Personal Relationships, 19 (2012), 37–50. Printed in the United States of America.Copyright © 2011 IARR; DOI: 10.1111/j.1475-6811.2010.01336.x

Getting over you: Contributions of attachmenttheory for postbreakup emotional adjustment

CHRISTOPHER P. FAGUNDES

The Ohio State University College of Medicine

AbstractThis study used an attachment theoretical framework to identify factors that contribute to postbreakup emotionaladjustment. Individuals who recently experienced a breakup were examined twice over a 1-month period. Greaterdesire to utilize an ex-partner as an attachment figure predicted less emotional adjustment. Higher levels ofattachment anxiety were associated with less emotional adjustment immediately after the breakup. People whoreflected more about the breakup exhibited less emotional adjustment immediately after the breakup compared topeople who reflected less, and also less emotional adjustment over the ensuing month if they also reported moreattachment anxiety. The article highlights the importance of normative attachment processes, in addition toattachment orientations and coping strategies, for understanding postbreakup adjustment.

It is well documented that romantic breakupsare associated with emotional distress andplace people at a heightened risk for a vari-ety of poor mental health outcomes (Bonanno,2004; Sbarra, 2006; Treynor, Gonzalez, &Nolen-Hoeksema, 2003). For example, ado-lescents who recently experienced a breakupare at a heightened risk of suicide and firstonset of major depressive disorder (Monroe,Rohde, Seeley, & Lewinsohn, 1999). Giventhe emotional devastation frequently triggeredby breakups, it is important to understand the

Christopher P. Fagundes, Institute for BehavioralMedicine Research, The Ohio State University Collegeof Medicine.

This research study was supported by the ClaytonResearch Fellowship for Graduate Student Excellence atthe University of Utah. I gratefully acknowledge SusanVecchi, Anna Adams, Kendrick Allen, Nick Ward, KyleMurdock, Dylan Camp, Lina Ramirez, Jon Butner, andLisa Diamond for their assistance with this project.Selected findings from this study were presented at the2008 biennial meeting of the International Association forRelationship Research, Providence, Rhode Island, and the2008 annual meeting of the Society for Personality andSocial Psychology, Albuquerque, New Mexico.

Correspondence should be addressed to ChristopherP. Fagundes, Institute for Behavioral Medicine Research,The Ohio State University, 460 Medical Center Drive,Room 144B, Columbus, OH 43210, e-mail: christopher.fagundes@gmail.com.

basic psychological mechanisms involved indissolving a romantic relationship and deter-mine whether certain factors promote opti-mal adjustment to a breakup. The presentinvestigation adopted an attachment theoret-ical approach to romantic relationship disso-lution. By assessing people who experienced aromantic breakup immediately after the eventand 1 month later, the goal of this studywas to examine how maintained attachmentto one’s ex-partner and individual differencesin attachment orientation and reflecting aboutone’s breakup, predicted postbreakup emo-tional adjustment.

Normative attachment and adjustment

Attachment theory suggests that people selec-tively orient attachment-related functions to-ward one primary attachment figure who isthe most salient and influential person in theirlives, as well as a number of tertiary attach-ment figures that are organized hierarchi-cally (Bowlby, 1982; Trinke & Bartholomew,1997). Based on Bowlby’s (1982) behavioral-based definition of attachment, these func-tions are proximity seeking, safe haven, andsecure base. Proximity seeking occurs when

37

38 C. P. Fagundes

an individual seeks and enjoys the proximityof his or her attachment figure, while activelyresisting separation. An attachment figurefunctions as a safe haven by alleviating dis-tress and providing support when difficultiesarise. Finally, an attachment figure functionsas a secure base by providing a core sense ofemotional and psychological security (Hazan,Gur-Yaish, & Campa, 2004). Bowlby arguedthat people do not successfully adjust to theloss of an attachment figure until they areable to reorganize their attachment hierarchyby no longer desiring to utilize the lost per-son for these functions (Bowlby, 1979; Hazanet al., 2004).

Considerable research suggests that adultromantic relationships are adult “versions” ofthe infant–caregiver bond (Fraley & Shaver,2000; Hazan & Shaver, 1987). Presumably,directing attachment-related desires awayfrom one’s ex-romantic partner is important toemotionally adjust to a romantic breakup. Yetthere has been little research directly inves-tigating this basic process. Even if adultsknow they cannot utilize their ex-partnerfor attachment-related needs, they likely stilldesire to do so, in the same way that a childwho loses an attachment figure continues toseek the lost attachment figure for comfortand security (Bowlby, 1980). For example,although an individual might stop seekingout the ex-partner when he or she needshelp or support (a classic example of thesafe haven component of attachment), he orshe might still experience strong desires todo so. Many people report struggling withsuch desires to continue directing attachment-related needs to their ex-partners, such asneeds for security, support, companionship,and reassurance, even when it might be inap-propriate to do so (Perilloux & Buss, 2008;Sbarra & Emery, 2005).

Such continued desire to use an ex-partnerfor attachment-related needs is arguably mal-adaptive, because by definition, the ex-partnercan no longer be relied upon (Sbarra & Hazan,2008). Continued desires to direct attachmentneeds to the ex-partner also suggest that theindividual in question has not effectively reor-ganized his or her attachment hierarchy suchthat he or she redirects attachment-related

needs to available attachment figures. Attach-ment theory suggests that continued desires todirect attachment needs to ex-partners shouldinterfere with postbreakup emotional adjust-ment. For the person who terminates the rela-tionship, the process of no longer desiring toutilize their romantic partner as an attachmentfigure likely begins before the relationshipends. Thus, by the time of the breakup, thesepeople may have already started redirectingattachment-related needs away from their ex-partner. Yet this is not likely to be the casefor the other partner (the one being “brokenup with”), which may be one reason why suchpeople typically exhibit more distress immedi-ately after a breakup (Sbarra & Emery, 2005).Surprisingly, no prior research has directlyexamined whether maintained attachment toone’s ex-partner after a romantic breakup isassociated with postbreakup emotional adjust-ment.

