Developmental Personality Types From Childhood to Adolescence: Associations With Parenting and...

16
Developmental Personality Types From Childhood to Adolescence: Associations With Parenting and Adjustment Amaranta D. de Haan Ghent University and Utrecht University Maja Dekovi c, Alithe L. van den Akker, Sabine E. M. J. Stoltz, and Peter Prinzie Utrecht University This study examined whether changes in childrens self-reported Big Five dimensions are represented by (developmental) personality types, using a cohort-sequential design with three measurement occasions across 5 years (four cohorts, 912 years at T1; N = 523). Correlates of, and gender differences in, type membership were examined. Latent class growth modeling yielded three personality types: Resilients (highest initial levels on all Big Five), Overcontrollers (lowest Extraversion, Emotional Stability, Imagination), and Undercontrollers (lowest Benevolence, Conscientiousness). Gender differences in type membership were small. Warm parenting, but not overreactive discipline, in childhood was associated with type membership. The types differed in adjustment problems by the end of middle adolescence. Personality change more likely occurs at the level of dimensions within types than in type membership. The life phase from late childhood into adoles- cence is a distinctive formative period, in which children need to adjust to the physical, hormonal, and psychosocial changes that accompany the onset of puberty (Galambos & Costigan, 2003; Smetana, Campione-Barr, & Metzger, 2006). Given the fact that personality is a developmentally dynamic construct (Roberts, Walton, & Vie- chtbauer, 2006), the pronounced developmental changes during this life phase are likely reected by personality change. Accordingly, research investigating personality development during childhood (De Fruyt et al., 2006; Prinzie & Dek- ovi c, 2008) or adolescence (Branje, van Lieshout, & Gerris, 2007; Klimstra, Hale, Raaijmakers, Branje, & Meeus, 2009) has been steadily accumulating. Research that examines personality development during the transition to adolescence is, however, very limited. Approaches to Studying Personality There are two approaches to the study of person- ality, the person-centered and variable-centered approaches (Mervielde & Asendorpf, 2000). The person-centered approach is based on the theoreti- cal rationale that complex systems, such as the human psyche, are best described by the patterning or conguration of their respective subsystems (Allport, 1937). A central goal of the person-centered approach is to identify groups or subsets of individ- uals who have similar congurations of personality characteristics and thus share the same basic per- sonality structure (i.e., personality types; Block, 1971). A well-known conceptualization of childhood personality types is based on the theory of ego- control and ego-resilience (Block & Block, 1980). Ego-resilience refers to the tendency to respond exibly rather than rigidly to changing situational demands, particularly stressful situations. Ego- control refers to the tendency to contain (overcontrol) versus express (undercontrol) impulses. Moderate ego-control is related to high resiliency, and high and low ego-control are both related to low ego-resiliency (Robins, John, Caspi, Moftt, & Stouthamer-Loeber, 1996). Combining ego-control and ego-resiliency results in three personality types: Resilients, who are well able to adjust their impulse control to environmental demands (high ego-resil- iency, moderate ego-control); Undercontrollers, who have low levels of impulse control (low ego-resil- iency, low ego-control), and Overcontrollers, who have high impulse control but low ability to adjust their impulse control to environmental demands Correspondence concerning this article should be addressed to Amaranta D. de Haan, Department of Developmental, Personal- ity, and Social Psychology (PP07), Ghent University, Henri Dunantlaan 2, B-9000 Ghent, Belgium. Electronic mail may be sent to [email protected]. © 2013 The Authors Child Development © 2013 Society for Research in Child Development, Inc. All rights reserved. 0009-3920/2013/xxxx-xxxx DOI: 10.1111/cdev.12092 Child Development, xxxx 2013, Volume 00, Number 0, Pages 116

Transcript of Developmental Personality Types From Childhood to Adolescence: Associations With Parenting and...

Developmental Personality Types From Childhood to Adolescence:Associations With Parenting and Adjustment

Amaranta D. de HaanGhent University and Utrecht University

Maja Dekovi�c, Alithe L. van den Akker,Sabine E. M. J. Stoltz, and Peter Prinzie

Utrecht University

This study examined whether changes in children’s self-reported Big Five dimensions are represented by(developmental) personality types, using a cohort-sequential design with three measurement occasions across5 years (four cohorts, 9–12 years at T1; N = 523). Correlates of, and gender differences in, type membershipwere examined. Latent class growth modeling yielded three personality types: Resilients (highest initial levelson all Big Five), Overcontrollers (lowest Extraversion, Emotional Stability, Imagination), and Undercontrollers(lowest Benevolence, Conscientiousness). Gender differences in type membership were small. Warm parenting,but not overreactive discipline, in childhood was associated with type membership. The types differed inadjustment problems by the end of middle adolescence. Personality change more likely occurs at the level ofdimensions within types than in type membership.

The life phase from late childhood into adoles-cence is a distinctive formative period, in whichchildren need to adjust to the physical, hormonal,and psychosocial changes that accompany theonset of puberty (Galambos & Costigan, 2003;Smetana, Campione-Barr, & Metzger, 2006). Giventhe fact that personality is a developmentallydynamic construct (Roberts, Walton, & Vie-chtbauer, 2006), the pronounced developmentalchanges during this life phase are likely reflectedby personality change. Accordingly, researchinvestigating personality development duringchildhood (De Fruyt et al., 2006; Prinzie & Dek-ovi�c, 2008) or adolescence (Branje, van Lieshout, &Gerris, 2007; Klimstra, Hale, Raaijmakers, Branje,& Meeus, 2009) has been steadily accumulating.Research that examines personality developmentduring the transition to adolescence is, however,very limited.

Approaches to Studying Personality

There are two approaches to the study of person-ality, the person-centered and variable-centeredapproaches (Mervielde & Asendorpf, 2000). Theperson-centered approach is based on the theoreti-cal rationale that complex systems, such as the

human psyche, are best described by the patterningor configuration of their respective subsystems(Allport, 1937). A central goal of the person-centeredapproach is to identify groups or subsets of individ-uals who have similar configurations of personalitycharacteristics and thus share the same basic per-sonality structure (i.e., “personality types”; Block,1971). A well-known conceptualization of childhoodpersonality types is based on the theory of ego-control and ego-resilience (Block & Block, 1980).Ego-resilience refers to the tendency to respondflexibly rather than rigidly to changing situationaldemands, particularly stressful situations. Ego-control refers to the tendency to contain (overcontrol)versus express (undercontrol) impulses. Moderateego-control is related to high resiliency, and highand low ego-control are both related to lowego-resiliency (Robins, John, Caspi, Moffitt, &Stouthamer-Loeber, 1996). Combining ego-controland ego-resiliency results in three personality types:Resilients, who are well able to adjust their impulsecontrol to environmental demands (high ego-resil-iency, moderate ego-control); Undercontrollers, whohave low levels of impulse control (low ego-resil-iency, low ego-control), and Overcontrollers, whohave high impulse control but low ability to adjusttheir impulse control to environmental demands

Correspondence concerning this article should be addressed toAmaranta D. de Haan, Department of Developmental, Personal-ity, and Social Psychology (PP07), Ghent University, HenriDunantlaan 2, B-9000 Ghent, Belgium. Electronic mail may besent to [email protected].

© 2013 The AuthorsChild Development © 2013 Society for Research in Child Development, Inc.All rights reserved. 0009-3920/2013/xxxx-xxxxDOI: 10.1111/cdev.12092

Child Development, xxxx 2013, Volume 00, Number 0, Pages 1–16

(low ego-resiliency, high ego-control). Althoughexamining personality types leads to a loss of infor-mation on interindividual differences (Asendorpf,2003), this approach is descriptively efficient andconceptually clear, and information about individu-als’ personality structures is (at least in part) pre-served (Donnellan & Robins, 2011; Robins & Tracy,2003).

In the variable-centered approach, the focus lieson relations among variables in a population, andreplicable personality dimensions across individualsare distinguished (Asendorpf & Denissen, 2006). Aninfluential variable-centered model of personality isthe Five Factor Model, or Big Five (e.g., Caspi &Shiner, 2006; Mervielde & De Fruyt, 1999, 2002).The Big Five describes personality dimensionsbased on five broad content domains: Extraversiondescribes sociability and agency; Agreeableness(labeled Benevolence in children) describes anempathic consideration of other people’s needs andemotions; Conscientiousness entails the ability tocontrol impulses and carry out plans and tasks;Emotional Stability (vs. Neuroticism) describes self-reliance and emotional balance, and Openness toexperience (labeled Imagination for children)entails an interest and willingness to try or con-sider new activities, ideas, and beliefs (Caspi &Shiner, 2006; Mervielde & De Fruyt, 1999). Thesefive dimensions are derived from lexical studies ofthe personality-descriptive language of adults andfrom the analysis of free parental descriptions ofchild personality. Whereas the variable-centeredapproach (i.e., Big Five dimensions) provides noknowledge regarding individuals’ personalitystructures, it has the advantage of organizing find-ings on personality development into a manage-able number of conceptually different domains(Donnellan & Robins, 2011).

