Journal of Personality and Social Psychology Copyright 2006 by the American Psychological Association

2006, Vol. 91, No. 3, 538 –552 0022-3514/06/$12.00 DOI: 10.1037/0022-3514.91.3.538

Five Types of Personality Continuity in Childhood and Adolescence

Filip De Fruyt Meike Bartels

Ghent University Free University of Amsterdam

Karla G. Van Leeuwen, Barbara De Clercq, Mieke Decuyper, and Ivan Mervielde
Ghent University

This study examines 5 types of personality continuity—structural, mean-level, individual-level, differ-

ential, and ipsative—in a representative population (N ⫽ 498) and a twin and sibling sample (N ⫽ 548)
of children and adolescents. Parents described their children on 2 successive occasions with a 36-month
interval using the Hierarchical Personality Inventory for Children (I. Mervielde & F. De Fruyt, 1999).
There was evidence for structural continuity in the 2 samples, and personality was shown to be largely
differentially stable. A large percentage had a stable trait profile indicative of ipsative stability, and
mean-level personality changes were generally small in magnitude. Continuity findings were explained
mainly by genetic and nonshared environmental factors.

Keywords: personality continuity and assessment, five-factor model, childhood, adolescence, behavior
genetics

In the past decades, there has been a wealth of studies on differential continuity and distinct mean-level change patterns can
personality continuity that were recently summarized in two meta- coexist, underscoring the independency of these two types of
analyses (MAs) of longitudinal data on differential (Roberts & continuity (Block, 1971). Moreover, the two MAs combined help
DelVecchio, 2000) and mean-level (Roberts, Walton, & Viecht- to advance a clear conceptual distinction between the two conti-
bauer, 2006) continuity. Differential continuity describes the de- nuity types and are further useful to solve apparent inconsistencies
gree to which the relative differences among individuals remains among studies, summarizing the available data collected by the
invariant across time, whereas mean-level stability refers to the different researchers involved in the personality continuity debate.
extent to which personality scores change over time. Longitudinal The two MAs raised new questions about additional forms of
designs are required to investigate differential stability, looking at personality continuity and about the potential moderators and
trait correlations across time, whereas mean-level stability can be antecedents of stability.
studied with the use of longitudinal data (Roberts et al., 2006). In Although the MA on rank-order continuity (Roberts & DelVec-
addition, mean trait scores from cross-sectional age cohorts are chio, 2000) provided stability estimates from 3 years to old age,
useful for mean-level stability comparisons (McCrae et al., 1999, the majority of studies in the MA for the youngest age groups
2000). relied on a limited set of temperament and Q-sort measures.
A MA on differential continuity (Roberts & DelVecchio, 2000) Comprehensive and hierarchically organized age-specific person-
showed that people acquire increasingly stable relative trait posi- ality measures might be more appropriate to study homo- and
tions with age, with largely linear increases in stability until a heterotypic developmental changes at young age (Caspi, 1998).
plateau is reached at least after age 50. A MA on mean-level
The MA on mean-level changes (Roberts et al., 2006) started from
stability (Roberts et al., 2006) showed that people tend to increase,
age 10 onward, because there is a dearth of studies on this
especially in their 20s to 40s, in social dominance (a facet of
continuity type for younger ages. Today, the MAs allow for
extraversion), conscientiousness, and emotional stability. People
estimation of the amount of mean-level change and the stability
further demonstrate increases on social vitality (another facet of
correlation between any two ages from age 10 onward, but there is
extraversion) and openness in adolescence, but decrease on both
less evidence on younger age groups.
traits in old age. Absolute changes for agreeableness were ob-
In addition to differential and mean-level continuity, other types
served only in old age. These two MAs convincingly illustrate that
of personality development have been examined, although less
frequently (Caspi, 1998). Means on trait dimensions cannot be
compared directly across measurement points when the covariance
Filip De Fruyt, Karla G. Van Leeuwen, Barbara De Clercq, Mieke structure varies across time. Structural continuity refers to the
Decuyper, and Ivan Mervielde, Department of Developmental, Personality, invariance of the covariance structure across time and is a neces-
and Social Psychology, Ghent University, Ghent, Belgium; Meike Bartels, sary requirement for the assessment of mean-level stability
Department of Biological Psychology, Free University of Amsterdam,
(Biesanz, West, & Kwok, 2003). In addition to the analysis of
Amsterdam, the Netherlands.
Correspondence concerning this article should be addressed to Filip De group means for different ages, one can also examine individual-
Fruyt, Department of Developmental, Personality, and Social Psychology, level change. Individual-level change refers to the magnitude of
Ghent University, H. Dunantlaan 2, B-9000, Ghent, Belgium. E-mail: increase or decrease exhibited by a person on any given trait.
[email protected] These changes may be masked in an analysis of mean-level con-
538
 PERSONALITY CONTINUITY IN CHILDHOOD AND ADOLESCENCE 539

tinuity, because equal numbers of individuals may increase or temperamental constructs such as Negative Emotionality (Thomas
decrease on a trait, resulting in no change for the entire group. & Chess, 1977) or Emotionality (Buss & Plomin, 1984), Sociabil-
Finally, ipsative stability refers to the continuity of the configura- ity (Buss & Plomin, 1984) or Surgency (Rothbart & Derryberry,
tion of traits within the individual and provides information on the 1981), Task Persistence (Thomas & Chess, 1977) or Effortful
stability of the patterning of traits within a person across time, Control (Rothbart & Derryberry, 1981), and Activity Level (Buss
hence facilitating a person-centered approach to personality devel- & Plomin, 1984; Goldsmith & Campos, 1982; Thomas & Chess,
opment (Robins & Tracy, 2003). These additional types of per- 1977). They assumed that temperamental differences are expres-
sonality continuity have been studied mainly from adolescence to sions of neurobiological mechanisms that have a strong genetic
adulthood (McCrae et al., 1999, 2000; Roberts & Del Vecchio, basis (Mervielde, De Clercq, De Fruyt, & Van Leeuwen, 2005). In
2000; Robins, Fraley, Roberts, & Trzesniewski, 2001), although a contrast, personality traits are mainly used to chart stable latent
few studies have addressed particular types of personality conti- differences in adults, presumed to be partly influenced by temper-
nuity in childhood (e.g., Van Lieshout & Haselager, 1994). Caspi, ament and interaction with the environment.
Roberts, and Shiner (2005) argued that there are relatively few The revival of trait psychology, and especially the preponder-
studies that assess a comprehensive set of personality variables to ance of the FFM (Digman, 1990), challenged these viewpoints.
track continuities and changes over time. No study has addressed First, personality psychologists tend to agree that five broad di-
all five types simultaneously across a substantial time interval in mensions, that is, Neuroticism, Extraversion, Openness to Expe-
childhood and adolescence using a comprehensive and hierarchical rience, Agreeableness, and Conscientiousness, can be considered
five-factor model (FFM) personality measure. the basic dimensions underlying adult personality. Personality
The present study examines all five types of continuity in two psychologists interested in the developmental antecedents of this
different samples of schoolchildren and adolescents assessed at FFM subsequently showed that these five are also useful to de-
two time points spanning a 36-month interval. To assess whether scribe individual differences in childhood and adolescence (Dig-
the observed continuity and change patterns generalize across man, 1963; Digman & Inouye, 1986; John, Caspi, Robins, Moffit,
studies, we examined continuity types in a representative popula- & Stouthamer-Loeber, 1994; Kohnstamm, Halverson, Mervielde,
tion sample and a genetic-informative sample of twins and sib- & Havill, 1998; Lamb, Chuang, Wessels, Broberg, & Hwang,
lings. The different nature of the samples further enabled us to 2002). John et al. (1994) introduced the basic personality dimen-
investigate important additional questions. The representative sam- sions to developmental psychologists and demonstrated that the
ple allowed the description of patterns of continuity and change “Little Five” predict externalizing problem behavior in children.
that are notable in the general population of children and adoles- The qualifier “Little” refers to dimensions denoting trait differ-
cents. The twin and sibling sample allowed the examination of ences in childhood, paralleling the label “Big” used to refer to
genetic and environmental influences on personality continuity dimensions characterizing adults.
and change, enabling a genetic– environmental decomposition of Second, McCrae and colleagues (2000) argued against the arti-
the personality trait variances cross-sectionally but also of the trait ficial distinction between temperamental constructs and personal-
covariance across assessment points. McGue, Bacon, and Lykken ity traits, because there are strong empirical and conceptual links
(1993) conducted a similar study of young adulthood, demonstrat- between the domains of temperament and personality. The defin-
ing that the stable core of personality is strongly associated with ing characteristics of temperament variables also apply to traits,
genetic factors and that personality change largely reflects envi- including early observability, genetic basis, and pervasive impact
ronmental factors. They reported that, on average, “over 80% of on a range of behaviors (McCrae & Costa, 1997). Indeed, behavior
the variance of the stable component of the Time 2 phenotype was genetic studies consistently document the strong genetic basis of
associated with genetic factors” but also that “a majority of the FFM traits (Jang, McCrae, Angleitner, Riemann, & Livesley,
Time 2 personality variance is unrelated to variance expressed at 1998), with heritability estimates ranging from .40 to .60 depend-
Time 1” (McGue et al., 1993, p. 105). As far as we know, the ing on the trait and measure. McCrae and Costa (1996) conceptu-
present work is the first behavioral genetic study seeking to iden- alized the FFM dimensions as basic tendencies shaping the inter-
tify and characterize genetic and environmental influences on actions with the environment, resulting in characteristic
individual differences in stability and change in children and (mal)adaptations such as interests, values, and attitudes but also
adolescents using a comprehensive FFM measure. problem behavior and psychopathology. Mervielde, De Clercq, et
We examined personality continuity for a set of basic person- al. (2005) compared the major models of temperament with the
ality dimensions and facets using a lexically based measure as- FFM, illustrating the similarities and the conceptual overlap be-
sessing a broad range of personality traits relevant for childhood tween temperament models and the FFM. Caspi et al. (2005)
and adolescence. Adopting a comprehensive, age-appropriate in- recently concluded that temperament and personality increasingly
ventory increases the likelihood of detecting changes and enhances appear to be more alike than different.
the generalizability of findings. Although parents are the primary
informants on children’s and adolescents’ personalities, the repre- Assessing Personality at Young Age
sentative population sample also completed a self-report person-
ality adjective measure so that we could examine shared method A comprehensive assessment of age-specific indicators of traits
variance. is crucial to studying personality at young age and especially its
development. Different approaches have been adopted to assess
Temperament and Personality FFM dimensions in children and adolescents. Often, FFM mea-
sures—initially developed for adults—are used to describe differ-
Developmental and child psychologists traditionally describe ences in younger age groups (see, e.g., studies using the Revised
stable and observable differences in young children by relying on NEO Personality Inventory [NEO-PI–R] to assess adolescents’
540 DE FRUYT ET AL.

