Download - Parenting and personality as predictors of child and adolescent internalizing and externalizing

Faculteit Psychologie en

Pedagogische Wetenschappen

Parenting and personality as predictors of child and adolescent

internalizing and externalizing problem behavior

Karla Van Leeuwen

Promotor: Prof. Dr. L. Verhofstadt-Denève

Copromotor: Prof. Dr. I. Mervielde

Proefschrift ingediend tot het behalen van de academische graad

van Doctor in de Psychologische Wetenschappen

2004

Dankwoord

DANKWOORD

Dit proefschrift is het resultaat van de inbreng van veel mensen. Mijn oprechte

dank gaat dan ook uit naar iedereen die mij begeleid en gesteund heeft tijdens de

periode dat ik aan het proefschrift heb gewerkt.

Mijn promotor, Prof. Dr. Verhofstadt-Denève, en copromotor, Prof. Dr.

Mervielde wil ik danken voor de mogelijkheid die ze mij boden om te werken rond

een boeiend onderwerp dat twee onderzoeksgebieden binnen ‘onze’ vakgroep

verenigt, de ontwikkelingspsychologie en de persoonlijkheidspsychologie. Leni, ik

ben jou in het bijzonder erkentelijk voor alle logistieke steun, de kansen die ik kreeg

tot bijscholing, deelname aan congressen, en het flexibel organiseren van mijn

werkzaamheden. Ivan, ik wil jou van harte danken voor de uitnodigingen tot

deelname aan congressen, intensive course en expertmeeting, voor de aanbreng van

nuttige literatuur en nieuwe ideeën, voor het nauwgezet nalezen van teksten, en niet in

het minst voor de motiverende feedback.

Daarnaast heb ik de zinvolle inhoudelijke suggesties op prijs gesteld afkomstig

van de leden van de doctoraatsbegeleidingscommissie enerzijds, met name Prof. Dr.

R. Claes, Prof. Dr. W. De Corte, en Prof. Dr. P. Van Oost, en van de anonieme

reviewers van de aangeboden manuscripten anderzijds. Ook Dr. Ad Vermulst wil ik

vermelden, met dank voor de leerrijke samenwerking!

Een gedeelte van de dataverzameling voor de studies in dit proefschrift

gebeurde in het kader van twee beleidsgerichte onderzoeken, financieel ondersteund

door de Vlaamse Gemeenschap. Deze mogelijkheid kreeg ik dankzij de promotoren

en copromotoren van de projecten, met name Prof. Dr. Caroline Braet, Prof. Dr. Leni

Verhofstadt-Denève en Prof. Dr. Ivan Mervielde. Heel graag wens ik ook alle

projectmedewerkers, Guy Bosmans, Wim De Mey, Jan De Weerdt, Thierry

Meerschaert, Els Merlevede, Ellen Moens en Anja Van Impe, te danken voor de hulp

bij de dataverzameling en de aangename samenwerking.

Het verzamelen van data kon enkel doordat honderden gezinnen bereid waren

om onbekende mensen in hun leven toe te laten en de tijd wilden nemen om

vragenlijsten in te vullen. Omdat ik dat niet als vanzelfsprekend beschouw, dank ik

Dankwoord

van harte alle ouders en kinderen die aan het onderzoek meewerkten. Een pluim ook

voor de (job)studenten die de gezinnen thuis hebben bevraagd en die gegevens hebben

ingebracht in een databestand.

Verder wil ik alle collega’s van de Vakgroep Ontwikkelings-,

Persoonlijkheids- en Sociale Psychologie bedanken, waartoe ik ook hen reken die

ondertussen een andere werkplek hebben gevonden. Ik zal jullie morele steun en jullie

prettige aanwezigheid tijdens informele gelegenheden niet licht vergeten. Speciale

dank gaat uit naar Prof. Dr. De Fruyt, Filip, voor de interesse in mijn onderzoek, de

vele ondersteunende suggesties en de ontspannende loopsessies. Barbara, onze fijne

samenwerking, de grappige momenten, babbels, en de steun bij moeilijke momenten

betekenen veel voor mij!

Ook mijn medestudenten van de opleiding Kwantitatieve Analyse in de

Sociale Wetenschappen aan de KUB, An Jacobs in het bijzonder, ben ik erkentelijk

voor het samen succesvol doorkomen van een lastig jaar.

Tenslotte zijn er nog familie en vrienden, die naast het werk voor de nodige

momenten van ontspanning zorgden. In het bijzonder wil ik mijn ouders vermelden

die steeds voor een warme opvoedingsomgeving hebben gezorgd, eerst voor hun

eigen kinderen, en nu ook voor hun kleinkinderen. Dank, ook aan mijn schoonouders,

omdat jullie altijd klaar staan voor mij en mijn gezin.

Björn, jou wil ik danken voor het scheppen van een prettige en rustige

thuisbasis, voor de tijd en ruimte die je me geeft voor werk en werkuitstappen, en niet

in het minst voor jouw fantastische manier van omgaan met onze dochtertjes. En ten

slotte wil ik Elena en Emmely zeggen dat hun vele lieve woordjes, tekeningen, leuke

verhalen, knuffels en zoentjes me heel blij maken. Dank voor alle fijne momenten,

jullie zorgden ervoor dat ik mezelf niet verloor in dit proefschrift!

Contents I

CONTENTS

Introduction 1 Method 3 Participants 3 Measures 6 Procedure 9 Statistical considerations 10 Overview of chapters 11 Chapter 1 11 Chapter 2 12 Chapter 3 13 Chapter 4 14 Chapter 5 15 References 17 Figures 22 Chapter 1: The Ghent Parental Behavior Scale: some psychometric properties 25 Abstract 25 Introduction 26 Study 1 29 Method 29 Participants 29 Materials 30 Procedure 31 Results 32 Study 2 35 Method 35 Participants 35 Materials 35 Procedure 35 Results 36 General discussion 38 References 44 Tables 50 Appendix 54 Chapter 2: A longitudinal study of the utility of the resilient, overcontrolled and undercontrolled personality types as predictors of children’s and adolescents’ problem behavior 57 Abstract 57 Introduction 58 Study 1 64 Method 64 Participants 64 Procedure 64

II Contents

Measures 65 Results 66 Conclusion 68 Study 2 69 Method 69 Participants 69 Procedure 69 Measures 69 Results 70 General discussion 72 References 76 Tables 80 Chapter 3: Child personality and parental behavior as moderators of problem behavior: a variable- and a centered approach 85 Abstract 85 Introduction 86 Study 1 94 Method 94 Participants 94 Measures 95 Procedure 97 Statistical analyses 98 Results 101 Discussion 105 Study 2 106 Method 106 Participants 106 Measures 106 Procedure 106 Results 107 Discussion 109 General Discussion 111 References 119 Tables 127 Figures 133 Chapter 4: Child personality and parental behavior as interacting predictors of child internalizing and externalizing behavior in clinically referred and non- referred children 135 Abstract 135 Introduction 136 Method 144 Participants 144 Measures 145 Procedure 147 Statistical analyses 147 Results 149 General discussion 155 References 163

Contents III

Tables 171 Figures 173 Chapter 5: Parent personality, child personality and parenting as predictors of child internalizing and externalizing behavior 179 Abstract 179 Introduction 180 Method 187 Participants 187 Measures 188 Procedure 191 Statistical analyses 191 Results 192 General discussion 198 References 204 Tables 212 Figures 215 General conclusions 219 The relative contribution of child personality, parent personality and parenting to child problem behavior 219 Clinical implications 224 Limitations and suggestions for further research 226 Final conclusion 229 References 230 Samenvatting 233 Hoofdstuk 1 234 Hoofdstuk 2 236 Hoofdstuk 3 238 Hoofdstuk 4 239 Hoofdstuk 5 241 Algemeen besluit 243 Klinische implicaties 246 Referenties 248

Introduction 1

INTRODUCTION

This doctoral dissertation focuses on the question why some children and adolescents

show problem behavior whereas others do not. It is examined to what extent this can be

explained by individual differences, i.e. child personality and parent personality, and

environmental influences, i.e. child-rearing behavior. This will enable us to find some

tentative answers on the following leading questions: does parental behavior only matter for

some kinds of children and not for others; do some child personality characteristics serve as a

protective or a risk factor in rather inadequate rearing environments; is it possible to identify

certain ‘types’ of children who are more or less vulnerable showing problem behavior in the

presence of certain parental behavior; do different forms of parental behavior elicit different

outcomes; do personality characteristics and parent behaviors differentially affect emotional

or behavioral problems in children; are child personality and parenting in the same way

related to problem behavior across referred and non-referred children; how is parent

personality related to aspects of parenting, child personality and child outcome behavior?

The majority of the participants in our studies are parents and children from the

general population. We concentrate on two developmental stages in the life course: childhood

and adolescence.

The principal outcome variable is the child’s (mal)adaptive behavior. Because the

samples predominantly consist of non-referred children, we utilize an empirically based,

dimensional approach, to assess behavioral and emotional problems, i.e. the Achenbach

System of Empirically Based Assessment (1991; 1995a; 1995b), instead of a categorical

taxonomy of psychopathology such as the DSM (Diagnostic and Statistical Manual of Mental

Disorders). More in particular we center our attention on the two broadband syndromes

internalizing and externalizing behavior, referring to emotional and behavioral problems

respectively.

The major predictors in our research are based on measures comprising items that

describe the normal or adaptive range of individual characteristics and parenting. It is

assumed that scoring at the extremes of these measures reflects maladaptive features. The

predictor variables are measures with a sound theoretical base. The social interactional theory

of Patterson and colleagues from the Oregon Social Learning Center (OSLC) serves as a

framework for the parenting variables. Patterson’s micro or ‘coercion’ theory states that

2 Introduction

maladaptive child behavior is the result of a series of social interactional processes, involving

the contingent use of aversive behavior combined with ineffective parent management

techniques, such as inconsistent punishment and reinforcement. The macro-level model (see

Figure 1) hypothesizes that parenting practices mediate the relationships between child

adjustment and family background contexts (Reid, Patterson & Snyder, 2002). Specific

family management practices such as monitoring, discipline (also referred to as ‘limit

setting’), positive reinforcement, problem solving and parental involvement, are considered

as crucial (Capaldi & Patterson, 1989; Patterson, Reid and Dishion, 1992). In our study the

personality variables are based on the Five Factor Model (FFM) of Personality. This robust

reference-model of personality can be used to represent personality not only in adults but also

in children and adolescents (Kohnstamm, Halverson, Mervielde & Havil, 1998; Shiner &

Caspi, 2003).

The field of developmental psycho(patho)logy, assumes that child developmental

outcomes can only be predicted by considering multiple determinants. From an ecological or

contextual perspective, the child is nested within a complex network of interconnected

systems (Belsky, 1984; Bronfenbrenner, 1986; Dishion, French & Patterson, 1995; Sameroff,

2000). These models suggest concurrent effects of the determinants, fostering the need for

exploratory research investigating interactions between predictor variables (Hinshaw, 2002).

In a heuristic model (see Figure 2) graphing the determinants of parenting and child

developmental outcomes, Belsky (1984), posits that three domains determine the quality of

parenting: (a) characteristics of the parent, (b) characteristics of the child, and (c) contextual

sources of stress and support, such as the marital relationship, the social network and work

experiences. Parent personality is considered as the most important determinant of parenting.

The nodes in the model are linked by specific pathways. Our dissertation is related to this

ecological perspective in the sense that it includes multiple determinants of child outcome

behavior, such as parent personality, child personality and parenting. In addition, we test

some of the assumptions of Belsky’s model.

According to Sameroff (1975) child behavior can be explained by three models. A

first model is the ‘main effects model’, explaining outcomes by either constitutional or

environmental factors. Several results of studies examining the main effects of child

individual differences and parenting will be discussed as part of the introduction of some of

the chapters. The second model, the ‘interactional effects model’, explains child behavior by

the statistical interaction of constitutional and environmental factors. Interactional effects are

suggested by Thomas and Chess’ goodness-of-fit-theory’ (1977), postulating that

Introduction 3

maladaptive child behavior is the result of a mismatch between a difficult child temperament

and parenting practices. In theory, temperament does not lead to behavioral problems by

itself, it has only an effect in conjunction with particular environments (Bates, Pettit, Dodge

& Ridge, 1998). Finally, the ‘transactional effects model’ examines developmental outcomes

in the recurrent reciprocal interchanges over time between the environment (parents and

others) and the child. Patterson’s ‘coercive cycles’ model (1982), is an example of the

transactional effects model, postulating bidirectional influences between children and parents

(Lytton, 1990). A child’s aggressive antisocial behavior is followed by aversive reactions by

the parent, which in turn escalates the child’s negative behavior. The focus of this dissertation

can be linked to the interactional effects model: in two chapters we investigate child

personality by parenting interactions to explain child behavior.

Method

Participants

The dissertation utilizes data from two samples consisting of non-referred children

and their parents, and one sample including referred children and their mothers. Sample 1

was measured at two assessment occasions separated by a 3-year interval1. Table 1 indicates

the use of the samples across the various chapters.

Table 1

Use of samples across various chapters

Chapter 1 Chapter 2 Chapter 3 Chapter 4 Chapter 5

Sample 1 (T1)

Sample 2

Sample 1 (T1)

Sample 1 (T2)

Sample 1 (T1)

Sample 1 (T2)

Sample 1 (T1)

Referred sample

Sample 1 (T1)

Sample 2

T1 = Time 1, T2 = Time 2 (3 year follow-up data )

1 The Sample 1 data at Time 1 were collected as part of a research project funded by the Flemish Community

(bel96/32), entitled “Deficits in parenting skills as an indicator of behavior problems with children and youth.

Development of a screening instrument for the Flemish community”, under the supervision of Prof. Dr. L.

Verhofstadt-Denève and Prof. Dr. I. Mervielde. Sample 1 follow-up data were collected as part of a research

project entitled 'Research on the effectiveness of an ecological intervention for children with conduct disorder'

funded by the Flemish Community (PBO99A/48-50/75), under the supervision of Prof. Dr. C. Braet and Prof.

Dr. L. Verhofstadt-Denève.

4 Introduction

Table 2 summarizes the characteristics of the non-referred samples. All children have

the Belgian nationality, and the parents are able to read and understand the Dutch language.

We examined if Sample 1 is representative with respect to family characteristics,

parents’ educational level and parents’ employment by comparing this sample with a

representative general community sample of Van den Bergh (1997), who investigated the

quality of life in school-aged children. The family composition was somewhat different, χ²(2)

= 11.95, p < .01, because of more single-parent families in Sample 1. The educational level

differed for mothers, χ²(5) = 12.37, p < .05, and fathers, χ²(5) = 25.46, p < .05, but not as a

consequence of an overrepresentation of the higher educational levels. There was a difference

in employment status for mothers χ²(2) = 9.75, p < .01, but not for fathers χ²(2) = 1.16, p >

.05. Although small differences between the two samples are present, it can be concluded that

the subjects of Sample 1 represent all the broad socio-economic strata. In particular the

sample characteristics do not corroborate the often-suggested concern that only middle- or

higher-class families voluntarily participate in research studies.

Second, we compared the characteristics of Sample 1 (Time 1) and Sample 2, because

these two samples are both used in chapters 1 and 5. The samples differed with respect to

family composition, χ²(2) = 27.62, p < .001, due to the absence of single-parent families in

Sample 2. There was also a difference in educational level of mothers, χ²(5) = 36.96, p <

.001, and fathers χ²(5) = 23.55, p < .001, due to an overrepresentation of the higher

educational level in Sample 2. The sampling method for Sample 2, i.e. students selected

families they knew well, presumably increased the chances for inclusion of middle- or

higher-class families. There were no differences between Sample 1 and 2 as regards the

employment status of mothers, χ²(2) = 1.24, p > .05, and fathers, χ²(2) = 3.34, p > .05.

Finally, we compared the characteristics of the two assessment occasions of Sample

1, because not all the subjects questioned at Time 1 continued participation at Time 2. Chi-

square statistics showed no significant differences for family characteristics, χ²(2) = 2.91, p >

.05, social indices for mothers, χ²(5) = 2.24, p > .05, and fathers, χ²(5) = 1.42, p > .05, and

employment status for mothers χ²(2) = 2.36, p > .05, and fathers χ²(2) = 0.24, p > .05. The

ratio boys/girls, χ²(1) = .07, p > .05, and mothers/fathers, χ²(1) = .03, p > .05 remained the

same over the two assessment moments. Hence, it can be concluded that dropouts did not

form a particular subgroup of the sample and that potential statistical differences between the

two assessment moments are not the consequence of socio-demographic differences.

Introduction 5

Table 2

Sample characteristics National

Sample Sample 1 Sample 1

3 year F.U. Sample 2

Time of data collection 1994-1995 1998-1999 2001-2002 2000 Selection of subjects via schools schools students Response rate of parents (%) 58 40 85 / N families 1789 600 512 175 n mothers n fathers N parents

1005 235 1240

596 535 1131

501 443 944

175 175 350

n male target children n female target children N children

871 918 1789

281 319 600

244 268 512

155 195 350

Children’s age range Mean age (SD)

6-12 9.2

7-15 11.0 (1.8)

10-18 13.9 (1.8)

7-14 10.6 (1.8)

Family characteristics (%) - original family - newly composed family - single-parent family

84.4 11.3 4.3

82.0 9.8 8.2

79.4 9.2 11.4

91.4 8.0 0.6

Mother’s age range Mean age (SD)

24-52 36.0

20-68 38.6 (4.7)

30-62 41.8 (4.5)

29-53 39.2 (4.0)

Father’s age range Mean age (SD)

22-66 39.0

26-63 40.6 (5.0)

33-66 43.7 (4.9)

26-60 41.1 (4.6)

Mother’s educational level (*) (%) 1. Elementary school 2. Secondary school : level 1 3. Secondary school : level 2 4. Secondary school : level 3 5. Higher education 6. University degree

13.4 18.1 16.2 12.9 31.3 8.1

12.9 17.6 21.8 9.1 29.7 8.7

12.8 14.9 21.1 10.7 31.6 8.9

4.0 9.1 16.0 10.3 51.4 9.1

Father’s educational level (*) (%) 1. Elementary school 2. Secondary school : level 1 3. Secondary school : level 2 4. Secondary school : level 3 5. Higher education 6. University degree

12.7 20.0 18.3 10.2 23.6 15.2

7.9 16.3 25.1 8.1 29.4 13.3

6.2 16.4 24.6 9.3 29.7 13.8

2.3 8.6 19.4 9.1 38.9 21.7

Mother’s employment status (%) - housewife - not employed - employed

17.3 9.6 73.1

19.3 5.5 75.2

16.1 6.6 77.3

16.0 4.6 79.4

Father’s employment status (%) - housefather - not employed - employed

0.4 5.2 94.4

0.6 4.1 95.3

0.9 4.1 95.0

0.0 1.7 98.3

Note: (*) The educational level refers to the highest level of education expressed in a hierarchical classification (elementary

school = elementary school level, lower vocational schooling and special education; secondary school level 1 = lower

secondary technical and higher vocational schooling ; secondary school level 2 = higher secondary technical and lower

general schooling; secondary school level 3 = higher general schooling)

6 Introduction

Measures

Table 2 summarizes the measures and informants used to assess the variables in the

dissertation. The collection of independent data from mothers, fathers and children was

supported by two motives: (a) the importance of consulting multiple sources of information

in order to increase the reliability of the information; (b) avoiding the phenomenon of

‘mother blaming’. Given that mothers are supposed to be more involved with primary child-

care, they are an easy target to blame when something goes wrong with their children

(Gerlsma & Emmelkamp, 1994). On the other hand, the role of the father has often been

neglected in research, or regarded as secondary (Perris, 1994).

Table 3

Overview of variables, measures and informants

Variable Measure Informant Child Psychopathology Child Behavior Checklist Mother

Father Youth Self Report (*) Child Child Personality Hierarchical Personality Inventory for Children Mother

Father Questionnaire Big Five (*) Child Parent Personality NEO-PI-R Mother

Father Parental behavior Ghent Parental Behavior Scale Mother

Father Child about mother Child about father

Parenting stress Parenting Stress Index Mother Father

Note: (*) Only measured in sample 1, Time 2

Child problem behavior - Parent ratings. The Achenbach Child Behavior Checklist

(CBCL) (Verhulst, Van der Ende & Koot, 1996) is a screening instrument for behavioral and

emotional problems in children. Parents rate the frequency of 113 problematic behaviors by

means of a 3-point-Likert scale. Two broadband syndromes can be derived: Internalizing,

with items referring to somatic complaints, social withdrawal, and anxiety/depression, and

Externalizing, including items related to aggression, hyperactivity, and delinquency. There is

evidence for good internal consistency of the scales, acceptable test-retest-reliability, and

cross-cultural construct validity (Dedrick, Greenbaum, Friedman, Wetherington & Knoff,

1997; De Groot, Koot & Verhulst, 1994; Verhulst et al., 1996). In this dissertation only the

broadband scales Internalizing and Externalizing behavior are examined as outcome

Introduction 7

measures. Some CBCL scales are not included: Social problems, Attention problems and

Thought problems.

Child problem behavior - Child ratings. Sample 1 children and adolescents were

administered at time 2 the Dutch version of the Youth Self Report (YSR; Verhulst, Van der

Ende & Koot, 1997), a parallel measure of the CBCL. This provided self-reports on

internalizing and externalizing problem behavior.

Parental behavior - Parent ratings. The Ghent Parental Behavior Scale (GPBS; Van

Leeuwen, 1999; 2000) consists of items expressing parental behavior. Parents rate the

frequency of each behavior towards one target child on a 5-point Likert scale ranging from

‘never’ to ‘always’. The items are assigned to nine scales: Positive parenting, Monitoring,

Rules, Discipline, Inconsistent discipline, Harsh punishment, Ignoring, Material rewarding

and Autonomy. The studies in this dissertation focus on two second-order dimensions of the

GPBS, i.e. ‘support or positive parenting’ and ‘negative control’. This is convenient because

(a) it reduces the number of scales by combining them into more parsimonious constructs,

and (b) it constitutes more reliable measures.

Because the evaluation of the psychometric properties of the GPBS is part of this

dissertation, a detailed description of the development and the characteristics of this measure

is given in chapter 1.

Parental behavior - Child ratings. The child version of the GPBS comprises the same

items as the parent version and allows children/adolescents to rate the parental behavior of

their mother and/or father.

Child personality - Parent ratings. The child’s Five Factor personality was assessed

with the Hierarchical Personality Inventory for Children (HiPIC; Mervielde & De Fruyt,

1999). Eighteen facets are hierarchically organized under the five domains, with (a)

Extraversion comprising the facets Shyness, Optimism, Expressiveness and Energy, (b)

Benevolence including Egocentrism, Irritability, Compliance, Dominance and Altruism, (c)

Conscientiousness consisting of Achievement motivation, Concentration, Perseverance and

Orderliness, (d) Emotional Stability based on Anxiety and Self-confidence, and (e)

Imagination containing Creativity, Curiosity and Intellect. The parents rate the items on a 5-

8 Introduction

point Likert scale. There is evidence for a highly replicable factor structure across samples of

children and adolescents (Mervielde & De Fruyt, 2002).

Child personality - Child ratings. Sample 1 children and adolescents completed at

Time 2 a Dutch shortened version of Goldberg’s (1992) hundred adjectives, entitled the

Questionnaire Big Five (QBF). The QBF assesses the Big Five personality domains, with

ratings on 7-point Likert scales for 30 adjectives, 6 per domain (Gerris, Houtmans,

Kwaaitaal-Roosen, de Schipper, Vermulst & Janssens, 1998). Previous research has

demonstrated that self-ratings on this adjective set provide a valid adolescent Big Five

personality profile (Dubas, Gerris, Janssens & Vermulst, 2002; Scholte, van Aken & van

Lieshout, 1997).

Parent personality - Parent ratings. To assess parent personality, the Dutch version of

Costa and McCrae’s NEO PI-R (Hoekstra, Ormel & De Fruyt, 1996) was used. This

questionnaire assesses the FFM domains Agreeableness, Conscientiousness, Neuroticism,

Extraversion and Openness. Domain scores are obtained by aggregating six domain facets,

and each facet is measured with eight items. Agreeableness is based on the facets Trust,

Straightforwardness, Altruism, Compliance, Modesty and Tender-Mindedness.

Conscientiousness comprises the facets Competence, Order, Dutifulness, Achievement

Striving, Self-Discipline and Deliberation. The facets Anxiety, Angry Hostility, Depression,

Self-consciousness, Impulsiveness and Vulnerability create the Neuroticism domain.

Extraversion includes the facets Warmth, Gregariousness, Assertiveness, Activity,

Excitement-Seeking and Positive Emotions. Openness is built on the facets Fantasy,

Aesthetics, Feelings, Actions, Ideas and Values. The Dutch NEO PI-R has satisfactory

psychometric characteristics (Hoekstra, Ormel & De Fruyt, 1996).

Parenting Stress - Parent ratings. The Dutch, experimental version of the Parenting

Stress Index (PSI) assesses stress in the pedagogical situation (‘Nijmeegse Ouderlijke Stress

Index’ (NOSI); de Brock, Vermulst, Gerris & Abidin, 1992). The development of this scale is

based on the assumption that stress is a multidimensional construct. Experience of stress in

child rearing can be evoked by characteristics of the parents, characteristics of the child or

contextual/demographical situations.

Parents rate 123 items by means of a 6-point-Likert scale. The ‘Parent domain’ refers

to stress evoked by characteristics of the parent, and includes seven scales: (a) feelings of

Introduction 9

incompetence in parenting, (b) role-restriction, (c) child-parent attachment problems, (d)

parental depression, (e) health problems, (f) social isolation, and (g) dissatisfaction with the

marital relationship. The ‘Child domain’ refers to feelings of stress as a consequence of

certain characteristics of a target child, comprising (a) low adaptability, (b) mood swings, (c)

distractibility/hyperactivity, (d) demandingness, (e) low positive reinforcement, (f) and low

acceptation of the child. A higher score on a scale indicates more feelings of stress. The

psychometric properties of the Dutch version of the PSI are acceptable to good (de Brock et

al.,1992).

Procedure

Sample 1. Subjects in Sample 1 were recruited in 1998-1999 and again questioned 3

years later. Families were selected via stratified random sampling of elementary and

secondary schools. An atlas outlining the socio-economic situation of the provinces in

Belgium (Mérenne, Van Der Haegen & Van Hecke, 1991), indicated that East- and West-

Flanders could be considered as representative Flemish provinces. For elementary schools the

sample was stratified by province, (East and West Flanders), region (rural or urban), school

type (public/private/catholic schools) and grade (third, fourth, fifth and sixth year of

elementary school). For secondary schools, sampling was based on province (East and West

Flanders), type of curriculum (vocational, technical and general education) and grade (first

and second year of secondary school). Schools were randomly selected (with a procedure in

SPSS) and asked for their cooperation with the study. About 80% of the primary schools and

60% of the secondary schools gave the researchers permission to contact parents via the

children. A letter addressed to the parents informed them about the goal and the procedure of

the research project. The response rate of parents with children in primary schools was 41%,

and 39% for parents with children in secondary schools. In most cases, there was no specific

reason why subjects did not want to participate in the study, 16% of the parents mentioned

lack of time. Family research has shown that in particular practical considerations, like time

and organisation, are the major reasons for not participating in studies, rather than

characteristics of the family that can bias the results (Spoth, Reyes, Redmond & Shin, 1999).

At Time two, about 85% of the families continued collaboration. At each assessment

period, a trained psychology student visited participating families at home. Students received

oral and written instructions, in order to standardize the data collection as much as possible.

Mother, father and child each completed independently the questionnaires. Visits lasted 90 to

180 minutes. Children who participated in the study received a small present for their

10 Introduction

cooperation (a box of colour pencils at Time 1, and a free film ticket at Time 2). Parents

received written feedback after the first measurement time.

Sample 2. The data of Sample 2 were collected in 2000. Students taking courses in

Developmental Psychology at Ghent University were instructed to solicit the cooperation of a

family of their choice for a study of 'Parental behavior in Flanders'. In order to standardize

the visits as much as possible, the research-assistants and students received oral and written

instructions. Both the parents (if they lived together) independently completed the

questionnaires at home, in the presence of the student.

Sample of referred children. Children of the referred sample were recruited from

various mental health services. Third year students were instructed to enlist clinically referred

children with emotional or behavioral problems as part of an assignment for the advanced

course on Personality Psychology at the Ghent University. The therapist or counselor and the

parents of the children gave informed consent. Mothers completed the CBCL, the HiPIC and

the GPBS. Exclusion criteria were the presence of a physical disability or a condition of

chronic disease.

Statistical considerations

It should be noted that the term ‘predictor’ in our study is used in a statistical sense. It

is common practice to use this terminology in the context of multiple regression analysis

even when the data are cross-sectional. We use ‘prediction’ and ‘effect’ to describe

associations between independent and dependent (or outcome) variables in both cross-

sectional and longitudinal data sets. However, we recognize that a cross-sectional study does

not permit conclusions about the causal nature of associations, because it does not comply

with one of the necessary conditions for causality i.e. measures of the independent variable

precedes in time assessment of the dependent variable.

We further use the statistical terms ‘moderated’ and ‘mediated’ effect. A moderated

or interaction effect (see Figure 3A) implies that the value of an outcome variable depends

jointly upon the value of two or more predictor variables; with moderated effects we specify

the conditions under which a relationship is weakened or strengthened. A moderator interacts

with another predictor variable while having an impact on a dependent variable. A mediated

effect (see Figure 3B) is found when an independent variable influences the mediatior which

in turn influences the outcome variable. Four conditions must be met for a variable to be a

Introduction 11

mediator (a) the predictor must be significantly associated with the mediator, (b) the predictor

must be significantly associated with the dependent measure, (c) the mediator must be

significantly associated with the dependent variable, and (d) the impact of the predictor on

the dependent measure is less, after controlling for the mediator Two types of statistical

strategies to detect moderated and mediated effects are commonly used: multiple regression

and structural equation modelling (Baron & Kenny, 1986; Holmbeck, 1997).

Overview of chapters

This doctoral dissertation consists of five chapters2. The following section describes

the goals of the different studies and their additional value compared to previous research.

The studies reported in chapters 1 and 2 provide the foundation for the subsequent research

reported in chapters 3, 4 and 5.

Chapter 1

In this dissertation, parental behavior is considered as one of the major factors related

to child problem behavior. The first chapter (Van Leeuwen & Vermulst, in press) describes

the development and refinement of a new instrument to measure parental behavior, the Ghent

Parental Behavior Scale (GPBS). Although it would have been more convenient to rely on an

existing instrument assessing child rearing, we nevertheless deemed the development of this

new measure as essential. An important motive was our interest in an instrument based on a

theoretical and/or empirical framework. One of the shortcomings in studies of parental

rearing is the inconsistency in definitions of parenting variables, compromising comparisons

between studies and generalizations over studies. A related problem is that of the type of

parental behaviors that should be included in research studies: every now and then the choice

of variables seems arbitrary and tends to produce bias (Perris, 1994).

The social interactional theory of Patterson and colleagues from the Oregon Social

Learning Center (OSLC) emerged as a reasonable candidate for a basic theory. The OSLC

conducted extensive research on both social and psychological processes related to child

(mal)adaptive development and family functioning, as well as on the utility of clinical

interventions. However, the available parenting measures are often only partially compatible

2 The chapters are manuscripts, some of them in press or submitted for publication.

12 Introduction

with Patterson’s theory. In addition, there are hardly any Dutch instruments assessing parental

behavior (cf. Langemijer, Pijnenburg & Veerman, 1997).

A pilot version of the GPBS was developed in 1998 and was evaluated in Van

Leeuwen (1999). A refined version of the GPBS is examined in the first chapter, by means of

two studies. The first study evaluates (a) the factorial validity, (b) the internal consistency, (c)

the degree of agreement between ratings of parents and children, (d) and the construct

validity, by relating the GPBS to various criterion variables. The second study examines

whether the findings of study 1 are replicable in an independent sample. Although this is

considered as a crucial step in the refinement of assessment instruments, it is often neglected

(Smith & McCarthy, 1995). A thorough evaluation of the GPBS is required, given the use of

this measure in the studies described in chapters 3, 4 and 5.

Chapter 2

A second major variable to explain child outcome behavior in this dissertation is child

personality. Two alternative perspectives can be adopted to describe child individual

characteristics: the variable- and the person-centered approach. The variable-centered

approach studies replicable broad categories of variables such as the dimensions of the Five-

Factor Model (FFM) of personality. This type of research focuses on the correlational

structure of the variables across persons within a particular population. The person-centered

approach identifies ‘types’, i.e. clusters of individuals with similar personality patterns or

typical configurations of variables within the person. The main advantage of the personality

type-approach is the classification of subjects with a given profile of scores on multiple

personality variables, instead of considering only one variable at a time.

This study (Van Leeuwen, De Fruyt & Mervielde, in press) joins a renewed interest in

the person-centered approach in personality (cf. Asendorpf, Caspi & Hofstee, 2002), and

evaluates the use of three personality prototypes based on the FFM of personality: resilient,

overcontrolled and undercontrolled.

Our research extends in several ways previous empirical studies documenting the

type-approach, because (a) it brings into play different informants, i.e. parental ratings and

adolescent self-ratings of personality, (b) includes two different FFM measures, i.e. the

Hierarchical Personality Inventory for Children and the Questionnaire Big Five, and (c)

examines two different age groups, i.e. 7-15 and 10-18. The longitudinal design with follow-

up data over a three-year interval further enables the investigation of type continuity and type

membership stability.

Introduction 13

In the chapter we describe the results of two studies. In the first study the replicability

of the types is examined. We use a standard statistical procedure to derive the types,

consisting of a particular sequence of cluster analyses outlined by Asendorpf, Borkenau,

Ostendorf & van Aken (2001). It is evaluated whether the clusters can be replicated across

different measures and informants, and over time. In the second study, the external validity of

the types is examined by relating the types to child and adolescent internalizing and

externalizing problem behavior. Furthermore, the incremental validity of the person-centered

approach is evaluated by comparing the relative utility of types and traits in predicting

adolescent problem behavior.

The outcomes of this chapter are important because the distinction between the

variable-centered and person-centered approach is also a major hallmark of the next chapter.

Chapter 3

Previous research extensively corroborated evidence for the main effects of child

individual characteristics and parental behavior on child behavioral outcomes. However,

main effect studies can lead to spurious correlations with problem behavior, and therefore it

is interesting to investigate interaction effects of parenting and individual characteristics on

child problem behavior. The third chapter (Van Leeuwen, Mervielde, Braet & Bosmans,

submitted) demonstrates the contribution of personality-environment interactions to the

development of child problem behavior. It has several strengths. First, previous studies have

predominantly focused on child temperament as part of the interaction with parenting,

whereas our research includes a broadband FFM measure of child personality. In addition,

the present study is the first to examine the interaction between child personality and

parenting from both a variable- and a person-centered approach. Second, previous studies

investigating parenting by child characteristics interactions are often restricted to assessment

of inadequate child parenting practices, such as restrictive control, psychological control or

coerciveness. However, it is equally valuable to consider parental behavior that may enhance

child adjustment or that may protect the developing child against emotional and behavioral

problems. Therefore, the present study includes positive parenting, a more effective

component of parenting, in addition to a negative control dimension. Third, we investigate

both internalizing and externalizing behavior as the outcome variables, whereas past research

has focused predominantly on externalizing behavior. Fourth, we make use of both a cross-

sectional and a longitudinal design.

14 Introduction

The chapter reports two studies. In the first study the interactions are investigated adopting a

variable-centered approach for the assessment of child personality, based on the Five Factor

Model, and parenting. Parenting and personality as well as their cross-products are entered as

(continuous) quantitative variables predicting child and adolescent problem behavior. Several

precautions are taken in order to guarantee sufficient power to detect interaction effects in the

hierarchical multiple regression analyses: the use of a large population sample, reliable

measures, and checks for cross-rater (parents and children) and cross-time (3-year time

interval) stability of the interaction effects. We also investigated whether the interaction

effects assessed at Time one predicted Time two problem behavior.

In the second study parenting by personality interactions are examined from a person-

centered perspective. A classic ANOVA design is used to look for interaction effects between

the three personality types (overcontrollers, undercontrollers and resilients) and discrete

categories of parents with high and low levels of positive parenting and negative parental

control. It is investigated if a priori defined types and type by parenting interactions predict

maladaptive child behavior and hence if types are a viable alternative for the variable-

centered approach in developmental and clinical research.

Chapter 4

The fourth chapter (Van Leeuwen, Mervielde, De Clercq & De Fruyt, submitted)

extends the research of the previous chapter by including both a large community sample and

a sample with children referred to an ambulant or residential clinical setting. This study is

innovative in several ways. First, in both samples the same measures are administered to

investigate parenting by child personality interaction effects on child problem behavior. It is

investigated whether there are mean-level effects of child personality and parenting on child

problem behavior in both the referred and non-referred sample. Second, including referred

and non-referred children allows a more sensitive test of moderator effects because a broader

spectrum of values for the predictor and outcome variables is available. Third, it is explored

whether the moderator effects can be generalized across the non-referred and the referred

group. The last research goal is of pivotal importance to inform clinical psychologists and

psychiatrists debating whether differences between normal and abnormal behavior should be

conceived as either qualitative or quantitative. Current research on personality disorders is

based on a continuity or spectrum model, postulating that differences between normal and

abnormal or clinical samples are constrained to quantitative or mean level differences on the

relevant domains of functioning (Costa & Widiger, 2002; Widiger & Clark, 2000). We

Introduction 15

extend this line of thinking in our research by proposing that, regardless of mean-level

differences, similar relationships between parenting, child personality and problem behavior

should be present in referred and non-referred samples of children. To our knowledge this is

the first study that examines this hypothesis.

Chapter 5

The goal of the study reported in the final chapter is to investigate the concurrent

effects of parenting, child personality and parent personality on child internalizing and

externalizing behavior. Although contextual or ecological models usually recognize that

parent personality is one of the nodes in the model determining child (mal)adaptive outcome,

few studies have addressed the combined effects of parent personality, child personality and

parenting. This study investigates two alternative models, by means of path analyses. The

models are extensions of Belsky’s model (1984). The first model includes direct paths from

each parent personality domain and each child personality domain to either Internalizing or

Externalizing behavior. In addition, this model includes indirect paths from the parent and

child personality domains to child problem behavior, via the parenting variables negative

control and positive parenting. Hence, this model contains both the direct effects of child and

parent personality on child externalizing and internalizing behavior and the indirect effects of

child and parent personality through parenting. The second (alternative) model includes

direct paths from the parent personality and parenting to child problem behavior and indirect

paths from the same variables via child personality. The major difference between both

models is that in the first model parenting is assumed to be the major mediator between

parent personality and problem behavior and child personality is conceived as influencing

parenting. In the second model child personality is the important mediator and hence

parenting is postulated to affect problem behavior directly but also indirectly by its effect on

child personality. Hence in the first model child personality is conceived as an antecedent or

cause of parenting whether in the second model the relationship is reversed, postulating a link

from parenting to child personality.

One of the major strengths of this study is the use of the FFM to measure both parent

and child personality, in order to increase the comparability of personality measurement for

parents and children. The models are tested separately for mother and father self-ratings of

personality and parenting. Child externalizing and internalizing behavior are independently

included as the outcome variables. Finally, the models are cross-validated in an independent

sample.

16 Introduction

Conclusion

With this dissertation we intend to contribute to the understanding of normal and

deviant child behavior by studying correlates of (mal)adaptive behavior. By improving the

understanding of the factors enhancing or diminishing the prevalence of child

psychopathology, we can gain insight in normal and abnormal developmental processes. In

addition, we can gather information to optimize prevention or intervention, one of the major

goals of developmental psychopathology (Sroufe & Rutter, 1984).

Introduction 17

References

Achenbach, T. M. (1991). Manual for the Child Behavior Checklist/4-18 and 1991

Profile. Burlington, VT : University of Vermont, Department of Psychiatry.

Achenbach, T. M. (1995a). Empirically based assessment and taxonomy:

applications to clinical research. Psychological Assessment, 7, 261-274.

Achenbach, T. M. (1995b). Developmental issues in assessment, taxonomy, and

diagnosis of child and adolescent psychopathology. In D. Cicchetti & D. J. Cohen (Eds.),

Developmental psychopathology. Vol 1: Theory and methods (pp. 57-80). New York: John

Wiley & Sons.

Asendorpf, J. B., Borkenau, P., Ostendorf, F., & van Aken, M. (2001). Carving

personality description at its joints: confirmation of three replicable personality prototypes

for both children and adults. European Journal of Personality, 15, 169-198.

Asendorpf, J. B., Caspi, A., & Hofstee, W. K. B. (Eds.) (2002). The puzzle of

personality types. [Special issue]. European Journal of Personality, 16.

Baron, R. M., & Kenny, D. A. (1986). The moderator – mediator variable distinction

in social psychological research: conceptual, strategic, and statistical considerations. Journal

of Personality and Social Psychology, 17, 725-739.

Bates, J. E., Pettit, G. S., Dodge, K. A., & Ridge, B. (1998). Interaction of

temperamental resistance to control and restrictive parenting in the development of

externalizing behavior. Developmental Psychology, 34, 982-995.

Belsky, J. (1984). The determinants of parenting: a process model. Child

Development, 55, 83-96.

Belsky, J., & Barends, N. (2002). Personality and parenting. In M. H. Bornstein (Ed.),

Handbook of parenting. Vol. 3. Being and becoming a parent (pp. 255-277). Mahwah, New

Jersey: Lawrence Erlbaum Associates.

Bronfenbrenner, U. (1986). Ecology of the family as a context for human

development. Developmental Psychology, 22, 521-530.

Capaldi, D., & Patterson, G. R. (1989). Psychometric properties of fourteen latent

constructs from the Oregon Youth Study. New York: Springer-Verlag.

Costa, P. T., & Widiger, T. A. (Eds) (2002). Personality disorders and the five-factor

model of personality (2nd ed.). Washington, DC, US: American Psychological Association,

493 pp.

18 Introduction

De Brock, A. J. L. L., Vermulst, A. A., Gerris, J. R. M., & Abidin, R. R. (1992).

NOSI: Nijmeegse Ouderlijke Stress Index. [NPSI: Nijmegen Parental Stress Index] Lisse:

Swets & Zeitlinger.

Dedrick, R. F., Greenbaum, P. E., Friedman, R. M., Wetherington, C. M., & Knoff,

H. M. (1997). Testing the structure of the Child Behavior Checklist/4-18 using confirmatory

factor analysis. Educational and Psychological Measurement, 57, 306-313.

De Groot, A., Koot, H. M., & Verhulst, F. C. (1994). Cross-cultural generalizability

of the Child Behavior Checklist cross-informant syndromes. Psychological Assessment, 6,

225-230.

Dishion, T. J., French, D. C., & Patterson, G. R. (1995). The development and

ecology of antisocial behavior. In D. Cicchetti & D. J. Cohen (Eds.), Developmental

psychopathology. Vol 2: Risk, disorder, and adaptation (pp. 421-471). New York: John

Wiley & Sons.

Dubas, J. S., Gerris, J. R. M., Janssens, J. R. M., & Vermulst, A. A. (2002).

Personality Types of Adolescents: Concurrent Correlates, Antecedents, and Type X Parenting

Interactions. Journal of Adolescence, 25, 79-92.

Gerlsma, C., & Emmelkamp, P. M. G. (1994). How larger are gender differences in

perceived parental rearing styles?: a meta-analytic review. In C. Perris, W. A. Arrindell a M.

Eisemann (Eds.), Parenting and psychopathology (pp. 55-74). Chisester: John Wiley & Sons.

Gerris, J. R. M., Houtmans, M. J. M., Kwaaitaal-Roosen, E. M. G., de Schipper, J. C.,

Vermulst, A. A., & Janssens, J. M. A. M. (1998). Parents, Adolescents and Young Adults in

Dutch Families: a longitudinal study. Nijmegen: Institute of Family Studies.

Goldberg, L. R. (1992). The development of markers for the Big-Five factor structure.

Psychological Assessment, 4, 26-42.

Hinshaw, S. P. (2002). Process, mechanism, and explanation related to externalizing

behavior in developmental psychopathology. Journal of Abnormal Child Psychology, 30,

431-446.

Hoekstra, H. A., Ormel, J., & De Fruyt, F. (1996). NEO Persoonlijkheidsvragenlijsten:

NEO PI R en NEO-FFI. Lisse: Swets & Zeitlinger.

Holmbeck, G.N. (1997). Toward terminological, conceptual, and statistical clarity in

the study of mediators and moderators: examples from the child-clinical and pediatric

psychology literatures. Journal of Consulting and Clinical Psychology, 65, 599-610.

Introduction 19

Kohnstamm, G. A., Halverson, C. F., Mervielde, I., & Havill, V. L. (1998). Parental

descriptions of child personality: Developmental antecedents of the Big Five? Mahway, NJ:

Erlbaum.

Langemeijer, K., Pijnenburg, H., & Veerman, J. W. (1997). Standaardisatie van de

gezinsdiagnostiek. Een inventarisatie van Nederlandstalige gezinsdiagnostische

instrumenten. [Family Assessment. Inventory of Dutch Family Assessment Instruments.]

Nijmegen: Stichting de Waarden, afdeling Onderzoek en Ontwikkeling.

Lytton, H. (1990). Child and parent effects in boys’ conduct disorder: a reinterpretation.

Developmental Psychology, 26, 683-697.

Mérenne, B., Van Der Haegen, H. & Van Hecke, E. (1991). België ruimtelijk

doorgelicht. Nationaal Instituut voor de Statistiek.

Mervielde, I., & De Fruyt, F. (1999). Construction of the Hierarchical Personality

Inventory for Children (HiPIC). In I. Mervielde, I. Deary, F. De Fruyt, & F. Ostendorf (Eds.),

Personality psychology in Europe. (pp. 107-127). Tilburg University Press.

Mervielde, I., & De Fruyt, F. (2002). Assessing children’s traits with the Hierarchical

Personality Inventory for Children. In B. de Raad & M. Perugini (Eds.), Big Five Assessment

(pp. 129-146). Göttingen: Hogrefe & Huber Publishers.

Patterson, G. R. (1982). Coercive family process. Eugene: OR Castalia.

Patterson, G. R., Reid, J., & Dishion, T. (1992). Antisocial Boys. A social

interactional approach. Volume 4. Eugene: Castalia Publishing Company.

Perris, C. (1994). Linking the experience of dysfunctional parental rearing with

manifest psychopathology: a theoretical framework. In C. Perris, W.A. Arrindell a M.

Eisemann (Eds.), Parenting and psychopathology (pp. 3-32). Chisester: John Wiley & Sons.

Reid, J. B, Patterson, G. R., & Snyder, J. J. (Eds.) (2002). Antisocial behavior in

children and adolescents. A developmental analysis and model for intervention. Washington,

DC: American Psychological Association.

Sameroff, A. J. (1975). Early influences on development: Fact or fancy? Merrill-

Palmer Quarterly, 21, 263-294.

Sameroff, A. J. (2000). Developmental systems and psychopathology. Development

and Psychopathology, 12, 297-312.

Scholte, R. H. J., van Aken, M. A. G., & van Lieshout, C. F. M. (1997). Adolescent

personality factors in self-ratings and peer nominations and their prediction of peer

acceptance and peer rejection. Journal of Personality Assessment, 69, 534-554.

20 Introduction

Shiner, R., & Caspi, A. (2003). Personality differences in childhood and adolescence:

measurement, development, and consequences. Journal of Child Psychology and Psychiatry,

44, 2-32.

Smith, G. T., & McCarthy, D. M. (1995). Methodological considerations in the

refinement of clinical assessment instruments. Psychological Assessment, 7, 300-308.

Spoth, R., Reyes, M., Redmond, C., & Shin, C. (1999). Assessing a public health

approach to delay onset and progression of adolescent substance use: Latent transition and

log-linear analyses of longitudinal family preventive intervention outcomes. Journal of

Consulting and Clinical Psychology, 67, 619-630.

Sroufe, L. A., Rutter, M. (1984). The domain of developmental psychopathology.

Child Development, 55, 17-29.

Thomas, A., & Chess, S. (1977). Temperament and development. New York:

Brunner/Mazel.

Van den Bergh, B. (1997). Kindertijd. Kinderen en ouders over de leefsituatie van

kinderen op lagereschoolleeftijd in Vlaanderen. [Childhood. Children and parents about the

quality of life of elementary school-aged children]. Leuven/Apeldoorn: Garant.

Van Leeuwen, K. (1999). Het meten van opvoeding met de Schaal voor Ouderlijk

Gedrag. [Measurement of parenting with the Parental Behavior Scale]. Diagnostiek-Wijzer, 3,

151-170.

Van Leeuwen, K. (2000). Handleiding Meetinstrumenten bij “Deficieten in

opvoedingsvaardigheden van ouders als indicatoren voor probleemgedrag bij jongeren. Een

screeningsstudie voor Vlaanderen (Impulsprogramma Humane Wetenschappen Vlaamse

Gemeenschap, Beleidsgericht onderzoek bel96/32)”. Universiteit Gent, Vakgroep

Ontwikkelings- en Persoonlijkheidspsychologie.

Van Leeuwen, K., De Fruyt, F., & Mervielde, I. (in press). A longitudinal study of the

utility of the resilient, overcontrolled and undercontrolled personality types as predictors of

children’s and adolescents’ problem behavior. International Journal of Behavioral

Development.

Van Leeuwen, K. G., Mervielde, I., Braet, C., & Bosmans, G. (submitted). Child

personality and parental behavior as moderators of problem behavior: a variable- and a

person-centered approach.

Van Leeuwen, K. G., Mervielde, I., De Clercq, B. J., & De Fruyt, F. (submitted).

Child personality and parental behavior as interacting predictors of child internalizing and

externalizing behavior in clinically referred and non-referred children.

http://www2.lib.rug.ac.be/cgi-bin/webspirs/3.1/webspirs.cgi?sp.usernumber.p=820268&url=yes&sp.nextform=urlsrch.htm&sp.dbid.p=PY84%2CPYFR%2CPYFS%2CPYFT%2CPYFU%2CPYFV%2CPYFW%2CPYFX%2CPYFY%2CPYGA%2CPYGB%2CPYGC%2CPYLC%2CPYLD%2CPYLE%2CPYNH%2CPYNI%2C



Introduction 21

Van Leeuwen, K. G., & Vermulst, A. A. (in press). The Ghent Parental Behavior

Scale: some psychometric properties. European Journal of Psychological Assessment.

Verhulst, F. C., Van der Ende, J. & Koot, H. M. (1996). Handleiding voor de

CBCL/4-18. Erasmus Universiteit Rotterdam, Afdeling Kinder- en Jeugdpsychiatrie.

Verhulst, F. C., Van der Ende, J., & Koot, H. M. (1997). Handleiding voor de Youth

Self-Report (YSR). [Manual for the Youth Self-Report (YSR)]. Rotterdam: Afdeling Kinder-

en Jeugdpsychiatrie, Academisch Ziekenhuis /Erasmus Universiteit.

Widiger, T. A., & Clark, L. A. (2000). Toward DSM-V and the classification of

psychopathology. Psychological Bulletin, 126, 946-963.

22 Introduction

Family Environment - discipline

- monitoring - parent involvement - positive reinforcement - problem solving

CONTEXTS - relationship

Settings

- neighbourhood Child Characteristics

- home - ADHD

- school - Social Information Processing

- temperament ANTISOCIAL - maturation BEHAVIOR Features - affective adjustment

- stigmatization - academic skills

- victimization - bonding

- economic resources

- social support

- behavioral norms Peer environment - deviancy training - network density - relationship

tentative linkages established linkages

Figure 1

Factors influencing the development and maintenance of antisocial behavior

(Dishion, French, & Patterson, 1995, p. 456)

Introduction 23

Marital relations

Social Network

Developmental history

Parent personality

Parenting Child characteristics

Work Child developement

Figure 2

A process model of the determinants of parenting (Belsky, 1984, p. 84)

24 Introduction

A.

Independent

variable

Dependent

variable

Moderator

B.

Independent

variable

Dependent

variable

Mediator

Figure 3

Models of moderated and mediated effects

The top panel shows a moderated effect, while the bottom panel graphs a mediated effect

Assessment of Parental Behavior 25

CHAPTER 1

THE GHENT PARENTAL BEHAVIOR SCALE: SOME

PSYCHOMETRIC PROPERTIES

Abstract

The theoretical basis of the Ghent Parental Behavior Scale (GPBS) originates from social-

learning theories. Parents rate the frequency of their behavior towards a target child between

8 and 14 years old and children rate the behavior of their parents.

Confirmatory factor analyses provide evidence for a solid factor structure in different

samples. Nine scales are distinguished: Positive parenting, Monitoring, Rules, Discipline,

Inconsistent discipline, Harsh punishment, Ignoring, Material rewarding and Autonomy. It is

also feasible to distinguish two second-order factors: support and negative control. The

internal consistency is acceptable to good. Correlations between ratings of parents and

children are positive and significant. We also find evidence for the assumption that positive

parenting is negatively associated with problem behavior and stress in parenting, and in

addition, that inadequate parenting is positively related to problem behavior and stress in

parenting.

26 Chapter 1

Introduction

Several studies demonstrate that parenting is a factor contributing to the development

of problem behavior in children and adolescents. Although it is acknowledged that parenting

can be a risk as well as a protective factor, most studies emphasize its influence on

externalizing behavior or antisocial and delinquent behavior (Chamberlain, 1994; Farrington,

1995; Loeber, 1990; Loeber & Stouthamer-Loeber, 1986; Patterson, 1982; Patterson,

Debarysche & Ramsey, 1989; Patterson, Reid & Dishion, 1992; Patterson & Stouthamer-

Loeber, 1984). This interesting field of research still offers many challenges, not in the least

because other current research questions the influence of parents on their children's behavior

and refers to heredity and peers as alternative sources of influence (Collins, Maccoby,

Steinberg, Hetherington & Bornstein, 2000). However, to tackle some of these assertions we

need adequate instruments assessing well-defined constructs and components of parenting.

Assessment of parenting is also important for prevention and intervention programs aiming to

reduce difficult, and in the end delinquent behavior, by enhancing parenting skills. The

assessment of parenting then serves (a) as a diagnostic tool, comparing parenting in a particular

family with 'typical' parenting in a representative group of families, and (b) as a tool for the

evaluation of intervention and prevention programs with parental training as a component.

In this article, we describe the development of the Ghent Parental Behavior

Questionnaire (GPBS), an instrument to measure parental behavior. This instrument is based

on the social interactional theory of Patterson and colleagues from the Oregon Social

Learning Center (OSLC). They developed the ‘coercion theory’ in order to explain children’s

aggressive behavior. The central idea is that maladaptive behavior of children is the result of

a developmental series of social interactional processes, characterized by the contingent use

of aversive behavior combined with ineffective parent management techniques, such as

inconsistent punishment and reinforcement. On a macro-level the model also includes the

impact of contextual factors on child outcomes, such as neighbourhood, socio-economic

status (SES), divorce and parental depression. In this model, it is hypothesized that parenting

practices mediate the relationships between child adjustment and family background contexts

(Reid, Patterson & Snyder, 2002). The OSLC staff emphasizes the importance of specific

family management practices such as monitoring, discipline (also referred to as ‘limit

setting’), positive reinforcement, problem solving and parental involvement (Capaldi &

Patterson, 1989; Patterson, Reid & Dishion, 1992). Similar constructs are also prominent in

(longitudinal) studies of delinquency. Limited monitoring or lack of supervision, harsh


discipline, rejection and little parental involvement are clear risk factors for later delinquency

(Deater-Deckard, Dodge, Bates & Pettit, 1998; Farrington, 1995; Loeber 1990).

It should be noted that observable parental behavior is distinctive from ‘parenting

styles’ of child rearing. The latter can be regarded as the general context, the climate in which

the more specific parenting practices are expressed (Darling & Steinberg, 1993). In this

respect, the typology of Maccoby and Martin (1983) distinguishes authoritarian,

authoritative, permissive and indifferent parenting based on two dimensions, i.e.

‘demandingness’ and ‘responsiveness’.

Parenting has been studied in various ways (Grotevant & Carlson, 1989; Langemijer,

Pijnenburg & Veerman, 1997). Interviews provide useful information about patterns of family

interaction (Stormshak, Speltz, DeKlyen & Greenberg, 1997), but they are time-consuming

and therefore more useful in a clinical intake, gathering information on parent-child

relationships in different contexts. Direct observation of parental behavior is definitely

valuable, because it facilitates individualized feedback and family tailored practice (Richman

et al., 1994). Observational studies show larger effect sizes for environmental influences on

social development, compared to parental reports (Collins et al., 2000). At the OSLC, the

emphasis is on sequential, micro-level analyses of observations of parent-child interactions in

their natural environment (Snyder & Stoolmiller, 2002). However, when time and money are

restricted, this is a rather inefficient method of data gathering. Another disadvantage of an

observational approach is that it demands intensive training of observers and a well-designed

rating system. Self-rating questionnaires have the advantage of being easy and inexpensive to

administer. However, a literature review of (self-rating) questionnaires leads to the

conclusion that the available measures suffer from several shortcomings. First, it is commonly

assumed that measures developed to quantify parental attitudes, behavioral intentions, beliefs,

self-perceptions or values are indicators of real parental behavior. However, according to some

authors these questionnaires show low reliability and questionable validity (Holden &

Edwards,1989). Some questionnaires are designed to test parents' knowledge of adequate

parental behavior, such as reinforcement techniques likely to enhance appropriate child

behavior and compliance (Frankel, 1993). It could be argued that this method of ‘testing’ may

be threatening to parents and therefore not advisable. Second, we are interested in the

assessment of each of the five parenting constructs specified by Patterson et al. (1992):

monitoring, discipline, positive reinforcement, problem solving and parental involvement.

Some questionnaires include one or more of these components of parental behavior, but none

includes all five (cf. Arnold, O’Leary, Wolff & Acker, 1993; Frick, 1991; Jacob, Moser,

28 Chapter 1

Windle, Loeber & Stouthamer-Loeber, 2000; Metzler, Biglan, Ary & Fuzhong, 1998;

Schaefer, 1965; Frick, 1991). Third, several studies do not report important psychometric

properties of parenting scales, such as their factorial validity (cf. Shelton, Frick & Wootton,

1996; Strayhorn & Weidman, 1988). Finally, some questionnaires assess parental self-ratings,

whereas others elicit child or adolescent perspectives on parenting (cf. Jacob et al., 2000;

Metzler et al., 1998; Schaefer, 1965). Measures that provide judgments of both parents and

children, are scarce.

To overcome the shortcomings of the existent instruments and the fact that there are

hardly any Dutch instruments assessing parental behavior (cf. Langemijer et al., 1997), a new

questionnaire was designed: the Ghent Parental Behavior Scale (GPBS). The GPBS aimes at

measuring five parental skills, proposed by Patterson et al. (1992): (a) Parental involvement,

i.e. showing interest in the child, (b) Monitoring, i.e. supervision, (c) Discipline, i.e. setting

rules, punishing unwanted behavior, (d) Positive reinforcement, i.e. rewarding adaptive

behavior, and (e) Problem solving, i.e. identifying and solving problems. Each of these skills

is operationalized by measures of general behavior (for instance: I punish my child when

he/she does not obey a rule), sometimes supplemented with recognizable examples (for

instance: ... when he/she comes home late without a valid reason). General behavior instead

of situation specific behavior was chosen in order to guarantee that items would be applicable

to most parents. Parents rate the frequency of each behavior towards one target child aged 8-

15 on a 5-point Likert scale ranging from ‘never’ to ‘always’. To avoid ambiguity items are

formulated in the first-person format (Holden & Edwards, 1989).

In a pilot study, a first version of the GPBS was evaluated3 in a sample of 258 non-

referred families, with 120 boys and 138 girls between 7 and 14 years old. Principal factor

analyses (PFA) with oblique rotation led to the conclusion that the factors did not completely

correspond with the hypothesized constructs. The content of the discipline (i.e. limit setting)

construct appeared to be multidimensional, including punishment after unwanted behavior

either in an effective or in an ineffective way (harsh punishment, inconsistent punishment)

and clarification of rules. In addition, the constructs problem-solving and positive

involvement and items referring to social rewarding were closely related. However, there was

structural evidence for monitoring and material rewarding. From these findings it was

concluded that a new version of the GPBS was required.