Attachment orientations and reactionsto separation

In addition to examining normative attach-ment processes, this study also investigatedwhether individual differences in attachmentorientations (sometimes called attachment“style”) are associated with breakup-relatedadjustment. Individual differences in attach-ment orientations were first outlined byAinsworth and colleagues with respect toinfants (Ainsworth, Blehar, Waters, & Wall,1978). Hazan and Shaver (1987) showedthat these same individual differences alsodescribe adults’ orientations toward their ro-mantic partners. Individual differences inattachment orientation are presently conceptu-alized in terms of the orthogonal dimensionsof attachment anxiety and avoidance (Fraley& Shaver, 2000). Specifically, the anxietydimension is characterized by a preoccupa-tion with partner’s accessibility and excessiveworry about rejection and abandonment. Theavoidance dimension is characterized by dis-comfort with closeness, emotional distancing,and a preference to remain highly independentand self-sufficient.

There have been a few notable studiesexamining associations between attachment

Attachment and postbreakup emotional adjustment 39

orientation and people’s emotional reactionsto breakups. Davis, Shaver, and Vernon (2003)found that more anxiously attached peopleretrospectively reported greater preoccupationwith ex-partner, more drug use, and greaterphysical and mental distress after a breakupcompared to less anxiously attached people.Furthermore, Sbarra (2006) illustrated thatmore anxiously attached people have moredifficulty recovering from postbreakup sad-ness than less anxiously attached people. Asfor avoidance, some research suggests thatmore avoidantly attached people have tenden-cies toward emotional distancing comparedto less avoidantly attached people, whichmay enhance their postbreakup adjustment(Davis et al., 2003). Yet other research sug-gests that the classic distancing strategies ofmore avoidant people might actually breakdown in the face of major attachment stres-sors such as relationship dissolution, allow-ing their underlying insecurity to surface.This is supported in work by Birnbaum, Orr,Mikulincer, and Florian (1997), who foundthat avoidantly attached people experiencedgreater distress following a divorce thansecurely attached people. Of course, a divorceis arguably a more significant attachment-related threat than the breakup of a datingrelationship; hence, it is not clear whetheravoidant people who have broken up witha dating partner will experience the breakupas a significant enough threat to erode theirdismissive defenses. Thus, we do not knowwhether more avoidantly attached people willshow enhanced or impaired adjustment to thedissolution of a dating relationship comparedto less avoidantly attached people.

Reflection and adjustment

The present research also considers howreflecting about one’s negative emotions aboutthe breakup is associated with postbreakupemotional adjustment. Although there is con-siderable work examining the impact ofreflecting on one’s negative feelings in re-sponse to the death of a loved one or a stress-ful life event (Nolen-Hoeksema, McBride, &Larson, 1997; Nolen-Hoeksema & Morrow,1991), no study has prospectively examined

the impact of reflecting on one’s nega-tive emotions immediately after a romanticbreakup. Within the bereavement literature,reflection has long been presumed necessaryto reorganize and eventually relinquish one’sprevious attachment bond to the lost part-ner in preparation for forming new attach-ment. In fact, traditional models of griefwork are based on the idea that it is impor-tant to reflect about one’s loss related nega-tive emotions in order to emotionally adjust(Nolen-Hoeksema, 2001; Stroebe, Schut, &Stroebe, 2005). However, there is an emerg-ing evidence to suggest that intensive process-ing of one’s thoughts and negative emotionsregarding a loss might actually be counterpro-ductive for many adults, because it retriggersand reinforces the negative feelings associ-ated with the loss (Nolen-Hoekema, Parker, &Larson, 1994; Nolen-Hoeksema et al., 1997;Nolen-Hoeksema, Morrow, & Fredrickson,1993). Furthermore, there is work to suggestthat the intense reflection is associated withmore severe negative mood after other stress-ful or traumatic life events (Nolen-Hoeksema& Morrow, 1991).

One possible explanation for these con-flicting perspectives is that people differ inthe degree to which reflection is adaptiveor harmful based on their attachment ori-entation. Research has consistently shownthat more anxiously attached people tend tobe less adept at down-regulating negativeemotions and often exhibit intensification ofemotion especially as a result of attachment-related threats compared to less anxiouslyattached people (Cassidy, 1994; Mikulincer &Shaver, 2007). They also tend to be highlypreoccupied with feelings of rejection andabandonment, rely on emotion focused copingstrategies, and overemphasize the importanceof their lost attachment figure as an elementof the self (Davis et al., 2003). Such peo-ple might find it difficult to reflect upon theirnegative emotions related to the breakup with-out becoming overwhelmed by the negativethoughts and feelings that it triggers. Hence,it is possible that reflection is not generallyharmful after a breakup (and possibly adap-tive), but it is harmful for more anxiously

40 C. P. Fagundes

attached people. The present research investi-gated this possibility.

Hypotheses

Based on the previous literature review,the current investigation tested the followinghypotheses:

1. People who reported higher levels ofdesired attachment immediately afterthe breakup will report less emotionaladjustment immediately after the break-up compared to people who reportedlower levels of desired attachment (H1).

2. People who reported their partner ter-minated their relationship will reportless emotional adjustment immediatelyafter the breakup compared to peoplewho chose to terminate the relationshipor viewed the termination as mutual(H2a). The association between termi-nator status and emotional adjustmentwill be mediated by desired attachment.People who reported their partner ter-minated the relationship will report sig-nificantly greater desires to continueutilizing their ex-partners as attachmentfigures compared to people who choseto terminate the relationship or viewedthe termination as mutual, which inturn will be associated with less emo-tional adjustment immediately after thebreakup (H2b).

3. More anxiously attached people willexhibit less emotional adjustment com-pared to less anxiously attached people(H3). As noted earlier, previous findingsregarding avoidance are mixed, somesuggesting enhanced adjustment andsome suggesting the opposite. Hence, anassociation between attachment avoid-ance and emotional adjustment will beinvestigated, but no specific predictionwas made.

4. The associations between reflection andemotional adjustment will be moderatedby attachment orientation. Specifically,reflection will prove more maladaptiveamong more anxiously attached peo-ple compared to less anxiously attachedpeople such that those who report

higher levels of reflection (compared tothose who report lower levels of reflec-tion) and are more anxiously attached(compared to those less anxiously at-tached) will exhibit less emotional ad-justment immediately after the breakupand less improved emotional adjustment1 month later (H4).

5. One month after the first assessment,decreased desire to utilize one’s ex-partner as an attachment figure willbe associated with increased emotionaladjustment (H5).

