Observational attachment theory based pa

Attachment & Human Development

Publication details, including instructions for authors and subscription information: http://www.tandfonline.com/loi/rahd20

Observational attachment theory-based parenting measures predict children’s attachment narratives independently from social learning theory-based measures

a b Carla Matias c , Thomas G. O’Connor , Annabel Futh & Stephen Scott cd

a Universidade Lusíada do Porto, Porto, Portugal

b Wynne Center for Family Research, University of Rochester Medical Centre, New York, NY, USA

c King’s College London, Institute of Psychiatry, London, UK

d National Academy for Parenting Research, London, UK Published online: 28 Nov 2013.

To cite this article: Carla Matias, Thomas G. O’Connor, Annabel Futh & Stephen Scott , Attachment & Human Development (2013): Observational attachment theory-based parenting measures predict children’s attachment narratives independently from social learning theory-based measures, Attachment & Human Development, DOI: 10.1080/14616734.2013.851333

To link to this article: http://dx.doi.org/10.1080/14616734.2013.851333

PLEASE SCROLL DOWN FOR ARTICLE Taylor & Francis makes every effort to ensure the accuracy of all the information (the

“Content”) contained in the publications on our platform. However, Taylor & Francis, our agents, and our licensors make no representations or warranties whatsoever as to the accuracy, completeness, or suitability for any purpose of the Content. Any opinions and views expressed in this publication are the opinions and views of the authors, and are not the views of or endorsed by Taylor & Francis. The accuracy of the Content should not be relied upon and should be independently verified with primary sources of information. Taylor and Francis shall not be liable for any losses, actions, claims, proceedings, demands, costs, expenses, damages, and other liabilities whatsoever or howsoever caused arising directly or indirectly in connection with, in relation to or arising

This article may be used for research, teaching, and private study purposes. Any substantial or systematic reproduction, redistribution, reselling, loan, sub-licensing, systematic supply, or distribution in any form to anyone is expressly forbidden. Terms & Conditions of access and use can be found at http://www.tandfonline.com/page/terms- and-conditions

ry 2014 anua

56 10 J t 08:

ondon] a y of L

it rs

ve ni

ay, U low

ol lH

d by [Roya

de oa

nl ow

Attachment & Human Development, 2013 http://dx.doi.org/10.1080/14616734.2013.851333

Observational attachment theory-based parenting measures predict children’s attachment narratives independently from social learning

theory-based measures

a b c Carla Matias c,d , Thomas G. O’Connor , Annabel Futh and Stephen Scott * a Universidade Lusíada do Porto, Porto, Portugal; b Wynne Center for Family Research, University

of Rochester Medical Centre, New York, NY, USA; c King’s College London, Institute of Psychiatry,

ry 2014

London, UK; d National Academy for Parenting Research, London, UK

anua

(Received 7 January 2013; accepted 13 June 2013) Conceptually and methodologically distinct models exist for assessing quality of

56 10 J

parent–child relationships, but few studies contrast competing models or assess their

t 08: overlap in predicting developmental outcomes. Using observational methodology, the

current study examined the distinctiveness of attachment theory-based and social learning theory-based measures of parenting in predicting two key measures of child adjustment: security of attachment narratives and social acceptance in peer nomina-

ondon] a

tions. A total of 113 5–6-year-old children from ethnically diverse families partici- pated. Parent–child relationships were rated using standard paradigms. Measures

y of L

derived from attachment theory included sensitive responding and mutuality; measures

it

derived from social learning theory included positive attending, directives, and criti-

rs

cism. Child outcomes were independently-rated attachment narrative representations

ve ni

and peer nominations. Results indicated that Attachment theory-based and Social Learning theory-based measures were modestly correlated; nonetheless, parent–child mutuality predicted secure child attachment narratives independently of social learning

ay, U

theory-based measures; in contrast, criticism predicted peer-nominated fighting inde-

low

pendently of attachment theory-based measures. In young children, there is some

ol

evidence that attachment theory-based measures may be particularly predictive of

lH

attachment narratives; however, no single model of measuring parent–child relation- ships is likely to best predict multiple developmental outcomes. Assessment in research and applied settings may benefit from integration of different theoretical and methodological paradigms.

d by [Roya

Keywords: parent–child relationships; attachment; parenting; internal working

de models; peer nominations oa

nl ow

D Several conceptual models exist to explain the nature of parent–child relationships and the mechanisms by which they influence children’s social and psychological development;

some models have formed the basis for clinical interventions. Leading models of parent– child relationships have been developed and operationalized largely in isolation, often without cross-fertilization or systematic efforts to distinguish what is particular about each model or its attendant measures. As a result, and despite calls for integrative research on parenting across multiple models (Grusec, 2011 ), we do not yet have a strong evidence base to inform choices made in research studies or clinical settings about whether or not there are substantive distinctions between measures of parent–child relationship quality

*Corresponding author. Email: stephen.scott@kcl.ac.uk

2 C. Matias et al.

derived from different theoretical models. The current study contributes to the limited comparative research by examining what may be particular to an attachment theory-based model of parent–child relationship quality as a predictor of a key attachment outcome, attachment representation, and a more general index of social competence derived from peer nominations.

