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

net/publication/229633680

Handbook of Child Psychology

Chapter · June 2007

DOI: 10.1002/9780470147658.chpsy0303

CITATIONS READS
1,210 17,452

2 authors, including:

Mary K Rothbart
University of Oregon
234 PUBLICATIONS 47,462 CITATIONS

SEE PROFILE

All content following this page was uploaded by Mary K Rothbart on 09 January 2016.

The user has requested enhancement of the downloaded file.

 Temperament 1

TEMPERAMENT

Mary K. Rothbart
University of Oregon

John E. Bates
Indiana University

Rothbart, M. K., & Bates, J. E. (in press). Temperament in children’s development. In W.

Damon, R. Lerner, & N. Eisenberg (Eds.), Handbook of child psychology, Sixth edition:
Social, emotional, and personality development (Vol. 3). New York: Wiley.
 Temperament 2

TEMPERAMENT

Recent years have witnessed major advances in our understanding of temperament in

childhood. Early views on temperament as unchanging and stable have been replaced by more
dynamic views of developmental change in temperament. An early emphasis on temperament in
infancy has been extended to the study of childhood and adolescence, and research on
temperament has burgeoned (Rothbart & Derryberry, 2002). In addition, rapid advances have
been made in our understanding of temperament-environment interactions. In this chapter, we
explore both historical influences and more recent advances in our understanding of individual
differences in temperament, differences observed by parents and physicians long before their
systematic study by students of human development.

A DEFINITION OF TEMPERAMENT

Gordon Allport defined temperament as “the characteristic phenomena of an individual’s

emotional nature, including his susceptibility to emotional stimulation, his customary strength
and speed of response, the quality of his prevailing mood, these phenomena being regarded as
dependent upon constitutional make-up and, therefore, largely hereditary in origin” (Allport,
1961, p. 34). Allport’s definition focused on individual differences in emotional reactivity.
Thomas and Chess (1977), however, took a broader approach to temperament, including
individual differences in attention and activity level as well as the emotions.
Taking into account both of these approaches, we have defined temperament as
constitutionally based individual difference in reactivity and self-regulation, in the domains of
affect, activity, and attention (Rothbart & Bates, 1998; Rothbart & Derryberry, 1981). By the
term constitutional, we refer to the biological bases of temperament, influenced over time by
heredity, maturation, and experience. “Reactivity” and “self-regulation,” are umbrella terms that
broadly organize the temperament domain. By reactivity, we refer to responsiveness to change
in the external and internal environment. Reactivity includes a broad range of reactions (e.g., the
emotions of fear, cardiac reactivity) as well as more general tendencies (e.g., negative
emotionality), and are not limited to general reactivity. Parameters of reactivity are measured in
terms of the latency, duration, and intensity parameters of motor, affective, and attentional
reactions (e.g., fear, anger, positive affect, orienting) (Rothbart & Derryberry, 1981). Emotional
reactivity also includes action tendencies. Thus, fear can predispose freezing, attack, and/or
inhibition, and positive affectivity can predispose approach. The expression or inhibition of
these behavioral tendencies can then feed back to influence the ongoing emotional reaction. By
self-regulation, we refer to processes such as effortful control and orienting that function to
modulate reactivity. We believe, as does Kagan (1998), that other important dimensions of
temperament are likely to be identified in the future.
Temperament describes tendencies or dispositions that are not continually expressed, but
require appropriate eliciting conditions. Fearful children, for example, are not continually
distressed and inhibited, but under conditions of novelty, sudden change in stimulation, or
signals of punishment, they may be particularly prone to a fearful reaction. Easily frustrated
children are not continually irritable or angry, but when their goals are blocked or there is a
failure of their expectations, they will be particularly prone to frustration.
 Temperament 3

Temperament and Personality

Temperament represents the affective, activational, and attentional core of personality,
whereas personality includes much more than temperament, particularly the content of thought,
skills, habits, values, defenses, morals, beliefs, and social cognition. Social cognition includes
the perception of the self, others, and the relation of self to objects, events, and others. Over
time, social cognition becomes increasingly important in eliciting and moderating temperamental
processes. This happens, for example, when anger comes to be elicited by judgments that others
have broken the rules when we have been following them. These perceptions can be influenced
by temperament (Derryberry & Reed, 1994), but they involve separable thought processes as
well. Personality traits have been defined as patterns of thoughts, emotion, and behavior that
show consistency across situations and stability over time, and that “affect the individual’s
getting along with other people and with himself” (Hilgard, 1962, p. 447). Temperament traits
similarly show consistency across situations and stability over time, but they are limited to basic
processes of reactivity and self-regulation, and do not include the specific content of thought or
the use of conceptually based defenses (e.g., paranoia).
In our view, temperament is the appropriate term for describing individual differences in
reactivity and self-regulation in non-human animals and young infants. Although some
researchers refer to individual differences in animals as their “personality” (e.g., Gosling & John,
1999), we find it helpful to consider the aspects of individuality we share with other animals
separately from those involving the content of thought that are more distinctly human. Animal
models of temperament allow investigations of affective and cognitive neuroscience that are not
possible in humans, and aid in the study of the neural underpinnings of temperament. Strelau
(1983) takes a similar position to ours, arguing that temperament results from biological
evolution, and is “peculiar to both humans and animals, which cannot be said of personality”
(Strelau, 1983, p. 258). In addition, “The individual has a temperament from the moment of
birth, since it is determined by inborn physiological mechanisms which, in turn, may be modified
under environmental influences” (p. 258).
We begin this chapter with a brief history of temperament research, considering its recent
history, its ancient roots, and its study in adulthood. In the second section, we examine the
structure of temperament as it has emerged from research on child development and from some
of the major theoretical models of the neurosciences. We also consider results of the search for a
taxonomic structure of adult personality traits, and relate temperament structure to the resulting
"Big Three" and "Big Five" factors of personality.
In the third section, we discuss methods and measures for the study of temperament,
considering both the benefits and liabilities of some of the major empirical approaches. Because
the use of parent-report in temperament research has been questioned (Kagan, 1994, 1998;
Kagan & Fox, this volume), we critically consider contributions of parent-report. In the fourth
section, we focus on temperament and development, considering issues of continuity and change.
In the fifth section, we discuss relations between temperament and behavioral adjustment. The
final section presents our conclusions and indicates future directions for the study of
temperament and development. Overall, we organize research findings on temperament within a
developmental framework. The literature review is not comprehensive, but we hope it captures
some of the major issues and approaches to the study of temperament in childhood. We now
begin with a historical review of temperament in childhood and adulthood.
 Temperament 4

HISTORY OF TEMPERAMENT

Research on Childhood
Normative Studies. Several lines of inquiry have contributed to contemporary
temperament research on children. One is the research of the normative child psychologists in
the 1920s and 30s, who observed large numbers of children to establish the normal sequences of
motor and mental development, and studied small samples of children intensively over time. In
doing so, they noted striking temperamental variability among the children they observed
(Gesell, 1928, as cited in Kessen, 1965; Shirley, 1933). Shirley’s intensive study of the motor
development of a group of infants during the first two years led to her observation of the infant’s
“core of personality.” Shirley noted that developmentally, “Both constancy and change
characterize the personality of the baby. Traits are constant enough to make it plausible that a
nucleus of personality exists at birth and that this nucleus persists and grows and determines to a
certain degree the relative importance of (other) traits” (1933, p. 56). She devoted a full volume
to these traits, even though her original intention had been to study only motor development.
Fifteen years later, Neilon (1948) located 15 of Shirley’s 25 babies, asking judges to match
Shirley's infant personality sketches to descriptions of the children as adolescents, based on
assessments by clinicians blind to their identity. These matches were more successful than
would have been expected by chance.
Gesell (1928, as cited in Kessen, 1965) identified the critical importance of the child’s
temperament within what he called the developmental “web of life.” His views of alternative
developmental pathways, so important to recent thinking, are illustrated in the case of C. D.:
This girl exhibited a striking degree of amenability, sociality, and good nature as early as
the age of nine months . . . She is now five years of age, and in spite of a varied
experience in boarding homes and institutions she has not lost these engaging
characteristics. They are part and parcel of her make-up quite as much as the lowered
tempo and the lowered trend of her general development. It can be predicted with much
certainty that she will retain her present emotional equipment when she is an adolescent
and an adult. But more than this cannot be predicted in the field of personality. For
whether she becomes a delinquent, and she is potentially one, will depend upon her
subsequent training, conditioning, and supervision. She is potentially also a willing,
helpful, productive worker. Environment retains a critical role even though heredity sets
metes and bounds (Gesell, 1928, p. 223 as cited in Kessen, 1965).
Three important concepts from Gesell and Shirley will be further elaborated in this
chapter. First, temperament traits are constitutionally based characteristics that provide the core
of personality and influence directions for development. Second, although some stability of
temperament is expected across age, developmental outcomes will also depend on the child’s
experience in the social context. Finally, as in the case of C. D., a given set of temperamental
characteristics allows for multiple possible outcomes. Different trajectories and outcomes may
occur for children with similar temperamental traits, and children differing in temperament may
come to similar developmental outcomes via different pathways (Kochanska, 1995).

Clinical Research
A second major line of research on temperament in childhood came from biologically
oriented clinicians. Bergman and Escalona (1949) identified children who were particularly
 Temperament 5

reactive to low intensities of stimulation in one or more sensory modalities. In Escalona's (1968)
groundbreaking book, The Roots of Individuality, she proposed the concept of "effective
experience," the idea that events in children's lives are experienced only as they are filtered
through the individual child's nervous system. A given event will thus differ in its effects for
children who differ in temperament. An adult's vigorous play, for example, may lead to pleasure
in one child and distress in another. Escalona noted that objective coding of environmental
events alone will not capture essential information about the individual child's reaction to them.
Infants observed by Escalona (1968) were followed in developmental studies of vulnerability,
resiliency, and coping by Murphy and Moriarty (1976). In other studies, Fries and Woolf (1954)
identified and studied congenital activity type, Korner (1964) studied neonatal individuality and
developed an extensive assessment schedule for the newborn, and Birns (1965) and her
associates developed and implemented some of the earliest standardized assessments of
temperament.
Among clinical investigators, Thomas, Chess, Birch, Hertzig, and Korn (1963) published
the first of their volumes on the extremely influential New York Longitudinal Study (NYLS).
Inspired by differences among their own children, Chess and Thomas studied individual
differences in what they called the "primary reaction patterns," collecting interviews from
parents of infants on repeated occasions. Beginning when their initial sample of 22 infants was
3-6 months of age, parents were extensively interviewed about their infants' behavior in varying
contexts. Each infant reaction and its context was then typed on a separate sheet of paper, and
Birch inductively sorted the descriptions into categories that came to represent to the 9 NYLS
temperament dimensions (Chess & Thomas, personal communication, 1992; Thomas et al.,
1963): Activity Level, Approach/Withdrawal, Adaptability, Mood, Threshold, Intensity,
Distractibility, Rhythmicity, and Attention Span/Persistence. Later, Michael Rutter suggested
the term "temperament" to describe their area of study, and this term was adopted by the NYLS
group (Chess & Thomas, personal communication, 1992).

Acceptance of Temperament Research

Reports from the NYLS arrived at an opportune time, when researchers in social
development were becoming increasingly aware of the contributions of individual children to
their own development. One influence was the burgeoning of infancy research, with researchers
studying the "initial state" of the individual and its subsequent adaptations (Osofsky, 1979).
Because the initial state varied from child to child (Korner, 1964), early differences could be
seen as providing the raw material for later development. Ideas originally put forward by Robert
Sears and associates (e.g., Sears, Maccoby, & Levin, 1957) were also re-emerging regarding bi-
directionality in the effects of socialization, from child to caregiver as well as caregiver to child
(Bell, 1968). Finally, cognitive approaches stressed children's influences on their own
development via their perceptual and cognitive mental representations of events (Kohlberg,
1969). Research on temperament would now introduce the idea that, in addition to individual
differences in thought patterns, individual differences in children's emotional processing could
bias their affective representations of experience, with important implications for their
development.

History of Temperament in Adulthood

Ancient Thought. Adult studies of temperament have a much longer history than
developmental work, much of it focused on biological aspects of personality. Temperament
 Temperament 6

ideas go back to Greco-Roman physicians over 2,000 years ago, and to even earlier traditions in
China and India (Diamond, 1974; Needham, 1973). In this thinking, psychological
characteristics were consistently linked to the physiology of the individual as it was understood
at the time. Thus, ancient Greco-Roman physicians identified the well-known fourfold typology
and linked it to the bodily humors: the sanguine individual, positive and outgoing, with a
predominance of blood; the melancholic person, prone to fear and sadness, with a predominance
of black bile; the choleric person, irritable and prone to aggression, with a predominance of
yellow bile; and the phlegmatic person, slow to excitation, with a predominance of phlegm
(Diamond, 1974). The fourfold typology was used throughout the Middle Ages, and in the
writings of Kant. By the time of Wundt (1903), however, a shift was made away from positing
temperamental "types" to studying dimensions of variability in temperament, a shift that has only
recently been reversed in Kagan's (1994, 1998) and others’ (Caspi, Sugden, et al., 2003) use of
temperament typologies.
Jung’s Theory. Psychological types were put forward by Jung (1923), but they differed
in important ways from current typologies. In Jung’s view, introverted and extraverted
tendencies were universal and reflected in thinking, feeling, sensation, and intuition. Jung
argued that introverted and extraverted tendencies are present in all persons, but that, for a given
person, one attitude becomes more elaborated and conscious, while the other is less elaborated,
more primitive, and, for the most part, unconscious. Differences among children in extraversion
and introversion, he wrote, can be seen early in life:
The earliest mark of extraversion in a child is his quick adaptation to the environment,
and the extraordinary attention he gives to objects, especially to his effect upon them.
Shyness in regard to objects is very slight; the child moves and lives among them with
trust. He makes quick perceptions, but in a haphazard way. Apparently he develops
more quickly than an introverted child, since he is less cautious, and as a rule, has no fear.
Apparently, too, he feels no barrier between himself and objects, and hence he can play
with them freely and learn though them. He gladly pushes his undertakings to an
extreme, and risks himself in the attempt. Everything unknown seems alluring (Jung,
1928, cited in Fordham, 1953, p 303).
Objects as described by psychoanalysts include both physical and social entities, so that
the more introverted child would be expected to dislike new social situations and to approach
strangers with caution and fear. Jung suggests that the introvert would also be inclined toward
pessimism about the future, and the extravert would show more ready approach and action upon
objects (impulsivity), greater sociability, and more optimism about the future (Jung, 1923).
Eastern and Western Schools of Temperament and Personality. In addition to Jung’s
theoretical model of introversion-extraversion, similar dimensions emerged from early factor-
analytic studies of temperament in adults. In Great Britain, Webb (1915) analyzed self-report
items referring to emotionality, activity, qualities of the self, and intelligence, and identified two
broad factors. One, labeled “w,” was defined as “consistency of action resulting from deliberate
volition or will” (Webb, 1915, p. 34). This factor bears similarities to temperamental Effortful
Control in childhood (Kochanska, Murray, & Harlan, 2000; Rothbart, Ahadi, Hershey, & Fisher,
2001), and to the higher-order personality factors more recently labeled Control, Constraint, or
Conscientiousness (Digman & Inouye, 1986). A second factor assessed distress proneness or
negative emotionality, sometimes labeled emotional stability-instability; Eysenck would later
call it Neuroticism (1967). By 1937, Burt had also identified the factor of Extraversion-
Introversion, and later factor-analytic research on questionnaire assessed personality has
 Temperament 7

repeatedly identified three factors: extraversion, neuroticism, and conscientiousness. These three
factors, sometimes called the “Big Three,” along with factors of Agreeableness and Openness to
experience, have been extracted from factor analyses of trait descriptive words and personality
items, and constitute what have been called the “Big Five” personality factors (Costa & McCrae,
1988, Goldberg, 1990).
Soviet and Eastern European research on temperament began with Pavlov's (1955/1935)
observations of individual differences among dogs in conditioning experiments, and led to an
active research tradition described by Strelau (1983). Pavlov linked temperamental differences
among the animals, which he also generalized to humans, to qualities of the central nervous
system, including strength of neural activation. Strength of activation was related to the "law of
strength" in classical conditioning, whereby increasing the intensity of a conditioned stimulus led
to the increased intensity of the animal's response. For some animals, however, increasing
stimulus intensity led to failure to respond. Pavlov described these animals as having a weak
nervous system; animals with a strong nervous system maintained the law of strength even at
high levels of stimulus intensity. Additional Pavlovian temperament constructs included strength
of inhibition, balance between activation and inhibition, and mobility, that is, flexibility of
nervous system adjustment to changing conditions. Soviet researchers began their work in the
laboratory, but a lack of generality of their measures across stimulus and response modalities, a
phenomenon they called partiality, led to a general shift in their focus from the laboratory to the
use of questionnaire methods (Strelau, 1983).
Although British and Soviet schools took different historical directions in the study of
temperament, with British researchers moving from questionnaires to the laboratory, and Eastern
European researchers from the laboratory to questionnaires, both schools remain actively
involved in the study of temperament, and both link individual differences in temperament to
hypothetical nervous system function We now review studies of the structure of temperament as
it has emerged from research on infancy and childhood, and describe neural models conceptually
related to this structure.

THE STRUCTURE OF TEMPERAMENT

In this section, we consider research on temperament that has led to revision of the
original list of nine NYLS dimensions (Tables 1 and 2). Much of this research employed factor
analysis of large sets of items within the temperament domain. Factor analysis allows
researchers to see simultaneously the relations and non-relations among large sets of behavior
descriptions. A major limitation of the factor analytic method is that the dimensions yielded by
the analysis depend on the descriptors included in the initial data matrix. It is nevertheless of
interest that several broad dimensions of temperament have consistently emerged from different
sets of data.

Infant Studies
In a review of the structure of temperament as indicated by infant studies (Rothbart &
Bates, 1998; Rothbart & Mauro, 1990), six dimensions were identified that provided a "shorter
list" of temperament variables for future researchers (Table 1). Individual differences in positive
emotionality were differentiated from negative emotionality, and two kinds of negative emotion
were identified, fear and anger/irritable distress.
 Temperament 8

Infant scales with different names often measure similar constructs. Goldsmith and
Rieser-Danner (1986) had both mothers and day-care teachers of 4-8 month old infants fill out
the Revised Infant Temperament Questionnaire (RITQ; Carey & McDevitt, 1978), the Infant
Characteristics Questionnaire (ICQ; Bates, Freeland, & Lounsbury, 1979), and the Infant
Behavior Questionnaire (IBQ; Rothbart, 1981). Distress to novelty was assessed by all three of
these instruments: IBQ Fear, ICQ Unadaptable, and ITQ Approach-Withdrawal scales.
Intercorrelations across these scales were high. For mothers, they ranged from .60-.69, with the
average r = .64; for day-care teachers, the intercorrelations ranged from .51-.73, with the average
r = .63 (Goldsmith & Rieser-Danner, 1986).
The second shared dimension was Irritable Distress, assessed by IBQ Distress to
Limitations, RITQ Negative Mood (which includes positive affect at one pole), and ICQ
Fussy/Difficult scales. Intercorrelations among these scales for mothers ranged from .44-.63,
with an average of .54; for day care teachers, the correlations ranged from .66-.74, with an
average of .71. The third shared temperament dimension was Activity Level, assessed only on
the RITQ and IBQ scales, where the correlation for both mothers and daycare teachers was .65.
Gartstein and Rothbart (2003) have recently studied the factor structure of expanded
scales measuring parent-reported infant temperament, as adapted from dimensions identified in
research on temperament in childhood (Table 1). Factor analysis of a large data set describing 3-
to 12-month-old children yielded three broad dimensions: Surgency/Extraversion, defined
primarily from scales of approach, vocal reactivity, high intensity pleasure (stimulation seeking),
smiling and laughter, activity level and perceptual sensitivity; Negative Affectivity, with loadings
from sadness, frustration, fear, and loading negatively, falling reactivity; and
Orienting/Regulation, with loadings from low intensity pleasure, cuddliness, duration of
orienting and soothability, and a secondary loading for smiling and laughter. As early as
infancy, there is thus evidence for three broad temperament dimensions.
_____________________
Table 1 about here
_____________________

A number of important conclusions have emerged from factor analytic studies on the
structure of infant temperament as reviewed by Rothbart and Mauro (1990) and Rothbart and
Bates (1998). First, the structure appears to correspond more to dimensions of reactivity in the
basic emotions and attention/regulation than to a general style. Second, bipolar constructs such
as approach versus withdrawal and good versus bad mood have not emerged from these
analyses; instead, unipolar constructs of infant temperament have gained support. Third, these
dimensions also correspond to individual differences emerging from studies of non-human
animals (Gosling & John, 1999), allowing links between temperament constructs in humans and
the psychobiology of individual differences.

