Attachment Avoidance and Anxiety as Predictors
of Response to Treatment with an Emotionally Focused Therapy
Program for Groups of Couples.
Magia Gkourtsogianni
University of Amsterdam
Master Thesis of Clinical Psychology Date: 11 February 2015
Student number: 6060404 Supervisor: Dr. H.J. Conradi
Table of Contents Abstract 3 Introduction 4 Method 10 Results 14 Discussion 24 References 30
Abstract
The current study examines the ability of adult attachment dimensions to predict response to treatment to an emotionally focused therapy program for groups of couples. It is hypothesized that when controlling for initial levels of relationship adjustment, higher levels of both actor attachment anxiety and avoidance can predict lower levels of relationship adjustment in response to the
program, for both men and women yet with a greater negative impact of avoidance for men. Both partner attachment anxiety and avoidance are also anticipated to have a negative influence on post-treatment relationship adjustment, with partner anxiety expected to have a greater effect regardless of gender. A sample of 83 adult heterosexual couples completed self-report measures of adult attachment dimensions and relationship adjustment before and after their participation to the couples program. Results indicate that the predictive value of attachment depended on gender and mental-health status. Increased levels of actor and partner attachment anxiety predicted a
substantial decrease in relationship adjustment at termination of treatment but only for women recruited from clinical (mental-health) practices. No effects were found for clinically recruited men. For non-clinically recruited men, higher actor avoidance predicted a moderate decrease in response to treatment. Unexpectedly, for non-clinically recruited women, higher actor attachment anxiety predicted a small increase in relationship adjustment at post-treatment. Theoretical and clinical implications of findings are discussed as well as methodological limitations.
Introduction
Emotionally Focused Therapy for couples (EFTC) is an evidence-based, systemic intervention that focuses on modifying distressed couples' inflexible interaction patterns and emotional responses and fostering the development of a secure emotional bond (Greenberg & Johnson, 1988; Johnson, 1966). Aside from EFT for individual couples, a protocolled psychoeducational program for groups of couples has been created, 'Hold Me Tight' (Johnson, 2008). The program is based on the theory and practice of EFTC and aims at educating couples on how to enhance their relationship by improving their emotional attachment connection. The Hold Me Tight (HMT) program has been translated and adapted in Dutch (Eekhoudt, Aarnoudse, & Nuland, 2010) and was recently
introduced in The Netherlands as a preventive group intervention through which couples can (re-) establish a firm emotional bond and prevent (further) disconnection.
A multitude of empirical studies has demonstrated the effectiveness of individual EFTC in reducing marital distress and enhancing relationship adjustment (for a review see Johnson, Hunsley, Greenberg, & Schindler, 1999; Johnson & Wittenborn, 2012), however to date no
research has been completed on either the effectiveness of, or response to treatment with the HMT course for groups of couples. The current study aims at investigating possible predictors of
response to treatment. Specifically, this study will assess whether individual differences in attachment anxiety and avoidance can predict levels of relationship adjustment at termination of treatment.
Both EFTC and the HMT program approach adult romantic relationships from an attachment perspective. Initially formulated by Bowlby (1982/1969) and empirically examined by Ainsworth
(Ainsworth, Blehar, Waters, & Wall, 1978) attachment theory states that seeking and maintaining close contact with a few irreplaceable and supportive others (one's attachment figures) is a primary motivating principle in human beings and an innate survival mechanism in a potentially dangerous world. Attachment figures ideally provide the individual with a physical and emotional safe haven by offering support and comfort during stress or danger as well as a secure base from which the individual can explore the world autonomously, trusting that care and support will be available if needed. While the attachment system plays a critical role during the early phases of life, the system is assumed to be active over the entire life span (Bowlby, 1988). Hazan and Shaver (1987) were the first to study adult romantic love from an attachment perspective, stimulating an extensive body of research which demonstrates that romantic relationships do function in psychological similar ways as infant-caregiver relationships (for a review see Fraley & Shaver, 2008).
Attachment relationship dynamics are best elucidated in Mikulincer and Shaver's (2003) model, which integrates recent research findings and earlier theoretical work by Bowlby (1973, 1980, 1982/1969), Ainsworth (1991), Cassidy and Kobak (1988) and Main (1995). According to this model the attachment system is an adaptive regulatory device that becomes activated when a person perceives a threat to his or her own security or to the security of the attachment bond. This activation leads the individual to adopt the primary attachment strategy, i.e. seeking emotional and/or physical proximity to the attachment figure. If the attachment figure is perceived to be available, responsive, and supportive, the individual feels secure and comforted, bonding is
strengthened and the attachment system eventually turns off, allowing the person to engage in non-attachment activities such as exploration or caregiving. Repetition of this interaction pattern leads to secure attachment. However, if the attachment figure is perceived to be unavailable,
unresponsive, or overprotective, the individual feels insecure and emotional distress accumulates. This leads to the adoption of one or both of the secondary attachment strategies aiming at
unsure option, the individual will adopt the hyperactivating secondary strategy, that can be seen as a “fight” fear response: the person will make very insistent and energetic attempts to attain support, will exaggerate the appraisal of threats and become hyper-vigilant regarding the (un-) availability of the attachment figure. When proximity seeking is not a viable option, the individual eventually gives up on the quest for proximity and may adopt the deactivating secondary strategy, i.e. a “flight” or “freeze” response that suppresses the attachment system. This includes blocking the monitoring of attachment-figure availability, denial of attachment needs, and the pursuit of excessive self-reliance. In the long run, secondary strategies may be consolidated as the main regulatory device accumulating in insecure attachment: hyperactivation into anxious (or
anxious-ambivalent or preoccupied) attachment and deactivation into avoidant attachment.
