Research Review: The relation between childand parent anxiety and parental control:

a meta-analytic review

Corine O. van der Bruggen, Geert Jan J.M. Stams, and Susan M. BogelsDepartment of Educational Sciences, Faculty of Social and Behavioral Sciences, University of Amsterdam,

The Netherlands

Background: There is growing research interest in the association between parental control and childanxiety. Parental control may enhance child anxiety and parents may exert control in anticipation oftheir childs anxiety-related distress. Moreover, high levels of anxiety in parents could influence thedevelopment of parental control. Whereas past reviews have solely examined the relation between childanxiety and parental control, this meta-analysis focuses on the associations between both child andparent anxiety and parental control. Methods: The associations of parent anxiety and child anxietywith observed parental control (k = 23 studies, N = 1,305 parentchild dyads) were investigated using ameta-analytic approach. Moreover, factors were identified that may function as moderators of theserelations, such as parent and child gender, family socioeconomic status, child age, and design andmeasurement characteristics. Results: A substantial association between child anxiety and parentalcontrol (d = .58) was found. Moderator analyses yielded the strongest effect sizes for studies with anoverrepresentation of girls, for school-aged children, for families from higher socioeconomic back-grounds, and for studies using a discussion task to assess parental control. Although a nonsignificantrelation was found for the relation between parent anxiety and parental control (d = .08), small butsignificant effects were found for school-aged children, for studies using a discussion task to assessparental control, and for samples with an overrepresentation of boys. Conclusions: As the directionof the association between child anxiety and parental control is unknown, future studies shoulduse experimental designs to further explore the causal link between child anxiety and parentalcontrol. Keywords: Anxiety, meta-analysis, parentchild interaction, parenting.

There is a growing interest in the associationbetween parental control considered as the pres-sure parents put on their children to think, feel orbehave in desired ways and child anxiety. There areat least three ways through which parental controlmay enhance child anxiety: (I) by increasing thechilds perception of threat (Rapee, 2001), (II) byreducing the childs perceived control over threat(Chorpita, Brown, & Barlow, 1998), and (III) by notproviding the child with occasions to explore theirenvironment and to develop new skills to cope withunexpected environmental events (Barlow, 2002).Rapee (2001) also discussed the effect of child

anxiety on parental control. He argued that parentsexert control in anticipation of their childsanxiety-related distress. The childs anxiety-relatedbehaviors alone, however, may not be a sufficientexplanation for the degree of control parents execute.In particular, the parents own anxiety level has beenproposed to influence parental control (e.g., Adam,Gunnar, & Tanaka, 2004; Bogels & Brechman-Toussaint, 2006; Ginsburg, Grover, & Ialongo, 2004;Turner, Beidel, Roberson-Nay, & Tervo, 2003).Ginsburg and Schlossberg (2002) suggested thathigh levels of anxiety in parents could interfere withthe development of parents adaptive coping skills,

which may lead to specific anxiety-enhancingparenting behaviors, such as modeling avoidance,rejection, and (over)control. High parental anxietylevel may also lead to avoidance of situations parentsperceive as threatening. Such avoidance is consid-ered to be a major symptom of anxiety. Anxiousparents tend to perceive challenging new situationsfor their children as threatening. They therefore mayexecute increased parental control in order to avoidthreatening situations. These parents control theirchildrens behavior by preventing them from facingnew situations (Wood, 2006; Woodruff-Borden,Morrow, Bourland, & Cambron, 2002). Moreover, ahigh anxiety level may also result in a reduction ofthe effort parents put into interaction with theirchild. This may lead to more inadequate parenting,such as a high degree of parental control (Woodruff-Borden et al., 2002).This meta-analytic review examines whether there

is empirical evidence to support the hypotheses thatchild and parent anxiety are related to parentalcontrolling behavior observed during parentchildinteraction. To our knowledge no review exists ofstudies examining the association between parentanxiety and parental control. Four narrative reviewsinvestigated the relation between child anxiety andboth self-reported and observed parental control(Ballash, Leyfer, Buckley, & Woodruff-Borden, 2006;Conflict of interest statement: No conflicts declared.

