life? Is such self-deception not essential to prevent one from giving up hope too
soon in the pursuance of the many goals that are not easily attained? Does self-
deception not enable one actually to achieve more than could be expected when
starting an undertaking?
Moreover, the findings concerning those problem groups, have not in all
cases indicated that they deviate from the general (self-esteem-enhancing) at-
tribution asymmetry (e.g. Schneider, 1977). An attribution pattern deviating
from the self-esteem-enhancing mode appears to be used more often subsequent
to failure than subsequent to success (e.g. Fitch, 1970; Halisch, 1983; Jopt &
Ermshaus, 1977). Only one success or failure may not elicit variations in attribu-
tion patterns. Halisch (1983) found that only after repeated experiences of failure
do failure-motivated subjects ascribe such failure to their own lack of ability.

INTERINDIVIDUAL CONTRAST IN ATTRIBUTION

PATTERNS: CONVERGENCE OF RESEARCH DOMAINS

This section reviews the various areas of research in which findings have high-
lighted an interindividual contrast, that is, in which different groups of re-
spondents used an attribution pattern that deviated from general self-esteem-
enhancing attribution.

Internal versus External Reinforcement Control

The personality construct postulated by Rotter (1954, 1966) is a precursor
of
attribution theory but it has not adequately been clarified in terms of its impor-
tance for attribution theory itself. For Rotter, internal versus external reinforce-
ment control meant individual differences in the belief that one’s own action and
its results can generate desired ends. Rotter’s construct involves the “instrumen-
tality,’” or ‘‘contingency,’’ between one’s own activity (and its results)
and its
contingent consequences mediated by others. However, this research
line has
never clearly and consistently kept separate the result of an action
and the
consequences of an action (see Heckhausen, 1980a). In light of
the distinctions
now current in attribution theory, Rotter’s usage of the expressio
n ‘‘external’’
refers to noncontingent (external) causal factors, such as
chance or the ar-
bitrariness of reinforcers. ‘‘Internal’’ on the other hand, refers
to factors relating
 5. CAUSAL ATTRIBUTION PATTERNS 149

to personal efficacy, such as ability in relation to task difficulty. The suitable

experimental conditions were made operative by means of chance versus skill-
dependent tasks. The experiments, however, involve attribution of action out-
comes and not, as the construct postulates, attribution of action consequences.
Even in a questionnaire on expectancy of control these two different attributions
are confounded. The content of the questions used shows a high dispersion
around very different areas of activity and life spheres; this weakens the validity
of Rotter’s trait construct (see Zuckerman & Gerbasi, 1977).
Distinguished in terms of this approach, the individual dispositions can best
be designated as a sense of high versus low self-efficacy. The belief in external
control, sensed as a feeling of low self-efficacy, thus appears to characterize the
type of persons who, in accounting for their achievement outcomes only rarely
use self-esteem-enhancing attribution patterns, (or even conversely, self-esteem-
reducing attributions). There are findings that confirm this hypothesis. External
control subjects, compared with internal control subjects, tend to attribute their
success less to ability/skills and their failure less to chance, especially when their
solution attempts are associated with a high expectancy of success (Gilmor &
Minton, 1974; Lefcourt, Hogg, Struthers & Holmes, 1975).
There are, however, also research findings that relate to the specific attribu-
tional dimension inherent in Rotter’s construct that is, the contingency of action
consequences (reinforcement). Contingency attribution plays a decisive part in
the experimental paradigm of helplessness research. In a training task, subjects
initially receive negative or positive feedback during their solution attempts, for
one group of subjects this instantiates objective noncontingency, for the other
group, contingency. In a subsequent task, either similar or different, the experi-
menters scrutinize the extent to which noncontingency during the training task
has impaired performance in the test task (i.e., to what extent noncontingency
actually induced ‘‘helplessness’’). To put it another way: Subsequent to the prior
experience of loss of control or of being confronted with noncontrollable factors,
it is assumed that the noncontingent group will exhibit greater helplessness in its
solution attempts for the following controllable (contingent) task. This occurs
because the previously acquired sense of loss of control is transferred to the new
situation. Accordingly, external control subjects are more likely than internal
control subjects to perceive noncontingent feedback as a noncontrollable factor
and in the test phase will therefore, become more markedly helpless in a more
generalized way.
Hiroto (1974) has provided evidence of the validity of this assumption in tasks
investigating avoidance learning (escaping an aversive tone by moving hands in a
“finger shuttle box’’). In a similar test task, external control subjects learned
more slowly than did internal control subjects or those in the control group, who
had had no prior experience or training. It was also possible to induce an
approximately equivalent helplessness effect when the noncontingency of the
outcome was pointed out to the subjects in an instruction period preceding the
150 HECKHAUSEN

test. However, ‘‘externals’’ appear to use not only attribution patterns, that more
markedly, induce helplessness, but also those that make helplessness more gen-
eralized. After their subjects had terminated a concept-formation task such as the
training task, Cohen, Rothbart, and Phillips (1976) administered two test tasks in
another room. One of the test tasks consisted of tracing the lines of a diagram,
without tracing any line twice and without lifting the pencil from the figure; the
diagram-tracing task posed an intellectual problem, as did the training task of
concept-formation. The second test task was a pure speed task (Stroop-test).
Both internal and external noncontingent subjects performed more poorly on the
more similar diagram-tracing task, with the externals suffering a stronger impair-
ment (as measured by less persistence at the insolvable task). At the less similar
speed task, only externals showed helplessness effects; they were slower.

Depression
The latest findings indicating that the helplessness effect will generalize across
situations and time in externally controlled subjects call attention to an attribu-
tional dimension that has long been neglected. That is the dimension of globality
(in contrast to specifity) (although the phenomenon itself was already incorporat-
ed by Kelley, 1967, in his conceptual schema as ‘‘entity.’’) Restating the origi-
nal helplessness theory, Abramson, Seligman & Teasdale (1978) made use of the
globality dimension in their reformulations of attribution theory (see also Miller
& Norman, 1979; Roth, 1980; Peterson & Seligman this volume). Given aver-
sive outcomes (failure), inhibitory causal factors in self-esteem-reducing attribu-
tion patterns are more global than in the event of success. Failure on a specific
task is much more readily ascribed to a general lack of ability and, in the end, to
lack of intelligence. To explain how reactive depression comes about, Miller &
Norman (1979) have focused on the self-esteem-reducing attribution pattern and
have given the following concise description:

Due to some combination of situational cues and repeated exposure to noncon-

tingent and nondesired outcomes, the individual’s attributions of these
outcomes
changes from external, variable, and specific causes to internal, stable, and
general
causes. This change results in a change in future expectancies, performances,
and
mood. Thus, in new situations, the individual expects noncontingency
and failure,
and when these congruent outcomes occur, they are attributed to
internal, stable,
and general causes, whereas discrepant outcomes of success and
contingency are
attributed to external, variable, and specific causes and do not
influence future
expectancies, performance, or mood. The individual is then depressed
and tends to
disregard outcomes of success and contingency while overgeneralizing
failure and
noncontingent outcomes. Thus, response initiation declines,
a greater number of
failure and noncontingent outcomes do occur, and the vicious
circle of depression
has begun. (p. 113/114)
 5. CAUSAL ATTRIBUTION PATTERNS 151

The original helplessness theory was simple: If specific outcomes proved

uncontrollable, they would produce an expectation of one’s own inefficacy; in
recurrent or similar situations this expectation would generate cognitive, emo-
tional, and affective ‘‘deficits’’ (cf. Seligman, 1975). However, what might
have seemed from the cognitivist perspective, a daring construction when this
theoretical framework to explain the findings of research conducted with in-
frahumans—namely, that the fatal effects of the expectation that outcomes
would prove uncontrollable—proved to be even more flimsy relative to the
effects of failure-related cognitions when humans were the research subjects.
Unlike dogs for instance, humans displayed neither motivational deficits nor, in
many cases, cognitive deficits, in the test task situation. Moreover, frequently
humans exhibited an improvement rather than a decrement in performance (see,
for example, Hanusa & Schult, 1977). Even ‘‘success,’’ when it is noncon-
tingent and is thus experienced as puzzling and enigmatic, can be perceived as a
failure (Griffith, 1977). The motivational effect of the cumulative experience of
failure depends on the causal attributions made, in particular on whether the scale
of causal attributions tips more toward one’s own lack of ability than toward task
difficulty (see discussion in Heckhausen, 1980a, p. 510f). Helplessness theory,
developed from investigations conducted with animals, had conceived the help-
lessness phenomenon to be a general fact of experimental psychology. Theory
construction, however, was soon confronted by phenomena belonging to differ-
ential psychology.
Depending on experimentally induced attributions (e.g. Tennen & Eller,
1977), on personal attribution patterns, or possibly on related differences of state
orientation versus action orientation (see Kuhl 1981; and this volume), only
some of the subjects were observed to display helplessness. Internal versus
external location of stable causes play a key role in generating feelings of help-
lessness. When in the training-task subjects ascribed failure to their own in-
ability, rather than to high task difficulty, helplessness was generalized to the test
task (Frankel & Snyder, 1978; Klein, Fencil-Morse, & Seligman, 1976; Tennen
& Eller, 1977). Frankel and Snyder introduced the tasks as being either easy or
difficult; Tennen and Eller manipulated them as involving either increasing or
decreasing levels of difficulty. When, in one way or another, an attribution of
very high task difficulty was suggested, the effect was not helplessness but
improved performance (a performance effect that in the Tennen and Eller investi-
gations outmatched even that of the ‘‘contingent’’ and control groups). Ob-
viously, these test results must be interpreted in terms of attribution theory: A
very high task difficulty attenuates the threatening inference that failure implies
lack of ability, even when considerable effort has been expanded.
Following the noncontingent series of failures, susceptibility to helplessness
was also shown more by subjects with higher scores of depression than by
nondepressives, high-depression scorers exhibited more depressive symptoms,
 152 HECKHAUSEN

less self-confidence and lower expectancies of success. No differences were

found for the noncontingent success series (Hammen & Krantz, 1976; Wener &
Rehm, 1975). In addition to negative affect changes and lowered self-esteem
(Blaney, 1977) was the characteristic asymmetry of attribution patterns (Rizley,
1978): Successes were ascribed to low level of task difficulty (external, stable,
specific), and failures to one’s own low ability (internal, stable, global).
The original concept that the experience of loss of control was a sufficient,
although not a necessary, condition for reactive depressions (Miller & Seligman,
1975; Seligman, 1975) was afflicted by the paradox in depression: How can
depressives blame themselves for outcomes they believe are uncontrollable
(Abramson & Sackeim, 1977)? This paradox can be resolved if it is assumed that
in addition to the expectancy of uncontrollability, depressives tend to ascribe
negative outcomes to internal, stable, and global causes. To be sure, such an
attribution pattern in itself does not make up depression, but it is a risk factor,
that could cause depression when adverse or aversive events occur and accumu-
late. On the other hand, even uncontrollable outcomes, whether actually or only
presumably uncontrollable, do not in themselves make people helpless or de-
pressive. Such events must also have an aversive function, so the action se-
quences that comprise high positive incentive values experienced so far appear to
be irretrievably lost. In cases where the attribution pattern itself was already
negative and experience of aversive effects had accumulated, the incentive-
disengagement-cycle (Klinger, 1975), will then no longer make possible the
upswing that usually follows temporary phases of depressive reaction.
This assumption has found support in the data presented by Seligman and his
associates (see Peterson & Seligman, this volume). In this study substantive
progress was made in the recording of attribution patterns. Instead of measuring
predetermined causal factors, Seligman, Abramson, Semmel, and von Baeyer
(1979) developed a scale for measuring attributional style based on their Attribu-
tional Style Questionnaire (ASQ). Subjects were given instructions for filling out
the attributional style scale. They were presented with a selection of “* good’’ and
‘“‘bad’’ outcomes, which included success and failure in achievement
situations.
The subjects were asked to try to imagine themselves in the situations specified,
and to pick the one major cause if this event happened to them. The subjects were
then instructed to rate each cause (on a 7 point scale) for degree of
internality,
stability, and globality. These attributional dimensions. separately
or together
but separated into good and bad outcomes, are the characteristic
values for
individual attribution patterns. They are relatively stable over time
and fairly
independent. The ‘‘insiduous’’ attributional style (internal, stable,
and global
attribution for bad outcomes) correlated with depression for
children, students,
and clinical patients. In the latter two groups (depressed students
and patients),
good outcomes were explained by means of the converse
attribution pattern
(even if to a lesser extent), that is, by external, variable, and
specific causes. The
 5. CAUSAL ATTRIBUTION PATTERNS 153

children, aged 9 to 10, had mothers for whom depression and insiduous attribu-
tional style were prominent; the results correlated with each other,
Peterson and Seligman (this volume) had also collected data showing that as a
risk factor the insiduous attributional style is a necessary, but not a sufficient,
condition for depression. In their study the negative event was the unsatisfactory
outcome of a midterm test. In cases where insiduous attributional style and
unsatisfactory test outcome coincided, depression symptoms (recorded on the
basis of questionaires) were displayed one week following and four weeks fol-
lowing the test. Separately, neither of these two factors gave rise to depressive
symptoms. A longitudinal study with the same children also demonstrated this
temporal sequence, and thereby the direction of causation. The insiduous attribu-
tion style proved to be the better predictor of phases of depression than, con-
versely, depressive states did for attributional style. There appear, therefore, to
be many factors indicating that a self-esteem-reducing attribution pattern pre-
disposes individuals to encounter difficulty in getting out of depressive phases
following a bad event.

Learned Helplessness in Schoolchildren

Dweck observed that after only a few failures, schoolchildren labeled as subject
to “‘learned helplessness,’’ rapidly lost the self-efficacy they usually displayed in
solving tasks. As Dweck and Repucci (1973) found the children were no longer
able to solve tasks they had demonstrably been able to solve previously. For
some of their fifth-grade subjects. The helpless students, for whom failure re-
sulted in decrement of performance and reduced persistence, also attracted atten-
tion by the peculiar pattern of answers they gave in the Intellectual Achievement
Responsibility (IAR) scale administered to them (Crandall, Katkovsky & Cran-
dall, 1965). Compared with their classmates who did not let themselves be
thrown off balance by failure, the subjects had a greater tendency to ascribe their
success to high effort and their failure to poor effort.
In a 25 day behavior change program for children characterized by “‘expecta-
tion of failure and deterioration of performance in the face of failure,’’ (12 out of
750, 10- to 13-year olds), Dweck (1975) demonstrated the decisive role of effort
attributions. After failing on a number of somewhat more difficult tasks, the
children were urged to ascribe failure to a lack of effort. In this way it was
possible to fully overcome the decrement in efficacy that had followed failure,
although while the comparative treatment program that comprised solely suc-
cess-only treatments showed no improvement.
In another study, Diener and Dweck (1978) traced the cognitive processes
displayed by helpless and mastery-oriented children following failure. On the
basis of the IAR Scale, fifth graders were divided into helpless and mastery-
oriented groups. Those children who, in determining their failure, tended to
 154 HECKHAUSEN

neglect the role of effort were placed in the helpless group. All the children were
asked to begin thinking out loud as they performed the task. The verbalization
procedure revealed a striking difference: helpless children tended to dwell exces-
sively on their failures, and attributed them primarily to lack of ability: mastery-
oriented children made very few attributions, but instead engaged in solution-
directed behavior to improve their performance. A first failure appeared to make
the problem more interesting for the mastery-oriented children and spurred them
on to search for new problem-solving strategies. Making attributions meant
letting attention wander from the criterial task demands on which one must focus
in such cognitive action. These findings are consistent with the distinction made
between state orientation and action orientation (Kuhl, 1981), and with the task-
irrelevant cognitions observed during an oral examination in subjects for whom a
failure-oriented motivational state was dominant (Heckhausen, 1982).

Achievement Motives

Individual differences in attributions were also first uncovered in achievement

motivation research using Crandall et al.’s (1965) IAR Scale. In correlation
studies both Weiner and Kukla (1970) and Weiner and Potepan (1970) found that
on the success-related items of the questionnaire, success-oriented subjects (with
a resultant high achievement-motivation) checked off internal causal factors,
primarily ability, more frequently than did the failure-oriented subjects. In a
series of experimental studies, in particular those undertaken by Meyer (1973),
after prior success or failure had been induced, subjects were asked to scale
directly ability, effort, task difficulty, and good or bad luck (see also
Halisch,
1983; Jopt & Ermshaus, 1977; Krug, 1972; Schmalt, 1976). Although the find-
ings from these series are not always identical or unambiguous, they consistently
show motive-linked attribution patterns that were apt to lower self-evaluation
for
subjects with a dominant failure-oriented motivation. This was especially
true
when the failure-oriented achievement-motive was rooted in a self-concept
im-
plying lack of ability rather than in fears of the social consequences
of failure
(Schmalt, 1976).
In spite of the more recent work in helplessness research, investig
ation of
attribution patterns in achievement motivation research still
leaves much to be
done. Thus, the direct scaling of the four factors mentioned
earlier, does not tell
the investigator to what extent individual subjects have logicall
y located one or
another factor into the dimensional structure of internality
or stability; further-
more, the dimension of globality is not represented. Nonethe
less, the results
obtained are sufficiently consistent to be summarized in terms
of specific criteria,
as done elsewhere (Heckhausen, 1980a). Accordingly:

In the case of success, the locus dimension is decisive. Success-motivated Ss

attribute success primarily to internal factors. especia
lly to high ability; failure-
 5. CAUSAL ATTRIBUTION PATTERNS 155

motivated Ss lay stress upon external factors, especially good luck, and sometimes
low task difficulty. In the case of failure, the stability dimension is decisive. For
success-motivated Ss, failure is attributable more to controllable or modifiable
factors, particularly to lack of effort, sometimes to bad luck. Conversely, failure-
motivated Ss tend to take their failure for granted and do not believe that this can be
modified because they attribute failure to lack of own ability or in some cases to
great task difficulty. Thus, while the success-motivated represent the image of the
usual self-esteem enhancing attribution asymmetry, failure-motivated subjects tend
toward an attribution pattern which does little to encourage them following the
experience of success, which discourages them following failure and which thus
has a self-esteem-reducing effect. (p. 524)

In light of these observations and findings, the achievement motive was

conceived as a self-evaluation system (Heckhausen, 1975, 1978). On the basis of
these different attribution patterns, information about otherwise identical
achievement outcomes is processed into different self-evaluations and success
expectancies. In this way, the attribution pattern of failure-motivated indi-
viduals, even after the experience of a series of successes, will hardly make them
feel happier or more self-confident, whereas, even after repeated failures, suc-
cess-motivated individuals will not give up hope. The instrumentality of the
motive-linked attribution pattern for self-evaluation and success expectancy has
become to the major point of attack for interventions intended to change the
achievement motive from fear of failure into hope of success. A direct form of
such intervention was applied by Dweck (1975) in the training program referred
to earlier. Krug and Hanel (1976) trained a motive-arousing attribution pattern by
means of model learning, self-observation, self-reports, and verbalizing
thoughts. In postexperimental tests they were able to demonstrate motivational
and attributional shifts in the direction, desired. The most successful procedures
used to improve the achievement motive have aimed intervention at the three
sensitive phases within the process of self-evaluation: more realistic goal setting,
a motive-arousing attribution pattern, and the resultant positive self-evaluation
(see Heckhausen & Krug, 1982).

Self-concept
The assumption seems justified that precisely those persons with low self-esteem
or a self-concept of low ability will succumb to the self-esteem-enhancing asym-
metry in their success and failure attributions. It appears, however, rather than
increasing self-regard those individuals prefer to maintain their negative self-
appraisal. Fitch (1970) found that to a greater (although not significant) degree
than persons with high self-esteem, subjects with low self-esteem ascribed their
failures to internal factors, particularly to lack of ability; they ascribed success to
good luck rather than to ability. Hence, they allocate attributions along the
stability dimension in such a way that success does not increase success expec-
 156 HECKHAUSEN

tancy to an extent inconsistent with self-esteem, and failure confirms failure

expectancy consistent with self-esteem.
Ames (1978) found a similar effect with fifth-grade children. The children
were randomly assigned first to outcome and then to reward structure (com-
petitive and noncompetitive conditions). The children were tested in same-sex
pairs, working simultaneously but independently on sets of achievement-related
puzzles. Low self-concept children tended to ascribe their success less to high
ability than did high self-concept children. This difference in ability attributions
was particularly marked under competitive conditions. Until the fifth grade, the
traditional instruction model (i.e., same amount of instruction time for all chil-
dren in the grade) seems to have flooded each child with such an abundance of
social comparison that a self-concept based on scholastic achievement appears to
be inescapable.
Nicholls (1976) investigated the attribution patterns of the self-concept
grounded in scholastic attainment. Again, the typical asymmetries of success and
failure attribution were evident. Remarkable, however, was that children low in
self-esteem did not attribute their success outcomes to high expenditure of effort,
but rather to good luck and to teacher’s assistance. It is not surprising, therefore,
that even success does not give rise to satisfaction with oneself and that effort
seems to be as ineffective with success as it is with failure. Nicholls (1979) found
a general age trend of an increasing covariation between attribution and scholas-
tic attainment for 6- to 12-year olds. Twelve-year-old children with high academ-
ic attainment ascribed their success to high ability and their failure to bad luck;
on the other hand, poorer performers attributed their success to effort expended
and good luck, and their failure to low ability.

Sex Differences
The attribution pattern a number of investigators have found for female subjects
is rather discouraging. In comparison to boys, girls ascribed success less to high
ability and failure less to low ability than to lack of effort or bad luck (Dweck
&
Repucci, 1973; Feather& Simon, 1975; Nicholls, 1975. 1978). Accordingly,
the
girls’ self-evaluations were less positive, their effort expenditure and persistenc
e
dropped off, and performance deteriorated. This is consistent with the typical
sex
differences found for success expectancy when girls are confronted
with new
tasks. Crandall (1969), in an investigation of 7 to 12-year-olds, found
that girls
had lower success expectancies than boys; a study by Dweck and Gilliard
(1975)
is consistent with these findings. During a failure series, subjects
were asked to
report their expectancies of success, either after each trial or prior
to the last trial
only. Boys showed higher persistence when they had made
their expectancy
statements prior to the trial; conversely, girls were more persistent
when they did
not have to make expectancy statements. If one assumes that
stating expectancies
 5. CAUSAL ATTRIBUTION PATTERNS 197

in the face of failure activates the personal attribution pattern, then the opposite
findings for persistence in both sexes are easily accounted for.
According to Dweck and Bush (1976), at first, it may be surprising that girls
are more inclined than boys, to adopt an attribution pattern that renders them
helpless: As a rule teachers regard girls as more hardworking and capable of
higher performance than boys (at least in the lower grades), girls receive more
approval than boys. They have resolved this seeming paradox by analyzing
teachers’ feedback in class. For boys, feedback refers to motivational deficits,
such as inattentive and disruptive behavior; for girls it refers almost exclusively
to achievement deficits. Hence, in the classroom boys experience more attribu-
tions to motivation and girls more attributions to ability. Therefore, it is an
unfortunate side effect of the more adapted behavior of girls that mediates a more
unfavorable attribution pattern for failures in mixed classes.
Obviously, prejudice concerning sex-typical differences in ability also plays a
part in generating differential attribution patterns. In a study entitled ‘‘What’s
skill for the male is luck for the female,’’ Deaux and Emswiller (1974), had
observer subjects attribute to ability or good luck successful male and female
performance on a visual discrimination task. Some of the objects to be discrimi-
nated were household articles belonging to the ‘*female’’ work environment);
others were tools (belonging to the ‘“‘male’’ work environment). For both sexes,
the observer subjects ascribed the causes of success in discriminating the
‘‘female objects’’ to ability. The equally successful discrimination of the ‘‘male
items’’ was attributed to ability for the males and to good luck for the females.
Similar sex-related attributional differences for equally successful performance
have been reported when males and females judge their own performances
(Deaux & Farris, 1977).

Summary
Taken together, the findings show a convergence, which is more remarkable in
view of the fact that the investigations stem from divergent theoretical traditions
comprise various tasks, differently manipulate success and failure, employ dif-
ferent causal factors, and use different techniques of assessment. In the final
analysis, the only element these studies have in common is that they are con-
cerned with mental tasks, the success or failure of which enables inferences of
intellectual ability. In all investigations, some subgroups with specific person-
ality traits tend to draw these inferences in a way that obscures the image of their
own abilities, calls it into doubt, and does not enhance self-esteem. Personality
characteristics linked with a negative attribution of ability cover a wide range of
phenomena: external control; depression; helplessness; fear-of-failure motive;
low self-concept, even being a female. The unfavorable effects of such an
attribution pattern have been amply demonstrated. Obviously negative attribu-
 158 HECKHAUSEN

tion is the determining mediator between success and failure on one hand and
between self-evaluative affects, success expectancy, persistence, and other
achievement outcomes on the other.
Within success and failure conditions favorable and unfavorable attribution
patterns are opposite. However, between success and failure conditions these
patterns are asymmetric. In the case of success, the opposition is primarily
between internal and external locus; in the case of failure, opposition is between
variable and stable causes. If one considers the findings, the data is revealing.
For instance, given success, an unfavorable attribution pattern is shown if the
individual does not make the obvious inference of high ability and attaches no
importance to high effort expenditure. Instead, he or she judges external factors,
such as good luck, support from others or (apparently less often) ease of task to
be the causes. Hence, in this case the preferred causes are pervasively external
and variable, rather than stable. Finally, they appear to be specific rather than
global (although there are scarcely any investigations into this last dimension.
Accordingly, success is something meted out to you, not something you can take
the credit for.
A favorable attribution pattern ascribes success to one’s own high ability.
High effort (as in the case of unfavorable attribution patterns) is hardly ever
perceived as a cause, probably because the inversely proportional compensation
of effort and ability would diminish the role of the latter. Primarily, all those
external causes, such as task ease, good luck or help from others, which might
jeopardize an attribution of success to own ability, are regarded as irrelevant. In
contrast to the unfavorable attribution pattern, the preferred causes are internal,
and they are stable and, presumably, global rather than specific. The contrary
nature of the two attribution patterns for success is centered on the question of
whether or not success is ascribed to internal and, at the same time. stable
causes, and to global rather than specific causes. The decisive causal dimension
for success attribution is locus of control (internal versus external).
For failure, if the individual tends to make the obvious inference of lack of
ability, the attribution pattern will be unfavorable; all other attributional alter-
natives that might weaken this inference are avoided. This applies especially
to
insufficient effort associated with ascriptions to internal factors; as well
as to
high level of task difficulty associated with external factors. The depressin
g and
helplessness-inducing consequences are still increased by the global nature
of the
internal and stable causes of deficiency. A favorable attribution pattern
attributes
failure to a still insufficient expenditure of effort, particularly if
the task is
perceived as difficult. In that event, the inversely proportional compensat
ion of
effort and ability is brought into play so that failure does not elicit
rash conclu-
sions implying lack of ability. Instead of, or in addition
to, lack of effort,
variable causes, including external ones such as bad luck, are invoked.
Variable
causes are preferred; stable causes—lack of ability, as well
as task difficulty—
 5. CAUSAL ATTRIBUTION PATTERNS 159

are excluded. The preferred variable causes are more likely to be specific ones.
Hence, the failure experience can be perceived as something that is not inevitable
and that can be overcome in the future.
The contrast between the two attribution patterns for failure consists in the
stability dimension; global, stable causes are set against specific, variable
causes. In both the favorable and unfavorable attribution patterns, high task
difficulty plays no differential role among the stable causes. The contrast is the
sharpest within the dimension of internal locus of control: lack of ability versus
lack of effort. In the first case, the prospects for success are negligible and
increasing effort expenditure is hardly worthwhile; in any case, the pay-off will
be low. In the second case everything remains open and there is no need to doubt
one’s own abilities; one could accomplish if one wanted to accomplish.

POSSIBLE TYPES OF ATTRIBUTION PATTERNS AND

THEIR EMERGENCE

The summary given presents a somewhat idealized picture of the opposing at-
tribution patterns for success and failure. The findings of a considerable number
of investigations have not pinpointed the contrast expected with regard to all the
relevant causes. Nonetheless, in order to infer, a maximally favorable or un-
favorable image of one’s own abilities, these findings have been assembled to
overall patterns that employ all possible attributions for success or failure. The
two opposite tendencies will henceforth be called ‘*Positive Attribution Pattern’’
(PAP) and ‘‘Depressive Attribution Pattern’? (DAP). Obviously, PAP and DAP
will not in all cases incorporate all the elements and relations outlined earlier, nor
will they always satisfy the requirements of a psychological logic.
In the case of PAP, failure is attributed either to insufficient effort or to bad
luck, depending on the situational factors; insufficient effort seems the more
likely inference for high task difficulty, and bad luck for moderate task diffi-
culty. Nor will PAP and DAP always be applied only to situational demands.
Both PAP and DAP may involve individual differences, that so far attribution
research seems to have overlooked. Above all, both patterns may take different
forms within a developmental progression before they assume the configuration
outlined earlier. In what follows, the various forms taken by PAP and DAP are
purely hypothetical constructions. The author’s only interest is in the question of
what hypothetically an individual should do in order to make own abilities
(competence) appear either in a favorable or an unfavorable light. Concerning
the various types of PAP and DAP analyzed, conjectures are made as to the stage
of development or age at which these emerge and about possible antecedents of
individual developmental differences.
 160 HECKHAUSEN

Type 1: Task Difficulty and Ability as Multiple

Sufficient Causes for Success and Failure

The simplest way to explain success and failure is by means of a causal schema
with only one covariation dimension, that is, with a single external or internal
causal factor: task difficulty or ability. Low task difficulty and high ability are
facilitating factors; high task difficulty and low ability are, inhibitory factors.
Covariation with task difficulty alone seems to have a developmental primacy
(Heckhausen, 1982; 1983). Given successful outcomes, the task must have been
easy, given unsuccessful outcomes, the task must have been difficult. Alter-
natively, the same covariation is applied to ability: given success, ability is high;
given failure, ability is low. When ability is discussed here it should be kept in
mind that in general ability cannot be conceived as a stable disposition prior to
age ten. For younger children, the concept of ability is still coupled with mo-
mentary intensity of effort; therefore, for them, ability is an intraindividual
variable dimension, derived primarily from the perceived degree of effort ex-
pended (see Heckhausen, 1983).
The two factors are mutually related—the more difficult the task, the more
ability is needed for a successful outcome. Conceivably the simplest combination
of two causal factors into one attribution pattern is the multiple sufficient causal
schema (Kelley, 1972). The schema presents three logically equivalent combina-
tions of causal factors that differ from each other in their psychological meaning.
Thus, given success: (a) the two facilitating factors (low difficulty and high
ability) need not coincide, but success attribution may be made; if only one
facilitating factor is present, either (b) low difficulty associated with low ability,
or (c) high ability associated with high level of difficulty. Conversely, given
failure, even one inhibitory factor will suffice to explain the outcome. Again
there are three conditions: (a) two of the inhibitory factors (high task difficulty,
low ability), however, only one inhibitory factor need be present, either (b) high
level of task difficulty, even though it is associated with high ability; or (c) low
ability associated with low difficulty. Such a schema of multiple sufficient causes
for failure is equivalent to a schema of multiple necessary causes for success.
This means that the outcome will be success only if the task is easy (not too
difficult) and ability is high.
The multiple sufficient causal schema is an ideal basis for PAP and DAP.
Logically equivalent alternatives of explanation differ in their psycholog
ical
meaning. Regardless of whether one intends to explain success or failure,
the
causal factors can be combined so that in all cases, either high ability
(PAP) or
low ability (DAP), is corroborated, or at least not called into question.
Of the
alternatives previously indicated, the preference for *‘high ability-hi
gh task diffi-
culty’’ always connotes PAP, and the combination ‘‘low ability-l
ow difficulty’
connotes DAP. Figure 5.1 contains the particular causal combinati
ons.
 5. CAUSAL ATTRIBUTION PATTERNS 161

Task Difficulty Task Difficulty

low high low high

high EI. high

Ability Ability

low low DAP

a) after success b) after failure

FIG. 5.1. Schema of multiple sufficient causes and preferred PAP and DAP
combinations of ability and task-difficulty attributions. According to the schema
of multiple sufficient causes a facilitating cause is sufficient for the occurrence of
success (high ability or low task difficulty), and an inhibitory cause for the occur-
rence of failure (low ability or high task difficulty). See framed quadrants. A
positive attribution pattern (PAP) is able to augment or at least to maintain a self-
image of high-level ability. It first picks a self-serving cause—i.e., high ability
after success, and high task difficulty after failure—and adds the other cause in a
countereffective direction: high task difficulty in case of success (underlining or
even augmenting high ability), and high ability in case of failure (high ability need
not to be excluded when level of task difficulty was very high). A depressive
attribution pattern (DAP) is able to maintain or even to further depreciate a self-
image of low-level ability. It first picks a self-depreciative cause—i.e., low task
difficulty after success, and low ability after failure—and adds the other cause in a
countereffective direction: low ability in case of success (low ability need not be
excluded when task difficulty is low) and low task difficulty in case of failure
(underlining or even further discounting low ability).

Type 2: Direct Compensation of the one Causal Factor

(internal or external) by the Complementary Causal Factor
Insofar as it is ability ascription that discounts the opposite effect of high task
difficulty, two combinations within the multiple sufficient causal schema stand
out. These combinations are PAP for success outcomes (high ability has mas-
tered high level of difficulty) and DAP for failure outcomes (low ability was
unable to master low difficulty). Furthermore, processes of augmentation or
discounting of one’s own ability may set in after having overestimated or under-
 162 HECKHAUSEN

estimated the level of difficulty. The augmentation and discounting of one’s own
ability is rooted not only in the causal schema of multiple sufficient causes, but
also in a causal schema that involves direct compensation of one causal factor
(facilitating or inhibitory) by another causal factor (inhibitory or facilitating).
Hence, the greater the master difficulty, the higher the ability; the lower the
unmastered difficulty, the lower the ability. Accordingly, PAP should tend to
overestimate task difficulty after success as well as after failure. For example,
the more extreme the level of difficulty appears to be, the less failure will
dampen the belief in one’s own high ability. On the other hand, DAP should tend
to underestimate task difficulty after success as well as after failure. For exam-
ple, the lower the difficulty appears to be, the less success would alter the belief
in one’s own low ability (see Fig. 5.1).
The relief afforded by the PAP combination for failure, when the level of
difficulty is very high has been demonstrated by a number of helplessness experi-
ments, some of which have already been referred to (Frankel & Snyder, 1978;
Klein et al., 1976; Snyder, Smoller, Strenta & Frankel, 1981; Tennen & Eller.
1977). In these studies, after the difficulty of the training task had been under-
scored, the test task elicited no helplessness effects. When difficulty is held to be
high, even failure does not call one’s own ability into question.
For this particular PAP combination, Frankel and Snyder (1978) and Snyder
et al. (1981) introduced the term of ‘‘Egotism’’. They have confronted help-
lessness theory with egotism as an alternative explanation. ““Egotism’’ brings to
mind the “‘self-handicap’” strategy (Berglas & Jones, 1978) discussed earlier. In
a highly effective, although self-deceiving way, personal ability could eventually
be augmented or discounted if after successful or unsuccessful outcomes. task
difficulty is either overrated (PAP) or underrated (DAP). To overestimate task
difficulty after the outcome makes success appear the more splendid, and failure
can be more readily excused; to underrate task difficulty lessens the value of
success and makes failure even more disheartening. No systematic investigations
have yet been undertaken of augmenting and discounting ability attributions
through over or underestimating task difficulty.

Emergence of Type 1 and Type 2 of PAP and DAP. Shultz and Butkowsky
(1977) found that children as young as 5 years of age mastered the schema
of
multiple sufficient causes, when in addition to the information about success
or
failure the cause given was either ability or difficulty (difficulty even
in the form
of consensus information: “Larry is solving a puzzle that four other
children
cannot solve’’). Correspondingly, when success occurred, in spite
of an inhibito-
ry factor (hard task or lack of ability), most of the children inferred
the other
cause as facilitory (high competence, easy task). The findings
of a study by
Falbo (1975) with 5-year-old children of high and low
IQ reveal attribution
patterns of the first type in terms of PAP and DAP.
Children of higher IQ
preferred to explain failure in terms of task difficulty; whereas,
children of low
 5. CAUSAL ATTRIBUTION PATTERNS 163

IQ preferred low ability. The two IQ groups came from different family back-
grounds: high-IQ subjects from middle-class homes; low-IQ subjects from fami-
lies receiving welfare benefits. In explaining their success (in terms of PAP): the
former used their (high) ability much more often, in explaining their failures (in
terms of DAP), the latter used their (low) ability more often.
These findings suggest that as early as toward the end of preschool age one
can expect to find not only cognitive prerequisites for the multiple sufficient
causes schema, but also individual developmental differences in terms of PAP
and DAP of the first type.
It seems that the prerequisites for the first appearance of the second type of
DAP and PAP, that is the augmenting of ability after success at a difficult task
and the discounting of ability after failure at an easy task, are hardly attained later
than the prerequisites for the first type patterns. The causal schema of direct
compensation underlying PAP and DAP of the second type seems to be derived
from an integration of the schema of simple covariation (of a facilitory cause
with an effect) and from the schema for multiple sufficient causes. Simple
covariation of the facilitory causes of ability or effort with the outcome has been
observed as early as by age 4 (see, overview in Heckhausen, 1983); and, as
reported earlier children as young as 5 years of age were found to dispose of the
multiple sufficient causes schema. Kun (1977) found direct compensation be-
tween ability (or effort) and difficulty among 5-year-olds, although the compen-
satory relationship was more evident in children in the higher grades (third and
fifth grades), and initially was inferred more frequently after success than after
failure. Similarly, Karniol and Ross (1976) found augmentation and discounting
in motivation attributions of 4-year-olds (on first sight this cannot be equated
with outcome attributions).
In light of the studies available, the assumption that there are substantial age
lags in the acquisition of the developmental prerequisites for PAP and DAP of
the first and second type does not seem justified. Individual differences may still
emerge prior to school entry, at this age, the developmental primacy of difficult
attributon is gradually disappearing. Experience of one’s own competence is
clearly evident and comprises an as yet diffuse concept of ability, still not
independent from the effort concept. In the first few years of school attendance,
both the ability and effort concepts are subject to the coupling schema (‘‘halo
schema’’ Kun, 1977; Surber, 1980; Tweer, 1976; overview, Heckhausen, 1983).
Because only differences in degree of effort but not of ability, can be seen in
oneself and in others, ability is derived from the perceived degrees of effort
expended (and not the reverse, that is, effort from ability), until the inverse-
compensatory causal schema for ability and effort is mastered. This means that
children from 3-4 years to 7-9 years experience themselves as having higher
ability (competence) in the performance of a task, the more effort they have
exerted to accomplish it. As long as the focus is only on effort expenditure, this
may protect the young child against the depressing effects of failure. Rholes,
 164 HECKHAUSEN

Blackwell, Jordan and Walters (1980) in a repeated failure series, tried in vain to
induce ‘‘helplessness’’ in children from 5 to 9 years. Neither the performance
nor the persistence of the subjects deteriorated. The authors were able to induce
helplessness in only 10 to 1 1-year-olds who had already overcome the coupling
schema and had mastered the inverse-compensatory schema. The coupling sche-
ma was most clearly in evidence among the 5 to 6-year-olds. Even after repeated
failure, this age group did not rate their ability as less high than a comparison
group that had consistently experienced success.
Because the child infers his own ability from the amount of effort he has
expended, one may conclude that the more opportunity to work on demanding
tasks a preschool age child (about 3 to 6 years) has and the more the covariation
between effort expenditure and outcome is a part of this experience, the more the
child will perceive himself as competent. Accordingly, he will, after both suc-
cess and failure, prefer PAP combinations of sufficient multiple causes, that is,
high ability and high level of difficulty not only after success but also after
failure. In contrast, DAP combinations will presumably be preferred by children
who were either overtaxed or undertaxed by the demands made upon their
competence in the preschool period. This is because, in either case, trying hard is
a rather rare experience that offers little opportunity to infer high ability. When
children’s competence is undertaxed, task demands are met too quickly and too
easily; when competence is overtaxed, children will stop trying too soon and
cannot perceive a connection between degree of expended effort and attained
outcome.
One must consider not only the beginning but also the phasing out of the
critical developmental span for PAP and DAP of the first and second types. After
a child enters school, cognitive developmental steps and new experiences in-
creasingly undermine the emergence and maintenance of both attribution pat-
terns. Cognitively, the coupling schema is replaced by the schema of an inverse
compensatory relation between effort and ability. More effort, associated with
the same outcome no longer leads to attribution of higher but rather to of lower
ability. At the same time, the children are increasingly able to grade task diffi-
culty, without respect to their own success or failure, on the basis of social
comparison information (consensus) concerning the performance distribution
within the class. This marks a change: a preconceived ability attribution made
within the schema of multiple sufficient causes gives way to self-ratings
based
more firmly on reality, a development that takes place at age 9 to 10 (Nicholls,
1978).
This is not to imply that with the progression of development, DAP
and PAP
are ultimately undermined and finally discarded completely. Rather,
it is more
probable that, as a result of developmental progression, PAP
and DAP are
transformed into attribution patterns of a higher order, within which
the essential
characteristics that mark PAP and DAP are retained in forms commens
urate with
the attained level of development. The schema of inversely
proportional compen-
 5. CAUSAL ATTRIBUTION PATTERNS 165

sation for effort and ability makes this possible (see Attribution Patterns: Type 4,
later).

Origin of Individual Differences. As mentioned, PAP will be fostered by

developmental environments in which 3 to 6-year-olds find opportunities and
incentives for coping with demanding tasks, to set goals for themselves that do
not overtax or undertax their developing competencies, to learn to be indepen-
dent, and to experience, through a variety of activities, the association between
effort and outcome. DAP would be facilitated by the inverse conditions, es-
pecially by demands that overtax or undertax the developing levels of children’s
competencies and that block covariation of effort and persistence with action
outcomes; therefore, the child will infer from the coupling schema, that his
ability is low. To what extent the constellation of the developmental environment
will tip the scales in favor of PAP or DAP will be determined substantially by
parental (particularly maternal) child-rearing patterns. The differences in family
background associated with PAP and DAP among 5-year-olds point, even if
somewhat ambiguously, in this direction (Falbo, 1975).
More illuminating are the many studies of motive origin that concern the
presumed influence of independence training on the formation of the achieve-
ment motive (see overview in Heckhausen, 1980a). If one holds success and
failure motivation to be equivalent to PAP and DAP, then, in keeping with the
findings on the influence of independence training, the following child rearing
patterns should foster PAP: Expectations of independence and ability to achieve
are commensurate to the child’s attained level of development, are child cen-
tered, and do not overtax the child. Freedom and opportunities for probing one’s
developing competencies are available and are made accessible.
DAP will be facilitated by parents who do not particularly value their child’s
successful performance or who are not particularly sensitive to the developmen-
tal adequacy of their demands and consequently expect too much or too little
from the child; freedom and opportunities for self-initiated development are
heavily restricted or are not made acessible. When a child’s competencies are
overtaxed, repeated failure combined with negative parental reactions will pro-
duce a depressing image of one’s own competence. Moreover, such children,
when confronted with difficulties, will precipitately give up trying, so that
covariation of effort and outcome cannot be experienced. The same deficiency
will occur if the child is permanently undertaxed by the demands made on him.
Even if the parents were constantly to praise the child for easily attained **suc-
cesses,’’ the effort-based inference of a high competence is denied the child. To
be sure, the child is able to some degree to compensate for such parent-mediated
deficits, provided freedom and opportunities for probing developing competen-
cies are somehow accessible.
However, when children can themselves choose the levels of difficulty of
their activities, under their parents’ control or within the freedom and oppor-
 166 HECKHAUSEN

tunities for development afforded outside the home, the question is: To what
extent have they been encouraged or discouraged in a preferred level of difficulty
that make demands on their attained competence (ability and effort, or per-
sistence)? As shown in Fig. 5.1, the PAP combination comprises high levels of
difficulty whereas the DAP combination entails low levels of difficulty. The
child’s preference for either of these attribution patterns may simply be due to a
difference in the parental reaction pattern to success and failure. The decisive
point is whether parental reactions are determined only by the outcome itself
(success or failure) or whether consideration is also given to level of task diffi-
culty that has been overcome and hence also to effort expended.
A reaction pattern that fosters preference of high levels of difficulty (PAP)
does not tally with simple reinforcement theory, always praises success and
makes light of failure. Rather, success should be praised only when the child has
had to try hard and has mastered a difficult task; failure should be criticized only
if the child has given up too soon, that is, has failed at a task that was no longer
too difficult. Such a reaction pattern enables the child to experience that action
outcomes are primarily competence-related, and that success and failure are
attributable less to task difficulty than to one’s own endeavors. The child will
prefer more demanding tasks, not only because he or she has learned that only for
difficult tasks will success be evaluated positively and failure not be negatively
assessed, but because accomplishing a more demanding task offers a stronger
sense of one’s own competence. According to the schema of multiple sufficient
causes, this becomes more apparent because success at a more demanding task is
in all cases attributable to high competence, but failure is not necessarily at-
tributable to lack of competence. This gives precisely the PAP combination
shown in Fig. 5.1: high task difficulty and high ability.
Conversely, DAP would be facilitated if the parents reinforced their child
only according to task outcome, not taking into account the level of task diffi-
culty the child has overcome (a practice indicative of the parents’ lack of sen-
sitivity for their child’s level of development). In this case, the parental reactions
do not covary with the effort expenditure indicative of competence, but rather
with the level of difficulty experienced. Easy tasks will bring success and ensure
praise, even if the child has made no effort and consequently has not experienced
a sense of personal competence. On the other hand, difficult tasks regardless
of
how hard the child has tried, elicit expectations of failure and blame. That is
why
that child prefers easily mastered tasks. Failure is always attributable to lack
of
competence, but success is not necessarily attributable to attained competence
.
This is the DAP combination: low difficulty and low ability.
The conclusion to be drawn from the diverging difficulty preferences
of the
children and the reinforcement patterns practiced by their parents
is that DAP
children not only encounter more success, but also manage to
get disproportion-
ately more praise from their parents than do PAP children. This
inflation of
 5. CAUSAL ATTRIBUTION PATTERNS 167

praise for too easily attained successes neither is associated with the experience
of competence based in real effort, nor does it allow this kind of inference.

Conclusions Concerning the Formation of Aspiration Level. When pre-

schoolers in task choice studies were asked to choose between visually repre-
sented degrees of difficulty of a task they had already practiced, they preferred to
resume on the difficulty level already mastered. However, when it was not a
matter of choice between tasks but rather of setting a goal along a uniform
continuum of achievement, the children preferred goals that went beyond what
had already been attained. This contradiction between the two experimental
paradigms of level-of-aspiration setting can easily be resolved if one considers
that in each case a different type of covariance information is offered; informa-
tion that makes salient the causal factor either of difficulty or of ability
(Heckhausen, 1982). In the case of task choice, distinctiveness information was
provided (Kelley, 1967) along with its level of difficulty (what can be solved and
what cannot); in the case of goal setting consistency information was provided
along with its level of ability (what one has done and what one has not yet been
able to do). A study repeated a few months later showed that the initial related-
ness between each of the experimental paradigms and one of the two dimensions
of covariance had become less close. In task choice, more children than pre-
viously preferred unsolved tasks, and in goal setting fewer children chose an
unrealistic level of aspiration (Heckhausen, 1980b).
The two contrasting developmental transitions should be associated with PAP
and DAP. In task choice, PAP children should show preference for unsolved
tasks earlier than DAP children because by exerting effort they have learned to
overcome difficult tasks and are more confident of their ability than are DAP
children. For goal setting, however, the reverse should apply: PAP children
should tend less, or later, than DAP children to lower the initially attempted high
level of task difficulty to the attained level or below it. Both conclusions remain
to be confirmed by experimental investigations:

Type 3: Discounting Success and Augmenting Failure

by Attribution to External, Task-Irrelevant Causes

Success at very difficult tasks and failure in very easy tasks are unexpected
events. They contradict one’s idea of what one can do (the assessed relation of
one’s own ability to task difficulty). After unexpected success, the emergent
cognitive dissonance can be reduced either by augmenting one’s own ability or
by discounting task difficulty. Correspondingly, after unexpected failure, one
can either augment task difficulty or discount one’s own ability. Whether one is
able to resist the temptation offered by the first alternative, to increase self-
esteem, or accepts the sobering realization of the second alternative, one will be
 168 HECKHAUSEN

acting in conformity with the two attribution patterns already discussed—PAP

and DAP, Type 2.
To attain dissonance reduction in this way may, however, seem too irrational
to be feasible. This will be the case particularly when consistency information
has been given that makes the unexpected event appear to be an exception. The
apparently inexplicable residue of other causes is then explained by attribution to
chance or to good or bad luck. Thus, it has been shown that the more a successful
outcome contrasts with previous outcomes the more success is attributed to luck
(Meyer & Plöger, 1979). On the contrary depending on the circumstances, other
variable causes for which one cannot be held responsible may be inferred,
especially psychophysiological states (such as fatigue) and support by or hin-
drance from others.
Accordingly, PAP and DAP strategies can be juxtaposed as follows. After
unexpected success, PAP rather than DAP will tempt actors to augment their
own ability—they will draw less on luck or on the help of others to explain the
event. When PAP is dominant, unexpected failure is played down by drawing on
bad luck, on being in bad form, or on hindrance by others, so that one will not
have to discount one’s own ability. As a precautionary step, again in conformity
with PAP, self-handicaps may be arranged to facilitate augmenting success and
discounting failure following the achievement outcome (Berglas & Jones, 1978).

First Appearance and Development of Individual Differences. There are as

yet no studies concerned with the development of attribution factors of ‘good
luck”’ or ‘‘bad luck’’ or support or hindrance from others. One of the cognitive
prerequisites for attributions of good luck or bad luck elicited by an unexpected
outcome is the ability of the child to form expectancies in which the preceding
feedback has been sufficiently integrated. To do this, the child must be capable
not only of sorting consistency information, but also of conceptualizing ability
and difficulty as stable factors. Before the first years of grade school, the concept
of ability is not clearly separated from fluctuating effort, expectancies until then
remain elastic (see Heckhausen, 1982). Only from about the age of 8 to 9 years
(for girls generally somewhat earlier) does prior success and failure systemat-
ically affect success predictions for boys (Crandall, 1969: McMahan. ie ee
Parsons & Ruble, 1977).
The formation of Type 3 attribution patterns, at least with respect to em-
ploying good luck or bad luck attributions in terms of PAP or DAP, can not be
expected to start before the age of 8 to 9. The developmental factors that produce
individual differences in the use of good luck or bad luck attribution are not
yet
known. But we may assume that once exceptional achievement outcomes have
been experienced as an unexpected event or as an event that needs to
be ex-
plained (because it is inconsistent with their stable concept of ability),
and once
they are able to employ the concept of chance, children elaborate
their so far
 5. CAUSAL ATTRIBUTION PATTERNS 169

developed PAP and DAP strategies (Type | and Type 2) by employing luck and
bad luck attributions.
The situation is different for the factors of social support or hindrance. From
babyhood onward children have assimilated many-sided experiences in relation
to both these factors. Thus, the refusal of offered support is observable already in
the ““wanting-to-do-it-oneself’’ attitude that emerges in the second year of life
(see Heckhausen, 1983). DAP is facilitated not only when (as is the case for
attribution pattern Type 1) children’s independence is not respected but also
when support is needlessly obstructed simply because the parents—or later the
teacher—is not willing to acknowledge the child’s incomplete or imperfect
achievement outcomes. Findings from a study by Nicholls (1976) are indicative
of the teachers’ influence in this respect. As reported earlier, 11-year-old stu-
dents with a low self-concept in terms of scholastic attainment, attributed their
success in the classroom to luck and to support from their teacher.