The EFTC model assumes that the powerful negative emotions and dysfunctional
interactional cycles that are typical of couples in distress reflect secondary strategies in the struggle of partners for attachment security (Johnson & Whiffen, 1999). In the face of separation distress, individuals may persistently attempt to control their partner's responses in the direction of increased availability and responsiveness (hyperactivation) or to rigidly manage overwhelming distress by withdrawing from attachment needs (deactivation). However when partners become increasingly distrustful, demanding, or detached, attachment security is further undermined in a cyclical fashion (Johnson & Whiffen, 1999). EFTC identifies three types of self-reinforcing, negative interaction patterns that are common in couples, termed the Demon Dialogues (Johnson, 2008). The first pattern is one of mutual blame and demand, reflecting the habitual use of hyperactivating strategies by both partners, the second is a pattern of demand-withdrawal where one partner typically
hyperactivates while the other deactivates, and the third pattern is one of mutual withdrawal, where both partners employ deactivating strategies. Empirical research (Gottman, 1991) has indeed demonstrated the power of negative affect and destructive interaction cycles in predicting long-term dissatisfaction and instability of romantic relationships. Importantly, withdrawal (Heavey,
Christensen, & Malamuth, 1995) and emotional disengagement (Gottman, 1994), cf. deactivation, appear to be more reliable predictors of marital dissatisfaction and divorce than negative
engagement and criticism, cf. hyperactivation.
The chronic consolidation of attachment dynamics tends to give rise to more enduring individual differences in attachment functioning. In the long run, a person’s habitual use of secondary strategies may be consolidated into an anxious and/or avoidant attachment disposition, which manifests itself across relationships with different partners (or other close relationships). An insecure attachment disposition is thought to bias interpersonal perception, leading to rigid
emotional responses and behaviors that create or intensify relational problems (Mikulincer & Shaver, 2003; Johnson & Whiffen, 1999). For example, research has shown that people with an avoidant attachment disposition tend to be distant, untrusting and low in emotional involvement in their relationships, while people with an anxious attachment have relationships defined by worry of abandonment, jealousy, anger and an intensification of negative emotions in general (Hazan & Shaver, 1987; Feeney, 1999). Similarly, when dealing with relationship conflict, anxious people tend to fight aggressively and to escalate conflict while avoidant partners tend to downplay the importance or presence of interpersonal problems (Mikulincer, 1998a).
Not surprisingly, individual differences in attachment quality influence adult relationship satisfaction. An extensive review (Mikulincer & Shaver, 2007) of previous studies on actor effects, i.e. the impact of one’s own characteristics on one’s own outcomes, has demonstrated a negative association between attachment anxiety and avoidance and relationship satisfaction, yet the association with avoidance appears to be more consistent for men. Considering the inherent systemic nature of attachment relationships, it makes sense to assume that the frequent display of negative emotions or avoidant behaviors characteristic of insecure attachment, are equally likely to have a negative impact on the satisfaction of one’s partner as well. Based on this assumption,
investigated (for a review see Mikulincer & Shaver, 2007). Taken together, findings suggest that individuals who have a partner with higher levels of either attachment anxiety or avoidance tend to be more dissatisfied in relationships. In contrast to actor effects however, a partner's anxiety appears to be a more consistent predictor of dissatisfaction than avoidance, regardless of gender. It is important to note, that the majority of studies on either actor or partner effects have been cross-sectional, with the exception of only a handful of prospective longitudinal studies. In the present study a prospective approach is applied, in the context of response to treatment.
Research on attachment as predictor of response to treatment with couple’s therapy is very scarce. One study (Conradi, De Jonge, Neeleman, Simons, & Sytema, 2011) found in response to eclectic couple’s therapy, a main effect of actor attachment anxiety on severity of psychological complaints, and actor attachment avoidance on relationship problem-solving capacity. Moreover, partners with higher levels of avoidance initially improved at their relational problem-solving capacity but showed a significant deterioration at follow-up.
A single study has investigated attachment as predictor for response to treatment to EFTC (Johnson & Talitman, 1997) measuring different aspects of attachment quality by means of the Attachment Questionnaire (AQ; West, Sheldon, & Reiffer, 1987). Interestingly, it was found that couples that made most gain in relationship satisfaction were those that included men who were initially reluctant to turn to spouses for contact and support (high on avoidance). Men with higher levels of proximity seeking (low on avoidance) at intake were also most likely to be classified as non-distressed at termination. Female attachment quality was unrelated to relationship satisfaction at post-treatment though the authors suggest that this might have been due to a limited range of females' scores on this measure. While these preliminary results are hopeful, high attachment vulnerability is considered a challenge for the process of EFTC. Highly insecure partners’ modes of information processing and affective expression appear to pose obstacles for the modification of attachment strategies through therapy (Johnson & Whiffen, 1999). EFT therapists are therefore
advised to adapt their interventions in various ways in order to meet the needs of such vulnerable clients. For example, therapists are advised to particularly focus on reassuring anxious partners by structuring sessions appropriately and by validating their experiences. Equally, it seems important to address explicitly the skepticism and reservations that avoidant partners tend to have, which restrict their full engagement in the therapeutic process (Johnson & Whiffen, 1999). The
educational, low-intensity and protocolled format of the HMT program however offers restricted opportunities for such targeted tailoring. Lacking adequate support, highly insecure partners are unlikely to change their affective responses and enhance their relationship adjustment while following the program. Higher attachment avoidance and anxiety of both actor and partner may therefore predict lower response to treatment to the HMT program.
In the present study the levels of relationship adjustment of a sample of adult couples were assessed before and after their participation to an 8-week Dutch HMT program, by means of a self-report questionnaire. Also, prior to treatment the self-self-reported attachment anxiety and attachment avoidance of both partners were measured. Because pre-treatment relationship adjustment has been found to be a consistent predictor of post-treatment relational satisfaction in behavioral marital therapy (Jacobson, Folette, & Pagel, 1986), insight oriented therapy (Snyder, Mangrum, & Wills, 1993) and EFTC (Johnson & Talitman, 1997), pre-treatment relationship adjustment will be
included in the analyses. In this way it is possible to estimate the unique contribution of attachment as predictor beyond that of initial relationship adjustment. It is hypothesized that higher levels of both actor attachment anxiety and avoidance can predict lower levels of relationship adjustment in response to the program for both men and women, yet with a greater negative impact of avoidance for men. Both partner attachment anxiety and avoidance are anticipated to have a negative
influence on post-treatment relationship adjustment as well, with partner anxiety expected to have a greater effect.
