Attachment and Loss

10.1177/0146167204264289 PERSONALITY Fraley, BonannoAND /ATTACHMENT SOCIALPSYCHOLOGY ANDLOSSBULLETIN

 Attachment and Loss:  A Test Test of Three Competing Models Models on the Association Between Attachment-Related Avoidance and Adaptation to Bereavement  R. Chris Fraley  University of Illinois at Chicago  George A. Bonanno Teachers College, Columbia University 

It is wid widely ely ass assume umed d tha thatt emo emotio tional nally ly avo avoida idant nt or def defens ensiveindi-  iveindi-  viduals will have a difficult time adjusting to the loss of a loved  one. However, However, recent research suggests that defensive individu-  als ten tend d to ada adapt pt qu quite ite wel welll to los loss. s. Suc Such h fin findin dings gs pos posee a nu numbe mber  r  of challenges to attachment theory—a theory that has tradition-  tradition-  ally held that emotional avoidance is indicative of poor psycho-  logical adjustment. In this article, the authors argue that con-  temporary tempo rary models of indiv individua iduall differe differences nces in adult attac attachment  hment  allow the derivation of at least three competing hypotheses re-  garding the relationship between avoidant attachment and  adaptation to loss. These hypotheses are tested using two-wave  data on 59 bereaved adults. Results indicate that whereas some  avoidant individuals (i.e., those who are fearfully avoidant)  have a difficult time adapting to the loss of a loved one, other  avoidant adults (i.e., those who are dismissingly avoidant)  show a pattern of resilience to loss.

Keywords:   attachment styles; psychological defense; bereavement;  emotion regulation 

Thelossofalovedoneisthoughttobeoneofthemost 

painful events a person can experience, one that often elicits powerful feelings of anxiety, hopelessness, and sorrow sor row.. Yet the there re is con consid sidera erable ble var variabi iabilit lityy in the deg degree ree to which people experience these feelings (Bonanno & Kaltman, 1999; Nolen-Hoeksema & Larson, 1999;  Wortman  Wor tman & Silver Silver,, 1989). Some people, for example, exhibit extreme signs of distress and may report heightened en ed le leve vels ls of an anxi xiet etyy or de depr pres essi sion on fo forr ye year arss af afte terr a lo loss ss.. Other people, in contrast, appear to be relatively unaffected by loss.

People Peop le wh who o do not sh show ow obv obvio ious us si sign gnss of di dist stre ress ss hav havee receiv rec eived ed a lot of the theore oretic tical al att attent ention ion by soc social ial,, per person sonalality,, and clinical psychologists (see Bonan no & Kaltman, ity 1999;; Wort 1999 ortman man & Sil Silver ver,, 1989 1989,, for rev review iews). s). It is gen genera erally  lly  beli be liev eved ed that that th thee fa fail ilur uree to ex exhi hibi bitt cl clea earr si signs gns of gr grie ieff is a maladaptive response to loss and that people who exhibit few grief symptoms immediately after a loss are likely to show signs of poor adjustment in the long run (Deutsch, 1937; Lazare, 1989; Lindemann, 1944; Osterweis, Solomon, & Green, 1984; Raphael, 1983; Sanders, 1993). This belief is a pervasive one, cutting across both lay and scientific theories of bereavement. Howeve How everr, onl onlyy rec recent ently ly have sci scient entist istss beg begun un to stu study dy the empiri emp irical cal rel relati ations onship hip bet betwee ween n theway peo peoplecope plecope wit with h loss and long-term adaptation (e.g., Bonanno & Kaltman, 1999; W. Stroebe & Stroebe, 1987; Wortman & Silver, 1989). One of the surprising findings from this resear res earch ch is tha thatt peo people ple who exh exhibi ibitt fewsympt fewsymptomsof omsof gri grief  ef   within the first few months following loss and, in particu-

 Authors’ Note:  We gratefully acknowledge the contributions of Ben

Hankin and Caroline Tancredy for reading drafts of this article and provid pro vidingus ingus wit with h fee feedba dback ck and sug sugges gestio tions. ns. Theresea Theresearchdescr rchdescribe ibed d in this article was supported by a grant from the National Institute of  Health, R29-MH57274 (George A. Bonanno). Please address correspondence to either R. Chris Fraley, Department of Psychology, Uni versi ty of Illin ois at Urba na- Cham paig n, 603 E. Dani el Street , Champai Cham paign,IL gn,IL 618 61820,or 20,or Geo GeorgeA. rgeA. Bon Bonanno anno,, Dep Depart artmen mentt of Cou Counse nselling and Clinical Psychology, 525 West 120th Street, Te Teachers achers College, Box 218, Columbia Un iversity iversity,, New York, NY 10027. PSPB, Vol. 30 No. 7, July 2004 878-890 DOI: 10.1177/0146167204264289 © 2004 by the Society for Personality and Social Psychology, Inc.

