THIS ARTICLE HAS BEEN CORRECTED.

SEE LAST PAGE

Psychotherapy © 2013 American Psychological Association

2014, Vol. 51, No. 1, 57– 65 0033-3204/14/$12.00 DOI: 10.1037/a0031100

Change in Attachment Insecurity Is Related to Improved Outcomes 1-Year

Post Group Therapy in Women With Binge Eating Disorder

Hilary Maxwell Giorgio A. Tasca, Kerri Ritchie, Louise Balfour, and

University of Ottawa Hany Bissada
The Ottawa Hospital, Ottawa, Ontario, Canada and the
University of Ottawa

An interpersonal model of binge eating disorder (BED) posits that difficulties with social functioning
precipitate negative affect, which in turn causes binge eating as a means of coping. Thus, long-term
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

decreases in attachment insecurity may be important for women with BED. No research has assessed if
This document is copyrighted by the American Psychological Association or one of its allied publishers.

long-term change in attachment insecurity is associated with sustained change in other outcomes. In the
current study, we hypothesized that changes in attachment anxiety and avoidance will decrease at
posttreatment and will be maintained up to 12 months after Group Psychodynamic Interpersonal
Psychotherapy (GPIP). We further hypothesized that long-term stability of these changes in attachment
insecurity will be related to other long-term outcomes. Women with BED (N ⫽ 102) attended 16 sessions
of GPIP. Measures were completed pretreatment, posttreatment, at 6 and 12 months follow-up. Attach-
ment anxiety, attachment avoidance, and the other outcome variables decreased significantly at 12
months posttreatment. Reductions in attachment anxiety and avoidance were significantly related to
decreases in interpersonal problems up to 12 months posttreatment, and reduction in attachment anxiety
was significantly related to decreases in depressive symptoms 12 months posttreatment. Further, the
significant relationship between reduced attachment avoidance and decreased interpersonal problems
strengthened over the long term. This is the first study to show an association between change in
attachment insecurity and change in other outcomes in the long term, and to show an adaptive spiral in
which greater reduction in attachment avoidance is increasingly associated with ongoing improvement of
interpersonal problems.

Keywords: binge eating disorder, attachment, group psychodynamic interpersonal psychotherapy

Attachment theory proposes that the availability or unavailabil- 1973, 1982; Shaver, & Mikulincer, 2002; Shorey & Snyder, 2006).
ity of an adult attachment figure during times of perceived or real Attachment styles can be categorized as secure and insecure.
threats affects the development of the child’s internal working Individuals with a secure attachment style develop a positive view
models of attachment (Bowlby, 1982). Internal working models of of the self and positive expectations about others’ availability, they
attachment are akin to cognitive-interpersonal schemas that are are able to express and share emotions, adaptively regulate affect,
responsible for the systematic patterns of emotion regulation, and use constructive means of coping (Mikulincer, Shaver, &
expectations, needs, and social behavior that people exhibit in Pereg, 2003; Mikulincer & Shaver, 2007; Shaver & Mikulincer,
close relationships, which in turn define attachment styles (Cobb & 2002). Attachment insecurity includes anxious and avoidant styles.
Davila, 2009). Attachment styles are the patterns of interpersonal Adults with attachment avoidance develop a positive view of
interactions and affect regulation in adulthood that describe how themselves and a negative view of others; these individuals tend to
individuals cope with distress and perceive others (Bowlby, 1969, deactivate their affect system, have difficulty experiencing or
expressing emotions, and have a preference for emotional distance
from others (Mikulincer et al., 2003; Shaver & Mikulincer, 2002).
Adults with attachment anxiety develop a negative view of them-
This article was published Online First February 11, 2013.
Hilary Maxwell, Department of Psychology, University of Ottawa, Ot-
selves and a positive view of others. They hyperactivate their
tawa, Ontario, Canada; Giorgio A. Tasca, Departments of Psychology and affect system, overemphasize negative emotions, have a strong
of Psychiatry, University of Ottawa and The Ottawa Hospital, Ottawa, need for closeness, and have a fear of being rejected (Mikulincer
Ontario, Canada; Kerri Ritchie and Louise Balfour, Department of Psy- et al., 2003; Mikulincer & Shaver, 2007; Shaver & Mikulincer,
chology, The Ottawa Hospital and University of Ottawa; Hany Bissada, 2002).
Department of Psychiatry, The Ottawa Hospital and the University of Attachment styles are relatively stable from childhood to adult-
Ottawa. hood (Shorey & Snyder, 2006; Waters, Merrick, Treboux, Crow-
The original treatment study was funded by a grant from The Canadian
ell, & Albersheim, 2000). However, negative life events or posi-
Institutes for Health Research, Institute of Gender and Health.
Correspondence concerning this article should be addressed to Giorgio tive interpersonal relationships may create change in internal
A. Tasca, Regional Center for the Treatment of Eating Disorders, Ottawa working models of attachment, influencing and ultimately chang-
Hospital–General Campus, Box 400 –501 Smyth Road, Ottawa, Ontario, ing one’s attachment style (Thompson, 2000; Waters et al., 2000).
Canada, K1H 8L6. E-mail: [email protected] Bowlby (1988) introduced the notion that change in one’s internal
57
 58 MAXWELL, TASCA, RITCHIE, BALFOUR, AND BISSADA

working model of attachment may occur through a safe and close 1993). BED is the most common eating disorder with a lifetime
client–therapist relationship. Brennan (1999) argued that this type prevalence of 3.5% among women (Hudson, Hipiri, Pope, &
of change in attachment working models is necessary for psycho- Kessler, 2007), and those with BED tend to have a high level of
therapy to have an impact on the client. Changes in attachment depressive affect (Stice, 1999).
representations may result in changes in relationship schemas and The interpersonal model of binge eating posits that interpersonal
self-concept, and these changes may be related to treatment out- problems lead to negative affect, which in turn triggers binge
comes (Tasca, Balfour, Ritchie, & Bissada, 2007). eating (Ansell, Grilo, & White, 2012). Steiger, Gauvin, Jabalpur-
Studies by Fonagy et al. (1995) and by Levy et al. (2006) found wala, Seguin, & Stotland (1999) found that interpersonal sensitiv-
that patients with borderline personality disorder who were as- ities and interpersonal stressors preceded negative affect and binge
sessed with the Adult Attachment Interview (AAI; George, Ka- eating. Ansell et al. (2012) found that interpersonal problems
plan, & Main, 1985) changed from an insecure to secure attach- resulted in negative affect, which triggered loss of control, binge
ment category after psychodynamic treatment. Travis, Binder, eating, and eating disorder psychopathology. These studies suggest
Bliwise, & Horne-Moyer (2001) measured pre- to posttreatment the potential importance of achieving long-term decreases in in-
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

changes in patient’s attachment styles with Bartholomew and terpersonal difficulties and affect dysregulation for women with
This document is copyrighted by the American Psychological Association or one of its allied publishers.

