Parenting and psychopathology: Differences in family members' perceptions of parental rearing styles

12
Pergamon Person. in&id. Diff: Vol. 23, No. 2, pp. 211-282, 1997 c(‘ 1997 Elsevier Science Ltd. All rights reserved Printed in Great Britain PII: SO191-8869(97)00033-O 0191-8869/97 $17.00+0.00 PARENTING AND PSYCHOPATHOLOGY: DIFFERENCES IN FAMILY MEMBERS’ PERCEPTIONS OF PARENTAL REARING STYLES Coby Gerlsma,‘* Tom A. B. Snijders2 Marijtje A. J. van Duijn2 and Paul M. G. Emmelkamp’ ‘Department of Clinical Psychology, University of Groningen, P.O. Box 30.001, 9700 RB Groningen and 2Department of Statistics and Measurement Theory, University of Groningen, Groningen, The Netherlands (Recrioed 19 October 1996) Summary-Psychiatric patients generally report more adverse recollections of their parents’ rearing behav- iour than individuals from the general community. It is, however, as yet unclear whether we can infer from this finding that the families of psychiatric patients differ from the families of healthy controls, that is, whether patients’ adverse views are shared by their family members. This issue bears on the construct validity of reports about parental rearing styles: should these reports be interpreted to reflect characteristics of the family, of the parent, of the parentchild relationship, or of the individual providing the reports? In this study, patterns of agreement and variability within families with regard to recalled parental behaviour were analysed in order to examine this aspect of the validity of parental representations. We examined whether families of psychiatric patients report less favourable parenting styles than families of healthy controls. Furthermore, we examined the level of agreement between all family members participating in the study, between the two members reporting on the same parentxhild relationship, between parents, and between siblings. Finally, we examined what factors might be accountable for differences of opinion between family members. Results suggested that perceptions of parental rearing styles are primarily tales by individuals, and to a much smaller extent tales about families, parents or relationships. The implications of these findings for research with regard to the relationship between parental rearing behaviour and adult psychopathology are discussed. 10 1997 Elsevier Science Ltd INTRODUCTION Families are often thought to exert a powerful influence on psychosocial development, for better or for worse. Among the many dimensions around which family life revolves, parental rearing behav- iour is traditionally regarded as one of the key issues to be considered when the offspring’s devel- opment deviates from what is expected or preferred. Studies investigating the relationship between parental rearing styles and adult psychopathology generally have to rely on self-report data, i.e. the individual’s subjective and usually retrospective perception of parental behaviour. Compared to individuals from the general community, psychiatric patients usually recall their parents’ behaviour as having been less affectionate, and more rejecting and overprotecting (e.g. Blatt & Homann, 1992; Burbach & Borduin, 1986; Gerlsma, Emmelkamp & Arrindell, 1990; Parker, 1988). These differences appeared also in the comparison of recovered patients with healthy controls (reviewed in Gerlsma rt al., 1990). Furthermore, memories of par- enting proved to be highly stable in time, even when the individual experienced substantial (that is, clinically relevant) changes in well-being and negative affect between measurement occasions (reviewed in Brewin, Andrews & Gotlib, 1993; Gerlsma, 1994). While the difference in recalled parental rearing styles between psychiatric patients and their healthy counterparts appears to be consistent and stable, there is no clarity as to its interpretation: to what extent does it allow the conclusion that the parenting styles prevalent in families of patients and healthy controls differ? Memories of parenting may be interpreted as reflections of actual parental behaviour (e.g. Brewin et al., 1993; Parker, 1989) or as reflections of the ‘phenomenological impact’ of parental behaviour (Arrindell, Emmelkamp, Brilman & Monsma, 1983; Gerlsma et al., 1990; Gerlsma, 1994) a label which aims to convey that memories are based on both true events *To whom all correspondence should be addressed. 271

Transcript of Parenting and psychopathology: Differences in family members' perceptions of parental rearing styles

Pergamon

Person. in&id. Diff: Vol. 23, No. 2, pp. 211-282, 1997 c(‘ 1997 Elsevier Science Ltd. All rights reserved

Printed in Great Britain

PII: SO191-8869(97)00033-O 0191-8869/97 $17.00+0.00

PARENTING AND PSYCHOPATHOLOGY: DIFFERENCES IN FAMILY MEMBERS’ PERCEPTIONS OF PARENTAL

REARING STYLES

Coby Gerlsma,‘* Tom A. B. Snijders2 Marijtje A. J. van Duijn2 and Paul M. G. Emmelkamp’

‘Department of Clinical Psychology, University of Groningen, P.O. Box 30.001, 9700 RB Groningen and 2Department of Statistics and Measurement Theory, University of Groningen, Groningen, The Netherlands

(Recrioed 19 October 1996)

Summary-Psychiatric patients generally report more adverse recollections of their parents’ rearing behav- iour than individuals from the general community. It is, however, as yet unclear whether we can infer from this finding that the families of psychiatric patients differ from the families of healthy controls, that is, whether patients’ adverse views are shared by their family members. This issue bears on the construct validity of reports about parental rearing styles: should these reports be interpreted to reflect characteristics of the family, of the parent, of the parentchild relationship, or of the individual providing the reports? In this study, patterns of agreement and variability within families with regard to recalled parental behaviour were analysed in order to examine this aspect of the validity of parental representations. We examined whether families of psychiatric patients report less favourable parenting styles than families of healthy controls. Furthermore, we examined the level of agreement between all family members participating in the study, between the two members reporting on the same parentxhild relationship, between parents, and between siblings. Finally, we examined what factors might be accountable for differences of opinion between family members. Results suggested that perceptions of parental rearing styles are primarily tales by individuals, and to a much smaller extent tales about families, parents or relationships. The implications of these findings for research with regard to the relationship between parental rearing behaviour and adult psychopathology are discussed. 10 1997 Elsevier Science Ltd

INTRODUCTION

Families are often thought to exert a powerful influence on psychosocial development, for better or

for worse. Among the many dimensions around which family life revolves, parental rearing behav- iour is traditionally regarded as one of the key issues to be considered when the offspring’s devel- opment deviates from what is expected or preferred.