We focus on measures derived from attachment theory (Ainsworth, Blehar, Waters, & Wall, 1978 ; Bowlby, 1988 ), an influential model of the parent–child relationship with a substantial research base and clinical application (Bakermans-Kranenburg, van, & Juffer, 2003 ), and social learning theory, an equally influential model with a substantial evidence base and clinical application (Dishion & Snyder, 2004 ; Patterson, DeBaryshe, & Ramsey, 1989 ; Webster-Stratton, Jamila Reid, & Stoolmiller, 2008 ). We examine these two models (among many that could be considered) because they are derived from distinct theoretical

ry 2014 traditions and have established observational methodologies.

A starting point for our investigation was to identify what may be distinctive about an anua

attachment theory-based approach to parenting assessment. One candidate for child out- come is the internal working model or attachment representations, a psychological

56 10 J mechanism that underpins attachment theory, its assessment, and its clinical application (Bowlby, 1982 ; Bretherton, 2005 ; Steele, Steele, & Johansson, 2002 ). In infancy, secure

t 08: internal working models are hypothesized to develop from sensitively responsive parental care. From the experience of parental sensitive care, the child develops a model of the attachment figure as available, protective, and caring, and of the self as worthy of love and

ondon] a protection. In contrast, children exposed to a pattern of unresponsive, insensitive care develop an insecure internal working model in which their models of others and self are

y of L it

unreliable, unlovable, and unprotective. The hypothesis that sensitively responsive care rs

leads the child to construct a secure internal working model is central to attachment theory ve ni

and is supported by many studies (e.g., De Wolff & van Ijzendoorn, 1997 ). However, empirical evidence that caregiving experiences are associated with an index of attachment

ay, U quality from separation-reunion procedures or narrative assessments past infancy is low

limited (Dubois-Comtois, Cyr, & Moss, 2011 ; Stevenson-Hinde & Shouldice, 1995 ; ol

Wong et al., 2011 ). That is in contrast to the comparatively large number of studies lH

showing reliable associations between doll-play narrative assessments of internal working models and behavior problems in pre-school and school-aged children (e.g., Barone & Lionetti, 2012 ; Bureau & Moss, 2010 ; Futh, O’Connor, Matias, Green, & Scott, 2008 ).

Studies are needed to examine the connection between the quality of the “real-life”

de experiences with the caregiver with the internal working model in school-aged children, oa

d by [Roya

both to extend this component of the attachment model past infancy and to examine this nl

connection amidst developmental changes in the child and the dyad. In particular, as ow

compared with infants, the association between parental sensitivity and children’s internal

D working model may differ in school-aged children. That is suggested from developmental

changes in the child’s affective expression that may elicit caregiving (e.g., Mesman, Oster, & Camras, 2012 ); additionally, models of sensitivity that underlie the parent’s ability to act as a safe haven for a distressed child and secure base for exploration have not been routinely applied past infancy. One developmental change that has been articulated is the “ goal-corrected partnership” (Bowlby, 1982 ; Marvin & Britner, 2008 ), or the notion that the child and parent are increasingly able to operate toward a shared set of negotiated goals. That is, experiences of a mutual partnership may also (i.e., in addition to sensitive- responsive parental care) lead to and correspond with the formation of secure internal models in the school-aged child. Several studies have examined problem-solving interac- tions in relation to adolescent attachment (Kobak, Cole, Ferenz-Gillies, Fleming, &

3 Gamble, 1993 ; Obsuth, Hennighausen, Brumariu, & Lyons-Ruth, 2013 ; Scott, Briskman,

Attachment & Human Development

Woolgar, Humayun, & O’Connor, 2011 ); we assess mutual or goal-corrected behavior in several settings in relation to secure internal models in young school-aged children.

Equally importantly, the study aimed to examine if observational measures derived from an alternative model of parent–child relationships, social learning theory, were as effective at predicting children’s attachment narratives. That is, we wished to discover whether or not observational parenting measures derived from social learning theory would predict the quality of attachment narratives as effectively as measures derived from an attachment theory tradition. This hypothesis has not yet been tested, but may hold important conceptual, methodological, and clinical applications. There is a solid basis for proposing that social learning theory-based parenting measures would equally predict children’s attachment narratives. Most fundamentally, social learning theory proposes that

ry 2014 parenting qualities are learned or internalized to shape children’s social-cognitive pro- cesses such as attributional biases; these social-cognitive processes, in turn, mediate social

anua and behavioral outcomes (Crick, Grotpeter, & Bigbee, 2002 ; Dodge, Pettit, Bates, & Valente, 1995 ).