Method
Participants
One hundred and three adult heterosexual couples participated in the study. Participants were required to master the Dutch language to be included. Exclusion criteria were a current DSM-IV axis I or II diagnosis. Couples were recruited through advertising, the Dutch EFT network website (www.eftnetwerk.nl) as well as clinical practices in the Dutch EFT network. Regarding demographic data, two participants did not fill in their age while four participants did not fill in all additional demographic information. For the rest of the sample, average age was approximately 43 years (M = 42.94; SD = 9.47; range = 22-68). Average relationship length was approximately 16 years (M = 15.93; SD = 9.59; range = 1.67-41.25). With regard to relationship status, 5.9% of participants were dating and living apart, 24.3% were living together and 69.8% were married. With regard to parenthood, 71.8% of participants had children while 28.2% did not. Considering educational attainment, 2% of participants had elementary school education, 15.5% had high school education (ranging form Dutch MAVO to VWO level), 21.4% middle-level applied education (Dutch MBO level), 31.3 % higher-level professional education (Dutch HBO level), 28.9% university education, and 1% reported ‘other’ education type.
Materials
Predictor measures: Relationship Adjustment.
Relationship adjustment was measured at pre-treatment as the total score of each participant on a Dutch translation (for the purpose of this study) of the Dyadic Adjustment Scale (DAS; Spanier, 1976). The DAS is a self-report measure and internationally the most widely used outcome
measure in couple’s therapy. It is composed of 32 items and 4 subscales: Dyadic Consensus, Dyadic Satisfaction, Affectional Expression, Dyadic Cohesion. The total score ranges from 0 to 151, with a higher score indicating greater relationship adjustment. The measure has excellent internal consistency as a recent meta-analysis (Graham et al. 2006) of the original measure reported a mean Cronbach’s α of .92 for the total DAS score.
Predictor measures: Attachment Anxiety & Avoidance.
Individual differences in attachment anxiety and avoidance were measured at pre-treatment as the total scores of the two respective subscales of the Dutch version of the Experiences in Close
Relationships questionnaire (ECR; Conradi, Gerlsma, Van Duijn, & De Jonge, 2006). The ECR is a
self-report questionnaire measuring adult attachment quality on the basis of both past and present romantic relationships. It contains 36 items in total and 2 subscales (with 18 items each) that reflect the two fundamental attachment dimensions: anxiety about rejection and abandonment (Anxiety) and avoidance of intimacy (Avoidance). A higher score on the Anxiety scale represents higher attachment anxiety, while a higher score on the Avoidance scale represents greater attachment avoidance. The two scales were originally constructed by means of a major factor analysis of all previously available self-report measures of adult attachment (ECR; Brennan, Clark, & Shaver, 1998). Conceptually, the Anxiety dimension reflects the degree of use of hyperactivating
attachment strategies while the Avoidance dimension indicates the use of deactivating attachment strategies. The internal consistency of the Dutch version of both subscales is good, with a
Cronbach’s α for Anxiety equal to .88 and for Avoidance equal to .86.
Outcome measure: Relationship Adjustment.
Relationship adjustment was measured as the total score on the DAS at post-treatment, i.e. right after the last session of the HMT course.
Procedure and Design
The Couples Intervention
The protocolled group intervention was based on the official Dutch translation and adaptation of the HMT program by Johnson (2008). The program was provided by Dutch EFT therapists.
Adherence to protocol was assessed by three independent ratters who scored random, therapy audio excerpts guided by structured checklists. The implementation check showed that therapists
demonstrated 90% adherence to protocol. Inter-ratter reliability was equal to 88%.
The HMT program is composed of 8 weekly 2-hour sessions and provides participants with psychoeducation, video examples of partner interactions and role-play exercises to be practiced both during sessions and at home.The main goal of the program is enhancement of partners' reciprocal: a) availability b) responsiveness regarding attachment anxieties and needs, and c) engagement. The following subjects are covered in the following order by the program: (1) psycho-education about love and attachment, (2) identification of dysfunctional interactional patterns between partners (termed the demon dialogues), (3) identification of reciprocal attachment
vulnerabilities underneath dysfunctional interactions (4) resolving misunderstandings and creating a secure attachment base, (5) rehearsing open and responsive communication, (6) forgiving injuries and restoring trust, (7) physical tenderness and sexuality (8) on-going care for the relationship.
Design
The current study utilized a within-subjects design that comprised of measurements at pre-treatment, i.e. right before session 1 of the eight-week HMT course, and a post-treatment measurement right after session 8, i.e. the last session of the couples program.The self-report
questionnaires were provided to participants by the trainers of the HMT program and participants were asked to fill in the questionnaires in private, i.e. without discussing the questions and answers with their partner.
Data Analysis
Taking into account the dependence of data within the couples of participants, as well as the hypothesis regarding gender differences for the effect of actor avoidance, data were divided by gender in order to be analyzed separately. Further, because certain participants were recruited from mental-health practices (N=80) it was assumed that they might be part of a distinct population characterized by vulnerability for psychopathology. Because of the well-documented association between insecure attachment and psychopathology in literature (for a review see Mikulincer & Shaver, 2007) it was suspected that in this group, attachment vulnerabilities might play a different, possibly more prominent role. To explore these possible differences, the sample was also divided in terms of mental-health status (non-clinically recruited vs. clinically recruited) next to gender, which resulted in four sub-groups to be analyzed separately. To test hypotheses, a series of hierarchical multiple linear regression analyses was conducted. This choice was based on the presence of more than two, continuous predictor variables: pre-treatment Relationship Adjustment, actor attachment Avoidance, actor attachment Anxiety, partner attachment Avoidance and partner attachment Anxiety, and one continuous outcome, i.e. post-treatment Relationship Adjustment. The partner attachment variables were created by matching, per couple, each participant’s actor score to that of their partner’s, which resulted in each actor score serving as the corresponding partner’s partner score. The IBM SPSS (20.0.0) statistics software was used for all analyses.
Results
From the 206 participants, forty did not complete all measurements required for the analyses, either because they dropped-out of the course, missed the session where the measurement took place or failed to fill in the questionnaires in an adequate manner. Specifically, 16 participants did not complete the measurements of attachment at pre-treatment; 19 did not fill in the questionnaires of relationship adjustment at pre-treatment while 38 did not complete the same measurement at post-treatment. The data of the remaining 166 participants were used for the main analyses.
