genes & criminals

Children’s Behavioral Styles at

Age 3 Are Linked to Their

Adult Personality Traits at Age 26

Avshalom Caspi

Institute of Psychiatry, King’s College London and

University of Wisconsin-Madison

HonaLee Harrington

University of Wisconsin-Madison

Barry Milne, James W. Amell,

Reremoana F. Theodore

University of Otago

Terrie E. Moffitt

Institute of Psychiatry, King’s College London and University

of Wisconsin-Madison

ABSTRACT We observed 1,000 3-year-old children who exhibited five temperament types: Undercontrolled, Inhibited, Confident, Reserved, and Well-adjusted. Twenty-three years later, we reexamined 96% of the

This work was supported by grants from the National Institute of Mental Health

(MH-49414, MH-45070), the Medical Research Council, and the William T. Grant

Foundation. Terrie Moffitt is a Royal Society-Wolfson Research Merit Award holder.

The Dunedin Multidisciplinary Health and Development Research Unit is funded by

the Health Research Council of New Zealand. We are grateful to David Lykken,

Richie Poulton, and Auke Tellegen for their helpful comments on this manuscript, to

Phil Silva, founding director of the Dunedin Study, and to the study members and

their families. Address correspondence to Avshalom Caspi, Social, Genetic and

Developmental Psychiatry Research Centre, 111 Denmark Hill, Institute of

Psychiatry, London SE5 8AF, England.

Journal of Personality 71:4, August 2003.

Blackwell Publishing 2003

children as adults, using multiple methods of comprehensive personality assessment, including both self- and informant-reports. These lon- gitudinal data provide the longest and strongest evidence to date that children’s early-emerging behavioral styles can foretell their characteristic behaviors, thoughts, and feelings as adults, pointing to the foundations of the human personality in the early years of life.

‘‘Show me the child and I’ll show you the man.’’ Alas, the folk belief embodied in this aphorism is not widely shared by behavioral

scientists. Classic theories of personality are divided by competing views about the nature of continuity and change in development,

and clashing opinions are often impervious to data (Block, 1984; Kagan, 1996). Although longitudinal evidence now points to the

long-term continuities of personality characteristics from late childhood and adolescence to adulthood (Caspi, 1998; Shiner,

1998), as well as across lengthy periods of adult life (Roberts & Friend-DelVecchio, 2000), some reviewers claim that it is nearly

impossible to predict the course of adult personality from early childhood (Lewis, 1997). This empirical question has taken on new immediacy because human personality traits have become a primary

target of the research of geneticists (Hamer, 1997), psychiatrists searching for pharmaceutical means to alter personality (Cloninger,

1999), as well as early-childhood interventionists (Bruer, 1999). Such research could benefit from a solid knowledge base about the

natural course of human personality development. The goal of this article is to test the links between behavioral qualities observed at

age 3 and personality functioning measured at age 26. Eight years ago, we provided evidence that behavioral qualities

observed at age 3 were linked to personality functioning, as

measured via self-reports, at age 18 (Caspi & Silva, 1995). Since that time, study members in our longitudinal panel have experienced

profound life changes; the period between ages 18–26 is one of the most dynamic periods in the life course (Rindfuss, 1991). During

their 18th year of life, the vast majority of our study members were leading an adolescent lifestyle: 85% of them were still living on the

South Island of New Zealand where they were born, 77% were residing in their parents’ house, and 70% were enrolled in school. By

age 26, our study members had left adolescence behind: 79% were working full time; 59% were living with an intimate partner; 32% had moved and worked overseas; 22% had become parents; 22%

496 Caspi et al.

had completed a university degree; and 21% had bought a house. By

reexamining connections between age 3 behavioral styles and personality functioning at age 26, we are on more solid grounds in

empirically testing the hypothesis that the ‘‘child is father of the man.’’

The research reported here was designed to overcome three practical and methodological obstacles to testing whether or not

children’s early-appearing behavioral differences can foretell their behaviors, thoughts, and feelings as adults (Magnusson & Bergman,

1990). First, a large sample is required to test hypotheses about the nature of connectedness in personality development because effect sizes associated with developmental continuities will be small to

moderate, given the many intervening experiences between infancy and adulthood. Second, it is imperative to sample the right people,

not just a lot of people, because biases in sample selection threaten both the internal and external validity of findings about longitudinal

associations. Although infrequently used in psychological research, general population samples offer important advantages for studying

developmental continuities because they yield accurate and general- izable estimates of the strength of associations across time. Third, it is necessary to wait many years for study members to grow up and

to ensure that, when followed up, the sample has not been ravaged by attrition. When individuals are non-randomly lost from a

longitudinal study, the validity and generalizability of conclusions about developmental continuities are compromised. We address

these three challenges in the context of the Dunedin Study, a longitudinal-epidemiological investigation of some one thousand

children born in 1972–1973. Since then, the Dunedin Study has suffered very little attrition, and 96% of the original study members

participated in the most recent assessment conducted when they were 26 years old.