personality: De Clercq & De Fruyt, 2003; De Fruyt, Mervielde, groups. In addition, organismic theories such as Piaget’s theory of
Hoekstra, & Rolland, 2000; McCrae et al., 2002). Other studies cognitive development (Piaget, 1983) suggest that newly acquired
adapted the phrasing of personality items for younger age groups cognitive structures influence the way children and adolescents
(e.g., the junior version of Eysenck’s Personality Questionnaire interact with their environment. Finally, personality theories such
measure, EPQ–J; Eysenck, 1963; Eysenck, Makaremi, & Barrett, as Cloninger’s theory of character development (Cloninger, Svra-
1994). However, it can be argued that these adapted measures are kic, & Przybeck, 1993) discern qualitatively different life stages
probably not suitable for a fine-grained assessment of childhood that an individual has to master before a more advanced develop-
and adolescent personality differences and especially not to assess mental phase or level can be achieved. All these theories empha-
developmental change (De Clercq, De Fruyt, & Van Leeuwen, size developmental discontinuities, but it remains to be established
2004). Therefore, an alternative approach that is more sensitive to whether these discontinuities reflect changes in basic tendencies or
subtle personality differences at young age should be developed on whether they are restricted to changing characteristic adaptations
the basis of the full range of personality differences observable (McCrae & Costa, 1996). The demonstration of different forms of
prior to adulthood. stability across the FFM trait hierarchy in childhood and adoles-
The lexical approach to personality description (De Raad, & cence would underscore McCrae and Costa’s Five-Factor Theory
Perugini, 2002) provides a convincing rationale for the develop- (McCrae & Costa, 1996) and hence requires that the previously
ment of a comprehensive child and adolescent personality taxon- reviewed theories of personality development account not only for
omy. Mervielde and De Fruyt (1999) adopted this approach to trait change but also for trait stability.
construct such a taxonomy for classification of a large pool of Roberts and DelVecchio (2000) recently conducted an MA
parental personality descriptions of Flemish children aged between review of differential stability and examined whether and when
6 and 13 years (Kohnstamm et al., 1998). They subsequently stability peaks during the life course, challenging Costa and Mc-
developed the Hierarchical Personality Inventory for Children Crae’s claim (1994; McCrae & Costa, 1984, 1990) that personality
(HiPIC), representing the content of parental descriptions in short is “set like plaster” after age 30. They analyzed 3.217 test–retest
sentence items referring to concrete and observable behavior. correlation coefficients from 152 longitudinal studies and demon-
Therefore, the HiPIC can be considered a lexically based measure strated an increase in stability from .31 during childhood to .54 in
of the active parental vocabulary, in contrast with the NEO-PI–R,
young adulthood, mounting to .64 by age 30, with the highest
in which the facets are not derived empirically but are selected
stability of .74 observed between 50 and 70 years, when the time
after a careful search of the adult personality literature.
interval was held constant at 6.7 years. These data suggest that
The HiPIC items span five broad domains, labeled as Extraver-
personality is less stable during the preadult years. The estimated
sion, Benevolence, Conscientiousness, Emotional Instability or
population correlations controlling for the time interval of the
Neuroticism, and Imagination. Some domain labels differ from the
longitudinal study are .31 for ages 0 –2.9, .49 for ages 3–5.9, .43
lexical adult Big Five (Goldberg, 1993). The HiPIC dimensions
for ages 6 –11.9, and .43 for ages 12–17.9, respectively.
Extraversion, Conscientiousness, and Emotional Instability refer to
McCrae and colleagues (2002) examined both mean-level and
content that is similar to the adult Big Five counterparts and hence
differential changes during adolescence in self-descriptions using
received the same label. The HiPIC Benevolence factor, however,
the NEO-PI–R, including an analysis of continuity at the level of
refers to a broader set of traits than the adult Big Five or FFM
Agreeableness factor because it includes traits linked to the “easy– the individual. They found that mean-level personality scores for
difficult” child concept described in the temperament literature Extraversion, Agreeableness, and Conscientiousness were stable
(Thomas, Chess, Birch, Hertzig, & Korn, 1963). The Benevolence between ages 12 and 18, but Neuroticism appeared to increase in
factor refers to differences in the manageability of the child from girls and Openness to Experience increased in both boys and girls.
the perspective of the parent-informant. The HiPIC Imagination A 4-year longitudinal study of intellectually gifted students
domain comprises both Intellect and Openness to Experience showed a considerable degree of rank-order instability for gifted
items, blending the two alternative labels for the fifth factor students across the two assessment points. Stability coefficients for
emerging from adult adjective-based lexical studies (Goldberg, boys across a 4-year interval ranged from .31 (Agreeableness) to
1993) and the questionnaire-oriented FFM approach (Costa & .49 (Conscientiousness) and for girls from .30 (Neuroticism) to .63
McCrae, 1992). Given its specific focus and comprehensiveness, (Conscientiousness). Individual-level continuity analyses relying
the HiPIC can be considered a more sensitive measure to assess on the reliable change index (RCI; Jacobson & Truax, 1991)
personality change at young age (De Clercq et al., 2004). indicated that about 60% of the sample did not change over the
4-year interval for each of the FFM dimensions. A similar analysis
for older college-age individuals (Robins et al., 2001) showed,
Personality Development in Childhood and Adolescence
however, that almost 80% of the individuals were stable on each of
Different developmental theories conceive puberty as a signif- the FFM dimensions, suggesting that differential stability is sig-
icant stage for social and personality development (Grotevant, nificantly lower during adolescence.
1998), involving changing social interactions with parents and The MA of Roberts et al. (2006) on mean-level change during
peers and increasing societal influences. The importance of ado- adolescence (10 –18 years) showed a small but insignificant in-
lescence as a key transitional phase is also acknowledged by crease for Openness to Experience (d ⫽ .23; K ⫽ 13, N ⫽ 2,911),
Erikson (1950, 1968), who considered adolescence a second indi- and significant increases for Social Dominance (d ⫽ .20; K ⫽ 5,
viduation stage in which change is likely. Similarly, behavioral N ⫽ 1,700) and Emotional Stability (d ⫽ 16; K ⫽ 23, N ⫽
theories (Robin & Foster, 1989) emphasize that during adoles- 10,557). No significant mean-level changes were reported for
cence, learning processes and contingencies are embedded in novel Social Vitality, Agreeableness, and Conscientiousness during
social networks and environments, including changing peer adolescence.
 PERSONALITY CONTINUITY IN CHILDHOOD AND ADOLESCENCE 541

Block (1971) and Hart, Hofmann, Edelstein, and Keller (1997) Complex and differential interdependencies existed within the twin and
used the California Q-set to examine within-person (ipsative) sibling sample at the level of the units of observation, that is, within
change, but to our knowledge no patterns of trait change in monozygotic (MZ) and dizygotic (DZ) twin pairs and between twins and
children and adolescents were examined with an inventory specif- siblings, but also at the level of the informants, that is, parents rating 2 (in
the case of twins only) or 3 (twins ⫹ 1 sibling) children of the same family.
ically designed to assess the FFM. Studies investigating stability of
The interdependencies at the level of the units of observation included a
personality prototype classification using three personality pro-
mixture of interchangeable (Griffin & Gonzalez, 1995) and noninter-
files, that is, resilient, under-, and overcontrolled children, pro- changeable (Gonzalez & Griffin, 1999) cases, with same-sex twins to be
vided partial evidence of ipsative stability (De Fruyt, Mervielde, & considered interchangeable and different-sex or twin-sibling pairs as dis-
Van Leeuwen, 2002). However, it was unclear whether the insta- tinguishable. Such interdependencies might affect both the nature of the
bility of prototype classification for some individuals should be relationships observed across time and the level of significance (Gonzalez
attributed to real developmental change measurement error or the & Griffin, 2000). Because very different and complex analyses per type of
procedure to derive prototypes (De Fruyt et al., 2002). Structural personality continuity would have to be conducted to control for all
continuity has not been studied explicitly in childhood and possible dependencies that exist within such a dataset, a random sample of
adolescence. 1 child or adolescent twin per family was extracted to make the results
directly comparable with the results obtained from the other samples
involved in the present research, but also in previous MA work. The
Method random twin sample included 208 twin children (boys, N ⫽ 106; girls, N ⫽
102), whose ages were on average 8.50 years (SD ⫽ 1.92).
Participants All twins and siblings for whom complete data records were available
across the two assessment moments were used for the genetic analyses,
Two longitudinal samples of children and young adolescents were taking into account their interdependencies. The twin sample included 35
available for the present analyses.1 MZ male pairs; 44 MZ female pairs; and respectively, 33 male, 35 female,
Representative population sample. The first sample included children and 56 mixed-sex DZ pairs of twins. The average age of the twins was 8.34
participating in a follow-up study investigating parenting, children’s per- years (SD ⫽ 1.94), and the mean age for the siblings was 9.79 years (SD ⫽
sonalities, and problem behavior at two assessment occasions separated by 2.34).
a 36-month interval. Participating families including 1 child were recruited
by random sampling of elementary and secondary schools. The sample was Questionnaires
stratified by province (East and West Flanders), region (rural or urban),
school type (public, private, or Catholic schools) and grade (3rd, 4th, 5th, HiPIC. Parents rated both samples at the two assessment occasions on
and 6th year) for elementary schools. For secondary schools, subject the HiPIC (Mervielde & De Fruyt, 1999). The HiPIC includes 144 items
sampling was based on province (East and West Flanders), type of curric- grouped into 18 facets that are hierarchically organized under the higher
ulum (vocational, technical, and general education), and grade (1st and 2nd order factors. Parents were instructed to “describe the child by the way he
year). Eighty percent of the elementary schools and 60% of the secondary or she has most often behaved over the last year” by indicating on a 5-point
schools granted permission to contact parents for this study. In case of Likert-type scale the degree to which each statement was characteristic of
refusal, schools were replaced with other randomly selected schools. A the child to be assessed, with scale anchors labeled as barely characteristic,
letter addressed to the parents informed them about the goal and the slightly characteristic, more or less characteristic, characteristic, and
procedures of the research project. The response rate for parents of primary highly characteristic. All items have a similar grammatical format and are
schoolchildren was 41%, and for parents with children in secondary formulated in the third-person singular, avoiding negations in the item and
schools, 39%. At Time 2, 82% of the families continued collaboration, excluding personality-descriptive adjectives.
12% refused, and 6% could not be reached. This sampling method resulted Facet labels directly reflect the nature of the parental free descriptions
in a well-balanced sample regarding socioeconomic status, gender, and age and are in some cases indicative of opposite poles of the domain scale they
(Van Leeuwen, 2004). are assigned to, requiring the computation of reversed scores (R) before
Families were visited at home by a trained psychology student at each facets can be aggregated into a domain score. The HiPIC structure outlines,
assessment point. The mother, father, and child were instructed to inde- as follows (a sample item is included between parentheses): Extraversion
pendently complete a series of questionnaires. Parents described their own consists of Energy (“bubbles with life”), Expressiveness (“shows feel-
personality using the authorized Dutch translation of the NEO-PI–R (Costa ings”), Optimism (“laughs through life”), and Shyness (R: facet score to be
& McCrae, 1992) and the personality of their child(ren) using the HiPIC reversed; “needs time to get used to peers”); Benevolence comprises the
(Mervielde & De Fruyt, 1999). Children also provided self-ratings of their facets Altruism (“defends the weak”), Dominance (R; “acts the boss”),
personality using an adjective inventory at Time 2. Only participants who Egocentrism (R; “takes him/herself into consideration first”), Compliance
participated at both Time 1 and Time 2 were included in the study, (“sticks to arrangements”) and Irritability (R; “is quick to take offense”);
resulting in a sample of 498 families, including 238 boys and 260 girls. Conscientiousness includes Concentration (“works with sustained atten-
There were no mean-level personality differences between the dropouts tion”), Perseverance (“keeps at it when the going gets tough”), Orderliness
and those who continued participation, although both mothers and fathers (“takes care of his/her possessions”) and Achievement motivation (“wants
who continued participation scored lower on Neuroticism, and fathers also to be among the best”); Emotional Instability measures Anxiety (“is afraid
scored higher, on average, on Extraversion, Openness to Experience, and of failure”) and Self-Confidence (R; “has confidence in own abilities”);
Agreeableness. Indeed, continuing participation depended more on the and finally, Imagination is composed of the facets Creativity (“can use
personality of the parent than on that of the child. The mean age of the
participants was 10.9 years (SD ⫽ 1.8 years; range 7–15) at Time 1, and
13.9 years (SD ⫽ 1.8 years; range 10 –18) at Time 2. 1
Both samples were already used for other research purposes. The
Twin and sibling sample. The second sample included parents and representative population sample was already used to study the interaction
children participating in a small-scale twin family study, providing HiPIC between child personality and parental behavior as predictors of problem
ratings of their twins and 1 sibling, if eligible, per family. HiPIC data on behavior (Van Leeuwen, Mervielde, et al., 2004), and the twin and sibling
two assessment points with a 36-month interval were available for 548 sample for a study on the personality type approach (De Fruyt et al., 2002).
children, all between the ages of 5 and 14 years, with a mean of 8.65 years However, for both these purposes no data were reported on the continuity
(SD ⫽ 2.11 years). The sample included 271 boys and 277 girls. of personality.
542 DE FRUYT ET AL.