3 Results of this study are reported in detail in Van Leeuwen (1999).


This article presents some psychometric properties of a refined version of the GPBS:

(a) the factor structure, determined by exploratory factor analyses, confirmatory factor

analysis and multi-group analysis based on LISREL; (b) the internal consistency of the

identified scales; (c) agreement between ratings of parents and children, and (d) the construct

validity, assessed by relating the GPBS to two criterion variables: the experience of stress in

parenting and problem behavior in children. Two studies are reported: (a) a study examining

the dimensional structure and other psychometric properties of the refined GPBS, and (b) a

confirmatory study in which the psychometric findings of the first study are tested.

Study 1

Overview

The purpose of this study was to evaluate a refined version of the GPBS, by

examining the factorial structure of the refined instrument and the internal consistency of the

constructs. Correlations between ratings of parents and children were calculated in order to

get some indication of interrater agreement. We also related the GPBS to two criterion

variables: the experience of stress in parenting and problem behavior in children. From the

literature (Belsky, 1984; Campbell, 1994), it can be expected that stress in parenting and

parental behavior will be associated. We hypothesize that parental stress is associated with

less provision of structure and autonomy (Grolnick, Weiss, McKenzie & Wrightman, 1996),

with less involvement in child rearing (Conger, Elder, Lorenz, Simons & Whitbeck, 1992)

and with more frequent use of punishment (Dielman, Barton & Cattell, 1977). We predict a

positive correlation between externalizing problem behavior and Harsh punishment, Ignoring

and Inconsistent discipline (Campbell, 1994; Campbell, Moore, Marakovitz & Newby, 1996;

Frick et al. 1992; Irvine, Biglan, Smolkowski, & Ary 1999; Jacob et al., 2000; Loeber, 1990;

Metzler et al. 1998; Patterson et al., 1992; Stormshak et al., 1997). Moreover, we expect a

negative correlation between problem behavior and Autonomy, Positive involvement,

Monitoring, Rules and Problem solving.

Method

Participants

Subject of this study was a non-referred random sample of 600 Flemish families and

one of their children between 7 and 15 years old (mean age = 11.0 years, SD = 1.8). The

sample included 596 mothers (mean age = 38.6 years, SD = 4.7), 535 fathers (mean age =

30 Chapter 1

40.6 years, SD = 5.0), 281 sons and 319 daughters. There were 82% original families, 9.8%

newly composed families and 8.2% single parent families. The sample was well balanced

regarding social-economic status of the families (Van Leeuwen, 2000).

Materials

Refinement of the GPBS. The results of the pilot study (Van Leeuwen, 1999) showed

that it was not possible to retrieve the five Patterson constructs from the initial set of items.

Four of the five original constructs were retained: parental involvement, monitoring, positive

reinforcement and problem-solving. Because the factor analyses indicated that the discipline

construct was multidimensional, it was differentiated into setting rules, punishment of

unwanted behavior, inconsistent discipline, harsh punishment, and ignoring. These scales are

all related to social interaction theory. The scale 'Ignoring' refers to neglecting unwanted

behavior of the child. It is akin to permissive discipline, allowing rules to go unenforced

(Arnold et al., 1993). Adding this scale completed the list of (dysfunctional) discipline

practices that were distinguished in the first version of the GPBS. Further differentiation

required developing new items, to reliably assess the content domains that could be

distinguished in the pilot study. Sources of inspiration for the new items were assessment

instruments from Frick (1991) and Schaeffer (1965). Apart from the nine scales based on

social learning theory, the scale 'Autonomy' was added, based on research at the University of

Nijmegen (Gerris et al., 1992, 1998; Vermulst, De Brock & Gerris, 2002). The concept of

autonomy is interesting from a developmental point of view. It is related to the notion of

'demandingness' (Maccoby & Martin, 1983), referring to behavior that taps parental

responsibility for the child’s compliance with certain rules and norms.

The revised questionnaire consists of 55 items: 25 new items and 30 items of the

experimental version, based on the retention criteria (a) primary loadings >.35, or (b) cross

loadings <.30. They are assigned to ten scales with at least five items per scale: Positive

involvement (making time for the child, showing interest; six items), Monitoring (supervision

of the activities of the child; seven items), Rules (teaching the child appropriate behavior;

seven items), Discipline (punishment of the child when it misbehaves; five items),

Inconsistent discipline (punishment in an inconsistent way; five items), Harsh punishment

(corporal punishment and verbal blaming; five items), Ignoring (neglecting unwanted

behavior; five items), Positive reinforcement (rewarding good behavior of the child; five

items), Problem solving (solving problems together with the child; five items), and

Autonomy (stimulating autonomous behavior of the child; five items). The format of the


GPBS was retained because it was positively evaluated in the pilot study. An adapted version

allows children to rate the behavior of their mother and father. The content of the items in the

child version is the same, however the items are formulated in the third person singular ("my

mother..." and "my father..."). One item of the parent version was omitted in the child version

(see Appendix, rul6), because it was too difficult to understand for children.

Instruments used for evaluation of the construct validity of the GPBS

The Dutch version of the Parenting Stress Index (PSI) is used to measure stress in the

pedagogical situation (De Brock, Vermulst, Gerris & Abidin, 1992). Parents rate 123 items

by means of a 6-point Likert scale. The ‘Parent domain’ refers to stress evoked by

characteristics of the parent, for instance feelings of incompetence in parenting, depression,

role-restriction, low attachment, lack of health, social isolation, and low satisfaction with the

marital relationship. The ‘Child domain’ refers to feelings of stress as a consequence of

certain characteristics of a target child, for instance low adaptability, mood swings,

distractibility/hyperactivity, demandingness, low positive reinforcement, and low acceptance

of the child. A higher score on a scale indicates more feelings of stress. The psychometric

properties of the Dutch version of the PSI are acceptable to good (De Brock et al., 1992).

The Achenbach Child Behavior Checklist (CBCL) (Verhulst, Van der Ende & Koot,

1996) is a screening instrument for behavioral and emotional problems in children. Parents

rate the frequency of problematic behaviors on a 3-point Likert scale. Two broadband

syndromes can be derived: Internalizing, grouping items referring to somatic complaints,

social withdrawal, and anxiety/depression, and Externalizing, including items related to

aggression, hyperactivity, and delinquency. There is ample evidence for good internal

consistency of the scales, acceptable test-retest-reliability, and cross-cultural construct

validity (Dedrick et al., 1997; Groot, Koot & Verhulst, 1994; Verhulst, Van der Ende &

Koot, 1996).

Procedure

Families were contacted via randomly selected schools in East- and West-Flanders.

Children from the third to the sixth year in elementary school and the first and second year in

secondary school were given a letter addressed to the parents, containing information about

the study and the procedure and requesting parents’ co-operation. This procedure led to a

sample of 600 Flemish families, with parents and children between 7 and 15 years old. The

response rate of the parents was 40%. Families were visited at home, and family members

32 Chapter 1

independently filled out the questionnaires in the presence of trained research-assistants or

students. In order to standardize the visits as much as possible, the research-assistants and

students received oral and written instructions. Both the parents (if they lived together) and

the target child, independently completed the Ghent Parental Behavior Scale. Items of the

GPBS were presented in a quasi random order, alternating items belonging to different scales.

Both parents filled out the Parenting Stress Index (De Brock et al., 1992) and the Child

Behavior Checklist (Verhulst et al., 1996).

Results

Factor Structure

Initially we used LISREL 8.52 (Jöreskog & Sörbom, 1996) to verify if the ten-factor

model was acceptable for each of the four groups (mothers, fathers, children about mothers

and children about fathers). The ten factors could not be confirmed. To explore the factor

structure further, we followed the strategy of Gerbing and Hamilton (1996), using exploratory

factor analysis (EFA) to avoid numerous analyses with LISREL. The final evaluation of the

factor models was done with LISREL. The purpose of this final evaluation is to have the

disposal of factor loadings and fit measures to compare with the results of the confirmatory

factor analyses in the second study.

The EFA's revealed that the scales Monitoring, Rules, Discipline, Inconsistent discipline,

Harsh punishment, Ignoring and Autonomy were retrievable. However, items referring to

problem solving, positive reinforcement (two social rewarding items) and to positive

involvement formed one factor. This new factor was labelled 'positive parenting'. The three

remaining items of positive reinforcement were labelled 'material rewarding'. One of the

items of inconsistent discipline showed a low loading on this factor but a high loading on the

discipline factor. This can be explained by the item content, i.e. complete lack of discipline,

and hence indicating the opposite of the other discipline items (see item dis6 in Appendix).

Because the same findings emerged in each of the four groups, we restarted factor analysis

with nine factors. The aim was to reach a simple factor structure: items with high loadings on

the target factor (> .40 in at least three of the four groups) and low cross-loadings on the

others (< .20 in at least three of the four groups). As a consequence some items had to be

removed, resulting in a nine-factor solution with 45 items for parents and 44 items for

children (item rul6 was not measured in the child version). The final results evaluated with

LISREL are reported in Table 1. The factor loadings were substantial and mostly above .40.


On a total of 178 primary factor loadings there were only three loadings lower than .40: two

were traced in the monitoring factor, the third in the discipline factor.

Table 1 also shows the fit measures for the factor models. The use of chi-square in large

samples is not adequate because "excessive test power (due to large N) may prompt the

rejection of acceptable models" (Hayduk, 1996, p. 197). Therefore, we decided to use two

measures recommended by several authors: (a) The root mean square error of approximation

(RMSEA; Byrne, 1998), and (b) The Comparative Fit Index (CFI) of Bentler (Marsh, Balla &

McDonald, 1996). RMSEA is utilized for assessing approximate fit (preferably with values

less than or equal .05, but values between .05 and .08 are indicative of fair fit (Kaplan, 2000,

p.113-114). CFI is a comparative fit index, values above .95 are preferred (Kaplan, 2000, p.

107), but should not be lower than .90 (Kline, 1998, p. 131). The RMSEA-values were lower

than .05 and CFI-values were all above .95 except for mothers (.92), indicating an acceptable

fit for all models. As already mentioned these results must be considered with care because

this LISREL evaluation is the endpoint of a series of explorative factor analyses and only

aimed to have a possibility to compare these results with the results of the second study.

The factor correlations (see Table 2) showed three groups of interrelated factors: (a) Positive

parenting, Monitoring, Rules and Autonomy (b) Discipline, Harsh punishment and Ignoring

and (c) Inconsistent discipline and Rewarding. Additional exploratory factor analyses to

explore the dimensional structure of the nine scales showed that a three-factor solution is not

tenable due to unacceptably low loadings on the third factor (for the scales Inconsistent

discipline and Rewarding). A two-factor solution is more appropriate if three scales are

removed: Monitoring, Inconsistent discipline and Rewarding. The first factor (labelled as

‘support’) contains the scales Positive parenting, Rules and Autonomy, the second factor

(label: ‘negative control’) the scales Discipline, Ignoring and Harsh punishment. The

explained variances were 35% (mothers), 40% (fathers), 51% (children about mothers) and

47% (children about fathers). This result is in line with the view that parenting behavior can

be subsumed into two broad dimensions warmth/support and control (Gallagher, 2002;

Maccoby & Martin, 1983). The two dimensions have almost zero correlations. The EFA

results were evaluated with LISREL to check whether they could be replicated in study 2.

The Chi-square values for the four groups ranged from 30.2 to 39.8 with df=8; the RMSEA

values between .09 and .11 and the CFI-values between .89 and .96. The completely

standardized factor loadings ranged mostly between 5. and .8. The fit of the models could be

improved by introducing one or two factor cross loadings. This resulted in small cross-factor

loadings that do not compromise the theoretical bases of the two-dimensional structure of the

34 Chapter 1

six scales. Cronbach’s alphas for the support dimension (with 21 items) varied between .86

and .90 and for the negative control dimension (14 items) between .80 and .87.

Internal Consistency

Cronbach’s alphas are reported in Table 1. Four scales show reliabilities of (mostly)

.80 or higher: Positive parenting, Rules, Discipline and Harsh punishment. The scales

Material rewarding and Ignoring have acceptable reliabilities of mostly .70 or higher.

Monitoring, Inconsistent discipline and Autonomy are less reliable but also have fewer items.

Agreement between Ratings of Parents and Children

Pearson correlations were computed between parental self-rating scores and

children’s ratings of maternal and paternal behavior. Correlations were positive and

significant for Positive parenting, with r = .22 (p < .001) and r = .30 (p < .001) for ratings of

maternal and paternal behavior respectively, for Monitoring with r = .19 (p < .001) and r =

.23 (p < .001), for Setting rules, with r = .12 (p < .01) and r = .19 (p < .001), for Discipline,

with r = .33 (p < .001) and r = .32 (p < .001), for Inconsistent discipline, with r = .17 (p <

.001) and r = .14 (p < .001), for Harsh punishment with r = .36 (p < .001) and r = .32 (p <

.001), for Ignoring with r = .19 (p < .001) and r = .12 (p < .01), for Material rewarding with r

= .32 (p < .001) and r = .28 (p < .001) and for Autonomy with r = .23 (p < .001) and r = .20

(p < .001).

Validity : Correlations with Criterion Variables

Ghent Parental Behavior Scale - Parenting Stress Index. The results largely

confirmed our hypotheses. Table 3 shows meaningful negative correlations. Increasing stress

in the Parent and Child domain is related to decreasing Positive parenting, Rules and

Autonomy and to increasing Inconsistent Discipline, Harsh punishing, Ignoring and

Discipline.

Ghent Parental Behavior Scale - Child Behavior Checklist. Table 3 shows several

significant correlations between parenting and problem behavior especially for Externalizing

behavior. The correlations of Internalizing behavior with the parenting scales are very low.

Increasing Internalizing behavior is related to less Positive parenting, Rules and Autonomy

(only for fathers) and to increasing Inconsistent punishing (mothers), Harsh punishing

(mothers) and Ignoring. Externalizing behavior is linked to less Positive parenting and Rules

and to more Discipline, Inconsistent discipline, Harsh punishment and Ignoring. Although


these results confirmed our expectations, the expected negative correlation between problem

behavior and Monitoring, was not found.

Study 2

Overview

This study evaluates whether the findings of study 1 can be replicated in a new

independent sample. Although this is considered as a crucial step in the refinement of

assessment instruments, it is often neglected (Smith & McCarthy, 1995). To verify whether

the factorial model of study 1 could be replicated in study 2, we used multi-group analysis

(Byrne, 1998; Jöreskog & Sörbom, 1996; Marsh, 1994). We further examined the internal

consistency of the constructs, correlations between ratings of parents and children and

correlations between scales of the GPBS, the PSI and the CBCL.

Method

Participants

Participants consisted of 175 mothers (mean age = 39.2 years, SD = 4.0), 175 fathers

(mean age = 41.1 years, SD = 4.6) of 175 families and two of their children between 7 and 14

years old (mean age = 10.6 years, SD = 1.8). The sample included 155 male target children

and 195 female target children. There were 91.4% original families, 8% newly composed

families and 0.6% single-parent families.

Materials

The same materials were used as in study 1: the Ghent Parental Behavior Scale, the

Dutch version of the Parenting Stress Index (De Brock et al., 1992) and the Achenbach Child

Behavior Checklist (Verhulst et al., 1996).

Procedure

Students selected 175 families as part of an assignment in (Clinical) Developmental

Psychology at the Ghent University. They were allowed to choose a family they knew well.

Data were collected from mothers and fathers and from two children in the family. Both

parents (if they lived together) independently completed the questionnaires at home, in the

presence of a student.

36 Chapter 1

Results

Factor Structure

The solution of study 1 was the target for confirmatory factor analyses with LISREL

8.52 within each of the four groups. For the model it was hypothesized that: (a) the 45 (44)

items should form nine prescribed factors; (b) each item should have a substantial nonzero

loading on the target factor and zero loadings on the non-target factors; (c) the factors were

free to correlate; (d) the unique variances of the items were uncorrelated. The results of the

factor analyses (completely standardized solution) are given in Table 1. The factor loadings

were substantial and mostly above .40 with some exceptions: eight loadings were between

.30 and .40 (three from positive parenting, four for the monitoring factor and one for the rules

factor). The values of RMSEA are around .05 and the CFI’s were all above .92, indicating

that the CFA-models show an acceptable fit.

Factorial Invariance

To test the stability of the factor model of study 1 a multi-group analysis was

conducted (Byrne, 1998; Jöreskog and Sörbom, 1996; Marsh, 1994). Because this analysis

requires independent samples, we can only test whether the factorial structure is invariant for

mothers of study 1 and 2, for fathers, children about mothers and children about fathers. The

first step created a baseline model with no equality constraints between parameter estimates

of study 1 and study 2 (i.e., the non-invariant model). The chi-square value and df of this non-

invariant model are equal to the sum of the chi-square values and df's of the two independent

groups. The second step was to constrain the factor loadings (lambda's) to be equal in both

groups (Λ invariant). The third step was to set both factor loadings and correlations between

factors (phi coefficients) equal in both groups (ΛΦ invariant) and in the fourth step factor

loadings, factor correlations and unique variances (theta-delta's) were set equal in both groups

(ΛΦΘδ invariant). The goodness-of-fit results are given in Table 4. The RMSEA values were

(with one exception) lower than .05 for all steps in each group. The values of CFI were all

above .90. In the last three columns, the changes in chi-square, df and the p-values are

reported, showing significant changes in several steps. For example, the change in chi-square

and df for mothers from a non-invariant model to a model with factor loadings constrained to

be equal in both samples was 87.1 with df = 45. This is a significant change and would lead

to rejection of the hypothesis that the factor loadings are invariant. For each of the four

models we tested the difference of the factor loadings between study 1 and study 2 by means


of chi-square difference tests. We found 4 significant differences for fathers, 5 for mothers, 0

for children about mothers and 3 for children about fathers. In fact we performed 45 + 45 +

44 + 44 (= 178) post hoc tests with 12 significant results (8 at the .05 level, 3 at the .01 level

and 1 at the .001 level). The items constituting the 12 significant results are mostly not

consistent across models. Because we did 178 post hoc tests, one may expect 9 loadings to be

significant (at a significance level of 5%) by chance. Techniques like MANOVA solve this

problem with correction procedures like Newman-Keuls, Scheffé or Bonferroni, implicating a

lowering of the significance level. In our case it is likely that all 12 significant results are in

fact not significant when corrected for capitalization on chance, and hence we better ignore

the significant chi-square results. This is in line with the conclusion by Marsh (1994) that

tests of statistical significance of the difference between two nested models have essentially

the same strengths and weaknesses as the chi-square test applied to any model. An apparently

more appropriate question is whether the lack of invariance is sufficiently small to justify the

conclusion that the parameter estimates are reasonably invariant across groups (Marsh, 1994).

The goodness-of-fit measures (RMSEA and CFI) showed hardly any differences between the

non-invariant models and the constrained models. Hence the conclusion is that the factor

loadings, factor correlations as well as unique variances are reasonably invariant between

study 1 and study 2 for each of the four groups.

The factor correlations (see Table 2) showed the same three groups of interrelated

factors: (a) Positive parenting, Monitoring, Rules and Autonomy (b) Discipline, Harsh

punishment and Ignoring and (c) Inconsistent discipline and Rewarding. We tested the two-

dimensional structure of the six scales as mentioned in study 1. Again the same factor

structure could be replicated. The completely standardized factor loadings varied between .5

and .8. Positive parenting, Rules and Autonomy loaded on the support factor and Discipline,

Ignoring and Harsh punishment on the negative control factor. Chi-square values for the four

groups ranged from 30.5 to 61.0 with df = 8, RMSEA varied between .09 and .13 and CFI

between .90 and .92. Cronbach's alpha for the support dimension (with 21 items) varied

between .85 and .89 and for the negative control dimension (14 items) between .83 and .84.

Internal Consistency

Cronbach’s alphas are mostly above .80 for Positive parenting, Setting rules,

Discipline and Harsh punishment. Ignoring and Material rewarding have alpha's around .70.

In accordance with study 1, Monitoring, Inconsistent discipline and Autonomy have the

lowest reliabilities.

38 Chapter 1

Agreement between Ratings of Parents and Children

As in study 1, Pearson correlations were positive and significant for Positive

parenting, with r = .23 (p < .001) and r = .28 (p < .001) for ratings of maternal and paternal

behavior respectively, for Monitoring with r = .13 (p < .05) and r = .22 (p < .001), for

Setting rules, with r = .14 (p < .01) and r = .27 (p < .001), for Discipline, with r = .46 (p <

.001) and r = .37 (p < .001), for Inconsistent discipline, with r = .16 (p < .001) and r = .17 (p

< .001), for Harsh punishment with r = .25 (p < .001) and r = .28 (p < .001), for Ignoring with

r = .22 (p < .001) and r = .20 (p < .01), for Material rewarding with r = .26 (p < .001) and r =

.29 (p < .001) and for Autonomy with r = .23 (p < .001) and r = .22 (p < .001).

Validity: Correlations with Criterion Variables

Ghent Parental Behavior Scale - Parenting Stress Index. In study 2, comparable but

somewhat weaker relationships existed as in study 1 (see Table 3), between the positive

parenting skills and stress in parenting (r ranging from .02 to r = -.26). These results

confirmed the findings of study 1.

Ghent Parental Behavior Scale - Child Behavior Checklist. The correlations in study

2 (Table 2) were similar to those in study 1. However, the negative correlation between

Positive parenting and Externalizing behavior of the child was somewhat weaker (r = .12, p <

.05). For fathers no significant relationship was found (r = -.05, p > .05). In study 1, we also

found that higher Material Rewarding is correlated with higher problem behavior (between r

= .12, p < .05 and r = .22, p < .001). The correlations had the expected signs: adequate

parental behavior was negatively related with problem behavior, negative parental skills had

a negative relationship with problem behavior. The relationships for Autonomy and

Monitoring with problem behavior were neither clearly negative nor positive.

General discussion

The Ghent Parental Behavior Scale is a questionnaire, designed to assess parenting

constructs based on Social Learning Theory. This theoretical framework enables to carve

parenting into well-defined constructs referring to observable parental behavior. In this article

we tested (a) the factorial validity of the GPBS, (b) the internal consistency of the identified

scales, (c) agreement between ratings of parents and children, and (d) the construct validity.


Regarding the factorial validity of the GPBS, an evaluation of an experimental

version (Van Leeuwen, 1999) led to the conclusion that the five Patterson constructs (Capaldi

& Patterson, 1989; Patterson et al., 1992) could not be replicated due to the fact that (a) the

content of the discipline (i.e. limit setting) construct appeared to be multidimensional, and (b)

problem-solving and positive involvement were closely related. Dishion, Li, Spracklen,

Brown and Haas (1999) also studied the convergent and discriminant validity of these five

parenting constructs. Their multitrait-multimethod analysis showed that the best-fitting model

allowed only four constructs (limit setting was eliminated) and three method factors (reports

from parents, children and trained observers) (Metzler, Biglan, Ary & Li, 1998).

Refinement of the constructs leading to the second version of the GPBS, enabled us to

find a solid factor structure with nine scales: Autonomy, Discipline, Positive parenting, Harsh

punishment, Monitoring, Rules, Ignoring unwanted behavior, Material rewarding, and

Inconsistent discipline. The nine components were largely replicable across different raters,

i.e. parents and children, and in independent samples by means of confirmatory factor

analyses. A second-order factor analysis detected a positive and a negative dimension in the

parent and child data. These dimensions are regularly mentioned in the literature about

parenting (Gallagher, 2002; Maccoby & Martin, 1983, Ten Haaf, Janssens & Gerris, 1994),

the first describing the affective nature of the parent-child relationship, indicated by showing

involvement and providing support. The second dimension refers to parental efforts to

influence the child’s behavior, such as setting and enforcing standards of behavior. We

labelled them ‘support’ (consisting of the scales positive parenting, rules and autonomy), and

‘negative control’ (consisting of the scales discipline, ignoring and harsh punishment).

The internal consistency, as indicated by Cronbach's alphas, was moderate to good for

the majority of the nine scales. However, scales with a small number of items, for example

autonomy and inconsistent discipline, had lower alpha values. Initially, each scale consisted

of at least five items, but this number diminished because only items with substantial primary

factor loadings were retained. Alpha values for the two dimensions support and negative

control indicate good internal consistency.

Our study shows positive correlations between parent and child ratings for all the

scales. The correlations are rather low, but this is consistent with the research literature. For

example, Shelton et al. (1996) report significant parent-child correlations only for

Involvement, Positive parenting and Corporal punishment (ranging from .08 to .28), not for

Poor monitoring/supervision and Inconsistent discipline. Schwarz, Barton-Henry and

Pruzinsky (1985) concluded that adolescents are biased towards viewing their parents as

40 Chapter 1

more similar than they actually are, and that parents are biased towards presenting an overly

positive image of their child-rearing behavior. However, Sessa, Avenevolli, Steinberg and

Morris (2001) could not replicate this: their data ruled out the possibility that parental biases

in self-report are due to a desire to present one’s self too favourably. This implies the increase

of the number of raters, and aggregation of ratings across judges. Consequently, this can be

considered as an argument favoring the use of the child-version of the GPBS. However, this

is only possible when children have acquired the cognitive skills needed to fill out the

questionnaire.

Evidence was provided for the construct validity of the GPBS. The results confirmed

the hypothesized relationships between stress in parenting and the rate of less adequate

parental behavior. Campbell et al. (1996) showed a relationship between on the one hand

maternal reports of negative discipline and negative control and on the other hand family

stress. Irvine et al. (1999) report correlations between depressive feelings of parents and

ineffective parenting practices like coercive, permissive and inconsistent discipline.

Associations between positive parenting and problem behavior were negative, whereas the

more inadequate parenting practices (like harsh punishment or inconsistent discipline) and

externalizing problem behavior were positively related. These results are consistent with

other studies, showing positive correlations between externalizing behavior and punitive

discipline, spanking, physical aggression (Campbell et al., 1996; Jacob et al., 2000; Lengua,

Wolchik, Sandler & West, 2000; Stormshak, Bierman, McMahon, Lengua & Conduct

Problems Prevention Research Group, 2000) and permissive and inconsistent discipline

(Irvine et al., 1999). Negative correlations have been reported between externalizing behavior

problems and (a) positive affect or warmth and involvement (Jacob et al., 2000; Stormshak et

al., 2000), (b) higher monitoring (Irvine et al., 1999; Jacob et al., 2000, (c) higher levels of

rule setting and (d) higher levels of positive reinforcement (Jacob et al., 2000). The lack of a

significant correlation between the Monitoring Scale and problem behavior in our study is

puzzling, and stands in contrast to findings from the literature (Loeber & Stouthamer-Loeber,

1986). However, Frick et al. (1992) also found a moderate association between parental

monitoring and conduct problems. We contend that our monitoring scale taps showing

interest in children for one set of parents, whereas for another set of parents it may reflect

controlling the child’s behavior ('supervision'). The positive correlations between the scales

Positive parenting and Monitoring support this explanation. Jacob et al. (2000) draw a similar

conclusion when discussing that their supervision scale loaded on the control dimension (as

expected) for parent reports, but loaded on the affect dimension for child reports. They


contend that the manner in which parents show supervision, can either be seen as

authoritarian (more structuring and controlling) or authoritative (more communicative).

Another viable explanation is that a positive correlation between problem behavior and

monitoring will be found when monitoring reflects lack of supervision, for instance when

children are playing in the streets, without an adult keeping track of them.

Limitations and future directions

Our initial efforts at developing an instrument to assess parental behavior are

encouraging, but the GPBS should be considered as an instrument ‘in progress’ as more

research is needed. First, some modification of the item-pool is needed in a future version of

the GPBS. Scales with only three items should be supplemented in order to increase their

internal consistency. The Monitoring scale needs more items assessing the act of ‘controlling’

a child’s behavior, in order to prevent the relationship with Positive parenting. It would also

be interesting to add an item asking how frequently a child is left unattended. Second,

consistency between several raters (parents, children, and observers) of parental behavior

should be further examined. Third, more research is needed on the construct validity of the

GPBS, preferably with other assessment methods than questionnaires, such as observations of

real parent-child interactions. Fourth, so far, the application of the GPBS has been limited to

a Dutch speaking, Belgian target population. More cross-cultural research is necessary to

identify if the GPBS is valid across different cultures. Fifth, The GPBS instrument was

initially designed for research purposes, but it could also be used in clinical practice. Future

research is necessary to investigate the discriminative validity of the GPBS for parents with

or without child-rearing problems, especially when used as a diagnostic tool. Sixth, the

number of respondents in the second study is fairly low in comparison to the number of

parameters to be estimated. This can lead to less robust estimates. But because of the stability

of the factor structure in study 2 (in comparison to study 1) for each of the four groups we are

optimistic about the stability in future research. Finally, it would also be interesting to design

a version of the GPBS for parents of younger children. Younger children form an important

target group for prevention of antisocial behavior. Furthermore, Yoshikawa (1994) concludes

from his overview of effective prevention programs for serious problem behavior that

interventions have the best results when children are of preschool age.

A critical note must be added. Although a relationship is suggested between parental

behavior and antisocial behavior of children, it is not the purpose of the GPBS to promote

far-reaching conclusions about the contribution of parents to delinquent or other problem

42 Chapter 1

behavior of their children. We do not want to 'detect' parents in order to 'blame' them for bad

parenting. In the development of antisocial behavior, the ecological point of view has to be

respected (cf. Belsky, 1984; Bronfenbrenner, 1986) with environmental and personal

characteristics in reciprocal interaction. This means that, for instance, personality, genetic

factors, socio-economic factors or stress variables also play an important role. If clinical or

social workers want to identify children at-risk for severe behavioral problems, a wide variety

of assessment measures must be included, a parenting measure being one of them.

It is also important to consider that high or low scores on a subscale do not necessarily

imply inadequate parenting (except for high harsh punishment). For instance, a low level of

monitoring can indicate too little control of the parent, but a high level of monitoring can be

experienced as suffocating for the child and is enhancing feelings of parental mistrust. For a

young adolescent, a high level of monitoring combined with little encouragement for

autonomous behavior, can be inadequate parenting. Therefore it is necessary not only to

consider individual parenting constructs, but also to look for certain patterns of parental

behavior (Vermulst et al., 2002). This also implies that the individuality of each family has to

be taken into account.

Advantages of the GPBS

With previous limitations and cautions in mind, we think that the GPBS is a

promising instrument for the assessment of parenting, with several advantages. First, the

questionnaire is built on social interactional learning theory, which provides a ‘practical’ way

to look at parenting. Items assess real parental behavior instead of the more unreliable

parental attitudes, behavioral intentions, beliefs, self-perceptions or values. The items are

specific enough to be used as part of the (longitudinal) evaluation of interventions such as

parent training. In this respect, we think that specific parenting constructs have an advantage

over 'dimensions' of parenting because they (a) allow more precise definitions, as they are

more uni- than multidimensional in nature, and (b) reflect behavior which is better adjustable.

Second, a parent version and a child version of the GPBS allow to examine parents’ as well

as children’s perceptions on parental behavior. To our knowledge, no other (Dutch)

questionnaires with this possibility are available. Another advantage is that the questionnaire

is easy to administer. And finally, the GPBS allows registering weaknesses as well as

strengths in parental behavior. Therapists, who want to involve parents in the treatment of

their children, can use this differentiated picture to motivate parental participation in therapy.


The above-mentioned advantages are mainly applicable in a clinical context. For

research purposes, the use of the second-order dimensions ‘support’ and ‘negative control’

might be more adequate because (a) they reduce the number of scales by combining them

into more parsimonious constructs, and (b) they constitute more reliable measures for both

parent and child ratings of parenting.

44 Chapter 1

References

Arnold, D. S., O’Leary, S. G., Wolff, L. S., & Acker, M. M. (1993). The Parenting

Scale: A Measure of Dysfunctional Parenting in Discipline Situations. Psychological

Assessment, 5, 137-144.

Belsky, J. (1984). The determinants of parenting: A process model. Child



development: research perspectives. Developmental Psychology, 22, 723-742.

Byrne, B. M. (1998). Structural equation modeling with LISREL, PRELIS and

SIMPLIS: Basic concepts, applications, and programming. Mahwah, NJ: Lawrence Erlbaum

Associates.

Campbell, S. B. (1994). Behavior problems in preschool children: a review of recent

research. Journal of Child Psychology and Psychiatry, 36, 113-149.

Campbell, S. B., Pierce, E. W., Moore, G., Marakovitz, S., & Newby, K. (1996).

Boys' externalizing problems at elementary school age: pathways from early behavior

problems, maternal control, and family stress. Development and Psychopathology, 8, 701-

719.

Capaldi, D., & Patterson, G. R. (1989). Psychometric properties of fourteen latent

constructs from the Oregon Youth Study. New York: Springer-Verlag.

Chamberlain, P. (1994). Family Connections. A social interactional approach.

Volume 5. Eugene: Castalia Publishing Company.

Collins, W. A., Maccoby, E. E., Steinberg, L., Hetherington, E. M., & Bornstein, M.

H. (2000). Contemporary Research on Parenting. The case for nature and nurture. American

Psychologist, 55, 218-232.

Conger, R. D., Elder, G. H., Jr., Lorenz, F. O., Simons, R. L., & Whitbeck, L. B.

(1992). A family process model of economic hardship and influences on adjustment of early

adolescent boys. Child Development, 63, 526-541.

Darling, N., & Steinberg, L. (1993). Parenting style as context: An integrative model.

Psychological Bulletin, 113, 487-496.

Deater-Deckard, K ., Dodge, K. A., Bates, J. E., & Pettit, G. S. (1998). Multiple risk

factors in the development of externalizing behavior problems : group and individual

differences. Development and Psychopathology, 10, 469-493.


Dedrick, R. F., Greenbaum, P. E., Friedman, R. M., Wetherington, C. M., & Knoff, H.

M. (1997). Testing the structure of the Child Behavior Checklist/4-18 using confirmatory

factor analysis. Educational and Psychological Measurement, 57, 306-313.

De Brock, A. J. L. L., Vermulst, A. A., Gerris, J. R. M., & Abidin, R. R. (1992).

NOSI: Nijmeegse Ouderlijke Stress Index. [NPSI: Nijmegen Parental Stress Index] Lisse:

Swets & Zeitlinger.

Dielman, T., Barton, K., & Cattell, R. (1977). Relationships among family attitudes

and child rearing practices. Journal of Genetic Psychology, 130, 105-112.

Dishion, T. J., Li, F., Spracklen, K. M., Brown, G., & Haas, E. (1999). The

measurement of parenting practices in research on adolescent problem behavior: A

multimethod and multitrait analysis. In R. S. Ashery (Ed.), Research meeting on drug abuse

prevention through family interventions, NIDA Research Monogram 177. (pp. 260-293).

Farrington, D. (1995). The development of offending and antisocial behavior from

childhood: Key findings from the Cambridge Study in Delinquent Development. Journal of

Child Psychology and Psychiatry, 36, 929-964.

Frankel, F. (1993). A brief test of parental behavioral skills. Journal of Behavior

Therapy and Experimental Psychiatry, 10, 29-34.

Frick, P. J. (1991). The Alabama Parenting Questionnaire. Unpublished instrument.

University of Alabama.

Frick, P. J., Lahey, B. B., Loeber, R., Stouthamer-Loeber, M., Christ, M. A. G., &

Hanson, K. (1992). Familial risk factors to oppositional defiant disorder and conduct

disorder: parental psychopathology and maternal parenting. Journal of Consulting and

Clinical Psychology, 60, 49-55.

Gallagher, K. C. (2002). Does child temperament moderate the influence of parenting

on adjustment. Developmental Review, 22, 623-643.

Gerbing, D. W., & Hamilton, J. G. (1996). Viability of exploratory factor analysis as a

precursor to confirmatory factor analysis. Structural Equation Modeling, 3, 62-72.

Gerris, J. R. M., van Boxtel, D. A. A. M., Vermulst, A. A., Janssens, J. M. A. M., van

Zutphen, R. A. H., & Felling, A. J. A. (1992). Child-rearing and family in the Netherlands.

Nijmegen: Institute of Family Studies, University of Nijmegen.

Gerris, J. R. M., Houtmans, M. J. M., Kwaaitaal-Roosen, E. M. G., Schipper, J. C.,

Vermulst, A. A., & Janssens, J. M. A. M. (1998). Parents, adolescents and young adults in

Dutch families: a longitudinal study. Nijmegen: Institute of Family Studies, University of

Nijmegen.

46 Chapter 1

Grolnick, W. S., Weiss, L., McKenzie, L., & Wrightman, J. (1996). Contextual,

cognitive, and adolescent factors associated with parenting in adolescence. Journal of Youth

and Adolescence, 1, 33-54.

Groot, A. de, Koot, H. M., & Verhulst, F. C. (1994). Cross-cultural generalizability of

the Child Behavior Checklist cross-informant syndromes. Psychological Assessment, 6, 225-

230.

Grotevant, H. D., & Carlson, C. I. (1989). Family assessment: a guide to methods and

measures. New York: The Guilford Press.

Hayduk, L. A. (1996). LISREL: Issues, debates and strategies. Baltimore: Johns

Hopkins University Press.

Holden, G. W., & Edwards, L. A. (1989). Parental attitudes toward child rearing:

instruments, issues and implications. Psychological Bulletin, 106, 29-58.

Irvine, A. B., Biglan, A., Smolkowski, K., & Ary, D. V. (1999). The value of the

parenting scale for measuring the discipline practices of parents of middle school children.

Behaviour Research and Therapy, 37, 127-142.

Jacob, T., Moser, R. P., Windle, M., Loeber, R., & Stouthamer-Loeber, M. (2000). A

new measure of parenting practices involving preadolescent and adolescent-aged children.

Behavior Modification, 24, 611-634.

Jöreskog, K., & Sörbom, D. (1996). LISREL 8. User's Reference Guide. Chicago:

Scientific Software International, Inc.

Kaplan, D. (2000). Structural Equation Modeling. Foundations and Extensions.

Thousand Oaks: Sage Publications.

Kline, R.B. (1998). Principles and Practice of Structural Equation Modeling. New

York: The Guilford Press.

Langemeijer, K., Pijnenburg, H., & Veerman, J. W. (1997). Standaardisatie van de

gezinsdiagnostiek. Een inventarisatie van Nederlandstalige gezinsdiagnostische

instrumenten. [Family Assessment. Inventory of Dutch Family Assessment Instruments.]

Nijmegen: Stichting de Waarden, afdeling Onderzoek en Ontwikkeling.

Lengua, L. J., Wolchik, S. A., Sandler, I. N., & West, S. G. (2000). The additive and

interactive effects of parenting and temperament in predicting problems of children of

divorce. Journal of Clinical Child Psychology, 29, 232-244.

Loeber, R. (1990). Development and risk factors of juvenile antisocial behavior and

delinquency. Clinical Psychological Review, 10, 1-41.


Loeber, R., & Stouthamer-Loeber, M. (1986). Family factors as correlates and

predictors of juvenile conduct problems and delinquency. In Morris, N., & Tonry, M. (Eds.).

Crime & Justice: An annual review of research (vol. 7, pp. 29-149). Chicago: University of

Chicago Press.

Maccoby, E. E., & Martin, J. A. (1983). Socialization in the context of the family:

parent-child interaction. In Mussen, P. H. (Ed.), Handbook of child psychology. Vol. IV.

Socialization, personality and social development (pp. 1-101). New York: Wiley.

Marsh, H. W. (1994). Confirmatory factor analysis models of factorial invariance: a

multifaceted approach. Structural Equation Modeling, 1, 5-34.

Marsh, H. W., Balla, J. R., & McDonald, R. P. (1996). Goodness-of-Fit indixes in

confirmatory factor analysis: the Effect of sample size. Psychological Bulletin, 103, 391-410.

Metzler, C. W., Biglan, A., Ary, D. V., & Li, F. (1998). The stability and validity of

early adolescents' reports of parenting constructs. Journal of Family Psychology, 12, 600-

619.

Patterson, G. R. (1982). Coercive Family Process. A social interactional approach.

Volume 3. Eugene: Castalia Publishing Company.

Patterson, G. R., Debarysche, B. D., & Ramsey, E. (1989). A developmental

perspective on antisocial behavior. American Psychologist, 44, 329-335.



Patterson, G. R., & Stouthamer-Loeber, M. (1984). The correlation of family

management practices and delinquency. Child Development, 55, 1299-1307.Reid, J.B.,

Reid, J. B, Patterson, G. R., & Snyder, J. J. (Eds.) (2002). Antisocial behavior in

children and adolescents. A developmental analysis and model for intervention. Washington,

DC: American Psychological Association.

Richman, G. S., Hagopian, L. P., Harrison, K., Birk, D., Ormerod, A., Brierly-

Bowers, P., & Mann, L. (1994). Assessing parental response patterns in the treatment of

noncompliance in children. Child and Family Behavior Therapy, 16, 29-42.

Sessa, F. M., Avenevoli, S., Steinberg, L., & Morris, A. S. (2001). Correspondence

among informants on parenting: preschool children, mothers and observers. Journal of

Family Psychology, 15, 53-68.

Schaefer, E. S. (1965). Children’s reports of parental behavior: An inventory. Child


48 Chapter 1

Schludermann, E., & Schludermann, S. (1970). Replicability of factors in children’s

report of parent behavior (CRPBI). Journal of Psychology, 76, 239-249.

Schwarz, J. C., Barton-Henry, M. L., & Pruzinsky, T. (1985). Assessing Child-rearing

Behaviors: a comparison of ratings made by mother, father, child and sibling on the CRPBI.


Shelton, K. K., Frick, P. J., & Wootton, J. (1996). Assessment of parenting practices

in families of elementary school-age children. Journal of Clinical Child Psychology, 25, 317-

329.

Smith, G. T., & McCarthy, D. M. (1995). Methodological considerations in the

refinement of clinical assessment instruments. Psychological Assessment, 7, 300-308.

Snyder, J., & Stoolmiller, M. (2002). Reinforcements and coercion mechanisms in the

development of antisocial behavior: The Family. In J.B. Reid, G.R. Patterson, & J. Snyder

(Eds.), Antisocial behavior in children and adolescents. A developmental analysis and model

for intervention (pp. 65-100). Washington DC: American Psychological Association.

Stormshak, E. A., Bierman, K. L., McMahon, R. J., Lengua, L. L., & Conduct

Problems Prevention Research Group (2000). Parenting practices and child disruptive

behavior problems in early elementary school. Journal of Clinical Child Psychology, 29, 17-

29.

Stormshak, E. A., Speltz, M. L., Deklyen, M., & Greenberg, M. T. (1997). Observed

family interaction during clinical interviews: a comparison of families containing preschool

boys with and without disruptive behavior. Journal of Abnormal Child Psychology, 25, 345-

357.

Strayhorn, F. M., & Weidman, C. S. (1988). A parent practices scale and its relation

to parent and child mental health. Journal of the American Academy of Child and Adolescent

Psychiatry, 27, 613-618.

Ten Haaf, P. G. J., Janssens, J. M. A. M., & Gerris, J. R. M. (1994). Child-rearing

measures: convergent and discriminant validity. European Journal of Psychological

Assessment, 10, 111-128.

Van Leeuwen, K. (1999). Het meten van opvoeding met de Schaal voor Ouderlijk

Gedrag. [Measurement of parenting with the Parental Behavior Scale]. Diagnostiek-Wijzer,

3, 151-170.

Van Leeuwen, K. (2000). Deficits in parenting skills as an indicator of behavior

problems with children and youth. Development of a screening instrument for the Flemish

community. Unpublished manuscript,Ghent University, Department of Psychology.


Verhulst, F. C., Van der Ende, J., & Koot, H. M. (1996). Handleiding voor de

CBCL/4-18. [Manual for the CBCL/4-18]. Erasmus Universiteit Rotterdam, Afdeling Kinder-

en Jeugdpsychiatrie.

Vermulst, A. A., de Brock, A. J. L. L., & Gerris, J. R. M. (2002). Nijmeegse

Opvoedingsvragenlijst (NOV). Handleiding. [The Nijmegen Child Rearing Questionnaire.

Manual]. Nijmegen: Institute of Family Studies, University of Nijmegen.

Yoshikawa, H. (1994). Prevention as Cumulative Protection: Effects of Early Family

Support and Education on Chronic Delinquency and Its Risks. Psychological Bulletin, 115,

28-54.

50 Chapter 1

Table 1

Primary Factor Loadings, Fit Indices and Cronbach's alpha's for Study 1 and Study 2 Study 1 Study 2 Factors + Items M F CM CF M F CM CF Positive parenting pos1 .53 .63 .58 .57 .61 .57 .50 .58

pos2 .72 .75 .72 .68 .70 .79 .64 .71 pos3 .56 .59 .58 .55 .49 .60 .51 .48 pos4 .72 .76 .71 .65 .62 .78 .62 .69 pos5 pos6

.57 .67 .59 .60 .53 .70 .55 .50

.43 .52 .49 .48 .35 .37 .57 .49 pos7 .49 .52 .53 .54 .42 .42 .50 .50 pos8 .68 .76 .62 .64 .65 .78 .64 .62 pos9 .49 .58 .58 .60 .46 .36 .64 .60 pos10 pos11

.58 .57 .63 .58 .42 .54 .54 .55

.59 .66 .64 .66 .56 .58 .58 .59 alpha .85 .87 .87 .86 .81 .86 .84 .85 Monitoring mon1 .71 .80 .51 .40 .72 .84 .57 .48 mon2 .83 .72 .71 .56 .91 .86 .72 .68 mon3 .34 .46 .62 .50 .38 .50 .49 .54 mon4 .33 .45 .43 .40 .34 .42 .38 .32 alpha .63 .72 .65 .52 .67 .74 .61 .56 Rules rul1 .49 .72 .81 .64 .67 .76 .76 .73 rul2 .53 .68 .78 .72 .52 .70 .75 .71 rul3

rul4 .61 .62 .68 .64 .56 .66 .60 .55

.68 .72 .80 .75 .80 .81 .72 .73 rul5 .62 .70 .69 .66 .54 .61 .68 .66 rul6

rul7 .65 .66 .56 .61 .55 .68 .35 .44

.62 .71 *) *) .70 .79 *) *) alpha .80 .85 .87 .83 .81 .88 .80 .80 Discipline dis1 .61 .70 .71 .71 .72 .75 .74 .70 dis2

dis3 .64 .60 .77 .70 .67 .72 .71 .74

.61 .68 .66 .66 .72 .67 .62 .61 dis4 .48 .57 .53 .48 .50 .63 .58 .51 dis5 .77 .80 .83 .83 .77 .84 .82 .78 dis6- .44 .47 .35 .47 .60 .42 .44 .44 alpha .76 .79 .81 .80 .82 .83 .81 .80 Inconsistent discipline inc1

inc2 .74 .69 .56 .59 .84 .75 .60 .64

.66 .59 .61 .45 .58 .62 .55 .54 inc3 .43 .43 .55 .52 .53 .53 .50 .55 alpha .64 .58 .60 .52 .68 .66 .56 .60

Harsh punishment har1 har2

.73 .78 .78 .76 .69 .73 .66 .71 .80 .81 .87 .83 .69 .81 .81 .87 har3

har4 .59 .57 .62 .53 .72 .58 .59 .67

.74 .73 .82 .75 .74 .75 .71 .78 alpha .81 .81 .85 .81 .80 .81 .78 .84 Ignoring ign1

ign2 .70 .63 .66 .54 .72 .69 .65 .48

.51 .46 .55 .51 .62 .52 .61 .60 ign3

ign4 .82 .85 .76 .66 .81 .82 .77 .60

.59 .50 .63 .53 .63 .62 .55 .64 alpha .75 .69 .75 .64 .78 .75 .74 .67 Material rewarding rew1

rew2 .68 .68 .80 .66 .82 .78 .72 .71

.57 .62 .60 .63 .50 .55 .56 .55 rew3 .75 .76 .82 .70 .79 .87 .79 .72 alpha .70 .72 .78 .70 .73 76 .72 69 Autonomy aut1

aut2 .57 .58 .69 .52 .70 .74 .54 .63

.80 .76 .82 .71 .78 .85 .75 .71 aut3 .40 .58 .54 .61 .50 .54 .51 .55 alpha .61 .65 .72 .65 .69 .74 .64 .66 Fit indices χ2 2119.8 1782.4 1923.2 1765.1 1725.5 1908.0 1472.2 1566.2 df 909 909 866 866 909 909 866 866 RMS AE

CFI .048 .043 .046 .044 .051 .055 .044 .047

.92 .96 .96 .96 .92 .93 .95 .94 Note. *) This item is not part of the child version; M=mothers; F = fathers; C = child

Table 2

Factor correlations for Study 1 (below diagonal) and Study 2 (above diagonal) POS MON RUL DIS INC HAR IGN REW AUT

M F CM CF M F CM CF M F CM CF M F CM CF M F CM CF M F CM CF M F CM CF M F CM CF M F CM CF

POS .22 .49 .52 .36 .46 .47 .65 .68 -.08 .11 .12 .10 -.23 -.08 .17 .13 -.35 -.12 -.12 -.11 -.20 -.15 .01 .21 .05 .01 .31 .36 .51 .52 .52 .71

MON

.29 .44 .41 .55 .19 .52 .31 .54 .24 .21 .31 .36 -.03 .01 .36 .23 .11 .01 .11 .08 .10 .15 .39 .41 .02 .14 .40 .46 .23 .31 .31 .44

RUL .56 .69 .67 .68 .41 .50 .46 .58 .11 .27 .26 .34 -.14 .01 .26 .10 -.10 -.07 -.06 -.02 -.01 -.01 .12 .20 -.02 .05 .18 .27 .35 .42 .71 .72

DIS -.01 .08 .10 .04 .18 .22 .40 .40 .29 .27 .27 .36 .08 .11 .17 .12 .37 .40 .52 .56 .39 .49 .45 .47 .33 .42 .17 .17 .09 .08 .14 .15

INC -.22 -.02 .27 .20 -.10 .10 .18 .33 -.05 .07 .20 .32 .07 .08 .13 .24 .21 .05 .09 .08 .23 .28 .31 .29 .18 .46 .31 .30 -.25 -.10 .10 .23

HAR -.19 -.20 -.08 -.08 .00 -.07 .26 .23 .04 -.06 .05 .12 .44 .38 .65 .59 .11 .06 .08 .10 .41 .48 .53 .47 .27 .15 .10 .03 -.09 -.01 -.25 -.17

IGN -.13 -.02 .00 -.03 .10 .18 .35 .22 .01 .08 .11 .16 .32 .34 .61 .45 .27 .31 .19 .40 .35 .26 .53 .46 .38 .29 .24 .37 -.05 -.08 .12 .20

REW .12 .17 .43 .51 .10 .13 .28 .43 .02 .13 .24 .41 .23 .34 .13 .20 .33 .37 .30 .28 .12 .03 .04 .11 .21 .32 .16 .20 -.02 -.10 .16 .15

AUT .37 .61 .58 .60 .20 .18 .33 .44 .28 .55 .61 .70 -.07 .12 .14 .18 -.24 .05 .22 .36 -.05 -.07 -.10 -.05 -.05 -.03 .12 .11 -.05 .09 .24 .36

POS = Positive parenting; MON = Monitoring; RUL = Rules; DIS = Discipline; INC = Inconsistent discipline; HAR = Harsh punishment;

IGN = Ignoring; MRE = Material rewarding; AUT = Autonomy; M = mothers; F = fathers; CM = children about mothers; CF = children about fathers

Table 3

Pearson Correlations between PSI, CBCL, and GPBS (Study 1 and Study 2)

Measure POS MON RUL DIS INC HAR IGN MRE AUT Study 1

M F M F M F M F M F M F M F M F M F

PSI

Parent domain -.35 a -.30 a -.02 .02 -.20 a -.22 a .13 b .05 .26 a .18 a .15 a .13 b .23 a .25 a .06 .04 -.14 a -.16 a

Child domain

-.34 a -.29 a -.01

.04

-.16 a -.17 a .19 a .16 a .24 a .14 b .23 a .19 a .31 a .27 a .09 c .02

-.09 c -.16 a

CBCLInternalizing -.17 a -.17 a -.01 .02 -.09 c -.10 c .07 .06 .13 a .08 .12 b .01 .12 b .15 a .05 .01 -.07 -.10 c

Externalizing -.28 a -.27 a -.03 .04 -.09 c -.10 c .23 a .25 a .24 a .12 b .26 a .21 a .21 a .22 a .08 c .06 -.03 -.10 c

Study 2

SI PParent domain -.26 a -.16 b .06 .00 -.16 b -.10 .13 c .18 a .31 a .22 a .18 a .17 b .36 a .30 a .15 b .27 a -.05 -.10Child domain

-.15 b -.14 c .09

.04

-.11 c .02

.23 a .18 a .27 a .22 a .20 a .18 a .29 a .30 a .25 a .19 a -.08

-.07 CBCL

Internalizing -.14 b -.09 -.06 -.06 -.12 c -.02 .13 c -.02 .16 b .17 b .15 b -.01 .22 a .09 .18 a .12 c -.07 -.04Externalizing -.12 c -.05 .05 -.01 -.05 -.01 .27 a .14 b .20 a .22 a .28 a .13 c .26 a .17 b .20 a .14 b -.04 .04

M = Mothers; F = Fathers; POS = Positive parenting; MON = Monitoring; RUL = Rules; DIS = Discipline; INC = Inconsistent discipline;

HAR = Harsh punishment; IGN = Ignoring; MRE = Material rewarding; AUT = Autonomy a p ≤ .001 ; b p ≤ .01 ; c p ≤ .05 (2-tailed)


Table 4

Goodness of Fit of CFA's of Study 1 and 2 for Each Group

χ2 df RMSEA CFI χ2diff

df diff p

Mothers

noninvariant model 3845.3 1818 .050 .921

Λ invariant 3932.4 1863 .049 .919 87.1 45 <.001

ΛΦ invariant 3993.0 1899 .049 .918 60.6 36 <.01

ΛΦΘδ invariant 4358.3 1944 .051 .905 365.3 45 <.001

Fathers


Λ invariant 3774.3 1863 .048 .948 83.8 45 <.001

ΛΦ invariant 3839.4 1899 .048 .947 65.1 36 <.01

ΛΦΘδ invariant 4031.9 1944 .048 .943 192.5 45 <.001

Children about mothers


Λ invariant 3457.3 1776 .045 .955 62.0 44 <.05

ΛΦ invariant 3506.1 1812 .045 .955 48.8 36 ns

ΛΦΘδ invariant 3601.1 1856 .045 .954 95.0 44 <.001

Children about fathers


Λ invariant 3416.8 1776 .045 .950 60.8 44 <.05

ΛΦ invariant 3479.3 1812 .045 .950 62.5 36 ns

ΛΦΘδ invariant 3517.3 1856 .044 .949 38.0 44 ns

54 Chapter 1

Appendix

Ghent Parental Behavior Scale - Parent version

(Translated from Dutch)

A : Instructions

On the following pages you will find some statements about handling your child. Read each statement

carefully. Indicate for each statement how frequent you use this way of handling your child. You can

choose from the following answer possibilities:

never rarely sometimes often always

Mark with a cross the answer category of your choice. You can choose only one answer for each

statement. For instance:

never rarely sometimes often always

I ask my child which books he/she likes to read

Keep in mind that your answer always is related to one and the same child.

It is possible that you think about some statements: 'I should like to do it different'. Nevertheless,

indicate how you act in reality. There are no good or wrong answers.

Please do not skip any items.

B : Items (normally presented in a quasi random order)

Positive parenting

pos1 I make time to listen to my child, when he/she wants to tell me something

pos2 When my child seems to have a problem, I discuss with him/her what is wrong

pos3 In the evening I talk with my child about the past and the coming day

pos4 When my child has a problem, we look together at different possible solutions

pos5 I ask my child about his/her hobbies and interests

pos6 I make excursions together with my child

pos7 I compliment my child when he/she spontaneously helps me out (for instance with laying the table)

pos8 When my child an I have a disagreement, we talk it over and we look together for a solution

pos9 I do activities together with my child, because I know that my child likes it (for instance playing a round game, shopping together)

pos10 I give my child a compliment, hug or a tap on the shoulder as a reward for good behavior

pos11 When I see my child after a day of school, I make it possible to spend some time with him/her

Monitoring

mon1 I keep track of the friends my child is seeing

mon2 I keep track of the neighbourhoods my child visits

mon3 When my child went out somewhere on his/her own, I inquire if he/she has actually been there

mn4 I ask my child how he/she spends his/her pocket money

Rules

rul1 I teach my child to be polite at school

rul2 I teach my child to obey rules

rul3 I teach my child to adapt to the habits in our family

rul4 I teach my child to adapt to rules at school or at work


rul5 I teach my child to handle his/her things with respect

rul6 * I teach my child respect for the authorities

rul7 I teach my child that it is important to behave properly

Discipline

dis1 When my child doesn't obey a rule (for instance: he/she comes home late without a valid reason; he/she has not completed a chore),

then I punish him/her

dis2 I punish my child, when he/she makes a nuisance of him/herself (for instance because he/she nags, contradicts me, lies, argues).

dis3 When my child has done something wrong, I punish him/her by taking away something nice (for instance the child can't watch TV,

isn't allowed to go out, has to be home earlier, has to go to bed earlier)

dis4 When my child has been misbehaving, I give him/her a chore for punishment

dis5 When my child does something that I don't want him/her to do, I punish him/her

dis6 (-) It happens that I don't punish my child after he/she has done something that is not allowed

Inconsistent discipline

inc1 When my child doesn't obey a rule, it happens that I threaten with a punishment, but that in the end I don't carry it out anyway

inc2 When I have punished my child, it happens that I let my child out of the punishment early

inc3 Before I eventually give a punishment, I have told my child many times that I would punish his/her behavior

Harsh punishment

har1 I slap my child when he/she has done something wrong

har2 I spank my child when he/she is disobedient or naughty

har3 I shake my child when we have a fight

har4 I spank my child when he/she doesn't obey rules

Ignoring

ign1 When my child does something that is not allowed, I give him/her an angry look and pretend he/she is not there

ign2 When my child does something that is not allowed, I only talk to him/her again when he/she behaves better

ign3 When my child does something that is not allowed, I give him/her an angry look and I ignore him/her afterwards

ign4 When my child does something that is not allowed, I don't talk to him/her until he/she says sorry

Material rewarding

rew1 I give my child money or a small present when he/she has done something that I am happy about

rew2 When my child has done his/her best, I allow something extra (for instance staying up later)

rew3 I let my child buy something when he/she has done something well

Autonomy

aut1 I teach my child to solve his/her own problems

aut2 I teach my child to take his/her own decisions

aut3 I teach my child that he/she is responsible for his/her own behavior

* This item is not included in the child version

The Utility of Resilient, Overcontrolled and Undercontrolled Personality Types 57 as predictors of problem behavior

CHAPTER 2

A LONGITUDINAL STUDY OF THE UTILITY OF THE

RESILIENT, OVERCONTROLLED AND UNDERCONTROLLED

PERSONALITY TYPES AS PREDICTORS OF CHILDREN’S AND

ADOLESCENTS’ PROBLEM BEHAVIOR

Abstract

This study addresses the utility of the Resilient, Overcontrolled and Undercontrolled

personality prototypes for predicting children’s and adolescents’ problem behavior in

a longitudinal general population sample (N=491). Both parents and one child

participated in the study at two measurement occasions separated by a three-year

interval. The major objective was to examine whether the prototypes predict different

clinical patterns, as reflected by mean-level differences on the internalizing and the

externalizing dimensions of the CBCL and the YSR. Prototype membership was

derived from cluster analyzing parental ratings of personality and adolescent self-

rated personality. All three types could only be recovered from the adolescent self-

ratings of personality. Although the prototypes showed clear differential relationships

with measures of internalizing and externalizing problem behavior, hierarchical

multiple regression analyses showed that personality types do not predict adolescent

problem behavior beyond what is predicted by personality dimensions.

58 Chapter 2

Introduction

From observations in everyday life, it can be concluded that children and

adolescents show striking differences in their tendencies to behave, think, and feel.

Some children are sociable, others shy, some are easily distressed, others even-

tempered. For psychologists it is interesting to study these differences in personality,

especially because they may have predictive validity for child and adolescent

development (Shiner, 2003). In this respect, it is useful to consider different

perspectives for describing individual differences. Research on individual differences

can be conducted in two ways, i.e. variable-centered or person-centered. The variable-

centered approach focuses on differences on dimensions across subjects, with

variables as analytical units, such as the Big Five personality dimensions. The person-

centered approach studies ‘types’, referring to patterns of characteristics within

individuals. In other words, types group individuals with similar personality patterns.