Moreover, the three personality types of Resi-lients, Undercontrollers, and Overcontrollers havebeen found to characterize specific constellations ofthe personality dimensions (Denissen, Asendorpf, &van Aken, 2008; Germeijs & Verschueren, 2011;Robins et al., 1996). Typically, Resilients have thebest adjusted personality profile (highest scores onExtraversion, Conscientiousness, Emotional Stabilityand Openness to Experience), Overcontrollers scorelowest on Extraversion and Emotional Stability, andUndercontrollers score lowest on Agreeablenessand Conscientiousness. Personality dimensions andpersonality types have been found to account formuch of the same variance in life outcomes (e.g.,Chapman & Goldberg, 2011), and should thus beconsidered complementary. The current study

examines whether configurations of Big Fivepersonality dimensions in childhood and adoles-cence can be meaningfully characterized by person-ality types.

Personality Development From Childhood toAdolescence

Previous research provides valuable knowledgeregarding how the Big Five dimensions, measuredat a particular moment in time, are represented bythe three personality types of Resilients, Overcon-trollers, and Undercontrollers. However, if changesin personality dimensions are not taken intoaccount, it may appear that specific personalitytypes are more prevalent in some developmentalphases than others or that individuals transitionbetween personality types, whereas such findingsmay reflect normative development in the dimen-sions (cf. Klimstra, Hale, Raaijmakers, Branje, &Meeus, 2010). To our knowledge, only two studies(using partly the same sample) examined whetherchanges in personality dimensions during adoles-cence (12–16 years; Klimstra et al., 2010) and earlyadulthood (12–20 years; Branje, Hale, Frijns, &Meeus, 2010) are represented by the three types.Both studies found the three personality types ofResilients, Undercontrollers, and Overcontrollers.Undercontrollers and Resilients showed stability inExtraversion and became more agreeable and opento experience, whereas Overcontrollers became lessextraverted and remained stable in Openness. Over-controllers became more agreeable between 12 and16 years (Klimstra et al., 2010), but remained stablein Agreeableness between 12 and 20 years (Branjeet al., 2010). All three types showed mean-level sta-bility in Conscientiousness and Emotional Stability.These results not only show that personality dimen-sions change over time but also that personalitytypes may show different mean-level changes in thepersonality dimensions. However, empirical workshould examine personality development in othersamples.

Moreover, research should investigate whetherdifferent types emerge at different stages of life(Donnellan & Robins, 2011). The inconsistenciesbetween the two mentioned studies (Branje et al.,2010; Klimstra et al., 2010) regarding Overcontrol-lers’ development in Agreeableness indicate thatchange in personality dimensions may be differentwhen examining development in one life phase(adolescence) compared to different life phases(from adolescence to early adulthood). Some empir-ical research indicates that development of person-

2 de Haan, Dekovi�c, van den Akker, Stoltz, and Prinzie

ality dimensions differ between developmentalphases. Research that examined personality devel-opment in adolescence from a variable-centeredapproach found that children’s self-reported Agree-ableness and Emotional Stability increased betweenages 11 and 17 years (Branje et al., 2007) andbetween ages 12 and 20 years (Klimstra et al.,2009). Both studies found that children’s self-reported Openness increased from ages 11 to20 years (Branje et al., 2007; Klimstra et al., 2009).However, a large cross-sectional study found thatself-reported Extraversion, Agreeableness, Conscien-tiousness, Emotional Stability, and Opennessshowed mean-level decreased between ages 10 and15 years (Soto, John, Gosling, & Potter, 2011). Onestudy, which examined personality development ina subset of the current sample using growth model-ing, similarly found that combined scores ofmother, father, and teacher ratings of child person-ality characteristics decreased between childhood(mean age = 8 years 10 months) and early adoles-cence (mean age = 13 years 9 months; van denAkker, Dekovi�c, & Prinzie, 2010). Possibly, the stres-ses of facing the changes that accompany the transi-tion from childhood to adolescence are reflected inmean-level decreases in personality characteristics(Galambos & Costigan, 2003), and this may be par-ticularly true for children with less adaptive person-ality structures (Overcontrollers, Undercontrollers).

The current study examines whether changes inBig Five personality dimensions from childhoodinto adolescence are represented by the three devel-opmental personality types. Because a cohort-sequential design was employed, with four cohortsthat were assessed three times across 5 years,personality development could be examined forages 9–17 years.

Developmental Personality Types and AdjustmentProblems

However, the classification of children intodevelopmental personality types gains substantiveusefulness only if the types are related to develop-mental outcomes (Branje et al., 2007; De Fruytet al., 2006; Klimstra et al., 2010; van den Akkeret al., 2010). Cross-sectional evidence typically findsthat Resilients display little problem behavior,Undercontrollers show higher externalizing (aggres-sive, rule-breaking) behavior and Overcontrollershave elevated levels of internalizing (anxious/depressed, withdrawn) problems (Akse, Hale,Engels, Raaijmakers, & Meeus, 2004; Robins et al.,1996). Longitudinal research has found that child

personality types, assessed at one time point,longitudinally predicted aggression and shyness(Denissen et al., 2008). The few studies that exam-ined relations between developmental personalitytypes and adjustment found that Undercontrollersshowed higher delinquency, and Overcontrollersreported higher depressive symptoms (Branjeet al., 2010; Klimstra et al., 2010). There is thusaccumulating empirical evidence for the predictivepower, both concurrent and longitudinal, of(developmental) personality types for adjustmentproblems.

The current study examines differences betweendevelopmental personality types in two types ofexternalizing problems, aggressive and rule-breakingbehaviors, and two types of internalizing problems,anxious/depressed and withdrawn symptoms. Assuch, it can be examined whether children, whosepersonality development is characterized as Resil-ient, Undercontrolling, or Overcontrolling, differ intheir adjustment problems.

Adjustment problems were rated by adolescents,mothers, fathers, and teachers. Adolescent self-reports are likely more valid than parental reports,because, as adolescents grow older and spendmore time unsupervised by parents, parents areincreasingly less knowledgeable of the behaviors,thoughts, and feelings of their children (see e.g.,Smetana et al., 2006). However, using adolescentself-reports for the assessment of both personalityand adjustment problems introduces shared methodvariance. Therefore, mother and father reports and,additionally, teacher reports were included. Theinformation obtained from several informants wascombined into a single factor score, yielding a reli-able and robust measure of adolescent adjustmentproblems across settings.

Parenting and Developmental Personality Types

If the developmental personality types differ inadjustment problems in adolescence, the questionarises whether the types are linked to contextualfactors in childhood. Contextual factors that aremost proximal to children, such as parenting behav-iors, are likely to be associated with developmentalpersonality type membership. In this study, twodistinctive parenting behaviors are examined. First,warmth comprises the extent to which parentsshow affection and are involved in their children’slives (Robinson, Mandleco, Olsen, & Hart, 1995)and is considered indispensable for the formationof secure attachments (Rothbaum & Weisz, 1994).Second, overreactive discipline is characterized by

Correlates of Developmental Personality Types 3

harsh, intrusive, and erratic responses to problem-atic child behavior (Arnold, O’Leary, Wolff, &Acker, 1993; Prinzie, Onghena, & Hellinckx, 2007)and thus describes an inadequate strategy for theenforcing of rule and standards.

Studies that examined longitudinal relationsbetween parenting and personality types found thatindividuals who were classified as Resilient inadulthood (age 27 years; Pulkkinen, 1996) oradolescence (ages 14–19 years; Dubas, Gerris,Janssens, & Vermulst, 2002) had received the mostsupportive, warm, and child-centered parenting inadolescence (age 14 years; Pulkkinen, 1996) andchildhood (age 9 years; Dubas et al., 2002). Under-controllers, in contrast, had been exposed to themost rigid, impulsive, and unjust parenting inadolescence and childhood, respectively. We knowof no research that examined associations betweenparenting and developmental personality types.However, one study that examined relationsbetween overreactive discipline and the develop-ment of children’s personality dimensions usinggrowth modeling (using a subset of the currentsample; van den Akker et al., 2010) found thatoverreactive discipline in childhood (mean age =8 years 10 months) was concurrently associatedwith lower Benevolence, Conscientiousness, andEmotional Stability and, somewhat surprisingly,with increases in Benevolence and Emotional Stabil-ity. Possibly, Undercontrollers, who initially scorelow on Benevolence and Conscientiousness, andOvercontrollers, who initially score low on Emo-tional Stability, receive more overreactive disciplinethan Resilients in childhood; Undercontrollers andOvercontrollers may also increase in Benevolence(Undercontrollers) and in Emotional Stability (Over-controllers) over time. Thus, a type approach topersonality development may provide a less ambig-uous picture of the parenting correlates of personal-ity development. The current study thereforeexamines relations between warmth and overreac-tive discipline in childhood and developmentalpersonality types.