Horowitz’s (1991) attachment style rating system. A significant BED (Ansell et al., 2012). Both interpersonal difficulties and affect
number of patients in their sample changed from an insecure to a regulation are key concepts related to attachment insecurity. We
secure attachment style. Several group therapy studies have also argue in this study that if attachment insecurity improved and
shown positive change in attachment scales from pre- to posttreat- continued to improve in the longer term, then this could be
ment (e.g., Lawson, Barnes, Madkins, & Francois-Lamonte, increasingly associated with ongoing symptom reduction among
2006). Tasca et al. (2007) examined improvement in attachment women with BED. This is akin to Yalom and Leszcz’s (2005)
insecurity (i.e., attachment anxiety and attachment avoidance) concept of an adaptive spiral in which change in one context fuels
from pre- to posttreatment while comparing two treatment modal- change in another, which in turn potentiates further change in the
ities, group cognitive– behavioral therapy (GCBT; Wilfley, Stein, first, and so on.
Friedman, Beren, & Wiseman, 1996), and group psychodynamic
interpersonal psychotherapy (GPIP; Tasca, Mikail, & Hewitt, Hypotheses
2005) for women with binge eating disorder (BED). They found
significant pre- to posttreatment changes in attachment insecurity The present study has three objectives and hypotheses: (1) to
in both treatment types. Further, changes in attachment anxiety examine whether decreases in attachment anxiety and avoidance is
following GPIP was associated with improved depressive symp- maintained up to 12 months following GPIP for women with BED,
toms, but this relationship was not found among those receiving (2) to examine whether these changes in attachment anxiety and
GCBT. Tasca et al. speculated that GPIP, with its focus on chang- avoidance are related to improvement in other outcomes including
ing relationship patterns and affect regulation, stimulated attach- binge eating, symptoms of depression, and interpersonal problems
ment internal working models and so its impact on attachment in the longer term, and (3) to examine whether the significant
representations might have resulted in a greater association with relationship between reduced attachment insecurity and other out-
depressive symptom outcomes. comes strengthen overtime, thus testing an adaptive spiral hypoth-
Few studies have provided evidence of change in insecure esis.
attachment over a longer period of time, such as 6 and 12 months
post psychological treatment. In a small study of undergraduates, Method
Kilmann, Laughlin, Carranza, Downer, Major, & Parnell (1999)
found that compared with a control group, those receiving a group Participants
intervention reported less fearful and more secure attachment
patterns up to 6 months post intervention. More recently, Kirch- Participants for the current study were 102 women who met
mann et al. (2012) examined changes in attachment characteristics diagnostic criteria for BED (Diagnostic and Statistical Manual of
of inpatients receiving group psychotherapy in routine care includ- Mental Disorders Fourth Edition; American Psychiatric Associa-
ing a nonrandomized control group. Twenty percent of patients tion, 1994). The mean age of this sample was 44.32 years (Stan-
improved from an insecure to secure attachment category at post- dard deviation; SD ⫽ 11.79). Regarding demographics, 89.20%
treatment, and 25% improved at a 1-year follow-up. Kirchmann et (n ⫽ 91) were Caucasian, 27.5% (n ⫽ 28) were single, 48.0%
al. also reported that posttreatment improvements in romantic (n ⫽ 49) were married or cohabiting, and 20.60% (n ⫽ 21) were
attachments were stable at the 1-year follow-up. No study that separated or divorced. The majority was employed, full-time
we are aware of reports an association between long-term reduc- (68.6%, n ⫽ 70) or part-time (14.7%, n ⫽ 15), had attended
tion in attachment insecurity and sustained improvements in other college or university (78.43%, n ⫽ 80), and the median family
outcomes. income was 60,000 to 69,000 in Canadian dollars (in which $1
In the current study, we assessed change in attachment insecu- US ⫽ $1 Canadian). Inclusion criteria for the original treatment
rity following GPIP, whether this change remains stable at 1-year study included the ability to speak and read in English and being
posttreatment, and whether longer-term reduction in attachment overweight (i.e., a body mass index [kg/m2] ⱖ27). Exclusion
insecurity is related to ongoing improvement in other outcomes in criteria included current or past compensatory behaviors (e.g.,
women with BED. BED is characterized by recurrent episodes of vomiting or diuretic use), diagnosis of bipolar or psychotic disor-
overeating during which the individual has a sense of loss of der, drug or alcohol abuse in the past 6 months, taking medication
control and experiences marked distress (Fairburn & Wilson, that may affect weight during treatment, being pregnant or plan-
 CHANGE IN ATTACHMENT INSECURITY 59

ning on becoming pregnant in the following year, and plans to produced (Brennan et al., 1998). The two factors were uncorre-
enroll or current enrolment in a weight-loss program. lated, r ⫽ .11 (Brennan et al., 1998). Shaver, Schachter, and
Six clinicians, including three PhD psychologists, two psychi- Mikulincer (2005) reported the means for a normative sample of
atrists, and one master’s level advanced practice nurse, were the 72 women, Anxiety Mean (M) ⫽ 3.55, SD ⫽ 1.10; Avoidance
therapists. Each therapist had at least 3 years of experience pro- M ⫽ 2.03, SD ⫽ 0.72. The mean interitem correlation for the
viding group psychotherapy and attended a 2-day GPIP workshop current study was 0.44 for the avoidance scale and 0.39 for the
and weekly supervision for quality assurance and manual adher- anxiety scale, indicating good internal consistency (Clark & Wat-
ence. A senior psychologist with 20 years experience in supervi- son, 1995).
sion, eating disorders, and group therapy supervised therapists. The ASQ (Feeney, Noller, & Hanrahan, 1994) was used to place
participants into low and high attachment anxiety conditions based
Procedure on their scores on the Need for Approval scale at pretreatment. The
ASQ is a self-report measure with 40 items on a Likert scale rate
Participants were either patients referred from a tertiary care 1 (totally disagree) to 6 (totally agree). The ASQ has two attach-
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

eating disorders program or self referred through local newspaper ment anxiety subscales, two attachment avoidance subscales, and
This document is copyrighted by the American Psychological Association or one of its allied publishers.