Type 4: Compensatory Causal Schemata for Ability

and Effort
In Type | and Type 3 attribution patterns the allocation of causes is arranged
among internal and external factors. However, for attribution pattern Type 4,
allocation of causes is based on the facilitative internal causes of ability and
effort. Within the causal schema of graded effects (Kelley, 1972, 1973) attained
outcomes are a direct function of both factors. On the other hand, the same
outcomes can result from entirely different inverse proportions of ability and
effort. Lower ability can be compensated by greater effort, less effort by higher
ability (effort compensation and ability compensation, respectively).
Again, the possible proportions of the two causal factors are also logically
equivalent in the Type 4 attribution pattern. They are not equivalents in psycho-
logical terms; ability attributions generate more subsequent affect (self-evalua-
tion), create more actionguiding cognitions (success expectancies), and lead to
more subsequent activities (persistence, effort, expenditure, task-choice, seeking
feedback, etc.). Although in evaluation by others, the attribution of responsibil-
ity and the reinforcement given depend on the attribution to effort, this does not
hold for self-evaluation. Here, ability attribution generates more affect and has
greater impact on action.
After success and failure, in order to make own ability appear greater, the
PAP strategy is to minimize the effect ascribed to effort. An extreme DAP
strategy would consist in the reverse procedure. In order to make one’s own
ability appear the lower, the effect ascribed to effort will be maximized. Because
the two factors can largely compensate for each other only after success, after
failure, the compensation schema is combined with the schema of multiple
necessary causes. To avoid failure, a certain minimum of either effort or ability
170 HECKHAUSEN

would have been required, even if the other factor was maximally strong. Within
this combination of schemata, failure is regarded as a less frequent, more ex-
treme event, as a rule, then, multiple necessary causes, rather than multiple
sufficient causes, may be assumed.
To what event an inverse relationship (i.e., compensation between ability and
effort) actually is employed will depend on the perceived level of task difficulty.
Success at a very difficult task and failure at a very easy task are infrequent
(unexpected) events. In both cases the respective presence or absence of internal
factors appears to be a multiple necessary schema, rather than one for which only
one of the two factors could provide a sufficient explanation.
Accordingly, to deal with such extreme effects, the scope for inverse-com-
pensatory arrangements is also narrowed down. This curtailment facilitates the
PAP and DAP strategy in explaining failure: Too little effort has been expended
to render one’s own high ability perceivable (PAP); when own ability is so low,
even great effort makes no difference (DAP). The opportunities for compensa-
tion have the widest scope in the broad field of intermediate difficulty. If one is
successful at tasks of intermediate difficulty, lack of effort makes own ability
appear more decisive as the causal factor; failure still does not cast doubt on
one’s own ability.

First Appearance. A basic prerequisite for Type 4 attribution patterns is

mastery of the compensatory causal schema for effort and ability. To test this
assumption, an effect and one of the two internal factors were given, and the
other factor was to be inferred (Kun, 1977; Surber, 1980). In a paired com-
parison of two actors with the same outcome, but with differential weighting of
the given factor, the subject was to determine the actor for which the factor to be
inferred was stronger (Surber, 1980; Tweer, 1976; see overview in Heckhausen.
1983). Compensation for a given degree of ability by effort to be inferred (effort
compensation) precedes the development of compensation for a given degree of
effort by ability that is to be inferred (ability compensation). Depending on the
kind of task and on memory load of how the information was presented, effort
compensation was used by some children as young as 5 years of age and by
others as old as 10. The majority of the children simply equated the degree of
effort with level of outcome (centered covariation). Ability compensation ap-
peared with an age lag of about one year between the ages of 6 to 11, after which
the majority of the respective age group no longer simply inferred degree
of
ability from degree of effort expenditure nor equated the two (coupling
or
‘‘halo’’ schema).
The earlier appearance of effort compensation is easy to understand.
One
takes into account that children can experience effort, which, in
contrast to
ability, is subject to volitional control—a factor that varies
for each individual.
Children learn early that some initially nonattained goals can be
reached by the
expenditure of additional effort (persistence). Later ability
compensation ap-
 5. CAUSAL ATTRIBUTION PATTERNS 74

pears, the question arises whether PAP and DAP Type 4 presuppose only effort
compensation. This question makes explicit the difference between the presenta-
tion of a task within an experimental setting and self-attribution in an everyday
context. In the latter case, neither effort nor ability is given as a known factor,
but both must be inferred and varied in opposite directions. We may assume that
not only ability compensation but even a higher order ability—for reversible
operations—is required; this is not the case for experimental tests that provide
the memory aids of fixed levels of outcome and one known factor. Conse-
quently, the inverse compensatory relation between ability and effort seems to be
mastered later than experimental studies have indicated—probably not earlier
than around the age of 10.
Use of the inverse compensatory relation between ability and effort is but one
of the cognitive prerequisites for the appearance of Type 4 PAP and DAP. Also
essential for their appearance is that a clear differentiation be made between the
concept of ability and the concept of effort and that ability as a personal charac-
teristic be conceived of as a stable trait. As long as the concept of ability is only a
stable trait, not an individual attribute of the self-concept, the ability concept is a
requirement only for the mastery of the compensatory causal schema in dealing
with both ability and effort (which itself is a cognitive prerequisite for Type 4
attribution patterns). Very few studies have been undertaken to tell from what
age onward ability is not only conceptualized as a stable factor but also perceived
as a self-concept of the level of one’s ability. In cross-sectional studies, Nicholls
(1975) investigated accurate perception of children’s own reading attainment
relative to all others in the class. Only from the age of 9 to 10 did the children
also assign to themselves the lower ranks in the distribution of reading perfor-
mance, and only between the age of 10 to 13 years did perception of their own
attainment correlate significantly with teacher ratings.
Like the first developmental prerequisite for Type 4 attribution patterns (mas-
tery of the compensatory schema), the second prerequisite (stable concept of
ability and individual self-concept of ability) indicates that only after the age of 9
to 10 will the first individual differences in terms of PAP and DAP Type 4
emerge.
The elementary school classroom, where children are constantly exposed to
social comparisons with their peers, appears to be the best training ground for
acquisition of these two cognitive prerequisites. In another study, Nicholls
(1979) examined age differences in relationship of attainment and perception of
their own attainment with attributions by having children consider occasions
when they performed well or poorly in reading. His subjects were 2nd, 4th, 6th,
and 8th graders. The self-ratings of the 8-12 year-olds correlated increasingly
with the teacher ratings of attainment. The emergence of individual differences
in self-perceived attainment seem to appear first in this age group. Only the 12-
year-olds’ (but not among the 10-year-olds) self-concepts of reading ability were
correlated with attribution patterns in terms of PAP and DAP Type 4. Nicholl’s
172 HECKHAUSEN

findings confirm the assumption that individual differences in Type 4 attribution

patterns develop only after the age of 10.
Finally, there is yet a third developmental prerequisite: Self-attribution of
ability must begin to generate relatively more affect than effort attribution. In
other words, a self-concept of high ability becomes the guiding principle for the
formation of PAP, a selfconcept of low ability, for DAP. In the first case,
positive self-evaluation results; the self-evaluation engendered in the second case
is by now a depressing one, unlike the earlier years, when low ability still
produced a positive self-evaluation, because on the basis of the coupling schema,
high effort still made it possible to infer high ability. In a study with 10- to 13-
year-old students that examined the influence that of self-evaluation on causal
attribution, only for failure-motivated subjects did effort attribution still generate
more affect than ability attribution, an attributional affect characteristic for eval-
uation by others (Heckhausen, 1978).
Considering all three cognitive prerequisites discussed we can say that Type 4
attribution patterns may be expected to emerge at the age of 11 to 12 years. It is
worth mentioning that in her helplessness studies Dweck chose her subjects from
this age group.

Origin of Individual Differences. One determinant for the development of

Type 4 PAP and DAP is the traditional way school instruction is organized: a
setting in which same-age classmates are constantly confronted by social com-
parison of achievement. Because the teachers’ rating of their students in their
everyday interaction with them is oriented primarily toward a social (and not an
individual) reference norm (e.g., Rheinberg, 1980), good students almost ex-
clusively experience success, bad students failure. Consequently, for high
achievers success becomes an event that can only be explained by a sta-
ble-internal cause (ability) and not by a variable-internal cause (effort). This
history of outcomes and their causal attribution prepares the ground for PAP. The
situation is reversed for low academic achievers. For them it is failure that is the
habitual and recurrent event to be explained only by a stable cause—lack of
ability. Their rare successes must stem from a variable cause—great effort, if not
good luck. The study (Nicholls, 1979) reported earlier underscores this influence
of school instruction on the genesis of individual differences. The differential
pattern of teacher feedback that boys and girls typically receive seems to make
girls more susceptible to DAP, discussed earlier (e.g., Dweck & Bush, 1976).
To be sure, it would be overestimating the influence of school, if in addition
to showing differences in ability, schoolchildren did not earlier, in the preschool
period, differ from each other in attribution patterns of a simpler type. Thus,
PAP Types I, 2 and 3 predipose children to a positive self-perception and
achievement behavior by the time they start school and thus favor positive
academic attainments. DAP Types 1, 2, and 3 mark out a more negative develop-
 5. CAUSAL ATTRIBUTION PATTERNS 173

ment. Thus, with cognitive development, the more simple attribution patterns are
transformed into more complex ones—a process in the course of which earlier
attribution patterns are supplemented by those assimilated later. Within this
process the weight given to the causal dimension shifts from internal versus
external causal attribution to stable versus variable attribution. The details of this
process and how it operates are not yet known. Longitudinal studies are needed.

Type 5 PAP and DAP: Specific versus Global Ability

As reported earlier, Abramson et al. (1978) maintain that the dimension of
globality plays a crucial role for the transfer of induced helplessness to other task
activities. The more that failure on a specific task is attributed to lack of global
ability (e.g., to “lack of intelligence”), the sooner the individual concerned will
give up. Peterson and Seligman (this volume) have highlighted the impact of the
globality dimension (besides internality and stability) for depression and its
aggravation following a midterm examination.
If after success and failure the causal factor of ability is placed at the opposite
endpoints of the globality dimension, a fifth and final type of attribution pattern
can be distinguished. In the case of PAP, given that success is ascribed to a
global (general) ability, and failure, possibly, to a highly specific lack of ability,
this fifth type of attribution pattern serves to maintain a self-concept of high
ability. For DAP, the ‘‘insiduous attribution style,’’ the situation is reversed:
success is ascribed to a highly specific partial competence; failure, to global lack
of ability. Unfortunately it is not yet known to what extent and under what
conditions attribution patterns of Type 5 appear as isolated, specific attributions.
Recent research efforts on helplessness start only from the total strategies of
“attributional styles.’’ The patterns “internal, stable, global’’ and ‘‘external,
variable, specific’’ alternatively for good and bad events, are placed in opposi-
tion to each other. These attribution patterns also circumscribe the four Types of
PAP and DAP already examined, namely, Types |, 2, and 3 in which internal
and external causes are played off against one another within the locus dimen-
sion, and Type 4, in which the inversely related proportions of the internal causal
factors are reallocated along the continuum of the stability dimension.
We are still in the dark about three questions: First, we still know nothing
about the frequency of use of the total strategy (i.e., comprising all five sub-
strategies), nor do we know anything about the possible combinations between
the five attribution patterns, which might be provided by samples from age
groups past the age of 10 or by subjects chosen by other sampling criteria. If the
episodes to be evaluated were tailored to fit the purpose, the closed- and open-
ended questionnaire format developed by Seligman et al. (1979) might yield data
on the basis of which it would be possible to construct an empirically well-
founded typology of attribution patterns and of their frequency. The findings
174 HECKHAUSEN

published so far only report mean scores of, and intercorrelations between, single
causality dimensions (e.g., Peterson & Seligman, this volume). They do not
bring us any closer to answering this question.
Second, if all the cognitive prerequisites are given, it can be assumed that as
more personality characteristics deviate toward psychopathological states—from
slight depressive symptoms to severe depression—a growing number of the five
different types of DAP will come to be combined into the total strategy, which
employs each of the logically inferrable attributions within a specific self-es-
teemreducing pattern. Possibly there are diagnostic patterns for differentiating
between particular personality traits (e.g., for type and degree of failure-oriented
achievement motives) or for the degree of risk when certain bad events occur.
There is much to be said for such a differential diagnostic approach. Each type
of attribution pattern focuses mainly on one of the three causal dimensions, and
each of the three dimensions generates specific effects. The dimension of locus is
decisive for self-evaluation; stability, for success expectancy and therefore, for
persistency of actions; and the globality dimension determines the extent to
which experiences of success or failure are transferred to other tasks and domains
of activity. For the last point, Peterson and Seligman (this volume) have pro-
vided interesting evidence. As indicated by lower performance scores, **global’’
subjects, not “‘specific’’ subjects, tended to show deficits on both the similar and
dissimilar test tasks.
Finally, the attribution patterns examined were ordered in their developmental
sequence. The first two types appear as early as in the preschool age; Type 3, at
elementary school age; and Type 4, after elementary school age. At what age
Type 5 first appears is an open question, as is the developmental sequence of the
decisive causal dimensions. Within attribution patterns Type | and 2, the pos-
sibility of direct compensation is employed for stable factors along the locus
dimension. To enable subsequent explanations of unexpected outcomes, Type 3
adds external-variable causal factors. Type 4 shifts attributions to the stability
dimension of internal factors and their inverse compensatory relation. When and
how globality is brought into play as an additional causal dimension and how it
relates to the other two dimensions are as yet not known; nor has any research
been done to answer these questions. It may be that globality combines with
stability rather than with the locus dimension. Stable causal factors tend to be
global, and variable factors tend to be specific, regardless of whether they may
be internal or external.
This is supported by evidence (Peterson & Seligman, this volume, Table 4)
concerning the intercorrelations between the three dimensions for children in 3rd
through 6th grades. For bad events, globality correlated with stability, but not
with locus (internality). Accordingly, the global causes attributed to bad events
could be both internal (e.g., low intelligence) and external (e.g., all task de-
mands were too high), but only the first of these alternatives could be subsumed
under DAP, Types | and 2. For good events, globality correlated with inter-
 5. CAUSAL ATTRIBUTION PATTERNS 175

nality, which could be in accord with PAP. To show more accurately how the
patterns underlying these attributional tendencies are related to each other, the
causal factors but not the causality dimensions ranked in a double classification
(according to stability and internality) would have to be correlated with globality
versus specificity. Moreover, the correlations reported are rather low. This was
to be expected because they relate to a sample in which subgroups representing
PAP and DAP attributional tendencies (whose magnitude is as yet unknown)
may be mixed to an unknown degree. It would be of great import to determine
the causal factor of ability in terms of the degree of its globality.

First Appearance and Development of Individual Differences. Nothing is

known about the emergence and development of the globality dimension or about
how it is used for causal attributions. One can only speculate about whether the
emergent concept of ability will, in the course of development, be attributed
more to global or to specific causal factors, or whether it will be differentiated
into global-specific levels of an hierarchical order. Most likely, the latter is the
case. Such differentiation at least would furnish optimal cognitive prerequisites
for PAP and DAP Type 5 combinations of attribution patterns.
Similarly, the development of individual differences in terms of PAP and
DAP Type 5 remains a puzzle. As mentioned earlier, Peterson and Seligman
found that depressive children in 3rd through 6th grades were apt to be found in
families in which the mother (but not the father) was also depressed and that
DAP for bad events (internal, stable, global) correlated for mothers and their
children. Unfortunately, no separate correlations for the dimension of globality
were reported. At what developmental stage in the first ten years of her child’s
life a mother becomes ‘“‘depressive,’’ or ‘‘depressogenous,”’ (i.e., she fosters
DAP in her child) and how she does this, are questions to be answered by future
research. However, the influence of school, especially elementary school,
should also be taken into account. It is very likely that experience at school leads
poor students to believe that their lack of ability, and the level of difficulty of the
demands learning makes on them, are attributable to global rather than to specific
factors.

CONCLUDING REMARKS

I have demarcated five different attribution patterns, each of which, after success
or failure, enables favorable (PAP) or unfavorable (DAP) inferences about one’s
own ability. Each of the five types of attribution patterns specified makes differ-
ent demands on cognitive development. The age at which these attribution pat-
terns appear for the first time varies and extends over a long period, ranging from
the preschool period into the second decade. By specifying the “‘critical’’ age
periods for the appearance of each type of attribution pattern, assumptions could
176 HECKHAUSEN

be made about the developmental factors that foster the emergence of individual
differences in terms of PAP and DAP. However, it is still open to question at
what age, based on the concept of global competence, PAP and DAP Type 5 first
appears, and when individual differences emerge.
The taxonomy of attribution patterns presented here has been elaborated from
a functional perspective. That is: After success or failure, in what different ways
can the image of own ability be seen in a favorable or an unfavorable light? This
taxonomy need not be used only for descriptive classification. We have at-
tempted to enanchor the particular types of attribution patterns in their develop-
mental progression. Because differential relations between these types and cer-
tain personality variables may be expected, the taxonomy can become a useful
instrument for empirical research. This is true is for the emergence of the particu-
lar attribution patterns, the combinations between different types in their devel-
opmental progression, the differential effects generated by them.
The analysis raised a number of questions that require further clarification.
The first question relates to the developmental sequence of the patterns of PAP
and DAP. In other words, does acquisition of PAP and DAP Type | predeter-
mine the successive order in which subsequent attribution patterns are acquired?
That is, will acquisition of Type | PAP and DAP lead to the acquisition of the
later types, so that PAP and DAP become progressively more radical? Or alter-
natively, will the progression of these attribution patterns produce various con-
tellations of their elements, so that total strategies that verge on the psycho-
pathological (at least in the case of DAP) will be developed only in exceptional
cases in the second decade? On the other hand, is the answer perhaps that
individuals do not, in fact, acquire any sharply defined attribution patterns but,
instead, only come to develop a strong or weak tendency toward either PAP or
DAP, a tendency which in given situations leads them to make use of that type of
the five attributions patterns specified (or of combinations between them) which
seems the most appropriate?
These questions can be answered only by longitudinal studies combined with
experimental variations of conditions. Such variations should not be confined
only to success and failure. It would in addition be possible to reduce the
attribution scope for particular causal factors. For such a purpose, the three
covariation dimensions of Kelley’s (1967) cube would be a suitable instrument.
Entity information can narrow the causal factors down to globality or specificity;
consensus information (which will be integrated only by older children; [e.g.
Heckhausen, 1982]), can reduce task difficulty only to high or low degree;
consistency information can imply only high or low ability or luck. This method,
if developed, would show whether or not I will have to rethink my admittedly,
rash tendency to consider the particular types of attribution patterns, once they
have developed, as personality-specific characteristics. Instead, more credit
given to situation-specific effects.
 5. CAUSAL ATTRIBUTION PATTERNS 177

There is an additional important issue; namely, the different consequences of

each type of attribution pattern. These consequences have been examined pri-
marily in terms of self-evaluation, a self-serving versus a self-derogative bias in
looking at one’s competence. No doubt, in view of its behavioral effects, this is a
central aspect but depending on the causality dimension in which they are based,
the various types of attribution patterns, also generate specific effects. The more
weight is given to internal causes, the greater the affect on self-esteem; the more
weight is given to stable causes, the more constant the ensuing success expectan-
cy and the longer will action tendencies (or nonaction tendencies) determined by
it persist. Finally, the more weight given to global causes, the more their effects
will be generalized across various subsequent task demands and task situations.
Concerning the various attribution patterns, the following points should be
considered. The two types of PAP and DAP that emerge earliest set ability and
effort within a preconceived relationship of direct compensation. The decisive
dimension for this is locus of control its effects self-evaluation in terms of PAP or
DAP. As both these causal factors are stable, success expectancy is also deter-
mined and stablized. In the case of PAP, success expectancy is positive, increas-
ing the intensity and persistence of efforts; in the case of DAP, it is negative,
lowering effort and persistence. PAP and DAP Type 3 make it possible to play
down unexpected or undesired outcomes by attributions to luck, so that self-
evaluation and success expectancy need not be modified. PAP and DAP Type 4
play their inverse-compensatory arrangements between stable ability and vari-
able effort. As a result if high ability is inferred after success and low effort is
inferred after failure (PAP), success expectations can be increased or maintained.
On the other hand, if high expenditure of effort is the inference made following
success and low ability the inference made following failure (DAP) expectations
can be retained unmodified or be reduced. These are both internal factors; al-
though ability generates more affects than does effort, Type 4 also serves to
immunize ability-related self-esteem against negative feedback. Finally, PAP
and DAP Type 5 will produce either a wider generalization of attributions to
other situations and task demands or will limit such generalization. This will
have an equal effect on behavior influenced by self-esteem and by success
expectancy (as well as on expectancy-related action parameters, such as
persistence).
Making the variety of effects of attribution patterns explicit may ultimately
make its payoff in the design of individualized intervention programs for motive
change (e.g., Krug & Heckhausen, 1982).
The types of effects outlined make it clear that in this context the conse-
quences of self-evaluation are crucial, and that these consequences are the unify-
ing thread winding its way through all types of attribution patterns and their
motivating effects. However, it is also conceivable that, at least in some cases, in
order to manipulate one’s own expenditure of effort and persistence, especially
 178 HECKHAUSEN

within attribution patterns Type 3 and 4, the prime factor is a volitional arrange-
ment of success expectancies. A certain degree of self-manipulated success
expectancy, rather than self-evaluation, may play a central role in attribution
biasis, this is because a given activity may be motivated by incentives of the
anticipated other consequences, such as evaluation by other persons, extrinsic
side effects, or superordinate goals (Heckhausen, 1977).
Finally, this discussion of the possible determinants of attribution patterns
returns to the question asked at the beginning: Why do a good many people seem
intent on preserving the depressive image they have of self and own abilities and
do not opt for more sanguine expectations, even when good opportunities are
offered? In my examination of the various types of attribution patterns, I made
the assumption that DAP drifts into self-derogative frames of mind with much
the same bias and lack of realism that PAP does. Perhaps I have become a victim
of the aesthetics of symmetric constructions. Possibly, as Alloy and Abramson
(1979) have put it, in reality, DAP people are ‘‘sadder but wiser’’; and possibly,
in spite of a proneness for DAP, more than PAP people, who keep so cheerful
and successful by looking at the world through rose-coloured glasses, they still
retain a capacity for self-criticism and for realism. This again is a question that
future research will have to clarify.

POSTSCRIPT

Since writing this, Kratz (1985), in her doctoral dissertation, examined early
positive and depressive attribution patterns. She compared first-graders (median
6:11) with third-graders (median 9:1) because it had proved difficult to isolate
consistent attribution patterns in preschoolers.
After they had first received success and later failure feedback on the task
(identifying missing parts of pictures), the children were first asked to give free
attributional causes for these respective outcomes. In neither case did they expe-
rience any difficulty in coming up with plausible causes. The 7-year-olds ex-
plained success predominantly with internal causes, and failure with internal as
well as external causes. The 9-year-olds demonstrated the self-serving attribution
asymmetry well known among adults. They predominantly attributed success to
high level of ability and failure to high task difficulty. However, in a forced-
choice situation, forcing a decision between attribution to difficulty vs. ability,
task difficulty was the preferred cause for both success and failure.
An attempt was then made to seek out individual differences related to PAP
and DAP. After the children had stated what they thought was the main cause for
the outcome, they were asked about the importance of the factor they had
not
named. After two success trials and two failure trials most of the children in both
age groups responded in terms of a multiple necessary causal scheme, i.e.
the
second factor also contributed to the respective outcome, be it success or failure.
 5. CAUSAL ATTRIBUTION PATTERNS 179

We refer to this as the ‘‘normal attribution pattern’? (NAP). The remaining

children were labeled DAP or PAP children, if they displayed the respective
response pattern on at least two of the four trials, and their responses in the
remaining trials related to multiple necessary causes. Whereas among the 7-year-
olds only 24% were PAP children and only 12% were DAP children, the ratio for
the 9-year-olds was reversed, 34% DAP vs. 17% PAP children. This suggests
that the school experience had brought about a developmental change in favor of
a depressive attribution pattern.
With the aid of further correlates, involving a different task (Matching Famil-
iar Figures, Kagan et al., 1964), it was also possible to establish the charac-
teristics of both subgroups’ attribution patterns apart from NAP. In order to first
insure that children were able to understand normatively defined differences in
task difficulty, they were shown a row of the same digit, representing the age of
children one year older, and some of these digits were crossed out to indicate that
these children had failed the task. The success vs. failure ratio of this reference
group could thus be varied. All 7-year-olds already possessed this capacity. If the
children were presented with normatively defined, differing levels of difficulty,
PAP children believed that they could solve more difficult problems than DAP
children, with NAP children falling in between. In other words, PAP children
had a self-concept of higher competence than DAP children. There was a corre-
sponding difference between the two groups in their confidence of success on a
task that they had been told was of intermediate difficulty for children who were
one year older. PAP children believed that they could solve problems at this
demanding difficulty level. But that was not true for the DAP children. They
already anticipated failure and capitulated when they were told that 70% of the
one-year-older reference group had succeeded on the task.
Kratz also obtained scales for the three causes (a) difficulty, (b) ability, and
(c) effort. The children were asked to mark one of five levels represented as
circles with increasing radii. Among the 7-year-olds there was no significant
difference between the PAP and DAP children (an indication of the expected
differences appeared only in the ability scaling). In contrast, among the 9-year-
olds, the PAP children assessed their ability after success higher than the DAP
children (again the NAP children fell in between).
Noteworthy is the stability of the DAP children’s self-concept of low ability in
both age groups (self-concept of ability being defined as confidence to succeed at
a task of normatively defined difficulty, as described previously). Whereas PAP
and NAP children assessed their ability as higher after success than after failure,
the DAP children scaled their ability no higher after success than after failure.
This high stability of the ability concept was clearly not due to a more advanced
developmental stage of the DAP children’s causal schemata of ability or effort
compensation. This was established with a physical effort task in which nearly
all the children were not yet able to comprehend that insufficient effort would be
compensated by a higher ability (ability compensation) and lower ability by
 180 HECKHAUSEN

greater effort (effort compensation). This forces one to conclude that a self-
depreciating attribution pattern is present prior to the development of a mature
ability concept—for example, that lack of ability can be compensated by in-
creased effort. However, even the PAP children were not arrogant in the assess-
ment of their ability. They did not believe that they could solve a task at which
70% of the one-year-older reference group had supposedly failed.
Moreover, it was not possible to attribute either the self-concept of lack of
ability or DAP to a lower stage of cognitive development. A test battery for
assessing level of cognitive operations in the Piagetian sense, did not differenti-
ate between DAP and PAP children.
Finally, an attempt was made to determine whether these children use the
Type 2 of PAP and DAP, i.e., whether they augment (PAP) or discount (DAP)
the difficulty of a task after success or failure, respectively. For this purpose,
children were asked to assess the difficulties of various tasks in the context of
normative cues (contrasting representations of two groups of children who were
or were not able to solve the particular version of a task). Subsequently, the
experimenter suggested that the child work on the intermediate difficulty task.
Each child experienced a success and a failure series. After each series the child
was asked to indicate the difficulty of the tasks for 10 same-aged children by
indicating the proportion of children who passed and failed on the task. (They
were given a set of cards each showing the same face on both sides. One of the
faces was crossed out to indicate failure.) Although the children’s judgment of
the reference group’s lesser or greater difficulty depended on their own experi-
ence of success and failure, respectively, there was no general augmenting or
discounting the difficulty in line with PAP and DAP of Type 2. However,
experimental short-comings cannot be ruled out completely.
Overall, it can be stated that individual differences. in attribution patterns of
the first type of PAP and DAP were already found among early grade-schoolers
at ages 7 and 9. The more frequent occurrence among PAP in first-graders
reverses in favor of DAP among third-graders. Compared with DAP, PAP was
associated with a higher self-evaluation of ability after solving a problem and
with greater expectation of success prior to tackling another problem. DAP
children had a more stable self concept of (lower) ability than PAP children,
which remained largely independent of antecedent success or failure. There was
no difference in the cognitive developmental stage of the two groups of children.
It was not possible to demonstrate an augmentation of task difficulty after suc-
cess or failure, among the PAP children.

REFERENCES
Abramson, L. Y., & Sackeim, H. A. (1977). A paradox in depression
: Uncontrollability and self-
blame. Psychological Bulletin, 84, 839-851.
Abramson, L. Y., Seligman, M. E. P., & Teasdale, J. D. (1978).
Learned helplessness in humans:
Critique and reformulation. Journal of Abnormal Psychology, 87,
49-74.
 5. CAUSAL ATTRIBUTION PATTERNS 181

Alloy, L. B., & Abramson, L. Y. (1979). Judgment of contingency in depressed and nondepressed
students: Sadder but wiser? Journal of Experimental Psychology: General, 108, 441-485.
Ames, C. (1978). Children’s achievement attributions and self-reinforcement: Effects of self-con-
cept and competitive reward structure. Journal of Educational Psychology, 70, 345-355.
Arkin, R. M., Appelman, A. J., & Burger, J. M. (1980). Social anxiety, self-presentation, and the
self-serving bias in causal attribution. Journal of Personality and Social Psychology, 38, 23-35.
Beck, A. T., Ward, C. H., Mendelson, M., Mock, J., & Erbaugh, J. (1961). An inventory for
measuring depression. Archives of General Psychiatry, 4, 561-571.
Berglas, S., & Jones, E. E. (1978). Drug choice as an internalization strategy in response to
noncontingent success. Journal of Personality and Social Psychology, 36, 405—417.
Blaney, P. H. (1977). Contemporary theories of depression: Critique and comparison. Journal of
Abnormal Psychology, 86, 203-223.
Bradley, G. W. (1978). Self-cerving biasis in the attribution process: A reexamination of the fact or
fiction question. Journal of Personality and Social Psychology, 36, 56-71.
Cohen, S., Rothbart, M., & Phillips, S. (1976). Locus of control and the generality of learned
helplessness in humans. Journal of Personality and Social Psychology, 34, 1049-1056.
Crandall, V. C. (1969). Sex differences in expectancy of intellectual and academic reinforcement.
In C. P. Smith (Ed.), Achievement-related motives in children. New York: Russell Sage.
Crandall, V. C., Katkovsky, W., & Crandall, V. J. (1965). Children’s beliefs in their own control
of reinforcements in intellectual-academic achievement situations. Child Development, 36, 91—
109.
Deaux, K., & Emswiller, T. (1974). Explanations of successful performance on sex-linked tasks:
What’s skill for the male is luck for the female. Journal of Personality and Social Psychology,
29, 80-85.
Deaux, K., & Farris, E. (1977). Attributing causes for one’s performance: The effects of sex,
norms, and outcome. Journal of Research in Personality, 11, 59-72.
Diener, C. I., & Dweck, C. S. (1978). An analysis of learned helplessness: Continuous changes in
performance, strategy, and achievement cognitions following failure. Journal of Personality and
Social Psychology, 36, 451-462.
Dweck, C. S. (1975). The role of expectations and attributions in the alleviation of learned help-
lessness. Journal of Personality and Social Psychology, 31, 674-685.
Dweck, C. S., & Bush, E. S. (1976). Sex differences in learned helplessness: 1. Differential
debilitation with peer and adult evaluators. Developmental Psychology, 12, 147-156.
Dweck, C. S., & Gilliard, D. (1975).Expectancy statements as determinants of reactions to failure:
Sex differences in persistence and expectancy change. Journal of Personality and Social Psychol-
ogy, 32, 1077-1084.
Dweck, C. S., & Repucci, N. D. (1973). Learned helplessness and reinforcement responsibility in
children. Journal of Personality and Social Psychology, 25, 109-116.
Falbo, T. (1975). Achievement attributions of kindergarteners. Development Psychology, 11, 529-
530.
Feather, N. T., & Simon, J. G. (1975). Reactions to male and female success and failure in sex-
linked occupations: Impressions of personality, causal attributions, and perceived likelihood of
different consequences. Journal of Personality and Social Psychology, 31, 20-31.
Fitch, G. (1970). Effects of self-esteem, perceived performance, and choice on causal attributions.
Journal of Personality and Social Psychology, 16, 311-315.
Frankel, A., & Snyder, M. L. (1978). Poor performance following unsolvable problems: Learned
helplessness or egotism? Journal of Personality and Social Psychology, 36, 1415-1423.
Gilmor, T. M., & Minton, H. L (1974). Internal versus external attribution of task performance as a
function of locus of control, initial confidence and success - failure outcome. Journal of Person-
ality, 42, 159-174.
Griffith, M. (1977). Effects of noncontingent success and failure on mood and performance. Jour-
nal of Personality, 45, 442—457.
 182 HECKHAUSEN

Halisch, F. (1983). VorbildeinfluB und Motivationsprozesse. Psychologisches Institut der Ruhr-

Universität Bochum.
Hammen, C. L., & Krantz, S. (1976). Effects of success and failure on depressive cognitions.
Journal of Abnormal Psychology, 85, 577-586.
Hanusa, B. H., & Schulz, R. (1977). Attributional mediators of learned helplessness. Journal of
Personality and Social Psychology, 35, 602-611.
Heckhausen, H. (1963). Hoffnung und Furcht in der Leistungsmotivation. Meisenheim: Hain.
Heckhausen, H. (1972). Die Interaktion der Sozialisationsvariablen in der Genese des
Leistungsmotivs. In C. F. Graumann (Ed.), Handbuch der Psychologie (Vol. 7/2, pp. 955-
1019). Göttingen: Hogrefe.
Heckhausen, H. (1975). Fear of failure as a self-reinforcing motive system. In I. G. Sarason & C.
Spielberger (Eds.), Stress and anxiety (Vol. 2, pp. 117-128). Washington, DC: Hemisphere.
Heckhausen, H. (1977). Achievement motivation and its constructs: A cognitive model. Motivation
and Emotion, 1, 283-329.
Heckhausen, H. (1978). Selbstbewertung nach erwartungswidrigem Leistungsverlauf: Einfluß von
Motiv, Kausalattribution und Zielsetzung. Zeitschrift für Entwicklungspsychologie und
Pädagogische Psychologie, 10, 191-216.
Heckhausen, H. (1980). Motivation und Handeln. Berlin: Springer.
Heckhausen, H. (1980b). Preschoolers’ risk-taking: A paradox resolved? Unpublished Manuscript.
Psychologisches Institut der Ruhr-Universität Bochum.
Heckhausen, H. (1981). Task-irrelevant cognitions during an exam: Incidence and effects. InH. W.
Krohne & L. Laux (Eds.), Achievement, stress, and anxiety. Washington, DC: Hemisphere.
Heckhausen, H. (1982). The development of achievement motivation. In W. W. Hartup (Ed.),
Review of Child Development Research (Vol. 6, pp. 600-668). Chicago: University of Chicago
Press.
Heckhausen, H. (1983). Entwicklungsschritte in der Kausalattribution von Handlungsergebnissen.
In D. Görlitz (Ed.), Kindliche Erklärungsmuster. Entwicklungspsychologische Beiträge zur
At-
tributionsforschung (Vol. 1, pp. 49-85). Weinheim: Beltz, 1983.
Heckhausen, H., & Krug, S. (1982). Motive modification. In A. J. Stewart (Ed.), Motivation and
Society. San Francisco, CA: Jossey Bass.
Hiroto, D. S. (1974). Locus of control and learned helplessness. Journal of
Experimental Psychol-
ogy, 102, 187-193.
Jopt, U.-J., & Ermshaus, W. (1977). Wie generalisiert ist das Selbstkonzept eigener
Fahigkeit? Eine
motivationspychologische Untersuchung zur Aufgabenabhängigkeit der Fähigkeits
wahrnehmung.
Zeitschriffür
t experimentelle und angewandte Psychologie, 24, 578-601.
Kagan, J., Rosman, B. L., Day, D., Albert, J., & Phillips, W. (1964).
Information processing in
the child: Significance of analytic and reflective attitudes, Psychological Monographs
, 78, (1,
Whole No. 578).
Karniol, R., & Ross, M. (1976). The development of causal attributions
in social perception.
Journal of Personality and Social Psychology, 34, 455-464.
Kelley, H. H. (1967). Attribution theory in social psychology.
In D. Levine (Ed.), Nebraska
Symposium on Motivation (pp. 192-238). Lincoln: University of Nebraska
Press.
Kelley, H. H. (1972). Causal schemata and the attribution process.
New York: General Learning
Press.
Kelley, H. H. (1973). The process of causal attribution.
American Psychologist, 28, 107-128.
Klein, D. C., Fencil-Morse, E., & Seligman, M. E. P. (1976).
Learned helplessness, depression,
and the attribution of failure. Journal of Personality and
Social Psychology, 33, 508-516.
Klinger, E. (1975). Consequences of commitment to and
disengagement from incentives. Psycho-
logical Review, 82, 1-25.
Kratz, H. (1985). Kausalattributionen und Attributionsmust
er fiir eigene Handlungsergebnisse im
Grundschulalter. Dissertation, Fakultät für Psychologie,
Ruhr-Universität Bochum.
 5. CAUSAL ATTRIBUTION PATTERNS 183

Krug, B. (1972). Mißerfolgsattribuierung und deren Auswirkungen auf Erwartungs- und

Leistungänderungen sowie auf Persistenz. Unpublished manuscript, Psychologisches Institut der
Ruhr-Universität Bochum.
Krug, S., & Hanel, J. (1976). Motivänderung: Erprobung eines theoriegeleiteten Train-
ingsprogrammes. Zeitschrift für Entwicklungspsychologie und Pädagogische Psychologie, 8,
274-286.
Krug, S., & Heckhausen, H. (1982). Motivförderung in der Schule. In F. Rheinberg (Ed.),
Bezugsnormen zur Schulleistungsbewertung: Analyse und Intervention: Jahrbuch 1982 für Em-
pirische Erziehungswissenschaft (pp. 65-114). Düsseldorf: Schwann.
Kuhl, J. (1981). Motivational and functional helplessness: The moderating effect of state versus
action orientation. Journal of Personality and Social Psychology, 40, 155-170.
Kun, A. (1977). Development of the magnitude-covariation and compensation schemata in ability
and effort attributions of performance. Child Development, 48, 862-873.
Lefcourt, H. M., Hogg, E., Struthers, S., & Holmes, C. (1975). Causal attributions as a function of
locus of control, initial confidence, and performance outcomes. Journal of Personality and Social
Psychology, 32, 391-397.
Luginbuhl, J. E. R., Crowe, D. H., & Dahan, J. P. (1975). Causal attribution for success and
failure. Journal of Personality and Social Psychology, 31, 86-93.
McMahan, I. D. (1973). Relationships between causal attributions and expectancy of success.
Journal of Personality and Social Psychology, 28, 108-114.
Meyer, W.-U. (1973). Leistungsmotiv und Ursachenerklärung von Erfolg und Mißerfolg. Stuttgart:
Klett.
Meyer, W.-U., & Plöger, F.-O. (1979). Scheinbar paradoxe Wirkungen von Lob und Tadel auf die
wahrgenommene eigene Begabung. In H. Filipp (Ed.), Selbstkonzept-Forschung: Probleme,
Befunde und Perspektiven (Pp. 221-235). Stuttgart: Klett-Cotta.
Miller, D. T. (1976). Ego involvement and attribution for success and failure. Journal of Person-
ality and Social Pscyhology, 34, 901-906.
Miller, D. T., & Ross, M. (1975). Self-serving biases in the attribution of causality: Fact or fiction?
Psychological Bulletin, 82, 213-225.
Miller, I. W. III, & Norman, W.H. (1979). Learned helplessness in humans: A review and attribu-
tion-theory model. Psychological Bulletin, 86, 93-118.
Miller, W. R., & Seligman, M. E. P. (1975). Depression and learned helplessness in man. Journal
of Abnormal Psychology, 84, 228-238.
Nicholls, J. G. (1975). Causal attributions and other achievement-related cognitions: Effects of
task outcome, attainment value, and sex. Journal of Personality and Social Psychology, 31, 379-
389.
Nicholls, J. G. (1976). Effort is virtuous, but it’s better to have ability: Evaluative responses to
perceptions of effort and ability. Journal in Personality Research, 10, 306-315.
Nicholls, J. G. (1978). The development of the concepts of effort and ability, perception of academ-
ic attainment, and the understanding that difficult tasks require more ability. Child Development,
49, 800-814.
Nicholls, J. G. (1979). Development of perception of own attainment and causal attributions for
success and failure in reading. Journal of Educational Psychology, 71, 94-99.
Parsons, J. E., & Ruble, D. N. (1977). The development of achievement-related expectancies.
Child Development, 48, 1075-1079.
Rheinberg, F. (1980). Leistungsbewertung und Lernmotivation. Göttingen: Hogrefe.
Rholes, W. S., Blackwell, J., Jordan, C., & Walters, C. (1980). A developmental study of learned
helplessness. Developmental Psychology, 16, 616-624.
Rizley, R. (1978). Depression and distortion in the attribution of causality. Journal of Abnormal
Psychology, 87, 32-48.
Ross, L., & Lepper, M. R. (1980). The perseverance of beliefs: Empirical and normative consid-
 184 HECKHAUSEN

erations. In R. A. Shweder & D. Fiske (Eds.), New directions for methodology of behavioral
science: Fallible judgment in behavioral research. San Francisco: Jossey-Bass.
Roth, S. (1980). A revised model of learned helplessness in humans. Journal of Personality, 48,
103-133.
Rotter, J. B. (1954). Social learning and clinical psychology. Englewood Cliffs, NJ: Prentice-Hall.
Rotter, J. B. (1966). Generalized expectancies for internal versus external control of reinforcement.
Psychological monographs, 80, (1, Whole No. 609), 1-28.
Schlenker, B. R. (1975). Self-presentation: Managing the impression of consistency when reality
interferes with self-enhancement. Journal of Personality and Social Psychology, 32, 1030-037.
Schmalt, H.-D. (1976). Leistungsmotivation und kognitive Zwischenprozesse im Erleben von Erfolg
und Mißerfolg. Unpublished manuscript, Psychologisches Institut der Ruhr-Universität Bochum.
Schneider, K. (1977). Leistungsmotive, Kausalerklarungen für Erfolg und Mißerfolg und erlebte
Affekte nach Erfolg und Mißerfolg. Zeitschrift für Experimentelle und Angewandte Psychologie,
24, 613-637.
Seligman, M. E. P. (1975). Helplessness: On depression, development, and death. San Francisco:
Freeman.
Seligman, M. E. P., Abramson, L. Y., Semmel, A., & von Baeyer, C. (1979). Depressive attribu-
tional style. Journal of Abnormal Psychology, 88, 242-247.
Shultz, T. R., & Butkowsky, I. (1977). Young children’s use of the scheme for multiple sufficient
causes in the attribution of real and hypothetical behavior. Child Development, 48, 464-469.
Snyder, M. L., Smoller, B., Strenta, A., & Frankel, A. (1981). A comparison of egotism, nega-
tivity, and learned helplessness as explanations for poor performance after unsolvable problems.
Journal of Personality and Social Psychology, 40, 24-30.
Snyder, M. L., Stephan, W. G., & Rosenfield, D. (1976). Egotism and attribution. Journal of
Personality and Social Psychology, 33, 435-441.
Stevens, L., & Jones, E. E. (1976). Defensive attribution and the Kelley cube. Journal of Person-
ality and Social Psychology, 34, 809-820.
Surber, C. F. (1980). The development of reversible operations in judgments of ability, effort, and
performance. Child Development, 51, 1018-1029.
Tennen, H., & Eller, S. J. (1977). Attributional components of learned helplessness and facilita-
tion. Journal of Personality and Social Psychology, 35, 265-271.
Tweer, R. (1976). Das ökonomieprinzip in der anstrengungskalkulation: Eine entwicklungspsychol-
ogische untersuchung. Unpublished manuscript, Psychologisches Institut der Ruhr-Universität
Bochum.
Weiner, B., & Kukla, A. (1970). An attributional analysis of achievement motivation. Journal of
Personality and Social Psychology, 15, 1-20.
Weiner, B., & Potepan, P. A. (1970). Personality characteristics and affective reactions toward
exams of superior and failing college students. Journal of Educational Psychology, 61, 144-151.
Wener, A. E., & Rehm, L. P. (1975). Depressive affect: A test of behavioral hypotheses. Journal
of Abnormal Psychology, 84, 221-227.
Wortman, C. B., Costanzo, P. R., & Witt, T. R. (1973). Effect of anticipated performance on the
attributions of causality to self and others. Journal of Personality and Social Psychology,
27,
372-381.
Zuckerman, M., & Gerbasi, K. C. (1977). Belief in internal control or belief in a just world: The
use and misuse of the I-E scale in prediction of attitudes and behavior. Journal of Personality, 45,
356-378.
 Helplessness and Attributional
Style in Depression

Christopher Peterson
Virginia Polytechnic Institute and State University

Martin E. P. Seligman
University of Pennsylvania

According to the reformulation of learned helplessness theory, a habitual insid-

ious attributional style is a risk factor for depression. The thinking leading up to
the reformulation is discussed and some recent research in support of it is de-
scribed. Finally, some potential points of departure for subsequent investigations
of attributional style and depression are outlined.

HELPLESSNESS AND ATTRIBUTIONAL STYLE IN

DEPRESSION

The learned helplessness model has recently been reformulated along attribu-
tional lines (Abramson, Seligman, & Teasdale, 1978; see also Miller & Norman,
1979; and Roth, 1980). According to this reformulation, uncontrollable negative
events do not automatically lead to pervasive deficits in learning, motivation,
and emotion. Instead, the individual’s causal attributions, concerning the uncon-
trollable events, govern the breadth and duration of these events, and the esteem
related nature of the deficits. Only in the case of certain attributions will wide-
spread helpless and depression result from uncontrollable events.
The reformulation was prompted by a growing number of laboratory studies.
These studies demonstrated that helplessness effects were not invariably pro-
duced following uncontrollable events, and that the magnitude of helplessness
effects, when they did occur, varied with factors not mentioned in the original
theory. The original helplessness theory accorded importance only to uncon-
trollability (see Abramson, et al., 1978, and Miller & Seligman, 1980, for
reviews of these studies). The reformulated model proposes that attributions

185
 186 PETERSON AND SELIGMAN

subsume the most important variables neglected by the original learned help-
lessness theory. The reformulation encompasses individual differences in attribu-
tional style; attributional style has been conceived as a risk factor for depression
(Seligman, Abramson, Semmel, & von Baeyer, 1979). The purpose of this
chapter is to review some of the new empirical studies generated by the refor-
mulation and, in light of these findings, to assess the reasonableness of the
reformulated model.

ORIGINAL HELPLESSNESS THEORY

A brief statement of the original helplessness theory, before its reformulation, is

presented (see also Alloy & Seligman, 1979; Maier & Jackson, 1979: Maier &
Seligman, 1976; Maier, Seligman & Solomon, 1969: Seligman, 1975; Seligman,
Maier, & Solomon, 1971). Overmier and Seligman (1967) and Seligman and
Maier (1967) observed that dogs given inescapable electric shock in a Pavlovian
hammock showed marked difficulty in escaping shock in a shuttlebox, 24 hours
later; indeed, two thirds of the animals tested failed to learn the response. This
failure to learn a response, readily acquired by dogs who had not experienced
previous shock, was accompanied by other surprising behavior. The helpless
dogs were extremely passive, sitting and enduring the shock without moving.
The occasional escape responses they did make were not reliably acquired, and
some dogs who escaped once or twice never escaped again.
It was proposed that the behavior of the animals resulted from their learning,
in the Pavlovian hammock, that their responses and outcomes (i.e., shocks)
were
independent of each other, and that nothing they did mattered. This learning
was
represented as an expectation of helplessness, generalized to the new
situation.
This expectation constituted the basic explanation of the phenomenon.
The ex-
planation, along with the phenomenon, was dubbed ‘‘learned
helplessness.”’
The expectation of helplessness was said to produce several deficits:
(a) the
animals were less likely to initiate responses, and hence, they
did not sample the
fact that they had control in the new situation (motivational
deficit); (b) if they
did respond, the animals were less likely to perceive that response
s and outcomes
were related and hence, were slower to learn what to do
in the new situation
(associative or cognitive deficit); (c) the animals also seemed
to show blunted
affect (emotional deficit). New work by Maier and Jackson
(1979) suggests that
helplessness involves an analgesic effect as well as
an associative deficit: thus,
the helpless animals were less motivated to escape an aversive
event because it
did not hurt them as much as it hurt nonhelpless animals.
A number of experiments (see Maier & Seligman,
1976) strongly suggest that
it is the uncontrollability of the aversive event and
not its traumatizing properties
that are responsible for the various helplessness effects
. Central to these demon-
 6. ATTRIBUTIONAL STYLE IN DEPRESSION 187

strations is the triadic design. In the correctly done helplessness experiment,

three groups are used: one group (the escapable group) is exposed to controllable
events, such as electric shocks terminable by some response; a second (the
inescapable group), is yoked to the first group: given the identical pattern of
physical events; crucially different, however, the offsets of the events are inde-
pendent of all responses; a third group (the no-pretreatment control) gets no
programmed events, controllable or uncontrollable. All three groups are later
tested on a task that involves the learning of some new response. Helplessness is
demonstrated when the inescapable group has more trouble than the escapable
and control groups. Learned helplessness theory for animals has been attacked by
theorists inclined to explain the phenomenon in peripheral or chemical terms;
many of the studies conducted by learned helplessness proponents have been
attempts to defend the cognitive account against these reinterpretations (for a
review of this controversy, see Maier & Jackson, 1979; Maier & Seligman,
1976, and Seligman, Weiss, Weinraub, & Schulman, 1980).
The checkered success of learned helplessness theory (as applied to animals)
in maintaining its cognitive emphasis attracted the attention of psychologists
interested in helplessness in people. Human experimental psychology was begin-
ning its cognitive revolution, when dog helplessness was first described (Neisser,
1967). The learned helplessness theory was seen as an instance of this perspec-
tive, which held that overt behavior was best understood by taking into account
thoughts, beliefs, and strategies. If a cognitive account of the behavior of dogs
(and mice and rats; Seligman, 1975) was viable, then surely a cognitive account
of human helplessness should be more viable.
The first studies of learned helplessness in humans were simply attempts to
demonstrate learning impairment following experience with uncontrollable
events (for a review of these early studies, see Wortman & Brehm, 1975). Some
of these studies were methodologically flawed, or unsuccessful in demonstrating
the predicted impairment; however, a number were successful (e.g., Hiroto &
Seligman, 1975). Most studies borrowed the triadic design from the animal
research. The controllable events were typically loud noises or shocks, escapable
by pushing a button, or solvable concept identification problems. The uncon-
trollable events were inescapable noises or shocks, or unsolvable concept identi-
fication problems.
These early demonstrations were followed by two distinct lines of work. In
the first, researchers interested in the application of the phenomenon accepted it
and its explanation, usually uncritically, and used it as the basis for their discus-
sions of an incredible variety of significant failures of human adaptation. Best
known is its use as a model of depression (e.g., Seligman, 1974; Journal of
Abnormal Psychology, 1978). However, learned helplessness was also impli-
cated in drug use (Berglas & Jones, 1978), heart attacks (Krantz, Glass &
Snyder, 1974), failure in school (Dweck & Reppucci, 1973), sex differences
 188 PETERSON AND SELIGMAN

(Dweck & Bush, 1976), aggression (Geen, 1978), and IQ (Smith & Seligman,
1980). At this writing, the number of articles applying helplessness to human ills
runs in the hundreds.
There are several possible reasons why future historians of psychology will be
better equipped than present psychologists to understand the popularity of help-
lessness theory. The possible reasons include: helplessness theory’s basis in
learning theory struck a responsive chord among psychologists interested in the
relationship between the animal experimentation and the needs of people and
society; the contribution of helplessness theory’s optimism, borne from appar-
ently ready solutions to helplessness (i.e., exposure to contingencies); the theo-
ry’s simplicity, straight-forward and testable predictions, made it stand in flatter-
ing contrast to the muddier theories usually proposed to account for failures of
adaptation; and learned helplessness theory linked experimental psychology, the
past as represented by learning theory and the future as represented by cognitive
theory, to a wide range of more applied areas of psychology. Learned help-
lessness theory was and is of considerable interest to social, clinical, and devel-
opmental psychologists, who attempt to understand complex human behavior
and the ways it can go awry.
At the same time that the applications of helplessness occurred, another line
of work was being conducted. A number of researchers and theorists were
interested in the helplessness phenomenon itself, and their attempts to study it in
detail raised serious questions about the validity of learned helplessness in
peo-
ple. A catalogue of anomalies accrued in the research literature: some of
them
were enumerated by Abramson et al. (1978), Miller and Norman (1979),
and
Wortman and Brehm (1975). Among the claimed anomalies were:

1. The motivational deficit described in the animal literature was not

always
present in humans. Although helpless subjects did not learn quickly,
it was not
because they failed to initiate responses. Indeed, the typical human
helplessness
experiment required a great deal of Cooperation and instrumental
responding,
such as following complex instructions.
2. Some studies failed to replicate helplessness, and some found
a facilitating
effect of uncontrollable events: following unsolvable problem
s, subjects some-
times did better than comparison groups.
3. Laboratory-produced helplessness sometimes generalized
beyond the spe-
cific pretreatment situation, and sometimes did not.
For instance, for some
subjects, changing the experimental room, between the
pretreatment and test task
was found, to wipe out helplessness effects.
4. Norms relating how other people had performed and
information about the
nature of the task affected the helplessness obtained.
5. A belief in response-outcome independence did
not necessarily produce
helplessness. Indeed, some writers argued that subject
s did not always learn that
responses and outcomes were independent (Peters
on, 1978).
 6. ATTRIBUTIONAL STYLE IN DEPRESSION 189

6. Competing explanations, such as expectation incongruency (Douglas &

Anisman, 1975), superstitious responding (Peterson, 1978), and egotism (Fran-
kel & Snyder, 1978), were proposed and supported by data.