Method

Participants

The participants for this study were 108undergraduate students attending a large pub-lic university in the Western United Stateswho had recently ended a romantic relation-ship (days from breakup at first assessment,M = 15.60, SD = 7.8 days). Participants wererecruited via email messages sent to all reg-istered undergraduate and graduate studentsat the university, as well as flyers postedaround campus. Participants were paid $12 foreach laboratory visit. Overall, the final sam-ple consisted of 66% females whose averageage was 21 years 3 months old (SD = 3 years3 months, range = 18–33 years). Thirty-threepercent of the sample indicated that theirpartner terminated the relationship. Five per-cent of the participants were Asian, 78%Caucasian, 7% Latino, 1% Native Ameri-can, and 9% declined to report their eth-nicity. The average relationship length priorto the breakup was 18.7 months (SD = 17.0,Mdn = 14.0, range = 4–90). A total of 108people completed the first assessment (T1),23 people chose not to return for the sec-ond assessment (T2) for undisclosed reasons,and 9 people completed the second assess-ment but reestablished a committed relation-ship with their partner. People who did notcomplete the follow-up assessment did notsignificantly differ on any of the T1 measuresfrom people who did complete the follow-upassessment. People who reestablished a com-mitted relationship with their ex-partner wereremoved after imputation (see below) as they

Attachment and postbreakup emotional adjustment 41

were from a different population that was notof interest, and they significantly differed onboth the adjustment and desired attachmentvariables; thus, the final sample consisted of96 people.

Measures

Attachment orientation

Romantic attachment orientation was assessedusing the Experiences in Close RelationshipsScale–Revised (Fraley, Waller, & Brennan,2000). The scale is a 36-item self-reportattachment measure containing two 18-itemsubscales: one assessing attachment anxietyand the other assessing attachment avoidance.The anxiety subscale includes items such as“I worry a fair amount about losing my part-ner” and “I do not often worry about beingabandoned” (reverse scored). The avoidancesubscale includes items such as “I prefer notto show a partner how I feel deep down”and “I feel comfortable depending on roman-tic partners” (reverse scored). Cronbach’s αswere .93 for attachment anxiety and .93 forattachment avoidance.

Reflection

The current investigation used the reflec-tion subscale of the Response Style Ques-tionnaire (Nolen-Hoeksema et al., 1993) toassess reflection. This self-report measureasks respondents to rate their responses tostatements when thinking about their breakupusing a Likert scale from 1 (almost never) to4 (almost always). The directions were mod-ified slightly to take into account the breakupcontext: “People think and do many differ-ent things when they think about a breakup.Please read each of the items below and indi-cate whether you never, sometimes, often, oralways think or do each one when you feeldown, sad, or depressed when thinking aboutyour recent breakup. Please indicate what yougenerally do, not what you think you shoulddo.” The reflection scale consists of five neu-trally valenced items that assess the degree towhich one thinks or ponders about their feel-ings related to the breakup (e.g., “Go awayby yourself and think about why you feel this

way”). Cronbach’s αs were .73 for the reflec-tion items, which is consistent with previousstudies using this scale.

Depressive symptoms

Depressive symptoms were assessed with the20-item Center for Epidemiological Studies-Depression scale (CES–D; Radloff, 1977).The CES–D is a widely used instrument, withprevious studies demonstrating strong relia-bility and validity. This self-report measureasks respondents to rate how they felt andbehaved during the past week using a Lik-ert scale from 1 to 4, in which 1 = rarely ornone of the time (< 1 day), 2 = some or a littleof the time (1–2 days), 3 = occasionally or amoderate amount of the time (3–4 days), and4 = most or all of the time (5–7 days). TheCES–D uses items such as “I felt that I couldnot shake off the blues even with the help ofmy family and friends” and “I felt depressed.”Cronbach’s αs for this measure were .91 forT1 and .91 for T2.

Positive and negative affect

Positive and negative affect were assessedwith the Positive and Negative Affect Sched-ule (Watson, Clark, & Tellegen, 1988). Thisscale is a 20-item self-report measure that hastwo 10-item subscales that are designed toprovide brief measures of both positive andnegative affect. Respondents were asked torate the extent to which they have experi-enced each particular emotion within the pastweek, with reference to a 5-point scale. Thisscale has been shown to be stable at appro-priate levels over a 2-month time period. Thescale points are: 1 (very slightly or not at all ),2 (a little), 3 (moderately), 4 (quite a bit)and 5 (very much). The positive affect scaleuses items such as “interested and excited”;the negative affect scale uses items such as“distressed and upset.” Cronbach’s αs for pos-itive affect were .91 for T1 and .91 for T2.Cronbach’s αs for negative affect were .89 forT1 and .89 for T2.

Composite index of adjustment

Given the collinearity between depressivesymptoms, positive affect, and negative affect

42 C. P. Fagundes

(depressive symptoms T1 and positive affectT1, r = −.70 p < .001; depressive symp-toms T1 and negative affect T1, r = .71,p < .001; positive affect T1 and negativeaffect T1, r = −.56 p < .001; depressivesymptoms T2 and positive affect T2, r =−.67 p < .001; depressive symptoms T2 andnegative affect T2, r = .73, p < .001; pos-itive affect T2 and negative affect T2, r =−.42, p < .001), a composite index for emo-tional adjustment was created, representingthe average of standardized positive affect,standardized and reflected negative affect (sothat higher scores represent less negativeaffect), and standardized and reflected depres-sive symptoms (so that higher scores representfewer depressive symptoms). The αs for thecomposite index were .85 at T1 and .86 at T2.Higher emotional adjustment scores representbetter adjustment.