Studies investigating the relationship between parental rearing styles and adult psychopathology generally have to rely on self-report data, i.e. the individual’s subjective and usually retrospective

perception of parental behaviour. Compared to individuals from the general community, psychiatric patients usually recall their parents’ behaviour as having been less affectionate, and more rejecting and overprotecting (e.g. Blatt & Homann, 1992; Burbach & Borduin, 1986; Gerlsma, Emmelkamp

& Arrindell, 1990; Parker, 1988). These differences appeared also in the comparison of recovered patients with healthy controls (reviewed in Gerlsma rt al., 1990). Furthermore, memories of par- enting proved to be highly stable in time, even when the individual experienced substantial (that is,

clinically relevant) changes in well-being and negative affect between measurement occasions (reviewed in Brewin, Andrews & Gotlib, 1993; Gerlsma, 1994).

While the difference in recalled parental rearing styles between psychiatric patients and their

healthy counterparts appears to be consistent and stable, there is no clarity as to its interpretation: to what extent does it allow the conclusion that the parenting styles prevalent in families of patients and healthy controls differ? Memories of parenting may be interpreted as reflections of actual parental behaviour (e.g. Brewin et al., 1993; Parker, 1989) or as reflections of the ‘phenomenological impact’ of parental behaviour (Arrindell, Emmelkamp, Brilman & Monsma, 1983; Gerlsma et al., 1990; Gerlsma, 1994) a label which aims to convey that memories are based on both true events

*To whom all correspondence should be addressed.

271

212 Coby Gerlsma et al.

and the highly individual perception thereof, including reconstructive elaboration (e.g. Greenwald, 1980; Ross, 1989). Furthermore, and irrespective of the controversy described above, it is as yet unclear whether patients’ negative views are shared by their family members, i.e. to what extent retrospective reports of parenting should be interpreted to reflect (actual or perceived) characteristics of the family, of the parents, of the parent-child relationship, or of the individual who provides the reports. This aspect of the validity of recalled parenting has theoretical as well as clinical relevance. High agreement between all family members would provide some support for the notion that families of patients are characterized by more adverse (perceptions of) parenting styles than families of healthy controls. In the case of low agreement (or high agreement only in particular pairs of family members) other possibilities arise. Studies on actual (as opposed to recalled) parental behav- iour within families (e.g. Dunn & Plomin, 1990; Dunn & Plomin, 1991) suggest that parental behaviour may vary from child to child since different children may elicit different feelings and behaviour from their parents. On the other hand, similar parental behaviour may be perceived

differently by different children. From this point of view, differences of opinion about the parental rearing styles prevalent in the family might be expected to be the rule rather than the exception. Examination of the pattern of agreement and of factors contributing to differences of opinion may further elucidate how the consistent link between parental representations and psychopathology

should be interpreted. In previous studies (reviewed in Gerlsma, 1994), correlations for pairwise agreement among

family members have been found to be as low as 0.08 and as high as 0.74, with the majority of findings hovering around 0.50. These correlations predominantly concerned agreement between healthy siblings (co-twins in approximately half of the studies); less is known about the level of agreement between psychiatric patients and their healthy siblings (see, however, Onstad, Ske, Torgensen & Kringlen, 1993; Robins, Schoenberg, Holmes, Benham & Works, 1985) or about the pattern of agreement when various family members are considered.

This study was designed to examine patterns of agreement and variability in the recollections of parenting provided by different family members, that is, father, mother, and two of their (adult) offspring. Children reported their memories of maternal and paternal rearing behaviour on a standardized questionnaire; parents reported memories of their own parenting behaviour towards the two children on an adapted version of the questionnaire. Research questions addressed were: (1) do families of psychiatric patients report less parental Affection and more Overprotection than families from the general community, and (2) to what extent is there agreement among family members, that is, is there agreement between all four members, between those members who report on the same (parentxhild) relationship, between parents, and/or between siblings. These analyses are expected to shed light on the question as to what extent reports of parental rearing style reflect characteristics of families, of particular parent-child relationships, or of particular individuals. Furthermore, we examined (3) to what extent differences of opinion within the family can be accounted for by differences in familial roles (father, mother, child), and position in the family, differences between specific relationships (e.g. same-sex or opposite-sex relationships), and charac- teristics of individual family members (e.g. level of current psychoneurotic complaints, social desirability response style). Families from the general community (‘healthy sample’) were compared with families with one child currently seeking treatment for psychological disorder (‘clinical sample’).

METHOD

Measures

The Egna Minnen Betrdffande Uppfostran (EMBU; Perris, Jacobsson, Lindstrom, Von Knorring & Perris, 1980; Dutch form by Arrindell et al., 1983) was used to assess memories of parental rearing styles. The EMBU is a 64-item questionnaire consisting of four subscales, i.e. Emotional Warmth, Rejection, Overprotection, and Favouring Subject. Subjects are requested to answer all items for both parents separately. The questionnaire’s psychometric qualities (in terms of internal consistency and replicability of factorial structure) have been extensively studied and invariably been found good (reviewed in Gerlsma et al., 1990).

The EMBU was adapted for use by the parents by reformulating those items of each subscale

Parenting and psychopathology 273

which had the highest factor loadings (Arrindell et af., 1983); for instance the original ‘My parents

used to hug me’ became ‘I used to hug [Mary]‘. This EMBU-Parental Form (EMBU-P), with a

total of 38 items for each child concerned, was printed for each family separately, with the names of the relevant child(ren) written out. Items referring to child1 and child2 were mixed and the order per child was randomized, in order to hinder the parent in comparing answers given for each child.