everyday things in a new way”), Intellect (“grasps the meaning of things mated at each measurement occasion in the first model, whereas in
quickly”), and Curiosity (“asks many ‘why’ questions”). Benevolence is the second model the intercorrelations were constrained to be
conceptually and empirically related to the adult FFM domain of Agree- equivalent across assessments. A significant difference in the fit
ableness, whereas Imagination is associated with the Openness to Experi- between these two models was considered indicative of structural
ence domain (De Fruyt et al., 2000).
change. The baseline model for the structural analysis at the Big
The HiPIC’s robust factor structure and high internal consistencies of
Five factor level was a single-indicator latent variable model, with
domains and facets have been documented in various studies with clinical
and nonclinical samples (Van Hoecke, De Fruyt, De Clercq, Hoebeke, & one latent variable associated with each of the 10 scores (five
Van de Walle, 2006; Van Leeuwen, De Fruyt, & Mervielde, 2004; Van dimensions ⫻ two assessment occasions). This is a fully saturated
Leeuwen, Mervielde, Braet, & Bosmans, 2004; Vollrath & Landolt, 2005). model, with the variances of the latent variables fixed to 1 and the
Domain scale reliabilities for the representative population and twin and variances of the residuals fixed to 0. The correlations among the
sibling samples ranged from .76 (Extraversion, Time 1; representative latent variables were freely estimated. In the second model, the
population sample) to .89 (Conscientiousness, Time 1; twin and sibling correlations between all pairwise traits across the two assessment
sample), and for the facet scales, from .77 (Self-Confidence, Time 2; points were constrained to be equal. For example, the correlation
representative population sample) to .91 (Intellect, Time 1; twin and sibling between Extraversion and Emotional Instability at Time 1 was
sample).
forced to equal the correlation between Extraversion and Emo-
Questionnaire Big Five. The population sample also provided self-
tional Instability at Time 2. Inspection of Table 2 shows that the
ratings on the Questionnaire Big Five (QBF; Gerris et al., 1998; Goldberg,
1992), a Dutch shortened version of Goldberg’s 100 adjectives, at the intercorrelations among the five HiPIC domains were invariant
second measurement point. The QBF includes 30 adjectives, 6 per FFM across measurement occasions for all age groups: age group 6 –7,
dimension, presented with a 7-point Likert-type scale. Scholte, van Aken, ␹2⌬(10, N ⫽ 88) ⫽ 15.69, p ⬎ .05; age group 8 –9, ␹2⌬(10, N ⫽
and van Lieshout (1997) and Dubas, Gerris, Janssens, and Vermulst (2002) 183) ⫽ 8.47, p ⬎ .05; age group 10 –11, ␹2⌬(10, N ⫽ 201) ⫽ 9.52,
have demonstrated that the QBF provides valid estimates of an individual’s p ⬎ .05; and age group 12–13, ␹2⌬(10, N ⫽ 210) ⫽ 11.45, p ⬎
standing on the FFM dimensions. Scale reliabilities were .77, .78, .67, .82, .05.
and .88 for Emotional Stability, Extraversion, Resourcefulness (the Open- The previous analyses at the domain level could also be ex-
ness scale of the QBF), Agreeableness, and Conscientiousness, respec- tended to examine structural invariance for the 18 facets across
tively (Van Leeuwen, 2004).
time. This very stringent test examining the invariance of 153
intercorrelations across time could not be done for the youngest
Informants age group because the number of parameters exceeds the number
In most cases, both parents provided HiPIC ratings of the children in the of participants. The 153 intercorrelations were not stable over time
population and the twin and sibling sample. Given their high intercorrela- for age group 8 –9, with ␹2⌬(153, N ⫽ 183) ⫽ 207.27, p ⬍ .01; or
tions, ranging from .57 (Emotional Instability, representative population for age group 10 –11, with ␹2⌬(153, N ⫽ 201) ⫽ 198.34, p ⬍ .01.
sample) to .77 (Conscientiousness, representative population sample), fa- Constraining the model for age group 12–13 showed no reduction
ther and mother ratings were averaged in both samples to obtain more in fit, with ␹2⌬(153, N ⫽ 210) ⫽ 175.31, p ⬎ .05, indicating that
reliable scores. Self-ratings on the QBF at follow-up were also available for the 153 facet intercorrelations were invariant for the oldest age
all children and adolescents of the representative sample, enabling an group across the 3-year interval.
examination of shared method effects. Factor structures of varimax-rotated principal component anal-
yses of the HiPIC facets could also be compared, calculating factor
Data Analytic Approach congruences across assessment occasions for each age group.
Table 3 shows that the factor congruence coefficients were all
The two samples are age heterogeneous (at Time 1, age ranged from
7 to 15 years in the representative sample and from 5 to 14 years in the higher than .90 except for the coefficients for Imagination from the
random selection of twins sample). The within-sample age range is thus age groups 8 –9 and 10 –11, which were .85 and .84, respectively.
much larger than the time interval of 36 months between the two
assessment points. To account for this large age range, we combined the Differential Continuity
representative sample and the random selection of twins (N ⫽ 682) and
then grouped in the following age groups at the first assessment: 6 –7 Differential continuity coefficients across the 36-month interval
years (N ⫽ 88), 8 –9 years (N ⫽ 183), 10 –11 years (N ⫽ 201), and (see Table 4), uncorrected for unreliability, were uniformly high
12–13 years (N ⫽ 210) to make the analyses more developmentally for all domains across the four age groups, with values ranging
informative. The five continuity types are examined primarily within
from .61 (age group 6 –7, Emotional Instability) to .86 (age group
these different age groups.
6 –7, Imagination), almost equaling test–retest reliabilities reported
in the manual (Mervielde, De Fruyt, & De Clercq, 2005). For the
Results first three age groups, coefficients for Emotional Instability were
smaller than those for the other FFM domains, and the magnitude
Structural Stability
of the correlations decreased slightly with increasing age for
The demonstration of invariance of the correlation matrix across Extraversion ( p ⬍ .01 between the youngest and the oldest age
the two measurement occasions was a conditio sine qua non before groups). Very similar conclusions could be drawn for the HiPIC
other forms of continuity could be examined. Table 1 reports the facets, with values ranging between .57 (age group 6 –7, Anxiety)
intercorrelations among the Time 1 and Time 2 domain scores and .83 (age group 6 –7, Altruism), suggesting that considerable
averaged across mothers and fathers for the four age groups. differential stability is manifested across all hierarchical levels of
Similarly to Robins et al. (2001), we examined structural continu- the FFM.
ity using structural equation modeling comparing the fit of two Cross-time correlations computed separately for maternal and
models. The correlations among the five factors were freely esti- paternal ratings (not reported in Table 4) were about .05 to .10
 PERSONALITY CONTINUITY IN CHILDHOOD AND ADOLESCENCE 543

Table 1
Intercorrelations Among the Five HiPIC Dimensions at Time 1 and Time 2 by Age Group

HiPIC dimension 1. Emotional instability 2. Extraversion 3. Imagination 4. Agreeableness 5. Conscientiousness

Age 6–7 years

1. Emotional Instability — ⫺.33** ⫺.41*** ⫺.05 .04

2. Extraversion ⫺.47*** — .28** .19 .00
3. Imagination ⫺.36*** .45*** — .14 .42***
4. Agreeableness .03 ⫺.04 ⫺.01 — .42***
5. Conscientiousness .02 ⫺.09 .38*** .39*** —

Age 8–9 years

1. Emotional Instability — ⫺.40*** ⫺.27*** ⫺.18* ⫺.04

2. Extraversion ⫺.40*** — .48*** .03 ⫺.04
3. Imagination ⫺.33*** .46*** — .15* .40***
4. Agreeableness ⫺.07 ⫺.02 .01 — .50***
5. Conscientiousness ⫺.03 ⫺.02 .35*** .44*** —

Age 10–11 years

1. Emotional Instability — ⫺.40*** ⫺.37*** ⫺.16* ⫺.20**

2. Extraversion ⫺.37*** — .48*** .03 .04
3. Imagination ⫺.43*** .45*** — .16* .48***
4. Agreeableness ⫺.10 ⫺.00 .07 — .47***
5. Conscientiousness ⫺.22*** ⫺.01 .45*** .42*** —

Age 12–13 years

1. Emotional Instability — ⫺.44*** ⫺.39*** ⫺.19** ⫺.17*

2. Extraversion ⫺.43*** — .52*** .13 .22***
3. Imagination ⫺.44*** .49*** — .22** .54***
4. Agreeableness ⫺.13 .06 .10 — .49***
5. Conscientiousness ⫺.20** .17* .52*** .33*** —