In short, the simple explanation of helplessness effects (i.e., uncontrollable

events — expectancy of response-outcome independence — interference with
learning) was not an account of the human laboratory helplessness. In contrast to
the criticisms of helplessness theory applied to animals, which took the model to
task for being too cognitive, the criticisms of the theory applied to people
charged that the theory was not cognitive enough! The critics of helplessness
theory applied to people called for a more complex and sophisticated account
(e.g., Wortman & Brehm, 1975). Most of these authors, in keeping with current
trends within social psychology, suggested an attributional reinterpretation of the
phenomenon.

THE REFORMULATION

In response to the anomalies, Abramson et al. (1978) proposed a reformulation

of learned helplessness. According to Abramson et al. (1978), more than an
expectation of response-outcome independence was necessary to account for
helplessness in the laboratory and depression outside of the laboratory. The
reformulation proposed that causal attributions were critical variables. Internal
attributions about negative events produce lowered self-esteem, and global and
stable attributions produce generalized helplessness, across situations and time.
Therefore, the reformulated model assumes that although the consequences of
doing so may be maladaptive, people act rationally, in accordance with their
attributions. The ‘‘man as scientist’? metaphor (Kelley, 1973; Kelly, 1955) of
attribution theory is embodied in the learned helplessness reformulation. Such
rationality is open to criticism, these criticisms are discussed as they apply to the
reformulation.
The reformulation accounted for most of the laboratory anomalies noted ear-
lier (see Abramson et al., 1978), and Miller & Seligman, 1980, for detailed
discussion of this point), as well as for several problems involved in applying
helplessness theory to depression. In passing, it is noted that if revisions in a
laboratory model of psychopathology, undertaken to improve the fit to laboratory
data, also clarify the psychopathology modelled, more confidence is justified in
the fit of the model to the pathology. This criterion for modelling may be added
to the more obvious ones of similar symptoms, causes, physiology, cures, and
prevention, discussed by Abramson and Seligman (1977).
One example of how the reformulation made the helplessness theory a better
explanation of depression has to do with bad events and depression. (“‘Negative
life events’’ are found to be jargon; therefore, hereafter they are called “bad
 190 PETERSON AND SELIGMAN

events.’’) Relationships have been demonstrated between bad events and depres-
sion (Brown & Harris, 1978; Paykel, 1974); however, they are not as robust as
the original helplessness theory suggested. The reformulated model explains
these weak relationships by saying that they do not reflect individual differences
in attributions about the causes of the bad events. Indeed, Semmel, Peterson,
Abramson, Metalsky, and Seligman (1980) demonstrated that the bad
event/depression link is increased if attributions are taken into account.
A second example of how the reformulated model allows better sense to be
made of what is known about depression is the depressive paradox (Abramson &
Sackeim, 1977; Janoff-Bulman, 1979; Peterson, 1979; Peterson, Schwartz, &
Seligman, 1981). Depressives tend to blame themselves about events over which
they feel no control; this is quite awkward for the original learned helplessness
theory to explain. However, it is quite easily accommodated by the reformulated
helplessness theory, which recognizes the difference between internal attribu-
tions (self-blame) and expectations of response-outcome independence
(helplessness).

ATTRIBUTIONAL STYLE

The investigations of the reformulation by our research group have focused

largely on individual differences in the tendencies to make internal, stable, and
global attributions. Although the reformulation was not presented exclusively in
terms of individual differences, such a strategy seemed necessary in order to
explain why some people become helpless following an experience with uncon-
trollable events and others do not. A reasonable assumption is that people have a
characteristic attributional style or a tendency to make similar sorts of attribu-
tions about a variety of events. Theoretically, a depressive attributional style is
one that habitually attributes bad events to internal, stable, and global causes.
Such an assumption does not minimize the role of situational evidence in deter-
mining what attributions people make; the vast social psychological literature
documents the importance of such evidence (e.g. Harvey, Ickes, & Kidd, 1976).
For instance, bad experiences are on the whole attributed to external causes,
while good experiences are on the whole attributed to internal causes (e.g.
Peterson, 1980; Weiner, 1974). One tends to explain one’s own behavior
by
external determinants; while one tends to explain the behavior of others
by
internal, dispositional determinants (e.g. Jones & Nisbett, 1971;
Monson &
Snyder, 1977), and so on. Given the constraints of situational
evidence, our
interest is in explaining why, in the same situation. people may
differ in their
helplessness and depression reactions. It is assumed, then, all
other determinants
of attributions being equal, that people have explanatory styles
that are applied to
the events they encounter.
 6. ATTRIBUTIONAL STYLE IN DEPRESSION 191

This assumption is in keeping with other cognitive accounts of individual

differences, such as theories of cognitive style (Goldstein & Blackman, 1978),
personal constructs (Kelley, 1955), locus of control (Rotter, 1966), negative
schemata (Beck, 1976), and so on. Put simply: It is assumed that people react to
events in accordance with their interpretaions; these interpretations are jointly
determined by the events themselves, and by the biases that people impose on the
events.
The frequent debates about whether personality characteristics operate across
situations usually do not focus on cognitive traits, such as attributional style.
Indeed, in light of evidence demonstrating that they have considerable cross-
situational consistency, some critics make an explicit exception for cognitive
traits (W. Mischel, 1968). Our assumption, by holding that the critical dif-
ferences are attributional (i.e., pertain to the causal inferences made by people),
differs from other cognitive accounts of individual differences. In contrast, other
accounts, focus on structural characteristics of cognition (Scott, Osgood, &
Peterson, 1979), on naive epistemologies other than causal ones (Kruglanski,
1980), and so on.
The research undertaken by our group has had several thrusts, which are now
explained. First, we were interested in whether there exists an attributional style
descriptive of depression. To this end, attributional style in several populations,
including depressed individuals, was assessed. It was hoped that a specific
attributional style would be found that: (a) characterized depressives as opposed
to nondepressives; (b) characterized different kinds of depressives; (c) varied
with severity of the depressive symptoms; and (d) did not characterize other
psychopathologies. Second, there was interest in determining if the depressive
attributional style caused depression. This question was attacked in two ways;
one strategy used causal modelling techniques to predict later depression from
earlier attributional style, the second strategy used laboratory experiments to
make the same inferences.

THE MEASUREMENT OF ATTRIBUTIONAL STYLE

Because the existence of attributional style was postulated, the first problem was
how it should be measured. A difficulty arose immediately: Are all causal at-
tributions accessible to a person’s awareness? When Heider (1958) first de-
scribed causal attributions, he made it clear—in branding his endeavor the study
of naive psychology—that he was talking about a person’s thoughts as they
could be communicated to others. Other authors have taken issue with this,
prefering instead to regard such cognitive constructs as tacit. The best known of
such critics are Langer (1978) and Nisbett and Wilson (1977).
 192 PETERSON AND SELIGMAN

Attributions are regarded as hypothetical constructs (MacCorquodale &

Meehl, 1948) akin to constructs such as hunger and expectation. This means that
attributions are not defined (and exhausted) by introspective evidence; otherwise,
they would be intervening variables. Rather, a variety of kinds of evidence bear
on whether or not an attribution is present—introspection, betting behavior,
voluntary action, and others—and a subjective report is neither necessary nor
sufficient. A deaf mute boxer who: can report no attribution, but bets his life
savings that he will win; trains 16 hours per day; increases the distance he runs
each day during training; and measures his lung capacity weekly can be said,
with some confidence, to attribute his victory to wind, endurance, and other
skill-effort factors. Therefore, to ask whether attributions are conscious or un-
conscious is to make a category mistake (Ryle, 1949). Attributions are not the
sort of entity that can be either conscious or unconscious; rather, evidence about
conscious experience bears on whether they are in play, just as does evidence
about behavior. However, the attribution is no more the conscious experience
than the subjective report of hunger is hunger. The introspection counts toward
the belief that hunger is in play, just as do the number of hours of deprivation, the
activity of the hypothalamus, the amount of adulteration of food that will be
tolerated; the introspection is neither necessary nor sufficient. The phe-
nomenological view of attribution, which holds the introspection is necessary
and sufficient for the attribution, is making the same serious error for cognition
that Skinner makes in the realm of behavior—that intervening variables, or
constructs whose meaning is exhausted by operational definition! of input-output
relations, are useful forms of scientific explanation.
Thus the dilemma becomes not whether attributions are conscious. but what
can best be ascertained about the attributions by talking to an individual, and
what other characteristics can best be ascertained by means other than talking.
Two extremes exist for ascertaining attributions. One is simply to ask people
in
rather abstract terms about them; their reports bear on their attribution
s. The
other is to observe people. One can observe people in situations in which
what
they do is ipso facto evidence that a given (tacit) attribution has been
held. This
latter strategy was followed by the experiments leading up to the reformulat
ion of
the learned helplessness model, and has been followed in new
tests of the
reformulation. People are put in laboratory-manipulated situations
in which the
alteration of an attribution could be inferred with some confidence
, and changes
in overt behavior observed.
The first sense of ascertaining attributions involves asking people
about their
Own attributions, and then seeing if attributions assesse
d in this manner bear

'We are not against operational definitions or well-def

inedness of terms in mapping how a
Concept is instantiated in a particular experiment. The
concept that underlies the particular operations
from experiment to experiment, however, cannot
be operationalized by necessary and sufficient
conditions in hypothetical construct science.
 6. ATTRIBUTIONAL STYLE IN DEPRESSION 193

reasonable relationships to other things these people do. Recent research along
this line and the second line is reported. This research was designed to see if, and
where, the two senses of assessing attributions converge.

THE ATTRIBUTIONAL STYLE QUESTIONNAIRE

In the research to be reported attributional style was measured with a paper-and-

pencil instrument called the ‘‘Attributional Style Questionaire’’ (ASQ). Several
considerations led to its format. Questions were needed that would measure the
degree to which subjects used attributions falling along the dimensions of inter-
nality, stability, and globality, as defined by the reformulated learned help-
lessness mode. Constraining subjects by producing them with possible attribu-
tions (e.g., ability, effort, task difficulty, luck, and so on), had to be avoided for
two reasons: First, there is no guarantee that an attribution regarded by an
attribution theorist as, for example, unstable is so regarded by all subjects; some
may believe that low effort is a stable characteristic of the individual (laziness),
others may perceive it as unstable (sloughing off). Second, Falbo and Beck
(1979) have shown that the attributions assumed by Weiner (1974) (ability,
effort, task difficulty, and luck) to exhaust the internal-external and stable-
unstable dimensions, neither occur preponderantly in free-responding nor cluster
as expected when they do occur.
On the other hand, Ross’ (1977) observation that the coding of causal attribu-
tions into abstract categories depends more on the (theoreticaly irrelevant) gram-
matical form of the attribution than on its actual meaning, argues against the use
of completely open-ended questions. The report by Elig and Frieze (1979) also
contends that open-ended attributional measures are not as reliable as fixed-
format procedures. However, in some research (e.g., Peterson, 1980), high
reliability in the coding of open-ended causal responses have been reported.
A reasonable compromise was employed between closed and open-ended
questions in the assessment of attributional style. For each of 12 (six good and
six bad) hypothetical events involving themselves, subjects were asked to gener-
ate a cause. They were then asked to rate their provided cause along 7-point
scales that corresponded to the internality, stability, and globality dimensions.
This format does not constrain or create the causal attributions, but it does allow
for simple and objective quantification. Further, this format allows the researcher
to content analyze the chosen attributions whenever it is desired.
The questionnaire (ASQ) is group administered; the following directions ap-
pear on the first page of the booklet:

Please try to vividly imagine yourself in the situations that follow. If such a
situation happened to you, what would you feel would have caused it? While events
may have many have many causes, we want you to pick only one—the major cause
 194 PETERSON AND SELIGMAN

if this event happened to you. Please write this cause in the blank provided after
each event. Next we want you to answer some questions about the cause.

Twelve different hypothetical events follow these instructions. Half are good
events (e.g., ‘““you do a project which is highly praised’’); while half are bad
events (e.g., “‘you meet a friend who acts hostilely toward you’’). The wording
of the various attributional dimension rating scale questions reflects the specific
event to be explained. However, in general, the subject is asked to circle a
number from 1 to 7 for scales anchored by: (a) totally due to other people or
circumstances totally due to me; (b) will never again be present/will always be
present; (c) influences just this particular situation/influences all situations in my
life. These questions pertain to the cause, and operationalize the externality-
internality, unstability-stability, and specificity-globality dimensions.
The rating scales associated with each event are scored in the directions of
increasing internality, stability, and globality. Composite scores are created sim-
ply by totaling the appropriate items and dividing the sum by the number of
items, in the composite. The practice has been to compute internality, stability,
and globality scores separately for good and for bad events; thus, each of these
six subscales consists of six items. Further the subscales are combined (internal
plus stable plus global) for good events by totaling them; the same is done for the
bad events. Table 6.1, presents the means, standard deviations, and incorrelatins
of the six subscales, based on a sample of 130 undergraduate (50 males, 80
females) enrolled in an abnormal psychology course at the State University of
New York at Stony Brook. The scale reliabilities, estimated by Cronbach’s
(1951) coefficient alpha, are also present. As can be seen, the scale reliabilities
mostly exceed the scale intercorrelations; however, the values are still relatively
low. Greater reliability is achieved when the good and bad subscales, respec-
tively, are combined; this yields an alpha of .75 for the good composite and .72
for the bad composite. In light of this increased reliability, unless the researcher
is interested in the individual dimensions, it is recommended that the composite
scales be used.
Also of note in Table 6.1 is the fact that for the positive events, the individua
l
attributional dimensions are intercorrelated at a level much nearer that of
their
reliabilities than are the dimensions for negative events. Thus, the present
instru-
ment did not discriminate among these dimensions for positive events
to the
degree that it did for negative events. Why should this be the case?
Perhaps
people make fewer discriminations among good events. They may
not spend as
much time ruminating over good events as they do over bad events;
therefore,
they may attend less to the causes of good events and are less
articulate (cf.
Scott, Osgood, & Peterson, 1979) in rating them (cf. Langer,
1978: Peirce,
1955; Ryle, 1949).
One hundred of the subjects who provided the above data also
completed the
attributional style questionnaire five weeks later. Table 6.2
summarizes the test-
 6. ATTRIBUTIONAL STYLE IN DEPRESSION 195

TABLE 6.1
Intercorrelations of the ASQ Subscale
a
Subscale 1 2 3 4 ES) 6 yi 8

Positive events
l. internality (.50)
2. stability 1622 (.58)
3. globality .36* .39* (.44)
4. composite — — — 79)

Negative events
5. internality 11 O01 —.03 — (.46)
6. stability =. 17 —.07 .03 _ allies (.59)
7. —globality —.15 .04 .24* — .28* 45” (.69)
8. composite i _ = = — — (72)

mean 5.28 336 SullR; _ 4.30 4.16 3.89

s.d. .79 .68 .80 -- .84 71 1.07
ss II ee
Note: Figures in parentheses are reliabilities, estimated by Cronbach’s (1951) coefficient alpha.
os OS

retest correlations of the attributional dimensions and the composites. As had

been hoped, the correlations are respectably high, because the scores are hypoth-
esized to reflect a ‘‘style.’’ It is worth observing that not all measures of cog-
nitive style proved to be so reliable, either internally or across time (Goldstein &
Blackman, 1978; Streufert & Streufert, 1978); therefore, this finding is
underscored.

TABLE 6.2
Test-Retest Correlations of the ASO

Subscale r

Positive events
internality 38:
stability 1095
globality 59,
composite 70"

Negative events
internality 64"
stability 69"
globality 57°
composite 047

*p < .001
 THE RELATIONSHIP OF ATTRIBUTIONAL STYLE TO
DEPRESSION

Among the first studies that followed the reformulation were several that corre-
lated individual differences in the attributional style with depressive symptoms.
Seligman et al. (1979) reported a study using undergraduates at the University of
Pennsylvania. These students completed the ASQ and two measures of de-
pressive symptoms: The Beck Depression Inventory (BDI; Beck, Ward, Men-
delson, Mock, & Erbaugh, 1961) and the Multiple Affect Adjective Checklist
(MAACL; Zuckerman & Lubin, 1965). The BDI is an often used and well
validated self-report questionnaire measuring the severity of 21 different symp-
toms of depression (see Beck, 1967, and Bumberry, Oliver, & McClure, 1978,
for validity evidence, both in clinical and undergraduate samples). The MAACL
asks subjects to circle various adjectives characterizing their mood; depression,
anxiety, and hostility scores are derived from this instrument.
Seligman et al. (1979) found, as predicted by the reformulated model, that an
attributional style in which bad events were attributed to internal, stable, and
global causes was correlated with the BDI and the MAACL, both for the indi-
vidual dimensions and for the composite. To a lesser degree, the opposite style
for good events was correlated with the BDI. Although this preliminary study
was an important one, it left several important questions unanswered. First,
would the correlation between attributional style and depressive symptoms ob-
tain for more severe depression? Second, is the attributional style particularly
characteristic of depression, as suggested by the reformulation, or is
it a more
general characteristic of psychopathology?
In order to answer both of these questions, a second study was undertak
en by
our research group (Raps, Peterson, Reinhard, Abramson, & Seligman
, 1982).
The second study administered the ASQ to unipolar depressed patients
hospi-
talized in a Veterans’ Administration Medical Center. Two nondepre
ssed groups
of patients also completed the ASQ, schizophrenics and surgical
patients. Both
comparison groups controlled for the effects of hospitalization
per se (because
other research suggests that hospitalization can result in helplessn
ess and depres-
sion; Raps, Peterson, Jonas, & Seligman, 1982); in addition,
the schizophrenic
group assessed whether the ‘‘depressive’’ attributional
style is specific to
depression.
Table 6.3 summarizes the means of the attributional dimens
ions and com-
posites for each of the three patient groups. As can be seen,
the predictions ofthe
reformulated learned helplessness model were
strongly supported; depressed
patients were more likely than surgical and schizophrenic
patients to attribute bad
events to internal, stable, and global causes. As with
the student study, although
to a lesser degree, the opposite style for good events
was also associated with
depression. For the most part, the schizophrenic
patients rated their provided
Causes similarly to the surgical patients; when
differences occurred, they were in

196
 6. ATTRIBUTIONAL STYLE IN DEPRESSION 197

TABLE 6.3
Attributional Style Means—Patients
me nn nn el hun sn EI DE eee
Style for Bad Events Style for Good Events

Group INT STA GLO COMP INT STA GLO COMP

Unipolar
Depressives (n=30)
mean 4.90 4.89 4.84 4.88 4.93 4.90 5.10 4.98
s.d. 1212 1.16 1.29 2.92 1833 1.39 1.41 3.81

Surgical patients (n=64)

mean 4.30 4.06 3.65 4.00 5.49 9293 Sl 5.44
s.d. 1.18 .90 1.37 213 .83 81 1.16 221]

Schizophrenics (n=15)
mean 3.51 4.01 4.00 3.87 5.67 7] S27) 5.50
s.d. 1.18 1.01 1219 226 1.19 1.02 I 3.02

p depressives vs. surgical

patients .001 .001 .002 .001 AON .008 ns .03
p depressives vs.
schizophrenics .001 .016 .069 .001 .07 .10 ns Al]
p surgical patients vs. .002 ns ns ns ns ns ns ns

Note: Abbreviations are as follows: INT=internality, STA=stability, GLO=globality, and COMP

=composite. The higher the number the more internal, stable, or global the score. Comparisons between
groups were made with r-tests; p levels are for two-tailed tests.

a direction away from the depressives. Thus, it can be concluded that the insid-
ious attributional style is not generally associated with psychopathology. Fur-
ther, it can be concluded that clinically depressed patients have an attributional
style like that of mildly depressed students; thus, support is given to the idea that,
at least cognitively, mild and severe depression may be continous. Although no
single study, or pair of studies, will decide the issue of continuity, once the
ground rules are established, it is an issue with an empirical answer. In order to
find the answer, it is suggested that the controversy be moved from the arm chair
or the couch to an explicit empirical domain.
Do children with depressive symptoms have this insidious attributional style?
Answering this question is a conceptual replication of the previous discussion,
however, because certain theorists doubt the very existence of childhood depres-
sion (e.g., Lefkowitz & Burton, 1978), is also of interest in its own right.
The advent of behavioral and cognitive formulations of depression, emphasiz-
ing overt manifestations, reopened the issue of childhood depression (e.g.,
Schuldenbrandt & Raskin, 1977). Children display many of the symptoms, that
 198 PETERSON AND SELIGMAN

are called depressive in adults. These symptoms are dismissed by some as

transient development by-products, or signs of inner turmoil; one of the things to
be tested was the duration of these symptoms. Depending on the degree that the
symptoms: cohere with each other, as they do in adults; are associated with
similar attributional style, are caused by similar attribultional style; and have
similar cure and prevention, they are considered depressive symptoms; it is
concluded that there is an entitiy of childhood depression.
The study of children begins to provide some evidence about the origins of
attributional style. The learned helplessness model has been silent on the source
of attributional styles; however, attributional literature suggests that such styles
might arise in at least two ways (Kelley, 1973). First, as the result of experiences
that suggest the same obvious kind of cause; people eventually come to apply
that attribution to events a priori (Peterson, 1980). Second, the style may be
learned abstractly as a maxim for living. With close analysis, distinctions be-
tween these two ways of acquiring attributional tendencies break down: even
inductive learning must be grounded in abstract beliefs about which causes are
relevant (Ajzen, 1977).
Parents are, of course, a reasonable source of attributional style; a child may
hear its parents habitually explaining events in a certain way and come to do the
same. Peterson recently witnessed a father and his young son at a basketball
game. In the first half of the game, the son watched with wide eyes as his father
attributed every call against the Celtics to the ““stinking referees’’; in the second
half, the son did the same thing, as his father nodded his agreement. The
son had
come to “‘understand’’ the causal structure of basketball (Berger & Luckmann,
1966). Of course, it is understood that Julius Erving was a much more
robust
cause of the Celtic ills than were the referees.
The first attempt to assess attributional style in children with the adult ASQ
failed, because the younger children in the pilot samples had trouble
using the
rating scales for stability and globality. Therefore, it was necessary
to retreat
from the strategy of compromising between open and closed-format
question-
naires; instead, the children were given a closed-format instrument,
which they
were able to understand and complete satisfactorily.
In the questionnaire (CASQ), the children were given a series of
hypothetical
events; for each event they chose between two possible causes.
The response
alternatives varied one of the attributional dimensions; the
other two remained
constant. For example, in this item:

You get an A on a math test.

(a) I’m smart.
(b) I’m smart at math.

Alternative (a) is a global attribution; alterna

tive (b) is a specific attribution.
Internality and stability are held constant. This
questionnaire has 48 items—
 6. ATTRIBUTIONAL STYLE IN DEPRESSION 199

eight items for each of the six attributional subscales (internality, stability, and
globality for good events versus bad events). Table 6.4 summarizes the intercor-
relations and internal reliabilities of these subscales for the sample described. As
can be seen, although they tend to exceed the scale intercorrelations, the reli-
abilities are more modest than for the adult scale. The composite scales for good
and bad events improved the reliabilities to .66 and .50, respectively at Time 1,
and .73 and .54 at Time 2.
Ninety-six children in grades 3 through 6 (in two Philadelphia elementary
schools) completed the questionnaires at two different times, separated by an
interval of six months. Table 6.5 summarizes the test-retest correlations, which
were substantial. It was concluded that even for children, there are stable attribu-
tional tendencies. The children also completed a self-report depression inventory
created by Kovacs and Beck (1977) and modelled after the BDI. At both admin-

TABLE 6.4
Intercorrelations of Attributional Style Dimensions Children (CASO)
ee ee I 7 de
Subscale 1 2; 2) 4 J 6 # &

Time 1
Positive Events
l. internality (.32)
2. stability 265 (.55)
3. globality 333% 41° (.40)
4. composite — — — (.66)

Negative events
5. internality 228 —.45* —.14 — (.43)
6. stability 227 Sal) = 02 _ 18 (.42)
7. globality = 0)! = OD He — 02 za“ (.31)
8. composite —_ u — — — — — (.50)

Time 2
Positive events
l. internality (.43)
2. stability 255 (.54)
3. globality 335 al“ (33)
4. composite — — a (G73)

Negative events
5. internality =,24" 415 =i 14 — (.56)
6. stability =al3 23 peril — 205 (413)
7. globality —.20 .03 .03 —_ .05 „237 (39)
8. composite == — a = — =< = (.54)

Note: Figures in parentheses are reliabilities, estimated by Cronbach’s (1951) coefficient alpha.
Sp 05
 200 PETERSON AND SELIGMAN

TABLE 6.5
Correlations Between Attributional Style
and Depression—Children (CASQ)

Correlation with CDI

Subscale Time 1 Time 2

Positive events
internality 34 =31**
stability SAT = 54
globality Oe 3
composite 932 =.54"5*

Negative events
internality .45* 28
stability ol 26%
globality 210 2266
composite Sn .40°**
ee Bee rn m 1, | Mn u rn
“p= .05
[op =O
= p< OO

istrations, the Children’s Depression Inventory (CDI) correlated with the attribu-
tional subscales and composites as predicted by the reformulation; these correla-
tions were observed both for the good and for the bad events.
More detailed analysis of the data revealed: (a) that girls were more depressed
than boys (p <.05) at both administrations: and (b) that there was no
tendency
for depressive symptoms to change with age. These findings indicate
that the
greater incidence of depression among adult females (e.g. Radloff,
1975) occurs
as early as nine years of age (cf. Dweck & Bush, 1976). Depressive
symptoms in
children are not ‘‘transitory developmental by-products’; They
are highly sta-
ble, and do not disappear with age. The mothers of 47 of these
children and the
fathers of 36 completed the BDI and the ASQ, and returned
them to the re-
searchers. The child measures were averaged for the two
administrations, and
correlations with those of the parents were computed. Briefly:
(a) mother’s
attributional style for bad events correlated with her child’s
attributional style for
bad events (.39) and with her child’s depressive symptoms
(.42); (b) mother’s
depressive symptoms correlated with her child’s depressive
symptoms (.37); (c)
father’s attributional style and depression were not related
to scores of his spouse
or their children. The intercorrelations of the mother
and child composite mea-
sures are presented in Table 6.6. More detailed analysis
suggested that the bulk
of the relationships among the composite attributional
measures resulted from the
contributions of the internality and globality subscales.
 6. ATTRIBUTIONAL STYLE IN DEPRESSION 201

It is speculated the cognitions of the depressive (Beck, 1967) may

be embed-
ded within an interpersonal vicious circle. The depressions of mother and
child
may maintain each other, particularly when each has the depressive attributi
onal
style. If so, then the currently popular individual therapies used with depressi
ves
(e.g., Beck, Rush, Shaw, & Emery, 1979) might be supplemented with
family
therapy; this has been done with other disorders in which family interacti
on
contributes. Most generally, the results suggest that the depressed child is apt to
be found in a family in which the mother is also depressed, and vice-versa.
Depressed mothers not only hurt themselves but their children as well. The fact
that fathers’ depression and attributional style were not associated with those of
their wives or children may be due to a characteristic of the modern family;
fathers spend less time with their children than do mothers.
Implicit in the reformulation is the notion that changes in attributional style
Should accompany successful therapy for depression (Seligman, 1981). Indeed,
an important aspect of the currently popular cognitive therapy for depression
(Beck et al., 1979) is an emphasis on changing the depressive’s worldview,
particularly the way in which the causes of an event are construed. To date, as a
consequence of therapy, changes in attributional style have not been docu-
mented; however, there is some case history evidence regarding such changes.
Nine transcripts of segments of therapy sessions with depressed patients were
blindly coded for depressive attributional style, by members of our research

TABLE 6.6
Intercorrelations of Depression and Attribution Style—
Mother and Child

Measure 1 2 5} 4 5)

Mother
1. ASQ composite positive events (.65)
2. ASQ composite—negative events al 2 (.79)
3. depression (BDI) ode CU (.92)

Child
4. CASQ composite—positive
events .08 —07 = 08 (.82)
5. CASQ composite—negative
events = (9) 295 Pal —26 (.76)
6. depression (CDI) a 42°” oy AB 4g" (.84)

Note: Figures in parentheses, on diagonal, are reliabilities, estimated by Cronbach’s (1951) coefficient
alpha.
*p < .05
DSO
7001
 202 PETERSON AND SELIGMAN

group. The transcripts were then rank-ordered by how bad the attributional style
was. It was then revealed that the coded transcripts did not belong to nine
different patients, but rather in the course of therapy to three different points of
only three patients. Thus, it was possible to test the prediction that as improve-
ment occurred, the patients would be less likely to explain bad events in terms of
internal, stable, and global causes (thus more likely to employ external, unstable,
and specific attributions).
The transcripts were edited excerpts from psychodynamic therapy sessions
(under the supervision of Mardi Horowitz) with three depressed women, all of
whom had recently experienced a traumatic loss, such as the death of a parent or
spouse. Three transcripts from each patient were available, from the beginning,
middle, and end of therapy. Each transcript was approximately four to five
double-spaced typed pages in length, and was edited to: protect the confiden-
tiality of the patient; to remove nonattributional clues about severity of depres-
sion, and when in therapy the session occurred; and the fact that each transcript
was paired with the transcripts of two other patients.
All phrases involving a specific bad event, for which a causal attribution was
made, were extracted from each transcript. In all, 78 such event-attribution units
were obtained; an average of 9.2 per transcript, with a range from 5 to 21. These
units were then randomized and presented to four judges, who using the three
dimensions from the ASQ (i.e., the 1 to 7 scales measuring internality, stability,
and globality) independently rated each cause.
Product-movement correlations between the ratings by all pairs of judges
were computed, as well as for a composite attributional score. The reliabilities
were quite high, indicating that the judges could agree about the designation of a
given attribution. For each of the nine transcripts, mean scores for each of the
dimensions and composite were formed. (The individual dimensions ordered the
transcripts in much the same way as did the composite.) For all three patients (all
of whom, it turned out, improved over therapy), the attributions changed in the
course of therapy. As hypothesized by the reformulation, the changes were away
from internal, stable, and global cuases toward external, unstable. and specific
causes. In addition, severity of depression across patients seemed roughly related
to insidiousness of the attribution style.
The procedures used here were conservative ones: coding was totally
blind,
relation to depression was blind, and the exact nature of the transcripts was
not
known to the group while they were coded. Further, therapy was conducte
d by a
therapist, not particularly sympathetic to our position. Indeed, because
his the-
oretical preference is to look beyond the manifest content of patients’
attributions
he presented the transcripts as a challenge. The very small sample
makes this
“‘evidence’’ highly tentative; however, this study may be added to
the others as
supportive of the idea that an insidious attributional style is
associated with
depression. Such an association was demonstrated with college
students, de-
pressed in-patients, and patients in analysis. Further, the findings
suggest that the
 6. ATTRIBUTIONAL STYLE IN DEPRESSION 203

insidious attributional style is a relatively stable characteristic, that style for

positive events is essentially independent of style for negative events, and that
there is a correspondence between the styles of mothers and their children.

ATTRIBUTIONAL STYLE AS A CAUSE OF

DEPRESSION

Several longitudinal studies of the development of depressive symptoms with

individuals’ initial attributional style as the risk factor are described. This is
followed by a description of several laboratory studies that examine the conse-
quences of helplessness induction for subjects with different attributional styles.
Both sets of studies test whether or not attributional style contributes causally to
depression.
The first study (Semmel et al., 1980) investigated both attributional style as a
risk factor in the development of depressive symptoms and the role of subsequent
bad events. None of the correlational studies described earlier assessed whether
or not actual bad events correlated with depressive symptoms, however, this is
critical to the reformulated model. The model does not claim that the insidious
attributional style in and of itself is sufficient for depression to develop; rather, a
bad event processed through the insidious attributional style is needed. This
hypothesis was investigated using a sample of college students at the State
University of New York at Stony Brook; it replicated a pilot study done at the
University of Pennsylvania.
Early in the semester 119 students in an abnormal psychology course had their
attributional style assessed (with the ASQ) along with their depressive symptoms
(with the short form of the BDI; [Beck & Beck, 1972]). An intervening event
also occurred—the midterm examination. According to the reformulation, de-
pressive changes should most likely occur among those students who possess
both a prexisting insidious attributional style and who in their own eyes, ‘‘fail’’
on the midterm. Failure on the midterm was determined by asking the students if
they were satisfied or dissatisfied with the grade they had received.
The students were devided according to whether they reported being satisfied
(n = 43) or dissatisfied (n = 76) with their midterm grade. It is important to note
that the two groups did not differ with regard to initial attributional style, nor
initial depression and grade aspirations (which were also assessed early in the
semester). However, the two groups, of course, differed in the actual grades they
received, which means that their satisfaction reflected somewhat uniform stan-
dards. The satisfied students averaged in the B/B+ range; those in the dissatis-
fied group scored in the C range. Overall, the two groups did not differ with
regard to depression, neither immediately following the return of the midterm,
nor four weeks later; meaning that a bad event in and of itself was not sufficient
for depresive symptoms to develop. Nor was the insidious attributional style
 204 PETERSON AND SELIGMAN

alone sufficient to bring about depressive symptoms. Among students satisfied

with the grade, no relationship was found between attributional style and de-
pressive symptoms (see Table 6.7). If attributional style, per se, were a cause of
depression, those students who succeed, but who have an insidious style, should
become more depressed than those who succeed but have a nondepressed style.
In contrast, when attributional style scores were correlated with ensuing de-
pressive symptoms for the dissatisfied group, statistically significant correlations
were found (see Table 6.7). For these students, a prior attributional style in
which bad events are attributed to internal, stable, and global causes correlates
with depressive symptoms one week and four weeks following the midterm; the
insidious style for good events (external, unstable, specific) correlated with
depressive symptoms four weeks following the midterm. Thus, both the insid-
ious attributional style and dissatisfaction with grade were necessary for later
depressive symptoms; these results support the reformulation. A depressive at-
tributional style puts a person at risk for later depressive symptoms, but only
when a bad event ensues. Neither the life event nor the style alone leads to
depressive symptoms. The correlational studies are clarified: Overall correlations
between attributional style and depression presumably result because part of the
sample has experiences recent bad events. The results were more pronounced
four weeks after the bad event than immediately after the event: this implies that
the depressive symptoms were not transient disappointment or bad mood.
A second longitudinal study used the children described earlier. The attribu-
tional style and depression of the 96 children were measured two times separated
by an interval of 6 months. Causal modelling was employed to evaluate the
hypothesis that attributional style is a ‘‘risk factor’’ for depression. It is possible
to decide questions of causal priority among nonmanipulated variables by using a
longitudinal design and cross-lagged panel correlational analyses (Campbell,
1963; Campbell & Stanley, 1963; Kenny, 1973, 1975, 1979). The correlations

TABLE 6.7
Correlations between Attributional Style and
Depression—Midterm Examination Study
——— EEE EE ee eee
Dissatisfied Satisfied
Correlation Students Students

One Week After Exam

positive composite with BDI 05 .04
negative composite with BDI 252 .04

Four Weeks After Exam

positive composite with BDI 242 .18
negative composite with BDI 28" = 06

p05
 6. ATTRIBUTIONAL STYLE IN DEPRESSION 205

TABLE 6.8
Cross-Lagged Correlations Between Attributional Style (CASQ)
Composites and CDI Scores—Children
ee

Bad events

DAS
saul .40*

Good events

Al—A2
-.47:
-.53* -.54*
DI— 53 D2
.80*
TG
Note: Al and A2 are attributions scores at Times | and 2, respectively;
D1 and D2 are depression scores at Times | and 2, respectively.
*p < .001

between attributional style and depression across time were computed. Because
causes work forward in time, differences between the correlations of attributions
at Time | and depression at Time 2 versus depression at Time | and attributions
at Time 2 suggest causal priority. (For examples of recent uses of this inferential
technique, see Kahle & Berman, 1979, and Knight, Roff, Barnett, & Moss,
1979). Because of the modest reliabilities of the attributional subscales, only the
composites were used. Table 6.8 summarizes the relevant correlations. It can be
seen that the correlation between attributional style for bad events at Time | and
depression at Time 2 exceed the correlation betwen depressive symptoms at
Time 1 and attributional style at Time 2 (p < .02). This supports the prediction
of the insidious attributional style about good events; the cross-lagged correla-
tions were essentially the same (—.47, —.54), implying no causal priority.
Golin, Sweeney, and Shaeffer (1981) have found similar results for adults.
These findings support the reformulated helplessness model, however, a
shortcoming should be mentioned. Depression, as operationalized here, was
more stable across six months than attributional style. The learned helplessness
model suggests that attributional style is more characteristic of a person than is
depression, which is sometimes a consequence of attributional style. According-
ly, the stability of depression should be lower than the stability of attributional
style. One possibility is that the lower reliabilities of attributional style accounts
for the discrepancy (the reliabilities of the CDI at the two administrations
 206 PETERSON AND SELIGMAN

were .86, and .80, respectively, estimated by Cronbach’s [1951] coefficient

alpha), because the reliability of a measure puts an upper limit on its stability.
Another possibility is that attributional style is not a particularly stable individual
difference, or that it may be more stable (and more predictive of depression) in
some individuals than in others (Abramson & Martin, 1980).
Several pilot laboratory investigations of the reformulation are discussed. A
number of studies, reviewed by Abramson et al. (1978), Miller and Norman
(1979), and Roth (1980), have shown that following helplessness induction,
situational manipulations of attributions determine the nature and extent of defi-
cits. Such studies are, of course, compatible with the reformulated model, and in
fact, they led to it; however, they do not speak to the role of attributional style in
helplessness and depression. In other words, do the attributions which people
habitually make their dispositional characteristics, affect helplessness deficits?
To answer this question, the authors are in the process of doing several
studies, all having essentially the same design: Subjects are divided on the basis
of their scores on the ASQ, they are given helplessness induction, and then a test
task. Performance on the test task is predicted to be jointly determined by the
pretreatment and attributional style. Such a design constitutes a strong and direct
test of the reformulation. The strength of the test stems from the fact that
individual differences in attributional style are predicted to operate in a situa-
tionally constrained setting (i.e., a laboratory helplessness experiment). The
learned helplessness model does not claim that dispositional attributional style
will always override situational determinants of attributions, so attention must be
paid to whether or not the situational constraints were weak enough to allow
attributional style to do its dirty work.
Following helplessness induction, each of the three attributional dimensions
have a specifically predicted property: (a) people with an internal attributional
style for bad events will show self-esteem deficits: (b) people with a stable
attributional style for bad events will show deficits across time: and (c) people
with a global attributional style for bad events will show deficits across
situations.
We have not yet investigated prediction (a) in the laboratory; however, (b)
and (c) have been addressed in pilot work. In one experiment, 96 subjects
designated as stable or unstable (on the basis of a median split on this dimension
of the ASQ) were given one of three pretreatments (escapable noise, inescapabl
e
noise, no pretreatment control) and then were tested to terminate noise on a
hand
shuttle-box, either immediately after the pretreatment or after a delay of three
days (see Hiroto & Seligman, 1975, for a description of this apparatus).
It was
predicted that only the stable subjects would show helplessness three
days later,
but that both stable and unstable subjects would show helplessness
on the imme-
diate test.
Table 6.9 summarizes the results for two dependent measures
: number of
successful escapes and mean latency of responses. For both measure
s, the same
 6. ATTRIBUTIONAL STYLE IN DEPRESSION 207

TABLE 6.9
Shuttlebox Performance—Stability Study
eee

Stable Subjects Unstable Subjects

Mean Mean
# Latency # Latency
Condition Successes (Seconds) Successes (Seconds)

Immediate Test
escapable 18.38 1.90 18.00 1.99
inescapable 9.00 8208 SD 4.18
control 115.73 2.68 15.88 2.38
Delayed Test (3 days later)
escapable 18.38 2.07 18.25 225
inescapable 6.25 4.16 10.75 3.09
control 13.62 2.99 10975 3.3333

results occurred: inescapable groups, except for the unstable subjects at three
days, did worse in the task than the escapable and control groups. This is exactly
what was predicted, and it was obtained even when the subjects’ internality and
globality scores (from the ASQ) were used as covariates in the appropriate
analyses of variance.
In the second experiment, 144 subjects designated as global or specific (again
by a median split on this dimension of the ASQ) were given one of the three
pretreatments and then tested on a task which was either similar or dissimilar to
the pretreatment task. The similar task was the hand shuttlebox (both the pre-
treatment task and the shuttlebox involved escaping noise); the dissimilar task
was a series of 20 anagrams (see Hiroto & Seligman, 1975). The prediction was
that global subjects would show deficits following inescapable noise on both the
similar and dissimilar test tasks; specific subjects would show deficits only on the
similar test task. The results of this study are in Table 6.10. Again, two depen-
dent measures were recorded for each task; each showed the same pattern of
results that were predicted by the reformulation. Inescapable subjects did badly
at the test task in all conditions, except for specific subjects with a dissimilar test,
even when internality and stability scores were used as covariates.
These two studies provide support for the reformulated model, however, they
are pilot work. Additional work, in which the sample size of the studies is
increased, and the parameters varied so that attributional style has more *‘room’’
to operate, will be done.
Nevertheless, there is reason to be optimistic about the hypothesis that attribu-
tional style “‘causes’’ later depressive symptoms and helplessness. The longitu-
dinal studies and the laboratory experiments provide converging evidence that a
person’s habitual attributional tendencies processing uncontrollable, bad events
 208 PETERSON AND SELIGMAN

TABLE 6.10
Globality Study

Global Subjects Specific Subjects

Mean Mean
# Latency = Latency
Condition Successes (Seconds) Successes (Seconds)

Similar Task
escapable 15.67 2.62 18.00 2.03
inescapable 1275 3522 12.58 3.17
control 14.58 EN 13.83 2.82
Dissimilar Task
escapable 18.18 17.85 127, 215
inescapable 14.91 39.42 18.08 19.53
control 17.67 23.47 17.58 22.58

produce depressive deficits. To a considerable degree, the reformulation has

been supported by the research undertaken to date. In the last section of this
chapter, some doubts about the reformulation are presented. Current controver-
sies in the social psychological literature on attribution theory are the point of
departure; and these suggest potentially fertile directions for future helplessness
research.

CONCLUSION: SUMMARY, CRITIQUE, AND NEW

DIRECTIONS

The empirical work described addressed two main questions. Does an insidious
attributional style characterize depression? Does this style cause depression?
Other questions addressed included: (a) Are mild and severe depression continu-
ous? (b) Is childhood depression a viable concept? (c) What are the origins of
attributional style? (d) Does attributional style as assessed by questioning a
person relate to behavior?
The authors are now in a position to offer some answers to these questions: A
tendency to attribute bad events to internal, stable, and global causes is associ-
ated with depressive symptoms. All of the studies point to this conclusion. Some
people have remarked: ‘‘Of course this attributional style characterizes depres-
sion—it is a symptom of depression.’’ In a certain sense, such a reduction of
our
research to the trivially obvious is flattering. However, the fact remains,
that
such a symptom of depression is not listed in any diagnostic manual. In addition,
nondepressed people have this alleged symptom; these are the individuals
who,
as predicted, are at risk for depression when bad events strike.
 6. ATTRIBUTIONAL STYLE IN DEPRESSION 209

Brief answers to the subsidiary questions addressed in the recent research may
also be provided. First, there is no evidence in the studies that shows mild
depression and severe unipolar depression to be anything but different in degree.
Still, it is recognized that much more extensive study of clinical populations is
needed, and we are turning our attention to such investigations. Second, child-
hood depression seems to be a meaningful concept; because it is not transient, it
has the same symptoms as adult depression, and it has the same cognitive risk
factors. Third, attributional style in children appears to derive, at least in part,
from mothers. Fine-grained study of mother-child interaction is recommended,
in order to discover how it is transmitted: the ultimate goal is to immunize
children against depressogenic parents. Fourth, there appears to be considerable
use to assessing attributional style by talking to a person. The Attributional Style
Questionnaire taps introspection about causes, and suggests that a person’s
causal inferences about the world are not entirely tacit; if they were, none of the
relationships found would be easily interpretable. In contrast to the vast social
psychological literature documenting discrepancies between what a person says
and what a person does (Wicker, 1969), research suggests that sensibly asked
questions are a fertile source of predictions about how a person will act.
There are several worries about the lines of research and theorizing under-
taken. First, the ‘'man as scientist’’ metaphor: such a model usually ignores
motivational constructs (although not always; cf. Heider, 1958). People hold
certain attributions, not because they are motivated to do so, but because—
granted their premises—it is rational to do so (Nisbett & Ross, 1980). This
notion contrasts starkly to psychoanalytic conceptions of depression that hold
depressives to be motivated to maintain their depression. At the present time,
there is insufficient evidence to decide if the attributional style shown by depres-
sion is at all motivated.
Another criticism of the ‘‘man as scientist’? metaphor has to do with the
“rationality’’ it assumes (Nisbett & Ross, 1980). The learned helplessness
model assumes that people operate logically with their attributional premises;
however, a great deal of recent research has demonstrated that normative models
of information processing are sometimes inaccurate descriptions of what people
actually do (e.g. Kahneman & Tversky, 1973; Nisbett & Ross, 1980; Ross,
1977; Slovic & Lichtenstein, 1971). Future work should be directed at uncover-
ing the heuristic bases of depressives’ judgments, perhaps taking as a point of
departure Beck’s (1976) errors of logic. Some theorists have argued that the so-
called causal attributions made by people are not technically causes but rather
justifications (Buss, 1978; Harre & Secord, 1971; T. Mischel, 1964). Answers
beginning with ‘‘because’’ need not refer to prior causes, but intead to goals that
rationalize the action in question. The upshot of this point is that the attributional
style of depressives may be forward looking, and not backward looking, as
implied by the reformulation. If you are hit by lightning and survive, the fact that
the cause of your bodily injuries is unstable, external, and specific is of no
 210 PETERSON AND SELIGMAN

consolation, if you have to face a year’s worth of complicated surgery and

convalescence. Attributional style in depression is probably related not just to
prior causes, but also to future consequences.
Related to this point is a question raised by Wortman and Dintzer (1978) and
Janoff-Bulman (1979) in their critiques of the reformulation. They doubted
whether the three attributional dimensions studied so far exhaust the critical
cognitive precursors of depression, and suggested that attributions of control-
lability were also crucial. According to Janoff-Bulman (1979), only if the inter-
nal attribution is about character will the attribution be depressive. In contrast,
internal attributions about behavior are antidepressive. According to Janoff-
Bulman (1979), people are powerless to change their character, but they can
modify and control their behavior.
To test this proposal, a sample of 87 undergraduates were asked to complete
an expanded version of the ASQ along with the BDI (Peterson, Schwartz, &
Seligman, 1981). The modification of the ASQ asked the subjects to quantify the
control they perceived over each of the events. The authors then blindly content
analyzed the attributions they provided as external, internal-behavioral, or inter-
nal-characterological. Briefly, the findings supported Janoff-Bulman (1979).
Only characterological attributions were associated with depressive symptoms:
both external and behavioral attributions were rated as more uncontrollable than
behavioral attributions. However, it was also found that characterological at-
tributions were rated as more stable and more global than were behavioral at-
tributions. In short, the factor of control could be reduced to the factors of
stability and globality, or vice versa. Thus, although a large number of charac-
teristics of depressive attributional style could be proposed, in all likelihood,
many of them would be redundant. For the time being, although recognizing that
other names could be attached to them, the authors will stay with the dimensions
of internality, stability, and globality. Other orthogonal dimensions might indeed
exist, however, researchers must demonstrate not just the conceptual dis-
tinctiveness of additional dimensions, but also their empirical distinctiveness.
A final issue, which should worry helplessness theorists, had to do with the
maladaptiveness of helplessness and depression. Because of the way help-
lessness experiments are set up and because the test tasks are solvable. a belief in
noncontingency between responses and outcomes, and the holding of the insid-
ious attributional style, are usually maladaptive. However, other views of
de-
pression see it as adaptive: a means of taking stock when surrounded by brick
walls; a signal to disengage from fruitless pursuits (Klinger, 1975); and a
mes-
Sage that one’s commerce with the world is not all right (Rippere, 1976).
All
failures of adaptation are not the fault of the individual: often it is the world
which is responsible for the ‘‘problem,’’ the ““maladaptive’’ cognitions of the
victim.
In summation, a good deal was gained by the reformulation of the
learned
helplessness model. It better explained the facts of helplessness and
depression.
 6. ATTRIBUTIONAL STYLE IN DEPRESSION 211

However, something was also lost—the “learning?” part of the original model.
The old model, coming out of the animal learning tradition, specified precisely
the environmental contingencies that brought about helplessness. The new
model, with its emphasis on the person’s cognition is less straight forward; the
transduction from environmental contingency to expectation is modified by other
cognitions, such as attributional style. Such a loss may be necessary, at least for
now; however, the final recommendation is that future researchers attempt to
understand how attributional style is acquired. It is hoped that this conference
provide the impetus for such an attempt.

ACKNOWLEDGMENTS

The research reported here was supported by PHS MH 19604 to M. Seligman. We would
like to acknowledge the help and advice of our colleagues, especially the members of the
Helplessness Seminar at the University of Pennsylvania.