Bergman and Magnusson, (1997) state that, although at a detailed level there is an

infinite variety of characteristics and states in a person, at a global level there are only

a small number of more frequently observed “typical patterns”. The profiles that

emerge from person-oriented studies are believed to contain extra information, above

the result of studying its parts separately (Bergman, 2001; Bergman & El-Khouri,

2001). Differentiating between a person-centered approach and a trait or variable

approach does not imply that these two perspectives are contradictory. Hofstee (2002)

and Mervielde and Asendorpf (2000) argue that they should be treated as

complementary.

Currently there is a renewed interest in the person-centered approach to

personality (cf. Asendorpf, Caspi & Hofstee, 2002), not because the variable-centered

approach has failed, but due to the recently emerging consensus about the variables

that should be included in a person-centered analysis (Mervielde & Asendorpf, 2000).

The present study corroborates to this line of research by (a) searching for replication

of three prototypes in children and adolescents, i.e. resilients, overcontrollers and

undercontrollers, (b) by investigating the external validity and utility of these

personality types as predictors of problem behavior in children and adolescents, and

(c) by examining the incremental validity of the type approach beyond what is gained

from the dimensional perspective.


Replication of the personality types Resilients, Overcontrollers and Undercontrollers

The question whether individuals from the population can be assigned to three

prototypes, depends on (a) the trait dimensions that are used, (b) the method for

deriving empirical types, (c) the operationalization of the traits (e.g. self-ratings

versus ratings by observers), and (d) the population characteristics (culture, age)

(Asendorpf, Borkenau, Ostendorf & van Aken, 2001).

Regarding the trait dimensions used to derive types, it can be argued that a

different number of types will be found when the patterns are based on many or on a

few traits, or that different types will emerge from different sets of trait dimensions

(Asendorpf et al., 2001). In line with the current trend in personality research, the

types in the present study are based on the Five-Factor Model (FFM) of Personality.

This model represents the personality traits that are most relevant to social adaptation

(Caspi, 1998) i.e. Neuroticism, Extraversion, Openness to experience, Agreeableness

and Conscientiousness. In various fields of research, this model has become

acknowledged as a robust reference-model of personality, substantially improving

communication among researchers (De Raad & Perugini, 2002). Moreover, the FFM

represents personality not only in adults but also in children and adolescents, and

hence facilitates comparisons across developmental periods (Shiner, 1998; Shiner &

Caspi, 2003). Finally, the FFM has proven to be clinically useful, showing meaningful

relationships with Internalizing and Externalizing problems (Caspi, 1998; John, Caspi,

Robins, Moffitt & Stouthamer-Loeber, 1994) and with the DSM-IV axis II personality

disorders (Costa & Widiger, 2002; De Clercq & De Fruyt, 2003).

The three types can be described in terms of FFM scores as follows: (a)

resilients score average on the socially adjusted Big Five characteristics

Agreeableness, Extraversion, Openness to experience and Conscientiousness, and low

on Neuroticism, (b) overcontrollers have high levels of Neuroticism and low levels of

Extraversion, and finally (c) undercontrollers score below the mean on Agreeableness

and Conscientiousness. The type labels are based on Block and Block’s (1980)

dimensions Ego control, i.e. the tendency to control or express motivational and

emotional impulses, and Ego resilience, i.e. the possibility to respond flexibly rather

than rigidly to situational demands.

In order to derive types, two empirical methods have been used: Q-factor

analysis and cluster analysis. In the first method, a judge sorts trait descriptions for

each subject into a category distribution, according to how well they fit the subject’s

60 Chapter 2

personality. Subsequently, a Q-factor analysis (also called inverse factor analysis)

classifies the Q-sort profiles into similar patterns (Asendorpf & van Aken, 1999).

Robins, John, Caspi, Moffitt and Stouthamer-Loeber (1996) were the first to derive

the three prototypes in 13-year old boys using this method. The types were further

replicated in several heterogeneous samples of children (Asendorpf & van Aken,

1999; Hart, Hofmann, Edelstein & Keller, 1997; Van Lieshout, Haselager, Risken-

Walraven & van Aken, 1995).

The second method derives types from individual’s ratings on personality

variables. Profile patterns of individuals are grouped by cluster analysis into relatively

homogeneous clusters, with the mean profile of the cluster members serving as the

prototypical pattern describing the cluster (Asendorpf & van Aken, 1999). Evidence

for the three personality types in adults has been found by Schnabel, Asendorpf and

Ostendorf (2002), with NEO-PI-R and NEO-FFI data in German samples, by Boehm,

Asendorpf and Avia (2002) in a Spanish student sample, and by Barbaranelli (2002)

in an Italian sample. Additional evidence has been found in Dutch and Flemish

samples of children and adolescents (Dubas, Gerris, Janssens & Vermulst, 2002; De

Fruyt, Mervielde & Van Leeuwen 2002) based on parental ratings or adolescent self-

reports of personality.

However, not all studies were successful in replicating the types. Costa,

Herbst, McCrae, Samuels and Ozer (2002) could only replicate the types in three of

four large and diverse American adult samples and concluded that the types are not

robust. Boehm et al. (2002) did not find the three prototypes in a Spanish general

population sample. Moreover, De Fruyt et al. (2002) failed to replicate the clusters

across different FFM measures in one and the same sample. Barbaranelli (2002)

further questioned the number of types suggesting a fourth cluster, i.e. a ‘non-

desirable cluster’, in which the negative aspects of both over- and undercontrolled

types are combined.

The third issue that determines the replicability of the types concerns the

operationalization of the traits. Until now, researchers diverged in their assessment of

the FFM by using different instruments, such as the NEO-inventories developed by

Costa and McCrae (1992), the California Child Q-set (Block & Block, 1978), the

Hierarchical Personality Inventory for Children (HiPIC; Mervielde & De Fruyt, 1999)

and lists of adjectives (Goldberg, 1992; Ostendorpf, 1990). Several studies have

provided evidence for the consistency of the types when different assessment


measures were used (Asendorpf & van Aken, 1999; Asendorpf et al., 2001;

Barbaranelli, 2002; De Fruyt et al., 2002; Dubas et al., 2002; Hart et al., 1997; Robins

et al., 1996; Schnabel, et al., 2002; Van Lieshout et al., 1995).

A subject that has received little attention in studying types, is the effect of the

informant on the personality ratings. Until now, most studies that replicated types

used independent self- and other-ratings (parents, teachers, …). As Asendorpf et al.

(2001) indicate, there may be differences in the prototypes with regard to self- versus

other-ratings. For example, low conscientiousness may be more pronounced in other-

ratings of undercontrollers than in self-ratings, because low conscientiousness may be

more salient and may have more negative consequences for others than for the self.

Recently, Rammstedt, Riemann, Angleitner and Borkenau (in press) examined the

replicability of the types across informants (self-, peer, and behavioral ratings) in an

adult twin sample. Only for self-reports the three types were identified, whereas for

peer and other ratings, only the resilient type was consistently identified across raters.

This study clearly demonstrates that the replicability of types may depend on the type

of informant.

Finally, only a few studies have investigated the continuity and stability of the

three types from a longitudinal perspective. Continuity refers to the constancy of the

prototype pattern across different measurement times in a particular sample, not to the

stability of the individual’s membership of a type category. Stability refers to the

constancy of type membership from one time to another and is measured by

comparing type classification at different assessment periods. Subsequently,

classifications of subjects are cross-classified, using Cohen’s kappa as a measure of

stability. It should be noted that stability can be low, despite a high continuity

(Asendorpf et al., 2001; Asendorpf & van Aken, 1999). There is some evidence for

short-term stability in adulthood (Asendorpf et al., 2001), indicating retest reliability

of the prototypes. For children, Asendorpf and van Aken (1999) found a high

continuity in types over a 5-year period, but a rather low stability, indicating that

many children changed from one type to another between ages 4 and 10. A

longitudinal twin study (De Fruyt et al., 2002) showed that the types were replicable

across a 3-year interval, i.e. providing evidence for continuity, but that prototype

membership was again rather unstable.

62 Chapter 2

Incremental Validity of Personality Types for the Prediction of Problem Behavior in

Children and Adolescents

External validity of the types has been investigated by relating types to

measures of social functioning or problem behavior. These studies are important

because they examine the utility of types as predictors of (mal)adaptive child and

adolescent development. Robins et al. (1996) described resilients as intelligent,

successful in school, unlikely to be delinquents and relatively free of

psychopathology, whereas overcontrollers were characterized as prone to

internalizing problems and undercontrollers as at risk for academic, behavioral and

emotional problems. Huey and Weisz (1997) found a positive association between

ego undercontrol and externalizing problems in children, and a negative association

with internalizing problems. Externalizers were characterized as impulsive extraverts

who were disagreeable but low in neuroticism, internalizers as overcontrolled

introverts who were agreeable and conscientious. Asendorpf et al. (2001) showed that

overcontrollers were more shy, less sociable, more lonely, had a lower self-esteem,

and fewer or less secure relationships with others; undercontrollers were judged as

less sociable (although more sociable than overcontrollers), more aggressive and as

experiencing more interpersonal conflicts.

Several arguments can be pointed out recommending the use of types in

research and clinical practice. First, types are conceptually convenient because they

summarize many traits in a single label (Costa et al., 2002), hence facilitating

communication in personality description (Asendorpf et al., 2001). They invite

researchers and professionals to think about the dynamic interplay of combinations of

variables, contrary to the traditional dimensional perspective considering only one

variable at a time. The person-oriented approach is further valuable for developmental

psychology. Adopting a theoretical rather than a methodological person-oriented

view, Bergman and Magnusson (1997), demonstrate this with an empirical example

regarding the evolution of children’s patterns of externalizing behavior. They show

that there is a strong tendency for problems to (a) not occur at all, or (b) for two or

more to occur together. Singular adjustment problems in childhood do not predict

adult problems. Developmental studies should be directed at detecting variables that

co-vary and at explaining developmental paths. Longitudinal trajectories permit to

assess continuity and discontinuity of individual differences over time. There is

evidence that type membership at age 3 predicts personality at age 19 (Caspi and


Silva, 1995) and interpersonal functioning in adulthood (Newman, Caspi, Moffitt, &

Silva, 1997). Hart et al. (1997) showed that each of the personality types has a

coherent pattern of relationships with adolescent development. Children initially

assigned in a longitudinal study to the resilient, overcontrolled, or undercontrolled

types, developed along different developmental trajectories. Finally, types are also

useful in clinical psychology, both at the level of prevention and intervention. For

purposes of prevention, the types can be used to screen at-risk children in a non-

referred population. It is important to note that maladaptive types (i.e. overcontrollers

and undercontrollers) are derived from raw scores on FFM measures comprising

items that describe the normal range of child behavior instead of maladaptive

behavior. The preponderance of items referring to maladaptive behavior in problem

behavior measures, such as the CBCL, sometimes provokes resistance in subjects or

informants to accurately complete these measures.

The distinction between the type- and the variable approach implies a choice

between two alternatives: losing information on interindividual variation between

types (in the type-approach) or preserving information on personality structure (in the

variable-approach). Consequently, it is interesting to study how the type and variable

approach concurrently influence outcome variables, i.e. the issue of incremental

validity (Asendorpf, in press).

Costa et al. (2002) were the first to compare the predictive power of the

categorical types versus the continuous FFM variables with regard to psychosocial

functioning and clinical symptoms. They entered dummy-coded types and variables in

different steps of an hierarchical regression analysis and examined both the overall

predictive power, and the incremental R² of the types over the dimensions and vice

versa. Their study showed that types did predict the outcome variables, but they

contributed nothing above or beyond what was already predicted by the FFM

dimensions, whereas the FFM variables had significant incremental validity over the

types. However, Asendorpf (in press) notes that a cross-sectional head-to-head

comparison of types and dimensions is often unfair, mainly because dichotomized

predictor variables lose information in comparison with the continuous personality

variables. Therefore, he contrasted three kinds of predictions, i.e. by type, by

continuous dimensions and by dichotomizing the dimensions through a median split.

This fairer comparisons revealed that the type approach, applied in a longitudinal

64 Chapter 2

study, was as predictive as a continuous dimensional approach but was generally less

effective in cross-sectional studies.

The Present Research

The present research examines the consistency, stability, and external validity

of the three types. In addition, the incremental validity of types beyond FFM

dimensions for predicting adolescents’ problem behavior will be examined. In Study

1, the replicability of the three prototypes is evaluated, (a) based on the dimensions of

the FFM in line with the current research trend (Asendorpf, 2002); (b) using the same

method for deriving types as proposed by Asendorpf et al. (2001), i.e. through a

particular combination of cluster analyses; (c) across different measures and

informants, i.e. parental ratings and adolescent self-ratings of personality, and (d) over

time, i.e. using follow-up measures over a three-year interval. In Study 2, the external

validity of the types will be examined through their association with internalizing and

externalizing problem behavior, and personality ratings provided by a different

informant. Finally, the incremental validity of the person-centered approach for

predicting problem behavior is investigated relative to the variable-centered approach.

Study 1

Method

Participants and procedure

The present study uses data from a follow-up study4 investigating the

relationship among parenting, personality characteristics and children’s problem

behavior, at two assessment occasions separated by a 3-year interval. Subjects were

parents and one, non-referred child, recruited via randomly selected elementary and

secondary schools. For elementary schools we stratified the sample by province, (East

and West Flanders), region (rural or urban), school type (public/private/catholic

schools) and grade (third, fourth, fifth and sixth year of elementary school). For

secondary schools, subject sampling was based on province (East and West Flanders),

type of curriculum (vocational, technical and general education) and grade (first and

4 Type classification of the sample at the first measurement moment (N = 539), has been reported in De

Fruyt, Mervielde & Van Leeuwen (2002)


second year of secondary school). Permission to contact parents for this study was

granted by 80% of the elementary schools and 60% of the secondary schools. Schools

that refused were replaced by additional randomly selected schools. All parents were

informed about the objectives and procedures of the research project. The response

rate for parents with children in primary schools was 41%, and 39% for parents with

children in secondary schools. At Time 2, 82% of the families continued

collaboration; 12% refused and 6% could not be reached. This sampling method

resulted in a well- balanced sample regarding social-economic status, gender and age

(Van Leeuwen, 2000). At each assessment time, families were visited at home by a

trained psychology student who instructed the mother, father and child to

independently complete a series of questionnaires. Parents described the child’s

personality and problem behavior at both measurement moments and the children

provided self-ratings of their personality and problem behavior at Time 2. Only

subjects who participated at both Time 1 and Time 2 were included for the present

analyses, resulting in a sample of 491 families with 234 boys and 257 girls. The mean

age of the children 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.

Measures

Parents rated their child’s personality on the Hierarchical Personality

Inventory for Children (HiPIC; Mervielde & De Fruyt, 1999) at both assessment

occasions. The HiPIC includes 144 items, assessing 18 facets hierarchically structured

under the Big Five personality domains Conscientiousness (including the facets

Achievement motivation, Concentration, Perseverance, Orderliness), Benevolence

(Egocentrism, Irritability, Compliance, Dominance and Altruism), Extraversion

(Shyness, Optimism, Expressiveness and Energy), Emotional stability (Anxiety and

Self-confidence) and Imagination (Creativity, Curiosity and Intellect). The

Benevolence domain is conceptually and empirically associated with Agreeableness

of the FFM, whereas Imagination is related to the Openness to experience domain of

the FFM (De Fruyt, Mervielde, Hoekstra, & Rolland, 2000). All HiPIC items were

rated on a 5-point Likert scale.

The robust factor structure and good internal consistency of the HiPIC have

been reported in previous studies with clinical and non-clinical samples (Mervielde &

De Fruyt, 2002). For the present analyses paternal and maternal ratings were

66 Chapter 2

averaged, because of their strong convergence for the 18 facet scales, with

correlations ranging from .51 to .77 (p < .001) and a median correlation of .60 at Time

1, and a range from .47 to .75 (p < .001) and a median correlation of .56 at Time 2.

Adolescents completed a Dutch shortened version of Goldberg’s (1992)

hundred adjectives, entitled the Questionnaire Big Five (QBF) at Time 2. The QBF

assesses the Big Five personality domains, with ratings on 7-point Likert scales for 30

adjectives, 6 per domain (Gerris et al., 1998). Previous research has demonstrated that

self-ratings on this adjective set provide a valid adolescent Big Five personality

profile (Scholte, van Aken & van Lieshout, 1997; Dubas et al., 2002).

Results

Structural validity and internal consistency of the personality measures

Research on replicable prototypes depends on the consistency at the level of

personality dimensions (Boehm et al., 2002), i.e. the different inventories should

provide valid ratings on the different FFM dimensions. In order to evaluate the

structural validity of the two personality questionnaires, we ran principal axis

factoring analyses followed by oblimin rotation.

Table 1 shows that at Time 1 and Time 2 the primary factor loadings for most

of the HiPIC facets were above .50. Some facets had high secondary loadings (equal

to or above .30), but these findings are consistent with other studies (Mervielde & De

Fruyt, 2002). The Cronbach alpha’s of the domain scales ranged from .76 to .88.

Table 2 shows that the QBF items loaded the five intended factors, with all secondary

loadings below .30. The Cronbach alpha’s of the QBF scales ranged from .68 to .88,

indicating moderate to good internal consistency.

Consistency and stability of prototypes across time and measures

In order to replicate the types, we followed the method described by

Asendorpf et al. (2001). In a first step, hierarchical cluster analyses according to

Ward’s method were conducted, using raw domain scores. The resulting three-cluster

solutions of these analyses were then used as initial cluster centers in a

nonhierarchical K-means clustering. Table 3 presents the cluster centers after

transformation to z-scores, allowing better comparability of the results. The

undercontrolled cluster, with expected below average mean scores (about one half of

a standard deviation below the mean) for Conscientiousness and Agreeableness, was


replicated for HiPIC measures at Time 1 and Time 2, and for the QBF data. The

overcontrolled cluster, with expected above average mean scores for Neuroticism and

below average mean scores for Extraversion, was clearly present in the QBF

personality reports of the children. A clear and pronounced overcontrolled pattern

could not be retrieved in the HiPIC data, although the Time 1 z-values resembled the

pattern better than the Time 2 z-values. The resilient cluster, with below average mean

scores for Neuroticism and average scores on the socially desirable traits

Extraversion, Imagination, Agreeableness and Conscientiousness, was also present in

the three datasets. The z-scores of the socially desirable traits however, were

substantially higher than average.

Gender differences were tested with the chi-square test. Based on the QBF

data, 50% of the boys were classified as undercontrolled, 27.6% as overcontrolled and

22.4% as resilient. Of the girls, 36.3% were identified as undercontrolled, 32.8% as

overcontrolled and 30.9% as resilient. Differences between boys and girls were

significant with χ² (2, N = 484) = 9.57, (p = .008). We further examined which

prototypes showed gender differences, by comparing the sex proportion per prototype

with the sex proportion of the aggregated remaining prototypes. There were more

resilient girls than boys (χ² (df = 1) = 4.43, p = .035), more undercontrolled boys than

girls (χ² (df = 1) = 9.21, p = .002), but there was no gender difference for

overcontrollers (χ² (df = 1) = 1.53, p = .216).

Consistency of the types across samples was assessed by cross-classification

of the original clusters in sample 1 and the original clusters of sample 2, evaluated

with Cohen’s Kappa coefficient. Kappa’s range between -1 and 1, with 0 referring to

judgmental independence. Kappa’s above .60 are considered as acceptable by

Asendorpf et al. (2001). Cross-classification of cluster assignment in parental ratings

at Time 1 and Time 2 (HiPIC measures), resulted in a kappa of .47. Cross-

classification of parental and adolescent cluster assignment at Time 1 demonstrated a

kappa of .15, and .12 at Time 2. These are the highest kappa’s when all possible

cross-classifications of clusters were tested. Kappa’s ranged from -.26 to .47 for

cross-classifications between parent ratings Time 1 and Time 2, from -.11 to .15 for

cross-classifications between parent ratings Time 1 and adolescent ratings Time 2;

and from -.09 to .12 for cross-classifications between parent ratings Time 2 and

adolescent ratings.

68 Chapter 2

The kappa value of .47 for the cross-classification of cluster assignment in

parental ratings at Time 1 and Time 2 (HiPIC measures), suggests low stability over

time. In contrast, the variable approach did show high stability over time. Pearson

correlations between HiPIC Time 1 and Time 2 measures were all significant, with r

= .76 (p < .001) for Conscientiousness, with r = .75 (p < .001) for Benevolence, with

r = .72 (p < .001) for Extraversion, with r = .68 (p < .001) for Emotional Stability and

with r = .73 (p < .001) for Imagination.

Comparison of the cluster centers of the HiPIC measures at Time 1 and Time

2 showed some similarities. In both samples, only the resilient and undercontrolled

types were replicated, suggesting partial continuity.

The cross-classifications of parental and adolescent cluster assignments at

Time 1 and Time 2 indicate low consistency across measures and judges. This was

also concluded from comparisons of cluster centers of the different measures. In

contrast, there was evidence for moderate agreement between different raters at the

variable level. Correlations between parental and adolescent ratings at Time 2 were all

significant, with r = .20 (p < .001) for Benevolence, r = .46 (p < .001) for

Conscientiousness, r = .37 (p < .001) for Neuroticism, r = .46 (p < .001) for

Extraversion and r = .34 (p < .001) for Imagination. Also, the correlations between

adolescent ratings and parental ratings at Time 1 were significant, with r = .13 (p <

.01) for Benevolence, r = .11 (p < .05) for Conscientiousness, r = .28 (p < .001) for

Neuroticism, r = .33 (p < .001) for Extraversion and r = .27 (p < .001) for

Imagination.

Conclusion

Although both the HiPIC and the QBF can be considered as good FFM

markers, the three prototypes were only partly replicable. Undercontrolled and

resilient types were found in parental HiPIC and adolescent QBF ratings, whereas the

overcontrolled type was only replicable in the adolescent personality reports. This

implies that for the study of the external and incremental validity of the types (Study

2), only the type classification based on the adolescent QBF ratings can be used.

Study 1 showed some continuity of the types, because the same types, i.e.

overcontrolled and resilient, were found in the parental ratings at Time 1 and Time 2.

Stability of the types was low, indicated by small kappa’s between parental ratings at

Time 1 and Time 2, contrary to the high variable-centered stability indicated by the


highly significant correlations between HiPIC Time 1 and Time 2 measurements.

Consistency of types between parental HiPIC and adolescent QBF ratings was low.

Study 2

In Study 2, the external validity of the three prototypes, derived from

adolescent personality reports, was investigated by examining how they are associated

with internalizing and externalizing problem behavior, based on adolescent self-

reports as well as parent reports on the one hand, and on the other hand with FFM

ratings provided by parents. The incremental validity of the person-centered approach

is evaluated by comparing the relative utility of types and traits in predicting

adolescent problem behavior.

Method

Participants, procedure, and personality measures were identical to those of

Study 1, but only those persons were retained who participated at both Time 1 and

Time 2 assessments.

Child problem behavior

Parents rated their child’s problem behavior with the Dutch version of the

Child Behavior Checklist (CBCL; Verhulst, Van der Ende & Koot, 1996), a screening

instrument for behavioral and emotional problems in children. Parents rated the

frequency of problem behaviors on a 3-point Likert scale. Two broadband factors can

be derived from the CBCL scales: Internalizing problem behavior, with items

referring to somatic complaints, social withdrawal, and anxiety/depression, and

Externalizing problem behavior, including items related to aggression, hyperactivity,

and delinquency. There is substantial evidence regarding the CBCL’s test-retest-

reliability and its cross-cultural construct validity (Dedrick, Greenbaum, Friedman,

Wetherington & Knoff, 1997; De Groot, Koot, & Verhulst, 1994; Verhulst et al.,

1996). Ratings provided by mothers and fathers were averaged for all analyses.

Correlations between maternal and paternal ratings were all significant (p < .001),

with .53 for internalizing behavior, and .58 for externalizing behavior.

70 Chapter 2

Children were administered the Dutch version of the Youth Self Report (YSR;

Verhulst, Van der Ende & Koot, 1997), a parallel measure of the CBCL, providing

self-reports on internalizing and externalizing problem behavior.

Results

Associations between prototypes and measures of problem behavior and personality

traits

ANOVA’s were run with the 3-cluster solution of the QBF data as the

independent variable and internalizing, externalizing and total problem score of the

CBCL (T1 & T2) and the YSR as the dependent variables.

Problem behavior. Table 4 shows several significant F-tests, indicating differences

between the three prototypes. Post-hoc analyses of parental ratings of problem

behavior showed that overcontrollers scored significantly higher on internalizing

problem behaviors than undercontrollers (p < .01) and resilients (p < .001) at Time 1

and at Time 2 (p < .001). These findings also extended to self-ratings of internalizing

behavior. When children rated their own internalizing behavior, all three subtypes

significantly differed from each other, with overcontrollers scoring higher than

resilients and undercontrollers (p < .001), and undercontrollers scoring higher than

resilients (p < .01).

For externalizing behavior, undercontrollers scored significantly higher than

resilients for CBCL reports, but only at Time 2 (p < .05). For YSR externalizing

behavior, undercontrollers scored significantly higher than overcontrollers (p < .05)

and resilients (p < .001).

For the total problem scale, undercontrollers scored significantly higher than

resilients for CBCL ratings (p < .05) at Time 2 and for YSR ratings (p < .001).

Overcontrollers scored significantly higher than resilients for CBCL ratings at Time 1

(p < .05), Time 2 (p < .01) and for YSR ratings (p < .001).

Big Five Personality. Post hoc tests for significant ANOVA results showed higher

scores on Neuroticism for overcontrollers than for undercontrollers (p < .001 at both

times) and resilients (p < .001 at both times), and significantly lower scores on

Extraversion for overcontrollers compared to undercontrollers and resilients (p < .001

in all cases) for parental ratings of children’s personality. Significantly lower scores

were obtained on Conscientiousness for undercontrollers compared to overcontrollers

(p < .001 at both times) and resilients (p < .01 at Time 1 and p < .001 at Time 2). The


expected lower scores on Agreeableness for the undercontrollers were only found at

Time 2 (p < .01). Resilients had the highest scores on the more socially desirable

domains, although differences were not always statistically significant.

For children’s self-ratings of personality the expected differences were found,

with all post hoc tests significant at the p < .001 level. Overcontrollers had

significantly lower scores on Emotional stability and Extraversion than

undercontrollers and resilients. Undercontrollers had significantly lower scores than

overcontrollers and resilients on Agreeableness and Conscientiousness. Resilients had

the highest scores on all the socially desirable trait dimensions and the lowest score on

Neuroticism. They always differed significantly from the under- and overcontrollers.

Incremental validity of the person-centered approach for the prediction of problem

behavior

The incremental validity of the person-centered approach in predicting

problem behavior was examined through hierarchical multiple regression analyses

(HMRA; see Table 5). Type membership, based on QBF adolescent ratings, was

translated in two dummy-coded variables. When three types are distinguished, two

dummy-coded variables capture all information on differences between the types

(Asendorpf, in press; Costa et al. 2002). Outcome variables were parental and

adolescent ratings of internalizing and externalizing problem behavior at Time 2.

When the outcome variable was rated by parents, parental five-factor measures

(HiPIC) were used in the HMRA; when the outcome variable was

externalizing/internalizing behavior as rated by the adolescents, adolescent five-factor

measures (QBF) were used. Two kinds of models were analyzed: in Model A, the two

dummy-coded types were entered in Block 1 and five-factor variables in Block 2; in

Model B, the five-factor variables were entered in Block 1 and the two dummy-coded

types in Block 2.

When entered as Block 1 of the HMRA (Model A), the types were significant

predictors of the outcome variables with values of R² between .02 and .13 for both

adolescent and parental ratings. In model A, the five-factor measures explained

additional variance (F with p < .001), with values of R² change between .20 and .54

for both parental and adolescent ratings.

When entered in Block 2 of the HMRA (Model B), the types did not explain

additional variance on top of the five-factor measures in the case of parental ratings.

72 Chapter 2

With adolescent self-ratings, the two dummy-coded types explained significant

additional variance for both externalizing and internalizing behavior with R² changes

of .01, although the increase was limited in magnitude.

General discussion

In this study we examined the replicability across time, raters and different

FFM measures of the resilient, undercontrolled and overcontrolled personality

prototypes in a general population sample of Flemish adolescents. In addition, the

external validity of the types was assessed with measures describing problem

behavior, as well as their incremental validity relative to trait dimensions as predictors

of clinical criteria. Compared to previous work on the type-approach, this study used

multiple FFM measures, adopting the same method for deriving types as

recommended by Asendorpf et al. (2001), but examining different age groups (7-15

and 10-18) and using different judges (parental versus adolescent self-ratings). The

longitudinal design with a considerable time interval of 3-years further enabled the

investigation of type continuity and type membership stability.

Type continuity across measures and informants

All three personality types could be derived from adolescent self-ratings on a

short adjective list, but only the undercontrolled and resilient prototypes could be

inferred from parental HiPIC ratings. The overcontrolled prototype was not clearly

identifiable at both Time 1 and Time 2. The resulting third clusters seem to group

both overcontrolled and average children, explaining the large number of adolescents

assigned to this cluster and the more extreme z-values of the socially desirable traits

of the resilient type. The resilient type grouped the smallest percentage of subjects

(about 27%), in contrast to other studies, where the largest percentage of individuals

is usually classified as resilient. Almost 50% of the subjects were typed as resilient in

the study of Asendorpf et al. (2001) and Dubas et al. (2002), while Boehm et al.

(2002) and Schnabel et al. (2002) found about 40% of resilients. Significantly more

girls than boys were typed as resilient in our study and significantly more boys than

girls were typed as undercontrollers. The results with respect to type-sex differences

reported in the literature are mixed. Asendorpf and Van Aken (1999) and Asendorpf

et al. (2001) report similar findings, but no sex differences were found in the study of


Boehm et al. (2002), whereas Schnabel et al. (2002) reported that more men than

women were typed as resilient, and more women than men were typed as

undercontrolled. In sum, our study contributed to the evidence with respect to the

replicability of the resilient and the undercontrolled prototypes across samples, FFM

measures and informants. The overcontrolled pattern could not be identified from

parental ratings in this sample. In addition, the number of subjects assigned to a

prototype and the sex-distribution across types turned out to be largely sample-

specific.

It is hard to disentangle in this study whether the difficulty to replicate types

can be attributed to differences in FFM measures or raters, because parents and

children in the present study also used different FFM indicators. However, failure to

replicate the overcontrolled type cannot be attributed to parental ratings, the HiPIC, or

the combination of both, because the three types were replicated in a larger dataset (N

= 539), available at the first measurement moment (De Fruyt et al., 2002). Boehm et

al. (2002) also concluded that clustering solutions are highly sensitive to sample

composition: adding as few as 26 subjects had notable effects on cross-study

consistency. Therefore it is likely that not the type of measure or informant, but rather

the composition of the sample determined the likelihood of finding the three

prototypes.

Type continuity across time and type membership stability

Parental ratings of children’s personality at both assessment occasions enabled

the evaluation of type continuity and type consistency across time. At both moments,

only the resilient and undercontrolled types were replicated, indicating moderate

continuity of types across time. However, our analyses indicate low stability in type

membership in childhood and adolescence, despite substantial correlations between

Time 1 and Time 2 HiPIC scales, indicating substantial differential stability over 3

years. Stability of trait configurations in individuals is also referred to as ipsative

stability. Our study suggests that personality during childhood and adolescence is

differentially stable, but that the evidence for ipsative stability is rather weak.

Asendorpf et al. (2001) do not expect high type stability between childhood

and adulthood, because the stabilities obtained in variable-centered analyses, although

significant, are usually also low. They also point out that low stabilities probably

indicate that the type-approach is sensitive to developmental changes, considering

74 Chapter 2

sufficient retest reliabilities. Bergmann and Magnusson (1997) even warn for the old

typological thinking in which persons are assigned to permanent classes and

emphasize that a type pattern can change in the developmental course. Perhaps a more

differentiated typology may eventually identify types with a greater cross-time

stability. Although Asendorpf et al. (2001) encourage the search for subtypes in order

to create a more differentiated classification system, it is not clear which method

should be used to reach this goal. A more detailed system that takes into account the

diversity of personalities in a sample also increases the chances for isolating small

groups that are hard to replicate in other samples (Caspi, 1998). Moreover, given that

the robustness of the three-type model is still debatable, it seems premature to try to

establish a consensus on a more fine-grained typology.

The method adopted in this study to assign individuals to types was chosen to

enhance comparability with previous work. However, this method has some

limitations. For example, it is typical for Ward’s method to produce clusters of

similar size. Bergman (2001) also remarks that the more conventional classification

methods classify all of the subjects, although a complete classification is unrealistic.

Bergman and Magnusson (1997) recommend searching for ‘dense points’ in the

multidimensional space, in order to identify a homogeneous class of subjects with a

typical pattern, without forcing all the subjects into an a-priori specified number of

types.

External and incremental validity

Predictive validity is ultimately one of the most important criteria to evaluate

the usefulness of psychological constructs, in this case personality types. In addition,

it is also essential to demonstrate that the type-approach predicts criteria of interest

beyond what can be predicted by means of already well-established perspectives, such

as the variable approach. The fact that the three prototypes show differential

relationships with measures of internalizing and externalizing problem behavior

confirms their external validity. The overcontrolled, undercontrolled and resilient

prototypes derived from adolescent ratings also demonstrated the presumed

relationships even when using parental ratings of problem behavior and five-factor

personality. Overcontrollers are characterized by internalizing behavior and higher

neuroticism, undercontrollers by more externalizing behavior, lower

conscientiousness and lower agreeableness, whereas resilients always scored low on


problem behavior and high on the socially desirable trait dimensions. The external

validity of the three prototypes derived from measures assessing the adaptive range of

individual differences seems to be robust and can be clearly linked to clinical profiles

and dimensions used by researchers and practitioners.

The second study demonstrated that overall five factor measures better

predicted adolescent problem behavior than types. Entered first or second in a

hierarchical multiple regression analysis, types only explained a small part of the

variance of the outcome variables. These findings suggest that the type perspective is

probably less interesting for researchers working with large samples and mainly

interested in prediction and in the covariance structure of variables. Choosing between

the type- and the variable approach implicates considering two alternatives: losing

information on interindividual variation between types (adopting the type-approach)

versus preserving information on personality structure (adopting the variable-

approach). A loss in interindividual variation, or restriction of range, is by definition a

disadvantage for predictive studies, and hence it is less likely that the person-centered

approach will outperform the variable-approach when accurate prediction is the main

objective. However, when professionals have to make decisions about an individual,

information of type membership may be necessary and indeed very informative and

useful. Professionals are often required to take decisions at the level of the individual,

judging a single case by combining and weighing information on several variables.

Knowledge about trait patterns within individuals is then indispensable for an

adequate interpretation of the available diagnostic information. De Fruyt (2002), for

example, in a follow-up study of college undergraduates entering the labor market,

demonstrated that some types were considerably more at risk to be unemployed. As it

stands now, there is evidence for three to four interesting prototypes, i.e. resilients,

over-, and undercontrollers, and individuals combining over and under-controlled

characteristics. Moreover several studies, including this one, have revealed clinical

correlates of the different prototypes for both children and adolescents. Given that the

replicability of the types tends to be sample dependent it is perhaps advisable to put

more effort in combining research on the replicability of types with assessment of

external validity within specific or clinical samples.

76 Chapter 2

References

Asendorpf, J. B. (2002). Editorial: The puzzle of personality types. European

Journal of Personality, 16, S1-S5.

Asendorpf, J. B. (in press). Head-to-head comparison of the predictive validity

of personality types and dimensions. European Journal of Personality, 17.

Asendorpf, J. B., Borkenau, P., Ostendorf, F., & van Aken, M. (2001).

Carving personality description at its joints: confirmation of three replicable

personality prototypes for both children and adults. European Journal of Personality,

15, 169-198.



Asendorpf, J. B., & van Aken, M. A. G. (1999). Resilient, overcontrolled, and

undercontrolled personality prototypes in childhood: replicability, predictive power,

and the trait-type issue. Journal of Personality and Social Psychology, 77, 815-832.

Barbaranelli, C. (2002). Evaluating cluster analysis solutions: an application to

the Italian NEO Personality Inventory. European Journal of Personality, 16, S43-S55.

Bergman, L. R. (2001). A person approach in research on adolescence: some

methodological challenges. Journal of Adolescent Research, 16, 28-53.

Bergman, L. R., & El-Khouri, B. M. (2001). Developmental processes and the

modern typological perspective. European Psychologist, 6, 177-186.

Bergman, L. R., & Magnusson, D. (1997). A person-oriented approach in

research on developmental psychopathology. Development and Psychopathology, 9,

291-319.

Block, J., & Block, J. H. (1980). The California Child Q-set. Palo Alto, CA:

Consulting Psychologists Press. (Original work published 1969).

Block, J. H., & Block, J. (1980). The role of ego-control and ego-resiliency in

the organization of behavior. In W. A. Collins (Ed.), Minnesota Symposia on Child

Psychology (Vol. 13, pp. 39-101). Hillsdale, NJ: Erlbaum.

Boehm, B., Asendorpf, J. B., & Avia, M. D. (2002). Replicable types and

subtypes of personality: Spanish NEO-PI samples. European Journal of Personality,

16, S25-S41.

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

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


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

Wiley.

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

personality traits in young adulthood: longitudinal evidence from a birth cohort. Child


Costa, P. T., Herbst, J. H., McCrae, R. R., Samuels, J., & Ozer, D. J. (2002).

The replicability and utility of three personality types. European Journal of

Personality, 16, S73-S87.

Costa, P. T., Jr., & McCrae, R. R. (1992). Revised NEO Personality Inventory

and NEO Five Factor Inventory professional manual. Odessa, FL: Psychological

Assessment Resources.

Costa, P. T., & Widiger, T. A. (Eds) (2002). Personality disorders and the

five-factor model of personality (2nd ed.). Washington, DC, US: American

Psychological Association, 493 pp.

De Clercq, B., & De Fruyt, F. (2003). Personality disorder symptoms in

adolescence: A Five-Factor Model perspective. Journal of Personality Disorders, 17,

269-292.

Dedrick, R. F., Greenbaum, P. E., Friedman, R. M., Wetherington, C. M., &

Knoff, H. M. (1997). Testing the structure of the Child Behavior Checklist/4-18 using

confirmatory factor analysis. Educational and Psychological Measurement, 57, 306-

313.

De Fruyt, F. (2002). A person-centered approach to P-E fit questions using a

multiple-trait model. Journal of Vocational Behavior, 60, 73-90.

De Fruyt, F., Mervielde, I., Hoekstra, H. A., & Rolland, J.-P. (2000).

Assessing adolescents' personality with the NEO PI-R. Assessment, 7, 329-345.

De Fruyt, F., Mervielde, I., & Van Leeuwen, K. (2002). The consistency of

personality type classification across samples and five-factor measures. European


De Groot, A., Koot, H. M., & Verhulst, F. C. (1994). Cross-cultural

generalizability of the Child Behavior Checklist cross-informant syndromes.

Psychological Assessment, 6, 225-230.

De Raad, B., & Perugini, M. (2002). Big Five factor assessment: introduction.

In B. de Raad & M. Perugini (Eds.), Big Five Assessment (pp. 1-26). Göttingen:

Hogrefe & Huber Publishers.

78 Chapter 2

Dubas, J. S., Gerris, J. R. M., Janssens, J. R. M., & Vermulst, A. A. (2002).

Personality Types of Adolescents: Concurrent Correlates, Antecedents, and Type X

Parenting Interactions. Journal of Adolescence, 25, 79-92.

Gerris, J. R. M., Houtmans, M. J. M., Kwaaitaal-Roosen, E.M.G., de Schipper,

J. C., Vermulst, A. A., & Janssens, J. M. A. M. (1998). Parents, Adolescents and

Young Adults in Dutch Families: a longitudinal study. Nijmegen: Institute of Family

Studies.

Goldberg, L. R. (1992). The development of markers for the Big-Five factor

structure. Psychological Assessment, 4, 26-42.

Hart, D., Hofmann, V., Edelstein, W., & Keller, M. (1997). The relation of

childhood personality types to adolescent behavior and development: a longitudinal

study of icelandic children. Developmental Psychology, 33, 195-205.

Hofstee, W. K. B. (2002). Types and variables: towards a congenial procedure

for handling personality data. European Journal of Personality, 16, S89-S96.

Huey, S. J., Jr., & Weisz, J. R. (1997). Ego control, ego resiliency, and the

five-factor model as predictors of behavioral and emotional problems in clinic-

referred children and adolescents. Journal of Abnormal Psychology, 106, 404-415.

John, O. P., Caspi, A., Robins, R. W., Moffitt, T. E., & Stouthamer-Loeber,

M. (1994). The “Little Five”: exploring the nomological network of the five-factor

model of personality in adolescent boys. Child Development, 65, 160-178.

Mervielde, I., & Asendorpf, J. B. (2000). Variable-centered and person-

centered approaches to childhood personality. In S. E. Hampson (Ed.), Advances in

Personality Psychology. Volume 1 (pp. 37-76). Philadelphia: Psychology Press.

Mervielde, I., & De Fruyt, F. (1999). Construction of the Hierarchical

Personality Inventory for Children (HiPIC). In I. Mervielde, I. Deary, F. De Fruyt, &

F. Ostendorf (Eds.), Personality psychology in Europe (pp. 107-127). Tilburg

University Press.

Mervielde, I., & De Fruyt, F. (2002). Assessing children’s traits with the

Hierarchical Personality Inventory for Children. In B. de Raad & M. Perugini (Eds.),

Big Five Assessment (pp. 129-146). Göttingen: Hogrefe & Huber Publishers.

Newman, D. L., Caspi, A., Moffitt, T. E., & Silva, P. A. (1997). Antecedents

of adult interpersonal functioning: effects of individual differences at age 3

temperament. Developmental Psychology, 33, 206-217.


Ostendorf, F. (1990). Sprache und persönlichkeitsstruktur: zur validät des

fünf-faktoren-modells der persönlichkeit [Validity of the five-factor model of

personality). Regensburg, Germany: S. Roderer.

Rammstedt, B., Riemann, R., Angleitner, A., & Borkenau, P. (in press).

Resilients, overcontrollers, and undercontrollers: the replicability of the three

personality prototypes across informants. European Journal of Personality, 17.

Robins, R. W., John, O. P., Caspi, A., Moffitt, T. E., & Stouthamer-Loeber,

M. (1996). Resilient, Overcontrolled, and Undercontrolled boys: three replicable

personality types. Journal of Personality and Social Psychology, 7, 157-171.

Schnabel, K., Asendorpf, J. B., & Ostendorf, F. (2002). Replicable types and

subtypes of personality: German NEO-PI-R versus NEO-FFI. European Journal of


Scholte, R. H. J., van Aken, M. A. G., & van Lieshout, C. F. M. (1997).

Adolescent personality factors in self-ratings and peer nominations and their

prediction of peer acceptance and peer rejection. Journal of Personality Assessment,

69, 534-554.

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

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

Shiner, R. L., & Caspi, A. (2003). Personality differences in childhood and

adolescence: measurement, development, and consequences. Journal of Child

Psychology and Psychiatry, 44, 2-32.

Van Leeuwen, K. (2000). Deficits in parenting skills as an indicator of

behavior problems with children and youth. Development of a screening instrument

for the Flemish community. Unpublished manuscript. Ghent University.

Van Lieshout, C. F. M., Haselager, G. J. T., Risken-Walraven, J. M., & van

Aken, M. A. G. (1995). Personality development in middle childhood. Paper

presented at the Biennial Meetings of the Society for Research in Child Development,

Indianapolis, IN.


CBCL/4-18. [Manual for the CBCL/4-18]. Rotterdam: Afdeling Kinder- en

Jeugdpsychiatrie, Academisch Ziekenhuis /Erasmus Universiteit.


Youth Self-Report (YSR). [Manual for the Youth Self-Report (YSR)]. Rotterdam:

Afdeling Kinder- en Jeugdpsychiatrie, Academisch Ziekenhuis /Erasmus Universiteit.

80 Chapter 2

Table 1

Factor Pattern matrix and Cronbach’s alpha coefficients for the HiPIC measured at

Time 1 and Time 2 Domain/facets CO BE EX ES IM T1 T2 T1 T2 T1 T2 T1 T2 T1 T2 Conscientiousness Achievement motivation .76 .75 .21 -.20 .11 .10 .24 .19 -.28 -.26 Concentration .84 .75 -.05 .10 -.07 -.12 -.18 -.18 -.09 -.15 Perseverance .77 .82 -.13 .06 -.03 .07 -.07 -.08 -.09 .00 Orderliness .80 .80 -.06 .11 -.02 -.02 .09 .07 .13 .11 Benevolence Egocentrism -.11 -.19 .82 -.73 -.12 -.04 .17 .23 .02 .02 Irritability -.20 -.25 .65 -.58 .02 .09 .37 .40 -.01 -.01 Compliance .47 .40 -.60 .66 .13 .02 .09 .01 .02 -.02 Dominance .09 .02 .80 -.76 .29 .38 -.17 -.15 -..04 -.05 Altruism .02 .02 -.47 .58 .53 .43 .22 .22 -.05 -.10 Extraversion Shyness -.05 -.02 -.06 -.08 -.71 -.72 .30 .34 -.05 -.05 Optimism -.01 -.06 -.26 .37 .63 .55 -.12 -.22 -.20 -.21 Expressiveness .00 .09 .13 -.09 .75 .77 .06 .09 -.05 -.03 Energy -.20 -.09 .19 -.18 .38 .47 .06 .03 -.25 -.17 Emotional Stability Anxiety .13 .06 .07 .00 .00 -.01 .92 .92 .00 -.04 Self-confidence .13 .06 .09 -.09 .28 .29 -.66 -.67 -.23 -.27 Imagination Creativity .01 .05 -.04 .13 .22 .25 -.08 .06 -.44 -.49 Curiosity .01 .01 -.04 .02 -.07 -.03 .07 .08 -1.04 -.97 Intellect .33 .20 .05 -.06 -.07 -.15 -.26 -.34 -.54 -.65 Cronbach’s alpha .88 .88 .83 .85 .76 .77 .79 .79 .77 .79 Total % explained variance at T1 = 68.95 Total % explained variance at T2 = 69.95 HiPIC= Hierarchical Personality Inventory for Children (Mervielde & De Fruyt, 1999)

CO= Conscientiousness ; BE = Benevolence ; EX= Extraversion ; ES= Emotional Stability ; IM =

Imagination


Table 2

Factor Pattern matrix and Cronbach’s alpha coefficients for the QBF

AG EX CO RE ES Agreeableness Cooperative .75 .10 -.06 -.06 .05 Kind .71 .03 .01 .02 .02 Helpful .66 .16 -.10 -.02 .08 Pleasant .58 -.24 -.04 .06 -.09 Sympathetic .56 -.19 -.05 .11 -.04 Agreeable .49 -.25 -.02 .11 -.02 Extraversion [-] Withdrawn -.04 .72 -.05 .05 -.13 [-] Reserved .03 .66 .04 .05 -.10 [-] Introverted -.05 .66 .06 .08 -.12 [-] Quiet .07 .60 -.07 -.01 .04 [-] Bashful .02 .48 .04 -.08 -.09 Talkative .16 -.46 .01 .12 -.10 Conscientiousness Organized -.02 -.04 -.90 .04 .00 Neat .03 -.05 -.88 -.05 -.04 [-] Sloppy .13 .11 .86 .09 -.02 Careful .13 .09 -.75 -.05 .01 Thorough .10 .07 -.60 .07 -.03 Systematic .02 .03 -.36 .13 -.01 Resourcefulness Creative .01 .04 -.06 .68 .04 Artistic .02 .11 -.10 .60 .09 Imaginative .08 -.05 .10 .52 .05 Innovative .14 -.15 -.03 .42 -.04 Deep -.11 .03 -.01 .37 -.05 Complex .08 -.09 .00 .34 -.05 Emotional Stability [-] High-strung -.06 -.05 -.08 -.14 -.83 [-] Nervous -.01 -.05 -.04 -.07 -.72 [-] Touchy -.02 .12 .01 .11 -.53 [-] Fearful -.06 .24 .01 .00 -.49 [-] Anxious .22 .15 .00 -.05 -.47 [-] Irritable -.07 -.02 .10 .14 -.42 Cronbach’s alpha .82 .78 .88 .67 .77 Total % explained variance = 43.61 QBF= Questionnaire Big Five (Gerris et al., 1998)

EX= Extraversion ; AG= Agreeableness ; CO= Conscientiousness ; ES= Emotional Stability ; RE=

Resourcefulness

[-] = item to be reversed

82 Chapter 2

Table 3

HiPIC and QBF final cluster centers (z-values) HiPIC time 1 (n=491) HiPIC time 2 (n=491) QBF time 2 (n=484) Clu1 Clu2 Clu3 Clu1 Clu2 Clu3 Clu1 Clu2 Clu3Neuroticism 0.34 0.01 - 0.70 0.29 0.09 - 0.45 - 0.20 0.83 - 0.62 Extraversion - 0.29 0.01 0.57 - 0.21 - 0.12 0.40 0.25 - 0.88 0.59 Openness to experience - 0.28 - 0.30 1.00 - 0.40 - 0.30 0.88 - 0.17 - 0.10 0.38 Agreeableness 0.30 - 0.95 0.75 - 1.14 0.27 0.76 - 0.40 - 0.03 0.67 Conscientiousness 0.15 - 0.98 1.10 - 0.95 - 0.12 1.16 - 0.82 0.35 0.91 N 224 157 110 140 215 136 207 147 130 Cluster % 45.6 32.0 22.4 28.5 43.8 27.7 42.8 30.4 26.8

Note: HiPIC time 1: Cluster 1 = not identifiable; Cluster 2 = Undercontrollers; Cluster 3 = Resilients;

HiPIC time 2: Cluster 1 = Undercontrollers; Cluster 2 = not identifiable; Cluster 3 = Resilients;

QBF time 2: Cluster 1 = Undercontrollers; Cluster 2 = Overcontrollers; Cluster 3 = Resilients


Table 4

Means and standard deviations of CBCL scales of the QBF-based prototypes Cluster 1

U Cluster 2

O Cluster 3

R

M (SD) M (SD) M (SD) F (2, 481) Post hoc a

CBCL - Time 1 Internalizing 53.06 (9.16) 56.22 (8.64) 51.22 (10.05) 10.53 *** U<O, O>R Externalizing 52.27 (9.92) 51.88 (10.29) 50.51 (9.60) 1.29 Total problem behavior 52.40 (10.22) 54.00 (9.95) 50.33 (11.43) 4.23 * O>R CBCL - Time 2 Internalizing 50.79 (8.83) 54.48 (9.96) 49.31 (8.49) 12.30 *** U<O O>R Externalizing 52.87 (10.19) 51.73 (10.05) 49.86 (9.13) 3.73 * U>RTotal problem behavior 51.73 (10.17) 52.87 (10.86) 48.57 (10.06) 6.38 ** O>R U>R YSR - Time 2 Internalizing 51.40 (9.54) 57.54 (10.55) 47.50 (9.91) 36.06 *** U<O O>R U>RExternalizing 53.49 (10.58) 50.68 (9.77) 48.12 (9.13) 11.82 *** U>O U>RTotal problem behavior 53.43 (9.82) 54.94 (10.65) 48.05 (9.67) 17.90 *** O>R U>R HiPIC - Time 1 Neuroticism 5.25 (1.13) 5.89 (1.20) 5.24 (1.12) 16.34 *** U<O O>R Extraversion 14.50 (1.84) 13.45 (2.10) 14.41 (1.69) 14.85 *** U>O O<R Openness to experience 11.05 (1.53) 10.62 (1.72) 11.23 (1.65) 5.23 ** O<R Agreeableness 17.52 (2.40) 17.55 (2.42) 17.83 (2.52) 0.73 Conscientiousness 12.27 (2.37) 13.13 (2.75) 13.69 (2.65) 13.08 *** U<O U<R HiPIC - Time 2 Neuroticism 4.84 (0.97) 5.68 (1.22) 4.79 (0.96) 34.35 *** U<O, O>R Extraversion 13.97 (1.81) 12.71 (1.92) 14.19 (1.63) 28.91 *** U>O,

10.56 (1.55) 10.11 (1.72) 10.98 (1.59) 17.86

11.91 (2.49) U<R

(0.81) O<R

(0.93) 132.93 (0.78)

5.12 (0.76) 5.41 (0.71) 5.97 (0.64) U<R(1.07) U<O

O<R Openness to experience 10.01 *** U>O, O<RAgreeableness 17.45 (2.61) (2.53) 18.34 (2.38) 4.98 ** U<RConscientiousness (2.22) 13.12 (2.71) 13.83 26.36 *** U<O QBF - Time 2 Emotional stability 4.38 (1.00) 3.22 4.85 (0.92) 117.97 *** U>O U<RExtraversion 5.18 3.90 (0.93) 5.56 (0.85) *** U>O O<R U<ROpenness to experience 4.51 (1.00) 4.57 5.04 (0.95) 13.92 *** O<R U<RAgreeableness 56.45 *** U<O O<RConscientiousness 3.01 (0.81) 4.60 5.35 (0.87) 291.69 *** O<R U<R

ypes differ at p <.05 in the Scheffé (equal variances) and Tamhane’s T2 (non equal variances)

post hoc comparisons

a subt

U = Undercontrolled; O = Overcontrolled; R = Resilient* p < .05. ** p < .01. *** p < .001

84 Chapter 2

Table 5

Hierarchical regression analyses with types as dummy-coded variables Model A Model B

R² ∆ R² ∆ F Df R² ∆ R² ∆ F Df

Dependent variable = Externalizing behavior (parental ratings)

2 dummy-coded types 1 .015 .015 3.73 * 2,481 A, C, N, E, I ² .557 .557 120.05*** .557 116.41 5,476 .000 2, 476

.049 A, C, N, E, I ² .461 5, 478A, C, N, E, O ² .461 .413 72.89*** 5, 476 2 dummy-coded types 1 .461 .000 0.02 2, 476

5, 478A, C, N, E, O ² .542 *** 2 dummy-coded types 1 .557 0.16

Dependent variable = Internalizing behavior (parental ratings)

2 dummy-coded types 1 .049 12.30*** 2, 481 .461 81.81***

Dependent variable = Externalizing behavior (adolescent ratings) 2 dummy-coded types .047 .047

25.57

Dependent variable = Internalizing behavior (adolescent ratings) 36.06 A, C, N, E, I ³ .368 38.61

1 11.82*** 2,478 A, C, N, E, I ³ .239 .239 29.89*** 5, 475A, C, N, E, O ³ .250 .203 *** 5,473 2 dummy-coded types 1 .250 .011 3.33* 2, 473

2 dummy-coded types 1 .131 .131 *** 2, 478 .368 55.37*** 5, 475A, C, N, E, O ³ .383 .252 *** 5, 473 2 dummy-coded types 1 .383 .015 5.64** 2, 473

Note: Model A = Types entered in Block 1 and Five Factor Measures entered in Block 2; Model B =

Five Factor Measures entered in Block 1 and Types in Block 2

A = Agreeableness; C = Conscientiousness; N = Neuroticism; E = Extraversion; O = Openness to

experience * p < .05. ** p < .01. *** p < .001 1 Types based on adolescent QBF ratings; ² Parental ratings of adolescent personality (HiPIC) ; ³ Self-

ratings of adolescent personality (QBF)

Child personality and Parental Behavior as Moderators of Problem Behavior 85 A Variable- and a Person-Centered Approach

CHAPTER 3

A VARIABLE- AND A PERSON-CENTERED APPROACH

CHILD PERSONALITY AND PARENTAL BEHAVIOR AS

MODERATORS OF PROBLEM BEHAVIOR:

Abstract

Parenting by child personality interactions in predicting child externalizing and

internalizing behavior were investigated in a variable- and a person-centered study.

The variable-centered study utilized data from a 3-year longitudinal study of 600

children, aged 7 to 15 at T1; and 512 children aged 10 to18 at T2. Parents rated Five

Factor Model child personality, Negative control, Positive parenting, child problem

behavior whereas children rated parental behavior. Hierarchical moderated regression

analyses showed significant parenting by child personality interactions, principally for

externalizing behavior. The interactions were largely replicable across informants and

across the 3-year time span. The most prominent personality domains featuring in the

interactions were Benevolence and Conscientiousness. The person-centered study,

classifying subjects in three types, showed that Negative parental control was more

related to externalizing behavior for undercontrollers than for resilients. Negative

parental control enhanced internalizing behavior for overcontrollers, whereas it did

not affect internalizing behavior of resilient children.

86 Chapter 3

Introduction

From an ecological or contextual perspective the child is nested in a complex

network of interconnected systems (cf. Belsky, 1984; Bronfenbrenner, 1986; Dishion,

French & Patterson, 1995; Sameroff, 2000), and therefore multiple sources may

contribute to the development of (problem) behavior in children. Besides individual

risk and protective factors such as intelligence, neuropsychological deficits and

temperament, variables such as parenting, family climate, marital relationship,

relationships with peers, and contextual factors (e.g. neighborhood, socio-economic

status) have been taken into account as determinants of problem behavior (Deater-

Deckard, 2001; Deković, 1999; Frosh & Mangelsdorf, 2001; Harris, 1998; Jessor,

Van Den Bos, Vanderryn, Costa & Turbin, 1995; Stouthamer-Loeber et al., 1993).

Deater-Deckard, Dodge, Bates and Pettit (1998) demonstrated that child factors

uniquely explain up to 19% of the variance in externalizing behavior, peer-related

factors up to 13%, parenting factors up to 6% and socio-cultural factors up to 4%.

Although the main effects of child personality and parental behavior on child

problem behavior have been extensively documented in past research, some of which

is shortly reviewed below, relatively few studies have addressed the combined or

interactive effects of personality and parental behavior as determinants of

internalizing and externalizing behavior (Barber, 1992; Collins, Maccoby, Steinberg,

Hetherington & Bornstein, 2000; O’Connor & Dvorak, 2001). However, ignoring

personality-environment interactions and considering only main effects, can lead to

spurious correlations with problem behavior. As O’Connor and Dvorak (2001) point

out, bivariate relationships may be hiding associations that are conditional, and not

universal. For example, negative parental control is related to externalizing behavior

but this overall relation can hide the fact that the relationship holds in particular for

children with a difficult temperament but not for resilient children. Hence, ignoring

this interaction may lead to unwarranted conclusions about the detrimental effects of

parenting behavior. The present study aims to examine the role of interactions

between parenting and child personality variables, as predictors of child and

adolescent externalizing and internalizing behavior. The study is innovative because it

is one of the first to use the Five Factor Model to assess the child’s personality,

adopting a longitudinal as well as a cross-sectional perspective to predict the effect of

interactions. Furthermore in order to assure sufficient power for the tests of interaction


effects, the study is conducted with a moderately large sample. In addition, it is the

first study to examine child personality by parenting interactions from a person-

centered approach, searching for reliable interaction effects between parenting and the

personality types resilient, undercontrolled and overcontrolled.

Personality and Temperament as Predictors of Child Problem Behavior

Both temperament and personality refer to individual differences that evince

some stability over time presumably because they have a biological basis (Eisenberg,

Fabes, Guthrie & Reiser, 2000). Temperament refers to “the constitutionally based

individual differences in emotional, motor and attentional reactivity and self-

regulation” (Rothbart & Bates, 1998, p. 109), whereas the construct of personality

refers to “individual differences in the tendency to behave, think, and feel in certain

consistent ways” (Caspi, 1998, p.312). Temperamental differences in infancy are

assumed to be the precursors of later personality differences (Caspi & Silva, 1995;

Eisenberg et al. 2000), however, they include only a subset of personality differences

in late childhood and adulthood (Shiner & Caspi, 2003). Several reviews have linked

the traditional dimensions of temperament to the Five Factor model (Caspi, 1998;

Mervielde & Asendorpf, 2000; Shiner, 1998; Shiner & Caspi, 2003). Recent

developments in personality research show a growing consensus about how individual

differences in children’s personality can be mapped in a comprehensive taxonomic

system, usually referred to as the ‘Big Five’ or the Five Factor Model (FFM). The five

broadband dimensions, Neuroticism, Extraversion, Openness to experience,

Agreeableness and Conscientiousness, forming the top level of this hierarchical

system, are derived from lexical studies of the personality descriptive language of

adults and from the analysis of free parental descriptions of child personality

(Kohnstamm, Halverson, Mervielde & Havill, 1998).