Note. HiPIC ⫽ Hierarchical Personality Inventory for Children. Intercorrelations at Time 1 are reported below the diagonal, and intercorrelations at Time
2 are reported above the diagonal; 6 –7 years: N ⫽ 88, 8 –9 years: N ⫽ 183, 10 –11 years: N ⫽ 201, and 12–13 years: N ⫽ 210.
* p ⬍ .05. ** p ⬍ .01. *** p ⬍ .001.

lower than the averaged parental ratings for domains and facets, single maternal ratings could be corrected for unreliability using
suggesting that the magnitude of the stability coefficients does not 12-week test–retest reliabilities described in the manual (Mer-
primarily result from the increased reliability to be expected from vielde, De Fruyt, & De Clercq, 2005) for maternal raters: .72
averaging across parents. The continuity coefficients computed for (Emotional Instability), .74 (Extraversion), .83 (Imagination), .78
(Benevolence), and .82 (Conscientiousness). Adopting these cor-
rections for unreliability provided averaged stability coefficients
Table 2 for maternal domain ratings of .93, .91, .89, and .86 for the age
HiPIC Domain and Facet-Level Structural Continuity Analyses groups 6 –7, 8 –9, 10 –11, and 12–13, respectively. Correcting the
by Age Group Across a 36-Month-Interval coefficients obtained for the facets showed averaged corrected
Age group Chi-square df p CFI facet stability coefficients of .92, .90, .88, and .85 for the age
groups 6 –7, 8 –9, 10 –11, and 12–13, respectively.
Domain level

6–7 years 15.69 10 .109 0.98 Table 3

8–9 years 8.47 10 .583 1.00
10–11 years 9.52 10 .483 1.00 HiPIC Factor Congruence Analyses Across a 36-Month Interval
12–13 years 11.45 10 .324 1.00 per Age Group

Facet level Age group EI E I B C

a
6–7 years 6–7 years .90 .95 .90 .98 .95
8–9 years 207.27 153 .00230 0.99 8–9 years .97 .96 .85 .98 .91
10–11 years 198.34 153 .00797 1.00 10–11 years .97 .95 .84 .99 .98
12–13 years 175.31 153 .10457 1.00 12–13 years .99 .98 .97 .99 .98

Note. HiPIC ⫽ Hierarchical Personality Inventory for Children; 6 –7 Note. HiPIC ⫽ Hierarchical Personality Inventory for Children. 6 –7
years: N ⫽ 88, 8 –9 years: N ⫽ 183, 10 –11 years: N ⫽ 201, and 12–13 years: N ⫽ 88, 8 –9 years: N ⫽ 183, 10 –11 years: N ⫽ 201, and 12–13
years: N ⫽ 210; CFI ⫽ comparative fit index. years: N ⫽ 210; EI ⫽ Emotional Instability; E ⫽ Extraversion; I ⫽
a
Total sample size is smaller than the number of parameters. Imagination; B ⫽ Benevolence; C ⫽ Conscientiousness.
544 DE FRUYT ET AL.

Table 4
HiPIC Differential Continuity Analyses by Age Group Across a 36-Month Interval

HiPIC domain and facet 6–7 years 8–9 years 10–11 years 12–13 years

Emotional Instability .61 .65 .63 .69

Anxiety .57 .62 .60 .65
Self-confidence .61 .63 .66 .66
Extraversion .83 .78 .76 .66
Energy .81 .71 .76 .63
Expressiveness .74 .75 .69 .67
Optimism .75 .64 .70 .59
Shyness .77 .78 .68 .70
Imagination .86 .69 .77 .69
Creativity .82 .72 .73 .65
Intellect .82 .70 .76 .74
Curiosity .80 .64 .74 .62
Benevolence .77 .71 .79 .75
Altruism .83 .73 .71 .69
Dominance .72 .61 .78 .70
Egocentrism .69 .63 .71 .69
Compliance .66 .69 .73 .66
Irritability .75 .69 .76 .70
Conscientiousness .76 .82 .77 .74
Concentration .77 .80 .75 .73
Perseverance .72 .74 .73 .67
Order .68 .80 .73 .73
Achievement Striving .72 .76 .72 .66

Note. HiPIC ⫽ Hierarchical Personality Inventory for Children; 6 –7 years: N ⫽ 88, 8 –9 years: N ⫽ 183,
10 –11 years: N ⫽ 201, and 12–13 years: N ⫽ 210. All correlations significant at p ⬍ .001.

To evaluate effects of having the same informants (parents) and the different informant and measures’ perspectives, with coeffi-
measures (HiPIC) across assessment occasions, we also examined cients, uncorrected for unreliability, varying between .21
differential stability using different measures and raters in the (Benevolence–Agreeableness) and .47 (Conscientiousness), with
representative population sample. HiPIC-averaged parental ratings all coefficients on the diagonal showing the largest magnitude.
obtained at Time 1 were correlated with QBF self-ratings assessed These coefficients are in line with other studies using self- and
3 years later. Coefficients are provided in Table 5. To enable an parental reports in children and adolescents (De Clercq, De Fruyt,
evaluation of the effect of having different raters (self vs. parents) Koot, & Benoit, 2004). The coefficients across the 36-month
and measures, we also describe the correlations between HiPIC interval were in general about .10 lower than the same-time coef-
and QBF at follow-up. The follow-up coefficients primarily reflect ficients, with again all coefficients at the diagonal being largest.

Table 5
Cross-Method/Informant Correlations Across 36 Months (Representative Population Sample)

HiPIC domain 1. QBF–EmS 2. QBF–Ext 3. QBF–Res 4. QBF–Agr 5. QBF–Con

3-year interval

1. HiPIC–EI ⫺.27*** ⫺.22*** ⫺.07 .02 .06

2. HiPIC–E .07 .33*** .19*** .11* ⫺.08
3. HiPIC–I .08 .09* .26*** .05 .02
4. HiPIC–B .10* ⫺.05 ⫺.01 .12** .11*
5. HiPIC–C .05 ⫺.01 .07 .08 .36***

Same-time assessment

1. HiPIC–EI ⫺.38*** ⫺.31*** ⫺.12** ⫺.02 .07

2. HiPIC–E .15*** .47*** .23*** .19*** ⫺.01
3. HiPIC–I .13** .13** .35*** .10* .10
4. HiPIC–B .13** ⫺.04 ⫺.03 .21*** .24***
5. HiPIC–C .03 ⫺.05 .07 .17*** .47***

Note. HiPIC ⫽ Hierarchical Personality Inventory for Children; QBF ⫽ Questionnaire Big Fire; QBF–EmS ⫽
Emotional Stability; QBF–Ext ⫽ Extraversion; QBF–Res ⫽ Resourcefulness; QBF–Agr ⫽ Agreeableness;
QBF–Con ⫽ Conscientiousness; HiPIC–EI ⫽ Emotional Instability; HiPIC–E ⫽ Extraversion; HiPIC–I ⫽
Imagination; HiPIC–B ⫽ Benevolence; HiPIC–C ⫽ Conscientiousness.
* p ⬍ .05. ** p ⬍ .01. *** p ⬍ .001.
 PERSONALITY CONTINUITY IN CHILDHOOD AND ADOLESCENCE 545

The differences in magnitude between the different-measures and tensen & Mendoza, 1986; Jacobson & Truax, 1991). Robins et al.
different-raters’ correlations across time—relative to the same- (2001) and Roberts, Caspi, and Moffitt (2001) used this RCI to
measure parental correlations across the 36-month interval—are examine reliable change across the FFM traits. The RCI has been
thus largely attributable to the different informants and measures, developed to assess the clinical significance of change after ther-
rather than testimony of poor differential continuity across this apeutic intervention and is computed as RCI ⫽ X2⫺X1/Sdiff, where
time lag. X1 represents a person’s score at Time 1, X2 represents that same
McCrae (1994) showed by path analytic arguments that the person’s score at Time 2, and Sdiff is the standard error of differ-
stability of the true score can be easily estimated by dividing the ence between the two test scores. Computing RCIs for a person’s
predictive correlation by the concurrent correlation. The concur- Big Five profile enables researchers to determine how many indi-
rent correlation between observers (self vs. parent) and/or different viduals remain stable on their five-trait pattern across time but also
instruments (HiPIC vs. QBF) sets an upper limit to the agreement provides information on the frequency of FFM individual-level
that can be expected, and the closer the cross-time correlation change patterns. A child can be stable on four of the FFM dimen-
across observers is to this concurrent correlation, the more stable sions but decline on Emotional Instability. This type of analysis is
the trait (McCrae, 1994, p. 163). Adopting this estimation method particularly useful to examine the kind and direction of individual
to the present data shows true score validity estimates of .71 changes but is also informative to describe the denser or more
(.27/.38), .70 (.33/.47), .74 (.26/.35), .57 (.12/.21), and .77 (.36/.47) frequent trait-change configuration patterns in a large sample.
for Emotional Instability, Extraversion, Imagination, Benevolence, Frequent individual trait-change configurations point to the kind of
and Conscientiousness, respectively. changes that can be expected during particular life stages, espe-
cially when replicable across samples of the same age range.
Mean-Level Continuity Difference scores per individual are compared with the distri-
bution of change scores that would be expected from error of
We examined mean-level stability using repeated measures measurement alone and hence separates true change from change
analysis of variance, including gender as a covariate (Costa, Ter- attributable to measurement error. Such analyses require indepen-
racciano, & McCrae, 2001; McCrae et al., 2002). The analyses of dent test–retest reliability estimates. For the present analyses,
mean-level domain differences presented in Table 6 demonstrate HiPIC test–retest estimates for maternal3 ratings across a 12-week
strong parallels across the youngest age groups, with no mean- interval reported in the manual (Mervielde, De Clercq, et al., 2005)
level differences reported for the age groups 6 –7 and 8 –9 for each were used, that is, .72 (Emotional Stability), .74 (Extraversion), .83
of the Big Five. Small mean-level decreases in Emotional Insta- (Imagination), .78 (Benevolence), and .82 (Conscientiousness).
bility were found for age groups 10 –11, F(1, 199) ⫽ 9.57, p ⬍ .01, Change scores exceeding a 95% confidence interval are assumed
ε2 ⫽ .05, and 12–13, F(1, 208) ⫽ 9.86, p ⬍ .01, ε2 ⫽ .05; and to represent true change (McCrae et al., 2002; Robins et al., 2001).
decreases in Imagination, F(1, 208) ⫽ 10.11, p ⬍ .01, ε2 ⫽ .05, The majority of participants did not change FFM positions
and Conscientiousness, F(1, 208) ⫽ 4.13, p ⬍ .05; ε2 ⫽ .02, were during the 36-month interval across the different age groups.
found for age group 12–13. No mean-level domain changes were Sixty-seven percent (age group 8 –9 years) to 77% (age group 6 –7
found for Extraversion and Benevolence. years) were ascribed similar FFM scores at the two assessment
Mean-level stability analyses at the fact level (not reported in a points. If children or adolescents changed on personality scores,
table) showed a mean decrease for Dominance, F(1, 86) ⫽ 5.27, change was usually restricted to one (about 20% across age
p ⬍ .05, ε2 ⫽ .06, Optimism, F(1, 86) ⫽ 6.42, p ⬍ .05, ε2 ⫽ .07, groups) or two (about 5%–10%) FFM domains. None of the
and Creativity, F(1, 86) ⫽ 6.89, p ⬍ .01, ε2 ⫽ .07, for age group participants in the total sample (N ⫽ 682) changed on each of the
6 –7; a mean increase for Altruism, F(1, 181) ⫽ 5.46, p ⬍ .05, FFM domains, and only 1 adolescent of age group 12–13 changed
ε2 ⫽ .03, for age group 8 –9; and mean decreases for Energy, F(1, scores on four of the FFM domains. Some trait change patterns
199) ⫽ 10.87, p ⬍ .001, ε2 ⫽ .05, Anxiety, F(1, 199) ⫽ 12.10, p ⬍ were denser across age groups, with a substantial number of
.05, ε2 ⫽ .06, and Creativity, F(1, 199) ⫽ 48.55, p ⬍ .001, ε2 ⫽ individuals (more than 10%) from age groups 8 –9 and 12–13 years
.20, for age group 10 –11. The largest number of changes was exhibiting decreased scores on Imagination.
found for age group 12–13, with mean decreases for facets across
all domains: Irritability, F(1, 208) ⫽ 11.52, p ⬍ .001, ε2 ⫽ .05; Ipsative Continuity
Achievement Striving, F(1, 208) ⫽ 13.31, p ⬍ .001, ε2 ⫽ .06;
Energy, F(1, 208) ⫽ 4.84, p ⬍ .05, ε2 ⫽ .03; Expressiveness, F(1, Ipsative continuity is usually examined with two methods. The
208) ⫽ 7.62, p ⬍ .01, ε2 ⫽ .04; Anxiety, F(1, 208) ⫽ 17.91, p ⬍ first approach relies on Cronbach and Gleser’s observation (1953)
.001, ε2 ⫽ .08; Creativity, F(1, 208) ⫽ 15.31, p ⬍ .001, ε2 ⫽ .07;
and Curiosity, F(1, 208) ⫽ 14.41, p ⬍ .001, ε2 ⫽ .07. In general,
2
the magnitude of these differences was small, except for the There has been some debate about the appropriateness of the RCI to
decrease in Creativity for age group 10 –11, and the majority of the account for regression to the mean effects when used to evaluate the effects
of therapeutic interventions. However, this criticism applies less to the
observed differences were not significant after application of the
present application of the RCI, because no explicit assumptions are made
Bonferroni correction for the number of statistical tests.
regarding the direction of the expected change. Individuals can become, for
example, more extraverted or more introverted.
Individual-Level Continuity 3
An anonymous reviewer pointed out that we do not have short-term
test–retest correlations for HiPIC ratings averaged across parents, which is
We also examined whether mean-level continuity extended to true. However, as we demonstrated, single parental ratings are only about
the individual level, examining the number of individuals showing .05 to .10 smaller across a 3-year interval than averaged ratings, so it can
decreased, equal, or increased trait scores, using the RCI2 (Chris- be expected that differences are even smaller across a shorter interval.
546 DE FRUYT ET AL.