REFERENCES

Abramson, L. Y., & Martin, O. (1980). Depression and the causal inference process. Unpublished
manuscript, State University of New York at Stony Brook.
Abramson, L. Y., & Sackeim, H. A. (1977). A paradox in depression: Uncontrollability and self-
blame. Psychological Bulletin, 84, 839-851.
Abramson, L. Y., & Seligman, M. E. P. (1977). Modeling psychopathology in the laboratory:
History and rationale. In J. D. Maser & M. E. P. Seligman (Eds.), Psychopathology: Experimen-
tal models. San Francisco, CA: Freeman.
Abramson, L. Y., Seligman, M. E. P., & Teasdale, J. D. (1978). Learned helplessness in humans:
Critique and reformulation. Journal of Abnormal Psychology, 87, 49-74.
Ajzen, I. (1977). Intuitive theories of events and the effects of baserate information on prediction.
Journal of Personality and Social Psychology, 35, 303-314.
Alloy, L. B., & Seligman, M. E. P. (1979). The cognitive component of learned helplessness and
depression. In G. H. Bower (Ed.), The psychology of learning and motivation (Vol. 13). New
York: Academic Press.
Beck, A. T. (1967). Depression: Clinical, experimental, and theoretical aspects. New York:
Hoeber.
Beck, A. T. (1976). Cognitive therapy and the emotional disorders. New York: International
Universities Press.
Beck, A. T., & Beck, R. W. (1972). Screening depressed patients in family practice: A rapid
technic. Postgraduate Medicine, 52, 81-85.
Beck, A. T., Rush, A. J., Shaw, B. F., & Emery, G. (1979). Cognitive therapy ofdepression. New
York: Guilford.
Beck, A. T., Ward, C. H., Mendelson, M., Mock, J., & Erbaugh, J. (1961). An inventory for
measuring depression. Archives of General Psychiatry, 4, 561-571.
Berger, P. L., & Luckmann, T. (1966). The social construction of reality. Garden City, NY:
Doubleday.
Berglas, S., & Jones, E. E. (1978). Drug choice as a self-handicapping strategy in response to
noncontingent success. Journal of Personality, 36, 405-417.
 212 PETERSON AND SELIGMAN

Brown, G. W., & Harris, T. (1978). Social origins of depression: A study of psychiatric disorder in
women. New York: Free Press.
Bumberry, W., Oliver, J. M., & McClure, J. N. (1978). Validation of the Beck Depression In-
ventory in a university population using psychiatric estimate as the criterion. Journal of Consult-
ing and Clinical Psychology, 46, 150-155.
Buss, A. R. (1978). Causes and reasons in attribution theory: A conceptual critique.
Journal of
Personality and Social Psychology, 36, 1311-1321.
Campbell, D. T. (1963). From descripton to experimentation: Interpreting trends
as quasi-experi-
ments. In C. W. Harris (Ed.), Problems in measuring change. Madison: University
of Wisconsin
Press.
Campbell, D. T., & Stanley, J. C. (1963). Experimental and quasi-exper
imental designs for re-
search on teaching. In N. L. Gage (Ed.), Handbook of research on teaching.
Chicago: Rand
McNally, 1963.
Cronbach, L. J. (1951). Coefficient alpha and the internal structure of tests. Psychometr
ika, 16,
297-334.
Douglas, D., & Anisman, H. (1975). Helplessness or expectation
incongruency: Effects of aversive
stimulation on subsequent performance. Journal of Experimental Psycholog
y: Human Perception
and Performance, 1, 411-417.
Dweck, C. S., & Bush, E. S. (1976). Sex differences in learned helplessness: 1. Differential
debilitation with peer and adult evaluators. Developmental Psycholog
y, 12, 147-156.
Dweck, C. S., & Reppucci, N. D. (1973). Learned helplessn
ess and reinforcement responsibility in
children. Journal of Personality and Social Psychology, 25,
109-116.
Elig, T. W., & Frieze, I. H. (1979). Measuring causal attributio
ns for success and failure. Journal
of Personality and Social Psychology, 37, 621-634.
Falbo, T., & Beck, R. C. (1979). Naive psychology and
the attributional model of achievement.
Journal of Personality, 47, 185-195,
Frankel, A., & Snyder, M. L. (1978). Poor performance
following unsolvable problems: Learned
helplessness or egotism? Journal of Personality and Social
Psychology, 36, 1415-1423.
Geen, R. G. (1978). Effects of attack and uncontrollable
noise on aggression. Journal of Research
in Personality, 12, 15-29.
Goldstein, K. M., & Blackman, S. (1978). Cognitiv
e style: Five approaches and research. New
York: Wiley.
Golin, S., Sweeney, P. D., & Shaeffer, D. E. (1981). The causality of causal attributi
ons in
depression: A cross-lagged panel correlational analysis.
Journal of Abnormal Personality, 90,
14-22.
Harre, R. M., & Secord, P. F. (1971). The explanation of social behaviour. Oxford:
Basil
Blackwell.
Harvey, J. H., Ickes, W. J., & Kidd, R. F. (1976).
New directions in attribution research (Vol. 1).
Hillsdale, NJ: Lawrence Erlbaum Associates.
Heider, F. (1958). The psychology of interpe
rsonal relations. New York: Wiley.
Hiroto, D. S., & Seligman, M. E. P. (1975).
Generality of learned helplessness in man. Journal
Personality and Social Psychology, 31, 311-32 of
7.
Janoff-Bulman, R. (1979). Characterological
versus behavioral self-blame: Inquiries into
sion and rape. Journal of Personality and Social depres-
Psychology, 37, 1798-1809.
Jones, E. E., & Nisbett, R. E. (1971). The
actor and the observer: Divergent percep
causes of behavior. New York: General Learni tions of the
ng Press.
Journal of Abnormal Psychology, (1978).
87(1).
Kahle, L. R., & Berman, J. J. (1979).
Attitudes cause behaviors: A cross-lagged
Journal of Personality and Social Psycho panel analysis.
logy, 37, 315-321.
Kahneman, D., & Tversky, A. (1973).
On the psychology of prediction. Psych
237-251. ological Review, 80,
 6. ATTRIBUTIONAL STYLE IN DEPRESSION 213

Kelley, H. H. (1973). The processes of causal attribution. American Psychologist, 28, 107-128.
Kelly, G. A. (1955). The psychology of personal constructs. New York: Norton.
Kenny, D. A. (1973). Cross-lagged and synchronous common factors in panel data. In
A. S.
Goldberger & ©. D. Duncan (Eds.), Structural equation models in the social sciences. New
York: Seminar, 1973.
Kenny, D. A. (1975). Cross-lagged panel correlation: A test for spuriousness. Psychological Bul-
letin, 82, 887-903.
Kenny, D. A. (1979). Correlation and Causality. New York: Wiley.
Klinger, E. (1975). Consequences of commitment to and disengagement from incentives. Psycho-
logical Review, 82, 1-25.
Knight, R. A., Roff, J. D., Barnett, J., & Moss, J. L. (1979). Concurrent and predictive validity of
thought disorder and affectivity: A 22-year follow-up of acute schizophrenics. Journal of Abnor-
mal Psychology, 88, 1-12.
Kovacs, M., & Beck, A. T. (1977). An empirical-clinical approach toward a definition of child-
hood depression. In J. G. Schulterbrandt & A. Raskin (eds.), Depression in Children. New York:
Raven, 1977.
Krantz, D. S., Glass, D. C., & Snyder, M. L. (1974). Helplessness, stress level and the coronary-
prone behavior pattern. Journal of Experimental Social Psychology, 10, 284-300.
Kruglanski, A. W. (1980). Lay epistemo-logic—Process and contents: Another look at attribution
theory. Psychological Review, 87, 70-87.
Langer, E. J. (1978). Rethinking the role of thought in social interaction. In J. H. Harvey, W. J.
Ickes, & R. F. Kidd (Eds.), New directions in attribution research (Vol. 2). Hillsdale, NJ:
Lawrence Erlbaum Associates.
Lefkowitz, M. M., & Burton, N. (1978). Childhood depression: A critique of the concept. Psycho-
logical Bulletin, 85, 716-726.
MacCorquodale, K., & Meehl, P. E. (1948). On a distinction between hypothetical constructs and
intervening variables. Psychological Review, 55, 95-107.
Maier, S. F., & Jackson, R. L (1979). Learned helplessness: All of us were right (and wrong):
Inescapable shock has multiple effects. In G. H. Bower (Ed.), The psychology of learning and
motivation (Vol. 13). New York: Academic Press.
Maier, S. F., & Seligman, M. E. P. (1976). Learned helplessness: Theory and evidence. Journal of
Experimental Psychology: General, 105, 3—46.
Maier, S. F., Seligman, M. E. P., & Solomon, R. L. (1969). Pavlovian fear conditioning and
learned helplessness. In B. A. Campbell & R. A. Church (Eds.), Punishment and aversive
behavior. New York: Appleton-Century-Crofts.
Miller, I. W., & Norman, W. H. (1979). Learned helplessness in humans: A review and attribution
theory model. Psychological Bulletin, 86, 93-119.
Miller, S. M., & Seligman, M. E. P. (1980). The attributional reformulation of helplessness:
Recent evidence. Unpublished manuscript, University of Pennsylvania.
Mischel, T. (1964). Personal constructs, rules, and the logic of clinical activity. Psychological
Review, 71, 180-192.
Mischel, W. (1968). Personality and assessment. New York: Wiley.
Monson, T. C., & Snyder, M. (1977). Actors, observers, and the attribution process: Toward a
reconceptualization. Journal of Experimental Social Psychology, 13, 89-111.
Neisser, U. (1967). Cognitive Psychology. New York: Appleton-Century-Crofts.
Nisbett, R., Ross, L. (1980). Human inference: Strategies and shortcomings of social judgment.
Englewood Cliffs, NJ: Prentice-Hall.
Nisbett, R., & Wilson, T. D. (1977). Telling more than we can know: Verval reports on mental
processes. Psychological Review, 84, 231-259.
Overmier, J. B., & Seligman, M. E. P. (1967). Effects of inescapable shock upon subsequent
escape and avoidance learning. Journal of Comparative and Physiological Psychology, 63, 23-33.
 214 PETERSON AND SELIGMAN

Paykel, E. S. (1974). Recent life events and clinical depression. In E. K. E. Gunderson & R. D.
Rahe (Eds.), Life stress and illness. Springfield, IL: Thomas.
Peirce, C. S. (1955). Philosophical writings of Peirce (J. Buchler, Ed.). New York: Dover.
Peterson, C. (1978). Learned helplessness following insoluble problems: Learned helplessness or
altered hypothesis pool? Journal of Experimental Social Psychology, 14, 53-68.
Peterson, C. (1979). Uncontrollability and self-blame in depression: Investigation of the paradox in
a college population. Journal of Abnormal Psychology, 88, 620-624.
Peterson, C. (1980). Attribution in the sportspages: An archival investigation of the covariation
hypothesis. Social Psychology Quarterly, 43, 136-140.
Peterson, C., Schwartz, S. M., & Seligman, M. E. P. (1981). Self-blame and depressive symp-
toms. Journal of Personality and Social Psychology, 41, 253-259.
Radloff, L. S. (1975). Sex differences in depression: The effects of occupation and marital status.
Sex Roles, 3, 249-265.
Raps, C. S., Peterson, C., Jonas, M., & Seligman, M. E. P. (1982). Patient behavior in hospitals:
Helplessness reactance or both? Journal of Personality and Social Psychology, 42, 1036-1041.
Raps, C. S., Peterson, C., Reinhard, K. E., Abramson, L. Y., & Seligmen, M. E. P. (1982).
At-
tributional style among depressed patients. Journal of Abnormal Psychology, 92, 102-108.
Rippere, V. (1976). Review of Seligman’s Helplessness. Behavior Research and Therapy,
15,
207-209.
Ross, L. (1977). The intuitive psychologist and his shortcomings: Distortions in the attribution
process. In L. Berkowitz (Ed.), Advances in Experimental Social Psychology (Vol.
10). New
York: Academic Press.
Roth, S. (1980). A revised model of learned helplessness in humans. Journal of Personality,
48,
103-133.
Rotter, J.B. (1966). Generalized expectancies for internal versus external control of reinforceme
nt.
Psychological Monographs, 80(1, Whole No. 609).
Ryle, G. (1949). The concept of mind. London: Hutchinson. 1949.
Schuldenbrandt, J. G., & Raskin, A. (1977). Depression in children. New York: Raven.
Scott, W. A., Osgood, D. W., & Peterson, C. (1979). Cognitive structure.
Washington, DC:
Winston.
Seligman, M. E. P. (1974). Depression and learned helplessness.
In R. J. Friedman & M. M. Katz
(Eds.), The psychology of depression: C ontemporary theory and
research. Washington, DC:
Winston.
Seligman, M. E. P. (1975). Helplessness: On depression, developme
nt, and death. San Francisco:
Freeman, 1975.
Seligman, M. E. P. (1981). A learned helplessness point of view.
In L. P. Rehm (Ed.) Behavior
Therapy for Depression: Present Status and Future Directions
. New York: Academic Press.
Seligman, M. E. P., Abramson, L. Y., Semmel, A., & von
Baeyer, C. (1979), Depressive attribu-
tional style. Journal of Abnormal Psychology, 88, 242-247.
Seligman, M. E. P., & Maier, S. F. (1967). Failure to escape
traumatic shock. Journal of Experi-
mental Psychology, 74, 1—9.
Seligman, M. E. P., Maier, S. F., & Soloman, R. L.
(1971). Unpredictable and uncontrollable
aversive events. In F. R. Brush (Ed.), Aversive conditio
ning and learning. New York: Academic
Press.
Seligman, M. E. P., Weiss, J., Weinraub, M.,
& Shulman, A. (1980), The Seligman-Weiss
debate. Behavior Research and Therapy, 18, 459-512
.
Semmel, A., Peterson, C., Abramson, L. Y., Metalsk
y, G. L., & Seligman, M. E. P. (1980). Pre-
dicting depressive symptoms from attributional style
and failure. Unpublished manuscript, Uni-
versity of Pennsylvania.
Slovic, P., & Lichtenstein, S. (1971). Compari
son of Bayesian and regression approaches to
study of human information processing in judgmen the
t. Organizational Behavior and Human Per-
formance, 6, 649-744,
 6. ATTRIBUTIONAL STYLE IN DEPRESSION 215

Smith, R., & Seligman, M. E. P. (1980). Black and lower class children are more vulnerable to
impairment of problem solving following helplessness. Unpublished manuscript, University of
Pennsylvania.
Streufert, S., & Streufert, S. C. (1978). Behavior in the complex environment. Washington, DC:
Winston.
Weiner, B. (1974). Achievement motivation and attribution theory. Morristown, NJ: General
Learning Press.
Wicker, A. W. (1969). Attitudes versus actions: The relationship of verbal and overt behavioral
responses to attitudinal objects. Journal of Social Issues, 25, 41-78.
Wortman, C. B., & Brehm, J. W. (1975). Response to uncontrollable outcomes: An integration of
reactance theory and the learned helplessness model. In L. Berkowitz (Ed.), Advances in experi-
mental social psychology (Volume 8). New York: Academic Press.
Wortman, C. B., & Dintzer, L. (1978). Is an attributional analysis of the learned helplessness
phenomenon viable? A critique of the Abramson-Seligman-Teasdale reformulation. Journal of
Abnormal Psychology, 87, 75-90.
Zuckerman, M., & Lubin, P. (1965). Manual for the Multiple Affect Adjective Check List. San
Diego: Educational and Industrial Testing Service.
 w—
ZA& (| Minis
» DB are iu) a Ba
u = u de i

u hen a ae ;
. & u u Eu I ur u

en me ee ee ae
i >= A - ae oe
e. vo u... (| aap ome -
) Ak ae a) a a a othe.
Vu ~ thet ah: te ua iu Aue
7 e - A vie me Aus ee ze ee ce
HH na 6 eh te eh u
= a =» . > are“ ow 3 ae t Bay
Ct oan se Eur u az
u om > eo Tr!
=> ¢ Ss S Ho ma)e & .
De >
0 e (aa ei (eee
Par

ph

vr
ae
pr
 Feeling versus Being Helpless:
Metacognitive Mediation of
Failure-Induced Performance
Deficits

Julius Kuhl
Max-Planck-Institute for Psychological Research
Munich

Some years ago, a newspaper printed an article about a gang of burglars who
robbed a warehouse and locked the two night watchmen in the trunks of two cars.
One of the men survived by breathing air from the spare tire. The next morning
he was able to attract the attention of some people passing by the car. The other
man had suffocated before the trunk was opened (cf. Atkinson, 1964, p. 74). If
one could have observed the less fortunate man shortly before his death, one
might have noticed that he did not try at all to find a solution to his problem. If
the man could have been asked why he did not try, he might have attributed his
lack of motivation to the hopelessness of his situation, he did not expect to be
able to control the situation, so he gave up. Motivational psychologists would be
delighted with this pattern of observations. They would conclude that the unfor-
tunate man’s behavior was perfectly consistent with an expectancy-value theory
of motivation, however, this insight would not have been of much help to the
poor man in the trunk.
This little story may illustrate some of the problems involved in cognitive
theories of action, in general, and the cognitive theory of learned helplessness, in
particular. In most studies, empirical tests of hypotheses regarding cognitive
mediators of behavior are based on correlational evidence. On the one hand,
there is some measure of the assumed cognitive mediator, such as causal attribu-
tion, self-efficacy, or belief in the uncontrollability of outcomes (cf. Bandura,
1977; Roth & Kubal, 1975; Weiner, Frieze, Kukla, Reed, Rest & Rosenbaum,
1971). On the other hand, there is some index of the behavior psychologists are
interested in, for instance, risk taking, behavioral change, or performance, re-
spectively. Of course, it is known that one may not draw causal inferences from

217
 218 KUHL

correlational evidence, but more often than not it is done, especially when
correlational data is all that is available.
To return to the unfortunate man in the trunk: Is it true that he does not find
the solution because he does not try? Is it true that he does not try because he
feels helpless? Or could the causal chain work just in the opposite direction?
Could it be that the man, in spite of initially trying very hard, actually was
helpless right from the beginning? In this case, the primary problem would be a
functional helplessness, that is, the man may have been so overwhelmed by the
dangerous situation that he could not think clearly. Motivational helplessness,
that is, giving up because of feeling helpless, may have developed as a result of
functional helplessness, rather than being its cause. Thus, one may ask whether
the man felt helpless as a result of being helpless or whether he became helpless
as a result of feeling helpless.

LEARNED HELPLESSNESS

Consider the typical laboratory experiment on learned helplessness. Subjects

who have been exposed to repeated uncontrollable failure on some training task
typically show impaired performance on a subsequent test task, compared
to a
control group that was allowed to succeed on the training task, or did not work
on
that task. The performance deficits in the failure group have been observed even
when the test task was totally different from the training problem (e.g.,
Hiroto &
Seligman, 1975). Seligman and his associates have attributed impaired
perfor-
mance following helplessness training to an expectational deficit.
According to
this theory, subjects develop a belief in the uncontrollability of outcomes
that
results in a motivational deficit. Deteriorated performance is considere
d a direct
consequence of reduced motivation.
An alternative explanation of helplessness phenomena is present
ed. This al-
ternative is consistent with what is known from theory and
research in achieve-
ment motivation. Unfortunately, this body of research within
the literature on
learned helplessness has been neglected; although the two
research traditions are
based on very similar experimental paradigms. It has been
theoretically expected
and empirically demonstrated that repeated exposure to uncontr
ollable failure is
associated with an increase rather than with a decrease
in the need for achieve-
ment (Atkinson, 1958; McClelland, Clark. Roby &
Atkinson, 1949). This phe-
nomenon has been called the motivational effect of
failure (Atkinson & Birch,
1974). In their discussion of an experiment on
learned helplessness, Miller &
Seligman (1975, p. 236) report an observation that is
consistent with that effect,
however, they do not emphasize the fact that it runs
counter to what should be
expected on the basis of their theory of helpless
ness. The subjects from the
failure group seemed to try hard when confronted
with the test task and later,
when they noticed how poor they were performing,
gave up. There is reason to
 7. METACOGNITIVE MEDIATION 219

believe that the need for or the personal value of achieving success does not
decrease following failure. But what about the assumed expectational deficit?
Do subjects transfer reduced expectancy of control from the training task to
the test task? The theory of learned helplessness is based on the assumption that
people generalize their experiences extensively (Seligman, 1975). The author
would like to emphasize the discriminative abilities of human subjects. Mischel
(1968; 1973) has collected a great amount of evidence suggesting that human
subjects may discriminate precisely, even between seemingly similar situations.
An extensive study was conducted on the degree to which cognitive variables
generalize across various situations. Among all the parameters included in the
study (for instance, expectancy, causal attribution, personal standards, perceived
value of success), expectancy variables turned out to be the most specific (Kuhl,
1977). Analyses of the data based on Rasch’s stochastic measurement model
(Rasch, 1960), revealed that concepts of one’s own ability to control outcomes
typically were not transferred from one task to another.
If neither the value nor the expectancy of succeeding at the test task is
reduced, as a result of failing on the training task, it is difficult to account for the
observed performance deficits on the basis of expectancy-value theory alone.
This theory may explain what people intend to do; however, it has to be ex-
tended, in order to explain the degree to which they actually accomplish their
goals. Discrepancies between motivation and performance have been attributed
to detrimental effects of overmotivation. This explanation actually begs the
question: What are the cognitive processes that intervene between the intention to
achieve a goal and the performance of the activities necessary to reach that goal?
The expectancy-value theory should be extended by a third factor, which the
author calls action-control. This factor presumably controls the extent to which
an intended action is actually performed.
Among the various cognitive processes controlled by that factor, selective
attention presumably plays a major role. Specifically, it is assumed that perfor-
mance snould be facilitated when attention is focused on action-related informa-
tion. Conversely, when attention focuses on state-related information, the perfor-
mance of an intended action should be impaired. The degree of state-orientation
may mediate failure-induced performance deficits. Subjects may be preoccupied,
focusing on the state created by the failure experience, trying, for instance, to
explain why they failed or trying to analyze their own emotional state. One
advantage of this interpretation of helplessness effects is that it explains the
generalization of performance deficits. This is done without making unnecessary
assumptions regarding a transfer of perceptions of uncontrollability to tasks that
are perceived as totally different from the task one has failed to accomplish. If
performance deficits are attributable to deteriorated cognitive functioning, as a
result of interfering state-related cognitions, they can transfer to any kind of
behavior, even to activities that are not achievement-related. A person who is
preoccupied with analyzing his or her state may show impaired performance on
 220 KUHL

any type of test, burn the dinner, or behave awkwardly toward his or her friends.
The theory of action-control is consistent with an attentional theory of anxiety
(Wine, 1971) as well as with Diener & Dweck’s (1978) empirical analysis of
helplessness effects. In this study, children who showed impaired performance
following helplessness training were more preoccupied with, what the author
calls, state-oriented thinking than children whose performance did not
deteriorate.

ACTION-CONTROL

Before the results from two experiments testing some of the assumptions are
presented, an attempt is made to specify, in greater detail than earlier theories,
the nature of the cognitions, which presumably interfere with performance, and
the functional significance of those cognitions. Action-control at first, was not
thought of as a means to explain helplessness effects: originally, the construct
was proposed to explain why people sometimes do what they intend to do and
why they sometimes fail to behave according to their intentions. Naturally,
action-control should be a function of personality and situational variables.
A
questionnaire was constructed to assess individual differences in action-control
(Kuhl, 1985). The validity of this questionnaire could be demonstrated
in a
recent study (Kuhl, 1982). Sixth graders (48) from a German school
were asked
to rate to what extent they intended to engage in various activities such as
playing
in a ballgame, watching TV, or reading a book, after school. The
next day,
subjects were asked how much time they had actually spent doing
those ac-
tivities. Correlations between reported degree of commitment and
reported time
spent were computed separately for each activity within two groups,
classified as
state-oriented or action-oriented, according to the median
of the action-control
scale. Figure 7.1 illustrates the results of this study. If the action-contr
ol scale is
a valid measure of a person’s tendency to behave according
to his or her inten-
tions, there should be a greater correlation between intention
and behavior within
the action-oriented group than within the state-oriented group.
This was actually
the case for 11 out of 13 free-choice activities. The
study also included a
situational variable affecting action-control. The list of activities
also described
nine forced-choice activities, which are usually carried
out under social pressure
in that age group. When the performance of an
activity is mainly externally
controlled, congruence between intention and performanc
e should not require a
strong degree of action-control; the results were also
in line with this prediction.
Correlations between degree of intention and actual
performance for the nine
externally controlled or routine activities, such as
brushing one’s teeth or doing
one’s homework, did not differ consistently between
action and state-oriented
subjects.
 7. METACOGNITIVE MEDIATION 221

Ta
08
7 N]state-oriented subjects

07 GZ >
7 2 action-oriented subjects
06 G Y
=
| Y eth?G Z
G Y
G
ap AZ
Y AY y mn YY
04
7
J 7
G
YY Yj
Yj
Yj
J A
AY
G Z
Aga 9 ZIG 9% AY G
o-
ZU
nu
U A
7gy „an
A\Y BUY
7G AY Y
Yew Y Yl Y
A\Y G AY
A TV
01 9 7 9 7 9
AA AIA ZIYI
9
IA MAAN
I
AIHIMIM
KXXYSKCGCCRL
SSS
7 y y
erce-crore
if u GY ACTIVITIES
= Z
$ gG M
-01 ni Z 2 2
2
So S+
eos6
=
=® Z
G
3= 8®
-02- 37 38 aa ® LZ ® > =
aR
ee >t NS= 5 Ss ® = = 2
c= i= > fm (o) E je] Dp Oo
-03- 22
Su
29
Ss
52
Eric
2 50.700059
2 =)
=c eo
=
eCda o
Ze °5 25 3 a os
E55 EN 2 seen
oO 2 o 2 E+
KK
for
essay-writing-test
Qn |
friends
with
play

FIG. 7.1. Rank correlations between degree of intentional commitment and

reported time spent for 13 free-choice afterschool activities, computed separately
within subjects groups classified as state-oriented or action-oriented.

What is the theoretical status of action control? How does it affect behavior?
Some possible answers to those questions are outlined in Fig. 7.2. Action-control
is thought of in terms of a metacognitive process aiming at the performance of an
intended action. This process is classified as metacognitive for two reasons: first,
it presumably organizes and controls cognitive operations, such as selective
attention and the amount and content of information processed, in a way that
maximizes the congruence between intention and performance; second, action-
control presumably relies heavily on knowledge about the effectiveness of vari-
ous cognitive operations for achieving that objective. The degree of action-
control aroused in a particular situation may vary between the two extremes
called state-orientation and action-orientation.
What could be a definitional basis for describing an organism as action-
oriented rather than state-oriented? The author would like to define the organism
as action-oriented when it successively or simultaneously attends to all of the
following elements: the present state; some future state; a discrepancy between
the two states; and action alternatives that may help transform the present into the
 $}9N4}SUOD
jo

222
sjUueperajuy UOIyDJUa!J0-3}DjS
(OS) UONDJUaIVO-UO!}DYV
(OW) SjUuepedajuy
WSINOSVLNV

aAipUB0ID}a;)
| UO!}!SOdSI
P4DMO} ezA)Duy,,
8}D}S a6uny9,, „Buiyyawos volsodsıPuDMo}
g
}2A2) OS Pauojs) LSIUOP,, abpa;mouy
yNoqo abpa};mouy
InogD OVW P2s0}S) (SIsNW,,
Ayiqo}}o4yuooun BuızAjDun
anıoa
jo Asoyoundasd
anıpa
JO pen
reel) ee SNOIDA SUONUBO9 SNOIIDA SUONUBOI SUON ;
pasno.o) („s}snw,,

ONIYO.LINOW AYOLVUVI3U4
SNOILINDOJI SNOILINSOO
aariubo5
| BDUAIIDS
JO sayojs SNd04-
UO $8}D}S SNd04-
UO LONDD
jana] aDUDAa]aIJ}
0/D - payosa66px3
-JOjul_ SOALNDUJ2}]D UOIDD SOADUJA}]D
PDo]JaAo
JO -JOJUI uoNDW Buissadoud- SNOIUOWISIDY
-JOJUI SIUDABJOY
0/D -IW!]
UONDW uoNDu BUISS3I0Jd UOND}
JO LONDUNOJUT
jDUCOW;
x| jpuoyowg
sand BuyDyIGeqG-
uojowe - GulyD41904
32940
GAI}DAj4JD3 |Duoyowa 2 BAI}DAjJD
ja1a] Ayiqosnosp Ayiqosnoso

UO! |3A3] Ayaıssog }>3J1P-]D09

Pa
Jo UOI}9D
Bulyosanasuad Buiwin)
yo Bulbupys
SOI}IAI}OD ay} Juasauyd3}D}S
ou) pepuaju! aBupyd PsDMO}
D Bunyn)(QUO
jo (27045
‘JO 9UI}NOJ
PUD UO1}9D Buiinbas
A}}DUsa}xe P31]044U09 xa}dwos Bumuup)d
UO!}ID

94 “ZZ Y DANTUSOSRIOW
[APOU
JO ‘JONUOS-UONOR
 7. METACOGNITIVE MEDIATION 223

intended future state. If one of those elements is missing, the organism is said to
be state-oriented.
Figure 7.2 outlines the theory of action-control. On the metacognitive level,
the theory assumes an antagonism between state-orientation and action-orienta-
tion. State-orientation is conceptualized as a metacognitive tendency controlling
cognitive operations that facilitate the analysis of some past, present, or future
state. Action-orientation is considered a metacognitive tendency that instigates
cognitive operations that, in turn, facilitate the arousal of goal-directed action
tendencies. Action-oriented cognitions may focus on expectancy, or value-relat-
ed information regarding the usefulness of various action alternatives. Examples
for state-oriented activities may be: ruminating about possible causes of the
present state, such as after a series of successive failures; or attending to one’s
own emotional state. State-oriented cognitions are considered not to instigate any
action tendency. They may even actively inhibit action tendencies, or block the
expression of action tendencies, as a result of the assumed partial incompatibility
between state and action-orientation. As the amount of state-oriented cognitions
aroused, increases, the probability that the person performs some intended action
decreases. Specifically, it is assumed that an action which requires complex
preparatory planning cannot be performed when the organism is state-oriented,
because part of the attentional capacity needed is used for state-oriented pro-
cesses. Routine activities or activities that are under strong environmental control
may be performed even when the organism is mildly state-oriented.
Among the assumed mediators of action-control described on the cognitive
level of the model (see Fig. 7.2), selective attention may be the most effective
one. Presumably, the metacognitive tendency toward action-orientation arouses
a tendency to selectively attend to stimuli that are relevant for the performance of
an intended action; whereas, state-orientation should be mediated by attention
focused on state-related information. Ach (1910) regarded selective attention as
the most important determinant of what he called the efficiency of the will.
Unfortunately, his early phenomenological analysis of the process intervening
between intention and performance—a process which he simply called the
will—was ignored in later theories of motivation. The main reason for this may
be Lewin’s (1935) criticism of the concept of volition and his attempt to replace
the problem of volition by the problem of motivation (cf. Kuhl, 1984b).
Besides selective attention, a second factor may mediate between intention
and performance. State-orientation may be associated with a greater amount of
information processed than is necessary for performing an action; action-orienta-
tion may be mediated by a strategy aiming at the most parsimonious processing
of action-related information. Also, the two antagonistic metacognitive pro-
cesses are considered to be associated with two opposing feelings. State-orienta-
tion should be associated with emotions that interfere with the instigation of
action tendencies. Pribram (1971) uses the word e-motion literally, referring to a
process that takes the organism ‘‘out of motion.’’ Conversely, the feelings
224 KUHL

associated with action-orientation may be called affects. Ach (1910) used this
term when referring to facilitating effects of the feelings aroused following an
unexpected failure outcome. As one can see from Fig. 7.2, the extent to which
each of the three assumed factors (selective attention, parsimony of information
processing, and emotional or affective feelings) mediates between the metacog-
nitive and the behavioral level should depend on respective antecedent conditions
(the salience of action vs. state-related cues, the amount of relevant or irrelevant
information available, and the arousability of debilitating or facilitating feelings
within the person).
It would be beyond the scope of the present contribution to report the em-
pirical evidence related to various aspects of the model (see Kuhl, 1985, for a
detailed summary). The results of two studies conducted to test selected aspects
of the model are summarized. In one study (see Fig. 7.3), the mediating role of
selective attention could be demonstrated (Kuhl, 1983, p. 269). Attention to
task-relevant or task-irrelevant stimuli was assessed by a method based on mem-
ory recall. The results indicated that action-oriented subjects recalled signifi-
cantly less task-irrelevant stimuli than did state-oriented subjects. However,
when the task was made uncontrollable the results were reversed. In an uncon-
trollable condition, action-oriented subjects appeared to pay less attention to
task-relevant stimuli than state-oriented subjects. This pattern of results is con-
sistent with the assumption of a metacognitive strategy aiming at the mainte-
nance of action-control. In order to stay active, the organism should selectively
attend to action-related stimuli, and stop attending to them when the situation
becomes uncontrollable. The latter strategy facilitates a change to another—
controllable—activity.

|] state-oriented
subjects

Z action-oriented
subjects
uo

WORDS
RECALLED
N

OF
NUMBER
TASK-IRRELEVANT
—_

FIG. 7.3. Number of task-irrelevant words re-

called in two experimental conditions (7 vs. 14
U 14 trials) and two groups of subjects (action vs.
N U MBER OF TRIALS state-oriented Ss)
 7. METACOGNITIVE MEDIATION 225

[|state-oriented subjects
>
=5
8 | 2 action-oriented subjects
Dub
3 state-related behavior | action-related behavior

se | GB
fo)

Y
> | y TAO Y Z
Pe25 : Aw AM 9
2 ZURZER Z. ||
G 7Z 7A\lA
A\A|P Q
Gg
919 191 9 |l51| 9 IA GY
919141919119
© G os G Z
l1AlA
G 7
Al 9
Y Y Z

at ano
3o> =e ee ileal sche 03
© ee) Poet Gis3
38= ee
ne rt ea
nee ey AEG
De ak
CS ee © N ee er a oOrcin’ Oo ae

ee
Een
enEen
ee a
95 Wen
ieaa Gare
SeSie
cn
ss
FIG. 7.4. Degree of endorsement of 5 state-related items and 5 action-related
items for two groups of subjects.

The second study (see Fig. 7.4) revealed some correlates of action-orienta-
tion, related to all three intervening processes postulated in the model (Kuhl,
1984a). Subjects who had undergone a hernia operation were selected from three
German hospitals. Two days after their operation, they were administered the
action-control scale and several other scales. State-oriented subjects rated the
intensity of their pain, as expected, significantly higher, and requested double
the amount of analgesics than was requested by action-oriented subjects. The
postulated debilitating effect of greater emotionalism in state-oriented indi-
viduals was also indicated. State-oriented subjects reported to have less fre-
quently left the room to see what was going on in the ward, and they reported to
have fewer plans for the time after leaving the hospital than was the case for
action-oriented subjects. Also, state-oriented subjects more frequently reported
to have been engaged in thinking about their situation and to have looked at their
incision than was reported by the action-oriented subjects.

ACTION-CONTROL AND LEARNED HELPLESSNESS

Note that the model expects that the amount of action or state-orientation aroused
on the metacognitive level should be a function of an interaction between a
personal disposition toward action or state-orientation, situational determinants,
226 KUHL

such as uncontrollability of outcomes, and commands or instructions, exerting

some pressure on the person to initiate some action (Fig. 7.2). In the present
context the controllability factor is of special interest. Recently (Kuhl, 1981), I
proposed a three-factor theory of learned helplessness that summarizes the the-
oretical points already made in this discussion. The three-factor model is based
on a dynamic theory of achievement motivation (Atkinson & Birch, 1974; Kuhl
& Blankenship, 1979a, 1979b). According to this model, repeated exposure to
uncontrollable outcomes first increases a (metacognitive) tendency toward ac-
tion-orientation that is replaced by a tendency toward state-orientation when a
perception of uncontrollability has developed (cf. Fig. 7.5, a & b). State-orienta-
tion may interfere with optimal cognitive functioning, especially when the task is
complex enough to demand full attention. The two lower curves in Fig. 7.5
describe the assumed changes in motivation during helplessness training.
In contrast to traditional helplessness theory (Abramson, Seligman & Teas-
dale, 1978; Seligman, 1975), a distinction is made between motivation for the
training task and motivation regarding the test task. This distinction has to be
made in situations in which subjects discriminate regarding the type of abilities
involved, between training and test task. The lower curve (Fig. 7.5, e, f),
referring to the subject’s motivation to work on the training problem, predicts
heightened motivation during the first phase of helplessness training and below
baseline motivation when outcomes are considered uncontrollable. This curve
represents Wortman and Brehm’s (1975) extension of the theory of helplessness,
including an initial phase of reaction. Increased motivation as an initial response
to failure is consistent with various developments within the theory of achieve-
ment motivation (Atkinson & Feather, 1966; Kukla, 1972; Meyer, 1973), as well
as with earlier theories of motivation (Ach, 1937; Hillgruber, 1912).
When training and test tasks are not perceived to be a function of similar
factors, motivation to engage in the test task is assumed to be inversely related to
the inclination to persist on the original problem. As long as the person is still
strongly motivated to overcome the difficulties on the training problem, moti-
vation to switch to a different task should be low. However, as soon as the
motivation to persist on the original problem ceases, an increase in the moti-
vation to demonstrate one’s ability on a different task is expected. This predic-
tion is based on the assumed substitutional relation between various achieve-
ment-related tasks. If the need for achievement cannot be satisfied on one task, a
tendency to demonstrate one’s ability on some substitute task should be aroused.
The traditional theory of helplessness represents a special case of the three-factor
model. When training and test tasks are identical or similar, the lowest curve in
Fig. 7.5 is sufficient to explain performance deficits, following prolonged ex-
posure to uncontrollable outcomes.
In two recent experiments, an attempt was made to discriminate between the
two theoretical explanations, which are at stake. In the first study, 36 subjects
were randomly assigned to one of six groups. The six groups resulted from
 227
“ssoussa[djoy pousesy Jo [apou! JO}ORJ-IIy] WY “GZ 914
MSVl ANV LV MSV tees Sebel Vv. ASVL IN3WLV3UL3U4 LV
5 29UDJID3Y |SSeUssa|djay adUD}JIDaY |SSeUSSea}d}aH 8DUD}IDaY [SSaussea|djayH
= |DUOIDUNZ OUOIIUN JDUOIDANOY JOUO!}DAIJOW JOUOHDAIJOW | JDUOINDAIJOW
—-—-— low importance

fo)
—
l

oO Ders)
za voul | |
| |
belt ey emer
S He
task

o 2CoOSolas
E
\
oO
®
co
Qa oO 59
of pretreatment

(SAS
I1zOoO A
=
ss
Ee
high importance

oO
2)
©
Cc
no)
5
©
®

LVAILON
IVNIOIUO
a
x
Lu

NOI
NOILVAILON

ASVL
NOILVLN31IUO
NOILDV SA 31V1S VNOILNLILSONS
 228 KUHL

crossing two types of pretreatment with three types of intervention following

pretreatment. The two pretreatment conditions were incontingent failure feed-
back at a Levine-discrimination task, and a control condition in which subjects
merely “‘had a look at the tasks.’’ The intervention conditions were (a) induction
of state-orientation; (b) induction of action-orientation; and (c) no-intervention
control. At the beginning of the experiment, subjects were introduced to both
tasks that were to be used in the experiment. Following the training phase,
induction of state-orientation was attempted by allotting the subjects 5 minutes to
write an essay about the present situation and about the reasons for their
performance.
Induction of action-orientation was attempted by having the subjects read a
story about a little child with eating problems. During the test phase, a standard
concentration test (“‘d2-test’’ Brickenkamp, 1962) was used. In this test, the
subject has to mark all the letters d having two apostrophes, from a random
sequence of d’s and p’s having 0, 1, or 2 apostrophes, above and/or below the
letter. Before and after pretreatment subjects responded to four scales assessing
expectancy of success and perceived controllability regarding both the training
and the test task.
In order to check the extent to which expectancy and controllability beliefs
had been transferred from the training to the test task, the two measures were
compared between the failure and the no-failure groups. There were no reliable
differences in expectancy of success or in perceived controllability, regarding
either task, prior to the training phase. Significant differences were found follow-
ing training between the failure groups and the no-failure groups both in expec-
tancy and in controllability ratings (Fig. 7.6), regarding the training task. How-
ever, a similar difference was not found, on the posttraining measures of
expectancy or controllability regarding the test task. The results based on expec-
tancy ratings are shown in Fig. 7.6. The results indicate that neither reduced
expectancy of success nor the reduced perception of controllability has been
transferred from the training to the test task. Also, no evidence was found
suggesting a motivational deficit regarding the test task. To the contrary, subjects
in the failure conditions reported significantly greater importance to performing
well on the test task than control subjects. Further, subjects from the failure
condition which included an induction of state-orientation, reported significantly
greater effort expenditure on the test task than did control subjects. Obviously,
these results are more in line with the three-factor theory than with traditional
helplessness theory emphasizing motivational deficits,
An analysis of performance data from the test task revealed a significant main
effect of failure induction. As can be seen from Fig. 7.7, the number
of symbols
checked was significantly lower in the failure groups. Planned comparis
ons of
performance scores, between each failure group and the no-failure-no-inte
rven-
tion control group, showed that only the failure group that was
instructed to
engage in a state-oriented activity displayed a significant performa
nce deficit.
 7. METACOGNITIVE MEDIATION 229

The fact that performance deficits were most dramatic in the state-oriented
failure group supports the alternative explanation proposed by the three-factor
model.
However, the induction of action-orientation does not seem to have been
successful. Although subjects in that condition did not have significant perfor-
mance deficits compared to the control group, they did not show the expected

Pretreatment:
uncontrollable
8 failure
2 7 [ |control
OSG
Ne
mes
Se
cc
ie
os
2.3
xo
See?
E

=° 1

' [e) riented action-oriented no induction

Ho ne ) Ont | ike OF Y © fete) Of

SQ
ao
(Sa)
Oo

at WwW
task
test

Mean
of
expectancy
success

stateworientad action-oriented no induction

Tnyapees Von im. EY Te Qin ae ©) i)

FIG. 7.6. Mean expectancy of success ratings, regarding training and test task,
obtained following pretreatment as a function of type of intervention (SO = state-
oriented, AO = action-oriented, C = no intervention) and type of pretreatment (S
= unsolvable task, C = ‘‘just looking at the tasks’’).
230 KUHL

—— Unsolvable SS
ode Control _Tetreatment

FIG. 7.7. Performance (number of symbols

CHECKED
SYMBOLS
OF
NUMBER
checked) at the test task as a function of pretreat-
so AO G ment and intervention. (SO = state-orientation;
INTERVENTION AO = action-orientation; C = no intervention).

reversal of performance deficits compared to the state-oriented failure-group.

The reason for this may be that, in order to prevent state-orientation, it may not
be sufficient to distract attention from the failure experience by having the
subjects read a nonachievement related story. A second experiment was designed
to study the effects of another procedure, aimed at the induction of action-
orientation.
Three groups were run in this experiment. Besides the usual no-failure control
group, there were two failure groups. In one of them, subjects were instructed,
prior to the training phase, to explicitly state the hypothesis they were testing.
The other group received the usual instructions suggesting implicit hypothesis
testing. The experiment also included a measure of the assumed personality
determinant of generalized helplessness effects. Each group was split in two
subgroups, according to the median of the scores on the action control scale. It
was hypothesized that performance deficits should be most pronounced in the
state-oriented failure group involving implicit testing of hypotheses. A reversal
of this helplessness effect was expected in the state-oriented failure group that
was instructed to explicitly state their hypothesis on each trial. Explicit hypoth-
esis testing was expected to prevent state-cognitions from occurring; it should
help subjects maintain a systematic problem-solving approach, which should
prevent them from engaging in state-oriented thinking. For action-oriented sub-
jects, no such difference between explicit and implicit hypothesis testing was
expected. Instead, action-oriented subjects were expected to show facilitation of
performance compared to baseline performance.
The results (Fig. 7.8) were in line with most of the predictions; state-oriented
subjects from the condition involving implicit testing performed significantly
worse than state-oriented subjects from the condition involving explicit testing.
Within the control group, action-oriented subjects performed significantly better
 7. METACOGNITIVE MEDIATION 231

than state-oriented subjects. Therefore, baseline performance was not computed

from this control group, but rather from control groups taken from two other
experiments. In those experiments, within the control groups there was no dif-
ference in performance between state-oriented and action-oriented subjects.
Compared to this baseline, state-oriented subjects, who had stated their hypoth-
eses, explicitly showed a significant facilitation effect; the performance deficit
found in state-oriented subjects from the implicit-testing condition approached
Statistical significance.
Analysis of expectancy and controllability ratings obtained after training,
revealed that those performance effects cannot be attributed to respective
changes regarding expectancy of success or perceived controllability (Fig. 7.9).
Although the two measures regarding the training task dropped in the two failure
groups, no such drop was found in ratings of expectancy and controllability of
the test task. Also, performance deficits cannot be attributed to a motivational
deficit; there were no significant differences between the groups with regard to
perceived importance of good performance or reported effort expenditure at the
test task.
In summary, these experiments suggest that performance deficits following
failure were not associated with a belief in uncontrollability, generalized from
the training task to the test task; in other words, people did not seem to become
helpless on the test task because they felt helpless when they began working on
that task. Subjects’ reports did not indicate any motivational deficit regarding the
test task that might have resulted from a generalized belief in a loss of control.
Thus, researchers are faced with a problem that cannot be solved on the basis of
an expectancy-value theory of motivation. Performance deficits were obtained

FIG. 7.8. Performance (number of symbols

checked) as a function of disposition for ac-
tion-orientation (AO), state-orientation (SO), SYMBOLS
OF
NUMBER
CHECKED
and type of pretreatment (EHT = explicit hy-
pothesis testing and incontingent failure; IHT
= implicit hypothesis testing and incontingent EHT IHT ®
failure;C = no-failure control). TREATMENT
232 KUHL

without respective deficits regarding expectancy or value of a good performance.

An attempt has been made to formulate a theory of action-control that proposes
an extension of expectancy-value theory. According to this extended theory, a
discrepancy between motivation and performance is expected when the subject is
state-oriented. The experiments show that significant performance deficits could
be found only when there was reason to believe that subjects were state-oriented,

Pretreatment:
ee failure and explicit hypothesis testing

FA failure and implicit hypothesis testing

wi] no failure control

@pyy
dee
I

(oy
Sa
O60
RR
°, ©$25
Mean
of
expectancy
N
task
training
at
success THN
CHS
o>
?,
IKÖL
0%$35
N

state-orientation action-orientation
aeOd Comme Cat, d..s.B
oO, S rom

an)

WwW

N
7
R
_
Mean
expectancy
of at
test
task
success 7
<>

state-orientation action-orientation
Tey. Dee) Gat SEAiL Out
FIG. 7.9. Mean expectancy-of-success ratings regarding training and test task
obtained following pretreatment as a function of pretreatment (EHT = incon-
tingent failure and explicit hypothesis testing; IHT = incontingent failure and
implicit hypothesis testing (= no-failure control) and state vs. action-orientation
(SO vs. AO).
 7. METACOGNITIVE MEDIATION 233

either as a result of experimental induction of or as a result of a disposition

toward state-orientation.
These results have important implications for applied purposes. Take, for
instance, underachievement or depression. To the extent that such disturbances
are attributable to functional helplessness, rather than to a motivational deficit,
an attributional intervention may not be the most effective strategy. An under-
achieving child or a depressive person may show impaired performance in a
variety of situations because he or she is too much concerned with explaining the
problem. An attributional therapy that reinforces this state-oriented attitude to-
ward the problem may even aggravate the disorder. An alternative approach may
focus on helping the client shift his or her attention away from the possible
causes of the problem, which may never be found with absolute certainty.
Instead the client should be encouraged to think about possible ways of changing
the situation; giving up unattainable goals and focusing on new attainable goals.
The client may learn that thinking about change is more helpful than ruminating
about the present state and its causes.
One final remark regarding the fruitfulness of the helplessness paradigm. The
demonstration of the transfer of performance deficits following failure across
considerably different tasks is of great theoretical significance. It calls attention
to a problem that has almost been forgotten since the time of Ach’s analysis of
the “efficiency of the will.’’ It may appear ironic that the transfer of helplessness
effects has not been discovered within the mainstream of traditional research on
achievement motivation. The discovery of generalized performance deficits fol-
lowing failure is a very important one, even though, or maybe because, the
proposed theoretical explanation may not be sufficient. The authors of the re-
vised theory of helplessness (Abramson et al., 1978) are aware of the fact that
helplessness phenomena cannot be fully explained on the basis of traditional
theories of motivation; they do not attempt to remedy this problem. This author
does not believe that the theory of action-control answers all the open questions;
it raises many questions to be answered in the future. This may, at least, prevent
researchers from getting caught in the present state of motivation theory, domi-
nated by expectancy-value conceptions. The remedy that is suggested to prevent
researchers from turning ‘‘state-oriented’’ is: Think of alternatives to the tradi-
tional state of the theory that suggests that people are helpless because they feel
helpless, it may be that they feel helpless because they are helpless.

REFERENCES
Abramson, L. Y., Seligman, M. E. P., & Teasdale, J. D. (1978). Learned helplessness in humans:
Critique and reformulation. Journal of Abnormal Psychology, 87, 49-74.
Ach, N. (1910). Uber den Willensakt und das Temperament. Leipzig: Quelle und Meyer.
Ach, N. (1937). Zur neueren Willenslehre. In Bericht über den 15 Kongreß der Deutschen
Gesellschaft für Psychologie in Jena, 1936 (pp. 125-156). Jena: Fischer.
Atkinson, J. W. (Ed.). (1958). Motives in fantasy, action, and society. Princeton, NJ: Van
Nostrand.
 234 KUHL

Atkinson, J. W. (1964). An introduction to motivation. Princeton, NJ: Van Nostrand.

