Attachment & Human Development

ISSN: 1461-6734 (Print) 1469-2988 (Online) Journal homepage: https://www.tandfonline.com/loi/rahd20

A validation of the Experiences in Close Relationships-Relationship Structures scale (ECRRS) in adolescents Dagmar Feddern Donbaek & Ask Elklit To cite this article: Dagmar Feddern Donbaek & Ask Elklit (2014) A validation of the Experiences in Close Relationships-Relationship Structures scale (ECR-RS) in adolescents, Attachment & Human Development, 16:1, 58-76, DOI: 10.1080/14616734.2013.850103 To link to this article: https://doi.org/10.1080/14616734.2013.850103

Published online: 12 Nov 2013.

Submit your article to this journal

Article views: 1259

View Crossmark data

Citing articles: 12 View citing articles

Full Terms & Conditions of access and use can be found at https://www.tandfonline.com/action/journalInformation?journalCode=rahd20

Attachment & Human Development, 2014 Vol. 16, No. 1, 58–76, http://dx.doi.org/10.1080/14616734.2013.850103

A validation of the Experiences in Close Relationships-Relationship Structures scale (ECR-RS) in adolescents Dagmar Feddern Donbaeka* and Ask Elklitb a Department of Psychology and Behavioural Sciences, Centre for Alcohol and Drug Research, Aarhus University, Aarhus, Denmark; bDanish Research Unit of Psychological Trauma, Institute of Psychology, University of Southern Denmark, Odense, Denmark

(Received 8 April 2013; accepted 7 June 2013) Emerging evidence points toward a two-dimensional attachment construct: avoidance and anxiety. The Experiences in Close Relationships-Relationship Structures scale (ECR-RS; Fraley, Heffernan, Vicary, & Brumbaugh, 2011) is a questionnaire assessing two-dimensional relationship-specific attachment structures in adults and, hence, moves beyond the traditional focus on romantic relationships. The present article explored the psychometric abilities of the ECR-RS across parental and best friend domains in a sample of 15 to 18-year-olds (n = 1999). Two oblique factors were revealed across domains, exhibiting satisfactory construct validity, including factorspecific links to the model of adult attachment (Bartholomew & Horowitz, 1991), and independent factor discrimination between subgroups. A robust validation supports the application of the ECR-RS to assessing relationship-specific adolescent attachment structures. Keywords: Experiences in Close Relationships-Relationship Structures questionnaire; self-report; attachment dimensions; test validity; adolescents

Introduction During adolescence, the attachment system undergoes a profound transformation. Emotional, cognitive, and behavioral development interact with one another to influence both the expression and meaning of attachment relationships as the adolescent prepares to function more independently of parents. Despite that the nature of adolescent attachment structures remains unclear; researchers within developmental psychology conceptualize the adolescent attachment system as expanding and evolving into multiple forms (Allen, 2008). Since the publication of the Strange Situation Classifications in 1978 (Ainsworth, Blehar, Waters, & Wall, 1978), there has been a wide interest in applying the study of individual differences in attachment patterns to populations beyond childhood. Within personality and social psychology over the last few decades, researchers have achieved increasing psychometric advances, utilizing multi-item scales based on a two-dimensional model of attachment for adult romantic relationships (Crowell, Fraley, & Shaver, 2008). The most recent attachment measurement within this line of scale development is the Experiences in Close Relationships-Relationship Structures (ECR-RS; Fraley, Heffernan, Vicary, & Brumbaugh, 2011), which is designed to assess attachment structures in multiple relationships using the same set of items. Validation of the ECR-RS is the focus of this article. *Corresponding author. Email: [email protected] © 2013 Taylor & Francis