The research reported here also included measurement strategies

to overcome a fourth obstacle in studying personality continuities. Many longitudinal studies rely on reports repeatedly obtained from

the same rater at different ages. For example, longitudinal studies of children often rely on reports provided repeatedly by mothers. Using

these reports may inflate estimates of continuity because observed continuities may reflect continuities in maternal characteristics as

well as in children’s characteristics (Bates, 1994). Other studies of personality continuity rely only on the repeated administration of

From Age 3 to Age 26 497

self-report inventories, which, because of self-presentational biases,

provide an incomplete perspective on continuities in personality development ( John & Robins, 1994).

Multimethod assessments are the preferred strategy for testing the coherence of personality across the life course. We address this

measurement challenge by studying the adult personality outcomes of distinct groups of children whose behavior was first observed and

recorded by examiners when the children were 3 years old. Twenty- three years later, we obtained self-reports and informant reports

about the children grown up, using two of the best-known contemporary trait models of personality: the Tellegen and the five-factor models of personality structure (Church & Burke, 1994;

McCrae & Costa, 1997). The present study thus offers one of the strongest tests to date of the continuity hypothesis and allows for

confident generalizations about the extent to which individual differences in personality are preserved across development.

METHOD

Sample

Participants are members of the Dunedin Study, a longitudinal

investigation of health and behavior (Silva & Stanton, 1996). The cohort of 1,037 children (52% male) was constituted at age 3 when

the investigators enrolled 91% of the consecutive births between April 1972 and March 1973 in Dunedin, New Zealand. Cohort

families represent the full range of socioeconomic status in the general population of New Zealand’s South Island, and they are

primarily white. Follow-ups have been carried out at ages 5, 7, 9, 11, 13, 15, 18, 21, and, most recently, at age 26 when we assessed 980

(96%) of the 1,019 study members still alive.

Measures

Temperament Types in Early Childhood. At age 3, each study child

participated in a 90-minute developmental testing session adminis- tered by examiners who had no knowledge of the child’s prior

behavior history. Following the testing, each child was rated by the examiner on 22 behavioral characteristics that were derived from

those used in the Collaborative Study on Cerebral Palsy, Mental Retardation, and Other Neurological Disorders of Infancy and

498 Caspi et al.

Childhood (see Goldsmith & Gottesman, 1981) and are similar in

scope to the behavior ratings contained in Bayley’s Infant Behavior Record (Matheny, 1980). Complete data were available for 1,023

children. We have been unable to locate information about the reliability of these single-assessment examiner ratings, completed in

1975. Rather than abandon these ratings, we carried out a contemporary study of interrater reliability on a sample of 83 4-

year-old children who, just like the Dunedin children, were administered various cognitive tests and whose behavior was then

rated using the same characteristics evaluated in the Dunedin study. Interrater reliabilities ranged from .7 to .9. This does not establish that the Dunedin ratings were reliable, but it does demonstrate they

could easily have been so. Factor and cluster analysis procedures were used to identify five

reliable types of children (for details, see Caspi & Silva, 1995). The Well-adjusted type (n 5 405; 48% male) included children who were

capable of self-control when it was demanded of them, were adequately self-confident, and who did not become unduly upset

when confronting the examiner in a novel situation. The Under- controlled type (n 5 106; 62% male) included children who were impulsive, restless, negativistic, distractible, and labile in their

emotional responses. The Confident type (n 5 281; 52% male) adjusted to the testing situation quickly; they were zealous,

exceptionally friendly, somewhat impulsive, eager to explore the testing materials, and displayed little or no concern about

separating from their caregiver. The Inhibited type (n 5 80; 40% male) included children who were socially reticent, fearful, and

easily upset by the examiner. The Reserved type (n 5 151; 48% male) were timid and somewhat uncomfortable in the testing session;

however, unlike inhibited children their shyness was not extreme and their caution did not interfere with their task orientation.

Personality Assessments in Adulthood. At age 26, we brought each

participant to the research unit within 60 days of their birthdays for a day-long health assessment, in which various interviews and

examinations (e.g., physical exam, personality assessment, dental exam) were conducted in counterbalanced order by different trained

personnel. Participants completed the Multidimensional Personality Questionnaire (MPQ; Tellegen et al., 1988; Patrick, Curtin, &

From Age 3 to Age 26 499

Tellegen, in press) which provides, for each person, a profile of

scores on 10 distinct personality traits that define three superfactors of personality: Negative Emotionality, Constraint (vs. Disinhibi-

tion), and Positive Emotionality (see Table 1). MPQ protocols were available for 975 of the 980 study members who participated in the

age 26 assessment. At age 26, we also asked the study members to nominate someone

who knew them well. These ‘‘informants’’ were mailed question- naires asking them to describe the study members using a brief

version of the Big Five Inventory (BFI; Benet-Martinez & John, 1998) which assesses individual differences on the five-factor model of personality: Extraversion (e.g., ‘‘Makes things exciting,’’

alpha 5 .79), Agreeableness (e.g., ‘‘Cold and distant with others,’’ alpha 5 .75), Conscientiousness (e.g., ‘‘Works until a thing is done,’’

alpha 5 .81), Emotional Stability (e.g., ‘‘Is relaxed, handles stress well,’’ alpha 5 .83), and Openness to Experience (e.g., ‘‘Likes to

reflect and play with ideas,’’ alpha 5 .85). Informant data were obtained for 946 (96%) of the 980 study members who participated

in the age 26 assessment. Most informants were best friends, partners, or other family members.