Journal of Child Psychology and Psychiatry 49:12 (2008), pp 12571269 doi:10.1111/j.1469-7610.2008.01898.x

2008 The AuthorsJournal compilation 2008 Association for Child and Adolescent Mental Health.Published by Blackwell Publishing, 9600 Garsington Road, Oxford OX4 2DQ, UK and 350 Main Street, Malden, MA 02148, USA

Bogels & Brechman-Toussaint, 2006; Rapee, 1997;Wood, McLeod, Sigman, Hwang, & Chu, 2003), butresults were inconsistent. Recently, McLeod, Wood,and Weisz (2007) conducted a meta-analysis ofstudies examining the relation between child anxietyand parental control and rejection, assessed bymeans of questionnaires and observational methods.Fifteen observational studies examining the relationbetween child anxiety and parental control or closelyrelated concepts were included. McLeod et al. founda medium overall effect size of r = .25 (d = .52) for therelation between child anxiety and parental control,and a somewhat smaller overall effect size of r = .20(d = .41) for the link between child anxiety andparental rejection. Relations, however, proved to bestronger for observed than self-reported parenting.The primary aim of our study is to examine the

relation between child anxiety and parental controlassessed by means of behavioral observation. Wechose to include only studies assessing parentalcontrol by means of behavioral observation, becauseMcLeod et al. showed that (I) parental control wasmore strongly associated with child anxiety than wasparental rejection, and (II) effect sizes were larger forstudies using observational measures of parenting.Moreover, observations of parenting are not subjectto the usual rater bias that is heir to parent self-reports of child-rearing practices, and which mayunduly inflate associations between perceived childand parent anxiety and self-report of parentingbehaviors. It should be noted that the associationbetween perceived and observed parenting tends tobe weak (e.g., Dishion, Li, Spracklen, Brown, &Haas, 1998), which indicates that perceived andobserved parenting constitute different constructs.Interestingly, observed parental control especially isa target of interventions aimed at reducing thechilds anxiety level (e.g., Bradley et al., 2003).McLeod et al. (2007) stressed the importance of the

role of parents own anxiety in the relation betweenchild anxiety and parental control. Whereas otherreviews have solely examined the relation betweenchild anxiety and parental control, the current meta-analysis focuses on the associations between bothchild and parent anxiety and parental control, thusextending McLeod et al.s meta-analysis.The present meta-analytic review is aimed at

identifying factors that may function as moderatorsin the association between child and parent anxietyand parental control, that is, variables having aneffect on the direction and strength of these rela-tions. We explore the role of specific measurementcharacteristics (e.g., measurement method, interac-tion task, measurement reliability) in moderating therelation between child and parent anxiety andparental control. Apart from these measurementcharacteristics, the factors investigated in our meta-analytic review include sample features (e.g., parentand child gender, socioeconomic status, child age),design characteristics (e.g., sample size), and

publication characteristics (year of the publication,impact factor of the journal).First, the possible moderating role of parent

gender is investigated. It has been proposed thatdifferences in parentchild interaction depend onparents gender. Fathers are more likely thanmothers to execute autonomy granting throughencouraging their children to be independent,adventurous and risk taking. Mothers, in contrast,stress emotional security and personal safety(Bogels & Phares, in press; Paquette, 2004;Popenoe, 1999). From this it can be derived thatfathers grant more autonomy and mothers executemore control. It is therefore possible that fathersand mothers being differently biased towardsexecuting lower and higher levels of control,respectively, may have an effect on the strength ofthe associations between child and parent anxietyand parental control. For instance, fathers andmothers could have different thresholds forchanging their levels of control in response to theirown or their childs degree of anxiety.There is empirical evidence showing that children

with a high anxiety sensitivity level are more sus-ceptible to parenting influences (Belsky, Hsieh, &Crnic, 1998; Kochanska & Aksan, 2006). As anxietysensitivity seems to be higher in girls than in boys(Silverman, Goedhart, Barrett, & Turner, 2003), it isimportant to account for gender differences whenexamining the relation between child anxiety andparental control.Parents may have the greatest influence on their