Descriptive statistics of major variables and exploratory pair-wise comparisons
Initially, the means and standard deviations of participants’ scores on pre-treatment actor Avoidance and Anxiety and pre- and post-treatment Relationship Adjustment were calculated, see Table 11. Next, a series of exploratory analyses of variance (one-way ANOVA’s) was conducted in order to compare these group means and detect any meaningful, statistically significant differences, see again Table 1. The significance level was equal to .05. Regarding the assumptions for ANOVA, for actor Avoidance variances were significantly different among the four groups, F(3, 162) = 2.644, p = .051 and the same was true for post-treatment Relationship Adjustment F(3, 162) = 3.821, p < .05. For these variables Welch’s (1951) F-ratios are reported instead, which are robust when homogeneity of variance is violated. Further, within non-clinically recruited men, the distribution of scores for actor Avoidance deviated from normality D (50) = .125, p = .051, while the same violation was observed for the scores of clinically-recruited men on post-treatment Relationship Adjustment, D (32) = .174, p < .05.
1The reader may want to notice that for clinically recruited participants, the mean actor scores of men on attachment are not identical to the mean partner scores of women and vice versa, due to the small difference in sample size
The results of the ANOVA showed no overall effect for participant group on actor Avoidance, yet this finding might be biased because of the previously mentioned violation of the assumption of normality. There was an overall effect found for participant group on actor Anxiety. A Tukey post-hoc test revealed that non-clinically recruited men scored significantly lower on actor Anxiety compared to each of the other three groups. The post-hoc test also showed a marginally significant difference between the actor Anxiety scores of the two groups of women: clinically recruited women scored higher than non-clinically recruited women. Further, no overall effect for participant group was found on Relationship Adjustment at pre- or post-treatment (yet the later result might be biased due to the violation of normality).
Table 1
Means and Standard Deviations of Major Variables and Exploratory Pair-wise Comparisons Between Groups
Non-clinically Recruited Clinically Recruited Men (1) N=50 Women (2) N=50 Men (3) N=32 Women (4) N=34 Variable M (SD) M (SD) M (SD) M (SD) ANOVA/ Welch’s F Post-hoc Tukey test Actor Avoidance (pre-treatment) 53.66 (14.57) 53.80 (17.39) 56.41 (13.83) 60.35 (23.46) F = .904 P = .443 Actor Anxiety (pre-treatment) 57.06 (16.81) 67.22 (15.17) 69.87 (18.71) 76.35 (16.62) F = 9.754 P = .000 1 < 2*; 1 < 3*; 1 < 4* 2 < 4† Partner Avoidance (pre-treatment) 53.80 (17.39) 53.66 (14.57) 61.08 (23.68) 54.91 (14.85) F = 1.430 P = .236 Partner Anxiety (pre-treatment) 67.22 (15.17) 57.06 (16.81) 75.88 (17.03) 68.69 (19.38) F = 8.635 P = .000 2 < 1*; 2 < 3*; 2 < 4* Relationship adjustment (pre-treatment) 101.63 (13.65) 95.83 (15.70) 97.19 (21.03) 92.63 (19.23) F = 2.046 P = .109 Relationship adjustment (post-treatment) 107.79 (12.39) 105.18 (12.37) 102.99 (17.49) 97.31 (20.28) F = 2.571 P = .06
Note. For the variables of actor Avoidance and post-treatment Relationship Adjustment, Welch’s F-ratios are reported due to a violation of homogeneity of variance.
Zero order correlations of study variables
Prior to running the main regression analyses, zero-order correlations were computed among all major variables for each participant group, see Tables 2 & 3. As expected within all four groups pre- and post-treatment Relationship Adjustment correlated highly positive. Further, in line with expectations, actor Avoidance at pre-treatment showed a statistically significant negative correlation with
Relationship Adjustment at post-treatment for both non-clinically recruited men and women and marginally for men, but not for women who were clinically recruited. ActorAnxiety correlated negatively with post-treatment Relationship Adjustment only for non-clinically recruited men as well as clinically recruited women. Further, partner Avoidance was negatively correlated with
post-treatment Relationship Adjustment only for non-clinically recruited men and women.Partner Anxiety showed a negative correlation with post-treatment Relationship Adjustment only for women of both mental-health statuses. Finally, correlations between predictor variables were examined. All
statistically significant correlations among predictor variables were found to be weak to moderate in strength, falling below r = 0.5. As a first inspection this observation suggests no violation of the assumption of multicollinearity.