Aims and Hypotheses

The current study aims to extend the knowledgebase on personality development, by examiningwhether changes in all Big Five personalitydimensions from childhood to adolescence can bemeaningfully represented by developmental person-ality types. Based on the view that the self hasgreater access to self-relevant information thanothers do (John & Robins, 1993), adolescent

self-reported Big Five questionnaire data were used.Given that studies have found several gender differ-ences in the prevalence of personality types, withboys more likely being classified as Undercontrol-ling, and girls as Overcontrolling and Resilient(Dubas et al., 2002; Klimstra et al., 2010), gender dif-ferences in personality type membership are exam-ined. Moreover, usefulness of the developmentalpersonality types is investigated by exploring differ-ences between the types’ externalizing (aggressive,rule-breaking) and internalizing (anxious/depressed, withdrawn) problems in adolescence.Further, associations between overreactive andwarm parenting in childhood and personality typemembership are investigated. Based on previousresearch (Branje et al., 2007; Klimstra et al., 2010;van den Akker et al., 2010), we expect that theinitial levels of, and changes in personality charac-teristics will be adequately represented by (at least)three developmental personality types (Resilients,Overcontrollers, and Undercontrollers). Boys areexpected more likely to be Undercontrolling, andgirls are more likely to be Overcontrolling andResilient (Branje et al., 2010; Dubas et al., 2002;Klimstra et al., 2010). Further, it is expectedthat Undercontrollers score highest on externalizingbehaviors and that Overcontrollers score higheston internalizing behaviors (Akse et al., 2004;Denissen et al., 2008; Klimstra et al., 2010; Robinset al., 1996). Additionally, based on empirical evi-dence (Block, 1971; Dubas et al., 2002; Pulkkinen,1996), parents of Resilients are expected to scorehighest on warmth, and parents of Undercontrol-lers are expected to score highest on overreactivediscipline.

Method

Participants and Procedure

This study is part of the longitudinal FlemishStudy on Parenting, Personality, and Development(see Prinzie et al., 2003). At Time 1 (2004), 523adolescents participated (261 boys, 262 girls). Of483 families (92%) both parents participated, of 36families only the mother (N = 518), and of 5 fami-lies only the father (N = 487) filled out the ques-tionnaires. At Time 1, mothers’ mean agewas 39 years 7 months (SD = 3.56, range = 30–55 years), fathers’ mean age was 41 years 6 months(SD = 4.25, range = 31–63 years). Mothers’ (M) andfathers’ (F) educational levels were as follows: ele-mentary school, M = 1%, F = 3%; secondary educa-tion, M = 41%, F = 43%; nonuniversity higher

4 de Haan, Dekovi�c, van den Akker, Stoltz, and Prinzie

education M = 45%, F = 34%; university M = 13%,F = 19%.

We used a cohort-sequential design and com-bined data from four cohorts who were aged 9, 10,11, and 12 years old at Time 1 (M = 10 years6 months, SD = 1.10) to approximate developmentaltrajectories for each of the Big Five personalitydimensions between ages 9 and 17 years. InTable 1, numbers of adolescents that were assessedat specific ages across the measurement waves areshown. At Time 2, 3 years later (2007), 459 familiesparticipated (444 adolescents, 449 mothers, 410fathers). At Time 3, 2 years later than Time 2(2009), 427 families participated (420 adolescents,418 mothers, 380 fathers). At Time 3, 290 teachersreported on children’s problem behaviors. Becausethe current study was designed to examinelong-term development, the spread of measurementwaves (3 and 2 years apart) were timed such thatthere were three overlapping data points betweencohorts, and to ensure that the youngest cohort wasolder at the third measurement wave than the oldestcohort was at the first measurement wave. As such,a proper balance was achieved between the twoprinciples that should be considered when design-ing a cohort-sequential study: efficiency of data col-lection and sufficiency of overlap (Anderson, 1995).

There were two forms of missingness in the data.First, there was planned missingness due to thecohort-sequential design. However, this form ofmissingness is not dependent on personality charac-teristics, so the assumption that the data weremissing at random was tenable. The second form ofmissingness in our data is due to attrition.Regarding attrition, missing values were found tobe completely at random, Little’s MCAR, v2(1,104) =1,141, p = .21. These results indicate a good fitbetween sample scores with and without imputation.Therefore, to maximize sample size at Time 2and Time 3, missing values were imputed usingthe expected-maximization algorithm (Schafer &Graham, 2002). As such, data from all participants

for whom valid data was available at Time 1 couldbe included in the analyses.

Measures

Child personality. Adolescents rated their ownpersonality characteristics at each measurementwave, using the Hierarchical Personality Inventoryfor Children (HiPIC; Mervielde & De Fruyt, 1999,2002). The HiPIC is a comprehensive personalityinventory that assesses individual differencesbetween children in 144 items. Findings concerningstructural replicability, and convergent, discrimi-nant and construct validity are reported elsewhere(Mervielde & De Fruyt, 1999; Prinzie & Dekovi�c,2008). The items are on a 5-point scale, rangingfrom 1 = almost not characteristic to 5 = very charac-teristic. Number of items for each dimension are:Extraversion (32), Benevolence (40), Conscientious-ness (32), Emotional Stability (16), and Imagination(24). Across dimensions and measurement waves,Cronbach’s alphas ranged between .82 and .92.

Adjustment problems. Externalizing and internal-izing symptoms were assessed by adolescents,mothers, fathers, and teachers at Time 3. Adoles-cents completed the Youth Self-Report (YSR; Achen-bach, 1991a), mothers and fathers filled out theChild Behavior Checklist (CBCL; Achenbach,1991b), and teachers completed the Teacher ReportForm (TRF; Achenbach, 1991c). The CBCL andaccompanying YSR and TRF are extensively vali-dated instruments with adequate reliability andvalidity (Achenbach, 1981a, 1981b, 1981c). Acrossmeasures, each item is rated as 0 = not true,1 = somewhat/sometimes true, or 2 = very/often true.To reduce rater bias, factor scores were computedusing exploratory factor analysis with the fourinformants’ scale scores as the indicators. Factorloadings and number of items (in parentheses) wereas follows: � .48 for aggression (73), � .57 forrule-breaking (61), � .38 for anxious/depressed(55), and � .56 for withdrawn problems (34).

Table 1Number of Children Assessed at Each Age

N per cohort

Age (years)Total Nper wave9 10 11 12 13 14 15 16 17

Time 1 (2004) 124 139 134 132 529Time 2 (2007) 101 116 117 110 444Time 3 (2009) 99 110 105 106 420Total N per age group 124 139 134 233 116 216 220 105 106

Correlates of Developmental Personality Types 5

Cronbach’s alphas for the composite scores rangedbetween .85 and .92.

Parenting behavior. Mothers and fathers ratedtheir own overreactive discipline at Time 1 with theParenting Scale (Arnold et al., 1993; Prinzie et al.,2007). The scale measuring overreactive disciplineconsists of nine items and measures parents’tendency to respond with anger, frustration, mean-ness and irritation, impatiently, and aversively toproblematic behavior of their children. Items pres-ent discipline encounters followed by two optionsthat act as opposite anchor points for a 7-pointscale where 1 indicates a high probability of usingan effective discipline strategy (e.g., “When mychild misbehaves…” “I speak to him calmly”) and7 indicates a high probability of making a disciplinemistake (“I raise my voice or yell”). The instrumenthas adequate test–retest reliability, distinguishesclinical from nonclinical samples, and has been vali-dated against behavioral observations of parenting(Arnold et al., 1993). In the current study, Cron-bach’s alphas were .77 for mothers’ and .76 forfathers’ ratings.

Mothers and fathers assessed their own warmthat Time 1 with the Parenting Practices Question-naire (Robinson et al., 1995). This scale comprises11 items and assesses the extent to which parentsexhibit warm parenting and are involved in theirchildren’s lives (e.g., “I show empathy when mychild feels hurt or frustrated”). Items were rated ona 5-point scale ranging from 1 = never to 5 = always.In this study, Cronbach’s alphas for mothers’ andfathers’ ratings of warmth were .82 and .86, respec-tively.