Attachment-related desires

To assess the degree to which people per-ceived and treated their ex-partners as attach-ment figures, participants completed amodified version of the WHOTO, a self-report measure based on Bowlby’s behavior-based definition of attachment, to assess howmuch they desire to utilize their ex-partneras an attachment figure compared to otherattachment figures (Hazan & Zeifman, 1994).This scale has been used in previous research(Fraley & Davis, 1997; Hazan, Hutt, Stur-geon, & Bricker, 1991) to assess the degreeto which people turn to various members oftheir social network for specific attachmentfunctions. Consistent with previous research(Fraley & Davis, 1997), the proximity, safehaven, and secure base scales were used.Each attachment scale uses three separateitems to assess where people perceive variousattachment figures on their attachment hier-archy for proximity seeking, safe haven, andsecure base functions. An example of prox-imity seeking item is “Who is the person youlike to be with as much as possible?” Anexample of safe haven item is “Who do youseek out when you’re worried about some-thing?” An example of secure base item is“Who do you know who will always be there

for you, no matter what?” Respondents areable to list up to five people in response toeach question, in order of importance, whichreflects the first five members on their attach-ment hierarchy. As noted above, the presentinvestigation focuses on where people desiretheir ex-partner to be on their attachment hier-archy even if there are self-imposed and/orsocially imposed reasons why they may not beable to use them for these functions. Hence,after each question on the WHOTO, peoplewere asked to report on where they desire theex-partner to be on their attachment hierarchy.Specifically, after each item participants wereasked the following question: “Even if youknow you cannot or should not, if you couldplace your former romantic partner anywhereon the above list, where would you desireto put him or her?” Depending on where anindividual desired to place their ex-partner foreach question on their attachment hierarchy,they received a score to reflect the position.If an individual desired for their ex-partnerto be first on a particular function, he or shereceived a rating of 5; if listed as second,he or she received a rating of 4, and so on.If an individual did not desire for their ex-partner to be on their hierarchy for a particularquestion at all, he or she received a scoreof 0. Given the collinearity between items ateach time point (proximity seeking T1 andsafe haven T1, r = .83, p < .001; proxim-ity seeking T1 and secure base T1, r = .62,p < .001; safe haven T1 and secure base T1,r = .75, p < .001; proximity seeking T2 andsafe haven T2, r = .88, p < .001; proxim-ity seeking T2 and secure base T2, r = .82,p < .001; safe haven T2 and secure baseT2, r = .84, p < .001), a composite three-item index for desired attachment was createdby combining desired proximity, desired safehaven, and desired secure base at each timepoint. This was calculated by summing scoresfrom each index and dividing by 3. Cron-bach’s αs were .88 for desired attachment atT1 and .90 for desired attachment at T2.

Procedure

Participants completed two laboratory visits,approximately 4 weeks apart (M = 33.1 days,

Attachment and postbreakup emotional adjustment 43

Mdn = 31 days). The night before their firstlaboratory visit (T1), participants filled out anonline version of the WHOTO questionnaire,and then completed additional questionnairemeasures during each of their visits.

Results

Multiple imputation was employed to imputeT2 missing data due to attrition following therecommendations of Graham (2009). Multi-ple imputation produces unbiased parameterestimates that appropriately reflect the truevariability of the missing data and has beenshown (through simulation studies) to be amore valid and less biased analytical approachthan listwise deletion. Multiple imputation hasbeen shown to perform well when data aremissing at random and even acceptable undersome case of nonrandom missingness. It isrobust to departures from normality assump-tions and performs well even with low samplesize. Following standard practice, the impu-tation procedure was repeated five times inorder to approximate the true measurementvariance represented in real data. All anal-yses were completed with each of the fullimputed data sets, and the coefficients gener-ated by each separate data set were averagedto produce final estimates. See Schafer andOlsen (1998) for more information on thisapproach. All reported analyses below reflectthese pooled estimates.1

Means and standard deviations for allstudy variables are presented in Table 1.2 All

1. Using the imputation approach to handle attritionyielded identical significance levels to traditional list-wise deletion with a small exception in Block 2 of theof the second regression analysis predicting change inadjustment (i.e., Table 4). Specifically, reflection wasassociated with T2 adjustment in the listwise approachat p < .05, while it was associated with T2 adjustmentin the imputation approach at p < .10. Consequently,the overall model in this same block yielded a sig-nificance level of p < .05 using listwise rather thanp < .10 using multiple imputation.

2. Given that the adjustment variables were standardizedbefore the composite index was formed, the mean is0 and therefore not interruptible. Therefore the meansof each component of the item index are presented inTable 1. It should be noted that the standard deviationsof composite index for adjustment variable were .86at T1 and .83 at T2.

Table 1. Means and standard deviations of allstudy variables

Variable M SD

Attachment anxiety 3.36 1.21Attachment avoidance 3.26 1.20Reflection 2.26 0.66Desired attachment T1 7.98 4.89Desired attachment T2 6.77 5.30Depressive symptoms T1 2.18 0.62Depressive symptoms T2 1.86 0.56Positive affect T1 1.07 1.48Positive affect T2 2.77 0.78Negative affect T1 2.51 0.85Negative affect T2 2.17 0.76

independent variables were centered.3 Zero-order correlations among study variables arepresented in Table 2. Examination of thecoefficients reveals that people who reportedhigher levels of initial desired attachmentreported less T1 emotional adjustment com-pared to people who reported lower initialdesired attachment (H1). People who reportedthat their partner had terminated their relation-ship reported lower T1 emotional adjustmentcompared to people who did not report theirpartner terminated the relationship (H2a).Consistent with H3, more anxiously attachedpeople reported less T1 emotional adjustmentcompared to less anxiously attached people.Furthermore, people who reported higher lev-els of reflection reported less T1 emotionaladjustment compared to people who reportedless reflection. After examination of thesezero-order correlations, a multiple regressionwas performed to examine the unique contri-bution of each of these variables in predictingT1 emotional adjustment, while controllingfor gender, which was previously shown tobe associated with breakup adjustment (Daviset al., 2003). As can be seen in Table 3, the

3. To rule out potential confounds, initial analysesincluded a number of additional variables that Iexpected might influence breakup adjustment: dayssince the breakup and prebreakup relationship length.These variables did not influence the strength or direc-tion of the effects of interest; thus, they were droppedfrom the final analysis to simplify presentation of theresults.

44 C. P. Fagundes

Table 2. Correlations among attachment orientation, coping strategies, normative attachment,emotional adjustment

Variable 1 2 3 4 5 6 7 8

1. T1 emotional adjustment2. T2 emotional adjustment .54∗∗3. T1 desired attachment −.27∗∗ −.084. T2 desired attachment .14 −.20† .50∗∗5. Attachment anxiety −.45∗∗ −.28∗∗ −.05 −.076. Attachment avoidance −.15 −.08 −.22∗ −.29∗∗ .28∗∗7. Reflection −.39∗∗ −.36∗∗ −.05 .03 .12 .028. Partner terminated −.25∗ .00 .28∗∗ .09 .19† −.23∗ .009. Gender (male = 0, female = 1) −.11 .02 .08 .15 .12 .12 .09 .00

†p < .10. ∗p < .05. ∗∗p < .01.