Table 6
HiPIC Domain Mean-Level Continuity Analyses by Age Group Across 36 Months

Domain and age group Time 1 Time 2 F p ε2

Emotional Instability
6–7 years 22.41 22.79 .36 ns
8–9 years 21.85 20.57 .76 ns
10–11 years 21.39 20.32 9.57 .01 .05
12–13 years 21.69 20.11 9.86 .01 .05
Extraversion
6–7 years 29.17 28.67 2.33 ns
8–9 years 28.85 28.15 .22 ns
10–11 years 28.70 27.63 1.29 ns
12–13 years 27.82 26.89 2.14 ns
Imagination
6–7 years 30.38 29.25 .92 ns
8–9 years 30.67 29.45 2.57 ns
10–11 years 29.86 28.59 3.31 ns
12–13 years 28.68 27.58 10.11 .01 .05
Benevolence
6–7 years 26.56 26.94 .83 ns
8–9 years 27.90 28.42 .03 ns
10–11 years 28.00 28.44 1.31 ns
12–13 years 27.90 27.92 1.88 ns
Conscientiousness
6–7 years 26.20 25.66 .01 ns
8–9 years 26.59 26.28 1.13 ns
10–11 years 26.24 26.02 .03 ns
12–13 years 25.80 25.22 4.13 .05 .02

Note. 6 –7 years: N ⫽ 88, 8 –9 years: N ⫽ 183, 10 –11 years: N ⫽ 201, and 12–13 years: N ⫽ 210; F according
to Wilks’s lambda; degrees of freedom for all analyses are (1, 86) for 6 –7 years, (1, 181) for 10 –11 years, and
(1, 208) for 12–13 years; ns ⫽ nonsignificant. HiPIC ⫽ Hierarchical Personality Inventory for Children.

that individual profiles can vary in three major ways: elevation (the participants, assuming no reliable changes in profiles across time:
average level of scores), scatter (the variability of scores), and that is, with similar scores for elevation, scatter, and shape per age
shape (the patterning of scores). Cronbach and Gleser (1953) group. This simulation produced distributions for the D2, D⬘2, and
developed three indices, D2, D⬘2, and D⬙2, for quantifying these D⬙2 indices per age group, with the 95th percentiles for age group
sources of variance. D2 is sensitive to differences in elevation, 6 –7 years at 98.65, 84.46, and 1.31, respectively; for age group
scatter, and shape and quantifies the squared differences between 8 –9 years at 89.62, 76.16, and 1.19, respectively; for age group
Big Five traits at two assessment occasions. D⬘2 is only sensitive 10 –11 years at 93.24, 79.11, and 1.20, respectively; and finally, for
to differences in scatter and shape and quantifies the squared age group 12–13 years at 108.60, 92.58, and 1.39, respectively.
differences between Big Five profiles after each profile has been Individuals with values beyond these 95th percentile values were
centered around its mean. Finally, D⬙2 only reflects differences in considered to have significantly changed profiles. Respectively,
shape and quantifies the squared differences between profiles after 9.1%, 14.8%, 14.9%, and 16.7% of the children per age group had
each profile has been standardized (Cronbach & Gleser, 1953; D2 values beyond the simulated cutoffs; 5.7%, 10.4%, 11.4%, and
Robins et al., 2001). Using these profile similarity indices, Robins
12.9% had D⬘2 values beyond the simulated cutoffs; and 9.1%,
and colleagues (2001) demonstrated that only 17% of their under-
8.2%, 6.0%, and 9.0% had D⬙2 values beyond the respective
graduate sample showed significant changes in the shape of their
cutoffs, suggesting that children’s and adolescents’ profiles pri-
profile across a 4-year interval. The majority of the changes in
marily reflected changes of elevation and scatter, but less so in
profiles were related to changes in elevation and/or scatter but not
terms of shape. Less than 10% of the individuals across all age
shape.
Parallel to the study by Robins et al. (2001, p. 626), probabilities groups exhibited changes in the shape of the profile.
were estimated by simulating trait scores on a sample of 50,000 A second but related approach to examining ipsative stability is
individuals with identical levels of elevation, scatter, and shape in to compute Q correlations, that is, within-person correlations
a person’s profile at the two measurement points and examining across the HiPIC domains or facets at Time 1 and Time 2. Robins
corresponding distributions. Simulated trait scores were based on et al. (2001) found that Big Five profile correlations ranged from
means, variances, and covariances estimated from the real data, .95 to .97 during college years, with a mean of .61 (SD ⫽ .39) and
and the test–retest reliability coefficients reported in the manual a median of .76. Within-person correlations have to be evaluated
(Mervielde, De Clercq, et al., 2005) were used to estimate the error against the distribution of within-person correlations that can be
variance. found in a sample with a similar mean and standard deviation, but
Across the four age groups, the D2 values ranged from 35.70 to in which profiles are randomly paired across assessment occasions.
45.68, the D⬘2 values from 22.88 to 30.46, and the D⬙2 from .20 to Robins et al. (2001) found an average value of .20 for a simulated
.22. To interpret these values, we simulated four samples of 50,000 data set.
 PERSONALITY CONTINUITY IN CHILDHOOD AND ADOLESCENCE 547

Q correlations across the 18 HiPIC facets were computed per ness at Time 1 and Time 2) and age (older children showed lower
age group and compared with the distributions of Q correlations levels of Emotional Instability and Extraversion at Time 1 and
for simulations randomly pairing Time 1 with Time 2 assessments Time 2) on the mean were found. Estimates for the full ACE model
(i.e., data were from the same group of persons, but scores were with the 95% confidence intervals are given.
paired randomly across time) per age group, having identical No significant differences in variances and covariances for boys
means and standard deviations at Time 1 and Time 2 as the real and girls were observed for any of the domains ( p saturated model
data. Across the four age groups, the median Q correlations ranged fitting is ⬎ .05 for all domains at the two time points), so for the
from .81 to .85, far above the median Q correlations in the remaining genetic analyses, MZ boys and girls were combined in
simulated data (.35 to .43), suggesting stability of the within- one group, and DZ boys and girls (including twins of opposite sex)
individual facet-trait profile for a large number of children and were combined in one group.
adolescents. Standardized estimates (with 95% confidence intervals) of ad-
ditive genetic, shared environmental, and nonshared environmen-
Behavior Genetic Determinants of Continuity tal influences on all domains at Time 1 and Time 2 and on the
covariance between Time 1 and 2 are given in Table 7. Heritability
Saturated model. Because of the small sample size, we com-
for the distinct domains varied from 7% for Emotional Instability
bined MZ twins and DZ twins across gender to increase the power
of the study. If nonsignificant effects of sex on variances and at the second assessment occasion and 43% for Imagination at
covariances are found, MZ male and MZ female twins can be Time 1. Relatively large influences of nonshared environment
combined in one group, and DZ male, DZ female, and DZ twins of were observed, partly representing measurement error. Shared
opposite sex can be combined in one group. To allow sex differ- environmental influences were small, except for Imagination, but
ences in mean, we implemented sex as a covariate in the genetic it should be kept in mind that this variance component is hard to
model fitting. Tests were conducted with a saturated model in Mx detect, and insufficient power gives rise to underestimation of the
(see http://www.psy.vu.nl/mxbib). effects of shared environment. Stability of the personality domains
Genetic model fitting. Cross-sectional heritabilities and ge- between the two assessment points was accounted for by additive
netic and environmental influences on stability of the domains genetic, shared environmental, and nonshared environmental in-
were estimated with the use of bivariate genetic analyses in Mx fluences (see the boldfaced items in Table 7). The additive genetic
(Neale, 2003). A Cholesky decomposition, with sex and age as standardized estimate for the trait covariance across time varied
covariates (definition variables), was used. The Cholesky decom- between .19 (Emotional Instability) and .51 (Benevolence). Esti-
position is descriptive and not driven by a specific developmental mates for the cross-time trait covariance for the shared environ-
hypothesis. It decomposes a covariance matrix into genetic and ment were usually small, except for Imagination (.34), and esti-
nongenetic covariance matrices and is a first approach to obtaining mates for the nonshared environment ranged from .31
genetic and environmental correlations across time in longitudinal (Imagination) to .68 (Emotional Instability).
datasets. Because of the small sample sizes, the analyses lacked Genetic and environmental correlations (i.e. overlap in influenc-
power to test for the significance of the distinct variance compo- ing genes or environmental factors) between the first and the
nents. However, significant effects of sex (girls showed higher second assessment points are given in Table 8. Genetic correla-
ratings of Emotional Instability, Benevolence, and Conscientious- tions and shared environmental correlations were 1.0 or almost 1.0