Childhood Studies
Factor analyses of questionnaire items based on the NYLS for older children have
similarly revealed a “shorter list” of broad temperament factors (Table 2). Analysis of mother
reports for 3- to 8-year-olds on the Thomas and Chess (1977) Childhood Temperament
Questionnaire in the Australian Temperament Project (ATP) yielded factors of Inflexibility
(irritability and uncooperativeness), Persistence, Sociability, and Rhythmicity (Sanson, Smart,
 Temperament 9

Prior, Oberklaid, & Pedlow, 1994); second-order factors extracted from the ATP data were
labeled Negative Emotionality, Self-Regulation, and Sociability.

___________________
Table 2 about here
___________________

The Children’s Behavior Questionnaire (CBQ; Rothbart et al., 2001) has also consistently
yielded three broad factors, found in U.S. replications and in research performed in the People’s
Republic of China and Japan (Ahadi, Rothbart, & Ye, 1993; Kochanska, DeVet, Goldman,
Murray, & Putnam, 1993; Rothbart et al., 2001). The first, called Surgency/Extraversion, is
defined primarily by the scales of Approach, High Intensity Pleasure (sensation-seeking),
Activity Level, and a negative contribution from Shyness. The second, called Negative
Affectivity, is defined by the scales of Discomfort, Fear, Anger/Frustration, Sadness, and loading
negatively, Soothability. The third factor, labeled Effortful Control, is defined by the scales of
Inhibitory Control, Attentional Focusing, Low Intensity Pleasure, and Perceptual Sensitivity.
These three factors map well upon the second order factors identified by Sanson et al. (1994):
Surgency/Extraversion upon Sociability; Negative Affectivity upon Negative Emotionality, and
Effortful Control upon Self-Regulation.
The first three factors emerging from a recent factor analysis of NYLS-inspired Middle
Childhood Temperament Questionnaire items (Hegvik, McDevitt, & Carey, 1982) for 8- to 12-
year-olds (McClowry, Hegvik, & Teglasi, 1993) also show similarity to these factors:
Approach/Withdrawal, Negative Reactivity, and Task Persistence. McClowry et al.’s (1993) two
smaller factors, Activity and Responsiveness, also parallel smaller factors in the ATP (Sanson et
al., 1994).
Presley and Martin’s (1994) analysis of teacher reports of 3- to 7-year-olds on the
Temperament Assessment Battery for Children yielded five factors demonstrating some overlap
with those described above. These include Social Inhibition, Negative Emotionality,
Agreeableness/Adaptability, Activity Level, and Task Persistence. In their review of factor
analytic studies on infant and child temperament, Martin, Wisenbaker, and Huttunen (1994) note
the robustness of the general temperament factors of Negative Emotionality, Task persistence,
Adaptability, and Social Inhibition. They see Activity Level as more problematic, because it is
related to both negative and positive affect early in life.
The factors emerging from research on temperament in childhood show strong
conceptual similarities with the “Big Three” factors and three of the “Big Five” or FFM factors
that have been extracted from analyses of self- and peer descriptions of personality in adults
(Goldberg, 1993) and children (Ahadi & Rothbart, 1994; Caspi, this volume; Digman & Inouye,
1986). The Negative Affectivity factor from childhood measures is conceptually similar to the
broad adult dimension of Neuroticism or Negative Emotionality. The Surgency and Sociability
factors are similar to the broad adult dimension of Extraversion or Positive Emotionality. The
Persistence, Self-Regulation, or Effortful Control factors map upon the adult dimension of
Control/Constraint (see Ahadi & Rothbart, 1994), and Martin et al.’s (1994)
Agreeableness/Adaptability factor onto the adult dimension of Agreeableness.
In our research on adults, we have found strong one-to-one correlations between factor
scores derived from factor analysis of temperament scales and Big Five measures (Rothbart,
Ahadi, & Evans, 2000), between Negative Emotionality and Neuroticism, Positive Emotionality
 Temperament 10

and Extraversion, and Effortful Control and Conscientiousness. In addition, Perceptual

Sensitivity was related to Openness, and temperamental affiliation to Agreeableness.
Neuroticism, however, often contains negative judgments about the self that may be strongly
related to an individual’s experiences with others; these may or may not have a strong
temperamental base, and research on temperament and personality in childhood becomes very
important.
In research with the Combined Temperament and Personality Scales (CTPQ) describing
565 children between the ages of the ages of 3 and 12, we have extracted factors that included
positive emotion, gregariousness, warmth, and soothability (Sociable Extraversion); anxiety,
self-consciousness, dependency, and depression/sadness (Internalizing Negative Emotionality);
inhibitory control, order, diligence, self-discipline, attentional focusing, and low distractibility
(Conscientiousness); aesthetics/creative ideas, intellect, and perceptual sensitivity (Openness);
and excitement seeking, assertiveness, self-centered, non-compliance/aggression, manipulative,
activity, and anger/hostility and impulsivity (Unsocialized Stimulation Seeking) (Rothbart &
Victor, 2004). This research differentiates fearful and angry negative affect, and links aspects of
extraversion with both unsocialized stimulation seeking and sociable extraversion. Longitudinal
research will be important in identifying the experiential and self-regulatory temperament
influences on these two outcomes.
Shiner (1998) has also recently contributed an important review of the structure of
personality in middle childhood. Her preliminary taxonomy includes these dimensions:
sociability, social inhibition, prosocial disposition, dominance, aggressiveness, negative
emotionality, mastery motivation, inhibitory control, persistence/attention, and activity level, and
Shiner indicates the importance of placing these dimensions within a developmental context. We
can, for example, study irritability in infants, but aggression cannot be observed until later in the
development. The hierarchical structure of Shiner’s (1998) additional dimensions will also be of
interest. Does dominance, for example, relate more to extraversion/surgency, or to
agreeableness/affiliation? Does anger proneness fit within negative emotionality or
agreeableness/affiliation (Shiner, 1998)? In Rothbart and Victor’s (2004) research, it is loading
on the Unsocialized Stimulation Seeking factor.
Summary. As noted above, work to date on temperament structure in infancy and in
childhood suggests revisions of the original NYLS nine dimensions to include (with broad,
aggregated constructs in parentheses): Positive Affect and Activity Level,
(Approach/Extraversion), Fearful Distress, Irritable Distress (General Negative Emotionality),
Effortful Control/Task Persistence, and Agreeableness/Adaptability. In our next section, we
make tentative links between these constructs and models developed in the neurosciences.

NEURAL MODELS OF TEMPERAMENT

We now describe neural models developed to enhance our understanding of

temperament. Cloninger (1986), Gray (1991), LeDoux (1989), Panksepp (1998), and Zuckerman
(1984) have all made contributions to the development of neural models for temperament (see
review by Rothbart, Derryberry, & Posner, 1994). We begin by describing models for positive
emotionality and approach (Surgency/Extraversion), and fear. Irritability/anger is also discussed,
and it is seen, along with fear, discomfort, and sadness, to represent part of a general construct of
susceptibility to negative affect or negative emotionality. The emotion-based dimensions are
related to differences observed in non-human animals (Gosling & John, 1999; Panksepp, 1998),
 Temperament 11

and to factor structures extracted from studies of personality in adults and temperament in
childhood. The fourth dimension, effortful control, will be further discussed in connection with
neural models for individual differences in executive attention (Posner & Fan, in press).
Emotion as an Information Processing System
Emotions can be seen as broadly integrative systems that order feeling, thought, and
action (LeDoux, 1989). They also represent the output of information processing networks
assessing the meaning or affective significance of events for the individual (LeDoux, 1989).
Whereas object recognition systems and spatial processing systems address the questions, “What
is it?” and “Where is it?”, neural emotion processing networks address the questions, “Is it good
for me?”, “Is it bad for me?”, and “What shall I do about it?” When there are individual
differences in temperamental emotionality, there are thus differences in the object perception as
well.
In neural processing of emotion, thalamic connections route information about object
qualities of a stimulus through sensory pathways (LeDoux, 1989). Simultaneously, information
is routed to the limbic system and the amygdala, where memories of the affective meaning of the
stimulus further influence the process. Later object processing can update the emotional
analysis, but in the meantime, back projections from the amygdala can influence subsequent
sensory processing. Output of the amygdala to organized autonomic reactions via the
hypothalamus and to motor activation via the corpus striatum reflect the motivational aspect of
the emotions (LeDoux, 1989).

Attention as a Control System

Neuroimaging studies demonstrate connections between emotional processing networks
and the executive attention system that allow the selection of emotional information for
conscious processing, so that we may or may not be aware of our emotional evaluations (Bush,
Luu, & Posner, 2000; Posner & Rothbart, 1991). Attentional systems can select for conscious
processing aspects of emotional analyses, and emotion can also influence the focusing and
shifting of attention (Derryberry & Reed, 1996, 2002; Gray, 1991). An important aspect of
social adaptation involves the appropriateness of a child’s social interaction and the related
acceptance of the child by others (Parker & Asher, 1987). Information about the state of others
will thus be an important contributor to appropriate social action, and failure of this information
to access action and consciousness can be a critical element in the development of disordered
functioning. When attention is focused on threatening stimuli or on the self, access to
information about others is likely to be less accessible. These are important examples of
information processing aspects of temperament that have major implications for social
development, and we discuss them again in the section on temperament and the development of
personality.

Positive Emotionality/Approach and Extraversion

We now briefly review neural models developed to describe a physical substrate for
approach and fearful inhibition. Based on animal research, Gray (1991) described the
Behavioral Activation System (BAS), involving sensitivity to rewards, and the Behavioral
Inhibition System (BIS), involving sensitivity to punishment, nonreward, novelty, and innate fear
stimuli (see applications to children by Blair, 2003). These two systems are seen as mutually
inhibitory, with their balance determining degrees of extraversion-introversion. Gray also
posited a fight-flight system that moderates unconditioned punishment (Gray, 1991). According
 Temperament 12

to Gray’s BAS model, reward-related projections from the amygdala to the nucleus accumbens
activate a motor program that increases proximity to the desired stimulus, and facilitate goal-
oriented behavior (Gray & McNaughton, 1996).
In a broader “Behavioral Facilitation System (BFS),” Depue and Collins (1999) proposed
a circuit involving the nucleus accumbens, ventral pallidum, and dopaminergic neurons that
codes the intensity of the rewarding stimuli, with related circuits involving the medial orbital
cortex, amygdala, and hippocampus integrating the “salient incentive context.” Individual
differences in the functioning of this network are thought to arise from variation in the
dopaminergic projections that encode the intensity of incentive motivation. With development,
dopaminergic facilitation can enhance responsivity to positive incentive stimuli (Depue &
Collins, 1999), and provide a neural basis for a positive feedback system. This system can lead
initially approaching children to become even more approaching, and temperamental differences
in extraversion to become greater with time.
Depue and Iacono (1989) used the BFS to account for initiation of locomotor activity,
incentive-reward motivation, exploration of environmental novelty (if stronger than opposing
fear reactions), and irritable aggression. Panksepp (1982, 1986a, 1998) concluded that “the
general function of DA activity in appetitive behavior is to promote the expression of
motivational excitement and anticipatory eagerness – the heightened energization of animals
searching for and expecting rewards” (1986a, p. 91). Cloninger (1986, 1987) also specified a
novelty-seeking dimension related to DA functioning, as did Zuckerman (1984) in his dimension
of sensation seeking (see review by Rothbart, 1989a).
A broad approach dimension has also been linked to positive affect. Tellegen’s (1985)
research on personality yielded a broad factor of Positive Emotionality, including pleasure and
positive anticipation, and Watson and Clark (1997) have argued that positive affect is the core of
individual differences in Extraversion. The children’s temperament factor of
Surgency/Extraversion (Rothbart et al., 2001) fits well with these models. In research on infants
(Rothbart, 1988; Rothbart, Derryberry, & Hershey, 2000), expressions of smiling and laughter
were related to their rapid latency to approach objects, and predicted their anticipatory eagerness
about upcoming positive events at the age of seven years.
In recent fMRI research with adults, Canli et al. (2001) found that persons higher in
extraversion showed greater brain response to positive than negative stimuli in widespread
frontal, temporal, and limbic activation of both hemispheres. Those higher in neuroticism
reacted more to negative than to positive stimuli, and showing more circumscribed activation
(fronto-temporal on the left side) and deactivation in a right frontal area. In a follow-up study
focusing on the amygdala, extraversion was correlated with left amygdala activation to happy,
but not to fearful, faces (Canli, Sivers, Whitfield, Gotlib, & Gabrieli, 2002). Neuroticism was
not correlated with activation to any of the emotional faces, except for significant amygdala
activation for the fearful expression.

Fear and Behavioral Inhibition

Fear is an emotional response activated in the presence of threat or signals of upcoming
danger, and its function appears to be a defensive one. Fear activation is accompanied by
inhibition of ongoing motor programs and preparation of response systems controlling coping
options such as fleeing, fighting, or hiding (see review by Rothbart et al., 1994). In Gray’s
(1991) model of the Behavioral Inhibition System (BIS), the fear-related BIS is based upon
circuits including the orbital frontal cortex, medial septal area, and the hippocampus. However,
 Temperament 13

the amygdala has been more often identified as the critical structure in the processing of
conditioned fear (LeDoux, 1989). Emotional networks involving the amygdala also appear to
respond more strongly to novel than to familiar stimuli (Nishijo, Ono, & Nishino, 1988).
Amygdala lesions in rodents disrupt autonomic and cortisol reactions, behavioral freezing and
fear vocalizations, and similar findings have been reported in primates (Lawrence & Calder, in
press). In humans, functional neuroimaging studies by Calder, Lawrence, and Young (2001) and
others support involvement of the amygdala in both acquiring and expressing fear, although not
in the voluntary production of facial expressions of fear (Anderson & Phelps, 2002). The
amygdala also is involved in the recognition of fear in the human face (Calder et al., 2001), and
there is evidence in humans that amygdala damage is related to reduced fear experience
(Adolphs et al., 1999; Sprengelmeyer et al., 1999).
Projections from the amygdala implement autonomic and behavioral components of fear,
including startle, motor inhibition, facial expression, and cardiovascular and respiratory changes
(Davis, Hitchcock, & Rosen, 1987). Individual variability in the structure and functioning of any
of these subsystems may be related to variations in behavioral expressions of fear, and multiple
components of other affective motivational systems such as approach/positive affect and
anger/irritable distress would also be expected.
The amygdala also appears to affect information processing within the cortex. For
example, the basolateral nucleus of the amygdala projects to frontal and cingulate regions
involved in the executive attention system (Posner & Petersen, 1990), as well as to ventral
occipital and temporal pathways involved in processing object information. These connections
are consistent with findings that anxious individuals show enhanced attention to threatening
sources of information (e.g., Derryberry & Reed, 1994). A more detailed analysis of structures
related to behavioral inhibition can be found in Kagan and Fox's chapter, this volume.
A psychobiological analysis of fear suggests there may be less disagreement about
temperament variables than had been previously thought. Fear as we have described it includes
arousal, felt emotion, motor response preparation for flight and/or attack (with responses often
inhibited), and attention toward the fear-inducing stimulus and/or possible escape routes (Davis
et al., 1987). When temperament researchers study aspects of this construct, they sometimes
stress the motivational aspects of the individual's response (e.g., Thomas & Chess's [1977]
Approach/Withdrawal; Kagan's [1994] behavioral inhibition), the distress proneness aspects
(Buss & Plomin's [1975] Emotionality; Goldsmith & Campos' [1982] Fear), its duration and
susceptibility to interventions (Rothbart's [1981] Soothability), its relation to arousal (Strelau's
[1983] reactivity), or multiple components of response (Rothbart & Derryberry's [1981] Fear). If
we take the broader view of fear suggested by neuroscience work, agreement is more evident,
and intercorrelations among scales measuring the differently named constructs as discussed
above further support this contention (Goldsmith, Rieser-Danner, & Briggs, 1991).
We have now touched upon possible neural substrates for approach/positive affect
systems related to reward-seeking, and for fear, linked to the inhibition of approach and of
sensation seeking that might lead to punishment ("harm avoidance"). An important aspect of
these constructs for students of social development is that they describe individual differences in
susceptibility to reward and punishment, suggesting that some children will be more activated by
reward, and some children will find stopping an activity easier when there is a high likelihood of
punishment (Rothbart, Ahadi, & Hershey, 1994). When a situation involves both potential
rewards and punishments, such as interactions with a stranger, the balance between approach and
 Temperament 14

fear tendencies based on temperament and previous experience will be critical to behavioral
outcomes.
This model has direct applications to child socialization. If we consider a toddler
performing an enjoyable act, such as shredding the pages of a book, the child's initial activities
will be influenced by the approach or extraversion system. Now the parent gives a sharp,
punishing command for the child to stop. Will the child's activity be inhibited? Patterson (1980)
found that parents of non-problem children were effective in stopping their children's aversive
behavior on three out of four occasions when they punished. When parents of problem children
used punishment, however, children were likely to actually continue the punished behavior
(Patterson, 1980; Snyder, 1977). Although parenting skills are also involved, children's
temperament is likely to make a basic contribution to this situation.
Individual differences in risky behaviors and accident-proneness (Matheny, 1991;
Schwebel, 2004), mastery motivation (Rothbart & Hwang, in press), and affective
representations of the environment (Derryberry & Reed, 1994) are all likely to be influenced by
temperamental approach and inhibition tendencies in interaction with past experience. The
coping strategies children use will also be influenced by their tendencies to approach or inhibit
action. Quay (1993) has employed Gray's constructs of the BAS and BIS to analyze the
development of undersocialized aggressive conduct disorder. In the next section, we describe
temperament systems related to distress to overstimulation, irritability and anger, and possible
controls offered by affiliative tendencies.