Attachment & Human Development

59

In recent years, psychometric improvements in multi-item scales have enabled these scales to account for increased individual variability in adult attachment patterns; as a result, researchers have become interested in applying these adult attachment scales to adolescent samples. For instance, the Relationship Questionnaire (RQ; Bartholomew & Horowitz, 1991) was validated in a clinical sample of adolescents (Scharfe, 2002), and the Experiences in Close Relationships-Revised (ECR-R; Fraley, Waller, & Brennan, 2000) was validated for use with 8 to 14-year-olds (Brenning, Soenens, Braet, & Bosmans, 2011). To our knowledge, there is currently no self-report measure relying on a two-dimensional model of attachment that has been validated with a non-clinical sample of adolescents between 15 and 18 years of age. Similarly, there currently exists no scale based on a two-dimensional model of attachment with the psychometric ability to assess multiple adolescent relationships, which has been accentuated specifically for adolescents both theoretically (Allen, 2008) and in attachment scale development (Crowell et al., 2008). This article addresses these two gaps by exploring the psychometric properties of the ECR-RS in a sample of Danish adolescents, with the objective of adapting the latest psychometric advances in adult attachment scale development to the study of adolescent attachment organization. Attachment Theory (Bowlby, 1969, 1973, 1980) is one of the most dominant frameworks for understanding early socio-emotional development. Bowlby proposed that, in learning to understand and interpret the external world, infants internalize their first social experiences, resulting in the development of Internal Working Models (IWM) of self and significant others. These socio-cognitive structures represent one of the core tenants of Attachment Theory and are believed to underlie whether proximity is sought or avoided in close relationships throughout the life-span (Bowlby, 1988). The development of more advanced cognitive skills in adolescence enables the conceptualization of attachment experiences and relationships, making adolescence the first period during which internal states of attachment organization become assessable (Allen, 2008). This level of formal operational thinking therefore allows self-report assessments of “systematic patterns of expectations, needs, emotions, emotionregulation strategies, and social behavior that result from the interaction of an innate attachment behavioral system” (Shaver & Mikulincer, 2002). While adult and adolescent attachment self-report measures have developed simultaneously, they appear to be grounded in conceptually distinct frameworks. Researchers in both domains have shown interest in relationship-specific attachment structures for decades. With roots in child clinical research, self-report measures of adolescent attachment have focused on the quality of parental and peer relationships (Armsden & Greenberg, 1987; Greenberg, Siegel, & Leitch, 1983; West, Rose, Spreng, & Adam, 2000; West, Rose, Spreng, Sheldon-Keller, & Adam, 1998) by assessing the specific insecure attachment constructs described in the first two volumes of Attachment Theory (Bowlby, 1969, 1973). Researchers within personality and social psychology, in contrast, have applied the Strange Situation Classifications to explore individual differences in adult romantic relationships (Brennan, Clark, & Shaver, 1998; Collins, 1996; Collins & Read, 1990; Fraley et al., 2000; Hazan & Shaver, 1987; Simpson, 1990) or general relationship orientations (Bartholomew & Horowitz, 1991). The Experiences in Close Relationships (ECR; Brennan et al., 1998) scale emerged from a 1996 study that explored 14 adult attachment self-report measures, together comprising 323 items all reworded to encompass a romantic relationship orientation. The results indicated that each measure, regardless of its initial conceptual framework, was best explained by two global dimensions: anxiety and avoidance. Hence, findings suggested that anxiety and avoidance dimensions corresponded to IWM of self and others (Crowell et al., 2008), underlying both Bowlby’s clinically derived insecure constructs