Table 2
Zero Order Correlations Among Major Variables for Non-Clinically Recruited Participants
Non-Clinically Recruited Men (N=50) Pre-treatment Rel. Adjustment Post-treatment Rel. Adjustment Actor Avoidance Actor Anxiety Partner Avoidance Partner Anxiety Pre-treatment Rel. Adjustment - .722*** -.332** -.441** -.225† .114 Post-treatment Rel. Adjustment - -.509*** -.332** -.305* .169 Actor Avoidance - .109 .216† .048 Actor Anxiety - .306* -.149 Partner Avoidance - -.031
Non-Clinically Recruited Women (N=50) Pre-treatment Rel. Adjustment Post-treatment Rel. Adjustment Actor Avoidance Actor Anxiety Partner Avoidance Partner Anxiety Pre-treatment Rel. Adjustment - .735*** -.389** -.182 -.379** -.301* Post-treatment Rel. Adjustment - -.327* .086 -.311* -.355** Actor Avoidance - -.031 .216† .306* Actor Anxiety - .048 -.149 Partner Avoidance - .109 †p < .10; *p < .05; **p < .01; ***p < .001
Table 3
Zero Order Correlations Among Major Variables for Clinically Recruited Participants
Clinically Recruited Men N=32 Pre-treatment Rel. Adjustment Post-treatment Rel. Adjustment Actor Avoidance Actor Anxiety Partner Avoidance Partner Anxiety Pre-treatment Rel. Adjustment - .906*** -.216 .319 -.262 -.166 Post-treatment Rel. Adjustment - -.271† -.138 -.144 -.176 Actor Avoidance - .319* .060 .048 Actor Anxiety - .186 -.247† Partner Avoidance - -.119
Clinically Recruited Women N=34 Pre-treatment Rel. Adjustment Post-treatment Rel. Adjustment Actor Avoidance Actor Anxiety Partner Avoidance Partner Anxiety Pre-treatment Rel. adjustment - .489** -.295* -.267† -.349* -.209 Post-treatment Rel. adjustment - -.130 -.506** - .014 -.303* Actor Avoidance - -.133 .083 .243† Actor Anxiety - -.002 -.262† Partner Avoidance - .337* †p < .10; *p < .05; **p < .01; ***p < .001
Variance explained by pre-treatment Relationship Adjustment and Attachment
To determine the degree in which variance in post-treatment Relationship Adjustment could be explained by the attachment variables over and above Relationship Adjustment at pre-treatment, a series of hierarchical multiple regression analyses were conducted for each sub-sample of participants. In the first series of analyses, pre-treatment Relationship Adjustment was entered in the first step and the attachment variables in the second step, see Table 4. An examination of case-wise diagnostics
revealed no outliers or otherwise influential cases. All assumptions for multiple regression were met2. The results of the analysis indicated that pre-treatment Relationship Adjustment could explain a large and significant amount of the variance in post-treatment Relationship Adjustment. Adding the
attachment variables in the second step, resulted in a significant increase in variance explained for non-clinically recruited men and clinically recruited women. The attachment variables appeared to add nothing to the amount of variance explained for non-clinically recruited women and clinically
recruited men. In a second series of multiple regressions the order of the steps was reversed, see again Table 5. It should be noted that here the assumption of normality was violated in the regression model of the first step (attachment variables) for certain participant groups: the distribution of standardized residuals deviated from normality for non-clinically recruited women, D (50) = .144, p < .05 and clinically recruited men, D (32) = .162, p < .05.Regarding the results, the attachment variables entered alone in the first step could explain a considerable significant amount of the variance in the outcome, for all participant groups except for clinically recruited men. As expected, adding pre-treatment Relationship Adjustment in the second step resulted in an additional increase of explained variance for all four groups. Summed up, results show that within non-clinically recruited participants, both pre-treatment Relationship Adjustment and attachment could explain a substantial amount of the variance in the outcome. Within clinically-recruited participants on the other hand appears a contrast: for clinically-recruited men the attachment variables as a whole appeared to have no predictive value, while for clinically recruited women, the attachment variables accounted for a larger amount of the variance in the outcome compared to pre-treatment Relationship Adjustment.
2An initial visual inspection of relevant plots (standardized residuals against standardized predicted values) for both
groups of females suggested possible heteroscedasticity in the data. Running a Koenker–Basset (1982) test of
Table 4
Proportion of Variance Explained in Stepwise Regression on Relationship Adjustment at Post-Treatment
Non-Clinically Recruited Clinically Recruited
Men N=50 Women N=50 Men N=32 Women N=34 Predictors stepwise (step 1 and 2) R2 ΔR R2 ΔR R2 ΔR2 R2 ΔR2
1a: Relationship Adjustment (pre-treatment) .521** .540** .821** .239**
2a: Attachment variables (pre-treatment) .625** .105* .600** .060 .837** .016 .572** .333**
1b: Attachment variables (pre-treatment) .375** .236* .129 .493**
2b: Relationship Adjustment (pre-treatment) .625** .251** .600** .364** .837** .708** .572** .079* model 1a, predictor: Relationship Adjustment.
model 2a, predictors: Relationship Adjustment, actor Avoidance, actor Anxiety, partner Avoidance, partner Anxiety. model 1b, predictors: actor Avoidance, actor Anxiety, partner Avoidance, partner Anxiety.
model 2b, predictors: actor Avoidance, actor Anxiety, partner Avoidance, partner Anxiety, Relationship Adjustment. R2 = R square; ∆R2 = change in R square; †p < .10 *; p < .05; ** p < .01
Individual predictors of post-treatment Relationship Adjustment
In order to test hypotheses, the value and significance of the beta’s of each predictor were regarded, when all five predictors were entered into the model, see Table 5. As expected, initially higher levels of Relationship Adjustment predicted higher levels of the same measure at post-treatment for all participant groups. In line with hypotheses, for men who were non-clinically recruited, higher levels of actor Avoidance at pre-treatment predicted lower levels of Relationship Adjustment at termination, over and beyond the effect of pre-treatment Relationship Adjustment.Contrary to hypotheses however no other attachment variables had any individual predictive value for these male participants.For non-clinically recruited women, none of the hypotheses were supported yet there was an unexpected significant finding: higher levels of actor Anxiety at pre-treatment predicted higher levels of Relationship Adjustment at post-treatment. Findings offered no support for hypotheses regarding clinically recruited men. For clinically recruited women, actor and partner Anxiety predicted lower levels of Relationship Adjustment at termination of treatment with both predictors having similar
predictive value, in line with predictions. Attachment Avoidance had for this participant group no effect.
Additionally, the beta’s of the attachment predictors were explored for the model (outlined in Table 4, step 1b) in which pre-treatment relationship adjustment was not included, see Table 6. Here, a similar pattern of significant predictors can be observed as in the main five-predictor model, except for the group of non-clinically recruited women, where partner Avoidance and Anxiety emerge as
marginally significant predictors with a negative value, while actor Anxiety is not significant.
