Handbook of Understanding and Measuring Intelligence 0761928871, 9780761928874

Table of contents :
Preface -- 1. Intelligence : a diva and a work horse / Oliver Wilhelm, and Randall W. Engle -- 2. Assessing Problem Solving in Context / Patrick C. Kyllonen and Soonmook Lee -- 3. Mental speed : on frameworks, paradigms, and a platform for the future / Vanessa Danthiir, Richard D. Roberts, Ralf Schulze, and Oliver Wilhelm -- 4. Cognitive mechanisms underlying intelligence : defense of a reductionist approach / Andrew R.A. Conway -- 5. Working memory capacity, attention control, and fluid intelligence / Richard P. Heitz, Nash Unsworth, & Randall W. Engle -- 6. Emotional intelligence : an elusive ability / Gerald Matthews, Moshe Zeidner, and Richard D. Roberts -- 7. Metacognition and intelligence / Christopher Hertzog, and A. Emanuel Robinson -- 8. Knowledge and intelligence / Phillip L. Ackerman, and Margaret E. Beier -- 9. Full frontal fluidity? Looking in on the neuroimaging of reasoning and intelligence / Michael J. Kane -- 10. Behavioral genetics and intelligence / Stephen A. Petrill -- 11. A dialectical constructivist view of developmental intelligence / Juan Pascual-Leone, and Janice Johnson -- 12. Development of intellectual abilities in old age : from age gradients to individuals / Martin Lövdén and Ulman Lindenberger -- 13. Group differences in intelligence and related measures / Werner W. Wittmann -- 14. Modeling structures of intelligence / Ralf Schulze -- 15. Item response theory and the measurement of cognitive processes / Florian Schmiedek -- 16. g factor : issues of design and interpretation / Lazar Stankov -- 17. Capturing successful intelligence through measures of analytic, creative, and practical skills / P.J. Henry, Robert J. Sternberg, and Elena L. Grigorenko -- 18. Faceted models of intelligence / Heinz-Martin Süss and André Beauducel -- 19. Assessing intelligence : past, present, and future / Richard D. Roberts, Pippa M. Markham, Gerald Matthews, and Moshe Zeidner -- 20. The role of domain knowledge in higher-level cognition / David Z. Hambrick -- 21. Measuring reasoning ability / Oliver Wilhelm -- 22. The measurement of working memory capacity / Klaus Oberauer -- 23. Working memory, intelligence, and learning disabilities / H. Lee Swanson -- 24. Cognitive ability in selection decisions / Denise S. Ones, Chockalingam Viswesvaran, and Stephan Dilchert -- 25. Understanding intelligence : a summary and an adjustable-attention hypothesis / Nelson Cowan -- 26. To g or not to g : that is the question / Nathan Brody -- About the editors -- About the contributors -- Author index -- Subject index

Citation preview

Handbook of m

f-'*.

i f^

itedby

RANDALL W. ENGLE

Boston Public Library Boston iviA U2>16 »

i

Handbook of

Understanding and Measuring Intelligence

Handbook of

Understand ng and Measuring i

Edited by

OLIVER WILHELM • RANDALL W. ENGLE Humboldt-University, Berlin,

Germany

Georgia Institute cf Technology

®SAGE Publications New Thousand Oaks

London

Delhi

© 2005 by Sage Publications,

Copyright

All rights reserved.

No

part of this

Inc.

book may be reproduced or

utilized in

any form or by any means, electronic

or mechanical, including photocopying, recording, or by any information storage and retrieval system, without

permission

in writing

from the publisher.

For information:

Sage Publications,

Inc.

2455 Teller Road

Thousand Oaks, California 91320 E-mail: [email protected] Sage Publications Ltd. 1

Oliver's Yard

55 City Road

London EC lY ISP United Kingdom

Sage Publications India

Pvt. Ltd.

B-42, Panchsheel Enclave Post

Box 4109

New

Delhi 110 017

India

Printed in the United States of America

Library of Congress Cataloging-in-Publication Data

Handbook of understanding and measuring

intelligence

/

edited by Oliver

Wilhelm, Randall W. Engle. p.

cm.

Includes bibliographical references and index.

ISBN 0-7619-2887-1 1.

Intellect. 2.

(pbk.) Intelligence tests.

I.

Wilhelm, Oliver.

II.

Engle,

Randall W.

BF431.H3 188 2004 153.9—dc22 2004013896

04

05

06

07

Acquiring Editor:

10

9

7

6

5

3

Jim Brace-Thompson

Editorial Assistant:

Karen Ehrmann

Production Editor:

Diana E. Axelsen

Copy

Gillian Dickens

Editor:

4

Typesetter:

C&M Digitals (P) Ltd.

Indexer:

David Luljak

Cover Designer:

Michelle Lee Kenny

2

1

Contents

Preface

1.

vii

Intelligence:

A

Diva and a Workhorse and Randall W. Engle

1

Oliver Wilhelm 2.

Assessing Problem Solving Patrick C. Kyllonen

3.

Mental Speed:

in Context and Soonmook Lee

On Frameworks,

1

Paradigms,

27

and a Platform for the Future Vanessa Danthiir, Richard D. Roberts,

Ralf Schulze, and Oliver Wilhelm 4.

Cognitive Mechanisms Underlying Intelligence:

47

Defense of a Reductionist Approach Andrew R. A. Conway 5.

Working Memory Capacity, Attention Control, and Fluid Intelligence Richard P. Heitz, Nash Unsworth, and Randall

61 W. Engle

6.

Emotional Intelligence: An Elusive Ability? Gerald Matthews, Moshe Zeidner, and Richard D. Roberts

1

Metacognition and Intelligence Christopher Hertzog and A. Emanuel Robinson

101

Knowledge and

125

.

8.

Phillip L. 9.

Intelligence

Ackerman and Margaret

Full Frontal Fluidity?

Looking

in

E. Beier

on the Neuroimaging

of Reasoning and Intelligence

Michael 10.

J.

11.

A

141

Kane

Behavioral Genetics and Intelligence Stephen A.

79

165

Petrill

Dialectical Constructivist

View

of Developmental Intelligence

Juan Pascual-Leone and Janice Johnson

177

12.

Development of

Intellectual Abilities in

From Age Gradients

Old Age: 203

to Individuals

Martin Lovden and Ulman Lindenberger 13.

Group Differences

in Intelligence

and Related Measures

223

Werner W. Wittmann 14.

Modeling Structures of Intelligence

241

RalfSchulze 15.

Item Response Theory and the Measurement of Cognitive Processes

265

Florian Schmiedek 16.

g Factor: Issues of Design and Interpretation

279

Lazar Stankov 17.

Capturing Successful Intelligence Through Measures of Analytic. Creative, and Practical Skills P. J.

18.

Henry; Robert

J.

Sternberg,

and Elena

295 L.

Grigorenko

Faceted Models of Intelligence

313

Heinz-Martin Sufi and Andre Beauducel 19.

Assessing Intelligence: Past. Present, and Future Richard D. Roberts, Pippa M. Markham,

333

Gerald Matthews, and Moshe Zeidner 20.

The Role of Domain Knowledge David

21.

Z.

in

Higher-Level Cognition

361

Hambrick

373

Measuring Reasoning Ability Oliver Wilhelm

22.

The Measurement of Working Memory Capacity

393

Klaus Oberauer 23.

Working Memory, H. Lee

24.

and Learning Disabilities

Cognitive Ability in Selection Decisions Deniz

25.

Intelligence,

409

Swanson

S.

431

Ones, Chockalingam Viswesvaran, and Stephan Dilchert

Understanding Intelligence:

A Summary

and an Adjustable-Attention Hypothesis

469

Nelson Cowan 26.

To g

or

Not

to

g

—That

Is the

Question

489

Nathan Brody

About the Editors

503

About the Contributors

504

Author Index

509

Subject Index

529

Preface

The

test

in the

of a first-rate intelligence

mind

same

at the

time,

and

is

the ability to hold

still

two opposed ideas

retain the ability to fimction. Scott Fitzgerald

F.

Although there has been great progress in intelHgence research over the past years,

and although there

and what this

book,

it

we want

might

is

some consensus on how

reflect, there are

many

and measuring intelligence. Besides an overview of the

we

perspective on prospective

work

approach serves two goals.

On one

in

both understanding

status

quo of knowledge

also want to provide a comprehensive

remains to be accomplished. This

in the field that

hand, intelligence assessment

making

tance for students interested in

100

measure intelligence

important questions unanswered. In

been achieved

to highlight progress that has

about individual differences in intelligence,

to

is

of major impor-

a career in psychology, education, or other

behavioral sciences. Being familiar with fundamental issues of understanding and

measuring intelligence

is critical

construct of intelligence

is

and outside of psychology.

emerging problems

Some of outline how

for

many

practical activities.

A

the chapters focus

be helpful for

many

applied and scientific decisions.

on the theoretical and construct

the understanding of intelligence can be

logical

in

where we

felt

it

Proposals for cal

intelligence,

was

These chapters

and how dealing with these aspects helps us to measure statistical

intelli-

and pragmatic

measuring intelligence and outline methodological and techno-

improvements and how these contribute

ligence. In this preface,

level.

improved by incorporating certain

gence more thoroughly. Other chapters focus on fundamental assessment problems

the other hand, the

basic understanding of the established consensus and

in the field will

theoretical perspectives

On

of relevance for several other scientific disciplines, within

we

briefly

go though

to

all

an improved understanding of

the chapters, grouping

intel-

them together

possible, to give an idea about the scope and content of this book.

new

constructs such as critical thinking, situational judgment, practi-

and emotional intelligence need

to

be evaluated thoroughly (see

Chapter 2 by Kyllonen and Lee; Chapter 17 by Henry, Sternberg, and Grigorenko: and

Chapter 6 by Matthews, Zeidner. and Roberts). Assessing and predicting these constructs

can be a promising new

problems

to

field.

As with

all

new developments,

be solved, on both the theoretical and operational

going to be very hard because established intelligence

been investigated for a long time.

We

maintain our standards before accepting

there are difficult

levels. Initially, this is

tests are very successful

and have

should applaud and support these efforts but

new

constructs as reflecting intelligence.

iii

.

HANDBOOK OF UNDERSTANDING AND MEASURING INTELLIGENCE In our efforts to understand intelligence,

we

are interested in unequivocally assigning

psychological meaning to terms such as intelligence, abstract thinking, reasoning, and so on. However, there

currently no clear consensus of just

is

exactly. Constructs such as mental speed (see

what these terms mean

Chapter 3 by Danthiir. Roberts, Schulze,

and Wilhelm), controlled attention (see Chapter 5 by Heitz, Unsworth, and Engle), and meta-cognition (see Chapter 7 by Hertzog and Robinson) are all candidates for important intelligence-related constructs

when

and candidates for explanations of what happens

test. The basic assumptions and the explanatory mechanisms proposed through such constructs to explain identical, similar, or related intelligence variables need to be explicit and falsifiable (see Chapter 4 by Conway). Focusing strictly on intelligence as cognitive processes and general resources seems to have implicitly excluded the area of knowledge. The measurement of knowledge is crucial because there can be no doubt that knowledge is required to succeed in specific (real-life) domains or in complex problem solving by virtue of an interaction with attentional capabilities. Knowledge is a key concept in understanding real-life performance, and its measurement has long been neglected in intelligence research. Focusing on domain-specific knowledge and its acquisition as an important component of intelligent behavior is likely to become more important in the future (see Chapter 8 by Ackerman and Beier, as well as Chapter 20 by Hambrick). The physiological substrate thought to be most crucial for intellectual functioning is

cognitively

taking an intelligence

the frontal lobes. Relating the functioning and malfunctioning of the frontal lobes to intelligence into

human

is

a specific

form of reductionism and can provide important new insights

cognition in general and individual differences in particular (see Chapter 9

by Kane).

The promise of fundamental new controversial technology of genetics intelligence (see Chapter 10

by

insights gained through the rapidly developing is

likely to

Petrill).

of the most promising completely

have a big effect on

how we

Research on the genetics of intelligence

new developments

and

think about is

one

in intelligence research.

Intelligence research has always profited from developmental contributions to the field. Intelligence testing, as

gogy.

On

Ebbinghaus and Binet invented

the other side, Piaget's developmental theory has

it.

its

has

its

origins in peda-

roots in the analysis of

errors of subjects in Binet's laboratory. Central concepts such as fluid

and crystallized

intelligence gained initial support largely by demonstrating different developmental trajectories in older age.

Two

chapters (Chapter

1 1

by Pascual-Leone and Johnson;

Chapter 12 by Lovden and Lindenberger) address the major recent psychological and methodological contributions of developmental psychology to our understanding of intelligence.

The observafion of group differences in intelligence is an important and pervasive The implications are far-reaching and of great importance in public

issue in testing. policy.

It

is

essential to develop a scientifically

sound understanding and provide

professional advice for policymakers. Although very controversial,

it

is

necessary to

discuss the issue, address scientifically established causes, and outline possible inter-

ventions (see Chapter 13 by Wittmann).

The methodological fundamentals of intelligence

testing

must be understood

any conclusions about theories of intelligence. The methods for dealing with traditionally twofold.

On

a test or subtest level, relations

interest in theories of intelligence structure (see

observational level, item response theory ric

is

between specific

Chapter 14 by Schulze).

crucial in test construction

to

draw

tests are

tests are

On

of

the lowest

and psychomet-

analysis (see Chapter 15 by Schmiedek). In both these areas, there has been sub-

two decades, and this is already affecting how we test our a basic knowledge of these two areas, it is pointless to try and

stantial progress in the past

theories.

Without

at least

Preface

work through many of

the chapters of this

book or any other advanced

treatise

on

intelHgence.

One

particularly controversial

problem

importance of the so-called general abstraction of a general factor, it

factor.

good

tests to

are necessary prerequisites to assess the

research

in intelligence

is

the nature and

Besides assigning meaning to the

statistical

and adequate methods

to extract

measure

it

magnitude and relevance of

this

concept (see

Chapter 16 by Stankov). Faceted models of intelligence,

initially

research, have developed into an arsenal of

intended as a specific theory of intelligence

methods and procedures

that are particularly

measurement approaches. Faceted models helped to put the focus of research efforts on the psychological level of intelligence research in that these models provided the methodological background for discussing intelligence tasks with respect to their psychological demands (see Chapter 18 by SuB and Beauducel). The influence of available intelligence tests on research can hardly be overestimated. useful in conceptualizing

Within intelligence research, as well as when intelligence measures are used areas of psychological research, the availability of measures constraint that has a strong impact on research

However, there

is

tests

established tests actually measure

Reasoning

outcomes as well as applied work.

is

of cognitive

abilities.

intelli-

Gaining insight into what

crucial in understanding

what

test results

mean

by Roberts, Markham, Matthews, and Zeidner).

ability is included in

every structural theory. In hierarchical, higher order,

and topographical theories of intelligence, reasoning ment. Similarly,

in other

an important pragmatic

an enormous gap between theoretically established models of

gence research and widely used (see Chapter 19

is

in

ability is central to the ability arrange-

applied settings, reasoning ability

is

usually the most potent predictor.

The structure of reasoning tests is debatable, however (see Chapter 21 by Wilhelm). Working memory is currently the most prominent candidate for the explanation of many crucial intelligence aspects. Working memory has been particularly fruitful in predicting and explaining intelligence. However, currently there is an inflation of working memory measures, and many of these measures (old and new) are psychometrically unsound and in desperate need of improvement (see Chapter 22 by Oberauer). Learning disabilities are an applied area of intelligence research with rapidly growing importance and relevance. The complicated problem of understanding and measuring specific learning disabilities can contribute importantly to our understanding of

unimpaired intelligence,

to distinguishing learning disabilities

from other cognitive

impairments, and to providing insight into details of the measurement and diagnosis of learning disabilities (see Chapter 23 by Swanson).

Another applied area deals with the validity and implications of intelligence. The contribution of intelligence to the explanation and prediction of real-life

been demonstrated field

in

many

fields.

phenomena has

These contributions are specifically important

in the

of personnel selection (see Chapter 24 by Ones, Viswesvaran, and Dilchert).

We

are particularly glad that

in the field

we were

able to convince

two outstanding researchers

of cognition and intelligence to be discussants for this book. The various

approaches to the understanding of intelligence are discussed and assessed

from Nelson Cowan (Chapter the presented approaches

is

25).

The focus here

likely to

in a

chapter

on the identification of which of deepen and improve our understanding of intelliis

The chapter discusses the contributions of Roberts, Markham, Matthews, and Zeidner; Conway; Heitz, Unsworth, and Engle; Kane; Oberauer; Wilhelm; Hertzog and Robinson; Ackerman and Beier; Hambrick: Pascual-Leone and Johnson; Lovden and Lindenberger; and Swanson. The contributions to the measuring of intelligence are discussed and assessed by Nathan Brody (Chapter 26). The focus here is on gence

in the future.

what we have learned from the measurement of intelligence so

far,

how we

should do

•

ix

HANDBOOK OF UNDERSTANDING AND MEASURING INTELLIGENCE it.

and how we can improve

it.

This chapter discusses the contributions of Danthiir.

Roberts. Schulze. and Wilhelm: Ones, Viswesvaran. and Dilchert; Petrill; Schulze;

Schmiedek: Stankov; Wittmann; SuB and Beauducel: Kyllonen and Lee; Henry, Sternberg, and Grigorenko; and Matthews, Zeidner, and Roberts. These two discussions conclude the book. Primarily, the

book

is

intended for use by advanced undergraduate and graduate

students, researchers, and professionals in psychology

meant

to provide a

background on recent research

and education. The book

in intelligence

develop a sound understanding of results and perspectives

More

generally, the

book

is

supposed to provide

and

is

to help readers to

in intelligence research.

scientists with

broad interests

in indi-

vidual differences, cognitive abilities, intelligence, educational measurement, thinking,

reasoning, or problem solving with a comprehensive description of the status quo and

prospects of intelligence research.

A number of individuals have helped

us substantially in the task of editing this book.

Vanessa Danthiir made important contributions

in skillfully

proofreading

all

chapters

and providing constructive suggestions well beyond simple orthography, layout

for

improvements where necessary, and careful copyediting. Jim Brace-Thompson. Alison Mudditt. Karen Ehrmann, and Gillian Dickens reacted flexibly and supportively to suggestions

we made

all

an effort to improve the book. Their constant support and great

in

editorial service helped us in focusing

on the content side of the book.

Editing this handbook has been an interesting, thought-provoking, and rewarding

experience.

We

would

like to express

of-the-art science to a broad audience to

our gratitude to the authors for providing

and for helping

be a lasting contribution to intelligence research.

this

volume helpful

human

intelligence.

in

to

We

make hope

this

state-

book what we hope

that all readers will find

gaining valuable insights into the theory and measurement of

We

will

be rewarded

provokes new and innovative research on

if this

book advances our

field

and

if

it

this topic.

— Oliver Wilhelm and Randall W. Engle

I

1 Intelligence:

A Diva and a Workhorse Oliver Wilhelm

Randall W. Engle

This

introduction will try to set the stage

for the subsequent chapters.

We

first dis-

cuss a few issues related to the measure-

ment of

intelligence before

commenting on

understanding of intelligence.

We

and behaviors.

and by no means exhaustive

A comprehensive of adjectives

list

includes the following: abstract, accurate, agile,

the

analytic, artistic, astute, attentive, aware, block-

do so because

headed, brainy, bright, broadminded. bubble-

intelligence testing has always been a very prag-

matic job. In an attempt to satisfy the ability testing, the

attributes, states,

demand

for

development, use, and evalua-

measurement tools has been a focal activof researchers working in the area. In the

headed,

captivating,

clumsy,

complex, confused, contemplative,

cerebral,

clever,

clod,

crafty, creative, cultured, deep, detail-oriented,

tion of

distractible,

ity

dull,

ditsy,

flawless,

doddery, doltish, dreamy,

exacting,

fanciful,

foolhardy,

early years of empirical psychological research,

foolish, foresighted, forgetful, fuddled, gifted,

was so success-

hot-headed, ignorant, imaginative, impercepti-

the

measurement of

ful,

with approaches related to

intelligence

its

understanding

so far from being satisfactory, that the

latter

was

overly neglected for a long period. Following these historical lines, efforts to the

we postpone more

recent

toward the understanding of intelligence

end of

ble,

imprudent,

instinctive, tive,

able,

impulsive,

innovative,

nious,

intellectual,

intuitive, inventive,

learned,

inattentive,

inquisitive,

logical,

intelligent,

inge-

insightful,

introspec-

jumbled, knowledge-

madcap, meditative,

mental, meticulous, mindless, musical, oafish,

this introduction.

original,

perceptive,

philosophical,

plodding,

poetic, pondering, pragmatic, precise, prudent,

The Measurement of Intelligence

questioning, quick-witted, reasonable, reckless, refined, responsive,

no lack

In English, as in other languages, there

is

of adjectives to describe intelligent

activities.

AUTHORS' NOTE:

Correspondence regarding

of Psychology. Humboidt-University.

this

scholarly, shallow, sharp-

witted, slow, sly, smart, sophisticated, talented,

uncreative,

uninhibited,

unreflective,

vigilant.

chapter should be addressed to Oliver Wilhelm. Department

Rudower Chaussee

18.

12489 Berlin. Germany. E-mail: oliver.wilhelm@

rz.hu-berlin.de. 1

2

HANDBOOK OF UNDERSTANDING AND MEASURING INTELLIGENCE

.

visional^', well-read, wise,

are just a 100 of

and

many more

in ever>'day

be used for self-reports

These

v\orldly.

adjectives that can

language as

well as in scientific studies of personality to

this

is

intuitively

Cronbach (1949)

plausible.

introduced the distinction between measures of typical

and maximal behavior. The distinction

between both forms of measurement

is

strongly

We

some more or

less dispositional fea-

associated with the content of a measure.

tures of individuals that

somehow have some-

usually assess typical behavior with self-reports

describe

of preferences and valences.

thing to do with intelligence.

As with many assessment procedures, we can use factor analysis and other multivariate

methods

to

summarize data on such

and we would

self-reports,

we

likely find various factors

might label motor

abilities,

precision in working

and thinking, knowledge, reasoning, memory, and the

attention,

on

self-ratings

like. Similarly,

collect

how

comother respondents, they would do on

to provide us with estimates on

pared to

we can

by asking respondents

abilities

We

certain tests.

well,

can help respondents provide

more accurate estimates by explaining ability constructs and giving them examples from actual

ability

We

tests.

to provide estimates of

can ask respondents

how many

ability test they solved correctly.

rely

items of an

We

can also

on the confidence with which individuals

respond to individual items of ability (Pallier et

al..

tests

We

ior

A

behavior,

with measuring

achievement, and proficien-

abilities, aptitude,

cies.

Maximal

side, is associated

prototypical measure of

maximal behav-

can be characterized by several features. The

assessed person

aware of the performance

is

appraisal and willing to perform as well as he

The standards

or she can.

for evaluating the

performance of a person are

explicit,

and the

assessed person has to exert effort to succeed

&

(see Sackett. Zedeck. lar

conceptualization).

Fogli, 1988. for a simi-

The pervasive

distincfion

between measures of maximal and measures of typical

behavior remains despite attempts to

assess constructs from one side of the divide

with methods from the other side of the divide.

There are other fundamental differences between measures of maximal and typical behavior.

Once

is

it

ensured that participants are

&

motivated to do well and have understood what

can also ask individuals to

they are asked to do. precision of measurement

2002; Stankov, 2000: Stankov

Crawford. 1997).

on the other

provide us with self-reports about their typical

engagement (Ackerman 1994: Goff & Ackerman, 1992), need

&

intellectual

Goff.

in the ability

tion,

there

domain can be very high. In addino need to be afraid of faked

is

for cogni-

answers after taking care of routine precautions

&

Jarvis,

because you can hardly pretend to be more able

1996). and openness for new ideas (Costa & McCrae. 1992: Saucier. 1992: Trapnell, 1994). However, empirical data suggest that none of

than you actually are by doing better than you

the above approaches provides an acceptable

and vulnerability

approximation to the actual intelligence

bias, is threatened so strongly that they

tion

as

(Cacioppo, Petty, Feinstein,

measured by standardized

correlations

level,

ability tests.

The

and standardized

ability tests are

tests.

well they do in a broad \ariety of ability

Apparently,

intellectual

asking

for

preferences in

engagement, self-ratings of

ability,

use of

proficiency measures

on the

how

such as response

is

jusfified

validity

and

ability, apfitude,

by the abundant evidence ufility

of such measures.

There can be no serious doubt that

ability tests

provide incrementally useful information that

would improve judgment and decision making. Still, ability tests

are not used for

all

decisions in

which they could provide useful informadon. For example, there are important cultural

from intelligence

ences

does not bridge the gap

betw een putati\e personalit}' and putative measures.

From

a psychometric

abilit>

perspective.

and

for practically important

or confidence rafings of responses to problems tests

cannot

Apart from the lack of vulnerability to some

decisions

to assess

to artifacts

critical biases, the

ronment constantly providing us with information about what we can accomplish and what we

humans do poorly w hen asked

the other hand, the structure of

be applied sensibly for many practical purposes.

usually disappointingly low. Despite our envi-

cannot,

On

self-report measures, as well as their validity

of self-reports of intellect, self-

estimates of ability, and confidence ratings for ability tests

actually can.

in the

differ-

frequency with which ability mea-

sures are used. These differences are not caused

by

cultural differences in the validity

and

utility

Intelligence:

A

Diva and a Workhorse

•

3

of intelligence measures: Those are quite stable.

possible to raise fundamental fairness problems

make use of intelligence measures

not only in the case of ability tests (Willingham,

It is

policy to

for high-stakes decisions in in

one culture but not

&

another (see Schuler, Frier,

Kauffmann,

1999) but also for other forms of information

used

high-stakes decisions affecting the lives

in

not clear, then, what causes

1993, for a comparison in the personnel selec-

of individuals.

tion context). Leaving the decision uninformed

and maintains the differences across countries

about the

of those

abilities

of the decision

likely to

is

who

are the subject

be less precise, less

with respect to the use of ability measures

social

norms and

it

may

be that cultural and

implicit policies place

more

importance on some aspects of selection than on abilities, the utility

of information provided by

can be considered a certainty across

ability tests

A somehow

It

frequently pointed out that predictions

is

based on ability

and

diction,

statement

is

it

tests

do not allow perfect pre-

important to realize that this

is

and

true

for a long time to

is

very likely to stay true

the frequency

is

and the validity of such measures. Rationally, is

—

ceteris paribus

to use the

to

most valid

you are

sions

—evident

in

that

you

it

want

will

of predictors for deci-

set

charge of Now, the knowledge

(set of) predictor(s) is actually

be maximally predictive

is

going

no secret science.

Psychologists around the world have argued and literally

thousands of studies demonstrate the

point over and over again:

You would

not want

to neglect intelligence as a predictor in almost

by such constructs as moti-

any selection decision because the validity of

vation and personality, all

problem

come. Because human behav-

ior is also affected

account for

related

with which various selection methods are used

of which

cultures.

in

high-stakes decisions.

dependable, subject to more biases, and less accurate. Although

It is

will

abilities

never

the \'ariance in performance. In

the prediction

is

going to be worse than

it

would

be with consideration of intelligence. Empiri-

gap between what

addition, to justify using ability measures, one

cally, there is a surprising

should demonstrate that they improve the pre-

should be done and what

diction of success, that they cannot be replaced

Leboyer, 1994; Schuler

by other predictors or combinations thereof,

most countries, most selection decisions are not

that

the costs are sufficiently smaller than the benefits,

and

that there

no fundamental problem

is

made using

done (Levy-

is

et al., 1993). In fact, in

the best predictors for success.

The most

surprising aspect about test use

with fairness in practical applications. Perceived

described above seems to be that no one

fairness of procedures in practical selection con-

surprised, worried, or shocked about those facts.

texts

might be the cause for differences

in the

use of psychometric ability measures across cultures. If this

were

should be a wide-

—a country

have not heard of

of "Test

initiatives

kid" in countries where ability testing

is

too

my

not a

default hurdle in access to higher education.

Nor

which

have we heard of fan clubs for college admis-

measures are used rarely for practical

sion tests or similar ability tests in countries

spread belief in ability

true, there

We

is

selection

—

that

Germany ability

in

measures are unfair

if

they are used in high-stakes decisions such as

admission to higher education or employment. Similarly, in a country with intense

and frequent

where

ability testing is a default

many critical decisions. From all the individual that

procedure for

differences constructs

have been investigated and established

use of ability measures for crucial decisions,

within the past 100 years, the constructs related

such as the United States, there should be a

to intelligence, ability, aptitude,

widespread belief to

that using ability is a fair

way

such decision procedures. Available

assist

ment have a

and achieve-

special status. Ability testing

of the big success stories of psychology.

is

It is

one not

evidence suggests that the cultural differences in

so easy to assign credit for "inventing" intelli-

the evaluation of various selection procedures

gence

either

from applicants or jobholders are too

small (Marcus, 2(X)3: Phillips Steiner

&

Gilliland,

huge differences between

&

Gully, 2002;

1996) to account for the

in the use

of ability

cultures. In addition,

it

is

tests

found

of course

to patriarchs or leaders in a field of intel-

ligence research. There are two famous and

well-acknowledged traditions

in

intelligence

research. First, the methodological innovation

of using the then rather

new

correlation coeffi-

cient to express the association

between two

4

HANDBOOK OF UNDERSTANDING AND MEASURING INTELLIGENCE

.

was

tests

a brilliant idea. Extending this idea to

tetrad analysis

and

later to factor analysis is cer-

one of the most crucial developments for psychology as a scientific discipline, and it can tainly

be traced to Charles Spearman most prominently. Second, the

initial

pragmatic use of

intel-

ligence measures (Binet, 1903) and, only a few

years

later,

method

in

widespread use as a selection

its

World War

I

are clearly the other side

of a unique success story. Both traditions later joined

when

it

came

to standardizing the condi-

tions of administration

and the comparisons of

individual test results with a group of reference subjects,

validating

other test value, and utility

by various means,

tests

mostly correlating some

test

when

value with

some

the demonstration of

of measures was warranted.

The successful use of practical purposes

constructs they are supposed to reflect,

panacea

to

the

raised above.

It

not a

is

problem of indetermination that

is

lack of precision in

deriving measures that caused the inflation of available tasks.

As

a result, frequently the psy-

chological interpretation of the meaning of a specific test score takes not

than two or three lines.

much more

space

A related problem is that

our interpretation of intelligence constructs, as assessed by several intelligence

tests,

frequently

does not go beyond paraphrasing supposed

communality of

content.

test

Although the

mathematics of factor analysis has developed rapidly

and there

decent software widely

is

good factor analysis remains an art much more than a technology because the

available,

composition and nature of the included variables

ability tests for

many

are so crucial.

and the lack of unequivocal

guidelines in deriving indicators from a theory are responsible for the manifold of measures

available today.

thought to

elicit

The number of

responses that allow an evalua-

tion of the observed behavior as

more or

intelligent is hard to estimate, but

it

rising. Carroll's

less

high and

is

(1993) review of individual dif-

ferences research in intelligence

lists

and

classi-

good proportion of tasks proposed so far. One apparent problem with almost all of those tasks is that the gap between what the task is supposed to measure and what it actually measures is not bridged by strict derivation of the

fies a

For example, specification of the measure-

task.

ment

The Understanding of Intelligence

distinct tasks

intention for a task supposed to

measure

mental speed usually leaves countless degrees

The

rapid development of statistical methods

and the overwhelming demand

for the wide-

spread use of intelligence tests caused an imporneglect:

tant

cognitive

trying

vidual differences initial

to

understand the basic

mechanisms responsible

we

label

for the indi-

intelligence.

The

roots of intelligence research are cogni-

tive in nature. Specifically.

Ebbinghaus (1896-

1897) and, to a lesser degree, Binet and Henry

(1898) were fundamentally interested

in

cogni-

processes and their implications. These

tive

were almost completely neglected

roots first

in the

eight decades of the 20th century.

The

of freedom for

rediscovery of cognition in intelligence research

tion takes

how exactly the operationalizaplace. Many of the decisions that need

and the related developments caused a number

be made before having a task that can be used

of important changes that might not be apparent

to

to assess individuals are likely to affect the indi-

vidual differences that can be observed. Thus, is

not clear what exactly individual differences

on the task

reflect.

One can

then ask, "Speed of

what?" With any individual

task, there is thus

always a problem of identifying exactly what the task actually measures.

using

multiple

certainly

is

measures

are

communality of several pretty

personnel selection. There are. however,

more is

many

areas in which the assessment of abilities

a crucial aspect. Granting access to higher

education, college and university admissions,

construct

diagnosis of learning disabilities, diagnosis of

in

a

addressing the

psychometric problems associated with the use of single indicators. However, analyzing the

The validity of ability assessment has many The most prominent examples come from

faces.

The approach of for

major progress

to laypersons.

it

things, all of

which

fuzzy operationalizations of the

all

forms of dementia, identifying

intellectual

and giftedness, diagnosis of attentional disorders, and diagnosis of mental retardation talent

and specific disorders associated with tual

intellec-

problems are just some of the most

Intelligence:

prominent applications. Corresponding with this

manifold of applications

of

mental

is

the

prominence

throughout

abilities

Intelligence

is

our

condition for success in school and college. Intelligence

is

the job, and

it

name

is

associated with income, health,

of daily li\ing. to

activities

few of the abundant correlates.

just a

However,

these valuable data

all

what we do when we behave is

on

the best predictor for success

and instrumental

do not

tell

intelligently,

us

what

required to solve a specific problem, what

can be done to do better on some form of prob-

some

lem, what will help to improve

how our

ability,

brain operates to solve problems, what

processes are accomplished while solving intelligence problems, what

makes a problem harder

than another one. what

we

can accomplish and

why. what makes one individual smart and another one not. and

many more

efforts (Carpenter. Just.

questions that

Consider a widely used form of an

ability test

such as matrices items. This form of

been used for a long time and

is

A

group of researchers took a more

larger

explain fluid intelligence. Early vidual differences in working strated a strong link

and correlational work, individual in working memory have been

mental

differences

established as a crucial construct of

(Conway,

intelligence

tive resources that are

The developments we

just

sketched for

matrices items can be found for

many forms of

traditional intelligence assessment. Categorical

syllogisms or relational inferences, for example,

factor theory

that items of this type required eduction of rela-

tions

and of correlates as pivotal and

role

of

that items of this

that the

specifically,

prior

Cattell thought

type required dealing with

novel situations, although he had

some sympa-

thy for Spearman's notion of eduction. Multivariate analysis unequivocally demonstrates that

items of this

typ)e assess

something

that is at the

core of established intelligence tests (Carroll.

& Snow, 1983; Marshalek. 1984). However,

1993; Marshalek. Lohman.

Snow, Kyllonen.

&

fundamental for human

have a long and scattered histor} in intelligence

Spearman (1938) thought

—would be minimal.

human

Kane, Miyake,

intelligence.

research, too (Baddeley. 1968; Carter. Kennedy,

Werdelin. 1958).

knowledge

Jarrold,

& Towse. in press; Miyake & Shah. 1999). This body of research regresses fluid intelligence and other intelligence constructs on functions of working memor>' and provides interesting evidence about cognitive processes and cogni-

& Bittner.

—

Christal,

highly recom-

test

the construct of fluid intelli-

abilities

&

1990). Subsequently, in a mixture of experi-

subsumed under

other

indi-

fluid intelligence

and working memory (Kyllonen

(Jensen. 1998; Spearman. 1938) but can also be

(Cattell. 1963).

between

work on

memory demon-

has

mended by proponents of a general

gence

Shell. 1990) arrived at

construct-oriented approach in their attempts to

directly relate to the cognition involved in ability testing.

&

similar conclusions.

lives.

a necessary though not sufficient

A Diva and a Workhorse

on

1981; Sternberg. 1980; Storing. 1908;

We

interesting

this

describe two

more

refrain

from elaborating

work and want \er\'

briefly

to

contemporar\ appro-

aches toward the understanding of intelligence.

The

first

approach

is

through recent work on

interference control and behavioral and cognitive inhibition.

The second approach is through neuropsychology and brain

evidence from imaging.

Beginning with task classes well known cognitive

in

psychology, individual differences

were investigated

in

an attempt to find

promising constructs of human

new and

abilities.

that cause conflicts in the cognitive

Tasks

system and

its widespread use. many things were unknown about such matrices items a few years ago. It is in more recent efforts that some deter-

require cognitive processes to cope with these

minants of difficulty have been established

observed with such tasks can be considered to

Other work on matrices items has

reflect interference control, cognitive inhibition,

despite

(Primi, 2001

).

conflicts have tions.

The

been used

in several investiga-

individual differences that can be

Among

used specific models of item response theory

and behavioral inhibition (Nigg. 2000).

(Embretson, 1995) and provided us with

these tasks are the stop-signal paradigm, anti-

pretations

of two-person

parameters

inter-

(general

control processing and working memor>' capacity) for

such measures. Earlier, more experimental

saccadic eye movements, switching a task

set,

the flancker task, the Stroop test, directed forgetting,

and proactive interference,

to

name just

6

HANDBOOK OF UNDERSTANDING AND MEASURING INTELLIGENCE

.

and the

More

important, however,

a few. The results of investigations analyzing

tailored,

individual differences on such tasks have been

the appearance of established intelligence tests

disappointing so differences

are

far.

not

The observed individual very

across

stable

task

classes, or they are not correlated sufficiently

&

with relevant criteria (Friedman

Miyake.

&

2004; Kramer, Humphrey, Larish, Logan,

Miyake

Strayer, 1994;

&

Atkinson,

et al..

2000; Salthouse,

Berish, 2003; Shilling,

& Rabbitt, 2002).

It is

hoped

Chetwynd,

that future investi-

gations will provide hints for communalities

among

individual differences in coping with

cognitive conflicts, as assessed by various task classes. If intelligence research

successful,

we can add

on

this topic is

an important ability to

our set of usual suspects

important

in predicting

like).

might change because ordinary intelligence tests are

hard to develop, and what they assess

is

hard to describe. Forms of assessment that are

&

easier to compile (Kyllonen

2002)

Irvine.

because determinants of difficulty and other item

known and

statistics are

subject to experi-

manipulations might be established.

mental

Similarly, taxonomies of

human

might

abilities

change substantially over the years because we succeed

developing a more profound under-

in

standing of constructs such as fluid intelligence, crystallized intelligence, or mental speed.

These

efforts are not likely to provide us with tests that

beat existing

ones hands down. They just

replace traditional forms of assessment that had

criteria.

The second approach focuses on logical level.

the physio-

Fundamental changes for

intelli-

emerge from

some

many

basic flaws but

applied victories

with forms of assessment that are more thor-

the

oughly developed, psychologically profound,

rapidly developing fields of neuropsychology

but as successful as their ancestors. In the con-

and brain imaging. Not only are we

text of this

gence research are

likely to

likely to

work,

will

it

be desirable to do some

more about the neural basis of general intelligence (Duncan et al., 2000), but there is

fine-tuning on constructs that count, by and

also neuropsychological evidence about the dis-

measurement and

learn

sociation of

more

1996) or on

et al.,

specific abilities (e.g..

how emotion and

Awh

cognition

interact in decision-making tasks (e.g., Sanfey, Rilling.

Aronson. Nystrom.

These new developments

more appropriate and

&

Cohen. 2003).

tive processes. little

by

and profound as

we

should have clear expec-

and similar

issues,

tations.

not very likely that

It is

we

will

improve

of ordinary and available intelligence tests in

to investigate cogni-

These methods are

informed

structs are as refined, solid,

they should be. In our attempts to address these

human

likely to

be of

use for our understanding of intelligence

unless

details of

the predictions achieved through the application

correct models of

methods

all

interpretation of these con-

will help us to invent

intelligence and provide an important addition to our arsenal of

because not

large, as established

psychological

theory.

many

applied settings to a large degree through

improving the form of traditional assessment.

Many

outstanding psychologists have put a

of effort into establishing what will work certain applied context

and what

lot

in

a

will not. This is

Without solid psychological theory, there is no way to understand cognitive processes. Just

not to say that modifying the appearance of

recording physiological data that accompany

slightly

some

tional

cognitive process will not be instrumental.

We do

need psychological theory and a psycho-

logical understanding before

we can make

sense

of some physiological correlate of intelligence behavior. This objection does not

make physiowe need to

logical data any less interesting, but

be careful and sensible It is

in their interpretation.

possible that, in the long run, the appear-

ance of intelligence stantially.

tests

will

change sub-

This could be attributed to new

computerized forms of

new forms of

test

test

presentations and

compilation

(adaptive.

intelligence tests

not worthwhile.

is

improve the

validity

intelligence

acceptance

tests,

in fields

and

We

utility

might

of tradi-

we might improve

where intelligence

testing

was not accepted as useful information in decision making, and we might have sounder instruments

that are easier to

It is

oping

develop and maintain.

we succeed in develnew abilities. There are establish new constructs of

also possible that

tests that assess

several attempts to intelligence,

and they are addressed

chapters of this book.