Plan of Analyses

First, means, standard deviations, and intercorrela-tions between the measures are presented. Then, toexamine whether there were potential subgroupswithin our sample and to determine what kind ofgrowth (linear, quadratic) described our data best,we assessed univariate latent growth models(LGMs; e.g., Duncan, Duncan, & Strycker, 2006),using Mplus (Muth�en & Muth�en, 1998–2007). InLGM, it is assumed that growth trajectories arisefrom a single distribution, and individual differ-ences are captured as random effects (variances)around common latent growth factors (intercept,slope). A latent intercept factor and latent linearslope factor (and, where supported, a latent qua-dratic slope factor) are indicated by the observedpersonality scores at the nine age points. The factorloadings of all measured variables on the intercept

factor were specified to be 1. The factor loadings onthe slope factor carry the information of the under-lying time metric. A trajectory corresponding tolinear change was evaluated by fitting a model thatspecified the slope factor loadings to be 0, 1, 2, 3, 4,5, 6, 7, and 8, for ages 9, 10, 11, 12, 13, 14, 15, 16,and 17, respectively. A trajectory corresponding toquadratic change was evaluated by fitting a modelthat specified the slope factor loadings to be 0, 1, 4,9, 25, 36, 49, and 64 for ages 9–17 years, respectively.The LGMs also allowed for testing cohort effects foreach personality dimension separately. For this, wecompared a LGM in which the growth parameters(latent intercept and slope means and variances, andintercept–slope covariance) were estimated freelyacross cohorts with a model in which these growthparameters were constrained to be equal acrosscohorts. A nonsignificant increase in chi-squarebetween these two nested models indicates that theunderlying growth parameters are not significantlydifferent across the four cohorts (Duncan et al.,2006).

Subsequently, group-based latent class growthmodeling (LCGM) was employed. LCGMs areaimed at finding the smallest number of classescapturing the most variance among individualswith regard to initial levels and change on variablesincluded in the model. In LCGMs, variances of theestimated means of intercepts and linear as well as,where applicable, quadratic growth factors are con-strained to zero. When compared to related proce-dures in which variances of the growth factors arefreely estimated (e.g., growth mixture modeling[GMM]; Muth�en & Muth�en, 2000), LCGMs result ina somewhat less optimal model fit, but producemore clearly distinguishable classes (Nagin, 2005).GMMs were further too complex to estimate, giventhe large amount of planned missingness that isinherent in a cohort-sequential design. All modelswere estimated with multiple random sets of start-ing values to prevent convergence on local maxima.

We used several criteria, outlined by Muth�enand Muth�en (2000), to determine the number oflatent classes (i.e., developmental personality types)that best characterized our data. First, we used thesample size adjusted Bayesian information criterion(SABIC) and the Lo–Mendell–Rubin likelihood ratiotest (LMR-LRT; Lo, Mendell, & Rubin, 2001). Theoptimal model has the lowest SABIC, and a signifi-cant LMR-LRT indicates that the model with k clas-ses has a better fit than the model with k�1 classes.The second consideration regards classificationquality, which was determined by examining theentropy and the posterior probabilities. Entropy

6 de Haan, Dekovi�c, van den Akker, Stoltz, and Prinzie

indexes classification accuracy by averaging theposterior probabilities after individuals have beenassigned to their most likely class, with valuescloser to 1 indexing greater precision. The averageposterior probability for each class for individualswhose highest probability is for that class shouldexceed a threshold of 0.7. Third, theoretical mean-ingfulness of classes in the various solutions wasconsidered. If a class found in a solution withk classes was found to be a slight variation of aclass already found in the solution with k�1 classes,the more parsimonious solution was chosen(Muth�en & Muth�en, 2000). Gender differences inthe prevalence of developmental personality typeswere explored using Pearson’s chi-square test.

Substantive usefulness of the classes was exploredby examining differences between types in levels ofadjustment problems in adolescence, using a multi-variate general mixture model, with personalitytype as the factor and child gender and age as co-variates. Further, personality type membership wasregressed on levels of overreactive and warm par-enting in childhood, using multinomial logisticregression in Mplus 5.0 (Muth�en & Muth�en, 1998–2007). The Mplus multinomial regression assignseach individual fractionally to all classes using theposterior probabilities. Additionally, we examinedwhether the logistic regression coefficients between

overreactive/warm parenting and personality typemembership differed for the parenting of mothersversus fathers by comparing the log likelihoods of amodel in which the paths from mothers’ versusfathers’ parenting to personality type membershipand adjustment problems were freely estimated,with a model in which these paths were con-strained to be equal for mothers versus fathers.

Results

Descriptive Statistics

Descriptive statistics (Table 2) indicate that gen-erally, mean levels of the personality dimensionsdecreased between the first and second measure-ment waves, with the exception of very minor(.01) increases of Emotional Stability in Cohorts 1and 3. Between the second and third measurementwaves, Extraversion continued to decrease inCohorts 1 and 3 (although for Cohort 1, thischange was minimal, .01), and showed very minorincreases (.01/.02) in Cohorts 2 and 4. Benevolenceand Conscientiousness showed continued mean-level decreases in all cohorts except Cohort 4(mean-level increase). Emotional Stability contin-ued to decrease in all cohorts except Cohort 1(mean-level increase). Imagination decreased in

Table 2Mean Levels of the Big Five Dimensions at Each Measurement, for Each Cohort Separately

Cohort

Age 9 years Age 10 years Age 11 years Age 12 years

M SD M SD M SD M SD

ExtraversionTime 1 3.62 0.47 3.53 0.44 3.57 0.41 3.52 0.52Time 2 3.53 0.40 3.43 0.48 3.50 0.41 3.41 0.56Time 3 3.52 0.50 3.44 0.56 3.42 0.52 3.43 0.52

BenevolenceTime 1 3.52 0.44 3.52 0.42 3.49 0.36 3.46 0.40Time 2 3.46 0.39 3.47 0.42 3.45 0.35 3.42 0.36Time 3 3.42 0.36 3.43 0.43 3.43 0.36 3.45 0.33

ConscientiousnessTime 1 3.37 0.45 3.33 0.44 3.37 0.43 3.20 0.44Time 2 3.27 0.42 3.28 0.46 3.20 0.45 3.08 0.53Time 3 3.24 0.44 3.21 0.52 3.16 0.47 3.19 0.48

Emotional StabilityTime 1 3.47 0.67 3.56 0.59 3.54 0.59 3.50 0.60Time 2 3.48 0.61 3.47 0.71 3.55 0.56 3.43 0.67Time 3 3.50 0.64 3.43 0.71 3.45 0.63 3.34 0.67

ImaginationTime 1 3.75 0.52 3.66 0.46 3.69 0.54 3.49 0.52Time 2 3.59 0.47 3.52 0.48 3.49 0.51 3.34 0.55Time 3 3.51 0.48 3.45 0.48 3.48 0.44 3.39 0.49

Correlates of Developmental Personality Types 7

Cohorts 1, 2, and 3 (although for Cohort 3, thischange was only .01) and increased in Cohort 4.However, given the very small sizes of changes inseveral instances, the extent to which the dimen-sions showed change over time should be deter-mined with latent growth modeling, in whichchange is modeled while taking into account initiallevels of the dimensions as well as interindividualdifferences in rates of change.

Between-time correlations for the personalitydimensions ranged from .51 to .67 for adjacentwaves (T1-T2, T2-T3), and from .38 to .46between T1-T3. Adolescent personality thusshowed moderate to large stability for each suc-cessive time point and small to moderate stabilitybetween first and third measurement moments.Table 3 shows the between-construct associationsfor the full sample (without distinction betweencohorts). Benevolence and Conscientiousness wererelated to lower externalizing (aggressive, rulebreaking) problems. Extraversion, Benevolence,Emotional Stability, and Imagination were allrelated to lower withdrawn problems and Emo-tional Stability was also related to lower anxious/depressed problems. Mothers’ and fathers’warmth were concurrently related to higher scoreson each of the Big Five. Mothers’ and fathers’overreactivity were related to lower Benevolence.Mothers’, but not fathers’, overreactivity wasrelated to lower Conscientiousness and Imagina-

tion and fathers’, but not mothers’, overreactivitywas related to lower Emotional Stability. Mothers’and fathers’ warmth were related to lower rule-breaking behaviors. Mothers’ and fathers’ overreac-tive discipline were related to more externalizingproblems, and mothers’, but not fathers’, over-reactive discipline was also related to moreinternalizing problems.

Shape of Growth of Personality Characteristics

Table 4 presents the model fit indices of theLGMs representing no-growth (intercept-only),linear change, and quadratic change. Results showthat for Extraversion and Emotional Stability linearmodels fit the data best, whereas for Benevolence,Conscientiousness, and Imagination models thatincluded linear and quadratic slopes fit the databest. Moreover, growth parameters of the LGMssuggested that there was significant variancearound intercepts, linear slopes, and (where appli-cable) quadratic slopes of the personality dimen-sions. Together, these results indicate that differentclasses may be identified that follow specificdevelopmental trajectories in each of the Big Fivedimensions.