The EMBU-P, as the original version, has four subscales which we labelled Affection (nine items from the original 18-item Emotional Warmth scale; range from 9 to 36) Rejection (12 items from the original 25-item Rejection scale; range from 12 to 48) Protection (12 items from the original 16-item Overprotection scale; range from 12 to 48) and Favouring Subject (all five items from the original Favouring Subject scale; range from 5 to 20). In all scales, higher scores indicate higher levels of the concept’s label (a>0.70). Children’s EMBU scores were based on the same 38 items which constituted the EMBU-P (a>0.70). In order to save space, only the data for Affection and Protection will be reported here; data on the Rejection and Favouring Subject scale are available upon request.

The total score of the Symptom CheckList- (SCL-90; Derogatis, 1975; Dutch form by Arrindell & Ettema, 1986) was used to estimate the amount of psychological and physical complaints

(a = 0.97). As a measure of the tendency to present family affairs in an unrealistically positive or negative

light we used the Social Desirability subscale of a Dutch version (Buurmeyer & Hermans, 1985; Buurmeyer & Hermans, 1988) of the Family Adaptation and Cohesion Scales (FACES I and II) by Olson, Russell, and Sprenkle (1983). This Social Desirability subscale consists of eight items (for example: ‘There were problems in our family’, ‘It happened that family members were dissatisfied with each other’) with a four-point answering format (‘never true’, ‘sometimes true’, ‘usually true’,

and ‘always true’), thus allowing for a minimum score of eight and a maximum of 32 (tix>O.81). Only the two extremes are considered to constitute dysfunctional scores, i.e. a tendency to present

an extremely negative image of the home situation (scores 8815) and a tendency to view the home situation in impossibly perfect terms (scores 2432; cf. Buurmeyer & Hermans, 1988, norm tables). The middle region of social desirability scores is considered to reflect a functional style, representing quite realistic appraisal of the home situation. For the analyses of our data, we constructed two dummy variables, one indicating whether the rater showed a dysfunctional negative presentation of the family or not, and one indicating whether the rater came up with a dysfunctional positive presentation or not.

Subjects andprocedure in the ‘healthy sample’

Subjects. Fifty (healthy community) families, consisting of father, mother, and two children responded to advertisement and pamphlets distributed in various educational institutes, and met our criteria for inclusion (parents living in the proximity of our research department, and par- ticipating children older than 16 years of age). Seventeen additional families of alternative com- position (i.e. one parent and two children, or two parents and one child, etc.) were also included. In families with more than two children the alphabetical order of the children’s first names was used to select which two children were asked to participate in the study. The distinction between these

two children in the study in ‘childl’ and ‘child2’ (or ‘sibling’) is arbitrary. In all, there were 58 fathers, 60 mothers, 60 child1 (35% male, 65% female), and 56 siblings (32.5% male and 67.5% female) in the final sample. Table 1 summarizes the biographical data for all types of raters.

Procedure. The children were mailed questionnaire booklets with the explicit instruction not to discuss the material with their siblings or parents. Most children (90%) no longer lived in the parental home.

The parents were visited at home and answered the EMBU-P, the SCL-90, and the Social Desirability scale in separate rooms, both in the presence of one of the research assistants.

Subjects and procedure in the clinical sample

Subjects. The sample of patients (n = 47; 62% female, 38% male) consisted of 17 outpatients seeking treatment for anxiety disorders (DSM-III-R Axis 1 diagnoses of social anxiety, agoraphobia, panic disorder, and obsessive compulsive disorders) and 30 inpatients referred to the Psychiatric ward of the Academic Hospital with various DSM-III-R Axis 1 and/or Axis 2 diagnoses. Only

274 Coby Gerlsma et al.

Table 1. Summary of biographical data and covariate scores for all raters in the healthy and clinical sample (standard deviations in parentheses)

Healthy sample Clinical sample

Number of family members participating -fathers -mothers --childl/patients -siblings

Mean age -fathers -mothers +hildl/patients -siblings

Sex +hild I /patients -siblings

Left parental home +hildl/patients -siblings

Eldest child --childl/patients -siblings

Youngest child +hildl/patients -siblings

Neuroticism score -fathers -mothers +hildl/patients -siblings

Social Desirability score -fathers -mothers +hildl/patients -siblings

58 21 60 30 60 41 56 40

55 58 54 56 26 29 2s 30

65% female 62% female 61.5% female 57.5% female

90% 90%

43% 13%

19% 38% 28% 18%

115 (22) 126 (30) 120 (30) 135 (39) 126 (34) 214 (62) 118 (30) 125 (32)

21 (4) 21 (4) 19 (5) 20 (4)

80% 15%

33% 15%

19 (4) 19 (4) 17 (6) 18 (5)

patients were included of whom at least one family member agreed to participate in the study. Furthermore, patients were excluded if they were older than 45 years of age, if they were currently experiencing psychotic symptoms, and if they were mentally disabled and could not fill out the various questionnaires.

In the clinical sample, 21 fathers, 30 mothers, and 40 siblings (57.5% female, 42.5% male) agreed

to participate. In Table 1 biographical data are summarized for all types of raters. Procedure. Inpatients were contacted about 2-3 weeks after admission, thus allowing for crisis

intervention and, in general, for some period of acclimatization. Inpatients answered the ques- tionnaires in a separate room at the Department of Clinical Psychology, in the presence of a research assistant. Outpatients were asked to fill out the questionnaires before treatment started.

We asked the patients concerned for permission to contact their family (father, mother, one sibling). If permission was granted, family members were informed of the research programme and

asked to participate.* Subsequently, family members filled out questionnaire booklets at their home address. It should be noted, however, that quite a few patients (20 out of 67) refused to let us contact their family at all. One of the reasons brought forward for this refusal was the disturbed relationship between them.

Statistical analyses

In order to allow for comparison with former studies on pairwise agreement in perceived PRS, Pearson’s Product-Moment correlations were computed. In order to examine patterns of agreement and variability among the four family members’ reports, hierarchical linear modelling (HLM; Bryk

*All patients participating in the clinical sample read and signed an informed consent letter

Parenting and psychopathology 215

& Raudenbush, 1992; Goldstein, 1995) was chosen as the means of analysis, to account for the nested structure of the data (raters within relationships within families). A full description and technical details of this particular HLM application can be found in Snijders and Kenny (submitted); a summary will be provided below. The analyses were done separately for the two dependent variables, parental Affection and parental Protection.