Table 5
Betas of All Predictors at Pre-Treatment on Relationship Adjustment at Post-Treatment
Non-Clinically Recruited Clinically Recruited
Men N=50 Women N=50 Men N=32 Women N=34 Predictor β p β p β p β p Relationship Adjustment .591 .000 .730 .000 .912 .000 .336 .031 Actor Avoidance -.296 .005 .001 .993 -.075 .386 -.012 .927 Actor Anxiety .012 .915 .205 .045 -.014 .875 -.537 .001 Partner Avoidance -.109 .278 -.034 .747 .101 .240 .257 .074 Partner Anxiety .114 .231 -.101 .337 -.012 .887 -.457 .003 Table 6
Betas of Attachment Predictors at Pre-Treatment on Relationship Adjustment at Post-Treatment
Non-Clinically Recruited Clinically Recruited
Men N=50 Women N=50 Men N=32 Women N=34 Predictor β p β p β p β p Actor Avoidance -.464 .000 -.193 .174 -.224 .251 -.108 .435 Actor Anxiety -.217 .090 .053 .691 -.091 .652 -.652 .000 Partner Avoidance -.134 .295 -.243 .076 -.138 .460 .162 .259 Partner Anxiety .155 .202 -.262 .065 -.204 .286 -.502 .002
Explained variance and associations with pre-treatment Relationship Adjustment
To close, a series of exploratory, cross-sectional regression analyses was conducted, with attachment variables at pre-treatment as predictors (entered simultaneously) and Relationship Adjustment at pre-treatment as the outcome, see Table 7. The assumptions for linear regression were met with one exception: the standardized residuals of non-clinically recruited women were non-normally distributed, D (56) = .160, p < .05.The results of the analysis indicated that, with the exception of clinically recruited men, for all participant groups attachment could explain a
significant amount of the variance in pre-treatment relationship adjustment. The beta coefficients (see Table 8) revealed that for non-clinically recruited men, higher levels of both actor Avoidance and Anxiety predicted lower levels of Relationship Adjustment measured concurrently. For non-clinically recruited women, higher levels of actor Avoidance predicted lower levels of Relationship Adjustment. No other predictors were significant at a .05 level, although for women recruited from clinical practices, marginally significant Beta’s with a negative value were observed for actor Avoidance, actor Anxiety and partner Avoidance.
Table 7
Proportion of Variance Explained in Regression on Relationship Adjustment at Pre-Treatment
Non-clinically recruited Clinically recruited Men N=56 Women N=56 Men N=37 Women N=38 Predictors R2 R2 R2 R2 Attachment variables (pre-treatment) .227* .168* .158 . .254*
Predictors: actor Avoidance, actor Anxiety, partner Avoidance en partner Anxiety. R2 = R square; ∆R2 = change in R square; †p < .10 *; p < .05; ** p < .01.
Table 8
Betas of Attachment Predictors at Pre-Treatment on Relationship Adjustment at Pre-treatment
Non-clinically recruited Clinically recruited
Men N=56 Women N=56 Men N=37 Women N=38 Predictor β p β p β p β p Actor Avoidance -.272 .037 -.284 .038 -.234 .182 -.278 .085 Actor Anxiety -.319 .015 -.136 .300 -.053 .775 -.282 .079 Partner Avoidance -.071 .584 -.181 .176 -.254 .142 -.268 .098 Partner Anxiety .082 .514 -.062 .641 -.161 .345 -.127 .454 Discussion
This study investigated whether romantic partners’ initial levels of attachment anxiety and avoidance, could predict response to treatment to a Dutch EFT program for groups of couples. Results revealed that when controlling for pre-treatment levels of relationship adjustment,
attachment dimensions could predict relationship adjustment at post-treatment, yet the presence and nature of this effect depended on the gender and mental-health status of participants.
The strongest effects for attachment were found regarding women who were recruited from mental-health practices. In line with predictions, higher levels of actor and partner anxiety, predicted lower levels of relationship adjustment at termination of the couples program. These effects where observed beyond the effect of initial levels of relationship adjustment and were in fact stronger. For all participants in the study there was a statistically significant, positive relationship observed between levels of pre- and post-treatment relationship adjustment, yet for clinically recruited women, actor and partner anxiety appeared to have a suppressing effect on this relationship.
relationship adjustment at post-treatment (SD=20.28). Contrary to hypotheses no effect was found for attachment avoidance. Similar patterns of results could also be traced when looking at the
correlations between predictors and outcome independently (zero-order correlations), were actor and partner anxiety were the only variables that showed a negative correlation with post-treatment relationship adjustment.
In contrast and contrary to hypotheses, no effects were detected for any of the attachment variables for the men who were recruited from mental-health practices. The zero-order correlations showed a similar absence of expected effects within this group, except for a marginally significant, negative correlation between actor avoidance and outcome. Hypotheses were on the other hand partially supported for non-clinically recruited men where higher levels of actor avoidance were found to predict lower levels of relationship adjustment at post-treatment, beyond the effect of initial relationship adjustment. No effect was found for the other attachment variables. Specifically, an increase of one standard deviation of actor avoidance (SD=14.58) corresponded to a decrease of approximately 0.3 standard deviations in relationship adjustment (SD=12.39), an effect medium in size. Interestingly, if we examine in comparison the predictive strength of the attachment variables independently (zero-order correlations) we see that for this group, next to actor avoidance also partner avoidance and actor anxiety correlated negatively with the outcome, yet again these effects disappeared in the context of response to treatment (and comparison among predictors).
For non-clinically recruited women no support was found for hypotheses, yet there was a surprising finding: higher levels of attachment anxiety were found to predict higher levels of relationship adjustment at termination of treatment. This effect was small, i.e. an increase of one standard deviation of actor anxiety (SD=15.17) was found to predict an increase of approximately 0.21 standard deviations of post-treatment relationship adjustment (SD=12.37). These findings stand also in contrast to the zero-order correlations for this participant group, where actor avoidance and actor and partner anxiety were all found to negatively correlate to the outcome, in line with
predictions. Once more, the comparison of all predictors in the main analyses rendered these effects insignificant.
The results concerning women recruited from mental-health practices clearly stand out, both in comparison to those of clinically recruited men and non-clinically recruited women. Results may be better interpreted in light of the fact that from the clinically recruited participants, the majority who had been under psychological treatment were women. Specifically, from the clinically recruited men (N=32) less than half were clients themselves (N=15; 47%), while more than three quarters (N=28; 82%) of clinically recruited women (N=34) had been under treatment. This observation may reveal an association between vulnerability for psychopathology and attachment insecurity, an association highlighted extensively in previous literature, both theoretically and empirically. Attachment theory was from the start (Bowlby, 1969/1982, 1973) a theory about psychopathology, arguing that
attachment insecurities, due to their cognitive, emotional and social-interactional aspects, form risk factors for psychopathology. A significant body of both cross-sectional and longitudinal studies have revealed a clear association between attachment anxiety in adults and psychiatric symptomatology, while the association with attachment avoidance is also demonstrated but much less consistently (for a review see Mikulincer & Shaver, 2007). There is also evidence that the relationship is
bi-directional: recent psychological distress appears to aggravate already present attachment insecurities, as a source of stress that naturally activates the attachment behavioral system (for a review see Mikulincer & Shaver, 2007).