Further, we tested whether the different cohortscould approximate one longitudinal curve. Compari-son of the unconstrained and constrained linearLGMs showed that cohorts did not differ significantly

Table 3Bivariate Intercorrelations Between the Measures, Across Cohorts

Time 1 (2004) Time 3 (2009)

1 2 3 4 5 6 7 8 9 10 11 12 13

Time 1 (2004)1. Overreact mother2. Overreact father .283. Warmth mother �.26 �.104. Warmth father �.12 �.12 .365. Extraversion �.02 �.05 .20 .136. Benevolence �.26 �.15 .20 .17 .167. Conscientiousness �.10 �.04 .11 .13 .19 .528. Emo. Stability �.06 �.09 .08 .10 .32 .25 .309. Imagination �.09 �.04 .21 .14 .48 .24 .46 .32

Time 3 (2009)10. Aggression .26 .20 �.07 �.10 .11 �.25 �.12 �.02 .0511. Rule-breaking .21 .17 �.14 �.13 .06 �.23 �.24 .03 .02 .6812. Anxiousness .14 .05 .06 �.05 �.07 .03 �.07 �.33 �.07 .25 .0113. Withdrawn .12 .01 �.11 �.09 �.26 �.11 �.10 �.22 �.16 .14 .16 .64

Note. Intercorrelations in bold are significant (p < .05).

8 de Haan, Dekovi�c, van den Akker, Stoltz, and Prinzie

on the latent growth factors, Extraversion, Dv2(15)= 24.95, p = .05; Benevolence, Dv2(15) = 16.53, p =.35; Conscientiousness, Dv2(15) = 24.41, p = .06; Emo-tional Stability, Dv2(15) = 14.17, p = .51; Imagination,Dv2(15) = 24.30, p = .06. The cohorts could thus becombined to approximate a longitudinal curve forages 9–17 years.

Developmental Personality Types

Based on the univariate LGMs, we specified amultivariate LCGM that included linear slopes forExtraversion and Emotional Stability, and linear aswell as quadratic slopes for Benevolence, Conscien-tiousness, and Imagination. A three-class solutionappeared as the optimal solution, based on infor-mation criteria, classification of individuals intoclasses, and the LMR-LRT (Table 5). Although theSABIC indicated that a four-class solution was alsoreasonable, inspection of the growth parametersindicated that the fourth class was merely a slightvariation of another class (Resilients). The posteriorprobabilities ranged between .84 and .93 and thealternative probabilities of group membershipranged from .001 to .126. This indicates that typemembership was clear for most individuals in thesample. Entropy was very high, .91, yielding fur-ther confidence in correct classification of the per-sonality types.

The personality profiles of the latent classesresembled the developmental personality types ofResilients, Undercontrollers, and Overcontrollers.Number of adolescents in, and growth parametersof the types are displayed in Table 6 and are plot-ted in Figure 1. Confidence intervals revealed thatResilients had the highest initial levels on all BigFive dimensions. Overcontrollers had the lowest

initial levels on Extraversion, Emotional Stability,and Imagination. Undercontrollers had the lowestinitial levels on Benevolence and Conscientious-ness, but these intercept scores were not signifi-cantly different from Overcontrollers’ interceptscores.

In Figure 1, a graphical presentation of thedevelopment of the Big Five dimensions in each ofthe types is plotted. It can be seen that all typesdecreased in Extraversion (linear change). Only Res-ilients decreased in Benevolence and in EmotionalStability (linear change), the other types showedstability in these dimensions. Resilients andOvercontrollers decreased in Conscientiousness untilaround age 14, after which the decrease flattened

Table 5Model Fit Indices of the Latent Class Growth Models

1class

2classes

3classes

4classes

5classes

Loglikelihood

5,556 5,068 4,941 4,838 4,765

AIC 11,174 10,226 9,999 9,823 9,703SABIC 11,208 10,275 10,063 9,902 9,797Entropy a .94 .90 .89 .89Smallestclass%

a 45 19 11 10

LMR-LRT b a 945*** 248*** 191 151

Note. Model fit indices in bold indicate best fitting models.AIC = Aikaike’s information criterion; SABIC = sample-size-adjusted Bayesian information criterion; LMR-LRT = Lo–Mendell–Rubin likelihood-ratio-test.aCannot be estimated for one-class models. bLMR-LRTs cannotbe computed in a cohort-sequential (multigroup) design. Restruc-turing the data set, such that variables represented ages at timeof measurement instead of measurement waves, allowed for esti-mation of the LMR-LRTs.***p < .001.

Table 4Model Fit Indices of the Cohort-Sequential Univariate Latent Growth Models: Assessing Shape of Growth

Model fit indices Model comparison

No-growth model(intercept only)

Linear growthmodel

Quadratic growthmodel

Linear vs.no-growth

Quadratic vs.linear

v2 df p v2 df p v2 df p Dv2 df p Dv2 df p

Extraversion 76.73 22 < .001 28.81 19 .07 21.91 15 .11 47.92 3 < .001 6.90 4 .14Benevolence 47.57 22 .001 19.35 19 .43 7.00a 15 .96 28.22 3 < .001 12.35 4 .01Conscientiousness 119.39 22 < .001 45.19 19 .001 34.69a 15 .03 74.20 3 < .001 10.50 4 .03Emotional Stability 63.14 22 < .001 15.78 19 .67 9.86 15 .83 47.36 3 < .001 5.92 4 .21Imagination 134.92 22 < .001 52.91 19 < .001 25.32 15 .05 82.01 3 < .001 27.59 4 < .001

Note. Fit indices in bold represent the best fitting models.aAlthough latent growth models including quadratic slopes fitted the data significantly better than linear growth models, the quadraticslope means were not significant for the models describing the development of Benevolence and Conscientiousness.

Correlates of Developmental Personality Types 9

out and, for Overcontrollers, Conscientiousnessincreased. Undercontrollers showed a linear declinein Conscientiousness. Regarding Imagination, Un-dercontrollers and Overcontrollers decreased untilaround age 14, after which this decrease flattenedout. Resilients showed a linear decrease in Imagina-tion. Undercontrollers decreased more in Imagina-tion than Resilients, and Resilients decreased morein Benevolence than Undercontrollers.

Gender differences in the prevalence of thepersonality types were found, v2(2) = 7.90, p < .01.Boys were somewhat more likely to be classified asUndercontrolling (45.6% vs. 35.5%) and girls weresomewhat more likely to be Resilient (46.2% vs.34.5%). Boys and girls were similarly likely to beclassified as Overcontrolling (19.9% and 18.3%).

Given the possibility that the classification ofchildren was based more on initial levels of person-ality characteristics than on changes over time, weadditionally examined whether similar types werefound if the intercept was modeled at age 17instead of age 9. Results of these analyses, whichcan be obtained from the corresponding author,indicated that only two children were classified dif-ferently if the intercept was modeled at age 17.Thus, although the types differed more on levelsthan on changes, including changes over timeyielded robust classification of children into devel-opmental personality types.

Correlates of Developmental Personality Types

Differences in average levels of adjustment prob-lems between developmental personality typeswere next examined, controlling for child genderand age. Differences in levels of adjustment prob-lems across types were all in the expected direction(Table 7). Undercontrollers scored significantlyhigher on aggressive and rule-breaking (externaliz-ing) problems than the other two types, but Over-controllers did not show different levels ofexternalizing problems than Resilients. Overcontrol-lers scored highest on anxious/depressed andwithdrawn (internalizing) problems. Undercontrol-lers scored lower than Overcontrollers on anxious/depressed problems, but higher than Resilients.Resilients and Undercontrollers did not differ onlevels of withdrawn problems.

We then examined whether the parenting ofmothers versus fathers was similarly related to per-sonality type membership. Comparison of the loglikelihoods of the baseline and the constrainedmodels suggest that mothers’ and fathers’ parentingwere similarly (i.e., not statistically different) relatedto developmental personality type membership,v2(4) = 0.80, p = .94. Higher levels of warmthdecreased the probability that children were classi-fied as Undercontrolling (b = �0.52, SE = 0.26,p < .05; OR = 0.57) or Overcontrolling (b = �0.59,

Table 6Growth Parameters and Confidence Intervals Around Growth Parameters of the Big Five Dimensions in Each Latent Trajectory Class

Resilients (n = 211) Undercontrollers (n = 212) Overcontrollers (n = 100)

M 95% CI M 95% CI M 95% CI

InterceptsExtraversion 3.77O (3.70, 3.84) 3.63O (3.50, 3.75) 3.14RU (3.02, 3.27)Benevolence 3.77UO (3.69, 3.85) 3.31R (3.18, 3.45) 3.41R (3.23, 3.58)Conscientiousness 3.68UO (3.59, 3.76) 3.20R (3.11, 3.28) 3.23R (3.04, 3.42)Emotional Stability 3.82UO (3.73, 3.90) 3.51RO (3.40, 3.62) 3.04RU (2.71, 3.37)Imagination 4.01UO (3.93, 4.09) 3.69RO (3.55, 3.83) 3.38RU (3.23, 3.52)Linear SlopesExtraversion �0.02* (�0.03, �0.002) �0.02** (�0.04, �0.01) �0.04** (�0.06, �0.01)Benevolence �0.03***U (�0.04, �0.02) 0.004R (�0.01, 0.02) �0.01 (�0.02, 0.01)Conscientiousness �0.05** (�0.08, �0.02) �0.04*** (�0.06, �0.03) �0.08* (�0.15, �0.004)Emotional Stability �0.03* (�0.04, �0.01) 0.00 (�0.02, 0.03) �0.02 (�0.06, 0.01)Imagination �0.04***U (�0.06, �0.02) �0.10***R (�0.16, �0.06) �0.10** (�0.16, �0.04)Quadratic SlopesConscientiousness 0.004* (0.001, 0.01) a 0.01* (0.001, 0.01)Imagination a 0.01* (0.001, 0.01) 0.01* (0.002, 0.02)

Note. Superscripts indicate that the given group differs significantly from other groups: R = Resilients; U = Undercontrollers; O = Over-controllers.aGrowth parameter was not estimated (i.e., constrained to be zero for the respective class).*p < .05. **p < .01. ***p < .001.