We

supportive with such efforts because difficult to satisfy the

in various

should be patient and it

is

truly

necessary requirements

.

A

Intelligence:

before a skeptical crowd of scientists

new

is

willing

possible with these

Diva and a Workhorse

•

7

new developments. As

a

in the field

consequence, the basic science of cognition and

of intelligence. To meet the set of requirements,

the technology of intelligence testing are not

to accept that there is a

we

player

need to have close collaborations

will

between cognitive psychology,

differential psy-

connected

properly

been connected

— indeed,

at

have only

they

very beginning.

their

The

chology, psychometrics, and possibly additional

rediscovery of cognition in intelligence research

disciplines.

is

The various

correlates and possible conse-

quences of individual differences

in intelligence

have provoked many heated debates and intense political controversies.

enough

own

sake

is

important

what we do, what awareness we

for

develop, what knowledge

we

important to note

is

It

that intelligence for its

we

compile, and what

understand, discover, and create throughout

our lives to justify almost any level of attention

much more

than

fashion.

a

Without an

informed cognitive understanding of intelligence as a construct, the technology of intelligence testing

going to make

is

to

little

On

no progress.

the other side, testing of cognitive abilities

is

one

of the most important applied fields for cognitive

psychology, and the neglect of this aspect likely to

is

be causal for the lack of implications

cognitive psychology has in applied settings.

Hence,

we would

encourage psychol-

like to

realize that the individual differences in intelli-

more psychometric background to gain some detailed knowledge about the cogni-

gence and their relevance do not simply disap-

tive

devoted to the construct.

It is

also important to

ogists with a

processes underlying intelligent behavior.

with causes, conse-

we would like psychologists with a more cognitive or experimental background to make more use of applied knowledge from

quences, and correlates of individual differences

psychometric research, including individual-

pear

if

we remove

attention

from them. The

opposite can be. and frequently there

is

a need to deal

true.

is,

Hence,

so that policies regarding

in intelligence

how

to use such information can be discussed most

The

coherently.

future

is

Similarly,

book

We

methodology.

differences

hope

that

this

contributes to serve this purpose.

likely to create prob-

lems with an even higher propensity to trigger fundamental conflicts. What would our advice

be when

it

comes

to

problems related

ing for genes associated with different levels of

Would we recommend

intelligence?

the use of

devices and drugs that augment intelligence?

There are many similar questions ahead, and a profound understanding of intelligence

is

a

the

one of

a substantial irony in that

is

most successful constructs of psychological

research resists a consensual definition. Similarly,

there

is

Ackerman,

P.

L..

& Goff, M.

no broad consensus

in the scien-

(1994). Typical intellec-

engagement and personality: Reply

tual

to

Rocklin (1994). Journal of Educational Psycho150-153.

logy, 86,

Awh,

E., Jonides, J.,

Koeppe, R. A.,

prerequisite for sensible advice.

There

References

to screen-

Smith, E. E., Schumacher, E. H.,

&

Katz, S. (1996). Dissociation

and rehearsal

of storage

in

verbal

working

memory: Evidence from positron emission tomography. Psychological Science, Baddeley, A. D. (1968).

A three

7,

25-3 1

minute reasoning

test

community about the conceptualization of intelligence and measurement methods for intelligence testing. This is why we would label intelligence a workhorse and a diva: The construct is extremely useful, but we do not have a

Binet, A. (1903). L' Etude experimentale de I'intelli-

proper definition of what

Binet, A.,

tific

it

is

and what

it

is

not.

Despite this lack of a consensual definition,

our understanding of the principles by which

human minds is

rapidly

advancing

operate, develop, or

progressing. at

Cognitive

malfuncdon science

is

a very fast rate-much faster than

the practical realization of

what

is

or could be

based on grammatical transformation. Psycho-

nomic Science,

10,

341-342.

gence [The experimental study of intelligence]. Paris: Schleicher.

&

Henry, V. (1898). La fatigue intellectu-

elle [Intellectual tiredness]. Paris: Schleicher.

Cacioppo,

W.

J.

T,

Petty, R. E., Feinstein,

J.

A.,

& Jarvis,

B. G. (1996). Dispositional differences in

cognitive

motivation:

individuals

varying

The in

life

and times of

need for cognition.

Psychological Bulletin, 119, 197-253.

8

.

HANDBOOK OF UNDERSTANDING AND MEASURING INTELLIGENCE

Carpenter,

A., Just,

P.

M.

A..

&

Shell,

one intelligence measures:

processing in the Raven

of the

Matrices

account

Progressive

Review;

Psychological

test.

What

(1990).

P.

A theoretical

97.

404-431. Carroll.

A

Human

(1993).

B.

J.

New

survey of factor-analytic studies.

Cambridge University

Grammatical yardstick. Cattell. R. B.

Press.

Factors, 23. 587-591.

A

critical

experiment. Journal of

Educational Psychology. 54. 1-22.

Conway, A. R. A.,

& Towse, J.

J.,

Miyake,

(in press). Variation in

working

memory. Oxford. UK: Oxford University Press. Costa.

P T,

NEO

&

Jn,

McCrae. R. R. (1992). Revised

personality and five factor inventory-

professional manual. Odessa. FL: Psychological

Assessment Resources. Cronbach, L.

New York:

testing.

Duncan, H.,

Seitz, R.

J..

Ahmed,

Harper

&

Kolodny,

J..

A., et

al.

Row. J.,

(2000).

neural basis for

general intelligence. Science. 289.

Ebbinghaus,

Methode zur ihre

bei Schulkindern

for testing mental abilities

with schoolchildren]. Zeitschrift

[On and

fiir

new

a its

use

Psycho-

und Physiologic der Sinnesorgane.

logic

13.

401^57. Embretson.

S. E. (1995).

The

role of

intelligence. Intelligence, 20.

among

working

mem-

&

P.,

in

inhibition

and interference control functions:

A latent variable analysis. Journal of Experimental &

methods:

A partial

replication and extension

German sample. Applied Psychology: An

in a

International Review, 52, 515-532.

Marshalek. B.. Lohman. D. F.

&

Snow. R. E. (1983).

The complexity continuum models of

hierarchical

in

radex and

the

intelligence. Intelligence,

107-127.

7,

P..

Emerson, M.

& Wager. T D.

A. H.. Howerter. A..

J.,

Witzki,

(2000).

The

their contributions to

complex

frontal lobe tasks:

A latent variable analysis. Cognitive Psychology, 49-100.

& Shah, P.

Miyake. A.. ory.

Nigg.

(

999).

1

Models of working mem-

Cambridge. UK: Cambridge University Press.

T

J.

(2000).

On

inhibition/disinhibition

in

developmental psychopathology: Views from cog-

and personality psychology and a working

nitive

taxonomy. Psychological Bulletin.

inhibition

Wilkinson. R.. Danthiir. V. Kleitman.

G.

Stankov.

L.. et al. (2002).

The

S.,

role

Journal of General

confidence judgments.

Psychology. 129. 257-299. Phillips.

J.

M.. to

&

Gully. S.

personnel

M.

(2002). Fairness reac-

selection

techniques

in

Assessment of typical

Journal of

London: Praeger

Strayer. D. L. (1994).

Beyond

a

unitary

in attention.

Human

Resource Management,

13,

1186-1205. Primi. R. (2001). Complexity of geometric inductive

reasoning tasks: Contribution to the understanding of

Kramer. A. F. Humphrey, D. G.. Larish,

9,491-512.

G,

Knezevic.

relations:

ability.

processing

Pallier.

Singapore and the United States. International

Psychology. 84. 537-552.

&

ogists Press.

L. (1992). Personality-

Jensen, A. R. (1998). The g factor: The science of

G. D.,

ed.. Vol. 4,

CA: Consulting Psychol-

Marcus, B. (2003). Attitudes towards personnel selec-

tions

engagement. Journal of Educational

intellectual

bition:

and organisational psychology (2nd

&

industrial

P

Ackerman,

intelligence

mental

Handbook of

(Eds.).

of individual differences in the accuracy of

169-189.

Miyake, A. (2004). The relations

Psychology: General, 133. 101-135. Goff. M.,

M. Hough

126. 220-246.

ory capacity and general control processes

Friedman. N.

Europe. In H. C. Triandis, M. D. Dunnette,

L.

41.

Priifung geistiger Fahigkeiten und

Anwendung

method

457^60.

Uber eine neue

(1896-1897).

H.

Levy-Leboyer, C. (1994). Selection and assessment

unity and diversity of executive functions and

Bor, D., Herzog,

A

Mahwah. NJ: Lawrence

development.

test

389^33.

& Irvine, S. H. (2002). Item generation

Miyake, A., Friedman. N.

(1949). Essentials of psychological

J.

more than) working-memory

Erlbaum.

tion

C. Kane, M.

A., Jarrold,

C,

P.

Reasoning

Christal, R. E. (1990).

(little

pp. 173-190). Palo Alto,

and crystallized

fluid

is

capacity?! Intelligence, 14,

Kyllonen,

in

S.,

Theory of

1963).

(

R.

Human

intelligence:

York:

& Bittner, A. C. (1981). reasoning: A stable performance

C. Kennedy,

Carter. R.

ability

for cognitive abilities:

&

P C,

Kyllonen.

J. F..

Logan.

Aging and

inhi-

view of inhibitory

Psychology and Aging,

human

intelligence.

Intelligence,

30,

41-70. Sackett.

P

tions

R..

Zedeck.

S..

&

Fogli. L. (1988). Rela-

between measures of typical and maximal

job performance. Journal ofApplied Psychology, 73,

482-486.

Intelligence:

&

Salthouse. T. A., Atkinson. T. M..

Berish. D. E.

Executive functioning as a potential

(2003).

mediator of age-related cognitive decline

normal

in

Journal of Experimental Psy-

adults.

chology: General, 132, 566-594. Sanfey. A. G.. Rilling, L. E..

K., Aronson,

J.

& Cohen. J.

A., Nystrom.

Openness versus

Personalauswahl

Much

&

Kauffmann, M. (1993).

itn

europdischen Vergleich

Chetwynd.

A..

&

Rabbitt,

inconsistency

Individual

An

sures of inhibition:

R M.

A.

across mea-

investigation of the con-

of inhibition

in

older adults.

Kyllonen,

R C., & Marshalek,

The topography of tions.

In

R.

J.

abilities.

W. (1996). Fairness

Gilliland. S.

Sternberg. R.

J.

81,

134-141.

(1980). Representation and process in

linear syllogistic reasoning. Journal

of Experi-

ability

B. (1984).

and learning correla-

Sternberg (Ed.), Advances

the psychology of

human

pp. 47-103). Hillsdale. NJ:

in

iiber

einfache

Schlussprozesse

investigations of simple

[Experimental

inference processes].

Trapnell,

A

lexical left turn.

nality, 8,

Werdelin,

11, 1-27.

D. (1994). Openness versus

P.

I.

intellect:

European Journal of Perso-

273-290. (1958).

The mathematical

ability:

Lund:

Gleerups.

Willingham. W. W. (1999). test fairness. In S. J.

A

Messick

systematic view of (Ed.),

Assessment

in

intelligence (Vol. 2,

higher education: Issues of access, quality, student

Lawrence Erlbaum.

development, and public policy (pp. 213-242).

Spearman, C. (1938). Measurement of intelligence. Scientia, 64. 75-82.

Storing, G. (1908). Experimentelle Untersuchungen

Experimental and factorial studies.

Neuropsychologia, 40. 605-619. E.,

of cognitive

Archiv fiir die gesamte Psychologic,

Gottingen: Hogrefe.

Snow, R.

&

Psychology intellect:

[Personal selection in European comparison].

validity

Steiner, D. D.,

tests

93-109.

mental Psychology: General, 109, 119-159.

Schuler. H., Frier, D..

struct

and performance on

France and the United States. Journal of Applied

381-386.

(2002).

121-143.

& Crawford, J. D. (1997). Self-confidence

ultimatum

the

in

ado about nothing. European Journal of Person-

Shilling, V. M.,

fluid intelligence. Intelligence, 28,

Stankov. L.,

reactions to personnel selection techniques in

game. Science. 300. 1755-1758.

ality.: 6.

9

D. (2003). The neural basis of

economic decision-making

Saucier, G. (1992).

•

Stankov, L. (20(K)). Complexity, metacognition, and

Intelligence, 25, J.

A Diva and a Workhorse

Mahwah, NJ: Lawrence Erlbaum.

2 Assessing Problem

Solving

Context

in

Patrick C. Kyllonen

SooNMOOK Lee

test-taking

Introduction

more than

skills

kinds

the

of

real-world learning and performance activities Last year, one of us received a call from a

graduate school colleague

who was working

for

an international strategic consulting firm (hereafter.

ISCF).

to help

if

ETS might

(e.g.,

but recent college graduates from a vari-

increasingly viewed as a social activity operat-

system for

their selection

who work on

small teams and help

solve problems for business clients.

It is

essen-

problem-solving job, ISCF staff told

us.

said the job did not require

knowledge

of standard solutions but instead,

flexibility,

They

"aptitude"

creativity,

and the

through

ability to think things

and devise novel solutions. They also told us

that

19, this

volume), but the tide

within

ing

a

social-cultural-political

trial

testing circles, this has led to a

closely reflect the

activities that test scores are

designed to predict.

develop

tests that

These include measures of such constructs as practical intelligence (Sternberg, 1997; see also

volume), situational judgment

with a standardized assessment that "looks like

(Motowidlo, Dunnette,

an IQ

critical

had served

their purposes, but they

wanted something more up-to-date, and, more important, they wanted a test that

was more

directly relevant to the business analyst job itself.

A common

criticism

of intelligence

and

aptitude tests used for personnel selection and

program admissions abilities out

lack

is

that

they

of context. The view

authenticity

and

that

is

measure that

they

such

reflect

movement

more

to

Chapter

test." It

context

(Gipps, 1999). In both educational and indus-

previously they had been screening applicants

tests

more basic

Lohman, 1993; or even Chapter

is moving toward more contextualized assessments. Assessment is

analysts.

ety of majors

tially a

be able

ing out the virtues of measuring

Business analysts are not

them redesign

business

MBAs

He asked

they are intended to assess (Frederickson, 1984;

Wiggins, 1993). There are counterviews point-

17, this

thinking

&

(Derry,

Carter,

Levin,

&

1990), and

Schauble,

1995). These are constructs reflecting problem-

solving competencies in

compared tude

to traditional

more

practical settings,

intelligence and apti-

tests.

In this chapter,

we

review these constructs

along with methods for measuring them.

We

focus on their definitions and comparisons, as well as address the issue of

why

there

is

so

11

HANDBOOK OF UNDERSTANDING AND MEASURING INTELLIGENCE much

interest in

problem solving

context in

in

both industry and higher education.

We

also dis-

with ways to teach mathematics concepts based

on understanding rather than

rote memorization.

cuss the process of developing and scoring mea-

Polya suggested

sures of such constructs and their statistical and

broken down into four phases: understanding

psychometric properties.

that

problem solving could be

problem, devising a plan, carrying out

the

the plan, and looking back.

associated with

Problem Solving, Situational

approach

Judgment, Practical Intelligence,

course,

is

more

it

not

Each phase has

specific issues. Polya's

limited

to

mathematics, of

and others have suggested similar schemes for problem solving in general. For

AND Critical Thinking as Constructs

example. Bransford and Stein (1984) proposed a

We

used a standard problem-solving approach

(e.g.,

Polya, 1957) for thinking about the job of

the business analyst and designing the screening

assessment. But in the course of thinking about relevant problem-solving constructs,

we

judgment, practical intelligence, and

critical thinking. All of these approaches have

been used

to assess

problem solving

in

reflect different perspectives.

A

two frameworks. Table 2.1

Problem-Soh ins Schemes

Brandford and Stein (1984) -IDEAL"

Pdlxa {1957)

Identify the problem

Understand the

Define and represent

Devise a plan

one way

or another. But they stem from difterent origins

and

(Identify. Define.

also con-

sidered research done under the headings of situational

IDEAL

scheme they called

Explore. Act. Look). Table 2.1 presents these

question

is,

are these constructs or approaches?

problem Explore possible strategies

Act on the chosen strategy

Carr\ out the plan

Look back and

Look back

Problem Solving

From

The construct of problem solving in psychology is often aligned with thinking and contrasted with memorization. An influential view was

evaluate

interviews

analysts from ISCF.

we conducted with business we found that they \ iewed

problem solving on

the

remarkably close

We

their jobs

in

terms

to the Polya-Bransford-Stein

present the categories and vari-

expressed by the mathematician Polya (1957).

framework.

who, following the psychologists Duncker (1935) and Wertheimer (1945). was concerned

ous subcategories emerging from our interviews

Table 2.2

in

Table

2.2.

Problem-Solving Framework

"How

important to know prior to employment?"

(40 statements

in the following

seven categories)

Assessing the nature/scope of a problem

(2/4)

Identifying issues and generadng hypotheses

(3/3)

Structuring the problem and developing plan of action

(3/4)

Gathering information

(10/10)

Undertaking analyses

(8/10)

Synthesizing findings

(2/5)

Implementing a solution

N=

96 (21Tf return

rate

[Web

survey]).

The

importance rating of 2.5 or more (on a scale from within that category. analysts'

The percentages

recommendations

21% 14%

(0/4)

(40)

Total Note:

13% 15% 14% 17%

for

1

ratio indicates the

=

number of statements within

not important to 5

=

6%

100%

the category receiving an

very important), over the total

number of statements

are the target content coverage for the assessments, based on importance ratings, and on

how many

questions from each categor>' there ought to be on the assessment.

3

Assessing Problem Solving in Context

With

Because of the apparent agreement on what constitutes

problem solving, as seen by compar-

ing Tables 2.1 and 2.2.

it

seems plausible

that

would be a problem-solving construct

there

identified in the individual-differences literature.

However,

review of the abilities

in his definitive

literature. Carroll (1993) found no evidence for

such a

acknowledge

factor. Carroll did

blem solving

is

frequently mentioned in discus-

sions of what intelligence

Detterman. 1986) and

is

difference

Sternberg

is (e.g.,

&

the

between reasoning and

psychology,

in

practical use today

ing classroom examinations cations).

Bloom proposed

One

still

in

creat-

(among other appliknowledge

that one's

could be assessed

at

from knowledge

(e.g., recall)

various levels of abstraction

through compre-

hension, application, analysis, synthesis, and

Bloom

words

also provided probe

knowledge,

define,

list,

or label something; for evaluation, summarize,

judge, or explain something).

A more

problem

that

is

solving implies reasoning within a domain. For

is

by college instructors

for each level (e.g.. for

that

difficult to

is

critical thinking.

taxonomy of educational objectives

evaluation.

believe

it

1

view was Bloom's (1956), whose

influential

We

problem solving, as these terms are most com-

monly used

essence of

often treated synony-

mously with reasoning. essential

that pro-

these perspectives,

all

sort out the

•

was

recent treatment of the topic

provided by Facione (1990),

who was commis-

example, domain-specific analogies are com-

sioned by the American Philosophical Associa-

monly employed to achieve inductive reasoning in real problem solving (see Holyoak &

tion

Thagard. 1989. 1995). Reasoning, or general

cal thinking experts

fluid ability, tent,

report

on the assessment of

critical

46

criti-

from philosophy and other

can be measured with minimal con-

academic disciplines. Through the use of the

a domain-

Delphi procedure, the experts argued for and

in a specific

achieved some consensus on the makeup of a taxonomy of critical thinking. The results of their efforts are summarized in Table 2.3.

but problem-solving ability

specific proficiency, best

domain of

to

thinking. Facione assembled a panel of

activity

is

measured

such as psychology, biology,

mathematics, physics, or business. Examples of this relationship are seen in the work of

Wittmann and SuB (1999; see Chapter

volume),

13, this

also SuB, 1996;

who showed

the role

Table 2.3

Critical

Thinking Skills

Critical Thinking Skills

of general reasoning in domain-specific problemsolving activities. Thus, to think

it

may

not be as useful

of someone possessing problem-solving

ability as

possessing problem-solving

Identify relationships

Interpretation

Comprehend

Evaluation

Assess others' credibility

Inference

Query evidence, draw

Explanation

State

Thinking

movement gained

critical

Analyze and evaluate

thinking

one's judgments

force as a reaction to the

There has been some effect from

(perhaps caricaturized) educational practice of

memorization and regurgitation. Cri-

thinking

tion circles.

topic

is

a popular topic in higher educa-

There are numerous books on the

(Amazon bookstore

books with

and justify

one's reasoning Self-regulation

tical

meaning

conclusions

Like problem solving, the

uncritical

the

skills in

some domain.

Critical

Analysis

lists

critical thinking

close to 6,000

in the

title)

and

related efforts

on

critical

this

thinking that one can find on college

campuses. But

this

work

is

almost completely

divorced from psychological studies of abilities.

and

on courses and programs of study

For example, there

is

no

human

critical think-

even a foundation (www.criticalthinking.org).

ing factor represented in Carroll's (1993) or any

sometimes contrasted with

other abilities model or taxonomy. Also, large-

in that critical thinking involves

scale systematic efforts to assess critical think-

CriUcal thinking

problem solving "reasoning unlimited

in

is

an open-ended manner, with an

number of solutions"

of Education. 1995), but this

agreed-upon distinction.

is

(U.S. Department

not a universally

ing,

although

at

ETS developed which

is

one time showing promise "Tasks

in

(e.g.,

Critical Thinking,"

being used in college outcomes studies;

see Erwin,

1998),

seem not

to

have thrived.

14

.

HANDBOOK OF UNDERSTANDING AND MEASURING INTELLIGENCE Sample Item From

Table 2.4

the Educational Testing Service's

MFT-MBA Test

Bertha Sunshine entered the travel business after having traveled regularly to the Caribbean and becoming familiar with the businesses in the region.

She discovered a market need and,

Sunny Fun

in the 1960s, started

Destinations, a travel agency that specializes in providing air and ground transportation, as well as hotel and for locations in the Caribbean.

meal arrangements

...

Bertha

is

may be more vulnerable than wonders how responsible the tour operator should be

faced with an ethical dilemma. Her attorney has cautioned her that she

she realizes,

at least

from

a publicity perspective. Bertha

when she only provides arrangements Fun Destinations does not check

for transportation, lodgings,

and food. However, she admits

the age of travelers or whether minors are in the

that Sunny company of chaperones.

Bertha has decided to use demographic segmentation to divide her market before she gives any additional consideration to redesigning her product offerings or communications strategies.

Which would be

the

most useful demographic segmentation variable, given Bertha's present situation?

(A) Religion (B) Education

(C)

Age

(D) Socioeconomic status

Perhaps

due

this is

to the

expense of assessing

The key idea underlying

and scoring the typically open-ended

critical think-

practical intelligence

due

to the lack of

from standard analytic

ing examinations; perhaps

it

is

consensus on what constitutes effective

critical

thinking. It

gence that

be that

it

is

simply an item type. For

example, ETS's Major Field Test for Masters of

(MFT-MBA)

Business Administration

is

an out-

comes assessment measure (Gonzales, 2002). Each item is said to measure critical thinking, as contrasted with recall. One way it does that is by

tacit

knowledge and

it is

distinguishable

that

intelligence, the intelli-

measured by conventional

is

and academic achievement

tests

may

is

ways.

First,

it

is

tests,

ability in

two

typically not taught explicitly

but must be inferred

—

think of "street smarts."

To get a sense for what this is, consider McCormack's (1994) catchy-titled best-seller. What They Don 't Teach You at Harvard Business School, which, according to the author, covers

presenting business situations requiring exami-

topics important to success in business that

nees to apply domain knowledge

finance,

outside the standard curriculum, such as reading

develop solutions to

people, running meetings, negotiating, and cre-

management, marketing) problems presented item

is

shown

in

to

(e.g.,

in the scenario.

An example

Table 2.4.

ating

good

first

impressions. Second, unlike typ-

ical intelligence measures, tacit

practical intelligence tive,

Practical Intelligence

skills

The concept of

practical intelligence

grew

fall

may

knowledge and

include social, affec-

and conative aspects. This could include such as managing your time, your career,

yourself,

and others (Wagner

may

&

Sternberg,

out of research conducted on tacit knowledge

1990), which

by Wagner and Sternberg (1986). Practical intelligence is a major component of Sternberg's

a motivation component.

(1985, 1997) triarchic theory of intelligence, or

was conducted by Hedlund

successful intelligence, which divides intelli-

first-year

gence into analytic, creative, and practical

University of Michigan's Masters of Business

ties.

A

abili-

comprehensive account of practical

intelligence, including

its

and methods for measuring Sternberg et

al.

(2000).

history, it,

motivation,

can be found

in

A

well involve a social

skill,

or

large-scale study on practical intelligence

business

Administration

et al.

school

(MBA)

(2001

students

program. As a

)

on 422 in

test

the

of the

viability of the practical intelligence concept,

Hedlund

et al.

— —based on

compared a new measure

Successful Intelligence Assessment

the

5

Assessing Problem Solving in Context

•

1

"Personnel Shortage" Scenario (One of Six Scenarios)

Table 2.5

Personnel Shortage" Scenario

The

setting

is

a manufacturing plant facing a personnel shortage.

of overtime and morale

is

Employees

low. (5-15 pages of further description.)

You

are

are the

working excessive amounts

Human

Resource Manager.

What do you do? I.

Situational

Judgment Problems (Abbreviated Paraphrases)

For the "personnel shortage" scenario,

A.

1.

1

to 7 scale.

Immediately hire 200 new workers Meet with unit heads to discuss strategies possible responses

Dealing With Employee Turnover

C.

of the following solutions on a

Dealing With Understaffing

2.

B.

rate the quality

possible responses

Dealing With Employee Morale

possible responses

(Responses are scored by correlating the mean vector of 30 experts' responses with the examinee's response vector across

all

30

to

60 responses, yielding a single score

II.

[a correlation coefficient] for

each scenario.)

Case Scenario Problems (Abbreviated Paraphrases)

For the "personnel shortage" scenario, write a response to the following questions:

1.

What

2.

What would you do

3.

What information

4.

What outcome do you expect? What

is

the

main problem? to address the

main problem?

did you focus on? obstacles?

(Responses are rated by 8 to 10 experts on overall quality and also along 10 dimensions, such as problem identification, solution generation, information use,

Source: Hedlund et

practical

al.

(2001).

intelligence

currently

and outcome identification.)

ideas with the measure

used for admissions decisions, the

General Management Aptitude Test

(GMAT),

to those problems. Table 2.5 presents

from a sample scenario and types and

how

an excerpt

illustrates

they are scored.

A

both item

total

of six

which consists of items commonly found on

scenarios were developed, covering problems

higher educations admissions assessments, such

related to a personnel shortage, strategic deci-

and general logical

sion making, dealing with a poorly performing

as reading comprehension,

and quantitative reasoning. Hedlund

(2001)

subordinate, a consulting challenge, interdepart-

developed two kinds of items, a "situational judg-

mental negotiations, and project management.

ment problem" and a "ca.se scenario problem." The situational judgment problem required

Scenario settings varied and included a bever-

examinees to

to rate the quality

problems presented

the case scenario

in

et al.

of various solutions

business scenarios, and

problem required examinees

write short essays

to

on how they would respond

age company, a nonprofit organization, and a communications technology firm. Examinees were given four of the six scenarios, two in the situational

the case

judgment

(rating) format

and two

problem (short essays) format.

in

16

HANDBOOK OF UNDERSTANDING AND MEASURING INTELLIGENCE

.

Correlation Matrix of Predictor Measures and Graduate (First- Year Business School) Grades

Table 2.6

GMAT U-GPA

U-GPA

SJT-1

.18

SJT-1

.08

.04

SJT-2

-.01

.13

.26

G-GPA

.44

.30

.20

.23

G-Project

.06

.14

.18

.24

Source: Hedlund

Note:

G-GPA

SJT-2

.13

(2001).

et al.

N = 422. U-GPA

= undergraduate grade-point average; SJT-1 = situational judgment problem score based on profile = case scenario problem score based on holistic scoring; G-GPA = graduate (first-year business average; G-Project = graduate (t'lrst-year business school) project rating (1—5 scale).

correlation scoring: SJT-2

school) grade-point

Table 2.6 presents findings. There were three

key ones.

two new measures were

the

First,

independent of

fairly

they were getting

at

GMAT,

suggesting that

different kinds of

skills.

intelligence proves to be predictable

duplicative

established

other

of)

from

(or

cognitive

constructs, such as general fluid and crystallized abilities,

may

it

nevertheless be useful both as

GMAT correlated higher with

an approach to developing assessments and for

first-year business school grades

(G-GPA) than new measures, a multiple regression suggested that the new measures did add to GMAT

uses other than higher education admissions or

did the

job selection, such as job preview and outcomes

Second, although

(e.g.,

from Rr =

.27 with

GMAT and undergrad-

uate grade-point average to R-

adding the situational judgment ilar results for the

GMAT

=

when

.30,

test score;

-

did {r

=

on

findings,

develop inexpensive simulation

judgment and

a

results

Hedlund

et al.

But the

is

psychology

at least

of job tasks, such as typing tests for clerical

workers or "in-basket"

tests for office

man-

presenting their

agers to select those most qualified for the job.

(2001) did not adjust

Motowidlo et al. (1990) suggested the idea of a situational judgment test as an inexpensive simulator, in which situations would be described

in

results are at least

promising

(in

high

reflect skills

impor-

success in higher education settings

The most important

is

that there is

evidence for the independence of practical

fidelity.

situational

Since then, there have been

judgment

tests

developed

in

many

in areas as

intellittle

paper-and-pencil to video-based tests (Weekley

assessments.

There have been criticisms of practical

words) rather than acted out or simulated

diverse as crew resource management (Hedge, Hanson, Borman, Bruskiewicz, & Logan, 1996) and army leadership (Legree, 1995), from

that are not reflected in conventional cognitive

ligence.

personnel

within

for using fairly simple, inexpensive simulations

must be interpreted cautiously

GMAT scores, and

long history

judgment approach may to

tests for select-

.18, .24 for situational

in suggesting that a scenario-based situational

tant

grew out of psychology to

judgment

ing job applicants in selected fields. There

for this range restriction effect due to direct selection.

testing

in industrial-organizational

Situational

work

new measures

because examinees were selected based partly

Judgment

.08), but the

at all (r

case problem scores, respectively).

These

aid.

Situational

sim-

case problem scores). Third,

did not predict first-year project quality

very well

assessment

grades

in predicting first-year business .school

intel-

ligence from other established cognifive factors

& Jones,

1997).

Situational

judgment and

tests

have been used

and they cover

such as general cognitive ability or job knowl-

for both selection

edge

both cognitive and socio-affective dimensions

(e.g.,

Brody, 2003; Gottfredson, 2003;

Jensen, 1993;

Hunter,

Ree & Earles, 1993; Schmidt & However, even if practical

1993).

(McDaniel, Hartman,

Nguyen, 2001).

training,

& Grubb, 2003; McDaniel &

In general, criterion validities

7

Assessing Problem Solving in Context

1

Sample Item From ISCF Assessment

Table 2.7

Headliners Concerts seeks your advice on creating a

Concerts

•

the middle of negotiating with

is in

Web

agreed that for any

divide the remainder.

its

Web

site to

transaction, Headliners Concerts gets

Company

promote

its

30 franchises. Headliners

The company and

franchises.

35%

the franchises have tentatively

of the profits, and the franchises evenly

representatives are worried that this arrangement might

franchises think that the gain in visibility

is

The

fall apart.

small compared to the potential lost revenues from ticket sales and

concessions.

Which of the following (a)

(b)

Can you help the Can the Web site

(d)

How How

(e)

Can

(c)

for

questions are most important for this case? (Choose the best two)

parties negotiate an

can you include the franchises can you publicize the the

company

Web

in decisions

site for

general

mately

p

cognitive

=

.34

tend

tests

to

be

ability

[see

(approxi-

tests

McDaniel, Morgeson,

& Braverman, 2001] vs. Schmidt & Hunter, 1993). Some of

Finnegan, Campion,

p

= .51

[see

the difference

may be due

to the relatively

of situational judgment

reliabilities

tests,

low

on

average (Legree, 1995). As context (situation)-

dependent measures,

made regarding

Web

the design of the

it

may

site?

concert fans to generate interest?

moderately high but perhaps not as high as those for

will like?

acquire the technical capabilities needed to build and maintain a

judgment

situational

agreement the franchises

generate substantial revenue?

also be due to mul-

we made

items, and

Web

three forms of such items

from the information. Each form

B

(34 items),

C

site?

—A (35

(35 items), and

D

items),

(comprising

32 items from the three forms for equating purposes)

—consisted of

Each

three scenarios.

scenario consisted of approximately 12 items.

A

sample scenario and items are presented

Table 2.7

(this is

in

not an actual item but has a

similar look and feel to actual items used in the

A

ISCF's assessment). Forms

D

through

were

tidimensionality in situational judgment tests,

administered to 159, 58, 163, and 153 appli-

with each

situation

cants, respectively.

Situational

judgment

introducing tests are

uniqueness.

thought to have

lower disparate impact than general cognitive ability

tests

(e.g.,

judgment

particularly

istered

tests are primarily

cogni-

lower (22.8 out of 35 items, standard deviation

rather than "what's the best

course of action" (McDaniel Ployhart

We in

&

et

al.,

2003;

Ehrhart, 2003).

developing the assessment for the

inter-

national strategic consulting firm (ISCF).

We

ISCF consultants to find when working for a client, and

interviewed

out what they did

that resulted in the stages of

problem-solving

framework described above (see Table 2.2). We identified actual problems and solutions from further staff

interviews.

ETS

working with ISCF

tions of

The key

[SD]

test

development

staff turned those descrip-

problems and solutions

into actual test

finding

Form

= 4.8)

A

{n

is

that the applicants

=

admin-

159) scored significantly

than the incumbent sample

(27.0 out of 35 items,

SD =

4.6).

evidence for the validity of the

(n-

This

test;

is

191)

some

we were

not

able to gather any additional evidence. Also, as is

adopted a situational judgment approach

initially

and incumbents.

1996;

framed with a prompt of "'what

would you do"

to

2001).

tive ability tests, but they also reflect personality,

was administered

comparing applicants

& Schmitt, & Kabin,

Pulakos

Sackett. Schmitt, Ellingson,

Situational

A

Form

191 incumbents also for

et

common al.,

moderate A,

with tests of this type (see McDaniel

2003, Table

B,

(r^^.

C,

=

1),

reliabilities

.73, .64, .67,

were only

and .64 for Forms

and D, respectively). Examinees

reported liking the assessment, finding

it

quite

or extremely interesfing (68%), and almost

thought

it

all

served as a good job preview (94%).

Situational

judgment

tests

such as

this are

not limited to business contexts or even to prob-

lem solving. For exainple, we are currently developing a video-based communication

skills

assessment for the Association of American

18

.

HANDBOOK OF UNDERSTANDING AND MEASURING INTELLIGENCE

Medical Colleges

which the context

in

&

medical setting (see Ettienne

The assessment

is

is

the

Problem Solving

Julian, 2001).

similar to the

ISCF

Despite the fact that problem solving

assess-

domain-specific activity

ment,

in that

it

consists of several scenarios,

each of which

accompanied by several

is

lems

items of varying types. material

—both

It

differs in that the

the scenario and the problems

and possible responses

—

is

presented in video

squabbles

elements, as

and sis,

format.

we

— there

are

illustrated

2.2). In his reanalysis

above (see Tables

However, he did not even attempt validity

for

Instead, he argued that

Are the constructs

just reviewed here

—problem

solving, critical thinking, practical intelligence,

and situational judgment

—

in fact constructs, or

are they something else, such as labels for a set

of loosely related tasks, or even assessment

methods? One something

which

it

validity.

for judging

criterion

a valid construct

is

is

whether

the degree to

possesses convergent and discriminant

Convergent validity

among

correlations

is

indicated by high

various, disparate measures

of the construct. For example, sures of problem solving

—

if

disparate

mea-

from multipleof business problem solving to

choice tests

say,

Likert scale items of problem solving in a

—

problem-solving construct. Discriminant validity is

measurement independence construct from other established ability

indicated by

of the

constructs, such as general cognitive ability (g).

For example,

if

measures of problem solving

were more highly correlated among themselves than they were with measures of something else such as inductive reasoning or learning

—

ability

— then

that

would

constitute evidence for

is

The higher

this context.

(Campbell

Fiske,

This suggests that problem solv-

ability factor.

ing

is

not a valid construct per se but rather a

description of a cognitive activity associated

with the general fluid ability factor

—one

pos-

sessing the features listed in Tables 2.1 and 2.2.

This description

is

consistent with other non-

individual-differences approaches

psychol-

in

ogy in which reasoning ability is seen as being employed in problem-solving tasks or activities (Holyoak

&

Critical

Thinking

Thagard, 1989, 1995).

There are domain-independent measures of critical thinking,

such as ones by Watson and

Glaser (1980). Ennis (1996). and Facione and

Facione (1990). There are also domain-specific thinking, such as

ETS's

Tasks of Critical Thinking and ETS's

MFT-

measures of

MBA

critical

programs. Neither kind of measure

domain

specific or general

—has been submitted

to the extensive construct validity analysis that its

con-

among

Carroll (1993) did not even include measures of

the correlations

lower the

evidence for construct validity

&

599) was rather a description of the

kind of task characterizing the general fluid

vergent or discriminant validity. Consequently,

correlations with other-construct measures, the

the

with induction, reasoning, and visual perception; see p.

a relative expression in

same-construct measures and the

greater

problem solving.

problem solving (along

would warrant making statements on

discriminant validity.

Highly correlated

valid-

to identify

med-

communications context were highly correlated, that would be evidence for a broad ical

2.1

based on factor analy-

Carroll (1993) suggested that problem-solving

discriminant

Constructs or Methods?

a

human relasome common

measures did indeed possess convergent ity.

is

math prob-

very different from solving

is

tions

— solving

1959).

particularly confirmatory

Factor analysis,

factor analysis, can

critical

thinking in his review. There

considerable interest in critical

shown by

is

certainly

thinking,

number of books and Web

the

sites

as

on

the subject, the existence of an annual international conference

on

critical

thinking (now in

its

be treated as a formal method for evaluating

24th year), and the number of tests being sold to

construct validity via convergent and discrimi-

measure

nant validity

sus on what critical thinking

criteria.

Let us consider each in turn.

of the

constructs

therefore

it.

There also

may

is

somewhat of is.

a consen-

Critical thinking

perhaps be most usefully thought

of as a set of intellectual activities, such as

9

Assessing Problem Solving in Context

and evaluation, and perhaps an

critical analysis

approach

as well as a set of ideas for, guid-

to.

ing the development of an assessment. This

how we

in fact,

MFT-MBA — as

oping the

is,

treated critical thinking in devel-

an approach to guide

item writing.

or so, there

is

1

a sufficient database to have

conducted meta-analyses on them. The key findings have been that, validity

their criterion

first,

reasonably high, almost as high as

is

that of general cognitive ability. is

Second, there

no consensus on the construct intelligence

practical

validity of

judgment. The construct could be

situational

Practical Intelligence

now

•

&

(Wagner

Sternberg,

knowledge for success (Oswald, Schmitt, Ramsay, Kim, & Gillespie, 2002), or multiple dimensions (Chan & Schmitt, 1997). Some contend that situational judgment tests could just be a measurement method rather than a specific construct (McDaniel et al., 2001; Weekley & Jones, 1997). Third, there is a smaller difference between performances of Black and White examinees than is found on 1993), basic

There are now numerous measures of

practi-

cal intelligence specially tailored to occupations

as diverse as military

commander, salesperson,

school principal, and college student (Sternberg 2000). Extensive studies also have been

et al..

done on the convergent and discriminant

validity

of practical intelligence measures. For example, Sternberg et lations

al.

(2000) reported fairly low corre-

between

Knowledge of

their

measure

Military Leaders

.18-25

(pp.

Tacit

1996; Sternberg, 1985), but these

but the situational specificity of such measures

=

.14-.42 with

that these findings

have is

independent of general cognitive abihty or that a general practical intelligence fac-

tor (Gottfredson, 2003).

And

indeed, the

(2001) study presented above

is

Hedlund

not clear-

cut in that respect. Multidimensional scaling has

been applied Horvath

et al.,

tive ability

a link with

—

analyses have been mostly limited to total scores

must be considered an important

and have not sufficiently determined whether

Different studies

practical intelligence

is

uni- or multidimensional.

Further research will be needed to settle the issue.

For example,

it

would be useful

conduct a

to

structs

number of

gence measures and batteries

diverse practical intelli-

(b) several cognifive ability

consisting of both

fluid

(reasoning,

In that

difficulties

even

summary, there may be some evidence

measures of situational judgment and pracdistinguishable from mea-

far

from being clearly established. However, if

such measures are not clearly distinct

from other established measures by standard

pendent of the already established factors of

validity criteria, they are

general fluid and crystallized abilities.

able by the

way

in

somewhat

Judgment

Because numerous situational judgment have been developed over the past decade

tests

distinguish-

which they are developed.