Main-effect-type studies show relations between temperament and problem

behavior, with negative emotionality as a general risk factor (Eisenberg et al., 2000).

Externalizing behavior including hyperactivity, attention problems, antisocial

behavior and conduct disorder, has been related to lack of control (Caspi, Henry,

McGee, Moffitt & Silva, 1995; Eisenberg et al., 2000). Other studies found evidence

for the association between personality and delinquency (Krueger, Schmutte, Caspi,

Moffitt, Campbell & Silva, 1994; Ruchkin, Eisemann & Hägglöf, 1999). Internalizing

behavior, indicating anxiety and depression, has been associated with behavioral

88 Chapter 3

inhibition (Eisenberg et al., 2000), flat affect and passivity (Caspi et al., 1995). Finch

and Graziano (2001) showed that the influence of temperament on depression in

adolescents is entirely mediated by the personality dimensions agreeableness,

extraversion and neuroticism.

Child Personality from a Variable- or a Person-centered Perspective

Mervielde and Asendorpf (2000) identified and compared two empirical

approaches for studying individual differences in children’s dispositions to behave,

think and feel. The variable-centered approach distinguishes replicable broad

categories of variables across individuals, such as the dimensions of the Five Factor

Model (FFM) of personality. Variable-centered research assesses the correlational

structure of the variables across persons within a particular population. The person-

centered approach studies ‘types’, identifying clusters of individuals with similar

personality patterns. The person-centered approach delineates typical configurations

of variables within the person and hence assesses the common within-person structure

of variables. Mervielde and Asendorpf (2000) attempted to clarify the distinction

between the two approaches by referring to the card game as an analogue. The

variable-centered approach delineates the features (e.g. color, suit and values) of the

cards that are used in the game, whereas the person-centered approach studies the

typical hands that are received by the players. From this analogue it is clear that

variable- and person-centered methodologies should not be conceived as competing

approaches but rather as complementary methods that together provide a more

comprehensive picture of how individual differences can and should be addressed.

The detection of reliable and replicable types depends on the identification of

adequate dimensions to represent individual differences. Hence, the growing

consensus about the utility of the Five Factor Model as a model to represent

individual differences from childhood to adulthood, not only sets the stage for the

discovery of reliable personality types but also for the study of the interaction between

types and dimensions on the one hand and environmental variables (e.g. parenting) on

the other hand.

Several researchers studied the replicability of three types: resilients,

overcontrollers and undercontrollers (cf. Asendorpf, Caspi & Hofstee, 2002). In terms

of FFM scores, the three types can be described as follows: (a) resilients score average

on the characteristics Benevolence, Extraversion, Openness to experience and


Conscientiousness, and below the mean on Neuroticism, (b) overcontrollers receive

high ratings on Neuroticism and low ratings on Extraversion, and (c) undercontrollers

have below mean scores on Agreeableness and Conscientiousness. So far, attempts to

replicate the three types across heterogeneous samples, time, informants, methods and

variables for clustering, have produced mixed results (for an overview, see Asendorpf,

Borkenau, Ostendorf & van Aken, 2001, and Asendorpf et al., 2002). This has led to

the conclusion that although the three prototypes are frequently recovered, they are

not necessarily the prototypes that best describe each particular sample (Asendorpf,

2003; Van Leeuwen, De Fruyt & Mervielde, in press).

The use of personality types in developmental psychology has several benefits.

Instead of considering different features independently, types specify a more

parsimonious configuration or combination of characteristics. As such, types facilitate

communication among researchers and clinicians interested in describing personality

in applied settings and identifying types at risk for developing psychopathology

(Asendorpf et al., 2001; Costa, Herbst, McCrae, Samuels, & Ozer, 2002). Types have

utility as predictors of child and adolescent (mal)adaptive development. Several

studies provided evidence for the link between externalizing behavior and the

undercontrolled type, and between internalizing and the overcontrolled type

(Asendorpf et al., 2001; De Fruyt, Mervielde & Van Leeuwen, 2002; Van Leeuwen,

De Fruyt & Mervielde, in press).

Parental Behavior and Child Problem Behavior

Although they are often used as interchangeable concepts, it should be noted

that ‘parental behavior’ is distinctive from ‘parenting styles’ as defined by, e.g. the

typology of Maccoby and Martin (1983), whereby authoritarian, authoritative,

permissive and indifferent parenting are based on two dimensions, demandingness

and responsiveness. Parenting styles can be regarded as the general context, the

climate in which the more specific parenting practices or behaviors are expressed

(Darling & Steinberg, 1993).

Most studies on risk and protective factors for the development of problem

behavior have focused on externalizing behavior, probably because it is more visible

and has more negative social consequences (Deković, 1999). The relationship of

parenting to child externalizing behavior has been documented both in clinical and

non-referred samples (Belsky, Hsieh & Crnic, 1998). Lack of parental involvement or

90 Chapter 3

poor acceptance-responsiveness, lack of supervision or poor parental monitoring,

harsh and inconsistent punishment and insufficient rewarding of positive behavior

have been identified as predictors of externalizing behavior (Deater-Deckard &

Dodge, 1997; Forehand, Miller, Dutra & Chance, 1997; Haapasalo & Tremblay,

1994; Loeber & Dishion, 1983; Loeber & Stouthamer-Loeber, 1986; Patterson &

Stouthamer-Loeber, 1984; Rothbaum & Weisz, 1994; Ruchkin, Eisemann & Hägglof,

1999; Stormshak et al., 2000; Wakschlag & Hans, 1999; Weiss, Dodge, Bates &

Pettit, 1992). Parenting has also been associated with child internalizing behavior,

such as anxiety (Gruner, Muris & Merckelbach, 1999; Siqueland, Kendall &

Steinberg, 1996) and depression (Muris, Schmidt, Lambrichs & Meesters, 2001;

Richter, 1994).

Child Personality and Parental Behavior as Moderators of Child Problem Behavior

According to Thomas and Chess’ goodness-of-fit-theory’ (1977), the

development of problem behavior has its origins in the child’s temperament and in its

interaction with the socializing environment. When there is a mismatch between a

difficult temperament and parenting practices, behavioral disturbances may develop.

In theory, temperament does not lead to behavioral problems by itself; it only has an

effect in conjunction with particular environments (Bates, Pettit, Dodge & Ridge,

1998). At present, there is some empirical evidence underscoring the importance of

interactions between child characteristics and parenting in the prediction of child

problem behavior. These studies vary in (a) design, i.e. cross-sectional versus

longitudinal (e.g. Bates et al., 1998; Belsky et al., 1998; Rubin, Burgess, Dwyer &

Hastings, 2003; Stoolmiller, 2001) versus experimental (e.g. Anderson, Lytton &

Romney, 1986), (b) sample composition, i.e. non-referred versus referred (e.g.

Anderson et al., 1986; Colder, Lochman & Wells, 1997), (c) age, i.e. preschool (e.g.

Paterson & Sanson, 1999; Rubin et al., 2003), school-aged (e.g. Lengua, Wolchik,

Sandler & West, 2000; Stoolmiller, 2001; Wootton, Frick, Shelton &

Silverthorn,1997) and adolescent (e.g. Carlo, Roesch & Melby,1998; Olweus, 1980),

and (d) gender, for example some studies only include boys (e.g. Anderson et al.,

1986; Belsky et al., 1998; Colder et al., 1997; Olweus, 1980; Stoolmiller, 2001). The

studies also differ in how parental behavior and child temperament are assessed. Most

studies linking parenting and child characteristics have been restricted to the

assessment of the child’s temperament, leading to the general conclusion that children


with a difficult temperament are particularly vulnerable when also exposed to punitive

parenting or negative control. Few studies have examined personality-environment

interactions from the perspective of the Five Factor Model. O’Connor and Dvorak

(2001) showed that parental behaviors only matter for some kinds of children and not

for others, and hence that specific combinations of personality characteristics and

parenting variables operate as protective or risk factors. Prinzie et al. (2003) also used

the FFM in their investigation of interaction effects between personality and parenting

on externalizing behavior. They showed (a) that benevolence functions as a protective

factor, buffering the effect of maternal or paternal overreactivity, i.e. the tendency by

parents to respond with irritation and/or anger to problematic behavior of their

children, and (b) that coercive parental behavior is more strongly related to

externalizing problems for children low on conscientiousness. The present study

extends previous research on child personality and parental behavior as moderators of

child problem behavior by (a) adopting the Five Factor Model for the assessment of

child personality in the prediction of problem behavior, instead of examining

temperament, (b) including both a more affective and positive component of parenting

as well as a control dimension, whereas most studies solely concentrate on a

detrimental form of parenting, (c) investigating both internalizing and externalizing

behavior as the outcome variables whereas past research has focused predominantly

on externalizing behavior, (d) utilizing a large population sample including both boys

and girls (e) adopting a longitudinal design as well as a cross-sectional design for

assessment of the effects of parenting by child personality interactions.

Problems with Moderated Regression Research

Although interaction effects are frequently reported in experimental studies,

field researchers often experience that moderator effects are extremely difficult to

detect (McClelland & Judd, 1993). This can be attributed in part to statistical

problems with moderated multiple regression (MMR) research (cf. Aguinis, 1995;

Chaplin, 1991; McClelland & Judd, 1993). In general, tests of hypotheses regarding

the effects of moderators often have very low statistical power and a high risk of Type

II errors. This leads researchers to prematurely dismiss theoretical models that include

moderating effects.

Several specific factors have been identified as determinants of the statistical

power of moderator research: (a) variable distributions, (b) operationalizations of

92 Chapter 3

predictor and criterion variables, (c) sample size, and (d) multicollinearity of predictor

variables (Aguinis, 1995). Problems of variable distributions occur when there is

predictor variable range restriction, due to nonrandom or biased sampling, such as in

clinical samples (Aguinis, 1995; Holmbeck, 1997). Second, inaccurate

operationalization of predictor and criterion variables can result in high measurement

error. Low reliability of the predictor variables adversely affects the reliability of the

product term. When the rating scale of the criterion variable does not include

sufficient scale points, referred to as ‘scale coarseness’, this results in information loss

and impedes the detection of moderator effects (Aguinis, 1995). Aiken and West

(1991, p.160-164) show that small effect sizes need large samples to be detected and

that the required sample size to produce power of .80 at α = .05 dramatically increases

as reliability decreases. The necessity for large samples also increases when the

proportion of variance accounted for by the first order effects is large and when the

interpredictor correlation is small. The final problem with MMR concerns predictor

multicollinearity: when predictor scores and the interaction term are highly correlated,

regression coefficients will be unstable due to larger error terms. However, Cronbach

(in Aguinis, 1995, p. 1149) argues that multicollinearity is not detrimental to the

power of MMR, in particular when there are only two predictors.

Several precautions were taken in the present study to reduce statistical

problems with MMR. First, we used reliable measures. Both personality and parenting

measures are second order factors. In order to assess child and adolescent personality,

a psychometrically well-validated instrument based on the Five Factor Model was

used. Moreover, the use of these five broadband personality domains facilitates

replicability and assures a comprehensive test (Caspi, 1998, De Raad & Perugini,

2002). The parenting variables are two second-order factors, i.e. ‘positive parenting’

and ‘restrictive or negative control’, derived from a measurement instrument based on

social learning theory. These factors are analogous to two dimensions of parenting

that are regularly mentioned in the literature (Gallagher, 2002), the first describing the

affective nature of the parent-child relationship, as indicated by involvement and

providing support (‘warmth’ or ‘responsiveness’) and the second referring to efforts

of parents to influence their child’s behavior, such as setting and enforcing standards

of behavior (‘control’ or ‘demandingness’) (Maccoby & Martin, 1983). To reduce

measurement error, we used a cross-rater strategy (Chaplin, 1991) and created

aggregated scores based on ratings provided by different informants (cf. infra).


Second, a non-restricted large general population sample (N= 600) was utilized. The

size of this sample approximates the range of N required to produce power of .80 at α

= .05 for small effect sizes and a reliability of .80 (cf. Aiken & West, 1991, p. 164).

To eliminate multicollinearity, two solutions are proposed in the literature, i.e.

centering of the predictor variables (putting predictors in a deviation score form so

that their means are zero) and the use of principal-components regression (PCR)

(Aguinis, 1995). The use of PCR has not been widely accepted as a method for

improving the tests of interactions (Aiken & West, 1991, p. 168-169). Centering of all

predictors, however, reduces multicollinearity, and in addition has also

interpretational advantages, and hence is therefore strongly recommended (Cohen,

Cohen, West & Aiken, 2003, p. 267).

Finally, to reduce the likelihood of capitalizing on chance, we checked for cross-rater

stability, i.e. parent versus child ratings of parental behavior and for cross-sample

stability, i.e. by replicating the findings across a three-year interval. We also

investigated whether the interaction effects assessed at Time one predicted Time two

problem behavior.

The Current Research

Interactions between parenting and child personality as predictors of child and

adolescent externalizing and internalizing behavior will be investigated in two studies.

In the first study, the parenting by personality interaction is examined from a variable-

centered approach. In order to assure a comprehensive test of the role of various

personality traits as moderators of problem behavior, a broadband Five Factor Model

instrument is used to assess children’s personality. Moreover, to guarantee sufficient

power to detect interaction effects the hypotheses are verified on a large sample and

checked for cross-rater and cross-time stability. In line with the variable-centered

methodology both parenting and personality are entered as (continuous) quantitative

variables predicting child and adolescent problem behavior. As the variable-centered

approach emphasizes interactions between variables and does not take into account

the within-person personality structure, the first study is supplemented with a second

study, adopting a person-centered approach. In the second study, utilizing a classic

ANOVA design, we look for interaction effects between the personality types

resilient, undercontrolled and overcontrolled (derived from clustering subjects based

94 Chapter 3

on their Five Factor personality profile) and discrete categories of parents with high

and low levels of positive parenting and negative parental control.

Study 1

Method

Participants

The present study is based on data from a longitudinal study investigating

parenting, parental and child personality characteristics and children’s problem

behavior, at two assessment periods separated by a 3-year interval. Subjects were

parents (both mothers and fathers) and one, non-referred child. At Time one, 600

families were included (N of mothers = 596, N of fathers = 533). The target children,

281 boys and 319 girls, had a mean age of 10.9 (SD = 1.8; range 7-15). The age

distribution is as follows: 1.3% of the children was 7 years old, 11.5% 8 years, 12.5%

9 years, 15.8% 10 years, 11.8% 11 years, 22.3% 12 years, 19.7% 13 years, 4.5% 14

years and 0.5% 15 years. Of the families 9.8% was not the original family and 8.2%

of the families included a single parent. For mothers and fathers respectively, the

mean age was 38.6 (SD = 4.7; range 20-68) and 40.6 (SD = (5.0; range 26-63). Both

mothers and fathers had representative levels of education: the highest level of

education was elementary school for 12.9% of the mothers and 7.9% of the fathers,

48.5% of the mothers and 49.5% of the fathers completed secondary education, 29.7%

of the mothers and 29.4% of the fathers finished higher education, and 8.7% of the

mothers and 13.3% of the fathers obtained a university degree. Of the mothers 75.2%

was employed whereas of the fathers 95.3% was employed.

At Time two about 85% of the families continued collaboration, with 512

families (N of mothers = 501, N of fathers = 443). The remaining 244 boys and 268

girls had a mean age of 13.9 (SD = 1.8; range 10-18). The age distribution is as

follows: 0.8% of the children was 10 years old, 10.6% 11 years, 13.7% 12 years,

16.8% 13 years, 13.3% 14 years, 22.3% 15 years, 18.0% 16 years, 3.9% 17 years and

0.6% 18 years. The ratio boys/girls, χ²(1) = .07, p > .05, and mothers/fathers, χ²(1) =

.03, p > .05 was equal across the two assessment moments. Chi-square statistics

showed no significant differences between the Time one and Time two samples for

family characteristics, χ²(2) = 2.91, p > .05, social indices for mothers, χ²(5) = 2.24, p


> .05, and fathers, χ²(5) = 1.42, p > .05, and employment status for mothers χ²(2) =

2.36, p > .05, and fathers χ²(2) = 0.24, p > .05. Hence, it can be concluded that the

Time two sample characteristics did not differ from the Time one sample

characteristics as a consequence of attrition.

Measures

Parental behavior. The Ghent Parental Behavior Scale (GPBS, Van Leeuwen

& Vermulst, in press) is a questionnaire, designed to assess parenting behavior based

on concepts from Social Learning Theory (Capaldi & Patterson, 1989; Patterson, Reid

& Dishion, 1992). This theoretical framework organizes parenting into five well-

defined constructs derived from observable parental behavior. Validation of a pilot

version led to the conclusion that the five Patterson constructs (positive involvement,

monitoring, problem solving, structure and positive reinforcement) appeared to be too

heterogeneous. Refinement of the constructs resulted in a new questionnaire with nine

scales: Autonomy, Discipline, Positive parental behavior, Harsh punishment,

Monitoring, Rules, Ignoring unwanted behavior, Material rewarding, and Inconsistent

discipline. Evidence for the factorial validity and for a moderate to good internal

consistency of the majority of the scales is provided by Van Leeuwen and Vermulst

(in press). Subjects rate the frequency of each behavioral item on a 5-point Likert

scale ranging from ‘never’ to ‘always’. The parent version provides self-ratings for

parental behavior and the child version, with the same items as in the parent version,

allows children/adolescents to rate the parental behavior of their mother and/or father.

In order to reduce the number of scales of the GPBS, factor analyses with

principal axis factoring and oblimin rotation were conducted for four groups of raters:

mothers, fathers, children about their mothers and children about their fathers. A two-

factor solution was most appropriate. Because the scales monitoring, material

rewarding and inconsistent discipline did not consistently load on the same

dimensions for each of the groups, they were dropped. In each of the four groups of

raters, two dimensions consistently emerged: ‘positive parenting’ (consisting of the

scales positive parental behavior, teaching rules and autonomy), and ‘negative

control’ (consisting of the scales discipline, ignoring of unwanted behavior and harsh

punishment). Table 1 shows factor loadings, percentages of explained variance and

Cronbach alpha’s for the two dimensions. The intercorrelations between the two

96 Chapter 3

factors are .03 and .13 for both assessment moments, indicating independency of the

two parenting dimensions. The correlations between the ratings provided by parents

and children on the two parenting dimensions were all positive and significant with p

< .001, ranging from .19 to .38 for Time one and ranging from .21 to .42 for Time

two.

Child Personality. The Hierarchical Personality Inventory for Children

(HiPIC; Mervielde & De Fruyt, 1999) was used to assess the child’s personality. Five

personality domains are measured, based on 18 facets hierarchically organized under

the five domains, i.e. Extraversion (based on the facets Shyness, Optimism,

Expressiveness, Energy), Benevolence (based on Egocentrism, Irritability,

Compliance, Dominance, Altruism), Conscientiousness (based on Achievement

motivation, Concentration, Perseverance, Orderliness), Emotional Stability (based on

Anxiety and Self-confidence) and Imagination (based on Creativity, Curiosity and

Intellect). Parents rated the 144 items on a 5-point Likert scale. The factor structure

proves to be highly replicable across both childhood and adolescence (Mervielde &

De Fruyt, 2002). In our study, Cronbach alpha’s (for maternal and paternal ratings at

the two measurement moments) ranged from .93 to .94 for Benevolence (N of items =

40), from .94 to .94 for Conscientiousness (N of items = 32), from .86 to .88 for

Emotional Stability (N of items = 16), from .89 to .91 for Extraversion (N of items =

32) and from .91 to .92 for Imagination (N of items = 24).

Child problem behavior. The Dutch version of the Achenbach Child Behavior

Checklist (CBCL; Verhulst, Van der Ende & Koot, 1996) is used as an instrument for

screening children with behavioral and emotional problems. Parents rate the

frequency of 113 problematic behaviors on a 3-point Likert scale. Two broadband

syndromes can be derived: Internalizing, with items referring to somatic complaints,

social withdrawal, and anxiety/depression, and Externalizing, including items

indexing aggression, hyperactivity, and delinquency. Cronbach alpha’s for the scale

Internalizing behavior (N of items = 31), ranged from .86 to .88 for maternal and

paternal ratings over the two measurement moments, and from .90 to .91 for the scale

Externalizing behavior ((N of items = 33).


Procedure

Families were recruited via stratified random sampling of elementary and

secondary schools. For elementary schools the sample was stratified by province,

(East and West Flanders), region (rural or urban), school type (public/private/catholic


secondary schools, sampling was based on province (East and West Flanders), type of

curriculum (vocational, technical and general education) and grade (first and second

year of secondary school). A letter addressed to the parents informed them about the

goal and the procedures of the research project. The response rate of parents with

children in primary schools was 41%, and 39% for parents with children in secondary

schools.

Given this modest response rate, we compared the socio-economic status of

the present sample with the characteristics of a representative general community

sample (N families = 1789) used to investigate the Quality of Life in Flemish school-

aged children (Van den Bergh, 1997). The family composition was somewhat

different, χ²(2) = 11.95, p < .01, mainly due to the higher proportion of single-parent

families in the present study. The educational level differed for mothers, χ²(5) =

12.37, p < .05, and fathers, χ²(5) = 25.46, p < .05, but not as a consequence of an

overrepresentation of the higher educational levels in the present study. There was a

difference in employment status for mothers χ²(2) = 9.75, p < .01, but not for fathers

χ²(2) = 1.16, p > .05. Although small differences between the two samples can be

observed, it can be concluded that the present sample represents the broad socio-

economic strata.

As regards the presence of problem behavior, we compared the prevalence of

children scoring above the 90th percentile (T-score > 63) on the Total Problem

Behavior Scale with an epidemiological study in Flanders, investigating emotional

and behavioral problems in children aged 6 to 12 (Hellinckx, De Munter, & Grietens,

1991). In the present study the percentage of children showing serious emotional

and/or behavioral problems was 13.8, whereas the percentage in the epidemiological

study was 15.4. Hence it can be concluded that the sample in the present study does

not consist of children with less problem behavior than might be observed in the

target population. Moreover the present study does not focus on assessment of mean

98 Chapter 3

levels or prevalence rates but rather on the relationships among variables and hence

the issue of representative sampling carries less weight.

At each assessment period, a trained psychology student who instructed the

mother, father and child to independently complete a series of questionnaires, visited

the families at home. Both parents filled out the HiPIC, GPBS and CBCL, whereas

children were administered the GPBS at both measurement moments. In addition,

teachers filled out the Dutch version of the Achenbach’s Teacher Report Form (TRF;

Verhulst, Van der Ende & Koot, 1997a) at Time one and children completed the

Youth Self Report (YSR; Verhulst, Van der Ende & Koot, 1997b) and a Dutch

shortened version of Goldberg’s (1992) hundred adjectives, entitled the Questionnaire

Big Five (QBF; Gerris et al., 1998) at Time two, to provide self-ratings of problem

behavior and personality respectively. In this study we only used measures that were

presented at both measurement moments, in order to facilitate comparison of the

results. We did not opt to query the children about their problem behavior and

personality at the first measurement moment (a) because the YSR is intended to asses

self-reported problem behavior from age 11 onwards, but almost half of the children

were younger than 11 years at the first assessment period; (b) because at present there

is little evidence for the validity and in particular for the discriminant validity of self-

rated personality elementary school children in general; (c) for practical reasons: we

did not want to overload the children with questionnaires at the first assessment

moment, bearing in mind their age and the length of the questionnaires (e.g. the

HIPIC consists of 144 items). The Ghent Parental Behavior Questionnaire was

specifically designed to obtain ratings of both parents, and children with reading

abilities (aged 7-8 years and older).

Statistical analyses

Parallel cross-sectional hierarchical multiple regressions were carried out on

the Time one and Time two data in order to detect significant interactions between

child personality and parenting. In order to check cross-rater stability, we conducted

at both times analyses for (a) the parent sample, with parental self-ratings of Negative

control and Positive parenting, and (b) the child sample, with child ratings of Negative

control and Positive parenting. Child gender (boy is coded as 1 and girl is coded as 2)

and age were consistently entered in Step 1 as control measures. In Step 2, one of the

five child personality domains (i.e. Benevolence, Conscientiousness, Emotional


stability, Extraversion and Imagination) and one of the two parenting constructs

variables (i.e. Positive parenting and Negative control) were entered, and in Step 3 the

cross-product of the personality variable and the parenting variable. Multiplying the

two centered predictors formed the interaction term. Evidence for a moderator effect

is found when there is a significant increase in the multiple R² after entering the

interaction term, as indicated by a significant incremental F test. In order to prevent

chance capitalization in finding moderator effects we applied the Bonferroni

correction. In this study, the alpha level was set at .05, with forty tests explaining

Externalizing behavior and forty tests explaining Internalizing behavior. Application

of the Bonferroni correction indicated that the alpha level for each individual test

needed to be lowered to .001 in order to adjust the overall alpha level to .05.

For testing and interpreting interactions, we followed the guidelines proposed

by Aiken and West (1991) and Cohen, Cohen, West and Aiken (2003). First,

significant interactions were interpreted by plotting simple regression lines for high (=

1 SD above the mean), mean and low (= 1 SD below the mean) values of the

moderator variables. Second, the significance of the slopes for these simple regression

lines and the difference between the slopes was tested with t-tests. Finally, interaction

patterns were identified, based on the signs of the regression coefficients of the two

independent (B1 and B2) and the interaction (B3) variables. Three theoretically

meaningful interaction patterns can be distinguished (Cohen et al., 2003, p. 285-286):

(a) synergistic or enhancing interactions, in which all three regression coefficients

have the same sign, indicating that combining predictors produces an incremental

effect beyond the additive effects; (b) buffering interactions, in which the two

predictors have regression coefficients of opposite sign, meaning that one predictor

weakens the effect of the other predictor. For buffering interactions one predictor

represents a risk factor, while the other predictor acts as a protective factor; (c)

interference or antagonistic interactions, in which both B1 and B2 have the same sign

and B3 is of opposite sign, indicating a compensatory or ‘either-or’ effect of B1 and B2

on the criterion. Interactions are sometimes described as ordinal (non crossing) or

disordinal (crossing). This distinction is less useful because it is mainly determined by

the strength of the first-order effects.

100 Chapter 3

Constructing composite scores

For the present analyses, aggregated scores for predictor and criterion

variables were created to reduce the potential number of analyses, generated by the

numerous combinations of types of raters (i.e. mothers, fathers, children rating

mothers and fathers), parenting variables (i.e. Positive parenting and Negative

control), personality domains (i.e. Benevolence, Conscientiousness, Emotional

stability, Extraversion and Imagination), dependent variables (Internalizing and

Externalizing behavior) and measurement moments. This was accomplished by

extracting a common factor score from each pair of ratings provided by two different

informants by means of principal component analyses: for mother and father ratings

of their own parental behavior, for ratings of child personality and problem behavior,

and for child ratings of maternal and paternal behavior. The score of the other rating

substituted the missing value in case only one of the two ratings was available (for

example in single-parent families). Thus, the common factors reflect (a) self-reported

parenting by two different informants (mother and father), (b) other-reports of child

personality and problem behavior by two different informants (mother and father),

and (c) other-reports of maternal and paternal behavior by one informant (the child).

The use of composite scores is supported by the meaningful and significant (p

< .001) correlations between the paired ratings composing the aggregated scores: the

correlation between mothers and fathers was .75 for externalizing behavior, .59 for

internalizing behavior, .70 for Benevolence, .81 for Conscientiousness, .69 for

Emotional stability, .69 for Extraversion, .69 for Imagination, .22 for Positive

parenting and .44 for Negative control. The correlation between child ratings for

mothers’ and fathers’ Positive parenting was .74, and .70 for Negative control. The

explained variances across the two time-points ranged from 79.50% to 87.63% for the

two dependent variables, from 82.46% to 90.76% for the five personality variables,

and from 61.17% to 89.53% for the two parenting variables.

Parent and child ratings of parental behavior were not aggregated because

principal component analyses at both measurement moments distinguished two

separate factors for Positive parenting, i.e. a factor including ratings of mothers and

fathers, and a factor including child ratings of maternal behavior and child ratings of

paternal behavior. Not aggregating scores of parents and children also enables the

cross-validation of interaction effects across ratings of parents and children.


Moreover, using child ratings of parenting in analyses with parent measures of

problem behavior and personality, decreases shared method variance.

The use of factor scores is valuable because they reflect the common core of

variables, reduce measurement error and hence increase the power of statistical tests

for interaction (Chaplin, 1991; Jaccard & Wan, 1995). In addition they center the

predictor variables, as the mean equals zero and the standard deviation equals one.

The present study partially controls for shared method variance, by combining

parental ratings of personality and child problem behavior with child ratings of

parenting in Study 1. We acknowledge that it would be more appropriate to use

ratings supplied by different informants for each of the measures. However, we prefer

to use ratings that were presented at both measurement moments, in order to facilitate

comparison of the results. Another way to address the problem of shared method

variance is to assess the convergent validity of the data supplied by the informants

against external criteria. Correlations between Time one TRF and CBCL ratings were

all significant at p < .001 with values of .44 (father-teacher) and .46 (mother-teacher)

for externalizing behavior and .18 (father-teacher) and .16 (mother-teacher) for

internalizing behavior. Correlations between Time two CBCL and YSR ratings were

all significant at p < .001 and with values of .41 (father-child) and .41 (mother-child)

for externalizing behavior and .35 (father-child) and .40 (mother-child) for

internalizing behavior. Correlations between Time two parental and QBF adolescent

ratings were all significant at p < .001 and with r = .20 for Benevolence, r = .46 for

Conscientiousness, r = .37 for Neuroticism, r = .46 for Extraversion and r = .34 for

Imagination. This type of analysis illustrates that the data provided by the informants

exhibit a reasonable degree of convergent validity and hence that the results of the

present study are not seriously encumbered by problems of shared method variance.

Results

Shared Method Variance

Cross-sectional Analyses at Time one

Gender and age effects. Results of the hierarchical regressions analyses are

shown in Table 2 and 3. Of the two control measures Gender and Age, entered in Step

1, Gender accounted for a significant proportion of the variance in Externalizing

behavior (R² = .039).

102 Chapter 3

Interaction effects. Three moderator effects were predictive of child

externalizing behavior, when parents supplied ratings of their own parental behavior.

Significant interactions for Negative control were found with the personality domains

Benevolence (R² = .025 and F change (1, 577) = 30.61, p < .001) and

Conscientiousness (R² = .014 and F change (1, 577) = 11.95, p < .001). A significant

interaction between Positive parenting and Benevolence predicted child externalizing

(R² = .010 and F change (1, 577) = 11.17, p < .001). No interactions significantly

predicted child internalizing behavior.

An interference or antagonistic interaction is found for the Positive parenting

X Benevolence interaction (See Figure 2). Here, the interaction term has the opposite

sign of both predictors. This suggests that both positive parenting and child

personality are negatively related to externalizing behavior, but that the importance of

positive parenting is lessened by Benevolence or vice versa.

Testing and interpreting interaction effects. The tests on the significance of the

difference between simple slopes (see Table 5) show that the slopes for children rated

low (1 SD below the mean) or around the mean on the personality characteristics

Benevolence and Conscientiousness are significant. For these children, parenting has

a significant effect on child problem behavior. When rated high (1 SD above the

mean) on one of these personality domains, the slopes are not significant. This

indicates that for children characterized by highly adaptive personality characteristics,

parenting does not predict externalizing behavior.

Based on the signs of the regression coefficients (see Table 2 and 3), two

interaction patterns can be identified, (a) buffering interactions and (b) interference or

antagonistic interaction patterns. For the Negative control X personality interactions

(see Figure 1), the coefficients of the independent predictors have opposite signs,

indicating that one predictor diminishes problem behavior while the other predictor

enhances it. Hence, negative control can be considered as a risk factor for

externalizing behavior in particular for children rated low or around the mean on

Benevolence and Conscientiousness. On the other hand, the impact of low scores on

Benevolence and Conscientiousness will be diminished for children experiencing low

parental Negative control. Benevolence and Conscientiousness can be regarded as

protective factors: children scoring high on these personality domains do not show

problem behavior at all, even when faced with Negative parental control.


The figures show that most interactions are ordinal: the rank order of the

outcomes of one predictor is maintained across all levels of the other predictor, within

the observed range of the second predictor. This appearance is influenced by the

strength of the first-order effects (cf. Cohen et al., 2003, p. 286). In this study, the

strongest independent effects are found for the personality variables.

Effect sizes and statistical power of interactions. In order to permit

comparison across studies, effect size measures are reported in Table 5. Effect sizes

for the interactions at Time one range from .01 to .05, and can be regarded as ‘small’

(Aiken & West, 1991, p. 158). The statistical power for the interaction terms does not

equal, but approximates the standard of .80. Taking the Negative control X

Benevolence as an example, we note reliabilities of .80, an interpredictor correlation

of -.30 and a squared multiple correlation of .50 for the main effects. According to

Aiken and West (1991, p.164) these values correspond with a power of .39-.44 at N =

392. A power of .80 with these values requires a sample size between 909 and 1056.

Replication of the cross-sectional analyses at Time two

Gender and Age effects. The control measure Gender, entered in Step 1,

significantly explained about 3% of the variance of externalizing behavior.

Interaction effects. The significant interactions with Negative control at Time

one for externalizing behavior, were replicated at Time two (see Table 2) both with

Benevolence (R² = .042 and F change (1, 501) = 51.60, p < .001) and

Conscientiousness (R² = .019 and F change (1, 501) = 14.13, p < .001). In addition the

Negative control X Benevolence interaction (R² = .024 and F change (1, 498) = 28.15,

p < .001) and the Negative control X Conscientiousness interaction (R² = .030 and F

change (1, 498) = 21.84, p < .001) were also replicated when children rated parenting

behavior. At Time one these interactions were only significant when the alpha level

was set at .01. The Positive parenting X Benevolence interactions was not replicable.

No interaction terms significantly predicted child internalizing behavior at Time two

(see Table 3).

Testing and interpreting interaction effects. The same interaction patterns of

Time 1, were replicated at Time 2. For the Negative control X Benevolence or

104 Chapter 3

Conscientiousness interactions, the interactions were of the buffering type. The effect

sizes of the interaction terms are also small, ranging from .01 to .10.5

Time one variables predicting Time two problem behavior

Table 4 reports the results of HMRA with Time one variables predicting Time

two externalizing and internalizing behavior. When parents provided self-ratings of

parental behavior the Negative control X Benevolence interaction (R² = .024 and F

change (1, 465) = 18.84, p < .001) and the Negative control X Conscientiousness

interaction (R² = .022 and F change (1, 465) = 13.95, p < .001) predicted Time two

child externalizing behavior. The results also show that Time one child personality is

a predictor for Time two externalizing behavior, in particular Benevolence and

Conscientiousness. Time two internalizing behavior is predicted by each of the five

personality dimensions, with Emotional stability and Extraversion as the strongest

predictors. Considering the parenting variables, shows that Time one Negative control

is a predictor of both Time two externalizing and internalizing behavior for parental

self-ratings, and also a predictor of externalizing behavior when child ratings are used

in the analyses. Time one Positive parenting is not a strong predictor of time two child

problem behavior.

5 We also checked for nonlinear relationships between independent and dependent variables, although

(a) the inclusion of curvilinear relationships in this study is not driven by theoretical assumptions and

(b) there is an ongoing debate whether one should include curvilinear effects if the central interest is in

the interaction term. Curvilinear effects show lower reliability than the cross-product terms and the

inclusion of multiple higher order terms introduces multicollinearity and instability of the regression

equation (Cohen et al., 2003, p. 299-300). Testing quadratic relationships was limited to models with

significant parenting by personality interactions. For the time 1 data, significant quadratic relationships

were present for benevolence and conscientiousness in the prediction of child externalizing behavior

and for benevolence predicting child internalizing behavior. These quadratic effects did not influence

the parenting by personality interactions for the majority of the models, indicating that the interactions

were not the result of the curvilinear nature of the regression model. Only two interactions could be

attributed to the curvilinear effects: the benevolence by positive parenting (parental ratings) interaction

and the benevolence by negative control (child ratings) interaction in the prediction of externalizing

behavior. For the time 2 data, the conscientiousness by positive parenting (child ratings) interaction

was not a significant predictor of externalizing behavior when the quadratic effect of conscientiousness

was included in the regression model.


Discussion

In this study we investigated parenting by child personality interactions for the

prediction of child and adolescent problem behavior. The current study examines

interactions detected in previous studies, for example between negative parental

control and low agreeableness as predictors of externalizing behavior (e.g. O’Connor

& Dvorak, 2001; Prinzie et al., 2003). However, integrating all Big Five personality

domains in this study, i.e. agreeableness, conscientiousness, emotional stability,

extraversion and imagination, enabled us to test for interactions between parenting

and each of the five broad child personality domains as predictors of both

externalizing and internalizing behavior.

HMRA analyses showed prominent interactions between negative parental

control and the personality domains benevolence and conscientiousness, predicting

externalizing behavior, and hence corroborating the findings of previous studies.

These interactions were found for parental self-ratings of parenting behavior and were

partly replicable across raters and a three year interval. Moreover, the Time one

interactions predicted Time two externalizing behavior. The results indicate that

negative control is a risk factor for externalizing behavior in particular for children

rated low or around the mean on benevolence and conscientiousness. On the other

hand, children rated high on these personality domains are not likely to develop

externalizing behavior even when exposed to restrictive parental control.

Significant interactions were also found between positive parenting and

benevolence.

From these results it can be concluded that children rated low on Benevolence are

likely to show externalizing behavior in particular when parents are not supportive.

When children show high levels of benevolence they are not affected by low levels of

positive parenting. However, this positive parenting by benevolence interaction is not

stable across judges and across time.

With internalizing behavior as the outcome variable, there were no significant

interactions. Our findings suggest mainly independent or additive contributions of

parenting and child characteristics to internalizing behavior. The significant main

effects suggest that in particular child personality is an important independent

predictor of internalizing behavior.

The first study adopted a variable-centered approach on child personality,

considering the effects of single Big Five domains and parenting variables as well as

106 Chapter 3

the effects of personality-parenting interactions. The variable-centered approach

identifies the important personality and parenting dimensions but it does not allow us

to target particular groups of children at risk. Moreover, it does not take into account

the prevalence and relevance of particular combinations of risk factors such as a low

score on both benevolence and conscientiousness and therefore it is not clear to what

extent such configurations constitute an additional risk factor for developing

externalizing or internalizing problem behavior. The second study, adopting a person-

centered approach, groups children into types based on their personality profile. The

same sample of children will be assigned to three replicable personality types, i.e.

resilient, overcontrolled and undercontrolled, and hence the next study will focus on

the interactions between these types and the parenting dimensions.

Study 2

In order to derive the overcontrolled, undercontrolled and resilient personality

types, we followed the method described by Asendorpf et al. (2001). In a first step,

Method

Participants

The sample in study 2 is part of the Time one sample of the longitudinal

dataset as described in study 1, and consists of 539 families with 251 boys and 288

girls. Subjects with missing variables were dropped, accounting for the reduced

sample size in comparison with the Time one sample of study 1.

Measures and Procedure

The results of this study are based on the same questionnaires administered in

study 1: the Ghent Parental Behavior Scale (GPBS; Van Leeuwen & Vermulst, in

press), the Hierarchical Personality Inventory for Children (HiPIC; Mervielde & De

Fruyt, 1999) and the Achenbach Child Behavior Checklist (Verhulst et al., 1996). The

data collection procedure is the same as described in study 1.

Assessment of Personality types 6

6 A full description of the derivation of the three types for this sample is given in De Fruyt et

al. (2002).


hierarchical cluster analyses with Ward’s method were conducted, using raw HiPIC

domain scores. The resulting three-cluster solutions of these analyses were then used

as initial cluster centers in a nonhierarchical K-means clustering procedure.

The continuous variables positive parenting and negative control, rated by the

parents, were recoded into three categories: scores of one SD ≤ the mean, scores

around the mean and scores of one SD ≥ the mean. In the analyses only subjects

scoring low or high on parenting variables were included. The number of children

exposed to high positive parenting was 105, whereas for low positive parenting N =

100. The category high negative control included 97 subjects and the category low

negative control grouped 91 subjects.

Personality types

Parenting by Personality type Interactions

Delineation of parenting categories

Results

The final three-cluster solution obtained after applying the two-step procedure

(Ward’s method and K-means clustering) clearly resembled the resilient,

undercontrolled and overcontrolled prototypes. The first cluster grouped 185 children,

with above average cluster centers for the personality domains benevolence,

extraversion, conscientiousness, emotional stability and imagination. This group was

identified as the resilient type. The second cluster included 162 children, scoring low

on emotional stability and extraversion, and resembled the overcontrolled type. The

third cluster with 192 children, characterized by low scores on conscientiousness and

benevolence and around the mean scores for emotional stability, extraversion and

imagination, was designated as the undercontrolled type.

GLM univariate analyses of variance were conducted to detect significant

parenting by personality type interactions. In the analyses with externalizing behavior

as the dependent variable, the dichotomy undercontrollers versus resilients was

entered as one of the independent variables, whereas in the analyses with internalizing

behavior as the dependent variable, the dichotomy overcontrollers versus resilients

was entered as one of the independent variables. This choice is based on previous

research, showing a meaningful relationship between the undercontrolled personality

108 Chapter 3

type and externalizing behavior and a relationship between the overcontrolled

personality type and internalizing behavior (Asendorpf et al., 2001; De Fruyt et al.

2002; Van Leeuwen et al., in press).

Table 6 shows the category means and the results of the analyses of variance.

Significant parenting by personality type interactions were only found when negative

control was included as the parenting variable, both for externalizing and internalizing

behavior. The results indicate that undercontrolled children are rated significantly

higher on externalizing behavior than resilient children, with the greatest difference

for children exposed to high levels of parental negative control. This suggests that

especially undercontrolled children, subjected to negative parental control, are at risk

for externalizing behavior. The difference in problem behavior between

undercontrollers and resilients fades, when the level of received negative control is

low.

The overcontrolled children are rated significantly higher on internalizing

behavior than resilient children, with the overcontrolled children in negative control

families showing the highest levels of internalizing problem behavior. The results also

indicate that high or low negative parental control does not make a difference for the

resilient children: the mean scores on internalizing behavior are almost equal.

7

The value of eta-squared (η², see Table 6), which is the regression coefficient

(R²) for a non-linear regression curve, can be used as an estimate of effect size. For the

interaction effects in this study, we found values of .04, which resemble the effect

sizes of the interaction effects in study 1 (see Table 5). Values of the observed

7 Parenting by type interactions were also examined with HMRA, with types as dummy-coded

variables (resilients versus undercontrollers and resilients versus overcontrollers) and parenting

dimensions as continuous variables. This revealed one significant interaction effect, i.e.

undercontrolled/resilient types by negative control, in predicting child externalizing behavior (R²

change = .027; F change (1,369) = 15.77, p < .001). This interaction was of the ‘synergistic or

enhancing’ type (Cohen et al., 2003, p. 285), with both predictors affecting externalizing behavior in

the same direction and their combination producing an additional effect. The effect size of this

interaction was .04, which equals the effect size of the GLM interaction effect (see Table 6), although

the number of included subjects was larger in the HMRA analysis (N = 374) than in the GLM analyses

(N = 121). The overcontrolled/resilient types by negative control interaction effect on internalizing

behavior, which was significant in the GLM analysis (N = 110), showed a statistical trend (p = .085) in

the HMRA analysis (N = 345).


statistical power of the interaction effects equal .59 and .65 and are also comparable

to the power estimations of study 1.

No significant interactions were found between the personality types and

positive parenting. However, there were significant main effects for both positive

parenting and personality type. Undercontrollers are rated significantly higher on

externalizing behavior than resilients, and overcontrollers are rated significantly

higher on internalizing behavior than resilients. Children, who are supported by their

parents, show significant lower levels of problem behavior in comparison with

children receiving low levels of positive parental behavior.

GLM analyses showed two significant interaction effects. Undercontrolled

children, (characterized by below average levels on both conscientiousness and

benevolence) showed significantly more externalizing behavior than resilient children,

especially when exposed to highly negative parental control. Overcontrolled children

(scoring low on emotional stability and extraversion) exhibited higher levels of

internalizing behavior than resilient children, with the highest levels of internalizing

behavior for children experiencing high negative parental control. Resilient children

were not affected by negative control and showed no problem behavior at all. In

Time one categorization predicts Time two problem behavior

Table 6 reports how Time one categorization predicts Time two problem

behavior. The Time one Negative control by Undercontrollers versus Resilients

interaction predicted Time two externalizing behavior. The results further illustrate

that the Time one personality types significantly predict Time two externalizing and

internalizing behavior. The Negative control categories also predicted Time two

externalizing and, to a lesser extent, internalizing behavior, whereas the Positive

parenting categories mainly predicted externalizing behavior.

Discussion

Study 2 adopted a person-centered approach and examined the interactive

effects of parenting and the personality types resilient, undercontrolled and

overcontrolled. The types were assigned according to the standard procedure outlined

by Asendorpf et al. (2001). Children were assigned to groups experiencing differences

in parenting based on their scores on the parenting variables: children experiencing

below or above average positive parenting and children exposed to below or above

average negative control.

110 Chapter 3

addition the results showed that the strongest Time one interaction effect still

predicted Time two externalizing behavior. The Time one main effects of personality

types and parenting categories largely held up over time.

The results of study 2 are partially in line with the results of study 1. Children

scoring below the mean on the dimensions benevolence and conscientiousness in

study 1 correspond to the undercontrolled children in study 2. These children are at

risk for externalizing behavior when exposed to high levels of negative parental

control. Children with above average scores on benevolence and conscientiousness in

study 1 resemble the resilient children in study 2. These children are not affected by

restrictive control. However in contrast to study 1, positive parenting, benevolence

and conscientiousness were not identified as moderators in the prediction of

externalizing behavior, because there was no significant interaction between positive

parenting and the dichotomy undercontrolled/resilient.

This study demonstrates the utility of personality types as moderators of the

relationship between environmental variables such as parenting and problem

behavior. Groups of children with similar configurations of personality characteristics

were a priori identified. Notice that to group children in a variable-centered approach,

researchers or clinicians have to define arbitrary ad hoc cut-off scores on the

dimensions or divide the children into groups based on a median-split. The present

study illustrates that a priori defined types have predictive validity in explaining

maladaptive child behavior and therefore they should be considered as a viable

alternative for the variable-centered approach in developmental and clinical research.

However, it has to be recognized that identifying the relevant personality dimensions

(adopting the variable-centered approach) remains important because the scores on

the dimensions are indispensable to group children into the types and to describe the

typical personality profiles of the resilient, undercontrolled and overcontrolled

children. The knowledge that undercontrolled children are at risk for developing

externalizing behavior, in particular when they have experienced a highly negative

parental style may have practical utility when one has to decide whether they are at

risk. However in order to explain why they are at risk it may be useful to refer to the

dimensions defining the particular personality profile of each personality type.

Therefore, types and dimensions should be considered as useful classes of information

generated by two complementary research strategies, i.e. the variable- and the person-

centered approach.


Past moderator research has been hampered by various methodological and

statistical shortcomings. To reduce these problems we used (a) a non-restricted large

population sample, (b) reliable measures, (c) more than one informant, and (d) a

follow-up measurement within a three year interval. The hierarchical moderated

regression analyses provide evidence for several significant parenting by child

personality interactions. Six major conclusions can be drawn: (a) several significant

parenting by personality interactions were identified, but they were more prominent

for externalizing than for internalizing behavior; (b) two types of interactions can be

identified, i.e. buffering interactions and interference interactions; (c) the interactions

are largely replicable across informants, i.e. parents and children, and across a three

year interval, i.e. for children and adolescents; (d) the most prominent personality

domains that figure in the interactions are Benevolence and Conscientiousness, (e) the

person-centered study also showed meaningful personality by parenting interactions,

General discussion

A comprehensive test of personality as a moderator

The present research extends in several ways previous empirical studies

documenting the role of personality-environment interactions for the development of

child problem behavior. First, other studies linking parenting and child characteristics

have been restricted to the assessment of the child’s temperament or various specific

personality characteristics. Our study includes a broadband measure of child

personality based on the Five Factor Model (FFM), that can be regarded as a robust

reference-model (De Raad & Perugini, 2002), applicable to both children and

adolescents (Shiner, 1998). In a second study, we consider child personality from a

person-centered approach, studying the role of personality types in interaction with

parenting. Second, review of the literature indicates that most studies investigating

parenting by temperament or personality interactions, are limited to negative child

parenting practices, such as coerciveness or restrictive control. The present research

includes both a positive and a negative dimension of parenting behavior. Finally, both

internalizing and externalizing behavior are studied as outcome variables, whereas

past research has focused mostly on externalizing behavior.

Reliable moderator effects

112 Chapter 3

and (f) interactions measured at Time one predict problem behavior across a three

year interval.

The present research identified moderating effects predominantly for

externalizing behavior, whereas for internalizing behavior parenting and child

personality turned out to be primarily additive effects. Strong negative first order

effects of extraversion and emotional stability on internalizing behavior were clearly

present: children scoring low on these adaptive traits are more prone to internalizing

behavior. Independent effects of extraversion and emotional stability on externalizing

behavior were much smaller. This corroborates previous research findings by Finch &

Graziano (2001) showing that the personality dimensions benevolence, extraversion

and neuroticism are strongly related to depression.

The weaker associations between the parenting variables and internalizing

behavior, in comparison with externalizing behavior, could be explained by the

theoretical and empirical grounds on which the parenting questionnaire is based, i.e.

the heavy emphasis in social learning theory on antisocial behavior and parenting. For

example, the majority of the behavior control items in the questionnaire

predominantly describe how a parent reacts when a child misbehaves, and hence there

is a limited emphasis on parental behavior that may reduce internalizing behavior. A

study of Barber (1996) showed that behavioral control, referring to the attempts to

control or manage children’s behavior, is uniquely associated with externalizing, and

not with internalizing adolescent problem behavior. However, there was no evidence

for a particular association between internalizing behavior and psychological control,

aimed at the psychological and emotional development of the child.

Types of interaction

A buffering interaction pattern was identified for the negative control by

personality interactions. Negative parental control tends to be a risk factor for

externalizing behavior, especially for children rated low or around the mean on

benevolence and conscientiousness. Hence, these personality domains function as

protective factors: children rated high on these domains do not show problem

behavior at all, even in the presence of parental restrictive control. Interference or

antagonistic interactions were present for positive parenting and benevolence. Again,

Differences in the prediction of externalizing and internalizing behavior


the results show that children rated high on benevolence do not show problem

behavior at all, even when they are deprived of positive parenting. The combination of

above mean level positive parenting and below mean level benevolence diminishes

externalizing behavior.

The replicability of the interactions

Some of the findings in this study were replicable across judges and across a

three year interval. Our study included two measurement moments: the first when

most of the subjects were school-aged children, the second when 88% of the sample

were adolescents. The benevolence by negative control and the conscientiousness by

negative control interactions predicting externalizing problem behavior for the school-

aged children remained three years later. These interactions measured at Time one,

were also predictive for Time two externalizing behavior. The first-order effects of

personality showed great stability over time, whereas the effects of parenting

behavior, especially on internalizing behavior, slightly diminished over time.

The role of benevolence and conscientiousness

The most prominent interactions include the personality domains benevolence

and conscientiousness. This finding is consistent with Prinzie et al. (2003) and related

to studies investigating parenting by temperament interactions. The interactions of

parenting and benevolence and conscientiousness may reflect ‘evocative person-

environment transactions’ (Caspi, 1998, p. 357): on the basis of their unique

personality characteristics, individuals act, the environment reacts, resulting in

mutually interlocking evocative transactions. Patterson (1982) also described this

process as part of the coercive family process.

The analyses showed no significant interactions for imagination, extraversion

and emotional stability, and hence the present results are consistent with Prinzie et al.

(2003).

The importance of interaction effects

The emphasis in the present study on the importance of parenting by child

personality interaction effects is not meant to minimize or to cloud the main effects of

the child’s personality or the effects of parenting. The importance of main effects is

clearly documented in tables 2,3,5 and 6 and is also recognized and stressed in several

114 Chapter 3

studies referred to in the introduction to this study. Although it is evident that main

effects explain a greater proportion of the variance of externalizing and internalizing

behavior, documenting reliable interaction effects essentially qualifies the main

effects.

The study of interaction effects is important both from a theoretical and from a

practical perspective. Reliable interaction effects qualify theories that are based on

straightforward and unqualified effects of parenting and child personality. Moreover,

because most theories are tested by adopting a variable-centered approach, the

implications at the individual or person level are usually underestimated. The practical

implications at the individual level can be judged more appropriately by adopting a

person-centered approach because it specifies the type of subjects for whom the

general rule does or does not apply e.g. negative control has detrimental effects on

(Time one) internalizing (see table 6). For instance, high versus low negative parental

control does make a difference for the degree of internalizing of the 162

overcontrollers but the same differences in parenting do not affect the internalizing

scores of the 185 resilient children. Hence although this interaction effect only

explains four percent of the variance it nevertheless shows that parenting is an

additional risk factor for less than half of the sample. Moreover in this case the main

effect of parenting explains less of the internalizing variance than the interaction

effect.

A similar procedure can be applied to the data reported in the variable-

centered study. Although this require setting arbitrary ad hoc cut-off points to

calculate the number of subjects to whom the general rule does or does not apply, the

logic is essential the same. In table 5 we report the interaction between personality

and parenting on externalizing as well as the simple slopes for the relationship

between parenting and externalizing for three (ad hoc) groups: those scoring high on

benevolence, moderate and low on benevolence. Figure 1A shows that the parenting –

externalizing relationship is significant for the group scoring below and around the

mean but not for the group scoring above the mean. Moreover calculation of the R2’s

for each of these groups shows that the effect of parenting explains 11% of the

variance of externalizing for the low benevolence group and 10% for the medium

group where it only explains 5% for the high benevolence subjects. This comparison

illustrates that the overall main effect of parenting underestimates the importance of

parenting for children low or around the mean on benevolence but overestimates it for


the other ad hoc personality group. Finally this sort of data not only has theoretical

significance but it is also important for clinical researchers who want to screen the

children and parents who will benefit from treatment such as enrollment in a parenting

training program.

Variable- versus person-centered approaches

Including personality types in interactions with parenting, instead of

personality variables, enriched the findings of the first study. Neither the number of

significant effects nor the effect sizes are remarkable in comparison with the findings

of study 1. After all, it was not our goal to enhance the likelihood of finding

interaction effects by using categorical variables, a strategy that is dissuaded by

McClelland and Judd (1993). The main advantage of the personality type-approach is

that it, instead of considering only one variable at a time, combines children with a

given profile of scores on multiple personality variables into one category. From study

1 we concluded that children rated low on conscientiousness or low on benevolence

were at risk for externalizing behavior when exposed to parental negative control.

Including personality types in the interactions, leads to the conclusion that children

rated below the mean on both conscientiousness and benevolence, i.e. the

undercontrolled children, are at risk for externalizing behavior when exposed to

parental negative control in comparison with resilient children. This study further

showed that children scoring low on emotional stability and low on extraversion, i.e.

the overcontrolled children, are at risk for internalizing behavior when exposed to

parental negative control. Resilient children are protected against possible negative

effects of parental negative control. With the variable-centered approach in study 1,

we only found evidence for main effects of personality and parenting in the prediction

of internalizing behavior. However studying overcontrolled children with a profile

indicating low scores on the two relevant dimensions (emotional stability and

extraversion) provides extra information on moderators of internalizing behavior.

These results emphasize the relevance of using personality types in

developmental research and clinical practice. Types clearly have predictive validity in

the study of maladaptive child behavior and should be further integrated in research

about personality-environment interactions. Types also have practical value because

they combine information on several personality domains in a single case. This

facilitates for example the diagnostic process as well as clinical decision making.

116 Chapter 3

To our knowledge there is only one other study, which has investigated

interactions between parenting and personality types in predicting problem behavior.

Dubas, Gerris, Janssens, and Vermulst (2002) found that undercontrolled adolescents

exposed to high levels of restrictive control, scored higher on internalizing behavior

than resilients and overcontrollers in less restrictive families. This finding is in

contrast with previous studies evaluating the external validity of the types. Dubas et

al. attribute this to the co-morbidity of externalizing and internalising problems.8 Our

study corroborates the externalizing tendencies for undercontrollers and the

internalizing tendencies for overcontrollers.

Limitations of this Study and Future Research

A study with biologically related parents and children, does not exclude that

associations between parental behavior and child measures are due to common factors

instead of environmental influences. Environmental influences, like parenting, partly

reflect genetic influences, i.e. genotype-environment covariance (Lahey, Waldman &

McBurnett, 1999). An example of such a passive type of genotype-environment

interaction is a child diagnosed as having a conduct disorder reared by parents

showing aggressive behavior. In such a case parental and child behavior are correlated

(Rowe, 2003).

The present study found evidence for a moderator effects model, explaining

child outcomes by the simple interaction of constitutional vulnerability (child

personality), and environmental factors (parenting). However, such a model does not

imply that conclusions can be drawn about reciprocal and recurrent interactions over

time between organism and the environment, as postulated by the transactional effects

model (Lytton, 1990).

The reliance in the present study on a questionnaire measure to assess

parenting is a potential limitation. Studies investigating effects of parenting on social

development, based on direct behavioral observations usually show larger effect sizes

than studies centering on parental reports (Collins et al., 2000). However, the need for

a large sample in order to maximize statistical power forced us to use self-rating

questionnaires. The GPBS was developed with great care, and has good psychometric

8 In an additional GLM analysis of variance, we checked if this interaction could also be

found in our data, but this was not the case.


properties. Besides, the questionnaire was administered to two informants, parents and

children. The combination of child ratings of parenting with parental ratings of

personality and problem behavior reduces shared method variance and allows

replication of the results found for parental ratings. It should be acknowledged that the

use of aggregated scores for parental behavior might mask differential effects of

maternal or paternal parental behavior, in particular for positive parenting because the

correlation between ratings supplied by mothers and fathers was below .3. We also

recognize the possibility that parental disagreement about child rearing also might

predict problem behavior.

Although we used measures at two assessment moments in order to replicate

the basic findings, these follow-up measures are not independent and hence

replication in independent samples is warranted. Nevertheless, the five broadband

measures of personality and the two parenting dimensions used in this study, can be

considered as useful measures to detect personality by parenting interactions.

Main effects of parental behavior on problem behavior of the child often

conceal interactions with personality of the child. This often leads to the erroneous

conclusion that some parenting practices (e.g. restrictive parenting) are detrimental or

bad and should be avoided at all costs. The present study documents that the effects of

parenting behavior should be qualified and hence that it may be premature to blame

all parents adopting a common parenting style for the problem behavior of children

without taking into account the moderating effects of the child’s personality.

Although in our analyses we consistently controlled for effects of child gender

and age, we did not look for differential effects of age and gender. Further research

could examine whether parenting by personality interactions are different for girls and

boys or depend on age groups, e.g. school-aged children versus adolescents. The

importance of this differentiation can be illustrated by research on externalizing

behavior showing evidence for a developmental typology of delinquency, i.e. the

childhood-onset category and the adolescent-onset category, characterized by

different etiologies (Moffitt, Caspi, Harrington & Milne, 2002). Three-way

interactions in hierarchical multiple regression analyses could verify if parenting by

personality interactions differ for boys and girls and according to age.

The finding that stable personality characteristics play a major role in child

problem behavior should not be equated with the pessimistic or deterministic view

that changing problematic behavior is impossible, but rather should be taken as

118 Chapter 3

evidence for a more realistic perspective (Lytton, 1990). Parents can have an impact

on their children’s behavior, but they will not be able to change major personality

traits, such as extraversion and neuroticism (Rowe, 1990). The present studies support

the view that parents may receive too much blame for the behavioral difficulties of

their children because the moderating role of the child’s personality is often ignored.

On the other hand, environmental factors combined with certain personality types

may increase the likelihood for the development of problem behavior.


References

Aguinis, H. (1995). Statistical power problems with moderated multiple

regression in management research. Journal of Management, 21, 1141-1158.



Aiken, L. S., & West, S. G. (1991). Multiple regression: testing and

interpreting interactions. Newbury Park: Sage Publications.