The correlations between self- and informant-reports of person-

ality are shown in Table 2 (cf. John & Robins, 1993). The pattern of convergent/discriminant correlations provides evidence regarding

the validity of the personality reports.

RESULTS

We examined personality differences among the five groups of

children using analysis of variance. (We examined temperament group � gender interactions for each of the age 26 MPQ and Big Five scales. Only one of 18 interaction terms was statistically significant at conventional levels. Because of this gender similarity, we report the analyses for males and females together.) All adult

personality variables were standardized to the same scale using the z- score transformation. Each variable thus had a mean of 0 and a

standard deviation of 1; effect sizes, expressed in standard-deviation unit differences, can be calculated from the presented data, where

d 5 .2 is a small effect size, d 5 .5 a medium effect, and d 5 .8 a large effect (Cohen, 1992).

500 Caspi et al.

Table 1 Multidimensional Personality Questionnaire (MPQ) Scale

Descriptions

MPQ Scale Alpha Description of a High Scorer

Negative Emotionality

Alienation .83 Views the world in malevolent

terms; expects mistreatment and

betrayal; feels a victim of bad luck.

Stress Reaction .83 Tends to experience frequent and

intense negative emotions,

including anxiety, distress, and

anger; overreacts to minor events.

Aggression .81 Willing to hurt others for own

advantage; enjoys frightening and

causing discomfort to others.

Constraint vs. Disinhibition

Traditionalism .74 Conventional, moralistic; expresses

authoritarian beliefs and attitudes;

desires a conservative social

environment.

Harm Avoidance .79 Avoids excitement, thrills, and

danger; prefers safe activities even

if dull or tedious.

Self-Control .81 Is reflective, cautious, careful,

rational, planful; not impulsive.

Positive Emotionality

Social Potency .78 Is forceful, decisive; dynamic, fond

of influencing others and of

leadership roles.

Achievement .75 Works hard; ambitious; enjoys

demanding projects.

Well-Being .75 Has a happy, cheerful disposition;

feels good about self and sees a

bright future.

Social Closeness .80 Is sociable, likes people, and turns

to others for comfort.

The Absorption scale was not included in the version of the MPQ administered in

the Dunedin Study.

From Age 3 to Age 26 501

T a

b le

2 C

o rr

e la

ti o

n s

B e

tw e

e n

S e

lf -R

e p

o rt

s o

n th

e M

u lt

id im

e n

si o

n a

l P

e rs

o n

a li

ty Q

u e

st io

n n

a ir

e (M

P Q

) a

n d

In fo

rm a

n t

R e

p o

rt s

A b

o u

t th

e B

ig F

iv e

P e

rs o

n a

li ty

F a

c to

rs .

B o

th S

e lf

- a

n d

In fo

rm a

n t

R e

p o

rt s

W e

re C

o ll

e c

te d

W h

e n

th e

S tu

d y

M e

m b

e rs

W e

re A

g e

2 6

Y e

a rs

(N 5

9 4

2 )

B ig

F iv e S c a le s

E x tr a v e rs io n

A g re e a b le n e ss

C o n sc ie n ti o u sn e ss

N e u ro ti c is m

O p e n n e ss

to E x p e ri e n c e

M P Q

P ri m a ry

T ra it s

A li e n a ti o n

� .1 2

� .2 7

� .2 6

.2 7

� .1 4

S tr e ss

R e a c ti o n

� .0 5

� .0 9

� .0 7

.4 1

� .0 2

A g g re ss io n

� .0 7

� .3 2

� .2 1

.0 6

� .0 3

T ra d it io n a li sm

.0 3

.1 7

.2 0

� .0 4

� .1 6

H a rm

A v o id a n c e

� .0 8

.1 4

.2 2

.1 2

� .0 8

S e lf -C

o n tr o l

� .0 5

.1 2

.3 8

� .0 3

� .0 6

S o c ia l p o te n c y

.3 1

� .0 1

� .0 1

� .1 1

.2 5

A c h ie v e m e n t

.0 6

� .0 1

.1 7

� .0 6

� .0 8

W e ll -B e in g

.1 7

.1 3

.0 6

� .2 3

.1 3

S o c ia l C lo se n e ss

.2 8

.2 5

.1 5

� .0 8

� .0 5

M P Q

S u p e rf a c to rs

N e g a ti v e E m o ti o n a li ty

� .1 0

� .2 8

� .2 2

.3 7

� .0 7

C o n st ra in t v s. D is in h ib it io n

.0 2

.1 9

.3 6

.0 3

� .1 3

P o si ti v e E m o ti o n a li ty

.3 3

.1 4

.1 4

� .1 9

.1 7

Table 3 shows the personality profiles of the age 3 temperament

groups when they were assessed at age 26 with the MPQ. The upper panel of the table reveals that the temperament groups differed

significantly on 5 of the 10 MPQ primary traits and, marginally, on a sixth trait. Each of these findings was followed by multiple

comparison tests to highlight where the significant group differences occurred. We describe these differences for each MPQ scale.