Indeed, Hanson, Horgen, and

Situational

of meta-

sures of general cognitive ability, but the case is

factor inde-

Jones, 1997), suggesting

interpretation

tical intelligence are

measures. Such a study could establish whether

was a general

&

likely cause.

different con-

analyses.

problem-solving) and crystallized (knowledge)

practical intelligence

may measure

(Weekley

possible

large study involving the administradon of both (a) a large

be a

to practical intelligence data (e.g.,

Some have argued

et al.

13, this

to

measures show (there may even be emodonal intelligence; see Chapter 6, this volume). These findings, which can be viewed as providing evidence for discriminant validity of situational judgment tests, support their construct validity. However, as with practical intelligence, there is a weakness in the evidence for convergent validity McDaniel et al. (2001) found a great deal of variability in the validity coefficients of situational judgment tests. This may be due to a number of causes,

/"

not proven the case that practical intelligence

is

Chapter

196-197). indicating discriminant

and correlations of

there even

(cf.

and fourth, there seems

volume),

greater correlation with personality than cogni-

with r

ability,

effectiveness ratings, indicating convergent validity.

cognitive ability measures

=

measures of general verbal

validity,

of

(TKML) and

Borman (1998)

concluded that situational judgment "best viewed as a

tests were measurement method, rather

than measures of a distinct individual differ-

ences construct."

20

HANDBOOK OF UNDERSTANDING AND MEASURING INTELLIGENCE

.

Methods and Item Types for Measuring Them

which we referred

experts),

we

workshops. For these,

to as item-writing

(the item writers

and a

psychologist) reviewed the transcripts and pre-

Developing contextualized assessments

—whether

they be problem solving, critical thinking, situational

at the transcripts

judgment, or practical intelligence

involves interacting with subject matter experts or people

who do

knowledge,

the activities requiring the

skills,

and

designed to measure.

abilities

the

test

is

necessary to interact

It is

with such experts to get a sense for what they do, as well as the words they use in describing

what they do, and

to write actual items. This is

required for content validation of the assess-

many different approaches to doing this, but here we describe the one we used in developing the ISCF assessment. Figure 2.1 presents the method we used. We began with some test specifications in this ment. There are

Table 2.1. 8

ISCF

We

framework

listed in

then presented this framework to

consultants in two-at-a-time interviews

and asked them whether

main phases of

this characterized the

their job and. if not. to correct

and express the phases of words. This resulted

their

job

in their

own

a modified problem-

in

included

several

listed), for a total

more

Each of these phases

specific

phases (not

of 40 specific phases.

We

then

taxonomy out to 468 consultants and received back 96 completed forms. The question we asked them concerned the degree to which they engaged in (or the amount of time they spent on) each of the 40 specific phases of the job. We then retained the phases for which more than 50*7^ of the consultants said they regularly engaged in that sent the modified problem-solving

activity

and used those

to guide the

We

as our test specifications

then interviewed 25 consultants on the

telephone (the third set of experts). For each interview,

worked and

to

we asked them

on, to describe

to recall a case they

what the case was about,

walk through the choices they were con-

set

of

that the

This

for.

where they

fourth group of experts suggested

thought the key problems and decision points

were

in the scenario,

and what they thought the

would

best approach to those problems

some

be. In

cases, their suggestions followed the item

writers'

decisions:

in

other cases, the experts

made new suggestions both about where sion points were and

were.

The

the deci-

what the best responses

fourth group of experts also suggested

foils in addition to the correct responses.

During

would

this time,

we

also considered

We made

deliver the items.

and-pencil and

Web

versions of the

considered the various item types

and

settled

on three

—multiple

how we

both papertest.

We

we would

also

use

choice, multiple

response (multiple choice with more than one correct answer), and rank ordering (the

examinee

ranks options rather than selecting the best one). quite

is

It

common

for situational

judgment

and practical intelligence item development

to

include a fourth item type, the Likert scale, in

which examinees

on a 4-

indicate,

to 9-point

how effective a particular response would be (or how likely it would be for them to select scale,

a particular response).

(Open-ended or con-

structed response formats are also possible, of

course.) However,

we

did not include the Likert

scale item type in the

ISCF assessment. We

believe that Likert scale items can be quite useful in there to

development of items.

come up with and

item writers had already written items

solving taxonomy, the major phases of which are presented in Table 2.2.

of scenarios that the third

experts had already

—

case, the problem-solving

The

liminary items with another set of experts.

process involved the experts giving a second look

is

many

assessments, particularly

less than a

when

consensus among experts as

what the key (correct answer) ought

to be. In

ISCF assessment, there agreement among experts (and

the particular case of the

was

sufficient

between experts and the keys,

ETS

and we thought

item writers) about it

therefore useful to

include only non-Likert scale item types.

fronted with and the decisions they made. These

cases were then transcribed, and ers

(who

specialized in business,

ing, or verbal reasoning)

from the third

ETS

item writ-

math item

wrote preliminary items

situation transcripts.

We

Scoring

writ-

then began a

round of interviews (with a fourth group of

An

important theme

lem solving

is

in

that experts

contextualized prob-

may

best response to a situation.

disagree on the

Making a decision

1

Assessing Problem Solving in Context

•

2

Item-Type \

Compendium

\\

Item Types

&

Web

Delivery

(or Other)

Delivery '' 1

—

Test

—

Content

»•

Specifications

*

—

Keying

Test

>

Assembly

Elicitation

Use Construct Frameworks

"

t

1

Expert

L

Expert Polls

Reviews & Revisions

Interviews 1

1

Modify Wording

Credit

Model

Through Expert Interviews 1

Figure

The Test Development Process

2.1

way

about which call

to

respond

no clear

with

is

or

right

Rank-ordering items present similar credit

may

pre-

sent difficulties not typically encountered with

multiple-choice knowledge and

There may be several useful dimensions characterizing scoring issues for these kinds of

measures.

One dimension

with the

fore-mentioned (and omitting con-

structed response items

types

are

is

item type: Keeping

from

multiple

this discussion),

multiple

choice,

response, rank order, and Likert scale.

dimension

"d'"

is

assigned?

credit

assignment problems.

For the ISCF assessment, keys were unanimous (agreed upon by item writers and

and so multiple-choice item scoring was straightforward. However, with the other two item types rank ordering and multiple response we had to deal with the issue of partial credit. In practice, various schemes we experts),

ability tests.

item

how

"c" and

wrong answers.

Consequently, scoring such measures

conventional

— then

often a judgment

the degree of consensus

is

A

second

on the key,

—

—

experimented with did not make much

differ-

we

tested.

ence in the scores of the samples Thus,

we

opted for a binary scoring scheme in

which one

or,

in

some

cases,

more than one

ranging from unanimity to variability. Multiple-

response pattern (either ranking or set of multiple-

choice items for conventional ability tests are

response selections) was given

unanimously keyed. But they do not

typically

have to be.

If

experts disagreed, then

it

might

make sense to give partial credit in proportion to the number of experts who chose a particular key. For example, if

"a" and

40%

ing "a" selecting

60%

of the experts chose

chose "b," then examinees select-

would get .6 credit, and examinees "b"' would get .4 credit. Multiple-

response and rank-order item types present even greater

challenges.

is

there

is

example, "b,"

20%

and

The

Likert scale item type will almost always

have the problem of key variability

—

it

would

be rare to have experts unanimously agree to a rating of a response to a 7-point scale. Thus, various,

problem on,

complex

schemes have become commonplace ments using Likert

scales.

say, a

partial-credit in assess-

The schemes can be

divided into two general approaches, distancescoring schemes and profile conelation schemes.

Distance-scoring schemes involve scaling the

"a" and "b" and the examinee selects

difference between experts' ratings and exami-

"a" and "c," if

multiple-response

full credit,

other responses were given no credit.

the issue of partial matching. If

items, there the key

With

all

is

how much

variability in

70%

credit

is

warranted? Or,

what experts

select

—

for

of the experts selected "a" and

selected "a" and "c," and

10%

selected

nees" ratings. For example, one can take the

absolute value of the difference or the square root of the difference, the square root of the

sum

of the squared differences, and so forth. For set

22

HANDBOOK OF UNDERSTANDING AND MEASURING INTELLIGENCE

.

problems

—

that

ones

is,

which several Likert

in

and the examinee's

scale items follow a prompt,

task

is

each response

to rate

tions can also be used.

the correlation

—

These involve treating

between the pattern of expert

score for that item

due

set.

rat-

Profile correlations elim-

to experts

recruitment tools.

Another advantage

profile correla-

ings and the pattern of examinee ratings as the

inate bias

previews" and can therefore serve as valuable

and examinees using

assessments aids

Teaching to the

how

can be valuable training

learning or job situation

for the

nees

to highly contextualized

that they

is

amounts

test

to teaching

itself.

exami-

respond to the kinds of problems

to

they will actually face on the job or in the learning situation. Contextualized assessments such

different parts of the scale.

another dimension along which scoring

as these can also be used as self-assessments,

schemes can be arrayed concerns which group actually does the keying. Thus far, we have con-

allowing a student or prospective employee to

Still

sidered experts as the

key (perhaps

But

in consultation

some

in

ones who

capabilities in the domain, enabling the obser-

with item writers).

vation of increased efficacy over time. Further-

when

circles, particularly

there

is

considerable variability in experts' judgments as to the correct key, the population itself

feedback on his or her problem-solving

get

determine the

can determine the key,

more, because contextualized measures such as this are

a result of training or experience, such measures

a "majority

can serve as outcome measures as well as

in

rules" kind of scheme. This approach could be

predictor measures.

applied to any of the scoring schemes discussed thus

far,

although typically

this is

thought to reflect knowledge acquired as

of examinees

only applied

Still

another advantage of some of the con-

textualized problem-solving assessments

is that

to certain items in the emotional intelligence

they appear to be associated with smaller differ-

unusual judgments, such as

ences between subgroups, particularly male-

literature requiring

the emotional expression of a piece of art (for a discussion, see Davies. Stankov,

1998; see also

&

Roberts,

MacCann, Roberts, Matthews,

&

female and Black-White.

sample size

sufficient

ences

in

to

We

did not have a

examine these

differ-

our studies with the ISCF. However,

reduced subgroup differences have been found

Zeidner, 2004).

repeatedly with both situational judgment tests

(McDaniel

et

2001) and practical

al.,

intelli-

Advantages and Disadvantages

gence measures (Hedlund

Even if contextualized assessments did not show increased validity (or even incremental

tages to these kinds of assessments as well.

On

validity)

over conventional ability measures,

they possess several advantages over conven-

measures

tional

could warrant their use

that

operational settings. that there is a strong

what

One major advantage is user demand for tests with

traditionally has

ity," that is,

in

been called "face valid-

for tests that closely resemble the

et al.,

the other hand, there are

major one

2001).

some disadvanThe

point appears to be that their

at this

reliability is consistently

found

to

be substan-

lower than the reliability of conventional

tially

ability

measures. This means that contextual-

ized measures will take

ventional

measures

same precision

much

longer than con-

provide scores

at

the

level as conventional scores.

For

to

example, using the Spearman -Brown formula, a with a reliability of .60 would have to be

subsequent learning, training, or job perfor-

test

mance

longer by a factor of 6 to reach a reliability of

situations that successful

soon face. In our

ISCF

staff,

it

was

initial

examinees

will

discussions with the

clear that they

wanted an

assessment that consisted of problems that

—

.90

for example, a 20-minute test

to be

120 minutes.

A

second disadvantage, and perhaps one

closely resembled the kinds of problems that

related to the

business analysts would later encounter on the

worked out

job.

There simply was no

abstract,

"IQ

bulk of the

interest for the kind of

test-like" tasks that abilities literature.

make up

the

Particularly in

business contexts, such tests double as "job

first, is

As

simply not

schemes

for these

illustrated above, there are

possibilities, but there is very little research

on the topic thus been

that the field has

the best scoring

kinds of measures.

many

would have

tried

far.

Scoring schemes that have

have ranged from simple

to

complex.

Assessing Problem Solving in Context

we

but

simply do not have enough experience

with these approaches to say which ones might

work

which ones may be more or

best,

less sus-

ceptible to clever coaching strategies (e.g., for

Likert scale items, answering at the extremes

lower scores), and

likely to result in

how

it

be to communicate the scoring schemes to

will

examinees

to

make

the assessment

that

the balanc-

ing of multiple considerations rather than simply the recall of predetermined answers. Still,

number of challenges remain. The

a

mary one

pri-

that using contextualized assess-

is

ments may require considerably more testing time than

fair.

judgment and

requires thinking and

is

easy

and fuzzier side of decision making

softer

23

•

tings

due

allocated in current selection set-

is

lower

to their

reliability.

Perhaps

this

increase can be justified as a result of the addi-

Conclusions and Future Directions

tional benefits of such assessments

— such A

"job preview" and training benefits. In recognition of the validity ceiling

from con-

as the

second

major challenge concerns exploring issues that

ventional cognitive assessments for personnel

come along with widespread

selection and admissions, researchers have sug-

such as their suscepfibility to coaching.

new

gested considering various

constructs to

Among

supplement current measures.

the ones

have received the most attention are mea-

that

sures of problem solving, critical thinking, situ-

operational use, Sfill

another major challenge concerns developing robust and fair scoring schemes for assessments in

which experts themselves might not agree on

what constitutes a correct answer.

and practical intelligence. All

ational judgment,

these stem from different research traditions but

how

overlap considerably in

they are used and

what they mean. Based on the

of possessing convergent and discriminant

ria

validity,

it

is difficult

to

make

the case that these

constructs are in fact valid constructs in the psy-

chological sense. sures possess a

On

the other hand, these

number of

References

traditional crite-

mea-

distinct advantages

over their traditional counterparts.

Among

these

are that because of their contextualized nature,

they possess face validity and therefore are appeal-

Bloom, B.

S. (Ed.). (1956).

goals:

Taxonomy of educational

The classification of educational

objectives:

Handbook I,

Bransford,

D.,

J.

&

Stein, B. S. (1984).

problem solver: learning,

Brody,

N.

and

A

creativity:

New York:

Construct

(2003).

Sternberg Triarchic Abilities Test: reanalysis. Intelligence, 31,

for multiple purposes besides

most notably

tion

as educa-

and training materials and as outcome

measures to evaluate the effects of various ventions.

may

such measures have

Importantly,

shown fewer subgroup

inter-

differences and therefore

see increased use as a

means

to reduce

disparate impact of existing selection procedures.

For these reasons,

we

are likely to see grow-

ing use of contextualized assessments of types, even

if

all

such measures do not carry with

them any increase

in predictive validity. All else

Campbell, D.

T.,

B. (1993).

J.

And one

Human

Cambridge University Chan, D.,

&

cognitive abilities:

New

Press.

paper-and-pencil method of assessment tional test

A

York:

Schmitt, N. (1997). Video-based versus

judgment

tests:

in situa-

Subgroup differences

in

performance and face validity perceptions.

Journal of Applied Psychology, 82, 143-159. Davies, M., Stankov.

construct.

potential users.

& Fiske, D. W. (1959). Convergent and

survey of factor-analytic studies.

tages of contextualized assessments over con-

many

Comment and

319-329.

matrix. Psychological Bulletin, 56, 81-105. Carroll,

tional

ventional ability assessments will outweigh their

of the

discriminant validation by multitrait-multimethod

being equal, or even slightly unequal, the advan-

disadvantages for

IDEAL

Freeman.

validation

as doubling as '"job previews." Also, they can

more easily be used

The

guide for improving thinking,

ing to clients and users for several reasons, such

selection or admissions,

New York:

cognitive domain.

Longmans, Green.

S.,

&

Roberts, R. (1998).

In

search

Emo-

of an elusive

Journal of Personality and Social

Psychology Derry,

L.,

intelligence:

75,

Levin,

J.

989-1015. R.,

&

Schauble,

L.

(1995).

can argue that the unexplored territory remaining

Stimulating statistical thinking through situated

research concerns the increasingly

simulations. Teaching of Psychology, 22, 51-57.

in

abilities

24

.

HANDBOOK OF UNDERSTANDING AND MEASURING INTELLIGENCE

Duncker. K.

(

1935). Ziir Psychologie des produktiven

Denkens [To

psychology of productive

the

Ennis. R. H. (1996). Critical thinking.

Upper Saddle

and assessment

Definitions

thinking, problem solving,

methods of

critical

and

Washington. DC: U.S. Department

writing.

R

M..

&

Julian. E. R. (2001). Identifying

behaviors of successful medical students and

Washington, DC: Association of

residents.

American Medical Colleges. Facione,

P.

A

(1990). Critical thinking:

assessment and instruction.

Facione,

P.,

statement of

Academic

Millbrae,

CA:

& Facione, N. (1990). California Critical Thinking Skills

Academic

California

Frederickson. N. (1984).

Horvath,

Thagard,

MIT

A..

J.

Millbrae,

Test.

CA:

J..

Forsythe. G. B..

Bullis. R.

C.

Williams, W. M.,

knowledge

in military leader-

J..

et al. (1996). Tacit

Army Technical Report 1042). Washington. DC: Army Research Institute for the Behavioral

U.S.

and Social Sciences.

knowledge." Current Directions Science.

Legree.

P.

2.

real test bias: Influences

ment. Review of Research in Education, 24.

unveils industry's first out-

comes assessment for MBA programs. Princeton, NJ: ETS.

develop fluid

abilities.

intelli-

claims and evidence. Intel-

343-397.

M.

E.,

&

Borman, W. C.

November). Situational judgment: An

and at

(1994).

Hanson, M. A., Borman, W. C,

&

Logan, K. K. (1996,

May). Predicting the crew resource management skills

of Air Force pilots (Institute Report

No. 283). Tampa, FL: Research Hedlund,

J.,

Personnel

Decisions

Institutes, Inc.

Plamondon.

Ashford, Practical

S.,

&

McDaniel. M. A., Hartman, N.

K..

Wilt.

intelligence for

J.

J..

Annual Society

for Industrial

Nebel.

K..

(2001, April).

business:

beyond the GMAT. Paper presented

Going

at the

18th

and Organizational

Psychology Conference, San Diego.

S.,

&

Grubb, W.

(2003, April). Situational judgment

L..

tests,

knowledge, behavioral tendency. Paper presented at the

1

8th Annual Conference of the Society

and Organizational Psychology.

Orlando, FL.

McDaniel, M. A., Morgeson,

&

A.,

situational

performance:

M.

McDaniel,

Situational

A

F. P.,

Finnegan, E. B.,

Braverman, E.

judgment

R

(2001).

job

tests to predict

clarification of the literature.

A.,

&

judgment

Nguyen, tests:

A

N.

T.

(2001).

review of practice

and constructs assessed. International Journal of Selection

Motowidlo. (1990).

Sternberg, R.

they don't teach you

Journal of Applied Psychology, 86(4). 730-740.

Testing Association. Pensacola. FL.

Bruskiewicz. K. T,

What

Hanard Business School. New York: Bantam.

Use of

W.,

Personality

tests.

Individual Differences, 36, 645-662.

ment (Symposium AB-34D). Paper presented

J.

&

weighting of performance-based

option

Campion, M.

the annual meeting of the International Militar\

Matthews, G.,

Emotional Intelligence (EI)

alternative approach to selection test developat

D.,

(2004). Consensus scoring and empi-

for Industrial

Hanson, M. A., Horgen. K.

to

Educational Researcher,

MacCann. C. Roberts, R.

III.

Gottfredson, L. S. (2003). Dissecting practical Its

Teaching and testing

(1993).

F.

McCormack, M.

355-392.

gence theory:

lAl-lbd.

testing procedure. Intelligence, 21,

Lohman. D.

rical

ETS

Psychological

(1995). Evidence for an oblique social

J.

Psychologist. 39. 193-202.

Gonzales, K. (2002).

in

9-10.

22. 12-23.

Press.

The

Gipps, C. (1999). Socio-cultural aspects of assess-

Hedge,

(1995). Mental leaps.

P.

Press.

Sternberg. R.

P

Sweeney,

Zeidner,

(1998,

&

J..

of testing on teaching and learning. American

ligence, 31,

Cognitive

intelligence factor established with a Likert-based

Press.

Thinking Dispositions Inventory and California Critical

(1989). Analogical

satisfaction.

Jensen. A. R. (1993). Test validity: g versus "tacit

expert consensus for purposes of educational

California

R

Thagard.

ship: Supporting instrument development (U.S.

of Education. Ettienne.

Holyoak. K.

Cambridge:

River, NJ: Prentice Hall.

(1998).

&

J..

mapping by constraint Science, 13, 295-355.

thinking]. Berlin: Springer.

Erwin. T. D.

Holyoak, K.

and Assessment,

S. J.,

An

Dunnette,

M.

9(1-2), 103-113. D..

&

Carter, G.

alternative selection procedure:

low-fidelity

simulation.

W.

The

Journal of Applied

P.'

activity occurs in other brain regions, too, but that activity

is

quickly supplanted by an inter-

fering stimulus. Thus, a pattern of activity, sus-

PFC. can represent task goals and.

tained in the

in turn, bias activation in

which a weaker neural path-

tant in situations in

way must be

particularly impor-

is

activated to achie\e a goal and a

more frequently used pathway, must

stronger,

be suppressed. According to Miller and Cohen's (2001) model, the biasing influence of the

and

functional

PFC

fMRI environment. They back

of stimuli

the respondent

Thus,

in the

stream

long-term

memory). As an

this biasing

can affect behavior,

When

consider the Stroop color-naming task.

presented with a conflicting stimulus, such as

word red printed

the

response

habitual

respond

"red."'

in

green ink, the stronger,

word and

read the

to

is

but the goal

to

is

name

the ink

must be repre-

color. Therefore, the task goal

M

and

task,

B

the one that

three back. Lures typically

is

the letter either n

Duncan, Burgess,

control,

with

&

Freer, 1996). Interestingly, patients

PFC damage,

lesser

WMC,

(Kane

&

tend to have less difficulty in the

when

Stroop task

as well as individuals with

the trials are incongruent

all

2003) than when the task

Engle,

includes congruent

trials.

When

incongruent, the task goal color)

is

all

(i.e..

the trials are

name

reinforced by the environment

seeing a conflicting stimulus on every therefore performance

is

sustained activation of a tion. In contrast,

tion

when

of congruent

less

the ink (i.e..

trial),

by

and

dependent on the

PFC

goal representa-

there are a high propor-

trials,

respondent can

the

temporarily forget the task goal and "get

away

with" responding on the basis of habit

(i.e.,

reading the word). ily forgotten,

If the task

and the

PFC

goal

is

temporar-

representation

is

-

1

back or n +

more

trials are

et al.

et al..

1

match

difficult to reject

2003).

(2003) reasoned that cognitive

and therefore prefrontal

activity, will

because there

is

a potential conflict between

and the task goal. They therefore

familiarity

examined the relationship between PFC and general

for lure trials

was a

there

activity

fluid ability. Indeed,

significant correlation

(r

=

ciated

with

lure

trials

and Raven's scores.

Importantly, this relationship

when PFC trials

was

was not diminished

activation associated with nonlure

partialled out.

Note how

this project

introduced both experimental control vs.

nonlure

(i.e.,

trials)

may

of individual

as well as statistical control

partialling out nonlure activity).

Genetic Mechanisms

no

differences

It

has been clear for quite

some time now

to

variation

in

to

80%;

50%

see Plomin

(estimates range from

&

that

g,

is

40%

Spinath, 2002). In other

PFC

words, genetic variation accounts for approxi-

behavioral

mately half of the variance in general intelligence

intelligence suggests that the

contribute

lure

(i.e.,

approximately

study

.54)

between the event-related signal change asso-

incorrectly to a conflicting stimulus.

recent

be

particularly important for lure trial performance

the heritability of general intelligence, or

A

it

back, and

longer active, the individual will then respond

in general

the

matches a recently presented stimulus but not

Gray

Johnson,

in a

B-R-X-M-B,

considered a "lure" because

is

than nonlure foils (Gray

Duncan. Emslie, Williams.

more

foils are

For example,

three-back task, in the stream

uation referred to as "goal-neglect"), then an

1995:

some

difficult to reject than others.

second

B

a match (target),

is

(foil).

Within the /!-back

task. If the goal is temporarily "forgotten" (a sit-

Emslie.

and

three-back task, the second

B-R-X-B-M

nonmatch

a

is

in a

indeed, these

&

letters),

must indicate whether the cur-

sented and must be sustained throughout the

incorrect response occurs (see

words,

(e.g.,

stream (n typically varies from one to

three).

therefore

in the

matches the one presented /7-back

rent stimulus

and the hippocampus

(and

outside the

In the A2-back task, a series

presented

is

therefore attention), motor systems (and there-

example of how

is,

then performed the n-

measure of working memory,

task, a

fMRI environment.

in the

fore response selection),

magnetic resonance

imaging (fMRI) measures. Subjects performed

pervasive, moderating sensory systems (and

is

assessed using

activity,

the Raven's task offline, that

other brain regions.

This top-down biasing

PFC

sive Matrices,

event-related

58

HANDBOOK OF UNDERSTANDING AND MEASURING INTELLIGENCE

.

An

population of healthy adults.

in the

new development

exciting

Conclusion

in the field of behavioral genet-

ics is multivariate genetic analysis

examines the extent

to

(MGA), which

which the variance

that

been a great deal of

In conclusion, there has

work aimed

determining

at

the

cognitive

determines heritability of one cognitive task cor-

and biological mechanisms

relates with the variance that determines the her-

performance on cognitive tasks and

itability is

of another cognitive task. For example,

it

two

possible for the heritability estimate for

understand

the

to better

between cognitive

relation

mechanisms and

that contribute to

This chapter

intelligence.

tasks to be equivalent, but that says nothing of the

reviews some general strategies that should be

"genetic correlation" between the two tasks. That

adopted

is,

the genetic basis for one heritability estimate

may be completely independent

of the genetic

which

basis for another heritability estimate (in

one

case, the variance that determines

heritability

in this endeavor.

research

It

also calls for future

both experimental and

exerts

that

and for future research

statistical control

that

examines not only cognitive mechanisms but neurological and genetic mechanisms as well.

estimate will be unrelated to the variance that

determines the other heritability estimate). In

MGA,

between two tasks

the genetic correlation

can vary from

MGA,

to

&

see Plomin

Spinath. 2002).

examine

therefore possible to

It is

correlation

among measures

Ackerman. the genetic

of cognition and

between measures of cognition and

intelligence.

For example, one could examine the genetic correlation Similarly, lation

among

WMC.

measures of

different

one could examine the genetic corre-

between measures of

intelligence.

WMC and measures no such studies

Unfortunately,

have been conducted.

However,

it

are

higher

&

(Plomin

when

the

task

is

tests

more complex

Spinath. 2002). For instance, the

heritability estimates for a Sternberg

scanning task were 0%, 35%, and sizes

that heri-

on elementary cognitive

tability estimates

of one.

and

three,

five,

memory-

47%

for set

respectively.

Similarly, the heritability estimate for a Posner

was 24% when matching was 60% when based on their name identity,

letter-matching task

physically identical letters but

matching

letters

despite the letters' physical difference. Interestingly, a parallel result

that

the

from psychometrics

tasks and general intelligence task

is

more a

is

higher

more complex (Ackerman.

Perdue, 2002). task

It

is

is

between so-called speed

correlation

when

the

Beier,

&

therefore possible that the

demands executive

attention, the

greater the heritability estimate, and the greater

between the task and

the correlation

remains to be seen with complexity

is

is if

g.

What

the rise in heritability

related to the rise in correla-

tion with complexity.

M.

Beier,

P. L.,

E.,

& Perdue, M. B. (2002).

Individual differences in working a nomological tual

speed

memory

within

network of cognitive and percep-

abilities.

Journal of Experimental

Psycholo}iy: General, 131. 567-589.

Anderson,

R..

J.

&

Lebiere, C. (1998). The atomic

components of thought. Mahwah, NJ: Lawrence Erlbaum. Cantor,

J..

&

Engle. R. W. (1993).

capacity as long-term

worth noting here

is

References

(for a detailed description of

1

individual-differences

Working-memory

memory

activation:

An

Journal of

approach.

Experimental Psychology. Learning, Memory,

and Cognition, Conway. A. R.

Cowan.

A..

Therriault. D.,

1101-1114.

19,

&

S. (2002).

variable analysis of working short term

memory

and general

Bunting.

N..

Minkoff,

memory

M. F,

A

latent

capacity,

capacity, processing speed,

fluid intelligence. Intelligence, 30,

163-183.

Conway. A. R.

A..

memory and inhibition

&

W.

Engle, R.

retrieval:

model.

A

(1994).

Working

resource-dependent

Journal of Experimental

Psychology: General, 123. 354-373.

Cowan. N. (1988). Evolving conceptions of memory storage, selective attention, straints within the

and

their

mutual con-

human information processing

system. Psychological Bulletin. 104, 163-191.

Cowan, N. (1995). Attention and memory: An integrated framework. Oxford, UK: Oxford University Press.

Cronbach. L. tific

J.

(1957).

The two

disciplines of scien-

psychology. American

671-684.

Psychologist,

12,

Cognitive

&

Daneman. M..

Carpenter,

dift'erences

A. (1980). Individual

P.

working memory and reading.

in

Journal of Verbal Learning and Verbal Behavior.

J.

I.

(2001).

Human

intelligence differences:

Towards a combined experimental-differential Trends

approach.

Cognitive Sciences,

in

5,

Dennett, D. C. (1995). Danvin's dangerous idea:

Evolution and the meaning of

life.

New

York:

Duncan,

J..

Burgess,

intelligence

&

P.,

Emslie, H. (1995). Fluid

lobe

frontal

after

lesions.

Neuropsychologia, 33, 261-268.

Duncan,

J.,

Emslie, H., Williams,

Freer, C.

P.,

(1996). Intelligence

and the

&

frontal

Cognitive Psychology, 30, 257-303.

Engle, R. W., Cantor,

&

J.,

Carullo.

A

18,

memory and

972-992. J.

E..

&

short-

general fluid intelligence:

P. S.

A

Journal of Experimental

477^85.

14,

R., Chabris. C. F.,

&

Braver, T. S. (2003).

Neural mechanisms of general fluid intelligence.

Nature Neuroscience,

Hambrick, D.

Z.,

&

6,

Engle, R.

W.

(2002). Effects of

and age on cognitive performance: of the

An

capacity. investiga-

knowledge-is-power hypothesis.

Cognitive Psychology, 44, 339-384.

Resnick (Ed.), The nature of intelligence

(pp.

Kane, M.

237-260). Hillsdale, NJ: Lawrence Erlbaum. J.,

Engle,

Bleckley, R.

&

J.,

memory

Engle,

M.

K.,

W. (2001).

Conway. A. R.

A

637-671.

9,

W. (2003). Working-

R.

capacity and the control of attention:

and task

petition,

set

neglect, response

com-

Stroop interference.

to

Journal of Experimental Psychology: General,

Kane, M.

Hambrick, D.

J.,

Wilhelm, O., Payne,

T.

Z.,

Tuholski.

W.,

&

S.

Engle, R.

W..

W.

(2004). The generality of working memory

A

latent-variable approach to verbal

133, 189-217.

Lachman,

Lachman.

R.,

An

processing:

L.,

J.

&

Butterfield, E. C.

and information

introduction.

A..

&

controUed-attention

Hillsdale,

NJ:

Lawrence Erlbaum. Miller, E. K.,

&

Cohen,

J.

D. (2001).

An

integrative

of prefrontal cortex function. Annual

theory

&

Miyake, A.,

Shah,

P

24, 167-202.

(1999).

Models of working

memoiy: Mechanisms of active maintenance and executive control. New York: Cambridge University Press. J.

S.

term:

(2002).

The case

Remembering over against

The blank

Pinker, S. (2002).

of human nature.

&

slate:

New York:

Spinath,

F.

M.

general cognitive ability Sciences,

Turner.

M.

L.,

6,

the short-

model.

standard

the

Annual Review of Psychology,

Plomin, R.,

Hunt. E. (1976). Varieties of cognitive power. In L.

Kane, M.

Naime,

316-322.

domain knowledge, working memory

tion

individual-differences perspective. Psyclumo-

Review of Neuroscience,

(1995). Cellular basis of work-

memory. Neuron,

J.

An

(1979). Cognitive psychology

Psychology: General, 128, 309-331.

Goldman-Rakic,

role

capacity,

Journal of Experimental Psychology: General,

Conway, A. R. A. (1999). Working memory,

latent variable approach.

The

(2002).

working-memory

memory and

Engle, R. W., Tuholski, S. W., Laughlin,

Gray,

W.

R.

in

executive attention, and general fluid intelligence:

capacity:

of four hypotheses.

test

Memory, and Cognition,

ing

Engle,

and visuo-spatial memory span and reasoning.

J.

Journal of Experimental Psychology: Learning,

term

&

J.,

of prefrontal cortex

(1992).

J.

Individual differences in working

comprehension:

130,

132, 47-70.

Johnson. R.,

lobe: The organization of goal-directed behavior.

59

169-183.

The contributions of goal

Allen Lane.

•

of Experimental Psychology: General,

mic Bulletin and Review,

164-170.

Intelligence

view of working-memory capacity. Journal

Kane, M.

450-466.

19,

Deary.

Mechanisms Underlying

53, 53-81.

The modern denial Viking.

(2002). Genetics and

(g).

Trends

W.

(1989).

in

Cognitive

169-176.

&

Engle, R.

Is

working

memory capacity task dependent? Journal of Memory and Language, 28, 127-154.

5 Working Memory

Capacity,

Attention Control,

AND Fluid Intelligence Richard

P.

Heitz

Nash Unsworth Randall W. Engle

Or take the power of attention. Is this wholly, or partly, or not at gence? All three views are widely held in the literature.

all the

same as

intelli-

Spearman (1927,

The

idea that attention

intelligence

length. tion

is

not

is

important to

novel.

Indeed,

Spearman (1927) discussed this issue at However, the relationship between atten-

and intelligence was contemplated even

earlier,

with some of the

first

empirical evidence

provided by Burt (1909). Binet (1903), the father

of intelligence testing,

AUTHORS' NOTE:

also

recognized the

p. 13)

importance of attention to general intelligence (Sternberg, 1982). William too,

attend to the

James (1890/1981),

wrote that "the number of things is

we may

altogether indefinite, depending on

power of

the individual intellect" (p. 405).

However, despite nearly a century of research, the

above question

of mental

ability,

still

pervades the science

and Spearman's comment

This research was supported by Grants F49620-97-1 and F49620-93- 1-0336 from the Air

Force Office of Scientific Research. Correspondence concerning

this

Atlanta,

GA

R NW.

chapter should be addressed to Richard

Heitz or Randall W. Engle, School of Psychology, Georgia Institute of Technology, 654 Cherry Street.

30332. E-mail: [email protected], [email protected], or randall. engle

@psych. gatech.edu. 61

62

HANDBOOK OF UNDERSTANDING AND MEASURING INTELLIGENCE

.

appears to be as tme today as

it was in 1927. However, converging evidence from both exper-

and factor-analytic methodologies

imental

intelligence

—

is

fluid

the ability to reason abstractly in

difficulty

studying the complex

in

attention-intelligence relationship

is

one

partly

of semantics. Both constructs are latent factors,

elude direct observation and must

in that they

Thus, one must cau-

be measured

indirectly.

tiously select

measurement

Because

and intelligence

of attention

theory

most

tools that are

sensitive to the constructs of interest.

one's

guides the selection of measures, observed cor-

depend on one's under-

relations will partially

lying theory. For this reason,

we

take particular

care in defining our conception of attention and intelligence. In the sections that follow,

and

just that,

theory,

attention

is

we do

related

one's

to

potential

for

There are three general types of theories the psychometric study of intelligence.

The

idea of "positive manifold"

observation that individuals

one intelligence intelligence

as the

cognitive

tasks;

test

tests.

But

structure).

extraction of ^ it

is

of

need not be the case. The

this is

just as

statistical

cal

methodologies, one can support either

Spearman's g factor or Thurstone's specific intelligences. It is important, therefore, to have

sound theoretical grounds for proposing the existence of

g,

Gf, Gc, or any other manifes-

tation of intellectual factors.

who

in

first

—

Measuring Intelligence AND Attention

the

score high on

Exactly what constitutes a good measure of been,

specific

intelligences

others hold a

Still

(e.g.,

somewhat

Sternberg. 1985).

One

such view that has become widely accepted

model of

developers

is

is

Cattell (1943, 1963).

that

g

is

actually

com-

it

who

validate their

one's ability to reason abstractly and perform

well in novel environments, whereas

Gc

corre-

and knowledge accumulated

over time, as might be measured by tests of fac-

knowledge. Evidence

performance

tasks (e.g.. Engle.

leads

us

to

Gf

consistently

in higher-order cognitive

Kane,

believe

that

that

&

Tuholski,

attention

1999)

exerts

its

is

This,

what

much

of

some form.

In

This circular logic appears in

the current literature, at least in

examining the relationship between working

memory (WM) and

intelligence

experimental research),

it is

not

(at

least

uncommon

in

that

researchers use a single intelligence measure

Raven Progressive Matrices) as the cri(e.g.. Brewin & Beaton, 2002). This begs

(e.g., the

the following question:

reflects

tests.

has been measured by past intelligence tests"

terion

Gf

admonishes measure by

with other intelligence

(Gc) and general-fluid (Gf) intelligence (see 1993). Accordingly,

earlier

in

he says, "takes the view that intelligence

posed of two components: general-crystallized

to abilifies

particularly

intelligence research. Cattell (1943)

of

Cattell's viewpoint

has

intelligence

research, rather post hoc. Criticizing practices in

test

intermediate view

relates to

a part of theory as

tend to score high on other

(Thurstone, 1938).

tual

much

methodology. For example,

Cattell (1943) notes that, using different statisti-

(p. 159).

sponds

generally

are

This idea, originated by

variety

also Carroll,

factors

measures showing the highest loadings (simple

proposing no single psychometric factor but

the hierarchical

lesser

manner consistent with those

interpreted in a

correlating

a

principal

first

Spearman (1904, 1927), suggests that there is a single intellectual ability on which people differ. At the other end of the spectrum are theories instead

factor-

in

Commonly, g emerges component or the highest

analytic methodology.

that one's ability to control

abstract reasoning.

stresses the

also disagreement as to

is

what constitutes psychometric g

our guiding

interest of

the

in

we show

our discussion

limit

order factor extracted from a battery of broad

novel environments and situations.

The

on Gf, and we

This aside, there

is

beginning to support the notion that attention

one of the major determinants of general

strongest effect

to this construct.

What

is

special about

Raven test that other measures lack? The answer to this question lies in the fact that the measure consistently loads highly on g and the

appears to be quite sensitive to individual differ-

ences (Marshalek. Lohman.

Although

this

is

appropriate

&

Snow. 1983).

for

small-scale

experimental research, more sensitive involve the extraction of a this reason,

we have used

Gf the

tests

latent factor.

Raven

test

For

alone

— Working Memory Capacity, Attention Control, and Fluid Intelligence in

experimental research and. in addition, have

included multiple measures of

Gf

in

our factor-

A

similar

problem

exists in defining atten-

William James's (1890/1981) adage

tion.

"everyone knows what attention true

on the

is" is certainly

level of introspection, but deriving a

on proves

definition that all can agree

Some

that

difficult.

only to main-

attest that attention serves

tain information through activation (Anderson.

Reder,

&

Lebiere. 1996).

some renounce

&

(Hasher

tion in favor of inhibition

&

1977).

Shiffrin.