Atkinson, J. W., & Birch, D. (1974). The dynamics of achievement-oriented activity. In J. W.
Atkinson & J. O. Raynor (Eds.), Motivation and achievement (pp. 271—326). Washington, DC:
Winston.
Atkinson, J. W., & Feather, N. T. (1966). A theory of achievement motivation. New York: Wiley.
Bandura, A. (1977). Self-efficacy: Toward a unifying theory of behavioral change. Psychological
Review, 54, 191-215.
Brickenkamp, R. (1962). Test d-2. Göttingen, FRG: Hogrefe.
Diener, C. I., & Dweck, C. S. (1978). Analysis of learned helplessness: Continuous changes in
performance, strategy and achievement cognitions following failure. Journal of Personality and
Social Psychology, 36, 451-462.
Hillgruber, A. (1912). Fortlaufende Arbeit und Willensbetätigung. Untersuchungen zur Psychol-
ogie und Philosophie, 1(6).
Hiroto, D. S., & Seligman, M. E. P. (1975). Generality of learned helplessness in man. Journal of
Personality and Social Psychology, 21, 311-327.
Kuhl, J. (1977). Meß- und prozeßtheoretische Analysen einiger Person und Situationsparameter
der Leistungsmotivation. Bonn: Bouvier.
Kuhl, J. (1981). Motivational and functional helplessness: The moderating effect of state versus
action orientation. Journal of Personality and Social Psychology, 40, 155-170.
Kuhl, J. (1982). Handlungskontrolle als metakognitiver vermittler zwischen Intention und Handeln:
Freizeitaktivitäten bei hauptschülern. Zeitschrift für Entwicklungspsychologie und Pädagogische
Psychologie, 14, 141-148.
Kuhl, J. (1983). Motivation, Konflikt und Handlungskontrolle. Heidelberg: Springer.
Kuhl, J. (1984a). Motivationstheoretische Aspekte der Depressionsgenese: Der Einfluß der
Lageorientierung auf Schmerzempfinden, Medikamentenkonsum und Handlungskontrolle. In W.
Wolfersdorf, R. Straub, & H. G. Hole (Eds.), Der depressive Kranke der psychiatrischen linik:
Theorie und praxis der diagnostik und therapie. Regensburg, FRG: Roderer.
Kuhl, J. (1984b). Volitional aspects of achievement motivation and learned helplessness: Toward a
comprehensive theory of action control. In B. A. Maher (Ed.), Progress in Experimental Person-
ality Research (Vol. 13, pp. 99-171). New York: Academic Press.
Kuhl, J. (1985). Volitional mediators of cognition-behavior consistency: Self-regulatory processes
and action vs. state orientation. In J. Kuhl & J. Beckmann (Eds.), Action control: From cognition
to behavior. New York: Springer-Verlag.
Kuhl, J., & Blankenship, V. (1979). Behavioral change in a constant environment: Shift to more
difficult tasks with constant probability of success. Journal of Personality and Social Psychology,
37, 549-561.
Kuhl, J., & Blankenship, V. (1979). The dynamic theory of achievement motivation: From
epi-
sodic to dynamic thinking. Psychological Review, 86, 141-151.
Kukla, A. (1972). Foundations of an attributional theory of performance. Psychological Review,
79, 454-470.
Lewin, K. (1935). A dynamic theory of personality. New York: McGraw-Hill.
McClelland, D. C., Clark, R. A., Roby, T. B., & Atkinson, J. W. (1949). The
projective ex-
pression of needs: 4. The effect of need for achievement on thematic apperception.
Journal of
Experimental Psychology, 39, 242-255.
Meyer, W. U. (1973). Leistungsmotiy und Ursachenerklärung von Erfolg und Mißerfolg.
Stuttgart,
FRG: Klett, 1973.
Miller, W. R., & Seligman, M. E. P. (1975). Depression and learned helplessnes
s in man. Journal
of Abnormal Psychology, 84, 228-238.
Mischel, W. (1968). Personality and assessment. New York: Wiley.
Mischel, W. (1973). Toward a cognitive social learning reconceptualization
of personality. Psycho-
logical Review, 80, 252-283,
Pribram, K. H. (1971). Languages of the brain. Englewood Cliffs,
NJ: Prentice-Hall.
 7. METACOGNITIVE MEDIATION 235

Rasch, G. (1960). Probalistic models for some intelligence and attainment tests. Kopenhagen:
Nielson & Lydicke.
Roth, S., & Kubal, L. (1975). Effects of noncontingent reinforcement on tasks of differing impor-
tance: Facilitation and learned helplessness. Journal of Personality and Social Psychology, 32,
680-691.
Seligman, M. E. P. (1975). Helplessness: On depression, development, and death. San Francisco:
Freeman.
Weiner, B., Frieze, H. J., Kukla, A., Reed, L., Rest, S., & Rosenbaum, R.M. (1971). Perceiving
the causes of success and failure. New York: General Learning Press.
Wine, J. (1971). Test anxiety and direction of attention. Psychological Bulletin, 76, 92-104.
Wortmann, C. B., & Brehm, J. W. (1975). Responses to uncontrollable outcomes: An integration
of reactance theory and the learned helplessness model. In L. Berkowitz (Ed.), Advances in
experimental social psychology (Vol. 8, pp. 277-336). New York: Academic Press.
 ~ ewer RT

BE u er
8 or i 77 &

a > oa Seer | De oy 7 5
- 2 A NET ns en A a Aa eee
Dad
eh "
a Pride te ee Er ah u w > a
. 4. > my amply Yee ern
eo. 6 aly = a « Sn zus ii) Nt. 4 Se se An Bes
Pern oo Cm % ») do ©, A ee.
oes. / pow Sm oye ah bas ww = ote Ie
oie = ri 4 te Peerage
=
e u > 5
u ~ ©
: u
7 A os =>
u 5 ~ ae

‘ - até i .

®
| a = u

[22 m a f ws » be A !
@ | f = —_*
u A —_— 2 = » ~ En

os. 7
a =>

rs
tivre
‘bd
 UNDERSTANDING
AND LEARNING

In this section, learning refers to the acquisition of new

knowledge, understanding conceptual relationships, and
problem solving. The chapters by Chi, Glaser and Pel-
legrino, and Greeno and Riley stem from an information
processing perspective that is frequently represented in
terms of computer analogies. The results of theoretical and
empirical work in this field have suggested that two types of
knowledge are necessary for understanding: factual and pro-
cedural knowledge appropriate to a particular content do-
main in which an individual must solve problems and take
action, and metacognitive knowledge concerning how to
direct and organize cognitive activities. Metacognitive
knowledge was discussed earlier by Brown, Flavell, and
Kluwe.
Chi’s discussion is directly related to the chapters in the
first section of this volume. Chi demonstrates the relevance
of domain-related knowledge to developmental research
concerning cognitive abilities, especially metamemory
skills. Using the notion of production rules, Chi provides a
framework for distinguishing among declarative and pro-
cedural ‘‘domain knowledge,’’ “‘strategic knowledge,”’
and ‘‘metacognitive knowledge,’’ which includes pro-
cedural and declarative components as well. Chi discusses
metamemory within this framework, and points out that the
use of the prefix “‘meta’’ may cause confusion. In addition,

237
238 PART Ill: UNDERSTANDING AND LEARNING

she notes that before we can expect insights about metacognitive development,
we require greater knowledge about the development of cognitive strategies.
The chapter by Robert Glaser and James Pellegrino provides a comprehensive
analysis of inductive thinking. They describe a research project designed to allow
a thorough analysis of the abilities that are central to inductive thinking.
Glaser and Pellegrino’s results show that subjects who differ with regard to
the success of their inductive inferences may be distinguished according to three
features: management of memory load, knowledge of task constraints, and the
accuracy of their knowledge representation. These results are used to formulate
guidelines that may be useful for adapting instruction to the cognitive abilities of
the learner, and for improving those cognitive abilities.
The development of a taxonomy of different types of learning is a central
concern of the chapter by Greeno and Riley. They address two concerns: the
extent to which metacognitive knowledge is conscious, and the relationship
between metacognitive knowledge and planning knowledge. Of particular in-
terest is the distinction between explicit and implicit understanding and how this
distinction relates to verbal reports concerning one’s own thinking. The experi-
mental procedures used with children often reveal that they have an implicit,
although not verbalizable understanding of the problems they are given. Under-
standing, according to Greeno and Riley, entails the elaboration of an appropri-
ate internal representation of a problem (what they call a problem schema) and
the deployment of actions adequate for solving the problem (action schema). In
this sense, understanding is an analogue to knowledge as addressed earlier by
Brown and Kluwe. Greeno and Riley provide empirical results that illustrate the
function of an appropriate internal representation to problem solutions.
 Representing Knowledge and
Metaknowledge: Implications
for Interpreting Metamemory
Research

Michelene T. H. Chi
Learning Research and Development Center
University of Pittsburgh

Children’s memory development has traditionally been interpreted in terms of

the development of control processes or strategies. However, experimental re-
sults that purport to show strategic deficiencies in young children often require
implicit or explicit assumptions about the nature of a child’s knowledge base. If
careful consideration of the knowledge base is made, it becomes clear that such
results may be due to an inadequate knowledge representation that may not
permit the application of a mature strategy. This alternative interpretation would
point to the importance of examining the knowledge base in determining memory
performance.
In order to discuss knowledge organization, a representational framework is
needed. Procedural knowledge (general strategies and domain-specific pro-
cedures) may be represented as production rules, declarative knowledge may be
embodied in a node-link network. Such a representation is useful not only for
explaining some of the well known developmental phenomena, but can also be
extended to represent metaknowledge. This extension leads to a taxonomy of
various types of metaknowledge.
By using such a representation and classification of metaknowledge, meta-
memory research can be examined. Such an analysis shows that some meta-
memory research may not tap metaknowledge at all, or that it may not discrimi-
nate between knowledge and metaknowledge. In any case, it is clear that before
progress is made in this area, the concepts of metacognition and metaknowledge
need to be defined and examined more closely.
This chapter is composed of four parts. The first section briefly describes why
a knowledge emphasis is important in cognitive developmental psychology. The
second section postulates a representational framework for the discussion of

239
240 CHI

various forms of knowledge: declarative, procedural, strategic, and meta-

knowledge. Within the framework proposed, the third section evaluates the term
metamemory and related research. The final section concludes with a general
discussion.

KNOWLEDGE APPROACH TO MEMORY

DEVELOPMENT

Within the last decade, there has been a trend in both cognitive psychology and
artificial intelligence research toward exploring the knowledge that must be
possessed by an individual and/or an intelligent system. The new theory of
artificial intelligence sees an intelligent system as one that can use and express
different forms of knowledge. The change in focus, from programming a system
with powerful search heuristics to programming a system to possess a large
quantity of organized knowledge, was necessitated by the parallel findings in
cognitive psychology concerning humans’ dependence on a large and well struc-
tured knowledge base. For example, a chess master’s expertise derives not from
his powerful search heuristics, but rather from the large storage of meaningful
patterns in long-term memory. This large storage of patterns is manifested when
the chess master is able to reproduce a board position after viewing it for only 5
seconds (Chase & Simon, 1973). A computer with powerful search heuristics has
yet to match the chess master’s capabilities in finding good moves; however, the
greater the amount of knowledge that is programmed into the computer, the
closer is its approximation to a human player. (For a more extensive review and
discussion, see Chi, Glaser, & Rees, 1982.)
At about the same time, the idea that knowledge may be a fundamental source
of developmental differences in memory was entertained by Brown (1975), Chi
(1976), Huttenlocher and Burke (1976), and others. To appreciate the distinction
of this outlook, one has to understand, up to that point in time, the theoretical
orientation of developmental psychology. One could say that developmental
psychology in the late 1960’s was influenced predominantly by Piaget’s theory,
and attempts in the American literature were characterized by the question of
stages and transition between stages. The research focused not only on norming
when a child enters each stage, but also on showing how the stage theory is really
transitory. For example, it was discovered that children of younger ages are
really capable if one provides hints, such as attending to the ‘‘relevant’’cues, of
doing tasks at a more sophisticated level, as exemplified by the influential thesis
of Rochel Gelman (1969).

Evidence for Strategy

The onset of the Atkinson and Shiffrin (1968) process model of memory, caused
American developmental psychologists to wonder whether or not the child is
inferior in the control processes used to manipulate information. These control
 8. REPRESENTING KNOWLEDGE AND METAKNOWLEDGE 241

processes are usually called strategies. Rehearsal, labeling, elaboration, and

categorization are a few of the strategies that have been identified. Hence,
research in the first half of the decade of the 1970’s was aimed at demonstrating
the limitations of young children in their use of control processes; in fact, the
demonstration was quite simple, straightforward, and clear. In general, young
children (around the age of 5) were found not to use strategies; the appearance of
active use of strategies was found to develop with age. Finally, even at times
when the routine of the strategy is available (such as knowing how to repeat a
string of items), the strategy does not necessarily mediate performance.
The importance of strategic deficiencies was further confirmed when it was
shown that the lack of a strategy hinders performance, and the possession of a
strategy facilitates performance. For example, the amount of recall has been
shown to correlate directly with the amount of rehearsal. If six and seven-year-
old nonrehearsers were induced to rehearse, their recall would be elevated to the
level of children who normally rehearsed (Keeney, Cannizzo, & Flavell, 1967).
Hence, because most developmentalists implicitly assume that there is a ‘‘suf-
ficiency criterion’’ involved, particularly for simple tasks, there has been no
need to bring in the knowledge base as a potential source for age differences. For
example, for digit recall, as in a digit span task, the assumption is that being able
to identify the digits by name is a sufficient criterion for maximal performance in
that task. This assumption is usually made as a result of generalization from adult
research. In the adult research, one generally assumes that individual differences
are minimal, and that all adults have approximately the same knowledge about
digits; therefore, performance in tasks requiring digits can and should only
reflect other processing differences. However, this implicit assumption simply
cannot be made in developmental research. The following example should illus-
trate the point.
A common task used in developmental research to assess memory perfor-
mance is free recall. Typically, approximately 20 items are presented to the child
to free recall. These 20 items fall into four or five taxonomic categories, such as
Furniture, Clothing, Animals, etc. One common performance measure is the
amount of recall. For children of around age 5, the amount of recall is usually
around four items; for older children, it can vary from 9-12 items, depending on
the age. Eleven-year-olds typically remember at most, eight items (Tenney,
1975).
The interpretation of the performance difference between age groups is that
younger children do not apply efficient organizational strategies. How is such an
interpretation substantiated? By looking at the recall output (that is, the pattern of
response sequence), one can seek some measure of organization, such as the
amount of clustering, subjective organization, chunking, or whatever. Typically,
younger children exhibit less organization in their recall output than older chil-
dren (Moely, 1977). Their organization also tends not to be consistent from trial
to trial. The lack of an orderly and consistent output is interpreted to mean that
young children do not impose an organization on the material to be recalled.
242 ~~ CHI

Inadequate use of organizational processes can also be shown to occur during the
encoding of the stimulus material, independent of a recall task (Ornstein &
Corsale, 1979). To do so, one typically assesses the degree to which children can
sort the presented items into their taxonomic categories prior to recall. The
typical findings mirror those found in the organization of the recall output. That
is, again, younger children’s sorting is less reliable, more random, and shows a
weaker organization; they tend not to put items into the taxonomic categories that
they belong in, and so on.
However, such differences in organization are seldom attributed to a knowl-
edge-based problem (although currently, investigators such as Ornstein & Cor-
sale, 1979, and Lange, 1978, are orienting toward a knowledge interpretation).
Because the child, when asked to do so, can easily identify the categories to
which these items belong, the investigator generally assumes that these items and
their membership in the taxonomic categories are known to the child. For exam-
ple, the child may be asked explicitly to put all the animals together, and he or
she will generally be able to do so. Alternatively, a child can produce a set of
‘core’? members when a category label, such as ‘‘animal,’’ is presented
(Nelson, 1974).
The foregoing analysis illustrates why the interpretation preferred by develop-
mentalists is dictated by the observation of the child’s inadequate use of retrieval,
or encoding strategies, to facilitate recall. Primarily, this is because such inade-
quate strategy usage is accompanied by a seemingly adequate knowledge base.
However, using the given example, it is asserted that, to be able to recognize that
cats and dogs are both animals does not require the same knowledge structure as
to be able to group them together without prompting. Figure 8.1 shows two tree
structures, depicting how that knowledge can be represented. In the first case,
members of the taxonomic categories are represented in the way they are envi-
sioned by the adult experimenter: a hierarchy with class members associated with
their superordinate nodes. In the other case, each instance of the category simply
has attached to it a set of features, one of which is the nature of the item. For
example, cat is seen as an animal, just as cat is also known as having four legs,
being furry, and so on. But somehow, the data base is not organized so that cats
and dogs are embedded under the superordinate Animal node. Such a structure
would produce exactly the results obtained: Children do not spontaneously
organize the items according to the taxonomic categories, but will organize the
items when prompted; this is because they can recognize that a cat is an animal
and a dog is an animal. Therefore, the interpretation presented is that the devel-
opmental findings reflect, in part, the organization of knowledge, rather than
strictly the application and use of strategies. A test of this interpretation would be
to first ask each individual child to produce his or her ‘‘core’’ category members,
and then to ask each child to sort or recall them. Presumably, these ‘‘core’’
members would be represented as a group, and organization during input and
output should be facilitated by the existing cohesive structure in memory. Some
 8. REPRESENTING KNOWLEDGE AND METAKNOWLEDGE 243

Animal

FIG. 8.1. Two possible semantic structures for concepts in memory.

evidence of this interaction of the structure of semantic memory and organization

is provided in Chi (1985).
There are additional reasons for taking a knowledge approach. The most
blatant one is that there are times when age differences in memory performance
are not accompanied by strategy differences, suggesting that some other sources,
such as knowledge, must have produced the difference. One kind of evidence is
provided by the study of Huttenlocher and Burke (1976); they specifically pre-
sented the stimulus items in a memory span task so as to induce grouping. They
reasoned that because adults grouped digits in three’s, children might benefit
from such grouping; therefore, to facilitate encoding, they presented the digit
244 CHI

string in groups of three. They found that children between the ages of 4 and 11,
all benefited from such grouping; however, differences were still maintained
across the age groups, suggesting that it is not a simple matter of a difference in
the use of grouping strategy.
A similar type of evidence can be found in training studies. That is, if training
in the use of strategies is provided for children of both age groups, both groups
will improve their performance so that the initial age difference is still main-
tained (e.g., Storm, 1978). A third kind of evidence concerns studies where very
young children (in the 2-4 age range) are tested; it is assumed that children in
this age range do not actively use strategies (Myers & Perlmutter, 1978), yet
their memory recall does improve with age. Although each of these three kinds
of evidence may be subjected to alternate interpretation, taken as a whole, they
suggest that age differences in memory performance are not always accompanied
by observable strategy differences. Instead, an alternative source of develop-
ment, knowledge, must at times be taken into consideration. (For an extensive
review of this interpretation, see Chi & Ceci, in press).

Evidence for Knowledge

Beyond the studies cited, there is actually direct support for the notion that
knowledge may be an important component of memory development. One way
to test this hypothesis is to seek evidence where age differences are minimized
because knowledge differences are reduced. Two related experiments illustrate
that approach. Dempster (1978) showed that when one controls for stimulus
familiarity in a memory span task, by using consonant letter strings that have
little structural similarity to English, age differences disappear between first and
sixth graders. Similarly, Boswell (1974) manipulated the letter strings for a span
task with very brief exposure durations, and found that the further away the letter
Strings were in their approximation to English, the more reduced the age dif-
ferences were (between the second graders and adults).
Another set of studies also implicated the role of knowledge in affecting
children’s performance. Richman, Nida and Pittman (1976) found that in a
repeated free recall learning task, older children generally learned more rapidly
than younger children. However, when words were used so that the mean-
ingfulness values were held constant across grades, between-grade learning dif-
ferences were minimal. Holding the meaningfulness constant implies that there
were no knowledge differences; therefore, age differences in learning perfor-
mance were minimized. Similarly, when the amount of semantic knowledge
children had of words in a cued recall task was controlled, no differences in
recall scores were observed between 7 and 12-year-olds (Ceci & Howe, 1978).
Both of these studies further point to the importance of having the appropriate
semantic knowledge related to the words used in the experiment. To a large
extent, the existence of the relevant knowledge determined whether or not age
differences were found.
 8. REPRESENTING KNOWLEDGE AND METAKNOWLEDGE 245

To further assess the importance of the role that knowledge plays in children’s
memory performance, one could presumably find situations where children actu-
ally have more domain knowledge than adults. If domain knowledge is an
important source of development, perhaps a reversal of the age trend could be
obtained. Precisely this manipulation was used by Chi (1978) in comparing the
recall of chess positions by 10-year-old children who had greater chess knowl-
edge than adults. A cross-over effect was obtained; the children could recall a
greater number of chess pieces than adults, yet at the same time, they could recall
fewer digits than the adults. The interpretation is that the results reflect the
children’s greater knowledge of chess, but more limited knowledge of digits, as
compared to the adults. This finding has been replicated in other laboratories.
For example, using words from the Battig and Montague norms for adults, and a
children-generated list of cartoon names and games for children, the same cross-
over effect in recall was obtained by Lindberg (1980). The results suggest that
having domain-related knowledge can actually overcome children’s limitation in
the use of strategies.
This section began by questioning whether the role of strategy usage is all-
encompassing in determining age differences in memory and cognitive perfor-
mance. Evidence for the possibility of another source of development—knowl-
edge—was proposed, by citing studies whose findings cannot be accounted for
by postulating an improvement in the use of strategies. Finally, more recent
evidence was provided which points directly to the importance of the role that
knowledge plays, and further, shows that the often observed traditional age trend
(performance improvement with age) can sometimes be reduced, eliminated, or
even reversed. This suggests that to completely understand age effects, one must
necessarily examine the knowledge differences between age groups. In the next
section, what it means to have more knowledge, that is, to have the relevant
knowledge that is organized and accessible, is examined more microscopically.

A FRAMEWORK FOR REPRESENTING KNOWLEDGE

In the previous section, the terms strategy and knowledge were loosely defined
and used dichotomously. The purpose of this section is to elaborate and describe,
in one possible way, the different forms that knowledge can take, as well as how
the concept of metaknowledge applies. The author borrows heavily from the
work of Anderson (1976), Greeno (1978), and Norman and Rumelhart (1975);
however, any deviations and modifications of their work are the results of intui-
tion and preconceived biases, suitable for explaining developmental findings.

Domain Knowledge
Domain knowledge can take two forms: declarative; and procedural. Declarative
knowledge is factual in nature, such as knowing about animals, knowing about
number facts, and knowing about bicycles. For example, knowing that a dog is
246 CHI

can be
an animal is declarative knowledge. Furthermore, declarative knowledge
e by the way it is represent ed. De-
distinguished from procedural knowledg
clarative knowledge can be represented in terms of a semantic propositi onal
network, where a concept (such as a dog) may be represented as a node, and
links specify the relationships among the nodes. Hence, the proposition that “‘a
dog is an animal’’ could be represented by two nodes, with an isa link between
them (see Fig. 8.2). A child may also know that a dog has four legs, and so on;
these pieces of information can simply be represented as additional related nodes
linked to the ‘‘dog’’ node. The degree of complexity of the semantic network
should correspond to the amount, elaborateness, and organization of a child’s
declarative knowledge (see Chi & Koeske, 1983).
Procedural knowledge is knowledge about how to do things. For example,
knowing how to ride a bicycle, or knowing how to do multiplication, would be

Meta-Strategies Procedural Knowledge

P1: IF P2 is used rash. WE the goal is to memorize and there

THEN study fewer trials is a string of digits
THEN repeat the string of digits

Strategies [tle ile the goal is to memorize and there

is a string of words
P2: IF the goal is to memorize and THEN repeat the string of words
there is a string of items
THEN repeat the string of items PS:£IE the goal is to memorize and there
is a string of letters
THEN repeat the string of letters

Declarative Knowledge PG: IF

there is a string of digits
THEN repeat the string of digits

Pre Ne the goal is to memorize

THEN repeat the string of items

PS IE
THEN repeat the string of items

BIZZIE the goal is to evaluate the

available knowledge about
animals
THEN activate animal nodes and
quantify how much knowledge
there is

FIG. 8.2. Representations for declarative, procedural, strategic, and meta-strate-

gic knowledge.
 8. REPRESENTING KNOWLEDGE AND METAKNOWLEDGE 247

considered procedural knowledge. Furthermore, procedural knowledge can be

represented as a set of production rules, which are condition-action pairs (New-
ell, 1973). The condition of a production rule specifies a feature or set of features
that must match either the content of working memory (such as stimulus inputs
that are stored there temporarily) or the structure of the activated portion of the
semantic network. Hence, in a way, the condition side of a production rule takes
as argument the structure of the declarative knowledge, and the action side
constitutes procedures that either modify or add to the semantic structure, or
manipulate an external environment. For example, Fig. 8.2 lists several different
productions (P3, P4, PS) that may be analogous to the procedures of rehearsing a
set of digits or a string of words.
Although there is an active debate concerning whether or not there should be a
distinction in representation between declarative and procedural knowledge (see
Anderson, 1976; Winograd, 1975), for developmental purposes, it is useful to
make a fundamental distinction between the two. Procedural knowledge and
declarative knowledge are also often considered to be domain-specific because
they generally refer to a domain specific knowledge. For example, declarative
knowledge can also be general world knowledge; in some sense, all knowledge is
classifiable into some type of domain. However, a more important distinction,
although still rather arbitrary, is the difference between procedural knowledge
and strategies.

Strategies
Strategies, as the concept is used in developmental research, seem to have a
special status; they have the property of being general and global, rather than
domain-specific. To capture that property, here a strategy is defined simply as a
general procedure, in the sense of application to several domains. P2 in Fig. 8.2
is an example of a strategy. In this representation, a strategy has four properties,
some of which have been implicitly assumed in the literature. First, and most
important, a strategy is general and domain-independent. In the example pro-
vided by P2, that production should be used whether the “‘items’’ are digits,
words, or letters. Second, a strategy as do all procedures, has a goal. Third, a
strategy as do all procedures, can have several components; that is, the action of
“‘repeat’’ can itself be a call to other procedures. The final property of strategies,
as distinct from procedural knowledge, is that the total number of strategies in
memory must be finite, and in fact, should be relatively small. That is, the total
number of rules, such as P2, should be fewer than the total number of rules, such
as P3—P9. This last property is an implicit assumption made by most researchers
(see Hayes, 1985), and necessarily underlies the philosophy of training studies.
If one wishes to distinguish between procedural knowledge and strategies the
finiteness assumption is also necessary.
The operational definition given to strategies, has three implications. The first
is that strategies do have many of the properties of procedural knowledge; basi-
248 CHI

cally, they are represented in the same way, as production rules. Second, be-
cause the difficulty of separating general from domain-specific declarative
knowledge has already been discussed, strategy is clearly artificially distinct
from procedural knowledge, in the sense that it is defined to be general rather
than domain-specific. For the time being, however, because it is possible to
artificially segregate the two, the distinction might serve a useful purpose. Final-
ly, even with the simplistic and global representation described one can already
begin to offer explanations for a variety of developmental phenomena in the
context of this representation. One common and most serious finding is that for
the purpose of generalization, strategy training has been unsuccessful. Belmont
and Butterfield (1977) have reviewed a number of training studies and found
only a few to have succeeded in obtaining generalization from one task situation
to another. It is conjectured that one possible reason for the failure to transfer,
is that perhaps what was taught was a procedure, and not a strategy; that
is, a domain-specific procedure, such as P3 or P4, may have been taught, and
what was required to generalize would be a production rule, such as P2 (see Fig.
822):
Another common finding is that younger children can often learn to use a
strategy, but unless explicitly told to use it, will not apply the strategy. This can
be interpreted to mean that perhaps during instruction the appropriate goal was
not properly encoded. That is, with the goal missing (in the case of P6), a child
will not know that the rehearsal procedure should be activated, when the task
instruction to memorize is presented. (This is one possible interpretation for
production deficiency.)
A third common finding is that whenever an instructional approach is used to
teach or induce the use of strategies, older children often will learn and adopt it
more readily than younger children. One possible explanation for this finding,
according to the framework used, is that the older child may already have a
number of specific versions of a production (corresponding to different domains,
such as P3, P4, and P5) available in memory, so that acquisition of the general
form of a strategy production (P2) would be easier. For example, the older child
may already know how to repeat a string of words, a string of letters, etc., when
he or she sees one. Therefore, it is easier to acquire a new production because he
or she already has in store all the other approximations to the final general form
of the production. Although how a learning mechanism operates has not been
postulated, one can clearly accept the proposition that a new production can be
learned faster and easier if there are others like it already in memory. For
example, one mechanism that comes to mind, is learning by analogy (Rumelhart
& Norman, 1980), another mechanism is generalization (Anderson, Kline, &
Beasley, 1978).
Hence, the foregoing interpretation, although speculative, suggests that with a
concrete representation of what different forms of knowledge look like, one may
begin to understand developmental performance.
 8. REPRESENTING KNOWLEDGE AND METAKNOWLEDGE 249

Metaknowledge
What is metaknowledge in the context of the present representational frame-
work? The term meta can be used in two ways; the straightforward way is to
interpret meta as a reference to cognition. Thus, metaknowledge is knowledge of
cognition or memory. In this sense, one could substitute the word meta by the
word cognition; therefore, metaknowledge is cognitive knowledge, and meta-
memory would be memory knowledge. A more complex interpretation is to
consider the term meta as a reference to second-order knowledge. Thus, it is
somewhat analogous to the concept of second-order function, as in second-order
derivatives in calculus. With this conception, because there are several forms of
knowledge, there necessarily must be several varieties of metaknowledge. An
examination of what representation the different varieties of second-order knowl-
edge might take, as well as the more direct and simplistic view of metaknow-
ledge, follows. Knowing the precise representation should elucidate the possible
acquisition processes that develop with age and learning.

Meta-declarative Knowledge. There might be meta-declarative knowledge,

for instance, that would be knowing what I know about animals. How would one
represent meta-declarative knowledge? There are two ways. One way is to treat
the term meta as a reference to cognitive knowledge; meta-declarative knowl-
edge would be factual knowledge about cognition, much like factual knowledge
about any other domain, such as animals. Hence, I could have stored in memory
that I am a chess expert, and therefore I know a great deal about chess; similarly,
that I am a cognitive psychologist and I know a great deal about cognition and
memory. In both cases, these are prestored declarative knowledge; the latter
example illustrates metaknowledge only in the sense that the content of the
knowledge is about cognition.
Alternatively, if the necessary knowledge about cognition, or any other do-
main, is not there, then a procedure is needed to assess what and how much one
knows about a domain (whether the domain is cognition or animals). In this
sense, the term meta refers to second-order knowledge, or a function. In repre-
sentation, meta-declarative knowledge would be the same thing as a procedure or
a strategy; it would be a rule that takes as inputs (the arguments) the declarative
knowledge, and produces some action on the inputs, such as some kind of
evaluation. Hence, theoretically, meta-declarative knowledge would take the
same form as either procedural knowledge or strategies. What appears to be
awareness of one’s declarative knowledge may be nothing more than an activa-
tion of the declarative knowledge in the context of the experimenter’s or self-
inquiry. For example, if the experimenter asks me if Iknow about animals, or
how much I know about animals, that would constitute meta-questions. To
answer the first type of question one needs to search or activate the relevant
nodes in semantic network, and answer when the available structure is activated.
250 CHI

If the question requires an evaluation of how much knowledge, then in addition,

some kind of assessment of quantity has to be made. The complexity of the
processes needed to answer meta-declarative type of questions depends on
whether or not the question addresses a prestored declarative knowledge, or if a
procedure is needed to assess existing declarative knowledge about cognition.
To recapitulate, according to the formulation, the kind of knowledge used for
answering meta-type questions depends entirely on the questions the experiment-
er posed, and whether or not the specific answer is already prestored as a
proposition. Thus, if the experimenter asked ‘‘What do you know about ani-
mals?’’ the child can answer simply by activating the animal nodes, and assum-
ing that the activated portion of the semantic network is the content of working
memory, give a core-dump of the contents of a working memory. Likewise, if
the experimenter asks ‘‘Are you a good chess player?’’ the response may also be
a prestored proposition—a simple declarative statement that can be retrieved.
But if the experimenter asks, ‘‘How much do you know about animals?’’ it may
require not only the activation of the animal nodes, but also some evaluation and
quantification of the activated portion of the semantic network. In Figure 8.2 P9
may be an example of such a procedure. For developmental interests, the ques-
tion boils down to whether or not children possess the additional evaluation
function, such as P9; the answer is probably not. They probably acquire them
with development, in the same manner that other complex procedures are ac-
quired, including quantification processes.
In sum, it has been proposed that meta-declarative knowledge can be of two
kinds. First, what is often meant by metaknowledge (or knowledge about cogni-
tion) may be nothing more than prestored declarative knowledge. For example,
knowing what task and strategy variables influence memory performance may be
the same kind of knowledge as knowing what climate an animal prefers to live
in. They are both declarative knowledge; it is only meta when one is discussing
cognitive knowledge. The second kind of metaknowledge uses the term meta in
the sense of a function or second-order operation. A function or a procedure is
used when an evaluation or any other form of action is taken on the existing
declarative knowledge. Therefore, in order to answer a question about some
content domain, whether it is memory or animals, an evaluation procedure is
needed; this usually occurs if a response does not already exist in memory.

Meta-strategies. What about meta-strategies? Meta-strategies would be like

knowing that my rehearsal production is efficient. Again, there are two ways to
represent meta-strategic knowledge. One way is direct prestored declarative
knowledge. For example, to answer the question ‘‘Do you remember better than
your friends?’ (Kreutzer, Leonard, & Flavell, 1975) may consist of nothing
more than retrieval of past encodings of those remembering events. In other
words, every time a situation arises in which my memory is pitted against my
friends’ memories, I might encode, on a relative scale, how good my remember-
 8. REPRESENTING KNOWLEDGE AND METAKNOWLEDGE 251

ing skills are. In this sense, the term meta is used only to refer to knowledge
about strategies or activities of remembering.
Alternatively, one can also view meta-strategic knowledge as second-order
operations. In this case, meta-strategy would be a rule that evalutes another rule.
Meta-strategies would take the entire strategy production as inputs and output
some evaluation of the production (see Pl, in Fig. 8.2). For instance, meta-
strategies may be needed to answer the same question: ‘‘Do you remember better
than your friends?’’ if the answer to such a question is not already prestored. In
which case the response requires an assessment of what kind of remembering
Strategies are available to oneself, and how well those strategies facilitate re-
membering. The assessment is quite complex, requiring two evaluations: one at
the level of the output of each individual memory strategy; the other at the level
of the sum of all the available strategies for remembering.
Hence, in the present conceptualization, meta-strategic knowledge can take
two forms: a simple interpretation is that it is declarative knowledge about
procedures of remembering; a complex interpretation is that it is a rule evaluating
another rule. In the latter sense, meta-strategies are distinctly different from
strategies, and may be conceived of as having special status (Chi, 1981). Howev-
er, because meta-strategies can themselves be represented as production rules,
they are not different in form or quality from other production rules, such as
strategies and procedural knowledge. Hence, there is agreement with Flavell’s
(1979) assumption that “‘metacognitive knowledge and metacognitive experi-
ences differ from other kinds only in their content and function, not in their form
or quality’’ (p. 906).

Meta-procedural Knowledge. Meta-procedural knowledge would be very

much the same as meta-strategies, with the exception that the inputs to the meta
rules would be procedural knowledge rather than general strategies; they both
take the same form. Meta-procedural knowledge would be rules that evaluate
other domain-specific rules. For example, the processes of predicting how well
one can come up with a good chess move could be considered meta-procedural
knowledge; whereas, the processes of assessing how well one can rehearse could
be considered meta-strategy.

Summary and Discussion

The proposition of this section is that if there are different forms of knowledge
(declarative, procedural, and strategic), then there must necessarily be different
types of metaknowledge (such as meta-declarative, meta-procedural, and meta-
strategic). Possible representations for these different types of metaknowledge
were discussed. Of the three types, meta-strategy and meta-procedural knowl-
edge take a slightly different representation from the other knowledge: It is a rule
of a rule.
252 ~—sCHI

It is suggested that the term meta can be used in two contexts. In one sense,
the term meta may refer only to knowledge that is cognitive in nature; meta-
declarative knowledge would refer to declarative knowledge about memory
facts. Meta-strategic knowledge would refer to declarative knowledge about
memory processes, much like what Flavell and Wellman (1977) refer to as
knowledge about strategy variables. In another sense, the term meta may refer to
second-order knowledge. In this case, meta-declarative knowledge would be a
procedure that takes as input declarative knowledge, and the action would be an
evaluation or some kind of assessment. Similarly, meta-strategic knowledge
would be a rule that takes as input another rule, and output some appropriate
evaluation or action.
Postulating the different possible representations of metaknowledge is useful,
not only for the purpose of understanding which category of knowledge a given
meta-phenomenon might fit, but also, it will help to understand how the mecha-
nism for learning and acquisition of that knowledge might develop. In the next
section, the different varieties of meta-phenomena are discussed, and the catego-
ry of metaknowledge each fits is explored.
So far, the different types of knowledge, which refers to information that is
stored and can be retrieved from memory, have been discussed. In this context,
processes are the execution of a sequence of procedures, which may be produc-
tion rules of any of the varieties that have been mentioned—procedural knowl-
edge, strategic knowledge, or meta-strategies. Beyond these rules, there should
be a set of executive rules, controlling the entire system’s on going activities
(whether the system is a child, a program, or an adult). Executive rules are
analogous to an interpreter that determines the sequence of processes to be
executed (Newell & Simon, 1972). A variety of processes, such as monitoring
one’s own state of processing and allocating ongoing attention, has now been
given a new status of meta-processes. Evidence that shows that these meta-
processes also develop with age, is now accumulating. Why these executive
processes have been given a new name, other than the possibility that they
control memory processes, is unclear. This is discussed later, in greater detail.

METAMEMORY RESEARCH

In this section, selected aspects of metamemory research in the literature, and

how they can be conceived in the theoretical framework proposed in the second
section of this paper, are summarized. It is difficult to integrate the large varieties
of this research, therefore, only a small sample is discussed: namely, meta-
knowledge of person variables; sensitivity; monitoring, and checking. This is a
somewhat random selection, simply because it is nearly impossible to be ex-
haustive for a paper of this length. The selected examples serve to illustrate the
complexity and fuzziness of some of the metamemory research.
 8. REPRESENTING KNOWLEDGE AND METAKNOWLEDGE 253

Metaknowledge of Person Variables

Flavell (1979) and Flavell and Wellman (1977) have proposed at least three
categories of metaknowledge that one can acquire about one’s ability to memo-
rize. One of them, known as knowledge of person variables, is a person’s
knowledge about intrinsic and stable characteristics about him or herself and
others. For example, a child’s knowing that he or she ‘‘can learn things better by
listening than by reading,’’ or that one of his or her friends is ‘‘more socially
sensitive than another,’’ (Flavell, 1979, p. 907), would be considered knowledge
of person variables. This knowledge is meta because it concerns knowledge
about one’s cognitive functions, such as memorization. Younger children have
been shown to possess less accurate metaknowledge of person variables than
older children. One such demonstration is to show that children are unrealistic
when it comes to estimating their own memory performance, such as in digit
span task, or other types of serial recall tasks. For example, comparing a child’s
prediction of how many digits he or she can recall with how many he or she can
actually recall, shows that young children are less accurate than older children in
their estimation (Flavell, Friedrichs, & Hoyt, 1970; Yussen & Levy, 1975).
According to the conceptual framework used, the examples of meta-
knowledge of person variables illustrate the two kinds of meta-declarative
knowledge discussed in the previous section. In the one case, asking a child
whether he or she can learn better by listening than by reading could be consid-
ered declarative knowledge, that is meta only in so far as it refers to cognitive
abilities. It is very much like one’s knowledge about any other domain, such as
the eyes and how they see, and the nose and how it smells. Similarly, Johnson
and Wellman’s (1982) research on what the child knows about the brain, is also
declarative knowledge of this variety. The reason it is classified as meta is
because it is about cognition; however, the form of the knowledge is no different
than one’s knowledge of the eyes or any other domain of knowledge, such as
animals.
The knowledge of person variables is declarative because, as Flavell and
Wellman (1977) have already noted, a large proportion of this person variable
knowledge is acquired through experience. The child could have encoded from
perceptions of past experiences whether he or she is better at listening than
reading. For example, after the first encounter of this kind, the child may decide
that he or she is relatively better at listening than reading (let’s say 3 on a 5 point
scale). The next occurrence of such an event could either reaffirm the initial
impression or change the evaluation so that now the child thinks he or she is
considerably better at listening than at reading (let’s say the rating has changed
from a 3 to a4). The mechanism that allows a child to decide what he or she can
learn better by listening than by reading could be the retrieval of this single
proposition, encoded as a result of past experiences in which listening and
reading outcomes were compared. In this sense, whether or not an age effect is
 254 CHI

manifested probably depends on a number of factors, such as: (a) the total
amount of experiences of this kind; (b) whether or not greater amounts of experi-
ence would produce a more accurate and robust assessment; and (c) the con-
sistency of the early encounters. For example, if a minimal amount of experience
is sufficient to produce a robust assessment, then a stable knowledge of person
variables could probably be reached early in life; consequently no age effects will
be observed. Likewise, if the first few encounters are relatively consistent, that
factor may also lead to the establishment of an early stable assessment. All these
factors, prior to the event when the meta-question was asked, may contribute
toward the mechanism by which a decision is made and prestored.
Hence, from this analysis, the interpretation of the results from such studies
depends on how knowledge of person variables was derived by the child from
past experiences. The multitude of factors affecting that decision (prior to an-
swering the meta-question) may contribute to the mixed age effects obtained in
the literature, regarding why some aspects of metaknowledge are available to a
young child and others are not.
A second way of deciding if one could learn better by listening than by
reading, if such knowledge was not prestored and known, would be to evaluate
and compare the outcome of reading and listening processes, and then decide
which one is more optimal. This strategy of responding would require several
meta-strategic rules, such as evaluating the outcome of comprehension processes
or the amount of comprehension in listening versus reading, as well as the
complexities of decoding in listening and reading. Such a decision requires more
processing than a retrieval of a stored proposition. It reflects meta-declarative
knowledge of the second kind; a procedure or several procedures are needed to
evaluate either declarative knowledge or other procedural or strategic knowl-
edge. In sum, it is proposed that to answer a simple meta-question, such as the
one posed about listening and reading, either a simple retrieval of previously
encoded proposition can occur, or a complex set of decision processes can occur.
In the former case, the process of retrieval itself may be simple; however, the
processes leading to the formation of the prestored decision may be quite com-
plex. In the latter case, the decision processes themselves are complex. In either
case, the complexity may contribute to the mixed results concerning age changes
in metaknowledge of person variables.
One aspect of metaknowledge of person variables that shows a clear develop-
mental trend, is the memory estimation paradigm. Because of the potentiall
y
complex decisions required of the task, the interpretation from these
results—
younger children have poorer metaknowledge—may be premature. It
is conceiv-
able that the developmental trend reflects the same kind of difficulti
es that
younger children encounter in any complex task. There may
not be any
uniqueness to the ‘‘meta-ness’’ of the task, nor to the interpreta
tion of the
results. Consider the processes involved in making a decision of
the kind usually
requested in an estimation task. Young children are presented
with a stimulus
 8. REPRESENTING KNOWLEDGE AND METAKNOWLEDGE 255

array of n items. After a brief exposure, they are asked to predict how many of
the n items they think they can recall in correct serial order. There are at least two
ways to answer such a question. One way is again the simple direct retrieval
method, which relies on past experiences; that is, if you remember that you
usually recall four digits, then you report four. Even though this explanation has
been dismissed by findings that show that young children, in view of obvious
failures, will not change their outlandish answers (Yussen & Eevy, 1975). one
could argue that temporary short-term feedback does not provide convincing or
sufficient evidence to the child; and moreover, probably can not be encoded and
integrated with existing knowledge in memory.
The second method requires an evaluation of the recall procedures. One
technique is to do an immediate simulation (i.e., self-testing), then report the
number of items recalled in the simulation. If immediate simulation is the mecha-
nism by which subjects derive an answer, then because recall per se is affected
by familiarity with the stimulus material, presumably, subjects’ estimation
should somehow be affected by familiarity with the stimulus material. A study
was conducted in which this mechanism of estimation had to be used, rather than
the one dependent on past experiences; because the task was such that no prior
recall experiences could have been possible. The task was to ask subjects to
estimate, after viewing it for 5 seconds, the number of pieces on the chess board
that they thought they could reconstruct (Chi, 1978). This task was particularly
suitable for the present purposes because, presumably, the chess players had
never had experience memorizing chess board positions before (unless they had
played blindfolded chess). A second task was also administered. From intuition,
this task seemed more complicated, and responses to the task do not seem to be
susceptible to the processes of an immediate simulation. The task required the
subjects to estimate how many trials of viewing they expected they would need in
order to perfectly reconstruct the board position. Hence, it was hypothesized that
the first task, pieces estimation, is susceptible to a simulation strategy, but not a
direct memory retrieval strategy; the second task, trials estimation, is susceptible
to neither a simulation nor a direct retrieval strategy.
The subjects were 18 adults of varying skill levels. Their chess abilities were
determined by the amount of time it took each one to complete a task called the
Knight’s Tour (Chase & Simon, 1973). The 18 subjects were then ranked ac-
cording to their Knight’s Tour times; the six fastest subjects were classified as
Good; the six slowest were classified as Poor. Prediction accuracies (determined
by the differences between predictions and actual recall) for two types of board
positions (middle and end games) for both tasks are shown in Fig. 8.3. Middle
games have 23 pieces and end games have 15 (Chi, 1978). In general (ignoring
for the moment the dotted lines in Figure 3a), there was increased accuracy of
prediction with skill for both tasks. This suggests that in general, the amount of
domain-specific declarative and procedural knowledge facilitated estimation per-
formance. To extrapolate, this suggests that, when younger children are less
 256 Hi

13 a——4A MIDDLE GAME (23 pieces) | 4—wa MIDDLE GAME (23 pieces)
12 @——® END GAME (15 pieces) @———-® END GAME (15 pieces)

NUMBER
OF
PIECES

wow
©
DO
D
Fanon

EXPERTS INTERMEDIATES NOVICES EXPERTS INTERMEDIATES NOVICES

FIG. 8.3. Prediction accuracies as a function of skill, determined by taking the

absolute differences between predictions and actual recall, for pieces and trials
estimation tasks.

accurate in an estimation task, in some sense it may reflect their limited knowl-
edge of the stimulus domain. Because the exact processes used in an estimation
task are unknown, one cannot rule out the influence of domain knowledge.
The increased prediction accuracy did not, however, reach significance for
the pieces estimation task. This probably was due to the small sample size; three
subjects’ data—the dotted lines—had to be excluded. However, the increased
accuracy of the trials estimation task was highly significant. The greater effect of
skill, in what was presumed to be the more complex task (trials estimation), even
when age is not considered, points to the complex interaction of skill with task
difficulty.
The possibility that direct memory retrieval of past experiences can determine
one’s estimation decision was suggested by the peculiar data of the three expert
Subjects (the dotted lines in Fig. 8.3a). Having been overly confident that they
could recall all the pieces on the chess board, they were extremely outlandish in
their predictions. Responses on a questionnaire indicated that they were the only
three subjects who had played blindfolded chess. It is possible that they were
misled by their blindfolded chess playing ability, assuming that it must indicate a
superior memory ability. That is, in playing blindfolded chess, one probably has
to memorize the sequence of moves, that can easily be reconstructed by a good
chess player. Such memories require different demands than remembering the
exact location of each individual chess piece. In the former case the three expert
Subjects’ superior memory probably misled them into assuming that their memo-
ry skill is also superior in the latter case, resulting in overly confident predic-
 8. REPRESENTING KNOWLEDGE AND METAKNOWLEDGE 257

tions. This particular finding also shows that, in the estimation of their memory
abilities, young adults much like young children, can be extremely unrealistic.
In sum, it is suggested that knowledge about person variables, which presum-
ably is tapped by an estimation task, may be no different in form from other types
of declarative knowledge, such as one’s chess playing abilities, or one’s driving
abilities. It is meta only in the sense that it is about cognition. As an illustration,
in the study described, the questionnaire asked subjects some factual questions
(such as: At what age did you start playing chess?), and some questions requiring
them to assess their chess playing abilities (such as: Are you a good, mediocre,
or poor player? How often do you beat your opponents?). The results show that
subjects of different skill groups were quite accurate at rating their chess playing
abilities: the Good players rated themselves 4.5 on a 5 point scale; the Mediocre
players rated themselves 3.7; and the Poor players rated themselves 2.5. Further-
more, they were only slightly less accurate about the frequency with which they
beat their opponents; the Good players thought they beat their opponents between
50-75% of the time, so did the Mediocre players; the Poor players said that they
only beat their opponents about 50% of the time. Finally, the Good players
claimed that they started playing chess at an earlier age (8.3 years) than the
Mediocre players (11.0 years) and the Poor players (12.3 years), on the basis of
their skills, this seems logical. In general, in these types of questions, which
required the retrieval of past perceptions, there were no skill differences. Be-
cause these questions are not about cognition, they are not meta in nature.
On the other hand, knowledge about persons variables, as exemplified by an
estimation task, may be meta when it requires an evaluation of existing de-
clarative as well as procedural knowledge. The chess results show that when
such processing is required, even when age is not manipulated, skill differences
can be exhibited. These results indicate that interpretations of studies where age
is manipulated require caution.
In this study, no skill differences were obtained in the subjects’ abilities to
assess their own chess playing abilities and yet some differences in their actual
abilities to estimate their memorizing abilities were obtained. This occurred even
though such a pattern of results should not have been obtained for a homoge-
neous age group—adults; they should have been as accurate in estimating their
memorizing abilities as they were in estimating their chess playing abilities.
However, such results can be interpreted by assuming that a simple memory
retrieval of past experiences was used to assess one’s chess playing abilities; a
complex set of decsion processes was used to assess one’s memorizing abilities
in chess (because even if the subjects have had the experience of playing blind-
folded chess, they should not have had experience memorizing locations of chess
pieces). The more complex the decision processes were (comparing pieces to
trials estimation), the more pronounced the effects of experience and skill. Thus,
many factors can influence such decision processes, including familiarity with
the stimulus domain and complexity of the task. Furthermore, such results also
258: CHI

suggest that the estimation paradigm need not reflect knowledge and processes
that have any special status, such as being meta; however, such tasks may reflect
the complexities of this type of decision process.

Sensitivity
Sensitivity, according to Flavell (1978), is the awareness that certain situations
require intentional memory-related behavior and that other situations do not. For
example, a child may be aware that he or she needs to prepare for effective future
recall, and therefore, should deliberately engage in preparation activities. The
most relevant data are those showing that the young child does not realize that
when an explicit request is made to memorize a set of items for future recall, it
implies that he or she should engage in some kind of activity, such as rehearsal.
This is taken to mean that the young child has not acquired ‘‘the metamnemonic
knowledge that the memorization instruction is a tacit invitation to be planful and
goal-directed’’ (Flavell, 1978).
In the conceptual framework used, the behavior exhibited by the young child
can be interpreted by postulating that he or she has not acquired the mature
memory strategies (such as for rehearsal (P2)), where the goal of remembering is
attached to the strategy. The interpretation is that a young child has in memory,
only production rules that approximate the final mature rules (such as P6), but
not the mature rules themselves (such as P2). Therefore, the overt request to
memorize is not a sufficient cue to trigger the strategic rule P2 and other similar
tules. One way to test the hypothesis is to teach children strategic rules that
incorporate the goals (of remembering), and then retest them in the same meta
questioning situation.
In a sense, the lack of sensitivity on the part of the young child is not different
from Flavell’s (1970) original notion of production deficiency, which denoted
the child’s inadequate use of strategies to facilitate remembering. The explana-
tion offered is also applicable to the phenomenon of production deficiency.
Therefore, even though sensitivity is given a special meta status, the interpreta-
tion would strip it of its meta status; in other words, in the present conceptualiza-
tion, sensitivity refers to the incompleteness of the strategic rules. This is one
possible interpretation. Sensitivity, therefore, need not be meta in either sense of
the word: whether it refers to cognitive knowledge or to second-order operations.
Sensitivity may reflect only a deficit in existing cognitive rules.