Data on perceived PRS from different family members may be considered as repeated measure- ments on families. A complete data set in this study consists of the reports given by four individuals each about the two relationships in which they are involved:

(1) mother about the relationship with childl: M+Cl;

(2) mother about the relationship with child2: M+C2; (3) father about the relationship with child]: F-+Cl; (4) father about the relationship with child2: F+C2; (5) child1 about the relationship with mother: Cl+M; (6) child1 about the relationship with father: Cl -+F; (7) child2 about the relationship with mother: C2+M; (8) child2 about the relationship with father: C2+F.

These eight reports, indicated as eight directed relationships are organized in four pairs of reports on unique dyadic relationships (e.g. M-Cl, and Cl +M are two reports about the unique dyadic relationship between mother and childl, M++C 1) within each family. Note, that in these unique dyadic relationships the two raters involved have, in principle, the same sample of parental behaviour

at their disposal to base their reports on. Furthermore, the eight reports can be organized into four clusters each of four reports in which one particular relation partner is involved (e.g. M+Cl, M+ C2, Cl +M, and C2+M are four reports in which the mother is involved).

HLM allows modelling of aJixed part (representing the average pattern within families) and a random part (representing the variability in patterns between families). To indicate the relevant characteristics that differentiate the eight directed relationships necessary for this modelling, dummy variables were constructed. These dummies are a function of the rater, the ratee, or the relationship (e.g. in our study ‘mother as a rater’ is a dummy variable with the value 1 in M+Cl and in M+ C2, and the value 0 in the remaining six reports).

Model of the$xedpart. For each dependent variable, the eight directed reports yield eight averages that are used in the fixed part of the model. This fixed part can be modelled so as to express the average effects of the roles of father, mother, and child. We chose to estimate the rater effects for fathers and mothers (with the children as raters as the reference point), thus investigating whether fathers and mothers, on average, differ as raters from children. Furthermore, we estimated whether relationships involving the father were rated differently from relationships involving the mother. In

the clinical sample two more effects were added to model the fixed part: the difference between relationships involving the patient and those involving the healthy sibling was estimated as well as the effect of the patient as a rater.

In addition, ratings within the family might be influenced by variables such as social desirability, amount of current psychoneurotic complaints, position in the family. These covariates are also represented in the fixed part of the model. Covariates tested were:

-Psychoneurotic complaints of the rater; -Psychoneurotic complaints of the ratee; -Dysfunctional negative presentation of the family’s functioning by the rater (extremely low

Social Desirability scores); -Dysfunctional positive presentation of the family’s functioning by the rater (extremely high

Social Desirability scores); -whether or not the relationship involved the eldest child of the family, or the youngest child of

the family, and/or a child who had already left the parental home; -whether or not rater and ratee were of the same sex.

Model of the random part. Analogous to generalizability theory (Shavelson, Webb & Rowley, 1989), the variances and covariances of the eight reports about a particular parental rearing style

276 Coby Gerlsma et al.

Table 2. Correlations between family members’ perceptions of parental rearmg styles

Paternal Maternal

Affectmn Protection Affection Protection

Agreement in healthy sample -siblings -parent and child1

--parent and child2

0.54 0.18 0.17 0.37” 0.36 0.29’ 0.19 0.08 0.33” 0.26” 0.18 -020

Agreement in climcal sample -siblings -parent and patient m-parent and child2

0.33 0.17 0.33 0.37” 0.17 0.34 -0.03 0.09 0.15 0.10 0.22 0.24

“PiO.05.

can be modelled to express a pattern in the differences between families. In this study we distinguished the following five groups of variability sources:

the family: agreement or shared experience in the entire family;

the person involved in the relationship (relation partner Mother, Father, Child): the extent to which there is consistency in the four reports about relationships involving the mother, and similarly for the father, for childl, and for child2; this consistency is measured to the extent

that it exceeds the agreement between all in the family; the unique dyadic relationships (Mother ++Childl, Mother+-+Child2, Father+-+Childl, Father-

Child2): the consistency in the two reports about each particular dyadic relationship (e.g. M+Cl and Cl -‘M) which is not already a consequence of the effects of the family and of the two individuals involved in the relationship; the person rating: the extent to which rater biases exist without being common to the entire family and their magnitudes for different roles within the family;

the covariance between father and mother as raters, and the covariance between the children as

raters.

Associated with the first four variability sources is a random effect, i.e. an effect that varies randomly among families. The amount of variability is expressed in the variance of the random effect, also

called variance component, that indicates to which extent families differ with respect to these aspects. For methods of testing and estimation we refer to the literature (e.g. Goldstein, 1995; Bryk &

Raudenbush, 1992); in this study we used the PC program ML3 (Prosser, Rasbash & Goldstein, 1991). The parameters in the hierarchical linear model, as presented in Tables 3 and 4 in the ‘Results’ section, are as follows. For the fixed part, including the effects of the covariates, the parameters can be regarded as regression coefficients in the usual way. The significance of these effects is tested by approximate f-tests, with t being equal to the estimate divided by the standard error: if the estimate is at least twice its standard error, the effect is significant (P<O.O5). For the random part, the

parameters are the variances of the random effects and the covariances of selected relevant pairs of random effects. The significance is tested as a difference in fit between various models by means of

the likelihood ratio X2-test (e.g. Bryk & Raudenbush, 1992, pp. 55-56; Prosser et al., 1991, pp. 19-

20). Data of the healthy and clinical sample were analysed separately and in the same way but for one

exception. Contrary to the clinical sample, the two children participating in the healthy sample are not meaningfully distinguishable on relevant a priori grounds (e.g. because of psychiatric history). Therefore, the parameters associated with child1 and child2 were restricted to be equal in the healthy sample.

In general, it should be noted that the data patterns are unbalanced in our study because of incomplete data, that is, not every family unit held the complete number of reports about parental behaviour (e.g. as in the families from which only one parent and two children participated). The HLM analysis used does, however, allow for such unbalanced designs (cf. Bryk & Raudenbush, 1992).