56 10 J On the other hand, there are important differences between these models in how the parent–child relationship is conceptualized and measured through observational research.

t 08: Specifically, whereas attachment theory focuses on sensitive responding to the child’s needs, social learning theory models emphasize different parenting dimensions, including (1) positive attention and praise as reinforcers of positive child behavior; (2) clear

ondon] a directions and consequences contingent on undesirable behavior; and (3) criticism con- tributing to coercive cycles that engender antisocial behavior through evoking increas-

y of L it

ingly hostile responses in an interactional build-up (Patterson, 1982 ). Of course, even rs

though these parenting measures differ from those proposed by attachment theory, they ve ni

might still predict security of children’s attachment narratives. Accordingly, we examined if measures that operationalize attachment theory and social learning theory models were

ay, U equally predictive of children’s attachment narratives. low

A second index of social and emotional development to be predicted from observa- ol

tional parenting measures is the ability to make harmonious relationships with other lH

children. We focus here on peer competence indexed by peer nominations, a leading measure of the construct. Both attachment theory and social learning theory predict a link between parenting and peer relations. Children’s experiences of sensitivity and the mutual quality of the goal-corrected partnership with their attachment figures are proposed to be

de carried forward into other relationships, accounting for within-individual stability in oa

d by [Roya

relationship patterns (Sroufe, 2005 ); empirical evidence links parental sensitive respond- nl

ing or attachment security to competence with peers (Sroufe, 2005 ; Thompson, 2008 ). ow

Similarly, social learning theory proposes parent–child relationship mechanisms that lead

D to good peer relations, with the child being rewarded for prosocial overtures and dis-

ciplined for antisocial acts. Empirical research in this tradition provides considerable evidence that positive parenting and disciplinary practices as measured from a social learning standpoint predict peer pro-social behavior and social competence (Sandstrom & Coie, 1999 ); in contrast, harsh parenting is linked with antisocial behavior, peer rejection, and social-skills deficits (Patterson, Dishion, & Yoerger, 2000 ).

Whether or not these alternative models – with their particular hypothesized mechan- isms and assessed dimensions of parenting – differentially predict peer competence has received markedly little attention, and relevant research provides mixed results. Fagot’s ( 1997 ) findings indicated that attachment classification did not provide additional infor- mation to negative parenting behavior when predicting peer negative reciprocity in the

4 C. Matias et al.

playgroup; however, Rose-Krasnor ( 1996 ) found that attachment security was associated with positive social engagement in group play whereas maternal directiveness was associated with aspects of children’s problem-solving behavior. The current study seeks to expand this line of inquiry by examining the overlap and distinctiveness of the measures of the two different models for predicting peer nominations of social competence.

It is important to distinguish between a theory and its operationalization in practice. This study made use of measuring approaches that, for each model, are widely used. For attachment related constructs, global scales were used to assess the key dimensions of sensitivity and mutuality, as an index of the goal-corrected partnership. Attachment-based dimensions were rated during everyday tasks (i.e., not paradigms particular to attachment theory), although there is ample evidence that attachment-related caregiving is observable

ry 2014 outside of attachment paradigms such as separation-reunion situations (Ainsworth et al., 1978 ; Pederson & Moran, 1996 ). For social learning theory-related constructs, event

anua counts of parental behaviors during the interaction with the child were used, the dominant approach taken in the last three decades (Dishion & Snyder, 2004 ; Patterson, 1982 ). We

56 10 J focused on contingent responses to child behavior as well as general behavior (e.g., criticism) of the parent. That is, each set of measures used to index attachment and social

t 08: learning theory in the current study were derived or adapted from existing measures that offer a reasonable representation of their underlying theories.

In summary, our goal was to examine the extent to which widely-used, good quality ondon] a

observational measures derived from attachment and social learning theory overlapped, and whether they differentially predicted two independently-rated key outcomes in young

y of L it

school-aged children, one particular to attachment theory, attachment narratives, and one rs

that is more general, peer-nominated social competence. ve ni

ay, U Methods low ol

Sample and procedure lH

The study is based on the first (pre-treatment) wave of data from the Primary Age Learning Skills (PALS) study, a preventive parenting intervention for at-risk, inner-city families (Scott et al., 2010 ). The study took place in four primary schools in the most disadvantaged ward in one of the most deprived inner-city London boroughs. Only

de children from the second and third cohorts were involved because the attachment narrative oa

d by [Roya

measure was not included in the first cohort. The primary caregiver of 151 children nl

(n = 70 higher risk, n = 81 lower risk) were approached and 75% agreed to take part and ow

were available for data collection (n = 51 higher risk, n = 62 lower risk). A total of 47%

D of mothers were Black African (first generation immigrants), 20% were of African

Caribbean origin, 20% White British/European, and 11% “Other”. Children’s mean age was 5 years 4 months; 49% were male ( Table 1 ). Written consent was obtained from each participant; the local research ethics committee approved the study.

Observational data were gathered in the home by videotaping the child with his/her primary caregiver (the vast majority, 86%, of primary caregivers were mothers). Three standard observational tasks were used (O’Connor, Matias, Futh, Tantam, & Scott, 2013 ): (a) free-play with a designated set of toys (10 minutes), (b) structured-play task where the parent and child worked collaboratively to recreate a pictured Lego structure (10 minutes); (c) tidy-up task (3 minutes). The latter two scenarios create moderate distress due to the difficulty in accurately assembling the Lego structure, and the time-pressure imposed in

5 Table 1. Participant characteristics.

Attachment & Human Development

Socio-demographic characteristics

Study sample

Mean values for England

% (n)/Mean (SD)

Child age (months)