In order to investigate the association between psychopathology and attachment insecurity in this study further, it is useful to look at the exploratory pair-wise comparisons between group means on attachment at pre-treatment, and also to compare these means to available norms from a highly representative sample of n=1,019 of the Dutch general population (Conradi, 2010). As already reported, the four participant groups did not differ significantly from each other in terms of actor avoidance. In line with this finding, the norms reveal that all four groups scored above average on
attachment avoidance (stanine 6; 17% of the general population). Regarding actor anxiety on the other hand, the pair-wise comparisons have exposed significant differences between groups. Firstly, non-clinically recruited men scored significantly lower on actor anxiety compared to all the other participant groups. Second, clinically recruited women scored higher on actor anxiety than non-clinically recruited women, and this difference was marginally significant. Looking at the norms we see that non-clinically recruited men scored average on actor anxiety (stanine 5; 20% of the general population), non-clinically recruited women and clinically recruited men scored above average (stanine 6; 17% of the general population), while clinically recruited women scored relatively high (stanine 7; 12% of the general population).
In line with theory, these additional comparisons reveal that clinically recruited women were indeed notably more anxious (with regard to attachment) than the majority of the general population. Therefore, it may be concluded that the relatively high levels of attachment anxiety hindered their response to treatment. The HMT course, being a ‘light’, protocolled group intervention was probably unable to offer these vulnerable women the right tools to deal with their anxiety and improve
relationship adjustment. Additionally, these women had male partners (the clinically recruited men) who were significantly more anxious with regards to attachment compared the partners of non-clinically recruited women. Past research has shown that partner anxiety is a rather consistent predictor of relationship dissatisfaction, more so than avoidance (for a review see Mikulincer & Shaver, 2007) which helps interpret the finding that partner anxiety was almost equally detrimental for the response to treatment of these female clients, as their own anxiety. Also, according to EFT theory, increased levels of anxiety of both partners are assumed to lead to a vicious circle of mutual blame and demand that undermines attachment security further. Under this light, the higher anxiety of both actor and partner put the clinically recruited women in a far more vulnerable position, both on an intrapersonal and dyadic level, compared to the non-clinically recruited women.
It is difficult to interpret the fact that none of the expected effects were found for clinically recruited men, especially since these men had highly anxious female partners which, as explained above, should affect their own response to treatment as well. One possible reason for the lack of observable effects is the small sample size (N=32) combined with the heterogeneity of this male group with regards to mental-health status. Future research with adequate sample size and a clear view on participants’ psychiatric history is needed in order to assess how attachment variables affect the response to treatment of clinically recruited men.
Another unexpected finding concerns non-clinically recruited women, where higher levels actor anxiety actually predicted higher levels of post-treatment relationship adjustment. One possible explanation is that in the beginning of the course, above-average levels of anxiety acted as a
motivating force for these women, fueling their engagement with their partner and the treatment process and leading to increased adjustment by the end of the course. Attachment anxiety is subject of intervention in HMT. One may think that moderate levels of anxiety can be dealt with
successfully in the course but higher levels, as those of the clinically recruited females, may have been too high for the light intervention to tackle. Again, another important element might be that the less vulnerable women had also the least anxious partners compared the rest of the sample.
Relatively secure partners probably offered these women valuable support during the course, and may have helped them to channel their attachment anxiety in healthy and constructive ways,
eventually strengthening the relationship bond. As logical as these speculations may sound however, they remain speculations. Future research could verify such claims and possibly gather additional valuable information via qualitative research methods. Open-ended questions could complement quantitative questionnaires in order to gather information on participants’ experience of the course, including subjective difficulties and helping factors in the treatment process.
A final discussion point flows out of a global comparison between results from the zero-order correlations, the longitudinal analyses, and the exploratory cross-sectional analyses at pre-treatment.
As already mentioned, although certain attachment variables could predict the outcome
independently (for certain participant groups), they could not predict the actual change between pre- and post-treatment relationship adjustment in the main regression analyses. Similarly, when
comparing the results of the main longitudinal analyses with all five predictors, to those of the longitudinal analyses with only the attachment predictors, we observe a change in the pattern of significant predictors for one of the groups (non-clinically recruited women). Even more interesting is to compare the former longitudinal model (attachment predictors on post-treatment relationship adjustment) to the cross-sectional model (attachment predictors on pre-treatment relationship adjustment) where for most participant groups, additional or different attachment predictors emerge as significant. These comparisons should be made with caution however, since, as it has been reported in the results section, the assumption of normality had been violated for some of the exploratory regression analyses. Taken together however, these observations may show that attachment predictors behave differently in the context of a longitudinal design and particularly in the context of response to treatment, compared to cross-sectional designs that form the majority of past research on which the hypotheses of the current study were based. Clearly, future research is needed to further investigate how individual differences in attachment impact the outcome as well as the process of therapeutic treatment.
To summarize, in the current study it has been demonstrated that for psychologically vulnerable women, relatively high anxiety about rejection and abandonment of both actor and
partner, seem to form a rather large hindrance for response to treatment with the Dutch HMT couples program.For men with no recent clinical history, above average levels of avoidance of intimacy appear to pose a moderate obstacle for response to treatment.It may be advised that men and women with such individual characteristics need more intensive treatment such as regular couples EFT therapy, where the therapist can address attachment vulnerabilities more adequately.In contrast, for
women with no recent history of mental-health treatment, mildly elevated levels of attachment anxiety may lightly boost response to treatment with the HMT course.
References
Ainsworth, M. D. S. (1991). Attachment and other affectional bonds across the life cycle. In C. M. Parkes, J. Stevenson-Hinde, & P. Marris (Eds.), Attachment across the life cycle (pp. 33–51). New York, NY: Routledge.
Ainsworth, M. D. S., Blehar, M. C., Waters, E., & Walls, S. (1978). Patterns of attachment:
Assessed in the strange situation and at home. Hillsdale, NJ: Erlbaum.
Bowlby, J. (1973). Attachment and loss: Vol. 2. Separation: Anxiety and anger. New York, NY: Basic Books.
Bowlby, J. (1980). Attachment and loss: Vol. 3. Sadness and depression. New York, NY: Basic Books.