Optimal Levels of Stimulation and Distress to Overstimulation

We have now described individual differences in systems of fear and
approach/extraversion. However, processes related to arousability have also been proposed to
underlie approach and withdrawal. One model incorporates the idea of "optimal level" of
stimulation. This approach derives temperament from general arousability or “strength of
nervous system,” put forward in the Soviet and Eastern European schools, Eysenck's (1967)
model, and ideas developed by Bell (1974). In the theory of Berlyne (1971), arousal potential,
created by stimulus intensity, novelty, and surprise, activates two motivational systems, one
related to pleasure and approach, elicited at lower levels of arousal potential, the other to distress
and withdrawal, elicited at higher levels of arousal potential. The two systems oppose each
other. Individual differences in the strength of each of these two systems support a level of
optimal arousal, that is, the point where approach and pleasure are at their highest, but
withdrawal processes do not yet dominate.
Schneirla (1959) put forward similar ideas, describing Approach and Withdrawal systems
across species related to the intensity of stimuli. Eysenck (1967) argued that introverts are more
arousable and sensitive to stimulation at low intensity levels than extraverts, linking this arousal
to the Ascending Reticular Activation System. Introverts were seen to experience both pleasure
and discomfort at lower levels of stimulus intensity; this lower optimal level of arousal would
lead them to seek lower stimulus levels. In Strelau's (1983) model, more strongly reactive
individuals engage in self-regulatory activities to maintain their optimal levels of stimulation.
Soviet researchers' concept of nervous system strength of activation or endurance under high-
intensity stimulation, and Strelau's reactivity construct are dimensions involving both sensitivity
and susceptibility to distress that might be present in early life. More recently, Aron and Aron
(1997) have also linked sensory sensitivity and reactivity to distress.
 Temperament 15

Developmental research on temperament show mixed support for a positive relation

between sensitivity and susceptibility to distress. Miller and Bates (1986) found significant
positive correlations between temperature sensitivity and susceptibility to the negative emotions.
In Keogh's (Keogh, Pullis, & Caldwell, 1982) Teacher Temperament Questionnaire, the third
factor, labeled "reactivity," included both negative affect and sensitivity items. In research with
the Children's Behavior Questionnaire (CBQ), positive correlations between perceptual
sensitivity and discomfort are regularly obtained (Goldsmith, Buss, & Lemery, 1997; King &
Wachs, 1995; Rothbart, Posner, & Hershey, 1995), but in the CTPQ, perceptual sensitivity was
moderately related to openness and social extraversion (including positive affect), and slightly
negatively related to the internalizing negative affects. Andersson, Bohlin, and Hagekull (1999)
found that parent reported reactivity to sensory stimulation at 10 months was related to greater
stranger wariness at 10 months and social inhibition at 25 months. Martin et al. (1994) have also
reviewed three factor analytic studies finding sensitivity items to load with items assessing
negative emotionality. Laboratory research assessing children's sensory thresholds along with
their thresholds for pleasure and discomfort would be helpful in further testing this theoretical
relationship.

Anger/Irritability
In Gray's (1991) model, the Fight/Flight system is constituted by circuits connecting the
amygdala, ventromedial nucleus of the hypothalamus, central gray region of the midbrain, and
somatic and motor effector nuclei of the lower brain stem processing information involving
unconditioned punishment and non-reward. When there is detection of painful or frustrating
input, the brain stem effectors produce aggressive or defensive behavior. Individual differences
in reactivity of this fight/flight system are also thought to underlie aggressive aspects of
Eysenck's general Psychoticism dimension, and Panksepp (1982) describes similar neural
circuitry in connection with a "rage" system (see review by Rothbart, Derryberry et al., 1994).
Important distinctions have more recently been made, however, among varieties of
aggression and anger, and their underlying neural systems. Aggression as a self-defense reaction
seems to be based upon the functioning of the same amygdala circuits as involved in the
production of fear (Blanchard & Takahashi, 1988). Aggression linked to protection of resources,
competition, and offensive aggression, on the other hand, involves a different system based on
the monoamine Dopamine (DA; Lawrence & Calder, in press). The DA system has been linked
to both the production of offensive aggression (Smeets & González, 2000), and to the
recognition of anger in the human face (Lawrence, Calder, McGowan, & Grasby, 2002). In
Lawrence et al.’s study, DA blockade impaired the recognition of anger, while spanning
recognition of other emotions and of facial identity.
We have noted above that Depue and Iacono (1989) suggest the dopaminergic system
may facilitate irritable aggression aimed at removing a frustrating obstacle, consistent with
findings that DA agonists (e.g., amphetamine) can enhance aggressive behaviors. Their view
suggests links between approach and frustration/anger, and, in fact, children’s activity level and
anger have been consistently positively related in parent-reported temperament (Rothbart &
Derryberry, 2002). In addition, infant activity level predicts not only positive emotionality at age
seven, but also higher anger/frustration and low soothability-falling reactivity (Rothbart et al.,
2000). Together with findings relating seven-year Surgency to aggression (Rothbart et al.,
2000), this suggests that strong approach tendencies may be linked to negative as well as positive
emotions (Derryberry & Reed, 1994; Rothbart et al., 2000). Children who showed a short
 Temperament 16

latency to grasp objects at 6.5, 10, and 13.5 months showed high levels of positive anticipation
and impulsivity at age seven, as well as high anger-frustration and aggression, again suggesting
that strong approach tendencies can contribute to later anger-related negative emotions as well as
to positive emotionality.

Negative Emotionality or General Distress Proneness

Negative Emotionality or distress proneness is often viewed as a general dimension
subsuming emotions of fear, anticipatory anxiety, sadness, frustration/anger, guilt, and
discomfort. For example, the “Five-Factor” model of adult personality includes negative
emotions as components of the Neuroticism superfactor. Neuroticism/Negative Emotionality has
been found to be orthogonal to Extraversion/Positive Emotionality (Eysenck & Eysenck, 1985;
Tellegen, 1985, Watson & Clark, 1992). As evident in our previous discussion, however, the
positive relationship between anger/frustration and strong approach tendencies suggests that a
more differentiated model is needed.
As noted above, separable neural systems have been found to be related to different
forms of negative affect. There are nevertheless several possibilities for identifying higher order
negative emotion reactions. One is the link between systems supporting fear and defensive
aggression (Blanchard & Takahashi, 1988). Defensive aggression in animal models seems to be
based on the same amygdalar circuits as fear, and in humans, anger in response to threat may
also be linked to fear.
Negative affect systems are also regulated by more general neurochemical systems
including dopaminergic and serotonergic projections arising from the midbrain, and by
circulating gonadal and corticosteroid hormones (Rothbart, Derryberry et al., 1994; Zuckerman,
1995). Neurochemical influences may thus also provide coherence of emotional states within an
individual, and support more general factors of temperament such as negative emotionality. For
example, serotonergic projections from the midbrain raphe nuclei appear to moderate limbic
circuits related to anxiety and aggression (Spoont, 1992). Low serotonergic activity may thus
increase an individual's vulnerability to both fear and frustration, contributing to a general factor
of negative affectivity, including depression. Gonadal hormones are related to both positive
affect and aggressiveness (Zuckerman, 1991), possibly influencing individual differences in
positive and angry states. Neural structures can thus can support variability at broad as well as
specific levels, although more research in the area is definitely needed.

Affiliativeness/Agreeableness
We share with other animals, including mammals, birds, and fish, systems of affiliation
that support pair bonds and the care of the young (Insel, 2003). Panksepp (1986c) indicates that
affiliative and prosocial behaviors may depend in part on opiate projections from higher limbic
regions (e.g., amygdala, cingulate cortex) to the ventromedial hypothalamus, with brain opiates
promoting social comfort and bonding, and opiate withdrawal promoting irritability and
aggressiveness. Because ventromedial hypothalamic lesions dramatically increase aggression,
Panksepp (1986b) also suggests that this brain region normally inhibits aggressive behaviors
controlled by the midbrain's central gray area. Hypothalamic projections can allow for friendly,
trusting, and helpful behaviors between members of a species by suppressing aggressive
tendencies. Mechanisms underlying prosocial and aggressive behaviors would in this way be
reciprocally related, in keeping with the bipolar Agreeableness-Hostility dimension found in Five
Factor Models of personality. Panksepp (1993) has also reviewed research suggesting links
 Temperament 17

between social bonding and the hypothalamic neuropeptide oxytocin (OXY), involved in
maternal behavior, feelings of social acceptance and social bonding, and reduction of separation
distress. OXY is also released during sexual activity by both females and males.
Agreeableness, including at the high end, the prosocial emotions and behaviors and
affiliative tendencies, and at the low end, aggression and manipulativeness, has been increasingly
studied in childhood (Graziano, 1994; Graziano & Eisenberg, 1997; Graziano & Tobin, 2002).
Like other originally bipolar dimensions, prosocial and antagonistic dispositions have also been
studied separately (Bohart & Stipek, 2001), and Graziano and Eisenberg (1997) suggest that the
two dispositions may be separable, even though negatively related. On a related issue, Shiner
and Caspi (2003) point out that antisocial and prosocial behavior have different etiologies
(Krueger, Hicks, & McGue, 2001). Any temperamental predisposition to pro-social behavior
needs to be seen as an open system, interacting with social experience for its outcomes. In
research described above linking temperament to personality in early and middle childhood, two
forms of extraversion/surgency have been identified; one linked to prosocial behavior and the
other to antisocial behavior and aggression (Victor, Baker, & Rothbart, 2004), again suggesting
the importance of socialization to pro- or anti-social behavior, and reminiscent of Gesell’s (1928,
as cited in Kessen, 1965) comment earlier in our chapter that CD could become a delinquent or a
willing and responsible worker.

Attentional Networks
Functional neuroimaging has allowed many cognitive tasks to be analyzed in terms of the
brain areas they activate, and studies of attention have been among the most often examined
(Corbetta, Kincade, & Shulman, 2002; Driver, Eimer, & Macaluso, in press; Posner & Fan, in
press). Imaging data support the presence of three networks related to different aspects of
attention, carrying out the functions of alerting, orienting, and executive control (Posner & Fan,
in press). We discuss orienting and executive control in this section, although alerting is also
likely to prove of interest in future temperament studies.
Orienting involves aligning attention to a source of signals. It may be overt, as in eye
movements, or covert, occurring without any movement (Posner, 1980). Orienting can be
manipulated by presenting a cue indicating where in space an event will occur, thereby directing
attention to the cued location (Posner, 1980). Orienting to visual events has been associated with
posterior brain areas, including the superior parietal lobe and temporal parietal junction as well
as the frontal eye fields (Corbetta et al., 2002). Lesions of the parietal lobe and superior
temporal lobe have been consistently related to difficulties in orienting (Karnath, Ferber, &
Himmelbach, 2001).
Orienting early in life is a reactive aspect of attention, and children differ both in their
latency to orient and their duration of orienting (see review by Ruff & Rothbart, 1996). In the
IBQ, individual differences in duration of orienting in infancy are positively related to smiling
and laughter and vocal activity, suggesting that orienting may be part of an early positive
reactivity or interest system (Rothbart, 1988; Rothbart, Derryberry et al., 2000).
The second major control system over reactive approach and action (the first is fearful
inhibition as discussed above) is that of effortful control, supported by development of the
executive attention network (Posner & Rothbart, 2000). Executive attention and effortful control
are related to volition and awareness of input (Posner & Rothbart, 1991). There are limits on
how much we can simultaneously attend to in directed thought or action. Areas of the midfrontal
lobe, including the anterior cingulate gyrus, appear to underlie a general executive attentional
 Temperament 18

network (Vogt, Finch, & Olson, 1992), in combination with lateral prefrontal areas. The anterior
cingulate represents the outflow of the limbic system, and is thus closely tied to emotion. It also
has close connections to adjacent motor systems. Activity of the anterior cingulate is modified
by dopamine input from the underlying basal ganglia. The anterior cingulate structure consists
of alternating bands of cells with close connections to the dorso-lateral frontal cortex and to the
posterior parietal lobe (Goldman-Rakic, 1988), suggesting a highly integrative role. The anterior
cingulate thus appears to provide an important connection between widely different aspects of
attention (e.g., attention to semantic or emotional content, or visual location).
Persistence, a dimension of personality conceptually related to effortful control in
temperament, has been related to brain activation (Gusnard et al., 2003). The effects of
persistence act strongly on midline and lateral prefrontal areas that are quite different than those
found active for positive and negative affect, suggesting regulatory aspects of persistence. An
increasingly accepted view (Posner & Rothbart, 2000) is that effortful control, represented in
midline frontal areas, acts to regulate brain areas like the amygdala that are more clearly related
to reactive aspects of negative affect.
In additional research, children who showed rapid approach as infants tended to be low in
attentional control and inhibitory control at age seven, consistent with findings of a negative
relation between Surgency/Extraversion and Effortful Control (Rothbart, Derryberry et al.,
1994), and suggesting that strong approach tendencies may constrain the development of
voluntary self control. If approach tendencies are viewed as the “accelerator” toward action, and
inhibitory tendencies as the “brakes,” it is not surprising that stronger accelerative tendencies
may weaken the braking influence of inhibitory control (Rothbart & Derryberry, 2002).
Executive control of attention is often studied by tasks that involve conflict, such as
varieties of the Stroop task, where subjects are asked to respond with the color of ink (e.g., red)
while ignoring the color word name (e.g., blue) (Bush et al., 2000). Resolving conflict in the
Stroop task activates midline frontal areas (anterior cingulate) and lateral prefrontal cortex
(Botvinick, Braver, Barch, Carter, & Cohen, 2001; Fan, Flombaum, McCandliss, Thomas, &
Posner, 2003). There is also evidence for activation of this network in tasks involving conflict
between a central target and surrounding flankers that may be congruent or incongruent with the
target (Botvinick et al., 2001; Fan, McCandliss, Sommer, Raz, & Posner, 2002).
Regulatory Functions of Executive Attention. The anterior cingulate gyrus, one of the
main nodes of the executive attention network has been linked to a variety of specific functions
(Posner & Fan, in press), including working memory (Duncan & Owen, 2000), emotion (Bush et
al., 2000), pain (Rainville, Duncan, Price Carrier, & Bushnell, 1997), monitoring for conflict
(Botvinick et al., 2001), and monitoring for error (Holroyd & Coles, 2002). In emotion studies,
the cingulate is often seen as part of a network involving orbital frontal and limbic (amygdala)
structures. The frontal areas seem to have the ability to interact with the limbic system
(Davidson, Putnam, & Larson, 2000), fitting well with the idea of their supporting self-
regulation.
A self-regulatory role for the cingulate has been identified in imaging studies with adults.
Cingulate activity was greater when subjects were instructed to control the amount of negative
affect felt in viewing a picture (Ochsner, Bunge, Gross, & Gabrieli, 2002). When hypnotism was
used to control the perception of pain due to heat, cingulate activity also reflected the perceived
degree of pain rather than the physical intensity of the heat stimulus (Rainville et al., 1997).
Finally, large lesions of the anterior cingulate either in adults or children result in great difficulty
in regulating behavior, particularly in social situations (Anderson, Damasio, Tranel, & Damasio,
 Temperament 19

2001). Smaller lesions may produce only a temporary inability to deal with conflict in cognitive
tasks (Turken & Swick, 1999; Ochsner et al., 2002). These results link the anterior cingulate to
regulation of neural activity related to emotion and behavior, and provide evidence for a role of
the cingulate as a part of a network involved in regulation, cognition, and affect (Bush et al.,
2000; Rueda, Posner, & Rothbart, 2004).
Development of executive attention between 2 and 7 years is indexed by marked changes
in the ability to deal with conflict, as well as the ability to detect and slow subsequent response to
errors (Mezzacappa, 2004: Rueda, Fan, et al., 2004; see reviews by Rothbart & Rueda, in press;
Rothbart, Posner, & Kieras, in press). Between 30 to 36 months, children are able to perform a
variant of the Stroop task, the Spatial Conflict key press task, which in adults is related to
activation of the anterior cingulate. In this task, conflict occurs when a stimulus is presented on
the side of the screen opposite its corresponding key. The dominant response is to press the key
consistent with the object's spatial location; the subdominant response is to press the key that
matches the stimulus. At 24 months, children are unable to properly perform this task, but by 30
months, most children can perform it, but are slowed by the conflict, as are adults (Gerardi-
Caulton, 2000; Rothbart, Ellis, Rueda, & Posner, 2003). Children performing more efficiently
on spatial conflict were rated by their parents as having relatively higher levels of effortful
control and lower levels of negative affectivity.
Summary. Models from neuroscience have been developed related to general dimensions
of Approach, Fearful Inhibition or Harm Avoidance, Irritability (Fight/Flight or Rage), and
Affiliativeness or Social Reward Dependence, Orienting, and executive attention as a basis of
Effortful Control. Optimal level models proposing a link between sensitivity and affect have
also found some support in the developmental literature. These dimensions offer a beginning for
future work that will more finely differentiate the temperament domain, its development, and
relation to the development of personality. In our review, we now move to considering
measurement issues in the study of temperament in childhood, providing extensive evaluation of
parent-report. It will be helpful for the reader to consider this material in addition to that put
forward by Kagan and Fox in this volume. We then discuss recent genetics research and other
psychobiological approaches to the study of temperament.

MEASUREMENT APPROACHES

Approaches to measuring temperament in children have included caregiver reports, self-

reports for older children, naturalistic observations, and structured laboratory observations (see
Table 3). Each of these approaches offers relative advantages for temperament study. Caregiver
report questionnaires, for example, can tap the extensive knowledge base of caregivers, who
have seen the child in many different situations over a long period of time. Questionnaires are
________________________
Table 3 about here
________________________

also convenient – they are relatively inexpensive to develop, administer, and analyze, and allow
the study of multiple variables (Bates, 1989a, 1994). Naturalistic observations, on the other
hand, can possess high degrees of objectivity and ecological validity; whereas laboratory
observations allow the researcher to precisely control the context and specific elicitors of the
child's behavior, as well as the time course and intensity of the child’s reaction.
 Temperament 20

In addition to their respective advantages, however, each technique also is subject to

error. In caregiver report measures, there may be perceptual or response biases in the informant.
Naturalistic observations are expensive, and often show relatively low day-to-day reliability so
that it becomes difficult for researchers to collect an adequate sample of relevant behaviors.
Laboratory procedures often limit the kinds of behavior that can be elicited, and repeated testing
necessary for measuring a complex trait in the laboratory may be unfeasible or involve carryover
effects. More detail on measurement issues can be found in Bates (1987, 1989a, 1994),
Goldsmith and Rothbart (1991), Rothbart and Goldsmith (1985), and Slabach, Morrow, and
Wachs (1991). In the present chapter, we focus on the issue of the scientific acceptability of
caregiver reports. Although the literature often provides admonitions about parent report, it
often does not analyze the strengths and limitations of each measurement approach. We have
recognized limitations of caregiver report (e.g., Bates, 1980), but nevertheless have found that
caregiver reports have broadly established validity (Rothbart & Bates, 1998).