Table 3. Summary of regression analysis pre-dicting T1 emotional adjustment

Variable β

Partner terminated −.14Gender (male = 0, female = 1) −.02Attachment anxiety −.35∗∗Attachment avoidance −.14Reflection −.36∗∗T1 desired attachment −.30∗∗R2 .42∗∗∗

∗∗p < .01. ∗∗∗p < .001.

simultaneous inclusion of these independentvariables using multiple regression did notchange the significance levels of the coeffi-cients, with the exception of terminator status,which was no longer significant. This indi-cates the possibility of mediation.

To determine whether or not T1 desiredattachment mediated the relationship betweenterminator status and T1 emotional adjustment(H2b), Baron and Kenny’s (1986) regres-sion procedures for testing mediation wereemployed. Regression coefficients for thismediating test are presented in Figure 1. Gen-der, attachment orientation, and reflectionwere included as covariates in all regressionequations (Baron & Kenny, 1986; MacKinnon,Lockwood, Hoffman, West, & Sheets, 2002).As previously mentioned, people who reportedtheir partner terminated the relationship

Table 4. Summary of hierarchical regressionanalysis predicting T2 emotional adjustmentcontrolling for T1 emotional adjustment(higher scores indicate better emotionaladjustment)

Variable �R β

Step 1 .29∗∗∗Emotional adjustment T1 .54∗∗

Step 2 .11†

Emotional adjustment T1 .47∗∗Partner terminated .13Gender (male = 0,

female = 1).10

Attachment anxiety −.10Attachment avoidance .05Reflective rumination −.17†

T1 desired attachment .10T2 desired attachment −.20∗

Step 3 .05∗Emotional adjustment T1 .50∗∗Partner terminated .12Gender (male = 0,

female = 1).15

Attachment anxiety −.08Attachment avoidance .04Reflective rumination −.17†

T1 desired attachment .09T2 desired attachment −.21∗Attachment Anxiety ×

Reflective Rumination−.19∗

Total R2 .45∗∗∗

†p < .10. ∗p < .05. ∗∗p < .01. ∗∗∗p < .001.

Attachment and postbreakup emotional adjustment 45

Figure 1. The mediating role of desiredattachment in the association between ter-minator status (1 = partner terminated therelationship, 0 = I terminated the relation-ship or breakup mutual ) and T1 emotionaladjustment. Beta coefficients are standardized.∗p < .05. ∗∗p < .01.

reported less T1 emotional adjustment com-pared to people who did not report theirpartner terminated the relationship, whichestablished that there was a relationship tobe mediated. Second, the hypothesized media-tor, T1 desired attachment, also predicted T1emotional adjustment such that people whoreported higher levels of desired attachmentreported less emotional adjustment comparedto people who reported less desired attach-ment. Third, people who reported that theirpartner terminated their relationship reportedhigher levels of desired attachment comparedto people who did not report their part-ner terminated the relationship. Finally, whenT1 emotional adjustment and terminator sta-tus were added simultaneously, the associa-tion between terminator status and emotionaladjustment was reduced to nonsignificance. ASobel test confirmed this reduction was signif-icant (Z = 2.15, p < .05). Thus, all of Baronand Kenny’s criteria for mediation were metand H2b was confirmed such the associationbetween “terminator status” and T1 emotionaladjustment was mediated by the fact that peo-ple who had partners that terminated the rela-tionship had significantly greater desires tocontinue utilizing their ex-partners as attach-ment figures.4

4. After the all independent variables were simultane-ously included in the regression analysis, all two-wayinteractions were included in a subsequent block, fol-lowed by all three-way interactions. None of theseinteractions reached significance; thus, I chose to omitthem from Table 3 for presentation clarity.

Next, a test of whether or not thesehypothesized variables predicted change inemotional adjustment from T1 to T2 wasconducted. A hierarchical multiple regres-sion was employed. In the first step, T2emotional adjustment was regressed on T1emotional adjustment (to reflect residualizedchange from T1 to T2). In the second step,gender, terminator status, attachment orienta-tion, T2 desired attachment, reflection, and T1desired attachment were added (T1 desiredattachment was added so that T2 desiredattachment would reflect residualized changein desired attachment). The two-way interac-tions were entered in the third step.

In accord with H5, decreased desiredattachment was associated with higher lev-els of emotional adjustment. Higher lev-els of reflection were associated with lessincreased emotional adjustment, albeit at trendlevel. Gender, terminator status, attachmentanxiety, and attachment avoidance were notsignificantly associated with T2 emotionaladjustment.5 As predicted (H4), attachmentanxiety interacted with reflection in predict-ing emotional adjustment from T1 to T2. Theaddition of this interaction term accountedfor a significant amount of additional vari-ance in the overall model. Figure 2 repre-sents this interaction. The outcome is T2emotional adjustment, controlling for T1 emo-tional adjustment, so that higher scores rep-resent better emotional adjustment from T1to T2. Reflection was significantly associatedwith less improved emotional adjustment atT2 when attachment anxiety was high (1 SDabove the mean), t = −4.09, p < .001; how-ever, this was not the case when attachmentanxiety was low (1 SD below the mean),t = .31, ns. None of the other nonhypothe-sized two-way interactions were significant.

5. Given that previous retrospective work (Saffrey &Ehrenberg, 2007) demonstrated that rumination medi-ated the association between attachment anxiety andbreakup adjustment, I conducted additional analy-ses to explore whether the lack of an associationbetween attachment anxiety and change in adjust-ment was attributable to the possible mediating roleof coping strategies. When attachment orientation wasalso entered into the model without reflection and itremained nonsignificant, as did avoidance. Thus, nosign of mediation was detected.

46 C. P. Fagundes

-0.3

-0.2

-0.1

0

0.1

0.2

0.3

0.4

Em

otio

nal A

djus

tmen

t--> Low Anx

High Anx

Low Reflection High Reflection

Figure 2. The moderating role of attachmentanxiety in the relationship between reflectionand T2 emotional adjustment (controlling forT1 emotional adjustment).Note. The graph represents predicted valuesfor T2 emotional adjustment for people 1 SDabove and below the mean for both reflectionand attachment anxiety.