878

Fraleyy, Bonanno / ATT Frale TTACHME ACHMENT NT AND LOSS lar, people who exhibit seemingly defensive or avoidant  lar, approaches to loss suffer less in the long run than those  who actively express or “work through” their grief (see Bonanno & Field, 2001; Bonanno & Keltner, 1997; Bonanno, Keltner, Holen, & Horowitz, 1995; Bonanno, Znoj, Siddique, & Horowitz, 1999). These findings pose a number of questions for classical and contemporary  theori the ories es of ber bereav eaveme ement nt tha thatt hol hold d thatputat thatputative ive avo avoida idant  nt  and defensive reactions to loss have negative implicati impl ications ons for adjustment and adapta adaptation tion (Fraley & Shaver,, 1999). Shaver 1 999). The objective of this article is to explore the implications tio ns of suc such h fin findin dings gs for one of the mor moree vis visibl iblee cont contememporary theories of bereavement—attachment theory  (seee Bowl (se Bowlby by,, 1980 1980;; Fra Fraley ley & Shav Shaver er,, 1999 1999;; Par Parkes kes & Weis eiss, s, 1983). Specifically, we will address the question of how  people who defensively organize their attachment  behavior (i.e., avoidant adults) adjust to the potentially  disruptive impact of the death of a loved one. We argue that contemporary attachment-theoretical perspectives allow the derivation of three alternative hypotheses regarding the association between avoidant attachment  and adaptation to loss. We will empirically test those hypotheses by studying patterns of adaptation in a sample of bereaved adults. Doing so should allow us to resolve ambiguities concerning the association between attachment patterns and bereavement as well as to address contemporary debates about the role of adaptation in processes that have been viewed historically  as defensive and maladaptive in nature. We begin by  reviewing revie wing class classic ic and contem contemporary porary attac attachment hment theor theoretetical views on the role of individual differences in attachment organization (i.e., attachment style) in the regulation of grief-related feelings, thoughts, and behaviors. CHRONIC GRIEF, ABSENT GRIEF,  AND ATTACHMENT ATTACHMENT ORGANIZATION

One of Bowl Bowlby’ by’ss imp import ortant ant cont contrib ributi utions ons to the lit litera era-ture on bereavement was a normative account of attachment and loss, one that drew on insights from ethology  and was rooted in a lifespan perspective (see Fraley & Shaver Shav er,, 1999 1999). ). In add additi ition on to off offeri ering ng a fun functi ctional onal anal analyysis of normative reactions to loss, Bowlby (1980) proposed pos ed a fra framew mework ork for con concep ceptua tualiz lizing ing aty atypic pical al for forms ms of  mourni mou rning. ng. Hisanalys Hisanalysis is of the these se dis disord ordere ered d for forms ms of gri grief  ef  suggested that they could be arrayed along a single co nceptual dimension, anchored by chronic mourning on one extreme and prolonged absence of o f conscious grieving in g on th thee ot othe herr (B (Bowl owlby by,, 19 1980, 80, p. 138 138). ). Ch Chro roni nicc mo mour urnning is characterized by protracted grief and prolonged difficulty in normal functioning. Individuals who suffer from chronic mourning may find themselves overly pre-

879

occupied with thoughts of their missing partners and unable unab le to ret return urn to nor normal mal fun functi ctioni oning ng for mon months ths aft after er thee lo th loss ss.. In co cont ntra rast st,, an ap appa pare rent nt abs absen ence ce of gr grie ieff is ch char ar-acterized by a conspicuous lack of conscious sadness, angerr, or dis ange distre tress. ss. Acc Accord ording ing to Bow Bowlby lby (19 (1980) 80),, ind indivi ividuduals exhibiting an apparent absence of grief may express relatively little distress following the loss, continue in their the ir jobs or act activi ivitie tiess wit withou houtt any not notice iceabl ablee dis disrup ruptio tion, n, and see seek k lit little tle sup suppor portt or sol solace ace fro from m fri friend endss and fam family ily.. It was Bow Bowlb lby’ y’ss be beli lief ef th that at th this is ma manne nnerr of re reac acti ting ng to lo loss ss may lead to difficulties in long-term adjustment.  A key component component of Bowlby’s Bowlby’s theoretical analysis was that th at th thee way pe peop ople le mourn mourn,, and wh whet ethe herr it is ad adap apti tive ve in the long run, can be understood partly as a function of  their attachment histories. Specifically, Bowlby (1980) argued that adults with anxious attachment histories (i.e., (i. e., his histor tories ies of ins insecu ecurit rityy, inc incons onsist istent ent car care, e, and per persis sis-tent frustration of attachment-related needs) would be more likely to exhibit prolonged or chronic grief,  whereas individuals for whom attachment-related needs had been consistently rebuffed or rejected would be more likely to express few overt signs of grief. Although Bowlby was fairly explicit about the ways in which these attach att achmen mentt exp experi erienc ences es bec become ome orga organiz nized ed int into o a cohe coherrent pat patter tern n of rel relati ating ng to oth others ers,, he didnot exp explic licitl itlyy lin link k his ideas about grief to a theoretical model of individual differences in attachment organization. In the late 1980s, a number of researchers began to develop taxonomies of individual differences in attachment organization for adults (e.g., Hazan & Shaver, 1987; Main, Kaplan, & Cassidy, 1985) based on the patterns of attachment that had been documented by   Ainsworth and her colleagues in infant research using the strange situation (see Ainsworth, Blehar, Waters, &  Wall, 1978). Hazan and Shaver (1987), for example, proposed a three-group taxonomy involving security, security, avoidance, and anxious-ambivalence (see Fraley & Shaver, 2000, for a review). When Bowlby’s ideas are considered in the context of Hazan and Shaver’s (1987) threecategory model, the theoretical linkages between attachment organization and grief are clear: Secure indi viduals should be distressed by the loss of a loved one but  should find it easier than others to adapt to the loss.  Anxious-ambivalent or preoccupied individuals should exhibit exhi bit a pat patter tern n of chr chroni onicc gri grief, ef, wher whereas eas avoi avoidant dant ind indii viduals should show an apparent absence of overt grief  symptoms (see Field & Sundin, 2001; Hazan & Shaver, 1992; Shaver & Tancredy, 2001; Wayment & Vierthaler Vierthaler,, 2002). 2002 ). Des Despit pitee thi thiss rel relati ative ve abs absenc encee of ove overt rt ind indica icator torss of  grieving, grievi ng, howeve howeverr, avoida avoidant nt indiv individual idualss shoul should d have diffi diffi-culties recovering from the loss—difficulties that may  manifest themselves later (see Shaver & Tancredy, Tancredy, 2001, for further discussion).