childrens behavior during early childhood, as thereare fewer non-family influences in the lives of youngchildren. Moreover, because early childhood is atime of tremendous learning and growth, facilitatedby the social environment, younger children may bemore susceptible to parental behavior than olderchildren (Connell & Goodman, 2002). Therefore, ageof the child should be included as a moderator of therelation between child anxiety and parenting. Age ofthe child could also moderate the relation betweenparent anxiety and parental control, as parent anxi-ety and ineffective parenting may be mutually rein-forcing. If this assumption holds, the effect size forthe relation between parent anxiety and parentalcontrol should be larger with increasing age of thechild.It has been demonstrated that parents from lower

socioeconomic backgrounds execute more parentalcontrol relative to parents from middle or highersocioeconomic backgrounds (Hoff, Laursen, & Tar-dif, 2002). Possibly, parents from lower socioeco-nomic backgrounds often experience aversiveenvironmental influences that may result inunvarying high levels of parental control. Conse-quently, lower-class parents may be less receptive totheir childrens anxiety-related behaviors (see Mar-tini, Root, & Jenkins, 2004), and may react less totheir own internal states. Therefore, the relation

1258 Corine O. van der Bruggen, Geert Jan J.M. Stams, and Susan M. Bogels

2008 The AuthorsJournal compilation 2008 Association for Child and Adolescent Mental Health.

between child and parent anxiety and parentalcontrol may be weaker in families from lower socio-economic backgrounds than in families from middleor higher socioeconomic backgrounds.Silk, Morris, Kanaya, and Steinberg (2003) showed

that parental control and granting of autonomyshould be considered distinct child-rearing dimen-sions rather than two opposite ends of a continuum.Therefore, the strength of the association betweenchild and parent anxiety and parental control may bedifferent in studies investigating parental controlthan in studies examining autonomy-grantingbehaviors. Silk et al. (2003) found that parentalcontrol, but not lack of parental granting of auto-nomy, was related to a high level of child anxiety. Incontrast, the meta-analysis of McLeod et al. (2007)demonstrated that the granting of autonomy wasmore strongly related to child anxiety than parentalcontrol.Specific features of the coding system for

assessing parental control or granting of autonomymay function as moderators of the relation betweenanxiety and parental control. First, the wayparental control is operationalized could have amoderating influence. In some coding systems thefrequency of certain behaviors was scored, whereasin other coding systems the degree of parentalcontrol/granting of autonomy was assessed. Sec-ond, the quality of the coding system (inter-raterreliability) may be a moderator. Another potentialmoderator is the type of task that is executedduring parentchild interaction. Working on aperformance task (e.g., Tangram puzzle), a dis-cussion task (e.g., conflict conversation), or inter-acting in an unstructured situation may elicitdifferent levels of controlling behavior. Pastresearch indicated that emotionally arousing tasksare more likely to elicit parental control than moreneutral tasks (Donenberg & Weisz, 1997; Gins-burg, Grover, Cord, & Ialongo, 2006). Furthermore,type of anxiety (social anxiety versus unspecifiedanxiety or internalizing symptoms) is included as amoderator.The final moderator concerns the studys design.

Some studies have compared the degree of parentalcontrol between extreme groups (clinically anxiousversus normal samples), whereas other studieshave investigated associations between the level ofanxiety and parental control in general populationsamples. McLeod et al. (2007) found that studiescomparing extreme groups showed larger effectsizes between child anxiety and parenting thanstudies investigating the association in generalpopulation samples.In sum, the first goal of this meta-analytic review

is to provide a clear answer on whether child andparent anxiety are related to observed parentalcontrol. The second purpose is to identify factorsthat may function as moderators of the relationsbetween child and parent anxiety on the one hand

and parental control on the other hand, such asparent and child gender, family socioeconomicstatus, child age, and design and measurementcharacteristics.