Male gender

51% Primary caregiver ethnicity: White British

Black African

Black African Caribbean

9% Lone parent

Total minority

22% Mother ended education by age 16 yrs

ry 2014 13%

State assisted housing

anua

5% Verbal IQ

Household income £175 per week or less

Parenting variables

56 10 J

CARP Sensitive Responding

t 08:

Mutuality

PBCS Criticism

Directive factor

ondon] a

Positive Attending factor

y of L

Notes: Socio-demographic data for England (where relevant) are from Office of National Statistics (2000).

it rs CARP scales are based on a 1–7 scale; PBCS scales are behavior counts per minute for each scale or summed

across scales for each factor.

ve ni

ay, U the tidy-up. Coders underwent extensive training in the two rating systems described below. Given the limited number of coders available, the same coder used each coding

low ol

system to rate a particular dyad, but the ratings were conducted at least 4 weeks apart and lH

coders were blind to outcome measures. In the 30 cases that were double coded for reliability purposes, the degree of overlap between measuring approaches was similar when coded by the same or the separate coder, suggesting the two approaches were not conflated when both were rated by one coder. Attachment narrative data and peer

d by [Roya nominations were gathered in the school and coded independently of parent–child inter-

de actions by different researchers; demographic data were collected at the home visit along oa

nl with the observational data. ow

D Measures

The Coding of Attachment-Related Parenting (CARP) (Bisceglia et al., 2012 ; O’Connor et al., 2013 ) is a global measure of parent–child interaction quality that was derived from attachment theory and related assessments in young school-aged children (Kochanska & Murray, 2000 ; Stevenson-Hinde & Shouldice, 1995 ). The CARP places conceptual emphasis on patterns of sensitivity, emotional attunement, and bi-directional dyadic processes such as mutuality. Two attachment-related parenting behaviors are the focus of this study; each uses a 7-point Likert scale (1 = No evidence; 7 = Pervasive/extreme evidence). Sensitive Responding assesses the degree to which the parent shows awareness of the child’s needs and sensitivity to his/her signals, promotes the child’s autonomy,

6 C. Matias et al.

adopts the child’s psychological point of view, and physically or verbally expresses positive emotion and warmth towards the child. Mutuality is conceptually compatible with the notion of the “goal-corrected partnership” (Bowlby, 1982 ) and reflects the degree to which parent and child in the dyad accept and seeks the other’s involvement in a joint activity, build on each other’s input and coordinate their efforts/actions while conducting a task together, maintain shared attention and fluid conversation, reciprocate positive affec- tionate behaviors, and keep physical proximity/closeness when interacting with each other. Means (SD) of the two CARP scales are provided in Table 1 . Inter-rater reliability of 30 parent–child play observations coded by two independent raters was good; intraclass correlations (ICC) were .73 for Sensitivity and .81 for Mutuality. Evidence for the validity of measure is found in a recent treatment study showing change in response to parental intervention (O’Connor et al., 2013 ).

ry 2014 The Parent Behavior Coding Scheme (PBCS) is an event-based observational measure deriving from existing social learning observational measures adapted from the widely used

anua Behavior Coding Scheme (BCS) (Forehand & McMahon, 1981 ). Conceptually, the PBCS categories heavily draw on the association between specific disciplinary parenting practices

56 10 J and problem behavior in children (Dishion & Snyder, 2004 ; Webster-Stratton et al., 2008 ). The measure assesses contingent maternal utterances. Child-centered verbalizations include

t 08: descriptive commenting of the child’s actions, positive and facilitative remarks about his/her achievements, and praising the child’s behavior. Child-directive verbalizations include parental commands and prohibitions. Critical verbalizations have both negative content

ondon] a and negative tone. Principal components analysis of the scales and prior research and conceptual models led us to derive three factors for analysis (details from last author)

y of L it

which we refer to as Positive Attending (identified by five scales: attending, positive rs

attending, praise, seeking cooperation, facilitating), Directive parenting (indicated by ve ni

three scales: vague commands, clear commands, and prohibitions) and Criticism. Means (SD) for the PBCS Criticism scale and Directive and Positive Attending factors are

ay, U presented in Table 1 ; these descriptive data are presented for illustrative reasons; we use low

standardized scores in analyses. Whereas Positive Attending and Directive parenting are ol

contingent parenting behaviors (i.e., coded in response to child behavior and the interac- lH

tion), Criticism may be contingent or non-contingent (i.e., global critical comments). The median intraclass correlation across codes, based on a set of 30 tapes, was .75.

The Manchester Child Attachment Story Task (MCAST) (Green, Stanley, Smith, & Goldwyn, 2000 ) is a narrative story stem task to elicit attachment representations in

de school-age children (Colle & Del Giudice, 2011 ; Futh et al., 2008 ) . Using dyadic play oa

d by [Roya

scenarios with a target child doll, mother doll, and dollhouse, the child’s attachment nl

representations are evaluated from doll characters’ behavior and the organization, content, ow

and coherence of the child’s narrative to four story-stems (nightmare, hurt knee, feeling ill,