Meanings of Parent-Reports
Parent-reports have been extensively used in personality, clinical, and developmental
research, including the study of temperament. At the same time, the validity of parent-reports
about children's temperament has been particularly questioned especially by Kagan (1998;
Kagan & Fox, this volume). We provide an alternative view to the one he has presented.
Digital versus Analog Validity. Determining the validity of parent-reports has often been
framed in an absolute or "digital" way, leading to a simple judgment of whether parent-reports
are valid or not. Thomas et al. (1963) asked whether a significant correlation existed between
parent-reports and independent ratings, and finding significant correlations, concluded that
parent-reports were valid measures of temperament. More typically, however, when statistically
significant correlations between parent ratings and independent ratings have been fairly small in
size, the conclusion has been that parent-reports are not valid. Any low correlation, of course,
could be due to problems with observer ratings as well as parent ratings, but this issue is seldom
addressed.
Early in the discussion of the meaning of parent-reports of temperament, Kagan (1982)
advocated a digital view of validity, and has continued to elaborate this view. In his earlier
writings, Kagan (1994; 1998) argues that parent-reports are not worthy of use in scientific
studies of temperament. In Kagan and Fox (this volume), it is concluded that parent report data
should be supplemented by observation. Our own position, reached more or less independently
(Bates, 1994; Rothbart, 1995; Rothbart & Hwang, 2002), is also that temperament research
benefits from the use of multiple measures. We would agree with the statement of Vaughn et al.
(1992), discussing measures of attachment security; that, “it would be most unfortunate if
pretensions to methodological rigor forced investigators to ignore sources of relevant
developmental information” (p. 470).
Caregivers’ Vantage Point versus Bias and Inaccuracy. One argument for the continued
use of parent-reports of temperament is that they provide a useful vantage point for observations.
Temperament dimensions are by definition general patterns of responses by the child, and
parents are in a good position to observe the child's behavior on multiple occasions, including
infrequently occurring behavior that nevertheless may be critical to defining a particular
temperament dimension. Most families minimize noxious stimulation for their babies, for
example, so that it is difficult to observe such situations naturalistically. Parents, in contrast, can
describe an infant's response to a variety of naturally occurring stimuli, like being given a
 Temperament 21

shampoo, or hearing the vacuum cleaner start up. Parent observations also meet concerns about
ecological validity, as well as ethical constraints about creating aversive situations in order to
assess temperament in the laboratory.
Kagan (1994, 1998), on the other hand, has argued that parent-reports have problems
with bias and inaccuracy. Bias and inaccuracy are real concerns, but in our view, they are not as
great a problem as Kagan suggests. Similar concerns have been extensively dealt with in
personality research, and the dominant conclusion has been that traits can be reliably assessed by
ratings of knowledgeable informants, including the self, friends, and parents (Kenrick & Funder,
1988; Moskowitz & Schwarz, 1982). In addition, validity is a problem for structured and
naturalistic observational measures of temperament as well as for parent-reports, and we have
summarized potential sources of measurement error in three temperament assessment methods in
Table 3.
Behaviors reliability coded in precisely defined situations have a high degree of
objectivity. This does not mean, however, that the observations also have high validity.
Observational research needs to demonstrate the same kinds of validity (content, construct,
convergent, discriminant) required of a parent-report measure (Bates, 1989a; Rothbart &
Goldsmith, 1985). There are some very promising laboratory assessments of temperament (e.g.,
Garcia-Coll, Halpern, Vohr, Seifer, & Oh, 1992; Goldsmith & Rothbart, 1991; Kagan, Reznick,
& Snidman, 1988; Matheny, Wilson, & Thoben, 1987), but none has become so established that
it can be seen as the gold standard for temperament measurement.
Kagan (1994) has also argued that the language of an individual item on a temperament
questionnaire is subject to multiple interpretations. This ambiguity, however, is the main reason
why researchers use scales of items rather than individual items to measure temperament.
Attempts are made to write the best possible items, but the researcher does not expect that by
doing this, all sources of error will be eliminated. Basic psychometric theory holds that the
reason a set of convergent, but imperfectly correlated items tends to have better test-retest
reliability, better stability over time, and better validity is that the error components of individual
items tend to cancel each other out when the item scores are added to each other, yielding a
closer approximation to a "true" score. This principle is true of aggregation across multiple
observations as well as multiple items. Fortunately, one need not be limited to simply adding
items and hoping that error is thereby reduced. With analytical tools such as LISREL and EQS,
one can also explicitly model linkages between items' and scales' error components, creating
latent constructs that more precisely control for measurement error. Other approaches to
reducing concerns about validity are to use validity scale filters as in the MMPI, and to study the
ways parents construe child behavior and the items researchers present to them (Bates, 1994).
A Components-of-Variance Approach to Validity. We prefer to frame the question of
validity of parent-reports in terms of components of variance, judging a measure’s validity on a
continuum rather than an absolute judgment. Bates and Bayles (1984) asked how much variance
in parent-reports could be explained by reports of independent observers in a series of second-
order empirical analyses on data from their longitudinal study. They found that: (1) mother
ratings of their children on an array of temperament and non-temperament traits showed
appropriate convergent and discriminant relations on similar sets of scales from 6 months to 3
years of age; (2) fathers and mothers agreed at generally moderate levels; (3) mothers and
observers (in both naturalistic and structured contexts) agreed at generally modest, but
significant levels; and (4) factors such as anxiety or tendency to describe oneself in socially
desirable ways, which could reflect subjective biases, accounted for only modest portions of the
 Temperament 22

variance. Measured subjective factors thus did not overshadow measured objective factors as
explanations of differences in parents' perceptions of their children. In addition to these
components of parents' perceptions, there remained error components.
Matheny et al. (1987) provided independent support for this model, using an array of
laboratory measures. Their aggregated maternal report scores correlated moderately to strongly
with laboratory scores of temperament: r = .52 at 12 months, .38 at 18 months and .52 at 24
months. Their conclusion was that "the objective component of maternal ratings was clearly
demonstrable and prominent" (Matheny et al., 1987, p. 324). They also showed that maternal
personality characteristics were not only correlated with mothers' perceptions of the child; they
were also correlated with their children's behavior as independently observed in laboratory
situations, a finding congruent with genetically based similarities between mother and child.
A pattern of moderate to strong validity correlations for parent-report can now be found
in a number of places in the literature. One important requirement for ascertaining construct
validity is that both measures demonstrate adequate reliability, and often the observational or
mechanical measures, not the parent-report measures, are deficient in this regard. To produce
adequate reliability, aggregation across multiple measures is often necessary (Rushton, Brainerd,
& Pressley, 1983). Eaton (1983) recorded activity level from actometers worn by preschoolers
over repeated nursery school free play sessions. Reliability of the actometers was .13 within a
single session, but rose to .75 when multiple sessions were aggregated (Eaton, 1994).
Aggregated scores also correlated .75 with parent report temperament ratings using the CCTI
activity level scale and .73 with composite staff ratings of child activity level.
Asendorpf (1990) used multiple measures on multiple occasions to assess children's
behavioral inhibition to strangers (shyness) across a four-year period beginning at age 3.
Measures included a parent-report assessment as well as observations of children's behavior with
strange adults and children. Of all the measures taken by Asendorpf, parent-report consistently
showed the strongest relations with the other measures; for example, parent-reported shyness
predicted latency to talk to a stranger at 3 years with r = .67; the overall average r between
parent-report and other shyness measures across the four years ranged from .43 to .53. Bishop.
Spence, and McDonald (2003) also reported convergence between parent ratings on a new
behavioral inhibition questionnaire and both teacher questionnaires and structured observation
measures.
Laboratory measures have also been found to be positively related to the IBQ and TBAQ
(see Goldsmith & Rothbart, 1991), and between temperament measures and model tasks
designed to reflect underlying brain function; for example, a positive relation was found between
a laboratory Spatial Conflict task designed to assay executive attention and the CBQ measure of
inhibitory control (r = .66) for 36-month-old children (Gerardi-Caulton, 2000). Rothbart, Ellis,
and Posner (2004) reported further relations between effortful control and scales and spatial
conflict scores. These findings provide further validational support for parent-reports of
temperament, but agreement is not always found (see Kagan & Fox, this volume). A
comprehensive review of validity studies in the measurement of temperament is needed.
Goldsmith et al. (1991) correlated mother reports with those of daycare teachers in
samples of preschoolers, toddlers, and infants. Using a variety of standard temperament scales,
they found strong convergence between scales from different questionnaires measuring the same
construct, and generally acceptable divergence between scales expected to differ. Correlations
between mother and daycare teacher for two older groups were in the typical range for
correlations between parents and other observers (.11-.50 for preschoolers, with the highest
 Temperament 23

correlation on activity level, and .00-.35 for toddlers, with the highest correlation also for
activity), and perhaps a little above this range for infants (.21-.60, with the highest correlation on
one of the measures of approach-sociability). Daycare teachers would presumably be well
acquainted with the children, although Goldsmith et al. did not report the degree of acquaintance.
However, Goldsmith et al. (1991) emphasize the difference in contexts between mother and
teacher observations.
As evidence for potential impact of failure to control context, they cite the Hagekull,
Bohlin, and Lindhagen (1984) study. Hagekull et al. (1984) asked parents to directly record
infant behavior, such as infants' reactions to loud sounds, over extended periods in specific
situations. Parent data converged strongly with independent observers' data: Correlations
between parents’ and observers’ direct observation data for two, 4-hour visits ranged from .60
(for attentiveness) to .83 (for sensory sensitivity). This suggests, contrary to the argument of
Seifer, Sameroff, Barrett, & Krafchuk (1994), that parents are not necessarily deficient or
strongly biased in their powers of observation, especially since their training for the task was
minimal. Hagekull et al. (1984) also found that open-time-frame, general questionnaire scales
completed by the parents converged to a modest to moderately strong degree with scales based
on independent direct observation, with correlations ranging from .21 to .63. We attribute the
apparent improvements in observer-parent agreement coefficients more to the study's careful
effort to observe sufficiently large numbers of key events than to any conceptual or psychometric
advantages in the questionnaire they used (the Baby Behavior Questionnaire or BBQ). Although
BBQ scales were developed though factor analysis, some of the scales in this instrument have
some difficulties in interpretation, related to apparently heterogeneous content.
Prenatal Perception Studies. Researchers have also studied parent expectations of
temperament before the child is born, which are often significant predictors of postnatal ratings
of temperament (Diener, Goldstein, & Mangelsdorf, 1995; Mebert, 1991), and we discussed
these in detail in our 1998 handbook chapter. Diener et al. (1995), for example, looked at mother
and father agreement over time. Prenatally, mothers' and fathers' temperament expectations were
only modestly to moderately correlated, and the correlation pattern was generally non-
differentiated. Mothers' expectations of unadaptability to novelty, for example, were more
highly related to fathers' fussy/difficult expectations than to their unadaptability expectations.
Postnatally, however, mother-father convergence was considerably stronger, and there was also a
strong pattern of discriminant validity. For example, mothers' ratings of unadaptability were
correlated with fathers' ratings of unadaptability .67, and with fathers' ratings of difficultness
only .28.
Mebert (1991) and Diener et al. (1995) speculated that prenatal expectations reflected a
vague internal working model of the infant before birth, which might influence temperament
through expectancy confirmation processes (Darley & Fazio, 1980). However, the fact that
mother and father perceptions of the infant become so much closer in both a convergent and a
discriminant sense from before to after their actual experience with the baby can be interpreted
as evidence for an objective component in the ratings.
Recent Attempts to Make Parent and Observer Perspectives More Similar. Although
some writers have argued that modest parent-observer agreement is simply a product of low
validity in the parents' reports, this is not a necessary conclusion. Modest correlations could, for
example, reflect observers simply not seeing the behavior parents based their reports on.
Naturalistic and structured observation measures are often based on between 30 minutes and 4
 Temperament 24

hours observation, and only a few are based on as much as 6 hours total. There is also little
evidence that these measures show high test-retest reliability.
Two extensive home observation studies attempted to address such problems (Bornstein,
Gaughran, & Segui, 1991; Seifer et al., 1994). In our 1998 handbook chapter, we described
these studies in detail, indicating limitation in the designs, and concluding that the ideal large
scale study for the limits of the objective component of parent-reports of temperament has not
yet appeared (Rothbart & Bates, 1998). Studies approaching the ideal would require
conceptually well developed measures with careful attention to both parent-report and observer
report forms. Extra attention would be devoted to validating observer as well as parent report
measures, testing for convergence and divergence among measures, as well as studying relations
between the measures and alternate ways of observing (e.g., summary ratings versus
independently recorded molecular behavior frequencies, or naturalistic versus structured
observations). One recent example of such a study is that of Forman et al. (2003), who showed
that aggregating laboratory measures of temperament across multiple tasks enhanced
convergence with mother report questionnaire measures.
The design would also pay greater attention to assessing contexts of temperament-
relevant behavior. Although the concept of temperament implies some degree of cross-
situational consistency, there is no reason to suppose that any given trait should be equally well
revealed in all contexts, given variability in instigation of the response. The issue of context is
crucial to all forms of temperament assessment, and will prove important as early temperament
characteristics are linked to developing coping strategies across varying situations.
Shall We Use Parent Reports? As we concluded in 1998, evidence to date supports the
careful use of parent-report measures of temperament. Two basic reasons to use parent-report
measures are a) that they provide a useful perspective on the temperament of children, since
parents can see a wide range of child behaviors, and b) that they have established a fair degree of
objective validity. In addition, parent-report measures have contributed to substantial empirical
advances, including our understanding of the structure of temperament in relation to the Big Five
or Big Three models, as described above, and their parallels in psychophysiological systems
(Bates, Wachs, & Emde, 1994; Rothbart, Derryberry et al., 1994). A further reason for using
parent-reports is that the social relationship aspects of temperament elicited from parents may in
themselves be important to understanding development.
Although we draw the conclusion that caregiver reports are useful in research, we share
the concerns expressed in the literature about their measurement issues. Caregiver reports must
be carefully interpreted as reflecting a combination of subjective and objective factors (Seifer,
2003). Improvements in caregiver report measures should recognize possible sources of bias,
such as parents’ tendency to contrast one child with another in rating temperament (Saudino,
2003) on some, but not all parent report questionnaires (Hwang & Rothbart, 2003). We should
also develop sub-scales to detect specific biases in reporters and improve the generalizibility of
the observational measures we use to validate caregiver reports (Goldsmith & Hewitt, 2003).
Observational and laboratory measures are appealing, and should also be employed in
temperament studies; however, they should not at this time be the sole measure of temperament.
The primary arguments for this position are that (a) the validity of a number of such measures of
temperament is not strongly established, and (b) even if the measures were well-validated, they
would often be awkward and highly expensive to use. Improvements in both parent-report and
observational measures are needed. It is also likely that parent reports can be made more
objective, and that the subjective components can be modeled more accurately and even
 Temperament 25

controlled for (see Bates, 1994). The construct validity of observational measures can also be
improved. We now turn to a review of research on additional temperament-related measures
assessing the neural substrates of temperament-related behavior.

PSYCHOBIOLOGICAL RESEARCH APPROACHES

Gunnar (1990) describes five assumptions guiding psychobiological research on

temperament (all quotations from p. 393): (1) "the assumption that temperament variation is
regulated by the central nervous system"; (2) "the assumption that measures of peripheral
systems inform us about the physiological bases of temperament because peripheral activity is
regulated centrally," allowing the use of non intrusive measures, such as heart rate or
electrodermal response; (3) "the assumption that fundamental temperament and emotional
processes reflect a common mammalian heritage (Panksepp, 1982)," allowing research on animal
models; (4) "the assumption that the aspects of central functioning related to temperament
variation are those linked to broad or general behavioral tendencies"; and (5) "Finally, as
reflected in Rothbart and Derryberry's theory (1981), concepts such as reactivity or arousal and
self-regulation or inhibition are central to most physiological theories of temperament" (Gunnar,
1990, p. 393). We have already adopted a number of these assumptions in the course of this
review, and they are further illustrated in this section.

Behavioral Genetics
One reason for adopting a psychobiological approach to temperament is the considerable
body of research indicating genetic contributions to the development of temperament and
personality. Results of this work are reported extensively by Caspi and Shiner (this volume),
Goldsmith (1989; Goldsmith, Losoya, Bradshaw, & Campos, 1994), Plomin (Plomin, Chipuer, &
Loehlin, 1990), Bouchard and Loehlin (2001), and for animal studies, Wimer and Wimer (1985).
Because these reviews are available elsewhere, we briefly review findings that appear promising
for an understanding of temperament and social development. Heritability estimates from
behavioral genetics studies calculate the proportion of phenotypic (observable) variance in a
characteristic attributable to genetic variation within a population, and heritability has proven to
be substantial for most broad temperament and personality traits. In the area of personality,
broad traits tend to show heritability estimates in the vicinity of .50 (Bouchard & Loehlin, 2001).
Tellegen et al. (1988) reported studies of adult MZ and DZ twins who had been reared
either together or apart. Overall correlations of traits for MZ twins reared apart were surprisingly
of a magnitude usually found for identical twins raised together (average r = .49), with
heritability estimates of about .50. Correlations for MZ twins raised apart were .61 for stress
reaction, .48 for sense of well-being, .50 for control, .49 for low risk taking, and .46 for
aggression. For the three superfactors of Positive Emotionality (extraversion), Negative
Emotionality (neuroticism) and Constraint (effortful and fearful control), only Positive
Emotionality showed evidence of higher correlations for MZ and DZ twins raised together in
comparison with twins raised apart (MZ apart r = .34, together = .63; DZ apart r = -.07, together
= .18).
Goldsmith et al. (1997) have reviewed developmental behavioral genetics research and
presented their own findings. Reviewing major twin studies, Goldsmith et al. (1997) reported
that parent-report measures yield monozygotic (MZ) twin correlations ranging from .50 to .80,
 Temperament 26

with dizygotic (DZ) correlations ranging from zero to .50. For scales based on Buss and
Plomin's (1984) EAS measure, DZ correlations are typically less than half MZ correlations and
often near zero, creating problems for heritability estimates. Although this pattern could also be
due to the interactive effects of multiple genes, it seems more likely to be due to perceptual
tendencies to contrast fraternal twins (Saudino, 2003). Despite the tendency for contrast effects
to spuriously inflate heritability estimates, there is substantial evidence of heritability. For
example, Silberg et al. (2004) showed heritability for ICQ traits in infants, even controlling for
contrast effects.
Evidence for larger DZ correlations that are more in line with expected values has been
found using observational methods (see Kagan and Fox's discussion of the genetics of behavioral
inhibition, this volume), and for parent-report studies employing the IBQ (Goldsmith, 1993) and
the TBAQ, but not the CBQ (Goldsmith et al., 1997). Research employing these last three
measures also suggests shared family influence, as did the MacArthur Longitudinal Twin Study
(Plomin et al., 1993) for parent-reported positive affect and approach (Goldsmith et al., 1997).
Goldsmith and Gottesman (1977) have also found evidence for genetic and shared family
influences on CBQ effortful control scales.
In a study involving 3- to 16-month-old infants using the IBQ and a laboratory
supplement at 9 months, genetic and shared environmental effects depended on the particular
dimensions of temperament studied (Goldsmith, Lemery, Buss, & Campos, 1999). Additive
genetic effects accounted fully for Distress to Limitations, Fear, and Activity Level measures,
shared family effects accounted for Soothability, and genetic, shared, and non-shared
environmental effects accounted for Smiling and Laughter and Duration of Orienting. The
covariation of mother and father report and lab measures of stranger distress reflected chiefly
genetic influences. Arsenault et al. (2003) similarly combined data across multiple reporters and
settings, finding stronger genetic effects for antisocial behavior across than within situations in
five-year-old children. Saudino and Cherny (2001) found that covariation between mother and
father reports on the CCTI shyness scale in infants from 14 to 36 months was mediated by
genetic factors, but to a lesser extent than it was mediated by non-shared environment factors.
Goldsmith's positive affect/approach findings are congruent with Tellegen et al.'s (1988)
study of MZ and DZ twins raised together and apart, in that shared family effects are found for
positive emotionality. The shared environmental effect on effortful control reported by
Goldsmith et al. (1997) requires replication, but shared family experience may also prove to be
important in the development of attentional control. These findings may stimulate research into
conditions that promote approach, orienting, positive affect, and self-control in the child’s early
social environment. Silberg et al. (2004) support the finding of a shared environment effect in
addition to genetic influence on an aggregated sociability and positive affect scale on the ICQ.
They also found evidence for a shared environment effect in unadaptability (novelty distress), as
well as genetic heritability.
Although behavior genetics research indicates strong heritability of individual differences
in temperament in the populations studied to date, these findings are based on the usual
environmental circumstances experienced by developing children, and any heritability estimates
reflect genes and environment operating together. The results do not tell us what might be
accomplished via environmental intervention. They also do not reveal the specific
developmental processes involved in temperament and personality outcomes. To learn more
about the latter questions, studies furthering our understanding of temperament and development
 Temperament 27

are essential. Zuckerman (1995) addressed the question of "What is inherited?" and proposed
this answer:
We do not inherit personality traits or even behavior mechanisms as such. What is
inherited are chemical templates that produce and regulate proteins involved in building
the structure of nervous systems and the neurotransmitters, enzymes, and hormones that
regulate them . . . How do these differences in biological traits shape our choices in life
from the manifold possibilities provided by environments? . . . Only cross-disciplinary,
developmental, and comparative psychological research can provide the answers (pp.
331-332).
We now recognize that experiential and environmental processes themselves build
changes in brain structure and functioning (Posner & Raichle, 1994), both before and after birth
(Black & Greenough, 1991). This situation is a far cry from the view that genetics gives us hard-
wiring that determines our future temperament and personality, and it demands developmental
research. An exciting recent approach takes advantage of new methods of molecular genetics,
and we review briefly some of the research in this area.