Anderson, K. E., Lytton, H., & Romney, D. M. (1986). Mothers’ interactions

with normal and conduct-disordered boys: who affects whom? Developmental

Psychology, 22, 604-609.

Asendorpf, J. B. (2003). Head-to-head comparison of the predictive validity of

personality types and dimensions. European Journal of Personality, 17, 327-346.

Asendorpf, J. B., Borkenau, P., Ostendorf, F. & van Aken, M. (2001). Carving

personality description at its joints: confirmation of three replicable personality

prototypes for both children and adults. European Journal of Personality, 15, 169-

198.

Barber, B. (1992). Family, personality, and adolescent problem behaviors.

Journal of Marriage and Family, 54, 69-79.

Barber, B. K. (1996). Parental psychological control: revisiting a neglected

construct. Child Development, 67, 3296-3319.



externalizing behavior. Developmental Psychology,34, 982-995.



Belsky, J., Hsieh, K., & Crnic, K. (1998). Mothering, fathering, and infant

negativity as antecedents of boys’ externalizing problems and inhibition at age 3

years: differential susceptibility to rearing experience? Development and

Psychopathology, 10, 301-319.



Capaldi, D., & Patterson, G. R. (1989). Psychometric properties of fourteen

latent constructs from the Oregon Youth Study. New York: Springer-Verlag.

120 Chapter 3

Carlo, G., Roesch, S. C., & Melby, J. (1998). The multiplicative relations of

parenting and temperament to prosocial and antisocial behaviors in adolescence.

Journal of Early Adolescence, 18, 266-290.




Wiley.

Caspi, A., Henry, B., McGee, R. O., Moffitt, T. E., & Silva, P. A. (1995).

Temperamental origins of child and adolescent behavior problems: from age three to

age fifteen. Child Development, 66, 55-68.

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

personality traits in young adulthood : longitudinal evidence from a birth cohort.


Chaplin, W. F. (1991). The next generation of moderator research in

personality psychology. Journal of Personality, 59, 143-178.

Cohen, J., Cohen, P., West, S. G., & Aiken, L. S. (2003). Applied multiple

regression/ correlation analysis for the behavioral sciences (3th ed.). Mahwah, New


Colder, C. R., Lochman, J. E., & Wells, K. C. (1997). The moderating effects

of children’s fear and activity level on relations between parenting practices and

childhood symptomatology. Journal of Abnormal Child Psychology, 25, 251-263.

Collins, W. A., Maccoby, E. E., Steinberg, L., Hetherington, E. M. &

Bornstein, M. H. (2000). Contemporary Research on Parenting. The case for nature

and nurture. American Psychologist, 55, 218-232.

Costa, P. T., Herbst, J. H., McCrae, R. R., Samuels, J. & Ozer, D. J. (2002).

The replicability and utility of three personality types. European Journal of


Darling, N. & Steinberg, L. (1993). Parenting style as context: An integrative

model. Psychological Bulletin, 113, 487-496.

Deater-Deckard, K. (2001). Annotation: Recent research examining the role of

peer relationships in the development of psychopathology. Journal of Child



Deater-Deckard, K., & Dodge, K. A. (1997). Externalizing behavior problems

and discipline revisited: nonlinear effects and variation by culture, context, and

gender. Psychology Inquiry, 8, 161-175.

Deater-Deckard, K, Dodge, K. A., Bates, J. E., & Pettit, G. S. (1998). Multiple

risk factors in the development of externalizing behavior problems: group and

individual differences. Development and Psychopathology, 10, 469-493.

Deković, M. (1999). Risk and protective factors in the development of

problem behavior during adolescence. Journal of Youth and Adolescence, 28, 667-

685.

De Fruyt, F., Mervielde, I. & Van Leeuwen, K. (2002). The consistency of



De Raad, B. & Perugini, M. (2002). Big Five factor assessment: introduction.




ecology of antisocial behavior. In D. Cicchetti & D.J. Cohen (red.), Developmental

Psychopathology. Vol. 1 Theory and Methods (pp. 421-471). New York: John Wiley

& Sons, Inc.

Dubas, J. S., Gerris, J. R. M., Janssens, J. R. M. & Vermulst, A. A. (2002).

Personality Types of Adolescents: Concurrent Correlates, Antecedents, and Type X

Parenting Interactions. Journal of Adolescence, 25, 79-92.

Eisenberg, N., Fabes, R. A., Guthrie, I. K., & Reiser, M. (2000). Dispositional

emotionality and regulation: their role in predicting quality of social functioning. Journal


Finch, J. F., & Graziano, W. G. (2001). Predicting depression from temperament,

personality, and patterns of social relations. Journal of Personality, 69, 27-49.

Forehand, R., Miller, K. S., Dutra, R., & Chance, M. W. (1997). Role of

parenting in adolescent deviant behavior: replication across and within two ethnic

groups. Journal of Consulting and Clinical Psychology, 65, 1036-1041.

Frosch, C. A., & Mangelsdorf, S. C. (2001). Marital behavior, parenting

behavior, and multiple reports of preschoolers' behavior problems: mediation or

moderation? Developmental Psychology, 37, 502-519.

122 Chapter 3

Gallagher, K. C. (2002). Does child temperament moderate the influence of

parenting on adjustment? Developmental Review, 22, 623-643.

Gerris, J. R. M., Houtmans, M. J. M., Kwaaitaal-Roosen, E. M. G., de

Schipper, J. C., Vermulst, A. A., & Janssens, J. M. A. M. (1998). Parents,

Adolescents and Young Adults in Dutch Families: a longitudinal study. Nijmegen:

Institute of Family Studies.

Gruner, K., Muris, P., & Merckelbach, H. (1999). The relationship between

anxious rearing behaviours and anxiety disorders symptomatology in normal children.

Journal of Behavior Therapy and Experimental Psychiatry, 30, 27-35.

Haapasalo, J., & Tremblay, R. E. (1994). Physically aggressive boys from

ages 6 to 12: family background, parenting behavior, and prediction of delinquency.

Journal of Consulting and Clinical Psychology, 62, 1044-1052.

Harris, J. R. (1998). The nurture assumption: why children turn out the way

the do. New York: Free press.

Hellinckx, W., De Munter, A., & Grietens H. (Eds.) (1991). Gedrags- en

emotionele problemen bij kinderen. Een epidemiologisch onderzoek bij 6- tot 12-

jarigen in Vlaanderen. Deel 1: Beschrijving en screening van probleemgedrag.

[Behavioral and emotional problems in children. An epidemiological study of 6- to 12

year olds. Part 1: Description and screening of problem behavior.] Leuven-Apeldoorn:

Garant.

Holmbeck, G. N. (1997). Toward terminological, conceptual, and statistical

clarity in the study of mediators and moderators: examples from the child-clinical and

pediatric psychology literatures. Journal of Consulting and Clinical Psychology, 65,

599-610.

Jaccard, J., & Wan, C. K. (1995). Measurement error in the analysis of

interaction effects between continuous predictors using multiple regression: multiple

indicator and structural equation approaches. Psychological Bulletin, 117, 348-357.

Jessor, R., Van Den Bos, J., Vanderryn, J., Costa, F. M., & Turbin, M. S.

(1995). Protective factors in adolescent problem behavior: moderator effects and

developmental change. Developmental Psychology, 31, 923-933.

Kohnstamm, G. A., Halverson, C. F., Mervielde, I., & Havill, V. L. (1998).

Parental descriptions of child personality: Developmental antecedents of the Big

Five? Mahway, NJ: Erlbaum.


Krueger, R. F., Schmutte, P. S., Caspi, A., Moffitt, T. E., Campbell, K., & Silva,

P. A. (1994). Personality traits are linked to crime among men and women: evidence

from a birth cohort. Journal of Abnormal Psychology, 103, 328-338.

Lahey, B. B., Waldman, I. D., & McBurnett, K. (1999). Annotation: The

development of antisocial behavior: an integrative causal model. Journal of Child


Lengua, L. J., Wolchik, S. A., Sandler, I. N., & West, S. G. (2000). The

additive and interactive effects of parenting and temperament in predicting problems

of children of divorce. Journal of Clinical Child Psychology, 29, 232-244.

Loeber, R., & Dishion, T. (1983). Early predictors of male delinquency: a

review. Psychological Bulletin, 94, 68-99.


predictors of juvenile conduct problems and delinquency. In M. Tonry & N. Morris

(Eds.) Crime and Justice: A review of research (Vol.7, pp. 29-149). Chicago: University

of Chicago Press.

Lytton, H. (1990). Child and parent effects in boys’ conduct disorder: a

reinterpretation. Developmental Psychology, 26, 683-697.

Maccoby, E. E., & Martin, J. A. (1983). Socialization in the context of the

family: parent-child interaction. In Mussen, P.H. (Ed.), Handbook of child

psychology. Vol. IV. Socialization, personality and social development (pp. 1-101).

New York: Wiley.

McClelland, G. H., & Judd, C. M. (1993). Statistical difficulties of detecting

interactions and moderator effects. Psychological Bulletin, 114, 376-390.




Mervielde, I., & De Fruyt, F. (1999). Construction of the Hierarchical

Personality Inventory for Children (HiPIC). In I. Mervielde, I. Deary, F. De Fruyt, & F.

Ostendorf (Eds.), Personality psychology in Europe. (pp. 107-127). Tilburg University

Press.




124 Chapter 3

Moffitt, T. E., Caspi, A., Harrington, H., & Milne, B. J. (2002). Males on the

life-course-persistent and adolescence-limited antisocial pathways: Follow-up at age

26 years. Development and Psychopathology, 15, 179-207.

Muris, P., Schmidt, H., Lambrichs, R., & Meesters, C. (2001). Protective and

vulnerability factors of depression in normal adolescents. Behaviour Research and

Therapy, 39, 555-565.

O’Connor, B. P., & Dvorak, T. (2001). Conditional associations between

parental behavior and adolescent problems: a search for personality-environment

interactions. Journal of Research in Personality, 35, 1-26.

Olweus, D. (1980). Familial and temperamental determinants of aggressive

behavior in adolescent boys: a causal analysis. Developmental Psychology, 16, 644-660.

Paterson, G., & Sanson, A. (1999).The association of behavioural adjustment to

temperament, parenting and family characteristics among 5-year old children. Social


Patterson, G. R. (1982). Coercive family process. Eugene, OR: Castalia Press.




management practices and delinquency. Child Development, 55, 1299-1307.

Prinzie, P., Onghena, P., Hellinckx, W., Grietens, H., Ghesquière, P., &

Colpin, H. (2003). The additive and interactive effects of parenting and children’s

personality on externalising behaviour. European Journal of Personality, 17, 95-117.

Richter, J. (1994). Parental rearing and aspects of psychopathology with

special reference to depression. In C. Perris, W. A. Arrindell a M. Eisemann (Eds.),

Parenting and psychopathology (pp. 235-251). Chisester: John Wiley & Sons.

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

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

and personality development (pp. 105-176). New York : Wiley.

Rothbaum, F., & Weisz, J.R. (1994). Parental caregiving and child externalizing

behavior in nonclinical samples: a meta-analysis. Psychological Bulletin, 116, 55-74.

Rowe, D. C. (1990). As the twig is bent? The myth of child-rearing influences

on personality development. Journal of Counselling and Development, 68, 606-611.


Rowe, D. C. (2003). Assessing genotype-environment interactions. In R.

Plomin, J. C. Defries, I. W. Craig, & P. McGuffin, (Eds.), Behavioral Genetics in the

Postgenomic Era (pp. 71-86). Washington DC: American Psychological Association.

Rubin, K. H., Burgess, K. B., Dwyer, K. M., & Hastings, P. D. (2003).

Predicting preschoolers’ externalising behaviors from toddler temperament, conflict

and maternal negativity. Developmental Psychology, 39, 164-176.

Ruchkin, V. V., Eisemann, M., & Hägglöf, B. (1999). Coping styles in

delinquent adolescents and controls: the role of personality and parental rearing.

Journal of Youth and Adolescence, 28, 705-717.

Sameroff, A. J. (2000). Developmental systems and psychopathology.

Development and Psychopathology, 12, 297-312.


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

Shiner, R., & Caspi, A. (2003). Personality differences in childhood and



Siqueland, L., Kendall, P. C., & Steinberg, L. (1996). Anxiety in children:

perceived family environments and observed family interaction. Journal of Clinical

Child Psychology, 25, 225-237.

Stoolmiller, M. (2001). Synergistic interaction of child manageability

problems and parent-discipline tactics in predicting future growth in externalizing

behavior for boys. Developmental Psychology, 37, 814-825.

Stormshak, E. A., Bierman, K. L., McMahon, R., Lengua, L. J., & Conduct


behavior problems in early elementary school. Journal of Clinical Child Psychology,

29, 17-29.

Stouthamer-Loeber, M., Loeber, R., Farrington, D. P., Zhang, Q., van

Kammen, W., & Maguin, E. (1993). The double edge of protective and risk factors for

delinquency: interrelations and developmental patterns. Development and



Brunner/Mazel.

Van den Bergh, B. (1997). Kindertijd. Kinderen en ouders over de leefsituatie

van kinderen op lagere schoolleeftijd in Vlaanderen. [Childhood. Children and

126 Chapter 3

parents about the quality of life of elementary school-aged children].

Leuven/Apeldoorn: Garant.

Van Leeuwen, K. G., & Vermulst, A. A. (in press). The Ghent Parental

Behavior Scale: some psychometric properties. European Journal of Psychological

Assessment.

Van Leeuwen, K., De Fruyt, F., & Mervielde, I. (in press). A longitudinal

study of the utility of the resilient, overcontrolled and undercontrolled personality

types as predictors of children’s and adolescents’ problem behavior. International

Journal of Behavioral Development.


CBCL/4-18. [Manual of the CBCL/4-18.] Erasmus Universiteit Rotterdam, Afdeling

Kinder- en Jeugdpsychiatrie.

Verhulst, F. C., Van der Ende, J., & Koot, H. M. (1997a). Handleiding voor de

Teacher’s Report Form. [Manual of the TRF.] Erasmus Universiteit Rotterdam,

Afdeling Kinder- en Jeugdpsychiatrie.

Verhulst, F. C., Van der Ende, J., & Koot, H. M. (1997b). Handleiding voor

de Youth Self Report. [Manual of the YSR.] Erasmus Universiteit Rotterdam,


Wakschlag, L. S., & Hans, S. L. (1999). Relation of maternal responsiveness

during infancy to the development of behavior problems in high-risk youths.


Weiss, B., Dodge, K. A., Bates, J. E., & Pettit, G. S. (1992). Some

consequences of early harsh discipline: child aggression and a maladaptive social

information processing style. Child Development, 63, 1321-1335.

Wootton, J. M., Frick, P. J., Shelton, K. K., & Silverthorn, P. (1997).

Ineffective parenting and childhood conduct problems: the moderating role of callous-

unemotional traits. Journal of Consulting and Clinical Psychology, 65, 301-308.

Table 1

Factor Pattern matrix for the GPBS dimensions Positive parenting and Negative control across a three year interval Time 1 Time 2 M F CM CF M F CM CF POS CON POS CON POS CON POS CON POS CON POS CON POS CON POS CONPOS .69 -.12 .78 -.11 .73 -.07 .73 -.12 .56 -.18 .65 -.22 .56 .16 .53 -.22RUL .72 .17 .75 .11 .76

.63.09 .81 .17 .81 .18 .78 .13

.01 .87 -.26 .88 .11

AUT .33 .02 .53 -.01 .00 .61 .01 .28 -.03 .55 .51 .01 .63 .07DIS

.15 .60 .17 .66 .09 .73 .09 .65 .19 .34 .27 .44 .19 .67 .15 .50IGN -.05 .45 .00 .44 -.12 .76 .02 .49 -.11 .52 -.18 .40 -.05 .61 -.03 .25HAR -.08 .63 -.14 .50 .04 .62 -.11 .73 -.15 .59 -.01 .53 -.11 .31 -.11 .72% Variance 35.47 39.94 50.69 47.07 32.06 35.67 39.82 39.54Cronbach α .82.86 .80 .91 .80 .90 .87 .90 .86 .77 .90 .75 .90 .91 .81 .80N of items 21 14 21 14 20 14 20 14 21 14 21 14 20 14 20 14 GPBS = Ghent Parental Behavior Scale

M = ratings by mothers ; F = ratings by fathers ; CM = children rating mothers ; CF = children rating fathers

POS = Positive parenting ; CON = Negative control

POS = Positive parental behavior ; RUL = Setting rules ; AUT = Autonomy ; DIS = Discipline; IGN = Ignoring ; HAR = Harsh punishment

128 Chapter 3

Table 2

Moderators of Externalizing behavior Time 1 Time 2 Parental ratings of

parenting Child ratings of parenting

Parental ratings of parenting

Child ratings of parenting

∆F ∆F B B ∆F B ∆F B Negative control Sex, Age 11.63 a -.13 b, .02 11.70 a -.16 b, .02 8.14 a -.07, .00 8.10 a -.11, .00 BE, CON

271.23 a -.57 a, .21 a 238.44 a -.64 a, .08 b 294.03 a -.66 a, .12 a 280.61 a -.69 a, .06 BE x CON 30.61 a -.15 a 8.44 b -.08 b 51.60 a -.18 a 28.15 a -.15 a Sex, Age 11.63 a -.19 b, .00 11.70 a -.20 b, .02 8.14 a -.10, .01 8.10 a -.13, .01 CO, CON 105.14 a -.35 a, .30 a 76.50 a -.40 a, .18 a 101.81 a -.42 a, .25 a 85.19 a -.41 a, .16 a CO x CON 11.95 a -.12 a 8.74 b -.11 b 14.13 a -.14 a 21.84Sex, Age 11.63 a -.32 a, .03 -.29 a, .02 8.10

a -.18 a 11.70 a -.34 a, .05 c 8.14 a a -.31 a, .03

ES, CON 58.73 a -.15 a, .37 a 26.44 a -.18 a, .23 a 45.08 a -.17 a, .33 a 33.93 a -.19 a, .27 a ES x CON 0.05 .01

8.14 0.03 -.01 0.04 -.01 0.13 .01

Sex, Age 11.63 a -.30 a, .03 11.70 a -.32 a, .04 c a -.24 b, .03 8.10 a -.25 b, .03 EX, CON 36.02 .29EX x CON 0.04 .01 Sex, Age 8.10

49.85 a .03, .37 a 15.32 a .04, .21 a a .02, .35 a 22.82 a .02, a 0.01 .00 .01 0.04 0.41 -.03 11.63 a -.31 a, .01 11.70 a -.33 a, .02 8.14 a -.26 b, .01 a -.28 a, .01

IM, CON 56.59 30.86 1.71

a -.13 a, .37 a 23.41 a -.17 a, .21 a 41.84 a -.14 a, .33 a a -.16 a, .27 a IM x CON -.05 2.94 -.07 0.71 -.03 2.26 -.06 Positive parenting Sex, Age 8.14 8.10 11.63 a -.18 b, .02 11.85 a -.19 b, .02 a -.11, -.01 a -.13, -.01 BE, POS 232.02 a -.65 a, -.03 233.18 a -.67 a, .02 277.42

11.17 0.82 .02 8.10

a -.73 a, .04 275.07 a -.73 a, .02 BE x POS a .10 a 0.54 -.02 .02 0.71 Sex, Age 11.63 a -.24 a, .01 11.85 a -.25 a, .01 8.14 a -.15, .00 a -.15, .01 CO, POS 64.47 a -.40 a, -.09 c 60.78 a -.42 a, .02 74.82 a -.49 a, -.07 72.61

a -.48 a, .00 CO x POS 7.58 b .10 b 0.33 .02 8.65 b .11 b 4.59 c .09 c Sex, Age 11.63 a -.39 a, .03 11.85 a -.39 a, .04 8.14 a -.42 a, .02 8.10 a -.39 a, .03 ES, POS 20.20 a -.16 a, -.19 a 9.00 a -.17 a, -.03 17.49 a -.19 a, -.13 b 13.72 a -.21 a, -.07 ES x POS 0.01 .00 0.63 -.03 1.49 -.05 2.02 -.06 Sex, Age 11.63 a -.38 a, .04 11.85 a -.39 a, .04 8.14 a -.36 a, .03 8.10 a -.32 a, .03 EX, POS 15.51 a .11 b, -.22 a 1.46 .07, -.02 8.17 a .06, -.18 a 2.30 .04, -.09 c

EX x POS 0.09 -.01 3.85 c .08 c 0.40 -.03 0.03 -.01 Sex, Age 11.63 a -.37 a, .02 11.85 a -.39 a, .01 8.14 a -.36 a, .00 8.10 a -.34 a, .01 IM, POS 17.11 a -.13 b, -.18 a 8.25 a -.17 a, -.01 14.42 a -.17 a, -.14 b 11.30 a -.19 a, -.06 IM x POS 4.74 c .09 c 0.59 .03 3.57 .08 0.64 .03

Note. With multiplicative terms, neither traditional unstandardized nor standardized regression coefficients are

appropriate to report. However, when the crossproduct is based on z-scores, which is known as Friedrich’s

procedure, it is appropriate to use the unstandardized solution with interaction terms (Aiken & West, 1991, p. 43-

44). In our study, all predictors are standardized, because they are factor scores.

BE = Benevolence; CO = Conscientiousness; ES = Emotional Stability, EX = Extraversion; IM = Imagination;

POS = Positive parenting; CON = Negative control

p ≤ .001 ; p ≤ .01 ; p ≤ .05 a b c

Child personality and Parental Behavior as Moderators of Problem Behavior 129 A Variable- and a Person-centered Approach

Table 3

Moderators of Internalizing behavior Time 1 Time 2 Parental ratings of




∆F B ∆F ∆F B ∆F B B Negative control Sex, Age 2.17 .01, .03 2.20 -.03, .03 1.34 .12, .03 1.49 .10, .03 BE, CON 24.38

1.00 Sex, Age 1.34

a -.22 a, .09 c 23.70 a -.28 a, -.08 c 36.56 a -.32 a, .08 33.91 a -.34 a, .01 BE x CON 9.44 b -.12 b -.04 2.27 -.06 3.15 -.07

2.17 .00, .02 2.20 -.03, .02 .11, .03 1.49 .10, .03 CO, CON 23.84 a -.23 a, .11 b 20.49 a -.26 a, -.04 22.14 a -.23 a, .13 b 17.94 a -.23 a, .05 CO x CON 2.55 -.06 0.56 -.03 0.01 .00 2.41 -.07 Sex, Age 2.17 -.17-.15 c, .05 b 2.20 -.17 c, .05 b 1.34 -.17 c, .04 1.49 c, .04 c ES, CON 160.73 a -.57 a, .12 a 150.31 a -.58 a, .01 175.83 a -.63 a, .09 c 170.95 a -.63 a, .06 ES x CON

-.03, .01 0.26 -.02 0.27 .02 0.10 .01 2.72 .05

Sex, Age 2.17 2.20 -.06, .02 1.34 .05, .01 1.49 .04, .01 EX, CON 54.59 a -.36 a, .19 a 40.98 a -.35 a, .02 59.58 a -.41 a, .18 a 52.76 a -.41 a, .13 b EX x CON 0.25 -.02 0.06 .01 1.52 -.05 0.06 Sex, Age 2.17 -.09, .00 2.20 -.13, .01 1.34 .01, .01 1.49

-.01 -.01, .01

IM, CON 28.67 a -.26 a, .14 a 22.93 a -.28 a, -.03 24.84 a -.25 a, .14 a 21.45 a -.26 a, .09 c IM x CON 1.32 -.04 1.09 -.04 0.04 -.01 2.87

-.07

Positive parenting Sex, Age 2.17 -.03, .03 2.23 -.03, .03 1.34 .09, .02 1.49 .10, .02 BE, POS 25.56 a -.24 a, -.10 c 22.09 a -.27 a, -.01 35.45 a -.34 a, -.05 34.32 a -.35 a, -.04 BE x POS 1.08 .04 1.09 -.04 0.11 .00 0.20 .02 Sex, Age 2.17 -.03, .02 2.23 -.02, .02 1.34 .07, .02 1.49 .10, .03 CO, POS 23.38 a -.23 a, -.10 b 19.89 a -.26 a, .00 a, -.10 c 19.59 a -.25 17.75 a -.25 a, -.05 CO x POS 1.31 .04 0.13 .02 1.19 .05 1.51 .05 Sex, Age 2.17 -.17 b, .05 b 2.23 -.17 b, .05 b 1.34 -.17 c, .04 c 1.49 -.18 b, .04 c ES, POS 23.38 a -.58 a, -.13 a 150.16 a -.58 a, -.02 171.47 a -.64 a, -.04 169.51 a -.63 a, -.04 ES x POS 1.31 .06 0.02 -.01 5.93 c .08 c 2.53 .05 Sex, Age 2.17 -.07, .02 2.23 -.07, .02 1.34 -.01, .01 1.49 .01, .01 EX, POS 44.26 a -.33 a, -.09 c 40.92 a -.35 a, .01 48.82 a -.39 a, -.06 48.14 a -.39 a, -.05 EX x POS 0.20 .02 2.03 .06 2.61 .07 3.47 .08 Sex, Age 2.17 -.12, .01 2.23 -.12, .01 1.34 -.04, .01 1.49 -.02, .01 IM, POS

1.74 26.66 a -.25 a, -.11 b 22.56 a -.27 a, .00 21.31 a -.25 a, -.08 20.14 a -.26 a, -.05

IM x POS 0.07 -.01 0.78 .04 -.05 1.97 .06 BE = Benevolence; CO = Conscientiousness; ES = Emotional Stability, EX = Extraversion; IM = Imagination;

POS = Positive parenting; CON = Negative control a p ≤ .001 ; b p ≤ .01 ; c p ≤ .05

130 Chapter 3

Table 4

Time 1 variables predicting Time 2 problem behavior Externalizing Time 2 Internalizing Time 2 Parental ratings of




Time 1 variables ∆F B ∆F B ∆F B ∆F B Negative control Sex, Age 5.76 b -.05, .03 5.78 b -.10, .03 0.80 .13, .03 0.80 .09, .03 BE, CON 115.47 a -.45 a, .17 a 101.69 a -.51 a, .03 15.84 a -.17 a, .11 c 12.74 a -.21 a, -.02 BE x CON 17.84 a -.15 a -.05 1.23 -.05 0.07 -.01

2.53

Sex, Age 5.76 b -.11, .01 5.78 b -.13, .01 0.80 .10, .02 0.80 .08, .02 CO, CON 63.36 a -.31 a, .23 a 46.24 a -.35 a, .10 b 13.65

Sex, Age 0.80 -.02, .04

a -.15 a, .13 b 9.12 a -.18 a, .00 CO x CON 13.95 a -.14 a 8.18 b -.11 b 1.72 .05 0.00 -.00

5.76 b -.20 c, .03 5.78 b -.24 b, .04 0.80 .01, .04 ES, CON 32.64 a -.11 b, .31 a 10.18

5.76

a -.14 a, .15 a 71.61 a -.42 a, .15 a 62.30 a -.43 a, .04 ES x CON 0.86 .04 1.88 -.06 0.17 .02 0.96 -.04 Sex, Age b -.19 c, .03 5.78 b -.23 b, .04 0.80 .09, .01 0.80 .07, .02 EX, CON 28.56 a .04, .31 a 5.74 b .05, .14 b 25.48 a -.24 a, .17 a 16.49 a -.24 a, .05 EX x CON

0.05 -.01 0.27 -.02 1.83 -.05 0.43 -.03

Sex, Age 5.76 b -.21 c, .02 5.78 b -.25 b, .02 0.80 .04, .01 0.80 .01, .01 IM, CON 30.29 a -.09 c, .30 a 9.19 a -.14 b, .14 a 13.31 a -.15 a, .14 a 7.67 a -.18 a, .02 IM x CON 0.36 0.12

-.03 6.11 c -.10 c -.02 4.70 c -.09 c

Positive parenting Sex, Age 0.80 5.76 b -.11, .03 5.91 b -.12, .03 .09, .02 0.80 .10, .03 BE, POS 101.12 a -.51 a, -.01 102.33 a -.52 a, .00 15.62 a -.19 a, -.10 c 14.08 a -.21 a, -.06 BE x POS 0.04

.08, .02 0.79 .03 0.00 .00 0.16 .02 .01

Sex, Age 5.76 b -.17 c, .01 5.91 b -.18 c, .01 0.80 0.80 .08, .02 CO, POS 42.60 a -.36 a, .04 41.95CO x POS 2.59 .06 .04 1.32 -.05

-.03, .04

a -.36 a, .00 12.21 a -.16 a, -.11 c 10.13 a -.17 a, -.06 0.98 0.50 .03

Sex, Age 5.76 b -.28 a, .03 5.91 b -.28 a, .04 0.80 0.80 -.02, .04 ES, POS 7.80 a -.12 b, -.11 4.61 b -.12 b,-.03 67.48 a -.42 a, -.11 b 63.59 a -.42 a, .-06 ES x POS 2.50 -.07 1.30

b, .04 0.37 -.03 3.28 .07 .04

Sex, Age 5.76 b -.27 b, .03 5.91 b -.27 0.80 .06, .01 0.80 .06, .02 EX, POS 5.95 b .09, -.14 b 1.33 .07, -.02 18.66 a -.21 a, -.09 c 17.20 a -.23 a, -.06 EX x POS 0.21 -.02 6.24 c .11 c 0.02 .01 0.31 .02 Sex, Age .01, .01 5.76 b -.28 a, .02 5.91 b -.28 a, .02 0.80 .00, .01 0.80 IM, POS 7.06 a

.01 -.11 c, -.10 c 4.41 c -.13 b, -.02 11.43 a -.15 a, -.11 c 8.70 a -.16 a, -.06

IM x POS 0.02 0.16 .02 1.68 -.06 0.05 .01

BE = Benevolence; CO = Conscientiousness; ES = Emotional Stability, EX = Extraversion; IM = Imagination;

POS = Positive parenting; CON = Negative control a p ≤ .001 ; b p ≤ .01 ; c p ≤ .05

Table 5

Tests of significance of difference between simple slopes (T-values) and Effect sizes of interactions TIME 1 TIME 2 T-values Effect sizes T-values Effect sizes le slopes Simple slopes Interaction Simp Interaction BFM BFM BFM BFM

+ 1 SD Mean - 1 SD r²Y.MI r²Y.M f² + 1 SD BFM Mean BFM

- 1 SD r²Y.MI r²Y.M f²

Externalizing behavior Negative control: parental ratings Neg. control x Benevolence 1.30 6.95 a 8.80 a - 0.504 3.88 - 5.53 a 0.529 0.05 - 1.62 a 7.87 a 7.18 a 0.596 0.554 0.10Neg. control x Conscientiousness 3.72 a - - 6 a 0.330

8.57 a 8.66 a 3.46 a 0.309 0.295 0.02 1.98 c 6.53 a 7.48 a 3.7 0.311 0.03

Negative control: child ratings Neg. control x Benevolence - 0.02 2.58 - 0.01 1.76 5.19 a - 5.3 b 4.21 a 2.90 b 0.481 0.474 - 2.08 c 1 a 0.568 0.544 0.06Neg. control x Conscientiousness 1.32 4.95 - - Positive parenting: parental ratings

a 5.76 a 2.96 b 0.252 0.240 0.02 - 0.26 4.17 a 6.48 a 4.67 a 0.308 0.278 0.04

Pos. parenting x Benevolence 1.51 0.467 - 0.99 - 3.13 b 3.34 a 0.477 0.02

Note. Effect size (f²)= (r² - r² ) / (1-r² ) (Aiken & West, 1991, p. 156)

BFM = Big Five Measure

Y.MI Y.M Y.MI

r²Y.MI = the squared multiple correlation from combined predictors by two sets of variables, M (main effects) and I (interaction effect)

r²Y.M = the squared multiple correlation resulting from prediction by set Ma p ≤ .001 ; b p ≤ .01 ; c p ≤ .05

Table 6

Category means and results of GLM with parenting categories and personality types Resilients Overcontrollers Undercontrollers Negative control Personality type Parenting X personality

type Low

(N) η² F Power ηLow* (N)

High* (N)

Low (N)

High (N)

High (N) F Power η² F ² Power

Externalizing T1 -0.66 (38)

0.05 (16) -0.25 1.32

38.38 .247 20.97 .152 5.53 .045 (28) (39) a 1.00 a 0.99 c 0.65

Externalizing T2 -0.56 (35) -0.45

-0.31 (11) (23) (33) 17.84 a .154 .987 18.27 a .157 .988 8.49 b .080 .823

Internalizing T1 (38) -0.49 (16)

0.18 (20)

0.99 (36) 4.23

-0.30 1.07

c .038 0.53 30.85 a .225 1.00 4.92 c .044 0.59

Internalizing T2 -0.54 .594 .173 .025

Undercontrollers Parenting X personality type

(35) -0.39 (11)

-0.02 (18)

0.77 (25) 4.96 c .055 17.84 a .987 2.16 .306

Resilients Overcontrollers Positive parenting Personality type

Low* High* F (N) (N) Low (N)

High (N)

Low (N)

High (N) η² Power F η² Power F η² Power

Externalizing T1 10.67 0.90 .086 .001 -0.01 (19)

-0.55 (44)

0.71 (42)

0.02 (22)

a .080 11.61 a 0.92 0.15 0.07

Externalizing T2 .006

Internalizing T1 (19) (44) 0.74 (28)

0.28 (22) 4.26 c .038 0.53 25.99 a .193 0.99 0.33 .003 0.09

-0.22 (15) -0.25

-0.54 (32) -0.51

-0.41 (37)

-0.16 (17) 6.87 b .066 .738 9.08 b .086 .847 0.57 .117

Internalizing T2 -0.08 (15)

-0.50 (32)

0.30 (23)

0.24 (18) 1.63 .019 .243 9.19 b .099 .850 0.94 .011 .160

Note: * Low and high refer to categories of parentinga p ≤ .001 ; b p ≤ .01 ; c p ≤ .05 ; T1 = dependent variables Time one; T2 = dependent variables Time two

Child personality and Parental Behavior as Moderators of Problem Behavior 133 A Variable- and a Person-centered Approach

A: Benevolence

-1,00

-0,50

0,00

0,50

1,00

1SD- Mean 1SD+

Negative control

Exte

rnal

izin

g

1SD- BEMean BE1SD+ BE

B: Conscientiousness

-1,00

-0,50

0,00

0,50

1,00

1SD- Mean 1SD+

Negative control

Exte

rnal

izin

g

1SD- COMean CO1SD+ CO

Figure 1

Interaction between Personality and Negative control predicting Externalizing behavior at

Time 1. Panel A: Benevolence (BE); Panel B: Conscientiousness (CO)

134 Chapter 3

Benevolence

-1,00

-0,50

0,00

0,50

1,00

1SD- Mean 1SD+

Positive parenting

Exte

rnal

izin

g


Interaction between Benevolence and Positive parenting predicting Externalizing

behavior at Time 1

Figure 2

Child personality and Parental Behavior as Moderators of Problem Behavior 135 in Referred and Non-referred Children

CHILD PERSONALITY AND PARENTAL BEHAVIOR AS

INTERACTING PREDICTORS OF CHILD

IN CLINICALLY REFERRED AND NON-REFERRED

CHILDREN

CHAPTER 4

INTERNALIZING AND EXTERNALIZING BEHAVIOR

Abstract

Interactions between Five Factor Model child personality and two parenting

dimensions were investigated as predictors of child problem behavior. The same

measures were used in both a large community sample (N = 596) and a sample with

clinically referred children (N = 205). Ratings of the child’s personality, internalizing

and externalizing behavior and the mother’s positive parental behavior and negative

control were obtained by the mother. Results showed significant mean-level

differences for personality and parenting. Both personality and parenting predicted

child problem behavior, with some differences in strength of the effects for the two

samples. Parenting by personality interactions predicted both externalizing and

internalizing behavior, with benevolence and conscientiousness as the most prominent

moderators. These effects were largely replicable in both the referred and non-referred

children. The results corroborate the continuity or spectrum model, as differences

between clinical and non-clinical samples are mainly mean level differences and

differences in strength of the relationship, but not structural or qualitative differences.

136 Chapter 4

Introduction

A relatively new trend in research on developmental psychopathology is the

study of the contribution of person by environment interactions as predictors of

problem behavior, beyond what is predictable by the main effects of risk factors.

Child temperament or personality and parenting have been identified as respectively

individual and environmental risk factors associated with maladaptive behavior

(Gallagher, 2002). Our reasons to examine interactions between these variables are

based on theoretical grounds, previous empirical findings, considerations about

clinical practice and the lack of research comparing the effect of moderators in

samples with clinically referred and non-referred children.

First, contextual or ecological models emphasize that a complex network of

interconnected variables is responsible for child adjustment (Belsky, 1984;

Bronfenbrenner, 1986; Dishion, French & Patterson, 1995; Sameroff, 2000). An

important feature of these models is that the predictors do not operate in isolation

when affecting the outcome variable. Instead, the models suggest combined effects of

predictors, fostering the need for exploratory research investigating interactions

between predictor variables (Hinshaw, 2002). This is also suggested by Thomas and

Chess’ goodness-of-fit-theory (1977), postulating that maladaptive child behavior is

the result of a mismatch between a difficult child temperament and parenting practices

(Bates, Pettit, Dodge & Ridge, 1998).

Second, behavior-genetic studies have argued against the crucial role assigned to

parenting in explaining child development (Morris, Silk, Steinberg, Sessa, Avennevoli

& Essex, 2002). Until now, the most repeated conclusion of behavior-genetic studies

is that the shared environment plays a small and rather insignificant role in the

development of children (Collins, Maccoby, Steinberg, Hetherington, & Bornstein,

2000). However, recent studies investigating child characteristics by parenting

interactions show a different picture, suggesting that the way parents treat their

children does indeed have an effect on child outcome, but only for children with

certain individual characteristics (see infra). Research in developmental

psychopathology will benefit from examining the interplay between person and

environment, because these types of findings contribute to the understanding of the

etiology of maladaptive child behavior in general, and risk and protective factors in

particular.


A third argument to consider interactions between parenting and child

individual characteristics stems from psychological practice. Studies investigating

main effects of parenting on child adjustment often lead to the conclusion that parents

are responsible for the behavioral difficulties of their children and hence that the

parents should be subjected to intervention or treatment programs aimed at enhancing

parenting skills in order to reduce child and adolescent maladaptive behavior (Kalb &

Loeber, 2003). It is not our intention to question the effectiveness of parent training

because there is ample evidence that parent training is indeed an effective intervention

method (Kalb & Loeber, 2003), especially for reducing child and adolescent antisocial

behavior (Brestan & Eyberg, 1998). Rather we want to emphasize that findings of

moderator research point toward a more realistic perspective on changing child

problem behavior (Lytton, 1990; Rowe, 1990).

Finally it should be emphasized that previous moderator research focused on

the study of parenting-child interactions either in general population samples or in

samples with clinically referred children. However, because these studies generally

rely on different instruments to study child-parenting interactions, it is difficult to

assess the continuity of the moderator effects across referred and non-referred

samples. The present study compares the effects of the child-parenting interactions in

both types of samples with the same instruments and hence explores the continuity of

moderator effects across a broad spectrum of values for parenting variables, child

personality variables and problem behavior. If the moderator effects do not depend on

the type of sample, the present study supports the case for the continuity of processes

moderating internalizing and externalizing problem behavior in normal (non-referred)

and clinical (referred) populations.

The present study extends previous research through (a) the focus on broad-

band personality dimensions to measure child characteristics, (b) the selection of

parenting variables, (c) the choice of outcome variables, (d) introducing various

methodological refinements intended to increase the power to detect interaction

effects, and (e) assessing the continuity of moderator processes across referred and

non-referred samples of children.

Child personality versus child temperament

A review of the literature indicates that previous research investigating person-

environment interactions mainly focuses on the child’s temperament or various

138 Chapter 4

specific personality characteristics, rather than relying on a comprehensive personality

taxonomy, such as the Five Factor Model (FFM). Temperament refers to early

behavioral tendencies, which are frequently emotional in nature and have a presumed

biological basis (Shiner, 1998), whereas personality refers to individual differences in

the tendency to behave, think, and feel in certain consistent ways (Caspi, 1998, p.

312). Although temperament traits are believed to capture a broad range of individual

differences in infancy, those differences constitute only a subset of personality

differences in later childhood and adulthood (Shiner & Caspi, 2003). Especially

longitudinal research designs may benefit from using a structure of individual

differences applicable to both children and adults. The FFM personality taxonomy

may serve this purpose, as there is growing evidence for similarities between the adult

‘big five’ dimensions and child personality structure. For example, cross-cultural

studies of parents’ free descriptions of their own children’s personality revealed

substantial support for the link between five factor structure in childhood and

adulthood (Kohnstamm, Halverson, Mervielde & Havill, 1998). Moreover, several

authors have shown that the traditional dimensions of temperament are closely related

to the dimensions of the Five Factor Model (Caspi, 1998; Mervielde & Asendorpf,

2000; Shiner, 1998; Shiner & Caspi, 2003). The five broad dimensions of the FFM

are commonly labeled as: Agreeableness (describes a continuum from warmth and

compassion to antagonism), Conscientiousness (the extent and strength of impulse

control), Extraversion (the extent to which the person actively engages the world or

avoids social experiences), Neuroticism (the extent to which someone experiences the

world as distressing or threatening) and Openness to experience (the quality and depth

of a person’s mental and experiential life) (Caspi, 1998).

Several studies provide evidence for relationships between child or adolescent

problem behavior and temperament or personality dimensions. The temperamental

disposition ‘lack of control’ (i.e. emotional lability, restlessness, short attention span

and negativism) has been associated with externalizing behavior, such as

hyperactivity and attention problems, antisocial behavior, and conduct disorder

(Caspi, Henry, McGee, Moffitt & Silva,1995; Eisenberg, Miller, Dutra & Chance,

2000). Agreeableness was linked to externalizing problems in non-clinical and clinical

samples of children and adolescents (John, Caspi, Robins, Moffitt & Stouthamer-

Loeber, 1994; Laursen, Pulkkinen & Adams, 2002). Other studies offer evidence for

the association between personality characteristics and delinquent behavior (Krueger,


Schmutte, Caspi, Moffitt, Campbell & Silva, 1994; Ruchkin, Eisemann & Hägglöf,

1999). Flat affect, passivity and behavioral inhibition were identified as contributors

of internalizing problems (Caspi et al., 1995; Eisenberg et al., 2000) and

agreeableness, extraversion and neuroticism have been related to depression (Finch &

Graziano, 2001). Neuroticism was associated to both anxiety and depression

(Millikan, Wamboldt & Bihun, 2002). Extreme scores on the five domains have also

been associated with several personality disorders (Coker, Costa & Widiger, 2002).

Parenting dimensions versus parenting styles

Research linking parenting and child adjustment consistently refers to two

primary parenting dimensions (Gallagher, 2002; Maccoby & Martin, 1983)

comprising related parental behaviors: Parental support (or responsiveness) describes

the affective nature of the parent-child relationship and includes parental behaviors

like involvement, showing interest in the child; Control (or demandingness) refers to

efforts of parents to influence their child’s behavior, such as setting rules and

enforcing standards of behavior (Maccoby & Martin, 1983). The control concept has

been further differentiated into psychological control, referring to behaviors intruding

into the psychological and emotional development of the child, such as love

withdrawal and guilt induction, and behavioral control, referring to parental behaviors

that attempt to manage children’s behavior (Barber, 1996; Barber, Olsen & Shagle,

1994). Parenting behaviors or dimensions can be distinguished from typologies of

parenting, resulting from the combination of different parenting dimensions. For

example, combining high and low demandingness and responsiveness, Maccoby and

Martin (1983) created four types: authoritarian, authoritative, permissive and

indifferent parenting. Such a type-approach is very useful because it takes into

account the prevalence of particular blends of risk factors (e.g. low demandingness

and low responsiveness). However, most parenting typologies are not yet properly

validated and therefore it is suggested to examine the separate dimensions that

constitute overall parenting style (Galambos, Barker & Almeida, 2003).

Several main effects studies, including correlational and experimental designs

(Hinshaw, 2002) and clinical and non-referred samples (Belsky, Hsieh & Crnic,

1998), confirm that parenting behaviors are good predictors of child maladaptive

behavior. Externalizing behavior, such as aggressive or antisocial behavior and

conduct disorder, has been predicted by lack of parental involvement or poor

140 Chapter 4

acceptance-responsiveness, lack of supervision or poor parental monitoring, harsh and

inconsistent punishment and insufficient rewarding of adequate child behavior

(Forehand, Miller, Dutra & Chance, 1997; Haapasalo & Tremblay, 1994; Loeber &

Dishion, 1983; Loeber & Stouthamer-Loeber, 1986; Patterson & Stouthamer-Loeber,

1984; Rothbaum & Weisz, 1994; Ruchkin, Eisemann & Hägglof, 1999; Stormshak,

Bierman, McMahon, Lengua et al., 2000; Wakschlag & Hans, 1999; Weiss, Dodge,

Bates & Pettit, 1992). A smaller number of studies investigated main effects of

parenting on emotional problems, such as anxiety (Gruner, Muris & Merckelbach,

1999; Siqueland, Kendall & Steinberg, 1996) and depression (Garber, Robinson &

Valentiner, 1997; Muris, Schmidt, Lambrichs & Meesters, 2001; Richter, 1994).

Some of the studies investigating main effects of parenting included both

externalizing and internalizing behavior as outcome variables and confirmed the

hypothesis that psychological control is more predictive of adolescent internalizing

behavior, whereas behavioral control is more predictive of externalizing behavior

(Barber, 1996; Barber, Olsen & Shagle, 1994). However, parental support was not

identified as a significant predictor of externalizing nor of internalizing behavior

(Galambos, Barker & Almeida, 2003).

Most studies concentrate on the negative aspects of parenting, presented as

risk factors in the etiology of maladaptive child behavior. However, it is equally

important to examine parenting skills that stimulate child adjustment or protect the

developing child against emotional and behavioral problems. Therefore, the present

study includes both a more affective component of parenting and a control dimension.

The studies linking parenting and child characteristics in the prediction of

externalizing behavior are mainly restricted to the assessment of the child’s

Internalizing versus externalizing behavior as outcome variables

The current study focuses on both internalizing and externalizing behavior as

outcome variables, although a literature review regarding parenting by child

characteristics interactions shows that externalizing behavior is predominantly studied

as the outcome variable. This probably can be attributed to the fact that in main

effects studies parenting frequently has been linked to aggression, behavior problems

and antisocial behavior (see supra). Alternatively it is possible that parenting by child

characteristics interactions genuinely affect externalizing behavior more than

internalizing behavior.


temperament. These studies vary in (a) design, most studies being cross-sectional,

some being longitudinal (e.g. Bates et al., 1998; Belsky et al., 1998; Rubin, Burgess,

Dwyer & Hastings, 2003; Stoolmiller, 2001) or experimental (e.g. Anderson, Lytton

& Romney, 1986), (b) sample composition, i.e. clinically referred (e.g. Anderson et

al., 1986; Colder, Lochman & Wells, 1997) versus non-referred samples, (c) in child

age, i.e. preschool (e.g. Paterson & Sanson, 1999; Rubin et al., 2003), school-aged

(e.g. Lengua, Wolchik, Sandler & West, 2000; Stoolmiller, 2001; Wootton, Frick,

Shelton & Silverthorn,1997) and adolescent (e.g. Carlo, Roesch & Melby,1998;

Olweus, 1980), and (d) gender, for example some studies only include boys (e.g.

Anderson et al., 1986; Belsky et al., 1998; Colder et al., 1997; Olweus, 1980;

Stoolmiller, 2001). The most important finding of these studies can be summarized as

follows. A ‘difficult to manage’ child or a child characterized by behavioral-

emotional undercontrol, high impulsivity, high activity levels, low emotionality will

show aggressive or antisocial behavior, when faced with punitive, harsh or

inconsistent discipline or poor parental monitoring. This means that children with a

difficult temperament are particularly vulnerable when also exposed to punitive

parenting or negative control. Also interesting is the finding that children with high

levels of callous (i.e. lack of empathy, manipulativeness) and unemotional (i.e. lack of

guilt, emotional constrictedness) traits tend to show high rates of conduct problems,

regardless of the quality of parenting they receive (Wootton et al., 1997).

Few studies have focused on the Five Factor Model as a measure of child

personality in the prediction of problem behavior. O’Connor and Dvorak (2001)

found that specific combinations of personality characteristics and parenting variables

operate as protective or risk factors and hence that parental behaviors only matter for

some kinds of children and not for others. Prinzie et al. (2003) showed that children

rated high on benevolence are protected from the negative effect of maternal or

paternal overreactivity, i.e. the tendency by parents to respond with irritation and/or

anger to problematic behavior of their children. Children rated low on

conscientiousness were at risk for externalizing problems when exposed to coercive

parental behavior.

A few studies have included both internalizing and externalizing behavior as

outcome variables. Blackson, Tarter and Mezzich (1996) concluded that school-aged

boys with a difficult temperament and experiencing negative parental discipline

showed more externalizing and internalizing behavior. Colder et al. (1997) detected

142 Chapter 4

interaction effects of parenting and temperament predicting childhood depressive

symptoms. Children characterized by high levels of fear and exposed to high levels of

parental involvement (‘overinvolvement’) showed more depressive symptoms.

However, for children with moderate levels of fear, low levels of parental

involvement were associated with lower levels of emotional competency. Harsh

discipline was associated with depressive symptoms for children characterized by

high fear. Lengua et al. (2000) showed that for school-aged boys and girls low in

‘emotionality’, parental rejection was more strongly associated with depression after

divorce. Tschann, Kaiser, Chesney, Alkon and Boyce (1996) also found evidence for

the general finding that children with more difficult temperaments display more

internalizing and externalizing behavior than children with easier temperaments,

particularly when also exposed to high levels of family conflict.

A first methodological issue concerns data sampling. The sample studied

needs to be large enough to provide adequate power to test interaction effects. In

moderated multiple regression research, sample size is one of the most important

single factors affecting power (Aguinis, 1995). In order to produce a power of .80 for

detecting small interaction effects, a sample size in the range of 752 and 1056 is

needed, with estimated reliabilities of .80 for the predictors, interpredictor correlations

between 0 and .50 and a squared multiple correlation between .20 and .50 for the main

effects of the predictor variables (Aiken & West, 1991, p.164). In addition, the

detection of person-environment effects is difficult in population samples, because

they require variation in both the person and environmental variables, which is often

only found in specific subgroups (Rutter et al., 1997). Compared to experimental

studies, field studies are often less than 20% efficient in detecting interactions

(Mclelland & Judd, 1993). A possible solution for this problem is to oversample

extreme observations of the predictor variables (Deater-Deckard & Dodge, 1997;

Mclelland & Judd, 1993). In this study, a randomly drawn sample (N = 600) will be

supplemented with a clinically referred sample (N = 205). Both the large number of

Methodological precautions

A widely used method to investigate interactions is moderated regression

analysis. In order to enhance the probability of detecting interactions, some

precautions can be taken into consideration.


cases and the inclusion of more cases at the extremes of the joint distribution of the

predictor variables will increase the power for detecting interaction effects.

Continuity of moderator effects across referred and non-referred samples

Second, it is crucial to have measures with good psychometric properties. In

addition, the present study needed measures that could be administered to a large

sample, therefore the use of questionnaires was indicated. Although studies utilizing

direct behavioral observations of parenting usually report larger effect sizes than

studies relying on data from questionnaires (Collins et al., 2000), there are several

questionnaire studies reporting parenting effects (e.g. Barber, 1996; Galambos et al.,

2003; Garber et al., 1997; Stormshak et al., 2000). To reliably measure parenting, an

instrument with good psychometric properties was developed (Van Leeuwen &

Vermulst, in press) assessing two parenting dimensions - positive parenting and

negative control - based on second-order factors. In order to assess child personality, a

comprehensive and psychometrically well-validated instrument was used, measuring

five broadband personality domains (Mervielde & De Fruyt, 2002). The outcome

variables were assessed with a Dutch version of a standard instrument (Achenbach

Child Behavior Checklist; Verhulst, Van der Ende & Koot, 1996) assessing several

syndromes that are grouped in two broad-band factors: internalizing and externalizing

behavior.

Previous studies on how child personality moderates the effect of parenting on

problem behavior were conducted either with general population samples or clinical

samples. Because the sample choice usually entails a different selection of measures it

remains unclear to what extent the moderator effects can be generalized across

clinical and non-clinical samples. The present study assesses moderator effects with

the same measures (child personality, parenting and problem behavior) in both a large

community sample and a sample with children referred to an ambulant or residential

clinical setting. Such a study includes a broad spectrum of values for child personality

variables, parenting and problem behavior and allows a more sensitive test of

moderator effects that depends on the range of values and in particular on the

frequency of extreme cases (Mclelland & Judd, 1993). Furthermore, including

referred and non-referred children in the same study permits a test of the continuity of

moderator effects across clinical and non-clinical samples. Between and among

clinical psychologists and psychiatrists there is a long-standing debate about whether

144 Chapter 4

the differences between normal and abnormal behavior should be conceived as either

qualitative or quantitative. Recent research on the relationships between (normal,

adaptive) personality dimensions and (abnormal, maladaptive) personality disorders

(Costa & Widiger, 2002; O'Connor, 2002) provides evidence for the continuity or

spectrum model that conceives abnormal or maladaptive personality disorders as

extreme variants of normal personality traits. Hence the differences between clinical

and non-clinical samples are considered to be mainly main level differences and not

structural or qualitative differences. Likewise the present comparison of personality

moderator effects in clinical and non-clinical samples can be considered as an

extension of this type of research because it provides a test of the continuity of the

moderator effects across both types of samples. If the child personality by parenting

interaction does not depend on the sample, the relationships between parenting, child

personality and problem behavior are similar for referred and non-referred children

and presumably based on the same processes in both groups. Hence the differences

between the two types of samples can be conceived as quantitative and a common

model can be postulated to explain how child personality moderates effects of

parenting on problem behavior.

The present research

Method

The present study utilizes a sample consisting of both clinically referred and

non-referred children. First, we will check whether child personality and parenting

differ for clinically referred children and non-referred children, that is, are there

mean-level differences for child personality and parenting. Second, it will be

investigated if there is evidence for parenting by child personality interactions in the

prediction of child internalizing and externalizing behavior. Furthermore we will

check whether these interactions differ for referred and non-referred children. Finally

it will be explored if the effects of child personality and parenting differ for

internalizing and externalizing behavior.

Subjects

The clinically referred sample consists of 205 mothers and children, recruited

from various mental health services. Children in the sample are 118 boys and 87 girls,

5 to 14 years old (M = 9.9, SD = 1.9). The children were referred for a variety of


behavioral and emotional difficulties. Based on the primary intake problem, the

children were assigned to the following broad diagnostic categories: 21.3% showed

symptoms of anxiety/depression, 8.4% presented withdrawn behavior or deficits in

social skills, 6.9% was referred for eating/sleeping/psychosomatic problems, 25.2%

manifested externalizing behavior, such as lying, aggression and temper tantrums,

11.4% exhibited conduct disorders, related to developmental disorders such as

attention deficit hyperactivity disorder, Tourette syndrome or autism, 9.4% reported

attention or concentration problems, 15.8% suffered from adjustment problems due to

major life stressors, such as divorce or decease of a parent, 1% displayed

obsessive/compulsive behavior, and finally, 0.5% was referred for suicide attempt or

self-injurious behavior.

The non-referred sample of children is part from a longitudinal study

investigating parenting, personality characteristics and children’s problem behavior.

Subjects were parents (both mothers and fathers) and one of their children aged 7 to

15. For the present study only ratings of the mothers (N = 596) from the first

assessment period are included. The 276 boys and 320 girls, had a mean age of 10.9

(SD = 1.9; range 7-15). The sample is well balanced regarding social-economic status,

gender and age (Van Leeuwen, 2000).

Measures

A comparison between referred and non-referred children showed significant

differences in gender and age. The non-referred group included 46.3% boys and the

clinical group 57.6% (χ²(1) = 7.73, p = .005). The mean age for the non-referred

children was 10.96 whereas for the referred group this was 6.87, with t(799) = 6.87, p

< .001. Therefore, age and gender will be entered as control variables in further

analyses.

Parental behavior. Two parenting dimensions, i.e. positive parenting and

negative control, were measured with the Ghent Parental Behavior Scale (GPBS, Van

Leeuwen & Vermulst, in press). This questionnaire is based on concepts from Social

Learning Theory (Capaldi & Patterson, 1989; Patterson, Reid & Dishion, 1992),

organizing parenting into well-defined constructs derived from observable parental

behavior. Subjects rate the frequency of behavioral items on a 5-point Likert scale

ranging from ‘never’ to ‘always’. Validation of a pilot version of the GPBS led to the

conclusion that the five hypothesized constructs (positive involvement, monitoring,

146 Chapter 4

problem solving, structure and positive reinforcement) appeared to be too

heterogeneous. Refinement of the constructs resulted in a new questionnaire with nine

scales: Autonomy, Discipline, Positive parental behavior, Harsh punishment,

Monitoring, Rules, Ignoring unwanted behavior, Material rewarding, and Inconsistent

discipline. Factor analyses showed evidence for two higher-order factors: ‘positive

parenting’ (consisting of the scales positive parental behavior, teaching rules and

autonomy), and ‘negative control’ (consisting of the scales discipline, ignoring of

unwanted behavior and harsh punishment) (Van Leeuwen & Vermulst, in press).

Cronbach alpha’s in the present sample equaled .86 for positive parenting and .79 for

negative control.

Child Personality. The Hierarchical Personality Inventory for Children

(HiPIC; Mervielde & De Fruyt, 1999) was used to assess child personality. The

inventory comprises items that describe the normal range of child personality related

behavior. Five broad personality domains are measured, dispersed over 18 facets that

are hierarchically organized under the five domains, i.e. Extraversion (partitioned in

the facets Shyness, Optimism, Expressiveness, Energy), Benevolence (also known as

Agreeableness; further divided into Egocentrism, Irritability, Compliance,

Dominance, Altruism), Conscientiousness (split into Achievement motivation,

Concentration, Perseverance, Orderliness), Emotional Stability (divided into Anxiety

and Self-confidence) and Imagination (also known as Openness to experience;

partitioned into Creativity, Curiosity and Intellect). Mothers rated the 144 items on a

5-point Likert scale. The factor structure of the HiPIC proves to be highly replicable

across both childhood and adolescence (Mervielde & De Fruyt, 2002). In the present

study we found Cronbach alpha’s of .95 for Benevolence (N of items = 40), .94 for

Conscientiousness (N of items = 32), .89 for Emotional Stability (N of items = 16),

.91 for Extraversion (N of items = 32) and .91 for Imagination (N of items = 24).

Child problem behavior. The Dutch version of the Achenbach Child Behavior

Checklist (CBCL; Verhulst, Van der Ende & Koot, 1996) assesses behavioral and

emotional problems in children. Mothers rated the frequency of 113 problematic

behaviors on a 3-point Likert scale. Two broadband syndromes can be derived from

CBCL scores: Internalizing, with items referring to somatic complaints, social

withdrawal, and anxiety/depression; and Externalizing, including items indexing


aggression, hyperactivity, and delinquency. For Internalizing behavior (N of items =

31) the value of Cronbach’s alpha was .89, for Externalizing behavior it was .92 (N of

items = 33).

Procedure

Children of the referred sample were recruited from various mental health

services. Third year students were instructed to enlist clinically referred children with

emotional or behavioral problems as part of an assignment for the advanced course on

Personality Psychology at the Ghent University. Informed consent was given by the

therapist or counselor and the parents of the children. Mothers completed the CBCL,

the HiPIC and the GPBS. Exclusion criteria were the presence of a physical disability

or a condition of chronic disease.

Non-referred children were recruited via randomly selected elementary and

secondary schools. For elementary schools the sample was stratified by province,

(East and West Flanders), region (rural or urban), school type (public/private/catholic


secondary schools, sampling was based on province (East and West Flanders), type of

curriculum (vocational, technical and general education) and grade (first and second

year of secondary school). A letter addressed to the parents informed them about the

goal and the procedures of the research project. Families were visited at home by a

trained psychology student who instructed the family members to independently

complete a series of questionnaires. Both parents filled out the HiPIC, GPBS and

CBCL. In the present study, we only focus on the ratings provided by the mothers, to

enable a joint analysis with the data supplied by the mothers of the clinical or referred

group of children.