Parenting and psychopathology 271

RESULTS

Correlations between children, fathers, and mothers in perceptions of PRS

The correlations between different raters rating the same parent’s behaviour are shown in Table 2. Correlations between siblings’ ratings ranged between 0.17 and 0.54 in the healthy sample, and between 0.17 and 0.37 in the clinical sample.

Considering the correlations between parents’ and children’s ratings, the strongest agreement seemed to be found between fathers and children in the healthy sample, ranging between 0.26 and 0.36, all four coefficients being significant. In the clinical sample, only the father-patient correlation with regard to Protection was significant. Motherxhild agreement appeared to be marginal in both samples (ranging from -0.20 to 0.19 in the healthy sample, and from -0.03 to 0.22 in the clinical sample.

We also computed correlations between the two ratings each rater made, that is, the correlation between parental behaviour within different relationships but rated by the same rater. These relation- ships proved to be much stronger. In the healthy sample the children’s accounts of maternal rearing

behaviour correlated between r = 0.44 and r = 0.77 with their accounts of paternal behaviour; in the clinical sample the range was from r = 0.5 1 to r = 0.88. The correlation between fathers’ accounts of their own behaviour towards one child and their accounts of behaviour towards the other child

ranged from r=0.54 to r=0.83 in the healthy sample, and from r=0.53 to r =0.72 in the clinical sample. A similar pattern was found in the mothers’ accounts (range from r = 0.65 to r = 0.74 in the healthy sample, and from r = 0.8 1 to r = 0.90 in the clinical sample).

Table 3. Estimates (and standard errors) in the hierarchical linear model for Affection in the healthy and clinical sample

Healthy sample Climcal sample

Estimate S E. Estimate S.E.

Fixed ejfects Constant term Father as rater Mother as rater Difference between relationships involving the mother

and relationships involving the father Patient as rater Difference between relationships involving the patient

and relationships involving the sibling

Random effects Family variance Rater variance -father -mother --child -patient -healthy sibling Covariance between parents as raters Covariance between children as raters

Relation partner variance -father -mother --child -patient -healthy sibling

Relationship variance -father-child -father-patient -father-healthy sibling -mother-child -mother-patient -mother-healthy sibling Residual

21.6 0.5 24.0 1.3 1.8” 0.8 2.0 1.7 1.6 0.6 3.8 1.6

- 1.7 0.3 - 1.7 0.3

0

13.4” 5.1”

13.4

-0.8 2.3 3.1 2.8

13.8” 4.0 1.9 1.3 0 0

1.8

0

4.7

0

4.5 2.3 2.7

1.3

0

0.9

-2.8 1.6 0.1 0.3

2.0 5.1

19.7” 9.1 15.6 6.9

43.0” 11.5 42.8 13.7 11.1” 6.9 9.2 10.3

3.0 1.5

0 2.3

8.8 1.2

0 0 2.2

3.0 2.2

0 2.0

3.6 2.6

0 0 0.7

“P < 0.05.

278 Coby Gerlsma et al.

Table 4. Estimates (and standard errors) in the hierarchical linear model for Protection in the healthv and clmical samole

Fixed e&ts Constant term Father as rater Mother as rater Difference between relationships involvmg the mother

and relationsblps involving the father Patient as rater Difference between relationships involving the patient

and relationships involving the sibling

Random <ffects Family variance Rater variance --father

-mother xhild

-patient --healthy sibling Covariance between parents as raters Covariance between children as raters

Relation partner variance -father -mother -child

patient -healthy sibling

RelatIonship variance -father-child -father-patient --father-healthy siblmg -mother-child --mother-patient ~mother~healthy sibling Residual

Healthy sample Clinical sample

Estimate SE Estimate SE

26.2 0.6 1.3 0.9 0.9 0.8

~ I.34 0.3

I .o 2.3

17.0 5.4 13.0” 4.2 16.9 3.6

-4.7 3.8 5.4 3.8

2.1 I.8 3.1 1.9 2.2 1.2

0 0

0 0

4.6 0.8

24.9 09 2.4 13 1.2 1.3

- 1.6 0.3

1.5 I .2 0.5 0.3

1.9 3.6

13.8” 7.3 22.1” 8.6

26.0” 79 12.3” 5.5

- I.8 6.2 3.4 5.4

4.8 2.9 2.9 2.x

0.9 2.3 0 0

0 0

0 0

4.3 3.3 0 0 5.3 I.2

“P < 0.05

Multilevel analysis of the healthy and clinical sample

Results of the Multilevel analyses in both samples are summarized in Table 3 for parental Affection, and in Table 4 for Protection. Covariate effects are not presented in the tables: significant effects will be reported in the text.

Affection

There seemed on average to be markedly lower Affection scores in the clinical sample [effect

‘Constant’ (aggregate mean within families) 24.0 with S.E. I.31 than in the healthy sample (Constant 27.6, S.E. 0.5).

The fixed rater effects indicate that parents reported more Affection than their children did. In the healthy sample, where both children serve as point of reference, fathers and mothers rated themselves as significantly (1.8 and 1.6, respectively; P~0.05) more affectionate; in the clinical sample, where the healthy sibling is the reference point, mothers rated themselves as more Affec- tionate (estimate 3.8; P<O.O5), while the patients rated their parents as less affectionate (estimate -2.8; 0.05~ P<O.lO) than their healthy siblings did. Furthermore, the fixed part shows that relationships involving the mother were recalled as more affectionate than relationships involving the father in both samples (estimate - 1.7; P~0.05). In the clinical sample it should be noted that relationships involving the patient were not recalled as less affectionate than relationships involving their healthy siblings (estimate 0.1; P> 0.10).