Monitoring and Checking

Monitoring, another form of metamemory, is knowledge of the current state of
one’s memory. Monitoring may be viewed as an on going activity that occurs in
conjunction with the main activity. For example, memory monitoring (Flavell,
1978) would be the processes of checking oneself for the degree of memoriza-
 8. REPRESENTING KNOWLEDGE AND METAKNOWLEDGE 259

tion. Such metaknowledge has been shown to be influential in a child’s memory

behavior. It is reasoned that retrieval efforts should reflect the degree to which
one can assess one’s current state of knowledge. Presumably, if a child can
assess in advance that he or she does not know an item, then the child should
prolong the use of effective retrieval or study strategies; if the assessment is that
he or she does know it, then retrieval or study activities should terminate. A
variety of empirical findings of this particular phenomenon have shown that
younger children (such as kindergarteners) are less accurate at sensing when they
have studied a set of items well enough for recall (Flavell, Friedrichs, & Hoyt,
1970). Another study showed that third graders appropriately allocated their
study efforts on the basis of their estimates of which items they knew and which
they did not; whereas, first graders did not make such ‘‘intelligent’’ choices
(Masur, McIntyre & Flavell, 1973). These results are important; they show that
attention allocation may be inappropriate in the younger child because the young-
er child is not aware of which of the to-be-learned items require more attention
and effort.
However, there may be a number of reasons for the young child’s inability to
allocate attention or studying efforts appropriately. These reasons need not be
related to a child’s knowledge of studying efforts per se, instead they may be
related to the child’s ability to judge the difficulty of the to-be-learned items. One
interpretation of such results is that younger children have more trouble judging
the difficulty of an item, and thus are less accurate at predicting how well learned
the items are, or which ones need more attention. The younger children’s diffi-
culty may stem from less developed knowledge of the stimulus items.
The distinction between relevant domain knowledge and what Flavell and
Wellman (1977) referred to as knowledge of task variables (one of which is
stimulus items) needs to be clarified. Knowledge of task variables (such as
stimulus items) implied that sometimes subjects are aware that a more familiar
type of stimulus item would be easier to learn than less familiar items (Kreutzer
et al., 1975). This is knowledge of stimulus variables. What is proposed is that
stimulus knowledge per se, affects a person’s judgment of the difficulty of a
stimulus item, and therefore, distorts his or her ability to allocate appropriate
amounts of studying efforts. A study that has been conducted concerning adult
expert and novice physicists illustrates the point (Chi, Glaser & Rees, 1982). In
this study, six expert and six novice physicists were asked to judge (by rating) the
difficulty of a set of physics problems. In general, the experts were more accu-
rate than the novices at judging the difficulty of the problems. The more impor-
tant finding, however, is the basis with which such judgments were made (see
Table 8.1). The experts based their judgments on either the underlying physics
principle governing the solution to the problem, or the physics components
entailed in the problem (the first two categories in Table 8.1). The novices,
however, based their decisions predominantly on superficial features of the prob-
lems (the third and fourth categories in Table 8.1). In fact, they are often reasons
2602 2:CH]

TABLE 8.1
Proportion of Response Types

Novices Experts

Abstract Principle 9% 30%

“straightforward application of Newton’s Second Law”
“collision problem, use Conservation of Momentum”
“no friction, no dissipative forces, just apply Energy
Conservation”

Problem Characteristics 33% 35%

“frictionless, problem is simplified”
“massless spring simplifies problem”
“pulley introduces difficulty”

Nonphysics Related Characteristics 40% 28%

“problem is difficult to visualize”
“easy calculations but hard to understand”
“many factors to consider, make problem difficult”

Nonproblem related Characteristics 18% 7%

“never did problems like this”
“numbers instead of symbols”
“must consider units”
“diagram distracting”
ee

that have nothing to do with physics at all, and should not have any influence on
the difficulty of the problems. Therefore, it is suggested that when a subject
(whether a child or a novice adult) is either inaccurate at judging the difficulty of
a problem, or uses inappropriate reasons for his or her judgments, it follows that
he or she will have more difficulty judging how well prepared he or she is, either
at memorizing an item or solving a problem. Therefore younger children’s in-
ability to accurately monitor their current state of knowledge (such as prepared-
ness for recall), as well as their inadequate allocation of attention, is attributed to
an inadequacy in part of their domain knowledge related to the stimulus items,
rather than strictly underdeveloped monitoring processes.
Checking, a related notion to monitoring, is the activity of checking whether
the outcome of one’s problem solving attempts is correct. Even though older
children and adults tend to be more accurate at monitoring and checking, Baker
and Brown (1984) have already argued that monitoring may not be an age related
phenomenon. One concrete example can illustrate what might produce checking
in some individuals and not in others. Again, the explanation centers on the pre-
sence of relevant domain knowledge. The example concerns checking the solu-
tion to a physics problem at the end of the problem solving processes. The study,
conducted by Andrew Judkis in our laboratory, was concerned with another
 8. REPRESENTING KNOWLEDGE AND METAKNOWLEDGE 261

matter. The task consisted of the presentation of a physics problem, one cue at a
time, to two physics professors (the experts) and two A students (the novices). The
question of interest was to see how well they could anticipate the final form of a
problem, starting with partial cues, and how many cues were required by each skill
group to reach that decision. At the end of the presentation of the entire problem
statement, the subjects were required to solve the problem. By mistake, the
numerical value for one of the known variables in the problem was too large, so
that the actual numerical solution was unusually small. At the end of the solution,
both experts noticed the discrepancy (in the sense that the answer was smaller than
they would expect). One of them, apparently confident of his calculation, merely
noted the discrepancy. The other, however, double checked and noticed the
unusually large value for the given, and only then was convinced that his answer
was right. On the other hand, the one novice who was able to solve the problem
actually commented that he thought the answer should be and was fairly large.
One possible interpretation for such results is that over years of experience,
the experts have acquired schemata of problem types, and one of the slots of a
schema is potential values of the solution. Hence, when a value does not corre-
spond to the range dictated in the schema, it signals to him that he should double-
check. According to this interpretation, checking is totally an outcome of the
presence of the relevant domain knowledge in memory, and not a meta-strategy
that some individuals have and others do not. Of course when age differences
were found, this explanation would be offered as an extrapolation to develop-
mental results. The fact that expertise and experience are often the relevant
factors has also been suggested by Brown and DeLoache (1978).

Summary and Discussion

This section reviewed a very small sample of metamemory research within the
framework of understanding: (a) how the phenomena under study are considered
meta; (b) how the phenomena fit in with the theoretical conception of meta-
knowledge proposed in the prior section; and (c) how labelling it meta facilitates
the understanding of its acquisition. The general conclusion is that calling a
variety of phenomena meta sometimes adds confusion to the overall picture.
Often, all the term meta implies is a reference to the fact that the domain being
discussed is cognition. Therefore, in discussing knowledge about task variables
and person variables, it is more elucidating to simply refer to them as such, so
that one can easily see the analogy to other types of similar knowledge (knowl-
edge about animals, knowledge about the nose and the ears). If one understands
how knowledge about the nose is acquired, then similarly, one should postulate
the same mechanism for the acquisition of metaknowledge about person or task
variables.
Because a clear conception of metaknowledge of person variables has not
been forwarded, the tasks designed to tap this knowledge may be misleading. For
262 CHI

example, the often used memory span estimation task may be influenced by a
variety of factors, making it difficult to: interpret the results; to tease apart the
aspects of the task that are meta, the aspects that are simply memory retrieval,
and the aspects that are complex decision processes that a child would not
normally possess, much like their lack of other complex processes.
Sensitivity, a new notion in metamemory, is postulated to be somewhat
related to the older notion of production deficiency. It is assumed here that the
same mechanism involved in general strategy learning would explain their
acquisition.
Finally, it was suggested that monitoring, checking, and other types of execu-
tive control processes, need not have any special status and need not be called
meta. It is not surprising that younger children may have deficiencies in these
types of processes, as they do in many other types of processes. These processes
deserve some attention in developmental studies; however, other than the fact
that executive processes control other memory and cognitive processes, why they
need to be studied under the topic of meta is unclear. It was shown that executive
processes are quite complex, and may be influenced by experience and learning;
therefore, their acquisition may not be any different from the acquisition of other
types of complex processes.

GENERAL DISCUSSION

This paper began by documenting the trend in cognitive developmental research

from centering on the child’s use and acquisition of strategies to focusing on the
child’s acquisition of knowledge. Because strategies are also one form of knowl-
edge, it is important to understand their distinction. A distinction was made by
specifying how knowledge (both declarative and procedural) and strategies can
be represented. Using this representation to serve as a conceptual framework,
metaknowledge is seen as either knowledge about factual events of cognition, or
knowledge about strategies and/or procedures. Of these, only the latter, knowl-
edge about strategies and procedures, because they are comparable to second-
order knowledge, may be seriously considered as having the special status of
meta; that is, they are rules about rules. In the last section, some of the
current
research concerning metamemory and metacognition was discussed. In particu-
lar, it was suggested that some metamemory research need not be called
meta,
other results may actually be artifacts of other types of knowledge. In order
to
untangle the somewhat mixed, irregular, and unclear results, a clearer
conceptual
framework is obviously needed to guide future empirical studies.
With a possible representational framework of what some of these meta-
phenomena might be, there appears to be less mystery about
the concept of
metacognition. For instance, it has been assumed that knowledge about one’s
cognitive processes (such as knowing about how much one can
generally recall,
 8. REPRESENTING KNOWLEDGE AND METAKNOWLEDGE 263

knowing about what task variables are the most difficult) probably is predomi-
nantly declarative knowledge, acquired with experience. The ability to respond
appropriately to the experimenter’s questions probably depends on how compli-
cated the question is, and whether the question requires a simple direct retrieval
of stored information, or whether it requires one or more transformations of the
stored information. The greater the amount of transformation required, probably,
the more salient the age effects; presumably, younger children have not acquired
the complex rules for these transformations. Sensitivity is defined as the lack of
relevant goals attached to the mnemonic rules that may be used with memorizing
tasks. Again, the goal is simply another component of the mnemonic rule that
may be acquired with experience. Finally, monitoring is seen as an activity much
analogous to the executive control processes of a system, whether it is an indi-
vidual or a computer. It is believed that even a young child has a sufficiently
sophisticated control structure to guide his or her processing. Some of the by
products of the executive control structure (such as checking), can be explained
in terms of the complexity and structure of the domain-specific declarative
knowledge base. It was conjectured that the reason that children may not, for
example, check their solution as readily as adults, reflects not so much deficits in
their control or monitoring processes, but rather, the lack of a relevant schema in
the declarative knowledge base to tell them that the answer was inappropriate.
The initial impetus for examining meta-phenomena was for the purpose of
explaining why children do not spontaneously use effective strategies in a memo-
rizing situation. It had occurred to Flavell (1971) that perhaps children are not
aware that they need to use strategies. If “‘the child is not aware of his own
limitations as a learner or the complexity of the task at hand, then he can hardly
be expected to take preventive actions,’’ in order to maximize performance
(Baker & Brown, 1984). To make this statement more concrete, in the digit span
task discussed earlier, this would imply that because the younger child is less
accurate at estimating his or her own digit span, it must mean that the child is not
aware of his or her own limitation. Therefore, the child is less likely to use a
strategy, such as rehearsal, to help remember. This would explain why young
children are so unintentional in their use of strategies; they do not realize that
they need them.
The research on this topic as a whole, however, has shown little relation
between memory and metamemory performance. It is believed that one reason is
that a precise definition of what metamemory and metaknowledge are has not
been forwarded. An attempt has been made to argue that: (a) a variety of meta-
type tasks are really not meta, (that is the tasks do not necessarily capture
knowledge about cognition or memory) therefore, one would not expect perfor-
mance on those tasks to correlate with memory performance; (b) metaknowledge
is not unlike knowledge of any other domain, and it is conceivable that knowl-
edge of cognition and knowledge of noncognitive domains develop simul-
taneously and independently, so that one need not bear on the other; and (c) the
 264 CHI

only kind of processes that must have some relation to cognition are those that
evaluate other strategies or rules. Now the problem is how meta-rules are ac-
quired. In a sense, the question of development has been pushed back one step:
How are meta-strategies or meta-rules learned and acquired with development?
Because even the initial question of how rules are learned in the first place cannot
be answered, will it not be more difficult to understand how meta-rules are
learned? The outlook is promising, however. If researchers worry more seriously
about the precise definitions and differentiations between cognition and metacog-
nition, they may advance the understanding of rule acquisition in general. The
purpose of this paper was to take one step in that direction.

ACKNOWLEDGMENTS

Preparation of this manuscript was supported in part by funds provided by the Learning
Research and Development Center, University of Pittsburgh, which is funded in part by
the National Institute of Education (NIE). The opinions expressed do not necessarily
reflect the position or policy of NIE, and no official endorsement should be inferred. The
author is also grateful for funds provided by the Spencer Foundation in support of this
research. The paper benefited from critical comments provided by Carl Johnson and
Catherine Sophian. (A translated version of this paper was published in German, in F. E.
Weinert & R. Kluwe (Eds.), Metakognition, Motivation und Lernen. Stuttgart West
Germany: Kohlhammer.)

REFERENCES

Anderson, J. R. (1976). Language, memory and thought. Hillsdale, NJ: Lawrence Erlbaum
Associates.
Anderson, J. R., Kline, P. J., Beasley, C. M. Jr. (1978). A general learning theory
and its
application to schema abstraction. ONR Technical Report.
Atkinson, R. C., & Shiffrin, R. M. (1968). Human memory: A proposed system
and its control
processes. In K. W. Spence & J. T. Spence (Eds.), The psychology of hearing and
motivation:
Advances in research and theory (Vol. 2). New York: Academic Press.
Baker, L., & Brown, A. L. (1984). Metacognitive skills of reading. In D. Pearson (Ed.), A
Handbook of reading research. New York: Longman.
Belmont, J. M., & Butterfield, E. C. (1977). The instructional approach to development
al cognitive
research. In R. V. Kail & J. W. Hagen (Eds.), Perspectives on the development
of memory and
cognition. Hillsdale, NJ: Lawrence Erlbaum Associates.
Boswell, S. L. (1974). The development of verbal and spatial organizatio
n for materials presented
tachistoscopically. Unpublished doctoral dissertation, University of Colorado.
Brown, A. L. (1975). The development of memory: Knowing,
knowing about knowing, knowing
how to know. In H. W. Reese (Ed.), Advances in child development
and behavior (Vol. 10).
New York: Academic Press.
Brown, A. L., & DeLoache, J. S. (1978). Skills, plans, and self-regulation. In R. Siegler (Ed.),
Children’s thinking: What develops? Hillsdale, NJ: Lawrence
Erlbaum Associates.
 8. REPRESENTING KNOWLEDGE AND METAKNOWLEDGE 265

Ceci, S. J., & Howe, M. J. (1978). Semantic knowledge as a determinant of developmental dif-
ferences in recall. Journal of Experimental Child Psychology, 26, 230-245.
Chase, W. G., & Simon, H. A. (1973). Perception in chess. Cognitive Psychology, 4, 55-81.
Chi, M. T. H. (1976). Short-term memory limitations in children: Capacity or processing deficits?
Memory & Cognition, 4, 559-572.
Chi, M. T. H. (1978). Knowledge structures and memory development. In R. Siegler (Ed.), Chil-
dren's thinking: What develops? Hillsdale, NJ: Lawrence Erlbaum Associates.
Chi, M. T. H. (1981). Knowledge development and memory performance. In M. Friedman, J. P.
Das, & N. O’Connor (Eds.), Intelligence and learning. New York: Plenum Press.
Chi, M. T. H. (1985). Interactive roles of knowledge and strategies in development. In S. Chip-
man, J. Segal, & R. Glaser (Eds.), Thinking and learning skills: Current research and open
questions (Vol. 2). Hillsdale, NJ: Lawrence Erlbaum Associates.
Chi, M. T. H., & Ceci, S.J. (In press). Content knowledge: It’s representation and restructuring in
memory development. In H. W. Reese & P. Lipsitt (Eds.), Advances in child development and
behavior (Vol. 20). New York: Academic Press.
Chi, M. T. H., Glaser, R., & Rees, E. (1982). Expertise in problem solving. In R. Sternberg (Ed.),
Advances in the psychology of human intelligence. Hillsdale, NJ: Lawrence Erlbaum Associates.
Chi, M. T. H., & Koeske, R. D. (1983). Network representation and recall: A case study ofa
child’s dinosaur knowledge. Developmental Psychology, 19, 29-39.
Dempster, F. M. (1978). Memory span and short-term memory capacity: A developmental study.
Journal of Experimental Child Psychology, 26, 419-431.
Flavell, J. H. (1970). Developmental studies of mediated memory. In H. W. Reese & L. P. Lipsitt
(Eds.), Advances in child development and behavior (Vol. 5). New York: Academic Press.
Flavell, J. H. (1971). First discussant’s comments: What is memory development the development
of? Human Development, 14, 272-278.
Flavell, J. H. (1978). Metacognitive development. In J. M. Scandura & C. Brainerd (Eds.), Struc-
tural/process theories of complex human behavior. Alpha a.d. Rijn, The Netherlands: Sijtoff &
Noordhoff.
Flavell, J. H. (1979). Metacognition and cognitive monitoring: A new area of cognitive-develop-
mental inquiry. American Psychologist, 34, 906-911.
Flavell, J. H., Friedrichs, A. G., & Hoyt, J. D. (1970). Developmental changes in memorization
processes. Cognitive Psychology, 1, 324-340.
Flavell, J. H., & Wellman, H. M. (1977). Metamemory. In R. V. Kail & J. W. Hagen (Eds.),
Perspectives on the development of memory and cognition. Hillsdale, NJ: Lawrence Erlbaum
Associates.
Gelman, R. (1969). Conservation acquisition: A problem of learning to attend to relevant attributes.
Journal of Experimental Child Psychology, 7, 167-187.
Greeno, J. G. (1978). Natures of problem solving abilities. In W. K. Estes (Ed.), Handbook of
learning and cognitive processes (Vol. 5). Hillsdale, NJ: Lawrence Erlbaum Associates.
Hayes, J. R. (1985). Three problems in teaching general skills. In S. Chipman, J. Segal, & R.
Glaser (Eds.), Thinking and learning skills: Current research and open questions (Vol. 2).
Hillsdale, NJ: Lawrence Erlbaum Associates.
Huttenlocher, J., & Burke, D. (1976). Why does memory span increase with age? Cognitive Psy-
chology, 8, 1-31.
Johnson, C. N., & Wellman, H. M. (1982). Children’s developing conceptions of the mind and
brain. Child Development, 53, 222-234.
Keeney, T. J., Cannizzo, S. R., & Flavell, J. H. (1967). Spontaneous and induced verbal rehearsal
in recall task. Child Development, 38, 953-966.
Kreutzer, M. A., Leonard, S. C., & Flavell, J. H. (1975). An interview study of children’s
knowledge about memory. Monographs of the Society
for Research in Child Development, 40 (1,
Serial No. 159).
266 CHI

Lange, G. (1978). Organization-related processes in children’s recall. In P. A. Ornstein (Ed.),

Memory development in children. Hillsdale, NJ: Lawrence Erlbaum Associates.
Lindberg, M. A. (1980). Is knowledge base development necessary and sufficient condition for
memory development? Journal of Experimental Child Psychology, 30, 401-410.
Masur, E. F., McIntyre, C. W., & Flavell, J. H. (1973). Developmental changes in apportionment
of study time among items in a multi-trial free recall task. Journal of Experimental Child
Psychology, 15, 237-246.
Moely, B. E. (1977). Organizational factors in the development of memory. InR. V. Kail & J. W.
Hagen (Eds.), Perspectives on the development of memory and cognition. Hillsdale, NJ: Law-
rence Erlbaum Associates.
Myers, N. A., & Perlmutter, M. (1978). Memory in the years from two to five. In P. A. Ornstein
(Ed.), Memory development in children. Hillsdale, NJ: Lawrence Erlbaum Associates.
Nelson, K. J. (1974). Variations in children’s concepts by age and category. Child Development,
45, 577-584.
Newell, A. (1973). Production systems: Models of control structures. In W. G. Chase (Ed.), Visual
information processing. New York: Academic Press.
Newell, A., & Simon, H. A. (1972). Human problem solving. Englewood Cliffs, NJ: Prentice-
Hall.
Norman, D. A., & Rumelhart, D. E. (1975). Explorations in cognition. San Francisco: Freeman.
Ornstein, P. A., & Corsale, K. (1979). Organizational factors in children’s memory. In C. R. Puff
(Ed.), Memory organization and structure. New York: Academic Press.
Richman, C. L., Nida, S., & Pittman, L. (1976). Effects of meaningfulness on child free recall
learning. Developmental Psychology, 12, 460-465.
Rumelhart, D. E., & Norman, D. A. (1980). Analogical processes in learning (Rep. No. 8005). La
Jolla, CA: Center for Human Information Processing.
Storm, C. (1978). Acquiring principles of semantic organization. Journal of Experimental Child
Psychology, 25, 208-223.
Tenney, Y. T. (1975). The child’s conception of organization and recall. Journal of Experimental
Child Psychology, 19, 100-114.
Winograd, T. (1975). Frames and the procedural-declarative controversy. In D. G. Bobrow & A.
M. Collins (Eds.), Representation and understanding: Studies in cognitive science. New York:
Academic Press.
Yussen, S. R., & Levy, V.M. Jr. (1975). Developmental changes in predicting one’s own span of
short-term meory. Journal of Experimental Child Psychology, 19, 502-508.
 Aptitudes for Learning and
Cognitive Processes

Robert Glaser
University of Pittsburgh

James W. Pellegrino
University of California, Santa Barbara

This chapter discusses intelligence—an aspect of human performance that ap-

pears to encompass all three of the topics of this symposium: metacognition,
cognitive style, and learning. In particular, we consider how intelligence, de-
fined as aptitude for learning, can be investigated in terms of current cognitive
theory. At the present time, cognitive psychologists are studying individual
differences in intelligence and aptitude in terms of the structures and processes
hypothesized in the study of cognition and cognitive development. The research
carried out along these lines is predicated on the assumption that aptitude tests
should be viewed as more than primarily predictors of achievement. Rather, such
tests should assist in identifying the processes involved in intellectual compe-
tence, and further indicate how these processes can be influenced and utilized to
benefit learning.
In order to contribute to this purpose, we have undertaken a research program
in an attempt to identify directly the cognitive processing components of perfor-
mance on tasks used to assess aptitude. The immediate goal is to analyze test
tasks, develop process models of task performance, and utilize these models as a
basis for individual difference analysis. The ultimate goal is to use the knowl-
edge gained to design conditions for learning that could be adjusted to these
individual characteristics, or design instruction to directly or indirectly teach the
processes that facilitate learning. The objective is not to train individuals to score
higher on tests of mental ability, but to directly or indirectly improve the cog-
nitive processes that underlie successful performance, both on tests of aptitude
and in instructional settings.
As a general background for this summary report of the research, four operat-
ing constraints that are important in a task-analytic effort aimed at understanding

267
 268 GLASER and PELLEGRINO

individual differences in aptitude processes are briefly mentioned. First, a partic-

ular task or set of tasks chosen for analysis should have a strong history of
reliable association with an aptitude construct that is of reasonable generality; it
also should have a consistent predictive validity with respect to a criterion perfor-
mance of significant interest. Second, in the analysis of any particular aptitude
construct, it is important to simultaneously consider the various tasks that define
an aptitude factor. The analysis of multiple tasks should enable one to differenti-
ate general and more content specific cognitive processes. Third, the analysis of
a particular task must also be explicitly concerned with explicating the sources of
item difficulty that provide the basis for individual variation in test performance.
An understanding of individual differences in task performance must include a
process theory of item difficulty. For this purpose, the processes specified as
components of performance must explain individual item characteristics, indi-
vidual subject performance as a function of ability and developmental level, and
the interaction of the two. A fourth constraint is dictated by the goal of develop-
ing a theory of individual differences in aptitude processes that can be translated
into instructional investigation. The empirical and theoretical results of any
particular analysis of the cognitive components of a task should be evaluated by
asking whether such results suggest testable instructional hypotheses. Such inter-
nal tests of the task-analytic effort can be sobering indications that a research
program has yet to achieve a sufficiently useful level of analysis. As this chapter
indicates, we are at the beginning of an ambitious research program that attempts
to comply with the previous constraints.

THE RELEVANCE OF RULE INDUCTION

Research that our group and others have conducted on a class of tasks that is
presumed to assess a psychological capacity for rule induction and is commonly
found on tests of aptitude and intelligence is now discussed. This set of intercor-
related tasks involves several task forms such as classification, series extrapola-
tion, analogy, and matrix tasks. These task forms simultaneously vary along
content dimensions that include letters, numbers, words, and geometric figures
as shown in Fig. 9.1, which shows series completion and analogy problems.
Spearman (1923) considered such tasks as measures of g, and viewed
them as an
index of the capacity to engage in intellectual processes that he referred
to as the
“eduction’’ of relations and correlates. Thurstone and Thurstone
(1941) treated
these tasks as representative of a primary mental ability called Induction
(I), and
Suggested that, as a second-order factor, rule induction might be
identical with
Spearman’s g. In more recent hierarchical aptitude models, such
tasks have been
treated as measures of gf or fluid analytic ability (Brody &
Brody, 1976). It
seems clear that such rule induction tasks assess basic reasoning
abilities com-
prising a robust aptitude construct that has relevance
for a larger domain of
 9. APTITUDES FOR LEARNING 269

SERIES COMPLETION PROBLEMS

Number Series
ZERKESSGIST
SIE I i Shh eet oe
712543590571 54785970) 5180

Letter Series
card Ged cd
fke-
a" tanks |

ANALOGY PROBLEMS

Verbal (A:B : : C:D’)

Sugar : Sweet : : Lemon :
Yellow Sour Fruit Squeeze Tea
Abate : Decline : : Wax :
Increase Improve Blemish Polish Wane

Numerical (A:B : :C:D::

VAT ST ERS
15:19:52 8:12: 5
10:402:76:3572:
28:21. ) 24182

Geometric g 8 % an N a

AAAAAK

FIG. 9.1. Task forms associated with rule induction.

 270 GLASER and PELLEGRINO

human performance. It has been argued that rule induction processes are similar
to those demanded in concept formation, and that they are related to a major form
of human problem solving that results in the acquisition of knowledge.
Our tentative view of such inductive reasoning tasks is that they sample the
way in which an individual makes use of existing knowledge to solve problems
where the solution depends on an analysis of the underlying relations (or concep-
tual similarity) among a set of problem elements. Within important limitations,
performance on these tasks has consistently correlated with academic achieve-
ment; individual differences in the capacity to engage in such analyses appear to
have direct implications for commonly required classroom learning processes.
A number of research findings on various inductive reasoning test tasks are
now described. This research is presented in the form of succinct descriptions
and conclusions, with little of the experimental detail and caveats found in
various reports of research (see references cited at the end of the chapter).

Series Completion Problems

Series completion items, as shown in Fig. 9.1, are found at several developmen-
tal levels on many standardized aptitude tests. Such items may be represented as
letter series, number series, picture series, or geometric figure series problems.
In all cases, the task structure is the same: elements comprising the series are
ordered according to some specific interitem relationship; and the individual’s
task is to extract the basic relationships and generate, predict, or select the next
item(s) in the series. The acquisition of serial pattern concepts has an extensive
history of psychological investigation. Of particular interest in this discussion is
Simon and Kotovsky’s (1963) and Kotovsky & Simon’s (1973) work on the
analysis of letter series problems of the type developed by Thurstone
and
Thurstone for their Primary Mental Abilities test battery. Simon and Kotovsky
developed a computer simulation model to represent the component processes
necessary for solution.
One important aspect of the analysis of this task is the distinction between
the
declarative knowledge and the procedural knowledge or processes
necessary for
the task. The declarative knowledge base for such letter series
problems is
limited to knowledge of alphabetic order and to relational concepts,
such as
identity (same letter), next (the next letter), and backwards-next (or reverse
ordering). Obviously, letter series problems do not involve
an extensive de-
clarative knowledge component, and it would not be expected
that individual
differences would arise from declarative knowledge deficienc
ies.
Given that the appropriate declarative knowledge is available,
the completion
of any letter series problem requires a set of basic procedu
res that are hier-
archically organized. In the Simon and Kotovsky (1963)
simulation model, there
are two basic routines: a pattern generator and a sequence
generator. The first of
these routines, pattern generation, can be broken
down into three processes: (a)
detection of the interletter relations for the given problem
elements, (b) use of the
 9. APTITUDES FOR LEARNING Zu

relational information to extract the period length of the pattern within the prob-
lem, and (c) generation of a pattern description or rule involving both the rela-
tions and the periodic structure of the problem. This rule, specifying the pattern
description, serves as input to the sequence generator, which applies it to the
current state of the problem and then extrapolates the pattern to generate the
additional elements required for problem solution. Differences in item difficulty
and potential individual differences in problem solution can result from the
application of any or all of these specific processes.
Concerning sources of task difficulty, a number of systematic properties of
the individual items determine the difficulty of a problem. One aspect of the
probability of error is the type of relation involved. Identity relations are easier to
detect than next relations, which, in turn, are easier than backwards-next rela-
tionships. The difference in difficulty between extrapolating identity and next
relationships also varies as a function of the position of the relationship within a
period.
These sources of error are readily explainable if one considers the require-
ments of working memory. Identity relationships do not place demands on work-
ing memory, whereas successive nonidentity relationships involve accumulating
placekeepers. The longer the period length, the greater the memory demands of a
problem, and the greater the likelihood that working memory limits may be
reached. For example, the letter and number series problems in Fig. 9.1 differ in
both period length and pattern complexity. Thus the overall pattern description
that constitutes a problem is related to problem difficulty. The length of the
pattern description, which is a function of period length, the types of rela-
tionships involved, and the resulting working memory requirements, are highly
correlated with problem errors. These errors can arise from performance inad-
quacies, which differ among individuals in relation detection, discovery of the
periodic structure, completion of a pattern description, or in the extrapolation
process.
Consider instructional experimentation based on this model. It has been dem-
onstrated that Simon and Kotovsky’s (1963) model of serial completion items
provides a reasonable account of performance in this task; their simulation of
human protocols provides a partial validation of the model. Given a concern with
the criterion of instructional tractability, another way in which such models can
be validated can be considered. If the processes embodied in a simulation model
are similar to those used by humans, then individuals whose performance repre-
sents a low or intermediate level of task competency might be trained in these
processes. However, if the processes specified by the model are incompatible
with human cognitive structures, then there will either be difficulty in training
these processes, or they will have no positive, and perhaps a negative, influence
on performance.
In an attempt to provide such an instructional test, a study was conducted that
involved direct and independent training in discovery of relations and discovery
of periodicity with a sample of children, from grades | through 6 (Holzman,
 272 GLASER and PELLEGRINO

Glaser, & Pellegrino, 1976). Both the training group and a control group were
given a pre and posttest set of letter series problems, identical in rule structure,
but initialized at different points in the alphabet. The results showed a significant
gain in performance for the training group. In particular, the training group
showed a percentage reduction in errors, over twice the number shown by the
control group (32% vs. 13%). Furthermore, the training group showed signifi-
cantly greater gains on problems based on more difficult patterns, whereas the
control group remained the same or reduced errors only on the easier items.
Thus, the training appropriately functioned where it was most needed—that is,
when individuals encountered more difficult relations and problems. The
qualitative difference between control and experimental conditions also suggests
that explicit training on the identified component processes may have provided
an information management strategy that facilitated pattern description and
extrapolation.
Related instructional studies conducted on series completion performance
showed an interaction between type of training and performance level. Sheer
practice on the test items was sufficient to produce significant gains in perfor-
mance at some levels of initial competence, however, at lower levels of initial
ability, explicit process training was more effective than practice. The source of
this interaction needs to be more precisely determined by explicit analysis of the
process differences among individuals that define different levels of task compe-
tence and intellectual development.

Analogical Reasoning Problems

We now present another task form presumed to assess the capability for inductive
reasoning. Of the many tasks that are assumed to assess this capability
, the
analogy problem is the most pervasive. Over the entire course of the testing
movement, analogy items, as shown in Fig. 9.1, have constituted a significan
t
portion of intelligence tests. Sternberg (1977a, 1977b) has provided
a detailed
review and discussion of the importance of analogical reasoning within
the field
of differential psychology. The centrality of this type of reasoning in
the concept
and measurement of intelligence can be found in the writings of such
individuals
as Spearman (1923) and Raven (1938).
In the past few years, Spearman’s theory has been expanded
and refined in the
more precise, experimentally founded theory presented by Sternbe
rg. Sternberg
has proposed and tested a theory of analogical reasoning that
specifies several
processes intended to apply across all analogical reasoning
tasks. In Sternberg’s
analyses, emphasis is placed on developing general models
of analogy solution
and specifying individual differences, in terms of latency
parameters, for the
various processes involved. However, the qualitative
characteristics of these
processes remain largely unspecified, and there is still no
adequate understanding
of how the processes are executed, what information
or content must be pro-
 9. APTITUDES FOR LEARNING 213

cessed, or how such information contributes to differences in item difficulty and

performance errors.
In the analysis of individual differences in analogy solution, we have at-
tempted to specify the components of task performance that differentiate between
levels of aptitude and how these components interact with differences in item
structure or content. A general overview of the work is reported in the remainder
of this paper. For this purpose, and at the expense of oversimplification, we infer
from this work three interrelated elements that appear to differentiate high- and
low-skill individuals in analogical reasoning test tasks. These elements are the
management of memory load, procedural knowledge of task constraints, and
organization of an appropriate conceptual knowledge base. We discuss each with
reference to relevant studies of figural, verbal, and numerical analogy tasks, and
then suggest implications for improving these skills.

Figural Analogies (Memory Load)

We first investigated figural analogies (see Fig. 9.1) because the type of item
features in these problems seemed easier to analyze (see Mulholland, Pellegrino,
& Glaser, 1980). In contrast to the symbolic aspects of verbal and numerical
analogies, the information necessary for item solution is externally represented in
the physical problem array. The specific aspects of item content are the figural
elements that are used to construct the separate analogy terms, and the spatial and
logical transformations that are applied to the elements to construct overall rules.
The elements that comprise the terms are easily perceived plane geometric fig-
ures—lines, circles, triangles, and quadrilaterals. The basic transformations in-
clude: removing or adding elements; rotating, reflecting, and displacing ele-
ments; size changes; and variations in element shading.
A simplified model (see Fig. 9.2) of the relationship between item content and
processing assumes an initial phase of pattern comparison and decomposition
process that yields units of information, representing the individual elements of a
pair of analogy terms. The complexity of such a process is a function of the
number of elements that must be isolated. The second stage of processing is
transformation analysis and rule generation. This phase of processing attempts to
determine the transformations that specify the rule for changing the A stimulus
into the B stimulus. The complexity of identifying and ordering a set of transfor-
mations is a direct function of the number of transformations required in an item.
Each operation associated with pattern decomposition and transformation
analysis of an A-B pair of terms yields a unit of information that is stored in
working memory, and each element-transformation proposition requires an indi-
vidual placekeeper or slot in working memory (e.g., Rumelhart, 1977; Simon &
Kotovsky, 1963). As a consequence of this, the largest single source of error is
multiple transformation of single elements. In this type of problem, the inter-
mediate results of the C-D rule generation inference process must be retained in
 PROCESS PRODUCT LATENCY
Pattern comparison and
decomposition of
Element List (geometric
patterns: lines, etc.) RT =eE
A and B

Transformation analysis Element-transformation

and rule generation for List (object-operation
A-B propositions) RT> =tT

Pattern comparison and

decomposition of Element List RT3 =eE
C and D

Transformation analysis
and rule generation for Element-transformation List RTy =tT
C-—D

RT total =xE+yl +k

FIG. 9.2. Simplified process model for figural analogy verification task.

memory and the entire transformation sequence inferred before the appropri-
ateness of the D terms can be judged. Thus, due to such additional demands on
working memory, it is possible for some of the original element-transformation
information to be lost or degraded.
To be somewhat more specific on this matter, consider that the time to
identify and order a set of transformations appears to be a function of the number
of elements and transformations that comprise an item; the probability of
error of
an item should increase as a function of the number of cognitive
operations
required to identify elements and the transformations applied to them.
These
general assumptions were tested in the Mulholland, et al. (1980)
experiment,
where figural analogies in a reaction time verification task were
presented to
university undergraduates. The analogies were generated from six
types of ele-
ments and six types of transformations that frequently occur on
items found on
aptitude tests. The elements were easily perceived geometric figures.
The trans-

274
 9. APTITUDES FOR LEARNING 275

formations included: rotation, reflection, size increase and decrease, doubling

and halving elements, and identity (no transformation). The number and type of
elements and transformations were systematically varied in the construction of
items, as shown in the examples of true and false items in Fig. 9.3.
Performance on these experimental items correlated highly (r = .69) with
scores on a figural analogy test, used as a subtest of a standardized test of
cognitive abilities. Individual latencies for solving the experimental items dif-
fered as a function of item structure in a systematic and reliable manner. Fig. 9.4
shows the latency data for various item types (i.e., combinations of number of
elements and transformations). The data are consistent with the assumption that
individuals decompose the terms of an analogy task in serial fashion; that is, they
isolate and identify the elements one by one, and with each additional element,
there is an increment in time. The processing time of transformations also repre-
sents a similar serial and additional processing mode.
As the effects of element and transformation requirements combine, that is, as
item complexity increases, there begins to be a problem of mental bookkeeping.
Each operation performed in decomposing the terms of an analogy and determin-
ing the individual transformations yields units of information that take up space
in memory. As more partial information is accumulated and entered into memo-
ry, one may begin to approach the limits of working memory: Processing time
and processing effort may have to be partially diverted to updating and maintain-
ing the accumulated contents of memory. The assumption of an increasing mem-
ory load in complex items suggests that it is a potential source of performance
errors. This is supported by the error data shown in Fig. 9.5, which shows that
errors increased as a function of the number of transformations and increased
most rapidly when several different transformations had to be performed on a
single element. The latter result suggests that, in items of this type, a special
cognitive demand is retaining and operating on the intermediate products of
transformation in memory. Thus, in the processing of such figural analogies,
young children and less proficient solvers appear to be particularly inefficient in
these aspects of performance (Pellegrino & Glaser, 1982).
As described earlier, the memory load influence on performance has been
verified by empirical studies of other test item forms (such as series extrapolation
problems), where individual letter series problems, which theoretically involve
several placekeepers in working memory, show the highest error rates (e.g.,
Holzman, Glaser, & Pellegrino, 1976; Kotovsky & Simon, 1973). A similar set
of processing and working memory assumptions may also be applicable to per-
formance differences across items on figural matrix problems, such as those in
the Raven’s Progressive Matrices Test (see Hunt’s, 1974, analysis of this task).
In all the work of this kind, the strategies that individuals use to manage memory
load, and the extent to which different individuals, in order to reduce memory
load, represent figural elements in memory in different ways, has not been
carefully studied.
 wan) sseI) anı] saıdojeuy asyey ssıbojeuy

276
| juawa)3
| LONEWJIOJSUBJ]

| Juawalz
€| suoNNeWIOJSUBI

Z sUawalJ
Z suoijewsOjsues)

E sjuawa|3
| | uoNeWOjsues

¢ sjuawa)3
E sUONEWJIOJSUBJ,

914 "6 sopdurxq

Jo ony pur Asfe} JLN9WO9Z sordorur
Pasn JOJ OY} UONBITJOA
SP)
 if = Number of Transformations

T-3

FIG. 9.4. Mean reaction times for true

analogies as a function of the number of
elements and transformations. Number of Elements

One Element

Two Elements

Three Elements

Percent
Errors

10) 1 2 3
Number of Transformations

FIG. 9.5. Mean errors for true analogies as a function of the number of elements
and transformations.

271,
 Verbal Analogies (Procedural Constraints)

The notion of differential procedural task constraints for high and low aptitude
individuals has been developed in studies of verbal analogy test tasks. On the
basis of an extensive protocol analysis study of multiple choice verbal analogy
items, Heller (1979) has proposed a general model of analogical reasoning per-
formance that attempts to describe the information processing differences be-
tween individuals of varying skill, in terms of behaviors that conform to or
violate analogy task constraints. Effective problem solution is characterized as a
series of steps directed toward the satisfaction of specific goals, and the more
constraints a solver is aware of, the more highly constrained will be the goals
pursued.
By considering the ‘‘syntax”’ or rules for interpreting the structure of analogy
items, solution protocols can be examined to determine whether or not they
include behaviors that violate task constraints. When this is done, three catego-
ries of solutions emerge: analogical, nonanalogical, and ‘“‘buggy’’ solutions.
Solutions that contain no violations of task constraints are categorized as analo-
gical. These analogical solutions are characterized by: (a) consistent attention to
the relations contained in two allowable word pairs, and (b) consistent attention
to the match between these pairs of relations.
Solutions that contain major violations of task constraints throughout a solu-
tion episode are categorized as nonanalogical. Nonanalogical solutions are char-
acterized by: (a) attention only to relations between illegal pairs of elements;
and/or (b) consistent attention to the match between inappropriately selected
pairs of relations in two word pairs; and/or (c) a consistent disregard for
the
match between relations contained in two word pairs.
In the category called ‘‘buggy’’ solutions, some solution behavior violates
task constraints, however, most is analogical. (The term ‘“buggy’’ is borrowed
from the computer programming notion of program *‘bugs,”’ correspon
ding
essentially to analogical procedures with subroutines that misfire
under certain
conditions, or analogical procedures with missing or faulty subroutin
es that
manifest themselves under certain conditions.) ““Buggy’’ analogica
l solutions
contained both of the types of behavior described for analogica
l and non-
analogical solutions. Sample solution protocols for analogical
and nonanalogical
types are shown in Table 9.1 (Heller, 1979).

Analogical Solutions. When the performance of individuals who did solve

items analogically is examined, a major distinction among
analogical solutions is
apparent in the development of the solvers’ understanding
of the analogical rule
in an item, solutions can be “conceptually driven’’
or ‘‘interactive.’’ In the
conceptually driven solutions (Protocol 1 in Table 9.
1), the solver’s initial under-
standing of the A-B rule drives evaluation of the optiona
l completion terms and is
sufficient for discrimination among the options.
However, in interactive solu-

278
 9. APTITUDES FOR LEARNING 219

tions (protocol 2), the A-B relation is either initially inaccessible, or does not
permit identification of a unique completion term from the option set; therefore
with reference to C-D relations, the rule is identified or modified. Because they
involve increasingly detailed specification of the analogical rule and/or consid-
eration of alternative conceptualizations of the rule, interactive solutions require
more extensive processing than conceptually driven solutions.
Individual differences in analogical reasoning ability appear to correspond to
the differential availability or utilization of these additional processes. Low-
ability solvers show an increased reliance upon conceptually driven solutions.
That is, when low-ability individuals use the analogical solution procedure, they
tend, more often than high-ability solvers, to evoke a sequence of processes
corresponding to initial identification of the analogical rule; however, they do not
subsequently modify that rule. Although low-ability solvers are also capable of
solving items interactively, they do so less often than higher ability solvers. On
more difficult items, which are less likely to be solvable in the conceptually
driven mode, low-ability solvers exhibit performance that violates task con-
straints. This observation is further apparent in the examination of nonanalogical
solutions.

Nonanalogical Solutions. Three types of nonanalogical solutions were iden-

tified: One type (Protocol 3) represents solutions in which no attempt was made
to identify the A-B relation— attention was paid only to the presence or absence
of C-D’ relations. (D’ indicates a multiple choice response.) A second type
(Protocol 4) represents solutions in which an attempt was made to identify both
A-B and C-D’ relations, but no apparent attempts were made to determine
whether any two relations matched. A third type (Protocol 5) represents solutions
in which all four analogy terms were considered, but attention was paid to the
interrelations among three or four terms rather than to the match between two
distinct relations within element pairs.
By definition, all three types of nonanalogical solutions include, a violation of
the central constraints of the analogy task—essentially that behavior should be
directed toward identifying two distinct relations that are analogous or matching.
However, to different extents, the three conform with task constraints and indi-
vidual differences were reflected in the ability to solve analogies in accordance
with task constraints. Nonanalogical solutions by the low-ability solvers were
primarily of the types where no attempt is made to identify the A-B relation or to
refer to two distinct relations. Higher ability solvers, on the other hand, consid-
ered all four terms in most of their nonanalogical solutions and attended to two
allowable relations.
In general, this research suggests that skilled analogy solvers are charac-
terized by more knowledge of task constraints, and by the ability to develop an
understanding of the analogical rule in response to the item stem, the rela-
tionships involved, and the response options. Conversely, less skilled solvers
 TABLE 9.1
Sample Solution Protocols (from Heller 1979)
—_—_:????}]} _—_—_——SSOSO ee m mmm—mm_m—n— nk —_

Analogy Elements Presented Solver's Response

Analogical Solution: Protocol 1

Conceptually driven; one opticn matches initial specifications

TEA:COFFEE::BREAD: Tea is to coffee as bread is to... . rolls because tea and

coffee, they’re both drinks, and they’re about the same
thing, just two different names for two different drinks,
and a bread and a roll would be about the same—two
different names for the same thing.

MILK (Reject) That doesn’t fit, it’s a drink.

BUTTER (Reject) Butter is something you put on bread, that doesn’t fit.
ROLLS (Accept) That’s good.
JAM (Reject) It’s like butter, something you put on bread. It wouldn’t fit
because you don’t put coffee on tea or in tea.

Analogical Solution: Protocol 2

Interactive; initial failure to identify A-B relation—analogical rule identified during
option
verification

ABATE:DECLINE::WAX: This is a good one, Oh Christ, I don’t know—I can’t say

anything yet because I don’t know what “abate” means.

POLISH (Accept) Well, wax and polish mean almost—well they're very
close, and maybe abate and decline are very close. I
don’t know, I’m just gonna put true.
INCREASE (Reject) I just don’t know.
WANE (Reject) To me, decline seems to have something to do with abate,
even though I don’t know what it means, but wane
doesn’t have anything to do with wax.
IMPROVE (Reject) I was thinking, maybe abate means “to decline” because
wax may mean “to improve.” And like before, it means
“to polish.” I like polish better, though.

Non-Analogical Solution: Protocol 3

Consideration of C-D’ relations only

LINE:RULER::CIRCLE: Ball. Because a ball is a circle, it’s round.

ee
ee See
ROUND (Accept) Yeah, a circle is round.
DRAW (Accept) No, because draw can’t be a circle. Oh! Yes, it could
be
because you draw a circle.
RADIUS (Accept) Radius is the numbers in the circle, that’s good.
COMPASS (Accept) Compass you use to go around—like you put your
pencil
and it’s a circle.
Sete
ee en

280
 TABLE 9.1 (Continued)
—_—_— ee m zn nase zz.., he

Analogy Elements Presented Solver's Response

(Which of these do you think Round, because a circle is round.

best completes the
analogy?)

Non-Analogical Solution: Protocol 4

Identification of A-B and C-D’ relations; no relational comparison

TELL:LISTEN::GIVE: Take. If you tell something, they’re like taking it in. If you
give something, they take it.

PRESENT (Accept) Tell is to listen as give is to present? Yeah, I’d go with

that! You give presents?
LOSE (Reject) No. Most people find something, they ain’t gonna give it
back.
GET (Accept) Yeah. If you get something, somebody gave it to you.
HAVE (Accept) When they give it to you, you have it. Yeah.

(Which of these do you think Present. Because you give presents.

is best?)

Non-Analogical Solution: Protocol 5

Consideration of A-B-C-D interrelations only

SUBJECT:CITIZEN::
KING: King—king—queen.
(Could you explain how you Well, subject to citizen—like the king is married to a queen
got that?) so I figured king and queen. They stay together.
(What about subject and cit- Well, citizen is a person and is like a subject. So I figured
izen made you think you'd that king and queen ought to fit into it. Same as subject
need something that went and citizen. If I hear you talking about a subject, then it’s
with a king?) probably the queen.

RULE (Accept) This one is a good one here because you’re describing the
rules. The king and rule is almost like the citizens and
rule and I think that, I guess this is a pretty good one.
It’s kinda hard.
(Could you explain a little bit Well, the subject is a type of one thing and a citizen is like
more what subject has to a person. So the king is a man who’s higher and the rule
do with citizen and king is—the king rules the citizen.
has to do with rule?)
KNIGHT (Reject) I don’t think so. Because knight—I can’t really say why.
(What is a knight?) A knight is a man that guards the king. That’s all I can
really say.

(Continued)

281
 282 GLASER and PELLEGRINO

TABLE 9.1 (Continued)

I ee a a T mn un 2 0 nungen
Analogy Elements Presented Solver's Response

PRESIDENT (Accept) This one’s all right. President—king’s almost the same
thing, and both of them are citizens and they’re subjects.
(What do you mean “They are Well, it’s something— subject to something—I can’t ex-
subjects?) plain. King and president are citizens and they’re subjects
to another person—they’re the subject of what other peo-
ple are talking about.
KINGDOM (Accept) This one’s all right because the kingdom’s where the king
lives. I guess it’s all right—I can’t go against it.

(You said that president, President. I like the king and the president because they re
kingdom, and rule are pos- almost the same persons, they both rule in different
sible. Which of those three places.
do you like the best?)
(And how do they connect Because they’re both citizens and are subject to a person.
with subject and citizen?)
eS

proceed analogically when they can easily identify an analogical rule, but if that
rule is initially inaccessible, or no C-D’ relation can be found to match
the
initially specified rule, they violate task constraints of appropriate analogical
syntax.

Numerical Analogies (Knowledge Base Influences)

Thus far in the discussion, the issue of the declarative or conceptu
al knowledge
base necessary for solving analogies, has essentially been ignored.
The discus-
sion now turns to the work on numerical analogies, where
the influence of
knowledge structure on solutions has been the focus of several
studies (Corsale &
Gitomer, 1979; Pellegrino, Chi, & Majetic, 1978). Althoug
h such knowledge is
relatively circumscribed for problems using numbers in
the typical range of 0—
1000, it is, nevertheless, variable across individuals, dependi
ng on their back-
ground and experiences, and differentially affects performa
nce. The study by
Corsale and Gitomer (1979) attempted to characterize the nature of the dif-
ferences in the knowledge bases of high- and low-ability
individuals, and then to
indicate how these knowledge differences influence
subjects’ problem solution
strategies. Two kinds of data were collected: an
initial set that was used to
characterize the knowledge representations of the
elementary school children
participating in the study, and a second set of protocol
data taken from their
problem solutions.
Two tasks were designed to tap the representati
ons of the children’s knowl-
edge of numerical relationships. The first task
was a grouping task; the children
 9. APTITUDES FOR LEARNING 283

were given a matrix of numbers from 0-32, and were asked to select groups of
numbers that went together, and then to justify their groupings. On the basis of
these justifications, the kinds of groups that the children made were classified
into four types: (a) abstract concepts, representing mathematically based group-
ings with superordinate labels, such as the set of primes and multiplicative or
exponential relationships; (b) operational concepts, which involved the stringing
together of numbers into number sentences; (c) nonmathematical concepts,
which were idiosyncratic groupings, or groups based on orthographic sim-
ilarities; and (d) digit based groupings, involving numbers that shared common
digits, the set of single digit numbers, etc. In the second knowledge representa-
tion task, children were presented with 20 pairs of numbers and were asked to
state as many relationships as they could for each pair.
Data from the grouping task and the pairs task were reduced by means of a
factor analysis. Three factor scores were derived for each child. The first factor
was readily interpretable as an estimate of the degree of abstractness found in the
children’s groupings and pair relationships. The second was a nonmathematical
factor, seeming to estimate the number of groupings and relations generated. The
third factor represented a preference factor, in which operational or computation-
based groups and relations were preferred or were more salient than abstract
groupings. Multiple regression analyses on these factor scores, using analogy
test performance as the criterion variable, indicated that the degree of ab-
stractness in number knowledge is a significant predictor of success in analogical
problem solving.
Having demonstrated that the form of knowledge representation is an influen-
tial variable, Corsale and Gitomer (1979) examined the interactions of knowl-
edge representation and strategy usage for children of different abilities. The
highest and the lowest scorers on the standardized analogy test were selected and
engaged individually in a session of oral problem solving. Analysis of these
protocols indicates that knowledge representation drives solution strategies by
defining the limits of the problem domain. Thus, high-ability individuals, who
have clear, high level number concepts that are abstract in nature, limit their
analogical hypotheses to a few plausible mathematical relationships. In contrast,
low-ability children have lower order number concepts, and their analogical
solutions indicate that they do not solve analogies with systematic, mathe-
matically based rules.
There were two categories of errors in analogy solution; these are listed in
Table 9.2. The first, mathematical errors, were of two types: (a) computation
errors; and (b) digit errors, in which the subject treated a number not as a total
number concept, but as a set of isolated digits (e.g., ‘'64 and 16 go together
because they both have a 6 in them’’). The second category of errors, analogical
ones, contained a variety of types: (a) nonrestrictive errors, where the rela-
tionship between the numbers in a pair was not specific enough to allow dis-
tinguishing among the options; (b) series errors, where analogy problems were
 284 GLASER and PELLEGRINO

TABLE 9.2
Errors in Numerical Analogy Solution
ee ee us nu nn mn nn nn
Mathematical Errors
1. Computation errors
2. Digit errors
Analogical Errors
1. Non-restrictive errors
. Series errors
. Single pair errors
. AC-BD errors
. Non-analogical computation errors
Au
RUN Errors where an individual applied the correct rule, but in the wrong direction
—— — ee III ee ee

turned into series problems; (c) single pair errors, where the subject adopted a
rule to apply to E that was true only of AB or CD, but not both: (d) AC-BD
errors, where children looked for relationships across pairs rather than within
pairs; (e) nonanalogical computation errors, where computations were analo-
gically inappropriate; and (f) errors where an individual applied the correct
rule,
but in the wrong direction.
Low-ability children committed more analogically inappropriate computa-
tion, nonrestrictive, and digit errors. The kinds of errors they made indicated
that
they do not restrict their hypotheses concerning an analogical rule to
mathe-
matical concepts (as noted by the digit errors), or to analogical concepts
(as noted
by analogically inappropriate computation). Apparently, low-skill
subjects have
more diffuse, less structured knowledge representations of number,
and this is
manifested in the kinds of errors they make.
Based on evidence of the less structured knowledge of low-skil
l children, the
lack of solution power can be seen particularly on the probabil
ity of success in
the interactive solution procedure. As previously indicate
d, this occurs in the
course of solving an analogy, when an individual does not
initially infer the AB
relationship and uses a backward inference strategy when
presented with the CD
pair. Both low- and high-skilled individuals use an interact
ive strategy with equal
frequency; however, its successful use is significantly
higher in high-aptitude
individuals.
The parallel between knowledge representation and
solution strategy can also
be seen in errors of the high-ability individuals. Analog
ies were turned into series
problems more frequently by highly skilled children.
This is a sophisticated kind
of error that involves the detection of mathematical
relationships across pairs and
that follows a constraint rule. In general, the errors
made by skilled individuals
show that the knowledge representation data, which
indicated constrained mathe-
matical concepts for these individuals. parallels
their use of that knowledge.
When they could not detect a rule, high-ability
children would “give up’’ and
not select a multiple choice answer, whereas
low-ability children would select an
 9. APTITUDES FOR LEARNING 285

answer—usually a wrong one—and justify it post-hoc. High-ability children

operate within both mathematical and analogical constraints in order to achieve
an analogically correct answer. They know when they are wrong and give up
rather than choose an answer that they know is wrong. Not only do low-ability
children choose the wrong answer rather than give up, but the protocol evidence
suggests that they justify their choices on nonanalogical and/or nonmathematical
grounds.
Considering both the knowledge representation data and the protocol data, the
study by Corsale and Gitomer (1979) suggests that the high-scoring individuals
on the number analogies aptitude test task have a greater degree of abstract
mathematical knowledge and show a greater salience of abstract over operational
concepts. This knowledge correlates with and predicts analogy performance.
High-ability subjects use their knowledge of abstract number relationships to
constrain the domain of permissible operations; this knowledge base determines
the appropriate use of strategies. In contrast, low-skill children, often engaged in
analogically inappropriate computation and nonrestrictive errors. They appar-
ently have not developed the highly constrained organizational structure of
knowledge that would provide them with constrained rules of operation.