We

ciation

more

between controlled and automatic pro-

cessing, with the former being effortful and

some type of (attentional) capacity And the manner in which control

subject to limitation.

operates

due

likely

is

to

specialized mecha-

Our

nisms, such as activation and inhibition. stance, then, untary,

is

that attentional control is a vol-

cognitive

effortful

act

serves to

that

maintain information through activation of relevant brain

circuitr\'.

task

ilance

is

respond

to quickly

to the

The attention required in a somewhat different than

vigthat

required on, say, a divided-attention task, in

which participants must perform two tasks simultaneously. is

another variety of attention

Still

manifest in selective-attention tasks, in

which

channels, of information, and they must select

a disso-

is

onset of targets.

participants are presented multiple sources, or

take a

comprehensive view. Certainly, there

are able to maintain attention to the task at

hand are more able

Zacks,

attention as simply controlled cognifion (e.g.,

Schneider

Those

activa-

and others remain agnostic, treating

1988),

the

in

service of detecting an infrequent target.

who

analytic studies, reported below.

must maintain attention

participants

63

•

and

inhibit the irrelevant

on us

one and disregard another. Although

it

might be

argued that attention serves different roles vigilance,

we would

attention tasks,

common

and

divided-attention,

argue that what

to all is the control

in

selectiveis

of attenfion. In

other words, attention can be put to use in a variety of ways, but the act of voluntarily

employing

on capacity-limited exec-

attention always relies

we

utive control. For example,

find that both

selective-attention tasks, such as dichotic listen-

Cowan.

ing (Conway,

&

Bunting. 2001). and

divided-attention tasks (often a cognitive task

incorporating a secondary load; e.g.,

Kane

&

Engle. 2000) reliably differentiate individuals

at

rated high or low in attentional control ability,

any one time, and suppress prepotent response

but tasks that can be performed automatically do

distracting information that impinges

tendencies that are task irrelevant.

keep

tant to

different

in

from a

mind

is

It

impor-

that this is qualitatively

situation in

which attention

is

automatically captured, such as by a flashing light or

theory

is

an unexpected sound. Note that far reaching.

type of information that

due

to

By our definition, any may be lost over time

decay or interference must be maintained

through attentional control. This

form of a grocery fic,

or

this

it

may be

list

may

take the

while driving in busy

the flashing

is

cue" when

the

&

Conway,

&

Engle. 2001;

Baylis, 2001).

apparent that individuals

who

ability to use controlled attention

From

this,

it

differ in their

may be

disso-

ciated using a variety of paradigms. In other

memory

words, working

by our

capacity

(WMC)

logic, attention control are not task

and.

depen-

dent but are a factor in any task that requires attention control (Turner

One

traf-

a response production such as

"Look away from

not (Kane. Bleckley.

Tuholski, Engle.

activity

individuals

is

that

&

Engle, 1989).

does not differentiate

automatic attention capture

certainly a special case of attentional

phenomena.

flashing cue tends to reflexively capture atten-

Here, an innate mechanism directs attention,

tion or the control of a socially unacceptable

effortlessly

implicit

attitude

to

prevent

it

from being

expressed in publicly observable behavior.

Such a stance

is

consistent with a

so-called "varieties of attention."

a

number of researchers

(e.g.,

number of

As noted by Parasuraman.

lus

and without

volition, to

some stimu-

such as the sudden appearance of some

object in the periphery or any feature such as flicker that suggests

movement. This

is

certainly

an adaptive response, as suddenly appearing objects

may be hazardous

or helpful, predator

1998), attention serves different types of func-

or prey. Although automatic attentional capture

depending on task demands. One variety

does not require control, resisting attention cap-

tions

of attention comes into play during sustained

ture

attention tasks, such as vigilance. In such a task.

(Kane

by a

salient

et al..

event does require control

2001). That

is.

executive control

64 is

HANDBOOK OF UNDERSTANDING AND MEASURING INTELLIGENCE

.

important for the endogenous prevention of

exogenous capture from

both internal

and exter-

nal distractors.

To understand

we must

issue,

attention-intelligence

the

first

two constructs, as the

these factor

is

how we measure

explain

validity of a latent

only as good as the manifest variables

comprising

it.

thought to tap

WMC

reliably

ability to control attention.

Following

As we

apparent that

some way

We

and

we sig-

fluid

illustrate in this chapter,

it is

WMC and fluid intelligence are in

related but clearly not isomorphic.

submit that the relationship

ability to control attention

by both

tion tapped

gence

this,

show a

WMC

between

one's

reflect

describe macroanalytic studies that nificant relationship

review

suggesting that tasks

microanalytic research

intelligence.

we

In the next section,

is

due

to the

—an aspect of cogni-

WMC

and

fluid intelli-

we

consider such

tests. In a final section,

alternative explanations as processing speed

and

WMC

by the number of items complex span task, the scores on such measures reflect controlled-attention abilquantitatively

recalled on

ity

rather than the

mation

that

Accordingly, it}'

number of

can be held

in a

we propose

that

"units'"

of infor-

short-term store.

WMC

is

an abil-

which an individual

reflecting the extent to

is

able to control attention, particularly in situations

involving

from competing

interference

information, activated representations, or task

demands. Before citing evidence for the controlled attention

WMC.

view of

how

clear exactly

it

WMC

is

necessary to

is

make

assessed and the

rationale for this particular methodology. In line

with the view that

WM consists of both domain-

specific storage buffers

central executive central

control

decrease

in the

be maintained.

and a domain-general

mechanism, tasks

that load the

mechanism should lead to a amount of information that can In

other

words, including a

heavy processing component should tax the

task complexity.

attentional system to such an extent that to-be-

recalled information

Microanalytic Studies OF Working

Memory

tions are often

AND Controlled Attention

is

to this end.

And,

inter-

appear to capture

ability that is related to g.

Fogarty and

tory tasks requiring maintenance and/or manip-

is

clear in recent

(e.g..

1996). but

also evident in earlier models. In particular.

Atkinson and Shiffrin (1968) noted that infor-

mation held

in

the

short-term

store

(STS)

decays over time, unless control processes are invoked to refresh these traces using rehearsal. Thus, attention needs to be directed to information held online to

view

this

is

keep

it

active. Admittedly,

quite different

from theories of

immediate memory focusing on the number of units of information that can be maintained over

G. A. Miller, 1956). But, as

time

(e.g.,

tend.

WMC

se.

is

not about a

number

or an

we

con-

amount

Using extreme group methodology, we

have observed numerous dissociations between individuals rated high and low in

simply cannot be explained tion

lost

working memory system

the

models proposed by Baddeley

per

be

Dual-task situa-

Stankov (1982) presented individuals with audi-

an attentional construct. This

it is

employed

likely to

important to realize that the central execu-

component of

tive

stores.

estingly, dual-task situations

some It is

more

is

from domain-specific

in

WMC

that

terms of a limita-

based on a number of units. The

differences

ulation of tones or letters. In one condition, participants

another,

performed these tasks alone:

in

these tasks were performed as dual

They found that the tasks correlated more when they were components of a dual task than when they were performed alone, suggesting that some additional ability became necessary when performing the two tasks simultaneously that was not required by either task alone. In addition, Fogarty and Stankov examined the first principal component (representing all of the tasks). As Fogarty and Stankov argue, the nature of the tasks employed suggests that the first principal component can be considered a Gf factor. They found that dual tasks load higher on the first principal component than single tasks, suggesting that they were better meatasks.

strongly

sures of

Gf

In other words,

showed higher

competing tasks

positive manifold than

single

are intrinsically qualitative rather than quantita-

tasks (Stankov. 1983; but see Stankov. 1989).

we measure

Recent neuroimaging research strengthens the

tive.

In

other words,

although

Working Memory Capacity, Attention Control, and Fluid Intelligence

view

dual

that

measure an important

tasks

between

Shin (1995) had participants perform single or

that

in a

functional magentic resonance

imaging (fMRl) scanner. They found

marked increases

tasks elicit

between

link

both complex span and reading comprehension

performance are predicated on word knowledge,

and executive

prefrontal cortex

Kane

&

Engle, 2002, for a review;

see also Chapter 9, this volume) and (b) our

view

dual-task

that

situations

Known

heavy

place

demands on executive control. The first widely accepted test of constructed by

WMC

was

(1980).

as the reading span task (RSpan), it

it is

consists of both

storage and processing components. In the orig-

RSpan

were asked

task design, participants

to six sentences (each set size presented

three times), participants last

were cued

word of each. The point

at

then the

were

that if this

WMC-reading comprehension

should

but

words

high-frequency

with

disappear

words. They found that the

true,

correla-

tion should be strong with low-frequency

WMC-comprehen-

was somewhat stronger with low-frequency words but still very much significant with high-frequency words. From this, Engle et al. (1990) argued that although word knowledge does play some role in the

to recall the

which the

partic-

WM WM

span-comprehension correlation, a general capacity limitafion

was

the important predictor.

Dual tasks such as RSpan and OSpan appear

from

to read sets of sentences out loud. After

two

leading to a correlation. Engle, Nations, and

Cantor (1990) reasoned

sion correlation

Daneman and Carpenter

inherently a dual task in that

inal

individuals simply have larger lexi-

cons than others. This hypothesis argued that

that dual

did not. This would be expected given (a) the

control (see

WMC and reading comprehension was

some

in prefrontal cortex

whereas either task performed alone

activity,

65

Another possible explanation of the conelation

executive ability. D'Esposito. Detre. Alsop. and

dual tasks

•

to

measure an important aspect of real-world

high-level cognition, as they are found to correlate

with such tasks as language comprehension,

ipants could accurately recall the final words

following directions, vocabulary learning, rea-

two out of three times was taken as their WMC. was significantly correlated

soning,

with higher-level comprehension measures,

between dual-task

such as reading comprehension and the ability

such as simple digit or word span,

Critically.

WMC

to identify a

noun

referent for a later occurring

Daneman and Carpenter

pronoun.

Daneman

&

Tardif,

(1980, 1983;

1987) argued that

WMC

reflects processing efficiency involved in reading. Accordingly,

cient

those individuals with

reading comprehension

available

more

residual

WMC

to use for

words.

taining

the

to-be-recalled

against

the

processing efficiency

from a study using a

WMC

effi-

processes have

idea

task in

ticipants solved arithmetic strings

main-

Evidence

came

which par-

and remem-

complex

(Engle. Kane, et

beyond

ing

it

more

However,

tasks.

The

that

some

by mak-

WM.

individuals

are simply better at multitasking in dual-task

being able to strategically allocate

situations,

resources toward the

primary and secondary

components of the task

in

an effort to maximize

span scores. These same individuals

may

also be

complex cognitive WMC-higher-order cognition

strategic during

lation.

tion hypothesis,

tasks,

corre-

This stance, termed the strategic alloca-

was evaluated by Engle, Cantor,

and Carullo (1992). To evaluate

(1980). indicating that the correlation between

Engle

WMC

assess the

not depen-

task,

that dual

difficult to refresh traces in

higher with measures of comprehension than

is

is

by simple span

might also be

it

(OSpan)' score correlated just as well and often

and higher-order cognition

critical difference

and a single

hypothesis, occupies the central executive,

leading to a

by Daneman and Carpenter

The

situations

that required

and Engle (1989) found

those reported

1999).

extra processing load incurred by these tasks,

more

span

and note taking

learning,

tasks require additional processing above and

bered the word that followed each string. Turner that this operation

al..

et al.

this hypothesis,

used a moving window technique to

amount of processing given

to different

WMC task, with amount of pro-

dent on the similarity of the processing com-

segments of the

ponent with the task being predicted; thus, the

cessing operationally defined as viewing time

RSpan-comprehension correlation could not

(VT).

be due to efficiency of reading processes, as

allocation,

Daneman and Carpenter

ficant negative correlation

asserted.

If

high spans are more strategic in resource

one would expect

to

observe a signi-

between

VT

on the

66

HANDBOOK OF UNDERSTANDING AND MEASURING INTELLIGENCE

.

processing component and span score. The

Analogous data are provided by Kane

et al.

authors found that viewing times did not correlate

(2001), using the anti-saccade task. In these

with span scores, leaving the strategic allocation

studies, participants

hypothesis untenable.

saccade (an overt eye movement) toward (pro-

Tasks such as RSpan and

OSpan might

appear to reflect individual differences

number of items

that can be

in the

maintained

in

an

attention-demanding dual-task situation. However,

view advanced

is

matically

P,

(anti-saccade) a flashing

to identify a briefly presented

or R. Because exogenous cues auto-

capture

attention,

anti-saccade

the

number of elements

response as well as a controlled saccade in the opposite direction. There are two ways perfor-

WMC

work

is

that indi-

reflect a relatively

mance is hindered in the anti-saccade condition. The first is through interference: Participants

stable characteristic of an individual's ability

who

to control attention to maintain information in

prepotent response will be slow to

an active or easily accessible

rect

state in interfer-

are not adept at inhibiting the automatic

saccade

(requiring that one look

distracting information.

cue).

Evidence for

this

claim

vided by

Conway

found

is

One

(2001

et al.

in a

number

elegant test )

is

pro-

using the vener-

The second

is

ing left look right")

tend to

make

away from

through a loss

tenance. For example,

make

anti-saccade

the

in

ence-rich contexts as well as block irrelevant,

of different paradigms.

a

misguided. Again, the

in the present

vidual differences in

B,

letter:

make

condition required the inhibition of a prepotent

focusing on a limitation in the

can be maintained

away from

exogenous cue

to

WMC

evidence makes clear that a theory of

that

saccade) or

were required

if

the goal

is lost,

in

a cor-

condition a flashing

goal main-

("When

flash-

participants should

a rapid error by looking toward the

phenomenon." Moray's

flashing cue, with their behavior guided by the

(1959) dichotic listening paradigm, originally

prepotent response tendency to look toward a

used by Cherry (1953). provided evidence that

flashing stimulus. Following from our theory,

able

party

"cocktail

particularly salient information

tended channel (see

filter

Moray's to

is

also

from an unat-

but should differ in anti-saccade performance:

In

were asked

a relevant channel while

point, the participants

own name was

spo-

unattended message. Moray found

ken

in the

that

approximately

33%

of participants reported

name when

it

was presented

to the

Conway et al. reasoned that if working memory capacity is related to the ability to

ear.

control attention in the service of selecting

relevant input and inhibiting irrelevant input,

then those rated high in

WMC

likely to report hearing their

should be less

own name

in the

irrelevant channel during a dichotic listening

The results confirmed that high spans were indeed more proficient in inhibiting the irrelevant channel: Low-span individuals were much more likely (65%) to report hearing their own

task.

name

pro-saccade performance

als

1995).

ignoring information in an irrelevant channel. At

unattended

it

and low-span individu-

allowed through an attentional

classic study, participants

hearing their

that high-

Wood & Cowan,

shadow words from

some

was expected

than were high-span individuals (20%).

This suggests that high-span individuals were at blocking distractors from the message they were instructed to ignore, whereas low-span individuals were less able to block the

indeed better

distracting information.

should not differ

When

in

an anti-saccade

performed correctly,

is

low spans should be slower, indicating a culty in resolving conflict.

commit more

Low

diffi-

spans should also

errors, indicative of goal neglect

(De Jong. Berendsen,

&

when an

should be

error occurs,

ing that the behavior

it

Cools,

1999).

And

fast, indicat-

was indeed automatic.

These hypotheses were confirmed: High- and low-span participants did not differ

in

pro-

saccade performance

in

both

but

did

differ

latency and error rate for anti-saccade perfor-

mance. fast,

And when

an error did occur,

it

was very

indicafive of goal neglect.

Interestingly,

in

an additional experiment,

Unsworth, Schrock, and Engle (2003) showed

do emerge in a pro-saccade you force individuals to perform them in a controlled manner. Unsworth et al. intermixed pro- and anti-saccade trials within the same block. Each trial began with a symbol that span differences

condition

if

that designated

or anti-saccade.

whether the

Under

trial

required a pro-

this situation,

low spans

were both slower and more error prone on both pro- and anti-saccade trials. Unsworth et al.

Working Memory Capacity, Attention Control, and Fluid Intelligence argue that this manipulation placed a premium

participants differ

on actively maintaining the goal of the creating a situation in which both pro- and

conflict

saccade

WMC

if

WMC

task, they

should show greater differences in the Stroop

when incongruent

task

are

trials

relatively

infrequent.

This hypothesis was supported.

based on the maintenance of

When 75%

of the

to

difficult

some number of items with such is,

resolve

to

ability

their

in

and maintain the goal of the

reconcile a

would be very

It

anti-

must be performed with control.

trials

theory of

task,

67

•

a finding. That

were due

limitations

to a lack of

trials

were congruent, highs

and lows differed markedly not

in

accuracy rate but

RT, suggesting that the manipulation

in

it

should

increased goal neglect for low spans. However,

not play a role in the anti-saccade task.

As we

the Stroop task

available "slots" to hold information,

common mechanism

argue, the

anti-saccade

mance tion.

is

performance and

condition likely to favor high

WMC

as previously discussed.

perfor-

employ controlled

the ability to

atten-

appears that in the anti-saccade task,

It

span differences

response time (RT) are due

in

or no compatible

RT. That

is,

reinforced

orienting response and execute a planned sac-

trials,

in the

ability to exert tion, as

we

of an

the ink,

endogenous control over

atten-

to

define

it.

Also

in

controlled attention theory

of low

ity

is

when

the task goal

was continually

by presenting many incongruent

presumably because

to say the color it

of

took them longer

resolve the conflict between the response

name

agreement with the

tendencies to say the word and to say the

an apparent inabil-

of the color. These findings illustrate two impor-

maintain the

spans to actively

response production "Look away from the flashing cue." Loss of this production

may

be the

WM:

tant functions of

relevant information

through

anti-saccade condition.

(incongruent Stroop

These observations led Engle and colleagues propose a two-factor theory of executive

maintenance of goal-

(i.e.,

name

the color

and

not the word) and dealing with interference

locus for span differences in error rate in the

to

were few

there

WM span groups did not

low spans were slower

—aspects

opposite direction

trials,

WMC individuals,

When

differ in accuracy rate but did differ slightly in

to a differential ability to inhibit the reflexive

cade

also rich in interference, a

is

affecting both

that

of

inhibition

controlled

require

responses

prepotent

trials).

In

summary, tasks

attention

to

maintain

information, inhibit distractor information, or

to increa-

block prepotent response tendencies seem to

sed latency and goal neglect leads to fast errors.

dissociate individuals scoring low or high on

Lending further credence

dual tasks such as the

attention,

study by

whereby interference leads

to this theory

Kane and Engle (2003).

is

performed a color-word Stroop task under ferent

cates

(e.g.,

that

Logan

&

individuals

incongruent Stroop

Zbrodoff, 1979) indi-

make more

trials

them

when

relative to

there

is

errors

on

word red naming the ink

(saying the

printed in green ink instead of

color only)

dif-

Previous

proportions of congruency.

research

a

a small proportion of

congruent Stroop

interpretation for this observation

The

Participants

One

trials. is

that

repeated presentation of an incongruent

the trial

dissociations

OSpan

or RSpan.

between high and low

WMC individuals described above employ relatively low-level cognitive tasks.

However, these

differences are true for higher-level tasks as well.

Low

spans suffer more than high spans in

a proactive interference task (Kane

2000), in a degree-of-fan task (Cantor 1993;

Conway

&

& &

Engle, Engle,

Engle, 1994), and in a cate-

gory generation paradigm that requires suppression (Rosen

&

Engle, 1997; see also Rosen

&

Engle, 1998).

(such as the word blue presented in red ink) reinforces the goal of the task. Conversely,

incongruent lose

trials

are

infrequent,

when

participants

Motivation

maintenance of the response production

"Respond to the color, not the word" because most of the time, they can perform well by relying on relatively automatic word reading. The prediction, then, following from the antisaccade findings,

is

that if high-

and low-span

The prevalence and leaves

open the

direction of span differences

possibility

that

some

factor

other than differences in ability to control attention differentiates the is

often

the

case

two groups. Namely,

that

it

high-span individuals

68

HANDBOOK OF UNDERSTANDING AND MEASURING INTELLIGENCE

.

outperform low spans; thus, the suggestion that

some dimension such

the groups differ along

as

"motivation" cannot be ignored, though there is

good evidence against such

example, high- and low-span groups differ theoretically meaningful

They do not

for both groups. In other words, monetary incentive encouraged both groups to "try" harder.

The above experimental

For

view.

a

amounts

in

and predictable ways:

differ in a typical pro-saccade task,

make

studies

WMC.

strong case for the hypothesis that

a as

measured by dual-task situations such as OSpan and RSpan. reflect one's ability to control atten-

when

of a proactive

tion.

This capability

interference task (before interference has built

tion

and interference place a premium on

in visual search, or in the first trial

up),

as mentioned

above.

If

low spans were

simply unmotivated, one would not expect them to

perform as well as high spans

A thorough

test

in

any condition.

of this explanation was

made

maintaining

is

important

information,

particularly

ence. If ity to

we

are correct in

control attention

is

assuming

that the abil-

made up of

WM

latent factor

strongly related to a latent variable

while individuals found to be high and low

Gf

spans (measured using OSpan) performed the

analytic methodology.

to

task. Pupillary dilation has

the

important to Gf, then a

by Heitz, Schrock. Payne, and Engle (2003).

RSpan

in

presence of strong internal or external interfer-

They measured task-evoked pupil

dilations

distrac-

measures. To

test this,

tasks should be

we

made up of

turn to factor-

been shown

be sensitive to mental effort and processing

load (Hess

&

1966). Thus,

Kahneman

Polt, 1964: if

&

Beatty,

span groups do differ along a

Macroanalytic Studies of AND Fluid Intelligence

WMC

dimension such as motivation, one would expect larger

task-evoked pupillary responses from

high spans,

who may

simply expend more effort

during task performance.

Of

course,

could

it

also be that high spans are simply

more

information

Daneman

processors

(e.g..

efficient

&

Carpenter, 1980) than low spans, which would predict smaller pupil dilations from high spans.

Ahern and Beatty (1979) found

Interestingly, that this

was

true for high-

and low-intelligence

One could pose regarding

memory and

a

number of

intelligence

fluid

questions

initial

between working

relationship

the

microa-

that

nalytic research cannot address. First,

WMC

various

tasks

reflect

a

do the

common

con-

Second, does that construct show con-

struct?

struct validity in predicting

some other

criterion

behavior as well as discriminant validity? third,

what

is

the relationship

between

And

WMC

individuals (high intelligence exhibiting smaller

and other constructs such as short-term memory

pupil dilations), although there are problems

(STM)? We argued

with their methodology (see Heitz et

situations tap an attentional control ability that

Heitz

et al.

al.,

span individuals consistently recall more correctly than

more mental

2003).

(2003) found that although highletters

do low spans, they do not exert

effort. Specifically,

high spans did

is

important to higher level cognition. If

indeed true that

and beyond

STM

and

ing of the

letter,

better

fit

critically,

when

recalling the letters.-

pupillary dilation

was

or.

However,

it

is

WMC tasks tap an ability above by simple span tasks, model including separable

that required

a structural equation

not exhibit larger pupil dilations during encod-

while reading the sentence,

previously that dual-task

WM

constructs

to the data than a

should provide a

model consisting of

a single latent factor. In addition,

show good discriminant

these con-

if

validity, a

STM

sensitive to processing

structs

load in this study, as there

latent factor should not correlate significantly

in pupil size as set sizes

was a linear increase grew larger. In addition,

with a

baseline pupil size measurements, recorded at the beginning of each

trial,

related with the set size

suggesting that

previously

how

affected

were strongly cor-

on the previous

trial,

"hard" one had to work one's

pupil

size

later.

Gf factor,

but a

WM latent factor should.

Obviously, no cognitive measure pure,"

meaning

or only

WMC.

that

is

Quite to the contrary,

that both simple

"process

no task measures only

we

STM assert

span tasks and complex span

tasks reflect similar constructs but in different

simple span or

STM

Furthermore, a monetary incentive reliably

proportions. That

increased baseline

measure, by and large, domain-specific storage

pupil

size

but

in

equal

is.

tasks

Working Memory Capacity, Attention Control, and Fluid Intelligence

.64

.63^^

.79

RSPAN

>

4Q*

WM

-f

^

CSPAN

^^^^^.^

.6r>^

^\

R .81*

y^A X_y f

.74

>

y^

BSPAN

67*

.41

RAVEN

29*/

^

common)—

gF

.67

.74*

CATTELL

^85* .79*

.61

V-

FSPAND .71*^^

.70

* FSPANS

Note: The latent

WM

,'-'.12

^ Model From Engle, Tuholski,

Structural Equation

Figure 5.1

69

.^^^77*

OSPAN

.78

•

variable

still

correlates with the latent

Gf

(1999)

et al.

variable after partialling the \ariance due to

represent manifest, or obser\ed. variables; circles represent latent factors. Paths

marked with an

STM. Boxes

asterisk are significant at

.05. WM = working memory: STM = short-term memory: Gf = general fluid intelligence: OSPAN = operation span: RSPAN = reading span: CSPAN = counting span: BSPAN = backwards digit span: FSPAND = forward digit span dissimilar: FSPANS = forward digit span similar: RAVENS = Ravens Progressive Matrices: CATTELL = Cattell's Culture-Fair Test.

p
six-back flict

event-related

for

activity. In the a priori search, lure trial activity

parietal,

activity.

in

a

— LPFC,

lateral posterior

anterior

cerebellum

cingulate,

—and,

and

second, via

lure

and target

activity

LPFC

trials, this

correlation with brain

was much weaker. Despite and

parietal

the fact that

areas generally

their activity in the /7-back task

increased

compared

to

baseline, none of this sustained brain activity

correlated significantly with Gf, thus recalling the

analogy of distinguishing basic braking

158

.

HANDBOOK OF UNDERSTANDING AND MEASURING INTELLIGENCE ability.

Consistent with

some work from

the

neuropsychological

tra-

0.8

dition, in

which behav-

ioral deficits are linked

0)

O)

c

with particular patterns

to

of brain damage,

LPFC

areas appear to be par55

0.0

-

important to

ticularly

Gf

the perfonnance of In

tasks.

both

verbal

and nonverbal domains, inductive,

and

relational,

sometimes gF (Raven's

APM

score)

(and

transitive

deductive)

problems

reasoning

LPFC

elicit

special

Of

activity.

importance, no

other brain regions have

been identified

High gF 0.2

literature

-

O)

c

across

the

so

involved

consistently

0)

in

are

that

of

variety

the

(0

reasoning and stimulus

o 15

domains reviewed

c

O)

0.0

here.

This research area

is

growing and producing interesting

ble

and

replica-

However,

results.

-0.2

from the perspective of

LowgF

trying

—

1

2

1

4

3

1

1

6

5

TR =

2.36

inform

understanding of

8

7

gence, Delay (scan frames,

to

our

r-

1

s)

it

intelli-

time

is

to

move beyond examining brain activity during

Figure 9.5

Neuroimaging Data From Gray, Chabris, and Braver (2003)

individual

reasoning

The

multivariate

tasks. Note: (A) Correlation between Raven's Advanced Progressive Matrices

(RAPM)

score and

functional magnetic resonance imaging (fMRI) signal from lateral prefrontal cortex

during correctly rejected three-back lure

LPFC

trials.

during correctly rejected three-back lure

mechanisms from sources of braking In

sum, only

WM

Time course of fMRI

signal

struct

from the

some connection be made between activity

high- versus low-Gf participants.

and performance

variation.

processes that were tied

directly to interference control

by Gf-related brain

(B)

trials for

were supported

task and activity

and performance

recent study by

Gray

field activity.

Gf condemands that

nature of the

(LPFC)

et al.

in

one

in others.

A

(2003) moves the

forward by linking reasoning performance

and brain

activity across multiple tasks

and by

using marker tasks of theoretically important

Conclusion

constructs, such as

WM

and attention control.

Their important findings support the behavioral

Neuroimaging research has begun

to

reveal

the brain structures important to general fluid

associations structs

among

and suggest

WM. that

attention,

PFC

and Gf con-

activity related to

Neuroimaging of Reasoning and Intelligence is what binds these constructs work that continues to combine Future

Tetris.

together.

and whole-brain metabolism during

and technique

theory

science

promise

great

advancing our understanding of not

in

how minds work

only

holds

how

but also

they differ

1

59

Although the correlation between Raven's score

attention control

cognitive-behavioral, psychometric, and neuro-

•

mance showed

=

tice (naive r

perfor-

Tetris

the predicted reduction following prac-

=

.77, practiced r

Raven's scores

.09),

continued to predict Tetris performance (naive r

=

practiced r

.41

LPFC

and

),

=

.36,

activity during Tetris did

not change with practice, remaining substantial at both

from one another and why.

time periods.

No

were reported between

correlations

Raven's performance and regional brain activity

Notes

in

either naive or practiced performers; only naive-to-

practiced change measures were reported. In parallel with the figural task,

1.

Duncan

et al.

(2000) also tested participants with an odd-man-out Service

modification of the

Educational Testing

(ETS)

(Ekstrom, French, Harman,

letter series test

Dermen. 1976). Here, four were presented

sets of letter

and

in a row,

same

structed according to the

but one

all

sequences

was con-

abstract rule. Again,

a control version of the task served as a baseline, in

which the to-be-discovered

rule

was

relatively obvi-

ous (three sequences were alphabetical). In contrast to the figural task, the letter task yielded only left

lateralized

acuvity (10/46. 46, 47). Although the

verbal nature of the task ization,

was

it

may account

its

lateral-

also true that the verbal reasoning and

more

control tasks had

similar g factor loadings to

each other than did the spatial tasks

and so the control condition

What

is

(LPFC)

less appropriate

that both the

activity similar to that seen in the

Raven's Progressive Matrices

The

2.

first

I

test

(Gur

(RPM)

The

et al., 1994).

34 adults

in

an analogies

spatial resolution for the

regional cerebral blood flow (rCBF) measurements

was

poor, but one detector appears to have been

placed over

LPFC Area

9; in fact, this area,

with other prefrontal cortex (PFC) structures,

compared 3.

showed elevated

Berman (1999) so

it

is

4.

and posterior

activity during the task

rCBF

Kirkby, Van

Horn, Ellmore. and

did not report a relative measure of

like that

from Haier

unclear whether that finding Haier.

Siegel.

et al. (1988). is

also replicable.

Tang. Abel, and

(1992) tested participants

in

and

Buchsbaum

the Raven's

Advanced

Progressive Matrices test in a session outside the scan-

M. O.

Boyle,

mem-

Journal of Exper-

abilities.

Acuna. B. D., Eliassen, Sanes,

J.

J.

C, Donoghue,

N. (2002). Frontal and

P.,

J.

&

parietal lobe

acdvation during transidve inference in humans.

Cerebral Cortex,

12,

1312-1321.

Baddeley, A. D. (1986). Working memory. Oxford,

UK: Oxford

University Press.

D.,

&

In G. A.

Hitch.

G.

Bower

(Ed.),

New York: Academic Barbas, H.,

Working

(1974).

The psychology

of learning and motivation (Vol.

8, pp.

47-89).

Press.

& Mesulam, M. M. (1985). Cortical afferent

input to the principalis region of the rhesus 15,

619-637.

&

Bechara, A., Damasio, H., Tranel, D., S.

W.

Anderson,

(1998). Dissociation of working

memory

from decision making within the human prefrontal cortex. Journal

Berker, E.,

of Neuroscience,

& Smith, A.

and other factors

in

18,

428^37.

(1988). Diaschisis,

site,

dme,

Raven performance of adults

with focal cerebral lesions. International Journal

of Neuroscience, 38, 267-285.

Berman, K. F, lllowsky, B. P,

to a resting baseline.

Esposito,

regional

along

&

E.,

imental Psychology: General, 131, 567-589.

monkey. Neuroscience,

test.

imaging study of verbal reasoning

am aware of examined

that

and perceptual speed

memory.

is

M.

Beier,

L.,

ory within a nomological network of cognitive

may

inductive reasoning tasks evoked lateral prefrontal

cortex

P.

(2002). Individual differences in working

Baddeley, A.

in the letter task

however,

clear,

Ackerman,

(in a pilot study),

have been more demanding and a baseline.

for

References

&

&

Weinberger. D. R.

(1988). Physiological dysfunction of dorsolateral

prefrontal

cortex

in

schizophrenia:

IV.

Further evidence for regional and behavioral specificity.

Archives of General Psychiatry, 45,

616-622. Braver, T. S., Cohen,

Smith, E.

E..

&

J.

D.. Nystrom, L. E., Jonides,

Noll, D. C. (1997).

A

ner,

and these participants were given extensive prac-

study of prefrontal cortex involvement

tice

on a computerized visuospatial-motor game called

working memory. Neurolmage,

5,

J.,

parametric in

49-62.

human

160

.

HANDBOOK OF UNDERSTANDING AND MEASURING INTELLIGENCE M., Colombo, N.. Delia Sala,

Brazzelli,

&

S.,

Spinnler, H. (1994). Spared and impaired cogni-

damage.

frontal

bilateral

after

abilities

tive

P.

M.

A., Just,

one intelligence

& Shell, P. (1990). What

A.,

test

A

measures:

account of the processing

in the

theoretical

Raven Progres-

Psychological Review, 97,

sive Matrices test.

404-^31.

&

Goldberg.

(1982).

J.

Operational efficiency and the growth of short-

memory

span. Journal of Experimental Child

Psychology, 33, 386-404.

&

Chorover, S. L, tion

(1966). Delayed alterna-

in

patients

with

cerebral

K.,

J.

Rostrolateral

Dorfman,

Holyoak, K.

Zhao,

J..

Z.,

(2001).

et al.

J.,

cortex involvement in

prefrontal

relational integration during reasoning.

Neurolm-

age, 14. 1136-1149. J.

B.

D.,

J.,

Forman,

S.

D., Braver, T. S., Casey,

&

D.,

D.

Noli,

C.

(1994). Activation of the prefrontal cortex in a

nonspatial working

memory

task with functional

MRI. Human Brain Mapping, A.,

&

Bulletin

A

meta-

Review,

3,

Cowan,

N.,

293-304.

1,

&

The cogni-

(Ed.),

MIT

Press.

&

Burgess, P,

J.,

Emslie. H. (1995). Fluid

intelligence after frontal lobe lesions. Neuropsy-

chologia, 33, 261-268.

Duncan,

Emslie, H., Williams,

J.,

The organization of

lobe:

Johnson, R.,

P.,

&

(1996). Intelligence and the frontal

goal-directed behav-

Cognitive Psychology, 30, 257-303.

ior.

J.,

Seitz, R.

Ahmed,

Kolodny,

J.,

A., et

Bor, D., Herzog, H.,

J.,

(2000).

al.

A

neural basis for

general intelligence. Science, 289, 457-460.

Duncan.

Williams.

J..

Brown,

(1993).

I.

Nimmo-Smith,

P..

The

&

I.,

control of skilled behav-

Learning, intelligence, and distraction. In

Meyer

D. E.

&

Komblum

S.

(Eds.), Attention

and performance XIV: Synergies

in

psychology, artificial intelligence,

experimental

and

cognitive

MIT

neuroscience (pp. 323-341). Cambridge: Press.

M.

Bunting,

M. Gazzaniga

neurosciences (pp. 721-733). Cambridge:

tive

ior:

Servan-Schreiber,

Conway. A. R.

Working

(1996).

(1995). Attention, intelligence, and the

J.

frontal lobes. In

Duncan,

lesions. Neuropsychologia, 4, 1-7.

Christoff, K., Prabhakaran, V.,

Cohen,

Duncan,

Freer, C.

M.

Cole.

performance

Kroger,

Psychonomic

analysis.

Duncan.

Case, R., Kurland, M. D..

term

P M.

Merikle.

memory and language comprehension: 422^33.

Cortex, 30, 27-51.

Carpenter,

&

Daneman. M.,

F.

Dureman,

&

I.,

Salde, H. (1959). Psychometric

and

phenomenon

revisited:

experimental methods for the clinical evaluation

The importance of working memory

capacity.

of mental functioning. Stockholm: Almqvuist

(2001).

The

cocktail party

Psychonomic Bulletin and Review;

Conway, A. R.

Cowan,

A.,

Therriault, D.,

&

Minkoff,

variable analysis of

memory

short term

and general

N.,

8.

331-335.

Bunting,

S. (2002).

M.

A

and Wiksell. F.,

latent

referenced cognitive

tional Testing Service.

fluid intelligence. Intelligence, 30,

A.,

Kane, M.

J.,

&

Engle, R. W.

Working memory capacity and

its

rela-

tion to general intelligence. Trends in Cognitive 7.

D'Esposito. M.. Ballard, D.. Zarahn. E..

&

Aguirre.

role of prefrontal cortex in sen-

memory and motor preparation: An eventrelated fMRI study. Neurolmage, II, 400-408. sory

D'Esposito, M., Detre, Atlas, S.,

&

J.

A., Alsop, D.

C.

Shin. R. K..

Grossman. M. (1995). The neural

basis of the central executive system of working

memory. Nature, 378, 279-281.

Daneman, M.,

&

Carpenter,

differences in working

P.

S.

450-466.

E.

The

(1995).

An

W.

M.

H.,

&

of factor-

Princeton, NJ: Educa-

of working

role

capacity and general control processes

169-189.

in intelligence. Intelligence, 20,

Engle. R.

Working memory and

(1996).

approach.

inhibition-resource

memory and

reading.

&

E. R. Stoltzfus,

R.

T

memory and human

New York: Engle, R.

W.

ity? In

In

Zacks

Oxford University

(2001).

What

(Eds.).

retrieval: J.

T.

E.

is

H. L. Roediger.

Working

(pp. 89-119).

cognition

Press.

working memory capacJ.

A. M. Suprenant (Eds.).

S.

Naime.

I.

Neath.

&

The nature of remem-

bering: Essays in honor of Robert G. (pp.

A. (1980). Individual

Journal of Verbal Learning and Verbal Behavior, 19,

Embretson,

tests.

kit

Richardson, R. W. Engle, L. Hasher, R. H. Logie,

547-552.

G. K. (2000). The

W., Harman,

J.

working memory capacity,

memory

Conway, A. R.

Sciences,

B.. French,

Dermen, D. (1976). Manual for

capacity, processing speed,

163-183.

(2003).

Ekstrom. R.

Crowder

297-314). Washington, DC: American

Psychological Association Press. Engle. R.

W.

(2002).

executive

Working memory capacity

attention.

Current

Directions

Psychological Science, 11. 19-23.

as in

Neuroimaging of Reasoning and Intelligence Kane. M.

Executive

(2004).

J.

working memory capacity, and

attention,

Ross

The psychology of learning and motiva-

New York: Academic

tion (pp. 145-199).

M.

Engle. R. W.. Kane.

&

J..

Individual differences in working ity

and w hat the\

W.

Tuholski. S.

Press.

(1999).

memory

capac-

us about controlled attention,

tell

general fluid intelligence and functions of the

&R

Miyake

prefrontal cortex. In A.

Shah

and

(Eds.),

NeuroReport.

reasoning.

inductive

Goel.

Gold. B.. Kapur,

v..

&

S..

Houle,

Journal of Cognitive Neuroscience.

ing.

P

Goldman-Rakic.

(1987). Circuitry of primate

S.

prefrontal cortex and regulation of behavior

memory.

representational

In

F.

Physiological Society.

University Press. E..

&

A

memory and general fluid

intelligence:

Journal of Experi-

latent variable approach.

mental Psychology: General, 128. 309-331.

P

Eslinger.

J..

&

disturbance of higher cognition after bilateral frontal lobe ablation: Patient

Neurology,

&

M..

S..

Van Horn.

Berman, K.

J.

D.. Ellmore.

Context-

(1999).

F.

dependent, neural system-specific neurophysio-

Mapping PET

logical concomitants of ageing:

correlates

during cognitive

activation.

Brain,

122, 963-979.

D.

&

London: Smith

&

vidual Dijferences,

Goel,

M.

J.

New

and Indi-

3,

The

(1980).

prefrontal

cortex.

C.

Frith.

gistic reasoning. v.,

C.

&

Dolan. R.

Neurolmage.

J.

(2000).

12.

504—514.

in

an inductive infer-

ence task. Journal of Cognitive Neuroscience,

Goel.

&

Dolan. R. (2001). Functional neuro-

anatomy of three-term

relational

reasoning.

tion

&

Dolan. R.

J.

of reasoning

v..

S.

M.. Skolnick. B.

L., et al. (1994). Lateral-

spatial tasks: Rela-

and Cog-

tionship with performance level. Brain

244-258.

nition, 24,

Haier, R. E.,

J.,

Siegel, B.

Wu.

C.

J.

V, Nuechterlein. K.

Pack.

J.,

et al.

H.. Hazlett.

(1988). Conical glu-

cose metabolic rate correlates of abstract reason-

Siegel.

J.,

Buchsbaum, M.

199-217.

12.

V. Tang. C. Abel.

B. S.

&

and

glucose metabolic

in regional cerebral

Hebb, D. O. (1945). Man's

L..

Intelligence

(1992).

415^26.

rate following learning. Intelligence. 16.

frontal lobes:

and

A

critical

Psychiatry,

54, 10-24.

& Penfield, W.

after extensive bilateral

(1940).

Human

behavior

removal from the frontal

and Psychiatry.

lobes. Archives of Neurology

Hitch, G.

Towse,

J.,

J.

limits children's retical

44.

N.,

& Hutton. U.

(2001).

What

working memory span? Theo-

accounts and applications for scholastic

ogy: General. 130. 184-198. Institute for Personality

Champaign,

Ability Testing.

belief.

Cognition,

87.

Jacobsen. C.

&

Houle.

An imagine

S.

( 1

997).

The

studv of deductive

F.

and Ability Testing. (1973).

intelligence with Culture Fair tests.

by

Gold. B.. Kapur. S.

seats of reason?

tomography. Intelligence,

(2003). Explaining modula-

B11-B22. Goel.

Muenz.

ized increases in cerebral blood flow during

Measuring

Neuropsychologia, 39. 901-909. Goel. v..