Alienation. A high score on this MPQ scale indicates a propensity to view the world in malevolent terms. High scorers on this scale believe that they are the victims of bad luck and that other persons

wish them harm; they feel mistreated, deceived, and betrayed by others and by life’s circumstances. The Undercontrolled children

scored highest on this scale.

Stress Reaction. A high score on this MPQ scale indicates a tendency to experience frequent and intense negative emotional

states, including anxiety, distress, and anger. The Well-adjusted children scored lowest on this scale and were most likely to quickly

get over upsetting experiences. The Undercontrolled children scored highest on this scale and acknowledged that they often responded with strong negative emotional reactions to many ordinary

circumstances.

Traditionalism. A high score on this MPQ scale is associated with conventional, moralistic, and authoritarian beliefs and attitudes.

For example, high scorers endorse strict child-rearing practices, are intolerant, and favor restrictions on freedom of expression. Under-

controlled children scored highest on this scale in adulthood, whereas the Confident children scored the lowest and grew up to be the least conventional adults, evincing relatively little concern about

what is ‘‘proper.’’

Harm Avoidance. A high score on this MPQ scale indicates a propensity to shun physically dangerous and thrilling experiences

and to prefer, instead, safe and dull activities that do not risk injury. The Inhibited children scored highest on this scale, followed by the

Reserved children, and both of these groups diverged emphatically from the Undercontrolled children, and, to a lesser extent, the

From Age 3 to Age 26 503

T a

b le

3 T e

m p

e ra

m e

n t

T y

p e

s Id

e n

ti fi

e d

a t

A g

e 3

H a

v e

D is

ti n

c t

P e

rs o

n a

li ti

e s

a t

A g

e 2

6 (B

o tt

o m

R o

w s)

a s

T h

e y

D id

a t

A g

e 1 8

(T o

p R

o w

s) ,

a s

S e

e n

T h

ro u

g h

S e

lf -R

e p

o rt

s o

n th

e M

u lt

id im

e n

si o

n a

l P

e rs

o n

a li

ty Q

u e

st io

n n

a ir

e (M

P Q

) P

ro v

id e

d b

y th

e C

h il

d re

n G

ro w

n U

p

T e m p e ra m e n t G ro u p s a t A g e 3

W e ll -A

d ju st e d

(n 5

3 6 6 – 3 8 0 )

C o n fi d e n t

(n 5

2 6 3 – 2 7 0 )

U n d e r-

c o n tr o ll e d

(n 5

9 2 – 9 7 )

R e se rv e d

(n 5 1 3 7 – 1 4 2 )

In h ib it e d

(n 5

7 2 – 7 6 )