Discussion

Drawing upon attachment theory as a frame-work for understanding loss, the article ex-tends recent efforts to determine predictors ofemotional adjustment after nonmarital disso-lution. This study is the first to examine hownormative attachment processes are associ-ated with postbreakup emotional adjustment.Results showed that people who reporteda greater desire to utilize their ex-partneras an important attachment figure exhibitedless emotional adjustment immediately aftera breakup compared to people who reportedless desire to utilize their ex-partner as anattachment figure. Furthermore, people whoexhibited less declined desire to utilize theirex-partner as an attachment figure 1 monthlater reported less improved emotional adjust-ment compared to individuals who exhib-ited more declined desire. People who didnot choose to terminate their relationshipreported less emotional adjustment immedi-ately after the breakup, and this associa-tion was mediated by their greater desireto utilize their ex-partner as an attachmentfigure. This study also extends work evaluat-ing how attachment orientations and reflectingabout a breakup contributes to postbreakup

emotional adjustment. Attachment avoidancewas not associated with postbreakup emo-tional adjustment. Results indicated that moreanxiously attached people exhibited lowerlevels of initial emotional adjustment com-pared to less anxiously attached people. Fur-thermore, people who reported high levels ofreflection about the breakup reported less ini-tial emotional adjustment compared to peoplewho reported low levels of reflection aboutbreakup. Furthermore, people who reportedhigher levels of reflection about the breakupreported less improved emotional adjustmentin the following month if they also reportedbeing more anxiously attached.

I began with the premise that in orderto emotionally adjust to a breakup, peoplemust stop desiring to utilize their ex-partner asan attachment figure. By systematically docu-menting that desired attachment is associatedwith emotional adjustment even after control-ling for attachment orientation and terminatorstatus, this study lends support to Sbarra andHazan’s (2008) recent theoretical argumentsuggesting that the degree to which one feelsdysregulation when separated from an attach-ment figure should be associated with thedegree to which one desires to use a partner(or ex-partner) when they cannot do so. Thefinding that the association between termina-tor status and initial postbreakup emotionaladjustment was mediated by desired attach-ment helps explain why people going througha breakup, who do not choose to terminate therelationship, experience more distress (Sbarra& Emery, 2005). Understanding mechanismsthat underlie why people who do not choose toterminate the relationship exhibit poorer emo-tional adjustment compared to people who ter-minate the relationship is important for boththeoretical and applied work.

The association between attachment anx-iety and initial emotional adjustment addsto previous work showing similar associa-tions using both prospective and retrospec-tive designs (e.g., Davis et al., 2003; Sbarra,2006). Previous studies examining the associ-ation between attachment avoidance and emo-tional adjustment after a loss have producedconflicting results—some showing that avoid-ance is associated with mild improvement

Attachment and postbreakup emotional adjustment 47

(Davis et al., 2003), while others showingit is not (Birnbaum et al., 1997). Giventhat more avoidantly attached people did notshow improved emotional adjustment com-pared to less avoidantly attached people, itappears that the strategies highly avoidantpeople typically employ to disengage fromattachment-related threats may not be fullyengaged in a breakup context (e.g., Fra-ley & Shaver, 1997; Mikulincer & Shaver,2005). As Mikulincer, Birnbaum, Woddis, andNachmias (2000) suggested, more avoidantlyattached people may be unable to employdefensive strategies when presented with starkattachment-related threats, such as the dissolu-tion of a romantic relationship. Yet contrary toprevious findings regarding avoidant people’sresponses to divorce (Birnbaum et al., 1997),I did not find that more avoidantly attachedpeople showed significantly poorer adjust-ment to their breakups than less avoidantlyattached people. This may suggest that non-marital breakups represent an intermediateform of attachment-related threat—not threat-ening enough to facilitate poor adjustment,but threatening enough to prevent the engage-ment of dismissive strategies that mightactually enhance adjustment. Investigatingspecific thresholds for attachment-relatedthreats among avoidant people may shedlight on this issue. One possibility is thatan avoidant individual’s previous history withromantic breakups will affect the likelihoodthat he or she will be able to engage dismis-sive coping strategies in the face of relation-ship loss.

Although traditional models of grief worksuggests that it is adaptive to reflect uponone’s emotions related to a loss in orderto “work through” the painful experience,recent work suggest that may not be the casegiven that reflecting upon one’s negative emo-tions may be harmful (see Wortman & Sil-ver, 2001, for a review of this controversy).This study suggest that reflecting about one’sbreakup-related emotions is associated withpoor emotional adjustment immediately afterthe breakup regardless of attachment anxiety;however, reflection was only maladaptive formore anxiously attached people 1 month later.Thus, processing one’s negative emotions

may be maladaptive regardless of attachmentanxiety immediately after the breakup becausethe loss is too salient and painful for everyone.However, over time, it is only maladaptiveto reflect upon one’s breakup if one is moreanxiously attached. This interaction supportsthe expectation that it is particularly maladap-tive for a more anxiously attached personto “work through their emotions” given theirinability to down-regulate or compartmental-ize the intense negative emotions likely trig-gered by reflection. It is possible that someof the inconsistency in the literature regardingwhether it is adaptive or maladaptive to reflectupon one’s emotions after a loss (Wortman &Silver, 2001) is attributable to variability inpeople’s capacities for effective emotion reg-ulation. Future studies should closely examineanxious people’s experiences with reflectionin other contexts such as bereavement.

Nonetheless, reflecting about one’s emo-tions associated with a loss may still proveto be an important part of long-term growthand recovery. Saffrey and Ehrenberg (2007)demonstrated that reflecting upon positiveaspects of a breakup experience was associ-ated with positive adjustment, but their par-ticipants were assessed at a considerably latertime than participants in this study (i.e., onaverage, 5 months after the breakup occurredvs. less than 1 month). This might be inter-preted to suggest that reflecting on a breakupcan be adaptive, but perhaps only after themore immediate, intense emotions associatedwith the loss have dissipated (see Sbarra,2006, for a review about the normative timecourse of nonmarital emotional adjustment).Thus, in addition to assessing the degree towhich one reflects on a breakup, researchersshould take into account when people reflect.It is notable that the final phases of manystage models of breakup and divorce high-light the importance of reflecting on one’sformer relationship in order to come up witha resolution that creates meaning and facil-itates growth (Rollie & Duck, 2006). How-ever, the present findings point to the fact thatexcessive reflection before emotional adjust-ment has taken place is likely unhelpful, espe-cially for people who are unable to down-regulate their negative emotions (e.g., people

48 C. P. Fagundes

high on attachment anxiety). Future studiesfocusing on the timing of adaptive versusmaladaptive forms of reflection can make animportant contribution to the development ofeffective clinical interventions aimed at pro-moting recovery from relationship loss.