J..

development. Journal of Experimental Psychol-

110-119.

v..

316-322.

421-438.

& Dolan, R. (2000). Anatomical segregation

of component processes

12.

E., Jaggi,

Hebb, D. O.,

Dissociation of mechanisms underlying syllo-

Goel,

6.

Ragland. D.. Resnick.

review. Archives of Neurology

AQ1-A21.

York: Raven.

Buchel.

v.,

C.

changes

Putnam.

Stankov. L. (1982). Competing tasks

as an index of intelligence. Personality

Fuster.

Gur. R.

Haier. R.

Elder.

The functions of the brain (2nd

New York:

ed.).

Fogarty. G..

Braver. T. S. (2003).

ing and attention studied with positron emission

The functions of the brain.

(1876).

Ferrier, D. (1886).

&

F.

R.. Chabris. C.

J.

Neural mechanisms of general fluid intelligence.

performance of verbal and

Esposito, G.. Kirkby. B. T.

EVR.

1731-1741.

35.

to 17.

187-208. Gray.

Nature Neuroscience,

Damasio. A. R. (1985). Severe

MD: American

damage. Journal of Psychology.

frontal lobe

Conway. A. R. A. (1999). Working memory, short-term

by

(Ed.),

The mental changes due

Goldstein. K. (1944). J.

Plum

Handbook of physiology: The nervous system (Vol. 5. pp. 373-417). Bethesda.

New York: Cambridge

10,

293-302.

maintenance and executive control

Engle. R. W.. Tuholski. S. W.. Laughlin.

S. (1998).

Neuroanatomical correlates of human reason-

Models of working memory: Mechanisms of active (pp. 102-134).

8,

1305-1310.

a two-

factor theory of cognitive control. In B. (Ed.).

PFerrier.

L '

&

Engle, R. W..

161

IL:

Institute

for

Personality and

(1935). Functions of the frontal asso-

ciation area in primates. Archives of

and

P.sychialiy. 33.

558-569.

Neurology

162

.

HANDBOOK OF UNDERSTANDING AND MEASURING INTELLIGENCE

Jacobsen. C.

F.

1936). Studies of cerebral function in

(

primates:

I.

tion area in

The functions of the

monkeys. Comparative Psychology-

Monographs, Kane. M.

K., Conway, A. R. A.. & A controlled-attention view

M.

W. (2001).

Marshalek. B.. Lohman. D. P.

models of

archical

mental activation Stroke, 11.

&

mem-

Engle, R. W. (2000). Working

ory capacity, proactive interference, and divided

on long-term memor>

attention: Limits

retrieval.

Journal of Experimental Psychology: Learning,

Memory, and Cognition, Kane, M.

&

J.,

Engle, R.

cortex

frontal

W.

26.

333-358.

(2002).

The

role of pre-

capacity,

May, C. P, Hasher,

McFie,

J.,

Kane, M.

J.,

&

Engle. R.

631-61

Revie^^•, 9,

W. (2003). Working-memor\

The

capacity and the control of attention:

L.,

& Kane, M. J. (1999). The role

contri-

Memory and

lesions. Brain, 75,

292-311.

Mcintosh, A.

Grady. C.

R..

L.

B.

6,

Cohen,

J.

An

D. (2001).

integrative

and task

Review of Neuroscience. 24, 167-202.

Stroop interference. Journal of

Kroger.

K.. Sabb,

J.

F

Cohen, M.

S. Y.,

W.. Pales, C. L., Bookheimer, S.,

&

Holyoak, K.

J.

(2002).

Recruitment of anterior dorsolateral prefrontal cortex in

human

reasoning:

A

parametric study

of relational complexity. Cerebral Cortex. 12,

477-485. Kyllonen,

R

Miller,

E.

&

K..

Desimone, R. (1994).

information processing:

A

test

of the four-

Science, 263, 520-522. Miller,

E.

K.,

memory

&

Erickson. C. A..

mechanisms of

(1996). Neural

16,

& Fazio, F.

S..

in

Neuropsychologia, 36, 369-376.

& Christal,

R C. is

(little

more

R. E. (1990). Reasoning than)

working-memory

capacity?! Intelligence. 14. 389—433.

Larson, G. E., Haier, R.

J.,

LaCasse,

L.,

K.

(1995). Evaluation of a "mental effort" hypothesis for correlations

between

cortical

metabolism

New York:

Luria. A. R.

New

(

1966).

Oxford University

in

man.

Perret. E. (1974).

J.

The

11,

(2001).

left

New

brain systems

reasoning.

probabilistic

954-965.

left frontal

lobe of

man and

the

suppression of habitual responses in verbal cate-

Pierrot-Deseilligny.

D.. Stenger, V. A.,

&

1991

(

C. ).

Rivaud.

V.,

Smith.

Glover. G. H.,

S..

Gaymard.

B..

&

Cortical control of reflexive visu-

ally-guided saccades.

Prabhakaran,

York: Basic Books.

MacDonald, A. W.. lU. Cohen.

Osherson, D.

versus

deductive

Cerebral Cortex,

Agid. Y.

Press.

Higher cortical functions

&

M.,

L.

evidence for distinct right and

gorical behavior. Neuropsychologia, 12, 323-330.

and intelligence. Intelligence, 21, 267-278. Lezak, M. D. (1983). Neuropsychological assessment.

deductive versus probabalistic reasoning.

Parsons,

for

& Hazen.

Schnur. T.

(1998). Distinct brain loci

120. 375-105.

ability

working

5154-5167.

Osherson. D.. Perani. D.. Cappa. Grassi, F,

Desimone. R. visual

of the macaque.

in prefrontal cortex

sources model. Journal of General Psychology,

Kyllonen,

Parallel

neuronal mechanisms for short-term memory.

Journal of Neuroscience.

C. (1993). Aptitude testing inspired by

faces.

of prefrontal cortex function. Annual

theory

set to

V..

571-584.

butions of goal neglect, response competition,

Experimental Psychology: General, 132, 47-70.

J.

(1996).

working memor\' task for

Cerebral Cortex.

&

Haxby,

L..

Horwitz,

and prefrontal functional

limbic

in

Miller, E. K.,

&

G..

interactions in a

1

J.

& Piercy, M. F. (1953). Intellectual impair-

Changes

individual-differences perspective. Psycho-

Risberg,

Cognition, 27. 759-767.

An

and

7,

normo- and hypercapnia.

in

of interference in memor>' span.

Ungerieider,

Bulletin

hier-

342-347.

executive attention, and general fluid intelligence:

nomic

&

I.,

ment with localized cerebral

working-memory

in

and

intelligence. Intelligence,

Cerebral hemodynamic response to

(1980).

imental Psychology: General, 130, 169-183. J..

Snow. R. E. (1983).

in the radix

Maximilian, V. A., Prohovnik,

of working-memory capacity. Journal of Exper-

Kane, M.

&

The complexity continuum 107-127.

13. 1-68.

Bleckley,

J..

Engle. R.

frontal associa-

&

114. 1473-1485.

fira/>j, J.

A.

L..

Gabrieli.

Desmond, J.

J.

E..

D. E. (1997).

An fMRI

Carter, C. S. (2000). Dissociating the role of the

Neural substrates of fluid reasoning:

dorsolateral prefrontal and anterior cingulate cor-

study of neocortical activation during f)erfor-

tex in cognitive control. Science, 288, 1835-1838.

mance of

Malmo.

R. B. (1942). Intert'erence factors in delayed

response

in

monkeys

after

removal of frontal

lobes. Journal of Neurophysiology, 5, 295-308.

the Raven's Progressive Matrices test.

Journal of Cognitive Psychology. 33. 43-63. Rainer.

G.

Asaad. W. F.

&

Miller. E. K. (1998).

Selective representation of relevant information

L

Neuroimaging of Reasoning and Intelligence by neurons

in

primate prefrontal cortex.

the

Nature. 393, 577-579.

Raven,

J.

C.

Raven,

J.

H. (1998). Pro-

J.

UK: Oxford

gressive matrices. Oxford,

Psychol-

Risberg,

&

J.,

&

Ingvar, D. H. (1973). Patterns of acti-

Schulze, R.

Working-memory

(2002).

capacity explains reasoning ability bit

—and

a

little

more. Intelligence. 30, 261-288.

Sweeney,

ogists Press.

163

SUB, H.-M., Oberauer, K., Wittmann, W. W., Wilhelm, O.,

& Court.

E.,

•

M.

M.

A., Mintun,

J.

Brown, D.

B.,

Kwee.

A.,

Wiseman.

S.,

Rosenberg, D. R.,

L.,

et al.

vation in the grey matter of the dominant hemi-

(1996). Positron emission tomography study of

sphere during memorizing and reasoning. Brain.

voluntary saccadic eye

Risberg,

J.,

Maximilian. A.

Changes of

cortical

V..

& Prohovnik,

I.

patterns

activity

habituation to a reasoning

test.

(1977).

during

Neuropsycholo-

793-798.

gia, 15.

C.

Roberts. R. D.. Beh. H.

&

Stankov. L. (1988).

Hick's law, competing task performance and inlelVigence. Intelligence, 12.

Ruff, C.

C.

111-131.

Knauff. M.. Fangmeier.

(2003). Reasoning and working

mon and

T..

&

&

J.

memory: Com-

distinct neuronal processes.

Neuropsy-

Miyake. A. (1996). The separability of

and language processing: approach.

An

Journal

spatial thinking

individual differ-

of Experimental

Psychology: General, 125, 4—27. Shallice, T..

strategy

damage

&

Burgess,

application in

M.

L..

&

P W.

frontal

lobe

man. Brain, 114, 121-1 A\.

(1989). Is working

Waltz,

J.

A..

Knowlton, B.

K. B., Mishkin,

in

(1999).

human

F.

A

J.,

S.,

Holyoak, K.

J.,

Boone,

de Menezes Santos, M.,

system for relational reasoning

prefrontal

Psychological

cortex.

Science, 10. 119-125.

Weinstein,

S..

&

Teuber, H.-L. (1957). Effects of brain

injury

on intelligence

test

scores. Science, 125, 1036-1037.

Wharton, C. M., Grafrnan, K.. Brauner,

J.,

J.,

Flitman, S.

Marks. A.,

S..

Hansen, E.

et al. (2000).

neuroanatomical models of analogy:

A

Toward positron

dimensionality of a task. Intelligence. 16.

Cognitive Psychology, 40, 173-197.

Wittenbom,

J.

R. (1943). Factorial equations for tests

of attention. Psychometrika,

Spilsbury. G. (1992). Complexity as a reflection of the

31^5.

Stankov, L. (1983). Attention and intelligence. Jour-

nal of Educational Psychology, 75,

W.

emission tomography study of analogical mapping.

(1991). Deficits in

following

Engle. R.

memory capacity task dependent? Journal of Memory and Language, 28, 127-154.

penetrating

working memory resources for

ences

spatial

75, 454--*68.

Turner.

et al.

Spreer.

chologia. 41. 1241-1253.

Shah. P.

movements and

working memory. Journal of Neurophysiology,

737-756.

96,

471^90.

Zysset,

S., Miiller, K.,

Lohman,

8,

G.,

19-35.

&

von Cramon,

D. Y. (2001). Color- word matching Stroop

task:

Separating interference and response conflict.

Neurolmage,

13.

29-36.

10 Behavioral Genetics

AND Intelligence Stephen A. Petrill

For

decades, theories of intelligence and

cognitive development have

swung wildly

DZ)

(dizygotic,

on average.

share

50%

If identical

of the same genes,

twins are more similar

nativism, followed

than fraternal twins on a particular outcome,

by eras of strict environmentalism. Over the past

then genetic influences are assumed. Similarly,

30 years, a more balanced view has emerged

if

recognizing that both nature and nurture are

logical parents

between eras of

important.

strict

genetic

Quantitative

have

studies

suggested that genes account for a substantial

of the variance

portion

throughout the that the

life

in

cognitive

ability

span but have also suggested

environment

is

also essential to under-

adopted children are more similar to their bio-

and siblings than

also implicated. Heritability (h-) measures the

extent to which individual differences on an out-

come measure

are influenced by genetic differ-

ences in the sample. Shared environment

when

among

The purpose of this chapter is to examine the ways in which behavioral genetics methods may be used to better

on a particular outcome

understand intelligence and

found among family members

standing

intelligence.

its

development. In

implied

to

DZ

the correlation

twins or

when

is

who

the proportion of variance in an

and

gene-environment processes on intelligence.

The chapter explores behavioral

genetic

the implications of the

literature

with

respect

to

molecular genetics, theories of intelligence, and

enced by environments held or attending the

same

nonshared environment genetic

The the cal

factors

that

of the

same genes, whereas

MZ)

influ-

in the

same

school. Finally, the

(e-)

measures non-

make family members

heritability

of intelligence

most consistent tlndings literature,

one

that

in the

has

is

one of

psychologi-

been replicated

across twin/adoption designs, across countries,

and across ages (Bouchard Identical twins (monozygotic,

outcome

common among

different, including error.

the assessment of intelligence.

Univariate Genetic Analyses

in

family members, such as growing up

home

is

are biologi-

environment measures

cally unrelated. Shared

genetic analyses of cogni-

twins

a significant correlation

particular, the chapter addresses the following

tive ability, longitudinal genetic analyses,

(c") is

MZ

similar in magnitude

topics: univariate genetic analyses of cognitive ability, multivariate

their adoptive

parents and siblings, then genetic influences are

share

fraternal

100%

Chipuer, Rovine,

twins

across

all

&

&

McGue,

1981;

Plomin, 1990). Collapsing

available studies, the heritability of

165

li

166

HANDBOOK OF UNDERSTANDING AND MEASURING INTELLIGENCE

.

intelligence

around

is

intelligence but also for

(Plomin,

abilities

.50,

both for general

most

specific cognitive

DeFries,

intelligence

is

&

McClearn,

McGuffin, 2001). However, the

heritability

not stable at .50 across the

^

of life

span. Genetic influences increase in importance

with the age of the sample

when examining cog-

Verbal Ability

development. In general, the heritability

nitive

Twin

Spatial Ability in

in

Twin 2

1

20% in 80% in old

of intelligence increases, from around infancy to

60%

in early

age (Boomsma, Bernieri,

adulthood to

1993; McCartney, Harris,

McGue, Bouchard,

1990;

lacono,

& &

Cross-Twin Correlation

Figure 10.1

Lykken, 1993; Plomin, 1986; Plomin, Fulker, Corley,

&

DeFries, 1997). The heritability of

intelligence remains substantial in old-old

ples

(McClearn

sam-

et al., 1997). In contrast, these

same twin and adoption

studies suggest that

shared environmental influences are important

two

this case,

cence.

Nonshared environmental influences

skills) is

differences in intelligence

become

is

the cross-twin correlation (see Figure 10.1). In

childhood but reduce to zero by adoles-

individual

and envi-

into genetic

dence. The essence of multivariate genetics

in early

(including error) remain significant. In general,

outcomes

different

ronmental sources of covariance and indepen-

Tw in

1

score on Variable

's

compared

to

Twin

2's score

1

(verbal

on Vari-

able 2 (spatial skills). If identical twins are

more

similar than fraternal twins, then genetic

increasingly influenced by genes as the age of

influences are hypothesized to influence the

the sample increases. This increase in heritabil-

correlation between verbal and spatial skills. If

also found in specific cognitive abilities

identical twins are equally similar to fraternal

ity is

such as verbal,

spatial, perceptual

memory. This increase

in heritability is

indirect evidence for the

fially

active

and/or

speed, and

evocative

poten-

emergence of

gene-environment

processes. This possibility

will

be discussed

are implied. Figure 10.2 demonstrates a bivariate genetic

example of such an approach. In between verbal skills

this case, the correlation

and

spatial

genetic

in

later in the chapter.

shared environmental influences

then

twins,

skills

(A),

is

influenced

shared (C),

by overlap

and nonshared

The independence of measured by residual shared environmental (c), and non-

(E) sources of variance.

Multivariate Genetic Analyses

spatial

skills

genetic

(a),

is

also

shared environmental

The

described

analyses

univariate

above

To

(e)

sources of variance.

the extent that the cross-twin covariance in

identical twins

is

greater than fraternal twins,

long-debated question of whether genes or

A will be

To

the extent that the cross-twin

environment

covariance

are

important

because

are

they

important

differences in intelligence.

address

to

the

individual

The answer

to this

twins.

C

large. is

similar in identical and fraternal

will be large.

To

the extent that verbal

correlate within individual

question appears to be that both genes and

and

environments are important for individual

twins but not across twins,

dif-

spatial

skills

E

will be large.

To

ferences in intelligence. Muldvariate genetics

the extent that spatial skills are independent

research has taken the next step to examine

from verbal skills, a, c, and e will be large. Although more sophisticated models have been postulated, they are based on this general approach. These models have been employed

how genes and environments affect opment of cognitive skills. Where genetic analyses

decompose

the devel-

univariate

the variance of a

outcome into genetic and environmental components of variance, multivariate genetic methods decompose the correlation between single

to

examine several theoretically

central issues

to the understanding of intelligence.

be discussed

in turn.

These

will

I

Behavioral Genetics and Intelligence

1

•

1

67

.OMZ, 0.5DZ

1.0MZ, 0.5DZ

Figure 10.2

Bivariate Genetic

Model

Molarity Versus Modularity

(as

measured by Bayley scores) are correlated

with their biological parents' general cognitive

A

longstanding and continuing debate

intelligence literature

is

best conceived as a molar or (see Chapters 14

and

in the

whether intelligence

16, this

is

modular construct volume).

A

molar

ability but not their parents' specific cognitive

These

abilities.

results suggest that genetic influ-

ences are mediated by a general

factor.

Reznick,

Corley, and Robinson (1997) found that the

among

verbal and nonverbal

system implies that a singular unitary process

intercorrelations

or processes function across a wide variety of

tests

cognitive tasks. Conversely, a modular system

environmental factors. Similarly, Price, Eley,

involves independent and uncorrelated cognitive

processing

units.

If

different

dimensions of

cognitive ability, such as verbal ability, spatial ability,

speed of processing, and memory, are

were driven by both genetic and shared

Stevenson, and Plomin (2000) found substantial

shared

and genetic

environmental

overlap

between language and non-language-based cognitive ability. Petrill, Saudino, Wilkerson,

and

influenced by the same set of genes and envi-

Plomin (2001) conducted a multivariate genetic

ronments, then a molar explanation best

analysis on the individual cognitive tests that compose nonverbal intelligence using the same sample as Price et al. The results of this study

data. If different

fits

the

dimensions of cognitive pro-

cessing are each associated with independent sets

of genes and environments, then a modular

explanation best

fits

Multivariate genetic studies suggest that the

covariance

among

different cognitive skills

is

influenced largely by shared genes, whereas the

discrepancy between cognitive

suggest

that,

similar to Price et

shared environment tend to

the data.

skills

is

ties similar,

al.,

genes and

cognitive abili-

whereas the nonshared environment

(and error) contributes to the discrepancy cognitive

among

skills.

Many more

influ-

enced largely by the nonshared environment. Put

make

studies

examine the issue of

molarity/modularity in early and middle child-

another way. the influence of genes on cognitive

hood. Rice, Fulker. and DeFries (1986) con-

molar, but the influence of non-

ducted a multivariate genetic analysis of verbal,

skills is largely

shared environments

is

ExamThompson,

largely modular.

ining infancy and early childhood,

spatial, perceptual speed, in

4-year-old adoptees.

and memory

abilities

The data suggest a

strong

Plomin, and DeFries (1985) suggest that cogni-

general genetic intelligence factor and a weaker

and 24-month adoptive children

shared environmental factor. Looking at these

tive skills in 12-

168

.

HANDBOOK OF UNDERSTANDING AND MEASURING INTELLIGENCE

same adoptees

in

middle childhood, molarity

becomes even more pronounced Alarcon. Plomin. (

&

& DeFries.

influenced

and almost

Adult Development (Finkel. Pedersen. McGue.

of the covariance

among

specific cognitive

lar

&

McCleam.

1995). Finally. Petrill et

al.

(1998)

found even more pronounced general genetic

abilities.

Twin

was

abilities

results were obtained when comparing Swedish data to the Minnesota Twin Study of

1998; Cardon, Fulker,

for a large portion of the variance all

cognitive

Plomin, 1992). Genetic g accounts

Fulker, Corley,

DeFries,

among

primarily by the nonshared environment. Simi-

studies in early and middle childhood

Thompson

yield similar results. Luo. Petrill. and

(1994) examined the importance of genetic g in a sample of 6- to 13-year-old twins. In this study, the covariance

among

Wechsler

1 1

(WISC-R)

ligence Scale for Children-Revised subtests

Intel-

effects

when examining nondemented

80-i-year-

old twins recruited as part of the Octotwin

Study (McCleam

et al., 1997).

Genetic molarity

accounted for a large proportion of the variance, as well as almost in specific

all

total

of the covariance

cognitive abilities.

mazes) was examined.

(excluding

Results suggested that the cross-twin covariance

among

was

these subtests

substantial. In fact,

Longitudinal Genetic Studies

genetic effects could largely be explained by

a single genetic factor across

all

1

subtests.

1

Shared environmental influences contribute

to

In addition to

examining the relationship among

cognitive skills at a given time point, behavioral

molarity, and nonshared environment contributes

genetic research has also examined the influ-

and Fulker 1995)

ence of genes and environments on the develop-

to modularity. Casto. DeFries.

obtained similar results

(

when examining

ple of 7- to 15-year-old twins

Colorado Reading

the

childhood suggest that molarity

middle

may be found

standardized measures of scholastic perfor-

mance

as well.

These studies suggest

that the

correlation between reading performance and

general cognitive ability

is

influenced solely

by shared genetic factors (Brooks. Fulker. DeFries. 1990; Thompson. Dettennan.

1991; Wadsworth,

obtained

1994).

when examining

the

&

Plomin.

comorbidity disability.

than 507c of the observed comorbidity

between reading and mathematical

may

&

Similar effects are

between reading and mathematical

More

Chapter

12.

approach has already been described

One

chapter the

—examining

heritability

across the

life

in

this

age-related differences in

of general cognitive ability span. Although this approach

examines whether the magnitude of genetic and environmental effects varies with age,

this

approach does not provide any data on whether

same genetic and environmental variance is life span or whether new

the

operating across the

sources of genetic and environmental variance

"tum on" or "tum

off" at different ages. This

general issue has been analyzed in two ways. First,

research has examined stability

instability of intelligence

— whether

and

the genetic,

&

shared environmental, and nonshared environ-

Genetic molarity appears to become even

genetic, shared environmental, and nonshared

(Knopik, Alarcon.

DeFries, 1997; Light

mental influences

DeFries, 1995).

more pervasive in adulthood and old age. Tambs, Sundet, and Magnus (1984) suggest that

back

when

is

between the same

(WAIS) subtests old Norwegian

McCleam

in a

sample of 30-

Scale

to 57-year-

Pedersen. Plomin. and

twins.

(1994) found that most but not

all

of

the genetic variance in specific cognitive abilities

can be attributed to effects

were

zero,

g.

Shared environmen-

whereas the discrepancy

one age are related at

later ages.

to the

Going

to cross-twin correlations, genetic stability

among Wechsler Adult

Intelligence

at

environmental influences

genes account almost entirely for the similarity

tal

skills (also see

volume, for a detailed discussion of ways

this

disabilities

be attributable to shared genetic influences

&

ment of cognitive

of examining cognitive development).

Project.

In addition, behavioral genetic data in

in

a sam-

drawn from

inferred

the test

remain stable Genetic

when

cross-twin correlations

measured across time instability

is

inferred

genetic influences can be found at a single

points in time but cross-twin correlations are low. In other words, there are genetic influences at a

given time, but they do not persist across

time. Studies have suggested that the genetic

Behavioral Genetics and Intelligence

among

covariance

cognitive skills across age

becomes increasingly important. Although older research alluded to this issue (e.g.. Cherny et al., 1994; Fulker, Cherny, & Cardon, 1993). more recent work has examined this issue in longitwin (Bartels, Reitveld. van Baal.

tudinal

&

Boomsma, 2002) and twin/adoption (Bishop 2003:

et al..

Petrill et al.. in press) studies

suring cognitive

mea-

from early childhood

skills

•

1

69

Endophenotypes and General Intelligence The

genetic

quantitative

described

results

above are important because they more comprehensively specify the phenotypic relationships

among

and longitudinally mea-

concurrently

sured cognitive

However, these data, by

skills.

how genes

themselves, do not explain

To address

influ-

through adolescence. In general, these studies

ence cognitive

suggest that the cross-twin correlations between

researchers have taken reductionistic approaches

the

same

tests

measured over time are largely

stable. In other

among

to

ability.

examine the pathways from genes

words, the genetic covariance

and consistent

between endophenotypes or more basic cogni-

cognitive skills

is

stable

processes thought to mediate the relation-

tive

Shared environmental covariance across age

ship between genes

important

also

One

to g.

approach has been to examine the relationship

though much of childhood and adolescence.

is

this question,

childhood but

early

in

is

Chapter

skills (see

3,

and complex cognitive this volume, for a more

detailed discussion). Researchers have

ultimately nonsignificant by adolescence.

exam-

approach

ined the genetic and environmental covariance

examines the magnitude of the covariance of

between "elementary" processes, such as reac-

Whereas

the

stability/instability

and decision

genetic and environmental influences at differ-

tion time, stimulus discrimination,

examined change as a phenotype. This approach examines the genetic,

time, and general cognitive ability. Although

shared environmental, and nonshared environ-

genetic influences

mental influences on the rate and trajectory of

tant for the overlap

change

and psychometric

ent ages, others have

in cognitive abilities across age. In his

"developmental

classic

synchronies" paper,

Wilson (1983) suggested that the trajectory of developmental milestones (physical, cognitive, etc.) in identical

twins

is

more

similar than in

fraternal twins, suggesting genetic influences

growth. ies

More

have used

on

recently, behavioral genetic studlatent

growth curve and multilevel

modeling strategies

to

test

more

explicitly

genetic and environmental influences on change (e.g..

McArdle. 1986: McArdle. Prescott. Ham-

agami,

&

Horn. 1998:

2002; Neale Finkel. Gatz.

McGue &

Christensen,

& McArdle. 2000; Reynolds, & Pedersen. 2002). These studies

there are fewer studies, like psychometric tests,

become

increasingly impor-

between elementary tasks

tests

of intelligence.

ined

the

between

relationship

mentary" cognitive stimulus

ability (e.g., reaction time,

sample

of

found

that

between psychometric and

ele-

discrimination)

in

the covariance

studies

of adulthood,

shared environmental

influences are zero, but genes remain important to

understanding the relationship between ele-

mentary cognitive tasks and psychometric

& Ho.

Somsen, 1991: Ho. Baker.

McGue. Bouchard. Lykken.

ies

important to note that these stud-

have typically examined very young children

(e.g.,

McArdle.

McGue & 2002).

1986) or older adults

Christen.sen, 2002:

Reynolds

(e.g..

et al.,

The magnitude of genetic influences on

change has not been systematically examined intermediate ages. Thus,

it

is

in

unclear whether

common

genetic and shared environmental factors. In

nongenetic influences are implicated it is

al.

mentary measures was influenced by

primarily responsible for the intercept, whereas

However,

a

6- to 13-year-old twins. Petrill et

ligence (Baker. Vernon.

change.

Wechsler

1 1

subtests (excluding mazes) and 6 tests of "ele-

have suggested that genetic influences are

in

Petrill,

Luo, Thompson, and Detterman (1996) exam-

1991:

intel-

Boomsma &

& Decker. & Feuer.

1988; 1984:

& & Boomsma,

Neubauer, Spinath. Riemann, Angleitner, Borkenau, 2000; Rijsdijk, Vernon,

1998: Vernon. 1989). Other studies have begun

examine

activity,

the genetic and environmental between brain structure, brain and intelligence (see Boomsma.

Anokhin,

&

to

covariance

de Geus, 1997; Pennington

et al.,

Boomsma, & de Geus, de Geus, & Boomsma, 1998:

genetics are important to change in later child-

2000; Posthuma, Neale,

hood, adolescence, and early adulthood.

2001; van Baal,

f.ji

170

HANDBOOK OF UNDERSTANDING AND MEASURING INTELLIGENCE

.

van

this

Molenaar,

Beijsterveldt,

Boomsma,

&

Geus,

de

1998a, 1998b; see also Chapter

9,

volume).

In general, these studies have suggested sub-

on elementary cog-

stantial genetic influences

and brain

nitive tasks, brain structure,

activity.

However, the conclusions one can derive from analyses of these endophenotypes are unclear.

some measures of "elementary" show moderate correlations

In particular,

cognitive

ability

with general cognitive ability (Neubauer

2000;

1996). However,

Petrill et al.,

known how "elementary" in

comparison

and

structure

solving, but

et al., is

not

these tasks really are

psychometric measures of

to

measures of brain

Conversely,

intelligence.

it

show very

these studies

phenotypic correlations between these

small

10.3

Genes,

Environments,

and Their

Influence on Cognitive Skills

do not require problem

activity

many of

Figure

measures and general cognitive

ability

(e.g.,

two more independent Chomey, Lubinski, Thompson, &

since failed to replicate in

samples

(Hill,

Plomin, 2002).

& Eysenck, 1992; Haier, Siegel, Tang, Abel, & Buchsbaum, 1992; Matarazzo, 1992; McGarry-Roberts, Stelmack, & Campbell, 1992; Posthuma eta!., 2001; Reed & Jensen, 1992; Rijsdijk & Boomsma, 1997; Vernon &

ated with g

Mori, 1992).

2003, for a discussion). Although

Barrett

What

is

emerging

from

this

may number

important

genes associ-

literature is a realization that the

in the several

hundreds

and may have very small effect sizes

below

1%

(at

or

of the variance per gene) (see Plomin, this is sober-

ing from a reductionistic, mechanistic perspective, the possibility

Molecular Genetics and General Intelligence addition

In

studying endophenotypes,

to

others have attempted to identify genes associ-

ated with cognitive replicated

findings

There are two well-

skills.

—an

association

dementia and the apolipoprotein Farrer et

al.,

between

E gene

et al.,

2002).

DNA

with normal variation ability are also

D

in

(e.g.,

markers associated general

cognitive

beginning to emerge: cathepsin

(CTSD; Jacobs

et al.,

2002),

cholinergic

explain from a mechanistic perspective. Intelli-

gence appears

be "general" because of

to

genetic variance, especially after adolescence.

Figure

10.3

depicts the relationship between

two cognitive measures. There of overlap

among

is

a high degree

individual sources of genetic

variance but a low degree of overlap

among

individual sources of environmental variance.

Molecular genetics research suggests variance

genetic

is

that this

composed of numerous

genes each with a very small effect

size.

Thus,

2002), and a functional polymorphism

cognitive processing are likely to be influenced

receptor

Malhotra

et al.,

potential findings, that

explain

any two behaviorally measurable components of

2

(VAL158MET) in the cathechol 0-methyltransferase (COMT) enzyme system. (Egan et al., 2001;

may

(CHRM2; Comings

muscarinic et al.,

(e.g.,

1997) and linkage between markers

located near 6p21 and reading disability

Kaplan

of hundreds of small-effect

why g is so easy to find from a psychometric perspective but so hard to size genes

it

is

2002).

Despite these

important to remember

replication has been elusive for general

by a

common

set

of genes, leading to a genetic

correlation. This multicomponential

model of

general intelligence has been postulated previ-

ously

at

cognitive

the level of independent phenotypic

processes

(e.g.,

Detterman,

1987).

study found evidence for an association between

However, the independent components of general intelligence may be even more basic, at the

IGF2R and

level of

cognitive ability. For example, one longstanding

general cognitive ability in two

separate samples

(Chomey

et al.,

1998) but has

hundreds of individual genes interacting

with one another

in probabilistic

and pleitropic

k

— Behavioral Genetics and Intelligence

Experience

An

we may

Etiology

Environment as Experience Versus Environment as Etiology

Figure 10.4

ways.

optimist

may

reasonably argue that

eventually understand the mechanisms

by which genes yield

g.

A

may argue we may never be

pessimist

with equal reasonableness that

complex and

able to characterize the infinitely

pathways through which individual

varied

171

•

differences

in

genes

individual

influence

differences in intelligence.

environments (see Plomin, Asbury,

&

Dunn,

2001). This approach has led to mixed results

environment (Reiss, Neiderhiser,

family

for

Hetherington,

&

Plomin, 2000) and

much

con-

troversy about the efficacy of examining the

nonshared environment as a useful predictor of developmental outcomes

2001 Turkheimer

et al.,

;

Plomin, Asbury,

(e.g.,

& Waldron,

2000).

One

of the major obstacles using current methods that the

error).

important questions concerning the

mechanisms of how genes influence

intelli-

gence, quantitative genetic results have shown that

genes are increasingly important, not only

for individual

is

is

defined as

anything that makes twins different (including

Implications Despite

nonshared environment

measures of cognitive

also for the links

between cognitive

ability but

At the

skills.

Thus,

it is

unclear

ences between cognitive

skills

currently or longitudinally

is

environment or measurement less,

understanding

the differ-

measured con-

"true" nonshared error.

Neverthe-

how environmental

ences affect development

in

the

influ-

context

increasingly large and stable genetic factors

behavioral level, these results have important

one of the central issues

implications for developmental theory, particu-

genetic research.

how we

how much

of is

in current behavioral

conceptualize the environments

However, one of the major limitations of the

associated with general cognitive ability. For

current debate on the genetics and the role of

example, because shared environmental

larly in

vari-

the environment has been a misunderstanding of

some have

the distinction between the environment as expe-

argued that early experiences are ultimately

rience and the environment as etiology. Figure

unimportant to the study of individual differ-

10.4 presents

ance

ences

is

negligible by adolescence,

in

development

Harris,

(e.g.,

Others have argued that what

more systematic examination of

is

1998).

necessary

is

a

the child-specific

an environmental experience

shared reading between parent and child. This

experience

is

clearly

environmental from a

phenotypic perspective. However, the individual

172

HANDBOOK OF UNDERSTANDING AND MEASURING INTELLIGENCE

.

differences in the

shared reading

amount of time and

may

quality of

through distinct etiolo-

arise

To the extent that shared reading is influenced by genes associated with child reading skills, shared reading will show a genetic component. If this is the case, parents will be more likely to spend similar amounts of time with gies.

their identical twins than their fraternal twins. In

addition, shared reading

may

operate through

to

be compatible with their genetically influenced

characteristics.

tions

starting

A Gene x Environment interaction is inferred when

there

a nonlinear association

is

Rowe, Jacobson, and Van den Oord (1999)

found

heritability of verbal

measured environments associated with cog-

cents.

environmental etiology.

have

DNA.

Over

sample of adoles-

of IQ was higher

in

An

SES

experience does not

coming is

low

families.

processes

may

explain the decrease in "shared

environmental" influences

two decades, behavioral

high

in

Developmentally. passive gene-environment

into

likely to

mediated by genotype. past

heritability

in a

socioeconomic status (SES) families than

but the probability of

the

IQ

a genetic as well as an

contact with a particular experience

be partially

The

moderated the

parental education

that

cognitive skills are genetically mediated, the

may have

between

genetic and environmental influences. For example,

nitive skills

come with new

school and puberty,

children and adults.

shared and nonshared environmental influences. that to the extent that

transi-

opportunities to meet and interact with other

The important point

is

The major developmental

childhood and adolescence, such as

in

same home

living in the

more

in childhood.

more

are

Children

likely to expe-

when

genetic studies have highlighted the importance

rience

of gene-environment (G-E) processes. There

younger and different environments when they

are

two general classes of G-E process

—gene-

studies have

ment

heritability.

(1977) described three types of Passive

G-E

correlation

is

G-E

and Loehlin correlation.

found when a child

provided an environment that

is

is

correlated with

the parents" (and. hence, the child's) genes.

An

example of passive G-E correlation is the number of books in the home and child IQ scores (Plomin & Bergeman. 1991). Parents with more education and higher cognitive performance scores provide more books for their children. Their biologically related children are

not only exposed to

more books but

are also

employed

a sibling design to derive

shared environmental estimates, and

nonshared environmental estimates, the sibling design confounds passive shared environment. for

why

G-E

One

correlation and

of the possibilities

shared environment becomes so low

because passive

G-E

correlation

is

childhood but declines as active and evocative

G-E correlation and interaction effects become so much stronger. Second, differential experience

in siblings

has been unnecessarily equated

with the nonshared environment.

experiences

Differential

may also be a funcG-E correlation. For

older children

in

tion of active or evocative

are contributing to higher cognitive test perfor-

example, the heritability of cognitive

mance

increases to

correlation exists

between child cognitive out-

comes and

the

A

G-E correlation is a second found when an environment is pro-

number of books

in the

home.

reactive or ex'ocative

type and

is

is

operating in

provided with some of the parents' genes that scores. Thus, a passive gene-environment

they are

Although many behavioral genetic

are older.

environment correlation and Gene x Environinteraction. Plomin. DeFries.

similar environments

adulthood

.80 in

(McGue

ability et al.,

1993). However, because these estimates are typically derived

from a

variate estimates of effects are

sibling design using uni-

heritability.

subsumed

nonpassive

G-E

in the heritabilitv estimate.

vided to an individual by others as a reaction to his or her genetically influenced behavior.

Active

G-E

correlation occurs

when

a child

Conclusion

seeks environments based on his or her geneti-

influenced characteristics. Scarr and McCartney (1983) proposed that these effects become increasingly salient over the course of development as children gain more control over situations and experiences that are more likely cally

It

is

difficult to seriously dispute that intelli-

gence has both genetic and environmental components. However, there is still much work to be

done

to delineate

what

this

for theories of intelligence

information means

and the assessment

Behavioral Genetics and Intelligence

we can draw

of intelligence. For now. general conclusions. First, strong genetic rently

component

that

and longitudinally. This

difficult to fact. It is

tation of

three

intelligence has a is

instead possible that g

makes

is

important (see Chapter 16. this

volume). Measured specific cognitive abilities

indeed correlated but yet are factorially

are

As described above,

distinguishable.

seem

distinctions

to be

References

these

shaped mainly by the

Alarcon. M.. Plomin, R.. Fulker. D. W.. Corley. R.. DeFries.

of specific cognitive abilities data

age

Genetics. 28(4), 255-263.

Baker. L. A.. Vernon. genetic

&

A..

P.

Ho. H. Z. (1991). The

between intelligence and

correlation

speed of information

Genetics. 2/(4). 351-368.

closer to the

below

the

level,

operate

—

of measured behavior.

level

another way. intelligence

some

may

of individual genes

far

Put

may be modular

at

but at the level of psychometric

12 years of

at

Colorado Adoption Project. Behavior

in the

reductionism required to find truly orthogonal processes intluencing intelligence

&

T.

Barrett. P.

Behavior

processing.

Eysenck. H.

J.

(1992). Brain evoked

and intelligence: The Hendrickson

potentials

paradigm. Intelligence. 76(3^), 361-382.

M.

Bartels. M.. Reitveld.

Boomsma. D.

van Baal. G. C. M..

J. J.,

mental intluences on the development of

molarity because of genes.

gence. Behavior Genetics. 32(A). 237-249.

Second,

this

genetic stability also has an

The

ences.

quantitative genetic data suggest that

genes are important for cognitive that

gene-environment effects are likely

operating

when examining

the

and

skills

to

be

measured en\

ronments associated with cognitive

i-

skills.

Quantifying the connections between gene-

environment process and development will provide a

more complete

that yield individual

picture of the

mechanisms

differences in cognitive

Third,

behavioral

genetics

has

important

implications for the assessment of intelligence. It is

a mistake to assume that just because gene-

important to intelligence that the envi-

ronment does not matter. The average intelligence

1998).

In

level of

has increased over time (Flynn, addition,

some

interventions

have

been shown to improve intellectual functioning

(Campbell 1991).

& Ramey,

However,

differences in

it

1994: is

intelligence

genetic differences. Thus, to

assume

that

Ramey

clear

it

& Campbell.

that

individual

are

influenced by

is

also a mistake

environments and interventions

&

C.

J.

intelli-

Corley. R.. Plomin. R..

S.,

Heuitt.

K. (2003). Develop-

J.

ment genetic analysis of general cognitive ity

from

to

1

1

abil-

2 years in a sample of adoptees,

biological siblings, and twins. Intelligence. 31.

31-19.

Boomsma.

D.

(1993). Current status and future

I.

prospects in twin studies of the development

of cognitive T. J.

Infancy to old age. In

abilities:

Bouchard

Jr.

&

Propping (Eds.). Twins

P.

as a tool of behavioral genetics (pp. 67-82).

UK:

Chichester.

ability.

tics are

Bishop. E. G., Chemy. S. DeFries.