F R a ti o fo r

T e m p e ra m e n t

A g e 1 8 to

A g e 2 6 S ta b il it y

C o rr e la ti o n b

M P Q

P ri m a ry

T ra it s

A li e n a ti o n

A g e 1 8

� .0 8 a

� .0 2 a

.3 8 b

� .0 9 a

.1 1 a b

3 .6 5 n n

.5 9

A g e 2 6

� .0 7 a

� .0 2 a

.3 4 b

� .1 0 a

.1 2 a b

4 .2 0 n n

S tr e ss

R e a c ti o n

A g e 1 8

� .0 5

� .0 5

.1 7

.0 8

� .0 3

1 .2 4

.5 2

A g e 2 6

� .1 0 a

.0 4 a b

.1 5 b

.0 3 a b

.0 8 a b

1 .9 0 w

A g g re ss io n

A g e 1 8

� .0 4 a

.0 8 a b

.2 2 b

� .0 9 a c

� .2 8 c

3 .4 1 n

.6 0

A g e 2 6

� .0 6

.0 5

.1 4

� .0 7

.0 2

1 .2 1

T ra d it io n a li sm

A g e 1 8

.0 6

� .0 9

� .0 2

.0 5

.0 8

.8 7

.5 5

A g e 2 6

.0 5 a

� .1 7 b

.1 5 a

.0 5 a

� .0 2 a b

2 .9 7 n

H a rm

A v o id a n c e

A g e 1 8

� .0 6 a b

� .0 1 a c

� .2 2 b c

.0 8 a

.4 0

4 .7 3 n n

.6 2

A g e 2 6

� .0 1 a b

� .0 8 a b

� .1 8 a

.0 7 b

.4 2

4 .8 1 n n

S e lf -C

o n tr o l

A g e 1 8

.0 8 a

� .1 4 b

� .1 9 b

.0 8 a

.1 8 a

3 .4 3 n n

.5 8

A g e 2 6

.0 8

� .0 8

� .1 3

.0 2

� .0 3

1 .5 4

S o c ia l P o te n c y

A g e 1 8

.0 8 a

.0 7 a

� .0 1 a b

� .1 5 b c

� .3 7 c

4 .2 0 n n

.5 6

A g e 2 6

.0 1 a

.1 6

� .0 5 a b

� .1 3 a b

� .3 1 b

4 .4 2 n n

A c h ie v e m e n t

A g e 1 8

.0 6

� .0 5

� .0 4

� .0 6

� .1 5

.9 5

.4 4

A g e 2 6

.0 5 a

.0 6 a

.0 1 a b

� .2 0 b

� .1 4 a b

2 .3 8 n

W e ll -B e in g

A g e 1 8

� .0 2

.0 7

� .0 9

.0 1

� .0 6

.6 5

.4 3

A g e 2 6

� .0 1

.0 6

� .0 3

.0 6

� .2 7

1 .8 7

S o c ia l C lo se n e ss

A g e 1 8

.0 5

� .0 3

� .0 3

� .0 4

� .0 2

.4 0

.5 1

A g e 2 6

.0 4

� .0 1

� .1 5

.0 1

.0 2

.7 1

M P Q

S u p e rf a c to rs

N e g a ti v e

E m o ti o n a li ty

A g e 1 8

� .0 7 a

.0 4 a

.3 3

� .0 3 a

� .0 6 a

3 .2 0 n

.6 0

A g e 2 6

� .1 1 a

.0 3 a

.2 7 b

� .0 5 a

.1 0 a b

3 .4 6 n n

C o n st ra in t v s

D is in h ib it io n

A g e 1 8

.0 3 a d

� .1 1 a b

� .2 1 b

.1 0 c d

.3 1 c

3 .9 5 n n

.6 7

A g e 2 6

.0 5 a

� .1 4 b

� .0 9 a b

.0 7 a

.1 8 a

2 .6 9 n

P o si ti v e E m o ti o n a li ty

A g e 1 8

.0 7 a

.0 5 a

� .0 7 a b

� .1 1 a b

� .2 7 b

2 .5 4 n

.5 4

A g e 2 6

.0 3 a b

.1 1 a

� .0 9 a b c

� .1 1 b c

� .2 9 c

3 .1 7 n

N o te : T h e lo w e r-

a n d u p p e r- b o u n d ra n g e s o n th e g ro u p s’ sa m p le

si z e s in d ic a te

th e n u m b e r o f st u d y m e m b e rs

a b o u t w h o m

w e o b ta in e d

M P Q

d a ta

a t a g e 1 8 a n d a g e 2 6 , re sp e c ti v e ly . T h e ta b le

sh o w s Z -s c o re s (M

e a n 5

0 ; S ta n d a rd

D e v ia ti o n 5 1 ), st a n d a rd iz e d o n th e fu ll

sa m p le . G ro u p s th a t sh a re

a su p e rs c ri p t d o n o t d if fe r fr o m

e a c h o th e r;

n n p o

.0 1 , n p o .0 5 , w p o .1 0 .

a C a sp i &

S il v a (1 9 9 5 ) re p o rt e d fi n d in g s b a se d o n 8 6 2 M P Q

p ro to c o ls

c o m p le te d a t th e a g e 1 8 a ss e ss m e n t.

A ft e r th e a ss e ss m e n t,

w e

o b ta in e d M P Q s fr o m

a n a d d it io n a l 7 6 st u d y m e m b e rs . T h e fi n d in g s re p o rt e d h e re

a re

b a se d o n th e 9 3 8 p ro to c o ls .

b A ll st a b il it y c o e ffi c ie n ts

a re

si g n ifi c a n t a t p o .0 0 1 . F o r fu rt h e r d e ta il s a b o u t p e rs o n a li ty

c o n ti n u it y a n d c h a n g e fr o m

a g e 1 8 to

a g e 2 6 , se e

R o b e rt s, C a sp i, &

M o ffi tt , 2 0 0 1 .

T a

b le

3 (c

o n

ti n

u e d

)

W e ll -A

d ju st e d

(n 5

3 6 6 – 3 8 0 )

C o n fi d e n t

(n 5

2 6 3 – 2 7 0 )

U n d e r-

c o n tr o ll e d

(n 5

9 2 – 9 7 )

R e se rv e d

(n 5

1 3 7 – 1 4 2 )

In h ib it e d

(n 5

7 2 – 7 6 )

F R a ti o fo r

T e m p e ra m e n t

A g e 1 8 to

A g e 2 6 S ta b il it y

C o rr e la ti o n b

Confident children, who, as adults, said they enjoyed dangerous and

exciting experiences and activities.

Social Potency. A high score on this MPQ scale indicates a propensity to be forceful and decisive. The Inhibited and Reserved

groups of children scored lowest on this scale and said they preferred others to take charge and that they did not enjoy being the center of

attention. The Confident children scored significantly higher than every other group of children on this scale, and, as adults, wished to

assume leadership roles and described themselves as vigorous, forceful, and dynamic.

Achievement. A high score on this MPQ scale indicates a tendency to work and drive oneself hard. The Inhibited and Reserved groups

of children scored lowest on this scale, suggesting that they avoid very demanding projects and that they are not terribly ambitious.