Limitations

This research is limited by the fact thatparticipants were students at a universitycampus. As a result, it is not possible todetermine the degree to which the findingsgeneralize to other populations such as mar-ried people and older adults. One particularconcern when assessing younger people inrelationship research is that they are typicallyinvolved in relatively new romantic relation-ships. Although the participants’ relationshipsin this study were relatively long for a collegesample, it is important to examine processesof relinquishing the attachment bond in muchlonger relationships. Given that the dissolu-tion of longer term relationships is likely toinvolve more severe and prolonged distress,it is possible that the association between theattachment and adjustment is more complexin these cases.

Another important limitation is that peo-ple’s levels of attachment anxiety may besomewhat elevated during these assessmentsgiven that an anxious individual’s “worstfears” are, in effect, coming true (especiallyfor those who were rejected by their part-ners). Yet other studies have shown simi-lar associations between attachment anxietyassessed before subsequent dissolution andinitial adjustment levels (Sbarra, 2006). Itwould be interesting for future research totreat attachment anxiety as a time-varyingcovariate to see if attachment anxiety increasesafter a breakup among people who are alreadymore anxiously attached.

Although study participants came in fortheir first assessment very soon after theirbreakup occurred, it is inevitable that to somedegree, changes in attachment and emotionaladjustment had already begun. It would beideal to have measures of normative attach-ment, relationship quality, and attachmentorientation immediately before the breakup.

This would allow researchers to investigatewhether people who view their romantic part-ner higher on their attachment hierarchy in anintact relationship have more difficulty emo-tionally adjusting to the loss of this person.Of course, the only way to collect such data isto follow a large number of people longitudi-nally, assessing their relationships on a regularbasis, and waiting until their relationships end.Ideally, such research could continue to fol-low people many months after the breakup toexamine longer term processes.

Finally, because these tests are correla-tional, causality cannot be determined. Inthe case of the relationship between desiredattachment and emotional adjustment, it can-not be determined whether desired attachmentdrives changes in adjustment, or vice versa.It is certainly possible that people’ desires todirect attachment functions toward their ex-partners are partially influenced by their feel-ings of distress. Hence, as people begin to feelbetter, it may become “easier” to no longerdesire to utilize an ex-partner as an importantattachment figure. Most likely, the associationbetween desired attachment and adjustmentis bidirectional, so that each process recipro-cally influences the other over time. Closerinvestigation of this possibility, using morefine-grained longitudinal analysis, is a criticaldirection for future research.

Conclusions

This study extends our understanding of whatpredicts emotional adjustment in the contextof a romantic breakup. This study sheds lighton whether reflecting about one’s breakup-related negative emotions is maladaptive byillustrating that immediately after a breakup,reflection is maladaptive regardless of attach-ment anxiety. Furthermore, people who reflectmore about a breakup exhibit less emotionaladjustment overtime if they also exhibit moreattachment anxiety. Furthermore, this studymakes a contribution to our understanding ofhow a continued desire to utilize an ex-partneras an attachment figure is associated with pooremotional adjustment to a breakup. Basedon Bowlby’s (1980) original observationsof children loosing their attachment figures,

Attachment and postbreakup emotional adjustment 49

one of the foundational theoretical tenets ofattachment theory is that people do not emo-tionally adjust to the loss of an attachmentfigure until they no longer desire to uti-lize them for attachment-related functions. Yetadult attachment researchers have devoted lit-tle attention to empirically examining whethera maintained desire to utilize an ex-partneras an attachment figure is associated withemotional adjustment after relationship dis-solution. By systematically documenting thatdesired attachment is associated with emo-tional adjustment even after controlling forattachment orientation and terminator status,this study suggests that dissolution researchersdevote increased attention to evaluating nor-mative attachment in addition to attachmentorientation.

References

Ainsworth, M. D. S., Blehar, M. C., Waters, E., &Wall, S. (1978). Patterns of attachment: A psycho-logical study of the strange situation. Hillsdale, NJ:Erlbaum.

Baron, R. M., & Kenny, D. A. (1986). The moderator-mediator variable distinction in social psychologicalresearch: Conceptual, strategic, and statistical consid-erations. Journal of Personality in Social Psychology,51, 1173–1182.

Birnbaum, G. E., Orr, I., Mikulincer, M., & Florian, V.(1997). When marriage breaks up: Does attachmentstyle contribute to coping and mental health? Journalof Social and Personal Relationships, 14, 643–654.

Bonanno, G. A. (2004). Loss, trauma, and human resil-ience: Have we underestimated the human capacityto thrive after extremely aversive events? Amer-ican Psychologist, 59, 20–28. doi: 10.1037/0003-066x.59.1.20.

Bowlby, J. (1979). The making and breaking of affec-tional bonds. London: Tavistock.

Bowlby, J. (1980). Attachment and loss: Vol. 3. Loss:Sadness and depression. New York: Basic Books.

Bowlby, J. (1982). Attachment and loss: Vol. 1. Attach-ment (2nd ed.). New York: Basic Books.

Cassidy, J. (1994). Emotion regulation: Influences ofattachment relationships. Monographs of the Societyfor Research in Child Development, 59, 228–283.

Davis, D., Shaver, P. R., & Vernon, M. L. (2003). Phys-ical, emotional, and behavioral reactions to breakingup: The roles of gender, age, emotional involvement,and attachment style. Personality and Social Psychol-ogy Bulletin, 29, 871–884.

Fraley, R. C., & Davis, K. E. (1997). Attachment forma-tion and transfer in young adults’ close friendshipsand romantic relationships. Personal Relationships, 4,131–144.

Fraley, R. C., & Shaver, P. R. (1997). Adult attachmentand the suppression of unwanted thoughts. Journal ofPersonality and Social Psychology, 73, 1080–1091.

Fraley, R. C., & Shaver, P. R. (2000). Adult romanticattachment: Theoretical developments, emerging con-troversies, and unanswered questions. Review of Gen-eral Psychology, 4, 132–154.

Fraley, R. C., Waller, N. G., & Brennan, K. A. (2000).An item response theory analysis of self-report mea-sures of adult attachment. Journal of Personality andSocial Psychology, 78, 350–365.