IMPLICATIONS

As is apparent from the work reported here, researchers are in the early stages of
the effort to understand and assess the nature and instructability of (learning the)
skills that comprise aptitude for learning. Nonetheless, our findings thus far (and
only their flavor has been described in this paper) permit some speculation.
Three interrelated factors emerge that appear to differentiate high- and low-skill
individuals. First, the management of memory difters, as is reflected by speed of
performance and handling of demands on working memory. Second, individuals
show differences in their knowledge of the constraints of problem-solving pro-
cedures—what we have called the syntax of analogical problem solving. Effec-
tive problem solution is characterized by problem-solving steps that satisfy goals
determined by problem constraints: The more constraints the solver is aware of,
the more highly constrained will be the goals pursued. Faced with a difficult
problem, a skilled individual generates subgoals that enable a return to higher
level goals. For the low-skill individual, solution difficulty results in violations
of problem-solving constraints, the imposition of procedural bugs, and the in-
ability to recover higher level goals when subgoals need to be pursued. Third, the
structure of the declarative-conceptual knowledge base and the level of represen-
tation of this knowledge differ as a function of ability. High-skill individuals
employ conceptual forms of knowledge that constrain their induction of rela-
tions; low-skill individuals encode their knowledge at more concrete surface
levels, as is manifested by their limited inferential power.
 286 GLASER and PELLEGRINO

What are the implications of these findings from the experimental analysis of
prototypical test tasks for a conception of academic learning skills? Of the three
differentiating aspects mentioned, the memory management component might
suggest some sort of processing facility and process training, such as the em-
ployment of rehearsal and organizational strategies of the kind studied in memo-
ry experiments. The other two components, however, which are concerned with
knowledge representation and problem-solving procedures, suggest a different
emphasis.
An emphasis on memory management leads one to consider influencing men-
tal processing skills, such as better methods for searching memory and elaborat-
ing connections, which would facilitate storage and retrieval. For example, an
individual could be taught to see fewer single elements and more wholistic
features in figural analogy problems. In contrast, an emphasis on the knowledge
base and its representation, and an emphasis on the knowledge of problem-
solving constraints suggest that one consider a knowledge strategy. In a knowl-
edge strategy, progress is seen in terms of improving the ways a knowledge base
is recognized and manipulated. When highly skilled individuals learn something
new or undertake a new problem of induction, they engage a highly organized
structure of appropriate facts and relationships, and associated procedures and
goal constraints. Skilled individuals are skilled because of their knowledge of
both the content involved in a problem and the procedural constraints of a
particular problem form, such as inductive or analogical reasoning. These two
kinds of knowledge interact so that procedural constraints are exercised in the
content knowledge base; the knowledge base enables procedural goals to be
attained.
This analysis leads to the suggestion that the improvement of the skills of
learning will take place through the exercise and development of procedural
(problem-solving) knowledge, in the context of specific knowledge domains.
The suggestion is that learning skills are developed when one teaches more than
mechanisms of recall and recognition for a body of knowledge. Learning skill is
acquired as the content and concepts of a knowledge domain are attained in
learning situations that constrain that knowledge in the service of certain pur-
poses and goals. The goals are defined by uses of that knowledge in procedural
schemes, such as those required in analogical reasoning and inductive inference.
At this point, it is difficult to say how this facility could actually be taught.
One might teach more of the knowledge base and its high level concepts, or one
might teach procedural knowledge, such as planning ahead and recognizing
when procedural constraints are violated. Teaching either separately would prob-
ably be unsuccessful, because each kind of knowledge facilitates the develop-
ment of the other. Such learning skills are probably developed through graded
sequences of experience that combine conceptual and procedural knowledge
.
(This is what must take place when a good instructor develops a
series of
 9. APTITUDES FOR LEARNING 287

examples that stimulate thinking.) Eventually, procedural schema for various

problem forms, such as analogical reasoning, might be abstracted from a variety
of knowledge bases, so that a generalized form of procedural knowledge can
transfer and be generally applied to new events.
Finally, there is the problem of diagnosing the weaknesses in individuals who
are unskilled in academic learning. When this is done, one generally finds that
their knowledge base is not rich and that their skill in maintaining directed use of
knowledge is not developed. Perhaps a reasonable tactic is to identify, in an
individual, some attained knowledge base where instruction can begin. Knowl-
edge developed in the course of an individual’s prior cultural experience can
provide knowledge representations and goal-directed behavior that can be ex-
ploited. As a result of prior experiences, these exist in varying degrees in indi-
viduals, and they can be transferred to domains of related knowledge that ap-
proximate more and more closely the formal abstractions and procedural
requirements necessary for school learning.

FURTHER RESEARCH

As indicated earlier in this paper, the work reported is just a beginning, and
further research needs to be done before theory and practice in the measurement
and training of aptitudes for learning can be significantly influenced. In our
laboratory, currently, research is contemplated to further understanding of analo-
gical reasoning problems. In work on figural analogies, there are plans to study
how the figures presented are encoded and represented, prior to the subsequent
processes of pattern comparison, transformation, and rule comparison. The anal-
yses of verbal analogies reported in this paper are essentially concerned with the
syntactic rules of the task. New research will consider the interaction between
syntactic constraints of the problem and the semantics of word meanings and
word relationships. The work on numerical analogy problems suggests a similar
investigation of mutual influence between conceptual knowledge of numbers and
the operation of procedural knowledge of task constraints. The results of the
numerical analogy study also suggest a training experiment in which children
learn certain levels of number knowledge, and the influence of these levels upon
problem solution is observed.

ACKNOWLEDGMENTS

Preparation of this paper was supported by funds provided by the Learning Research and
Development Center, University of Pittsburgh, which is supported in part by the National
Institute of Education, U.S. Department of Health, Education, and Welfare.
 288 GLASER and PELLEGRINO

REFERENCES

Brody, E. B., & Brody, N. (1976). Intelligence: Nature, determinants, and consequences. New
York: Academic Press.
Corsale, K., & Gitomer, D. (1979). Developmental and individual differences in mathematical
aptitude. Paper presented at the meeting of the Psychonomic Society, Phoenix, AZ, November.
Heller, J. T. (1979). Cognitive processing in verbal analogy solution. Unpublished doctoral disser-
tation, University of Pittsburgh.
Holzman, T. G., Glaser, R., & Pellegrino, J. W. (1976). Process training derived from a computer
simulation theory. Memory & Cognition, 4, 349-356.
Hunt, E. (1974). Quote the Raven? Nevermore! In L. W. Gregg (Ed.), Knowledge and cognition.
Potomac, MD: Lawrence Erlbaum Associates.
Kotovsky, K., & Simon, H. A. (1973). Empirical tests of a theory of human acquisition of concepts
for sequential events. Cognitive Psychology, 4, 399-424.
Mulholland, T. M., Pellegrino, J. W., & Glaser, R. (1980). Components of geometric analogy
solution. Cognitive Psychology, 12, 252-284.
Pellegrino, J. W., Chi, M. T. H., & Majetic, D. (1978). Ability differences and the processing of
quantitative information. Paper presented at the meeting of the Psychonomic Society, San An-
tonio, TX, November.
Pellegrino, J. W., & Glaser, R. (1982). Analyzing aptitudes for learning. In R. Glaser (Ed.),
Advances in instructional psychology (Vol. 2). Hillsdale, NJ: Lawrence Erlbaum Associates.
Raven, J.C. (1938). Progressive matrices: A perceptual test of intelligence, 1938, individual form.
London: Lewis.
Rumelhart, D. E. (1977). Introduction to human information processing. New York: Wiley.
Simon, H. A., & Kotovsky, K. (1963). Human acquisition of concepts for sequential patterns.
Psychological Review, 70, 534-546.
Spearman, C. (1923). The nature of intelligence and the principles of cognition. London:
Macmillan.
Sternberg, R. J. (1977a). Component processes in analogical reasoning. Psychologica
l Review, 84,
353-378.
Sternberg, R. J. (19776). Intelligence, information processing, and analogical reasoning:
The
componential analysis of human abilities. Hillsdale, NJ: Lawrence Erlbaum Associates.
Thurstone, L. L., & Thurstone, T. C. (1941). Factora studies of intelligence. Chicago:
University
of Chicago Press.
 Processes and Development
of Understanding

James G. Greeno
Mary S. Riley
University of Pittsburgh

This chapter has two main sections. In the first section some general remarks
about processes of understanding are presented. In these remarks two general
issues that are important to metacognitive and attributional processes, as well as
to understanding, are addressed. The first issue is whether metacognitive knowl-
edge, attributions, or understanding necessarily involves awareness. The second
issue is the relationship between various forms of metacognitive knowledge, in
particular, between knowledge that is involved in evaluating cognitive pro-
cedures and knowledge that is involved in planning and other strategic aspects of
performance.
In the second section the results of some empirical and theoretical research
regarding the development of children’s skill in solving a class of problems are
presented. These findings indicate that the major component that develops is
ability to understand the problems that are presented.

PROCESSES OF UNDERSTANDING

The ideas presented have been developed in the context of studies of cognitive
processes relevant to school mathematics. Specific analyses in this program have
been presented in previous reports (Greeno, 1978; 1980a; 1980b) and further
analyses are forthcoming.
In studying various phenomena involving understanding in mathematics, it
has become clear that the term “‘understanding’’ does not have a single referent.
Several quite different things are referred to when one speaks of understanding;
discussion is helped if these different referents are explicitly identified.

289
290 GREENO and RILEY

TABLE 10.1
Forms of Understanding

I. Theoretical: Related to General Principles

A. Explicit: Expressible in Language, Including Explanation

Example: Physics Experts (Chi, this volume)
B. Implicit
1. Evaluation of Performance
Examples: Geometry Proof Checking
Watch a Puppet Count (Mierkiewicz & Siegler, Note 2)
Adjust Counting Procedure (Gelman & Gallistel, 1978)
2. Analogical Use of Structure
a. Mapping of Procedures
Example: Subtraction with Numerals and Blocks (Resnick, 1982)
b. Transfer
Example: Learn Procedures in Relation to a Schema
(Wertheimer, 1945; Anderson, Greeno, Kline, & Neves, 1987)
c. Solving Problems of Application
Examples: Word Problems in Algebra (Paige & Simon, 1966)
Word Problems in Arithmetic
C. Conformity

HI. Intrinsic Structure of Performance

A. Knowledge for Planning

Examples: Abstract Planning Space (Newell & Simon, 1972)
Hierarchically Organized Actions (Sacerdoti, 1977)
B. Integrated Performance
Example: Independence from Visual Cues in Algebra (Davis, Jockusch, &
McKnight, 1978)
EE EE eee ee

One question of particular concern is understanding in the context of cognitive

procedures. This question of understanding a procedure is whether the person
performs with some understanding or whether performance is rote and mechan-
ical. Table 10.1, organized in two main categories, lists several forms of under-
standing that involve procedures. The forms of understanding that are called
theoretical involve some general principle or structure that is related to the
procedure; the forms that are called intrinsic involve factors that are included in
the procedure itself.

Theoretical Understanding
Many cognitive procedures, at least to some extent, are based on general princi-
ples. Many of the capabilities that are referred to as understanding relate to an
individual’s knowledge of such principles. For example, preschool children can
count small sets of objects correctly. It is theoretically and educationally impor-
 10. PROCESSES AND DEVELOPMENT OF UNDERSTANDING PAS

tant to consider whether this ability reflects more than the cognitive procedure of
counting. That is: Do preschool children understand principles of number and
quantity, such as the principle of cardinality and one-to-one correspondence? (cf.
Gelman & Gallistel, 1978). To address such questions, hypotheses about the
nature of understanding and about tasks for which children’s performance would
provide evidence of the extent of their understanding, are needed.

Explicit Knowledge. One form in which a person might understand a princi-

ple is to know how to state the principle in a language and to explain the
consequences of the principle to the procedure. This criterion of understanding is
often met by adult experts, such as the advanced physicists studied by Chi
(chapter 8, this volume). When individuals can state relevant principles and
explain why they are relevant, clearly, they are aware of the principles they
understand. On the other hand, one often attributes understanding, both to adults
and children, when individuals cannot state principles explicitly in a theoretical
language. This indicates that the criterion of explicit linguistic knowledge of the
principle, at least in some circumstances, is stronger than one wants to adopt. If
understanding is attributed to someone who is not able to state the principle, that
person may be unaware of the principle.

Conformity. Frequently, a lack of understanding is attributed to someone

whose performance does not conform to a principle. For example, if a child were
to perform a counting procedure with a set of objects, but then could not answer
the question, ‘‘How many are there?’’ one would probably conclude that the
child did not understand the principle of cardinality. When individuals’ perfor-
mance does conform to a principle, it is possible that they understand the princi-
ple, and that their understanding does not include awareness of the principle.
However, the attribution of understanding, simply on the basis of correct perfor-
mance, seems rather gratuitous, and would probably make it difficult to maintain
a coherent theoretical distinction between merely knowing a procedure and un-
derstanding the procedure. It can be concluded that the criterion of a procedure’s
conforming to a principle is not sufficient for attributing understanding of the
principle.

Implicit Understanding. The criterion of conformity is too weak; however,

the criterion of explicit linguistic knowledge is too strong for attributing under-
standing of a principle. The question that arises is whether or not there are
definite criteria that are stronger than conformity, but weaker than explicit
knowledge. We propose two such criteria; each criterion has a reasonable basis
in theory and the scientific literature. One of the criteria is an ability to evaluate
examples of performance according to the principle—that is, to judge whether or
not specific examples are consistent with the principle. Uhe other criterion is use
of the principle to identify structural properties shared between procedures in
 292 GREENO and RILEY

different domains, or to solve problems in a domain different from the one in

which the procedure was initially acquired.

Evaluation of Performance. There are two domains in which knowledge of

a concept is tested by determining whether or not an individual can correctly
distinguish between positive and negative examples of the concept. One is in
traditional laboratory studies of concept attainment (e.g., Bruner, Goodnow &
Austin, 1956), where subjects have the task of inducing a concept from informa-
tion about examples that are or are not members of a category, and demonstrate
that they have acquired the concept by correctly sorting a set of examples. A
second domain is in computational linguistics, where a knowledge of the syntax
of a language is represented by a recognition machine that successfully parses
sentences of the language, but fails to parse strings that are not well formed
sentences.
In procedural domains, one kind of understanding of a principle can be tested
by presenting examples of performance, and testing whether or not an individual
correctly discriminates between examples that are or are not consistent with the
principle. For example, in one current research project, Magone and Greeno
(Greeno, 1983) used the task of proof checking to test students’ understanding of
the principle of deductive consequence. It was found that students, who have
become quite competent in constructing proofs, often fail to detect errors in
proofs that are shown to them, especially when the errors involve an insufficient
basis in previous assertions for a step in the proof. There was also success in
instructing some students in a method of proof checking, designed to improve
their understanding of deductive consequence.
Counting by preschool children is another domain where understanding has
been tested by evaluation of examples. Mierkiewicz and Siegler (1980) recently
conducted a study in which children are shown a puppet counting a set of objects.
The puppet sometimes makes errors that violate general principles, such as the
principle of one-to-one correspondence. Children who identify these errors indi-
cate their knowledge of the relevant principle, especially when they also give
sound explanations supporting their evaluative judgements.
Ability to evaluate procedures is also tested when children are required to
adjust their method of counting. Gelman and Gallistel (1978) asked children
to
count a set of objects; designated one of the objects and said, ‘Make this the
ne
where n varied from one trial to the next. To accomplish this task correctly,
a
child must devise a new procedure for counting, different from the
preferred
method in which objects are counted in regular order from one end
to the other.
Children adjust their counting procedures in various ways, only some
of which
conform to general principles of counting. For example, if the designate
d object
is given a numeral that is smaller than it ordinarily would have, when
the number
is to be used, some children will skip to it, and not return to include
the skipped
objects in the count. This violates the principle of one-to-one correspon
dence and
 10. PROCESSES AND DEVELOPMENT OF UNDERSTANDING 293

results in an assignment of an incorrect cardinal number to the

set. However,
many children adjust their counting procedures in ways that conform
to the
general principles of counting. A reasonable inference is that, at least
by some of
the children, a process of choice is included in the adjustment. Children
who
choose adjustments that conform to the general principles perform an evaluatio
n
of alternative procedural components; those who choose correct procedures
indi-
cate that they understand the principles on which the choice is based.

Analogical Use of Structure. A problem is encountered in some domain,

and the person has not previously learned how to solve that problem. There are
problems in another domain, however, for which the person knows a solution
procedure; and a principle underlying that procedure applies to both of the
domains. If the procedure in the known domain can be used in solving the new
problem, or can be used to facilitate learning of a new problem-solving pro-
cedure, then the basis of that transfer, or new learning, can reasonably be
inferred to be understanding of the principle that is shared by the procedures in
the two domains.

Instructional Mapping of Procedures. An example of the understanding of a

principle in learning of a procedure is in research conducted by Resnick (1982).
Resnick has developed an instructional method for teaching children a correct
procedure for subtraction of multidigit whole numbers. The method uses place-
value blocks: small cubes correspond to units, bars (10 long) that correspond to
tens; and flat shapes (10 by 10) that correspond to hundreds.
Resnick identified children whose performance on subtraction problems was
incorrect in a systematic way, which is interpreted as indicating that the child
lacks understanding of some significant arithmetic principle. The procedure with
blocks was designed in a way that makes some important contraints quite salient.
For example, the procedure with blocks involves taking blocks away from a set
that is present; if the number to take away is greater than the number that is
present, there obviously is an impediment to performance. The procedure with
blocks requires that the child know a correspondence between the representations
of numbers with numerals on paper and with blocks.
When the procedure with blocks had been learned, Resnick instructed the
child in a set of correspondences between procedures with the blocks and with
numerals on paper. Each time a change is made in the problem with blocks, a
corresponding change must be made in the numeral-paper representation. For
example, when a child trades a ten-block for ten unit-blocks, there are two steps
(removing one ten-block and bringing in ten units); each of these steps corre-
sponds to a change that must be made in the numerals (crossing out the tens-
numeral, replacing it with the next-smaller numeral, and inserting a ‘‘one’’ next
to the units-numeral). When the child had learned to perform subtraction in the
blocks and numerals simultaneously, Resnick gradually removed the blocks from
294 GREENO and RILEY

the procedure, so that the child performed subtraction correctly with just
numerals.
Used in this way, the blocks provide a basis for some children acquiring
correct procedures for subtraction. There is suggestive evidence that this is
accomplished through acquisition of general principles of arithmetic that are
abstract enough to be applied both to numerals and to blocks. For example, one
child was asked how the procedure that had been learned differed from the one
the child had used previously. The child answered, ‘‘Before, I just took the
numbers apart. Now, I leave them together, and take them apart too.’’ This
provides evidence that the general principle shared across the two domains was
understood by the child.

Transfer. Ability to transfer procedures to novel problems provides rela-

tively strong evidence that a person understands general principles in the solution
(e.g., Gagne, 1966; Katona, 1940; Wertheimer, 1945/1959). Wertheimer’s best
known example is the problem of calculating the area of a parallelogram.
Wertheimer argued that the procedure of calculating, using the formula A = b X
h, could be mechanical or could involve understanding. He characterized under-
standing as including appreciation of spatial relations that involve the way a
parallelogram can be transformed into a rectangle. He described children’s per-
formance on various transfer tasks, as evidence of their understanding.
In a recent analysis of meaningful learning, Greeno (1983) has developed an
hypothesis about the cognitive process involved in learning that can be transfered
to novel problem situations. According to this analysis, transfer is enabled by a
schema that can be used to represent the structure of problems in the domain
where a procedure is learned, as well as in the domain of transfer. Simulations of
learning were programmed to clarify the difference between meaningful learn-
ing, with procedures embedded in a schema, and more mechanical learning, in
which procedures are associated with specific stimulus conditions of a problem
situation. The learning that was simulated involved geometry problems in which
line segments are related to one another as parts of larger units. If a learner
acquires proof procedures that are embedded in a schema of part-whole rela-
tionships, then those procedures can be used in another domain, such as adjacent
angles, where the objects can also be organized in terms of part-whole rela-
tionships. The simulation provides an explanation of transfer to a problem of
proving the congruence of vertical angles, discussed by Wertheimer
(1945/1959).

Problems of Application. A third kind of indication of understanding is

ability to solve problems of application, as when algebra formulas can be used to
solve word problems. Paige and Simon (1966) found that a variety of processes
involved in understanding problems led to different representations, some
of
which enabled subjects to detect inconsistencies in information in problem state-
 10. PROCESSES AND DEVELOPMENT OF UNDERSTANDING 295

ments. Mayer and Greeno (1972), and Egan and Greeno (1973) studied different
methods of instructing subjects about the binomial formula for calculating proba-
bilities of outcomes in random experiments. They found that more meaningful
instruction led to better performance in application problems, as well as in more
frequent detections of inconsistent or insufficient information in problem
statements.
Another example of problem solving mapping between a verbal and mathe-
matical domain is in story problems that are solved by primary grade children,
using simple addition and subtraction of numbers. Research based on problems
in this domain is discussed in the second part of this chapter.

Intrinsic Understanding
Most frequently when one speaks of understanding a procedure or solution of a
problem, one refers to a relationship to some general principle. However, one
can also speak of understanding of a less general kind, involving apprehension of
the intrinsic structure of a problem solution or a problem-solving procedure.
Greeno (1977) has proposed that understanding a problem solution is analogous
to understanding a sentence, where understanding consists of constructing a
representation that includes relations among the ideas that are mentioned in the
sentence. In this sense, understanding a problem solution or problem-solving
procedure would consist of constructing a representation of the solution or pro-
cedure, including meaningful relations among the various steps of the solution or
the various cocomponents of the procedure. Understanding of this kind may be
analogous to metacognitive knowledge, consisting of strategies for performance
(Brown, chapter 3, this volume; Kluwe, chapter 2, this volume).
Two forms of intrinsic understanding can be distinguished. The first form is
an aspect of strategic knowledge, involving knowledge for planning solutions of
specific problems. The second form, which may or may not be a consequence of
the first, consists of integrated knowledge that enables a person to perform a task
in a smooth, coherent manner.

Knowledge for Planning. Understanding of a sequence of things often in-

cludes realizing the way in which subsets of the things fit together in groups. In
language, understanding of a sentence includes realizing the sets of words that fit
together in phrases. In the solution of a problem, understanding includes realiz-
ing the sets of problem steps that fit together in meaningful units. It seems
reasonable to classify the ‘‘phrases’’ of a problem solution as sets of steps that
are related to a single major goal of the problem, or that are the details of a single
global action. Newell and Simon (1972) used the term ‘‘episode’’ to refer to such
units. If a person understands another’s solution for a problem, then the episodes
in the solution should be apparent, and the understanding should include the
subgoals that motivated the activity in episodes, as Brown, Collins, and Harris
296 GREENO and RILEY

(1978) observed. Similarly, if a person’s own solution of a problem is accom-

plished with understanding, then the various steps should be organized in a
meaningful way, with individual steps performed in relation to subgoals that
determine an organization of episodes in the solution.
Knowledge of the kind required for this kind of understanding has been
characterized as knowledge for planning. Newell and Simon (1972) hypoth-
esized that planning requires a representation of the problem that distinguishes
between essential and inessential problem features. A plan is constructed as a
sequence of subgoals that involve essential features; execution of the plan re-
quires accomplishing these subgoals, including working out details correspond-
ing to inessential features of the solution. A similar idea was developed by
Sacerdoti (1977), who hypothesized that knowledge required for planning con-
sists of information about global actions. These global actions correspond to sets
of more elementary actions that can be performed directly in the problem en-
vironment. Organized problem solving occurs when an individual constructs a
sequence of global actions that can lead to the main problem goal, and then, in
turn, works out the details of each of the global actions.
Knowledge used in planning is metacognitive, in the sense that it is about
cognitive procedures. In order to plan one’s own problem-solving activity, or to
understand the plans involved in another’s problem solutions, a person requires
knowledge about the individual procedural components in a solution that shows
how the components are grouped into episodes and related to subgoals. This
metacognitive knowledge does not necessarily include explicit representation of
general principles, nor is it necessarily accompanied by any self-conscious pro-
cesses that would constitute awareness of the cognitive organization that it
determines.

Integrated Performance. A kind of understanding is often attributed to

someone who has become highly skilled in a task. For example, a musical
performer, after studying and practicing it, may say that he or she understands a
piece; whereas, initially the piece could be played successfully, but without a
sense that it is understood. It seems reasonable to hypothesize that the under-
standing that results from practice includes appreciation of relationships within
the piece that were not appreciated initially. There relationships might well be
provided by a structure of knowledge involving global actions or subgoals like
those involved in planning.
One consequence of developing higher order cognitive units, needed to
organize problem solving in a skilled way, has been noted by Davis, Jockusch,
and McKnight (1978). This consequence, observed in students as they became
more skilled in algebra, was that their performance became less dependent on the
explicit intermediate results that they wrote out as they solved problems. Davis et
al. referred to performance in early stages of practice as visually modulated
sequences, in which intermediate results are written and used as cues for con-
 10. PROCESSES AND DEVELOPMENT OF UNDERSTANDING 297

tinuation of the performance. They hypothesized that practiced students have

acquired general schematic action structures that enable them to direct their
performance without using the visual cues provided by written intermediate
results. This kind of general schema might also be responsible for the ability to
perform planfully, and therefore would constitute understanding, in the same
way that understanding may be attributed to a person with knowledge for
planning.

DEVELOPMENT OF ABILITY TO UNDERSTAND

In this section some empirical findings regarding development of skill in problem

solving are presented. There is evidence that in the specific tasks studied, the
main developmental component that accounts for older children’s better perfor-
mance is acquisition of improved ability to understand problems.
The problems studied involve children’s knowledge of numbers and arith-
metic. The understanding that is required is one of the forms of implicit the-
oretical understanding, discussed earlier. Specifically, these problems require
mapping from a problem domain, described in some text, to the domain of
arithmetic or counting. Relationships described in the problem text must be
understood, and problem-solving procedures are chosen on the basis of the
understanding that is achieved.

Theoretical Framework and Issues

The framework used in this discussion is sketched in Fig. 10.1. Important com-
ponents of this analysis were contributed by Joan Heller (see Riley, Greeno &
Heller, 1983). The terms Problem Schemata and Action Schemata, shown in
brackets, refer to knowledge that the problem solver is required to have stored in
order to perform the tasks. Problem schemata are general conceptual structures
that are used in representing the information in problem texts. These are struc-
tures like those hypothesized in recent analyses of language understanding (e.g.,
Norman & Rumelhart, 1975; Schank & Abelson, 1977). Action schemata are
general procedural structures that function as plans for problem-solving activity.
These are knowledge structures, like those hypothesized in recent analyses of
knowledge for problem solving, involving knowledge about actions at different
levels of generality (e.g. Brown & Burton, 1978; Sacerdoti, 1977).
It is hypothesized that the process of solving a problem involves three main
processes, which do not necessarily occur in a strict sequence. One process is
comprehension of the text, resulting in a representation of the concepts and
relationships that are described. A second process, using the information in the
problem representation, is selection of an appropriate set of problem-solving
298 GREENO and RILEY

. | Problem Action
Schemata Schemata

Schematized Schematic
Problem (2) Action
Representation >? Representation

In |»

Problem Solution
Text

FIG. 10.1. Framework of a model of problem understanding and solution. Ar-

rows represent processes of (1) comprehension, (2) mapping from conceptual
relations to quantitative procedures, and (3) execution of procedures.

procedures. Finally, the procedures that have been chosen are executed to obtain
the problem solution.
This framework suggests alternative hypotheses about possible sources of
developmental improvement in problem-solving skill. If it is decided that some
task requires a specific kind of action schema, the older children perform better
on that task than younger children, one might conclude that the older children
have acquired that schema that younger children lack. This kind of inference was
typical in Piaget’s work.
However, another possibility is that younger children have the relevant action
schema for solving a problem, but that schema is not activated by the situation in
which the problem is presented. In principle, failure of activation could occur for
any of three reasons: first, linguistic processes needed to translate text into a
semantic representation could be lacking; second, connections between the prob-
lem representation and action schemata might be lacking; and third, the required
problem schemata might be lacking. Evidence is presented that seems to rule out
the hypothesis that children lack the action schemata needed for the problems
studied. We currently favor a hypothesis that combines the first and third alter-
native possibilities; that is, the knowledge that children acquire in becoming
skillful in simple arithmetic word problems includes procedures and general
conceptual knowledge that enables them to represent problem information
successfully.
The conclusion that we reach in our analysis puts us squarely in the middle
of
a Current issue in the literature of cognitive development. On the basis
of failures
in various tasks, Piaget and his associates concluded that children lack
under-
standing of some very important concepts: conservation of number;
class inclu-
sion; seriation; and so on. Evidence for these failures came from
performance
that is inconsistent with the general concepts. For example, when
a child sees
 10. PROCESSES AND DEVELOPMENT OF UNDERSTANDING 299

two sets with the same number of objects and says one set has more, that
performance is inconsistent with the concept of number conservation. On the
other side, numerous investigators have shown that children will show perfor-
mance that is consistent with thise principles in other circumstances. For exam-
ple, Gelman and Gallistel (1978) provided evidence that preschool children
understand principles of number, and Trabasso, Isen, Dolecki, McLanahan,
Riley, and Tucker (1978) summarized evidence that children, under appropriate
circumstances, show that they understand the principle of class inclusion.
Our conclusion concurs with recent investigators that the standard Piagetian
conclusion of structural deficiency is too simple; yet another instance is added, in
which an action schema that was thought to be lacking in young children has to
be present for them to perform as they do in a modified task. However, we also
conclude that the difference between younger and older children is more pro-
found than simply acquiring the meanings of a few linguistic terms. The most
plausible hypothesis, regarding the problems analyzed, is that younger children
have the general procedures needed to solve the problems, but they lack ability to
coherently represent problem information. The ability to construct those repre-
sentations constitutes a set of important conceptual advances that older children
have achieved.

Evidence for Action Schemata

Consider the following example: ‘‘Joan has $8 and Sue has $11. How many
more dollars does Sue have than Joan?’’ The answer is found by subtracting 8
from 11, but because quantities must be identified and the decision to subtract
numbers requires understanding of quantitative relationships, a more compli-
cated procedure is needed.
Figure 10.2 shows one procedure for solving problems that involve the com-
parative question, “‘How many more?’’ In thinking about this procedure, it is
convenient to think of a child using some blocks to work out the solution.
However, a similar procedure also may be used when blocks are not used, in
which case the operations involving sets refer to mental representations of
quantities.
Apply the procedure in Fig. 10.2 to the example as follows: X and Y are Sue’s
and Joan’s dollars, respectively; N1 is 11, N2 is 8. In the first two steps,
representations of these sets are formed; with blocks, two piles would be made;
11 in one pile and 8 in the other. The third step involves identifying a subset ofX
that has the same number of blocks as Y; in the description of the procedure this
subset is called Z. Physically, a child might form pairs of blocks, with one block
from X corresponding to each block in Y. In any case, Step 3 of the procedure
depends on the idea of one-to-one correspondence. In Step 4, the complement of
Z in X is formed and brought into attention. In working with blocks, the set W
would just be the blocks remaining, after Z was formed. Step 4 depends on the
 300 GREENO and RILEY

Compare X, More, Y
(X has N1, Y has N2)

— Makeset: X, N1
N Makeset: Y, N2
3. Make Match: Z in X match Y
a. Form subset of X equal to Y (one-to-one correspondence)
b. Z = matching subset
4. Make Remain: W in X from Z
a. Separate X: Z and remainder (class inclusion)
b. W = remainder
5. CountW

FIG. 10.2. Procedure for Answering, ‘‘How many more?”

idea of class inclusion. Finally, the answer to the question is obtained by count-
ing the number of blocks in W.
This problem is interesting, partly because it is quite difficult for children.
Table 10.2 shows data collected by Riley as part of a study of the development of
children’s ability to solve simple arithmetic word problems. Most of the kinder-
garten children were 6 years old, most of the first grade students were 7 years
old, and so on. The first two columns show data for problems asking, ‘‘How
many more in one set than another?’’ The next two ask, ‘‘How many less or
fewer in one set than another?’’ The problems in the last two columns gave a
quantity, and a number that was lost or taken away, and asked, ‘“‘How many
resulted?’” An example of the last kind involving money would be: ‘‘Sue had
$11 and spent $8. How many does she have now?’’ The columns marked ‘‘With-
out Blocks,’’ show data when the children were simply given the problems
orally. The columns marked ‘‘With Blocks,’’ show data when children were
given the same problems, but where there were small wooden blocks that the
children could use to represent the numbers in the problems. Note that questions
of *‘How many more?’ or ‘‘How many less?’ pose significant difficulty
for
children as old as 8 years; however, other problems solved by subtraction
are
quite easy for children, before they enter first grade.
Why are problems involving comparison so difficult for younger children?
It
is tempting to conclude that they lack the cognitive procedures required
to find
the answer. In Riley et al’s (1983) analysis, processing requires
formation of
subsets, one of which is formed by putting its members in one-to-one
correspon-
dence with members of another set. A conclusion that these processes
are not yet
in the repertoire of young children would be consistent with Piaget’s
(1952) well
known findings regarding class inclusion and conservation
of number.
This interpretation, however, is contradicted by data collecte
d by Hudson
(1980), who presented problems of the kind shown
in Firy103,26.12 nursery
school, 24 kindergarten, and 28 first grade children. Two
different questions
 10. PROCESSES AND DEVELOPMENT OF UNDERSTANDING 301

TABLE 10.2
Proportions of Correct Solution

How Many More? How Many Less? How Many Now?

School Without With Without With Without With

Grade Blocks Blocks Blocks Blocks Blocks Blocks

K 3 ele? “13 .04 61 1.00

l 33 .28 sll 22) 1.00 1.00
2 .65 .85 .65 75 1.00 1.00
3 1.00 1.00 1.00 1.00 1.00 1.00

were asked. One question was the usual comparative question, in this case,
“How many more birds than worms are there?’’ The other question was an
alternative that Hudson devised: ‘‘Suppose the birds all race over and each one
tries to get a worm? How many birds won’t get a worm?’’ The results were
striking; they are shown in Table 10.3. Hudson gave different questions of each
type, and the proportions indicated are for children who gave six or more re-
sponses of the same kind. Correct responses were the set differences—for exam-
ple, one more bird than worms, or one bird won’t get a worm. The most frequent
incorrect response was the number in the larger set—for example, “‘How many
more birds than worms?’’ ‘‘Five’’; or ‘‘Five birds.’’ Another frequent error was
to give both set sizes, for example: ‘Five and 4,”’ or ‘Five birds and 4 worms. 5
Very few of the nursery school or kindergarten children answered the ‘How
many more?’’ questions by giving the difference between the sets. However,
nearly all the children of all three ages answered the *‘How many won’t get?”
questions correctly.
Hudson’s results clearly contradict the hypothesis that young children lack a
procedure for finding a set difference. To answer ““Won’t get’’ questions, chil-
dren carried out a procedure essentially like the one in Fig. 10.2; forming the
correspondence of sets indicated in Step 3, and then counting the remaining
subset of the larger set. Many of the children did this overtly, pointing with their
fingers, either in counting equal subsets or in forming pairs of objects from the
two sets. Because children clearly have this procedure in their repertoire, one
must look elsewhere for an explanation of their difficulty with the *‘How many
more?’’ questions.
An alternative interpretation of young children’s difficulty is that they have
not yet acquired procedures for representing the information in the ‘“‘How many
more?’’ problem, including the meaning of the question; that is, they do not yet
have the schema for comparison needed to form representations that include
comparative relationships between quantities. If it is assumed that the restate-
ment of the question enables children to activate procedures for forming corre-
sche-
sponding sets and counting the remainder without forming an intermediate
 302 GREENO and RILEY

= e
= =
FIG. 10.3. Example Problem From Hudson’s Experiment

matized representation of the problem information, this interpretation is

consistent with Hudson’s findings. In Fig. 10.1, this would involve a process
going directly from problem text to schematic action representation—an arrow
diagonally from the lower left to the upper right of the diagram. This seems
a
reasonable hypothesis about the effect of Hudson’s form of the question. The
remark that birds will race over and each will try to get a worm, communicat
es a
strong suggestion about forming corresponding sets; the question How many
won’t get a worm? focuses directly on the complement of the set formed
by
correspondence.

Analysis of Developmental States

We propose a hypothesis that the main form of growth in problem-
solving skills
consists of processes of understanding. However, a hypothesis
that older chil-
dren solve problems more successfully because they understand
better does not

TABLE 10.3
Proportions of Children with Consistent Correct Responses

Grade How Many More? How Many Won't Get?

Nursery School sll .83

Kindergarten 23 .96
First .64
en 1.00
 10. PROCESSES AND DEVELOPMENT OF UNDERSTANDING 303

answer the question of developmental change adequately; it merely identifies a

domain in which to look for an answer. An adequate answer requires an analysis
of the ability to understand problems, as well as an analysis of the changes that
result in the cognitive abilities that are required in order for understanding to
occur. In this section the results of an empirical developmental study of skill in
solving arithmetic word problems are presented, and some theoretical conclu-
sions about the states through which children develop as they acquire stronger
ability to understand these problems, are described.

Empirical Analysis. An empirical study was conducted by Riley, in order to

provide a reasonably thorough descriptive account of children’s performance on
a variety of problem-solving tasks. The problems were selected on the basis of an
analysis by Heller and Greeno (see Greeno, 1980a; Riley, Greeno & Heller,
1983) of a set of semantic schemata that forms a sufficient basis for understand-
ing the variety of arithmetic word problems that are solved by addition or sub-
traction. The three general categories of problems are called Change, Combine,
and Compare. In a Change problem, there is a quantity that is either increased or
decreased by some event. For example: ‘‘Kurt had $11 dollars and he gave $8 to
John. How many dollars does Kurt have now?’’ In the second kind of problem,
called Combine, there are two quantities; neither quantity is changed; the prob-
lem solver must think of the combination of the quantities. For example, ‘‘There
are five boys and six girls. How many children are there?’’ Finally, in Compare
problems there also are two quantities that do not change; the problem solver
must think of the difference between them. For example: ‘“There are 5 girls and
2 boys. How many more girls are there than boys?’’ The problem types are listed
in Table 10.4.
In Riley’s experiment, all three kinds of problems were presented to children.
A further difference among problems also was used, as shown in each kind of

TABLE 10.4
Kinds of Problems

Change Compare

Direction Unknown Direction Unknown

1) Increase, Result 1) More, Difference

2) Decrease, Result Dismlsess® Difference
3) Increase, Change 3) More, Compared
4) Decrease, Change 4) Less, Compared
5) Increase, Start 5) More, Referent
6) Decrease, Start 6), Less; Referent
Combine, Unknown
1) Combination
2) Subset
ee en Dun er ee ET Ee
 304 GREENO and RILEY

problem—Change, Compare, and Combine—there are three items of informa-

tion. Different problems can be formed by varying the items of information
given, and by varying the item to be found by the problem solver. In Change
problems, the three items are the starting quantity, the amount of the change, and
the resulting quantity. Any of these can be found if the other two are given,
therefore, there are three different cases: the unknown quantity can be the start;
the change; the final result. Furthermore, the change can either be a decrease or
an increase; therefore, there are six kinds of Change problems. A similar set of
variations exists for Compare problems, where the direction of difference may be
more or less; the unknown quantity may be the amount of difference (How many
more? or How many less?), the quantity used as the referent (that is, in ‘‘A has 5
less than B,”’ B is the referent that A is compared to); or the quantity that is being
compared. In Combine problems there are fewer semantic categories—the un-
known is either the combined quantity or one of the subsets.
In her study, Riley used all these kinds of problems, including subsets of the
Combine category that varied in the order of presenting the information; this
resulted in a total of 18 problems. Eighty-one children were interviewed: 23 from
kindergarten, 18 from first grade, and 20 each from the second and third grades.
Each child was interviewed twice. In the first session, the 18 problems were
presented orally; the child was asked to give an answer and write an arithmetic
expression for the problem. In the second session the same problems
were
presented; the child was asked to repeat each problem and then solve it using
some wooden blocks.
Only a few summary findings from Riley’s experiment are presented. Table
10.5 shows average performance for children of different ages on the problems
in
the three general semantic categories. First, as one would expect, older
children
performed better on these problems than younger children. Second,
the three
categories of problems were not equally difficult. The Compare problems
were
harder, on the average, than the Combine or Change problems.
A reasonable hypothesis is that there are conceptual schemata
corresponding
to these three problem types—a schema of quantitative change,
a schema of
combination, and a schema of comparison. It might be temptin
g to speculate that
these three schemata emerge at different times in cognitiv
e development. For

TABLE 10.5
Proportion Correct by Grade and Problem Type
e ee
Grade Change Combine Compare

K .48 .42 hl
| .66 .60 21
2 .89 .83 .62
3 .90 295
m .88
 10. PROCESSES AND DEVELOPMENT OF UNDERSTANDING 305

TABLE 10.6
Proportion Correct by Unknown
Quantity

Unknown

Type 1 2 3

Change .95 sl 5
Combine .96 all
Compare 32 32 33

example, at a certain age, a specific child might have the schemata of change and
combination, but not the schema of comparison. However, data given in Table
10.6 show that such a view would be too simple. Table 10.6 shows average
performance for problems of the same general category, differing in the items of
information that were unknown. The data show that the different problems of
each semantic type differed in difficulty. For example, children had no difficulty
with change problems if the problem gave the start and change amounts and
asked for the result. If the result and change were given, with the starting
quantity unknown, most of the younger children had difficulty. The remaining
case, with the start and result given, was intermediate. These findings show that
one must consider specific features of problems in an effort to understand the
problem-solving skills of children at different ages.
Another result involves the difference between performance with and without
blocks as aids for solving the problems, shown in Table 10.7. On the average,
use of the blocks facilitated solution of problems, especially for the youngest
children. The facilitating effect of having objects available for representing as-
pects of the problems is not surprising, and should be taken into account in
developing a theoretical analysis of the development of problem-solving skill.
Theoretical Analysis. The framework used in this analysis was discussed
earlier. Recall that two main components of knowledge are postulated: schemata

TABLE 10.7
Proportion Correct by Problem-Solving
Situation

Grade Without Blocks With Blocks

K .26 .4l
l 50 49
> “fp 84
3 91 95
 306 GREENO and RILEY

for representing problems, and schematic representations of actions for planning

problem solutions.
It is assumed that there are three schemata used in understanding simple
arithmetic story problems: a schema for Change problems; a schema for Com-
bine problems; and a schema for Compare problems. Fig. 10.4 is a diagram of
the structure that is hypothesized as the schematic representation of one of the
forms of the Change type of problem; a problem involving a decrease
in a
quantity, with the result unknown. The problem might be, ‘‘Joe had 8 rabbits;
5
ran away; how many rabbits does Joe have now?’’ The model developed
to
simulate understanding of such problems builds the first component, labeled
(1)
in Fig. 10.4, when it receives ‘‘Joe had 8 rabbits.’’ Then when a
sentence such
as “Five ran away.”’ is received, it infers that the problem is about
a change, and
constructs the nodes labeled (2) in Fig. 10.4. Finally, when the question
‘‘How
many now?’ is received, the structure is completed by constructing
the nodes
labeled (3) in Fig. 10.4.
In the analysis of developmental change that was worked out
by Riley et al.
(1983), different stages of skill are represented by different
abilities, for repre-
senting problems and performing actions. Different schemat
a are required to
represent the three problem types. Change problems include
start, change, and
result quantities as in Figure 10.4. Combine problems
require representations
involving three quantities, two of which are parts of
the third quantity, the
whole. For Compare problems, a schema was hypothe
sized that includes a
difference between two quantities, one of which is identifi
ed as the referent and
the other as the comparison quantity.
The action schemata hypothesized include the follow
ing: (MAKESET X N),
which takes a number and creates a set of objects with
that number in it, called X:

(3)

Cause Result

(2) (3)
Quantity

State Action Identiity Je Amount

“ IS
(1) eon Rec Joe

Identity Amount Identity

a Amount
= — SN

FIG. 10.4. Schematized Representation of a Problem

 10. PROCESSES AND DEVELOPMENT OF UNDERSTANDING 307

(PUTIN X N), which adds a number of things to an existing set; (TAKEOUT X

N) which removes a number of things from an existing set; and (COUNT X),
which determines the number of objects in a set. Fig. 10.5 shows a problem-
solving plan for the problem whose representation is shown in Fig. 10.4. The
problem-solving system assembles this plan from the action schemata, when the
representation in Fig. 10.4 has been formed. In addition to the schemata men-
tioned, the repertoire of action schemata includes structures, such as those men-
tioned in Fig. 10.2; (MATCH Y in X) and (SEPARATE X: Z and remainder).
In Riley’s analysis of developmental states, different stages of skill are repre-
sented by different schemata for representing problems. Children with more
problem-solving skill are assumed to have an ability to represent problems more
completely; that is, they have more detailed schemata for representing rela-
tionships among quantities. Their representational skills also include capabilities
for representing some items of information internally, rather than depending on
external realizations of quantities in the form of sets of objects.
Riley’s analyses were based on systematic patterns of performance that she
observed in her developmental study. Within each set of problems—those in-
volving Change, Combine, and Compare problems—she was able to identify
two or three graded subsets of problems. The easiest subset was a group of
problems that were solved correctly by nearly all the children, including the
kindergarten children. A second subset of problems was solved correctly by
some, but not all of the children; in one case there were two intermediate subsets.
For many of the problems in which errors occurred, children tended to give
characteristic errors—that is, a particular incorrect response occurred with quite
a high frequency.
An example is in Table 10.8 which presents data from children’s perfor-
mance, when they had blocks available to aid in solving the problems; the
numbers given as errors are illustrative. Different children solved these problems
with different numbers; therefore, a characteristic error indicates the kind of
answer given. For example, in problem 3, the specific number that was given,
whatever that number was, was the final result.
When blocks were available, nearly all the children solved problems I, 2 and
4 correctly. One problem, number 3, was of intermediate difficulty, and prob-
lems 5 and 6 were the most difficult. Children who made errors on problem 3
usually said ‘‘8’’ (giving the final result rather than the amount of change); *'S””

Change set x, Decrease, 5

(set x has 8)

Makeset x, 8 Count x

FIG. 10.5. Schematic Plan for Problem Solution

 308 GREENO and RILEY

TABLE 10.8
Patterns of Performance on Change Problems
ps— su. eek

Levels of Performance
Example Problems 1
ee WA ee a 2 3
Wi ee Mn Sr en a
Result unknown
l. Joe had 3 marbles.
He found 5 more marbles.
How many marbles did Joe have then? + == +
2. Joe had 8 marbles.
He lost 5 marbles.
How many marbles did Joe have then? aS ae 3
Change unknown
3. Joe had 3 marbles.
He found some more marbles.
Then Joe had 8 marbles.
How many marbles did Joe find? us. == =:
4. Joe had 8 marbles.
He lost some marbles.
Then he had 3 marbles.
How many marbles did Joe lose?
F ++ an
Start unknown
5. Joe had some marbles.
He found 5 more marbles.
Then he had 8 marbles.
How many marbles did Joe have to begin with?
es: Se +
6. Joe had some marbles.
He lost 5 marbles.
Then he had 3 marbles.
How many marbles did Joe have to begin with?
NA NA +

was a characteristic error for problem 5 (giving the amount of

change instead of
the starting amount). The characteristic error in problem 6 was
to remain silent,
looking puzzled.
Riley developed hypotheses in the form of models that
represent skill at
different levels. The theoretical goal was to simulate
the different levels of
children’s performance on these problems. That is, she
developed hypotheses
about the kinds of information-processing components
needed to explain the
different patterns of children’s performance on the various
problems. The models
differ in the ways in which quantitative information
is represented.
Riley developed three models for different levels
of performance on the
Change problems. In the simplest model, the only proble
m schema for represent-
ing quantitative information, identifies an object
(e.g., Joe’s rabbits) with a
quantitative value. Action schemata that are
used include: MAKESET: PUTIN :
TAKEOUT; and COUNT. This model is sufficient to solve the first two prob-
 10. PROCESSES AND DEVELOPMENT OF UNDERSTANDING 309

lems correctly. A starting set is created, and then either increased by PUTIN or
decreased by TAKEOUT. When the question is asked, the resulting state is
created and counted.
The main difference between Riley’s models for Level | and Level 2 is that at
Level 2 it is assumed that additional information is represented internally. This
model can construct representations using the schema for change problems,
shown in Fig. 10.4, where a mental record is kept of the structural role of each
item of information. This additional structural information enables the model to
give the correct answer to Problem 3. The external actions with blocks are the
same on the first and third problems—first making a set of 3 and then increasing
the set to 8 (either with PUTIN or with MAKESET). The Level I model, without
a record of the structural information, cannot distinguish these two situations;
however the Level 2 model identifies the set of 8 as the result, and remembers
that initially that set had 3; therefore, it is able to infer (or remember) the amount
of the increase correctly.
Notice in Table 10.8 that although Problem 3 was answered incorrectly by
children at the lowest level of performance, typically, Problem 4 was solved
correctly by all children; this probably was fortuitous. When a set is made
initially and then reduced (in this case to 3) the removed set of blocks is still
available for inspection. Thus, children were able to give the correct answer to
Problem 4, even if they did not keep a memory record of the structural
relationships.
Finally, Riley’s Level 3 model employs a schema for combination of quan-
tities shown in Fig. 10.6, and constructs plans using the following action sche-
mata: (MAKESET X ?), which constructs a representation of an unknown quan-
tity; (JOIN S T U), which forms a superset by putting two subsets together; and
(SEPARATE U S T), which forms two subsets from a superset. Riley hypoth-
esized that problems with the start unknown are understood by a kind of transla-
tion into concepts of set combination. Consistent with this idea is the fact that the

Parts

Quantity

Part 1 Part 2 Identity Amount

i
Identity Amount fe y
Identit ul
Amount

FIG. 10.6. Schema for Combination of Quantities

310 = GREENO and RILEY

difficulty of problems 5 and 6 was very similar to the difficulty of Combine

problems with subsets unknown, for each group of children that Riley observed.
Riley also developed models of levels of performance on the Combine and
Compare problems, in which different levels of skill are simulated with represen-
tational components analogous to those described for the Change problems. At
the lowest level, Riley assumed that the child’s representations of problems are
limited to the external displays of blocks; at an intermediate level some relevant
additional information is represented internally; and at the most advanced level
there are features represented internally that direct the understanding and actions
in a top-down manner.
A test of these models was obtained by determining how many of the children
in the experiment performed in one of the ways consistent with a model. Table
10.9 shows the proportions of children in the various age groups whose perfor-
mance was consistent with each of the modes at each age level. The data are from
performance when blocks were available. Performance without blocks involves
retention in memory, which produces additional complications. The columns
labeled 3a for Change and Compare problems indicate children whose perfor-
mance exceeded Level 2, but included an error on a problem that would be
performed correctly by a child who had Level 3 knowledge. Keep in mind that
the models were developed after inspection of the data for patterns of systematic
performance. Even so, it is encouraging that explainable patterns could be found

TABLE 10.9
Proportions of Patterns Consistent with Models
ee ee ae
Level Grade

K 1 2 3
er re WIRD ul ne Bus
Change l 85 .39 0 0
Problems 2 .39 zZ .10 .05
3a 13 7 .30 15
3b 09 22 60 80
Residual .04 .05 0 0

Combine l 92 28 .05 0
Problems 2 22 .28 .05 0
3 al 22 .85 .95
Residual .09 22 .05 05

Compare l .48 0 0 0
Problems 22 22 AST .10 0
3a 04 an 45 45
3b 0 .05 20 SR
Residual .26 17, 29 0
 10. PROCESSES AND DEVELOPMENT OF UNDERSTANDING 311

that showed this degree of consistency with the children’s performance on these
problems.