Bowlby, J. (1982/1969). Attachment and loss: Vol. 1. Attachment (2nd ed.). New York, NY: Basic Books.
Bowlby, J. (1988). A secure base: Clinical applications of attachment theory. London, England: Routledge.
Brennan, K. A., Clark, C. L., & Shaver, P. R. (1998). Self-report measurement of adult attachment: An integrative overview. In J. A. Simpson & W. S. Rholes (Eds.), Attachment theory and
close relationships (pp. 46–76). New York, NY: Guilford Press.
Cassidy, J., & Kobak, R. R. (1988). Avoidance and its relationship with other defensive processes. In J. Belsky & T. Nezworski (Eds.), Clinical implications of attachment (pp.300–323). Hillsdale, NJ: Erlbaum.
Conradi, H.J. (2010). ECR voorlopige scoring en normering. Manuscript in preparation.
Conradi, HJ., De Jonge P., Neeleman A., Simons P., & Sytema, S. (2011). Partner attachment as a predictor of long-term response to treatment with couple’s therapy. Journal of Sex and
Marital Therapy, 37, 286–297.
Conradi, H.J., Gerlsma, C., van Dujin, M., & de Jonge, P. (2006). Internal and external validity of the experiences in close relationships questionnaire in an American and two Dutch samples.
European Journal of Psychiatry, 20, 258-269.
Eekhoudt, J. F., Aarnoudse B., & Van Nuland, H. (2010). Houd Me Vast Programma: Gesprekken
voor verbondenheid. Oisterwijk, Nederland: EFT Netwerk Nederland.
Feeney, J. A. (1999). Adult romantic attachment and couple relationships. In J. Cassidy & P. R. Shaver (Eds.), Handbook of attachment: Theory, research, and clinical applications (pp. 355–377). New York, NY: Guilford Press.
Field, A. (2009). Discovering statistics using SPSS (3nd ed.). London, England: Sage.
Fraley, C. R., & Shaver, P. R. (2008). Attachment theory and its place in contemporary personality theory and research. In O. P. John, R. W. Robins, & L. A. Pervin (Eds.), Handbook of
personality: Theory and research (pp. 518-541). New York, NY: The Guilford Press.
Gottman, J. (1991). Predicting the longitudinal course of marriages. Journal of Marital and Family
Therapy, 17, 3-7.
Gottman, J. (1994). An agenda for marital therapy. In S. M. Johnson, & L. S. Greenberg (Eds.), The
heart of the matter: Perspectives of emotion in marital therapy (pp. 256-296). New York,
NY: Brunner/Mazel.
Graham, J. M., Yenling, J.L., & Jeziorski, J.L. (2006). The Dyadic Adjustment Scale: a reliability generalization meta-analysis. Journal of Marriage and Family, 68, 701-717.
Greenberg, L. S., & Johnson, S. M. (1988). Emotionally focused therapy for couples. New York, NY: Guilford Press.
Hazan, C., & Shaver, P. R. (1987). Romantic love conceptualized as an attachment process. Journal
of Personality and Social Psychology, 52, 511–524.
Heavey, C. L., Christensen, A., & Malamuth, N. M. (1995). The longitudinal impact of demand and withdrawal during marital conflict. Journal of Consulting and Clinical Psychology, 63, 797- 801.
Jacobson, N. S., Follette, W. C., & Pagel, M. (1986). Predicting who will benefit from behavioral marital therapy. Journal of Consulting and Clinical Psychology, 54, 518–522.
Johnson, S. M. (1996). The practice of emotionally focused marital therapy: Creating connection. New York, NY: Brunner/Mazel.
Johnson S. M. (2008). Hold Me Tight®: Conversations for Connection Program. Ottawa: International Center for Excellence in Emotionally Focused Therapy.
Johnson, S. M., Hunsley, J., Greenberg, L., & Schindler, D. (1999). Emotionally focused couples therapy: Status and challenges. Clinical Psychology: Science and Practice, 6, 67–79. Johnson, S., & Talitman, E. (1997). Predictors of outcome in emotionally focused marital therapy.
Journal of Marital and Family Therapy, 23, 135-152.
Johnson, S. M., & Whiffen, V. E. (1999). Made to measure: Adapting emotionally focused couples therapy to partners’ attachment styles. Clinical Psychology: Science and Practice, 6, 366- 381.
Johnson, S. M., & Wittenborn, A. K. (2012). New research findings on emotionally focused therapy: Introduction to a special section. Journal of Marital and Family Therapy, 38, 18–22.
Koenker, R., and Bassett, G., (1982). Robust Tests for Heteroscedasticity Based on Regression Quantiles. Econometrica, 50, 43-61.
Main, M. (1995). Attachment: Overview with implications for clinical work. In S. Goldberg, R.Muir, & J. Kerr (Eds.), Attachment theory: Social, developmental, and clinical perspectives (pp. 407–474). Hillsdale, NJ: Erlbaum.
Mikulincer, M. (1998a). Attachment working models and the sense of trust: An exploration of interaction goals and affect regulation. Journal of Personality and Social Psychology, 74, 1209–1224.
Mikulincer, M., & Shaver, P. R. (2003). The attachment behavioral system in adulthood: Activation, psychodynamics, and interpersonal processes. In M. P. Zanna (Ed.), Advances in
experimental social psychology (Vol. 35, pp. 52–153). New York, NY: Academic Press.
Mikulincer, M., & Shaver, P. R. (2007). Attachment in adulthood: Structure, dynamics, & change. New York, NY: Guilford Press.
Snyder, D. K., Mangrum, L. F., & Wills, R. M. (1993). Predicting couples' response to marital therapy: A comparison of short- and long-term predictors. Journal of Consulting and Clinical Psychology, 61, 61–69.
Spanier, G. B. (1976). Measuring dyadic adjustment: New scales for assessing the quality of marriage and other dyads. Journal of Marriage and the Family, 38, 15–28.
Welch, B. L. (1951). On the comparison of several mean values: An alternative approach. Biometrika, 38, 330-336.
West, M., Sheldon, A., & Reiffer, L. (1987). An approach to the delineation of adult attachment: Scale development and reliability. Journal of Nervous and Mental Disease, 175, 738–741.