MOLECULAR GENETICS AND TEMPERAMENT

The mapping of the human genome has provided a promising new direction for studying
genes and environment in development (Carey, 2003; Plomin & Caspi, 1999). Genetic alleles
identified in previous adult research, for example, have been examined in children. An
association between the 7-repeat allele of the DRD4 and novelty seeking in adults was reported
in 1996 (Benjamin, Ebstein, & Belmaker, 1996; Ebstein et al., 1996), although replication of
these results has been inconsistent. In addition, variation in the 5-HTTLPR, a serotonin
transporter gene, had been associated with Five Factor Model Neuroticism scores and with
measures of fear and Harm avoidance (see review by Lesch, Greenberg, Higley, Bennett, &
Murphy, 2002). In a recent imaging study, presentation of fear stimuli also resulted in increased
activation of the right amygdala in adults with the l/s or s/s form of the 5-HTTLPR gene (Hariri
et al., 2002).
Ebstein and his colleagues used a longitudinal sample to investigate these two genetic
polymorphisms in relation to neonatal and later infant behavior (Auerbach et al., 1999; Ebstein et
al., 1998). Ebstein and Auerbach used the NBAS to measure temperament during the neonatal
period, and the IBQ to measure temperament at 2 months (Auerbach et al., 1999; Ebstein et al,
1998). In the newborn examination, the DRD4 long variant that has been linked to sensation
seeking in adults was associated with orientation, range of state, regulation of state, and motor
organization. In addition, an interaction was found between DRD4 and the 5-HTTLPR
polymorphisms. The serotonin transporter gene s/s polymorphism that has been linked to fear
and distress in adults was related to lower orientation scores, but only for neonates who did not
have the long repeat variant of DRD4. For those who did, presence of the 5-HTTLPR s/s
genotype had no effect.
Newborns who demonstrated high orientation and motor organization showed lower
negative emotionality at 2 months. In addition, 2-month-old infants with long repeat DRD4
alleles had lower scores on IBQ negative emotionality and distress to limitations. Infants with
the s/s 5-HTTLPR genotype previously related to fear and distress in adults had higher scores on
negative emotionality and distress, and infants who shared both short repeat DRD4 alleles and
short repeat 5-HTTLPR alleles showed the highest levels of negative emotionality and distress.
 Temperament 28

Thus, the balance between orientation and distress found at the behavioral level (Harman,
Rothbart, & Posner, 1997) may also be reflected at the genetic level. Finally, at 1-year of age,
infants with the long DRD4 allele had lower negative emotionality scores, and showed less fear
and less social inhibition (Auerbach et al., 1999).
Suomi and his colleagues have recently reported interactions between genes and
environment in rhesus monkey studies of the 5-HTTLR gene and development (Barr et al., in
press; Bennett et al., 2002). A standardized temperament assessment was carried out at 7, 14, 21,
and 30 days of age, and monkeys with the 5-HTTLR short repeat allele showed higher levels of
distress, as in human studies (Bennett et al., 2002). In human infants, the short allele was also
linked to lower orientation scores (Auerbach et al., 1999; Ebstein et al., 1998), but analysis of
monkey data instead revealed an environment by gene interaction. The short repeat allele was
related to lower orientation scores, but only for monkeys who had been nursery-reared with
peers, not for mother-reared monkeys. The authors note that a number of factors may account
for the interaction. One is that a general tendency to distress related to being reared in the
nursery may be related to their distractibility; another is that mothers may buffer their infants’
experience so as to moderate the expression of the genetic characteristic. Bennett et al. (2002)
further reported an interaction between the short repeat form of the 5-HTTLR genotype and
rearing condition in relation to CSF concentrations of 5-HIAA, a marker for a disposition to
aggressive behavior.
Barr et al. (in press) found that nursery-reared young monkeys engaged in more social
play than mother-reared monkeys, but if they had the short form of the distress-related 5-HTTLR
allele, nursery reared monkeys showed lower levels of social play, similar to the amount of play
of mother-reared monkeys. Nursery reared animals with the l/s genotype were more aggressive
than either the mother-reared or l/l nursery-reared animals, suggesting the involvement of
serotonin in development of aggression, but only in animals exposed to early-life stress and
maternal deprivation. The authors note that human research also supports both genetic and
environmental contributions to the etiology of aggressive disorders (e.g., Carey, 1996; Dodge &
Pettit, 2003).
In a study of four-year-olds, Schmidt and Fox (2002) found a relation between the long
repeat form of DRD4 and high scores on observed disruptive behavior and parent-reported
aggressive and delinquent behavior. Schmidt, Fox, Perez-Edgar, Hu, and Hamer (2001) also
found a link between the long repeat form and mothers’ reports of attention problems. No links
were found between serotonin transporter alleles and shyness, although they had been predicted.
Children with the long 7-repeat allele of DRD4 related to sensation seeking in adults show
behavioral aspects of ADHD, but do not demonstrate deficits in conflict performance as
measured by the color-word Stroop task (Swanson et al., 2000). Sensation seeking might well be
a prominent characteristic in at least some children diagnosed with ADHD. Evidence from
evolutionary studies suggests that the 7-repeat allele is under positive selective pressure (Ding et
al., 2002), which might be related to its association with sensation seeking, a possible advantage
during human evolution (Ding et al., 2002).
In research on individual differences in attention (Posner & Fan, in press), the anterior
cingulate, associated with executive attention, is only one synapse away from the ventral
tegmental area, a major source of dopamine (DA), and the five types of DA receptors are all
expressed within the cingulate. The Attention Network Task (ANT), which assesses efficiency
of alerting, orienting, and executive attention, was used in a small scale twin study (Fan, Wu,
Fossella, & Posner, 2001). In this study, the executive attention network score showed high
 Temperament 29

enough heritability (.89) to justify the search for specific genes. At least two candidate genes
were found to be related to executive attention (Fossella, Posner, Fan, Swanson, & Pfaff, 2002):
the dopamine D4 receptor (DRD4) gene and the MAOA gene, related to the synthesis of
dopamine and norepinepherine. In a neuroimaging study, these genes also were related to
differences in brain activation within the anterior cingulate gyrus (Fan, Fossella, Sommer, Wu, &
Posner, 2003). The presence of the more common 4 repeat allele, rather than the long repeat
allele of DRD4 that has been related to sensation seeking, was associated with greater difficulty
in resolving conflict (Fossella et al., 2002).
These findings are all recent and require further confirmation and extension, but they
indicate the possible utility of relating genetic differences to specific brain networks and
temperamental characteristics. It will be particularly interesting to look at relationships at
different ages and in connection with different life experiences. Human studies have also
identified significant interactions between gene and environment in maladaptive outcomes, and
we discuss these in the Temperament and Adjustment section below.

Approach/Withdrawal and Hemispheric Asymmetry

We now consider developmental research taking a psychobiological perspective and
employing psychophysiological indicators. In the first of these, differences in cerebral
hemispheric activation have been related to temperamental tendencies toward approach versus
inhibition-withdrawal. Evidence from electrophysiological (EEG) and lesion studies has related
higher anterior left hemisphere activation in response to stimulation to increased positive affect
and/or decreased negative affect (see reviews by Davidson et al., 2003; Davidson & Tomarken,
1989). The reverse relationships – higher anterior right hemisphere activation related to higher
negative affect and/or decreased positive affect – have also been reported. Resting EEG
asymmetries have also been related to positive and negative emotional reactivity (e.g., Davidson
& Fox, 1989; Tomarken, Davidson, Wheeler, & Doss, 1992). Harmon-Jones and Allen (1997),
for example, reported greater left than right-frontal cortical activity in women subjects with
higher scores on Carver and White’s (1994) BAS questionnaire. The BIS measure was not
related to asymmetry. Buss et al. (2003) found relations between extreme right EEG asymmetry
and high basal and reactive cortisol, replicating previous primate findings from Kalin and his
colleagues (Kalin, Larson, Shelton, & Davidson, 1998).
Fox, Calkins, and Bell (1994) reported that infants with stable right frontal EEG
asymmetry between 9 and 24 months of age displayed more fearfulness and inhibition in the
laboratory than other children. At four years, children who showed more reticence and social
withdrawal were also more likely to show right frontal asymmetry. Calkins, Fox, and Marshall
(1996) also found that children selected for high motor activity and negative affect to laboratory
stimulation at 4 months showed greater right frontal asymmetry at 9 months, greater mother
reports of fear at 9 months, and more inhibited behavior at 14 months. However, no concurrent
relation was found between behavior and frontal asymmetry at 9 and 14 months, and greater
activation of both right and left frontal areas was related to higher inhibition scores at 14 months.
The authors (Calkins et al., 1996) suggest a need to differentiate between fearful and angry
distress, as discussed above in the Structure of Temperament section. They also hypothesize that
high motor/high negative affect and high motor/high positive affect may be associated with later
different kinds of problems. For high motor/high positive affect, the problems would be
associated with difficulties in self-control.
 Temperament 30

Autonomic Reactivity and Self-Regulation

By assuming central controls on peripheral reactivity, psychobiological researchers have
developed models of centrally regulated systems that can be studied early in life. In this section,
we consider briefly some of the research on electrodermal responding, heart rate, and vagal tone.
Electrodermal Reactivity. Several early studies reported a relationship between
electrodermal response and introversion (see review by Buck, 1979). Jones (1960), for example,
compared the 10 highest and 10 lowest electrodermal responders age 11-18 in the Berkeley
Adolescent Growth Study. High electrodermal responders were described as showing high
emotional control, quiet, reserve and deliberation, and as being calm and responsible. Low
electrodermal responders were rated as more impulsive, active and talkative, more attention
seeking, assertive and bossy. Adult studies have also found stable individual difference in
electrodermal reactivity to be negatively related to measures of extraversion (e.g., Crider &
Lunn, 1971). Fowles (1982) reported that electrodermal responding, but not heart rate reactivity,
was related to measures of Gray's behavioral inhibition system (BIS).
Fabes et al. (1994) studied kindergarten and second grade children's facial expressions of
distress and skin conductance (SC) reactivity to a film about children being hurt in an accident.
SC variables were used along with maternal and other variables to predict children's helping, in
sorting crayons into boxes for hospitalized children. For both ages, SC reactivity, used as a
marker of personal distress, was positively related to facial distress and negatively related to
helping. Results were seen to reflect an interference of personal negative affect with children's
prosocial behavior. Concurrent heart rate measures were not related to other variables. In a
study of older children (third and sixth graders), SC was positively related to facial expressions
of distress to a film and negatively related to mothers' report of dispositional helpfulness, but for
girls only (Fabes, Eisenberg, & Eisenbud, 1993). Evidence has thus been found for
electrodermal response as both a sign of distress and behavioral inhibition. Lang and his
associates (Lang, Bradley, & Cuthbert, 1997), however, have found that adults' skin conductance
to viewing pictures increased for both aversive and pleasant stimuli, so that the sympathetic
response measured in SC may be more general than previously thought.
Heart Rate and Vagal Tone. A good deal of recent research has focused on heart rate and
on vagal tone as a measure of parasympathetic cardiac control. In her review of heart rate (HR)
research, Von Bargen (1983) reported heart rate reactivity to stimulation to be the most stable
and reliable of HR measures. As noted by Kagan (1998), HR variability has been linked to low
behavioral inhibition, in some but not all studies. Fabes, Eisenberg, Karbon, and Troyer (1994)
found HR variability to be positively related to kindergarten and second grade children's
instrumental coping to a baby's crying. Fabes et al. (1993) also found positive relationships
between HR variability and measures of sympathy (dispositional sympathy for girls, concerned
attention to others' distress for boys) in third and sixth grade children.
Cardiac vagal tone has also been related to temperament (Bornstein & Suess, 2000; El-
Sheikh, 2001). Respiratory Sinus Arrhythmia (RSA), the fluctuation in heart rate occurring at
the frequency of respiration, has been used to assess parasympathetic control via the vagal nerve
(Porges, 1986). There is an increase in heart rate during inspiration and a decrease during
expiration that is extracted from HR variation, and Porges argued that variability in RSA reflects
individual differences in tonic parasympathetic vagal tone. Higher baseline vagal tone is also
related to greater vagal suppression to stimulation, although some infants with regulatory
disorders show high RSA but do not demonstrate suppression of RSA with attention (DeGangi,
 Temperament 31

DiPietro, Greenspan, & Porges, 1991). Vagal suppression is often seen as reflecting attentional
strategies to cope with the environment or respond to stress (Huffman et al., 1998). Berntson,
Cacioppo, and Quigley (1993) note that RSA is not a direct equivalent to tonic vagal control of
the heart, because it is determined by multiple peripheral and central processes. They
nevertheless conclude that RSA is an important noninvasive measure that "shows a high degree
of sensitivity to psychological and behavioral variables" (p. 193).
Keeping these concerns in mind, we can consider some of the findings relating vagal tone
or RSA measures to temperament variables. Porges, Doussard-Roosevelt, and Maiti (1994)
reviewed studies relating newborn vagal tone to irritability, and found that young infants with
high baseline levels of vagal tone were also highly reactive and irritable. Later in development,
however, vagal tone has been found to be related positively to interest and positive
expressiveness and negatively to internalizing distress (for a more extensive review of these
findings, see Beauchaine, 2001; Porges, 1991). Thus RSA after 5-6 months tends to be
associated with positive emotionality and approach as well as irritability. Richards and Cameron
(1987) found that baseline RSA was positively correlated with parent-reported approach at 6-
and 12-months, and Fox and Stifter (1989) reported more rapid approach to strangers in infants
with higher RSA at 14 months. Stifter, Fox, and Porges (1989) found that 5-month-olds with
higher RSA looked away from the stranger more during a strangers' approach and showed higher
levels of interest and positive affect, although this pattern was not found at 10 months. Evidence
of stability of vagal tone is only found after about 9 months of age (Porges & Doussard-
Roosevelt, 1997).
Stifter and Corey (2001) found that greater suppression of vagal tone to cognitive
challenge in 12-month-olds was associated with experimenter ratings of approach, and Fox and
Field (1989) reported more rapid adjustment to preschool in 3-year-olds with higher vagal tone;
these children also showed higher positive affect and greater adaptability. Katz and Gottman
(1995) found that children with low vagal tone at age 5 showed a stronger correlation between
marital hostility at age 5 and problem behaviors at age 8 than children with high vagal tone (rs =
.65 and .25), although the interaction was not significant. Katz and Gottman saw their finding as
congruent with a buffering effect of higher vagal tone that might operate through attentional self-
regulation. El-Sheikh, Harger, and Whitson (2001) also found an interaction: 8- to 12-year-olds
with higher vagal tone appeared to be buffered against anxiety related to high verbal marital
conflict. Finally, El-Sheikh (2001) found that higher vagal suppression to a taped argument in 6-
to 12-year-olds was related as a protective factor to internalizing, externalizing, and other social
problems related to parental problem drinking, whereas negative affectivity was a vulnerability
factor in the effects of parental drinking
Beauchaine’s (2001) interpretation of developmental findings is that higher vagal tone is
associated with more adaptive functioning at any given age: in the neonate, it is linked to
irritability; in the older infant and child, to approach and positive affect as well as to irritability.
Later higher vagal tone is associated with more appropriate social behaviors and adaptation to
stressors, along with lower depressive and anxious psychopathology (see review by Beauchaine,
2001). However, the links between attentional regulation and vagal tone or suppression suggest
more specific interpretations of these findings. In infancy, irritability and anger are positively
related to approach, positive affect, and duration of orienting (Rothbart & Derryberry, 2002), and
infants high in vagal tone may be showing a stronger approach system linked to parasympathetic
function in these responses. Later, infants high in vagal tone and vagal suppression may be
showing stronger attentional regulation. At 9 months, Porges, Doussard-Roosevelt, Portales, and
 Temperament 32

Suess (1995) found vagal tone to be positively correlated with ICQ fussy/difficultness. For a
small longitudinal sample, however, even after partialling out 9-month ICQ difficulty, 9-month
vagal tone predicted lower difficulty at 3 years, possibly related to attention regulation and even
to the development of the executive attention system across this period. Research with adults has
in fact recently linked high heart rate variability to performance on tasks involving executive
attention (Hansen, Johnsen, & Thayer, 2003), and children with higher vagal tone have been
found to show greater ability to sustain attention (Suess, Porges, & Plude, 1994). In any case,
relations between vagal tone and temperament vary depending on the age of the child, and
include both reactive and self-regulative processes.