60

D.F. Donbaek and A. Elklit

(e.g., Armsden & Greenberg, 1987) and Ainsworth’s Strange Situation Classifications of individual differences (e.g., Hazan & Shaver, 1987). Based on the large item pool, the ECR was designed as a 36 multi-item two-dimensional scale representing the 18 highest factor-loadings on each dimension. To overcome classical test theory limitations believed to generate misleading inferences of individual differences, a 36-item revised version (ECR-R; Fraley et al., 2000) was developed, for which Item Response Theory was used to explore items from the four most commonly used attachment measures of romantic relationships (Brennan et al., 1998; Collins & Read, 1990; Griffin & Bartholomew, 1994; Simpson, 1990). The ECR and ECR-R possess superior psychometric properties (Fraley et al., 2000) and are currently the most valid and most commonly used self-report measures of adult attachment (Crowell et al., 2008; Ravitz, Maunder, Hunter, Sthankiya, & Lancee, 2010). There has been great interest in exploring the psychometric properties of the ECR and ECR-R even further, which has led to adaptations and several validations across populations and cultures (Alonso-Arbiol, Balluerka, & Shaver, 2007; Conradi, Gerlsma, van Duijn, & de Jonge, 2006; Fairchild & Finney, 2006; Kooiman, Klaassens, van Heloma Lugt, & Kamperman, 2012; Olssøn, Sørebø, & Dahl, 2010; Sibley, Fischer, & Liu, 2005; Sibley & Liu, 2004; Tonggui & Kazuo, 2006; Wei, Russell, Mallinckrodt, & Vogel, 2007; Wongpakaran, Wongpakaran, & Wannarit, 2011). To compare adult attachment structures within and across four different relationships, thus moving beyond the focus of romantic relationships, Fraley and colleagues (2011) recently developed the ECR-RS scale, which was based on the ECR-R item pool. The latest psychometric advances within adult attachment scale development, with the ability to explore relationship-specific orientations within a two-dimensional framework, are extended further in this study by exploring the validity of the ECR-RS for use in adolescent attachment research. Validating the ECR-RS with adolescents The present article explores the structure and underlying constructs of the ECR-RS in adolescents. Analysis of construct validity comprised examination of the independent psychometric abilities of underlying attachment dimensions to show convergent and discriminant correspondence with the model of adult attachment (Figure 1) and their ability to discriminate between subgroups. Fraley and colleagues (2011) recommended predefining hypotheses for dimension-specific discriminations to explore possible Model of self Anxiety High

Low

Secure

Preoccupied

Dismissing

Fearful

Model of other Avoidance

High

Figure 1. Model of adult attachment. © [Kim Bartholomew]. Reproduced with permission. Permission to reuse must be obtained from the rightsholder.

Attachment & Human Development

61

independent contributions of ECR-RS anxiety and avoidance attachment dimensions. Because the current study was exploratory, the following hypotheses were intended as preliminary indications to enhance the understanding of attachment as a dimensional construct, where current adolescent–parent contact and illegal substance use were hypothesized to influence the level of attachment security in adolescents. The first hypothesis was that a lack of current adolescent–parent contact or absence of parental figures would inhibit the maintenance of attachment security in adolescence. Attachment organization is believed to influence the formation of future relationships throughout the life-span (Bowlby, 1977). Fundamental Attachment Theory (Bowlby, 1969, 1973, 1980) states that attachment behavior toward parental figures profoundly transforms during adolescence, with the focus of attachment redirected to non-parental figures (Allen, 2008; Bowlby, 1988; Hazan, Hutt, Sturgeon, & Bricker, 1991; Weiss, 1982). Attempts to regulate the attachment and exploratory systems are manifested in increased negotiation with parents, facilitating autonomy while striving to preserve adolescent–parent relationships (Allen, 2008). Because transitional periods resulting in increased conflict are most frequently experienced in adolescence, the adolescents are vulnerable to feeling more insecure, making a well-organized adolescent–parental relationship especially crucial during these periods. The second hypothesis was that illegal substance use would be an additional risk factor for the development of insecure adolescent attachment. Although substance use is common in adolescence (Hibell et al., 2012) and may even be characterized as a normative adolescent-limited risk behavior (Moffitt, 1993), studies have shown that illegal substance use in adolescence increases the likelihood of insecure attachment structures. Illegal substance use has been specifically associated with insecure structures in adolescence in the form of negative expectations (Allen, Leadbeater, & Aber, 1994) and lack of emotional strategies in close relationships (Schindler et al., 2005) while also mediating the long-term relationship between a range of psychopathological disorders and the development of insecure attachment organization (Allen, Hauser, & Borman-Spurrell, 1996). Statistically based on the ECR-R item pool and specifically designed to assess relationship-specific structures, the ECR-RS has good psychometric potential and is highly applicable to adolescent attachment research. The present study explored relationship-specific attachment structures in adolescents across parental and best-friend domains. Because being in a current romantic relationship was a requirement for taking part in the first validation study in adults (Fraley et al., 2011), the present validation study in adolescents set out to explore the possible influence of romantic relationship experience on attachment organization. Previous research has indicated a relationship in adolescence between secure attachment and comfort with intimate emotional interaction (Allen, Porter, McFarland, McElhaney, & Marsh, 2007; Zimmermann, 2004). In light of the issues discussed above, the first aim of this article was to explore the structural validity of the ECR-RS in adolescents, but this article also aimed to examine convergent, discriminant, and discriminative validity, looking at possible factor-specific correspondence to the model of adult attachment (Figure 1) and differences between subgroups within current adolescent–parent contact, illegal substance use, and romantic relationship experience.