Table 7
Standardized Estimates of Additive Genetic, Shared Environmental and Nonshared Environmental Influences on Variances and
Covariances

HiPIC domains A at Time 1 A at Time 2 C at Time 1 C at Time 2 E at Time 1 E at Time 2

Emotional Instability
Time 1 .23 (.005–.43)a .11 (.00–.23)d .66 (.49–.87)g
Time 2 .19 (⫺.02–.46)c .07 (.00–.23)b .13 (.00–.27)e .06 (.00–.16)f .68 (.45–.90)h .87 (.71–.97)i
Extraversion
Time 1 .39 (.16–.56) .06 (.00–.16) .55 (.40–.76)
Time 2 .40 (.12–.62) .28 (.05–.49) .10 (.00–.23) .10 (.00–.23) .50 (.32–.76) .61 (.44–.83)
Imagination
Time 1 .43 (.28–.57) .29 (.17–.40) .28 (.20–.41)
Time 2 .35 (.18–.52) .29 (.11–.45) .34 (.21–.46) .30 (.20–.41) .31 (.20–.46) .41 (.29–.57)
Benevolence
Time 1 .38 (.17–.54) .00 (.00–.05) .61 (.46–.83)
Time 2 .51 (.24–.69) .40 (.16–.57) .02 (⫺.02–.10) .07 (.00–.16) .47 (.30–.74) .53 (.38–.76)
Conscientiousness
Time 1 .36 (.14–.53) .00 (.00–.07) .64 (.47–.86)
Time 2 .46 (.20–.65) .38 (.15–.55) .00 (.00–.07) .00 (.00–.06) .54 (.35–.80) .62 (.45–.85)

Note. Boldfaced values indicate covariances, and 95% confidence intervals are in parentheses. HiPIC ⫽ Hierarchical Personality Inventory for Children;
A ⫽ heritability; C ⫽ shared environment; E ⫽ nonshared environment.
a
Heritability Time 1; b Heritability Time 2; c Influence of A on covariance between Time 1 and Time 2.
d
Shared environment Time 1; e Shared environment Time 2; f Influence of C on covariance between Time 1 and Time 2.
g
Nonshared environment Time 1; h Nonshared environment Time 2; i Influence of E on covariance between Time 1 and Time 2.
548 DE FRUYT ET AL.

Table 8
Genetic and Environmental Correlations for the HiPIC Domains

HiPIC domain and

time A at Time 1 A at Time 2 C at Time 1 C at Time 2 E at Time 1 E at Time 2

Emotional instability
Time 1 1 1 1
Time 2 1.00 1 1.00 1 .57 1
Extraversion
Time 1 1 1 1
Time 2 .94 1 1.00 1 .67 1
Imagination
Time 1 1 1 1
Time 2 .82 1 .94 1 .74 1
Benevolence
Time 1 1 1 1
Time 2 1.00 1 1.00 1 .63 1
Conscientiousness
Time 1 1 1 1
Time 2 .98 1 .97 1 .67 1

Note. HiPIC ⫽ Hierarchical Personality Inventory; A ⫽ heritability; C ⫽ shared environment; E ⫽ nonshared

environment.

for all HiPIC domains, suggesting one underlying set of genes and invariance at the HiPIC facet level could not be demonstrated for
one underlying set of shared environmental influences for person- the 8 –9-year-olds and the 10 –11-year-olds, although the chi-
ality at both assessment occasions. Nonshared environmental cor- square values were small considering the large number of param-
relations varied from .57 to .74, suggesting that large parts overlap eters, and the CFI indices were 1.0 or close to 1.0.
but that there are also time-specific nonshared environmental in- The present analyses clearly underscored that the positioning of
fluences, presumably measurement error. the major dimensions of personality in childhood and adolescence
is stable across a substantial time interval for different ages in
Discussion childhood and adolescence. The demonstration of structural con-
tinuity for these age groups is a prerequisite before ipsative, mean-,
The present study examined five types of personality continuity and individual-level personality continuity data can be adequately
in children and adolescents as rated by their parents on a lexically examined and interpreted (Biesanz et al., 2003) and hence further
based measure specifically designed to assess personality at young legitimates the interpretation of MA findings reported for adoles-
age. The present work extends two recent MAs on differential cence (Roberts et al., 2006).
(Roberts & DelVecchio, 2000) and mean-level (Roberts et al.,
2006) personality continuity, assessing the two continuity types for
younger age groups and examining three additional types of con- Differential or Rank-Order Continuity
tinuity for children and adolescents. Moreover, genetic and envi-
ronmental influences on trait continuity were also estimated. The rank ordering of individuals across the 3-year interval was
demonstrated to be very stable, with coefficients usually beyond
.70 — uncorrected for unreliability—for the domains of Extraver-
Structural Continuity
sion, Imagination, Benevolence, and Conscientiousness. Differen-
The SEM analyses showed that the covariance among HiPIC tial continuity for Emotional Instability was somewhat lower but
domains was clearly invariant across a 36-month interval for the still above .60. These findings largely generalized across the
different age groups, providing evidence for structural stability different age groups, except for a decline in rank-order continuity
at the five-factor level from childhood to late adolescence. The for Extraversion with increasing age. Analyses at the facet-level
results from these analyses at the domain level were also produced very similar results. Slightly lower correlation coeffi-
confirmed by the congruency analyses comparing factor struc- cients were observed for maternal ratings. These could be further
tures for the different age groups across time points. All con- corrected for unreliability showing correlations, averaged across
gruence coefficients were larger than .85, a value considered FFM domains, ranging from .93 for the 6 –7-year-olds to .86 for
indicative of structural replicability (Haven & ten Berge, 1977), the 12–13-year-olds.
except for a minor deviation (.84) for Imagination in the 10 –11 The results obtained here differ in three respects from previous
age group. reports on rank-order continuity. First, much higher correlations
The most stringent test of structural personality continuity, were found than those described in the MA on rank-order stability
examining the longitudinal invariance of the 153 facet intercorre- of Roberts and DelVecchio (2000) and the self-report results
lations, was positive only for the oldest age group of 12–13 years, described by McCrae et al. (2002). The coefficients obtained in the
suggesting that structural invariance applies to both hierarchical present work double the correlations reported for this life stage in
levels of the HiPIC, at least from 12–13 to 15–16 years. Structural the MA of Roberts and DelVecchio (2000), estimated to be .43 for
 PERSONALITY CONTINUITY IN CHILDHOOD AND ADOLESCENCE 549

12-year-olds, and they are also much higher than the median nevolence. No mean-level changes for Extraversion were found in
4-year stability coefficient of .38 reported by McCrae et al. (2002) the present work, and mean-level Dominance scores slightly de-
in a sample of gifted youth. The present coefficients obtained from creased for the youngest age group in childhood but did not show
parental ratings are similar to the correlations reported for adults, normative changes thereafter. The majority of differences at the
underscoring that FFM traits—when rated by parents—show al- facet level were reported for the late adolescents in our sample,
ready differential stability at young age. although the magnitude of these differences was small, except the
Second, correlation coefficients were similar across age groups, decrease in Curiosity reported for the 10 –11-year-olds. Curiosity
except for a decrease of differential stability for Extraversion in the and also Creativity showed a slight decrease in the 12–13-year-
oldest group. These findings do not confirm the expectation (Rob- olds. These findings are counter to those of Roberts et al., who
erts & DelVecchio, 2000) that differential continuity will increase found an increase in Openness during adolescence. An anonymous
with age. According to the present data, individuals have a stable reviewer suggested that perhaps some of the HiPIC Curiosity and
positioning relative to each other in elementary and the first years Creativity items are less applicable for these older age groups and
of secondary school, except for Extraversion, at least when paren- are hence less frequently endorsed. However, inspection of the
tal reports are considered. content of the Curiosity and Creativity items suggests that all items
A final difference is that in adulthood, self- and peer-reported are applicable to a broader age range, including adolescence.
rank-order continuity provide more similar coefficients (Costa &
McCrae, 1988), whereas the findings from the present study— Individual-Level Changes
compared with those reported by Roberts and DelVecchio (2000)
and McCrae et al. (2002)—suggest large differences in rank-order The analysis of the individual trait-change patterns demon-
stability of adolescent self-reports versus parental reports. It could strated that two thirds (age group 8 –9) to three quarters (age group
be argued that the higher rank-order coefficients for parental 6 –7) of all individuals did not show reliable change on any of the
ratings are due to the tendency of parents to retain a lasting image FFM dimensions underscoring individual-level continuity for a
of their child despite real changes. Indeed, involving the same substantial number of individuals. If change occurred, it was
observers and measures across assessment occasions makes it usually restricted to one FFM domain. Only one trait-change
difficult to disentangle observer from developmental effects. How- pattern occurred more frequently; that is, children and adolescents
ever, the analysis correlating HiPIC parental assessments at Time showing decreased scores on Imagination, especially in the 8 –9-
1 with QBF self-ratings at follow-up showed that the lower cross- year-old group and the 12–13-year-old group. These observations
time correlations were mainly attributable to the different infor- are counter to the findings of Roberts et al. (2006) and McCrae et
mant perspectives (parent vs. self-rating) and the different inven- al. (2002).
tories (HiPIC vs. QBF), rather than questioning differential The implications of the individual-level findings for the person-
stability. Relying on path-analytic arguments (McCrae, 1994), we ality research agenda are twofold. First, the present findings un-
see that true estimates for all HiPIC domains were above .70, derscore the necessity to study the determinants of stability (Caspi
except for a lower value of .57 for Benevolence in a design with et al., 2005), because the majority of children and adolescents are
different raters but also different measures. These values are cer- ascribed stable positions on traits. Secondly, determinants of
tainly higher than those meta-analytically computed by Roberts change have to be investigated (Mroczek & Spiro, 2003), although
and DelVecchio (2000) but remain considerably lower than simi- the absence of dense trait-change patterns in the present samples
larly estimated true-score stabilities in adults as reported by Costa suggests that the nature and direction of trait changes vary widely
and McCrae (1988). The lower validities of the self-reports in within the population, underscoring the necessity to examine very
other studies might be alternatively explained by a limited insight large groups. If the divergent pattern of individual trait changes
in personality in early adolescence. would be replicated in such large samples, this would suggest that
trait change is largely specific for the individual or even random,
Mean-Level Stability rather than normative.