Cortisol Reactivity
Another approach that takes into account both reactivity and self-regulatory control is the
work of Gunnar and others on cortisol reactivity. During stress reactions, the adrenal cortex
secretes steroid hormones, including the glucocorticoids, cortisol, and corticosterone (Carter,
1986). These hormones increase blood glucose and work with catecholamines to produce
glucose from free fatty acids, also serving an anti-inflammatory function for injury and disease.
Gunnar and her associates have investigated cortisol reactivity in relation to individual
differences in temperamentally based self-regulation, and have reviewed links between stress
hormone activity and development (Gunnar & Cheatham, 2003). Decreases in cortisol reactivity
are found between 2 and 4 months and further decreases between 6 and 18 months (Gunnar,
Brodersen, Krueger, & Rigatuso, 1996; Lewis & Ramsay, 1995).
Both temperament and child care are related to cortisol levels. Dettling, Parker, Lane,
Sebanc, and Gunnar (2000) found that children high in temperamental negative emotionality and
low in self-regulation showed the greatest increase in cortisol levels when they were in less than
optimal childcare situations. Donzella, Gunnar, Krueger, and Alwin (2000) investigated stress
responses to competition in 3- to 5-year-olds. Children played against a familiar adult
experimenter, and initially won 3 games, then lost 3. Temperamental surgency assessed via
teacher report was related to both positive affect during winning, and to tense and angry affect
during losing. Although most children did not show increases in cortisol to competition, the
15% who did were higher in temperamental surgency and lower in effortful control. The authors
concluded that more extraverted, surgent children are most vulnerable to stress during
competition. This is an important finding, because it indicates links between stress hormones
and positive affect as well as negative affect systems.
Gunnar (1994) also found an initially surprising relation between cortisol levels and
preschool children's adjustment to a group setting. Rather than finding higher levels of cortisol
for 3- to 5-year-old inhibited (and presumably more stress-prone) children early in the school
year, Gunnar found that measures of cortisol reactivity to the school experience were related to
mother-report CBQ (Rothbart et al., 2001) surgency measures of high activity, stimulation
seeking, and impulsivity, with a trend toward less shyness. Teachers also reported fewer
internalizing problems, greater popularity, and independence for children with higher cortisol
levels. Later in the school year, however, higher cortisol reactivity was associated with teacher
reports of greater internalizing behavior and CBQ reports of sensitivity to discomfort.
Gunnar (1994) suggests that temperamentally linked coping activities of children may
mediate their cortisol reactions, so that more shy children will be less likely to experience
stressful interactions initially because of avoidant or inhibitory coping strategies. More outgoing
children will be more likely to seek out stress and show its effects in early, but not later group
 Temperament 33

experience, when they may have mastered the social challenge (Gunnar, 1994). Gunnar's work
indicates the importance of studying reactive measures in the context of regulatory coping. In
more recent research, peer reactions to a child’s behavior have also been found to be related to
cortisol function. Gunnar, Sebanc, Tout, Donzella, & van Dulmen (2003) reported that children
with higher cortisol levels in the nursery school classroom were those whose temperamentally
based behavior (high surgency; low effortful control) led other children to reject them. These
findings reflect recent increases in complexity of temperament-environment relationships that we
will discuss later in the chapter.
Summary. Behavior genetics research supports the idea that the chemical templates we
inherit are reflected in our temperament, social, and personality characteristics, and recent
research suggests that gene-environment interaction may be particularly important in relation to
personality and social development. More developmental research is needed, however, to
specify how developing brain mechanisms interact with environmental events to support these
outcomes. In research linking temperament to psychophysiology, recent investigations have
related tendencies toward approach and withdrawal to left and right hemisphere brain activity,
respectively. In addition, electrodermal responding has been linked to reserve and negative
emotionality. Heart rate (HR) variability and vagal tone have also been studied, with the latter
taken as a measure of parasympathetic function. HR variability has been linked to prosocial
responding and inversely, in some studies, to behavioral inhibition. Vagal tone has been linked
to behavioral irritability, approach, positive affect, and attention, with the direction of the linkage
varying depending on the age of the child. Both vagal tone and cortisol research stress the
importance of both reactive and self-regulative variables in the understanding of
psychophysiology and development, and all of these approaches are promising for tracing links
between genetic inheritance, experience, and behavioral outcomes.

TEMPERAMENT AND DEVELOPMENT

Stability and Change

Early theorists of temperament stressed the importance of finding stability of
temperament over time. Thus, for Buss and Plomin (1975), to qualify as a "temperament," a
characteristic must demonstrate stability from its early appearance to late in life. More recent
approaches to the field, however, have noted that temperament itself develops, and the study of
this development allows us a greater understanding of both normative and individual differences
(Goldsmith et al., 1997; Rothbart & Derryberry, 1981; Rothbart, 1989b). Temperamental
measures can fail to show normative stability, for example, but genetically related individuals
may show strong similarities in their patterns of change. These results have been found in
behavioral genetics work on both activity level (Eaton, 1994) and behavioral inhibition
(Matheny, 1989).
Even for dimensions showing normative stability, expressions of temperament are likely
to change over time. In measuring negative emotionality, for example, six-year-olds spend much
less time crying than do six-month-olds, but worry a good deal more. To appropriately assess
stability of temperamental characteristics, it is necessary to establish continuity in the
temperament constructs studied across time. Pedlow, Sanson, Prior, and Oberklaid (1993)
assessed the Australian Temperament Project (ATP) sample at intervals from infancy to 7-8
years of age. By using structural equation modeling, they identified factors that applied across
the entire age range (Approach/Sociability, Rhythmicity), or across several of the time intervals
 Temperament 34

studied (Irritability, Persistence, Cooperation-Manageability, and Inflexibility). A model

correcting for error of measurement was then used to assess individual stability on these factors
from year to year, and estimates were considerably higher than those previously reported, mostly
in the range of 0.7-0.8. Even with these levels of stability, however, there is considerable room
for individual change in their children’s relative position on these characteristics.
Since our last review, an important meta-analysis of studies on the stability of personality
traits, including temperament, has been carried out by Roberts and DelVecchio (2000). This
review was organized according to the Five-Factor model, and began with studies of
temperament and personality in infancy. Considerable stability was found in measures of these
variables after about the age of three years, with estimated cross-time correlations for 0-2.9 years
= .35; 3-5.9 years = .52; 6-11.9 years = .45; and 12-17.9 years = .47. The increase at 3 to 6 years
is of interest, given evidence for the development of executive attention and effortful control
during the first three years of life, possibly related to early instability. As attention systems
stabilize, controls over earlier more reactive tendencies may increase prospects for stability of
temperament and personality. Beyond childhood, levels of stability continue to increase through
adolescence and young adulthood, not peaking until after the age of 50 (Roberts & DelVecchio,
2000).
In the next section, we consider issues of temperament stability and change in relation to
social-emotional development and the development of personality. We review research
examining the development of temperament in the areas of positive affect/approach and
inhibition, distress proneness, activity level, and effortful control. Individual differences in
emotional and motor reactivity can be seen early in life, and they will be influenced over time by
the development of more regulatory systems, one of their emotionally based (fear or behavioral
inhibition), the other more directly self-regulative (effortful control). The first system develops
earlier than the second, and both are developing during the period when Roberts and DelVecchio
(2000) reported the lowest levels of normative stability.

Contributions of Temperament to Development

Temperament constructs are fundamental to thinking about trajectories of social-
emotional and personality development (Rothbart, Ahadi et al., 1994). As noted above,
temperament is implicated in social learning, with some children more responsive to reward,
others to punishment. Some children will be highly responsive to both. Temperament is also
closely linked to the development of coping strategies. If one child tends to experience high
distress to strangers, for example, and another child little distress, coping strategies involving
avoidance of strangers may be elicited and reinforced for the first child, but not for the second.
If the second child also experiences delight in the interaction with a stranger, more rapid and
confident approach to interactions with strangers is likely in the future. Thus, the practice and
reinforcement of children’s temperamentally based responses may serve to magnify initial
differences through a positive feedback process. Individual differences in temperament also
promote the child’s active seeking or avoiding of environments. Scarr and McCartney (1983)
describe these genotype/environment interactions as "niche picking." The child who stays at the
edge of a nursery school class or a party is selecting a different experience than the child who
goes directly to the center of social excitement.
In Gray's (1991) theory, extraverts, high in positive affect and approach (the BAS), are
seen as more susceptible to reward, and introverts, high in fear and shyness, to punishment (the
BIS). This model suggests that caregiver treatment may have differing developmental outcomes,
 Temperament 35

depending on differences among children. In other models, optimal level theories (e.g., Bell,
1974; Strelau, 1983) stress individual preferences for high or low levels of stimulation. A child
easily overwhelmed by stimulation will try to keep things quiet, whereas a child who requires
high levels of stimulation for pleasure will attempt to keep things exciting. Mismatches in
optimal levels between a parent and child, or among siblings, may require major adaptations
from one or more of the children or parents. Situational challenges, such as an intense daycare
experience for an easily over-stimulated infant, or demands for extended quiet time for a
stimulus-seeking older child, may lead to problems for both child and caregiver.
Scarr and McCartney (1983) also describe evocative interactions, where the child's
temperamental characteristics elicit reactions from others that may influence the child's
development. Thus, a positive and outgoing disposition may serve as a protective factor elicting
the support of others in a high-risk environment (Werner, 1985). Radke-Yarrow and Sherman
(1990) noted that in a high risk situation, a buffering effect can occur when the child's
characteristics meet the needs of the parent (these needs may be quite idiosyncratic). Acceptance
by adults can then lead children to feel there is something special or important about them
personally. This notion is very similar to Thomas and Chess's (1977) "goodness-of-fit"
argument.
Because temperament itself develops (Rothbart, 1989b), new systems of behavioral
organization (e.g., smiling and laughter, frustration, executive attention) will also come "on line"
over time. Any new systems that serve to regulate action and emotion will also come to
modulate characteristics that were previously present, yielding potential instability of
temperament across the developmental transition. In addition to the direct effects of developing
control systems of fear and effortful control, children who develop a given control system early
in life may have quite different experiences than children who develop the system later (Rothbart
& Derryberry, 1981). For example, the child who develops fear-related inhibition late is likely to
experience a greater number of interactions with potentially threatening objects or situations than
the child who develops fearful inhibition early. The child who is fearful and inhibited to
potential dangers early in development may spend more time watching and making sense of
events in the environment than the less inhibited child. We now consider some of the major
dimensions of temperament within a developmental context.

Extraversion/Surgency vs. Shyness and Behavioral Inhibition

By 2-3 months, infants show a pattern of smiling, vocalization, and motor cycling of the
limbs described by Kistiakovskaia (1965) and termed the “animation complex, including
smiling, quick and animated generalized movements with repeated straightening and bending of
hands and feet, rapid breathing, vocal reactions, eyeblink, etc.” (Kistiakovskaia, 1965, p. 39).
These reactions appeared to increase in duration and decrease in latency into the second and third
months of life (Kistiakovskaia, 1965). Werner (1985) reviewed cross-cultural evidence for both
an increase in smiling between 2 and 4 months, and an increase in vocalization at 3-4 months.
This cluster of intercorrelated behaviors (smiling and laughter, vocal and motor activity) is also
found in parents' reports of temperament and in home observations (Rothbart, 1986). Although
the pattern has been called "sociability," it is displayed toward exciting and novel objects as well
as toward people (Bradley, 1985). "Sociability" therefore may be too narrow a label for this
dimension, at least during early infancy.
Beyond 3-4 months, positive affect shows normative increases in probability and duration
across the first year of life, both in home observation and parent-report data (Rothbart, 1981,
 Temperament 36

1986). Stability has also been found for individual differences on a composite positive
emotionality measure including smiling and laughter, motor and vocal activity, as assessed by
parent-report and home observation between 3-9 months, and stability of a laboratory measure of
smiling and laughter between 3 and 13.5 months of age (Rothbart, 1986). Smiling and laughter
in infancy as observed in the laboratory also predicted both concurrent (Rothbart, 1988) and 6- to
7-year-old approach tendencies (Rothbart et al., 2001). Pedlow et al. (1993) also found stability
from infancy to 7-8 years on their dimension of Approach/Sociability.
Later in the first year, an important form of inhibition and control over approach
develops: some infants who were highly approaching at five or six months now come to inhibit
their approach responses when the stimuli are unfamiliar and/or intense (Rothbart, 1988;
Schaffer, 1974). In our laboratory, we found increases in infants' latency to grasp novel and
intense toys from 6.5 to 10 months of age (Rothbart, 1988). Infants' approach latency to low
intensity stimuli showed stability from 6.5 months to later ages (10 and 13.5 months), but to high
intensity stimuli, it did not. This finding is congruent with the idea that behavioral inhibition is
developing late in the first year, with the inhibitory reactions particularly evident in response to
high intensity stimuli. Once inhibition of approach is established, longitudinal research suggests
that individual differences in approach versus inhibition to novelty or challenge will be a
relatively enduring aspect of temperament. In familiar or low intensity situations, however,
chiefly positive activation will be evident. The inhibiting aspect of fear qualifies it as a control
system that modulates other response tendencies, and we elaborate this argument in discussing
Kochanska's (1993) research later in the chapter.
By early childhood, social inhibition with strangers shows moderate stability (Asendorpf,
1993; Gest, 1997). Honzik (1965) also noted that longitudinal Fels subjects' scores on
"spontaneity" versus "social interaction anxiety" were stable and predictive over long periods for
both males (the first three years to adulthood) and females (6-10 years to adulthood; Kagan &
Moss, 1962). Bayley and Schaeffer (1963) found their most stable and persistent category
between infancy to 18 years to be "active, extraverted" versus "inactive, introverted" behavior.
Tuddenham (1959) reported stability on scales indexing "spontaneity" versus "inhibition" for
subjects from 14 to 33 years in the Oakland Growth Study. Finally, Honzik (1965) found that
for the period between 21 months and 18 years of age, the two most stable dimensions were
"introversion" versus "extraversion" and "excessive reserve" versus "spontaneity."
These results can be added to evidence from Kagan (1998; Kagan & Fox, this volume) on
stability of behavioral inhibition, and to Caspi and Silva's (1995) and Pfeifer, Goldsmith,
Davidson, and Rickman’s (2002) recent work on stability of outgoingness and inhibition. In
Pfeifer et al.’s (2002) research, children were examined at 4 and 7 years with laboratory, TBAQ,
and CBQ assessments of behavioral inhibition and uninhibited behavior. At the younger age,
children were classified as extremely inhibited, extremely uninhibited, or intermediate. Close to
half the children remained in their original subgroup over the 3-year period. More than half
changed subgroup, but the change tended to be to the intermediate group rather than to the other
extreme. Caspi and Silva (1995) identified a group of children high on approach or confidence
at age 3-4, who were outgoing and eager to undertake tasks, and adjusted easily to challenging
situations. At age 18, these children were relatively low on self-reported Control (i.e., more
impulsive) and high on Social Potency (leadership and low shyness). Children identified in the
preschool period as inhibited (fearful, with problems in sustaining attention) were, at age 18,
high on Harm Avoidance, notably low on Aggression, and low on Social Potency. Caspi and
Silva’s (1995) finding that inhibition or fearfulness served as a protective factor against the later
 Temperament 37

development of aggression is also congruent with the positive correlations found between
temperamental fearfulness and the development of conscience described below.
In summary, evidence for approach tendencies related to positive affect can be seen early
in development. Later in infancy, behavioral inhibition related to fear develops. Once
established, tendencies toward approach versus inhibition demonstrate significant stability over
relatively long developmental periods, with important implications for social development.
Activity Level. Another major temperamental characteristic that can be measured early in
development is activity level. Using both ultra-sound imaging and mothers’ reports, activity
level can also be measured prenatally, and evidence for temperamental stability has been found
over the short periods that have been measured (Eaton & Saudino, 1992). In early research,
Fries (Fries & Woolf, 1954) and Escalona (1968) identified activity level as a major dimension
of individual differences among infants. Birns, Barten, and Bridger (1969) found no stability of
activity level from the newborn period to ages 3 and 4 months, but some stability was found
from 4 weeks to later assessments.
A possible explanation for instability of early activity level is the tendency for activity to
be linked to both negative and positive emotional reactivity. When high levels of activity occur
in the newborn, they are often linked to the expression of negative affect (e.g., Korner,
Hutchinson, Koperski, Kraemer, & Schneider, 1981). Escalona (1968) observed that newborns
engage in their highest motor activity during distress; positive states were associated with
quiescence. Later in development, however, the infant often becomes motorically aroused while
in an alert and non distressed state, as noted by Kistiakovskaia (1965), and activity frequently
occurs when the infant is orienting toward novel objects or receiving caregiver stimulation
(Wolff, 1965). Links between activity and newborn expression of negative affect may account
for its failure to predict later activity. Indeed, when Korner et al. (1985) measured non-distress
motor activity in the newborn, vigor of activity predicted high daytime activity and high
approach scores on the BSQ at ages 4-8 years. Another finding is that activity at 4 months,
coupled with negative affect, predicted later behavioral inhibition (Fox, Henderson, Rubin,
Calkins, & Schmidt, 2001; Kagan, 1998); activity coupled with positive affect predicted later
uninhibited behavior (Calkins & Fox, 1994; Fox et al., 2001).
Saudino and Eaton (1995), using actometer measures in a twin study, did not find
normative stability in activity level from 7 to 36 months. Nevertheless, in Saudino and Eaton's
(1995) study, MZ twins were more similar than DZ twins at both ages, and MZ twins were also
more concordant in their changes in activity from 7 to 36 months than were DZ twins. Lower
levels of stability of activity level from the first year to later periods have also been reported.
Roberts and DelVecchio (2000) found the lowest mean levels of stability for activity level
among other temperament/personality dimensions studied (r = .28); other dimensions ranged
from .35 to .47 in consistency over time.
At least two explanations are possible for early instability. One, presented above, is that
activity can be related to both positive and negative affect, so that the two kinds of activity
should be differentiated. Second, the onset of inhibition as an aspect of fearfulness late in the
first year may lead to lower activity for a number of children under conditions of novelty or high
intensity. A second form of control over impulsive activity will also be developing beginning
late in the first year and during the preschool years. This is the effortful control system, related
to the development of executive attention, to be discussed later in this section. Its development
coincides with normative decreases in activity level, which in a meta-analysis of activity level
studies showed a peak between 7 and 9 years (Eaton, McKeen, & Campbell, 2001).
 Temperament 38