Statistical analyses

Hierarchical multiple regression analyses (HMRA) were conducted to test the

main hypotheses. Several ‘functional sets’, i.e. groups of independent variables

grouped for logical reasons are included in different steps of the HMRA (Cohen,

Classic ANOVA’s were carried out in order to check mean-level differences

for group (clinically referred or non-referred children) and gender, and possible group

by gender interactions. These analyses were primarily carried out to verify if it would

be necessary to include interactions between gender and the other variables.

148 Chapter 4

Cohen, Aiken & West, 2003). In these analyses gender and age were consistently

entered in Step 1 as control measures. In Step 2, we entered ‘group’ as a variable, one

of the five child personality domains (i.e. Benevolence, Conscientiousness, Emotional

Stability, Extraversion or Imagination), and one of the two parenting dimensions (i.e.

Negative control or Positive parenting). This step enables us to investigate

independent or main effects of group, personality and parenting variables. In Step 3,

we entered the interactions between the three predictor variables, i.e. personality by

parenting, personality by group and parenting by group. The first interaction tests

whether the effects of parenting depend on certain values of child personality (or vice

versa) in the prediction of problem behavior. The personality by group and parenting

by group interactions indicate to what extent the relationship between personality or

parenting and problem behavior is different for the clinical and the non-referred

group. In step 4, a 3-way interaction term was entered, i.e. personality by parenting by

group. A significant three-way interaction term indicates that the parenting by

personality interaction is different for the clinical and the non-referred groups. That is,

it verifies whether the parenting by child personality interaction depends on or can be

generalized across the referred and the non-referred samples.

For testing and interpretation of interactions, we followed the guidelines by

Aiken and West (1991) and Cohen et al. (2003). To form an interaction term between

the continuous variables, the two centered predictors were multiplied. Centering the

predictor variables (i.e. putting predictors in a deviation score form, so that their

means are zero) is recommended to eliminate multicollinearity between predictor

variables. We used z-scores in the analyses, because with multiplicative terms, neither

traditional unstandardized nor standardized regression coefficients are appropriate to

report. However, when the crossproduct is based on z-scores (commonly known as

Friedrich’s procedure) it is appropriate to use the unstandardized solution with

interaction terms (Aiken & West, 1991, p. 43-44). Interaction terms between the

categorical variable ‘group’ and the continuous parenting and child personality

variables were formed by multiplying the dummy-code variable group (non-referred

group = 0 and clinical group = 1) with the centered continuous variable.

We further applied the step-down procedure proposed by Aiken and West

(1991, p. 105) and dropped non-significant interactions when strong theoretical

grounds for expecting an interaction were lacking. Evidence for a significant

interaction is found when there is a significant increase in the multiple R² after


entering the interaction term, as indicated by the incremental F test. When the three-

way interaction turned out to be not significant, it was dropped from the final model.

Results

Effects of group and gender

Table 1 shows means, standard deviations and results of GLM analyses of

variance. Significant main effects of group were present for each variable. Referred

children had significantly higher mean scores on both internalizing and externalizing

problem behavior compared to non-referred children. Referred children scored

significantly lower on the personality domains benevolence, conscientiousness,

emotional stability, extraversion and imagination than the children in the general

population sample. As regards the parenting variables, the mothers of the clinically

referred children reported significantly more negative control and less positive

parental behavior than the mothers of the non-referred children.

Significant gender effects were found for externalizing behavior, with higher

average scores for boys. Boys scored significantly lower than girls on the personality

domains benevolence and conscientiousness, whereas girls scored significantly lower

than boys on emotional stability. The effect sizes ranged from .01 to .17 for the effect

of group, and from .01 to .03 for the effect of gender.

Two significant interaction effects of group by gender were present. For

internalizing behavior children from the referred group had higher means than the

non-referred children, but this difference was greater for girls than for boys. A group

by gender effect was also found for the personality domain extraversion, with referred

children scoring significantly lower than non-referred children, but with a greater

difference for girls than for boys.

Personality as a predictor of child problem behavior

Main effects of personality. Table 2 reports, as part of the HMRA involving

either maternal negative control (top panel) or positive parenting (bottom panel), the

main effects on internalizing and externalizing problem behavior for each of the five

broad personality domains measured. The results across the two panels are entirely

consistent. Benevolence is significantly related to externalizing and to a lesser extent

to internalizing behavior, confirming previous research (John et al., 1994; Laursen et

al., 2002), indicating a negative relationship between this personality factor and

150 Chapter 4

problem behavior. A similar pattern emerges for conscientiousness. Children scoring

high on this personality factor tend to have far fewer externalizing problems and less

internalizing problems. Taking into account the prominent role of emotional stability

in the development of adult psychopathology, it is important to notice that also for

children and adolescents this personality factor significantly predicts internalizing

behavior and to a lesser extent externalizing behavior. Extraversion, as expected, is

significantly related to internalizing behavior but not to externalizing problem

behavior. Finally, smaller but significant effects of imagination on both internalizing

and externalizing behavior are reported in Table 2. From these analyses a clear and

consistent pattern for the relationship between personality and problem behavior

emerges. Benevolence and conscientiousness are negatively related to externalizing

behavior and to a lesser extent to internalizing behavior. Emotional stability and

extraversion are negatively related to internalizing behavior and emotional stability is

to some extent related to externalizing behavior. Finally, children scoring high on

imagination tend to present less internalizing and externalizing problems although this

effect is small compared to the effects of the four other broad-band personality

factors.

Consistency of personality effects across referred and non-referred groups.

The HMR analyses reported in Table 2 show several significant group by personality

interactions, indicating differences in the strength of the relationship between

personality and problem behavior for the clinical and non-referred group. Four of the

ten group by personality interactions reported in this table are significant in the top

panel (entering maternal control in the second step of the HMRA) as well as in the

bottom panel (reporting HMRA controlling for positive parenting in the second step).

In addition to these consistent group by personality interactions there is one small

(p<.05) group by benevolence interaction for internalizing when controlling for

positive parenting. Only one of the four consistent (across both panels) group by

personality interactions predicts externalizing behavior, showing that the relationship

between benevolence and externalizing depends on the group. The slopes in Figure 1

indicate that the relationship between benevolence and externalizing behavior is

stronger in the referred group, although the slopes are significant in both the non-

referred (t = -14.55, p < .001) and in the referred group of children (t = -15.3, p <

.001).


Group by personality interactions were more prominent for predicting

internalizing behavior. Figure 2 shows that, as expected, internalizing behavior is

more frequent in the referred than in the non-referred group but that in addition to this

main effect, the personality-internalizing relationship is stronger in the referred group

for emotional stability (panel A) and extraversion (panel B). Both for emotional

stability and extraversion the slopes of the non-referred (ES: t = -13.56, p < .001; EX:

t = -7.13, p < .001) and referred group (ES: t = -13.88, p < .001; EX: t = -7.33, p <

.001) are significant. Somewhat surprising is the result reported in panel C, showing a

negative relationship between conscientiousness and internalizing for the non-referred

group (t = -4.75, p < .001) but not for the referred group (t = 0.75, p = .455). Given

that this is the first study testing the consistency of personality-internalizing behavior

relationships across referred and non-referred samples, it remains to be seen whether

this effect can be replicated. In general the relationship between personality and

problem behavior tends to be somewhat stronger in the referred than in the non-

referred group and in particular for internalizing behavior. At first sight it may seem

reasonable to find that the relationship between personality and problem behavior is

stronger in the referred or clinical group because referred children as a group show

higher levels of problem behavior. However this main effect does not imply that the

relationship between problem behavior and its predictors (i.e. personality) should also

be stronger in the referred group, because the strength of this relationship is not

affected by mean level differences between the referred and the non-referred group of

children.

Although in less than half the cases the relationship between personality and

problem behavior is different in the referred and the non-referred sample, it is

debatable to what extent this evidence challenges the spectrum hypothesis. Strictly

speaking the spectrum hypothesis postulates a continuum from normal to maladaptive

or abnormal personality and hence only mean level differences between clinical and

non-clinical samples are to be expected. This implies that the relationship between

adaptive (normal) personality and problem behavior (e.g. internalizing or

externalizing) is the same in clinical and non-clinical groups. Three of the four

documented interactions indicate a personality-problem behavior relationship that is

similar in form but differs in strength. Moreover for three out of the four cases the

slopes of the personality-problem behavior regression are significant in both referred

152 Chapter 4

and non-referred groups and hence similar relationships emerge that tend to be

somewhat stronger in the referred than in the non-referred group.

Parenting as a predictor of child problem behavior

Main effects of parenting. Negative control as a dimension of parenting is

strongly related to the degree of externalizing behavior and to a lesser extent to

internalizing. As is evident from the top panel of Table 2, negative control

consistently predicts higher levels of both types of problem behavior, irrespective of

the effects of the personality variables that are entered in the second step of the

HMRA.

The bottom panel of Table 2 shows that positive parenting significantly decreases

both internalizing and externalizing problems, except when benevolence is entered

together with positive parenting in the second step of the regression. Overall the

effects of parenting on problem behavior tend to be smaller than the effects of

personality variables reported in the previous section. Nevertheless the present results

confirm the substantial detrimental impact of negative control but in addition show

the beneficial effect of positive parenting on the degree of internalizing and

externalizing problems presented by children and adolescents.

Consistency of parenting effects across referred and non-referred groups. The

top panel of Table 2 illustrates that the strength of the relationship between negative

control and externalizing behavior depends on the group. As show in Figure 3 (panel

A), negative control is more detrimental for externalizing behavior in the referred

group (t = 7.99, p < .001) than in the non-referred group (t = 6.16, p < .001) although

both slopes are significant. That is restrictive parenting generally predicts

externalizing behavior of all children and adolescents but negative control is

especially harmful for referred children.

The bottom part of Table 2 reports significant interactions between the group

and positive parenting for both internalizing and externalizing problems. Figure 3,

panel B illustrates that positive parenting has no effect on externalizing for the

referred group (t = -0.01, p = .990) but is a protective factor for non-referred children

(t = -3.59, p < .000). Moreover a similar pattern emerges from Figure 4, illustrating

that positive parenting has a positive effect on internalizing problems in the non-

referred group (t = -2.92, p < .01) whereas it has virtually no effect in the referred

group (t = 1.27, p = .206). In general it can be concluded that the relationship between


control or restrictive parenting and externalizing behavior is stronger in the referred

group than in the non-referred group and hence that this parenting style predicts

problem behavior for both referred and non-referred children and adolescents.

Positive parenting on the other hand predicts reduced problem behavior only for the

non-referred children and it decreases both internalizing and externalizing behavior.

However referred children do not seem to benefit from positive parenting.

Personality variables as moderators

Table 2 reports eight significant personality by parenting interactions. Four of

the five personality variables moderate the effects of parenting on either internalizing

or externalizing behavior. The parenting by personality interactions indicate that

benevolence, conscientiousness, emotional stability and imagination can be identified

as moderators of the effect of parenting on problem behavior.

In order to interpret the interaction effects, the signs of the regression

coefficients of the two independent predictor variables and the interaction variable

were inspected (Cohen et al., 2003, p. 285-286). This revealed two interaction

patterns: (a) buffering interactions, in which the two predictors have regression

coefficients of the opposite sign, meaning that one predictor weakens the effect of the

other, (b) interference or antagonistic interactions in which both the regression

coefficients have the same sign and the interaction coefficient is of the opposite sign,

indicating an either-or effect of the predictor variables on the criterion. The buffering

interactions were found for the interactions with parental negative control, the

interference/antagonistic interactions were found for the interactions with positive

parenting. Significant interactions were interpreted by plotting simple regression lines

for high (1 SD above the mean), mean and low (1 SD below the mean) values of the

moderator variables. The significance of the slopes for these simple regression lines

slopes was tested with t-tests.

Negative control showed a significant interaction with benevolence and

conscientiousness in predicting both externalizing and internalizing behavior. Figure 5

shows two examples of these interactions. Panel A represents the benevolence by

negative control interaction predicting externalizing behavior with significant slopes

for low (t = 6.17, p < .001) and mean (t = 5.08, p < .001) scores on benevolence and

non- significant slope for high (t = 0.98, p = .328) scores on benevolence. The

conscientiousness by negative control interaction predicting externalizing behavior

154 Chapter 4

has significant slopes for low (t = 6.25, p < .001), mean (t = 6.16, p < .001) and high (t

= 2.92, p < .01) scores on conscientiousness. Panel B graphs the conscientiousness by

negative control interaction predicting internalizing behavior, with significant slopes

for low (t = 3.21, p < .001) and mean (t = 2.76, p < .01) scores on conscientiousness

but not for high (t = 0.85, p = .396) scores on conscientiousness. The interaction effect

of negative control by benevolence predicting internalizing behavior is of the same

kind, with significant slopes for low (t = 4.40, p < .001) and mean (t = 3.28, p < .001)

scores on benevolence but not for high (t = 0.01, p = .988) scores on benevolence.

Positive parenting showed a significant interaction with benevolence,

conscientiousness and imagination in the prediction of externalizing behavior, and a

significant interaction with emotional stability in the prediction of internalizing

behavior. Figure 6 graphs two examples of personality moderating the effects of

positive parenting. Panel A shows the significant imagination by positive parenting

interaction, predicting externalizing behavior, with significant slopes for children

rated low (t = -3.89, p < .001) or around the mean (t = -3.59, p < .001) on imagination,

but not for children rated high (t = -1.63, p = .103) on imagination. Panel B shows the

emotional stability by positive parenting interaction, predicting internalizing behavior,

with significant slopes for low (t = -3.54, p < .001) and mean (t = -3.73, p < .001)

emotional stability, but not for high scores on this domain (t = -1.88, p = .060). In the

interaction with positive parenting predicting externalizing behavior, only the slope

for low (t = -2.79, p < .01) scores on benevolence is significant, not the slopes for

mean (t = -1.33, p = .183) and high (t = 1.05, p = .292) scores. For the positive

parenting interactions with conscientiousness, the slopes for low (t = -3.00, p < .01)

and mean (t = -2.10, p < .05) scores are significant, not for high (t = -0.19, p = .847)

scores.

From the results it can be concluded that parenting by personality interactions

are more frequently predicting externalizing (five) than internalizing behavior (three).

Benevolence is a prominent moderator variable, although it should be noted that only

children rated low or around the mean are at risk for problem behavior, when also

exposed to inadequate parenting. Conscientiousness is also an important moderator of

the effects of both dimensions of parenting on externalizing behavior. Children with

high ratings on benevolence, conscientiousness and imagination are protected against

inadequate forms of parenting.


Differences in parenting by personality interactions for referred and non-referred

children

Significant 3-way interactions indicate different parenting by personality

interactions for clinically referred and non-referred samples of children. Although we

have no reasons to expect different patterns for the referred and non-referred samples,

three significant 3-way interactions on a total of twenty tested were found. First, the

imagination by negative control interaction predicting externalizing behavior is not

significant for the non-referred group, but only affects externalizing behavior in the

referred group of children. Referred children scoring low on imagination are more at

risk for externalizing behavior when they are also exposed to high levels of negative

maternal control (t = 7.27, p < .001), whereas referred children scoring high on

imagination are protected against externalizing behavior, even when their mothers

show high levels of negative control (t = 2.81, p < .01). A significant

conscientiousness by negative control interaction predicting internalizing behavior is

only found in the non-referred group, with significant slopes for low (t = 4.40, p <

.001) and mean (t = 3.28, p < .001) levels, and not for high levels of conscientiousness

(t = 0.01, p = .988). A significant emotional stability by positive parenting interaction

predicting internalizing behavior was only found in the non-referred sample of

children. Non-referred children scoring low (t = -4.66, p < .001) or around the mean (t

= -4.55, p < .001) on emotional stability are at risk for internalizing behavior when

they are deprived from positive parenting. The slope for non-referred children scoring

high on emotional stability is not significant (t = -1.02, p = .309).

The finding that only three out of twenty possible three-way interactions are

significant, suggests that in general, there are few differences between referred and

non-referred groups of children in the way personality and parenting interact.

Moreover the three cases show inconsistent interaction patterns, one showing a

personality by parenting interaction for the referred group and the two others a

significant interaction for the non-referred group.

General discussion

The present study examined child personality by parenting interactions in both

a general population sample and a sample with clinically referred children.

Administering the same instruments in both samples enabled us to explore to what

156 Chapter 4

extent (a) the relationships between personality and problem behavior depend on the

group, (b) the parenting-problem behavior relationships differ for referred and non-

referred children, (c) personality moderates effects of parenting on externalizing and

internalizing behavior, and (d) the personality-parenting interaction can be

generalized across both samples. The results of these analyses are discussed in light of

the continuity or spectrum model which postulates that differences between normal

and abnormal or clinical samples are restricted to quantitative or mean level

differences on the relevant domains of functioning (Costa & Widiger, 2002; Widiger

& Clark, 2000).

Personality as a predictor of child problem behavior

In the present study, personality was measured with an instrument assessing

the five factor domains and comprising items that describe the normal range of

personality characteristics and behaviors (HiPIC; Mervielde & De Fruyt, 1999, 2001).

Morey (1997) has suggested that personality traits describing the normal variation in

personality can represent both clinical and normal subjects, whereas measures of

abnormal traits can only adequately describe clinical subjects. In contrast, evidence

was found for structural invariance of measures of psychopathology and normal

personality in referred (O’Connor, 2002) and non-referred (De Clercq & De Fruyt,

2003; O’Connor, 2002) groups of respondents, pointing towards a common

dimensionality and structure for clinical and non-clinical samples. However, the

equivalence of the dimensional structure of measures addresses only one important

facet in the comparative study of normal and abnormal functioning. It is equally

important to statistically compare the effects of predictors (independent variables) on

problem behavior (dependent variables) across referred and non-referred samples. The

present study confirms large mean level differences between the referred and non-

referred group on problem behavior and personality, indicating that the referred

children obtain higher scores on problem behavior and lower scores on the Big Five

personality dimensions as measured by the HiPIC.

Besides mean level differences between the referred and the non-referred

sample, the present study demonstrated that personality is a strong predictor of child

problem behavior, with differential effects of the five domains for internalizing and

externalizing behavior. In general, personality predicted problem behavior in both the

referred and non-referred groups of children but the relationship between personality


and problem behavior tended to be somewhat stronger in the referred than in the non-

referred group, particularly for internalizing behavior. Hence, it can be concluded that

the five-factor personality model is a useful model to study antecedents of problem

behavior in both referred and non-referred samples, showing continuity of effects

across non-referred and referred children.

Although strictly speaking the relationship between personality and problem

behavior differs, the difference pertains to the strength of the relationship and not to

the kind of relationship. If one postulates a weaker version of the spectrum hypothesis

allowing for mean differences and differences in strength but not for qualitative

differences in the relationship between personality and problem behavior, then this

weaker form of the spectrum hypothesis is strongly supported. It is clear that the

majority of personality-problem behavior relationships do not depend on the group

and that for three out of the four significant interaction effects the relationship is

similar in form but slightly stronger in the referred group. Hence the difference

between referred and non-referred groups is quantitative rather then qualitative. If one

wants to build a model for personality-problem behavior relationships for referred and

non-referred groups, the model would include the same variables, the path coefficients

would indicate the same direction but some of the path coefficients would turn out to

be slightly higher for referred than for the non-referred groups. Although strictly

speaking this is a deviation from the spectrum hypothesis, it is a minor deviation that

is a far cry from the often claimed qualitative differences between referred and non-

referred groups.

Parenting as a predictor of child problem behavior

Mean level analyses showed higher maternal negative control and lower

positive parental behavior in the referred group of children. Negative control was

identified as a strong predictor of externalizing behavior, confirming the results of

previous studies (Barber, 1996; Barber et al., 1994; Deater-Deckard & Dodge, 1997;

Stormshak et al., 2000; Weiss et al., 1992) and documenting the negative relationship

of restrictive parenting or negative control with problem behavior. The problem

behavior enhancing effect of negative control, was found in both the referred and non-

referred group of children, but the effect was somewhat stronger in the referred group.

A different result emerged, however, for the positive parenting dimension. A lack of

positive parenting predicted both externalizing and internalizing behavior, but only for

158 Chapter 4

children in the non-referred group. This is in contrast with Galambos et al. (2003),

who did not found evidence for parental support as a significant predictor of

externalizing nor of internalizing behavior.

In the discussion of the relationship between personality and problem behavior

it was emphasized that the form of relationships was similar for the referred and non-

referred groups but the associations tended to differ in strength. This pattern is

confirmed for the negative control parenting-problem behavior relationship but not for

the relationship between positive parenting and problem behavior. Positive parenting

predicts less problem behavior in the non-referred group but not in the referred group.

Because both types of parenting have a different relationship with problem behavior,

these analyses confirm post hoc the discriminant validity of the parenting dimensions.

Moreover the lack of continuity of effects of positive parenting across referred and

non-referred groups is a challenge for the spectrum hypothesis. A model for parenting

effects on non-referred children should include significant paths from negative control

and positive parenting to problem behavior. A model for effects of parenting in

referred or clinical children and adolescents could do without a path from positive

parenting to problem behavior.

Personality as a moderator in the relationship between parenting and problem

behavior

The results of this study support the differential susceptibility hypothesis of

Belsky (1997), which assumes that children scoring differently on specific personality

domains also vary in their sensitivity to parental influences. In particular, children

rated high on conscientiousness and benevolence are protected against the detrimental

effects of high maternal negative control and low positive behavior. Being rated low

or around the mean on benevolence or conscientiousness is consistently associated

with more problem behavior, when also exposed to inadequate parenting. These

findings corroborate the results of Prinzie et al. (2003) and O’Connor and Dvorak

(2001).

In a strict sense, the design of the present study does not allow to draw

conclusions about recurrent reciprocal interchanges over time between child and parent,

i.e. as specified by the transactional effects model explaining child development (Lytton,

1990). Our measures do not assess child behavior at time 1, eliciting a certain parental

response, which influences the child at time 2, and so on. These transactional interaction


sequence is suggested in Patterson’s coercion theory (Patterson, 1982; Patterson, Reid &

Dishion, 1992), which assumes that a child learns by negative reinforcement that his or

her aversive behavior is an effective strategy for controlling the behavior of others and

to escape from undesirable situations. This theory tends to focus on the parental response

to the child (Stoolmiller, 2001), and in particular on the fact that some parents are not

able to manage their child’s behavior. By including child individual characteristics in the

present study and investigating the statistical interaction between parenting and child

personality, we have shown that the effects of parental behavior on child problem

behavior depend on certain child characteristics or vice versa. Low benevolence and low

conscientiousness in particular, are related to child unmanageability, a concept which is

highly relevant to coercion theory (Stoolmiller, 2001). Therefore it can be hypothesized

that children rated low on benevolence and conscientiousness evoke increasing levels of

negative control from their mothers, which enhance in return the risk for developing

problem behavior.

Continuity of moderator processes across referred and non-referred samples of

children

The validity of the continuity or spectrum model was also tested by assessing

whether the interaction effects of child personality and parenting were sample

dependent. In line with the spectrum hypothesis we did not expect differences in

moderator effects across referred and non-referred samples of children. The results

showed some minor (three out of twenty) three-way interactions, indicating different

interactions for the referred and non-referred children. The significance of these third-

order interactions was due to the existence of a significant personality by parenting

interaction in one group, and a non-significant interaction in the other group. However

the three significant interactions explained only a minor part of the variance and the

pattern was not consistent across groups or personality dimensions. One analysis

showed a significant imagination by negative control interaction for the referred group

but not for the non-referred group whereas the conscientiousness by negative control

interaction was significant in the non-referred group but not in the referred group.

Hence before rejecting the spectrum hypothesis regarding personality by parenting

interactions, a replication of these findings in an independent sample is needed to

establish the robustness of these small and inconsistent effects. In sum we can

conclude that the personality by parenting effects are largely replicable in both the

160 Chapter 4

referred and non-referred group of children and thus corroborate the spectrum

hypothesis assuming that similar processes regulate problem behavior in both groups.

Internalizing versus externalizing behavior

The present study reveals some differential relationships between personality

and parenting on the one hand and internalizing and externalizing behavior on the

other hand. It was shown that benevolence and conscientiousness are negatively

related to externalizing problem behavior, whereas internalizing behavior was mainly

predicted by low emotional stability and low extraversion. These results are

compatible with the pattern emerging from studies adopting a person-centered

approach (De Fruyt, Mervielde & Van Leeuwen, 2002; Van Leeuwen, De Fruyt &

Mervielde, in press). These studies demonstrate that resilient children (scoring above

average on all personality domains) display the least amount of internalizing and

externalizing behavior. Undercontrolled children (scoring low on

benevolence/agreeableness and conscientiousness) tend to show externalizing

problems whereas overcontrollers (scoring low an emotional stability and

extraversion) mainly show internalizing problem behavior. In addition to these classic

personality-problem behavior relationships, the present study also provides evidence

for a significant but weaker negative relationship between imagination and both

internalizing and externalizing.

The relationships between the two parenting dimensions and child problem

behavior also reveal a differential pattern. Negative maternal control is a significant

predictor of externalizing behavior, with high levels of negative control predicting

high levels of externalizing behavior. Negative control also predicts internalizing

behavior, but the regression coefficients are smaller, indicating smaller effects.

Positive parental behavior on the other hand predicts both internalizing and

externalizing behavior but to a lesser extent than maternal negative control. Overall,

negative maternal control is the better predictor of externalizing behavior whereas

positive parenting is a weaker predictor of both internalizing and externalizing

behavior.

Limitations of this Study and Future Research

Because this study is the first to assess the consistency of moderator effects of

parenting and personality across a referred and a non-referred sample, replication of


this study in independent samples is needed in order to examine the robustness of the

effects. Moreover, the inclusion of longitudinal data would provide the opportunity to

compare effects over time, and hence permit stronger conclusions regarding the

direction of the effects and the causal status of the antecedents of child

psychopathology.

The results are based on measures that are completed by the same informant

(the mother) and as such this poses the problem of confounded method-variance.

Relying on a single reporter can exaggerate the true relations between parenting and

child adjustment (Galambos et al., 2003), and therefore it is worthwhile to obtain

ratings from more than one informant. Future research should address this issue by

collecting data from fathers in referred samples and or by aggregating scores over

parents.

The finding that stable child personality characteristics predict child problem

behavior should not be associated with the pessimistic or deterministic view that

problematic behavior is stable and difficult to change, but rather as evidence for a

more realistic perspective (Lytton, 1990). The present study supports the viewpoint

that parents often receive too much blame for the behavioral difficulties of their

children (Anderson, Lytton & Romney, 1986) or take too much credit for the lack of

problem behavior because they ignore the fact that effects of parenting interact with

child characteristics. For example Frick, Cornell, Barry, Bodin and Dane (2003)

suggest that children scoring high on callous-unemotional traits are less susceptible to

parental training than children low on these traits. Likewise, therapists and counselors

who steer clear of blaming parents for their children’s behavioral outcomes, may

uncover that parents become more eager partners for implementing effective

interventions (Rowe, 1990). On the other hand, environmental changes may reduce

the likelihood that predispositions become more manifest (Raine & Dunkin, 1990).

Given a mismatch between child characteristics and parental behavior, parents can be

instructed to use more effective parenting strategies to prevent the development of

problem behavior (Colder et al., 1997). Taking account of the child’s personality can

guide the development of more effective custom-made treatment programs. Center

and Kemp (2003, p. 83) note that: “Differentiating between parents of children with

typical and difficult personality profiles could possibly enhance the effectiveness of

Implications for Clinical Practice

162 Chapter 4

the approach. Parents of children with a difficult child profile probably require both

education about their child’s predispositions as well as more extensive training in

child management techniques.” Screening of child personality and parenting may

indeed be necessary to point out the ‘poorness of fit’ in parent-child interactions

(Ostergren, 1997) and to develop tailor-made and hence more realistic and effective

training programs (Kalb & Loeber, 2003; Sheeber & Johnson, 1994; Teerikangas,

Aronen, Martin & Huttonen ,1998).

Finally, the fact that the present study generally supports the spectrum

hypothesis, contending continuity for the personality-parenting-problem behavior

relationships across referred and non-referred samples, obviates the need to invest

resources in the development of sample specific models for understanding and

treatment of children’s internalizing or externalizing behavior.


References

Aguinis, H. (1995). Statistical power problems with moderated multiple

regression in management research. Journal of Management, 21, 1141-1158.

Aiken, L. S., & West, S. G. (1991). Multiple regression: testing and

interpreting interactions. Newbury Park: Sage Publications.






Barber, B. K., Olsen, J. E. & Shagle, S. C. (1994). Associations between

parental psychological and behavioral control and youth internalized and externalized

behaviors. Child Development, 65, 1120-1136.



externalizing behavior. Developmental Psychology, 34, 982-995.



Belsky, J. (1997). Variation in susceptibility to environmental influence: an

evolutionary argument. Psychological Inquiry, 8, 230-235.





Blackson, T. C., Tarter, R. E., & Mezzich, A. C. (1996). Interaction between

childhood temperament and parental discipline practices on behavioral adjustment in

preadolescent sons of substance abuse and normal fathers. American Journal of Drug

and Alcohol Abuse, 22, 335-348.

Brestan, E. V., Eyberg, S. M. (1998). Effective psychosocial treatments of

conduct-disordered children and adolescents: 29 years, 82 studies, and 5.272 kids.

Journal of Clinical Child Psychology, 27, 180-189.



164 Chapter 4

Capaldi, D., & Patterson, G. R. (1989). Psychometric properties of fourteen

latent constructs from the Oregon Youth Study. New York: Springer-Verlag.







Wiley.




Center, D., & Kemp, D. (2003). Temperament and personality as potential

factors in the development and treatment of conduct disorder. Education and

treatment of children, 26, 75-88.

Cohen, J., Cohen, P., West, S. G., & Aiken, L. S. (2003). Applied multiple

regression/ correlation analysis for the behavioral sciences (3th ed.). Mahwah, New


Coker, L. A., Samuel, D. B., & Widiger, T. A. (2002). Maladaptive personality

functioning within the big five and the five-factor model. Journal of Personality

Disorders, 16, 385-401.

Colder, C. R., Lochman, J. E., & Wells, K. C. (1997). The moderating effects

of children’s fear and activity level on relations between parenting practices and

childhood symptomatology. Journal of Abnormal Child Psychology, 25, 251-263.

Collins, W. A., Maccoby, E. E., Steinberg, L., Hetherington, E. M. &






Deater-Deckard, K., & Dodge, K. A. (1997). Externalizing behavior problems

and discipline revisited: nonlinear effects and variation by culture, context, and

gender. Psychology Inquiry, 8, 161-175.


De Clercq, B., & De Fruyt, F. (2003). Personality disorder symptoms in

adolescence: A Five-Factor Model perspective. Journal of Personality Disorders, 17,

269-292.

De Fruyt, F., Mervielde, I., & Van Leeuwen, K. (2002). The consistency of




ecology of antisocial behavior. In D. Cicchetti & D.J. Cohen (red.), Developmental


& Sons, Inc.









Frick, P. J., Cornell, A. H., Barry, C. T., Bodin, S. D., & Dane, H. E. (2003).

Callous-unemotional traits and conduct problems in the prediction of conduct problem

severity, aggression, and self-report of delinquency. Journal of Abnormal Child


Galambos, N. L., Barker, E. T., & Almeida, D. M. (2003). Parents do matter:

trajectories of change in externalizing and internalizing problems in early

adolescence. Child Development, 74, 578-594.



Garber, J., Robinson, N. S., & Valentiner, D. (1997). The relation between

parenting and adolescent depression : self-worth as a mediator. Journal of Adolescent

Research, 12, 12-33.

Gruner, K., Muris, P., & Merckelbach, H. (1999). The relationship between

anxious rearing behaviours and anxiety disorders symptomatology in normal children.

Journal of Behavior Therapy and Experimental Psychiatry, 30, 27-35.

166 Chapter 4

Haapasalo, J., & Tremblay, R. E. (1994). Physically aggressive boys from ages

6 to 12: family background, parenting behavior, and prediction of delinquency.


Hinshaw, S. P. (2002). Process, mechanism, and explanation related to

externalizing behavior in developmental psychopathology. Journal of Abnormal Child


John, O. P., Caspi, A., Robins, R. W., Moffitt, T. E., & Stouthamer-Loeber, M.

(1994). The “Little Five”: exploring the nomological network of the five-factor model

of personality in adolescent boys. Child Development, 65, 160-178.

Kalb, L. M., & Loeber, R. (2003). Child disobedience and non-compliance: a

review. Pediatrics, 111, 641-652.







Laursen, B., Pulkkinen, L., & Adams, R. (2002). The antecedents and

correlates of agreeableness in adulthood. Developmental Psychology, 38, 591-603.

Lengua, L. J., Wolchik, S. A., Sandler, I. N., & West, S. G. (2000). The

additive and interactive effects of parenting and temperament in predicting problems

of children of divorce. Journal of Clinical Child Psychology, 29, 232-244.






of Chicago Press.




family: parent-child interaction. In Mussen, P. H. (Ed.), Handbook of child


New York: Wiley.


McClelland, G. H., & Judd, C. M. (1993). Statistical difficulties of detecting

interactions and moderator effects. Psychological Bulletin, 114, 376-390.





Inventory for Children (HiPIC). In I. Mervielde, I. Deary, F. De Fruyt, & F. Ostendorf

(Eds.), Personality psychology in Europe. (pp. 107-127). Tilburg University Press.




Millikan, E., Wamboldt, M. Z., & Bihun, J. T. (2002). Perceptions of the family,

personality characteristics, and adolescent internalizing symptoms. Journal of the

American Academy of Child and Adolescent Psychiatry, 41, 1486-1494.

Morey, L. C. (1997). Personality diagnosis and personality disorders. In R.

Hogan, J. Johnson & S. Briggs (Eds.), Handbook of Personality Psychology (pp. 919-

946). San Diego: Academic Press.

Morris, A. S., Silk, J. S., Steinberg, L., Sessa, F. M., Avennevoli, S., & Essex,

M. J. (2002). Temperamental vulnerability and negative parenting as interacting

predictors of child adjustment. Journal of Marriage and Family, 64, 461-471.

Muris, P., Schmidt, H., Lambrichs, R., & Meesters, C. (2001). Protective and

vulnerability factors of depression in normal adolescents. Behaviour Research and

Therapy, 39, 555-565.

O’Connor, B. (2002). The search for dimensional structure differences

between normality and abnormality: A statistical review of published data on

personality and psychopathology. Journal of Personality and Social Psychology, 83,

962-982.

O’Connor, B. P., & Dvorak, T. (2001). Conditional associations between

parental behavior and adolescent problems: a search for personality-environment

interactions. Journal of Research in Personality, 35, 1-26.

Olweus, D. (1980). Familial and temperamental determinants of aggressive

behavior in adolescent boys: a causal analysis. Developmental Psychology, 16, 644-660.

Ostergren, C. S. (1997). Differential utility of temperament-based guidance

materials for parents of infants. Family Relations, 46, 63-71.

168 Chapter 4






Patterson, G. R. (1982). Coercive family process. Eugene, OR: Castalia Press.






Raine, A., & Dunkin, J. J. (1990). The genetic and psychophysiological basis

of antisocial behavior : implications for counseling and therapy. Journal of

Counseling and Development, 68, 637-644.

Richter, J. (1994). Parental rearing and aspects of psychopathology with

special reference to depression. In C. Perris, W.A. Arrindell a M. Eisemann (Eds.),

Parenting and psychopathology (pp. 235-251). Chisester: John Wiley & Sons.

Rothbaum, F., & Weisz, J. R. (1994). Parental caregiving and child externalizing



on personality development. Journal of Counseling and Development, 68, 606-611.







Rutter, M., Dunn, J., Plomin, R., Simonoff, E., Pickles, A., Maughan, B.,

Ormel, J., Meyer, J., & Eaves, L. (1997). Integrating nature and nurture: implications

of person-environment correlations and interactions for developmental

psychopathology. Development and Psychopathology, 9, 335-364.




Sheeber, L. B., & Johnson, J. H. (1994). Evaluation of a temperament-focused,

parent-training program. Journal of Clinical Child Psychology, 23, 249-259.


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




Siqueland, L., Kendall, P. C., & Steinberg, L. (1996). Anxiety in children:

perceived family environments and observed family interaction. Journal of Clinical

Child Psychology, 25, 225-237.



behavior in boys. Developmental Psychology, 37, 814-825.




29, 17-29.

Teerikangas, O. M., Aronen, E. T., Martin, R. P., & Huttonen, M. O. (1998).

Effects of infant temperament and early intervention on the psychiatric symptoms of

adolescents. Journal of the American Academy of Child and Adolescent Psychiatry,

37, 1070-1076.

Tshann, J. M., Kaiser, P., Chesney, M. A., Alkon, A., & Boyce, T. W. (1996).

Resilience and vulnerability among preschool children: family functioning,

temperament, and behavior problems. Journal of the American Academy of Child and

Adolescent Psychiatry, 35, 184-192.

Van Leeuwen, K. (2000). Deficits in parenting skills as an indicator of

behavior problems with children and youth. Development of a screening instrument

for the Flemish community. Ghent University, Department of Psychology.

Van Leeuwen, K., De Fruyt, F., & Mervielde, I. (in press). A longitudinal

study of the utility of the resilient, overcontrolled and undercontrolled personality

types as predictors of children’s and adolescents’ problem behavior. International

Journal of Behavioral Development.

170 Chapter 4



Assessment.










Widiger, T. A., & Clark, L. A. (2000). Toward DSM-V and the classification

of psychopathology. Psychological Bulletin, 126, 946-963.

Wootton, J. M., Frick, P. J., Shelton, K. K., & Silverthorn, P. (1997).

Ineffective parenting and childhood conduct problems: the moderating role of callous-

unemotional traits. Journal of Consulting and Clinical Psychology, 65, 301-308.

Table 1

Category means and results of GLM with group and gender Non-referred Referred Group Gender Group X Gender Bo ys Girls Boys Girls M (SD) N M (SD) N M (SD) N M (SD) N F η² F η² F η² Externalizing 9.01

(7.65) 275 6.43 (6.05) 320 16.99

(11.24) 118 12.51 (10.67) 87 112.68 a 0.12 28.50 a 0.03 2.06 0.00

Internalizing

87

118

0.00

319

0.15

8.09(6.84) 275 7.03

(5.90) 320 12.87 (8.62) 118 16.23

(9.51) 87 143.99 a 0.15 3.88 c 0.00 14.40 a 0.02

Benevolence -0.75(2.71) 268 0.14

(2.42) 314 -3.55 (3.00) 118 -2.60

(3.18) 87 156.33 a 0.17 17.39 a 0.02 0.02 0.00

Conscientiousness 12.45(2.63) 268 13.37

(2.65) 314 10.77 (2.61) 118 12.13

(2.71) 87 45.35 a 0.05 27.65 a 0.03 1.04 0.00

Emotional stability 0.58 (1.35) 268 0.40

(1.28) 314 -0.63 (1.30) 118 -1.14

(1.49) 160.24 a 0.17 9.94 b 0.01 2.22 0.00

Extraversion 8.15(2.20) 268 8.33

(2.13) 314 7.23 (2.19)

6.59 (2.66) 87 53.49 a 0.06 1.58 0.00 5.01 c 0.01

Imagination 11.10(1.68) 268 11.00

(1.80) 2.02

314 10.19 (1.86) 118 10.60

(1.91) 87 20.21 a 0.03 1.21 3.01 0.00

Negative control 2.10 (0.52) 274 (0.53)

2.20 (0.42) 118 2.12

(0.47) 87 6.02 c 0.01 3.40 0.00 0.01 0.00

Positive parenting 4.26 (0.40) 276 4.26

(0.38) 320 4.03 (0.48) 118 4.01

(0.45) 87 50.98 a 0.06 0.26 0.00 0.00

172 Chapter 4

Table 2

Moderators of Problem Behavior

Child Externalizing behavior Child Internalizing behavior ∆F B ∆F B Panel 1:Maternal Negative control Sex, Age 22.96 a -.14 b, .00 2.57 .09, .01 Group, BE, CON 271.21 a .13 c, -.52 a, .14 a 57.07 a .79 a, -.22 a, .09 c BE x CON, GR x BE, GR x CON 19.70 a -.11 a, -.20 a, .11 3.21 c -.08 c, .15, -.11 BE x CON x GR 0.01 1.60 Sex, Age 22.96 a -.20 a, -.02 2.57 .07, .01 Group, CO, CON 113.95 a .53 a, -.28 a, .20 a 53.20 a .90 a, -.19 a, .11 b CO x CON, GR x CO, GR x CON 13.30 a -.08 b, -.03, .34 a 4.58CO x CON x GR 1.24 6.67

a, .00 2.57

b -.11 b, .23 b, .04 b .20 b

Sex, Age 22.96 a -.33 -.08, .03 Group, ES, CON 77.72 a .58 a, -.10 c, .23 a 185.95 a .23 b, -.48 a, .10 b ES x CON, GR x ES, GR x CON 12.45

0.09 a .00, -.04, .47 a 8.89 b -.04, -.30 a, .10

ES x CON x GR 0.33 Sex, Age 22.96 a -.30 a, .00 2.57 .01, .00 Group, EX, CON 90.84

2.57

77.42 a .75 a, .05, .23 a a .63 a, -.28 a, .14 a EX x CON, GR x EX, GR x CON 11.03 a -.01, .06, .42 a 2.10 -.02, -.17 c, .01 EX x CON x GR 3.79 0.26 Sex, Age 22.96 a -.31 a, -.01 .03, .00 Group, IM, CON 55.18

1.42 6.93

78.68 a .64 a, -.09 c, .23 a a .81 a, -.14 a, .10 b IM x CON, GR x IM, GR x CON 11.87 a -.02, .07, .39 a IM x CON x GR b -.19 b 1.98 Panel 2: Maternal Positive parenting Sex, Age 22.82 a -.16 b, .00 2.74 .07, .01 Group, BE, POS 246.01 a .13, -.58

2.74

a, -.04 55.67 a .79 a, -.24 a, -.09 BE x POS, GR x BE, GR x POS 11.32 a .08 b, -.23 a, .11 2.52 .04, .16 c, .16 c BE x POS x GR 0.30 0.00 Sex, Age 22.82 a -.23 a, -.03 .06, .00 Group, CO, POS 80.25 a .56 a, -.32 a, -.08 c

4.90 0.22

50.66 a .89 a, -.20 a, -.09 c CO x POS, GR x CO, GR x POS 3.38 c .07 c, -.11, .09 b .06, .25 b, .14 CO x POS x GR 0.03 Sex, Age 22.82 a -.37 a, -.01 2.74 -.10, .02 Group, ES, POS 41.99 a .70 a, -.11 b, -.15 a

181.08 a .25 b, -.51 a, -.15 a ES x POS, GR x ES, GR x POS 1.54 9.26 a .11 b, -.27 a, .16 c ES x POS x GR 0.00 4.82 c -.14 c Sex, Age 22.82 a -.35 a, -.01 2.74 .00, -.01 Group, EX, POS 42.77 a .85 a, .07, -.17 a 84.25 a .66 a, -.28 a, -.12 b EX x POS, GR x EX, GR x POS 2.31 .03, .11, .16 c 4.56 b .06, -.15 c, .19 b EX x POS x GR 0.84 0.05 Sex, Age 22.82 a -.34 a, -.02 2.74 .02, -.01 Group, IM, POS 43.89 a .71 a, -.10 b, -.15 a 52.04 a .85 a, -.18 a, -.11 b IM x POS, GR x IM, GR x POS 2.70 c .07 c, -.03, .15 c 3.34 c .06, .14, .16 c IM x POS x GR 0.90 1.44 Note. BE = Benevolence; CO = Conscientiousness; ES = Emotional Stability, EX = Extraversion; IM =

Imagination; POS = Positive parenting; CON = Negative control; a p ≤ .001 ; b p ≤ .01 ; c p ≤ .05

Via a step-down procedure non-significant interactions were dropped, and therefore only steps with

significant coefficients are reported.

Child personality and Parental Behavior as Moderators of Problem Behavior in Referred and Non-referred Children

173

-1

-0,5

0

0,5

1

SD 1- mean SD 1+

Benevolence

Exte

rnal

izin

g

referrednon-referred

Figure 1

Interaction between Benevolence and Group predicting Externalizing behavior

Note: Figures 1 and 2 can be considered as representative figures for all significant personality by

group interactions, with different parenting variables included in the HMRA model. The graphing of

the interactions in Figure 1 and Figure 2, Panel B are based on HMRA models with Negative control

included; the graphing of the interactions of Figure 2, Panel A and Panel C are based on HMRA

models with Positive parenting included. Figures 1 and 4 can be considered as representative figures

for all significant parenting by group interactions, with different personality variables included in the

HMRA model. In figure 3, Panel A is based on a HMRA with Conscientiousness included, and Panel B

is based on a HMRA with Imagination included. Figure 4 is based on a HMRA model with

extraversion included.

174 Chapter 4

Panel A: Emotional Stability by Group

-1,00

-0,50

0,00

0,50

1,00

SD 1- mean SD 1+

Emotional stability

Inte

rnal

izin

g


Panel B: Extraversion by Group

-1,00

-0,50

0,00

0,50

1,00

SD 1- mean SD 1+

Extraversion

Inte

rnal

izin

g


Panel C: Conscientiousness by Group

-1,00

-0,50

0,00

0,50

1,00

SD 1- mean SD 1+

Conscientiousness

Inte

rnal

izin

g


Figure 2

Interactions between Personality and Group predicting Internalizing behavior

Panel A: Extraversion; Panel B: Emotional Stability; Panel C: Conscientiousness


175

A: Negative control by Group

0,00

0,50

1,00

1,50

SD 1- mean SD 1+

Negative control

Exte

rnal

izin

g


B: Positive parenting by Group

0,00

0,50

1,00

1,50

SD 1- mean SD 1+

Positive parenting

Exte

rnal

izin

g


Figure 3

Interactions between Parenting and Group predicting Externalizing behavior

Panel A: Negative control

Panel B: Positive parenting

176 Chapter 4

-1,00

-0,50

0,00

0,50

1,00

SD 1- mean SD 1+

Positive parenting

Inte

rnal

izin

g


Figure 4

Interactions between Positive parenting and Group predicting Internalizing behavior


177

A: Benevolence (BE) x Negative control predicting Externalizing behavior

-1,00

-0,50

0,00

0,50

1,00

1SD- Mean 1SD+

Negative control

Exte

rnal

izin

g


B: Conscientiousness (CO) x Negative control predicting Internalizing behavior

-1,00

-0,50

0,00

0,50

1SD- Mean 1SD+

Negative control

Inte

rnal

izin

g

1SD- COMean CO1SD+ CO

Figure 5

Negative control by personality interactions

Panel A: with Benevolence predicting Externalizing behavior

Panel B: with Conscientiousness predicting Internalizing behavior

178 Chapter 4

A: Imagination (IM) x Positive parenting predicting Externalizing behavior

-1,00

-0,50

0,00

0,50

1,00

1SD- Mean 1SD+

Positive parenting

Inte

rnal

izin

g

1SD- ESMean ES1SD+ ES

B: Emotional Stability (ES) x Positive parenting predicting Internalizing behavior

0,00

0,50

1,00

1SD- Mean 1SD+

Positive parenting

Exte

rnal

izin

g

1SD- IMMean IM1SD+ IM

Figure 6

Positive parenting by personality interactions

Panel A: with Imagination predicting Externalizing behavior

Panel B: with Emotional stability predicting Internalizing behavior

Parent Personality, Child Personality, Parenting and Problem Behavior 179

CHAPTER 5

PARENT PERSONALITY, CHILD PERSONALITY AND

PARENTING AS PREDICTORS OF CHILD

INTERNALIZING AND EXTERNALIZING BEHAVIOR

Abstract

The present study investigates the concurrent effects of parenting, child personality

and parent personality on child internalizing and externalizing behavior. Two

alternative models are tested by means of path analyses. The first model conceives

child and parent personality as antecedents of parenting, which in turn influences

child internalizing or externalizing behavior. In the second model the relationship is

reversed, postulating a link from parenting and parent personality to child personality,

which in turn, affects problem behavior. None of the parent personality domains

directly influenced child externalizing behavior. However, neuroticism in both

parents, and openness to experience in mothers were associated with higher levels of

child internalizing behavior. Parent personality moderately affected parental behavior.

The results showed indirect effects of negative control on child problem behavior. The

results showed evidence for both models and bi-directional influences of parenting

and personality.

180 Chapter 5

Introduction

Current research trying to explain child (mal)adaptive behavior assumes that

neither the family environment, nor the child, nor the broader social context can be

considered as the single significant determinant of child outcome. This idea is well

represented by contextual or ecological models, emphasizing that a complex network

of interconnected variables is responsible for child adjustment (Belsky, 1984;

Bronfenbrenner, 1986; Dishion, French & Patterson, 1995; Lerner, Rothbaum, Boulos

& Castellino, 1997; Sameroff, 2000). Although parent personality is usually

recognized as one of the nodes in the network determining child (mal)adaptive

outcome, studies assessing the combined effects of parent personality, child

personality and parenting are rare (Belsky & Barends, 2002). In a heuristic model

graphing the determinants of parenting and child developmental outcomes, Belsky

(1984) suggests that parents’ personality indirectly influences their children’s

development, through their parental behavior. In other words, parenting reflects

parents’ personality, and hence parenting intervenes as a mediator in the relationship

between parent personality and child behavior. According to Belsky, parent

personality can be considered as the most important determinant of parenting, next to

contextual sources of stress and support, and the child’s individual characteristics. The

model also posits a direct influence of child individual characteristics on child

developmental outcomes, but it does not include a direct influence of parent

personality on child personality or child outcome behavior. However, from empirical

evidence examining the relationship between parent personality and child personality

and between parent personality and child outcome behavior, it can be hypothesized

that parent personality also could have a unique effect on children’s personality and

outcome behavior. It can also be questioned whether the postulated relationship

between child individual characteristics and parenting can be reversed, with parenting

influencing child personality, instead of child personality affecting parenting.

The goal of the present study is to investigate the relationships between parent

personality, child personality, parenting and child problem behavior in two alternative

models that take into account the previous remarks. The hypothesized paths are based

on empirical findings, some of which are briefly reviewed below. Although the design

of the current study does not permit to disentangle genetic and environmental effects,

we will consider conclusions from behavioral genetic studies when discussing the


relationships between parent personality, child personality, parenting and child

problem behavior. In the past, social scientists and behavioral geneticists have offered

distinct views on the relative contribution of genes, the shared and the non-shared

environment. According to the behavioral geneticists, social scientists overestimate

the effects of the environment by regarding environmental influence as independent of

genetic effects (Rowe, 1997). Social scientists, on the other hand, criticize the method

of analysis used in behavioral genetic studies, pointing out that they overestimate the

effects of genes (Collins, Maccoby, Steinberg, Hetherington & Bornstein, 2000).

Child individual characteristics and parenting as predictors of child problem

behavior

Several studies have documented the impact of individual child characteristics

and parenting on child problem behavior. Most of these studies searched for

independent or main effects of these variables, others looked at additive or cumulative

effects and a third type of study investigated parenting or child characteristics as an

intervening variable, i.e. as a moderator or a mediator (Putnam, Sanson & Rothbart,

2002).

Studies investigating child main effects assume that a child’s individuality is

responsible for the child’s outcome. Externalizing problems have been linked to the

temperamental disposition ‘lack of control’ and to the personality domains

agreeableness and conscientiousness (Caspi, Henry, McGee, Moffitt & Silva, 1995;

Eisenberg, Miller, Dutra & Chance, 2000; Krueger, Schmutte, Caspi, Moffitt,

Campbell & Silva, 1994). Internalizing problems were associated with flat affect,

passivity and behavioral inhibition (Caspi et al., 1995; Eisenberg et al., 2000) and

with the personality domains agreeableness, extraversion and neuroticism (Finch &

Graziano, 2001; Millikan, Wamboldt & Bihun, 2002). Other studies have shown a

link between externalizing behavior and the undercontrolled personality type, i.e.

children scoring low on Agreeableness and Conscientiousness, and between

internalizing and the overcontrolled type, i.e. children scoring high on Neuroticism

and low on Extraversion (Asendorpf, Borkenau & Ostendorf, 2001; De Fruyt,

Mervielde & Van Leeuwen, 2002).

Main effects of parenting have been investigated in correlational as well as in

experimental studies (Hinshaw, 2002) and in clinical and non-referred samples

(Belsky, Hsieh & Crnic, 1998). In summary, externalizing behavior is predicted by

182 Chapter 5

lack of parental involvement or poor acceptance-responsiveness, lack of supervision

or poor parental monitoring, harsh and inconsistent punishment and insufficient

rewarding of desirable child behavior (Forehand, Miller, Dutra & Chance, 1997;

Haapasalo & Tremblay, 1994; Loeber & Dishion, 1983; Loeber & Stouthamer-

Loeber, 1986; Patterson & Stouthamer-Loeber, 1984; Rothbaum & Weisz, 1994;

Ruchkin, Eisemann & Hägglof, 1999; Wakschlag & Hans, 1999; Weiss, Dodge, Bates

& Pettit, 1992). It also has been shown that psychological control is more predictive

of adolescent internalizing behavior, whereas behavioral control is more predictive of

externalizing behavior (Barber, 1996; Barber, Olsen & Shagle, 1994).

Although results of main effect and cumulative risk studies underscore the

influence of both parenting and child characteristics on maladaptive outcomes, these

studies have been criticized, because main effects may reveal spurious correlations of

parenting or child characteristics with problem behavior. McCrae and Costa (1994)

doubt that parenting has a major influence on psychopathology, but yet they suggest

that personality and parenting may have interactive effects on psychopathology. The

general conclusion is that children with a difficult temperament are more at risk for

externalizing problems, when also exposed to negative, punitive parenting (e.g.

Anderson, Lytton, & Romney, 1986; Bates, Pettit, Dodge & Ridge, 1998; Belsky,

Hsieh & Crnic, 1998; Carlo, Roesch & Melby, 1998; Rubin, Burgess, Dwyer &

Hastings, 2003; Paterson & Sanson, 1999).

Recently some studies examined personality-environment interactions,

assessing personality from the perspective of the Five-Factor Model (FFM), including

the personality domains Neuroticism, Extraversion, Openness to experience,

Agreeableness and Conscientiousness. The FFM taxonomy has become

acknowledged as a robust reference-model of personality (De Raad & Perugini,

2002), representing personality not only in adults but also in children and adolescents,

and hence facilitating comparisons across developmental periods (Kohnstamm,

Halverson, Mervielde, & Havill, 1998; Shiner & Caspi, 2003). O’Connor and Dvorak

(2001) showed that children with particular personality characteristics are more at risk

than others for problem behavior when exposed to certain parenting practices. Prinzie,

Onghena, Hellinckx, Grietens, Ghesquière and Colpin (2003) found that children

rated high on benevolence were protected against parental overreactivity, i.e. the

tendency by parents to respond with irritation and/or anger to problematic behavior of

their children. Children low on conscientiousness were more at risk for externalizing


behavior when exposed to coercive parenting. A study by Van Leeuwen, Mervielde,

Braet, and Bosmans (submitted) included the interaction effects of parental ‘support’

and ‘negative control’ as well as the child’s five-factor personality profile as

determinants of internalizing and externalizing behavior. The results indicated that

children rated low or around the mean on benevolence and conscientiousness were at

risk for externalizing behavior in particular when they were exposed to parental

negative control or deprived from positive parenting. With internalizing behavior as

the outcome variable, only the negative parental control by benevolence interaction

was significant.

Parent personality as a predictor of parenting and child problem behavior

Although most research considers child personality and parenting as the major

determinants of child problem behavior, it is often ignored that parenting itself may be

determined by stable parental personality traits (Spinath & O’Connor, 2003). In spite

of that, there are few studies addressing the relationship between the normal variation

in personality and parenting. Belsky and Barends (2002) have summarized the results

of studies regarding the effect of parental Big Five personality on parenting.

Neuroticism, or facets of neuroticism, such as anxiety and irritability/hostility are

related to less competent parenting. More extraverted persons, who are sociable,

active, talkative, person oriented and affectionate, tend to be engaged in more

affective and more supportive parenting. Parents, who enjoy new experiences, have

broad interests and are imaginative, i.e. scoring high on Openness to experience, are

likely to engage in more positive parenting. Parents scoring high on Agreeableness,

being softhearted, good-natured, trusting, helpful, etc., show more positive affect and

sensitivity, and less negative control. Conscientious parents, characterized by

orderliness, responsibility and dependability, tend to show more supportive parenting

and less negative control. A review of research addressing the correlation between the

normal variation in personality and parenting shows a clear but modest link (Spinath

& O’Connor, 2003).

Behavioral genetic studies suggest genetic influences on parenting. Measures

of parenting, reflecting the behavior of individuals, may indirectly assess heritable

personality traits. Rowe (1997) found evidence for genetic variation in measures of

parental affection/warmth, but not for parental control. Spinath and O’Connor (2003)

tested competing theories about the origins of individual differences in parenting,

184 Chapter 5

using parenting data of an adult twin sample. They showed (a) that there was

moderate genetic influence on the parenting variables over-protectiveness,

authoritarianism, supportive/indulgent parenting, but not on parental rejection; (b) that

shared environmental influences on parental rejection support a cultural/familial

transmission of harsh and rejecting parenting; (c) that in general, the overlap between

personality and parenting is modest, and that the overlap, is mainly between negative

parenting behaviors and personality characteristics.

Regarding the link between parental characteristics and child problem

behavior, one research line has adopted a categorical approach by focusing on the

effect of parental psychiatric disorders, such as schizophrenia, depression, anxiety,

eating disorders, personality disorders, alcoholism and substance use, on child-rearing

behavior and child developmental outcomes (Zahn-Waxler, Duggal, & Gruber, 2002).

However, adopting a dimensional approach for assessing personality traits could

prove to be more sensitive to a wider range of functioning, because trait measures are

not confined to assessment of extreme forms of functioning (Nigg & Hinshaw, 1998).

Evidence for the relation of parent personality and parenting with child

internalizing problem behavior was found in a study of Brook, Brook and Whiteman

(2003). They showed that maternal personality has both direct and indirect effects on

a toddler’s insecure and dependent behavior. Child rearing had a direct effect on the

behavior of the toddler, but also served as a mediator in the relationship between

maternal personality and toddler behavior. Clark, Kochanska and Ready (2000)

observed the child’s negative emotionality and related it to the parents’ style of

control and responsiveness, and parental personality as measured with the NEO-FFI.

Parent personality, alone and in interaction with child emotionality, had significant

effects on parenting, although a substantial amount of the variance remained

unexplained. Parents scoring high on neuroticism and extraversion displayed a more

controlling parenting style, and parents scoring high on conscientiousness were more

responsive to their children. An interaction was found between low empathy/

extraversion of the mother and high child negative emotionality, predicting more

power-assertive parenting. Nigg and Hinshaw (1998) studied parent’s FFM

personality in relation to their children’s antisocial behavior. Higher rates of child

overt antisocial behavior were primarily related to maternal FFM measures, such as

higher neuroticism and lower conscientiousness. Boys with ADHD and comorbid


ODD or CD tend to have fathers characterized by lower levels of agreeableness and

higher levels of neuroticism.

The relationship between parent personality and child personality

Given that parental and child personality are correlated, one wonders to what

extent this relationship is based on genetic effects. Behavioral genetic studies have

documented that personality is partly heritable, and that, although environmental

influence is important, almost all the environmental variance is nonshared. Research

adopting the FFM shows heritabilities ranging from 30% to 50%, with extraversion

emerging as one of the most heritable traits, followed by neuroticism or emotional

stability (Ebstein, Benjamin & Belmaker, 2003; Loehlin, 1992; McCrae et al., 2000;

Rowe 1997).

Child personality as a predictor of parenting or vice versa?

The macromodel of Dishion, French and Patterson (1995) explaining child and

adolescent antisocial behavior is a typical example of a socialization model

postulating effects of parenting practices on child temperament. However, according

to behavior genetic studies, there is not much empirical evidence underscoring the

widespread and entrenched cultural belief that child rearing is responsible for a major

part of the individual differences in children’s personality traits (McCrae & Costa,

1994; Rowe, 1997). On the other hand, it should be acknowledged that extreme

environments, in particular negative control, elicit changes. However changes in the

developmental course of children seems to be probable when environmental

circumstances are outside the normal range (Scarr, 1992). Experimental research

designs, based on random assignment to treatment and control group, can reveal

effects of parenting interventions on children’s functioning (Vandell, 2000).