Method Study design and sample The current study was part of a larger cross-sectional study on substance use with 13,000 participants 15–65 years of age carried out by the Centre for Alcohol and Drug Research

62

D.F. Donbaek and A. Elklit

at Aarhus University in Denmark. Participants were randomly selected by the Danish Civil Registration System from four municipalities representing combinations of rich, poor, rural, and urban living areas. The 15 to 18-year-old age groups were stratified (n = 4269) and recruited for two independent online surveys. The current validation study was part of the second survey, which included a sample of 1999 15 to 18-year-olds (47.5% response rate) who completed the ECR-RS (Fraley et al., 2011) and RQ (Bartholomew & Horowitz, 1991) attachment scales (see Measures for details). The sample was 52% female and had an average age of 16.44 years (SD 1.11). The majority of respondents were Danish natives (87.5%) living with both parents (69%). Nearly the full sample reported to be currently attending school or serving an apprenticeship (95.5%).

Measures Experiences in Close Relationships-Relationship Structures The ECR-RS1 (Fraley et al., 2011) is a 9-item self-report questionnaire designed to assess two underlying attachment dimensions: avoidance (items 1–6) and anxiety (items 7–9) across mother, father, romantic partner, and best friend domains. A Likert scale ranging from 1 (strongly disagree) to 7 (strongly agree) was used to obtain a total score for the avoidance (range: 6–42) and anxiety (range: 3–21) dimensions separately. The first four items were reverse scored. Higher scores reflect greater levels of insecure attachment within each relationship domain. Fraley and colleagues (2011) have recommended using of any one or more relationship domains depending on the individual research purpose. An online adaptation of the ECR-RS was utilized, incorporating the parental figure and best friend domains (Appendix 1). Before answering items in the parental figure domain, respondents were asked to select a parental figure reference worded as follows: “select the parental figure you feel most closely attached to. If you feel equally attached to more than one of the alternatives, e.g., both of your parents, please select one at random.” Participants had the following options: mother, father, stepmother, stepfather, or other. Following data collection the parental figure domain comprised the entire sample, whereas mother, father, and other (including stepmother, stepfather, and other parental figure) domains were generated as independent groups.

Relationship questionnaire The RQ1 (Bartholomew & Horowitz, 1991) was designed to assess individual differences within a two-dimensional model of adult attachment, consisting of four paragraphs each describing prototypical attachment patterns: secure, preoccupied, dismissing, and fearful (Figure 1). Inspired by Bowlby’s (1969) conceptualization of IWM, high and low degrees of anxiety (model of self) and avoiding closeness (model of others) in relationships with significant others characterize the four prototypical attachment patterns within the twodimensional model. The RQ contains two parts. The first is a forced-choice paragraph in which respondents identified the single best-fitting attachment pattern. In the second part, participants used a Likert scale ranging from 1 (not at all like me) to 4 (somewhat like me) to 7 (very much like me) to rate each of the four paragraphs. The questionnaire was originally designed to measure general adult attachment orientations toward “others.” However, Bartholomew (n.d.) recommends rewording the RQ to adapt it to the individual relationship-specific orientations being studied. In the present study, we adapted the RQ to parental figure and best friend attachment orientations by applying the wording used in the