&

(2002). Genetic and environ-

I.

measures, there will likely be some degree of

impact on potential gene-environment influ-

&

C. (1998). Multivariate path analysis

J.

nonshared environment. However, the level of

level

problems

associated with cognitive functioning.

the manifes-

hundreds of genes of very small effect

abilities are not

treat

73

it

statistical arti-

This does not mean that specific cognitive

size.

and

detect, prevent,

1

stable concur-

stability

dismiss g as merely a

on how we

•

Boomsma, D.

Wiley.

Anokhin. A..

I..

&

de Geus. E. (1997).

Genetics of electrophysiology: Linking genes,

and behavior. Current Directions

brain,

in

Psychological Science. 6(4), 106-110.

Boomsma, D.

I..

&

Somsen. R.

measured

tion times

in a

J.

M.

(1991). Reac-

choice reaction time

and a double task condition:

A

small twin study.

Personalin and Individual Differences. 12(6),

519-522. Bouchard.

J.

T.

Jr..

&

McGue. M.

studies of intelligence:

A

(1981). Familial

review. Science. 212.

1055-1059. Brooks. A.. Fulker. D. W..

&

DeFries.

J.

C. (1990).

associated with intelligence occur in a genetic

Reading pert'ormance and general cognitive

vacuum. Instead, understanding the mechanisms through which genetics and environment

ability:

affect intelligence will

have important implications

data.

A

multivariate genetic analysis of twin

Personality

11(2). 141-146.

and Individual

Differences,

174

.

HANDBOOK OF UNDERSTANDING AND MEASURING INTELLIGENCE

CampbeU, F. A.,

& Ramey, C. T. (1994). Efifects of early

intervention on intellectual and academic achieve-

ment:

A

income

follow-up study of children from low

684-698.

families. Child Development, 65,

Comparison of Minnesota and

twins:

adult

Swedish

Behavior

data.

Genetics,

25(5),

421^31. Flynn,

R. (1998).

J.

IQ gains over

Toward

time:

find-

&

ing the causes. In U. Neisser (Ed.). The rising

Plomin, R. (1992). Multivariate genetic analysis

curxe (pp. 25-66). Washington, DC: American

Cardon, L. R., Fulker, D. W.. DeFries,

C,

J.

of specific cognitive abilities in the Colorado

Adoption Project

age

at

Intelligence,

7.

16,

Casto, S. D.. DeFries.

C,

J.

&

W.

Fulker, D.

(1995).

Multivariate genetic analysis of Wechsler Intelli-

gence Scale for Children-Revised (WISC-R)

Chemy.

A

Behavior Genetics, 25{ 1 ), 25-32.

S. S.. Fulker.

(1994).

S., et al.

J.

developmental genetic analysis of continuity

and change

in the

Bayley Mental Development

MacAnhur

Index from 14 to 24 months: The

Longitudinal Twin Study. Psychological Science, 5(6), 354-360.

LISREL

J.,

Daniels, (1998).

J..

&

Plomin. R. (1990).

modeling: Genetic and environmental

influences on

Chomey, M.

IQ

revisited. Intelligence. 14,

Chomey.

K., Seese, N.,

McGuffin.

J.,

A quantitative

cognitive

ability

P.,

Thompson,

trait

1-29.

L. A., et

J.,

al.

E.,

Wu.

McCleam

G. E.

J..

Siegel. B.. Tang.

baum. M.

C,

Association.

S. (1992). Intelligence

ing learning. Intelligence, 76(3-4), Harris,

way

children turn out the

they do.

Chomey, M.

L. A.,

&

Lubinski, D., Thompson,

J.,

A

Plomin, R. (2002).

quantitative

(2002).

S., et al.

Molecular Psychi-

10-11.

A

children:

failure to replicate. Psychological

Science, 7i(6), 561-562.

Ho, H.

&

Z., Baker, L. A..

Jacobs, N., Rijsdijk, F,

Decker,

ligence and mental retardation.

American Jour-

E.,

Kolachana. B.

S..

H., Mazzanti. C. M., Straub, R. E.,

COMT

Vall08/158 Met

genotype on frontal lobe function and risk for

Proceedings of the National

Academy of Science, B.,

A..

Mayeux,

J.

L.,

Hyman,

R., etal. (1997).

Effects of age. sex. and ethnicity on the association

between apolipoprotein E genotype and

Alzheimer disease:

A

R., et al. (2002).

population twin

meta-analysis.

Kaplan, D.

D., Olson,

R

J.,

Ahn.

J..

.K., et al. (2002).

7(4),

Won,

Pauls, link-

APOE

6p2 1.3-22. American Journal of Human Genetics,

70(5), 1287-1298.

Knopik, V.

S.,

Alarcon, M.,

Comorbidity deficits:

of

&

DeFries,

mathematics

J.

C. (1997).

and

reading

Evidence for a genetic etiology. Behav-

and

Light,

J.

G.,

&

DeFries.

J.

C. (1995). Comorbidity

of reading and mathematics

disabilities:

ing Disabilities, 28(2), 96-106.

Luo, D.,

Petrill, S. A.,

& Thompson, L. A.

analysis of cognitive

278(16), 1349-1356.

Reserve Twin Project. Intelligence,

& McCleam. abilities in

Genetic

and environmental etiologies. Journal of Learn-

Journal of the American Medical Association.

G. E. (1995). Heritability of cognitive

gen-

368-374.

T W.,

Evidence for

exploration

McGue, M..

A

study of the causes of

Molecular Psychiatry,

Gayan,

E..

Child psy-

ability:

Alzheimer Disease Meta Analysis Consortium.

Finkel, D., Pedersen. N. L.,

(1988).

ior Genetics, 27(5), 447-453.

9S(12), 6917-6922.

Farrer, L. A., Cupples, L. A., Haines,

Kukull, W.

M.

age and association with reading disability on

nal of Mental Deficiency, 92(1), 2-11.

schizophrenia.

S.

Derom. C, Danckaerts, M.,

Derom,

Thier)', E.,

association.

etal. (2001). Effect of

trait

locus not associated with cognitive ability in

influences. Behavior Genetics, 18(2). 247-261.

in cognition.

T

Why York:

cognitive processing: Genetic and environmental

Detterman, D. K. (1987). Theoretical notions of intel-

J.

New

Free Press.

eral

Callicott.

in

415^26.

R. (1998). The nurture assumption:

J.

(CHRM2)

Egan, M. F, Goldberg,

Buchs-

and changes

regional cerebral glucose metabolic rate follow-

chopathology and lower cognitive

7,

&

Abel. L..

Role of the cholinergic muscarinic 2 receptor gene

(Eds.), Nature,

psychology (pp. 77-97). Washington,

Psychological

Rostamkhani, M., McGue,

S..

&

Covariation between intelligence and speed of

M., Lacono, W. G.. Cheng, L.

atn,;

L. R. (1993).

locus associated with

children.

in

1

Owen, M.

Science, 9(3), 159-166.

Comings, D.

& Cardon.

S. S..

DC: American Psychological

Hill, L.,

M.

Chipuer. H. M., Rovine.

&

Plomin

In R.

nurture

Haier. R.

D. W.. Emde. R. N.. Robinson.

Corley, R. R, Reznick.

J.,

Chemy.

Continuity and change in cognitive development.

383-400.

factors.

Psychological Association. Fulker. D. W..

of genetic

Malhotra, A. K., Kestler, L. J.

A., Goldberg, T. E.,

g:

(1994).

Hierarchical

An

factor

data from the Western

J.,

&

18,

Mazzanti,

335-347.

C,

Bates,

Goldman, D. (2002).

.

Behavioral Genetics and Intelligence

A

polymorphism

functional

and performance on a

test

in the

COMT

gene

of prefrontal cogni-

American Journal of Ps\chiatr\\ 759(4),

tion.

J.

D. (1992). Biological and physiological

76(3^),

correlates of intelligence. Intelligence,

257-258.

McArdle.

(1994).

G

there

Is

P

Kennedy. D., Simon,

A

(2000).

A., Lefly, D., Chhabil-

MRI

twin

human

L. (1998).

J.

Hamagami,

Prescott, C. A..

A

&

F..

contemporary method

for

Petrill.

brain.

S.

Plomin. R.. Hewitt.

P.

Lipton.

A..

S.

Chemy.

Corley. R. P. et

S.,

al. J.

C.

Journal of Cognitive Neuro-

the

science. 27(1). 223-232.

J.,

H., et

J.

study of size variations in

76(1). 163-2(X).

Horn.

E.

genetic

(Is there

133-143.

abihty'?) Intelligence, 18,

Pennington. B. F. Filipek,

behavior genetic models. Behavior Genetics,

J.

McCleam, G.

beyond g?

75

influence on specific cognitive abilities indepen-

das. N..

(1986). Latent variable growth within

J. J.

McArdle,

1

dent of genetic influences on general cognitive

652-654. Matarazzo,

&

Pedersen, N. L., Plomin, R..

•

al.

(in

J..

press).

developmental-genetic analyses of age changes

Genetic and environmental contributions to gen-

Developmental Neuro-

eral cognitive ability through the first 16 years of

in

intellectual

abilities.

psychology. 74(1). 69-114.

McCartney.

M.

K., Harris,

A

(1990).

F.

Developmental Psychology.

Petrill. S. A.,

Luo. D., Thompson, L. A..

of general intelligence by elementary cognitive

E..

226-237.

tasks:

Johansson. B.. Berg.

Ahem. F.

L..

apart:

(1997). Sub-

on cognitive

80 or more years

in twins

Pedersen,

S..

Petrill. S. A., et al.

stantial genetic influence

abilities

old. Science, 276,

1560-1563. P.

A.. Stelmack, R. M.,

& Campbell,

K. B. (1992). Intelligence, reaction time, and event related potentials. Intelligence,

76(3^).

289-314.

Lykken, D. cognitive

T. J..

&

A

G. E.

Jr..

lacono,

W.

&

G.,

Behavioral genetics of

T. (1993).

ability:

R. Plomin

ton.

Petrill.

S. A..

Plomin. R.. Berg,

Pedersen, N., Ahern,

lifespan

McCleam

and psychology

perspective.

In

(Eds.), Nature,

(pp. 59-76).

DC: American Psychological

80 years and

older.

Psychological Science, 9(3),

Petrill. S. A..

Saudino. K.

Washing-

T.

J.,

Jr.,

Association.

Lykken. D. T.

twins reared apart. Intelligence. 76(3^).

289-314.

& Plomin. two-

in

Plomin. R. (1986). Development, genetics, and per-

Lawrence Erlbaum.

sonality. Hillsdale, NJ:

Plomin, R. (2003). Genetics, genes, genomics, and g.

Molecular Psychiatry,

&

8,

1-5.

Dunn.

J.

Nonshared en\ironment a decade

Why

(2001).

same family so

different?

later.

Cana-

dian Journal of Psychiatry. 46. 225-233.

& Christensen,

K. (2002). The heritabil-

of level and rate-of-change in cognitive func-

tioning in Danish twins aged 70 years and older.

Experimental Aging Research, 28(4),

&

M. C,

McArdle,

growth

Research,

Neubaurer. A.

3,

curves

J.

J.

for

435^5

1

(2000). Structured

twin

data.

Twin

165-177.

C.

Angleitner, A.,

Spinath.

&

Plomin. R..

&

Bergeman, C.

Borkenau,

M..

P

Riemann.

R..

(2000). Genetic

speed of information processing and their

psychometric intelligence:

rela-

Evidence

S.

(1991).

nurture: Genetic influence

The nature of

on "environmental"

measures. Behavioral and Brain Sciences.

14,

373^27. Plomin. R.. DeFries.

J.

C.

&

Loehlin.

J.

(1977).

Genotype-environment interaction and cortelation in the analysis of

F.

and environment influences on two measures of

to

Wilkerson. B..

year-old twins. Intelligence, 29, 31-43.

are children in the

tion

S.,

and modularity' of cognitive functioning

Feuer, D. (1984). Information processing abili-

latent

The

(1998).

183-189.

Plomin. R.. Asbury. K.,

McGue, M.,

Johansson, B..

S.,

et al.

genetic and environmental relationship between

&

McGue. M.. Bouchard. ties in

F.,

R. (2001). Genetic and environmental molarity

McGue, M.. Bouchard.

nurture,

Genetic and environmental influences.

Behavior Genetics, 26(2), 135-148.

general and specific cognitive abilities in twins

McGarry-Roberts.

Neale,

Detter-

meta-analysis of twin studies. Psychological

McCleam. G.

ity

&

man. D. K. (1996). The independent prediction

Bulletin. 107.

N.

Bemieri,

developmen-

Growing up and growing tal

life.

&

J.,

human

behavior. Psycho-

logical Bulletin. 84, 309-322.

Plomin, R.. DeFries,

McGuffin. ed.).

P.

J.

C, McCleam,

G.

E..

&

(2001). Behavioral genetics (3rd

New York:

Worth.

Plomin, R., Fulker, D. W., Coriey, R.,

& DeFries,

J.

C.

from the German Observational Study of Adult

(1997). Nature, nurture, and cognitive develop-

Twins. Intelligence. 28(4). 267-289.

ment from

1

to

16 years:

A

parent-offspring

I

I

176

HANDBOOK OF UNDERSTANDING AND MEASURING INTELLIGENCE

.

adoption study. Psychological Science.

S(6),

Posthuma, D., Neale, M. C. Boomsma, D. Geus, E.

Child

moderator.

as

Development,

70(5).

How

make

1151-1162.

AA2-AA1.

C. (2001

J.

).

I..

&

de

Are smarter brains running

& McCartney.

Scarr, S.. their

own

faster? Heritability of alpha

peak frequency, IQ,

environment

and

Behavior Generics.

424-435.

intercorrelation.

their

Tambs.

31(6), 567-579.

C.

Price. T. S.. Eley. T.

Stevenson.

&

J..

A

&

M..

J.

people

theory of genotype-

Child Development.

effects.

Sundet.

K..

K. (1983).

environments:

54.

Magnus. R (1984).

Plomin. R.

Genetic and environmental contributions to the

(2000). Genetic and environmental covariation

covariation between the Wechsler Adult Intelli-

between verbal and non-verbal cognitive devel-

gence Scale (WAIS) subtests:

opment

Intelligence. 16{4).

infancy. Child Development.

in

77(4).

Thompson,

948-959.

Ramey, C.

&

T.,

Campbell.

early childhood education tence:

Reed, T.

E..

&

and scholastic achievement: Genetic overlap but

velocity in a brain nerve

158-165.

2,

pathway of normal

Thompson.

Plomin. R..

A.,

L.

&

and specific cognitive

gence. 76(3^). 259-272.

Adoption Project. Intelligence.

&

Hetherington, E. M..

Turkheimer.

&

E..

environment:

A

phering genetic and social influences on adoles-

quantitative

review.

MA:

Cambridge.

Harvard

&

Pedersen.

N. L. (2002). Sources of influence on

rate

of

cognitive change over time in Swedish twins:

An

application of latent growth. Experimental

Aging Research. 28(4). 407-433. Reznick.

A

J.

S..

Corley. R.,

Robinson,

J.

(1997). in

second year. Monographs of the Society

for Research

in

Child Development.

62{

1

).

1-154. Rice.

T..

&

DeFries,

J.

C. (1986).

Multivariate path analysis of specific cognitive abilities

the

in

Colorado Adoption Project.

Behavior Genetics. 76(1). 107-125. Rijsdijk.

E

V.,

&

Boomsma.

D.

I.

(1997). Genetic

nerve conduction velocity and

IQ.

Behavior

Genetics, 27(2). 87-98. Rijsdijk, F. V., Vernon.

The

P

(

&

Boomsma. D.

I.

genetic basis of the relation between

speed-of-information processing and IQ. Behavioural Brain Research, 95(1), 77-84.

Rowe, D. C, Jacobson. K. C.

& Van den

J.

C.

& Boomsma. D.

I.

1998). Longitudinal study of genetic influences

on ERP-P3 during childhood. Developmental Neuropsychology, 14(1). 19^5. van Beijsterveldt. C. E. M.. Moienaar.

C,

J.

&

C. M.. de

P.

Boomsma. D.

I.

(1998a).

EEG

Genetic and environmental influence on coherence. Behavior Genetics. 28(6).

van Beijster\eldt. C. E. M.. Moienaar.

Geus. E.

J.

C.

&

Boomsma. D.

I.

P.

443^53. C. M.. de

(1998b). Indi-

P300 amplitude:

A

genetic

study in adolescent twins. Biological Psychology. 47(2).

Vernon.

91 -\ 20.

A. (1989).

P.

The

and Individual

heritability of

measures

Vernon.

P.

tion

A..

&

time,

Differences. 10(5) 573-576.

Mori,

M.

Wadsworth, J.

S.

J.

J.

(1994).

76(3^). 273-288. School achievement. In

C. DeFries. R. Plomin.

&

D. W. Fulker (Eds.).

Nature and nurture during middle childhood (pp. 86-101).

Oord. E.

(1992). Intelligence, reac-

and peripheral nerve conduction

velocity. Intelligence.

A..

and

Psychological Bulletin.

of speed of information processing. Personality

mediation of the correlation between peripheral

(1998).

theoretical, methological.

vidual difference in

Fulker. D. W..

9(1). 1-13.

726(1). 78-108.

Geus, E.

twin study of intelligence

longitudinal

the

&

C.

Colorado

abilities in the

van Baal. G. C. M. de Geus. E.

Reynolds. C. A.. Finkel. D.. Gatz. M..

J.

Waldron. M. (2000). Nonshared

Plomin, R. (2000). The relationship code: Deci-

University Press.

DeFries.

(1985). Parent-infant resemblance for general

adults correlates with intelligence level. Intelli-

cent development.

Plomin. R.

environmental difference. Psychological Science.

University Press.

J.,

&

L. A.. Detterman, D. K..

(1991). Associations between cognitive abilities

Cambridge,

project.

Jensen. A. R. (1992). Conduction

Reiss. D., Neiderhiser,

study of twins.

and academic compe-

The Abecedarian

UK: Cambridge

A. (1991). Povern,

F.

A

475-487.

Cambridge,

MA:

Blackwell.

Wilson. R. S. (1983). The Louisville Twin Study:

C. G. (1999). Genetic and environmental influ-

Developmental synchronies

ences on vocabulary IQ: Parental education level

Development. 54. 298-316.

in behavior.

Child

11 A Dialectical constructivist view of

Developmental Intelligence Juan Pascual-Leone Janice Johnson

Dedicated to the

memory

of Robbie Case

Separate causal factors operate through development to affect somewhat separately cognitive processes

and produce

to elucidate the nature

individual differences.

.

.

.

Further evidence

is

needed

of these factors.

NollandHorn(1998, g stands unassailed as a big concretion of mental triumph and a cognitive enigma.

p.

280)

a psychometric

test variance. It is

Deary (2002.

p.

176)

In summary, empiricism begins with the record of plain facts, science denounces this

evidence to discover hidden laws. There

is

no science but of that which

Bachelard (1949.

William Stem (1914/1977), contemporary of Binet

person's

problem

J.

Pascual-Leone)

38; translation by

hidden.

situations.

This coping synthesizes

or constructs performance by

composing pre-

concept. defined intelligence as the

viously acquired processes (schemes) in novel

successfully coping with novel

ways. Thus, Burt and Stankov defined intelligence

skill in

AUTHORS" NOTE; two authors from

anonymous

and inventor of the IQ

p.

is

Preparation of this chapter was facilitated by an operating grant (#410-2001-1077) to the

the Social Sciences

referee,

whose

insightful

and Humanities Research Council of Canada.

comments guided us

We

thank the editors and an

in the final revision.

177

178

HANDBOOK OF UNDERSTANDING AND MEASURING INTELLIGENCE

.

as "the 'integrative function of the mind' that

encompasses processes 2002,

p. 25).

at all levels"

assume

all

and

adulthood.

of

beginning

its

third approaches.

that intelligence

increases with age in normal persons, up to

spite

explanation and process analysis in the second

Other definitions have been offered,

often less apt, but

In

cognitive science, which contributes to causal

6.

(Stankov,

Binet's

(in

and

Because intelligence as a construct is rooted in individual and age differences, psychometric intelligence

and developmental intelligence might

intel-

be the two main contenders. However, the other

ligence often suffer from lack of integration

four are essential for clarification and construct

Spearman's work), current conceptions of with cognitive-developmental research.

Ever

since Reuchlin (1962. 1964). researchers have

validity.

Many

believe that factor analysis

descripti\e tool, a

neo-Piagetian theory with psychometric con-

ture available in data (e.g.. Burt.

ceptions of intelligence

Demetriou,

Case,

Anderson. 1999;

(e.g..

&

Platsidou,

Pascual-Leone, 1969). The hope

mental theory and

its

Kazi,

is

2001;

that develop-

empirical methods can

2000;

is

way of condensing

recognized the need to integrate Piagetian or

Reuchlin.

1941; Deary,

and think

1962).

a general the struc-

other

that

approaches, such as development, serve to anchor causal interpretations of factors

&

Horn

(e.g.,

Horn, 1998;

Noll. 1994; Reuchlin. 1964). Develop-

help "to devise measures that are theoretically

mental intelligence addresses the ontogenetic/

grounded and can be used with equal

psychogenetic evolution of the person's informa-

all

age groups" (Anderson. 1999.

we summarize an

chapter,

toward

this

integration

and

p.

facility in

316). In this

attempt to work

progressive emergence of noninformational organi-

to other

smic processes (resources, factors of mamration),

relate

it

ongoing attempts.

all critical

A Theories. Models, and Varied

for causal theorizing.

ways

is

to express the

tural findings

construals of intelligence

descriptive when it phenomena and strucencountered. It is causal when its

theory or model

offers

constructs are distinct from the descriptive constructs to be explained,

The

first

obstacle to integration of different

views about intelligence

We

components) and the

tional processes (schemes,

is

epistemological.

can be independently

anchored on experience via experimentation,

and

the

causal

constructs

are

capable

of

distinguish six substantive viewpoints from

accounting for change that descriptive con-

which intelligence can be investigated (PascualLeone & Goodman, 1979);

experience, maturation, or organismic change.

undergo as a

structs (and data)

result of events,

Descriptive and causal theories or models, in 1

the innate potential or capacity for development (a genotype,

i.e..

Hebb"s [1949] Intelligence A);

turn,

can be either local or general (Pascual-

Leone,

1978.

2.

the high-cognitive

(i.e..

These

1980).

of theories/models

are

all

distinct

jointly

sorts

needed.

intellective/intellectual)

competence of a partly grown or mature person

Combined, they tion

(an intellectual phenotype.

i.e..

(i.e..

yield

two dimensions of

varia-

local/general and descriptive/causal)

Hebb's [1949] that

can be crossed. Thus. simplif> ing. there are

Intelligence B);

local 3.

psychometric intelligence, indexed by cognitive tests

4.

(Vernon's [1969] Intelligence

developmental intelligence,

and Piaget

Q:

causal

(Pepper's

[1942]

or models. This

is

complexity of

and

is.

the

more

distinctly

intelli-

normally achie\ing individuals (Pascual-Leone

& Goodman's

local

important because the more

which, as Binet

general a causal theory did. scales the

descriptive,

'"mechanistic"),

general causal (Pepper's "organismic") theories

gent performances by the age of the youngest

differentiated

it

will

structural theories

dinate,

it

and the more theories

be from the descriptive

means

to explain

and coor-

local, the less differentiated

[1979] Intelligence D);

causal 5.

general

descriptive,

will

neuroscience, which provides a new criterion

descriptive

of construct validitv; and

distinction to collapse.

theories,

tend to be from eventually

their

leading the

A

From

Gardner's (1993)

perspective,

this

Dialectical Constructivist View of Developmental Intelligence

multiple intelligences theory appears as a collection of local descriptive theories,

hierarchical

1998;

Carroll,

analysis

factor

models

process

descriptive

the

(e.g.,

schemes, working mental space, short-term

memory) on which

mechanisms

these causal

theories

(e.g..

apply (as Oberauer also suggests

Horn,

1998;

this

1998;

Cattell,

and various

with

179

•

Chapter 22,

in

volume).

Horn & Noll. 1994; Noll & Horn. 1998) appear as more sophisticated and better structured

intelligence has to be

general descriptive accounts. Causal theoreti-

processes distinct from psychometrics, or gen-

may be

interpretations

cal

theories, but often there

adjunct

to

these

no clear distinction

is

We

believe that a general causal theory of

grounded on organismic

developmental psychology, but

eral descriptive

These constructs should

interpretable in them.

between causal and descriptive constructs,

be molarly defined so that they can serve as

perhaps due to lack of organismic general causal

technical

models

analysis of tasks, or test items, in a principled

can generate theory-based {funda-

that

mental) measurement of key causal constructs

language

for

conducting

process

manner.

(Deary. 20(X)).

A tant

its

explain

to

Therriault,

Laughlin.

emerged

size has

the

intelligence factor

SiiB,

focused on work-

line of causal theorizing

memory and

ing

as impor-

of a general

existence

(Conway, Cowan, Bunting,

& Minkoff. & Conway,

1999; Kyllonen, 2002;

2002; see also Chapter

&

Schulze,

We

volume).

5, this

are

sympathetic with this interpretation, which in a different language agrees with an early predic-

tion of Pascual-Leone

1970; Pascual-Leone

(Pascual-Leone.

&

1969,

Effortful Structural Learning The most

memory (WM) Engle

(Pascual-Leone.

It

2003) exhibits

1970; Pascual-Leone

1969,

&

discussed below) executive

when

number of

first

units that

retain in terms of

WM

simultaneously can

schemes or chunks.

schemes or

structures that the individual possesses. Indeed,

Pascual-Leone (1969) was the

mem-

Following Miller (1956), these working

ory theories, as well as our own, count the

coupled with the repertoire of effortful (or LM, as

al.,

Smith, 1969) pioneered:

mental-attentional

of the person but only

capacity

(A/)

(Cowan, 2001;

as a construct

1999; Ruchkin et

four characteristics that our developmental theory

1.

maximum

corresponds to the

et al.,

conception of working

interesting

Goodman, 1979) with

developmental intelligence:

to

OF Mental Attention and

2(X)2; Engle, Tuholski,

Oberauer, Wittmann, Wilhelm,

reference

Intelligence as a By-Product

2.

Tacitly, they reject the idea

memory and

("central")

researcher to

of

WM as a separate WM to

instead posit

point out that neo-Piagetian theory might offer a

be a state of hyperactivation (of information-

"good platform" from which

carrying units) within long-term

tion of

to attempt integra-

developmental and psychometric

gence (Case

et al..

2001.

p.

intelli-

327). Considerable 3.

research has

shown

that

Piagetian and neo-

Piagetian tasks (examples are described later in this

Focal attention and

its

assumed

mechanism

be the

to

emergence of

chapter and in neo-Piagetians references

given) load on fluid general intelligence (Gf) or

memory (LTM

or repertoire of schemes).

4.

At

least

in

WM

limited capacity are that

causes

in the brain.

Cowan

(2001) and Engle

new conception

et al.

WM

general intelligence factors (see Lautrey, 2002;

(1999), this

Vernon, 1969). From our perspective, however,

from short-term memory (STM) by claiming

the construct of a separate ("central") working

that

memory

unclear under scrutiny and possibly

(as

Embretson,

expresses the impact of controlled attention on

false

is

(see

2(XX), in press;

1995;

Ruchkin, Grafman, Cameron,

Bemdt, 2003: SuB

et al.,

2002). Indeed,

conflate or confuse causal (e.g.,

Pascual-Leone,

it

&

STM

is

distinguishes

the set of activated units in

Hebb. 1949. thought of

it),

whereas

LTM

WM

STM.

tends to

dynamic mechanisms

mental or endogenous/executive attention)

In this section,

we show how our developmental

theory explicates these four points, the

first

two

180

HANDBOOK OF UNDERSTANDING AND MEASURING INTELLIGENCE

.

addressing the problem of a psychological unit

and the and

last

two

the construct of mental attention

STM

relation to

its

and

WM.

analytical schemes produced with mental effort (initially in the left tic

hemisphere), global or holis-

automatized schemes (right hemisphere), and

so forth.

Schemes as Psychological

A

Units.

chunk

is

Schemes

an

are

ill-defined but necessary psychological unit that

tems

can be explicated by the scheme concept.

representations

move from one memory

Information does not

store to another; rather,

by

it

is

carried or mediated

distinct collections of neurons, often distrib-

uted over the brain, that are cofunctional (vis-a-

and coactivated

vis certain activities)

We

tasks.

these

call

schemes. Schemes situated

beyond

collections

the organism are

in

(i.e..

determine

conjointly

self-propelling

dynamic

sys-

produce plans, patterns of action, or

when

they

They

apply.

are

abstracted across situations for a given sort of

They must be

praxis.

internally consistent to be

formed. Their conditions and/or effects can, in turn,

be constituted by (copies of) other schemes.

Experiences are only possible because schemes produce, assimilate, or structure them.

distinct.

situation specific) causal factors

(i.e.,

overdetermine

that

functional

some

in

that

the need) manifest pertbrmance. and their

The Hidden Operators of the Mind's Brain. With the theory of schemes alone, one cannot general organismic constraints

explain

(i.e..

proper recognition facilitates process and task

those applying across kinds of schemes) such

analysis (Pascual-Leone,

as

Leone

&

1995; Pascual-

1970,

Goodman. 1979; Pascual-Leone

&

Johnson. 1991. 2004; Pascual-Leone, Johnson. Baskind, Dworsky.

&

WM

"central"

capacity

mechanisms,

inhibitory

"central"

limits,

structural versus con-

schemes'

tent learning, "central" resolufion of

competition in the network, or the emergence

Severtson. 2000).

Schemes overdetermine performance because

of truly

performances via unplanned

novel

they are self-propelling (Piaget"s assimilation

dynamic syntheses.

function) and tend to apply to the situation

prefrontal lobes take care of this by controlling

fire as

(i.e..

collections of neurons) under minimal

processes elsewhere

—

conditions of activation. Every scheme has a

areas)

dual reality: as a functional package of qualita-

Ruchkin

tive characteristics that

form

into) experience

its

parameter

learned:

chies

recursive

etc.),

It

that

the

scheme

will

set

(via neuronal

we

procedures

and can be

common to have scheme hierarschemes of schemes of schemes,

is

(i.e..

which we

the coordinated

call structures or

schemas

"program" or "script"

that

(e.g..

gov-

erns our behavior in restaurants or within familiar sorts

of problem solving).

LTM

can be best

explicated as a manifold repertoire of schemes

or structures sorted in kinds: executive schemes (prefrontal), operative tive

schemes

or linguisdc

(posterior is

limited.

metaphor of a

attentional "pointers."

A

restricted

better for-

mulation might be to postulate that the brain, in

synchonization) to cause performance. Schemes like

that

(2003) and others have expressed

with the

idea

number of

cortex

the

power

addition to a repertoire of schemes, has a small

(i.e.,

become hyperactivated and apply are

this

et al.

thought that

activation

probability

the

tions

inject

(i.e..

applies and as a

it

in

controlling

is

neural network area) that condi-

\ariable quantitative

weight for

can inform

when

a

often

It

(i.e..

—

schemes

(frontal). ^^?//ra-

perceptual or representational

occipital, parietal, or temporal),

coordinated packages of automatized opera-

of general-purpose functional resources that

hidden or

call

Leone,

1987,

Johnson.

We

silent

1995.

1991.

operators (Pascual-

2000; Pascual-Leone

2004; Pascual-Leone

&

et al.,

them operators because they are mechanisms of brain "hardware," defined as molar procedures whose computational details are unspecified and that apply on (constrain) schemes to change or produce new schemes or to synthesize truly novel performances. These operators are hidden because they lack substantive content referents, which 2000).

call

functional

schemes have sentational,

(i.e..

etc.).

purely relational

perceptual,

Instead,

motor, repre-

operators

multivariate

express

constraints

surprising patterns or "anomalies" (exhibited

tive/motor schemes (frontal right hemisphere,

under specific empirical circumstances)

basal ganglia, and cerebellum), affective/emotion

pure theory of schemes cannot explain. They

schemes (broadly defined limbic

are our attempt to formulate, as categories of

system).

that a

— A

Dialectical Constructivist View of Developmental Intelligence

•

181

Table 11.1

TCOs

Operator

Description

Brain Rei^ion

A

Set of affective processes that intervene in motivation and attentive arousal

Limbic lobes

C

Both the process of content learning and the schemes derived from

Broadmann primary

Hidden Operators Listed

in

Order of Their Likely Evolutionary Emergence

and secondary

associative content learning

areas

The ^eW

F

which

operator,

mechanism manner

acts as the brain's binding

that brings

All areas

closure to mental representations in a neo-Gestaltist

LC

The process of automatized

logical-structural learning derived from

C

Right hemisphere

learning through overpractice

T

Temporarily and effortlessly collates sequences of schemes, thus facilitating

Occipitotemporal

the coordination that constitutes distal objects

S

Facilitates

emergence of spatial schemes or schemas by coordinating

relations of coexistence

among

activated

schemes

Occipitoparietal

effortlessly within the

situation

The

attentional interrupt,

which corresponds

unwanted schemes activated

inhibition of

power of

to the

Prefrontal

central active

in the situation

M

Mental

LM

Logical-structural learning caused by the effortful use of mental-attentional

attentional capacity of the individual

Prefrontal

Left hemisphere

capacity

tertiary areas

Executive schemes

in the

person's repertoire for the task

constructs within a psychological theory,

some

purely organismic constraints that the brain's cortical architecture

imposes on psychological

processes and behavior.

with permission, summarizes the operators that

we

10 hidden

currently consider. These 10

categories of operators (which

we

operators for short) are ordered

in

call

hidden

Table 11.1

according to their evolutionar}' emergence, as

we

speculate

to be.

it

We

have best methods of

hand

Prefrontal

&

1998: Johnson. Fabian.

Pascual-Leone,

&

Leone

1970,

Pascual-Leone. 1989:

1987,

Baillargeon,

1994;

2000). Development of

et al..

Table 11.1, modified from Arsalidou (2003)

at

endogenous growth of mental mental

theory,

attention

among

(i.e.,

2001;

Pascual-

Pascual-Leone

M

explains the

attention. In our

endogenous, executive)

from dynamic interactions

results

four different sorts of processes:

ator (the

scheme

M oper-

activation resource), / operator

(attentional interrupt, "central" inhibition

anism),

F operator

mech-

(neo-Gestaltist internal field

assessment for the eighth operator, mental (M)

mechanism, often known as minimum-principle

the key hidden oper-

or stimulus-response [S-R] compatibility), and

attentional capacity. This

ator

is

underlying Cowan's (2001)

et al.'s

and Engle

(1999) concept of working memory.

M

capacity appears empirically as a purely relational pattern relating task

viduals'

age.

That

is,

as

complexity to indi-

we

parametrically

E operator (the tive

allocates

functions.

M

.set

of execu-

or / to relevant/

M, and E are prefrontal We believe F results from local lateral

irrelevant schemes).

/,

inhibition processes in the cortex. This system

increment the number of aspects to be enter-

is

our model for mental attention.

that this mental attentional

system com-

age of children capable of passing the

bines capacity and other dynamic constructs to

M

explain what others would call the central exec-

task or item increases predictably. That

capacity

(often

neo-Piagetians)

called

is,

working memory by

increases

endogenously with

chronological age until adolescence

(e.g..

Case,

utive (see Chapter 4, this

attention

(.see

volume) or controlled

other chapters in this volume).

To

explain cognitive development, one needs other

182

HANDBOOK OF UNDERSTANDING AND MEASURING INTELLIGENCE

.

hidden operators as well. For instance, there content

or the

learning,

substantive)

(i.e.,

is

solving, cognitive development, and emotional/

C

interpersonal

operator, and logical-structural/relational learn-

L

ing, or the

The L operator encom-

operator.

development, and they exhibit

individual cognitive-style differences

by the

ability

indexed

cope well with them. For

to

two varieties: one obtained by automatization from C learning, called LC learning, and another obtained from repeated acts of mental attention (M), which we call

cope well, but field-dependent persons, despite having good developmental intelligence, often

LM learning.

2002; Pascual-Leone, 1989; Wapner

passes

at least

Witkin's field-independent persons

instance,

cope badly (Goode, Goddard, 1991; Witkin

Misleading Versus Facilitating Situations and

&

& Pascual-Leone, & Demick,

Goodenough, 1981).

Facilitating situations

those

are

in

which

Their Importance. The power of mental atten-

only task-relevant schemes are activated, so

by the number of organismically

automatized schemes are often sufficient to

tion is indexed

distinct aspects (task-required

schemes)

that the

person simultaneously can entertain within mis-

A

leading situations.

when

it

elicits

situation

schemes

misleading

is

that are unsuitable for

solve

the

Application of automatized

task.

schemes does not require use of mental

effort.

Furthermore, acquisition of automatized schemes is

much more dependent on

a suitable socio-

hand and. when they apply, lower the

cultural milieu (learning opportunities) than

on

probability of the subsequent activation of task-

maturational processes, such as growth of

M

schemes (Pascual-Leone, 1969, 1987, 1989; Pascual-Leone & Baillargeon, 1994;

capacity.

the task at

relevant

Pascual-Leone

& Johnson, 2004). In misleading

situations, a processing conflict usually

emerges

reliable

One

Strategy Y.

unsuitable but

is

well-learned

to

The other

which we

is

X,

is

by

cross-sectional developmental traces of perfor-

M and

/

as

&

ing the situation into a facilitating one. Stable

executive schemes and

Baillargeon, 1994). Because

in

misleading situations, the unsuitable Strategy

Y

is

more or

less

automatized or overlearned.

this is the first strategy to

be mobilized and

acti-

vated. Consequently, for the suitable Strategy to determine performance,

ing Strategy (i.e.,

X

one must avert apply-

Y by using active, central

mental-attentional

inhibition

interruption

operator), and concurrently. Strategy

X

—our

Y

embedded

in

it)

situations

are

pre-

this

leading (Pascual-Leone,

paradigm are mis-

1969,

1980) because

the appearance of a clay ball being compared, for instance, with a long sausage of clay (which

in

the intellective process that corrects this error

and breaking down

common

first

substance, weight, and

(i.e..

volume). The items of

when

Piaget's conservation

creating this illusion function as Strategy Y, and

the perception of the manifest drawing. Mis-

leading

of physical matter

is

to

the relevant figure. Strategy X,

partly inhibiting

sented to respondents,

hiding (has

process that succeeds in seeing the relevant

by

well-known developmental

most children the illusion that the ball has less amount (this is due to the greater perceptual surface of the sausage). The low cognitive schemes

{M

in contrast, is the executive mental-attentional

figure

A

must be

compels the respondent is

leading situations.

task paradigm exhibiting stages,

contains equal substance) perceptually suggests

operator). For instance, in an embedded-figures

see only the manifest drawing that

stages of development exist but (contrary to Piaget's claims) are found reliably only in mis-

/

boosted by mobilizing mental attention test item. Strategy

to disappear with repetitions

suitable but needs

capacities (Pascual-Leone. 1989, 1995;

Pascual-Leone

mance, which tend

as respondents learn to cope with the task, turn-

be effortfully boosted by mental-attentional

mechanisms such

dis-

opera-

field factor

strategy,

Misleading situations typically exhibit

continuous or stagewise trajectories in their

facilitated

—our F

only in mislead-

ing ones.

call

automatized schemes (often

or

congruent with the tor).

strategy,

domain-free stages of development

in facilitating situations but

between two or more different implicit or explicit strategies.

One, therefore, should not expect

in

problem

functions

shows the

as

Strategy X. This example also

that ability to interrupt (centrally inhibit)

misleading schemes does not suffice to

explain a developmental process

(i.e..

applica-

tion of Strategy X); sufficient mental-attentional

capacity also

is

needed.

When

Piaget increased

A

Dialectical Constructivist View of Developmental Intelligence

the task complexity by changing the

from a

paradigm

of conservation of substance to one

test

of weight and then of volume, the developmen-

should consult Engle

model also explains

theoretical

lem. Task analyses show (Pascual-Leone, 1976)

difi'erent interpretation.

difficulty increased

to 10 years

M

that

1

rently increased

meaning of

this notation is

Within I,

this

model of mental

attention

F>, working memory (WM),

stood by

Cowan

'?

Journal of Experimental Child Psychology.

77,

138-154.

M. Chapman

Pascual-Leone,

J.,

I.

E. Sigel (Ed.).

Mahwah.

A

rein-

permanence. In M. Candler

(Eds.), Criteria

for competence

Lawrence Erlbaum.

(1999).

A dialectical

Development of mental represen-

and

applications (pp. 169-200).

NJ: Lawrence Erlbaum.

Pascual-Leone.

&

J..

Johnson.

J.

Manual

(2001).

manuscript. York University. Toronto, Ontario,

Canada. J..

&

Johnson.

(2004). Affect, self-

J.

motivation, and cognitive development: tical

A dialec&R.

constructivist view. In D. Y. Dai

S.

J.

Integrative perspectives on intellectual

nition:

functioning and development (pp. 197-235).

Mahwah, NJ: Lawrence Erlbaum. Dworsky,

J..