Stoolmiller (2001) notes that critics of the parenting-practices perspective have never

dealt adequately with the strong evidence provided by randomized trials, indicating

the causal role of parenting in antisocial behavior and delinquency. For example

parent training proved to be effective as an intervention strategy aimed at reducing

child and adolescent antisocial behavior (Brestan & Eyberg, 1998), by enhancing

specific skills in parents and/or children to behave in particular situations. However, it

remains unclear whether these interventions primarily change the child’s maladaptive

behavior or also have an effect on the rather stable child personality characteristics.

186 Chapter 5

McCrae and Costa (1994) state that child-rearing practices have little effect on basic

tendencies in personality, such as perceptual and cognitive abilities, physical

characteristics and personality traits, but that they may have a major effect on the

individual’s characteristic adaptations such as the child’s self-concept.

Another method to study the effects of parenting on normal adult personality traits

makes use of memory based retrospective ratings of parenting. Reti, Samuels, Eaton,

Bienvenu, Costa and Nestadt (2002) found significant but modest correlations.

Subjects rating themselves as high on neuroticism, low on conscientiousness, low on

self-directedness and high on harm avoidance, experienced lower parental care, higher

parental behavior restrictedness and denial of psychological autonomy in childhood.

The effect of child personality or child temperament on parenting is a

cornerstone of Thomas and Chess’ goodness-of-fit-theory (1977). A child with a

difficult temperament elicits particular parental behavior, and a mismatch between

temperament and parenting practices may result in maladaptive child behavior.

Studies examining interactions between child characteristics and parenting as

predictors of child problem behavior also are based on the premise that children with

a particular temperament or personality evoke (mal)adaptive parental behavior, and

presume that a particular combination of child characteristics and parental behaviors

is associated with child problem behavior.

The present study

Recently, Prinzie, Onghena, Hellinckx, Grietens, Ghesquiere and Colpin (in

press) investigated the concurrent effects of Five-Factor child and parent personality,

and negative parenting on children’s externalizing behavior. They found that the

influence of parent and child personality characteristics on child externalizing

behavior is not exclusively mediated by negative parenting practices, such as laxness,

coerciveness, and overreactivity. The child personality domain benevolence

(agreeableness in the FFM) negatively affected the parenting variables, whereas

emotional stability positively affected overreactivity. The parent personality domains

agreeableness, emotional stability and autonomy were related to some of the parenting

variables. They also provided evidence for the direct influence of parent personality

domains on child externalizing behavior, because neuroticism showed a positive

effect and agreeableness a negative effect on externalizing behavior.


The current study is partly based on the findings of this study, and investigates

two different path models. Model 1 (see Figure 1) includes direct paths from each

parent personality domain (Agreeableness, Conscientiousness, Neuroticism,

Extraversion and Openness to experience) and each child personality domain

(Benevolence, Conscientiousness, Emotional Stability, Extraversion and Imagination)

to either Internalizing or Externalizing behavior. In addition, this model includes

indirect paths from each of the parent and child personality domains to child problem

behavior, passing through the parenting variables negative control and positive

parenting. Hence, this model includes both the direct effects of child and parent

personality on child externalizing and internalizing behavior as well as the indirect

effects of child and parent personality through parenting. This model is tested

separately for mother and father self-ratings of personality and parenting, and also for

child externalizing and internalizing behavior.

Model 2 (see Figure 1) is an alternative model including direct paths from the

parent personality variables and parenting variables to child problem behavior. The

model also postulates effects of parent personality and parenting on child personality,

providing estimates for the indirect effects of child personality on child problem

behavior. Again, this model will be tested separately for mother and father self-ratings

of personality and parenting, and separately for child externalizing and internalizing

behavior.

Method

Participants

The current study utilizes data from a general population sample (N = 600).

Due to list wise deletion of missing observations, the sample was reduced to 499

families. In these families data was available from at least one parent. Target children

were 225 boys and 274 girls with a mean age of 10.95 (SD = 1.87; range 7-15). Of the

families 9.1% was not the original family and 5.2% of the families included a single

parent. The effective size of the mother sample was 413, with 97% biological

mothers. The mean age of the mothers was 38.69 (SD = 4.39; range 20-59). Of the

mothers 74.6% was employed. The effective size of the father sample was 409, with

93.6% biological fathers. The mean age of the fathers was 40.73 (SD = 4.76; range

29-63) and 94.6% of the fathers was employed. Both mothers and fathers had

representative levels of education.

188 Chapter 5

The sample for the cross-validation of the models consisted of 175 families in

which two siblings were rated, both in the same age range as the initial data set. For

163 families and 310 children there was complete data for at least one of the parents.

The 310 target children included 137 boys and 173 girls with a mean age of 10.59 (SD

= 1.80; range 7-14). The majority of the families (8.6%) was original and 0.6% was a

single parent family. The effective number of children rated by their mothers after list

wise deletion of missing values was 247, whereas the effective number of children

rated by their fathers after list wise deletion of missing values was 286. The mean age

of the mothers was 39.2 (SD = 3.9; range 29-53), 99.4% were biological mothers and

81% of the mothers were employed. The mean age of the fathers was 41.2 (SD = 4.6;

range 26-60), 95.7% were biological fathers and 98.8% was employed.

A chi-square test, χ²(1) = 0.91, p > .05, showed no significant difference in the

number of boys and girls across the two samples.

Measures

Parental behavior

The Ghent Parental Behavior Scale (GPBS, Van Leeuwen & Vermulst, in

press) assesses parenting behavior derived from Social Learning Theory (Capaldi &

Patterson, 1989; Patterson, Reid & Dishion, 1992). The central idea of this theory is

that child problem behavior is the result of a series of interactional processes,

characterized by the contingent use of aversive behavior, combined with ineffective

parental management techniques. The Social learning theory focuses on observable

parental behavior and specific family management practices such as monitoring, limit

setting, positive reinforcement, problem solving and parental involvement.

Subjects rate the frequency of behavioral items on a 5-point Likert scale. Nine

scales can be distinguished: Autonomy, Discipline, Positive parental behavior, Harsh

punishment, Monitoring, Rules, Ignoring unwanted behavior, Material rewarding, and

Inconsistent discipline. Evidence for the factorial validity and for a moderate to good

internal consistency is provided by Van Leeuwen and Vermulst (in press). To limit

the number of variables to be included in the path analyses, two second-order factors

were computed and labeled as: ‘positive parenting’ (consisting of the scales positive

parental behavior, teaching rules and autonomy), and ‘negative control’ (consisting of

the scales discipline, ignoring of unwanted behavior and harsh punishment). These


dimensions resemble two parenting dimensions regularly mentioned in the literature

(Gallagher, 2002), the first referring to the affective nature of the parent-child

relationship (‘support’ or ‘responsiveness’), and the second describing the efforts of

parents to influence their child’s behavior, such as setting and enforcing standards of

behavior (‘control’ or ‘demandingness’) (Maccoby & Martin, 1983).

Cronbach alpha’s for positive parenting showed values of .86 and .91 for

mother and father ratings respectively, and .80 for both mother and father ratings of

negative control.

Parent personality

The Dutch version of Costa and McCrae’s NEO PI-R (Hoekstra, Ormel & De

Fruyt, 1996) was used to assess FFM domain scores for the parents: Neuroticism,

Extraversion, Openness to experience, Agreeableness and Conscientiousness. Domain

scores are obtained by aggregating six domain facets, and each facet is measured with

eight items. The facets Anxiety, Angry Hostility, Depression, Self-consciousness,

Impulsiveness and Vulnerability form the Neuroticism domain. Extraversion includes

the facets Warmth, Gregariousness, Assertiveness, Activity, Excitement-Seeking and

Positive Emotions. Openness to experience comprises the facets Fantasy, Aesthetics,

Feelings, Actions, Ideas and Values. Agreeableness is based on the facets Trust,

Straightforwardness, Altruism, Compliance, Modesty and Tender-Mindedness.

Conscientiousness comprises the facets Competence, Order, Dutifulness,

Achievement Striving, Self-Discipline and Deliberation. The Dutch NEO PI-R has

satisfactory psychometric characteristics (Hoekstra, Ormel & De Fruyt, 1996).

Internal consistency in the present study was sufficient, with Cronbach alpha values of

.92 and .92 for Neuroticism, .87 and .88 for Extraversion, .87 and .88 for Openness to

experience, .87 and .87 for Agreeableness, and .88 and .90 for Conscientiousness for

mother and father ratings respectively.

Child personality

The child’s personality was assessed with the Hierarchical Personality

Inventory for Children (HiPIC; Mervielde & De Fruyt, 1999). There is empirical

evidence for similarities between the adult ‘Big Five’ dimensions and child

personality structure (Kohnstamm, Halverson, Mervielde & Havill, 1998; Shiner &

Caspi, 2003). De Fruyt, Mervielde, Hoekstra and Rolland (2000) showed considerable

190 Chapter 5

convergence between the two measures included in the present study. A joint

principal component analysis of adolescent self-ratings on the NEO PI-R and the

facets of the Hierarchical Personality Inventory for Children underscored the assumed

hierarchical structure in both questionnaires.

The HiPIC includes eighteen facets hierarchically organized under the five

domains, with (a) Extraversion comprising the facets Shyness, Optimism,

Expressiveness and Energy, (b) Benevolence including Egocentrism, Irritability,

Compliance, Dominance and Altruism, (c) Conscientiousness consisting of

Achievement motivation, Concentration, Perseverance and Orderliness, (d) Emotional

Stability based on Anxiety and Self-confidence, and (e) Imagination containing

Creativity, Curiosity and Intellect. The parents rate the items on a 5-point Likert scale.

There is evidence for the replicability of the factor structure across samples of

children and adolescents (Mervielde & De Fruyt, 2002).

In the present study, Cronbach alpha’s for mother and father ratings

respectively, ranged from .93 to .94 for Benevolence (N of items = 40), from .94 to

.94 for Conscientiousness (N of items = 32), from .86 to .88 for Emotional Stability

(N of items = 16), from .89 to .91 for Extraversion (N of items = 32) and from .91 to

.92 for Imagination (N of items = 24). The correlations between ratings of mothers

and fathers showed values of .70 for Benevolence, .82 for Conscientiousness, .69 for

Emotional Stability, .69 for Extraversion, and .69 for Imagination (p < .001),

indicating a strong convergence between parental ratings.

Child problem behavior

Child problem behavior was measured using the Dutch version of the

Achenbach Child Behavior Checklist (CBCL; Verhulst, Van der Ende & Koot, 1996).

The two broadband syndromes that can be derived from the CBCL were used as

measures of problem behavior: Internalizing, including items referring to somatic

complaints, social withdrawal, and anxiety/depression, and Externalizing, comprising

items indexing aggression, hyperactivity, and delinquency. Cronbach alpha values for

the scale Internalizing behavior (N of items = 31) equaled .86 for maternal and .88 for

paternal ratings; for the scale Externalizing behavior (N of items = 33) Cronbach

alpha values were .90 for mother and .91 for father ratings. Correlations between

mother and father ratings were all significant (p < .001), with .53 for internalizing

behavior, and .58 for externalizing behavior.


Procedure

Six hundred families were randomly selected by contacting parents via

elementary and secondary schools. For elementary schools the sample was stratified

by province, (East and West Flanders), region (rural or urban), school type

(public/private/catholic schools) and grade (third, fourth, fifth and sixth year of

elementary school). For secondary schools, sampling was based on province (East and

West Flanders), type of curriculum (vocational, technical and general education) and

grade (first and second year of secondary school). The response rate of parents with

children in primary schools was 41%, and 39% for parents with children in secondary

schools. A trained psychology student visited families at home, and instructed parents

to independently complete all inventories.

The cross-validation data were collected by students taking courses in

Developmental Psychology at Ghent University. They were instructed to solicit the

cooperation of a family of their choice. In order to standardize the visits as much as

possible, the research-assistants and students received oral and written instructions.

Both the parents (if they lived together) independently completed the questionnaires at

home, in the presence of the student.

Analyses

A potential methodological problem is the reliance upon reports of the same

rater in one and the same model. To overcome this problem of shared method

variance, we used aggregated scores of mother and father ratings for child personality

and child problem behavior variables. This was accomplished by extracting a

common factor score from each pair of ratings from mothers and fathers.

We controlled for effects of gender and age by regressing the child personality

variables, the parenting variables and the problem behavior variables on age and

gender and saving the standardized residuals. The resulting matrix of standardized

residuals was further used as an input matrix to compute the covariance matrix in

PRELIS 2.53, with list wise deletion of missing observations. Because all the

variables in the model were continuous the Maximum Likelihood method was used to

estimate the model parameters in LISREL 8.53. To evaluate possible data-model

inconsistencies and overall data-model fit, we used the following fit indices: (a) chi-

square; (b) the root mean square error of approximation (RMSEA), with values below

.05 indicating a good fit, between .05 and .08 a reasonable fit, between .08 en .10 a

192 Chapter 5

mediocre fit and above .10 a poor fit (Byrne, 1998); (c) the square root of the average

squared residual (RMR), used to compare the fit of two different models for the same

data (Jöreskog & Sörbom, 1996) with small residuals indicating a well fitting model;

(d) the Comparative Fit Index (CFI) of Bentler, which is an incremental fit index with

values greater than .90 indicating an acceptable fit, and recommended by Marsh, Balla

and McDonald (1996); (e) Akaike’s information criterion, providing information

about the parsimony of the model, with smaller values indicating a more efficient

model. Paths with nonsignificant t-values were removed in a trimming process,

starting with deleting the path with the smallest t-value. In addition, non-significant

phi-coefficients were fixed to zero.

Results

Results of Model 1 testing

Model with Externalizing behavior as the outcome variable. The final model

(see Figure 3) with excellent fit indices (see Table 1) included some significant effects

of mother’s personality variables on parental behavior. Standardized total and indirect

effects are reported in Table 2. There were significant positive effects from mother’s

Conscientiousness (t = 3.74) and Openness to experience (t = 2.46) on Positive

parenting, positive effects from mother’s Extraversion (t = 3.60) and Neuroticism (t =

4.61) on Negative control and a negative effect from mother’s Openness to experience

(t = -2.99) on Negative control. Agreeableness was not significantly related to the

parenting variables nor to Externalizing behavior. None of the mother’s personality

variables was directly related to child Externalizing behavior. However, there were

some significant indirect effects of mother’s personality on externalizing behavior

through the parenting variable negative control: indirect effects of Extraversion (t =

2.86), Neuroticism (t = 3.29) and Openness to experience (t = -2.52).

The child’s personality variables were all related to Externalizing behavior,

except for Imagination, with the highest coefficient for Benevolence (t = -15.32) and

lower but comparable coefficients for Conscientiousness (t = -3.80), Emotional

stability (t = -3.00) and Extraversion (t = 3.60). Conscientiousness affected Positive

parenting (t = 4.08). Benevolence (t = -4.43) and Extraversion (t = 2.34) also affected

Negative control, leading to some significant indirect effects on Externalizing of

Benevolence (t = -3.22) and Extraversion (t = 2.10).


Of the parenting variables, only Negative control had a direct effect on

Externalizing behavior (t = 4.69). The squared multiple correlations for the structural

equations indicate the proportion of variance in the endogenous variables accounted

for by the variables in the structural equations. In this model 51% of the variance in

Externalizing behavior was explained, 13% of the variance in Negative control and

11% of the variance in Positive parenting.

The final model for fathers, after model trimming, showed no significant paths

from father’s personality to child Externalizing behavior, which w as also found in

the mothers’ data. However, there were two significant indirect effects of father’s

personality on Externalizing behavior through Negative control, i.e. Agreeableness (t

= -2.28) and Openness to experience (t = -2.48). Several significant effects were

found from father’s personality on father’s parental behavior: Conscientiousness (t =

5.39) and Openness to experience (t = 3.97) affected Positive parenting. These effects

were also present in the mother’s data. Agreeableness (t = -2.68), Conscientiousness (t

= 2.12) and Openness to experience (t = -3.02) affected Negative control.

The child personality domains Benevolence (t = -14.90), Conscientiousness (t

= -3.67), Emotional stability (t = -3.50) and Extraversion (t = 3.59) predicted

Externalizing behavior. Benevolence also predicted indirectly Externalizing behavior,

through the parenting variable Negative control (t = -3.32).

Of the two parenting variables, only Negative control affected directly

Externalizing behavior (t = 4.34). This model explained 54% of the variance in

Externalizing behavior, 19% of the variance in Positive parenting and 11% of the

variance in Negative control.

The model for father self-ratings of personality and parenting is largely

comparable to the model for mother self-ratings of personality and parenting. The

main difference is that mother’s Extraversion and Neuroticism predicted Negative

control in addition to Openness to experience, whereas father’s Agreeableness and

Conscientiousness explained Negative control in addition to Openness to experience.

Model with Internalizing behavior as the outcome variable. The final model

with maternal self-ratings of personality and parenting (See Figure 4) showed direct

paths from mother’s Neuroticism (t = 2.68) and Openness to experience (t = 2.94) to

child Internalizing problem behavior. This is in contrast with the model for

Externalizing problem behavior, where no evidence was found for direct effects of

mother’s personality on Externalizing behavior. Also in contrast with the previous

194 Chapter 5

model, is the number of indirect effects of mother’s personality on Internalizing

behavior: here, only the effect of Neuroticism (t = 2.04) through Negative control was

significant, whereas in the model for Externalizing behavior, Extraversion and

Openness to experience showed additional indirect effects through Negative control.

The effects of mother’s personality on parenting and the effects of child

personality on parenting are not further discussed here, because they are equal to the

effects in the model explaining Externalizing behavior.

The child personality domains predicted somewhat differently child

Internalizing behavior in comparison with Externalizing behavior. The highest effect

was for Emotional stability (t = -11.83), and there were additional effects of

Benevolence (t = -4.40) and Extraversion (t = -3.59). No effects were found for

Conscientiousness and Imagination. There were no significant indirect effects of the

child personality variables through parenting.

As regards the parenting variables, Internalizing behavior was only predicted

by maternal Negative control (t = 2.27), not by Positive parenting. The model

explained 43% of the variance in Internalizing behavior, 11% in Positive parenting

and 13% in Negative control.

The model with father self-ratings of personality and parenting showed a

direct effect of father’s Neuroticism on child Internalizing behavior (t = 2.69). The

effect of father’s personality and child personality on the parenting variables were the

same as in the model for Externalizing behavior. For the child personality domains

predicting Internalizing behavior, some differences in strength of the coefficients were

found: the strongest predictor of Internalizing behavior was Emotional stability (t = -

10.53), followed by Benevolence (t = -3.95), Extraversion (t = -2.60), and

Conscientiousness (t = -2.11). None of the father’s parenting variables predicted child

Internalizing behavior. The model explained 19% of the variance in Positive

parenting, 11% in Negative control and 42% in Internalizing behavior.

The main difference between the models for mother and father data is the lack

of significant effects of parenting on Internalizing behavior in the father model. Also,

in the mother’s model, both mother’s Neuroticism and Openness to experience

directly affected child Internalizing behavior, whereas in the father’s model only

father’s Neuroticism predicted Internalizing behavior.


Results of Model 2 testing

Model with Externalizing behavior as the outcome variable. After model

trimming by deleting the insignificant paths, the final model had a χ² = 36.21, df = 42,

p = .722. Significant paths were found between the comparable mother and child

personality domains Conscientiousness/ Conscientiousness (t = 4.94), Neuroticism/

Emotional stability (t = -6.63), Extraversion/ Extraversion (t = 3.86), and Openness to

experience/ Imagination (t = 5.54), but not for Agreeableness/ Benevolence. Mother’s

Positive parenting showed a significant path to child Benevolence (t = 2.82) and

Conscientiousness (t = 3.79), but not directly to Externalizing behavior. Mother’s

Negative control showed a significant path to child Benevolence (t = -4.76) and

Externalizing behavior (t = 4.71). There was also an indirect effect of Negative

control on Externalizing behavior (14, t = 4.91), The model indicated no direct, but an

indirect effect of Positive parenting on Externalizing behavior (t = -3.70). All of the

child personality domains, except for Imagination, affected Externalizing behavior

with the highest coefficients for Benevolence (t = -14.63), and comparable effects for

Conscientiousness (t = -3.83), Emotional Stability (t = -3.00), and Extraversion (t =

3.14). Significant paths were also found between several child personality variables,

with Benevolence affecting Extraversion (t = -4.97), Conscientiousness affecting

Emotional Stability (t = -2.92), Extraversion (t = 6.27) and Imagination (t = 9.38),

with Emotional stability affecting Conscientiousness (t = 4.48) and Extraversion (t =

2.12), with Extraversion affecting Benevolence (t = 4.57), Conscientiousness (t = -

5.82) and Imagination (t = 2.53), and with Imagination affecting Benevolence (t = -

3.19) and Emotional Stability (t = 5.53). Due to the inclusion of paths between the

different child personality variables, numerous indirect effects were present, that are

not further discussed.

The final model for fathers (χ² = 21.15, df = 36, p = .977) showed significant

paths between the comparable father and child personality domains Neuroticism/

Emotional stability (t = -5.15), Extraversion/ Extraversion (t = 3.07), and Openness to

experience/ Imagination (t = 2.65), but not between Agreeableness/ Benevolence and

Conscientiousness/ Conscientiousness. As expected from the results in Model 1 there

were no direct effects from father’s personality on Externalizing behavior. Father’s

positive parenting showed significant effects on child Benevolence (t = 4.91) and

Extraversion (t = 2.79) whereas Negative control affected Benevolence (t = -3.05),

196 Chapter 5

Emotional stability (t = -2.81) and Extraversion (t = 2.32). The model included a

direct effect (4.34) and an indirect effect (t = 5.09) from Negative control on

Externalizing behavior. There was no direct effect of father’s Positive parenting, but a

significant indirect effect on Externalizing behavior (t = -2.80). As in the previous

models, the child personality domains Benevolence (t = -15.03), Conscientiousness (t

= -3.68), Emotional stability (t = -3.52) and Extraversion affected (t = 3.61)

Externalizing behavior. Several significant paths were found between some child

personality domains with Benevolence affecting Conscientiousness (t = 6.83),

Emotional stability (t = -2.81) and Extraversion (t = 4.00), Conscientiousness

affecting Emotional stability (t = -2.39), Extraversion (t = -4.50) and Imagination (t =

7.61), Emotional stability affecting Benevolence (t = 3.94), Conscientiousness (t =

3.48) and Extraversion (t = -3.37), Extraversion affecting Benevolence (t = -4.38) and

Emotional stability (t = 4.28), and finally Imagination affecting Extraversion (t =

8.88).

A comparison between the mother and father model showed some differences.

There was no path between child and father Conscientiousness, as was the case in the

mother’s model. In the father’s model there was an additional path from Negative

control to child Extraversion.

Model with Internalizing behavior as the outcome variable. The final model

for mothers with χ² = 32.44, df = 41, p = .828, included significant paths from mother

Neuroticism (t = 2.68) and Openness to experience (t = 2.97) to Internalizing

behavior. Between the comparable mother and child personality domains, the same

significant paths were found as for Externalizing behavior, with similar t-values. The

paths between the child personality variables were also similar to those in the model

for Externalizing behavior. There were significant paths from the child personality

variables to Internalizing behavior, with the largest coefficient for Emotional stability

(t = -11.82) and comparable coefficients for Benevolence (t = -4.41) and Extraversion

(t = -3.59). Negative control directly affected Internalizing behavior (t = 2.29) and

child Benevolence (t = -4.76). Positive parenting affected both child Benevolence (t =

2.82) and Conscientiousness (t = 3.79), but not Internalizing behavior. In contrast to

the Externalizing model, there was no indirect effect of Positive parenting or Negative

control through the child personality domains.


The father model for Internalizing behavior, with χ² = 23.04, df = 36, p = .954,

showed the same effects (and the same t-values) of the father personality domains on

the child personality domains as in the model for Externalizing behavior. There was

also a significant effect of father Neuroticism on Internalizing behavior (t = 2.71). No

direct paths were present from the parenting variables to Internalizing behavior.

However, there was an indirect effect of Positive parenting (t = -2.26) through the

child personality domains, which was not the case for Negative control. Father’s

Positive parenting affected child Benevolence (t = 3.34) and Conscientiousness (t =

2.87), whereas father’s Negative control affected Emotional stability (t = -2.78) and

Extraversion (t = 3.34). The child personality domains Emotional stability (t = -

10.61), Benevolence (t = -3.96), Extraversion (t = -2.62), and Conscientiousness (t = -

2.13) predicted Internalizing behavior. Finally, the same paths were found between

the child personality domains as in the father model for Externalizing behavior.

Comparison of Model 1 and Model 2

In the present study, we specified two alternative or competing models, the so-

called AM situation (Jöreskog & Sörbom, 1993), and on the basis of an analysis of a

single set of data, one of the models should be selected. However, it should be noted

that the analyses in this study cannot be considered as strictly AM (alternative model

testing).They can be considered as explorative and hence as MG (model generating)

because data-fitting models were constructed, starting from almost saturated models,

and trimmed by a process of gradual deletion of non-significant paths.

The choice of one of the two models as the best fitting model cannot be made

straightforwardly in this study. From Table 1 it can be concluded that the model fit

indices RMSEA, RMR, CFI and AGFI are excellent for all the models. The AIC

measure showed smaller values in the second than in the first model, indicating

greater parsimony of the second model. From these findings, one should conclude that

the second model fits the data better, in particular the father data. However, the

second model includes reciprocal paths between the child personality variables,

showing instability indices higher than the rule of thumb value of 1 (Jöreskog &

Sörbom, 1993, p. 155). This is an argument against selecting model 2 as the preferred

model. However, it seems premature to reject model 2 only because of the fact that

198 Chapter 5

the five child personality variables are strongly related, and because they render the

model instable by admitting several paths between these variables.

Cross-validation of the models in an independent sample

The replication of the structure of Model 1 in an independent sample showed

reasonable results for mothers and excellent results for fathers (see Table 1). The p-

values associated with the chi-square values were above .050 for the mother model

explaining internalizing behavior as well as for the father models explaining both

internalizing and externalizing behavior, and above .010 for the mother model

explaining externalizing behavior. RMSEA values were all below .05. Replication of

the model with the path coefficients set equal to the path coefficients of the initial data

indicated that the strength of the relationships differed for the two data sets, with

significant chi-square values for both mothers and fathers and for both internalizing

and externalizing behavior.

Replication of the structure of Model 2 was successful for the mother data

explaining externalizing and internalizing behavior and for the father data explaining

externalizing behavior. The model did not converge for the father data explaining

internalizing behavior. Chi-square values of the constrained model (path coefficients

set equal to the path coefficients of the initial model) showed that the strength of the

relationships was significantly different across the two data sets.

The cross-validation of the models again does not permit to select one of the

two alternative models. In some cases Model 2 was better replicable than Model 1,

but in other cases Model 2 was more replicable than Model 1.

General discussion

Inspired by ecological or contextual models, and in particular by the heuristic

model of Belsky (1984), the present study examined the simultaneous effects of

parent personality, child personality and parenting on child internalizing and

externalizing behavior. The study should be considered as exploratory, because the

fitted models are data-driven rather than theory-driven. This is due to the limited and

inconsistent empirical evidence to derive the initial path models. In this respect, our

study can also be regarded as innovative and contributing to a still emerging body of

empirical findings integrating parent and child personality and parenting in the


prediction of child problem behavior. Although testing and modifying models with

the same data is not the most appropriate method, in practice is it the most common

method. A strength of this study is that to wipe out the problem that the final fitting

models were obtained by capitalizing on chance, we cross-validated them in an

independent sample. A second benefit is the use of a comprehensive personality

taxonomy, i.e. the Five-Factor Model, to assess both child and parent individual

characteristics. Utilizing the same taxonomy for measuring parent and child individual

differences, instead of focusing for example on child temperament, enhances the

comparability of measures across parents and children. A third strength of this study

compared to previous research is the inclusion of both ineffective parental

management techniques, i.e. negative control, as well as more ‘supportive’ parental

behavior. Until now, the role of positive forms of parenting in the prediction of child

problem behavior remains unclear.

As a general conclusion, the results of our study underscore the importance of

including parent personality, child personality and parenting in models predicting

child outcome behavior.

Parent personality as a predictor of parenting, child personality and child problem

behavior

We investigated the effects of parent personality on parenting, based on the

hypothesis that parenting behavior may reflect general and stable parental personality

domains. Mother’s higher self-ratings on Conscientiousness and Openness to

experience were associated with more Positive parenting. The same significant

relationships were found in the father data. These findings corroborate the results of

previous studies summarized by Belsky and Barends (2002). Higher mother self-

ratings on Neuroticism, Extraversion and lower ratings on Openness to experience

predicted more maternal Negative control. The father data also showed that lower

levels of Openness to experience predicted Negative control. In addition, lower levels

of Agreeableness predicted Negative control. Somewhat unexpected was the positive

relationship between father’s Conscientiousness and Negative control. Belsky and

Barends (2002) mention that it is not clear how Conscientiousness should relate to

parental behavior. It is possible that highly conscientious parents are too demanding

for their children, and hence impose a lot of restrictive control. The relationship

200 Chapter 5

between Neuroticism and Openness to experience and the parenting variable Negative

control was also reported in Prinzie et al. (in press).

Agreeableness did not show any significant relationships with parenting nor

with child problem behavior in the mother data, and this finding is in contrast with

Kochanska, Clark and Goldman (1997) who reported a relationship between low

Agreeableness and more power-assertive and less responsive parenting. In the father

data there was a path from Agreeableness towards Negative control. In contrast with

the results of Losoya, Callor, Rowe and Goldsmith (1997) there was no association

between self-reported Agreeableness and more positive parenting.

Previously, parenting behavior has been identified as an important behavioral

variable, that is only modestly associated with personality in non-referred samples

(Spinath & O’Connor, 2003). The finding that parent personality only partly explains

parental behavior, suggests that parental behavior is more than an fallible indicator of

parent personality. This corroborates evidence showing that parenting is multiple-

determined, for instance by sources of stress and support or by the parents’ own

childhood parenting experiences (Belsky, 1984; Spinath & O’Connor, 2003).

There was no evidence for a direct link between child externalizing behavior

and mother’s or father’s personality domains. This is in contrast with the results of

Prinzie et al. (in press), who reported a negative effect of Emotional stability and a

positive effect of Agreeableness on child externalizing behavior problems. Our results

indicated that children ‘s internalizing behavior is directly influenced by mother’s

Neuroticism and Openness to experience. Higher levels of Neuroticism and Openness

to experience predicted more child internalizing behavior. For fathers, only

Neuroticism directly influenced children’s internalizing behavior.

In the second model we investigated relationships between parent personality

and child personality. Significant relationships were found between the coincident

mother and child personality domains Conscientiousness, Neuroticism/ Emotional

stability, Extraversion, and Openness to experience/ Imagination. The same

relationships were found for fathers, except for Conscientiousness. Parents’

Agreeableness did not affect children’s Benevolence. For Model 1 some of the

coincident parent and child personality domains generated the same effects. Both

mother’s and child’s Conscientiousness affected mother’s positive parenting, whereas

both mother’s and child’s Extraversion affected maternal negative control. Mother’s

Neuroticism and the child’s Emotional stability were both related to the child’s


internalizing behavior. In the father model, father’s Agreeableness and the child’s

Benevolence both affected father’s negative control. Internalizing behavior was also

affected by both father’s Neuroticism and child’s Emotional stability.

Our results partly corroborate the ‘vulnerability-stress’ model, which

postulates that a predisposition to a disorder is inherited, and that this liability is likely

to be actualised in the presence of certain stressors in the family, such as poor

parenting (Goldstein, 1988; Lytton & Gallagher, 2002). In the present study for

example, highly neurotic mothers and children affect child internalizing behavior, but

there is also an indirect link between parental neuroticism and child internalizing (and

externalizing) behavior, through negative control. Moreover there is a relationship

between the child’s Emotional stability and parental Neuroticism.

Child personality and parenting as predictors of child problem behavior

The child personality domain Benevolence was the best predictor of child

externalizing behavior, which is in accordance with the study of Prinzie et al. (in

press). As regards internalizing behavior, Emotional stability was the strongest

predictor. Imagination neither predicted child internalizing or externalizing problem

behavior.

In general, our models indicate stronger effects of negative control than for

positive or supportive parental behavior. Only negative control directly affected child

problem behavior. In Model 1 the effects of negative control were partly determined

by effects of parent and child personality. In the mother data, parental Extraversion,

Neuroticism, Openness to experience, and child Benevolence and Extraversion

affected negative control. Mother’s negative control predicted both externalizing and

internalizing child behavior. In the father data, only child externalizing behavior was

related to negative control, which in return was affected by father’s Agreeableness,

Conscientiousness and Openness to experience, and children’s Benevolence. The

model suggests only partial mediation by the parenting variable negative control.

Researchers often inadvertently claim to have detected a mediational effect when in

fact they found evidence for an indirect effect (Holmbeck, 1997). For example, in the

mother data, negative control seems to mediate the association between the child

personality domain Benevolence and externalizing behavior. From the SEM approach

202 Chapter 5

to test mediated effects described in Holmbeck (1997, p.602-603)9, it can be

concluded that none of the parenting variables could be considered as a mediator in

the relationship between parent or child personality and child problem behavior.

Model 2 showed also indirect effects of negative control through the child personality

characteristics.

Positive parenting was not directly associated with child problem behavior.

This is in accordance with Galambos, Barker and Almeida (2003), who found that

parental support was neither a significant predictor of externalizing nor of

internalizing behavior. However, some indirect effects of positive parenting were

found through the child personality variables.

Child personality as a predictor of parenting or vice versa?

The two alternative models postulated different directions for the relationship

between child personality and parenting. For some of the variables, we found that

child personality domains affect certain parenting variables according to the first

model and vice versa via for the second model, hence suggesting bi-directional

influences of child personality and parenting. In Model 1 there is an effect of

Benevolence on Negative control and in Model 2 an effect of Negative control on

Benevolence, for both mothers and fathers. Low child Benevolence is related to

higher Negative control, but high Negative control also leads to lower benevolence in

children. The mother data further show that highly conscientious children elicit more

maternal positive parenting, but positive parenting also increases child

conscientiousness. For fathers, a similar association is found for Benevolence and

Positive parenting. In the father model explaining Externalizing behavior we also

found that Extraverted children elicit more Positive parenting and vice versa. In

9 The SEM approach to test mediated effects is described in Holmbeck (1997, p.602-603). First, in an

overall fitting model, the paths should be significant between the predictor variable (A) and the

mediator (B), between the mediator (B) and the outcome variable (C), and the predictor variable (A)

and the outcome variable (C). Next, one assesses the fit of the A→ B→ C model under two conditions.

In the first condition the A→ C path is constrained to zero, in the second condition this path is not

constrained. A significant improvement in fit over the first model, examined by a significance test on

the basis of the difference between the two model chi-squares, implies that there is no mediational

effect of variable B.


addition it should be noted that these ‘bi-directional’ effects were more stable for the

models explaining externalizing and internalizing behavior in the mother than in the

father data.

Limitations and suggestions for further research

A first limitation of the present study is the small size of the cross-validation

sample. The power of the final models for the initial samples (N = 413 for mothers

and N = 409 for fathers) ranges from .79 to .95, with RMSEA’s ranging from .00 to

.02 (see Hancock & Freeman, 2001). However, the models in the cross-validation

samples (N = 247 for mothers and N = 286 for fathers) have insufficient power with

RMSEA values ≥ .02, possibly resulting in an indeterminate conclusion. Only the

cross-validation of Model 1 explaining internalizing and externalizing behavior in the

father data showed sufficient power.

A second limitation is that the documented relationships do not necessarily

indicate causation. The data in this study are cross-sectional, and do not permit

definite conclusions about the development of problem behavior. Furthermore, some

of the relationships can be examined from a reversed direction, for example the

effects of child problem behavior on parenting behavior. The fact that child

personality affects parenting and vice versa, strongly suggests interactions between

these variables. A possible alternative is to estimate and test hypotheses about

interactions between continuous variables using structural equation interaction models

(cf. Moulder & Algina, 2002; Schumaker, 2002).

The present study tried to extend the limited empirical evidence regarding the

concurrent effects of parent personality, child personality and parenting on child

externalizing and internalizing. Our results suggest that these variables explain a

substantial part of the variance in problem behavior, but also that some variables seem

to be more important than others. Further research is necessary to replicate the

findings of the current study, but also to evaluate new, even more complicated

models.

204 Chapter 5

References




Asendorpf, J. B., Borkenau, P., Ostendorf, F. & van Aken, M. (2001). Carving

personality description at its joints: confirmation of three replicable personality

prototypes for both children and adults. European Journal of Personality, 15, 169-

198.



externalizing behavior. Developmental Psychology,34, 982-995.



Barber, B. K., Olsen, J. E. & Shagle, S. C. (1994). Associations between

parental psychological and behavioral control and youth internalized and externalized

behaviors. Child Development, 65, 1120-1136.



Belsky, J. (1997). Variation in susceptibility to environmental influence: an

evolutionary argument. Psychological Inquiry, 8, 230-235.

Belsky, J., & Barends, N. (2002). Personality and parenting. In M. H.

Bornstein (Ed.), Handbook of parenting. Vol. 3. Being and becoming a parent (pp.

255-277). Mahwah, New Jersey: Lawrence Erlbaum Associates.





Brestan, E. V., Eyberg, S. M. (1998). Effective psychosocial treatments of

conduct-disordered children and adolescents: 29 years, 82 studies, and 5.272 kids.

Journal of Clinical Child Psychology, 27, 180-189.




Brook, J. S., Brook, D. W., & Whiteman, M. (2003). Maternal correlates of

toddler insecure and dependent behavior. Journal of Genetic Psychology, 164, 72-88.

Byrne, B. M. (1998). Structural equation modeling with LISREL, PRELIS and

SIMPLIS: Basic concepts, applications, and programming. Mahwah, NJ: Lawrence

Erlbaum Associates.




Caspi, A., Harrington, H., Milne, B., Amell, J. W., Theodore, R. F., & Moffitt,

T. E. (2003). Children’s behavioral styles at age 3 are linked to their adult personality

traits at age 26. Journal of Personality, 71, 495-514.




Clark, L. A., Kochanska, G., & Ready, R. (2000). Mothers’ personality and its

interaction with child temperament as predictors of parenting behavior. Journal of

Personality and Social Psychology, 79, 274-285.

Collins, W. A., Maccoby, E. E., Steinberg, L., Hetherington, E. M., &



De Fruyt, F., Mervielde, I., Hoekstra, H. A., & Rolland, J.-P. (2000).

Assessing adolescents’ personality with the NEO PI-R. Assessment, 7, 329-345.

De Fruyt, F., Mervielde, I. & Van Leeuwen, K. (2002). The consistency of



De Raad, B., & Perugini, M. (2002). Big Five factor assessment: introduction.




ecology of antisocial behavior. In D. Cicchetti & D. J. Cohen (red.), Developmental


& Sons, Inc.

Ebstein, R. P., Benjamin, J., & Belmaker, R. H. (2003). Behavioral genetics,

genomics, and personality. In R. Plomin, J. C. Defries, I. W. Craig, & P. McGuffin

206 Chapter 5

(Eds.), Behavioral Genetics in the postgenomic era. (pp.365-388). Washington DC:

American Psychological Association.














Goldstein, M. (1988). The family and psychopathology. Annual Review of


Haapasalo, J., & Tremblay, R. E. (1994). Physically aggressive boys from ages

6 to 12: family background, parenting behavior, and prediction of delinquency.


Hancock, G. R., & Freeman, M. J. (2001). Power and sample size for the root

mean square error of approximation test of not close fit in structural equation

modeling. Educational and Psychological Measurement, 61, 741-758.

Hinshaw, S. P. (2002). Process, mechanism, and explanation related to

externalizing behavior in developmental psychopathology. Journal of Abnormal Child


Hoekstra, H. A., Ormel, J., & De Fruyt, F. (1996). NEO

Persoonlijkheidsvragenlijsten: NEO PI R en NEO-FFI. Lisse: Swets & Zeitlinger.

Holmbeck, G. N. (1997). Toward terminological, conceptual, and statistical

clarity in the study of mediators and moderators: examples from the child-clinical and

pediatric psychology literatures. Journal of Consulting and Clinical Psychology, 65,

599-610.


John, O. P., Caspi, A., Robins, R. W., Moffitt, T. E., & Stouthamer-Loeber,

M. (1994). The “Little Five”: exploring the nomological network of the five-factor

model of personality in adolescent boys. Child Development, 65, 160-178.

Jöreskog, K., & Sörbom, D. (1993). LISREL 8: Structural equation modeling

with the SIMPLIS command language. Chicago: Scientific Software International.

Jöreskog, K., & Sörbom, D. (1996). LISREL 8. User's Reference Guide.

Chicago: Scientific Software International.

Kochanska, G., Clark, L. A., & Goldman, M. S. (1997). Implications of mothers’

personality for their parenting and their young children’s developmental outcomes.

Journal of Personality, 65, 387-420.







Laursen, B., Pulkkinen, L., & Adams, R. (2002). The antecedents and

correlates of agreeableness in adulthood. Developmental Psychology, 38, 591-603.

Lerner, R. M., Rothbaum, F., Boulos, S., & Castellino, D. R. (2002).

Developmental systems perspective on parenting. In M. H. Bornstein (Ed.), Handbook

of parenting. Vol. 2. Handbook of Parenting (pp. 315-344). Mahwah, New Jersey:

Lawrence Erlbaum Associates.






of Chicago Press.

Losoya, S., Callor, S., Rowe, D., & Goldsmith, H. (1997). Origins of familial

similarity in parenting. Developmental Psychology, 33, 1012-1023.

Lytton, H., & Gallagher, L. (2002). Parenting twins and the genetics of parenting.

In M. H. Bornstein (Ed.), Handbook of parenting. Vol. 1. Children and Parenting (pp.

227-253). Mahwah, New Jersey: Lawrence Erlbaum Associates.

208 Chapter 5




New York: Wiley.

McCrae, R. R. & Costa, P. T. Jr. (1994). The paradox of parental influence :

understanding retrospective studies of parent-child relations and adult personality. In

C. Perris, W. A. Arindell, & M. Eisemann (Eds.), Parenting and Psychopathology

(pp. 107-125). New York : Wiley.

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

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

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

personality, and life span development. Journal of Personality and Social Psychology,

78, 173-186.







Millikan, E., Wamboldt, M. Z., & Bihun, J. T. (2002). Perceptions of the family,

personality characteristics, and adolescent internalizing symptoms. Journal of the

American Academy of Child and Adolescent Psychiatry, 41, 1486-1494.

Moulder, B. C., & Algina, J. (2002). Comparison of methods for estimating and

testing latent variable interactions. Structural Equation Modeling, 9, 1-19.

Nigg, J. T., & Hinshaw, S. P. (1998). Parent personality traits and

psychopathology associated with antisocial behaviors in childhood attention-deficit

hyperactivity disorder. Journal of Child Psychology and Psychiatry, 39, 145-159.










Prinzie, P., Onghena, P., Hellinckx, W., Grietens, H., Ghesquiere, P., &

Colpin, H. (in press). Parent and child personality characteristics as predictors of

negative discipline and externalising problem behaviours in children. European

Journal of Personality.

Putnam, S. P., Sanson, A. V., Rothbart, M. K. (2002). Child temperament and

parenting. In M. H. Bornstein (Ed.), Handbook of parenting. Vol. 1. Children and

parenting (pp. 255-277). Mahwah, New Jersey: Lawrence Erlbaum Associates.

Reti, I. M., Samuels, J. F., Eaton, W. W., Bienvenu, O. J., Costa Jr, P. T., &

Nestadt, G. (2002). Influences of parenting on normal personality traits. Psychiatry

Research, 111, 55-64.

Rothbaum, F., & Weisz, J. R. (1994). Parental caregiving and child externalizing



on personality development. Journal of Counselling and Development, 68, 606-611.

Rowe, D. (1997). Genetics, temperament, and personality. In R. Hogan, J.

Johnson, & S. Briggs (Eds.), Handbook of personality psychology (pp. 795-824).

Orlando, FL: Academic Press.









Scarr, S. (1992). Developmental theories for the 1990s: Development and

individual differences. Child Development, 63, 1-19.

Schumacker, R. E. (2002). Latent variable interaction modeling. Structural

Equation Modeling, 9, 40-54.

210 Chapter 5




Spinath, F. M., & O’Connor, T. G. (2003). A behavioral genetic study of the

overlap between personality and parenting. Journal of Personality, 71, 785-808.




29, 17-29.


Brunner/Mazel.

Van Bakel, H. J. A., Riksen-Walraven, J. M. (2002). Parenting and

development of one-year olds: links with parental, contextual, and child

characteristics. Child Development, 73, 256-273.

Van Leeuwen, K. G., Mervielde, I., Braet, C., & Bosmans, G.. Child

personality and parental behavior as moderators of problem behavior: a variable- and

a person-centered approach. Manuscript submitted for publication.

Van Leeuwen, K. G., Mervielde, I., De Clercq, B.J., & De Fruyt, F. Child

personality and parental behavior as interacting predictors of child internalizing and

externalizing behavior in clinically referred and non-referred children. Manuscript

submitted for publication.



Assessment.











Zahn-Waxler, C., Duggal, S., & Gruber, R. (2002). Parent psychopathology. In

M. H. Bornstein (Ed.), Handbook of parenting. Vol. 2. Social conditions and applied


Table 1

Goodness of fit measures Model 1 Chi-square Df P RMSEA RMR CFI AGFI AICMothers Externalizing behavior 26.34 34 .823 .000 0.019 1.00 0.97 140.34 Cross validation structure 51.38

34 .028 .047 0.037 0.97 0.92 165.38

Cross validation constraints 81.18 47 .001.931

.056 0.046 0.95 0.91 195.18137.84 Internalizing behavior 21.84 33 .000 0.019 1.00 0.98

Cross validation structure 46.35

33 .062 .041 0.035 0.98 0.92 162.35Cross validation constraints

83.58 47 .001 .057 0.048 0.94 0.90 199.58

FathersExternalizing behavior 36.95 33 .291 .017 0.025 1.00 0.96 152.95 Cross validation structure 24.03

33 .873 .000 0.024 1.00 0.96 140.03

Cross validation constraints 68.35 46 .018.300

.042

.017 0.058 0.96

1.00 0.930.96

184.35153.75 Internalizing behavior 35.75 32 0.025

Cross validation structure 24.46

32 .827 .000 0.024 1.00 0.96 142.46194.83 Cross validation constraints 76.83 45 .002 .051 0.058 0.96 0.92

Model 2 Mothers Externalizing behavior 36.21 42 .722 .000 0.026 1.00 0.97 134.21 Cross validation structure 59.03 .041

42 .042 0.046 0.97 0.92 157.03

Cross validation constraints 107.64 65 .001 .052 0.074 0.94 0.91 205.64Internalizing behavior 32.56 41 .824 .000 0.027 1.00 0.97 132.56 Cross validation structure 52.57

41 .106 .034 0.044 0.98 0.93 152.57Cross validation constraints

115.02

65 .000 .057 0.073 0.93 0.91 215.02 Fathers

Externalizing behavior 21.15 36 .977 .000 0.021 1.00 0.98 131.15 Cross validation structure 41.99

36 .227 .024 0.035 0.99 0.94

0.93151.99

Cross validation constraints 91.14 61 .007 .042 0.0590.022

0.961.00

201.14Internalizing behavior 23.04 36 .954 .000 0.98 133.04 Cross validation structure *

Cross validation constraints ** model does not converge

Table 2

Standardized Total and Indirect Effects of Model 1 Predictors NEO-A NEO-C NEO-E NEO-N NEO-O BE CO ES EX IM POS CONMothers I I I T I T I T I T T T I T I T I T I T I T I TPOS .18 a .12a .20a CON 2 - - 1

- - - - -

.20a . 3a .15a .21a .1 a

.14EXT .0

3a .03

a .04

a .04a

.03a .0

3a .6

0a .0

4a - .15

a .1

1a a .02

a .1

7a

Mothers POS a 1 .2 .18 . 2a 0a CON

- - 1

01 - a - - -

.20a .23a .15a .21a

.19 .1 a

INT .02

.02

.13a .02a .10a - . .02a .50a

.14a

.01

.09a

Fathers POS 0 .25a .18a .13a .2 a CON - -

- .02 - - - .0 - 1 ers

- .13a .11a .14a .25a EXTFath

- .02a .02a .02 - .02a .02a

.62a 4a - .14a .13a .13a . 5a

POS 20 .25a .18a .13a . a CON - .13a - -

11a - .17a - .09a 46a - .11a .14a .25a

INT . - . .11a

a p < .05; an alpha value of .05 was used for all statistical tests; T = standardized Total effects, I = standardized Indirect effects

Parental ratings: NEO-A: Agreeableness; NEO-C: Conscientiousness; NEO-E: Extraversion; NEO-N: Neuroticism; NEO-O: Openness to experience; POS: Positive

parenting; CON: Negative control; EXT: child Externalizing behavior; INT: child Internalizing behavior

Child ratings: BE: Benevolence; CO: Conscientiousness; ES: Emotional stability; EX: Extraversion; IM: Imagination

Table 3

Standardized Total and Indirect Effects of Model 2 Predictors

NEO-N CO IM NEO-C NEO-E NEO-O BE ES EX POS CONMothers I T I T I T I T I T I T I T I T I T T I T IBE .06 .06 .11 - - a - 5 a .2 a .20 a a .17 a .3 a - 2 a - a .30 a .1 a - 04 - a a .11 a - .05 a - .05 a - .05 a .05 a .45 .4 0 .17 5 .6 .25 5 . .19 a .16 a CO - .07 .09 - .28 - - - -

.01 - - 02 - 02 02 .14 a 07 -

-

rs

.20 a

.01 - .10 a .07 a -

- a - .09 a - .02 .02 a .28 a

.05.34 a .34 a .26 a

.01 --

.20 a .22 a .51 a .04

- .07.34 a

- .0

.072 a

.20 a

.- .02

.02 - .10 a

..10 a .ES

EX -

- .1

.014 a

.01

.11 a- .32 a .00

.09.07 a .07 a .05 -

-

.04 .22 a -

.01.

.04 - .22 - .09 a - a .07 a .07 a

- .42 a .87 a

.44 a - .63 a .27 a .67 a

- .67 a .32 a .32 a .02 .02 .15 a

.15 a

IM .13 a .13 a --

.02 --

.02 - .06 a -

.06 a .22 a .00.03

.08 .08 .42 a - .10 a .18 a .18 a - .06 --

.23 a .02 .02 .11 a .11 a -

.03 - .03

EXT Mothe

.05 a

- .05 a

.04 a

.04

.07 a .07 a .03 a

a - .39 a .17 a

- .15 a

.00 - .21 a

- .10 a

- .13 a

.25 a .15 a .15 a

- .11 a - .11 a .31 a

.14 a

BE .06 a .17 .62 .06 a .11 a .11 a - .05 a - .05 a - .05 a - .05 a - .45 a - .45 a .20 a .20 a a .17 a .35 a - a - .25 a .30 a .15 a - .04 - .19 a .16 a CO ES

- -

- 02 .02 . . .06 .11 08 - - .02

.20 a - .10 a .07 a - .07 a - .09 a - .09 a - .02 -

.02 .28 a .28 a --

.34 a - .34 a .26 a - .20 a - .22 a .51 a - .07 - .07 .20 a

- .02 - .10 a

.10 a -

.01 - .01 .01 .01 - .32 a .00 .07 a .07 a - .05 -

.05 .04 .22 a -

.01.27

- .01 .04 .67 a

.04

.67 a.34 a .02 a - .02

.- .02 .02 .02

EX .14 a .14 a .11 a - .22 a 02

- .09 a - .09 a .07 a .07 a 00

- .42 a .87 a .44 a - .63 a a .07.18

- .32 a .32 a .15 a .03

.15 a IM .13 a .13 a -

-.02 -

- -

.06 a -

.06 a .22 a .07

.08 .08 .42 a - .10 a .

.18 a a --

- .23 a

.02

.17.02.17

a .11 a

- - .03

INT - .02 - .02 .04 a

.04 a

.29 a .18 a

- .04 a

- .01 .16 a

- .08 - .56 a

.06 a .13 a

.02 - a - a - - .02 .09 a

.00 Fathers

BE - .04 .04 .19 .19 .02 a - .16 .16 a - a - a - a - - .02 a - .40 a - .40 a - .15 a .15 a .43 a - .26 - .21 a .57 a - a - a .17 a - .17 a - .24 a - .05CO

a -

.02 .02 - .19 a -

.19 a .01 .01 .21 a -

.29 a - .23 a - .23 a .45 a .00 .10 .10 .08 .08 .09 a .09 a - .13 a - .13 a ES .08 a .08 a - .22 a .21 a .04 a .04 a -

.20 a .32 a - .34 a .12 - .48 a - .48 a .46 a - .14 .37 a .37 a

.26.01 .01 - .02 .22 a

EX .12 a - .06 a .04 .04 .06 a .06 a .34 a -

.17 a -

.10 .48 a -

.10 .33 - .33 a - .33 a .53 a - a .23 a .04

.06 .06 - .09 a IM EXT

.01 .01 -

.09 a

.17-

.09 a .12 a .00 .10 a .10 a .36 a - .10 a .20 a .20 a

.26.05 .05 .04 .04 a .04 a -

.06 a .06 a

.02 a

.02 a

a .17 a

.01 .01

- .30 a .28 a

.01 .15 a - .39 a

- a .13 a

.00 .10 a .10 a

- .08 a

- .08 a

.32 a

.17 a Fathers

BE - .20 - .20 a - .02 a .02 a .08 a - .08 a - a - - .69 a - .69 a .17 a - .47 a .43 a .43 a - .32 a .89 a - .20 a - .70 a .16 a - .22 a - .21 a - .21 a CO ES

- - - a 03 - - - .05

.01 .01 - .23 a -

.23 a -

.01 -

.01 -

.20 a

.13-

.20 a - .46 a - .46 a .48 a - .46 a .03 .79 a - .06

.36 - .06 .05 - .18 a

- .10 a

.10 a .20 .07 a .07 a - .25 a .23 a .04 a .04 a a .64 a - .27 a .28 - .48 a

- .48 a .27 a .16 a .36 a -

.01 -

.01 -

.03 a

EX .08 a - .19 a - .01 - .01 .06 a .06 a .24 a

- 1.04 a .27 a .30

1.26 a .01 .84 a - .71 a

- .71 a .53 a

- .56 a

.22 a .22 a

.09 a .24 a

IM .00 .00 -

.13 a -

.13 a .11 a .00 - .11 - .11.03

- .26 a .27 a .27 a

.02 .02 --

.04

.19--

.04

.19 a. .03 - .06 a

.06 a

INT .03 a - .03 a .28 a .17 a - .02 a - .02 a - .14 a .02 .11 - .36 a .10 - .11 a .00 a a - .05 a .05 .05a p < .05; an alpha value of .05 was used for all statistical tests; T = standardized Total effects, I = standardized Indirect effects ;Parental ratings: NEO-C: Conscientiousness;

NEO-E: Extraversion; NEO-N: Neuroticism; NEO-O: Openness to experience; POS: Positive parenting; CON: Negative control; EXT: child Externalizing behavior; INT:

child Internalizing behavior ; Child ratings: BE: Benevolence; CO: Conscientiousness; ES: Emotional stability; EX: Extraversion; IM: Imagination


Model 1

Model 2

Parent

Personality

Child

Personality

Parenting Child Problem Behavior

Parent

Personality

Parenting

Child

Personality

Child Problem Behavior

Figure 1

Two alternative path models explaining child problem behavior

216 Chapter 5

Note: Model 1 for fathers includ

Agreeableness (-.13) and Conscie

and Extraversion (.19) to Positive

the father data: from father Ex

Conscientiousness to Positive pare

Agreeableness

Conscientiousness

Neuroticism

Extraversion

Openness to experience

Benevolence

Conscientiousness

Child

personality

Parent

personality

5

Emotional stability

Extraversion

Imagination

Figure 3

Model 1 for mothers explain

.18

.23

Positive parenting
.20
.12
-.1
es t

ntio

pare

trave

ntin

1

ing

-.2

he following extra

usness (.11) to Neg

nting. The followin

rsion and Neuroti

g.

Negative control

0

5

1

Externalizing beh

-.6

-.1

-.1

.14

paths: from the father per

ative control, from child B

g paths are present in the

cism to Negative control

avior

.17
.11
.20

sonality domains

enevolence (.13)

mother but not in

and from child

Externalizing behavior


Figure 4

Model 1 for mothers explainin

Note: Model 1 for fathers includes

Agreeableness (-.13) and Conscient

and Extraversion (.19) to Positive

behavior. The following paths are

Neuroticism to Negative control, fr

Extraversion to Negative control.

Agreeableness

Conscientiousness

Neuroticism

Extraversion

Openness to experience

Benevolence

Conscientiousness

Emotional stability

Extraversion

Imagination

Parent

personality

Child

personality

5

.18

e parentingPositiv

.20

g Internalizing behavio

the following extra path

iousness (.11) to Negative

parenting, from child Co

not present in the fathe

om child Conscientiousne

Negati

9

1

0

4

.13

.23

.12

-.1

.10

-.1

-.2
Internalizing .20
ve control behavior

-.5

-.1

.11

r

s: from the fath

control, from

nscientiousness

r data: from f

ss to Positive p

.09

er personality domains

child Benevolence (.13)

(-.09) to Internalizing

ather Extraversion and

arenting and from child

General Conclusions 219

GENERAL CONCLUSIONS

In this section the findings from the previous chapters are integrated and

general conclusions are formulated. Furthermore, we discuss the clinical implications

of our results. Finally, some general limitations of our research and promising

directions for future research are discussed.

The relative contribution of child personality, parent personality and parenting

to child problem behavior

This dissertation joins several contemporary research trends. First, research on

interactions between parenting and non-environmental factors, has been suggested as

one of the new directions in the study of child development (Collins, Maccoby,

Steinberg, Hetherington & Bornstein, 2000; Cummings, Davies & Campbell, 2000;

Gallagher, 2002). This approach addresses criticism by behavioral geneticists (e.g.

Rowe, 1997) and others (e.g. Harris, 2000), stating that the effects of parenting are

often overestimated, e.g. by studying only main effects of parenting. Our research

investigated the effects of interactions between child personality and parenting on

child outcome (chapters 3 and 4). The psychometric qualities of the questionnaire

used to assess parental behavior were evaluated in chapter 1. Second, the present

dissertation joins the current enthusiasm for the person-centered and variable-centered

approach to personality (cf. Asendorpf, Caspi & Hofstee, 2002), by examining the

replicability of types and their incremental validity (chapter 2) and by assessing

interactions between personality dimensions and types and parenting behavior as

predictors of child outcome behavior (chapter 3). A third, innovative approach

included an investigation of assumptions of a continuity or spectrum model of

individual differences (Costa & Widiger, 2002; Widiger & Clark, 2000), postulating

that differences between normal and clinical samples are quantitative rather than

categorical. This was examined by comparing parenting by child personality

interactions in a referred and a non-referred sample of children (chapter 4).

The last chapter of the dissertation focuses on ecological models of child

development (Belsky, 1984; Lerner, Rothbaum, Boulos & Castellino, 2002), and takes

220 General Conclusions

into account multiple determinants of child outcomes, by examining with path

analysis, the concurrent effects of parent personality, child personality and parenting

on emotional and behavioral outcomes of children (chapter 5). The major findings

from our studies are summarized in the following section.

Which child personality characteristics serve as a protective or a risk factor in rather

inadequate rearing environments?

The most prominent child personality domains interacting with parenting

(chapters 3 and 4) were benevolence and conscientiousness. Benevolence refers to the

broad trait underlying egocentrism, irritability, compliance, dominance and altruism,

whereas conscientiousness combines the facets achievement motivation,

concentration, perseverance and orderliness. Children rated low or around the mean

on benevolence and conscientiousness were at risk for externalizing behavior when

parents showed high levels of negative control in the interaction with their child.

Negative control refers to discipline, harsh discipline, and ignoring unwanted child

behavior. High scores on this dimension indicate inadequate efforts of parents to

influence the child’s behavior. In contrast, children rated high on benevolence and

conscientiousness were not likely to show externalizing behavior even when exposed

to negative control. In other words children rated high on these personality domains

are protected against the effects of detrimental parenting. With internalizing behavior

as the outcome variable interactions are almost nonexistent.