Attachment & Human Development

63

ECR-RS, “this person.” While the RQ remains one of the most frequently used adult attachment measures (Ravitz et al., 2010), in recent years researchers have found that categorical models cannot account for much variation in adult attachment structures. The RQ is currently recommended in combination with multi-item measures (Fraley & Waller, 1998) and was therefore selected for construct validation purposes in this study. Illegal substance use A single-item dichotomous question was used to measure current use of illegal substances: “have you used illegal drugs including cannabis in the last two months.” Participants indicated either yes or no. Current adolescent–parent contact Two single-item questions measured current contact with mother and father separately, asking: “have you had contact with your mother/father in the past six months.” For both items, participants indicted either yes, no, or no mother/father. Relationship experience Lifetime romantic relationship experience was assessed by a single-item question, “have you ever had a boy-/girlfriend,” which requires a yes/no response. An additional singleitem yes/no question was used to assess current romantic relationship involvement: “have you currently got a boy-/girlfriend.” Procedure In cooperation with Statistics Denmark, participants were given written information on sample selection, the study’s purpose, contact information for the researchers, and a link to the online survey with a personalized password. On the first page of the survey, participants gave informed consent to participate. The survey included a large battery of self-report questionnaires comprising 288 items, including demographic details. A selection of questionnaires was used for the current study, described elsewhere. Implemented filtrations ensured that the individual participant was only confronted with relevant questions. Participants were given a gift voucher for participation. All items were forced choice; therefore, no missing data were present. Ethics Prior to data collection, the Danish Data Protection Agency granted permission to conduct the study. Danish research ethics committees only cover biomedical research and do not apply to research within social sciences or questionnaire surveys (Hartlev, 1996). Methods of analysis Data were collected using the online survey program SurveyXact. Statistical analyses were performed with IBM Statistical Package for Social Sciences (SPSS) 20. Structural validity of the ECR-RS was examined with exploratory factor analysis, which is recommended when applying a scale to a new population or using it in a new language (Terwee

64

D.F. Donbaek and A. Elklit

et al., 2007). Pearson product-moment correlation coefficients were calculated to examine the relationship among ECR-RS subscales within and between domains. Standard multiple regression analyses were used for convergent and discriminant validity, measuring the contribution of RQ attachment patterns when accounting for ECR-RS subscale scores in separate models. Multiple regression analyses were also used for testing the discriminative influence of demographic details (gender and age), illegal substance use, current adolescent–parent contact, relationship experience, and parental figure reference on ECR-RS avoidance and anxiety levels.

Results ECR-RS dimensionality and internal consistency across domains Factor analysis Exploratory factor analysis with the extraction method of principal axis factoring was used to examine the dimensionality of the ECR-RS in a Danish sample of adolescents (n = 1999). The vast majority of the current sample selected their mother (70.2%) as the parental figure reference for the parental figure domain, followed by their father (25.2%), other parental figure (3.8%), stepfather (.5%), and stepmother (.3%). The factor structure underlying the nine items was examined separately within each of the following domains: best friend, mother, father, and combined parental figure. The Kaiser-Meyer-Olkin measure of sampling adequacy (>.79) and Bartlett’s test of sphericity (p < .05) indicated good factor analyses across domains. The initial three eigenvalues were 4.09, 1.87, 1.00 for best friend, 3.88, 1.95, .89 for mother, 3.51, 2.11, .99 for father, and 3.80, 2.00, .93 for parental figure domains. The first two factors accounted for 66.29%, 64.79%, 62.43%, and 64.43% of the total cumulative variance in the best friend, mother, father, and parental figure domains, respectively. Based on the Kaiser Criterion and Cattell’s scree tests, a two-factor structure solution was favored. Direct Oblimin rotation was used for further interpretation of factor loadings on the two extracted factors. The nine items showed similar discrimination between the two factors across domains (Table 1), with items 1–6 showing moderate to high loadings on the first factor (avoidance) and items 7–9 showing high loadings on the second factor (anxiety). A few exceptions for this pattern were found within the best friend domain, in which items 5 and 6 showed low to moderate cross-loadings. This will be discussed later. Because these two items were not cross-loaded within other relationship domains, all nine items were retained. Factor loadings for the mother, father, and parental figure domains were highly similar. The structural validity of the ECR-RS corresponded well to its first validation study in adults (Fraley et al., 2011), reflecting specific attachment avoidance and anxiety items.