D lost in a store). Emotional mood induction is used to generate mild stress and increase the

likelihood of evoking attachment feelings and cognitions. The interview is videotaped for later coding. The MCAST was administered at school to avoid possible bias from having the caregiver in the immediate vicinity. Content codes (rated using a 9-point scale unless otherwise noted) were: Maternal Responsiveness, Warmth, Assuagement evident in the child’s narrative and by observation, Affect modulation (6-point scale), and Mentalizing (of self and mother, which were combined; rated 0–3). The scales were standardized through Z-scores. Attachment Disorganization was coded as any evidence of bizarre, unusual features in the child’s narrative using a 9-point scale. Factor analyses (details available from the authors) supported the compositing of the content codes to form a continuous scale of attachment security; for a priori reasons, the Disorganization scale

7 was considered separately. Reliability was demonstrated in two phases. First, the main

Attachment & Human Development

attachment coder, who was blind to all identifying information and observational ratings, passed a reliability assessment for continuous and categorical codes on an independent sample of 10 cases as part of the MCAST training (80% agreement for classifications). Second, on 20 randomly selected tapes from the current sample, ICCs across the scales averaged .67 and there was 80% agreement on the 4-way classification (kappa = .66, p < .01). We use a continuous scale of attachment security because of the statistical advantages over categorical ratings.

Peer nominations This is a well-validated method of measuring social adjustment (Miller-Johnson, Coie,

ry 2014 Maumary-Gremaud, & Bierman, 2002 ). Three attributes were obtained: peer-rated popu-

anua larity (“like most”), peer rejection (“like least”), and fighting behavior (“fights”). The standard peer nomination methodology was used, whereby each child in the classroom is asked to nominate three children for “like most”, “like least”, and “fights” categories;

56 10 J children made selections while viewing a school photograph of their classmates. Scores

t 08: were standardized within class to allow comparisons across classes. All peer nomination ratings were made privately at school. Because the principal in one school was unwilling to allow peer nominations, fewer children had data on this measure.

ondon] a The British Picture Vocabulary Test (Dunn & Dunn, 1981 ) is a standardized measure of child verbal intelligence, adapted from the Peabody Picture Vocabulary Test. Testers y of L

it were trained by a professional authorized by the test developer and certified as competent rs

to administer the test. ve

ni ay, U

Socio-demographic factors low

We measured parental education (categorized as low = left school by 16 yrs; ol

mid = secondary/technical qualification; high = degree), income (categorized as lH

low = <£175 p/week; mid = £176 to £325 p/week; and high = >£326 p/week), single-parent status, and housing type (categorized as council house/flat and owned/ rented property); and self-designated ethnicity of the parent (see Table 1 ).

d by [Roya

de oa

nl Results ow

D The MCAST measures of Security and Disorganization were moderately negatively correlated (r(110) = –.57, p < .01) and both were correlated with verbal IQ (r(112) = .26, p < .01; r(109) = –.22, p < .05, respectively). Females were rated more secure than males (standardized means across the scales were .12 (.77) for females and –.27 (.86) for males, t(111) = 2.53, p < .05). Gender and verbal intelligence were therefore included as covariates when predicting attachment security. The only socio-demographic variable associated with Peer nominations was gender; compared to females, males received more nominations for Dislike (3.27 (2.52) vs 2.12 (1.93), t(100) = –2.61, p < .05) and Fights (5.29 (4.07) vs 1.14 (1.18), t(100) = –7.00, p < .01). Gender was therefore included as a covariate when predicting peer status.

8 C. Matias et al.

Table 2. Correlations between attachment theory-based and social-learning theory-based parenting measures.

Sensitive Responding Directive

Positive Attending

Sensitive Responding

.81*** Note: *p < .05; **p < .01; ***p < .001.

Overlap between attachment and social learning theory measures ry 2014 Correlations between parenting measures within each theory indicated substantial overlap

anua between Mutuality and Sensitive Responding ( Table 2 ). Within the social learning measures, there was modest overlap between Positive Attending and Directive

56 10 J Parenting; Criticism was unrelated to Positive Attending but modestly associated with Directive Parenting ( Table 2 ). Correlations between measures from each theory indicated

t 08: the highest degree of overlap for positive parenting measures but no relationship between attachment measures and Directive Parenting, and a modest negative relationship between the two attachment measures and Criticism ( Table 2 ). It therefore appears that the

ondon] a measures of positive parenting from either theory may be tapping into somewhat similar processes, whereas the social learning theory-derived measures of Directive Parenting and

y of L it

Criticism are indexing processes that are distinct from the attachment theory-derived rs

measures. ve ni

ay, U Prediction of child social adjustment from parenting measures low

There were modest to moderate correlations between the two attachment-based measures ol

of parenting and attachment security and peer nominations ( Table 3 ). In contrast, none of lH

the social learning theory measures was significantly associated with the attachment narrative measures; however, both Positive Attending and Criticism were associated with peer nominations of fighting. Also, as the findings in Table 3 show, none of the attachment theory-based or social learning theory-based measures of parenting was sig-

de nificantly associated with Disorganization in the attachment narrative. oa

d by [Roya

nl ow

Table 3. Correlations between parenting measures and child outcomes.

Attachment narrative

Peer nominations

Liked Disliked Fighting Attachment theory-based measures

Sensitive Responding

.25* −.23* −.30** Social learning theory-based measures Positive Attending

−.07 .19 .45*** Note: n’s range from 97 to 108; *p < .05; **p < .01; ***p < .001.