10 de Haan, Dekovi�c, van den Akker, Stoltz, and Prinzie

SE = 0.20, p < .01; OR = 0.55) as compared to Resil-ient, but higher levels of warmth did not distin-guish Undercontrollers from Overcontrollers(b = �0.08, SE = 0.34, ns; OR = 0.93). Overreactivediscipline did not distinguish Resilients fromUndercontrollers (b = 0.10, SE = 0.12, ns; OR = 1.10),from Overcontrollers (b = 0.17, SE = 0.12, ns;OR = 1.21), or Undercontrollers from Overcontrollers(b = 0.08, SE = 0.16, ns; OR = 1.08).

Discussion

Using children’s self-reported Big Five personalitydimensions in a cohort-sequential design, thecurrent study examined whether changes in person-ality dimensions between childhood and adoles-cence can be meaningfully characterized bydevelopmental personality types. Additionally, thedescriptive power and utility of the types wasexamined by exploring differences in multi-infor-mant externalizing (aggressive, rule-breaking) andinternalizing (anxious/depressed, withdrawn) prob-lems at age 17 years (Time 3), and by relating par-ents’ self-reported parenting in childhood (Time 1)to type membership. Further, gender differences inthe prevalence of personality types were explored.

Overall patterns of change indicated thatbetween the first and second measurement waves,the personality dimensions showed mean-leveldecreases in nearly all cohorts, although severalchanges were minor. Our finding that the personal-ity dimensions generally decreased in early adoles-cence are consistent with other cross-sectional (Sotoet al., 2011) and longitudinal (van den Akker et al.,

2.5

2.8

3.1

3.4

3.7

4.0

9 10 11 12 13 14 15 16 17

Ben

evol

ence

Age (Years)

UNCOVCRES

2.5

2.8

3.1

3.4

3.7

4.0

9 10 11 12 13 14 15 16 17

Emot

iona

l Sta

bilit

y

Age (Years)

UNCOVCRES

2.5

2.8

3.1

3.4

3.7

4.0

9 10 11 12 13 14 15 16 17

Extra

vers

ion

Age (Years)

UNCOVCRES

2.5

2.8

3.1

3.4

3.7

4.0

9 10 11 12 13 14 15 16 17

Con

scie

ntio

usne

ss

Age (Years)

UNCOVCRES

2.6

2.9

3.2

3.5

3.8

4.1

9 10 11 12 13 14 15 16 17

Imag

inat

ion

Age (Years)

UNCOVCRES

Figure 1. Graphical presentation of the growth parameters(intercepts, linear slopes, quadratic slopes) of the Big Fivepersonality dimensions in the three latent trajectory classes ofUndercontrollers (UNC), Overcontrollers (OVC), and Resilients(RES).

Table 7Mean Levels of Adjustment Problems Across Trajectory Groups

F(2, 520)

Levels of adjustment problemswithin classes

ResilientsUnder-

controllersOver-

controllers

Aggression 14.51*** �0.24U 0.26RO �0.05U

Rule-breaking 19.55*** �0.27U 0.29RO �0.05U

Anxiousness/depressed

45.57*** �0.32OU �0.07RO 0.81RU

Withdrawn 15.66*** �0.19O �0.06O 0.50RU

Note. Superscripts indicate the given group differs significantlyfrom other groups: R = Resilients; U = Undercontrollers;O = Overcontrollers. Levels of adjustment problems are devia-tions from the group mean factor scores (i.e., zero). Covariateswere gender child (0 = boy, 1 = girl) and age child in years.***p < .001.

Correlates of Developmental Personality Types 11

2010) research that examined personality develop-ment across the transition to adolescence, and offerfurther evidence that the physical, hormonal, andpsychosocial changes during the transition to ado-lescence may result in (temporarily) less maturepersonality profiles. Between the second and thirdmeasurement waves, patterns of changes differedbetween cohorts, with cohorts showing continuedmean-level decreases, stability, or increases in thepersonality dimensions. Given the absence of cohortdifferences in the latent growth modeling growthfactors (intercept and slope means, variances, andintercept–mean covariances), however, conclusionscan be drawn for personality change spanning ages9–17 years (Duncan et al., 2006).

In line with our hypotheses and with past longi-tudinal research (Branje et al., 2010; Klimstra et al.,2010), the personality types of Resilients, Under-controllers, and Overcontrollers were found to rep-resent changes in the personality dimensionsacross childhood and adolescence. In line withother cross-sectional (Dubas et al., 2002) and longi-tudinal (Branje et al., 2010; Klimstra et al., 2010)studies, a substantial minority of children in oursample was classified as Resilient (highest initiallevels on all personality characteristics). Overcon-trollers had lowest initial levels on Extraversionand Emotional Stability (and Imagination), andUndercontrollers had lowest initial levels onBenevolence and Conscientiousness (cf. Akse et al.,2004; Denissen et al., 2008; Robins et al., 1996).However, unlike other cross-sectional (Robinset al., 1996) and longitudinal (Branje et al., 2010;Klimstra et al., 2010) studies, Undercontrollers inour sample did not score significantly lower onBenevolence and Conscientiousness than the othertypes. This difference between our study and otherwork is likely due to relatively small variance inthe personality variables in our sample (Donnellan& Robins, 2011). Moreover, the types differed moreon initial levels than on changes, a finding that ismost likely due to the relative large amounts ofvariance around the intercept, rather than slopegrowth factors. Given the highly consistent classifi-cation of the personality types when the interceptwas modeled at age 17 years, results indicate thatpersonality types are able to meaningfully repre-sent interindividual differences in personalitybetween childhood and adolescence, also whennormative changes in personality are taken intoaccount (cf. Block, 1971; Donnellan & Robins,2011). Overall, this study shows that similar typesemerge at different stages of life (Donnellan &Robins, 2011) and thus that changes are more

likely to occur for the personality dimensionswithin personality types, than for prevalence ratesof personality types.

Regarding changes in personality characteristicsfor the personality types, we found that all typesbecame less extraverted over time. Resilientsbecame less benevolent and emotionally stable,whereas the other types showed stability in thesepersonality characteristics. All three types becameless conscientious, but for Resilients, the decreaseslowed down and for Overcontrollers, this decreaseturned into an increase from around 14 yearsonward. All three types became less imaginative,but for Undercontrollers, this increase slowed downand Overcontrollers started to increase again fromaround age 14 years onward. The findings, that thevariances around the slopes for Conscientiousnessand Imagination in conventional growth analysestranslated into different shapes of growth in the dif-ferent classes (even if the quadratic growth factormean for Conscientiousness was not significant inthe full sample), illustrate the descriptive power oflatent class growth analyses. Overall, results areconsistent with the view that although personalitychange is in the direction toward maturity frommiddle adolescence onward, it may actually be inthe direction of less maturity between childhoodand early adolescence (Soto et al., 2011). It is impor-tant to note that the findings of the growth analyseswere not due to model misspecification, as thedescriptive statistics showed similar patterns ofchange. Together, these results show that the per-sonality types meaningfully represent interindividu-al differences in normative changes in personalitydimensions across life phases.

We found two small gender differences in theprevalence of developmental personality types.Boys were somewhat more likely to be Undercon-trollers, whereas girls were somewhat more likelyto be Resilients. These findings are in line withother cross-sectional (e.g., Dubas et al., 2002) andlongitudinal (Klimstra et al., 2010) work on person-ality types in adolescence, although the genderdifferences we found were less pronounced than inprevious studies. Further, in contrast to past stud-ies, girls in our sample were not more likely to beOvercontrollers than boys. Possibly, this inconsis-tency compared to other work is due to our exami-nation of developmental personality types fromchildhood into adolescence. Thus, the increasedprobability of girls to be Overcontrolling may sur-face in adolescence, but much less so when examin-ing changes in girls’ personality development fromchildhood into adolescence.