Internal consistency The two attachment factors, avoidance and anxiety, were tested for internal reliability separately within each relationship domain. High Cronbach’s alpha coefficients were found for both the avoidance (>.81) and anxiety (>.86) subscales. Pearson correlation coefficients were conducted to test the relationship between the two attachment subscales within and across parental figure and best friend domains (Table 2). Moderate correlations were found between the two subscales within each domain (e.g., parental figure avoidance and anxiety) as well as between the same attachment subscale across domains

.80 .89 .93 .52 .43 .51 .03

−.04 −.12 −.07 .23 .23 .32 .87 .82 .86

.90 .92 .95 .52 .22 .34 −.01 .00 −.09

.75 .90

−.14 .22 .10 .15 .80

−.02 −.18

.02 −.06

.91 .49 .33 .46 .03

.81 .83

.73 .91

−.16 .19 .18 .17 .83

−.07 −.16

Anxiety Avoidance Anxiety

Father (n = 504)

.04 −.05

.93 .50 .39 .49 .03

.81 .89

Avoidance

.76 .90

−.15 .24 .14 .17 .80

−.03 −.19

Anxiety

Parental figure (n = 1999)

Notes: *Based on Exploratory factor analysis using principal axis factoring with Direct Oblimin rotation. Factor loadings greater than .32 are in bold to indicate possible cross-loading (Costello & Osborne, 2005).

.06 −.05

Avoidance

Mother (n = 1403)

Anxiety

Avoidance

Best friend (n = 1999)

Pattern matrix* for ECR-RS factors across best friend, mother, father, and parental figure domains.

1. It helps to turn to this person in times of need. 2. I usually discuss my problems and concerns with this person. 3. I talk things over with this person. 4. I find it easy to depend on this person. 5. I don’t feel comfortable opening up to this person. 6. I prefer not to show this person how I feel deep down. 7. I often worry that this person doesn’t really care for me. 8. I’m afraid that this person may abandon me. 9. I worry that this person won’t care about me as much as I care about him/her.

Table 1.

Attachment & Human Development 65

66 Table 2.

D.F. Donbaek and A. Elklit ECR-RS dimensions within and across best friend and parental figure domains. Best friend Avoidance

Best friend Avoidance Anxiety Parental figure Avoidance Anxiety Descriptives Min/max Mean Standard deviation Skewness (SE)

Parental figure Anxiety

.83* .40**

.88*

.34** .22**

.21** .40**

6/42 13.98 7.36 .95 (.06)

3/21 6.12 4.31 1.50 (.06)

Avoidance

.83* .31** 6/42 15.85 7.59 .76 (.06)

Anxiety

.86* 3/21 4.26 3.19 3.18 (.06)

Notes: *Cronbachs alpha; **Pearson’s correlation coefficient r is significant at the .01 level (two-tailed).

(e.g., best friend and parental figure avoidance subscales). Small correlations were obtained between subscales across domains (e.g., parental figure avoidance and best friend anxiety). The attachment anxiety and avoidance subscales seemed to exhibit clear independence; the evidence does show them to be interrelated constructs, but with little within-person variability across best friend and parental figure domains. Significant positive skewness for both subscales across domains indicated a general lack of normal distribution of scores, suggesting that the majority of participants had secure attachment structures. This observed pattern corresponds theoretically and empirically well to earlier studies in non-clinical adult samples (Brennan et al., 1998; Fraley et al., 2000, 2011).

Convergent and discriminant validity of the ECR-RS The remaining construct validation was based on the avoidance and anxiety dimensions from the parental figure domain. The ECR-RS dimensions were expected to demonstrate both convergent and discriminant relationships with RQ (Bartholomew & Horowitz, 1991) prototype ratings. Because ECR-RS dimensions were moderately correlated, standard multiple regressions were used to assess the contribution of the four RQ pattern ratings, which were entered in the model as independent variables (secure, preoccupied, dismissing, and fearful) to account for ECR-RS avoidance (Table 3) and anxiety (Table 4) in separate analyses, while controlling for the other ECR-RS dimension. The regression models for avoidance Table 3.

Multiple regression model: RQ patterns predictors of ECR-RS avoidance.*

Secure Preoccupied Dismissing Fearful Anxiety Constant

B

SE

t

p

−1.91 −.21 .41 .72 .33 21.98

.09 .11 .07 .11 .05 .68

−22.47 −1.96 5.56 6.61 7.09 32.16

A validation of the Experiences in Close Relationships-Relationship Structures scale (ECR-RS) in adolescents.

Emerging evidence points toward a two-dimensional attachment construct: avoidance and anxiety. The Experiences in Close Relationships-Relationship Str...
386KB Sizes 0 Downloads 0 Views