9 Regression analyses were used to examine the uniqueness of the different approaches

Attachment & Human Development

to parenting measurement. Given the high degree of overlap between Mutuality and Sensitive Responding, we focus on the former in analyses below (the pattern of results with Sensitive Responding was similar; details available from the authors).

Regression analyses were performed to test the hypothesis that Mutuality was significantly associated with secure attachment narratives independent of socio- demographic variables and three parenting measures from the social learning theory model: Positive Attending, Directive Parenting and Criticism. Although the bivariate associations between social learning theory-based measures and attachment narrative were not significant, we nevertheless wanted to ensure that they did not mediate the prediction from the attachment based Mutuality measure. As might be expected from the pattern of correlations in Table 3 , regression results indicated that observed

ry 2014 Mutuality was significantly associated with secure attachment narrative independent of all covariates (B = .16; SE = .07; t = 2.34, p < .05); of the other variables in the model,

anua only child gender (B = –.36; SE = .15; t = –2.36, p < .05), and verbal IQ (B = .02; SE = .01; t = 2.47, p < .05) were also significantly associated with attachment security.

56 10 J Peer nominations of fighting were significantly associated with both attachment and social learning theory measures in bivariate analyses; accordingly, we conducted a

t 08: regression analyses with both sets of parenting measures as predictors. Regression results ( Table 4 ) indicated that attachment-based ratings did not predict significant independent variance in peer nominations whereas the social learning theory construct of Criticism did;

ondon] a male gender was also a significant predictor. Regression models for peer nominations of Liked and Disliked indicated that neither attachment nor social learning theory measures

y of L it

predicted peer ratings independently, despite the significant bivariate correlations reported rs

in Table 3 ; this implies that the attachment theory-based and social learning theory-based ve ni

measures account for overlapping variance. Given the ethnically diverse nature of the sample, we carried out a series of supple- ay, U

mentary analyses to examine possible moderator effects by ethnicity or social class low

ol lH

Table 4. Prediction of peer nominations from parenting scales.

d by [Roya β

B (SE)

B (SE)

B (SE)

de Step 1. oa

nl

ow .49***

D Verbal IQ

R 2 = .35; R 2 ∆ = .35*** F(2,94) = 4.22*

R 2 = .08; R 2 ∆ = .08*;

R 2 = .13; R 2 ∆ = .13***

F(2,94) = 25.25*** Step 2.

F(2,94) = 7.30***

−.52 (.46) .10 Attending Directive

R 2 = .48; R 2 ∆ = .13*** F(6,90) = 2.32

R 2 = .13; R 2 ∆ = .05;

R 2 = .16; R 2 ∆ = .03;

F(6,90) = 13.96*** Notes: Regression estimates reported from the final model. *p < .05; ***p < .001.

F(6,90) = 2.84

10 C. Matias et al.

indicators. We found no reliable evidence that any of the demographic factors in Table 1 moderated the associations between parenting/attachment measures and child outcomes (details available from the last author).

Discussion It might be presumed that, because of their different conceptual origins, parent–child

measures derived from attachment and social learning theories would not overlap and would differentially predict developmental outcomes. This important hypothesis has received surprisingly little clinical or empirical attention, however. As a result, rather than ascribing particular significance to one model or another, the contrary position might also be proposed: that is, there is substantial overlap between and predictive power from

ry 2014 alternative parent–child measures because the core elements of “good quality” parenting are equally detected by measures derived from different theories.

anua We did not set out to provide a definitive test of the distinctiveness of attachment theory-and social learning theory-based measures of parenting. Instead, we aimed to begin

56 10 J to fill in the gap in research that translated different conceptual models into practical measures for research and practice. The current study yielded two particularly novel

t 08: findings. The first is that observer ratings of parent–child interaction quality during standard interaction tasks predicted the quality of children’s attachment narratives in young school-aged children. A second related, and more novel, finding is that observa-

ondon] a tional measures of parent–child interaction quality predicted children’s attachment narra- tives independent of observational measures derived from an alternative model, social

y of L it

learning theory. Somewhat less novel but notable is the finding that peer nominations rs

were associated with multiple measures of parent–child relationship quality; peer-reported ve ni

fighting behavior was particularly associated with critical parenting behavior. We discuss the implications of the findings after noting the study strengths and weaknesses.

ay, U There were a number of limitations. First, given the inclusion of high-risk, ethnically low

diverse families, these findings may not generalize to other samples – although the ol

findings might be expected to generalize to high-risk samples in general, which tend to lH

be ethnically diverse. We found no evidence for moderation by ethnicity, but the study was not powered to test ethnicity as a moderator. We also focused on a single caregiver, which is inevitable given the high rate of single-parent families in moderate-high risk settings. Also, it is possible that recording parenting behavior under more stressful

de situations such as when the child is worried, ill or unexpectedly separated would have oa

d by [Roya

led to stronger associations with child narratives. Furthermore, whereas attachment theory nl

measures were coded on global scales, social learning theory measures were based on ow

event sampling. Differences between measures and their predictions might reflect this