advertisements to a group treatment study for BED. Participants one attachment security subscale. Need for Approval is one of the
were screened over the phone by a research coordinator who attachment anxiety scales and it assesses the need for others’
assessed for exclusion criteria, described the study, and assessed confirmation and acceptance. This scale loaded highly on the
for BED using the Eating Disorder Examination (EDE; Fairburn & attachment anxiety factor of the ECR (Brennan et al.). In the
Cooper, 1993) interview. In all, 230 individuals contacted the current study, participants with scores on Need for Approval
study, 81 were not eligible because of exclusion criteria, and an ⬍3.59 were assigned to the low attachment anxiety condition and
additional 29 declined an in-person assessment. Of the 120 indi- participants with scores of ⱖ3.59 were assigned to the high at-
viduals who met with a research coordinator or doctoral student for tachment anxiety condition. The cutoff of 3.59 was based on
an in-person assessment, five participants met exclusion criteria, findings from a previous study that the interaction between Need
and 13 declined treatment, leaving 102 participants who began for Approval and treatment type (GCBT vs. GPIP) predicted
treatment. Participants provided informed consent to participate in change in posttreatment days binged (Tasca et al., 2006). The point
the research, as well as for the use of research materials to be used of intersection of the regression lines for the two treatment types
for secondary research purposes. This study received Research was 3.59 on Need for Approval. The mean interitem correlation for
Ethics Board approval. the Need for Approval for the current study was 0.34, indicating
All participants were assessed on pretreatment measures, re- good internal consistency (Clark & Watson, 1995).
ceived one pregroup preparation session and 16 weeks of GPIP.
Each therapy group had 5 to 10 members and sessions lasted 90
minutes. As a part of the study design, participants completed the Interpersonal Problems
Attachment Style Questionnaire (ASQ; Feeney, Noller, & Hanra- We used the mean total score of the 64-item Inventory of
han, 1994), and were then assigned to a treatment group based on Interpersonal Problems (IIP; Horowitz, Rosenberg, Baer, Ureño, &
their scores on the Need for Approval scale (see a full description Villaseñor, 1988) to assess distress from interpersonal sources.
of the ASQ and cutoff scores in the Measures section below). Each The IIP is a self-report measure that contains statements describing
of the six therapists conducted two groups, and therapists crossed common interpersonal problems (e.g., It is hard for me to let other
over conditions such that each therapist led both a high and low people know when I am angry). Participants respond using a Likert
attachment anxiety group. Therapists were blind to study condi- scale ranging from zero to four with higher scores reflecting more
tion. Participants completed the same measures at pretreatment, at severe interpersonal problems. Hansen and Lambert (1996) re-
posttreatment, and at 6 months follow-up. At 12 months follow-up, ported means of 0.97 (SD ⫽ 0.48) for a normative sample and 1.48
a research coordinator mailed the same questionnaires and as- (SD ⫽ 0.56) for a clinical sample. The mean interitem correlation
sessed for days binged and self-reported weight over the phone. for the current study was 0.42 indicating good internal consistency
(Clark & Watson, 1995).
Materials
Depressive Symptoms
Attachment
The Beck Depression Inventory-II (BDI-II; Beck, Steer, &
The Experiences in Close Relationships Scale (ECR; Brennan, Brown, 1996) was used to assess depressive symptoms. This
Clark, & Shaver, 1998) is a self-report measure containing 36 self-report measure consists of 21 symptoms and attitudes re-
items about feelings and experiences in romantic relationships. lated to depression that participant’s rate on a four-point Likert
Half of the items measure attachment anxiety and half measure scale ranging from zero to three with higher scores indicating
attachment avoidance. The ECR scales were used in the current greater depressive symptoms. The BDI-II has an alpha reliabil-
study as dependent variables. Participants respond using a Likert ity of 0.92 for outpatients and a 1-week retest correlation of
scale ranging from 1 (disagree strongly) to 7 (agree strongly) based 0.93 (Beck et al., 1996). The total mean score for a normative
on their experiences in their current relationships. Sixty subscales sample was 12.56 (SD ⫽ 9.93) and 22.45 (SD ⫽ 12.75) for a
from different measures of adult attachment were factor analyzed clinical sample (Beck et al.). In the current study, the mean
to create the ECR (Brennan et al., 1998). A two-factor model with interitem correlation was 0.36 indicating good internal consis-
high factor loadings within each respective factor, ⬎0.40, was tency (Clark & Watson, 1995).
 60 MAXWELL, TASCA, RITCHIE, BALFOUR, AND BISSADA

Binge Eating ment ⫽ 2, 6 months follow-up ⫽ 3, and 12 months follow-up ⫽

4) was log transformed into .00, .30, .48, and .60, respectively, to
The EDE (Fairburn & Cooper, 1993) and a calendar recall model rapid change from pre- to posttreatment, and less rapid
method were used to assess days binged in the past 28 days. The change from posttreatment to 12 months follow-up (Tasca et al.).
reliability and validity of the EDE to assess binge eating is well To account for the relationship between initial scores and rate of
documented (Fairburn & Cooper, 1993). Ten percent of recorded change in the growth models, pretreatment scores were included as
interviews were reassessed by a blind independent rater to assess control variables at level 2 in each model (Tasca & Gallop, 2009;
inter-rater reliability for the current study. Intra-class correlation Tasca, Illing, Ogrodniczuk, & Joyce, 2009). The Hierarchical
coefficient (␳) using a two-way random effects model (Shrout & Linear Modeling program version 7 with the full maximum like-
Fliess, 1979) indicated a high degree of agreement for assessing lihood method of estimation was used.
days binged the past 28 days, ␳ ⫽ .98. To test whether there were changes in attachment anxiety and
attachment avoidance up to 12 months follow-up, we ran two-level
Group Psychotherapy MLMs in which repeated measurement occasions were nested
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

within individuals (see Model 2, Appendix). To test whether the

This document is copyrighted by the American Psychological Association or one of its allied publishers.

GPIP (Tasca et al., 2005) was the group treatment used in this long-term change in attachment anxiety and avoidance was asso-
study. GPIP combines principles from theories of psychotherapeu- ciated with change in binge eating frequency, depressive symp-
tic treatment including psychodynamic, interpersonal, and group toms, or interpersonal problems, we first ran a time-varying cova-
therapies (Malan, 1979; Tasca et al., 2005; Yalom & Leszcz, riate model that examined the association between change in each
2005). GPIP theory describes BED symptoms as a means of attachment scale and change in each of the other outcome variable
maladaptive coping in response to negative affect precipitated by (see Model 3, Appendix). If this time-varying effect was signifi-
interpersonal conflict and unmet attachment needs. GPIP uses cant, we then examined if this relationship strengthened over time
interactions that occur during group psychotherapy to model and by including an interaction term (attachment ⫻ logtime) at level 1
modify participants’ Core Relational Patterns (CRPs; Strupp & of the time-varying covariate model (see Model 4, Appendix). The
Binder, 1994). These interpersonal interactions that affect and variance terms associated with the main effect and interaction
define one’s self concept are identified during the early stage of terms were fixed as suggested by Singer and Willet (2003), who
therapy. More adaptive interactions within the group are encour- indicated that with few measurement occasions per participant,
aged during the middle stage of therapy, and participants are also there is often insufficient data to estimate additional variance
encouraged to generalize these changes to relationships outside of components at level 2. To assess improved model fit, we compared
therapy. The final stage of therapy focuses on consolidating the deviance statistics from the model with the variable of inter-
changes and helping individuals internalize shifts in their CRPs ested to a previous model without that variable [⌬␹2(df) ⫽ devi-
(Tasca et al., 2005). A previous study that assessed the effective- ance statistic from previous model⫺deviance statistic from current
ness of the treatment for symptom reduction in this sample (Tasca model]. Degrees of freedom were calculated as the difference in
et al., in press) reported that therapists in the current study the number of parameters estimated in each model.
achieved acceptable adherence to the treatment manual.