Method

Sample of studies

The primary search method involved inspection of thecomputerized databases Psycinfo and ERIC (Educa-tional Resources Information Centre). No specific yearwas indicated, and the following key words were usedfor searching in varying combinations: parentchildinteraction, motherchild interaction, fatherchildinteraction, observation, control, autonomy granting,intrusiveness, overinvolvement, overprotection,parentinfant interaction, motherinfant interaction,fatherinfant interaction, parenting, parental, mother,father, maternal, paternal, rearing style, rearing prac-tices, parenting style, anxiety, anxiety disorder, andinternalizing problems. In the second step, the refer-ence lists from previous reviews and the only meta-analysis were used. The third step included a search inthe reference sections of the studies drawn from thedatabases to identify citations that had not appeared sofar. Studies were inspected and included if they met thefollowing criteria: (a) studies measuring parent and/orchild anxiety in terms of anxiety (e.g., Hudson & Rapee,2002) or internalizing problems (e.g., Mills & Rubin,1998), but not studies measuring predisposition toanxiety such as behavioral inhibition (e.g., Kagan,Snidman, Arcus, & Reznick, 1994) and anxious tem-perament (e.g., Kochanska, 1995; Stevenson-Hinde &Glover, 1996); (b) studies assessing parental control(Woodruff-Borden et al., 2002), granting of autonomy(e.g., Siqueland, Kendall, & Steinberg, 1996), psycho-logical control (e.g., Caron, Weiss, Harris, & Catron,2006), negative control (e.g., Booth, Rose-Krasnor,McKinnon, & Rubin, 1994), involvement (e.g., Hudson& Rapee, 2002), intrusiveness (e.g., Feldman, Green-baum, Mayes, & Erlich, 1997), command (e.g., Hummel& Gross, 2001), watching and controlling (e.g., Donen-berg & Weisz, 1997), or restriction (Krohne & Hock,1991).The large majority of the studies examining the

relation between child anxiety and parental control orbetween parent anxiety and parental control werecross-sectional. In three of the four longitudinalstudies (Feldman et al., 1997; Mantymaa, Puura,Luoma, Salmelin, & Tamminen, 2004; Rubin, Bur-gess, & Hastings, 2002) concurrent statistics were notavailable. In the case of Feldman et al. (1997), weexamined the association between mothers traitanxiety and observed maternal intrusiveness sixmonths later, as more tasks were used to assessmaternal intrusiveness at this time, which providesincreased validity and reliability. In the case of Man-tymaa et al. (2004) and Rubin et al. (2002), only dataon the relation between maternal intrusiveness andchild internalizing problems were available after twoor four years. Booth et al. (1994) examined the rela-tion between maternal negative control and exclus-ively child internalizing problems at age 4, but not

Child and parent anxiety and parental control 1259

2008 The AuthorsJournal compilation 2008 Association for Child and Adolescent Mental Health.

when the children were 8 years old. Therefore, thedecision was made to include data obtained when thechildren were 4 years old.No studies were excluded from the meta-analysis on

the basis of flawed designs. Instead, impact factor andpublication status were included as moderators in theanalyses as indicators of study quality (Mullen, 1989).Only studies of new samples that had not beenexamined in previous studies were included in themeta-analysis. The study by Whaley, Pinto, and Sigman(1999), however, was excluded in favor of the largerstudy by Moore, Whaley, and Sigman (2004). The finalsample of the current meta-analysis included 17 stud-ies examining the association between child anxietyand parental control and 11 studies investigating therelation between parent anxiety and parental control.From this sample, 6 studies examined both childand parent anxiety in relation with parental control.A detailed list of all included studies is provided inTable 1. (*These studies are marked with an asterisk inthe References section).

Coding the studies

Each study was coded using a detailed coding systemfor recording sample, design and measurement, andpublication characteristics. All studies were coded bytwo coders and the inter-rater reliabilities were satis-factory, mean Kappa for categorical variables = .94(.711.00), mean ICC for continuous variables = .99(.951.00). Sample characteristics coded were parentgender (mothers/primary caregivers or fathers/bothparents), child gender (percentage of boys), familysocioeconomic status (lower, middle, or higher class),and mean age of the children. The following design andmeasurement characteristics were included: studydesign (group contrasts between clinical and normalsamples versus correlational designs using generalpopulation samples), type of anxiety (social anxietyversus unspecified anxiety or internalizing symptoms),sample size, parenting concept (assessing parentalcontrol or mixed parental control and granting ofautonomy), measurement methods (assessing fre-quency or degree of parental behaviors), interactiontask (using a performance, discussion, or combination/unstructured task), and measurement quality (inter-rater reliability). Year of publication and impact factorof the journal (ISI Journal Citation Reports, ScienceEdition, 2005) were included as publication character-istics.Parenting was split up into a parental control cate-