No mean-level changes in childhood (age groups 6 –7 and 8 –9 Ipsative Stability

years) were reported at the HiPIC domain level. In line with the
findings of Roberts et al. (2006), mean-level decreases in Emo- Although both individual-level and ipsative continuity indices
tional Instability were reported from age 10 onwards, and no focus on the individual, they are psychometrically very different.
changes were reported for Benevolence. Contrary to the Roberts et The computation of the RCI to inspect individual-level changes
al. (2006) findings, Conscientiousness and Imagination showed requires a comparison with the standard deviation of the popula-
small decreases in age group 12–13, whereas Roberts et al. found tion, contrary to the analysis of ipsative stability that relies exclu-
no change for Conscientiousness and found increased scores, al- sively on the individual’s scores (De Fruyt, Van Leeuwen, Bagby,
though insignificant, for Openness to Experience. Rolland, & Rouillon, 2006). Ipsative stability was examined using
Roberts et al. (2006) distinguished two facets of Extraversion— two procedures applied to different hierarchical levels of the FFM.
Social Vitality and Social Dominance—and reported a significant Q correlations were computed across the individual’s facet scores
increase for Social Dominance in adolescence. Social Vitality at the two assessment points, whereas Cronbach and Gleser’s
reflects traits such as sociability, positive affect, gregariousness, (1953) D2, D⬘,2 and D⬙2 indices were used to examine differences
and energy level and is probably closely related to the HiPIC’s among HiPIC domain-trait profiles across time.
Extraversion factor, including Energy, Shyness (R), Expressive- The results from both types of analyses strongly converged.
ness, and Optimism, whereas their Social dominance category is Substantially higher Q correlations within individuals across time
probably best reflected by the HiPIC’s Dominance facet of Be- were observed than for simulated ratings, assuming a random
550 DE FRUYT ET AL.

pairing of Time 1 and Time 2 scores. Using Cronbach and Gleser’s volving teachers as informants for young children and asking
(1953) D indices, we found that less than 10% of the sample adolescents to provide self-ratings. Third, the genetic-informative
exhibited changes in the shape of their domain profile, suggesting sample is relatively small, hampering the statistical power of the
that if change occurs, it is mostly change of elevation or scatter. analyses. The analyses that were conducted are informative, con-
Similar ipsative analyses conducted across a 4-year interval in sidering the dearth of genetic-informative studies on individual
undergraduate students (Robins et al., 2001) showed that 17% differences in personality development in childhood and adoles-
demonstrated changes in the shape of their profile, suggesting that cence, but larger samples would enable a series of more detailed
individual-level changes are observed more frequently in young and powerful analyses, including the exploration of the FFM lower
adulthood than in childhood or adolescence. level traits, and behavior genetic analyses of different types of
Behavior genetic determinants of continuity. The twin and personality continuity, for example, the trait pattern across time.
sibling sample, although small in size, provided a unique oppor- Fourth, a potential threat for longitudinal research is that more
tunity to decompose trait variances cross-sectionally as well as stable individuals do not drop out, and hence personality continuity
trait covariances across assessment points. In line with previous is more likely to be demonstrated than change. However, continu-
behavior genetic studies (Caspi et al., 2005; Loehlin, 1992), we ity is demonstrated across all traits not only for traits related to
estimated the additive genetic variance for the HiPIC domains to dropout in longitudinal research, such as Conscientiousness (De
be around .30, varying between .07 (Emotional Instability, Time 2) Fruyt & Mervielde, 1999). In addition, prolonged participation in
and .43 (Imagination, Time 1). Smaller estimates were obtained for studies with children depends more on parental decisions. In that
the shared environment, except for Imagination, and the remaining case, parents who are more likely to continue participation de-
variance was explained by the nonshared environment, including scribe their children as more consistent, which is unlikely to be the
measurement error. The small variance explained by the shared case. Finally, the present study included only two assessment
environment for the HiPIC domains (except for Imagination) has occasions. Additional assessment points, preferably across a
to be interpreted with caution, because the small sample size longer time span, would allow the use of latent-growth curve
makes it difficult to detect shared environmental influences. These modeling and provide better ways to handle measurement error
results are much in line with the patterns observed in other behav- and offer opportunities to address new questions.
ior genetic studies on individual differences in personality in In conclusion, the evidence for different types of personality
adulthood (Jang et al., 1998). continuity supports and extends previous research demonstrating
Most interesting were the analyses on the genetic– that the level of continuity in childhood and adolescence is higher
environmental decomposition of the trait covariances (representing than often expected (e.g., compared with young adulthood; Rob-
stability), showing very similar patterns for Extraversion, Benev- erts et al., 2006). Caspi and colleagues (2005) argued in this
olence, and Conscientiousness, with 40% (Extraversion) to 51% respect that personality trait development is not a continuity-
(Benevolence) of the covariance across time accounted for by versus-change proposition, but that continuity and change coexist.
genetic factors, 0% to 10% explained by the shared environment, The major challenge for developmental theories will be to account
and around 50% by the nonshared environment. The covariance of not only for trait changes but also for trait continuity.
Emotional Instability was largely explained by the nonshared
environment and to a moderate extent by genetic factors and 13%
by the common environment. Finally, the three variance compo-
References
nents explained an almost equal amount of covariance for Imagi- Biesanz, J. C., West, S. G., & Kwok, O.-M. (2003). Personality over time:
nation. The magnitude of the genetic and shared environmental Methodological approaches to the study of short-term and long-term
correlations (1.0 or near 1.0) suggests that the same set of genes development and change. Journal of Personality, 71, 905–941.
and shared environmental influences determines the continuity Block, J. (1971). Lives through time. Berkeley, CA: Bancroft.
observed across time, whereas the lower values for the nonshared Buss, A. H., & Plomin, R. (1984). Temperament: Early developing per-
environment suggest that different environmental factors, which sonality traits. Hillsdale, NJ: Erlbaum.
are unique to an individual, operate across time. Caspi, A. (1998). Personality development across the life-course. In W.
Strengths and limitations. The present study has a number of Damon & N. Eisenberg (Eds.), Handbook of child psychology: Social,
emotional, and personality development (pp. 311–388). Wiley: New
strengths, including the use of a hierarchical and comprehensive
York.
personality measure specifically designed to assess traits at young Caspi, A., Roberts, B. W., & Shiner, R. L. (2005). Personality develop-
age, as well as the availability of a population-based representative ment: Stability and change. Annual Review of Psychology, 56, 453– 484.
sample and a genetic informative sample. Christensen, L., & Mendoza, J. L. (1986). A method assessing change in a
However, this work also has a number of limitations that should single subject: An alteration of the RC index. Behaviour Therapy, 17,
be taken into account when interpreting the results. First, continu- 305–308.
ity was examined across a limited time lag with the use of only one Cloninger, C. R., Svrakic, D. M., & Przybeck, T. R. (1993). A psychobi-
inventory. It could be argued that more change is to be expected ological model of temperament and character. Archives of General
across longer time intervals, and the observed trait-change pat- Psychiatry, 50, 975–990.
terns— or the absence of change— could be specific for the HiPIC. Costa, P. T., Jr., & McCrae, R. R. (1988). Personality in adulthood: A
six-year longitudinal study of self-reports and spouse ratings on the NEO
For example, the decrease in Curiosity in adolescence might be
Personality Inventory. Journal of Personality and Social Psychology,
particular for the HiPIC. It cannot be excluded that more changes 54, 853– 863.
will be observed across a longer time interval and for different Costa, P. T., Jr., & McCrae, R. R. (1992). Revised NEO Personality
personality measures. Second, the continuity types were examined Inventory and Five-Factor Inventory Professional Manual. Odessa, FL:
with parents as primary informants on children’s and adolescents’ Psychological Assessment Resources.
personality. Useful extensions of this research could include in- Costa, P. T., Jr., & McCrae, R. R. (1994). Set like plaster? Evidence for the
 PERSONALITY CONTINUITY IN CHILDHOOD AND ADOLESCENCE 551