D different coding strategy. However, this difference is consistent with how measures are

conventionally coded within each model, and the data presented here suggest that related constructs can be highly correlated despite different coding strategies. Given the practical limits and demands of coding, the same rater coded all parenting measures (in separate coding sessions); shared rater variance between attachment theory-based and social learning theory-based measures would bias against finding distinct predictions. Finally, given the cross-sectional nature of the design, we cannot draw casual conclusions about direction of effects. Counter-balancing these limitations were several strengths to this study, including a reasonably large sample for intensive, direct observational methodol- ogy; the inclusion of hard-to-reach families that are often not well-represented in research; high quality direct observational measures of parent–child interaction assessed from

11 alternative conceptual-methodological schemes; and outcome measures that were derived

Attachment & Human Development

from independent sources using very distinct methodologies. An important finding in this study was the link between observer ratings of parent– child interaction quality during standard assessment procedures and the quality of school- aged children’s attachment narratives derived from the story-stem methodology. That is an important complement to existing research on attachment narratives in school-aged children, which has tended to focus on adjustment outcomes. Indeed, empirical evidence linking observational measures of caregiver behavior to attachment security is substantial in infancy (Belsky, 1999 ; NICHD & E. C. C. R. N, 2006 ); but limited past infancy (Stevenson-Hinde & Shouldice, 1995 ). The observational coding system used in the current study was designed specifically to build on the work in younger children, and provides some of the only observational evidence in school-aged children that caregiver

ry 2014 sensitivity and a mutually responsive interaction pattern predict how children talk about attachment experiences using the very distinct methodology of the narrative assessment

anua (and coded by independent raters). Furthermore, this study found that there is something unique about the ability of observational ratings of parenting derived from an attachment

56 10 J theory model to predict attachment narratives. Specifically, Mutuality, the manner in which caregivers and children negotiate play, tasks, and clean-up – and analogous to

t 08: the goal-corrected partnership notion from attachment theory – predicted narrative attach- ment security after accounting for alternative observational measures and covariates; analyses of the Sensitive Responding construct were parallel but somewhat weaker. In

ondon] a contrast, measures of positive attention derived from a social learning model (Patrick, Snyder, & Schrepferman, 2005 ) were weakly and non-significantly associated with

y of L it

attachment narratives. The implication is that there may be value in operationalizing rs

caregiving behaviors along the lines developed by attachment theory rather than using a ve ni

more general “positive” style of interacting for understanding the security/insecurity of children’s attachment narratives. In addition to contributing to the validity of the observa-

ay, U tional and narrative measures, these findings also support the value of ascertaining low

attachment-related parenting in naturalistic observational situations rather than only in ol

separation-reunion procedures (Pederson & Moran, 1996 ). The distinctive benefit of lH

attachment theory-based parenting measures were limited to attachment-specific out- comes, attachment representations, and not to another key index of social and emotional development, peer nominations.

We did not find reliable links between either attachment or social learning measures of

de parent–child interaction quality and attachment Disorganization in the story stem narra- oa

d by [Roya

tive. That may be because parenting behaviors associated with disorganization are less nl

likely to be observed in brief observations in everyday settings, or because the observa- ow

tional ratings did not include relevant dimensions. Adding dimensions such as parental

D helpless states of mind, abusive or fear-inducing behavior (Lyons-Ruth, Yellin, Melnick,

& Atwood, 2005 ), and various ways in which maternal awareness of infant states has been coded (Fonagy & Target, 2005 ; Meins, Fernyhough, Fradley, & Tuckey, 2001 ) may have improved prediction of disorganization. It is possible that attachment disorganization was not adequately assessed in the narrative measure, although that explanation is challenged by previous analyses linking this measure of disorganization to behavioral and emotional problems (Futh et al., 2008 ).

Both attachment and social learning theory hypothesize that parent–child relationship quality would predict peer competence, although perhaps from different parenting dimen- sions and/or through different mechanisms. Results from this investigation indicated that measures from both models predicted peer nominations of aggression (fighting).

12 C. Matias et al.

However, social learning theory based ratings of parent–child interaction (criticism) added additional prediction of peer nominated fighting behavior independent of attachment- based ratings, which did not predict unique variance. That is similar to Fagot’s ( 1997 ) findings, and suggests that there may be a particular value in assessing conflict and criticism specifically when predicting negative child behavior that is carried forward to other settings and relationships, perhaps because of a direct behavioral learning (e.g., modeling) process. Critical and coercive parenting, which were directly captured in the social learning theory-based measures (and arguably less so in the attachment theory- based measures), provides the means and the opportunity to teach a child a repertoire of aversive and aggressive behaviors. It may be noteworthy that none of the parenting measures predicted unique variance in the other dimensions of peer ratings. It is possible that the associations with Fighting were stronger than for measures of acceptance/rejection

ry 2014 because it is a more readily observable outcome to young children or because it has greater salience in this high-risk setting. It is also important to note that conceptions of

anua “ like” and “dislike” or popularity may be confounded in high-risk settings because of social norms that may reward disruptive behavior.