12 de Haan, Dekovi�c, van den Akker, Stoltz, and Prinzie

Moreover, the developmental personality typesshowed different levels of externalizing and inter-nalizing problems at Time 3, which were all in theexpected direction. Given that the differencesbetween children that were found from age 9 yearsonward were related to adjustment problems at age17 years, identification of developmental personal-ity types have the potential to aid prevention andtreatment of externalizing and internalizing prob-lems. The current study extends existing knowledgegained from conventional growth modeling, whichshowed that differences between children in per-sonality development were related to adjustmentproblems (van den Akker et al., 2010). Further, ourfinding that Undercontrollers showed highest levelsof externalizing problems, and Overcontrollersscored highest on internalizing problems replicateresults of the two studies that examined differencesbetween developmental personality types in adjust-ment problems (Branje et al., 2010; Klimstra et al.,2010). The current study’s results showed that rela-tions between personality and adjustment problemsare not only concurrent, but also longitudinal.Moreover, differences between children (i.e., vari-ances around growth parameters) can be meaning-fully translated into developmental personalitytypes that can be used as descriptive and, possibly,predictive tools for the explanation of externalizingand internalizing problems across the transitionalperiod between childhood and adolescence.

Developmental personality type membershipwas also associated with warm, but not overreac-tive, parenting at Time 1. Given the different infor-mants that were used for the assessment ofpersonality (adolescents) and parenting (parents),these associations were not due to shared methodvariance. Previous work similarly found consistentassociations between warm or supportive parentingand personality types (Dubas et al., 2002; Pulkkinen,1996). Possibly, our measure of warmth taps intooverall family climate, and parental warmth andsupport can be seen as being related to how parentsthink they should parent their children. Overreac-tive discipline, at the other hand, reflects harsh, butunintentional, responses to problematic childbehaviors. Possibly then, the development ofpersonality throughout childhood and adolescenceis more strongly affected by an overall supportivefamily climate than by (ineffective) disciplinarytechniques. Moreover, personality is substantiallyinfluenced by genetics and behavioral-genetic stud-ies have suggested evidence for the heritability ofsupport (warmth) but not ineffective (overreactive)behavioral control (Kendler & Baker, 2007). The dif-

ferential associations found for overreactivity andwarmth and children’s personality characteristicsmay thus have arisen because warmth is affectedby genetics, as is child personality, whereas overre-activity is less genetically influenced (Kendler &Baker, 2007). Given the design of the current studythe argument of genetic influence is, of course,speculative, and future research could assess thispossibility with a genetically informed design.

Limitations and Future Directions

Several limitations warrant caution in the inter-pretation of these results. First, due to the largesample all measures were assessed by question-naires. Fortunately, because of our multi-informantdesign, we are confident that the results are notdue to rater bias alone. Nonetheless, a multi-method measurement strategy (e.g., inclusion ofobservations) may more accurately assess parent-ing, child personality, and child behaviors. Further,it is difficult to fully capture personality changewith widely spread out assessments (3 and 2 yearsapart), particularly during developmental periodswith multiple changes. Additionally, three measure-ment waves should be considered the minimumrequirement for assessing (nonlinear) change.Future studies should examine whether develop-mental personality types can be replicated usingmore measurement occasions that are more denselydistributed. Furthermore, our choice of predictors isnot exhaustive and we urge researchers to includea broad range of risk factors, covering both proxi-mal (siblings, peers, school setting) and distal (eth-nic or cultural context, socioeconomic factors)processes. Including more contextual factors mayextend the generalizability of our results beyondour sample of White, middle-class families. More-over, we examined relations with levels of adjust-ment problems (at T3) and of parenting (at T1) onlyand thus, results of this study do not yield knowl-edge regarding changes in parenting or adjustmentproblems, or regarding direction of effects. That is,as children grow older and gain more behavioralautonomy from parents, parenting behaviors like-wise change, and changes in parenting have beenfound to be affected by child personality (Roskam& Meunier, 2012). Because initial levels of adjust-ment problems behaviors were not controlled for,the possibility that associations between develop-mental personality types and adjustment problemsreflect cross-sectional associations in childhood can-not be ruled out. Although personality and adjust-ment problems likely show associated change, this

Correlates of Developmental Personality Types 13

study cannot provide knowledge regarding theextent to which developmental personality typesare related to changes in adjustment problems overtime or, conversely, whether levels of adjustmentproblems in childhood relate to changes in person-ality dimensions (within types). In sum, relationsbetween personality, parenting, and adjustmentproblems should be viewed as dynamic and contin-uously interacting, and research should investigatelongitudinal dynamics between child personality,parenting, and adjustment problems.

Conclusions

Notwithstanding these limitations, the currentstudy made several important contributions toexisting knowledge on personality developmentand its correlates. First, it was shown that personal-ity types meaningfully described individual differ-ences in the development of personality dimensionsfrom childhood into adolescence. The consistency ofthe classification of children into types when takinginto account normative changes in the personalitydimensions further suggest that, rather than differ-ent prevalence rates of types across life phases, per-sonality development is more likely to occur at thelevel of dimensions. The current study’s results alsoaid developmental theory and interventions thatare aimed at reducing adjustment problems, byshowing that children whose constellation of per-sonality characteristics make them more vulnerablefor adjustment problems in adolescence can beidentified already in childhood. Moreover, includ-ing parenting and in particular parental warmth, inaddition to child characteristics may substantiallyimprove the effectiveness of interventions. In sum,personality is a developmentally dynamic construct,which can be adequately described by developmen-tal personality types. Moreover, warm parentinghas predictable relations to personality typesthroughout childhood and adolescence and, con-versely, personality types measured at any agehave consistent associations with externalizing andinternalizing adjustment problems by the end ofmiddle adolescence.

References

Achenbach, T. M. (1991a). Manual for the Youth Self-Reportand 1991 profiles. Burlington: Department of Psychiatry,University of Vermont.

Achenbach, T. M. (1991b). Manual for the Child BehaviorChecklist/4–18 and 1991 profiles. Burlington: Departmentof Psychiatry, University of Vermont.

Achenbach, T. M. (1991c). Manual for Teacher Report Formand 1991 profiles. Burlington: Department of Psychiatry,University of Vermont.

Akse, J., Hale, W. W., III, Engels, R. C. M. E., Raaijmak-ers, Q. A. W., & Meeus, H. J. (2004). Personality, per-ceived parental rejection and problem behavior inadolescence. Social Psychiatry and Psychiatric Epidemiol-ogy, 39, 980–988. doi:10.1007/s00127-004-0834-5

Allport, G. W. (1937). Personality: A psychological interpre-tation. New York: Holt, Rinehart, & Winston.

Anderson, E. R. (1995). Accelerating and maximizinginformation in short-term longitudinal research. InJ. Gottman (Ed.), The analysis of change (pp. 139–164).Hillsdale, NJ: Erlbaum.

Arnold, D. S., O’Leary, S. G., Wolff, L. S., & Acker, M. M.(1993). The Parenting Scale: A measure of dysfunctionalparenting in discipline situations. Psychological Assess-ment, 5, 137–144. doi:10.1037//1040-3590.5.2.137

Asendorpf, J. B. (2003). Head-to-head comparison of thepredictive validity of personality types and dimensions.European Journal of Personality, 17, 327–346. doi:10.1002/per.492

Asendorpf, J. B., & Denissen, J. J. A. (2006). Predictivevalidity of personality types versus personality dimen-sions from childhood to early adulthood: Implicationsfor the distinction between core and surface traits.Merrill-Palmer Quarterly, 52, 486–513. doi:10.1353/mpq.2006.0022

Block, J. (1971). Lives through time. Berkeley, CA: BancroftBooks.

Block, J. H., & Block, J. (1980). The role of ego-control andego-resiliency in the organization of behavior. In W. A.Collins (Ed.), Development of cognition, affect, and socialrelations (Vol. 13, pp. 39–101). Hillsdale, NJ: Erlbaum.

Branje, S. J. T., Hale, W. W. H., III, Frijns, T., & Meeus,W. H. J. (2010). Longitudinal associations betweenperceived parent-child relationship quality and depres-sive symptoms in adolescence. Journal of AbnormalChild Psychology, 38, 751–763. doi:10.1007/s10802-010-9401-6

Branje, S. J. T., van Lieshout, C. F. M., & Gerris, J. R. M.(2007). Big Five personality development in adolescenceand adulthood. European Journal of Personality, 21,45–62. doi:10.1002/per.596

Caspi, A., & Shiner, R. L. (2006). Personality develop-ment. In W. R. M. Lerner, N. Eisenberg, & W. Damon(Eds.), Handbook of child psychology: Social, emotional andpersonality development (6th ed., pp. 300–365). Hoboken,NJ: Wiley.