Results
Data Analysis Plan
Our data were nested within the 12 psychotherapy groups; Preliminary Analyses
therefore, observations may not have been independent possibly
resulting in inflated Type I error (Luke, 2004; Tasca, Illing, Joyce, Univariate outliers, multivariate outliers, and multicolinearity
& Ogrodniczuk, 2009). We assessed for within-group dependence and singularity were all assessed and no violations were noted
for each outcome variable by calculating the intra-class correlation (Tabachnick & Fidell, 2007). Days binged at 6 months follow-up
coefficient (ICC), ␳, by the method described by Tasca, Illing, was skewed; however, as the degree of skewness was not severe,
Joyce, and Ogrodniczuk (2009). For each outcome variable, we z ⬍ ⫺3.87, p ⬍ .01 (Tabachnick & Fidell, 2007) and the variable
used a three-level multilevel model (MLM) in which pretreatment, was normally distributed for the other three time points, we did not
posttreatment, 6 months follow-up, and 12 months follow-up transform this variable. A missing data analysis revealed that data
scores at level 1 were nested within individuals at level 2, and were missing at random (see Tasca et al., in press). Table 1 shows
individuals were nested within groups at level 3 (see Model 1, participants M and SD of each variable at each time point.
Appendix). Using the equation described by Tasca, Illing, Joyce, Dependence in the data required adjustment of Type I error rate
and Ogrodniczuk (2009), we used the between-groups variance for the two-level MLM of two of the measures. For the IIP, ␳ ⫽
components for the slope (␶10j) of the models with (conditional .12, so we adjusted the Type I error to .024. For days binged, ␳ ⫽
model) and those without (unconditional model) attachment anx- .10, so we adjusted Type I error to .026.
iety condition, and the between-person variance component (␶1ij) Tasca et al. (in press) reported significant decreases in days
to calculate the ICC, ␳ ⫽ ␶10j(conditional)/(␶10j[unconditional] ⫹ binged, BDI scores, and IIP scores up to 12 months post GPIP, all
␶1ij). For dependent variables in which ␳ ⬍ .05, the dependence in ps ⬍ .001. There were no significant between-group differences on
the data was considered ignorable for groups of this size (Kenny, any of the outcome measures between participants in the high
Kashy, & Bolger, 1998). For dependent variables that had a ␳ ⬎ attachment anxiety condition (i.e., high need for approval) versus
.05, the Type I error rate was adjusted using tables provided by those in the low attachment anxiety condition (i.e., low need for
Kenny et al. (1998). The time metric (pretreatment ⫽ 1, posttreat- approval) (Tasca et al., in press).
 CHANGE IN ATTACHMENT INSECURITY 61

Table 1
Means and Standard Deviations of Outcome Variables

Pretreatment Posttreatment 6 months follow-up 12 months follow-up

Variables M SD n M SD n M SD n M SD n

Days binged 15.25 5.72 102 6.82 7.32 86 6.48 8.51 69 4.78 5.54 55
BDI 20.93 11.49 100 12.72 10.21 81 10.56 8.66 63 13.53 10.62 44
IIP total 1.47 0.54 99 1.36 0.52 80 1.24 0.51 60 1.20 0.56 43
Avoidance 3.50 1.29 99 3.24 1.22 82 3.05 1.06 63 3.04 1.20 44
Anxiety 4.07 1.24 99 3.70 1.09 82 3.54 1.20 63 3.66 1.13 44
Note. BDI ⫽ Beck Depression Inventory; IIP ⫽ Inventory of Interpersonal Problems; days binged ⫽ past 28 days; Avoidance ⫽ Experience in Close
Relationships Scale Avoidance Scale; Anxiety ⫽ Experiences in Close Relationships Anxiety Scale. Data for outcome variables (days binged, BDI, and
IIP) were previously reported by Tasca et al. (in press).
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
This document is copyrighted by the American Psychological Association or one of its allied publishers.

Long-Term Change in Attachment fit was significantly improved for attachment anxiety, p ⫽ .005,
and attachment avoidance, p ⫽ .008. There was also a significant
To test the first hypothesis, we examined whether GPIP resulted
relationship between change in depressive symptoms and change
in significant change in attachment avoidance and anxiety up to 12
in attachment anxiety, p ⫽ .04 (see Table 3). As depressive
months posttreatment (see Table 2). Both attachment anxiety and
symptoms decreased, so did attachment anxiety scores. The model
attachment avoidance decreased significantly. Because time was
fit with the added parameter was significantly improved, p ⬍ .001.
log transformed, the significant slopes reflected a faster rate of
These relationships were not moderated by attachment anxiety
change from pre- to posttreatment and a slower rate of change
condition (i.e., these relationships did not differ across attachment
from post to 6 to 12 months follow-up. The difference in deviance
anxiety condition to which participants were initially assigned).
statistics indicated that the MLM with the logtime parameter fit the
data better than the unconditional MLM, p ⬍ .001, for both To assess hypothesis 3, we examined whether the relationship
attachment anxiety and attachment avoidance. Long-term change between change in both attachment insecurity scales and change in
in attachment anxiety and avoidance were examined by the study IIP scores, and the relationship between change in attachment
condition. Results indicated that change in attachment insecurity anxiety and change in depressive symptoms strengthened over
was not moderated by attachment anxiety condition (i.e., high vs. time by assessing the interaction term in the time varying covariate
low need for approval) to which participants were initially as- model (see Model 4, Appendix). The relationship between reduced
signed. attachment avoidance and improved IIP scores significantly
strengthened over time, p ⬍ .001 (see Table 4). The model with the
Long-Term Change in Attachment Associated With interaction term fit the data significantly better than the uncondi-
tional model, p ⬍ .001. Attachment anxiety condition did not
Change in Other Outcomes
moderate this interaction (see Table 4).
To test the second hypothesis, we examined whether there was
a relationship between reduced attachment anxiety and attachment
avoidance with improved days binged, scores on the BDI, and Discussion
scores on the IIP in separate time varying covariate MLM (see Consistent with the first hypothesis, attachment anxiety and
Model 3, Appendix). There was a significant relationship between attachment avoidance decreased significantly over time to 12
change in IIP scores and change in both attachment anxiety, p ⫽
months posttreatment for women with BED receiving GPIP. This
.01, and attachment avoidance, p ⫽ .02. As IIP scores decreased
extends the results of previous research that showed that attach-
attachment insecurity scores also decreased (see Table 3). Model
ment insecurity improved from pre to post GPIP (Tasca et al.,
2007). Further, these findings are consistent with recent results
Table 2 reported by Kirchmann et al. (2012) about the long-term mainte-
Fixed and Random Effects for Logtime (␤10) of Pretreatment to nance of change in attachment insecurity. Together these studies
Posttreatment to Six Months Follow-Up to 12 Months Follow-Up provide further evidence that psychological interventions, includ-
From the Multilevel Models (MLM) of Attachment Scales ing group therapy, can have a lasting impact on self-reported
attachment insecurity.
Fixed effects Base Logtime One of the main reasons for assessing change in attachment
Attachment scale ␤10 t p D D insecurity in participants with BED was to examine the notion that
change in interpersonal functioning and affect regulation, concepts
Avoidance ⫺.71 ⫺4.04 ⬍.001 735.23 438.34 that are central to attachment, would lead to improvement in other
Anxiety ⫺.89 ⫺4.16 ⬍.001 800.77 550.83
outcomes for women with BED. As the interpersonal model of
Note. Fixed effects df ⫽ 99, N ⫽ 101. D ⫽ deviance statistic, with Base BED suggests (Ansell et al., 2012), difficulties in interpersonal
parameters ⫽ 2 and Logtime parameters ⫽ 4; Base ⫽ MLM with no
relationships lead to negative affect that result in binge eating.
predictors in the model; Logtime ⫽ MLM with only logtime as a level 1
parameter. Attachment scale was the dependent variable and logtime was Therefore, helping those with BED adaptively regulate affect and
the independent variable. Parameters are from Model 2 in the Appendix. improve relationships in the long-term may be important for better
 62 MAXWELL, TASCA, RITCHIE, BALFOUR, AND BISSADA