Attentional Orienting and Effortful Control

Attention has both reactive and self-regulative aspects. In reactive attention, that is,
orienting to exogenous stimulation, consistency of rates of infant looking have been found across
three quite different measures in 3-month-olds: a visual discrimination paradigm, an auditory
discrimination paradigm, and rate of looking toward the mother in social interaction (Coldren,
Colombo, O'Brien, Martinez, & Horowitz, 1987). Byrne, Clark-Touesnard, Hondas, and Smith
(1985) also reported stability from 4 to 7 months in average looking time and duration of first
look in visual habituation tasks.
A developmental shift in visual orienting appears to occur late in the first year of life.
Kagan, Kearsley, and Zelazo (1978) noted a U-shaped developmental pattern of fixation times to
clay faces with scrambled and unscrambled features in both North American and Guatemalan
children. From 4 to 8 months, there is a steep decline in the amount of time children spend
looking at both kinds of faces. Between 13 and 36 months, however, there is an increase in
looking time that is stronger for scrambled than for unscrambled faces. Kagan et al. (1978)
argue that “stability of duration of orienting from 8 to 13 and 13 to 27 months, without
comparable 8 to 27 month continuity, suggests that determinants of fixation time change between
8 and 27 months” (Kagan et al., 1978, p. 81). No stability was found between 4 months and later
measures. These changes are in keeping with findings that signs of executive attention begin to
emerge toward the end of the first year, allowing increased attentional control and planning, and
presumably changing the meaning of individual differences in looking at objects (see also
discussion in Ruff & Rothbart, 1996).
The development of effortful control, that is, the efficiency of executive attention,
including the ability to inhibit a dominant response and/or to activate a subdominant response, to
plan, and to detect errors, also appears to be linked to the child's developing ability to maintain a
focus of attention over an extended period. Sustained attention and the ability to delay are
positively related, and both develop over the preschool years. Krakow, Kopp, and Vaughn
(1981) studied sustained attention to a set of toys in 12- to 30-month-old infants. Duration
increased across this period, with stability of individual differences between 12 and 18 months,
and between 24 and 30 months. Sustained attention was also positively related to self-control
measures, independent of developmental quotient, at 24 months. Children who showed high
sustained attention at 12 months were described by their mothers as more quiet and inactive at 24
and 36 months.
As in the case of behavioral inhibition, once normative changes in executive attention
have occurred, additional controls over more reactive behavior will have been added. Children
may now demonstrate increased flexibility in their deployment of attention from one situation to
another. With low effortful control in an older child, however, the child's activity may continue
be driven by the intensity, novelty or discrepancy of the stimulus, or by its associations with
previous reward and punishment. Orienting may then be of long or short duration, but it will not
show the flexibility of response that is possible when executive attentional control has been
added.
Krakow and Johnson (1981), using measures of self-control under verbal instructions
with children age 18-30 months, found large age effects in inhibitory control. They also found
moderate levels of stability of inhibitory self-control across the 12-month period. Vaughn,
Kopp, and Krakow (1984) reported on two aspects of self-control: delay and compliance. Cross-
task consistency and coherence across the two broader measures increased across age, and the
 Temperament 39

authors concluded “that individual differences in self-control emerge and are consolidated during
the 2nd and 3rd years of life” (p. 990). Reed, Pien, and Rothbart (1984) found strong age effects
in two measures of self-control (a pinball game and Simon-says game) in a cross-sectional study
of children aged 40-49 months. These studies together indicate increases in self-regulation
across 18-49 months of age. In our research using the Stroop-like Spatial Conflict tasks
described above, children began to demonstrate effective management of conflict at 30 months,
and 36-month-old children who showed greater interference in reaction time for conflicting
responses were reported by their mothers as exhibiting lower levels of inhibitory control
(Rothbart et al., 2004). Less accurate children were also reported as showing higher levels of
anger/frustration in the IBQ (Gerardi-Caulton, 2000), suggesting attentional control over emotion
as well as action. Additional research with conflict tasks indicates development of conflict
performance between the ages of 2 and 7 (see review by Rothbart & Rueda, in press).
Murphy, Eisenberg, Fabes, Shepard, and Guthrie (1999) studied children longitudinally at
4-6, 6-8, 8-10, and 10-12 years of age, using parent and teacher report. Children were reported
as increasing control over attentional shifting and inhibitory control across age, and girls also
decreased in impulsivity. With the exception of attention shifting, the measures of self
regulation showed considerable stability across this period of development.
Attention and Distress. A consistent theme of the relation between attention and distress
is that the two mutually influence each other. In a study of attentional orienting and soothing in
three to six month old infants (Harman et al., 1997), infants were first shown a sound and light
display; about 50% of the infants became distressed to the stimulation. They then strongly
oriented to interesting visual and auditory soothing events when these were presented. While the
infants oriented, facial and vocal signs of distress disappeared. However, as soon as the
orienting stopped, for example, when the object was removed, the infants’ distress returned to
almost exactly the levels shown prior to its presentation. Apparently, the loss of overt signs of
distress is not always accompanied by a genuine loss of distress. Instead, some internal system,
which we termed the distress keeper, appears to hold a computation of the initial level of
distress. Repeating the soothing stimulus also appeared to reduce its soothing effectiveness
(habituation) at six months, but not at three to four months.
Regulatory behaviors were studied in a longitudinal study of 66 children seen at 3, 6, 10,
and 13 months of age (Rothbart, Ziaie, & O’Boyle, 1992). The infants were presented with
stimuli that varied in intensity and predictability, and children showed considerable active coping
with their own distress and excitement. At 6 months, children’s disengagement of attention
could be reliably coded. Overall disengagement of attention was not stable from 6 to 10 months,
but from 10 to 13 months, children demonstrated stability in their tendency to disengage from
distress-producing visual stimuli such as masks and mechanical toys. Infant disengagement was
also related to lower levels of negative affect at 13 months. Stability from 10 to 13 months was
also found in infants’ use of mouthing, hand to mouth (e.g., thumb sucking), approach, and
withdrawing the hand, suggesting that some of the infants’ self-regulation strategies were
becoming habitual by this time.
Direct links have also been found between children’s disengagement of attention and
decreases in negative affect (Stifter & Braungart, 1995). Correlations also have been found
between infants’ use of self-regulation in anger inducing situations and their early childhood
ability to delay responses (Calkins & Williford, 2003), suggesting that mechanisms used to cope
with negative emotion may later be transferred to control of cognition and behavior, as suggested
by Posner and Rothbart (1998). Further support of this idea was found by Mischel and his
 Temperament 40

colleagues (Sethi, Mischel, Aber, Shoda, & Rodriguez, 2000). Toddlers were briefly separated
from their mothers and children’s coping strategies coded. Later, at age five, their behavior was
observed in a situation where they could delay gratification for a more valued reward. Children
who used more distraction strategies during the maternal separation at the younger age were later
able to delay longer.
Long-term stability in the ability to delay gratification and later attentional and emotional
control has been reported (Mischel, 1983). In Mischel's work, the number of seconds delayed by
preschool children while waiting for rewards that were physically present (a conflict situation)
significantly predicted parent-reported attentiveness and ability to concentrate when the children
were adolescents (Shoda, Mischel, & Peake, 1990). Children less able to delay in preschool
were also reported as more likely to go to pieces under stress as teenagers, and to show lower
academic competence in SAT scores, even when controlling for intelligence (Shoda et al., 1990).
In follow up studies, preschool delay predicted goal setting and self-regulatory abilities when the
participants reached their early thirties (Ayduk et al., 2000), suggesting remarkable continuity in
self-regulation.
In Caspi and Silva's (1995) study, preschool children characterized as "well adjusted"
were described as flexible in orientation, and “capable of reserve and control when it was
demanded of them" (p. 492). These children’s flexibility of responsiveness may have been
linked to greater executive attention and effortful control, as well as to higher ego resiliency
(Block & Block, 1980) as described below. At age 18, children earlier identified as "well
adjusted" by Caspi and Silva had high scores on Social Potency, including leadership and low
social shyness. Interesting positive links have been found between activity level as assessed
through actometers and children’s performance on motor conflict tasks (Campbell, Eaton, &
McKeen, 2002). Active four to six year old children, especially the younger children, showed
better performance on tasks that required them to inhibit a habitual response in order to perform
a non-habitual response. Their “paper highlights . . . the potentially functional, yet much
neglected role that physical movement may play in young children’s development” (p. 295).
Attention thus shows major developments over the first years of life, with a more self-
regulative system added to a more reactive one (Rothbart, Posner, & Rosicky, 1994). As noted
above, Caspi and Silva's (1995) Factor 1 (Lack of Control), including a combination of
irritability and lack of self-regulation at age 3-4 years, was strongly related to negative
emotionality at 18. Studies are now underway exploring contributions of both temperament and
parent treatment to the development of self-control, as in Silverman and Ragusa's (1992) study
predicting 4-year-old self-control from 24-month child temperament and maternal variables.
Olson, Bates, and Bayles (1990) have also found relationships between parent-child interaction
at 13 months and 2 years (but not at 6 months) and children's self-control at age 6.
Two Control Systems. Early individual differences in motor and emotional reactivity thus
appear to be influenced by development of at least two temperament-related control systems, one
of them part of an emotional reaction (fear and behavioral inhibition); the other more completely
self-regulatory (attentional control), with the first system developing earlier than the second.
This view is related to the theory of ego-control and ego-resiliency developed by Jean and Jack
Block (Block & Block, 1980). The Blocks posited two control systems: one, ego control,
involved fearful or inhibitory control over impulsive approach. The second, ego-resiliency, was
defined in terms of flexible adaptation to changing circumstances. The latter system is related to
the temperamental characteristic of attentional effortful control, and research by Eisenberg et al.
(1996) supports the predicted relationship between ego resiliency and CBQ attentional control in
 Temperament 41

kindergarten to third grade children. Resiliency was also related to social status and to teacher-
reported socially appropriate behavior. Effects of self-regulation were also significantly stronger
for children who were high in negative emotionality (Eisenberg et al., 1996).
Eisenberg et al. (2004) have studied parent- and teacher-reported effortful control and
impulsivity in relation to ego resiliency in children 4.5 to 8 years, with a 2-year follow-up. At
both ages, effortful control and impulsivity predicted unique direct variance in resiliency and
externalizing, and they also predicted internalizing problems indirectly, through resiliency. A
moderating effect was also found on teacher-reported anger and the relation between effortful
control, impulsivity, and externalizing. All relations held in predictions from Time 1 to Time 2,
except the path from impulsivity to externalizing.
In the Blocks' theory, resiliency or flexibility contributes to the development of adaptation
and mental health. As Block and Kremen (1996) put it:
Adaptability in the long-term requires more than the replacement of unbridled impulsivity
or under-control, with categorical, pervasive, rigid impulse control. This would be over-
control of impulse, restriction of the spontaneity that provides the basis for creativity and
interpersonal connection. Instead and ideally, dynamic and resourceful regulation and
equilibrium of impulses and inhibitions must be achieved. It is this modulation of ego-
control that we more formally mean by the construct of ego-resiliency. It can be said that
the human goal is to be as under-controlled as possible and as over-controlled as
necessary. (p. 351)
In the ego control construct, when fear and its correlates develop within a relatively
constricted life, approach tendencies are strongly opposed, and rigid functioning may result.
Ego-resiliency, on the other hand, is strengthened by a set of life experiences that build upon
capacities for both expression and control of impulses. Effortful control appears to provide an
important underlying system for the development of ego resiliency, with impulsivity also related.
The Blocks' theory stresses the importance of experience in the development of adaptation, with
endogenous control systems allowing cultural influence on the behaviors, thoughts, and emotions
that are controlled, as well as on the particular self-regulatory capacities and strategies used by
the child.
Summary. Because temperament systems themselves develop, in this section we have
presented a brief account of the early development of aspects of positive affect and approach,
activity level, and distress in relation to attentional control. (For further information on the
development of temperamental distress, see Rothbart & Bates, 1998) Some of these
developmental changes lead us to expect temperamental stability within only limited time
windows. Early reactive systems of emotionality and approach become overlain by the
development of at least two temperamentally linked control systems. The first, fearful inhibition,
is linked to developments in fearfulness late in the first year of life. The second, effortful
attentional control, develops across the preschool period and shows considerable stability.
Another likely control mechanism for the support of socialization is the development of a social
reward system, connected with children's desires to please and to refrain from hurting their
parents and other persons, likely linked to temperamental affiliativeness. Any failure of these
controls may be linked to the development of behavior problems. Because these temperamental
systems are open to experience, appropriate socialization will be necessary for positive
outcomes.
 Temperament 42

TEMPERAMENT AND THE DEVELOPMENT OF PERSONALITY

Life experiences influence connections between children’s emotional dispositions, their

conceptual understanding of events, and their use of coping strategies to deal with these events.
These “mental habits” are influenced by the child’s temperament, expectancies, beliefs, values,
goals, self-evaluations, appraisals, as well as understandings of the situation, the self, and/or
others (Mischel & Ayduk, 2004; Teglasi & Epstein, 1998). In Mischel and Ayduk’s (2004)
model, individuals differ in the “ease of accessibility” of cognitive-effective units and in the
organization among them.
Coping strategies, which may have been originally based on temperamental
predispositions, become part of these units, and may be consolidated or inhibited depending in
part on their consequences. Mischel and Ayduk (2004) give the example of individual
differences in rejection sensitivity (RS):
RS is a chronic processing disposition characterized by anxious expectations of rejection
[Downey & Feldman, 1996] and a readiness to encode even ambiguous events in
interpersonal situations [e.g., partner momentarily seems inattentive] as indicators of
rejection that rapidly trigger automatic hot reactions [e.g., hostility-anger, withdrawal-
depression, self-silencing]. Probably rooted in prior rejection experiences, these
dynamics are readily activated when high RS people encounter interpersonal situations in
which rejection is a possibility, triggering in them a sense of threat and foreboding.
(Mischel & Ayduk, 2004, p. 118)
As rejection sensitivity becomes habitual, the person’s attention may become quite
narrowly focused on the likelihood of rejection, and defensive behaviors (e.g., anger or
preventative rejection of the other) may develop to fend off the expected rejection. Different
levels of generality of such a disposition are also possible. Rejection sensitivity, for example,
might extend to a wide range of human relationships, but the sensitivity may also be more
specific, so that only rejection by the child’s peers, but not by adults, has been sensitized. RS
may be so specific that it is limited to a single person in a single kind of situation. Mental habits
are particularly likely to develop in connection with intimate relationships, as in the family, but
they may be carried over to new relationships when more positive expectations and coping
methods are lacking. Thus, the experience of early criticism and rejection, which may have its
strongest impact on children prone to distress, may have long-term consequences for problems in
development.
Mischel and Ayduk’s (2004) analysis of rejection sensitivity describes an anxious or
defensive set, but alternatively, children’s experiences with others may be generally of
acceptance. If so, the child will be less likely to be on guard about rejection or to show a
defensive perceptual set. Instead, the child’s attention can be directed more broadly, allowing
greater conscious awareness of the state and needs of others. More distress prone, fearful, and
irritable children may be more likely to develop such habits as rejection sensitivity, but after
experiencing high levels of rejection, even a low distress-prone child would be likely to develop
sensitivity to rejection. More surgent and approaching children may also be more likely to
expect acceptance, but even the more distress prone child may lack the conditions for becoming
sensitive to rejection when others are not critical and rejecting. This model stresses
temperament-environment interactions for a number of social-emotional processes that are likely
to be differentiated by context, and we review some of these important interactions in the next
 Temperament 43

section.
When repeatedly exercised, habitual activations of clusters of thoughts, emotions, and
action tendencies to a particular stimulus or situation become very likely to occur and difficult to
change. When mental habits involve distress, how might they be weakened or disconnections
achieved within the habit? In Eastern traditions, this is done partly through diminishing the role
of the ego so that situations can become less threatening to the self. Mental discipline and
meditation also allow weakening of links between thoughts and emotions or thoughts and action
tendencies. Western therapy similarly works through the clients’ patterns of reaction, attempting
to reconstruct previously consolidated patterns and provide new frameworks for meaning.
Taking a developmental view, however, one would wish to give the child the kinds of
experiences that will form favorable and non-injurious mental habits in the first place.
Socialization in the U.S. and other western cultures often strongly emphasize habits
related to the individual or ego, promoting the pursuit of individual security, satisfaction of
individual desires, and achievement of a positive self-concept. In other cultures, the shaping of
the child’s mental habits can be quite different. Mascolo, Fischer, and Li (2003) suggest, for
example, that the biological mechanisms on which pride and shame are based are similar across
cultures, while the responses can be shaped in quite different directions:
For example, in American dyads, pride experiences develop as socialization agents praise
children’s accomplishments; shame experiences develop in social contexts in which
children are made aware of their flawed identities. In contrast, in China, modest self-
harmonization develops as parents efface their children’s accomplishments while
relatives and other significant others praise them; shame is a normative emotion that
develops as parents use explicit shaming techniques to socialize filial piety in children.
(Mascolo et al., 2003, pp. 401-402)
In this view, the biological equipment or temperament is similar across cultures, but the
mental habits and representations of self, the world and other, will vary from culture to culture,
and, we would add, context to context. By the time a child is a well-socialized member of the
society, more biologically based temperament will have been shaped into a set of values, goals,
and representations of the self and others that specify what is good and bad for the person. Even
for children who are not well-socialized, values stressed by the culture may nevertheless have an
effect. Children in the United States, for example, may still attempt to promote a positive self-
concept, and pursue it though a delinquent peer group, even when the goals and values followed
to achieve the positive concept may not be socially acceptable ones.
Shiner and her colleagues have recently been studying the continuity of personality from
the period 8-12 to 20 years (Shiner, Masten, & Tellegen, 2002) and 30 years (Shiner, Masten, &
Roberts, 2003). The 8-12 year variables, taken from parent and child interviews and teacher
questionnaires, included measures of mastery motivation, academic conscientiousness, surgent
engagement, agreeableness, and self-assurance versus anxious insecurity. Adaptation in
childhood and adulthood was assessed in terms of academic achievement, rule abiding conduct
versus antisocial behavior, and social competence. Adult measures employed self- and parent-
report questionnaires, including the Multidimensional Personality Questionnaire (MPQ;
Tellegen, 1985), as well as data on academic achievement, rule compliance, social competence,
job competence, and romantic competence.
Tellegen’s self-report MPQ contains three broad personality factors. Positive
Emotionality (PEM) includes scales for well-being, achievement, social potency, and social
closeness. Negative Emotionality (NEM) includes scales for stress reaction, alienation, and
 Temperament 44

aggression. Constraint (CON) includes scales for control, harm avoidance, and traditionalism.
PEM was moderately predicted at age 20 by mastery motivation, surgent engagement, and self-
assurance in middle childhood (Shiner et al., 2002). PEM was also related to concurrent social
and romantic competencies at 20 years, but adult positive emotionality was not linked to any of
the childhood measures of adaptation. Negative Emotionality (NEM) at 20, on the other hand,
was related to low adaptation in all areas in childhood, and to all concurrent adaptation measures
except romantic competence. Even controlling for childhood personality, lower academic
achievement and greater conduct problems in childhood continued to predict adult NEM.
Childhood mastery motivation and surgent engagement were also inversely related to NEM in
adulthood. At age 20, Constraint (CON) was predicted by earlier lower self-assurance and
higher academic competence, but when childhood personality was controlled, it was not related
to childhood adaptation.
Shiner et al. (2002) suggest that positive emotionality may be more closely linked to
current adaptation, whereas negative emotionality shows more continuity with earlier adaptation.
In our section on adjustment, we will note strong links between negative emotionality and
psychopathology, both in childhood and adulthood. Negative emotionality is also particularly
linked to behavior problems when effortful control is low (Caspi, 2000; Eisenberg, Fabes,
Guthrie, & Reiser, 2000). Shiner et al. (2002) note that adults high on negative emotionality tend
to be particularly upset by daily problems (Suls, Martin, & David, 1998). How might the
development of mental habits contribute to these findings? First, it is likely that habits related to
distress may have involved attempts to decrease distress through mental processing, which may
have included repeatedly thinking about problematic events. Positive experiences would be less
of a challenge and tend to be less intense, so they are likely to have been less rehearsed. When
mental habits have been tied to difficult and painful situations in the past, one faces not only
current problems, but also representations in memory that bring forward the mental habits linked
to similar situations. Well-practiced associations may make negative affect, cognition, and
action links stronger. Thus early failure, e.g., poor achievement in school, may create the
possibility of long term negative affect or neuroticism that extends to achievement situations
later in life.
Caspi, Harrington et al. (2003) linked observations of 1000 children at age 3 to their self-
reported personality at age 26 (96% of the original sample). Undercontrolled children (10% of
the sample) had been temperamentally impulsive, restless, distractible, and negativistic at age 3;
Confident children (28%) were friendly, eager, and somewhat impulsive; Inhibited children (8%)
were fearful, reticent, and easily upset; Reserved children (15%) were timid but not extreme in
shyness; Well-adjusted children (40%) appeared to be capable of self control, adequately self-
confident, and did not become upset during testing. At age 26, previously Undercontrolled
children were higher in negative emotionality, more alienated, and subject to stress reactions.
They also tended to follow a traditional morality. Formerly Inhibited and Reserved children
were high in harm avoidance, low in social potency (less vigorous, dynamic, forceful), and low
in achievement. Both previously Undercontrolled and Confident children were low in harm
avoidance. Confident children were high on social potency as adults; Inhibited children were
high in Constraint and low in Positive Emotionality.
Caspi, Harrington et al.’s (2003) findings provide evidence that the temperament of the
child truly provides the core of aspects of the developing personality. Undercontrolled children,
who combined extraversion/surgency, negative affect, and low attentional control at age 3,
showed neurotic and alienated tendencies as adults. Confident extraverted children were
 Temperament 45

confident and unfearful as adults. More shy and fearful Inhibited and Reserved children
maintained their caution and harm avoidance into adulthood and were low in social potency,
whereas the more extreme Inhibited children were also high in Constraint (a mixture of
fearfulness and self-control) and low in Positive Emotionality and social support. The most
interesting aspect of the results, however, goes beyond temperament to touch upon alienation,
traditional values, and social support.
Kubzansky, Martin, and Buka (2004) related children’s personality/temperament at age 7
as derived from observer ratings to self ratings at age 35. Children’s behavioral inhibition did
not predict adult functioning, but their anger proneness (Distress) predicted adult
Hostility/Anger, and inappropriate interpersonal self-regulation in childhood predicted adult
Interpersonal Sensitivity. Strong relations were found between child Distress Proneness and
adult somatization, another very intriguing finding. Overall, we expect that these studies will
inspire more contributions to the longitudinal literature related to these variables in the future.