gory and a mixed category of parental control andgranting of autonomy. Only one study (Moore et al.,2004) exclusively measured the degree of parentalgranting of autonomy and therefore this study wasassigned to the mixed category. The measurementmethods applied in Donenberg and Weisz (1997) andRubin et al. (2002) were a combination of assessing thefrequency and the degree of parental behaviors. Thesestudies were assigned to the frequency category, whichappears more suitable than the degree category, asfrequency counts of parentchild behaviors were theprimary unit of analysis. Type of parent anxiety couldnot be included as a moderator, as all studies examinedrelations with unspecified parent anxiety. Some studies

(e.g., Feldman et al., 1997) reported more indicators forinter-rater reliability of the coding system. In thesecases, the strictest measure was used. Socioeconomicstatus was coded as lower, middle, or higher class. As arule we followed the socioeconomic status classificationreported in the studies. When studies (e.g., Feldmanet al., 1997; Mills & Rubin, 1998) did not mention meanage of the child, the mean was estimated from the agerange.

Data analysis

The outcomes of all studies were transformed intoCohens d, the standardized association between childanxiety and parental control, and between parentanxiety and parental control. In most cases, effect sizeswere computed on the basis of reported means andstandard deviations or correlations. Where only the p,k2, or F was reported, these statistics were used. When astudy did not find a significant result, effect sizes werecalculated using the reported raw data available in thestudys report. Next, in two cases, authors did notprovide the statistical information necessary to calcu-late an effect size. These authors were contacted to gainthe required information. But when an attempt failed,which happened in two cases, we assigned an effect sizeof zero based on a one-tailed p of .50 (Z = .00). This is acommonly used and conservative strategy, which mayproduce an underestimation of the true magnitude ofan effect size (Durlak & Lipsey, 1991). The exclusion ofnonsignificant results from meta-analyses, however,would result in an overestimation of the magnitude ofthe combined effect size estimate (Rosenthal, 1995).Some studies examined both parental control andgranting of autonomy (e.g., Ginsburg et al., 2004;Siqueland et al., 1996). When this was the case, esti-mated effect sizes were averaged, resulting in oneoverall effect size for a combination of parental controland granting of autonomy.Next, the data across studies were analyzed, with the

goal of estimating the population effect sizes for therelation between child anxiety and parental control andfor the association between parent anxiety and parentalcontrol. We also computed significances of the effectsizes and confidence intervals around the point estim-ate of the effect sizes. Homogeneity analyses andanalysis of variance were conducted. When there ishomogeneity, estimates of effect sizes will differ onlybecause of unsystematic sampling error. In case ofheterogeneity, however, one needs to be careful wheninterpreting the results. Heterogeneity indicates thatone should look for moderators explaining variability inthe sample of studies. For the categorical variables, weconducted analysis of variance. A series of regressionanalyses was carried out to compute the individualstandardized regression coefficients for the continuousvariables. Subsequently, multiple regression analysiswas conducted, in order to establish whether significantmoderators were unique.A common problem in conducting meta-analysis is

that many studies remain unpublished because ofnonsignificant findings. To examine whether suchpublication bias or file drawer problem exists one cancalculate the minimum number of studies with nullresults that are needed to reduce significant meta-

1260 Corine O. van der Bruggen, Geert Jan J.M. Stams, and Susan M. Bogels

2008 The AuthorsJournal compilation 2008 Association for Child and Adolescent Mental Health.

Table1Thestudiesincludedin

themeta-analysis:Samplecharacteristicsandeffectsizes

Study

Samplecharacteristics

NGender

parent

%Boys

Agechild

(years)

SES

Type

anxiety

Design

Parentingconcept

d

Childanxiety

Booth

etal.(1994)