Table 3
Time-Varying Effect of Attachment Scales (␤10) on Each Outcome

Fixed effects Unconditional Conditional

Outcomes ␤10 t p n df D D

Attachment avoidance
Days binged ⫺.19 ⫺1.02 .31 101 205 1,915.33 1,735.44
BDI .54 1.14 .25 100 207 1,822.85 1,804.12
IIP total .04 2.30 .02 98 199 3.63 ⫺4.13
Attachment anxiety
Days binged ⫺.37 ⫺1.56 .12 101 205 1,915.33 1,734.07
BDI .37 2.02 .04 100 207 1,822.85 1,803.36
IIP total .03 2.55 .01 98 199 3.63 ⫺3.34
Note. D ⫽ deviance statistic, with Unconditional parameters ⫽ 8 and Conditional parameters ⫽ 9. Unconditional ⫽ multilevel model (MLM) without
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

the attachment term in the model; Conditional ⫽ MLM with the attachment as a level-1 parameter; BDI ⫽ Beck Depression Inventory; IIP ⫽ Inventory
This document is copyrighted by the American Psychological Association or one of its allied publishers.

of Interpersonal Problems. Parameters are from Model 3 in the Appendix.

outcomes related to interpersonal functioning, depressive affect, The GPIP model posits that attachment needs lead more directly to
and binge eating. negative affect, which then reduces one’s need to cope by binge
We found that decreases in attachment anxiety and avoidance eating (Tasca et al., 2005). Hence, the relationship between attach-
scores were each associated with improvements in interpersonal ment insecurity and binge eating may be more complex. In previ-
problems, and that these associations were sustained up to 12 ous research with a mixed eating-disorder sample, Tasca, Szad-
months posttreatment. Decreases in attachment anxiety was also kowski et al. (2009) found that the association between specific
associated with reduced symptoms of depression up to 12 months attachment insecurities and eating disorder symptoms were medi-
after treatment. Although we cannot infer a causal relationship ated by different and specific affect regulation strategies. Upregu-
from these associations, we argue that changes in attachment lation of affect or downregulation of affect may mediate the
functioning co-occurred and perhaps led to these symptomatic relationship between change in attachment anxiety or attachment
changes. GPIP works to modify participants’ CRPs, which ulti- avoidance, respectively, and change in binge eating.
mately includes changing one’s introject or self-concept, and this As predicted in hypothesis three, the association between re-
change underlies improved interpersonal functioning and coping duced attachment avoidance and improved interpersonal problems
with negative affect. Therefore, one could argue that as women became stronger overtime to 1 year posttreatment. Those with
with BED experienced changes to their introject as a result of attachment avoidance are known to have difficulty expressing their
GPIP, internal working models of attachment were altered, leading emotional needs in relationships and to have interpersonal prob-
to more satisfying relationships and more adaptive affect regula- lems characterized as cold and distant (Horowitz et al., 1988;
tion. Tasca, Foot, Leite, and Maxwell (2011) provide applied case Tasca, Balfour, Presniak, & Bissada, 2012). Women with BED and
examples that present typical therapist behaviors in GPIP over the higher attachment avoidance receiving GPIP in this study may
course of therapy. have become increasingly better at experiencing and expressing
Despite these positive associations, there were no direct associ- their attachment needs over time. They also likely experienced a
ations between change in attachment insecurity and change in concurrent greater improvement in interpersonal warmth and
binge eating. It may be that the relationship between change in closeness in relationships as attachment increasingly improved
attachment working models and change in binge eating is indirect. post GPIP. This result is similar to Yalom and Leszcz’s (2005)
notion of an adaptive spiral in which improvement in one area of
functioning results in positive change in another related area,
Table 4 which in turn feeds back to further change in the first area of
Fixed and Random Effects for Attachment by Time Interaction functioning. For those with BED, these results highlight the im-
(␤30) From the Time-Varying Covariate Model portance of focusing on decreasing attachment avoidance during
therapy and after therapy as a way to increasingly improve inter-
Fixed effects Unconditional Conditional
personal relationships in the long run. These results provide addi-
Outcomes ␤30 t p D D tional support for using attachment theory to inform the treatment
IIP of eating disorders. Tasca, Ritchie, and Balfour (2011) provide
Avoidance by logtime .35 3.90 ⬍.001 ⫺4.13 ⫺30.19 case examples of how to use attachment theory inform assessment,
Anxiety by logtime .13 1.65 .10 ⫺3.34 ⫺8.36 case formulation, and group psychotherapy of those with an eating
BDI disorder.
Anxiety by logtime .55 .52 .60 1,803.36 1,803.20
Note. Fixed effects IIP df ⫽ 198, N ⫽ 98; BDI df ⫽ 206, N ⫽ 100. D ⫽
deviance statistic, with Unconditional parameters ⫽ 9 and Conditional Limitations and Future Research
parameters ⫽ 10; Unconditional ⫽ MLM without the interaction term in
the model; Conditional ⫽ MLM with the interaction term as a level 1 There are some limitations to this study. First, self-report meth-
parameter; IIP ⫽ Inventory of Interpersonal Problems; BDI ⫽ Beck ods of measuring attachment style may be limited to information
Depression Inventory. Parameters are from Model 4 in the Appendix. that is consciously available to participants. Future research might
 CHANGE IN ATTACHMENT INSECURITY 63