In addition to these significant differences, several trends bear

mention. The Confident and Undercontrolled children obtained the lowest scores on Self-control, whereas the Undercontrolled children

also scored highest on Aggression and lowest on Social Closeness. The Inhibited children, meanwhile, scored significantly lower on Well-being than all other groups with the exception of the

Undercontrolled children. To summarize the adult personality correlates of early childhood

temperament, we examined differences between the five temperament groups along the MPQ’s three superfactors. The bottom panel of

Table 3 shows that the five temperament groups differed significantly on each superfactor. Undercontrolled children grew up to be

characterized by the highest trait levels of Negative Emotionality. Confident children grew up to be the most disinhibited (low

Constraint) adults and differed significantly in this regard from the Well-adjusted, Reserved, and Inhibited children, but not from their Undercontrolled counterparts. Inhibited children grew up to have the

highest trait levels of Constraint and the lowest trait levels of Positive Emotionality. The Reserved children were likewise characterized by

low Positive Emotionality and differed significantly from Confident children who had the highest scores on this personality superfactor.

In this article, we report links between age 3 behavioral styles and the age 26 MPQ profiles of the study members. Eight years ago, we

From Age 3 to Age 26 507

published an article showing links between age 3 behavioral styles

and the MPQ profiles of the study members when they were age 18 (Caspi & Silva, 1995). In order to facilitate comparisons between the

associations reported in the two articles, Table 3 also shows the age 18 MPQ profiles of the five age 3 temperament groups. We thus

report two means for each age 3 temperament group: each group’s MPQ trait mean at age 18 and again at age 26, yielding a total of 65

mean pairs. Comparing these pairs, we see that in 58 of the 65 pairs, the means did not change by more than 1/10 of a standard deviation.

The only changes of note occurred among the Inhibited children. Their relative scores on MPQ Well-being were much lower at age 26 than at age 18, although at both ages their overall relative Positive

Emotionality scores were lowest. In addition, their relative scores on Aggression increased, and the Inhibited children appeared to be

characterized by relatively more Negative Emotionality at age 26 than at age 18.

The age 26 MPQ results, derived from self-reports, are further confirmed by informant reports. Table 4 shows that the five groups

of children differed in adulthood on each of the Big Five dimensions. The Confident children were rated as the most Extraverted adults and the Inhibited children as the least

Extraverted. The Undercontrolled children were regarded as the least Agreeable and Conscientious members of the sample, and they

were also rated as more tense and anxious (high Neuroticism). Finally, the Well-adjusted and Confident groups were rated as

significantly more Open to Experience than the Undercontrolled, Reserved, and Inhibited children; the latter three groups were

described as relatively closed to new experiences.

DISCUSSION

That children’s temperamental qualities should presage their adult personalities is not a novel proposition. What is novel is its empirical

demonstration. Although the longitudinal connections reported in this article represent small-to-moderate effect sizes, the connections

are striking in that they span three different data sources across 23 years; from observer ratings after a 90 min. exposure to the children

at age 3 to self-reports and informant reports of the children as 26- year-old adults (for further discussion of effect sizes in longitudinal

508 Caspi et al.

T a

b le

4 T

e m

p e

ra m

e n

t T

y p

e s

Id e

n ti

fi e

d a

t A

g e

3 H

a v

e D

is ti

n c

t P

e rs

o n

a li

ti e

s a

t A

g e

2 6

,a s

S e

e n

T h

ro u

g h

P e

rs o

n a

li ty

R e

p o

rt s

A b

o u

t th

e B

ig F

iv e

P e

rs o

n a

li ty

F a

c to

rs P

ro v

id e

d b

y In

fo rm

a n

ts W

h o

K n

o w

th e

S tu

d y

M e

m b

e rs

W e

ll

T e m p e ra m e n t G ro u p s a t A g e 3

B ig

F iv e S c a le s

W e ll -A

d ju st e d

(n 5 3 7 0 )

C o n fi d e n t

(n 5

2 6 0 )

U n d e r- C o n tr o ll e d

(n 5 9 0 )

R e se rv e d

(n 5 1 4 0 )

In h ib it e d

(n 5

7 4 )

F R a ti o fo r

T e m p e ra m e n t

E x tr a v e rs io n :

.0 2 a b

.1 3 b

� .0 8 a c

� .1 1 a c

� .2 3 c

2 .7 5 n

A g re e a b le n e ss :

.0 7 a

.0 3 a

� .3 4

.0 3 a

� .0 4 a

3 .2 3 n n

C o n sc ie n ti o u sn e ss :

.0 4 a

.0 4 a

– .3 5

� .0 4 a

.1 7 a

3 .6 9 n n

N e u ro ti c is m :

� .0 6

� .0 4 a

.3 0 b

� .0 3 a

.0 4 a b

2 .5 7 n

O p e n n e ss

to E x p e ri e n c e :

.1 3 a

.0 8 a

� .3 4 b

� .1 8 b

� .1 5 b

6 .3 2 n n

N o te : T h e ta b le

sh o w s Z -s c o re s (M

e a n 5

0 ; S ta n d a rd

D e v ia ti o n 5

1 ) st a n d a rd iz e d o n th e fu ll sa m p le . G ro u p s th a t sh a re

a su p e rs c ri p t d o

n o t d if fe r fr o m

e a c h o th e r;

n n p o .0 1 , n p o .0 5 .