Considering the random part of the mode1 in the healthy sample, the large and significant variance components for individual raters, and the small and nonsignificant components for relationships, relation partners, and the family as a whole suggest substantial influence of the individual rater. With the exception of the father as a relation partner, all other variance components were very

Parenting and psychopathology 279

small, indicating that recollections of parental Affection are primarily tales by individuals, and tales

aboutfathers as relation partner. The small residual variance shows that the effects included in the

model yield a fairly complete picture. The pattern of variances in the clinical sample was quite similar; here, too, recollections of

parental Affection were dominated by large and significant rater effects. Other than in the healthy sample’s families, however, not the father as a relation partner, but the specific father-patient relationship showed a large and significant variance component. Furthermore, contrary to the healthy sample, parents in the clinical sample appeared to agree about the level of Affection they had shown their children (evident from the significant father-mother covariance effect).

Of the various covariates entered subsequently in the analysis of the healthy sample, the regression coefficient for Dysfunctional Negative presentation of the rater, as reflected in an extremely low Social Desirability score, was significant (estimate -2.3, S.E. 0.9; PcO.05). This indicates that a

Dysfunctional Negative presentation of the family’s functioning was related to lower Affection scores. Furthermore, relationships involving an eldest child yielded higher Affection scores (estimate

1.0, S.E. 0.4; P<O.O5). In the clinical sample a significant regression coefficient was found for the psychoneurotic com-

plaints score of the patient-as-a-rater, with higher level of distress being related to lower Affection scores (estimate - 0.04, S.E. 0.01; P < 0.05).* For the other family members no significant covariate effect for psychoneurotic complaints was found (estimate 0.01, SE. 0.04). Furthermore, as was the

case in the healthy sample, an effect was found for the rater’s Dysfunctional Negative presentation of the family. The coefficients were significant for both patients (estimate - 5.8, S.E. 1.8; P-C 0.05) and the other family members (estimate - 2.7, S.E. 1.3; P < 0.05) but the effect size was much larger

in the case of the former. Contrary to the healthy sample, inclusion of the covariates reduced some of the variance components in the random part of the model, in particular the variance component

of the patient as a rater (from 43 to 24) and, to a somewhat lesser extent, the component of the healthy sibling as a rater (from 43 to 37) so that these covariates appear to explain part of the differences in Affection scores between patients and between siblings. The variance components of the fathers and mothers as raters were not affected in this way by the inclusion of the covariates.

In sum, these results suggest that recollections of parental Affection are primarily tales by individuals and tales about fathers (in the healthy sample) or about the father-patient relationship (in the clinical sample). These individuals’ reports (and patients’ reports in particular) were affected by their tendency to present the family in an extremely undesirable light, and by their current level of psychological and physical well-being. Nevertheless, it should be noted that the aggregate mean (the mean Affection score across all four raters) was substantially lower in the clinical sample than

in the healthy sample.

Protection

On average, families in the clinical sample reported slightly less Protection than families in the healthy sample. Relationships involving the mother were recalled as being more protective in both samples (estimate - 1.3 in the healthy sample, and - 1.6 in the clinical sample; P<O.O5). Within the clinical families, relationships involving the patient were rated on average somewhat more protective than relationships involving the healthy sibling, but the difference was not significant.

Results for the healthy and the clinical sample were strikingly similar where the random part was concerned. As was the case with regard to the Affection ratings, the rater effects were by far the strongest (variances between 13 and 17 in the healthy sample and between 12 and 26 in the clinical sample); all rater effects were significant. The covariance between the parents’ accounts and that between children’s accounts was not significant, indicating little agreement between parents and between siblings. Furthermore, in both samples small and nonsignificant variance components were found for the relation partner variables, as well as for the unique dyadic relationships and families

as a whole.

*The size of the psychoneurotic complaints effect seems deceptively small. For purposes of interpretation it should be noted that the amount of psychoneurotic complaints was measured on a 90-item scale, yielding a score range from 90 to 450, with a healthy population mean of 123.05 (Arrindell & Ettema, 1986). In this study, a range was observed from 90 to almost 400, so that the range of the effect of psychoneurotic complaints on Affection scores is 310 * 0.04= 12.4.

280 Coby Gerlsma et al.

Of the covariates entered subsequently, only the rater’s Dysfunctional Negative presentation of family functioning proved to be significant. In the healthy sample such extremely negative social desirability scores were related to increased Protection scores (estimate 2.7, S.E. 0.9; P< 0.05). The same held true for Dysfunctional Negative presentation by the patient as a rater (estimate 4.6, S.E. 1.7; P < 0.05), but not for the other family members as raters. Except for a decrease in the variance component of the patient as a rater, inclusion of this covariate did not affect the model to any great extent.

Overall, these results justify the conclusion that reports about parental Protection are mainly

tales by individuals, to a much smaller extent tales about relation partners and not at all tales about unique dyadic relationships. Contrary to our findings with regard to parental Affection, families from the clinical sample reported on average approximately similar levels of Protection as families from the healthy sample. Overall, relationships involving the patient were rated only slightly higher in Protection than relationships involving the healthy sibling. The tendency to portray the family’s goings-on in an extremely negative light was the only significant covariate effect found. Such a tendency was related to an increase in Protection ratings.

DISCUSSION

Psychiatric patients generally have more adverse recollections of their parents’ rearing behaviour than individuals from the general community. This study was designed to examine to what extent this view is shared by their family members, and to examine what factors might be accountable for differences of opinion within the family. We found mixed results in this regard: while the average level of Affection reported in families of patients was lower than that reported in the families of the healthy sample, there were considerable differences of opinion within the families of both samples. In fact, examination of the pattern of agreement among family members suggests that memories of parental rearing style are dominated by individual rater bias, indicating that such memories primarily tell us about the individual rater, rather than the parent+child relationship, the parent as a relation partner or about the families rearing climate. For instance, if memories of parental behaviour were related to shared perceptions of the family (as in ‘we’re a close family’; see also McCrae and Costa (1988) discussion of ‘family mythology’), one would have expected the family component to be much larger, that is, to constitute a major source of variability between families. Furthermore, the

covariances between parents as raters and between siblings as raters were generally small, indicating that shared beliefs hardly influenced accounts of parenting behaviour. Finally, with few exceptions, the variance components for the family members as relation partners as well as for the parent-child relationships were equally small as those for the family components. Instead, the largest part of the variability in ratings of parental behaviour was accounted for by the individual rater effects. Hence, if actual parental rearing behaviour is dominated by a particular style in families, as is commonly assumed in family research (cf. Dunn & Plomin, 1991) different members of the family seem to recall it quite differently. With regard to our findings on parental Affection and psychopathology, a cautious conclusion might be that the parenting climate in our patients’ families may, on average, have been less affectionate which the mothers seemed least, and the patients seemed most prone to emphasize.