measure attachment states of mind using alternative methods to Bartholomew, K., & Horowitz, L. (1991). Attachment styles among young
self-report such as the AAI (Main, Goldwyn, & Hesse, 2002). adults: A test of a four-category model. Journal of Personality and
Second, there were missing data over the four time points, poten- Social Psychology, 51, 1173–1182.
tially biasing results and limiting generalizability to those with Beck, T. A., Steer, A. R., & Brown, G. K. (1996). The Beck Depression
available data. A previous treatment study using the same data Inventory-II professional manual. San Antonio: The Psychological Cor-
found the data to be missing at random (Tasca et al., in press), poration.
Bowlby, J. (1969). Attachment and loss: Vol 1. Attachment. London, UK:
indicating that the results were not biased by attrition or missing
Pimlico.
data. Third, the characteristics of our sample are common among
Bowlby, J. (1973). Attachment and loss: Vol 1. Separation: Anxiety and
BED treatment-seeking female patients (Striegel-Moore, Wilson, anger. New York: Basic.
Wilfley, Elder, & Brownell, 1998); however, they are not repre- Bowlby, J. (1982). Attachment and Loss. Vol 1: Attachment (2nd ed.). New
sentative of the larger population, and therefore, results may not York: Basic Books.
generalize to a sample with more diverse characteristics or with Bowlby, J. (1988). A secure base: Parent-child attachment and healthy
male patients with BED. Future research should replicate these human development. New York: Basic Books.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

findings with other group therapy samples. Fourth, the design of Brennan, K. A. (1999). Searching for secure bases in attachment-focused
This document is copyrighted by the American Psychological Association or one of its allied publishers.

the study placed participants into groups homogenously composed group therapy: Reaction to Kilmann, et al. (1999). Group Dynamics:
of participants who scored either high or low on an anxious Theory, Research, and Practice, 3, 148 –151. doi:10.1037/1089-2699.3
attachment measure. Because of the study design, the results may .2.148
be limited to similarly constructed psychotherapy groups. Re- Brennan, K. A., Clark, C. L., & Shaver, P. R. (1998). Self-report measure-
searchers should replicate these findings in groups that are heter- ment of adult attachment: An integrative overview. In J. A. Simpson &
ogeneous with regard to attachment anxiety. Fifth, the results W. S. Rholes (Eds.), Attachment theory and close relationships (pp.
46 –76). New York: Guilford Press.
indicated an association between attachment insecurity and out-
Clark, L. A., & Watson, D. (1995). Constructing validity: Basic issues in
comes over time but the correlational design of the study does not
objective scale development. Psychological Assessment, 7, 309 –319.
allow us to infer a causal relationship. Future studies that specif-
doi:10.1037/1040-3590.7.3.309
ically test causal relationships between change in attachment and Cobb, R. J., & Davila, J. (2009). Internal working models and change. In
change in outcomes are necessary. Nevertheless, the results of this J. H. Obegi & E. Berant (Eds.), Attachment theory and research in
study are consistent with attachment theory and the GPIP model of clinical work with adults (pp. 209 –233). New York, NY: Guilford Press.
change. Finally, as indicated, change in attachment insecurity was Fairburn, C. G., & Cooper, Z. (1993). The eating disorder examination
not related to change in binge eating, which is a key feature of (12th ed. ). In C. G. Fairburn & G. T. Wilson (Eds.), Binge eating:
BED. Future research may also focus on testing models that Nature, assessment, and treatment (pp. 317–360). New York: Guilford
include mediators, such as type of affect regulation strategy to Press.
further explore possible indirect relationships between change in Fairburn, C. G., & Wilson, G. T. (1993). Binge eating: Nature, assessment,
attachment insecurity and improved binge eating. and treatment. New York, NY: Guilford Press.
Feeney, J. A., Noller, P., & Hanrahan, M. (1994). Attachment in adults.
New York: Guilford Press.
Conclusion Fonagy, P., Steele, M., Steele, H., Leigh, T., Kennedy, R., Mattoon, G., &
This study adds to emerging evidence that attachment insecurity Target, M. (1995). Attachment, the reflective self, and borderline states:
decreases up to 12 months after group psychological treatment. The predictive specificity of the adult Attachment Interview and path-
This is the first study to show that these changes are related to ological emotional development. In S. Goldberg, R. Muir & J. Kerr
(Eds.), Attachment theory: Social, developmental, and clinical perspec-
long-term decreases in interpersonal difficulties and in depressive
tives (pp. 233–279). Hillsdale, NJ: Analytic Press.
symptoms. Clinically, these findings help inform clinicians of the
George, C., Kaplan, N., & Main, M. (1985). Adult attachment interview.
importance of focusing on changing insecure attachment represen-
Unpublished manuscript. University of California, Berkeley, CA.
tations for women with BED. Targeting affect-regulation strategies Hansen, N. B., & Lambert, M. J. (1996). Brief report: Assessing clinical
and interpersonal functioning in groups may result in change in significance using the inventory of interpersonal problems. Assessment,
individuals’ introjects and internal working models of attachment. 3, 133–136.
These may in turn lead to improved relationships and lower Horowitz, L. M., Rosenberg, S. E., Baer, B. A., Ureno, G., & Villasenor,
depression outside of group that likely persists well beyond the V. S. (1988). Inventory of interpersonal problems: Psychometric prop-
treatment. The results also provide evidence for an adaptive spiral erties and clinical applications. Journal of Consulting and Clinical
(Yalom & Leszcz, 2005) in which attachment avoidance change Psychology, 56, 885– 892. doi:10.1037/0022-006X.56.6.885
and improved interpersonal problems increasingly influence each Hudson, J. I., Hiripi, E., Pope, H. G., & Kessler, R. C. (2007). The
other in the long-term post group therapy. prevalence and correlates of eating disorders in the national comorbidity
survey replication. Biological Psychiatry, 61, 348 –358. doi:10.1016/j
.biopsych.2006.03.040
References Kenny, D. A., Kashy, D. A., & Bolger, N. (1998). Data analysis in social
American Psychiatric Association. (1994). Diagnostic and statistical man- psychology. In D. T. Gilbert, S. T. Fiske, & G. Lindzey (Eds.), The
ual of mental disorders (4th ed.). Washington, DC: American Psychiat- handbook of social psychology (Vols. 1 and 2, 4th ed., pp. 233–265).
ric Publishing. New York: McGraw-Hill.
Ansell, E. B., Grilo, C. M., & White, M. A. (2012). Examining the Kilmann, P. R., Laughlin, J. E., Carranza, L. V., Downer, J. T., Major, S.,
interpersonal model of binge eating and loss of control over eating in & Parnell, M. M. (1999). Effects of an attachment-focused group pre-
women. International Journal of Eating Disorders, 45, 43–50. doi: ventive intervention on insecure women. Group Dynamics: Theory,
10.1002/eat.20897 Research, and Practice, 3, 138 –147. doi:10.1037/1089-2699.3.2.138
 64 MAXWELL, TASCA, RITCHIE, BALFOUR, AND BISSADA