CONCLUSIONS

Uniqueness cannot be claimed for the models developed in these analyses. In-
deed, there are some redundant features in the models; somewhat simpler ac-
counts could be given to explain the observed improvements in skill. Even so,
we are quite confident that the children’s improvement in skill in these problems
involves something along the lines of these models. The more skillful models
have their skill because they understand problems better. That is, their represen-
tations of problems include the relevant features of the problem more com-
pletely, and in ways that lead to choice of appropriate problem-solving actions.
In the final section the results of Hudson’s (1980) and Riley’s developmental
studies in relation to the ideas about understanding, presented earlier, are
discussed.
In the tasks studied by Hudson and Riley, solutions for problems are obtained
by procedures for making quantitative inferences. In analyses presented here
procedures have been characterized as counting operations; however, similar
analyses would result if more advanced arithmetic operations, such as addition
and subtraction of numbers, we were considered. The problems that children are
asked to solve are presented verbally, or with pictures. Thus, in order to choose
an appropriate problem-solving procedure, the child is required to make some
kind of translation from the verbal or verbal/pictorial presentation. There is
evidence that this translation is the major source of difficulty encountered by
young children for these problems, and that the major developmental changes
that enables older children to perform successfully is acquisition of procedures of
understanding that construct representations of problem situations.
Strong evidence for the importance of understanding is provided by Hudson’s
findings, which show that a change in the wording of problems greatly facilitates
young children’s ability to solve them. Similar findings have been obtained for
other problem domains. For example, Markman (1979) has shown that use of a
class term, rather than a collection term (e.g., ‘‘forest’’ rather than ‘“trees’’)
greatly facilitates children’s performance on class-inclusion problems. It may be
that in many domains, young children have acquired general procedures for
making inferences before they have been thought to do so, and, because of
undeveloped skills in understanding, perform weakly on cognitive tasks.
In our analysis of the understanding process that children acquire, based
largely on Riley’s developmental study, we hypothesized that older children
have procedures for constructing general representations of quantitative rela-
tionships. These schematized representations mediate between problem texts and
problem-solving procedures. They represent general structural features of prob-
 32 GREENO and RILEY

lem situations; therefore, the ability to form these representations constitutes an

important aspect of understanding, referred to as implicit theoretical
understanding.
Regarding young children’s difficulty with problems, we conclude from Hud-
son’s results and similar findings that children probably have acquired pro-
cedures that conform to general principles, such as one-to-one correspondence.
However, procedural conformity is different from stronger forms of understand-
ing. The question whether children have acquired something like implicit the-
oretical understanding of a concept, such as one-to-one correspondence, is not
answered by showing that they have a procedure that is consistent with that
concept. In general, the question whether a child understands some general
principle is probably not well formed. Instead, we expect that a detailed analysis
of just what cognitive processes and structures a child has, usually will benecess-
ary for an adequate assessment. In other words, we analysis of what and how a
child understands is preferable to arguments about whether understanding has
been achieved.

ACKNOWLEDGMENTS

The research reported herein was supported by the Learning Research and
Development
Center, supported in part by funds from the National Institute of Education
(NIE), United
States Department of Health, Education, and Welfare. The opinions
expressed do not
necessarily reflect the position or policy of NIE, and no official endorseme
nt should be
inferred.
We are grateful to Tom Hudson for discussions and helpful criticisms
of this chapter.

REFERENCES
Brown, J. S., & Burton, R. R. (1978). Diagnostic models
for procedural bugs in basic mathe-
matical skills. Cognitive Science, 2, 155-192.
Brown, J. S., Collins, A., & Harris, G. (1978). Artificial
intelligence and learning strategies. In H.
F. O’Neil (Ed.), Learning strategies. New York: Academi
c Press.
Bruner, J. S., Goodnow, J. J., & Austin, G. A. (1956).
A study of thinking. New York: Wiley,
1956.
Davis, R. B., Jockusch, E., & McKnight, C. (1978).
Cognitive processes in learning algebra.
Journal of Children’s Mathematical Behavior, 2, 5-320.
Egan, D. E., & Greeno, J. G. (1973). Acquiring
cognitive structure by discovery and rule learning.
Journal of Educational Psychology, 64, 85-97.
Gagné, R. M. (1966). Human problem solving:
Internal and external events. In B. Kleinmuntz
(Ed.), Problem solving: Research, method, and
theory. New York: Wiley.
Gelman, R., & Gallistel, C. R. (1978). The child’
s understanding of number. Cambridge, MA:
Harvard University Press.
Greeno, J. G. (1977). Process of understanding
in problem solving. In N. J. Castellan, D. B.
 10. PROCESSES AND DEVELOPMENT OF UNDERSTANDING 313

Pisoni, & G. R. Potts (Eds.), Cognitive theory (Vol. 2). Hillsdale, NJ: Lawrence Erlbaum
Associates.
Greeno, J. G. (1978). A study of problem solving. In R. Glaser (Ed.), Advances in instructional
psychology (Vol. 1). Hillsdale, NJ: Lawrence Erlbaum Associates.
Greeno, J. G. (1980a). Analysis of understanding in problem solving. In R. H. Kluwe & H. Spada
(Eds.), Developmental models of thinking. New York: Academic Press.
Greeno, J. G. (July 1980b). Forms of understanding in mathematical problem solving. Paper pre-
sented at the meeting of the 22nd International Congress of Psychology, Leipzig.
Greeno, J. G. (1983). Forms of understanding in mathematical problem solving. In S. G. Paris,
C.-M. Olson, & H. W. Stevenson (Eds.), Learning and motivation in the classroom. Hillsdale,
NJ: Lawrence Erlbaum Associates.
Hudson, T. (1980). Young children’s difficulty with ‘‘How many more than . . . are there?’’ ques-
tions (Doctoral dissertation. Indiana University). Dissertation Abstracts International, 41, 01.
Katona, G. (1940). Organizing and memorizing. New York: Columbia University Press.
Markman, E. M. (1979). Classes and collections: Conceptual organizations and numerical abilities.
Cognitive Psychology, 11, 395-411.
Mayer, R. E., & Greeno, J. G. (1972). Structural differences between learning outcomes produced
by different instructional methods. Journal of Educational Psychology, 63, 165-173.
Mierkiewicz, D. B., & Siegler, R. S. (August 1980). Preschoolers’ abilities to recognize counting
errors. Paper presented at the meeting of the Fourth International Conference for the Psychology
of Mathematics Education, Berkeley, California.
Newell, A., & Simon, H. A. (1972). Human problem solving. Engelwood Cliffs, NJ: Prentice-
Hall.
Norman, D. A., & Rumelhart, D. E. (1975). Explorations in cognition. San Francisco: W. H.
Freeman.
Paige, J. M., & Simon, H. A. (1966). Cognitive processes in solving algebra word problems. In B.
Kleinmuntz (Ed.), Problem solving: Research, method, and theory. New York: Wiley.
Piaget, J. (1952). The child’s conception of number. London: Routledge & Kegan Paul.
Resnick, L. B. (1982). Syntax and semantics in learning to subtract. In G. Carpenter, J. Moser, &
T. Romberg (Eds.), Addition and subtraction: Developmental perspective. Hillsdale, NJ: Law-
rence Erlbaum Associates.
Riley, M. S., Greeno, J. G., & Heller, J. I. (1983). Development of word problem solving ability.
In H. P. Ginsburg (Ed.), Development of mathematical thinking. New York: Academic Press.
Sacerdoti, E. D. (1977). A structure for plans and behavior. New York: Elsevier-North Holland.
Schank, R. C., & Abelson, R. P. (1977). Scripts, plans, goals, and understanding:
An inquiry into
human knowledge structures. Hillsdale, NJ: Lawrence Erlbaum Associates.
Trabasso, T., Isen, A. M., Dolecki, P., McLanahan, A. G., Riley, C. A., & Tucker,
T. (1978). How do children solve class-inclusion problems? In R. S. Siegler (Ed.), Children’s
thinking: What develops? Hillsdale, NJ: Lawrence Erlbaum Associates.
Wertheimer, M. (1945). Productive thinking. New York: Harper & Row. (Enlarged edition, 1959)
 : = ‘
- «6 SE an HO Hn
—
. 7 7 7 =

en 1 ee ce ee
GI leer Ong m ww Wh
> ‘ ivé

rea
Ines De EEE en (hee

Pe
5 ees

eye
= od ii? s ' or Piel i.
— > 4 Bm Br ie Di ee u
Zr iw u 5 * Ka 2ER nV ieee “BR A »
. ap ay er El Zu ann
: 5 u A ca Ky pig ae erae

may a

et, te res 6 Aare ba M- cea eis AA in

Bes a er f vo 5 pane i
au i) Fun!

' > 71 an pw u 7 i. ur ET, Br ” i

. > im IN Zu zn 2 a. Zu
u > =@ 4 = } Ib > » A

‘ aff =) ab iver 6 Mi gaia? ie

» , ei ze
ate wr u n q 4 > A, «@

' .

5 Ay u ny Hi =

te cs a ne . DEREN le ' wi wes) a pol Ea, Un

u Te are 1 u m
a ae le N ghee? rn vom Shy oe Tee Te ww
al mi el Bartle ee eb
2 er . >
= 7 ps
ar-_
‘eo
re
Ara
a ZI
Aa
ué

we
nar

K N Sa Pe
gegen .. “on — i ‘(+ seh @ t en 1,¢é 2 ®

7 = 5 LE 42" Aen Ge “ D Wem + ’ 2

ar ee Awen Van ih f mie roy VB 0
E We ” Ku @
-
5
ga UG u > ‘ rT “id ‘ ound”
| ‘ f bu ae 7) ic Qe)er perso
:
vin, er 77 4877
Br net cio HG * ? “76 su
w aa wel @ Gl Os lott, ren Me sl Eb? Va dA
 Author Index

A Baldwin, M. J., 67, 109

Bandura, A., 6, 1/4, 217, 234
Abelson, R. P., 297, 3/5 Barann, G., 9, 11, 13, 16
Abramson le Ye. 141. 147-150. 1522 173: Barclay, C. R., 38, 47, 62, 124-25, 131,
178, 180, 181, 184, 185, 186, 188-90, 132, 134, 138
196, 206, 211, 214, 226, 233 Barnett, J., 205, 213
Ach: Ni; 223, 2247226, 233 Barstis, S., 45, 6/
INTZEn ISDN Bartleit IP Oa, BIEL is), JOP)
Albert#32.17952182 Beal, C2 R:, 24,29
Alderman, L. B., 42, 62 Beasley, C. M., Jr., 248, 264
Allenwv.re 102-109 Beck, A. T., 146, 181, 191, 196, 199, 201,
Alloy, Le Be, 147,178, 181, 211 AAW), ADS), HOE, Alls}
Ames, C., 156, 181 Beck RG alo ssecu,
Anderson, J. R., 245, 247, 248, 264 Beck, R. W., 203, 211
Anisman, H., 189, 2/2 Becker, 79
Anzai, Y., 99, 100, 109 Belmont, J. M., 39, 62, 88, 109, 122, 124,
Appelman, A. J., 147, 181 [28129 SI a7 sO SO ee 8 7204
Arbitman-Smith, R., 135, /38 Bereitenn €.1272138
Arkin, R. M., 147, 181 Bergen be 19872017
Atkinson, J. W., 217, 218, 226, 233, 234 Berglas, S., 147, 162, 168, /8/, 187, 211
Atkinson, R. C., 240, 264 Berman, J. J., 205, 212
Au, K., 103, 109 Bernstein, B., 103, /09
August, D. L., 24, 29, 88, 111 Binet, A., 89, 103, 109, 110, 118, 138
Austin, G. A., 292, 3/4 Birch, D., 218, 226, 234
Bisanzea @ 1235365507
Bjork, R. A., 79, 112
B
Blackman, L. S., 128, 138
Baillargeon, R., 89, 1/2 Blackman, S., 191, 195, 212
Baker, L., 66, 73, 86, 88, 109, 118, 137, Blackwell, J., 164, 183
260, 263, 264 Blaney, P. H., 152, 181

315
 316 AUTHOR INDEX

Blankenship, V., 226, 234 121121182 00122267 297 ISO 732%

Bobrow, D. G., 42, 63, 79, 80, 118, 138 134,135, 138,139
Boden, M. A., 79, 82, 110 Cannizzo, S. R., 241, 265
Boggiano, A., 40, 64 Carey, M. H., 76, 111
BOrKOWSK EEG IE aoe Garver E57 0814
137, 139 Cayanauch, Jn @ SAT viola 113:
Boswell, S. L., 244, 264 128, 139
Bovet, M., 69, 92, 112 GEIST RAAN 62794492065
Bowerman, M., 93-97, 110 Chase, W. G., 240, 255, 265
Bradley, G. W., 145, 1/8] Chi9M2 32 Hee 52 107742 407541762736:
Bransford, J. D., 47, 63, 73, 82, 87, 107, 111, 237, 240, 243-46, 251, 255, 259,
LOO SLO MI Spel OSeel os 265, 282, 288, 291
Bray, N., 40, 41, 63, 72, 112, 122, 138 Childs, C. P., 102, 108, 711
Brehm, J. W., 187-89, 215, 226, 235 Clark, E. V., 89, 95, 96, 111
Brewer, W. F., 80, 111 Clark, R. A., 218, 234
Brickenkamp, R., 228, 234 @obenS 21500187
Broadbent, D. E., 76-77, 110 Collins, A., 70, 80, 111, 295, 314, 1111
Brody, E. B., 268, 288 Corsale, K., 242, 266, 282, 283, 285, 288
Brody, N., 268, 288 Costanzo, P. R., 147, 184
Broudy, H. S., 68, 710 Crandall, V. C. 153, 154, 156, 168, 181
Brown, A. L., 6, 8, 9, 17-18, 32, 36, 38, 40, Crandall, V. J., 153, 154, 181
AT AD NOU O2 1055160,108 09. 71273. Crane N. 1.4262
71-82, 85-89, 95-97, 101-3, 107, 108, Crisafi, 78
10 ONE =? 6a 20236. Cronbach, L. J., 194, 195, 199, 201, 206,
137, 138, 139, 240, 260, 261, 263, 264 212
Brown, G. W., 190, 212 Crowe, D. H., 145, 183
Brown, J. S., 295, 297, 314
Bruner, ].S., 89, 97, 107, 110, 113, 292
314 D
Bryant, N. R., 101, 771
Bubrow, G. A., 114 Dahan, J. P., 183
Bugelski, B. R., 3, /4
Dallenbach, K. M., 4, 14
Bühler, K., 70, 110
Datkovsky, 153
Bumberry, W., 196, 2/2 Days sl zo
Burger, A. L., 128, 138
Davis, R. B., 296, 3/4
Burger, J. M., 147, 181 Day, J. D., 108, 170, 125,126. 132-34. 136,
Burke, D., 240, 243, 265
138, 139, 179, 182
Burton, N., 197, 213
Deaux, K., 146, 157, 181
Burton, R. R., 297, 314 DeLoache, J. S., 32, 49, 62, 68, 87, 89, 95—
Bush, E. S., 157, 172, 181, 188, 200, 212
IT 110117261264
Buss, A. R., 209, 2/2 Dempster, F. M., 244, 265
Butkowsky, I., 162, 184
Dennett) =Can oral
Butterfield, E. C., 39, 62, 88, 109, 122. 124,
Dewey, J., 67, 111
128, 129, 131, 137, 138, 139, 248 264
Dickson, W. P., 24, 29
Diener, C. I., 146, 153, 181, 220, 234
Dintzer, L., 210, 215
(©
Dixon, R. A., 11, 14
Dolecki, P., 299, 315
Campbell, D. T., 204, 212
Donaldson, M., 27, 29
Campione, J. C., 9, 17-18, 19, 32, 38,
62, Dörner, D., 33, 36, 44, 47, 62
69, 71-73, 87, 88, 101, 107-9, 110,
Douglas, D., 189, 2/2
 AUTHOR INDEX 317

Duncker, K., 55, 62 Friedrichsen, G., 34, 35, 63

Deck @ Se] 4015 30155 STATS aya, Ele AU, 2525)
187, 188, 200, 212, 220, 234 Frieze le il OSe 22
Fromkin, V. A., 96, 112

E
G
Ebbinghaus, H., 2, /4
Egan, D. E., 294, 314 Gagné, R. M., 76, 112, 294, 314
Bliea WW 1937272 Gallistel, C. R., 290, 292, 299, 314
Eller Sail 5 llelG2 764 GardnensE0971,88.22112120139
Elliot-Faust, D., 9, 15 Gardner, W. P., 85, 114
Emery, G., 201, 271 Geen„ Ra Ga, 188.212
Emswiller, T., 157, 18] Gelman, R., 89, 112, 240, 265, 290, 292,
Erbaugh, J., 146, 181, 196, 211 299, 314n
Ericsson, K. A., 69, 72-78, 111 Gerbasi, K. C., 149, 185
Ermshaus, W., 148, 154, 182 Gholson, B., 39, 64
Fraser Go WER SSH Gilliard, D., 156, /8/
Eysenck, H. J., 3, 14 Gilmor, T. M., 149, 181
Gitomer, D., 282, 283, 285, 288
Glaser, R., 4, 5, 10, 74, 78, 86, 114, 238,
240, 259, 265, 271-75, 288
Glasses Ce 97187213;
Fabricius, W. V., 11, /4 Gleitman aa leo Deel OOM Ete
Ealb0410.21622165,181,.1937212 Gleitman, L. R., 95, 112
Farris, E., 146, 157, 181 Goldin, S. E., 42, 62, 83, 86, 112
Feathers Nis, 156, 181,226, 234 Goldstein, K. M., 191, 195, 212
Feldman, N., 40, 64n Golin, S., 205, 212
Feltovich, 86 Goodnow, J. J., 292, 314
Fencil-Morse, E., 151, 162, 182 Green alee 2429 Sols
Herrara, R. Ar, 73, 87, 101, 107, 1097710; (Greene Pawar 82 12,
LES, 25 Ss4 S55 139 Greenfield, P. M., 102, 108, 111
Henettm RR. 1297138 Greeno, Je Gr, 0 l0, 498 790125 238, 245.
Feuerstein, R., 102, 111, 115, 135, 136, 139 265, 289, 292, 294, 295, 297, 300, 303-
Fitch, G., 146, 148, 155, 181 5, 307-12, 314, 315
Flanagan, J. C., 75, 1/11 Griffith, M., 151, /8/
Flavell, J. H., 9, 14, 17-18, 21, 23, 24, 29, Guilford, J. P., 4, /4
31792730318.022.03,.00,08: 09,12,
WS, Ws, ile 88 LUBE IPAS 113,118 20;
12712991311329139724125053, H
258, 259, 263, 265, 266
lores, Gy Er 82,112 Hacker, W., 6, 1/4
Foote, K., 42, 63 Hagen, J. W., 11, /4
Ford, L. H., 45, 6/ Halasz, F. G., 80, 81, 113
Horsys) G5 83,2113. Hale, G. A., 42, 62
Forrest-Pressley, D., 9, 15 Halisch, F., 148, 154, 182
Foxman, P. N., 76, /// Hammen, C. L., 152, 182
Frankel, A., 151, 162, 181, 184, 189, 212 Hanel, J., 155, 183
French, L. A., 101, 102, 108, 1/0 Hanusa, B. H., 151, 182
Friedrichs, A. G., 31, 37, 62, 66, 69, I11, Harre, R. M., 209, 212
118-412041270131139725525397209 Harris, G., 295, 3/4
 318 AUTHOR INDEX

Harris, T., 190, 2/2 Johnson, J. W., 40, 48, 63

Harvey, J. H., 76, 190, 212 Johnson-Laird, P. N., 74, 112
Hasher, L., 34, 42, 62, 80, 81, 112 Jonas, M., 196, 2/4
Hayes, J. R., 247, 265 Jones, E. E., 145, 147, 162, 168, 181, 184,
Hayes-Roth, B., 82-86, //2 187902118272
Hayes-Roth, F., 82-85, 112 Jopt, U. J., 148, 154, 182
Heckhausen, H., 6, 12, 14, 39, 62, 141-42, Jordan, C., 164, 183
144, 146, 148, 151, 154, 155, 160, 163, Judkis, A., 260
165, 167-70, 172, 176-78, 182, 183 Justice, E. M., 40, 41, 63, 72, 112
Heider, F., 191, 209, 212
Heller, J. I., 278, 288, 297, 300, 303-5,
307-12, 315 K
Hertzoener md
Hesse, F. W., 33, 43, 47, 62, 63 Kagan, J., 45, 64, 145, 179, 182
Hillgruber, A., 226, 234 Kahle, L. R., 205, 2/2
Hiroto, D. S., 149, 782, 187, 206, 207, 212, Kahneman, D., 209, 2/2
218, 234 Karmiloff-Smith, A., 45, 63, 69, 87, 89, 91-
Hogg, E., 149, 183 95, 98, 99, 105, 107, 112
Holmes, C., 149, 183 Karniol, R., 163, /82
Holmes, M., 128, /38 Karph, D. A., 76, 113
Holzmantwie G2711° 72992759238 Katkovsky, W., 154, 181
Horowitz, M., 202 Katona, G., 294, 315
Howard, D., 42, 62 Keeney, T. J., 241, 265
Howe, M., 44, 62, 244, 265 Kelley, H. H., 150, 160, 167, 169, 176, 182
ston, I ID)... Sl, Sik, 2, Ce, @®, WT, Wi 189, 191, 198, 273
ZO 2a Se 18.9 95352598265 Kelly, G. A., 189, 2/3
Hudson, 301-2, 313, 315 Kelly, M., 37, 40, 48, 63
HucyssEBr.67, 172 Kemler, D. G., 42, 63
Ill, En leg, De HE! Kendall? Gane 72. 3.12339
Humphrey, G., 70, //2 Kennedy, B. A., 128, 139
Hunt, E., 275, 288 Kenny, D. A., 204, 213
Huttenlocher, J., 240, 243, 265 Ketron, J. L., 74, 115
Kidd, R. F., 190, 272
Kleiman, G. M., 26, 29, 102, 108, 115
| Klein, D. C., 151, 162, 182
Kline, P. J., 248, 264
Ickes, W. J., 190, 212 Klingen se 07710113, 152. 1825210, 273
Inhelder, B., 69, 89, 91-93, 99, 107, 112 Kluwe, R. H., 7, 17-18, 31, 34-36, 39, 40,
non MN. Wenn Sh, Gh, Tan 44, 63, 87
Isen, A. M., 299, 315 Knight, R. A., 205, 2/3
Knopf, M., 9, 11, 13, 16
Kobasigawa, A., 45, 63
J Koeske, R. D., 246, 265
Koestler, A., 82, 113
Jackson, R. L., 186, 187, 213 Korkel, J., 9, 10, 14, 16
James, W., 80, 112 Koslowski, B., 89, 97, 107, 113
Janoff-Bulman, R., 190, 210, 2/2 Kotovsky, K., 270, 271, 273, 275, 288
Jockusch, E., 296, 3/4 Kovacs, M., 199, 2/3
Johnson, C. N., 253, 265 Krampen, G., 6, /4
Johnson, J. A., 36, 37, 64, 74, 114 Krantz, D. S., 187, 213
 AUTHOR INDEX 319

Krantz, S., 152, 182 McDermott, J., 83, 113

Kratz, H., 178, 179, 182 McDermott, R. D., 103, 113
Kreutzer Wil, ANS. TOT: Mace, W., 82, 115
250, 259, 265 McGeoch, J. A., 3, 4, 15
Kreuzig, H. W., 33, 47, 62 MelniyreK@aWE sins iscsi S298
Krug, B., 154, 183 139, 259, 266
Krugss lool ie 1820183 McKnight, C., 296, 3/4
Kruglanski, A. W., 191, 2/3 McLanahan, A. G., 299, 315
Kubal ai 4 217235 McMahan, I. D., 168, 183
QUO, Is 7G WS By sr, el, Se GEE McNemar, Q., 4, /5
2192207223 9224 22063234 Maier, S. F., 186, 187, 213, 214
Kukla, A., 154, 184, 217, 226, 234, 235 Majetic, D., 282, 288
Kun, A., 163, 170, 183 Mandler, G., 70, 113
KRurtzı BIER N Mandler, J. M., 70, 113
Markman, E. M., 42, 63, 67, 88, 113, 118,
1397313, 315
L Marshall, J. C., 65, 66, 88, 113
Martin, O., 206, 211
LaBerge, D., 80, 113 Masur, ES 370056332115 118912981391
Lachman, J., 41, 63 259, 266
Lachman, R., 41, 63 Mayer, R. E., 294, 315
Lange, G., 242, 266 Meacham, J. A., 6, /5
Langer, E. J., 191, 194, 273 Mech PEO 273
Lawton, S. C., 88, 110, 124, 138 Meichenbaum, D., 129, 139
Lefcourt, H. M., 149, 183 Mendelson, M., 146, /8/, 196, 211
Lefkowitz, M. M., 197, 213 Metalsky, G. L., 190, 2/4
Beonand 54 Cola ai Os Sau teal se 2505 Meyer, W. U., 147, 154, 168, 183, 226, 234
259, 265 Mierkiewicz, D. B., 292, 315
Lepper, M. R., 146, 183 Miller, D. J., 128, 139
BeyyaaYy MM 221720253, 2535260 Miller, D. T., 145, 183
Lewin, K., 223, 234 NWN Ses IB, So, 712, 1018)
Lichtenstein, S., 208, 2/4 Mill ete lea Veen IS Oe Sonmllscree 00m 15)
Lindauer, B. K., 8, 75) Miller, S. M., 185, 189, 2/3
Lindbergh, M. A., 245, 266 Miller, W. R., 152, 183, 218, 234
Lindsay, P. H., 82-83, 1/3 Milner, B., 86, 113
locker Ir. 10; L153: Minton, H. L., 149, 181
Loebl, J., 40, 64 Mischel, T., 209, 2/3
Logan, G. D., 80, 113 Mischel, W., 6, 15, 191, 219, 234
Lompscher, J., 37, 63 Mock, J., 146, 181, 196, 211
Luckmann, T., 198, 211 Moely, B. E., 241, 266
lie Gp. SB 4's). ee: Monson, T. C., 190, 213
Luginbuhl, J. E. R., 145, 183 Moore, J., 120, 139
ana Ae Rees Onell Morgan, J. L., 80, 111
Morris, D., 47, 63
Morton, J., 65, 66, 88, 113
M Mosse Jia 205,213
Mulholland, T. M., 273, 274, 288
MacCorquodale, K., 192, 2/3 Murphy, M. D., 124, 126, 134, 138
McClelland, D. C., 218, 234 Myers, M., 67, 73, 113
McClure, J. N., 196, 2/2 Myers, N. A., 244, 266
 320 AUTHOR INDEX

N Phillips, W., 179, 182

Piaget, J., 6, 15, 46, 63, 68, 69, 71, 74, 76,
Naus, M. J., 80, 81, //3 89, 90-91, 95-96, 99, 100, 114, 120,
Neisser Why SA, SBCs, WE ZS: 139, 240, 298, 301, 315
Nelson, K., 94, 113 Pittman, L., 244, 266
Nelson, K. E., 94, 113 Plenderleith, M., 3, /5
Nelson, K. J., 242, 266 Plöger, F. O., 168, 183
Netschajeff, H., 3, /5 Posner, M. I., 80, 114
Newcombe, N., 45, 64 Postman. L.,3, 19
Newell, A., 82, 83, J//, 114, 120, 125, 139, Potepan, P. A., 154, 184
247, 252, 266, 295, 296, 315 Powell 17 Se 74115
Nicholls, J. G., 40, 63, 156, 164, 169, 171- Pressley, M., 9, 15
12:83 Pribram, K. H., 233, 234
Nida, S., 244, 266 Putz-Osterloh, W., 33, 43, 63
Nisbett, R. E., 72, 74, 114, 190, 191, 209, Pylyshyn, Z., 69, 71, 114, 118, 120, 139
DIAS:
Nooteboom, S. C., 96, 97, 114
Norman, DIA, 42) 63..10%179233#85, R
SO LI 45a 18h 135,245,00486 260,
297313 Radloff, L. S., 200, 2/4
Norman, W. H., 150, 183, 188, 206, 213 Raps, C. S., 196, 2/4
Rasch, G., 219, 234
Raskin, A., 197, 2/4
O Raven, J. C., 272, 288
Reason, J. T., 81, 114
Oliver, J. M., 196, 212 Reeda 1229170233
Olson, D. R., 27, 29 Reed, S. K., 36, 64, 74, 114
Ornstein, P. A., 242, 266 Rees, E., 240, 259, 265
Osgood, D. W., 191, 194, 2/4 Reese SH Wi Ono. Wann
Overmier, J. B., 186, 2/3 Rehnaa Pes load
Owings, R., 47, 63 Reinhard, K. E., 196, 2/4
Reither, F., 33, 47, 64
Reitman, W., 32, 64
Reppucci, N. D., 153, 156, 181, 187, 212
P
Resnick, L. B., 78, 114, 293, 315
ResıeSs 72179985
Paige, J. M., 294, 315 Rheinberg, F., 172, 183
Palinscar, A. S., 103, 108, 114 Rholes, W. S., 164, 183
Baus S, (G55, 8, 1S, Ol, Tas MS)
Richman, C. L., 244, 266
Parsons, J. E., 40, 41, 63, 64, 168, 183
Richman, S., 39, 64
Pawlow, J. P., 3, 15 Riegel, K. F., 89, 114
Paykel, E. S., 190, 2/4
Rifkin, B., 44, 64
Bette L&S 1040274 Riley, C. A., 299, 315
Pellegrino, J. W., 4, 5, 10, 78, 114, 238, Riley, M. S., 5, 10, 49, 238, 297, 300, 303-
271-75, 282, 288
5, 20712531385 375
Perlmutter, M., 8, 14, 37, 41, 63, 72. 13%
Ringel, B. A., 38, 64
111, 114, 244, 266 Rippere, V., 210, 274
Petersen, G., 47, 63
Rizley,R21522785
Pefersonm Cerone aon ISO5 152 10317 Robinson, F. P., 3, 75
188-91, 193, 194, 196, 198, 210, 2/4
Phillips, S., 150, 787
 AUTHOR INDEX 321

Rogoff, B., 45, 64, 85, 114 Shatz, M., 87, 115
Rosenbaum, R. M., 217, 235 Shaw, B. F., 201, 217
Rosenfield, D., 145, 184 Shawahano2 5)
Rosman, B. L., 179, 182 Shaw, R. E., 82, 115
Ross, J., 40, 64 Shepp, B., 42, 63
Ross, L., 146, 783, 193, 209, 213, 214 Shi szene 2
Ross, L. E., 42, 62 Shiffrin, R. M., 64, 77, 80, 115, 240, 264
Ross, M., 145, 163, 182, 183 Shipley, E. F., 95, 7/12
Roth, S., 150, 184, 185, 206, 214, 217, 235 Shulman, A., 187, 2/4
Rothbart, M., 150, 181 Shultz, I. R., 162, 184
Rotter, J. B., 146, 148-49, 184, 191, 214 Siegler, R. S., 43, 64, 292, 315
Rozin, P., 69, 71, 1/4, 120, 140 Simon, D. P., 33, 64, 86, 115
Rubinstein, S. L., 6, 75 Simon, H. A., 33, 39, 64, 69, 72-78, 83, 86,
Ruble, D. N., 40, 41, 63, 64, 168, 183 SO DD ILE, JHE 5, WA. Ds.
Ruger, H. A., 77, 114 253, 20992007 2109211271372759288:
Rumelhart, D. E., 80, 114, 245, 248, 266, 294-96, 315
2.13, 288, 297, 315 Simon, J. G., 156, 181
Rushey As Js, 201. 217 Sinclair, H., 69, 92, 112
Ryle, G., 192, 194, 214 Singer, J. B., 24, 29
Skinner, B. F., 81, 82, 115
Sladewski-Awig, L., 41, 64
S Slovic, P., 209, 214
Smiley, S. S., 88, 1710
Sacerdoti, E. D., 83, 115, 296, 297, 315 Si WER CHE Woy JU
Sackeim, H. A., 152, 180, 190, 211 Srartin, 1 Ro, 72.118)
Scardamalia, M., 127, 138 Smith, R., 188, 2/5
Schadler, M., 78, 114 Smoller, B., 162, 184
Schallert, D. L., 26, 29, 102, 108, 115 Snyder, C. R. R., 80, 1/4
Schallice, T., 77, 81-83, 85, 86, 114, 115 Snyder, M., 190, 213
Schanke Re Ge. 2975315 Snyder, M. L., 145, 151, 162, 181, 184, 187,
Scheier, M. F., 6, /4 ICEIWA EA bs}
Schlenker, B. R., 147, 184 Solomon, R. L., 186, 213, 214
Schmalt, H. D., 146, 154, /84 Speannan, Crs) (O15 2085 2727288
Schneider, K., 148, 184 Speer ER 245 29788, 717,
Schneider, W., 8-11, 15, 64, 77, 80, 115 Spinoza, B., |
Scholnick, E., 37, 40, 48, 63 Springer, C. J., 38, 64
Schuldenbrandt, J. G., 197, 214 Stanley, J. C., 204, 212
Schulz. R151.2182 Stasz, C., 33, 43, 47, 64
Schwartz, S. M., 190, 210, 2/4 Stäudel, T., 33, 47, 62
Scott, W. A., 191, 194, 2/4 Stavey, R., 93, 115
Secord, P. F., 209, 212 Stein, B., 47, 63
Selfridge, O., 82-84, 115 MANE 183, Sion Hel, (sre!
Seligman, M. E. P., 6, 11, 141, 142, 150-53, Sans), ING IL Lily hey LS)
162, 173, 174, 1807 182,183, 184, 185— Stephan, W. G., 145, 184
SON 196, 2012067 2072 21072119212, Sternberg, R., 32, 33, 36, 39, 44, 48, 64
DIS DUES, Pikes, 2:8 7219929267 253,7252, Sternberg, R. J., 74, 115, 118, 140, 272, 288
235 Stevens, L., 145, /84
Selz, O., 36, 64 StUpek Ds Jay ule Lo)
Semmel, A., 152, 173, 184, 186, 190, 196, Storm, C., 244, 266
214 Strauss, S., 93, 115
Shaeffer, D. E., 205, 212 Strenta, A., 162, 184
 322 AUTHOR INDEX

DMTCULer too) WwW

SIeUrert, SG 95s 15)
Struthers, S., 149, 183 Walters, C., 164, 183
Sugarman, S., 96, 97, 111 Wambold, C., 124, /39
Surber, €. F., 163, 170, 184 Ward, C. H., 146, 181, 196, 211
N75) 1c) i aD 022 Weil, E., 39, 48, 64
Weiner, B., 154, 184, 190, 193, 215, 217,
235
T
Weinert, E. E2.4,59°11.,13216, 70
Weinraub, M., 187, 2/4
har, ID) 18), 96), JM Weisss) 21872214
dicasdale- 5 Dy 15051730180) 185.083. \Weisz,. 12 Rule 15
189, 206, 211, 226, 233 Wellmans EI Mis 8. 1182164318 02° 72723
Tennen, H., 151, 162, 184 75478,.106; 112.116, 119, 139.252.
Tenney, Y., 124, 140, 241, 266 2332 259) 265
Thorndike, E. L., 67, 115 Wener, A. E., 152, 184
Thorndyke, P. W., 33, 43, 47, 64, 83, 112 Werner, H., 107, //6
Thornquist, K., 32, 64 Wertheimer, M., 294, 315
Thronesbery, C., 41, 63 Wertsch, J., 26, 29, 46, 64, 101, 103, 108,
Thurstone, L. L., 4, 15, 268, 288 116
Thurstone, T. C., 268, 288 White HP Pe Gs
Tornquist, 73 Wicker, A. W., 209, 215
Trabasso, T., 67, 88, 115, 299, 315 Wilcox, B. L., 122, 138
Travers, S., 37, 40, 48, 63 Wilder, 76
Trudewind, C., 6, 15 Wilson, tabs wi2s 7421147191223
Tucker, T., 299, 315 Wimmer, H., 32, 64, 73
Turvey, M. T., 79, 82, 115 Wine. Jie 2205230)
Tversky, A., 209, 212 Winograd, T., 82, 112, 247, 266
Tweer, R., 163, 170, 184 WittsT. Re; 147, 184
Woodrow, H., 2, 16
Worden, P., 41, 64
U Wortman, C. B., 147, 184, 187-89, 210,
213, 220.235
Ueckert, H., 33, 64 Wundt, W., 2, /6

Vaughn, E. D., 1, 15
Vesonder, G De. 38) 67
Volpert, W., 6, 14
Von Baeyer, C., 152, 173, 184, 186, 196,
214
Von Cranach, M., 6, 14 Zacks, R. T., 34, 42, 62, 80, 81, 112
Vosse). Bes5s267
Zeigarnik, B., 3, 16
WS, ING Ik, 185, JGR
Zetlin, A., 128, 138
Vygotsky, L. S., 69, 100, 103, 107, 116
Zimmer, J., 40, 64
 Subject Index

A Attributional style, 81-83, 190-211

Attributional Style Questionnaire (ASQ),
Ability 152, 193-95
compensatory causal schemata for, 169-73 measurement of, 191-93
self-concept of, 179-80 depression and, 150-53, 185-86, 189-90,
specific vs. global, 173-75 196-208
success and failure and, 160-61 See also Causal attribution patterns for
Accessibility theory, 71-72, 120 achievement outcomes
Achievement motives, interindividual contrast Autocriticism, 89
in attributional patterns and, 154—55 Automatic processing, 80-81
Action-control, 219-33 Autonomous regulation, 89, 90
Action schemata, 297-303, 308, 310
Active regulation, 89, 90, 96
Affects, 224 B
Analogical reasoning problems, 272-73
Analogical use of structure, 293 Beck Depression Inventory (BDI), 196
Analogies, 273—85
Analogical use of structure, 293
Analogies, 273-85 c
Aptitudes for learning, 267—88
analogical reasoning problems, 272-73 Causal attribution patterns for achievement
figural analogies, 273-77 outcomes, 143—84
numerical analogies, 282—85 age and, 174-76
relevance of rule induction to, 268—85 interindividual contrast in, 148-59
series completion process, 270-72 intraindividual asymmetry of, 145—48
verbal analogies, 278-82 normal attribution pattern, 179
Aspiration level, formation of, 167 possible types and emergence of, 159-75
Asymmetric attribution patterns, 145—48 Change problems, understanding, 304—12
Attention, selective, 223 Checking, 37-39, 260-61, 262

323
 324 SUBJECT INDEX

Child rearing patterns, 165—67 Effort, compensatory casual schemata for,

Children’s Depression Inventory (CDI), 200 169-73
Classification, 36-37 Egotism, 162
Cognitive dissonance, 167-69 Error correction, developmental data on, 96—
Combine problems, understanding, 304—12 97
Compare problems, understanding, 304-12 Evaluation, 39-40
Competence, learning effectiveness and 4-7 of performance, 291—93
Computer planning models, 83-85 Executive decisions, 31-64, 84, 124-25
Concurrent verbalizations, 74, 75 executive control, 36-41, 46-48, 79-88
Condition-action-connections, 35 executive regulation, 35, 41-46, 48-61
Conformity, 291 framework for, 34-46
Conscious access, 71-72 role of, 32-34
Conscious regulation, 89, 99-100 Expectancy-value theory of motivation, 231—
Control 33
executive, 36-41, 46-48, 79-88 Experiences, metacognitive, 24-28
reinforcement, internal vs. external, 148-50 Explicit knowledge, 291
of rules, 133
Controllability, attributions of, 210
Controlled processing, 80-81 F
Correction, error, 96-97
Creative simplification, 93-94 Failure
motivational effect of, 218-19
possible attribution patterns for, 159-75
D
See also Causal attribution patterns
Failure-induced performance deficits,
Debugging state, 80
metacognitive mediation of, 217—35
Decisions, categories of, 84
Figural analogies, 273-77
See also Executive decisions
Declarative knowledge, 5, 31, 245-46, 270
Depression, attributional style and, 190, 196—
208 G
helplessness and, 185-86
interindividual contrast in attributional pat- Global ability, 173-75
terns, 150-53 GPS (General Problem Solver), 83, 85
reformulation of helplessness theory and,
189-90
Depressive Attribution Pattern (DAP), 159-75, H
176, 178-80
Developmental data Helplessness
on children’s verbal reports, 77-79 attributional style in depression and, 185—86
on executive control, 85-88 theory, 150-52
on other-regulation, 103-5 See also Learned helplessness
on self-regulation, 96-100 Heterarchies, attribution and, 82-83
Domain knowledge, 245—47

Implicit understanding, 291, 313

Efficiency, learning, 2-3 Individual differences, 3, 4-7, 165-69, 172
Efficiency of the will, 223 Tig Ma
 SUBJECT INDEX 325

Inductive reasoning test tasks, 268-85

Information-processing models, 79-80
Information processing, speed of, 45—46
Learned helplessness, 186-90, 218-20
Insidious attributional style, 197-98, 202-3,
action-control and, 225-33
208
original theory of, 187-89
Instructional mapping of procedures, 293-94
reformulation of, 189-90, 205-8, 210-11
Instructional research, status of, 121-24
in schoolchildren, 153-54
See also Problem learners, instructional re-
three-factor model of, 227
search with
Learning
Instrumental Enrichment program, 135
activity, metacognition, and motivation as
Integrated performance, 296-97
determinants of individual, 7-13
Intensity, regulation of processing, 44—45
effectiveness, competence and performance
Interactive learning processes, 101-5, 108
as determinants of, 4-7
Interindividual contrast in attribution patterns,
efficiency, 2-3
148-59
general laws of, 2-3
Interindividual variables, 22
processes, interactive, 101-5
Internalization, Vigotsky’s theory of, 100
See also Aptitudes for learning
Intraindividual asymmetry of attribution pat-
terns, 145—48
intraindividual variables, 22
Intrinsic understanding, 295-97 M
IQ, attribution patterns and, 162-63
Irreversibility vs. reversibility of actions, 49— “Man as scientist’? metaphor, 189, 209-10
61 Memory
development, knowledge approach to, 240—
45
K load, 273-77
monitoring, 37-39, 258-59
Knowledge performance, metamemory and, 9-10
approach to memory development, 240—45 Memory estimation paradigm, 254-57
base, differences in, 5 Mental processing skills, 286
base influences, 282—85 Metacognition
about cognition, 67—68 defined, 8, 21, 65-69, 119
declarative, 5, 31, 245-46, 270 as determinant of individual learning ac-
domain, 245—47 tivity, 7-13
evidence for, 244—45 forms of, 67-68
explicit, 291 motivation and, 11-13
framework for representing, 245-52 questions, problems, and issues concerning,
instructional attempts with problem learners 25-28
using, 126-27 roots of, 69-105
metacognitive, 21, 22—23 executive control, 36-41, 46-48, 79-88
meta-declarative, 249-50, 253-54 other-regulation, 100-105
metaprocedural, 251 self-regulation, 88-102, 119, 129-33,
about one’s own cognitive resources, 119, 135-36
120 speculations about nature and develop-
for planning, 295-96 ment of, 21-29
procedural, 5, 246-47, 270, 286 status as concept, 105-9
stimulus, 259 verbal reports as data, 69-79
strategy, 286 Metacognitive experiences, 24-28
of task variables, 259 Metacognitive knowledge, 21, 22-23
 326 SUBJECT INDEX

Metacognitive mediation of failure-induced integrated, 296-97

performance deficits, 217-35 relation of verbal reports to actual, 74-76
Meta-declarative knowledge, 249-50, 253-54 ‘*Persistence’’ of regulatory activity, 59
Metaknowledge, 249-51, 253-58, 261-62 Person variables, 22, 253-58, 261-62
Metamemory, 8-10, 66 Plan and plan-abstraction decisions, 84
research, 352-62 Planning
Metaplan decisions, 84 deficits, 86-88
Metaprocedural knowledge, 251 knowledge for, 295-96
Metaprocedural reorganization, 91-95, 97-99 models, computer, 83-85
Meta-strategies, 250-51 Positive Attribution Pattern (PAP), 159-75,
Monitoring, 37-39, 258-60, 262 178-80
Mother-child interaction, attributional style Prediction, 40-41
and, 200-201, 209 Predictive verbalizations, 74
Motivation Problem learners, instructional research with,
as determinant of individual learning ac- 117-40
tivity, 7-13 age and, 134
expectancy-value theory of, 231-33 metacognition and instructional attempts,
metacognition and, 11-13 124-33, 135-36
Motivational effect of failture, 218-19 status of instructional research, 121—24
Multiple access, 71-72 Problem schemata, 297—98
Multiple necessary causes for success, 160-61 Problem solving
Multiple sufficient causes for failure, 160-61 children’s regulation of, 49-61
computer planning models of, 83-85
development of skill in, 297-312
N of problems of application, 294-95
processes, 33-34
NOAH (Nets of Action Hierarchies), 83-84, skills, generality/specificity of, 125
85 See also Executive decisions
Normal attribution pattern (NAP), 179 Procedural constraints, 278—82
Numerical analogies, 282-85 Procedural knowledge, 5, 246-47, 270, 286
Processing
automatic and controlled, 80-81
O capacity, regulation of, 42-43
intensity, regulation of, 44-45
Opportunistic planning model (OPM), 84—85 Process models, 27-28
Other-regulation, 100-105

R
P

Reading skills, metacognitive, 66—67

Parents
Reasoning, inductive, 268—85
attributional style and, 198 Reflection, 70-71
child-rearing patterns fostering PAP and Reflective access, 71-72, 120
DAP, 165-67 Regulation
metacognitive development and, 26 of cognition, 68
Pattern generation, 270-71
executive, 35, 41-46, 48-61
Performance
other-, 100-105
deficits, metacognitive mediation of failure- self-, 88-102, 119, 129-33, 135-36
induced, 217-35 Rehearsal, 241
executive control and, 46-48 Reinforcement control, internal vs. external,
individual learning effectiveness and, 4—7
148-50
 SUBJECT INDEX 327

Reorganization, metaprocedural, 91—95 Success, types of attribution patterns for, 159—

Resources, awareness of, 119, 120 75
Retrospective verbalization, 74 See also Causal attribution patterns
Reversibility vs. irreversibility of actions, 49—
61
Rule induction, relevance of, 268-85 T
Rules, 132—33

Task difficulty, 160-61

S Task variables, 22-23, 259
Teachers, self-regulation and, 102
Theoretical understanding, 290-95, 313
School, metacognitive development and, 26-
Transfer, ability to, 294
27
Triadic design, 187
Selective attention, 223
Self-concept, 155-56, 179-80
Self-criticism, attribution patterns involving,
146-47 U
Self-management, instruction in, 132, 133
Self-reflection, 70-71 Understanding, 289-315
Self-regulation, 88-102, 119, 129-33, 135-36 development of ability, 297-312
Self-testing, 255 implicit, 291
Sensitivity, 258, 262 processes of, 289-97
Series completion problems, 270-72 intrinsic understanding, 295-97
Sex differences, interindividual contrast in at- theoretical understanding, 290-95, 313
tributional patterns and, 156-57 Universal variables, 22
Speed of information processing, 45-46
State-orientation, 221-23
Stimulus knowledge, 259 V
Strategy(ies), 247-48
instructional attempts with problem learners Verbal analogies, 278-82
using, 127-33 Verbal reports as data, 69-79
meta-strategies, 2350-51
properties of, 247
evidence for, 240-44 W
knowledge, 286
variables, metacognitive knowledge of, 23 Will, efficiency of the, 223
Structure, analogical use of, 293 World-knowledge decisions, 84
 = a
7 es
| +a 12 u

7 —
u re = „NS ow oe al “> is

=) 5
,
a
Da I @ 0 ; > am ay mi a
Pe hu Bug
RZ t——. eel 4 ae Pe 1 - = 2

Hott. |‘ orem,
. 7
“edi. HOt hate of

aoe) on. 36 ing Be a Ir ze

pe Fr
= :

; Jar. Ai mae ei Bek

| 4 (ie art
u ui ae =) 26026 MV £8) @ Aue
weu
- u) :
I oa 2 - on om Ait ate
bb —

. Mil ne «fad
a » ie te e
z i le I tame oe
rail! in San,
>

é
- ah
itl. 4 je f
a= Fe We

u
“a *
3 5282 00113 1567
LB1065
M3813
1987
 STACKS LB1065.M3813 1987

al a]

|
HH} |
JERE

5282 00113 1567

ISBN 0-89859-569-X