studies, see Caspi, 2000). More importantly, the connections make

psychological sense. When observed at age 3, children classified as Undercontrolled

(10% of the sample) were rated as irritable, impulsive, emotionally labile, and impersistent on tasks. At age 26, they were intolerant and

scored high on traits indexing Negative Emotionality; they were easily upset, likely to overreact to minor events, and reported feeling

mistreated, deceived, and betrayed by others. This profile was corroborated by other people, who described Undercontrolled

children grown up as antagonistic, unreliable, tense, and narrow- minded. Elsewhere, we have shown that Undercontrolled children were also the most likely to become enmeshed in an antisocial

lifestyle (Moffitt et al., 1996). When observed at age 3, children classified as Inhibited (8% of

the sample) were shy, fearful, and socially ill at ease. At age 26, they were characterized by an overcontrolled and nonassertive person-

ality style; they expressed little desire to exert influence over others and reported taking little pleasure in life. This profile was

corroborated by other people, who described Inhibited children grown up as less affiliative and lacking lively interest and engagement in their worlds. Elsewhere, we have shown that

Inhibited children lacked social support and were prone to internalising psychiatric problems (Caspi et al., 1996). As adults,

Inhibited children appear to be the kind of people who fail to take on the world in ways that will produce joy.

The remaining three temperament groups did not display such dramatic personality profiles as adults, but continuity was dis-

cernible in each group. Confident children (28% of the sample) were characterized by a zealous and outgoing approach to the testing

session when observed by examiners at age 3. At age 26, they were the least conventional and most agentic members of the sample, and other people described them as extraverted. Reserved children (15%

of the sample) were characterized as apprehensive in the novel testing session when they were observed at age 3. At age 26, they

described themselves as unassertive and diffident and were described by people who knew them well as introverted and less open to

experience. Finally, the Well-adjusted type (40% of the sample) included children whose behavior at age 3 was characterized as age-

and situation-appropriate; their style of approach and response to the testing session was regarded as expectable by the examiners and

510 Caspi et al.

made for smooth testing. This style was still discernible at age 26;

statistically, Well-adjusted children defined average adults. The Dunedin study members previously completed the MPQ

when they were 18 years old, at which time we reported that temperamental qualities at age 3 predicted those personality scores.

The present article shows that, eight years later, the continuities from age 3 are still evident and the connections with age 26

personality scores are, in some instances, even stronger and more psychologically coherent. How can it be that personality continuities

increase or, at least, retain their strength despite the passage of time? The period between ages 18–30 is brimming with niche-pick opportunities when young adults can create their own environments

in ways that are correlated with their dispositional tendencies (Scarr & McCartney, 1983): They can choose what to do, where to live, and

with whom. It is thus possible that becoming an adult actually gives individuals the opportunity to better express their true personalities.

We noted in the introduction that, historically, debates about personality continuity versus change have been contentious. Equally

so are debates about nature versus nurture (McCrae et al., 2000). The findings reported in this article address the former debate but are silent about the latter. Unfortunately, these two debates are often

conflated. Because evidence points to the considerable heritability of many psychological characteristics—including the temperamental

qualities (Goldsmith & Gottesman, 1981) and personality traits (Plomin & Caspi, 1999) studied in this article—it is frequently inferred

that genetic effects also explain how continuity is achieved. This is not necessarily so. Behavioral genetic studies are concerned primarily with

the origins of individual differences, but genetic sources of influence on individual differences in personality do not preclude the possibility

that continuities in these differences are mediated environmentally. The Dunedin study’s behavioral data were first collected only at age

3, after a lot of development already occurred. There is much to learn

about the origins of these individual differences, and it also remains to be seen whether it is possible to foretell adult life patterns from

individual differences measured before age 3 (Rothbart & Bates, 1998). What this study does establish are theoretically meaningful connec-

tions between 3-year-old children’s behavioral styles and their adult personalities. There is more to establishing this answer than satisfying

intellectual curiosity. If early-emerging behavioral differences did not predict outcomes, behavioral scientists, parents, and teachers could

From Age 3 to Age 26 511

safely ignore such individual differences. However, because such

differences do shape the course of development, information about these individual differences can be harnessed to design parent-training

programs and school-based interventions to improve children’s development. Ironically, although demonstrations of continuity are

often viewed as deterministic and pessimistic, such findings provide the strongest support for the urgency of early intervention.

REFERENCES

Bates, J. E. (1994). Parents as observers of their children’s development. In S.

Friedman, & H. C. Haywood (Eds.), Developmental follow-up: Concepts,

domains, and methods. San Diego: Academic Press.

Benet-Martinez, V., & John, O. P. (1998). Los cinco grandes across cultures and

ethnic groups: Multitrait multimethod analyses of the Big Five in Spanish and

English. Journal of Personality and Social Psychology, 75, 729–750.

Block, J. (1984). From infancy to adulthood: A clarification. Child Development,

51, 622–623.

Bruer, J. T. (1999). The myth of the first three years. New York: Free Press.

Caspi, A. (2000). The child is father of the man: Personality continuities from

childhood to adulthood. Journal of Personality and Social Psychology, 78,

158–172.