Children rated low on benevolence were also more prone to externalizing

behavior when parents exhibited little positive parental behavior. This parenting

dimension refers to parental involvement, social reinforcement, problem solving,

stimulation of autonomous behavior and rule setting, with parents scoring high on this

dimension being supportive in the relationship with their child. The interactions

between positive parenting and personality domains turned out to be less stable across

informants and time, compared to the negative parental control by personality

interactions. By taking into account multiple determinants of problem behavior (in

chapter 5), positive parenting no longer predicted child externalizing or internalizing

behavior across informants.


Is it possible to identify ‘types’ of children which are more or less vulnerable of

showing problem behavior in the presence of certain parental behavior ?

Our results show some evidence for three types of children: the overcontrolled

children, scoring low on emotional stability and extraversion, the undercontrolled

children, rated low on both benevolence and conscientiousness, and the resilient

children, scoring above the mean on conscientiousness, benevolence, emotional

stability, extraversion, and imagination. Our findings demonstrate that parental

behavior only affected particular groups of children. Undercontrolled children showed

more externalizing problems when also exposed to parental restrictive control,

whereas overcontrolled children exhibited more emotional problems when

experiencing negative control. Resilient children did not show problem behavior, even

when the child-rearing environment was not optimal, i.e. children rated high on

socially desirable personality domains were not affected by high levels of negative

control. Finally, positive parenting coded as a categorical variable showed no

significant interactions with personality types in the prediction of child problem

behavior.

Are there different antecedents for behavioral problems and emotional problems?

Previous studies examining the effects of parenting and child individual

characteristics have predominantly focused on externalizing behavior as the outcome

variable. The present research also included child internalizing behavior, allowing to

study the relative importance of the predictor variables for both types of

developmental problems.

Our results suggest that child personality characteristics and parenting

differentially affect emotional versus behavioral problems in children. Benevolence

and conscientiousness were the main predictors of externalizing, whereas emotional

stability and extraversion were predominantly associated with internalizing behavior.

Different forms of parental behavior also produced different outcomes.

Negative parental control was primarily predictive of externalizing behavior, but also

had a minor effect on child internalizing behavior. The effects of positive parenting

were less stable, often showing smaller effects than negative control and with diverse

results depending on the kind of analysis.

Moderating effects were predominantly found for externalizing behavior.

Internalizing behavior was primarily determined by main effects: both parenting and


child personality were independent predictors of internalizing behavior that did not

interact. Internalizing behavior increased with more negative parenting and with

below average emotional stability and extraversion.

Are the effects of child personality and parenting on problem behavior the same for

referred and non-referred children?

Extending of our general community sample with a sample of children

referred to an ambulant or residential clinical setting (see chapter 4) allowed us to

explore mean level differences in parenting and personality, differences in main

effects of personality and parenting, and differences in parenting by personality

interactions.

Mean level analyses showed that the referred group of children obtained

higher scores on emotional and behavioral problems and lower scores on

benevolence, conscientiousness, emotional stability, extraversion and imagination.

Furthermore, children in the referred group experienced less positive parenting and

more negative control.

It was shown that personality was a strong cross-sample predictor of both

externalizing and internalizing behavior but with a somewhat stronger effect in the

referred than in the non-referred group, particularly for predicting internalizing

behavior. However, the analyses pointed towards a quantitative rather than to a

qualitative difference. Parental negative control was the best predictor of behavioral

problems in both the referred and non-referred group. Moreover, lack of positive

parenting predicted problem behavior, but only for children in the non-referred group.

In line with the continuity or spectrum hypothesis we did not expect to find

differences in moderator effects across referred and non-referred samples of children.

In other words, it was shown that problem behavior is predicted by the same variables

and presumably based on similar processes in both groups. Hence, our results largely

confirmed the spectrum or continuity hypothesis.

Should parenting be considered as a moderator or as a mediator of the relationship

between personality and problem behavior?

Theoretical models (Belsky, 1984) and previous research (Prinzie, 2002)

suggest that parental behavior can be considered as mediating the relationship

between individual characteristics and problem behavior of the child. In the fifth


chapter it was investigated whether the parenting dimensions negative control and

positive parenting can account for the relationship between personality characteristics

(of parents and children) and child problem behavior. The results did not provide

strong evidence for this hypothesis; only indirect effects of parenting were present. In

the mother data self-rated extraversion, neuroticism, openness to experience, and child

benevolence affected negative control, which in turn predicted both child

externalizing and internalizing behavior. In the father data, self-rated agreeableness,

conscientiousness, openness to experience and child benevolence predicted negative

control, which only predicted child externalizing behavior. The results reported in

chapters 3 and 4 provide evidence for parenting as a moderator, i.e. under specific

conditions of parenting a given effect of child personality occurred. Our results further

point towards bi-directional influences of some personality and parenting variables,

such as benevolence and negative control.

What is the relative importance of parent personality, parenting and child personality

as determinants of child outcome behavior?

The fifth chapter included parent personality, in addition to parenting behavior

and child personality as predictor of child problem behavior. None of the father or

mother personality domains was directly related to child externalizing behavior.

Higher levels of parental neuroticism were associated with higher levels of

internalizing behavior for both mothers and fathers, whereas higher levels of maternal

openness to experience predicted more internalizing behavior . It was also shown that

parent personality moderately affected parental behavior. For both mothers and

fathers higher self-ratings on conscientiousness and openness to experience were

related to more positive parenting. Higher self-ratings on neuroticism and extraversion

and lower self-ratings on openness to experience predicted more negative control for

mothers, whereas lower openness to experience, lower agreeableness and higher

conscientiousness induced paternal negative control. Our results suggest that parental

behavior is more than a reflection of parent personality alone and hence that parenting

probably has multiple determinants. For instance it is likely that parenting is affected

by sources of stress, such as marital relationship problems and by a disadvantaged

socio-economic background.


Is the dimensional approach more suitable for research purposes and the typological

approach more suitable for professional practice?

One important feature of this dissertation is the comparison of the utility of the

dimensional versus the categorical or typological approach. Relying on the results of

the present studies it can be concluded that the dimensional approach is particularly

interesting for research purposes. The second chapter showed that the Five-Factor

Model of personality had more predictive power than the type approach. The third

chapter also provides evidence for the usefulness of the dimensional approach to

study the antecedents of problem behavior. It was pointed out that the dimensional

approach, compared to the type approach, does not suffer from a reduction of

interindividual variation or from a restriction of range. Therefore, adopting a

dimensional approach for assessing personality traits is probably a more sensitive

strategy for assessment of a broader spectrum of psychological functioning.

The second and third chapter of this dissertation revealed clinical correlates of

the different prototypes for both children and adolescents. This indicates the

advantage of the type-approach in particular for clinical psychologists. Types are

convenient because they consider several traits concurrently. The use of a single label

for the interplay of traits within an individual facilitates communication among

practitioners. For purposes of prevention, the types can be used to screen at-risk

children in a non-referred population. For purposes of diagnosis, types are interesting

because personality domains are considered in conjunction with each other.

Moreover, the types are derived from FFM ratings on questionnaires describing the

normal range of child behavior instead of maladaptive behavior. Subjects or

informants might feel more willing and cooperative to describe individual differences

using an adaptive item pool.

Clinical implications

Implications for diagnosis

This research suggests that assessment of both parental behavior and child

personality is advisable when children are referred for behavioral or emotional

problems. Assessing child personality and parenting are necessary to examine parent-

child interactions (Ostergren, 1997). The measures used in our dissertation are

suitable for this purpose. The Ghent Parental Behavior Scale is easy to administer and


allows parents and children to rate both strengths and weaknesses in parental

behavior. The Hierarchical Personality Inventory for Children is a robust measure of

child personality based on the FFM. For both measures extensive normative data are

available, facilitating a more effective delineation of the borders between adaptive and

maladaptive behavior.

Implications for treatment programs

The present study supports the viewpoint that parents should not be blamed

for the behavioral difficulties of their children or deserve too much credit for the lack

of problem behavior because such a stance ignores the fact that effects of parenting

interact with and hence depend on child characteristics. Likewise, therapists and

counselors who take the blame for children’s problem behavior off the parents’ back,

may find out that these parents will become more willing co-therapists and active

agents in the therapeutic process (Rowe, 1990).

Some clinicians hesitate to acknowledge that temperament or personality

predicts child problem behavior. Because child personality characteristics are

relatively stable, it is presumed that this stability implies that problematic behavior

is also stable and hence difficult to treat. We hope that this dissertation will

encourage parents and clinicians to adopt a more realistic perspective, in stead of a

pessimistic or deterministic point of view. It is true that it will not be possible to

fundamentally change some child predispositions. Although a highly active child

may be able to expend some energy in sports or outdoor activities, this will not turn

him/her into a calm and composed child. However, by enhancing some specific skills

and providing information on how to behave in specific situations, the child may

gradually increase the level of adaptive behavior. On the other hand, environmental

changes may reduce the likelihood that predispositions become more manifest (Raine

& Dunkin, 1990). Given a mismatch between child characteristics and parental

behavior, parents can be instructed to use more effective parenting strategies in order

to prevent or alter problem behavior. For example our results show that a lot of

negative control is more detrimental than beneficial for the child’s development.

Parents can learn to adopt alternative parenting strategies, such as the rewarding of

prosocial behavior, to mould the child’s behavior.


Addendum: temperament-focused parent training

According to Putnam, Sanson and Rothbart (2002) programs aimed at

enhancing parenting skills will be deficient if parents attention is not directed to the

child’s individuality and to the need to be flexible in their approach to childrearing.

Taking the child’s individual characteristics into account can guide the development

of more effective individually-tailored treatment programs. The educational

component of such a program familiarizes parents with the concept of child

temperament and helps them understand how the child’s temperament contributes to

child behavior. This is followed by a training in parent management skills that are

compatible with the child’s temperament (Sheeber & Johnson, 1994). Evaluation of

the benefits of this kind of training shows that participants report greater satisfaction

in the relationship with their children, feel more competent as parents, and experience

more secure attachment. Teerikangas, Aronen, Martin and Huttonen (1998) have

shown that the psychosocial prognosis of children at temperamental risk (i.e. being

fussy/demanding) can be improved, by home-based interventions focusing on parent-

child interaction during the first five years of their life. Our research investigating

parenting and child personality relationships may be indicative for the development of

this kind of programs.

Limitations and suggestions for further research

Assessment of parental behavior

Although the Ghent Parental Behavior Scale is a promising instrument, based

on a sound theoretical framework, showing reasonable psychometric properties and

being practical to use, further research is necessary, with a focus on examining the

utility of the questionnaire for clinical practice by investigating the discriminative

validity for parents with or without child-rearing problems. Our study was only based

on parental and child ratings of parental behavior. However, studies utilizing direct

behavioral observations usually show larger effect sizes (Collins et al., 2000). One

possibility is to relate observations of parent-child interactions to the GPBS scales, in

order to further investigate its construct validity.

In line with the type-approach to personality assessment, one wonders whether

it would be useful to identify types of parenting. Parenting typologies result from the

combination of different parenting dimensions or behaviors. For example Maccoby


and Martin (1983) created four parenting types, i.e. authoritarian, authoritative,

permissive and indifferent parenting, by combining high and low demandingness and

responsiveness. However, because most parenting typologies are not yet properly

validated it is suggested to further examine the dimensions that constitute overall

parenting style (Galambos, Barker & Almeida, 2003). Previous research adopting the

typological approach in parenting has several limitations, such as: (a) the lack of

ecological validity, resulting in difficulties to properly classify all of the parents; (b)

the assignment of parents to types based on sample-specific median splits rather than

on the basis of well-validated cutoff points; (c) the limited information on the relative

contribution of the dimensions constituting the types (Jackson & Foshee, 1998).

Nevertheless, a type-approach to parenting can be considered as viable because it

integrates multiple indicators into a in single type (e.g. high negative control and low

positive parenting), which is convenient for both researchers and clinicians.

Therefore, the development of a valid parenting typology might be an interesting

future research line.

Transactional effects

One of the shortcomings in our research is that we cannot draw conclusions

about transactional effects (Lytton, 1990; Sameroff, 1975), pertaining to the recurrent

reciprocal interchanges over time between the environment (parents and others) and

the child. This is due to the strong demands of a research design investigating these

transactional effects. However, our study includes data measured at two assessment

occasions separated by a 3-year interval. Further analyses of these follow-up data are

feasible, allowing more firm conclusions than cross-sectional analyses, and hence

enhancing our understanding of the determinants of child problem behavior. For

instance, we could use cross-lagged panel analysis to examine relationships between

parenting at Time 1 and children’s behavior at Time 2, controlling for children’s

behavior at Time 1 (Vandell, 2000). An interesting alternative, suggested by

Stoolmiller (2002), is to study the effect of the interaction of child problem behavior

and parenting at Time 1 on child behavior at Time 2. This is based on the idea that

unmanageability interacts synergistically with parenting behavior in the prediction of

later antisocial behavior, hence providing support for the theory of coercive family

cycles.


Age and gender

In our analyses we consistently controlled for effects of child gender and age

instead of looking for differential effects of age and gender. Further research could

examine whether parenting by personality interactions are different for girls and boys

or depend on age groups, e.g. school-aged children versus adolescents. The follow-up

data further permit to investigate developmental changes, such as changes in parental

behavior over time. From a developmental point of view individuals need to

accomplish certain ‘developmental tasks’ across different life-span periods, in order to

reach global competence or successful adaptation (Masten & Coatsworth, 1995). For

example, adolescents need to reach a certain degree of independence, but this is only

possible when parents stimulate autonomy. It could be examined if parents adapt their

parental behavior according to the lifespan or developmental stage of the child.

Within-family analysis

Another promising direction for future research is the analysis of within-

family effects. Studying families with at least two siblings living in the same

household, can provide additional information on effects of parenting (Vandell, 2000).

In behavioral genetics, parenting has been identified as a shared environmental

influence, i.e. the experiences shared by children growing up in the same family.

Several studies indicate that in particular non-shared environmental influences, i.e. the

unique experiences of each individual, also predict child adjustment. The effects of

the shared environment, including the family environment, are thought to be minimal.

However, differential treatment by parents of children in the same family might entail

different outcome behavior. Likewise it is possible that children with different

personalities elicit different parental behavior, leading to differential developmental

outcomes. One convenient way to measure differential parenting consists of

computing difference scores for parental and maternal parenting. An alternative and

more sophisticated approach would be the use of multilevel modeling allowing

simultaneous estimation of shared environmental effects and child-specific effects (cf.

Jenkins, Rashbash & O’Connor, 2003).

Effects of the broader family context

As mentioned before, multiple sources may contribute to the development of

(problem) behavior in children. Intelligence, neuropsychological deficits,


temperament, parenting family climate, marital relationship, relationships with peers,

and contextual factors (e.g. neighbourhood, socio-economic status) have been taken

into account as determinants of problem behavior. Ecological models of child

development suggest that the broader social environment, or the ‘macrocontext’ is an

important determinant of child functioning. Although not included in the present

research, we obtained measures assessing the socio-economic status of the family,

marital conflict and stressful life-events, which permit us to investigate hypotheses

regarding the stressful family environment. Another significant predictor of child

problem behavior, which was not included in this study, is the influence of peers

(Harris, 2000).

Final Conclusion

In conclusion, the research reported in this dissertation can be considered as

innovative and adding new information to the field of developmental

psycho(patho)ology. From a theoretical point of view, we incorporated some actual

research ideas, by studying interactions between environmental, i.e. parenting, and

non-environmental factors, i.e. personality, by adopting both a person-centered and a

variable-centered approach in our studies, and by investigating assumptions of a

continuity or spectrum model of individual differences.

The present research further provided empirical evidence that improves our

understanding of the antecedents of child emotional and behavioral problems. The

studies identified some important predictors of problem behavior. It was also

demonstrated that problem behavior is predicted by the same variables in a referred

and a non-referred group of children and presumably based on similar processes.

Finally, our results yielded some implications for clinical practice, such as the

importance of combining the assessment of parental behavior and child personality.

Our findings may be indicative for the development and improvement of therapeutic

programs aimed at preventing and reducing problem behavior in children and

adolescents.


References

Collins, W. A., Maccoby, E. E., Steinberg, L., Hetherington, E. M., &






Cummings, E. M., Davies, P. T., & Campbell, S. B. (2000). Developmental

Psychopathology and Family Process. Theory, research, and clinical implications.

New York: The Guilford Press.

Harris, J. R. (2000). Socialization, personality development, and the child’s

environments: comment on Vandell (2000). Developmental Psychology, 36, 711-723.

Masten, A. S., & Coatsworth, J. D. (1995). Competence, Resilience, and

Psychopathology. In D. Cicchetti & D. J. Cohen (Eds.), Developmental

psychopathology. Vol 2: Risk, disorder, and adaptation (pp. 421-471). New York:

John Wiley & Sons.

Jackson C., & Foshee, V. A. (1998). Violence-related behaviors of

adolescents: Relations with responsive and demanding parenting. Journal of

Adolescent Research, 13, 343-359.

Jenkins, J. M., Rasbash, J., & O’Connor, T. G. (2003). The role of the shared

family context in differential parenting. Developmental Psychology, 39, 99-113.






Lerner, R. M., Rothbaum, F., Boulos, S., & Castellino, D. R. (2002).

Developmental systems perspective on parenting. In M. H. Bornstein (Ed.), Handbook

of parenting. Vol. 2. Handbook of Parenting (pp. 315-344). Mahwah, New Jersey:

Lawrence Erlbaum Associates.







New York: Wiley.

Ostergren, C. S. (1997). Differential utility of temperament-based guidance

materials for parents of infants. Family Relations, 46, 63-71.

Prinzie, P. (2002). Coercive family processes, parent and child personality

characteristics as predictors of antisocial behavior in 4-to-9-year-old children.

Unpublished doctoral dissertation, Katholieke Universiteit Leuven, Belgium.

Putnam, S. P., Sanson, A. V., Rothbart, M. K. (2002). Child temperament and

parenting. In M. Bornstein (Ed.), Handbook of parenting. Vol. 1. Children and


Teerikangas, O. M., Aronen, E. T., Martin, R.P., & Huttonen, M.O. (1998).

Effects of infant temperament and early intervention on the psychiatric symptoms of

adolescents. Journal of the American Academy of Child and Adolescent Psychiatry,

37, 1070-1076.

Raine, A., & Dunkin, J. J. (1990). The genetic and psychophysiological basis

of antisocial behavior : implications for counseling and therapy. Journal of

Counseling and Development, 68, 637-644.

Rowe, D.C. (1990). As the twig is bent? The myth of child-rearing influences

on personality development. Journal of Counseling and Development, 68, 606-611.

Rowe, D. (1997). Genetics, temperament, and personality. In R. Hogan, J.

Johnson, & S. Briggs (Eds.), Handbook of personality psychology (pp. 795-824).

Orlando, FL: Academic Press.

Sheeber, L. B., & Johnson, J. H. (1994). Evaluation of a temperament-focused,

parent-training program. Journal of Clinical Child Psychology, 23, 249-259.



behavior in boys. Developmental Psychology, 37, 814-825.

Vandell, D. L. (2000). Parents, peer groups, and other socializing influences.


Widiger, T. A., & Clark, L. A. (2000). Toward DSM-V and the classification

of psychopathology. Psychological Bulletin, 126, 946-963.

Samenvatting 233

SAMENVATTING

Dit proefschrift bestaat uit vijf hoofdstukken die kaderen binnen de

vraagstelling waarom sommige kinderen emotionele of gedragsproblemen vertonen

en andere niet. Er wordt onderzocht in welke mate dit kan verklaard worden door

individuele verschillen, met name de persoonlijkheid van het kind en de

persoonlijkheid van de ouder(s), en omgevingsinvloeden, met name het

opvoedingsgedrag van de ouders. Het algemeen theoretisch kader waarop het

proefschrift is gebaseerd, is het ecologisch of contextueel perspectief, waarbij men

aanneemt dat kindgedrag bepaald wordt door een complex netwerk van factoren. Met

de studies in onze verhandeling pogen we een antwoord te vinden op een aantal

pertinente vragen: heeft ouderlijk gedrag enkel een effect op kinderen met bepaalde

eigenschappen; kunnen bepaalde persoonlijkheidseigenschappen dienen als

beschermende of risicofactoren in de aanwezigheid van eerder onaangepast of positief

ouderlijk gedrag; is het mogelijk om bepaalde types van kinderen te identificeren die

min of meer kwetsbaar zijn voor het ontwikkelen van probleemgedrag; leiden

verschillende ouderlijke gedragingen ook tot verschillende uitkomsten bij kinderen;

zijn ouderlijk gedrag en persoonlijkheidskenmerken van kinderen op dezelfde manier

gerelateerd aan emotionele als aan gedragsproblemen; spelen persoonlijkheid en

ouderlijk gedrag een gelijkaardige rol voor klinische en niet-klinische kinderen; wat is

het aandeel van de ouderlijke persoonlijkheid in het netwerk van relaties tussen de

persoonlijkheid van het kind, ouderlijk gedrag en probleemgedrag van het kind?

We focussen in dit proefschrift op twee brede leeftijdsgroepen, met name

schoolkinderen en adolescenten. Subjecten in de studies zijn ouders en kinderen, voor

de meerderheid ad random geselecteerd uit de algemene populatie. Eén van de studies

onderzoekt ook kinderen die in behandeling zijn voor psychologische problemen.

De afhankelijke variabele in de studies is probleemgedrag bij kinderen,

opgedeeld in gedrags- of externaliserende problemen, en emotionele of

internaliserende problemen. Om deze variabele te meten is gebruik gemaakt van het

Achenbach System of Empirical Based Assessment, met vragenlijsten voor ouders,

leerkrachten en jongeren (Verhulst, Van der Ende, & Koot, 1996; 1997a; 1997b). De

onafhankelijke variabelen zijn geoperationaliseerd aan de hand van vragenlijsten die

234 Samenvatting

adaptief gedrag meten. De vragenlijsten hebben een theoretische basis en goede

psychometrische eigenschappen. Om persoonlijkheid bij kind en ouders te meten is

gebruik gemaakt van het robuuste Vijf-factorenmodel voor persoonlijkheid. Bij het

kind zijn de vijf domeinen Emotionele stabiliteit, Extraversie, Openheid,

Welwillendheid en Consciëntieusheid gemeten met de Hiërarchische Persoonlijk-

heidsvragenlijst voor Kinderen (Mervielde & De Fruyt, 1999). Bij de ouders zijn de

overeenkomstige vijf domeinen Neuroticisme, Extraversie, Openheid, Altruïsme en

Consciëntieusheid gemeten met de NEO PI-R (Hoekstra, Ormel & De Fruyt, 1996).

Om ouderlijk gedrag te meten, is in het kader van dit proefschrift een vragenlijst

ontwikkeld, de Schaal voor Ouderlijk Gedrag (SOG), waarvan de bespreking deel

uitmaakt van het eerste hoofdstuk.

Hoofdstuk 1

Psychometrische eigenschappen van De Schaal voor Ouderlijk Gedrag

De Schaal voor Ouderlijk Gedrag (SOG) is gebaseerd op het sociaal

interactioneel denkkader van Patterson en collega’s. Volgens hun ‘coercion theory’ is

onaangepast gedrag van kinderen het gevolg van een opeenvolgende reeks van

interacties, waarbij kinderen leren dat onaangepast gedrag effectief kan zijn om

onaangename situaties te ontwijken, doordat ouders met ineffectief ouderlijk gedrag

reageren op dit onaangepast gedrag. In een macromodel stellen zij dat ouderlijke

vaardigheden mediëren in de relatie tussen enerzijds contextuele factoren zoals socio-

economische omstandigheden, de buurt, familiale context en anderzijds de aanpassing

van het kind. Zij leggen hierbij de nadruk op vijf ouderlijke vaardigheden: ouderlijke

betrokkenheid, monitoring (supervisie van het kind), discipline (ook benoemd als

grenzen stellen), positieve bekrachtiging, en probleem-oplossen.

Een inventaris van de nederlandstalige vragenlijsten met betrekking tot gezin

en opvoeding (cf. Langemijer, Pijnenburg & Veerman, 1997), leerde ons dat er geen

vragenlijsten bestaan met de meetpretentie die wij beogen. Tenslotte liet een

screening van de internationale vragenlijsten verschillende beperkingen zien op

psychometrisch en praktisch vlak.

De resultaten van een pilootversie van de SOG (Van Leeuwen, 1999) toonde

aan dat het niet mogelijk was om de vijf Patterson constructen structureel te

onderscheiden binnen de initiële set van items. Vier van de vijf constructen zijn

Samenvatting 235

behouden in een nieuwe versie: ouderlijke betrokkenheid, monitoring, positieve

bekrachtiging en probleem-oplossen. Omdat factoranalyse aangaf dat het construct

disciplinering multidimensioneel was, is dit gedifferentieerd in: regels stellen, straffen

van ongewenst gedrag, inconsistent straffen, hard straffen en negeren. Al deze schalen

zijn gerelateerd aan de Sociale Interactie Theorie. Verder is de vragenlijst aangevuld

met items die verwijzen naar het aanleren van autonoom gedrag, wat vanuit

ontwikkelingspsychologisch oogpunt een interessant concept is. De gereviseerde SOG

bestaat uit 55 items waarbij de frequentie van een bepaald ouderlijk gedrag wordt

beoordeeld op een vijf-puntenschaal. Ouders kunnen hun eigen gedrag beoordelen en

er is een versie voor kinderen met dezelfde items waarbij afzonderlijk het gedrag van

vader en moeder beoordeeld wordt.

Door middel van twee studies zijn de psychometrische eigenschappen van de

SOG geëvalueerd. De eerste studie onderzoekt in een ad random samengestelde

steekproef van 600 gezinnen met kinderen tussen 7 en 15 jaar: (a) de factoriële

structuur van de SOG, (b) de interne consistentie van de subschalen, (c) de correlaties

tussen beoordelingen van ouders en kinderen, en (d) de constructvaliditeit door

correlaties te berekenen tussen de subschalen van de SOG en enerzijds het ervaren

van stress in de opvoeding door de ouders, gemeten met de Nijmeegse Ouderlijke

Stress Index (NOSI; de Brock, Vermulst, Gerris & Abidin, 1992), en anderzijds

probleemgedrag bij het kind, gemeten met de Gedragsvragenlijst voor Kinderen

(CBCL; Verhulst, Van der Ende & Koot, 1996). De tweede studie onderzoekt de

repliceerbaarheid van de resultaten uit de eerste studie binnen een onafhankelijke

steekproef van 175 gezinnen met gegevens over twee kinderen.

De factoranalyses leverden negen schalen op: Autonomie, Positief ouderlijk

gedrag (met items refererend naar positieve betrokkenheid, sociaal belonen en

probleem-oplossen), Straffen, Hard straffen, Inconsequent straffen, Negeren,

Monitoring, Regels stellen en Materieel belonen. Een confirmatorische factoranalyse

toonde aan dat de negen componenten repliceerbaar zijn over verschillende

beoordelaars (ouders en kinderen). Een tweede-orde factoranalyse gaf twee dimensies

aan die benoemd zijn als (a) Positief ouderlijk gedrag, gebaseerd op de schalen

Positief ouderlijk gedrag, Regels stellen en Autonomie, en (b) Negatieve controle,

gebaseerd op de schalen Straffen, Hard straffen en Negeren. De interne consistentie

was matig tot goed voor het merendeel van de schalen. De correlaties tussen

beoordelingen van ouders en kinderen waren positief en significant. De verwachte

236 Samenvatting

relaties tussen de schalen van de SOG en de criteriumvariabelen waren aanwezig.

Meer opvoedingsstress was gerelateerd aan minder positief ouderlijk gedrag, regels en

autonomie, en aan meer straffen, inconsequent straffen, hard straffen en het niet

optreden bij ongewenst gedrag. Wat de relaties met probleemgedrag van het kind

betreft, waren er vooral significante correlaties met externaliserend probleemgedrag.

Minder positief ouderlijk gedrag en regels en meer straffen, inconsequent straffen,

hard straffen en negeren van ongewenst gedrag waren geassocieerd met meer

externaliserend gedrag. Enkel de verwachte negatieve correlatie tussen

probleemgedrag en monitoring werd niet gevonden.

Hoofdstuk 2

Een longitudinale studie naar de bruikbaarheid van de

persoonlijkheidstypologie ‘overcontrollers, undercontrollers en resilients’

in de predictie van probleemgedrag bij kinderen en adolescenten

Het beschrijven van individuele karakteristieken bij kinderen kan vanuit twee

perspectieven: de variabele- en de typebenadering. De variabelebenadering bestudeert

categorieën van variabelen zoals de dimensies van het Vijf-factorenmodel (FFM) voor

persoonlijkheid. Hierbij onderzoekt men de repliceerbaarheid van de structuur van de

variabelen over personen binnen verschillende steekproeven. Bij de typebenadering

trachtt men types binnen een groep subjecten te identificeren, dit zijn clusters van

individuen met dezelfde kenmerken of persoonlijkheidspatronen. Via deze benadering

is het mogelijk om verschillende persoonlijkheidsdimensies te combineren binnen één

persoon, en krijgt men een categoriale indeling. Deze studie onderzoekt de

repliceerbaarheid van drie persoonlijkheidstypes, die meermaals empirisch zijn

geïdentificeerd volgens een analytische standaardprocedure, een combinatie van

clusteranalyses: de ‘resilients’ of veerkrachtige kinderen, die gemiddeld scoren op de

vijf domeinen, de ‘undercontrollers’ of kinderen die onder het gemiddelde scoren op

welwillendheid en consciëntieusheid, en de ‘overcontrollers’ of kinderen die laag

scoren op emotionele stabiliteit en extraversie.

De studie sluit aan bij de huidige trend in onderzoek naar

persoonlijkheidstypes, maar is vernieuwend in een aantal opzichten: (a) er wordt

rekening gehouden met verschillende informanten (beoordelingen van ouders en

adolescenten); (b) er worden twee verschillende maten gebruikt om persoonlijkheid te

Samenvatting 237

meten, namelijk de HiPIC (Mervielde & De Fruyt, 1999), afgenomen bij ouders, en een

verkorte versie van Goldberg’s adjectievenlijst, de Questionnaire Big Five (Gerris et

al., 1995), afgenomen bij adolescenten; (c) het onderzoeksdesign omvat follow-up

data met een interval van drie jaar tussen de twee meetmomenten, waardoor twee

verschillende leeftijdsgroepen – kinderen en adolescenten – worden onderzocht.

Volgende onderzoeksdoelen zijn met deze studie beoogd: (a) de repliceerbaarheid van

de types, (b) de continuïteit van de types over twee meetmomenten met een interval

van drie jaar, (c) de stabiliteit van het individueel lidmaatschap tot één van de types

over de twee meetmomenten en (d) de relatieve predictieve waarde van types versus

variabelen voor probleemgedrag.

De drie persoonlijkheidstypes konden gerepliceerd worden bij

zelfbeoordelingen van adolescenten op een lijst met adjectieven, maar enkel de types

undercontrolled en resilient konden afgeleid worden uit ouderbeoordelingen. De derde

cluster groepeerde zowel het overcontrolled type als kinderen die gemiddeld scoren

op de vijf persoonlijkheidsdimensies. De continuïteit van de types over de twee

meetmomenten was matig: op beide meetmomenten konden enkel de types

undercontrollers en resilients gerepliceerd worden. Verder toonde de analyses een

lage stabiliteit van het individueel behoren tot een bepaald persoonlijkheidstype bij de

overgang van kindertijd naar adolescentie. Wat betreft de predictieve validiteit van de

types kon worden aangetoond dat de types differentieel geassocieerd zijn met de vijf

persoonlijkheidsdomeinen alsook met internaliserend en externaliserend

probleemgedrag van kinderen. Overcontrollers waren gekenmerkt door een hoge mate

van neuroticisme en internaliserend gedrag, undercontrollers door weinig

consciëntieusheid en welwillendheid en veel externaliserend gedrag. Resilients

werden altijd positief beoordeeld op sociaal wenselijke persoonlijkheidsdimensies en

hadden gemiddelde scores voor probleemgedrag. De studie toont echter ook aan dat

de drie types geen betere voorspellers waren van probleemgedrag ten opzichte van de

vijf dimensies. De resultaten impliceren dat de keuze tussen een variabele- of een

typebenadering afhangt van de doelstelling. Voor empirisch onderzoek lijkt de

variabelebenadering voordelig omdat men informatie behoudt over de interindividuele

variatie en niet onderhevig is aan range-restrictie. Voor praktisch gebruik lijkt de

typebenadering handig, zeker wanneer men beslissingen moet nemen over individuen

en men de informatie afkomstig van verschillende variabelen tegenover mekaar moet

afwegen.

238 Samenvatting

Hoofdstuk 3

Persoonlijkheid van het kind en ouderlijk gedrag als moderatoren van

probleemgedrag: een variabele- en een typebenadering

Relatief weinig studies hebben de gecombineerde of interactie-effecten van

ouderlijk gedrag en persoonlijkheid onderzocht als determinanten voor internaliserend

en externaliserend probleemgedrag. In dit manuscript zijn interacties tussen de

persoonlijkheid van het kind en ouderlijk gedrag onderzocht in twee studies. Om

voldoende power te garanderen voor de detectie van interactie-effecten via

hiërarchische multiple regressie-analyse, is gebruik gemaakt van een vrij grote

steekproef (N = 600) en betrouwbare maten. Om uit te sluiten dat interacties enkel op

basis van het toeval gevonden worden, is er gekeken naar de stabiliteit van de effecten

over beoordelaars, met name ouders en kinderen, en over tijd, via een cross-sectioneel

en longitudinaal design.

De eerste studie onderzocht interactie-effecten vanuit een variabelebenadering,

wat inhoudt dat zowel ouderlijk gedrag als persoonlijkheid als (continue)

kwantitatieve variabelen in de analyses zijn ingevoerd. De analyses toonden

interacties tussen ouderlijke negatieve controle en de persoonlijkheidsdomeinen

welwillendheid en consciëntieusheid in de predictie van externaliserend gedrag. Deze

interacties waren repliceerbaar over beoordelaars en over tijd. Bovendien bleken de

interacties gemeten op tijdstip 1 ook bij te dragen tot externaliserend gedrag op

meetmoment 2. De resultaten suggereren dat negatieve controle een risicofactor is

voor externaliserend gedrag, en meer bepaald voor kinderen die weinig of gemiddeld

welwillend of consciëntieus zijn. Een hoge mate van welwillendheid of

consciëntieusheid lijkt een protectieve factor te zijn, want deze kinderen vertonen

weinig externaliserend gedrag, zelfs als de ouders veel negatieve controle uitoefenen.

Predictieve interacties voor externaliserend gedrag werden ook gevonden tussen

positief ouderlijk gedrag en welwillendheid. Weinig welwillende kinderen vertoonden

meer externaliserend gedrag wanneer ouders weinig ondersteunend waren. Kinderen

die beoordeeld werden als zeer welwillend toonden weinig probleemgedrag, zelfs bij

geringe ouderlijke ondersteuning. De interacties met positief ouderlijk gedrag waren

echter minder stabiel over beoordelaars en tijd dan de interacties met negatieve

controle. Bij de analyses van internaliserend gedrag waren geen interacties significant.

De resultaten suggereren vooral onafhankelijke effecten van persoonlijkheid en

Samenvatting 239

ouderlijk gedrag. Een lage mate van vooral emotionele stabiliteit en extraversie,

weinig positief ouderlijk gedrag en uitgesproken negatieve controle waren

geassocieerd met meer internaliserend gedrag.

De tweede studie vertrok vanuit een typebenadering voor wat de

persoonlijkheid van het kind betreft. Kinderen konden toegewezen worden aan drie

persoonlijkheidstypes, de ‘resilients’, de ‘undercontrollers’ en de ‘overcontrollers’.

Omdat tot op heden weinig evidentie is gevonden voor prototypische

ouderschapsstijlen is ouderlijk gedrag ingedeeld in de categorieën weinig (- 1 SD

onder het gemiddelde), matig, of veel (+1 SD boven het gemiddelde). De resultaten

waren vergelijkbaar met de eerste studie. De undercontrollers vertoonden meer

externaliserend gedrag dan de resilients, en overcontrollers meer internaliserend

gedrag dan de resilients vooral wanneer ze blootgesteld werden aan veel negatieve

ouderlijke controle. De minder stabiele effecten uit de eerste studie, werden niet

gerepliceerd in de tweede studie.

Concluderend kan gesteld worden dat deze studie het belang aantoont van

interacties tussen de persoonlijkheid van kinderen en ouderlijk gedrag in de predictie

van probleemgedrag. Ouderlijk gedrag kan een impact hebben op probleemgedrag van

het kind, maar enkel bij kinderen met bepaalde persoonlijkheidskenmerken. Een

persoonlijkheid van het kind, gekenmerkt door vooral sociaal aanvaarde kenmerken,

kan een protectieve rol spelen bij minder adequaat ouderlijk gedrag, terwijl kinderen

met sociaal ongewenste eigenschappen risico lopen op probleemgedrag bij inadequaat

ouderlijk gedrag. Verder toonde deze studie het belang aan van zowel de variabele-

als de typebenadering in de voorspelling van probleemgedrag.

Hoofdstuk 4

Persoonlijkheid van het kind en ouderlijk gedrag als interagerende voorspellers

van emotionele en gedragsproblemen bij kinderen

in een klinische en niet-klinische steekproef

Deze studie is een aanvulling op de vorige studie door de steekproef uit de

algemene populatie (de “niet-klinische” groep) uit te breiden met een steekproef van

kinderen die in ambulante of residentiële behandeling zijn voor emotionele of

gedragsproblemen (de “klinische groep”). Ten eerste wordt onderzocht of er

verschillen zijn in gemiddelden voor de twee groepen op vlak van persoonlijkheid,

240 Samenvatting

opvoedingsgedrag en probleemgedrag. Ten tweede wordt nagegaan of interacties

tussen persoonlijkheid van het kind en ouderlijk gedrag probleemgedrag verklaren.

Immers, de inclusie van een klinische groep kinderen verhoogt statistisch gezien de

power om interactie-effecten te detecteren. Tenslotte wordt gekeken of de interactie-

effecten kunnen gegeneraliseerd worden over de klinische en de niet-klinische groep.

Dit onderzoeksdoel is belangrijk binnen het debat dat klinisch psychologen en

psychiaters voeren over de vraag of het verschil tussen normaal en abnormaal gedrag

als kwalitatief of kwantitatief moet gepercipieerd worden. In onderzoek naar

persoonlijkheidsstoornissen gaat men uit van een continuïteits- of spectrummodel,

waarbij geponeerd wordt dat verschillen tussen normale en klinische steekproeven

bepaald worden door verschillen in gemiddelden op relevante domeinen van het

functioneren. In de studie wordt deze visie geëxponeerd naar de hypothese dat tussen

klinische en niet-klinische groepen verschillen tussen gemiddelden waarneembaar

zijn, maar dat gelijkaardige relaties tussen persoonlijkheid, opvoeding en

probleemgedrag bij de twee groepen kunnen geobserveerd worden.

De resultaten toonden aan dat er verschillen in gemiddelden zijn tussen de

klinische en niet-klinische groep voor wat betreft probleemgedrag van het kind, de

vijf persoonlijkheidsvariabelen, en opvoedingsgedrag van de moeder. De kinderen uit

de niet-klinische groep hadden gemiddeld lagere scores op internaliserend en

externaliserend probleemgedrag en gemiddeld hogere scores op welwillendheid,

consciëntieusheid, emotionele stabiliteit, extraversie en vindingrijkheid, in

vergelijking met de klinische groep kinderen. Moeders rapporteerden gemiddeld meer

positief ouderlijk opvoedingsgedrag en minder negatieve controle in de niet-klinische

groep.

De persoonlijkheid van het kind verklaarde probleemgedrag van het kind in

beide groepen, maar de sterkte van de relatie verschilde enigszins voor de twee

groepen en bleek ietwat sterker te zijn in de klinische groep, vooral voor

internaliserend gedrag. Dit verschil is echter eerder kwantitatief dan kwalitatief van

aard. Wat de relatie tussen probleemgedrag en opvoedingsgedrag betreft, toonden de

analyses dat negatieve controle samenhing met meer probleemgedrag en dat die

samenhang iets sterker was in de klinische groep. Voor positief ouderlijk gedrag bleek

de relatie verschillend voor de klinische en de niet-klinische groep: weinig positief

ouderlijk gedrag was geassocieerd met zowel internaliserend als externaliserend

gedrag, maar enkel in de niet-klinische groep. Verder bleken ook interactie-effecten

Samenvatting 241

voor beide groepen probleemgedrag te verklaren. Kinderen die gekenmerkt werden

door een lage welwillendheid of consciëntieusheid vertoonden probleemgedrag bij

blootstelling aan negatief ouderlijk gedrag, vooral negatieve controle. Voor drie van

de twintig onderzochte interacties bleken de interacties voor klinische en niet-

klinische groepen te verschillen, in die zin dat bepaalde interacties wel in de ene groep

maar niet in de andere groep voorkwamen. Deze verschillen waren echter niet

consistent over de groepen en over persoonlijkheidsdimensies. Samengevat kunnen

we besluiten dat onze resultaten de spectrumhypothese ondersteunen. Verschillen

tussen een klinische en niet-klinische groep zijn eerder kwantitatief van aard, en

gelijkaardige processen liggen aan de basis van probleemgedrag in de klinische en

niet-klinische groep.

Hoofdstuk 5

Ouderlijk gedrag en persoonlijkheid van ouder en kind als predictoren van

internaliserend en externaliserend gedrag

Het doel van deze studie is de gelijktijdige effecten te onderzoeken van

ouderlijk gedrag, persoonlijkheid van het kind en persoonlijkheid van de ouder binnen

een padmodel. Contextuele of ecologische modellen (cf. Belsky, 1984) erkennen dat

de persoonlijkheid van de ouder een rol speelt binnen de ontwikkeling van het kind,

maar er zijn weinig studies die effectief persoonlijkheid van de ouders als verklarende

variabele voor probleemgedrag bij het kind hebben onderzocht. Binnen deze studie

worden zowel persoonlijkheid van de ouder als persoonlijkheid van het kind volgens

het Vijf-factorenmodel geoperationaliseerd. Twee alternatieve padmodellen worden

onderzocht, afzonderlijk voor moeders en vaders, en afzonderlijk voor externaliserend

en internaliserend probleemgedrag. Het eerste model omvat paden vanaf elk van de

vijf ouderlijke persoonlijkheidsvariabelen en vanaf elk van de vijf kind

persoonlijkheidskenmerken naar het probleemgedrag van het kind. Daarnaast zijn er

ook indirecte paden van de persoonlijkheid van kind en ouder naar probleemgedrag

van het kind via de ouderschapsdimensies negatieve controle en positief ouderlijk

gedrag. Het tweede padmodel onderzoekt de directe effecten van de vijf ouderlijke

persoonlijkheidsdomeinen en de opvoedingsdimensies op het probleemgedrag van het

kind alsook de indirecte effecten van deze domeinen via de vijf

persoonlijkheidsdimensies van het kind.

242 Samenvatting

Het effect van ouderlijke persoonlijkheid op ouderlijk gedrag is onderzocht

vanuit de hypothese dat ouderlijk gedragingen stabiele persoonlijkheidskenmerken

van ouders kunnen reflecteren. De resultaten van het eerste model toonden aan dat een

hogere mate van consciëntieusheid en openheid bij moeders en vaders samenhing met

meer positief ouderlijk gedrag. Meer neuroticisme, extraversie en minder openheid bij

moeders was gerelateerd aan meer negatieve controle, terwijl deze dimensie bij vaders

voorspeld werd door minder altruïsme en openheid. Ook meer consciëntieusheid bij

vaders was geassocieerd met meer negatieve controle. Ouderlijke persoonlijkheid

droeg slechts gedeeltelijk bij tot ouderlijk gedrag, wat doet vermoeden dat ouderlijk

gedrag door meerdere determinanten kan verklaard worden, zoals bronnen van stress.

Er is geen evidentie gevonden voor een direct verband tussen ouderlijke

persoonlijkheid en externaliserend gedrag van het kind, noch bij moeders, noch bij

vaders. Internaliserend probleemgedrag daarentegen werd wel direct verklaard door

neuroticisme en openheid van de moeder, en door neuroticisme van de vader.

Het tweede model toonde associaties tussen de vergelijkbare moeder en

kinddomeinen consciëntieusheid, neuroticisme/emotionele stabiliteit, extraversie en

openheid. Dezelfde relaties waren aanwezig in de vaderdata, behalve voor het domein

consciëntieusheid. Altruïsme bij ouders bleek niet samen te hangen met

welwillendheid van kinderen. De samenhang tussen persoonlijkheidskenmerken van

ouders en kinderen bleek ook de resultaten van het eerste model, omdat sommige van

de overeenkomstige persoonlijkheidsdomeinen dezelfde effecten sorteerden. Zo

bleken consciëntieusheid van zowel moeder als kind samen te hangen met positief

ouderlijk gedrag van moeders, terwijl extraversie van zowel moeder als kind

geassocieerd waren met negatieve controle. Neuroticisme bij zowel moeders als

vaders en emotionele stabiliteit bij kinderen waren ook beiden gerelateerd aan

internaliserend gedrag van het kind. In het model voor vaders bleken altruïsme van de

vader en welwillendheid van het kind negatieve controle bij vaders te beïnvloeden.

Model 1 toonde slechts een gedeeltelijk mediërend effect van ouderlijk gedrag binnen

de relatie persoonlijkheid kind/ouder en probleemgedrag. In de padmodellen had

ouderlijk gedrag eerder een indirect effect op probleemgedrag, via de effecten van de

persoonlijkheid van ouder en kind op probleemgedrag.

De keuze voor één van beide modellen ligt niet voor de hand, noch vanuit

statistisch oogpunt, noch vanuit theoretisch oogpunt. Beide modellen lijken

Samenvatting 243

aanvaardbaar en suggereren een bidirectionele invloed van ouderlijk gedrag en

persoonlijkheid in de ontwikkeling van probleemgedrag.

Algemeen besluit

In dit algemeen besluit komen we terug op de leidende onderzoeksvragen die

bij de inleiding zijn gesteld en bespreken we de relatieve bijdrage van de

persoonlijkheid van kind en ouders en ouderlijk opvoedingsgedrag tot emotionele en

gedragsproblemen bij het kind.

Ook positief ouderlijk gedrag is in interactie met de persoonlijkheid van het

kind, vooral met welwillendheid, geassocieerd met gedragsproblemen. Deze dimensie

van ouderlijk gedrag verwijst naar ouderlijke betrokkenheid, sociale bekrachtiging,

probleem-oplossen, het stimuleren van autonomie bij het kind en het stellen van

regels. Bij kinderen die weinig welwillend zijn, gaat een gebrek aan positief ouderlijk

Welke persoonlijkheidskenmerken van het kind vervullen de rol van risico- of

protectieve factor in een eerder inadequate opvoedingsomgeving?

De belangrijkste persoonlijkheidskenmerken die interageren met ouderlijk

gedrag zijn welwillendheid en consciëntieusheid. Welwillendheid is het brede

persoonlijkheidsdomein dat egocentrisme, irriteerbaarheid, gehoorzaamheid,

dominantie en altruïsme overkoepelt, terwijl consciëntieusheid verwijst naar de

onderliggende facetten prestatiemotivatie, concentratie, doorzettingsvermogen en

ordelijkheid. Kinderen met lage scores op welwillendheid en consciëntieusheid

vertonen meer externaliserend probleemgedrag wanneer de ouders veel negatieve

controle uitoefenen in de interacties met die kinderen. Negatieve controle verwijst

hier naar straffen, hard straffen en het onbestraft laten van ongewenst gedrag. Een

hoge mate van negatieve controle refereert dus naar inadequate pogingen van ouders

om ongewenst gedrag te beïnvloeden. Kinderen die gekenmerkt worden door een

hoge mate van welwillendheid en consciëntieusheid, vertonen geen

gedragsproblemen, zelfs als ze worden blootgesteld aan negatieve controle. Bijgevolg

kunnen een hoge mate van welwillendheid en consciëntieusheid als beschermend

worden beschouwd tegen negatieve ouderlijke invloeden. De interacties tussen

negatieve controle met welwillendheid en consciëntieusheid zijn repliceerbaar over

informanten en over de tijd.

244 Samenvatting

gedrag samen met meer externaliserend probleemgedrag. Voor kinderen die juist wel

welwillend zijn, houdt het al dan niet blootgesteld worden aan positief ouderlijk

gedrag geen verband met probleemgedrag. De interacties tussen kindkarakteristieken

en positief ouderlijk gedrag zijn wel minder stabiel dan de interacties met negatieve

controle. Voor emotionele problemen zijn de interacties tussen de persoonlijkheid van

het kind en het opvoedingsgedrag van de ouder minder belangrijk als verklarende

factor.

Is het mogelijk om bepaalde types van kinderen te identificeren die meer of minder

probleemgedrag vertonen bij bepaalde opvoedingspraktijken?

Onze resultaten tonen evidentie voor een indeling van kinderen in drie

persoonlijkheidstypes: de ‘resilients’ of veerkrachtige kinderen, die gemiddeld scoren

op de vijf persoonlijkheidsdomeinen, de ‘undercontrollers’, die onder het gemiddelde

scoren op welwillendheid en consciëntieusheid, en de ‘overcontrollers’, die laag

scoren op emotionele stabiliteit en extraversie. Het ‘undercontrolled’ type vertoont

meer externaliserend gedrag wanneer er tevens ouderlijke negatieve controle is, het

‘overcontrolled’ type vertoont meer internaliserend gedrag bij negatieve controle. Bij

veerkrachtige kinderen wordt geen probleemgedrag vastgesteld, zelfs als de

opvoedingsomgeving niet optimaal is, zoals bij een hoge mate van negatieve controle.

Interacties tussen de persoonlijkheidstypes en veel of weinig positief ouderlijk gedrag

hangen niet samen met probleemgedrag bij het kind.

Zijn ouderlijk gedrag en persoonlijkheidskenmerken van kinderen op dezelfde manier

gerelateerd aan emotionele als aan gedragsproblemen?

De resultaten tonen een differentiële samenhang tussen kindkenmerken,

ouderlijk gedrag en externaliserend en internaliserend probleemgedrag. De

persoonlijkheidsdimensies welwillendheid en consciëntieusheid zijn de belangrijkste

voorspellers voor externaliserend gedrag, terwijl emotionele stabiliteit en extraversie

meest samenhangen met internaliserend gedrag. Wat ouderlijk gedrag betreft, blijkt

negatieve ouderlijke controle vooral geassocieerd te zijn met externaliserend gedrag

en in mindere mate met internaliserend gedrag. De effecten van positief ouderlijk

gedrag zijn minder stabiel over studies heen en zijn meestal minder sterk dan

negatieve controle. Interacties tussen ouderlijk gedrag en persoonlijkheid van het kind

zijn vooral geassocieerd met externaliserend gedrag, terwijl internaliserend gedrag

Samenvatting 245

vooral verklaard wordt door onafhankelijke of hoofdeffecten van persoonlijkheid en

ouderlijk gedrag.

Spelen persoonlijkheid en ouderlijk gedrag een gelijkaardige rol voor klinische en

niet-klinische kinderen?

Op basis van gemiddelden kan er een duidelijk onderscheid gemaakt worden

tussen kinderen die wel en niet in behandeling zijn voor psychologische problemen:

de klinische kinderen vertonen hogere scores voor internaliserend en externaliserend

probleemgedrag, worden gekenmerkt door lagere scores op de persoonlijkheids-

domeinen welwillendheid, consciëntieusheid, emotionele stabiliteit, extraversie en

openheid, en zijn blootgesteld aan minder positief ouderlijk gedrag en meer negatieve

controle. Voor beide groepen blijkt persoonlijkheid bij te dragen tot internaliserend en

externaliserend probleemgedrag, al zijn er kleine verschillen in de sterkte van de

relaties. Ook negatieve ouderlijke controle blijkt voor de beide groepen op dezelfde

manier bij te dragen tot probleemgedrag. Een gebrek aan positief ouderlijk gedrag is

enkel voor kinderen in de niet-klinische groep geassocieerd met probleemgedrag.

Interacties tussen ouderlijk gedrag en persoonlijkheid als verklarende variabelen voor

probleemgedrag blijken nauwelijks te verschillen tussen een klinische en niet-

klinische groep. Samengevat kunnen we stellen dat onze resultaten lijken aan te geven

dat dezelfde antecedenten van belang zijn voor het verklaren van probleemgedrag in

een klinische en niet-klinische groep kinderen, en dat de verschillen eerder

kwantitatief dan kwalitatief van aard zijn.

Wat is de bijdrage van persoonlijkheid van ouders binnen het netwerk van relaties

tussen persoonlijkheid van kinderen, ouderlijk gedrag en probleemgedrag?

Persoonlijkheid van ouders verklaart gedeeltelijk probleemgedrag bij kinderen,

al zijn de verbanden eerder indirect dan direct. Geen van de

persoonlijkheidskenmerken van moeder of vader houdt direct verband met

externaliserende problemen. Voor internaliseren is er wel een direct verband met

neuroticisme bij beide ouders en ook met openheid bij moeders. Sommige

persoonlijkheidskenmerken van ouders zijn geassocieerd met ouderlijk

opvoedingsgedrag. Meer positief ouderlijk gedrag hangt samen met meer

consciëntieusheid en openheid. Meer negatieve controle wordt voorspeld door meer

zelfgerapporteerd neuroticisme, en een lagere mate van extraversie en openheid bij

246 Samenvatting

moeders, evenals door een mindere mate van openheid, altruïsme en meer

consciëntieusheid bij vaders. De resultaten tonen een indirect verband aan tussen

persoonlijkheid en probleemgedrag. Negatieve ouderlijke controle wordt zowel bij

vaders als moeders verklaard door bepaalde persoonlijkheidskarakteristieken, maar

beïnvloedt zelf ook probleemgedrag.

Klinische implicaties

De onderzoeksresultaten suggereren dat assessment van zowel ouderlijk

gedrag als persoonlijkheid van het kind aan te bevelen is als kinderen voor gedrags- of

emotionele problemen worden doorverwezen. Dit kan het inzicht in de interacties

tussen ouders en kind vergroten. De meetinstrumenten die in het proefschrift zijn

gebruikt, zijn geschikt voor dit doel. De Schaal voor Ouderlijk Gedrag (SOG) is

eenvoudig om af te nemen, laat toe dat zowel ouders als kinderen ouderlijk gedrag

beoordelen, en geeft een indicatie over sterkten en zwakten van ouderlijk

opvoedingsgedrag. De Hiërarchische Persoonlijkheidsvragenlijst (HiPIC; Mervielde

& De Fruyt, 1999) is een instrument met goede psychometrische eigenschappen. Voor

beide instrumenten zijn normen beschikbaar die toelaten om normaal en afwijkend

gedrag vast te stellen.

Sommige praktijkmensen staan huiverachtig tegenover de evidentie dat

kindkarakteristieken een rol spelen bij probleemgedrag van kinderen. Omdat

temperament en persoonlijkheid worden beschouwd als relatief stabiel over tijd neemt

men aan dat ook probleemgedrag stabiel is en bijgevolg moeilijk behandelbaar.

Andere hulpverleners creëren dan weer onrealistische verwachtingen, door geen

rekening te houden met bepaalde individuele eigenschappen. We hopen met onze

resultaten vooral een realistische visie op probleemgedrag aan te moedigen in plaats

van een pessimistische of deterministische visie, of anderzijds een te optimistische

visie. Door therapie zal men bepaalde eigenschappen van een kind niet fundamenteel

kunnen veranderen. Toch kan men gewenst gedrag doen toenemen door het geven van

specifieke informatie. Ook kan beïnvloeding door de omgeving voorkomen dat

bepaalde predisposities meer manifest worden. Wanneer er een disharmonie is tussen

kenmerken van het kind en ouderlijk gedrag, kan men ouders instrueren om meer

effectieve ouderlijke strategieën te gebruiken om probleemgedrag te voorkomen of te

Samenvatting 247

verminderen. De resultaten suggereren bijvoorbeeld dat veel en hard straffen bij een

kind dat weinig welwillend of consciëntieus is eerder schadelijk is. Een alternatieve

aanpak lijkt aangewezen, zoals bijvoorbeeld het belonen van prosociaal gedrag.

Het onderzoek ondersteunt verder de visie dat ouders niet moeten

verantwoordelijk worden gesteld voor de gedragsmoeilijkheden bij hun kind, of

anderzijds te veel krediet moeten krijgen voor de afwezigheid van probleemgedrag.

Indien men dit wel doet, negeert men de interacties tussen ouderlijk gedrag en de

persoonlijkheid van het kind. Therapeuten die de schuldvraag bij ouders omtrent het

probleemgedrag van hun kind wegnemen, zullen die ouders als co-therapeuten kunnen

gebruiken in het therapeutisch proces. Er zijn reeds therapeutische programma’s

ontwikkeld waarbij men als educatieve component uitleg verschaft omtrent

temperament en persoonlijkheid van het kind, wat vervolgens wordt aangevuld met

een training in specifieke ouderlijke vaardigheden, aanleunend bij de individuele

karakteristieken van het kind.

Concluderend kan gesteld worden dat het onderzoek een bijdrage biedt op

zowel theoretisch, empirisch als praktisch vlak. Op theoretisch vlak zijn een aantal

actuele en/of vernieuwende ideeën in het proefschrift opgenomen, zoals het

bestuderen van interacties tussen omgevingsfactoren, met name de opvoeding, en

individuele karakteristieken, met name de persoonlijkheid, in het verklaren van

probleemgedrag. Daarbij zijn zowel een type- als een variabelebenadering gebruikt.

Verder zijn assumpties getest van een spectrummodel voor individuele verschillen.

Op empirisch vlak is evidentie geleverd voor het belang van de studie van interacties

van persoonlijkheid en ouderlijk gedrag als verklaring voor probleemgedrag bij

kinderen. Daarbij zijn enkele prominente persoonlijkheidsdomeinen en ouderlijke

gedragingen naar voor gekomen als voorspellers van probleemgedrag bij kinderen. Op

praktisch vlak kunnen onze bevindingen richtlijnen geven voor de ontwikkeling en

verbetering van therapeutische programma’s gericht op preventie en verminderen van

probleemgedrag bij kinderen en adolescenten.

Verder onderzoek is echter zinvol en kan focussen op ondermeer: assessment

van ouderlijk gedrag, geslachts- en leeftijdsverschillen voor (interactie-)effecten van

persoonlijkheid en ouderlijk gedrag, verder longitudinaal onderzoek, analyses omtrent

verschillende kinderen binnen een gezin en de inclusie van variabelen omtrent de

bredere familiale context.

248 Samenvatting

Gerris, J. R. M., Houtmans, M. J. M., Kwaaitaal-Roosen, E. M. G., de

Schipper, J. C., Vermulst, A. A., & Janssens, J. M. A. M. (1998). Parents,

Adolescents and Young Adults in Dutch Families: a longitudinal study. Nijmegen:

Institute of Family Studies.

Hoekstra, H. A., Ormel, J., & De Fruyt, F. (1996). NEO

Persoonlijkheidsvragenlijsten: NEO PI R en NEO-FFI. Lisse: Swets & Zeitlinger.

Langemeijer, K., Pijnenburg, H., & Veerman, J. W. (1997). Standaardisatie

van de gezinsdiagnostiek. Een inventarisatie van Nederlandstalige

gezinsdiagnostische instrumenten. Nijmegen: Stichting de Waarden, afdeling

Onderzoek en Ontwikkeling.

Van Leeuwen, K. (1999). Het meten van opvoeding met de Schaal voor

Ouderlijk Gedrag. Diagnostiek-Wijzer, 3, 151-170.




Verhulst, F. C., Van der Ende, J., & Koot, H. M. (1997a). Handleiding voor de

Teacher’s Report Form. [Manual of the TRF.] Erasmus Universiteit Rotterdam,


Referenties



De Brock, A. J. L. L., Vermulst, A. A., Gerris, J. R. M., & Abidin, R. R.

(1992). NOSI: Nijmeegse Ouderlijke Stress Index. Lisse: Swets & Zeitlinger.




Verhulst, F. C., Van der Ende, J., & Koot, H. M. (1997b). Handleiding voor de

Youth Self Report. [Manual of the YSR.] Erasmus Universiteit Rotterdam, Afdeling