It should be mentioned in this context that the clinical sample in the study was self-selected, and

therefore biased. As was noted before, quite a few patients refused permission for us to contact their family members for the very reason that they had ceased to maintain contact with one particular or even all family members because of disturbed relationships in the family. Hence, the high percentage of nonresponse (approx. 30%) interferes with the research questions addressed in this study. Indeed, given the issues researched, selective nonresponse seems hardly avoidable. The implications for interpretation of the present results regarding this biased sample seem to be twofold. First of all, the clinical sample turned out somewhat smaller than the healthy sample. Statistically, therefore, one would expect to find less significant effects in the clinical sample than in the healthy sample. In fact, this was not the case. Overall, the results for the two samples were quite comparable, but the clinical sample yielded more significant effects, particularly covariate effects. Secondly, the clinical sample is biased in a positive direction, that is, families with severely disturbed relationships

Parenting and psychopathology 281

were probably underrepresented; most families in our sample were still on speaking terms with each other. We expect that the selection bias resulted in more favourable reports of parental behaviour than could be expected in the total population of psychiatric patients. This implies that the averages and fixed effects found are probably positively biased estimates (e.g. higher Affection ratings than one would expect in a representative sample). Furthermore, extremely negative ratings of parental rearing behaviour are underrepresented in our sample, thus reducing the variability in reports. In a representative sample all variance components may be expected to be somewhat larger, presumably without altering the general pattern. In sum, the differences found between the healthy and clinical

sample, and between patients and their healthy siblings are probably underestimates which may turn out to be even larger in a representative sample.

There were some intriguing exceptions to the overall pattern. Fathers’ and mothers’ rater biases in their accounts of the Affection shown to their children covaried significantly and positively, but only in the clinical sample. A tentative explanation might be that lack of parental affection, being an almost stereotypically suspected precursor of all kinds of mishap, had become a topic of discussion between the parents when they realized that one of their children experienced severe problems.

The other exceptions to the dominating role of rater effects concern the significant variance components found for the father as a relation partner (healthy sample) and for the father-patient relationship (clinical sample) in accounts of Affection. It seems odd that these exceptions concern the role of the father, rather than the role of the mother. After all, often being considered the

primary caregivers, mothers were frequently blamed or scapegoated for their supposedly especially important role in the development of psychopathology in their offspring (e.g. Caplan & Hall- McCorquodale, 1985; Chess, 1982). We are fairly certain that the majority of mothers in our sample were, in effect, the primary caregivers in the families concerned but as relation partners they did not contribute to reports about their caregiving behaviour, suggesting that as relation partners the mothers in our sample were perceived as quite interchangeable: compared to the father’s role, the mother’s role seems to be so specific as to render her personal contribution aspecific. One might speculate that the mother role is perceived as being so fundamental as to have become taken for granted (‘mothers will be mothers’). If one is willing to interpret parental representations as reflec- tions of actual parental behaviour (e.g. Brewin et al., 1993) our findings imply that most mothers actually behave similarly. A speculative explanation would be that mothers are reared to do so, as was proposed by Chodorow (1978) analysis of the reproduction of mothering. It is interesting to note in this context that mothers are generally recalled as more affectionate and

protective than fathers (reviewed by Gerlsma & Emmelkamp, 1994), as was also evident in the significant difference between relationships involving mothers and relationships involving fathers in this study. These findings suggest that research on the link between parenting and psychopathology may benefit from studies on (perceived as well as actual) paternal rearing behaviour.

In search for factors possibly explaining some part of the differences of opinion we found that child characteristics, such as position in the family according to birth order and gender, did not appear to be strongly related to reports of parental behaviour. We found only one covariate effect related to birth order, i.e. relationships involving the firstborn child were recalled as more Affectionate. On the other hand, reports of parental behaviour were related to social desirability of the rater, in particular the tendency to present the family in an extremely negative light, and the amount of psychoneurotic complaints reported by the rater. In general, these covariates related to more negative accounts of parenting, i.e. less Affection and more Protection, as would be expected

on the basis of common sense as well as previous research on the relationship between perceived parental rearing styles and psychopathology. Of all types of raters, the patients’ memories seemed to be most strongly related to these covariates. In terms of causal relationships there are several possibilities. On the one hand, the amount of psychoneurotic complaints experienced may give rise to a ‘plaintive set’ (cf. Parker, 1989) which might have induced the patients (and other types of raters for which these covariate effects were found) to give negative accounts of both the family’s functioning and their parents’ behaviour. On the other hand, a negative appraisal of the family and parental behaviour may play a role in the development of the psychoneurotic distress.

While negative presentation and psychoneurotic complaints did seem to explain some of the variability in reports of parental behaviour in this study, future research is needed to address these

282 Coby Gerlsma et al

issues more specifically. Such future studies on the relationship between parental rearing styles and adult psychopathology may benefit from within-family comparisons of both actual and perceived parental rearing behaviour, in order to shed more light on the question as to why some individuals

report childhood adversity and psychiatric disorder, while others, who grew up in an apparently very similar environment, report neither.

Ackno~~lr~9ements-Completion of this study was facilitated by Grant 900-557-008 from the Foundation for Medical and Health Research MEDIGON. For helping to assemble the data for the study we are very grateful to Drs Cornelis van Houwelingen, Harm Jan Pot, Rento Heins, Jan de Vries. Fabiola Schuurmans, Els de Boer. Wim Dijkstra of the Psychiatric Clinic in ~~roningen; to Agnes Scholing of the Department of Clinical Psychology in Groningen, and to the research assistants Jellie Das, Esther Winter. Jeannette van der Meer, Leida Feringa, Marike Tazelaar, and Margot de Vlieger.