Kirchmann, H., Steyer, R., Mayer, A., Joraschky, P., Schreiber-Willnow, research using multilevel models. Group Dynamics: Theory, Research,
K., & Strauss, B. (2012). Effects of adult inpatient group psychotherapy and Practice, 13, 151–162. doi:10.1037/a0014837
on attachment characteristics: An observational study comparing routine Tasca, A. G., Mikail, F. S., & Hewitt, L. P. (2005). Group psychodynamic
care to an untreated comparison group. Psychotherapy Research, 22, interpersonal psychotherapy: Summary of a treatment model and out-
95–114. doi:10.1080/10503307.2011.626807 comes for depressive symptoms. In M. E. Abelian (Ed.), Focus on
Lawson, D. M., Barnes, A. D., Madkins, J. P., & Francois-Lamonte, B. M. psychotherapy research (pp. 159 –188). Hauppauge, NY: Nova Science
(2006). Changes in male partner abuser attachment styles in group Publishers.
treatment. Psychotherapy: Theory, Research, Practice, Training, 43, Tasca, G. A., Balfour, L., Presniak, M. D., & Bissada, H. (2012). Outcomes
232–237. doi:10.1037/0033-3204.43.2.232 of specific interpersonal problems for binge eating disorder: Comparing
Levy, K. N., Meehan, K. B., Reynoso, J. S., Weber, M., Clarkin, J. F., & group psychodynamic interpersonal psychotherapy and group cognitive
Kernberg, O. F. (2006). Change in attachment patterns and reflective behavioral therapy. International Journal of Group Psychotherapy, 62,
function in a randomized controlled trial of transference-focused psy- 197–218. doi:10.1521/ijgp.2012.62.2.197
chotherapy for borderline personality disorder. Journal of Consulting Tasca, G. A., Balfour, L., Ritchie, K., & Bissada, H. (2007). The relation-
and Clinical Psychology, 74, 1027–1040. doi:10.1037/0022-006X.74.6 ship between attachment scales and group therapy alliance growth
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

.1027 differs by treatment type. Group Dynamics: Theory, Research and

Luke, D. A. (2004). Multilevel modeling. Thousand Oaks, CA: Sage
This document is copyrighted by the American Psychological Association or one of its allied publishers.

Practice, 11, 1–14. doi:10.1037/1089-2699.11.1.1

Publications. Tasca, G. A., Foot, M., Leite, C., & Maxwell, H. (2011). Interpersonal
Main, M., Goldwyn, R., & Hesse, E. (2002). Adult attachment scoring and processes in psychodynamic-interpersonal and cognitive behavioural
classification systems. Berkeley, CA: University of California. group therapy: A systematic case study of two groups. Psychotherapy,
Malan, D. (1979). Individual psychotherapy and the science of psychody- 48, 260 –273. doi:10.1037/a0023928
namics. London: Butterworths. Tasca, G. A., & Gallop, R. (2009). Multilevel modeling of longitudinal
Mikulincer, M., & Shaver, P. R. (2007). Attachment in adulthood: Struc- data for psychotherapy researchers: The basics. Psychotherapy Re-
ture, dynamics, and change. New York: The Guilford Press. search, 19, 429 – 437. doi:10.1080/10503300802641444
Mikulincer, M., Shaver, P. R., & Pereg, D. (2003). Attachment theory and
Tasca, G. A., Illing, V., Joyce, A. S., & Ogrodniczuk, J. S. (2009).
affect regulation: The dynamics, development, and cognitive conse-
Three-level multilevel growth models for nested change data: A guild
quences of attachment-related strategies. Motivation and Emotion, 27,
for group treatment researchers. Psychotherapy Research, 19, 453– 461.
77–102. doi:10.1023/A:1024515519160
doi:10.1080/10503300902933188
Shaver, P. R., & Mikulincer, M. (2002). Attachment-related psychodynam-
Tasca, G. A., Ritchie, K., & Balfour, L. (2011). Implications of attachment
ics. Attachment and Human Development, 4, 133–161. doi:10.1080/
theory and research for the assessment and treatment of eating disorders.
14616730210154171
Psychotherapy, 48, 249 –259. doi:10.1037/a0022423
Shaver, P. R., Schachter, D. A., & Mikulincer, M. (2005). Attachment
Tasca, G. A., Ritchie, K., Conrad, G., Balfour, L., Gayton, J., Lybanon, V.,
style, excessive reassurance seeking, relationship process, and depres-
& Bissada, H. (2006). Attachment scales predict outcome in a random-
sion. Personality and Social Psychology Bulletin, 31, 343–359.
ized controlled trial of two group therapies for binge eating disorder: An
Shorey, S. H., & Snyder, R. C. (2006). The role of adult attachment styles
aptitude by treatment interaction. Psychotherapy Research, 16, 106 –
in psychopathology and psychotherapy outcomes. Review of General
Psychology, 10, 1–20. 121. doi:10.1080/10503300500090928
Shrout, P. E., & Fleiss, J. L. (1979). Intraclass correlations: Uses in Tasca, G. A., Ritchie, K., Demidenko, N., Balfour, L., Krysanski, V.,
assessing rater reliability. Psychological Bulletin, 86, 420 – 428. doi: Weekes, K., . . . Bissada, H. (in press). Matching women with binge
10.1037/0033-2909.86.2.420 eating disorder to group treatment based on attachment anxiety: Out-
Singer, J. D., & Willet, J. B. (2003). Applied longitudinal data analysis: comes and moderating effects. Psychotherapy Research.
Modeling change and event occurrences. New York: Oxford University Tasca, G. A., Szadkowski, L., Illing, V., Trinneer, A., Grenon, R., Demi-
Press. denko, N., . . . Bissada, H. (2009). Adult attachment, depression, and
Steiger, H., Gauvin, L., Jabalpurwala, S., Seguin, J. R., & Stotland, S. eating disorder symptoms: The mediating role of affect regulation strat-
(1999). Hypersensitivity to social interactions in bulimic syndromes: egies. Personality and Individual Differences, 47, 662– 667.
Relationship to binge eating. Journal of Consulting and Clinical Psy- Thompson, R. A. (2000). The legacy of early attachments. Child Devel-
chology, 67, 765–775. doi:10.1037/0022-006X.67.5.765 opment, 71, 145–152. doi:10.1111/1467-8624.00128
Stice, E. (1999). Clinical implications of psychological research on bulimia Travis, L. A., Binder, J. L., Bliwise, N. G., & Horne-Moyer, H. L. (2001).
nervosa and binge eating disorder. Psychotherapy in Practice, 55, 675– Changes in clients’ attachment styles over the course of time-limited
683. dynamic psychotherapy. Psychotherapy: Theory, Research, Practice,
Striegel-Moore, R. H., Wilson, T., Wilfley, D. E., Elder, K. A., & Training, 38, 149 –159. doi:10.1037/0033-3204.38.2.149
Brownell, K. D. (1998). Binge eating in an obese community sample. Waters, E., Merrick, S., Treboux, D., Crowell, J., & Albersheim, L. (2000).
International Journal of Eating Disorders, 23, 27–37. doi:10.1002/ Attachment security in infancy and early adulthood: A twenty-year
(SICI)1098-108X(199801)23:1⬍27::AID-EAT4⬎3.0.CO;2-3 longitudinal study. Child Development, 71, 684 – 689. doi:10.1111/1467-
Strupp, H. H., & Binder, J. L. (1994). Psychotherapy in a new key: A guide 8624.00176
to time-limited dynamic psychotherapy. New York: Basic. Wilfley, D. E., Stein, R. I., Friedman, M. A., Beren, S. A., & Wiseman,
Tabachnick, B. G., & Fidell, L. S. (2007). Using multivariate statistics (5th C. V. (1996). Group cognitive behavioral therapy for binge eating
ed.). Boston, MA: Pearson Education. disorder. Unpublished manuscript.
Tasca, A. G., Illing, V., Ogrodniczuk, J. S., & Joyce, A. S. (2009). Yalom, I. D., & Leszcz, M. (2005). The theory and practice of group
Assessing and adjusting for dependent observations in group treatment psychotherapy (5th ed.). New York: Basic Books.
 CHANGE IN ATTACHMENT INSECURITY 65