56 10 J There are obvious challenges in contrasting attachment and social learning theory models of the parent–child relationship for predicting social competence in school-aged

t 08: children. For example, there is similarity in how attachment theory-based and social learning theory-based measures are described, and some studies have begun to integrate concepts from attachment and social learning theory models directly (Kerns, Tomich,

ondon] a Aspelmeier, & Contreras, 2000 ; Scott et al., 2011 ; Sutton, 2001 ; Van Zeijl et al., 2006 ). Nevertheless, uncertainty about which models or measures to adopt for practical use in a

y of L it

research or clinical setting remains a common problem because of the general lack of rs

comparative research. The broader goal of this line of study is to test and clarify the limits ve ni

of distinct models and the reliability of their translation toward the formation of an integrative model and measurement (Grusec, 2011 ). One specific challenge for research

ay, U combining measures from alternative models is methodological. For example, whereas low

attachment studies have tended to focus on global measures, social learning theory ol

measures tend to focus on discrete behaviors (Forehand & McMahon, 1981 ; Patterson, lH

1982 ). That difference in methodology – molar versus molecular coding – creates an inevitable confound in comparing measures, and that may have influenced the results. However, the difference between frequency counts and rating scales should not be over- emphasized, as the correlation between the frequency of positive attending counts and

de sensitive responding rating was high. Our focus was in contrasting alternative models as oa

d by [Roya

they are typically applied because this would offer the most relevant test of our hypoth- nl

eses. Clearly, further progress in testing, contrasting, and integrating alternative models of ow

parent–child relationship quality will also depend on methodological refinements, which

D currently vary across conceptual traditions. Linking this kind of work to clinical contexts

is particularly needed given concerns – and so far, limited data – on how well traditional research measures inform decision-making about parenting and intervention in such settings as custody and visitation, foster care, and pediatric clinics (Byrne, O’Connor, Marvin, & Whelan, 2005 ; Joseph, O’Connor, Briskman, Maughan, & Scott, 2013; Marie- Mitchell & O’Connor, 2013 ).

Comparing alternative observational ratings scales is rarely executed in developmental research (Allen, Hauser, Bell, & O’Connor, 1994 ; McElhaney & Allen, 2001 ), but provides a particularly strong test of alternative models. This strategy may be especially valuable to address calls to consider how attachment, social learning, and other models of parent–child relationship are distinct or redundant (Goldberg, Grusec, & Jenkins, 1999 ;

13 O’Connor, 2002 ). Observational assessments are intensive and therefore expensive, but

Attachment & Human Development

are considered a gold standard for elucidating causal mechanisms and determining the effectiveness of clinical treatment. A key consideration is which conceptual model and set of associated observational measures may be most suited to the conceptual or clinical question of interest; these findings begin to address that issue. Lastly, our findings underscore the potential value in adopting a cross-theoretical, integrative approach to assessment for academic studies and applied settings. For example, young children exhibiting aggression towards their caregiver may be seen by some practitioners as suffering from an insecure attachment pattern and could be offered attachment-based psychotherapy (Speltz, DeKlyen, & Greenberg, 1999 ); other practitioners might interpret oppositional behavior as a consequence of reinforcement of antisocial acts and offer parent training . Whether both approaches are effective and, if so, whether or not it is

ry 2014 due to the same mechanism requires additional clinical research. Findings from these kinds of investigations could inform practitioners about how and when to apply different

anua models to treatment, such as when one approach is not working (Scott & Dadds, 2009 ), and help identify potential sources of variation in response to parenting interventions

56 10 J (Scott & O’Connor, 2012 ). t 08:

Funding This research was made possible by grants from the Foundation for Science and Technology –

ondon] a

Ministry of Science, Technology, and Higher Education, Lisbon, Portugal, the Joseph Rowntree Trust, the Psychiatry Research Trust, the Jacobs Foundation, and the Economic and Social Research

y of L

Council (UK).

it rs

ve ni

References Ainsworth, M. D. S., Blehar, M. C., Waters, E., & Wall, S. (1978). Patterns of attachment: A

ay, U

psychological study of the strange situation. Hillsdale, NJ: Lawrence Erlbaum.

low

Allen, J. P., Hauser, S. T., Bell, K. L., & O’Connor, T. G. (1994). Longitudinal assessment of

ol

autonomy and relatedness in adolescent-family interactions as predictors of adolescent ego

lH

development and self-esteem. Child Development, 65(1), 179–194. Bakermans-Kranenburg, M. J., van, IJzendoorn, M. H., & Juffer, F. (2003). Less is more: Meta- analyses of sensitivity and attachment interventions in early childhood. Psychological Bulletin, 129(2), 195–215.

Barone, L., & Lionetti, F. (2012). Attachment and social competence: A study using MCAST in

d by [Roya

low-risk Italian preschoolers. Attachment & Human Development, 14(4), 391–403.

de Belsky, J. (1999). Interactional and contextual determinants of attachment security. In J. Cassidy & oa

nl

P. R. Shaver (Eds.), The handbook of attachment: Theory, research and clinical applications (pp. 249–264). New York, NY: Guilford.

ow

D Bisceglia, R., Jenkins, J. M., Wigg, K. G., O’Connor, T. G., Moran, G., & Barr, C. L. (2012).

Arginine vasopressin 1a receptor gene and maternal behavior: Evidence of association and moderation. Genes, Brain, and Behavior, 11(3), 262–268.