73

M54

4.3

MU

CNegativecontrol

.01

Caronetal.(2006)

70

P64

9.7

LU

CPsychologicalcontrol

.19

Donenberg

&Weisz(1997)

34

P63

11.5

LU

GWatching&Controlling

.41

Dumasetal.(1995)

84

M48

4.2

HS

GCommand

.97

Dumas&LaFreniere

(1993)

60

M40

4.1

MS

GCommand

.59

Greco&Morris(2002)

48

F52

11.4

MS

GCommands&Physicalcontrol

.76

Hummel&Gross(2001)

30

Both

F&M

40

11.9

MS

GCommand

.53

Mantymaaetal.(2004)

50

M46

.17

MU

CHostility

&Intrusiveness

.00

Mills&Rubin

(1998)

35

M43

7.0

MS

GBehavioral&Psychologicalcontrol

.72

Rubin

etal.(2002)

72

M43

2.1

MU

CIntrusivecontrol

.20

Siquelandetal.(1996)

44

Both

F&M

+onlyForM

68

11.0

MU

GAutonomygranting

.61

Parentanxiety

Adam

etal.(2004)

102

M48

2.0

MC

Intrusiveness

.28

Feldmanetal.(1997)

47

M50

.72

MC

Intrusiveness

.12

Ginsburg

etal.(2004)

50

M48

5.9

LG

Overcontrol&Autonomygranting

.01

Turneretal.(2003)

81

OnlyForM

47

10.0

MG

Saysto

becareful,standsclose&

verballydirects

activity

.14

Woodruff-Bordenetal.

(2002)

51

OnlyForM

47

9.5

MG

Control

.11

Childanxiety

andparent

anxiety

Barrettetal.(2002)

44(Ch)

81(Pa)

Both

F&M

+only

ForM

(Ch)

M(Pa)

34

9.1

MG

Control

1.72(Ch)

.00(Pa)

Broome&Endsley(1989)

83

M47

5.1

C

Admonishment(including

Autonomygranting)

.26(Ch)

.00(Pa)

Hudson&Rapee(2001)

75

M49

10.7

MG(Ch)

C(Pa)

Involvement

.89(Ch)

.15(Pa)

Hudson&Rapee(2002)

57

Both

F&M

52

11.0

HG(Ch)

C(Pa)

Involvement

.57(Ch)

.32(Pa)

Krohne&Hock(1991)

47

M51

11.5

C

Restriction(includingAutonomy

granting)

.57(Ch)

.50(Pa)

Moore

etal.(2004)

68

M51

10.7

MG

Autonomygranting

1.14(Ch)

.57(Pa)

Note.Ch=child,Pa=parent.ForGenderparent,M

=mother,F=father,P=primary

caregiver.ForSES,L=low,M

=middle,H=high.ForTypeanxiety,S=social,U=unspecified.For

Design,G=groupcontrast,C=correlation.

Child and parent anxiety and parental control 1261

2008 The AuthorsJournal compilation 2008 Association for Child and Adolescent Mental Health.

analytic results to nonsignificance (Durlak & Lipsey,1991). Meta-analytic findings are considered to berobust if this fail-safe number exceeds the critical valueobtained with Rosenthals (1995) formula: 5 times thenumber of studies, plus 10. Appropriate fail-safenumbers (p < .05) were computed. Outlying effect sizeswere identified on the basis of z-values larger than 3.3or smaller than 3.3 (Tabachnick & Fidell, 2001).

Results

Table 1 presents the effect sizes of all studiesexamining the association between child anxiety andparental control and/or the relation between parentanxiety and parental control. No outlying effect sizeswere identified. In interpreting the magnitude ofeffect sizes, generally accepted conventions formu-lated by Cohen (1988) were used. Effect sizes ofabout d = .20, d = .50, and d = .80 were consideredas indices of small, medium, and large effects,respectively.

Child anxiety

Seventeen studies examining the link between childanxiety and parental control were included in thismeta-analytic review, and these studies reporteddata on N = 974 parentchild dyads. The meta-analysis yielded a medium-to-large and significantoverall effect size of d = .58, p < .001 (CI .51