stability of adult personality. In T. F. Heatherton & J. L. Weinberger Gonzalez, R., & Griffin, D. (2000). On the statistics of interdependence:
(Eds.), Can personality change? (pp. 21– 40). Washington, DC: Amer- Treating dyadic data with respect. In W. Ickes & S. Duck (Eds.), The
ican Psychological Association. social psychology of personal relationships (pp. 181–213). London:
Costa, P. T., Jr., Terracciano, A., & McCrae, R. R. (2001). Gender Wiley & Sons.
differences in personality traits across cultures: Robust and surprising Griffin, D., & Gonzalez, R. (1995). Correlational analysis of dyad-level
findings. Journal of Personality and Social Psychology, 81, 322–331. data in the exchangeable case. Psychological Bulletin, 118, 430 – 439.
Cronbach, L. J., & Gleser, G. C. (1953). Assessing similarity between Grotevant, H. D. (1998). Adolescent development in family contexts. In W.
profiles. Psychological Bulletin, 50, 456 – 473. Damon& N. Eisenberg (Eds.), Handbook of child psychology: Social,
De Clercq, B., & De Fruyt, F. (2003). Personality disorder symptomatol- emotional, and personality development (pp. 1097–1149). Wiley: New
ogy in adolescence: A Five-Factor Model perspective. Journal of Per- York.
sonality Disorders, 17, 269 –292. Hart, D., Hofmann, V., Edelstein, W., & Keller, M. (1997). The relation of
De Clercq, B., De Fruyt, F., Koot, H. M., & Benoit, Y. (2004). Quality of childhood personality types to adolescent behavior and development: A
life, perceived competence and traits in children surviving cancer: A longitudinal study of Icelandic children. Developmental Psychology, 32,
state-trait multiple-rater perspective. Journal of Pediatric Psychology, 195–205.
29, 579 –590. Haven, S., & ten Berge, J. M. F. (1977). Tucker’s coefficient of congruence
De Clercq, B., De Fruyt, F., & Van Leeuwen, K. (2004). A “little five” as a measure of factorial invariance: An empirical study (Tech. Rep. No.
lexically-based perspective on personality disorder symptoms in adoles- 290 EX).Groningen, the Netherlands: University of Groningen.
cence. Journal of Personality Disorders, 18, 479 – 499. Jacobson, N. S., & Truax, P. (1991). Clinical significance: A statistical
De Fruyt, F., & Mervielde, I. (1999). RIASEC types and Big Five traits as approach to defining meaningful change in psychotherapy research.
predictors of employment status and nature of employment. Personnel Journal of Consulting and Clinical Psychology, 59, 12–19.
Psychology, 52, 701–727. Jang, K. L., McCrae, R. R., Angleitner, A., Riemann, R., & Livesley, W. J.
De Fruyt, F., Mervielde, I., Hoekstra, H. A., & Rolland, J.-P. (2000). (1998). Heritability of facet-level traits in a cross-cultural twin sample:
Assessing adolescents’ personality with the NEO PI–R. Assessment, 7, Support for a hierarchical model of personality. Journal of Personality
329 –345. and Social Psychology, 74, 1556 –1565.
De Fruyt, F., Mervielde, I., & Van Leeuwen, K. (2002). The consistency of John, O. P., Caspi, A., Robins, R. W., Moffitt, T. E., & Stouthamer-Loeber,
personality type classification across samples and five-factor measures. M. (1994). The little 5—Exploring the nomological network of the
European Journal of Personality, 16, 57–72. 5-factor model of personality in adolescent boys. Child Development,
De Fruyt, F., Van Leeuwen, K. G., Bagby, R. M., Rolland, J. P., & 65, 160 –178.
Rouillon, F. (2006). Assessing and interpreting personality change and Kohnstamm, G. A., Halverson, C. F., Mervielde, I., & Havill, V. L. (1998).
continuity in patients treated for major depression. Psychological As- Parental descriptions of child personality. Developmental antecedents
sessment, 18, 71– 80. of the Big Five? The LEA series in personality and clinical psychology.
De Raad, B., & Perugini, M. (2002). Big Five factor assessment: Intro- Mahwah, NJ: Erlbaum.
duction. In B. de Raad & M. Perugini (Eds.), Big Five assessment (pp. Lamb, M. E., Chuang, S. S., Wessels, H., Broberg, A. G., & Hwang, C. P.
1–26). Göttingen: Hogrefe & Huber. (2002). Emergence and construct validation of the Big Five in early
Digman, J. M. (1963). Principal dimensions of child personality as inferred childhood: A longitudinal analysis of their ontogeny in Sweden. Child
from teacher’s judgements. Child Development, 34, 43– 60. Development, 73, 1517–1524.
Digman, J. M. (1990). Personality structure: Emergence of the five-factor Loehlin, C. J. (1992). Genes and environment in personality development.
model. Annual Review of Psychology, 41, 417– 440. London, UK: Sage.
Digman, J. M., & Inouye, J. (1986). Further specification of the 5 robust McCrae, R. R. (1994). The counterpoint of personality assessment: Self-
factors of personality. Journal of Personality and Social Psychology, 50, reports and observer ratings. Assessment, 2, 159 –172.
116 –123. McCrae, R. R., & Costa, P. T., Jr. (1997). Personality trait structure as
Dubas, J. S., Gerris, J. R. M., Janssens, J. R. M., & Vermulst, A. A. (2002). human universal. American Psychologist, 52, 509 –516.
Personality types of adolescents: Concurrent correlates, antecedents, and McCrae, R. R., & Costa, P. T., Jr. (1984). Emerging lives, enduring
type ⫻ parenting interactions. Journal of Adolescence, 25, 79 –92. dispositions: Personality in adulthood. Boston: Little, Brown.
Erikson, E. (1950). Childhood and society. New York: Norton. McCrae, R. R., & Costa, P. T., Jr. (1990). Personality in adulthood. New
Erikson, E. H. (1968). Identity: Youth and crisis. New York: Norton. York: Guilford.
Eysenck, S. B. G. (1963). Junior Eysenck Personality Inventory. San McCrae, R. R., & Costa, P. T., Jr. (1996). Toward a new generation of
Diego, CA: Educational and Industrial Testing Service and Human personality theories: Theoretical contexts for the five-factor model. In
Services. J. S. Wiggins (Eds.), The five-factor model of personality: Theoretical
Eysenck, S. B. G., Makaremi, A., & Barrett, P. T. (1994). A cross-cultural perspectives (pp. 51– 87). New York: Guilford.
study of personality—Iranian and English children. Personality and McCrae, R. R., Costa, P. T., Jr., de Lima, M. P., Simoes, A., Ostendorf, F.,
Individual Differences, 16, 203–210. Angleitner, A., et al. (1999). Age differences in personality across the
Gerris, J. R. M., Houtmans, M. J. M., Kwaaitaal-Roosen, E. M. G., De adult life span: Parallels in five cultures. Developmental Psychology, 35,
Schipper, J. C., Vermulst, A. A., & Janssens, J. M. A. M. (1998). 466 – 477.
Parents, adolescents and young adults in Dutch families: A longitudinal McCrae, R. R., Costa, P. T., Jr., Ostendorf, F., Angleitner, A., Hrebickova,
study. Nijmegen, the Netherlands: Institute of Family Studies. M., Avia, M. D., et al. (2000). Nature over nurture: Temperament,
Goldberg, L. R. (1992). The development of markers for the Big-Five personality and lifespan development. Journal of Personality and Social
factor structure. Psychological Assessment, 4, 26 – 42. Psychology, 78, 173–186.
Goldberg, L. R. (1993). The structure of phenotypic personality traits. McCrae, R. R., Costa, P. T., Jr., Terraciano, A., Parker, W. L., Mills, C. J.,
American Psychologist, 48, 26 –34. De Fruyt, F., & Mervielde, I. (2002). Personality trait development from
Goldsmith, H. H., & Campos, J. J. (1982). Toward a theory of infant age 12 to 18: Longitudinal, cross-sectional, and cross-cultural analyses.
temperament. In R. N. Emde & R. J. Harmon (Eds.), The development Journal of Personality and Social Psychology, 83, 1456 –1468.
of attachment and affiliative systems (pp. 161–193). New York: Plenum. McGue, M., Bacon, S., & Lykken, D. T. (1993). Personality stability and
Gonzalez, R., & Griffin, D. (1999). The correlational analysis of dyad-level change in early adulthood: A behavioral genetic analysis. Developmental
data in the distinguishable case. Personal Relationships, 6, 449 – 469. Psychology, 29, 96 –109.
552 DE FRUYT ET AL.

Mervielde, I., De Clercq, B., De Fruyt, F., & Van Leeuwen, K. (2005). differences in temperament. In M. E. Lamb & A. L. Brown (Eds.),
Temperament, personality and developmental psychopathology as child- Advances in developmental psychology (pp. 37– 86). Hillsdale, NJ: Erl-
hood antecedents of personality disorders. Journal of Personality Dis- baum.
orders, 19, 171–201. Scholte, R. H. J., van Aken, M. A. G., & van Lieshout, C. F. M. (1997).
Mervielde, I., & De Fruyt, F. (1999). Construction of the Hierarchical Adolescent personality factors in self-ratings and peer nominations and
Personality Inventory for Children (HiPIC). In I. Mervielde, I. Deary, F. their prediction of peer acceptance and peer rejection. Journal of Per-
De Fruyt, & F. Ostendorf (Eds.), Personality psychology in Europe. sonality Assessment, 69, 534 –554.
Proceedings of the Eight European Conference on personality psychol- Thomas, A., Chess, L. A., Birch, H. G., Herzig, M. E., & Korn, S. (1963).
ogy (pp. 107–127). Tilburg, the Netherlands: Tilburg University Press. Behavioral individuality in early childhood. New York: New York
Mervielde, I., De Fruyt, F., & De Clercq, B. (2005). Handleiding Hiërar- University Press.
chische Persoonlijkheidsvragenlijst voor Kinderen [Manual Hierarchi- Thomas, A., & Chess, S. (1977). Temperament and development. New
cal Personality Inventory for Children]. Ghent, Belgium: Ghent Univer- York: Brunner/Mazel.
sity, Department of Developmental, Personality, and Social Psychology. Van Hoecke, E., De Fruyt, F., De Clercq, B. J., Hoebeke, P., & Van de
Mroczek, D. K., & Spiro, A., III. (2003). Modeling intraindividual change Walle, J. (2006). Internalising and externalising problem behavior in
in personality traits: Findings from the Normative Aging Study. Journal childhood enuresis: A Five-Factor Model perspective. Journal of Pedi-
of Gerontology Psychology Science, 58B, 153–165. atric Psychology, 31, 460 – 468.
Neale, M. C. (2003). Mx: Statistical modeling (3rd ed). Richmond: Uni- Van Leeuwen, K. (2004). Parenting and personality as predictors of child
versity of Virginia. and adolescent internalizing and externalizing problem behavior. Un-
Piaget, J. (1983). Piaget’s theory. In P. H. Mussen (Ed.), Handbook of child published doctoral dissertation, Ghent University, Ghent, Belgium.
psychology: Vol. 1, History, theory, and methods (pp. 103–128). New Van Leeuwen, K., De Fruyt, F., & Mervielde, I. (2004). A longitudinal
York: Wiley. study of the utility of the resilient, overcontrolled and undercontrolled
Roberts, B. W., Caspi, A., & Moffitt, T. E. (2001). The kids are alright: personality types as predictors of children’s and adolescents’ problem
Growth and stability in personality development from adolescence to behavior. International Journal of Behavioral Development, 28, 210 –
adulthood. Journal of Personality and Social Psychology, 81, 670 – 683. 220.
Roberts, B. W., & DelVecchio, W. F. (2000). The rank-order consistency Van Leeuwen, K. G., Mervielde, I., Braet, C., & Bosmans, G. (2004). Child
of personality traits from childhood to old age: A quantitative review of personality and parental behavior as moderators of problem behavior:
longitudinal studies. Psychological Bulletin, 126, 3–25. Variable- and person-centered approaches. Developmental Psychology,
Roberts, B. W., & Walton, K., & Viechtbauer, W. (2006). Patterns of 40, 1028 –1046.
mean-level change in personality traits across the life course: A meta- Van Lieshout, C. F. M., & Haselager, G. J. T. (1994). The Big Five
analysis of longitudinal studies. Psychological Bulletin, 132, 1–25. personality factors in Q-sort descriptions of children and adolescents. In
Robin, A. L., & Foster, S. L. (1989). Negotiating parent–adolescent C. F. Halverson, Jr., G. A. Kohnstamm, & R. P. Martin (Eds.). The
conflict: A behavioral–family systems approach. New York: Guilford developing structure of temperament and personality from infancy to
Press. adulthood (pp. 293–318). Hillsdale NJ: Erlbaum.
Robins, R. W., Fraley, R. C., Roberts, B. W., & Trzesniewski, K. H. Vollrath, M., & Landolt, M. A. (2005). Personality predicts quality of life
(2001). A longitudinal study of personality change in young adulthood. in pediatric patients with unintentional injuries: A one-year follow-up
Journal of Personality, 69, 617– 640. study. Journal of Pediatric Psychology, 30, 481– 491.
Robins, R. W., & Tracy, J. L. (2003). Setting an agenda for a person-
centered approach to personality development. Monographs of the So- Received May 12, 2005
ciety for Research in Child Development, 68, 110 –122. Revision received April 4, 2006
Rothbart, M. K., & Derryberry, D. (1981). Development of individual Accepted April 6, 2006 䡲