Appendix
Multilevel Models

Model 1: Three-Level Model Used to Calculate ␲1i ⫽ ␤10

Intraclass Correlation Coefficients ␲2i ⫽ ␤20 ⫹ ␤21 (individual pretreatment scorei) ⫹ r2i

Level 1: Ytij ⫽ ␲0ij ⫹ ␲1ij (logtimetij) ⫹ etij

Level 2: ␲0ij ⫽ ␤00j ⫹ ␤01j (individual pretreatment scoreij) ⫹ r0ij Model 4: Two-Level Time Varying Covariate Model to
␲1ij ⫽ ␤10j ⫹ ␤11j (individual pretreatment scoreij) ⫹ r1ij Assess the Effect of the Interaction between Change in
Level 3: ␤00j ⫽ ␥000 ⫹ ␥001 (group pretreatment scorej) ⫹ ␥002 Attachment and Logtime
(conditionj) ⫹ u00j
␤01j ⫽ ␥010 ⫹ u01j
Level 1: Yti ⫽ ␲0i ⫹ ␲1i (attachment scale) ⫹ ␲2i (logtimeti) ⫹
␤10j ⫽ ␥100 ⫹ ␥101 (group pretreatment scorej) ⫹ ␥102 (condi-
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

tionj) ⫹ u10j ␲3i (attachment ⫻ logtimeti) ⫹ eti

This document is copyrighted by the American Psychological Association or one of its allied publishers.

␤11j ⫽ ␥110 ⫹ u11j Level 2: ␲0i ⫽ ␤00 ⫹ ␤01 (individual pretreatment scorei) ⫹
r0i
Model 2: Two-Level Model Used to Assess Change in ␲1i ⫽ ␤10
Attachment Scales up to 1 year ␲2i ⫽ ␤20 ⫹ ␤21 (individual pretreatment scorei) ⫹ r2i
␲3i ⫽ ␤30
Level 1: Yti ⫽ ␲0i ⫹ ␲1i (logtimeti) ⫹ eti
Level 2: ␲0i ⫽ ␤00 ⫹ ⫹ ␤01 (individual pretreatment scorei) ⫹ r0i
␲1ij ⫽ ␤10 ⫹ ⫹ ␤11 (individual pretreatment scorei) ⫹ r1i

Model 3: Two-Level Time Varying Covariate Model to

Assess Change in Attachment Scales Associated With
Change in Outcomes Received September 28, 2012
Level 1: Yti ⫽ ␲0i ⫹ ␲1i (attachment scale) ⫹ ␲2i (logtimeti) ⫹ eti Revision received October 22, 2012
Level 2: ␲0i ⫽ ␤00 ⫹ ␤01 (individual pretreatment scorei) ⫹ r0i Accepted October 23, 2012 䡲

See page 87 for a correction to this article.

This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
This document is copyrighted by the American Psychological Association or one of its allied publishers.

Correction to Maxwell et al. (2013)

The article “Change in Attachment Insecurity Is Related to Improved Outcomes 1 Year Post Group
Therapy in Women With Binge Eating Disorder” by Hilary Maxwell, Giorgio A. Tasca, Kerri
Ritchie, Louise Balfour, and Hany Bissada (Psychotherapy, Advanced online publication. February
11, 2013. doi: 10.1037/a0031100), contained errors in the interpretation of the time-varying
covariate analyses in the Online First version of the article. The authors had stated that: (a) there was
a significant relationship between change in Inventory of Interpersonal Problems (IIP) scores and
change in both attachment anxiety, p ⫽ .01, and attachment avoidance, p ⫽ .02; and (b) there was
also a significant relationship between change in depressive symptoms and change in attachment
anxiety, p ⫽ .04. The correct interpretation is: (a) attachment avoidance scores and attachment
anxiety scores are related to IIP scores across all time points; and (b) attachment anxiety scores are
related depressive symptoms across all time points.

In the corrected version of the article additional multilevel modeling analyses continue to partly
support hypothesis 2. That is, change in attachment avoidance scores and change in attachment
anxiety scores are related to change in IIP scores (but not change in depressive symptoms as
originally reported). Also incorrect was the interpretation that: the relationship between reduced
attachment avoidance and improved IIP scores significantly strengthened over time, p ⬍ .001. The
correct interpretation is that a lower score in attachment avoidance at any time point is related to
improvement in IIP scores at that time point. The correct interpretation remains consistent with
hypothesis 3. All versions of this article have been corrected.

DOI: 10.1037/a0034919