Baskind.

S.,

&

August). Misleading situa-

and

tions, active inhibition,

the fair assessment

Paper presented

of mental capacity:

at

the

Behaviour, and Cognitive

(2001). If the magical

number

for operations

is 4,

within

(in

J.

press).

Hidden operators of

mental attention applying on sion of a separate working

LTM

give the

illu-

memory. Behavioral

Pascual-Leone,

&

J.,

Johnson,

&

Baillargeon, R. (1994). Develop-

mental measure of mental attention. International

Journal of Behavioral Development, J.,

&

Goodman, D.

gence and experience: Instructional Science.

17,

(1979). Intelli-

A neoPiagetian 8,

161-200.

301-367.

approach.

J.,

Baskind,

S.,

Dworsky,

Severtson, E. (2000). Culture-fair assess-

ment and the processes of mental Kozulin

&

Rand

Y.

(Eds.),

attention. In A.

Experience of medi-

An impact of Feuerstein 's

theory

education and psychology (pp. 191-214).

New York: Pascual-Leone,

Pergamon. J.,

&

of water level: review.

Sciences.

Ottawa,

Science,

Ontario, Canada.

in

J.,

Johnson.

J..

S. (1998,

ated learning:

Pascual-Leone,

The psycho-

mental attention, executives, and symbols. In

working memory? Behavioral and Brain Sciences,

and Brain

(1991).

J.

role in task analysis:

& Johnson, J.

24, 136-138.

Pascual-Leone,

its

(pp. 155-187). Hillsdale. NJ:

S.,

how does one account

Pascual-Leone.

Johnson,

terpretation of object

&

35-66).

(pp.

Lawrence Erlbaum.

Hillsdale, NJ:

meetings of the Canadian Society for Brain,

843-844.

Pascual-Leone,

from

Pascual-Leone.

To appraise developmental

not enough. Behavioral

Pascual-Leone,

in

Pourchot (Eds.),

Sternberg (Eds.). Motivation, emotion, and cog-

processing capacity. International Journal of

Pascual-Leone,

& T.

Adult learning and development: Perspectives

Pascual-Leone.

(1996). Vygotsky, Piaget, and the

J.

problems of

Pascual-Leone,

models of learning

C. Smith

for FIT: Figural Intersections Test Unpublished

as dialectical factors in cognitive growth.

Pascual-Leone.

M.

tation: Theories

An experimentalist's underchildren. Human Development. 37,

J. (

Irwin, R. (1998). Abstraction,

constructivist view of representation: Role of

University Press.

Pascual-Leone,

adulthood. In

Pascual-Leone,

Press.

(1990b). Reflections on life-span

J.

intelligence, consciousness

stages of

that

Wisdom:

Sternberg (Ed.),

and development

on wisdom:

essay

Towards organismic processes ble. In R. J.

&

J.,

the will, the self, and

(1990a).

J.

Pascual-Leone,

Morra,

A

Advances

S. (1991). Horizontality

neoPiagetian developmental in

Child Development and

Behaviour, 23, 231-276.

Pascual-Leone,

J.,

&

Smith.

J.

(1969).

The encoding

and decoding of symbols by children: experimental

A new

paradigm and a neo-Piagetian

model. Journal of Experimental Child Psychology

8,

328-355.

— A

&

Pennings. A. H..

M. G.

Hessels.

measurement of mental

Dialectical Constructivist View of Developmental Intelligence

(19%). The

P.

attentional capacity:

of change

23, 59-78.

in children

Oxford University

neo-Piagetian de\elopmental study. Intelligence,

Stankov,

Pepper, S. C. (1942).

Emerging minds: The process

Siegler, R. S. (1996).

A

World hypotheses: A study

R.

L.

&

thinking. Oxford,

's

A

E. L.

diminutive general.

Grigorenko (Eds.),

of evidence. Berkeley: University of California

general factor of intelligence:

Press.

(pp. 19-37).

Piaget.

(1983). Piaget's theor\'. In

J.

(Series Ed.)

&

W. Kessen

of child psychology: Vol.

methods (4th

1.

New York:

John

the history of psychology

Brown

&

K.

J.

Thampy.

Trans.). Chicago:

University of Chicago Press.

(Original

work

ical theory':

A

New

systematizing attempt.

York:

M.

(1962). Les methodes quantitatives en

psychologic [Quantitative methods

in

psychol-

ogy]. Paris: Presses Universitaires de France.

Reuchlin,

M.

(1964).

L'intelligence:

genetique operatoire

et

Conception

conception factorielle

[Intelligence: Operational genetic

and

factorial

conceptions]. Revue Swiss de Psychologic. 23,

113-134.

Ruchkin. D. R.

S.

S.,

(in

systems:

A

SuB,

H.-M.,

Oberauer,

&

Wilhelm, O.,

and a

Tolman,

press). state

J.,

Cameron,

W. W.,

Wittmann,

capacity explains reasoning ability

little bit

E.

K.,

Schulze, R. (2002). Working-

&

C,

more. Intelligence, 30, 261-288.

Brunswik, E. (1935). The organism

Psychological Review, 42, 43-11. Vernon,

E. (1969). Intelligence

P.

K..

Working memory

&

Bemdt,

retention

of activated long-term memory.

Behavioral and Brain Sciences, 26. 109-111.

and

cultural envi-

ronment. London: Methuen.

Vuyk, R. (1981). Oven lew and critique of Piaget's epistemology

genetic

New Wapner.

1965-1980

(Vol.

1).

York: Academic Press.

S.,

& Demick,

J.

dence-independence.

Grafman,

(Series B.

and the causal texture of the environment.

International Universities Press.

Reuchlin.

1750-1920

Washington, DC: University Publications

memory

published 1975)

Rappaport. D. (1960). The structure of psychoanalyt-

Trans.). In

of America. (Original work published 1914)

(1985). The equilibration of cognitive struc-

J.

M. Whipple,

D. N. Robinson (Ed.), Significant contributions to

4{A)).

tures (T.

How general is it?

Mahwah, NJ: Lawrence Erlbaum.

testing intelligence (G.

and

Wiley. Piaget,

In 772^'

Stem, W. (1977). The psychological methods of

Handbook

History, theory,

103- 128).

ed.. pp.

Mussen

H.

P.

(Vol. Ed.),

UK:

Press.

(2002). g:

Sternberg

J.

201

•

(Eds.). (1991). Field

Hillsdale,

NJ:

depen-

Lawrence

Erlbaum. Witkin,

H.

A.,

&

Goodenough, D.

Cognitive styles: Essence

E.

(1981).

and origin. New York:

International Universities Press.

12 Development of Intellectual

Old Age

Abilities in

From Age Gradients

to Individuals

Martin Lovden

Ulman Lindenberger

The

present

chapter

objectives: (a) to ric

has

main

two

summarize psychomet-

theorizing and evidence about intel-

Cattell,

theor>';

1971:

Horn,

Cattell, 1966, 1967), the

PPIK

1996; see also Chapter

8, this

1982:

Horn

&

theory (Ackerman.

volume), and the

ligence in adulthood and old age and (b) to

decomposition of cognition into mechanics and

promote a person-centered (idiographic) approach

pragmatics (Baltes, 1987; Baltes. Lindenberger.

psychometric study of adult intellectual

to the

development. Both objectives can be framed the context of

two-component models of

span cognition. These models posit that

development of

life

in

life

span

intellectual abilities reflects

two

fundamental and dynamically interacting influences, the biological and the cultural. Historical

examples of such models can be found

in

&

Staudinger, 1998).

Though

the scope, termi-

nology, and details vary considerably across the different

they

versions,

all

share fundamental

assumptions beyond mapping intelligence onto

two underlying components

(see Baltes et

al..

1998; Lindenberger. 2001).

models assume

In general, these

ontogeny, there

is

component

that,

during

an "investment" of the bio-

Tetens's (1777) differentiation between relative

logical

and absolute mental capabilities and

Hebb's

tors related to neurophysiological status) into

(1949) distinction between intellectual power

bodies of biographical ly acquired knowledge

and

through processes of socialization, experience,

intellectual products

in

(Hebb, 1949). Typical

contemporary examples include the theory of fluid

and crystallized intelligence

AUTHORS' NOTE: Institute for

(i.e.,

Address correspondence

Human Development.

Gf-Gc

to

and education 1971).

(i.e..

(i.e.,

heredity and other fac-

investment theory; Cattell,

These investment processes lead

Martin Lovden, Center for Lifespan Psychology,

Max

to

Planck

Lentzeallee 94, D-14195 Berlin, Germany; e-mail: loevden@mpib-berlin.

mpg.de 203

204

HANDBOOK OF UNDERSTANDING AND MEASURE^G INTELLIGENCE

.

and intraindividual differences

inter-

in

the

acquisition and organization of procedural and

The

declarative knowledge.

nent and

its

biological

compo-

mental correlates are held to decline

after maturity.

The

component, however,

cultural

continues to increase over the

span as long as

life

knowledge maintenance and knowledge acquisition outweigh losses in the biological component.

Two-component

thus

theories

dispute

hypothesis

is

framed

we

Furthermore,

in terms of these interactions.

point out that theories about

development are generally anchored

intellectual

at the intraindividual level.

Therefore, the use of

methods relying on interindividual differences should be complemented by methods that are better suited for assessing

change

this intraindividual level.

We

in structure at

draw

attention to

the

a select toolbox of intraperson, or idiographic,

validity of a unitary general intelligence con-

psychometric methods and note that their appli-

struct

(i.e..

intellectual

g) in understanding the dynamics of life

span development: that

two broad categories of

least

deemed necessary

is,

at

ability factors are

to capture the basic proper-

of intellectual

first

category of ability factors represents mea-

life

surable outcomes of the influence of the biolog-

component on development.

itself in

It

manifests

novel information

and transformation of

(i.e.,

reasoning or Gf; Cattell,

&

1971; Horn, 1982; Horn

processes, including working

memory, process-

ing speed, and aspects of coordination and con-

et

of processing

al.,

(e.g..

1998; Horn,

Bakes,

1985).

1987; Baltes

Henceforth, these

processes are referred to as the "mechanics" of intelligence (cf. Baltes, 1987).

more disparate category of to procedural

mon

The second and

ability factors refers

and declarative knowledge com-

to a given culture

(i.e..

Gc), to specialized

knowledge such as occupational expertise Chapter

8,

this

(e.g..

volume), and to knowledge

about the meaning and conduct of Baltes et

life

(e.g.,

1998). Henceforth, these applica-

al.,

Cross-Sectional Evidence

The

Cattell, 1966, 1967)

or in broader ensembles of basic information

trol

The Age Gradients of the Mechanics and the Pragmatics

cognitive processes involving extrapola-

reorganization,

tion,

allow researchers to investigate the

amount and nature of heterogeneity in patterns of change and to gauge the validity of methods based on interindividual differences.-

span development. The

ties

ical

may

cation

is

Study

(SLS;

perhaps the most com-

prehensive source on adult age gradients of intelligence.

From young adulthood to old age,^ SLS display

the cross-sectional findings of the

continuous decrements for four mechanic abilconstructs:

ity

perceptual

speed,

inductive

reasoning, spatial orientation, and verbal ory.

In contrast, the

verbal

in

more pragmatic

mem-

abilities,

ability, show an young adulthood with an asymptote

knowledge and numerical

increase in

middle adulthood, followed by a plateau

modest decrements are discemable

The steady negative age for the

mechanics

tions are collectively referred to as the "prag-

rated by an

matics" of intelligence.'

evidence.

we summarize

Longitudinal

Seattle

Schaie. 1994, 1996)

in the

until

in old age.

differences observed

SLS

study are corrobo-

enormous mass of cross-sectional

In

their

large-scale

meta-analysis,

the available

Verhaeghen and Salthouse (1997) reported age

cross-sectional and longitudinal evidence per-

correlations of -.52 for processing speed, -.40

In this chapter,

taining to the

mechanic and pragmatic age gra-

dients in adulthood and old age. In this vein,

show

that

improvements

in the

we

understanding of

these gradients require efforts to

decompose

the

for reasoning, -.38 for spatial ability, -.33 for

episodic memory, and -.27 for short-term ory.

Importantly,

significant

were also observed,

mem-

nonlinear trends

at least for

reasoning, pro-

time/age dimension into constituent compo-

cessing speed, and episodic memory, suggesting

nents, such as terminal decline, selective attri-

accelerating age-related decline over the adult

tion, retest effects,

addition,

recent

interactions

and cohort differences. In

attempts to directly

model

between the mechanics and prag-

matics in old age are described (Ghisletta

&

Lindenberger, 2003), and the dedifferentiation

life

span. For the pragmatics, the cross-sectional

young adulthood followed by stability is present in most studies, but the age of peak performance differs somewhat pattern of growth during

across studies

(e.g..

Nilsson

et al..

1997: Park

Development of Intellectual Abilities 2002; Ronnlund, Nyberg, Backman,

et al.,

Nilsson, 2003). Figure

may

(2002),

12.1.

from Park

et

&

et al.,

al.

large

serve as an example of the typical

age gradients observed. The figure depicts cross-sectional age gradients based on multiple

Working memory, short-term memory, episodic memory, and processing speed display monotonic decline, whereas verbal knowledge shows increase indicators for five abilities.

•

205

2002) or are suggestive of relatively decrements appearing within, but not

&

before, old age (e.g., Baltes

1997; Lindenberger others at

Old Age

in

report

&

Lindenberger, 1994,

Baltes,

age-related

differences

about age 50 (Backman

&

Wechsler, 1997). In any case,

1997), starting

Nilsson, it

1996;

seems safe

to

conclude that the preservation of the pragmatics often observed for groups of young-olds does

or stability at least into the 70s. This inter-

not generalize to groups of old individuals

action constitutes the "classic aging pattern"

Backman

et al.,

(cf.

2000).

(Botwinick, 1977) or the "hold versus no-hold pattern" of adult intellectual development (e.g.,

W.

Jones

1949; Wechsler,

Hunt,

L.

&

Longitudinal Evidence

1955; see also

The

Conrad, 1933).

The major source of discrepancy cross-sectional

amount of

data concerns

decline in the

in available

onset

the

and

pragmatics during the to old age

longitudinal findings from the

show decelerating young adulthood, peaks in

construct

abilities

from young-old Backman, Small, Wahlin, & Larsson, 2000, for review; cf. Salthouse, 2003). Whereas some studies report small age differences during this

continuous decrements beginning

transition

period

(e.g.,

Baltes

&

Lindenberger,

Christensen, 2001; Nyberg et

al.,

1997;

2003; Park

study

mechanic and pragmatic age gradients. The latent

increases

(see

SLS

suggest a greater degree of similarity between

across

middle age, and accelerating declines

As

thereafter.

a notable exception, perceptual speed shows in early adult-

hood. The general pattern of rising and falling

span curves

life

is

discemable

in

most other

Working Memory Short-term

^—

—

*

Memory

Long-Term Memory

Speed

of

Processing

Verbal Knowledge

20

40

30

Figure 12.1

Cross-Sectional

Age

60

50

Age groups

(years)

Gradients for Composite Measures of Working Memory, Short-Term

Memory, Long-Term Memory, Speed of Processing, and Verbal Knowledge Source: Adapted with permission from Park

et al.

(2002). Copyright

(APA). Note: Each composite

is

constructed from three

tests.

©

2002 by

(n

the

= 345)

American Psychological Association

206

.

tt4NDBOOK OF UNDERSTANDING AND MEASURING INTELLIGENCE Zonderman.

Studies (e.g.. Giambra, .'Vrenberg.

Kawas.

&

Wilson

et al..

Costa. 1995: Ronnlund et

2002: Zelinski

&

2003:

al..

Stewan. 1998).

headings that have become increasingly familiar to researchers in the field: cohort differences,

and selective

retest effects,

attrition.

For the pragmatics, the bulk of evidence sug-

SLS

gests that the

observation of an onset of

decline in as early as middle adulthood atypical.

The dominant

pattern

is

is

one of remark-

ably stable, or even increasing, performance until

young-old or old age (Christensen. 2001:

Ronnlund

et

al..

2003; Singer, Verhaeghen.

& Baltes. 2003: Wilson

Ghislena. Lindenberger. et al..

At

2002). first

Cohort Differences

rather

glance, large amounts of evidence on

Substantial f)erformance improvements have

been observed on various

tests

functioning

past

example,

at

during

the

of intellectual century

— for

time of recruitment into military 1987:

1998: Raven.

service

(Flynn.

2000).

These "time-lagged" improvements,

Neisser.

together with the evidence of mostly positive

SLS

longitudinal developmental gradients of intelli-

cohort gradients provided the

gences are available. Consider, for example, an

1994. 1996). point to the potential severity of

and recent study by McArdle.

impressive

Hamagami. and Woodcock (2002). including individuals (n = 1.193) from early Ferrer-Caja.

By combining

childhood to old age. sectional latent

ics

and longitudinal information

in

a

growth cur\e modeling framework, the

authors to

cross-

showed

that different

growth cur\es had

be specified for the mechanics and pragmatof intelligence

(e.g.. for

Gf and Go.

This

evidence clearly underscores the multidirectionality

of intellectual de\elopment predicted by

tw o-component theories of cognition.

confounding

comparing age differences historical time.

later-bom cohorts

a single point in

they favor

if

(i.e..

a constant rate), cross-

at

may

sectional age gradients

overestimate the

magnitude of true age-related changes (see Baltes. Reese. illustration).

& Nesselroade.

Furthermore,

cohort differences

may

it

1977. for an early

conceivable that

is

originate

may

from multiple

operate additi\eh.

opposing directions, or interactively: and

they

may

differ

constellation

in

that

and nature

across cohorts and abilities. In the SLS. for

average

plotting

age

gradients,

cross-sectional or longitudinal, as a function of

chronological age ser\es imjX)rtant descripti\e

However,

purposes.

at

For example, when cohort gradi-

ents are linear and positive

in

Clearly,

w ithin-cohort de\ elopmental change secular trends when

between-cohort

with

sources: that the effects

Summary

(Schaie.

it

is

well

known

that

example, approximately continuous improve-

bom from 1907 to 1966 were found for inductive reasoning, verbal memory, and spatial orientation, amounting to ments across cohorts

standard deviation (SD) for the for-

a fallible and incomplete

around a

1.5

index of developmental change. To substan-

mer two

abilities. In contrast, for verbal ability,

chronological age

tively

interpret

is

and understand de\elopmental

of cognition, one must carefully

gradients

unpack the ingredients contributing

to a particu-

lar score, for a particular individual, at a partic-

ular point in time. Aside

from components with

a relatively strong intrinsic relation to chronological age, such as maturation

cognitive

and senescence,

developmental gradients are

influ-

numerical

ences

in

vein,

(

2(X)0 obser\ ed )

to

closely linked to chronological age. In the fol-

lowins.

we

discuss these influences under three

amount of data. effects

on

on the Ravens Progressive Matrices test Thomdike. 1977). Hence, if one were

distance from death, onset of pathology, and less

a great

modest cohort

the Mill Hill vocabulary scale but substantial

(see also

are

early in the

cohort effects, and then even disadvan-

when summarizing

Raven effects

of which

bom

tages for later-bom generations. In a similar

sources such as learning histon., health status.

all

and perceptual speed, there for cohorts

20th centur\. followed by a p)eriod of no differ-

enced by a wealth of additional de\ elopmental

nonnormative events,

ability,

was an advantage

adjust

cross-sectional

age differences for

cohort effects, these gradients would probably

change

in

ways

that

were more complex than a

constant reduction across abilities and ages.

Development of Intellectual Abilities (i.e.,

Retest effects denote the possibility that, in

repeated-measures designs, prior exposure to

may

alter

performance

at retest,

either through practicing task-relevant elements

of

or through a broad range of reactive

skill

effects such as general familiarization with the

negative) alter-

or (possibly

situation

testing

207

•

a composite of cued and free recall) in two population-based samples of adults from

Retest Effects

a test or task

Old Age

in

10 cohorts 35 to 80 years old at baseline (total A^

=

The measurement

1,788).

was

interval

5 years. Adjusting for retest effects critically

observed longitudinal gradients.

the

altered

Average

observed for episodic

effects

retest

memory were

reliable (.15 SD).

More

important,

of these effects accentuated the

statistical control

ations in motivation and interest. Thus, if retest

decline observed for older groups of individuals

operate in studies of intellectual age

and converted the continuous increments from

important to estimate their direc-

age 35 to age 60 to a pattern of stable perfor-

effects

gradients,

it is

magnitudes, and correlates. For example,

tions,

mance

across this age period. In contrast, retest

longitudinal findings showing a late onset of

effects for the semantic

decline in the mechanics might be a result of

ligible (.04

retest effects; that

decrements

in the

beyond

(e.g.,

the

may mask

these effects

is,

underlying ability dimensions level-practiced elements of

Retest effects have typically been regarded as threats to the internal validity of longitudinal studies.

However,

it

is

with

ated

6 Rabbitt,

1998; Wilson et

taking and

individual

differences,

may Lowe

to explicitly

together,

2002), and failure

al.,

instances (see also Rabbitt et

test

with

model them may underestimate the

magnitudes of average decline

designs.

problem structures may be associ-

systematic

may appear even

retest effects

long retest intervals such as 5 years, they

For

example, prior acquaintance with the nature of

raw longi-

tudinal curves of growth and decline in any major

way. Thus,

quite possible that they

operate in cross-sectional

also

were neg-

factor

affect the

vary in magnitude across abilities (see also

Salthouse, 1991, 2000).

skill; cf.

memory

SD) and did not

some

in

al.,

but not

evident that retest effects,

it is

may

unanalyzed,

all

2001). Taken

when

left

limit substantive conclusions

based on raw longitudinal data.

thereby constituting a major ingredient in cohort effects. In this respect, longitudinal

theoretically

superior because,

designs are

with proper

Selective Attrition

A common

design precautions and analysis techniques, retest effects

can be estimated and

statistically

controlled. In practice, however, this has rarely

been done (but see Lovden, Ghisletta, Lindenberger, in press; McArdle et Rabbitt. Diggle. Smith, Holland,

2001; Ronnlund

et

al.,

2002;

al.,

&

&

Mclnnes,

2003; Schaie,

1988,

1996; Wilson etal., 2002).

One method retest effects

on the sample

level is to

compare

the performance in a group of returnees with that

of a cohort-matched and not previously

tested

sample of individuals (Schaie, 1988).

If

attrition effects (see next section) are negligible or,

better

account,

still,

then

also estimated and taken into

differences

between samples

should reflect retest effects (and error). In a recent study using this method,

(2003) examined semantic

mance

(i.e.,

Ronnlund

memory

et al.

perfor-

a composite of verbal fluency and

knowledge) and episodic memory performance

who drop

is

out between measure-

ment occasions perform, on average,

at

lower

levels than returnees (e.g., Lindenberger, Singer,

& &

Baltes, 2002;

Bumight,

who drop decline

of estimating the magnitude of

finding in longitudinal studies

that individuals

to

during

2003; Zelinski

measurement intervals prior Bosworth & Schaie, 1999;

(e.g.,

& Wallace,

1979; Singer et attrition

et al.,

Furthermore, individuals

out often evince greater cognitive

dropout

Colsher

Ronnlund

1997).

effects

1991; Siegler

al.,

(i.e.,

variables of interest)

&

2003). Such attrition

is

may occur

Botwinick.

nonrandom related

to

for a variety

of reasons that can be grouped into two additive constituents:

mortality-related

experimental selectivity (Baltes

selectivity

& Labouvie,

and 1973;

for computational separation, see Lindenberger et al.,

2002).

Experimental selectivity occurs whenever individuals

who are alive but

unable or unwilling to

continue participation differ systematically from

208

HANDBOOK OF UNDERSTANDING AND MEASURING INTELLIGENCE

.

who do

those

participate. MortaUty-associated

on the other hand, occurs whenever

selectivity,

for old than for

individuals in close proximity to death (e.g..

who do

participants

not return because they are

young-old subgroups of the

total

sample. Clearly, selectivity limits the generalizability

of the results obtained

in longitudinal studies

deceased) differ on relevant attributes from

because the observ ed sample

individuals with a longer distance to death. In

sentative of the original sample. Experimental

may

principle, both types of selectivity attrition

refer to

selectivity poses a serious threat to the external

population validity of longitudinal findings as

selectivity poses an

and, subsequently, correlations of the variables

ability level, or both.

it

internal \alidity threat to longitudinal studies,

as

of the experimenter

reflects the inability

it

randomly assign surviving individuals

to

no longer repre-

change of

related to level of ability,

Whereas experimental

is

a returnee status, mortality-related selectivity

to is

To capture

a player in the population league.

KJeemeier

this population-heterogeneity aspect,

(1962) coined the term tenninal decline, argu-

may

affect the

under study 2000). tivity,

means, \ariances. covariances. Lindenberger.

(e.g.. Little.

The existence of

on the other hand,

found heterogeneity

&

Maier.

mortality-related selec-

suggestive of pro-

is

development and issues

in

warnings against the uncritical practice of averaging individual differences

in intelligence

Backman

a panicular age/time (cf

et al..

over

2000).

ing that aging-related changes in old age should

For example, when rigorous health screening

be distinguished from the changes associated

applied, old age groups

&

with impending death (see also Riegel

Riegel.

1972). Specifically, changes associated with ter-

minal decline

may

reflect other causal structures

than aging-induced changes

—

for example, spe-

cognition-influencing diseases and global

cific

may

of younger age groups, and the difference

tive

sample composition may

is

not be representa-

result in

in

an underesti-

mation of cross-sectional population gradients. That

is.

individuals sur\iving into late

life

who

are ready to participate in cross-sectional studies

may

constitute a select group in terms of multi-

1996:cf. Thaler. 2002).

ple,

correlated

Mounting evidence also points to an association between performance on intelligence tests

and cognitive

breakdowns of the biological system

and longevity

(e.g..

Singer

et al..

(e.g..

Berg.

2003: Small

Biickman. 1997). as well as between changes

and mortality

intelligence

Schaie.

&

1999: Maier

Bosworth

(e.g..

&

2003: see Berg, 1996; Small

et al..

(/?

=

that, in a

for the Berlin

the

first

Aging Study (BASE), those

in

when

(because mortality risk

al.

nme who

died within 6 years after cognitive performance

was

initially

initial

measured

(n

= 277) showed lower

performance compared to individuals

who were

still

alive

6 years

later (n

=

229).

Funhermore. greater a\erage decrements over 4 years

the

after the first

were observed for the

4

to

measurement occasion

indi\ iduals

6 years after the

initial

who

died during

measurement on

measures of perceptual speed and knowledge, as

compared

to those

who

survived and partici-

pated in the 6-year follow-up. Attrition due to factors other than mortality

was

(i.e..

related to only the initial le\

experimental)

el.

nitudes of both selectivitv effects

may

&

group of individuals

= 70-103) measured

516: age

the other hand,

also be overestimated, particularly in old age.

Backman.

1999. for overviews). For example. Singer et

(2003) reported

On

the magnitude of age-related decrements

&

1999: Singer

Smith.

socioeconomical. biological,

characteristics.

and the magwere greater

That

selectivity effects increase dramatically is

associated with age).

when averaging performance

is,

tively healthy individuals

of rela-

and individuals expe-

riencing terminal decline, aging-related changes

and dying-related changes are confounded. there

vein,

this

is

prolonged preclinical phase of dementia

when

(i.e.,

cognitive deficits are present but have

not yet have reached a diagnostic threshold: Elias et

al.,

Winblad.

&

cf.

2000: Small. Fratiglioni. Viitanen,

Backman. 2000). Because

dence of dementia age.

In

persistent evidence for a

including

an

is

the inci-

markedly higher

in old

unknown proportion of

individuals with preclinical signs of dementia

may

again

overestimate the

"pure"

aging-

intelligence relation (Slivvinski. Lipton. Buschke.

&

Stewart, 1996: but see

Wahlin. again, as ity

Small. is

&

Backman, Laukka,

Fratiglioni.

2002).

And

true for age-associated heterogene-

phenomena,

the admixture of individuals

with signs of terminal decline and preclinical

Development of Intellectual Abilities

dementia will increase variances and correlations, especially

in

Backman

age-heterogeneous samples

el al.,

2002; Sliwinski

(e.g..

sum, two opposing selectivity effects appear to operate in aging populations, one In

resulting in

tions in old age

owing

rela-

dementia

to preclinical

209

•

(differentiation) during maturation (e.g., Garrett,

becomes compressed

1946). and

(dedifferenti-

ated) again in old age (e.g.. Bakes, Cornelius,

et al., 1996).

exaggerated age-performance

Old Age

in

&

Nesselroade,

Spiro.

As

1970).

such,

1980; Reinert,

Willis,

hypothesis conveys

the

dynamic view of the structure of and

abilities

information-

underlying

the

a

intellectual

to

processing mechanisms. During childhood and

attenuated age-performance relations because

old age. the operations and expressions of the

and impending death and the other leading

may be more

individuals in older age groups

younger

positively selected than indi\iduals in

age groups

(cf.

Backman

2000). Clearly,

et al..

common

these considerations deconstruct the

assumed

diversity of cognitive abilities are

to

be subject to strong and fundamental systemic constraints

developmental status of

the

(i.e..

biological substrates of intelligence). In con-

assumption of a homogeneous population and

trast,

underscore the need to describe and explain the

common

during maturation and constraint

adulthood, this

in

relaxed, and other factors,

is

and occupational/

developmental characteristics of the individuals

such as

who compose

educational opportunities, occupy more promi-

the samples.

interest, motivation,

nent roles as determinants of intellectual devel-

opment, leading to greater diversity

Summary Apparently,

understanding

the

mental gradients of intellectual

abilities

is

a

impede

generalization across age groups and abilities,

and developmental changes appear

to

be hetero-

geneous by nature. Various influences may work simultaneously in opposing directions, and their

magnitude may

differ as a function of

study. Thus, an

improved understanding of the

age and

developmental gradients of intelligence requires consideration of

all

these factors in concert and,

as a best-case scenario,

latent

w ithin

method-

a unitar>'

framework. Recent developments

in

growth cur\e modeling and multilevel

modeling are promising candidates for such

frameworks (see Collins Liang,

&

Zeger,

& Sayer, 2001;

1994;

Little,

The dynamic view of

develop-

tricky business. Internal validity threats

ological

in levels

of

functioning in different abilities.

Diggle,

Schnabel.

&

early findings of a decrease in the

variance accounted for by g and decreases in

hood

to early maturity (e.g., Garrett, 1946), as

well as by corresponding findings of increases in old this

age

Reinert, 1970; see Chapter 14,

(e.g.,

volume, for a general treatment of factorial

invariance).

With respect

entiation in old age,

to

covariance dediffer-

more recent empirical

(e.g..

Bakes

&

Lindenberger,

Hertzog, Dixon,

Mitrushina Willis,

&

&

&

1991;

Satz,

Intrieri,

some

are

et al.,

1997; Hultsch,

Small, 1998; Li

&

1998; see Li

also

Lindenberger.

et al.,

(e.g.,

extent, the

somewhat mixed evidence methodological problems

ferentiation hypothesis

and the comparison of

Dedifferentiation Hypothesis

results across studies, including factors

Couched

restriction in range (e.g..

life

such as sample composition, unsuitable age groups, and

span intellectual

development, the differentiation/dedifferentiation hypothesis asserts that the functional organization of intellectual abilities

is

relatively

Juan-

2002). At least

surrounding the empirical testing of the dedif-

Old Age: The

terms of

2004;

Maitland,

some prominent

available

2002; Park

et al.,

Schaie,

may emanate from

to

in

evi-

dence has generally bolstered the early findings

Espinosa

in

from child-

correlations across different abilities

exceptions

Interactions Between Mechanics

of

structure

was sparked by amount of

hypothesis

dedifferentiation

1999, for an overview), but

Baumert, 2000).

AND Pragmatics

the

intelligence conveyed by the differentiation/

com-

pressed in childhood, undergoes decompression

An

Deary

et al.. 1996).

essential corollary of old-age dedifferen-

tiation is the hypothesis that the entire

intellectual abilities

by a

common

is

source

space of

increasingly dominated (or

an

ensemble of

210

.

HANDBOOK OF UNDERSTANDING AND MEASURING INTELLIGENCE

common

old-age decrements in pragmatic abilities are

performance levels for the pragmatics. However, in old age, negative age gradients generally prevail for both the mechanics and the pragmatics. Hence, in old age, the difference in

be induced by the mechanic decline

the directionality of the age gradients vaporizes

&

(for a longitudinal analog, see

Framed

sources) of intellectual development.

terms of the distinction between

in

mechanics and pragmatics of cognition,

the

assumed

to

(Ghisletta

Lindenberger,

2003).

other

In

Ronnlund

et al.,

words, the biological changes reflected by the

2003). This pattern, together with the findings

decline in mechanic abilities eventually limit

of covariance dedifferentiation,

the expression and accumulation of pragmatic

with the notion that the deteriorating functional

knowledge ties fall

in

old age

when

levels of

mechanic

the

below certain threshold

levels.

abili-

At higher

mechanic functioning, development of

the pragmatics

is

more dependent on and

experiential factors ical factors (but see

cultural-

dependent on biolog-

less

&

Hambrick

This dynamic account should perhaps be distinguished from the alternative nondynamic dedifferentiation notion of a

common

develop-

limit

to

starts

the

expression of the pragmatics in old age.

So

far,

evidence pertaining to the

direct

proposition that old-age decrements in prag-

matic abilities are driven by the mechanics

has

Engle. 2002).

mechanics

of the

level

consistent

is

been lacking.

Recently,

Lindenberger (2003)

dynamic

ing a recently developed structural equation

and

Ghisletta

lacuna by apply-

filled this

(i.e.,

lead-lag)

modeling method, the dual-

(DCSM; McArdle,

mental cause operating with constant force

change score model

throughout the adult

given that individual differences in rates of

McArdle & Hamagami. 2001; McArdle. Hamagami, Meredith. & Bradway. 2000). to

aging are present, rank ordering of individuals

combined longitudinal and cross-sectional data

should be

span.

life

increasingly

Specifically,

saturated

by aging-

related variance across age/time. Accordingly, if

age-related changes in different abilities are

common

driven by a

source

(i.e.. if

rates of aging

in different abilities are associated), correlations

across abilities should increase as a function of

time/age (Hofer

& Sliwinski, 2001; see also Li &

Schmiedek. 2002). Thus,

speaking, the

strictly

notion of an increased dominance of constraint

is

covariance

common

not necessary to explain a pattern of

However,

dedifferentiation.

the

BASE

from

and knowledge (Vocabulary and Spot-

Pictures)

a-Word) were used

DCSM

The

between

abilities

—

that

is,

among

participants

higher level of performance 1996; Spearman,

lower correlations

(e.g..

with a

Deary

et al.,

1927). Furthermore, as will

become evident below,

several

ings support the notion of a

more recent

common

find-

source that

linear models,

One

line of

ferentiation

evidence for the dynamic dedif-

account can be subsumed under

the heading of directionality dedifferentiation

(Baltes et

al..

2003). That

1998; Ghisletta is,

in the period

&

Lindenberger,

from young

to

considered as a variant

random

known

(LGM)

or

as hierarchical

coefficient models, or

mixed-effects models). Although these models

emanate from differ

different statistical traditions

in

down

to boil

to these

implementation

technical

to the

we

models as

same underlying

statistical

henceforth collectively refer

LGMs

(for detailed descrip-

tion of these models, see Collins

&

Sayer, 2001;

Diggle

et al.,

LGMs

estimate a time-based gradient for a

cific

1994; Little et

al.,

2000). In general,

group and represent individual

from

this gradient,

unobserved heterogeneity tories to error.

and

(e.g.,

& Ghisletta. in press), they all seem

model; therefore,

as deviations

varies in strength as a function of age.

may be

multilevel models (also

Lindenberger

non of divergence

mechanics and

growth curve models

of latent

being able to explain differentiation during mat-

2004) and the phenome-

to index the

the pragmatics, respectively.

may

al.,

= 516; age range = 70-104).

(n

Processing speed (Digit Letter and Identical

dynamic notion has the additional benefit of uration (e.g., Li et

2001;

sp)e-

trajectories

thereby allowing

in the individual trajec-

be represented rather than treated as

Extending these models to the multivariate

case allows for estimation of the extent to which levels

and changes are associated across

differ-

young-old age, the cross-sectional findings

ent variables, at the level of latent (true) scores.

summarized above reveal steady monotonic

The beauty of

decline in the mechanics but stable or increasing

extends the time/age-locked and symmetrically

the

DCSM.

however,

is

that

it

— Development of Intellectual Abilities

modeled associations between different variables (e.g., associations between slopes of change over 70

such as the one

60

conveyed by the dedifferentiation hypothesis.

DCSM,

other words, with the

low

that

the proposition

mechanic functioning

levels of

In

—and thus

temporally precede) decline is

amenable

answer

20

knowledge

different

.

80

by

from

that of

by

statis-

Knowledge

knowledge speed?

Ghisletta

-

60

-

"^"""^

-

and

o 50 o 2 40-

-

30 on 20

the lagger within this specific system of variat /-I

70

CD

cessing speed was the leader and knowledge was

processing speed

110

Age

Lindenberger (2003) clearly showed that pro-

is,

'

1

100

90

to the following

in processing

reported

results

ables; that

1

70

(i.e.,

speed on subsequent change

on subsequent change

The

-

30

the magnitude of the influence of

Is

level of processing in

''''''^'''^

-

pragmatics

to direct empirical testing

tically evaluating the

question:

in the

-

o 50 o ^ 40-

limit

drive

21

•

"I

the acquisition, expression, or maintenance of

the pragmatics in old age

Old Age

Speed

a certain time interval) to allow for empirical testing of lead-lag relations

in

'

70

80

100

90

time exerted

110

Age

a substantially stronger influence on change in

knowledge from

t-l to

r

than knowledge at t-\

did on subsequent change in processing speed

&

Gradients for Processing Speed and Knowledge

Lindenberger, 2003, for

statisti-

cal details). In other words, declines in

knowl-

(see Ghisletta

when and

edge,

if

they occur, are temporally

preceded by lower levels of processing speed (for similar results regarding younger participants,

McArdle

see

et al..

2000).

It

follows that, with

advancing age, interindividual variance

edge

increasingly

is

spjeed.

To

knowl-

by variance

saturated

in

suggesting that pragmatic abilities are

increasingly

ation.

in

composed of mechanic

we

return to the Singer et

al.

(2003) study,

516

sample was followed up two times, approx-

imately 4 and 6 years after the

Figure

12.2

depicts

initial

three

measure-

differently

calculated age gradients for processing speed

and knowledge, respectively. For our purposes, these measures

may

again serve as indicators of

the mechanics and pragmatics of cognition.

The

black lines represent the cross-sectional

gradient extracted

at

the

first

occasion for the longitudinal is,

Verhaeghen. Ghisletta,

Note: The thin black lines represent the cross-sectional gradient extracted

at the first

longitudinal sample (n

=

measurement CKcasion

132).

The dashed

for the

line describes the

corresponding cross-sectional age gradient for the initial

sample

-

{n

516).

The

the longitudinal convergence gradient

sectional

total

thick black line represents (i.e.,

combined

cross-

and longitudinal information) for the longitudinal

=

132).

further dissect the pattern of dedifferenti-

individuals ages 70 to 103 at baseline and that

thin

Singer,

Lindenberger. and Bakes (2003).

sample (n

Attrition." Recall that this study included

ment.

Source: Adapted from

variance.

described under the heading of "Selective

this

Three Differently Calculated Age

Figure 12.2

for those individuals (n

quently

survived

and

=

measurement sample

132)

who

participated

—

that

measurement

repeated individuals

are

highly

occasions.

select

owing

These to

both

experimental and mortality-associated reasons.

The dashed

line

describes the corresponding

cross-sectional age gradient for the total initial

sample (n = 516). This gradient presumably describes a sample including

more

participants

from cognition-associated health orders and terminal decline. The thick black suffering

disline

represents the longitudinal convergence gradient

(i.e.,

combined cross-sectional and

longitu-

dinal information) for the longitudinal sample. If

we

start

by considering the cross-sectional

gradient of the total sample,

it

is

clear that the

subse-

negative gradients prevail in both processing speed

the

and knowledge. In contrast, the cross-sectional

in

212

.

HANDBOOK OF UNDERSTANDING AND MEASURING INTELLIGENCE longitudinal

2003; see also Wohlwill. 1973: for a summary,

sample are more idiosyncratic: Knowledge remains stable, whereas processing speed

rumbles have been overheard by mainstream

decreases. In other words, there

developmental research.

gradients

describing

indicating a

select

the

is

Sample x Ability x Age

a pattern

interaction,

may be

suggesting that decline in the mechanics

normatively age related, whereas decline pragmatics, which presumably

in the

induced by

is

see Li. 2003). However, the echoes of their

cognitivist stereotypes,

In is

it

contrarv'

fact,

fair to

Skinnerian operant psychology, with

state its

to

that

careful

and manipulation of behavioral

description

repertoires o\ er time,

was much more germane

to

mechanic decline, may also be associated with

intraindividual research practice and theorizing

impending death (see also Small,

than most of today's developmental endeavors

&

von Strauss.