Caspi, A. (1998). Personality development across the life course. In W. Damon

(Series Ed.) & N. Eisenberg (Vol. Ed.), Handbook of child psychology, Volume

3: Social, emotional, and personality development (pp. 311–388). New York:

Wiley.

Caspi, A., Moffitt, T. E., Newman, D. L., & Silva, P. A. (1996). Behavioral

observations at age 3 predict adult psychiatric disorders: Longitudinal

evidence from a birth cohort. Archives of General Psychiatry, 53, 1033–1039.

Caspi, A., & Silva, P. A. (1995). Temperamental qualities at age 3 predict

personality traits in young adulthood: Longitudinal evidence from a birth

cohort. Child Development, 66, 486–498.

Cloninger, C. R. (1999). Personality and psychopathology. Washington, DC:

American Psychiatric Association Press.

Church, T. A., & Burke, P. J. (1994). Exploratory and confirmatory tests of the

Big Five and Tellegen’s three- and four-dimensional models. Journal of

Personality and Social Psychology, 66, 93–114.

Cohen, J. (1992). A power primer. Psychological Bulletin, 112, 155–159.

Goldsmith, H. H., & Gottesman, I. I. (1981). Origins of variation in behavioral

style: A longitudinal study of temperament in young twins. Child Development,

52, 91–103.

Hamer, D. (1997). The search for personality genes: Adventures of a molecular

biologist. Current Directions in Psychological Science, 6, 111–114.

John, O. P., & Robins, R. W. (1993). Determinants of interjudge agreement on

personality traits. Journal of Personality, 61, 521–552.

512 Caspi et al.

John, O. P., & Robins, R. W. (1994). Accuracy and bias in self-perception:

individual differences in self-enhancement and the role of narcissism. Journal

of Personality and Social Psychology, 66, 206–219.

Kagan, J. (1996). Three pleasing ideas. American Psychologist, 51, 901–908.

Lewis, M. (1997). Altering fate: Why the past does not predict the future. New

York: Guilford Press.

McCrae, R. R., & Costa, P. T., Jr. (1997). Personality trait structure as a human

universal. American Psychologist, 52, 509–516.

McCrae, R. R., Costa, P. T., Jr., Ostendorf, F., Angleitner, A., Hrebickova, M.,

Avia, M. D., Sanz, J., Sanchez-Bernardos, M. L., Kusdil, M. E., Woodfield,

R., Saunders, P. R., & Smith, P. B. (2000). Nature over nurture:

Temperament, personality, and life span development. Journal of Personality

and Social Psychology, 78, 173–186.

Magnusson, D., & Bergman, L. (1990). Data quality in longitudinal research.

Cambridge, UK: Cambridge University Press.

Matheny, A. P., Jr. (1980). Bayley’s Infant Behavior Record: Behavioral

components and twin analyses. Child Development, 51, 1157–1167.

Moffitt, T. E., Caspi, A., Dickson, N., Silva, P. A., & Stanton, W. (1996).

Childhood-onset versus adolescent-onset antisocial conduct in males: Natural

history from age 3 to 18. Development and Psychopathology, 8, 399–424.

Patrick, C. J., Curtin, J. J., & Tellegen, A. (in press). Development and validation

of a brief form of the Multidimensional Personality Questionnaire. Psycho-

logical Assessment.

Plomin, R., & Caspi, A. (1999). Behavioral genetics and personality. In L. A.

Pervin, & O. P. John (Eds.), Handbook of personality theory and research (2nd

ed.; pp. 251–276). New York: Guilford.

Rindfuss, R. R. (1991). The young-adult years—Diversity, structural-change, and

fertility. Demography, 28, 493–512.

Roberts, B. W., Caspi, A., & Moffitt, T. E. (2001). The kids are alright: Growth

and stability in personality development from adolescence to adulthood.

Journal of Personality and Social Psychology, 81, 670–683.

Roberts, B. W., & Friend-DelVecchio, W. (2000). Consistency of personality

traits from childhood to old age: A quantitative review of longitudinal studies.

Psychological Bulletin, 126, 3–25.

Rothbart, M., & Bates, J. E. (1998). Temperament. In W. Damon (Series Ed.) &

N. Eisenberg (Vol. Ed.), Handbook of child psychology, Volume 3: Social,

emotional, and personality development (pp. 105–176)). New York: Wiley.

Scarr, S., & McCartney, K. (1983). How people make their own environments: A

theory of genotype-environmental effects. Child Development, 54, 424–435.

Shiner, R. (1998). How shall we speak of children’s personalities in middle

childhood? A preliminary taxonomy. Psychological Bulletin, 124, 308–332.

Silva, P. A., & Stanton, W. (Eds.)(1996). From child to adult: The Dunedin study.

Auckland: Oxford University Press.

Tellegen, A., Lykken, D. T., Bouchard, T. J., Wilcox, K. J., Segal, N. L., & Rich,

S. (1988). Personality similarity in twins reared apart and together. Journal of

Personality and Social Psychology, 6, 1031–1039.

From Age 3 to Age 26 513

514