REFERENCES

Arrindell, W. A., Emmelkamp, P. M. G., Brilman, E. & Monsma. A. (1983). Psychometric evaluation of an inventory for assessment of parental rearing practices; a Dutch form of the EMBU. Acts Psvchiurrica ~~andjna~?~~a, 67, 1633177.

Arrindell, W. A. & Ettema, J. H. M. (1986). XL-90-R; ~and~eidir~g hlj een m~~~tidin~ensiarzele psyfftopathoiogie-itzrficutor (SCL-90-R; Manual far a ntuiiidimensionalps?,rhopathology checklist). Lisse, The Netherlands: Swets and Zeitlinger.

Blatt, S. J. & Homann, E. (1992). Parentchild interaction in the etiology of depression. Clinical Psychola,gy Recie~, 12,477 91.

Brewin, C. R., Andrews, B. & Gotlib, I. H. (1993). Psychopathology and early experiences: A reappraisal of retrospective reports. Psychological Bulletin, 113, 82-98.

Bryk, A. S. & Raudenbush, S. W. (1992). ~ierar~~i~a~ linear mo&le(s.for social and beharioural research: Applications and data annlysis 1~1e~~ods. Newbury Park: Sage.

Burbach, D. J. & Borduin. C. M. (1986). Parent-child relations and the etiology of depression: A review of methods and findings. Clinicai P.s~ycholagy Review, 6, 133- 153.

Buurmeyer, F. A. & Hermans, P. C. (1985). De Gezins Dimensie Schalen als hulpmiddel bij gezinsdiagnostiek (The Family Dimension Scale as an aid in family diagnostics). TQdschrifi “oar Psychntherapie, II, 336-346.

Buurmeyer, F. A. & Hermans, P. C. (1988). Gezins Dimensie Schalen; Hundleiding (Family dimension scales; manual). Lisse, The Netherlands: Swets & Zeidinger B. V.

Caplan, P. I_., Hall-McCorquodale, 1. (1985). Mother-blaming in major clinical journals. Anznerican Jaurnui uf Orfhop- sychiafry, 55, 3455353.

Chess, S. (1982). The “Blame the Mother” ideology. International Journal @“Mental Health, II, 955107. Chodorow, N. (1978). Tire reproduction a~mothering: Psychoanalysis and the sociology afgender. Berkeley: University of

California Press. Derogatis, L. R. (1975). XL-90: Administratian, scoring andprocedures Manual-Ifor the R(erised) version. Baltimore, MD:

John Hopkins School of Medicine. Clinical Psychometrics Research Unit. Dunn, J. B Plomin, R. (1990). Separate lives: Why siblings are so d$ferent. New York: Basic Books. Dunn, J. R: Plomin, R. (1991). Why are siblings so different? The significance of differences in sibling experiences within the

family. Family Process. 30, 271-283. Gerlsma, C. (1994). Parental rearing styles and psychopathology: Notes on the validity of questionnaires for recalled parental

behaviour. In C. Perris, W. A. Arrindell & M. Eisemann (Eds.), Parenting andpsychapafhology. Chichester: Wiley. Gerlsma, C. & Emmelkamp, P. M. G. (1994). How large are gender differences in perceived parental rearing styles: A meta-

analytic review. In C. Perris. W. A. Arrindell & M. Eisemann (Eds.), Purenfing and Ps~vchopathology. Chichester: Wiley. Gerlsma, C., Emmelkamp, P. M. G. & Arrindell, W. A. (1990). Anxiety. depression, and perception of early parenting: A

meta-analysis. Clinical Psychology Reriov, 10, 251-277. Goldstein, H. (1995). ~~uIiilez~ei s~ufis~ic~~ mo&s (2nd edn). London: Edward Arnold, Greenwald, A. G. (1980). The Totalitarian Ego: Fabrication and revision of personal history. American P.vycholagist, 35,

603-608. McCrae, R. R., Costa, P. T., Jr. (1988). Do parental influences matter? A reply to Halverson. Journal af’ Personality, 56,

445449. Olson, D. H., Russell, C. S. & Sprenkle, D. H. (1983). Circumplex model of marital and family systems: VI. Theoretical

update. Family Process, 22, 69-83. Onstad, S., Ske, I., Torgensen, S. & Kringlen, E. (1993). Parental representations in twins discordant for schizophrenia.

Ps~,ehalogi~al .~~,~~i~~ne, 23, 3355340. Parker, G. (1988). Parental style and parental loss. In A. S. Henderson&G. D. Burrows (Eds.), Handbook qfsaciaipsychiarry

(pp. 15-15). Amsterdam: Elsevier. Parker, G. (1989). The Parental Bonding Instrument: Psychometric properties reviewed. Psychiatric Denelopments, 4, 3177

335. Perris, C, Jacobsson, L., Lindstrom, H., Von Knorring, L. &Pert%, H. (1980). Development of a new inventory for assessing

memories of parental rearing behaviour. Acra Psychiatrica .Scandinaaica, 61, 265274. Presser, R.. Rasbash, J. & Goldstein, H. (1991). ML3; ~?~~~are.f~r three-Iewi ana@&. London: University of London,

Institute of Education. Ross, M. W. (1989). Relation of implicit theories to the construction of personal histories, P.&zoiugLal Rerie,v, 96, 341-

357. Robins, L. N., Schoenberg, S. P., Holmes, S. J., Ratcliff, K. S., Benham, A. & Works, J. (1985). Early home environment

and retrospective recall: A test for concordance between siblings with and without psychiatric disorders. American Journal of Orl/lop.sychiarrJ,, 55, 27-41.

Shavelson, R. L., Webb, N. M., Rowley, G. L.(I989). Generalizabiliry theory. American Psychologist, 44, 922932. Snijders, T. A. B. & Kenny, D. A. ~submitted). Multilevel models for family data.