Chapman, B. P., & Goldberg, L. R. (2011). Replicabilityand 40-year predictive power of childhood ARC types.Journal of Personality and Social Psychology, 101, 593–606.doi:10.1037/a0024289

De Fruyt, F., Bartels, M., Van Leeuwen, K. G., De Clercq,B., Decuyper, M., & Mervielde, I. (2006). Five types ofpersonality continuity in childhood and adolescence.Journal of Personality and Social Psychology, 91, 538–552.doi:10.1037/0022-3514.91.3.538

14 de Haan, Dekovi�c, van den Akker, Stoltz, and Prinzie

Denissen, J. J. A., Asendorpf, J. B., & van Aken, M. A. G.(2008). Childhood personality predicts long-term trajec-tories of shyness and aggressiveness in the context ofdemographic transitions in emerging adulthood. Journalof Personality, 76, 67–99. doi:10.1111/j.1467-6494.2007.00480.x

Donnellan, M. B., & Robins, R. W. (2011). Resilient, over-controlled, and undercontrolled personality types: Issuesand controversies. Social and Personality Psychology Com-pass, 4, 1070–1083. doi:10.1111/j.1751-9004.2010.00313.x

Dubas, J. S., Gerris, J. R. M., Janssens, J. M. A. M., &Vermulst, A. A. (2002). Personality types of adolescents:Concurrent correlates, antecedents, and Type 9 parent-ing interactions. Journal of Adolescence, 25, 79–92. doi:10.1006/jado.2001.0450

Duncan, T. E., Duncan, S. C., & Strycker, L. A. (2006). Anintroduction to latent variable growth curve modeling:Concepts, issues, and applications (2nd ed). Mahwah, NJ:Erlbaum.

Galambos, N., & Costigan, C. L. (2003). Emotional andpersonality development in adolescence. In I. B. Weiner(Series Ed.) & R. M. Lerner, M. A. Easterbrooks, &J. Mistry (Vol. Eds.), Handbook of psychology: Vol. 6.Developmental psychology (pp. 211–240). New York:Wiley.

Germeijs, V., & Verschueren, K. (2011). Indecisivenessand Big Five personality factors: Relationship andspecificity. Personality and Individual Differences, 50,1023–1028. doi:10.1016/j.paid.2011.01.017

John, O. P., & Robins, R. W. (1993). Determinants of inter-judge agreement on personality traits: The Big Fivedomains, observability, evaluativeness, and the uniqueperspective of the self. Journal of Personality, 61,521–551. doi:10.1111/j.1467-6494.1993.tb00781.x

Kendler, K. S., & Baker, J. H., (2007). Genetic influenceson measures of the environment: A systematic review.Psychological Medicine, 37, 615–626. doi:10.1017/S0033291706009524

Klimstra, T. A., Hale, W. W. H., III, Raaijmakers, Q. A.W., Branje, S. J. T., & Meeus, W. H. J. (2009). Maturationof personality in adolescence. Journal of Personality andSocial Psychology, 96, 898–912. doi:10.1037/a0014746

Klimstra, T. A., Hale, W. W. H., III, Raaijmakers, Q. A.W., Branje, S. J. T., & Meeus, W. H. J. (2010). A devel-opmental typology of adolescent personality. EuropeanJournal of Personality, 24, 309–323. doi:10.1002/per.744

Lo, Y., Mendell, N. R., & Rubin, D. B. (2001). Testing thenumber of components in a normal mixture. Biometrika,88, 767–778. doi:10.1093/biomet/88.3.767

Mervielde, I., & Asendorpf, J. B. (2000). Variable-centeredand person-centered approaches to childhood personal-ity. In S. E. Hamspon (Ed.), Advances in personality psychol-ogy (Vol. 1, pp. 37–76). Philadelphia: Psychology Press.

Mervielde, I., & De Fruyt, F. (1999). Construction of theHierarchical Personality Inventory for Children (HiPIC).In I. Mervielde, I. J. Deary, F. De Fruyt, & F. Ostendorf(Eds.), Personality psychology in Europe (pp. 107–127).Tilburg, Netherlands: University Press.

Mervielde, I., & De Fruyt, F. (2002). Assessing children’straits with the Hierarchical Personality Inventory forChildren. In B. De Raad & M. Perugini (Eds.), Big Fiveassessment (pp. 129–146). Seattle, WA: Hogrefe &Huber.

Muth�en, B. O., & Muth�en, L. K. (2000). Integratingperson-centered and variable-centered analyses:Growth mixture modeling with latent trajectory classes.Alcoholism: Clinical and Experimental Research, 24,882–891. doi:10.1111/j.1530-0277.2000.tb02070.x

Muth�en, L. K., & Muth�en, B. O. (1998-2007). Mplus user’sguide (5th ed.). Los Angeles, CA: Muth�en & Muth�en.

Nagin, D. S. (2005). Group-based modeling of development.Cambridge, MA: Harvard University Press.

Prinzie, P., & Dekovi�c, M. (2008). Continuity and changeof childhood personality characteristics through thelens of teachers. Personality and Individual Differences, 45,82–88. doi:10.1016/j.paid.2008.03.002

Prinzie, P., Onghena, P., & Hellinckx, W. (2007).Re-examining the Parenting Scale: Reliability, factorstructure, and concurrent validity of a scale forassessing the discipline practices of mothers and fathersof elementary-school-aged children. European Journal ofPsychological Assessment, 23, 24–31. doi:10.1027/1015-5759.23.1.24

Prinzie, P., Onghena, P., Hellinckx, W., Grietens, H.,Ghesqui�ere, P., & Colpin, H. (2003). The additive andinteractive effects of parenting and children’s personal-ity on externalizing behaviour. European Journal ofPersonality, 17, 95–117. doi:10.1002/per.467

Pulkkinen, L. (1996). Female and male personality styles:A typological and developmental analysis. Journal ofPersonality and Social Psychology, 70, 1288–1306. doi:10.1037//0022-3514.70.6.1288

Roberts, B. W., Walton, K. E., & Viechtbauer, W. (2006).Patterns of mean-level change in personality traitsacross the life-course: A meta-analysis of longitudinalstudies. Psychological Bulletin, 132, 1–25. doi:10.1037/0033-2909.132.1.1

Robins, R. W., John, O. P., Caspi, A., Moffitt, T. E., &Stouthamer-Loeber, M. (1996). Resilient, overcontrolled,and undercontrolled boys: Three replicable personalitytypes. Journal of Personality and Social Psychology, 70,157–171. doi:10.1037//0022-3514.70.1.157

Robins, R. W., & Tracy, J. L. (2003). Setting an agenda for aperson-centered approach to personality development.Monographs of the Society for Research in Child Development,68, 110–122. doi:10.1111/j.1540-5834.2003.06801012.x

Robinson, C. C., Mandleco, B., Olsen, S. F., & Hart, C. H.(1995). Authoritative, authoritarian and permissiveparenting practices: Development of a new measure.Psychological Reports, 77, 819–830. doi:10.2466/pr0.1995.77.3.819

Roskam, I., & Meunier, J. C. (2012). The determinants ofparental childrearing behavior trajectories: The effectsof parental and child time-varying and time-invariantpredictors. International Journal of Behavioral Develop-ment, 36, 186–196. doi:10.1177/0165025411434651

Correlates of Developmental Personality Types 15

Rothbaum, F., & Weisz, J. R. (1994). Parental caregivingand child externalizing behavior in nonclinical samples:A meta-analysis. Psychological Bulletin, 116, 55–74.doi:10.1037//0033-2909.116.1.55

Schafer, J. L., & Graham, J. W. (2002). Missing data: Ourview of the state of the art. Psychological Methods, 7,147–177. doi:10.1037//1082-989X.7.2.147

Smetana, J. G., Campione-Barr, N., & Metzger, A. (2006).Adolescent development in interpersonal and societalcontexts. Annual Review of Psychology, 57, 255–284.doi:10.1146/annurev.psych.57.102904.190124

Soto, C. J., John, O. P., Gosling, S. D., & Potter, J. (2011).Age differences in personality traits from 10 to 65: BigFive domains and facets in a large cross-sectionalsample. Journal of Personality and Social Psychology, 100,330–348. doi:10.1037/a0021717

van den Akker, A. L., Dekovi�c, M., & Prinzie, P. (2010).Transitioning to adolescence: How changes in childpersonality and overreactive parenting predict adoles-cent adjustment problems. Development and Psycho-pathology, 22, 151–163. doi:10.1017/S0954579409990320

16 de Haan, Dekovi�c, van den Akker, Stoltz, and Prinzie