Fratiglioni,

Biickman. 2003). The shape of

the longitudinal gradient

is

with this conclusion. Specifically,

in

very old

age (> 90 years), negative gradients are for both processing speed this age,

it

likely that,

is

the effects of

e\ ident

and knowledge. At

even

(cf.

in a select

impending death are

sample,

large, relative

Baer. 1970; Richelle. 1993).

Two

largely consistent

interrelated insights are at the core of

the intraperson developmental research

agenda

(cf Chapter 14. this \olume: see Borsboom.

& van Heerden, 2003. for a review

Mellenbergh.

from a general perspective). interindividual

analyses of

First,

differences do not necessarily

aging-induced changes. Taken together, then,

inform the study of how

the pressure of biology-related factors, such as

within individuals. Recently, Molenaar. Huizenga.

specific diseases, in

and terminal decline increase

old age and may, at least in part, act as driving

forces of directionality dedifferentiation in old age. This conclusion

is

further bolstered by find-

ings suggesting that health-related and biologylinked variables such as sensory functioning

may

\'ariables are related

and Nesselroade (2003) elegantly addressed matical-statistical theory

a process having the to intra-

—

ergodicity

same

this

from mathe-

issue by applying a concept taken

—denoting

structure with respect

and interindividual variances. As a

rule

of thumb, a process can only be ergodic when

be related more strongly to cognitive function-

means, variances, and covariance are invariant

ing in old age. especially pragmatic abilities

across time.

(e.g..

Bakes

&

Lindenberger, 1997).

Intraindividual Patterns of

Change

As should be evident from the findings summarized above, developmental processes

related

to

godic.

From

are

intelligence this point

inherently

of view,

ble, if not likely, that analyses

seems

it

nonerplausi-

of interindividual

differences deliver results that are unrelated to It

is

a truism that the primary objective for

the structure of intraindividual differences. This

analogous to the conclusion

research on the aging of intelligence, and for

point

developmental psychology

aging-related changes from cross-sectional data

in general, is to

under-

is

made

stand the nature of developmental processes at

is

the level of individuals. Yet. standard multivari-

confounds

applied to hypotheses concerning

selectivity.

ate statistics

concerned with associations among variables across, rather than within, individuals.

long time,

this blatant

For a

discrepancy between con-

by interindividual-differences

the

in

form of cohort

The second message

development, such as standard cross-sectional

and longitudinal factor analyses, are notoriously

difficult

that inferring

effects

pertains to a critical

assumption of standard multivariate that

structural

and

among

relations

statistics:

variables

are

invariant across individuals. In light of the sus-

picion that developmental processes

may

lead

cepts and methods has been noted by a loosely

to interindividual differences in intraindividual

connected family of approaches, such as advo-

patterns of change, this assumption strikes a dis-

cates of a person-oriented view on development

cordant note.

(e.g..

Block, 1971; Magnusson. 1998; Valsiner,

1984); by

life

span theoreticians and methodol-

ogists (e.g.. Baltes et al.. 1977);

&

Nesselroade. 1979: Baltes

and by developmental systems the-

oreticians (e.2..

Lemer. Dowling.

&

Lara Roth.

Some

classics are worth revisiting

in this context. Baltes et al. (1977), for

suggested that the aim of tal

psychology should be

life

example,

span developmen-

to describe

and explain

intraindividual differences in behavior,

accompa-

nied bv the studv of interindividual differences

Development of Intellectual Abilities

and commonalities

in

these

intraindividual

changes. The developmental path of each individual tal

depicted as a multivariate developmen-

is

trajectory or, put ditferently, as a multivariate

pattern of intraindividual change.

drawn between studying ries

A

sharp line

is

intraindividual trajecto-

as the analytical building blocks

and the

study of the differences between them, implying that these differences

step

down on

might be substantial. To

the abstraction ladder, a similar

message appears every now and then as

critiques

dedifferentiation; that

sive

which

common

1985).

Such

gradients, exemplified in this chapter

trajectory.

The age

at

vary from person to person. Thus, an age-based

may

analysis

mask developmental

actually

terns that are truly

common

pat-

across individuals

because the biological timing of transition peri-

may

ods

ences

Age

differ across individuals.

in the onset

differ-

of adolescent growth spurts

example

sort (Wohlwill. 1970).

growth (see Estes,

"normally aging" and

developmental transition begins will

this

are the classical

of developmental

213

•

groups but denote, to some degree, different

time periods of a

across indi\ iduals in studies of learning and in

1956. for an early appearance; see also Hertzog,

Old Age

"terminally declining" individuals are not exclu-

against the heavy reliance on simple aggregates

studies

is,

in

for

phenomena of

this

Thus, to discern develop-

mental commonalities across individuals, one

may need

to substitute

age with neurobiological

and cultural-social constructs

that are close to

by the longitudinal and cross-sectional gradients

the developmental process of interest (see, e.g.,

of intellectual de\elopment. contain a mix of

Li

and interindividual variances and

intraindividual

mask

patterns of growth that are unique to the

individual

(cf.

& Nesselroade.

Jones

1990). In the

&

Schmiedek. 2002; Nesselroade

&

Schmidt

McCollam, 2000; Wohlwill. 1973). heterogeneity in patterns of change exists,

If

why

did standard multivariate techniques such

extreme case, one might end up with a gradient

as standard factor analysis not reveal

describing none of the individual growth curves

ence? As shown by Molenaar (1999; see also

make up the gradient. To illustrate, let us return

Molenaar

that

(2003) study for the

As deduced from

last

Singer

to the

time

(i.e..

et al.

Figure 12.2).

the absence of directionality

is

et al..

relatively

its

pres-

2003). standard factor analysis

insensitive to substantial

hetero-

geneity in patterns of intraindividual change; that

is.

solutions based on interindividual differ-

may

yield acceptable

dedifferentiation for the select cross-sectional

ences

sample, the pattern of dedifferentiation often

although the data correspond to individuals

observed

are very heterogeneous in terms of the structure

cross-sectional

in

studies

might be

driven by a subset of individuals in close prox-

of a person entering the termi-

nal decline phase

may move toward

a one-factor

to the

intellectual aging,

like to see increased use

one would

of methods that sepa-

solution (e.g., of measures of the mechanics and

rate intraindividual variability

interindividual differences, as well as

structure of a "normally aging" person of the

that

same age may be more

and transition periods

in patterns

the risk of sidestepping

many

differentiated. Statistical

made

up of mixtures of these two types of individuals

we

in these direcfions.

at

best,

mixture. Thus, a focus on the individual, or at

on

relatively

individuals, (e.g.,

is

homogeneous subgroups of to capture the dynamic

of intellectual development. if

some of the causal

of aging. At

useful approaches,

describe three broad avenues that might lead

Person-Oriented Approach With Pattern-Based Methods

needed

nonstationary. time-dependent) properties

Of course,

methods

an unknown

proportion of the individuals constituting this

least

and change from

have the potential to reveal heterogeneity

yield a picture of intellectual development in old

age that approximates,

who

In conclusion, to enhance the validity of our

knowledge about

the pragmatics), whereas the intellectual ability

parameters obtained from samples that are

data,

of their intraindividual changes.

imity to death. Put differently, the intellectual ability structure

fits

Instead of focusing on associations between variables,

pattern-based

relationships structures are

to

among

methods focus on aim

individuals, with the

group individuals on the basis of

indeed mortality related, one would expect most

Similarities and dissimilarities

individuals to sooner or later evince a pattern of

als are

similarity.

among

individu-

based on the profile of values on the

214

HANDBOOK OF UNDERSTANDING AND MEASURING INTELLIGENCE

.

to turn the

manage with this approach (see Bergman. 1998; Bergman et al.. 2003). Typically, classifications

conventional orientation of the cross-sectional

are carried out at each time, followed by the

under study. In terms of Cattell's

variables

(1952) data box. the general idea

is

data matrix 90 degrees, factoring people over variables

Q

(i.e..

technique), rather than factor-

ing variables over people

R

(i.e.,

technique).

study of individuals' group membership over

time with the aim to find typical and ami typical (cf von Eye, 1990) transition paths across the

Popular examples of model-based variants

groups established

include categorical factors, such as latent class

Bergman

analysis and

longitudinal extension, latent

its

When

transition analysis.

the observed variables

are continuous, the corresponding technique

called

Exploratory

analysis.

profile

latent

is

et al.,

at

each cross section

(e.g.,

2003). Panel designs lend them-

selves to this kind of analysis; for example, group

may map

transitions

theoretical model.

onto a stage-sequential

However, often,

at least

with

regard to intellectual development in old age.

pattern-based methods include configural fre-

transition

quency analysis (von Eye. 1990) and variants

process that deserves more direct and intense

of cluster analysis. Bergman, Magnusson, and

El-Khouri (2003)

may

be consulted for a com-

study.

might be a relatively slow and gradual

The next approach seems promising

in

this regard.

prehensive overview of the methods and the

behind a pattern-oriented

rationale

theoretical

approach, accompanied by an accessible descrip-

LGMs Allowing for Differences Parameters Between Subgroups

in the tion

of the technical

exploratory

(e.g..

implementation of the

in a closer

match between the

theoretical

model

and the measurement model. Specifically, principle,

there

is

no variance accounted

by the pattern-oriented approach

that

in

for

cannot be

explained by a standard variable-oriented factor

approach

(e.g.,

based methods

This approach merges the variable-oriented

non-model-ba.sed) alternatives.

The advantage of using a pattern-based approach lies perhaps not so much in statistics as

Horn, 2000). However, pattern-

may

offer a

more

direct reflection

LGM the

approach used

model-based

to structure

change with

methods

pattern-oriented

described above. Categorical latent variables are

used to describe groups of changing individuals

homogenous within and heterogeneous The notion is that each group

that are

across groups.

corresponds to a subpopulation with set

its

own

of parameters. These features are then com-

bined with the conventional growth modeling

of the theoretical suspicion that development

approach, which provides the parameters. To

may produce

recapitulate,

not only quantitative interindivid-

LGM

is

an attractive approach to

ual differences but also qualitatively different

analyze longitudinal data, relating an observed

developmental paths for different individuals. In

variable to time or to

other words, these methods

may provide a more way of focusing on the individual and. therefore, a way of grasping the extent and

such as age.

direct

estimated under the assumption that

nature of sample heterogeneity in patterns of

population, and a continuous latent variable

change.

used to capture any heterogeneity

A

further advantage of this approach

is

the

some time-related variable However, a mean growth curve is

uals in the sample are

all

drawn from

individ-

the

same is

in individual

trajectories of growth.

standard multivariate approaches offer ways to

The combination of the two approaches way to gauge the validity of the mean growth curve. By using an outcome variable

represent heterogeneity by including interaction

measured

at

terms or by employing multigroup factor models,

model

formed

the pattern-based approach offers a viable alterna-

correspond to different growth curves of the

capacity to structure and explore unobserved population heterogeneity'. Specifically, whereas

tive for structuring

unknown

geneity and exploring

its

many

sources of hetero-

Although the pattern-based approach offers a subgroups, the analysis of change

is

outcome

multiple time points, a latent class in

which the

latent classes

variable. In other words, individuals

are assigned to different groups based

correlates.

natural avenue for establishing

offers a

homogeneous is

not easy to

on

their

longitudinal trajectory. For example, one group

may have a linear growth, and may have a quadratic growth. Individual

of individuals

another

Development of Intellectual Abilities differences in the trajectories within the groups

by growth factor variances for each

are captured

group

(as

LGM). Muthen and

traditional

in

see also Muthen. 2001) provide

Muthen (2000;

in

Old Age

215

•

differences, individuals with similar patterns of

change can be aggregated into

intraindividual

homogeneous groups on (Nesselroade

tests

&

the basis of statistical

Molenaar, 1999). In

this

accessible overviews of this framework, and

way, generalization across individuals can be

Raudenbush (2001) describes similar

approached, and interindividual differences

tech-

niques within the multilevel modeling tradition.

in

intraindividual processes can be explored.

Multivariate,

Concluding Remarks

Replicated, Single-Subject,

Repeated-Measures Designs The methods described so

To capture

far are useful tools

for analyzing typical data sets in studies of intellectual aging

—

individuals,

relatively

that

data sets including

is,

few variables, and,

in

stand

if

among each

they do not differ

other only because they happen to be

bom

at

a

similar point in time nor construe intelligence as

a unitary and static construct. In younger adult-

hood, the biology-based mechanics and the

research design, as well as in data-

culture-based pragmatics of intelligence display

analytical procedure.

may

cal shift

individuals as

old age, one can neither treat

contrast,

the approach described shift

at

in

requires a profound

few measurement occasions. In

best, a

many

complexities of intellectual

the

development

As argued above,

this radi-

well be necessary to better under-

dynamics of

structural

the

now

intellectual

aging. Specifically, the designs and data-analytic

relatively loose couplings

age gradients. In

mechanics

late

and display different

senescence, decline in

limits pragmatic functioning, perhaps

The

due, at least in part, to terminal decline.

techniques described in this section are based on

onset and nature of this transition are likely to

the intensive longitudinal study of single indi-

differ

The degree of

viduals.

generalizability across

from person

in patterns

ties

to person.

General regulari-

of change are likely to exist

individuals can be assessed after patterns of

but can only be discerned with certainty after

change have been analyzed

intraindividual time-dependent structures have

level.

Nesselroade

1990; Nesseb-oade

and Molenaar strong

(e.g..

it

Molenaar

(e.g..

a

offers

intraindividual

Jones

&

Nesselroade,

& Schmidt McCollam,

advocates of

Because

at the intraindividual

to

to collect

several

many

variables

and individual-oriented perspective.

for

a

many times within may be to explore

of variables are related across time

single

individual;

may

that

is,

P-technique

be employed (variables over

Notes 1.

among

examine lagged

variables within persons (see

relations

Kim

&

Nesselroade, 2(X)3, for a recent application).

When vidual

the individual's structure of intraindi-

change has been clearly understood

without being contaminated by interindividual

studies

on

age have used verbal fact

we

join

those

(e.g., E.

Hunt, 2000) arguing for the use of a

wider range of pragmatic markers. 2.

to

in old

knowledge as a marker of the pragmatics. This

ment

analysis,

development

will be reflected in this chapter, although

com-

P-technique factor analysis, called dynamic

The overwhelming majority of

intellectual

bination of multivariate time-series analysis and

occasions). Molenaar (1985) suggested a

factor

age requires a multivariate, dynamic,

separate

Factor analysis

performed on the longitudinal data

factor analysis

intellectual develop-

approach.

clearly

in old

and interindividual differences,

individuals.

how groups

improved understanding of

ment

more widespread applications are desirable. Empirically, the main feature of this approach is

been observed, analyzed, and compared. Thus,

2003) are

et al..

general

this

way

2000)

In this chapter, in

we

discuss intellectual develop-

adulthood and old age from a psychometric

perspective. This perspective

is

complementary rather

than opposed to other perspectives, such as cognitivebehavioral

Chapter

1

(e.g., 1

,

childhood

this

Zacks. Hasher,

&

Li,

2000; see

volume, for an example concerning

development),

neurocognitive

Reuter-Lorenz. 2002), or neurophysiological Raz, 2000).

(e.g.. (e.g.,

216

HANDBOOK OF UNDERSTANDING AND MEASURING INTELLIGENCE

.

3.

Because any valid developmental stages are

both generation and person specific Smith, 2003), partitioning the

life

we

likely to

is

when nothing

a hopeless endeavor. Therefore, explicitly stated,

Baltes

&

span into substan-

based on chronological age

tive stages

is

(e.g..

be

else

use the labels young (ages

20-40). middle age (ages 40-60), young-old age (ages

Baltes.

B.. Lindenberger, U.,

P.

Life-span

(1998).

developmental

1029-1 143).

R

Baltes,

B.,

& Nesselroade.

rationale

replace chronological age with words.

Nesselroade

R.

J.

of longitudinal

&

(

P

Baltes.

A

L. (1996).

theory of adult intellectual

development: Process, personality,

and

interest,

knowledge. Intelligence, 22. 227-257.

Backman,

Laukka, E.

L.,

J.,

& Fratiglioni. L. (2002).

J..

Influences of preclinical

dementia and impending death on the magnitude of age-related cognitive deficits. Psychology and Aging. 17.

Backman,

435^W2.

&

L.,

memory

functioning across the adult

European Psychologist,

Backman,

L.,

(1996).

Small. B.

J..

1.

Semantic span.

life

27-33.

Wahlin, A.,

Larsson,

M.

(2000). Cognitive functioning in very old age. In

F

M. Craik

I.

&

T.

A. Salthouse (Eds.), The

handbook of aging and cognition (2nd ed., pp. 499-558). Mahwah, NJ: Lawrence Erlbaum.

M.

Baer. D.

An

(1970).

age-irrelevant concept of

development. Merrill-Palmer Quarterly.

16,

B. (1987). Theoretical propositions of

P.

On

span developmental psychology: ics

J.

the

life-

dynam-

between growth and decline. Developmental

Psychology Baltes,

P.

23,

&

Willis. S. L. (1980). Integration versus

differentiation of tluid/crystallized intelligence in

old age. Developmental Psychology. Baltes.

P

B..

&

opment of

Labouvie. G. V. intellectual

explanation,

M.

P.

Lawton

(

(Eds.),

6,

625-635.

1973). Adult devel-

performance: Description,

modification.

opment and aging

In

C.

Eisdorfer

&

The psychology of adult de\'el-

(pp. 157-219).

Washington. DC:

P.

B..

&

Lindenberger, U. (1997). Emergency

of a powerful connection between sensory and cognitive functions across the adult

A

R.

J.

methods.

Monterey,

&

Smith.

young old

(2003).

J.

From

New

dilemmas of

to the

frontiers in the

successful aging of the the fourth age.

Gerontology 49. 123-135. Berg.

Aging, behavior, and terminal

(1996).

S.

&

E. Birren

J.

W. Schaie

K.

(Eds.),

Handbook of the psychology of aging (4th pp. 323-337). New York: Academic Press. L. R. (1998).

A

ed..

pattern-oriented approach

to studying individual

development: Snapshots

and processes. In R. B. Cairns, L. R. Bergman,

&

Kagan

J.

(Eds.),

Methods and models

for

.uudying the individual (pp. 83-121). Thousand

Oaks, CA: Sage.

Bergman.

L. R..

Magnusson.

D..

&

El-Khouri. B. M.

Studying individual development

(2003).

in

new window

to the

Psychology and Aging.

life

span:

study of cognitive aging? 12,

approach. Mahwah, NJ: Lawrence Erlbaum. Block,

(1971). Lives through time. Berkeley,

J.

1

2-2

1

CA:

Bancroft. D.. Mellenbergh, G.

(2003).

The

J.,

&

van Heerden.

J.

theoretical status of latent variables.

Psychological Review, 110, 203-219.

&

Bosworth. H. B..

Schaie. K.

W.

(1999). Survival

effects in cognitive function, cognitive style,

sociodemographic Longitudinal

Study.

and

the

Seattle

Experimental

Aging

variables

in

Research. 25, 121-139.

Botwinick. J.

J.

E. Birren

(1977).

&

Intellectual

abilities.

K. W. Schaie (Eds.).

In

Handbook

of the psychology of aging (pp. 580-605).

American Psychological Association. Baltes.

B.,

Borsboom,

611-626.

B.. Cornelius. S. W.. Spiro. A.. Nesselroade,

R..

Nesselroade,

an interindividual context: A person-oriented

238-245. Baltes.

P

future of aging:

Bergman.

&

&

research

to

and develop-

York: Academic Press.

CA: Brooks/Cole.

decline. In

Nilsson, L.-G.

R.

J.

development psychology:

Life-span

Introduction

Baltes.

Wahlin. A.. Small, B.

New

Reese, H. W.,

B..

In

B. Baltes (Eds.), Longitudinal

P.

(pp. 1-39).

(1977). P.

York:

1979). Histor>' and

research.

research in the study of behavior

ment

Ackerman,

New

John Wiley.

60-80). and old age (age > 80) only as means to

References

Handbook of

(Ed.),

Theoretical models of

1.

(pp.

M.

Staudinger, U. in

Lemer

psychology. In R. M.

child psychology: Vol.

human development

&

theory

New Cattell.

York: Van Nostrand Reinhold.

R.

B.

(1952).

The

three

basic

factor-

analytic research designs: Their interrelations

and

derivatives.

499-520.

Psychological Bulletin, 49,

Development of Intellectual Abilities R.

Cattell,

(1971).

B.

Their

Abilities:

striutiire.

growth, and action. Boston: Houghton Mifflin. Christensen. H. (2001).

What

cognitive changes can

be expected with normal ageing? Australian

and New Zealand Journal of

Psychiatry,

35,

768-775. Collins,

&

M..

L.

Sayer. A.

New

(2001).

(Eds.).

methods for the analysis of change. Washington,

DC: American Psychological Colsher,

P. L.,

& Wallace,

Association.

R. B. (1991). Longitudinal

defined population of community-dwelling

e\der^.

Deary,

J.,

I.

C. R.,

Annals of Epidemiology, Egan,

&

V.,

215-230.

1,

Gibson, G., Austin, E.

Kellaghen. T. (1996). Intelligence and

M. F.

R.

F.,

Zeger, S. L. (1994).

of longitudinal

Oxford University Elias.

&

Liang, K.-Y..

for assessing

the

interdependence of ageing-

A.,

Au,

R..

Horn,

L. (1982).

J.

A

intelligence

cohort.

J.

tests

Bulletin, 101,

F

M. Craik

I.

Ghisletta,

really

IQ gains

&

in 14 nations:

measure. Psychological

Horn,

A developmental theory of intel372-378.

1,

Lindenberger, U. (2003). Age-

in

the

Berlin Aging

Study: Direct evidence for ability dedifferentiation in old age.

Manuscript submitted for

publication.

&. Costa,

span changes

in

P T,

Jr.

(1995). Adult

B., life

immediate visual memory and

verbal intelligence. Psychology

and Aging,

10,

123-139.

Hambrick, D.

&

Engle, R. W. (2002). Effects of

domain knowledge, working memory and age on cognitive f)erfonnance: tion

of the knowledge

is

An

capacity, investiga-

power hypothesis.

Cognitive Psychology, 44, 339-387.

Hebb. D. O. (1949). The organization of behavior.

New

Horn,

Handbook of

York: John Wiley.

(2000).

L.

J.

New

York:

theory.

B. B.

In

Wolman

intelligence (pp. 267-300).

Comments on

integrating

924-930.

24,

Horn.

&

L.,

J.

and

R.

Cattell.

57,

Horn.

J.

(1966). Refinement

B.

of the theory of fluid and crystallized

test

of Educational Psychology,

253-270.

&

L.,

Cattell. R. B. (1967).

Age

differences

and crystallized intelligence. Acta

fluid

Psychologica, 26, 107-129.

C. Dixon, R. A., & Small, B. J. Memory change in the aged. New York:

Hultsch. D. F, Hertzog.

Cambridge University gence. Learning

Press.

for crystallized intelli-

it

and Individual

Difference, 12,

123-129. Hunt, W. L. (1949). The relative rates of decline of

Wechsler-Bellevue "hold" and "don't hold"

Journal of Consulting Psychology,

&

J.,

Nesselroade,

J.

tests.

440-443.

13,

R. (1990). Mulfi-

variate, replicated, single-subject, repeated

mea-

sures design and P-technique factor analysis:

A

review of intraindividual change studies.

Experimental Aging Research. Jones, H. E.,

Z.,

Aging and

York: John Wiley.

Jones, C.

Giambra, L. M., Arenberg, D., Zonderman, A.

Kawas, C,

(Eds.),

Remodeling old models of

Gf-Gc

Hunt, E. (2000). Let's hear

based structural dynamics between perceptual speed and knowledge

Trehub

S.

(1985).

L.

J.

intelligence:

(1998).

American Psychologist,

R,

&

cognitive processes (pp. 137-178).

in

171-191.

Garrett, H. E. (1946). ligence.

Psychological

134-140.

R. (1987). Massive

crystal-

concepts of

to

cognitive psychology and aging in adulthood. In

intelligence. Journal

W. K. (1956). The problem of inference from

What IQ

relation

in

and

fluid

Alcoholism: Clinical and Experimental Research,

22-year

Framingham

curves based orb group data.

Flynn,

The theory of

White.

Archives of Neurology, 57, 808-813.

Bulletin, 53,

341-352.

related changes. Gerontology, 47,

person-centered and variable-centered research

D'Agostino, R. B. (2000). The preclin-

prospective study of the

Estes,

Under-

evaluation of research designs

on problems associated with the use of alcohol.

P

phase of Alzheimer's disease:

ical

(2001).

J.

An

York:

Press.

Beiser. A.. Wolf.

&

New

data.

M.

Sliwinski,

7-45.

7,

standing aging:

New

J.,

&

Hofer, S. M.,

105-132.

F

differences

individual

Research on Aging,

in gerontology.

(Ed.),

Analysis

217

•

perspective: Implications for cognitive research

the differentiation hypothesis. Intelligence, 23,

Diggle,

An

(1985).

Old Age

Plenum.

Brand,

J.,

C.

lized

application of cognitive function measures in a

Hertzog,

in

& Conrad, H.

16,

S. (1933).

decline of intelligence:

A

171-183.

The growth and

study of a homoge-

neous group between the ages of ten and Genetic Psychology Monographs,

Juan-Espinosa, Rebollo,

I..

M.,

Garcia,

Colom.

R..

L.

13,

F, Escorial.

& Abad. F

J.

(2002).

dedifferentiation hypothesis: Evidence

WAIS

III.

Intelligence, 30,

sixty.

223-275.

395-408.

S..

Age

from the

218

.

Kim,

HANDBOOK OF UNDERSTANDING AND MEASURING INTELLIGENCE

J. E.,

& Nesselroade,

among

R. (2003). Relationships

J.

and well-

social support, self-concept,

being of older adults:

dynamic

A

study of process using

factor models. International Journal of

Behavioral Development, 27. 49-65.

L.

&

Backman.

New

Nilsson (Eds.),

L.-G.

frontiers in cognitive aging. Oxford.

UK: Oxford

University Press.

Lindenberger, U., Singer.

&

T..

Longitudinal selectivity

R

Baltes.

B. (2002).

aging populations:

in

Kleemeier, R. W. (1962). Intellectual changes in the

Separating mortality-associated versus experi-

senium. Proceedings of the American Statistical

mental components in the Berlin Aging Study

Association,

290-295.

1,

Lemer, R. M., Dowling,

(BASE). Journal of Gerontology: Psychological

&

E.,

Lara Roth,

S. (2003).

Contributions of lifespan psychology

of developmental

elaboration

future

M. Staudinger

theory. In U.

&

to

the

systems

U. Lindenberger

Understanding human development:

(Eds.),

The

interplay

plasticity across levels:

of biology and culture

behavior across the

&

Baumert

A

computational exploration of the

between deterioration of neurotransmittor

systems and dedifferentiation of cognitive abihties in

old age. In L.-G. Nilsson

(Eds.),

&

H.

J.

Markowitsch

of memory

neuroscience

Cognitive

(pp. 103-146). Gottingen:

Hogrefe

&

Huber.

S.-C, Lindenberger, U., Hommel. B.. Aschersleben. G.. Prinz. W..

&

R

Baltes.

B. (2004). Lifespan

transformations in the couplings between intellectual

abilities

and

processes.

constituent

&

Schmiedek.

sarily aging:

F.

(2002).

Age

is

not neces-

Another step toward understanding

the "clocks" that time aging. Gerontology, 48,

Lindenberger. U. (2(X)1). Lifespan theories of cogni-

development. In N.

J.

Smelser

&R

&

Schnabel. K. U..

Little. T. D..

Baumert.

J.

(2000).

Modeling longitudinal and multilevel data.

Mahwah, NJ: Lawrence Erlbaum. Cognition

press).

(BASE): The psychology.

Lowe, C,

&

first

&

of the

ropsychology cal issues.

Lindenberger. U. (in

the

Aging Study

Berlin

Aging,

years.

ten

Neuro-

Cognition.

Rabbitt.

reliability

&

P..

in

R M.

A.. (1998). Test-re-test

CANTAB

ISPOCD

and

batteries: Theoretical

and

neu-

practi-

Neuropsychologia. 36. 915-923.

Magnusson. D. (1998). The

logic

and implication of a

person approach. In R. B. Cairns, L. R. Bergman.

&

J.

Kagan

(Eds.).

Methods and models for

studying the individual (pp. 33-63). Thousands

Maier, H.,

B. Baltes

& Smith, J. (1999).

of mortality

Psychological predictors

Journal of Gerontology:

in old age.

Psychological Sciences. 54B, P44-P54. J.

J.

(2001).

A

latent

score

difference

approach to longitudinal dynamic structural equation analyses. In R. Cudeck.

S. duToit,

&

of the social

D. Sorbom (Eds.). Structural equation modeling:

behavioral sciences (pp. 8848-8854). Oxford.

Present and future (pp. 342-380). Lincolnwood,

(Eds.). International encyclopedia

&

Mahwah, NJ:

(pp. 187-200).

Lawrence Erlbaum.

McArdle,

5-12.

tive

&

Modeling longitudinal and

Oaks, CA: Sage.

Psychological Science, 15, 155-163.

S.-C.

(Eds.).

Lovden. M.. Ghisletta.

Lindenberger. U. (1999). Cross-level

unification: link

Psychological

span.

life

171-194.

Bulletin. 129,

S.-C,

shaping the mind and

in

longitudinal

in

of continuers and

effects

dropouts. In T. D. Little. K. U. Schnabel,

multilevel data

York: Kluwer Academic.

opmental

Li.

the

New

of devel-

Li,

On

research:

Maier, H. (2000).

and generalizability

Selectivity

J.

Li. S.-C. (2003). Biocultural orchestration

&

Lindenberger, U.,

Little. T. D..

Dialogues with lifespan psychology (pp. 4 3^22). 1

Li,

Sciences, 57B, P474-P482.

UK:

IL: Scientific Software International.

Elsevier.

Lindenberger. U..

&

R

Baltes.

B. (1994). Sensory

functioning and intelligence in old age:

connection. Psychology

Lindenberger. U..

&

Baltes.

and Aging,

R

9,

A

strong

339-355.

B. (1997). Intellectual

functioning in old and very old age: Crosssectional results

from the Berlin Aging Study.

Psychology and Aging. Lindenberger. U..

12.

410-432.

& Ghisletta. P. (in press). Modeling

McArdle.

J.

J..

Ferrer-Caja. E..

Hamagami. F.

&

Woodcock. R. W. (2002). Comparative longitudinal structural analyses of the

of multiple intellectual

growth and decline

abilities

over the

life

span. Developmental Psychology. 38, 115-142.

McArdle,

J.

J..

&

Hamagami.

E

(2001).

Linear

dynamic analyses of incomplete longitudinal data. In L.

M.

Collins

&

A. Sayer (Eds.),

New

longitudinal changes in old age:

From covariance

methods for the analysis of change. Washington,

dynamic systems.

In R. A. Dixon,

DC: American Psychological

structures to

Association.

Development of Intellectual Abilities McArdle,

Hamagami. F. Meredith. W..

J.,

J.

Bradway, K.

(2000). Modeling

P.

&

dynamic

the

hypotheses of Gf-Gc theoiy using longitudinal

and Individual

life-span data. Learning

Differ-

ences. 12. 53-80.

&

M..

Mitrushina,

Nessekoade,

Stability

(1991).

P.

of

(2000).

Development. 24. 295-300.

Betula health,

M.

C.

P.

(1985).

A

dynamic

factor

model

of multivariate time series.

analysis

the

Psychometrica. 50. 181-202.

Molenaar, H.

Ader

J.

M.

C.

P.

Mellenbergh (Eds.). Research

in the social, behavioral,

and

life

sciences (pp. 143-167). London: Sage.

Molenaar. J.

C. M.. Huizenga. H. N..

P.

R.

The

(2003).

A

variability:

&

U. M. Staudinger

Emgrund,

L.,

et

cohort

prospective

K..

Nyberg,

study:

L..

The

(1997).

al.

Memory,

and aging. Aging, Neuropsychology, and

Cognition,

Nyberg,

1-32.

4.

Maitland, S. B., Ronnlund. M., Backman, L.,

(2003). Selective

et al.

adult age differences in an age-invariant multifactor

model of declarative memory. Psychology

and Aging.

C,

Park, D.

149-160.

18,

Lautenschlager, G., Hedden. T, Davidson,

&

R

Nesselroade,

N.

between the

Models of visuospatial and verbal memory

and intraindividual

theoretical and empirical vindica-

developmental

of

tion

&

relationship

structure of interindividual

L..

Dixon, R. A., Wahlin. A.,

(1999). Longitudinal analysis. In

& G. J.

methodology

Backman.

Adolfsson, R., Bucht. G.,

individuals. Brain Dysfunction, 4, 174-181.

for

Schmidt McCollam, K. M.

processes. International Journal of Behavioral

cognitive functions in young-old versus old-old

Molenaar.

219

•

Putting the process in developmental

Nilsson, L.-G..

Satz.

&

R.,

J.

Old Age

in

systems

theory.

In

Smith, A. D.,

S..

across the adult

life

Smith.

K. (2002).

span. Psychology

and Aging.

299-320.

17.

R M. A.,

Rabbitt,

Diggle, R, Smith, D., Holland,

P.,

&

Lindenberger (Eds.),

Mclnnes, L. (2001). Identifying and separating

Understanding human development: Dialogues

the effects of practice and of cognitive ageing

with

New

U.

psychology

lifespan

(pp. 339-360).

Muthen, B. (2001). Second-generation

structural

equation modeling with a combination of cate-

and continuous

gorical

opportunities

for

latent

class/latent

&

New

variables:

latent

modeling. In L. M. Collins

growth

A. Sayer (Eds.).

methods for the analysis of change Washington, DC: American

A^^K-

(pp. 291-322).

&

B..

Muthen. L. K. (2000). Integrating

person-centered and variable-centered analyses:

Growth mixture modeling with classes. Alcoholism: Clinical

latent trajectory

and Experimental

Research. 24, 882-891. Neisser, U. (Ed.). (1998).

gains

in

IQ and

The

related measures. Washington,

DC: American Psychological Nesselroade,

J.

comes of

R., &. Ford.

Neuropsychologia,

Raudenbush,

work

S.

W.

(2001).

comparing

for

change. In L.

M.

Toward

39,

a coherent frame-

trajectories of individual

& A.

Collins

Sayer (Eds.),

methods for the analysis of change

DC:

Washington,

American

New

(pp. 35-64).

Psychological

Association. J.

(2000).

Change and

The Raven's Progressive Matrices: stability

over cultures and time.

Cognitive Psychology, 41, 1-48.

Raz, N. (2000). Aging of the brain and

its

impact on

cognitive performance: Integration of structural

and functional findings. rising cun'e: Long-tertn

study of elderly

longitudinal

residents.

532-543.

Raven,

Psychological Association.

Muthen.

during a large

community

York: Kluwer Academic.

T.

In F.

I.

M. Craik

&

A. Salthouse (Eds.), The handbook of aging

and cognition (2nd

ed., pp. 1-90).

Mahwah,

NJ:

Lawrence Erlbaum.

Association.

D. H. (1985). P-technique

age: Multivariate, replicated, single-

Reinert,

G. (1970). Comparative factor analytic

studies of intelligence throughout the life span.

&P

Bakes

subject designs for research on older adults.

In L. R. Goulet

Research on Aging.

developmental psychology: Research and theory

Nesselroade.

J.

R..

&

7.

46-80.

Molenaar,

R

C.

M.

(1999).

Pooling lagged covariance structures based on short,

multivariate

dynamic factor

time-series

analysis

for

H.

Hoyle

(Ed.). Statistical strategies for small

sample

analysis.

research (pp. 223-250).

Sage.

In

R.

Newbury

Park.

CA:

(pp.

476-484).

Reuter-Lorenz,

mind and

P.

New

B.

(Eds.), Life-span

York: Academic Press.

A. (2002).

New

visions of the aging

brain. Trends in Cognitive Sciences. 6,

394^00. Richelle,

M.

news on

(1993).

Du nouveau

sur I'esprit? [Any

the mind?]. Paris: Presses Universitaires

de France.

220

HANDBOOK OF UNDERSTANDING AND MEASURING INTELLIGENCE

.

&

Riegel, K. F.

Riegel, R.

M.

(1972). Development,

drop, and death. Developmental Psychology,

6,

306-319. L..

Backman,

&

L..

Nilsson,

and decline

L.-G. (2003). Stability, improvement,

adult life-span development of declarative

memory: Cross-sectional and longitudinal data

from a population-based

Manuscript

study.

Salthouse, T. A. (1991). Theoretical perspectives on cognitive aging. Hillsdale. NJ:

Lawrence Erlbaum.

Salthouse, T. A. (2000). Methodological assumptions

aging research. In

in cognitive

F.

1.

M. Craik

&

A. Salthouse (Eds.). The handbook of aging

and cognition (2nd

467^98). Mahwah.

ed., pp.

NJ: Lawrence Erlbaum. Salthouse. T.

A.

(2003).

knowledge,

&

of aging,

Interrelations

performance.

In

Lindenberger (Eds.).

U.

Understanding human development: Dialogues

W.

L.

(1988).

of adult

studies

M.

(pp.

265-288).

York: Kluwer Academic.

K.

Schaie,

psychology

lifespan

New

Howe

Internal

cognitive

threats

validity

development.

in

In

C.

in

adulthood: Progress in cognitive

Brainard (Eds.), Cognitive

J.

development research (pp. 241-272).

New

York:

(1994).

development.

Intellectual

304-313.

development

in

Willis.

B..

L..

S.

&

(

&

C.

Botwinick.

J.

and Aging, (1979).

A

13,

8-20.

long-term

The matter of

selective subject attrition.

Journal of Gerontology, 34. 242-245.

in

T, Verhaeghen. R.

&

Baltes.

R

Ghisletta.

in

Psychological Science,

Small. B.

8,

to death

168-172. Winblad.

Fratiglioni, L.. Viitanen, M..

J..

Backman.

L. (2000).

impairment

B..

The course of cognitive Alzheimer's disease:

in preclinical

and 6-year follow-up of a population-based

3-

sample. Archives of Neurology, 57, 839-844. Small,

B.

Fratiglioni,

J.,

Backman.

performance

nitive

von Strauss,

L.,

E.,

&

and cog-

L. (2003). Terminal decline

very old age: Does cause

in

18,

193-202.

Spearman. C. (1927). The

abilities

of man. London:

Macmillan. Tetens,

N. (1777). Philosphische versuche iiber die

J.

menschliche

und

natur

and

development].

its

entwicklung

ihre

human

Leipzig:

nature

Weidmanns

Erben und Reich. Thaler.

D.

Design of an aging

(2002).

S.

brain.

Neurobiology of Aging, 23, 13-15.

Thomdike. R.

L. (1977). Causation of Binet

14,

Valsiner,

IQ decre-

197-202. J.

(1984).

frameworks

in

Two

alternative epistemological

psychology: The typological and

modes of

variational

5.

&

Verhaeghen. R.

thinking. Journal of

Mind

449^70.

Salthouse. T. A. (1997). Meta-

B. (2003).

The

the Berlin

Stewart.

W.

(1996).

The

B..

in

aging.

in

cross-classification.

Wechsler. D. (1955).

Buschke.

H..

&

effects of preclinical

Journal of Gerontology:

Psychological Sciences. 5 IB. P217-P225.

to

configural

frequency analysis: The search for types and

fate of cognition

dementia on estimates of normal cognitive functioning

von Eye. A. (1990). Introduction

Intelligence

Cambridge,

University Press.

Manual for

New

Scale.

the Wechsler Adult

York:

The Psycho-

logical Corporation.

Wechsler. R.

models. Psychological Bulletin.

UK: Cambridge

Aging, 18. 318-331. Lipton.

structural

122. 231-249.

antitypes

Aging Study (BASE). Psychology and

M.,

and

Lindenberger.

P..

very old age: Six-year longitudinal findings in

Sliwinski,

Time

and cognitive performance. Current Directions

Estimates of linear and non-linear age effects

C.

longitudinal study of intellectual ability of older

U.,

309-313.

12.

L. (1999).

1998). Longitudinal invariance of

adults:

Singer.

a

adults.

psychometric ability factor structures

across 7 years. P.sychology I.

from

old

analyses of age-cognition relations in adulthood:

Intrieri, R.

adult

Backman.

and Behavior

York: Cambridge University Press.

Schaie. K. W.. Maitland. S.

Siegler.

adult intellectual

The Seattle Longitudinal Study.

adulthood:

New

The course of

/\/?7fnca/j Ps}chologi.