TEMPERAMENT AND ADJUSTMENT

In the preceding section, devoted to temperament and personality, we have begun to

consider temperament and some aspects of adjustment. In this section we consider in more detail
theoretical models and research findings relating temperament to individual differences in
adjustment. By adjustment we mean not only psychopathology, but also positive behaviors,
including the development of conscience. We are more interested in dimensions of adjustment
than with categorical diagnostic systems, and think of adjustment in terms of adaptation to
particular contexts. A child may carry temperament traits from one context to another, but their
implications for adjustment will depend on the specific context and expectations of the parent,
peer, or teacher (Chess & Thomas, 1984; Lerner & Lerner, 1994), in connection with
experiences and adaptations to specific situations as suggested in the mental habits model
described previously.

Does Temperament Predict Adjustment?

Meaningful patterns of relationship exist between constructs of temperament and
constructs of adjustment in the development of children. This was clear by the late 1980s (Bates,
1989b), and has became more firmly established since then, with many studies showing
temperament links with psychopathology (for example, see reviews by Eisenberg et al., 2000;
Lonigan, Vasey, Phillips, & Hazen, 2004; Rothbart & Bates, 1998; Rothbart & Posner, in press;
Sanson, Hemphill, & Smart, 2004; Wachs & Bates, 2001). In these studies, temperament and
adjustment have been measured in a variety of ways, including parent reports, teacher reports,
and direct observation, with adjustment assessed at home and at school, using both cross-
sectional and longitudinal designs. In this review, we focus primarily on patterns of relations
that indicate a differentiated view of how temperament might contribute to the child’s
adjustment. We mention two methodological and conceptual issues before describing findings.
The issue of measurement Acontamination@ (Sanson, Prior, and Kyrios, 1990) has been of
continuing interest to researchers. Sanson et al. argued that relations observed between a
temperament measure and an adjustment measure might be an artifact of content overlap
between the two supposedly distinct measures. Item content in a temperament scale, for
example, might concern behaviors that are the same as those in the measure of psychopathology
or vice versa. Bates (1990) argued that adjustment and temperament should actually have some
 Temperament 46

conceptual overlap. The child=s adjustment could reflect a component of temperament, and
psychopathology could be, at least in part, an extreme point on a temperament dimension. For
theoretical reasons, however, we tend to regard temperament and adjustment as separate
concepts.
Temperament characteristics may also be contributors to adjustment rather than
equivalent to adjustment. This possibility can be supported, if studies show that temperament-
adjustment links exist even after correcting for content overlap, or if studies show links between
temperament and adjustment that transcend the simple content overlap model. In fact, studies
where expert raters and psychometric principles are used to remove items with overlapping
content do demonstrate links between temperament and psychopathology even after
“decontamination” (Lemery, Essex, & Smider, 2002; Lengua, West, & Sandler, 1998;
Oldehinkel, Hartman, de Winter, Veenstra, & Ormel, 2004). The second way of supporting a
distinction between temperament and adjustment is shown by a variety of studies. One study, for
example, found that therapy led to changes in parents= descriptions of their children=s
psychopathology, but not their temperament (Sheeber, 1995). Other studies address the question
of the developmental processes through which temperament and adjustment are related, and
these will be discussed after considering the second methodological issue.
The second methodological issue is source bias. As we have argued, caregivers’ reports
do show validity. However, when a conceptual relation is inferred between two constructs
measured via the same source, e.g., parents, the possibility exists that relations are due to
preconceptions in the minds of the informant rather than in the behavior of the subject. Bates
and Bayles (1984) have nevertheless argued, on the basis of many different tests of subjective
and objective components in parents= perceptions of their children, that measures of subjective
bias do not account for more of the variance than measures of objective phenomena. In addition,
as will be detailed later, the different measures are related to one another within and across time
in a differentiated pattern, e.g., with early novelty distress predicting later novelty distress, and
predicting internalizing problems more than externalizing problems. Recent studies have also
shown credible levels of objectivity in caregivers= descriptions of children=s temperament, even
when subjective factors, such as depression, play some role (e.g., Bishop et al., 2003; Forman et
al., 2003). Thus, source biases are not as powerful as one might have feared, and caregivers
perceive children=s behavioral traits in relatively differentiated, rather than global or unitary
ways. This brings us to the central question:

How Does Temperament Predict Adjustment?

Temperament might be involved in the development of behavior problems in a number of
ways. Clark, Watson, and Mineka (1994) listed four ways in which mood and anxiety disorders
might be related to personality characteristics (also see Shiner & Caspi, 2003): (a) vulnerability
models, where there is a predisposition to the development of disorders (e.g., in response to
stressors); (b) the pathoplasty model, a variant of vulnerability in which personality shapes the
course of a disorder (e.g., by producing an environment that maintains the disorder); (c) the scar
hypothesis, in which a disorder produces enduring changes in personality (e.g., increased levels
of insecurity); and (d) the spectrum or continuity hypothesis, where the psychopathological
condition is an extreme manifestation of the underlying personality trait. Clark et al. (1994) point
out that the four models need not be mutually exclusive. These models may also extend to
behavioral disorders, and to positive outcomes as in: (a) protective models, where the person is
predisposed to deal adequately with challenging situations; (b) the boost from positive adaptation
 Temperament 47

model, in which the experience of overcoming challenge strengthens feelings of optimism and
well-being; or (c) the spectrum or continuity model, in which the positive outcome is itself the
manifestation of an underlying set of characteristics, such as a positive outlook on experience.
These and other possible processes linking temperament, risk conditions, and psychopathology
are listed in Table 4. Generally, available evidence does not allow for a choice among the
models, but in recent years, behavioral and molecular genetics research is offering the promise of
choices (e.g., Eaves, Silberg, & Erkanli, 2003).

_________________________
Table 4 about here
__________________________

Direct Linkage. Most studies of the relations between temperament and adjustment have
considered direct, linear effects, where a particular temperament trait contributes to the
development of an adjustment pattern. Additive effects of multiple temperament traits are also
possible, as when two or more temperament traits linearly increase the risk of some disorder,
such as negative affectivity and lack of impulse control predicting behavior problems (Eisenberg
et al., 1996), both negative emotionality and fearfulness predicting levels of young boys’
internalizing problems (Gilliom & Shaw, 2004), or both impulsivity and negative emotionality
associated with adolescents’ antisocial behavior (Stice & Gonzales, 1998).
In evaluating direct linkage models, studies considering multiple temperament traits in
relation to multiple dimensions of adjustment are critical. According to current theories of
psychopathology, individual differences in specific temperament-related brain circuits are linked
to specific forms of motivation or functioning (Bates et al., 1994; Clark et al., 1994; Fowles,
1994; Gray, 1991; MacDonald, 1988; Rothbart, Derryberry et al., 1994; Rothbart & Posner, in
press), as discussed above. There is some, but of course not complete agreement on the specific
systems and how they map onto behavioral traits. In previous sections we discussed systems
controlling inhibition to novelty and conditioned signals of punishment and non-reward, as well
as unconditioned fear, positive affectivity and reward seeking, sensitivity to social rewards, and
attentional control. We now use these systems as general constructs to organize the evidence on
temperament and adjustment. At this time, only a limited number of studies permit a
differentiated view of temperament-adjustment linkages, and none of the studies is
methodologically strong enough to stand alone in support or rejection of a psychobiological
systems model. However, enough convergence exists that we are confident about the broad
outlines of direct linkage models.
Theoretical Expectations. Direct linkage models will become more detailed as
neurobehavioral systems are better understood, and as measures of adjustment are meaningfully
differentiated. For now, one would expect early irritability, or general tendencies toward
negative affect, to predict a wide variety of adaptive difficulties, including both internalizing, or
anxiety problems, and externalizing, or conduct problems, as well as deficits in positive social
competencies. As measures of irritability are more finely differentiated, however, more clearly
defined pathways to later adjustment may be identified. For example, sensitivity to minor
aversive stimuli might predispose a child to both internalizing (e.g., whining and withdrawal)
and externalizing (e.g., reactive aggression) behavior problems, whereas irritability to frustration
of reward or of stimulation-seeking behavior (Rothbart, Derryberry et al., 1994) would likely
pertain more to externalizing tendencies than to internalizing ones.
 Temperament 48

Temperamental tendencies toward fearfulness in novel or potentially punishing situations

should predict internalizing-type adjustments most directly, although they may also serve to
predict externalizing problems in inverse or interactive ways, as discussed below. A finer
differentiation of fearfulness will ultimately be important for predicting different kinds of
internalizing adjustment. For example, separation distress may differ in some ways from novelty
fear (see Fowles= [1994] discussion of theories placing separation fear in a panic or fight/flight
brain system and novelty fear in a behavioral inhibition system, and see our sections on
Panksepp=s [1998] psychobiological theory).
Positive affectivity or surgency, involving activity, stimulation seeking, assertiveness,
and possibly some aspects of manageability, should be involved more closely in externalizing
than in internalizing problems, except that depression has a strong component of low positive
affectivity (Tellegen, 1985). On the other hand, a trait of prosocial tendency, affiliation and
agreeableness, perhaps involving sensitivity to social rewards (MacDonald, 1992), might prove
separable from the more general extraversion or surgency (positive affectivity) system, as
Rothbart and Victor’s (2004) findings suggest. Low levels of prosocial interest and concern
would be expected to be associated with the development of externalizing and not internalizing
problems, and perhaps with the failure to acquire positive social competencies independent of
behavioral problems.
Finally, systems controlling attention, especially the executive attention system described
above, would be expected to be related to both externalizing and internalizing, but to have more
to do with externalizing problems than with internalizing ones. As with fear systems, attentional
control should also play an additive or interactive role with other temperament characteristics. In
addition, a well-functioning set of attentional controls is likely to be linked to more positive
developmental outcomes.
Empirical Findings of Direct Linkage. A number of studies provide support for the
models just described. In general, predictive relations between temperament and adjustment are
of modest to moderate size. Correlations between infancy measures and adjustment in late
preschool and middle childhood tend to be smaller, and those between preschool or middle
childhood and later periods larger. Even though the correlations may be modest to moderate in
size, they have been well-replicated, and it is clear that they are not merely chance findings.
Moreover, the size of the relations is usually not less than and sometimes greater than predictions
from other theoretically linked variables, such as parenting quality. Lytton (1995), for example,
performed a meta-analysis of studies predicting conduct disorder (a diagnosis of extreme
externalizing problems) and criminality, finding child temperament variables to be the single
most powerful predictor of the outcomes, even in comparison with qualities of parenting.
In the Bloomington Longitudinal Study (BLS), infancy and toddlerhood ICQ
temperamental difficultness (frequent and intense negative affect and attention demanding)
predicted later externalizing and internalizing problems as seen in the mother-child relationship,
from the preschool to the middle-childhood periods (Bates & Bayles, 1988; Bates, Maslin, &
Frankel, 1985; Bates, Bayles, Bennett, Ridge, & Brown, 1991; Lee & Bates, 1985). Early
negative reactivity to novel situations (unadaptability) predicted less consistently, but when it
did, it predicted internalizing problems more than externalizing problems. Early resistance to
control (perhaps akin to the manageability dimension of Hagekull [1989], and perhaps at least
partly related to the construct of effortful control) predicted externalizing problems more than
internalizing problems. This was also found in predicting externalizing problems at school in
 Temperament 49

both the BLS and a separate longitudinal study, the Child Development Project (CDP; Bates,
Pettit, Dodge, & Ridge, 1998).
In a structural modeling analysis of CDP data, dealing with the overlap in externalizing
and internalizing symptoms, Keiley, Lofthouse, Bates, Dodge, and Pettit (2003) separated
mother and teacher reports of behavior problems across 5 to 14 years into pure externalizing,
pure internalizing, and covarying factors, and then considered early childhood predictors of each
of these factors. Resistant temperament (unmanageability) predicted the pure factors of mother-
and teacher-rated externalizing problems, but not the pure internalizing factors. Unadaptable
temperament (novelty distress) predicted positively both mother and teacher pure internalizing
factors, and to a lesser degree, and negatively, the pure mother and teacher externalizing factors.
That is, unadaptable temperament predicted higher levels of internalizing problems and, less
strongly, lower levels of externalizing problems. Although a disposition to fearfulness would not
necessarily constrain dispositions to aggressive and uncooperative behaviors, it is intuitively
reasonable that children who are fearful and sensitive to potential punishment would be likely to
inhibit externalizing behavior (Bates, Pettit, & Dodge, 1995). And finally, difficult temperament
(negative emotionality and demandingness) predicted, in this multivariate context, none of the
pure factors, but only the covarying externalizing plus internalizing factor in mothers’ reports.
These predictions are all consistent with models where temperament extremes either
constitute pathology dimensions or predispose to risk for these conditions. The linkages are of
modest size, but they obtain from early in life, and are not eliminated by the inclusion of family
and parenting characteristics in prediction, so they are not simply artifacts of family functioning.
Also supporting the general pattern, Gilliom and Shaw (2004) found, in a sample of preschool-
age boys from low-income families, that high levels of negative emotionality were associated
with initial levels of both externalizing and internalizing problems, whereas high levels of
fearfulness were associated with decreases in externalizing problems over time. High initial
levels of internalizing problems were associated with increases in internalizing problems over
time. In partial contrast, Russell, Hart, Robinson, and Olsen (2003) found that negative
emotionality as measured by parent report on the EAS did not predict preschoolers’ adjustment
as rated by teachers. However, EAS shyness was related to both lower prosocial behavior and
lower aggressive behavior at preschool.
Lemery et al. (2002) also provide support for a differential linkage model. Composited
mother CBQ ratings of child temperament at 3 and 4 years predicted both mother and father
reports of behavior problem symptoms at age 5, in a differentiated pattern. Whether or not the
temperament scales were Apurified@ by removing items overlapping with preschool age behavior
problems, early anger predicted later externalizing problems more strongly than it predicted later
internalizing problems, early fear and sadness predicted later internalizing problems more
strongly than they predicted later externalizing problems, and early inhibitory control inversely
predicted later externalizing or ADHD problems more strongly than it predicted later
internalizing problems.
The Dunedin Longitudinal Study (Caspi & Silva, 1995), mentioned previously in the
context of personality development, provides further support, and extends measures of
temperament from parent to experimenter ratings. Ratings based on the child=s behavior during
testing sessions, aggregated from 3 and 5 years, predicted aggregated ratings of parents and
teachers in late childhood (over ages 9 and 11) and early adolescence (over ages 13 and 15).
Early approach (outgoing responses to strangers and test materials–the inverse of inhibition)
predicted, inversely, internalizing problems better than externalizing problems for boys. It did
 Temperament 50

not predict either kind of problem for girls. Early sluggishness (a factor combining lack of
positive affect, passivity, and wariness/withdrawal from novelty) predicted later internalizing and
externalizing problems for girls, but not boys, as well as the relative absence of positive
competencies for both girls and boys. It is not clear how approach and sluggishness emerged
separately from a factor analysis describing similar dimensions, but whatever the underlying
distinction between the two dimensions, they predicted outcomes differently for the two genders.
A third temperament dimension, combining lack of control, irritability, and distractibility
(corresponding approximately to the resistance to control or manageability factors from parent-
report questionnaires), predicted, for both genders, externalizing problems more strongly than
internalizing problems or positive competencies. The discovery of differentiated patterns in
studies such as the BLS, CDP, and Dunedin study has occurred despite the tendency for
externalizing and internalizing adjustment scores to be somewhat correlated with each other,
making the pattern all the more remarkable.
In other recent studies, Morris et al. (2002) found that irritable temperament was
positively associated with first- and second-graders= externalizing and internalizing problems
equally, whereas effortful control was negatively associated with externalizing more strongly
than internalizing problems. Paterson and Sanson (1999) found that five-year-olds= low
persistence (attentive and on-task) was associated with externalizing, whereas low approach to
people and novel objects was associated with internalizing problems. Lengua, Wolchik, Sandler,
and West (2000) found 9- to 12-year-old impulsivity (CBQ) to be associated with conduct
problems and not with depression, and positive emotionality (DOTS) to be negatively associated
with depression, an association stronger than with conduct problems. In partial contrast to the
general pattern, however, Lengua et al. also reported that negative emotionality (DOTS) was
associated with depressive symptoms more strongly than with conduct problem symptoms.
Mun, Fitzgerald, Von Eye, Puttler, and Zucker (2001), also using the DOTS, reported
that withdrawal tendencies at 3-5 years were more strongly predictive of internalizing than
externalizing problems 3 years later. Their negative emotional reactivity scale also failed to
confirm the general pattern of roughly equal associations with both externalizing and
internalizing problems, instead predicting externalizing more strongly than internalizing
problems. Murphy, Shepard, Eisenberg, and Fabes (2004) found that both negative emotionality
and self regulation (whether assessed by teacher or parent questionnaires) in middle childhood
predicted social competence and behavior problems in early adolescence as rated by teacher or
parent reports. And Eisenberg et al. (2001) found that anger, impulsivity and low self-regulation
(observed and parent/teacher-rated) were more strongly associated with externalizing problems
(parent/teacher-rated), whereas sadness and low impulsivity were more strongly related to
internalizing problems.
Also notable is the study by Rothbart, Ahadi et al. (1994). Temperamental negative
affectivity (CBQ) was concurrently associated with a full range of social traits in 6- to 7-year-
olds, including aggressiveness, guilt, help seeking, and negativity (e.g., in response to suggestion
of a new activity). However, subcomponents of the general negative affect factor were
associated with the social traits in a more differentiated way: Fear and sadness were more related
to traits such as empathy and anger, and discomfort to aggression and help seeking. A small
subsample in the Rothbart, Ahadi et al. (1994) sample had been tested in the laboratory as
infants. Temperament as assessed in the laboratory five to six years earlier showed a somewhat
similar pattern of linkage with the social behavior outcomes: Infant laboratory activity (again,
usually regarded as part of surgency or positive affectivity) predicted aggressiveness and

View publication stats