Graham, J. W. (2009). Missing data analysis: Making itwork in the real world. Annual Review of Psychology,60, 549–576. doi: 10.1146/annurev.psych.58.110405.085530.

Hazan, C., Gur-Yaish, N., & Campa, M. (2004). Whatdoes it mean to be attached? In J. A. Simpson &W. S. Rholes (Eds.), Adult attachment: New direc-tions and emerging issues (pp. 55–85). New York:Guilford.

Hazan, C., Hutt, M. J., Sturgeon, M. J., & Bricker, T.(1991). The process of relinquishing parents as attach-ment figures. Paper presented at the biennial meetingsof Society for Research on Child Development, Seat-tle, WA.

Hazan, C., & Shaver, P. R. (1987). Romantic love con-ceptualized as an attachment process. Journal of Per-sonality and Social Psychology, 59, 511–524.

Hazan, C., & Zeifman, D. (1994). Sex and the psycholog-ical tether. In K. Bartholomew & D. Perlman (Eds.),Advances in personal relationships. Vol. 5. Attachmentprocesses in adulthood (pp. 151–177). London: Jes-sica Kinsley Publishers.

MacKinnon, D. P., Lockwood, C. M., Hoffman, J. M.,West, S. G., & Sheets, V. (2002). A comparison ofmethods to test mediation and other intervening vari-able effects. Psychological Methods, 7, 83–104.

Mikulincer, M., Birnbaum, G., Woddis, D., & Nachmias,O. (2000). Stress and accessibility of proximity-related thoughts: Exploring the normative and intrain-dividual components of attachment theory. Journal ofPersonality and Social Psychology, 78, 509–523.

Mikulincer, M., & Shaver, P. R. (2005). Mental repre-sentations of attachment security: Theoretical foun-dation for a positive social psychology. In M. W.Baldwin (Ed.), Interpersonal cognition (pp. 233–266).New York: Guilford.

Mikulincer, M., & Shaver, P. R. (2007). Attachment inadulthood: Structure, dynamics, and change. NewYork: Guilford.

Monroe, S. M., Rohde, P., Seeley, J. R., & Lewinsohn,P. M. (1999). Life events and depression in adoles-cence: Relationship loss as a prospective risk factorfor first onset of major depressive disorder. Journalof Abnormal Psychology, 108, 606–614.

Nolen-Hoeksema, S. (2001). Ruminative coping andadjustment to bereavement. In M. S. Stroebe, R. O.Hansson, W. Stroebe, & H. Schut (Eds.), Handbookof bereavement research: Consequences, coping, andcare (pp. 545–562). Washington, DC: American Psy-chological Association.

50 C. P. Fagundes

Nolen-Hoeksema, S., McBride, A., & Larson, J. (1997).Rumination and psychological distress among be-reaved partners. Journal of Personality and SocialPsychology, 72, 855–862. doi: 10.1037/0022-3514.72.4.855.

Nolen-Hoeksema, S., & Morrow, J. (1991). A prospec-tive study of depression and post-traumatic stresssymptoms after a natural disaster. Journal of Person-ality and Social Psychology, 100, 569–582.

Nolen-Hoeksema, S., Morrow, J., & Fredrickson, B. L.(1993). Response styles and the duration of episodesof depressed mood. Journal of Abnormal Psychology,102, 20–28.

Nolen-Hoekema, S., Parker, L. E., & Larson, J. (1994).Ruminative coping with depressed mood followingloss. Journal of Personality and Social Psychology,67, 92–104.

Perilloux, C., & Buss, D. M. (2008). Breaking up roman-tic relationships: Costs, experienced, and copingstrategies deployed. Evolutionary Psychology, 60,164–181.

Radloff, L. S. (1977). The CES–D Scale: A self-reportdepression scale for research in the generalpopulation. Applied Psychological Measurement, 1,385–401.

Rollie, S. S., & Duck, S. (2006). Divorce and dissolu-tion of romantic relationships: Stage models and theirlimitations. In M. A. Fine & J. H. Harvey (Eds.),Handbook of divorce and relationship dissolution(pp. 223–240). Mahwah, NJ: Erlbaum.

Saffrey, C., & Ehrenberg, M. F. (2007). When thinkinghurts: Attachment, rumination, and postrelationshipadjustment. Personal Relationships, 14, 351–368.

Sbarra, D. A. (2006). Predicting the onset of emotionalrecovery following nonmarital relationship dissolu-tion: A survival analyses of sadness and anger. Per-sonality and Social Psychology Bulletin, 32, 298–312.

Sbarra, D. A., & Emery, R. E. (2005). The emotionalsequelae of non-marital relationship dissolution: Anal-yses of change and intraindividual variability overtime. Personal Relationships, 12, 213–232.

Sbarra, D. A., & Hazan, C. (2008). Coregulation, dysreg-ulation, self-regulation: An integrative analysis andempirical agenda for understanding adult attachmentseparation, loss, and recovery. Personality and SocialPsychology Bulletin, 12, 141–167.

Schafer, J. L., & Olsen, M. K. (1998). Multiple impu-tation for multivariate missing-data problems: Adata analyst’s perspective. Multivariate BehavioralResearch, 33, 545–571. doi: 10.1207/s15327906mbr3304_5.

Stroebe, W., Schut, H., & Stroebe, M. S. (2005). Griefwork, disclosure and counseling: Do they help thebereaved? Clinical Psychology Review, 25, 395–414.

Treynor, W., Gonzalez, R., & Nolen-Hoeksema, S.(2003). Rumination reconsidered: A psychometricanalysis. Cognitive Therapy and Research, 27,247–259.

Trinke, S. J., & Bartholomew, K. (1997). Hierarchiesof attachment relationships in young adulthood.Journal of Social and Personal Relationships, 14,603–625.

Watson, D., Clark, L. A., & Tellegen, A. (1988).Development and validation of brief measures ofpositive and negative affect: The PANAS scales.Journal of Personality and Social Psychology, 54,1063–1070.

Wortman, C. B., & Silver, R. C. (2001). The mythsof coping with loss revisited. In M. S. Stroebe,R. O. Hansson, W. Stroebe, & H. Schut (Eds.), Hand-book of bereavement research: Consequences, coping,and care (pp. 405–429). Washington, DC: AmericanPsychological Association.