880

PERSONALIT PERSO NALITY Y AND SOCIA SOCIAL L PSYCHO PSYCHOLOGY LOGY BULLE BULLETIN TIN

CONTEMPORARY RESEARCH AND DEBATES ON ATTACHMENT, BEREAVEMENT, AND DEFENSE

 Although attachment researchers have assumed an association between avoidance and maladaptive or delayed grief, contemporary research on adaptation and loss has challenged this assumption. Bonanno and his colleagues, for example, have searched for, but not  found, empir empirical ical evide evidence nce for a patte pattern rn of delaye delayed d grief grief.. People Peo ple who exh exhibi ibitt few sig signs ns of gri grief ef sho shortl rtlyy aft after er the los losss of a pa part rtne nerr ra rare rely ly ex exhi hibi bitt a he heig ight hten ened ed in incr crea ease se in gr grie ief  f  symptoms months or years later (see Bonanno et al., 1999;; Bona 1999 Bonanno nno & Fie Field, ld, 2001 2001;; Bon Bonanno anno,, Wort ortman man,, et al. al.,, 2002). Moreover, the use of seemingly defensive or avoidan avoi dantt str strate ategie giess doe doess not conf confer er gre greate aterr ris risk k for dev develeloping depressive symptoms either in the early or later months following loss (Bonanno et al., 1995; Bonanno, Field, Kovacevic, & Kaltman, 2002). For example, Bonanno et al. (1995) examined emotional avoidance among recently conjugally bereaved individuals by contrasti tra sting ng the their ir sub subjec jectiv tivee rep report ortss of dis distre tress ss wit with h the their ir lev level el of autonomic arousal following an interview in which they discussed the death of their spouse. Emotionally  avoidant participants were defined as those who reported relatively little distress while concurrently showing elevated heart rates during the interview (see  Weinberger,  Weinber ger, Schwartz, & Davidson, 1979, for a similar conceptualization). Bonanno and his colleagues found that bereaved participants who exhibited a verbalautonom aut onomic ic dis dissoc sociat iation ion had the few fewest est gri grief ef sym sympto ptoms ms at  6, 14, 25, and 60 months after the loss (Bonanno et al., 1995, 1999; Bonanno & Field, 2001).  Although  Althoug h the research by Bonann Bonanno o and his colleagues has not explicitly addressed the relationship between reactions to loss and avoidant attachment as it  has bee been n def define ined d and mea measur sured ed by att attach achmen mentt res resear earchchers, these findings clearly pose some challenges to attachment theory. If emotionally avoidant adults are psychologically vulnerable, we would expect them to show an increase in symptoms of psychological distress over time. At the very least, we would expect them to be  just as poorly adjusted in the long run as those who more readily express their grief. Although we believe these findin fin dings gs pos posee a num number ber of pro proble blems ms forthe cla classi ssicc thr threeeecategory model of individual differences in attachment, recent advances in the study of adult attachment make the link between avoidance and grief more nuanced than what has been assumed previously. In the early  1990s, Barthol Bartholomew omew introduc introduced ed an alter alternative native model of  individual differences that makes a critical distinction among different kinds of avoidant strategies (see Bartholomew, 1990; Bartholomew & Horowitz, 1991). Specifically, Bartholomew distinguished between fearfulavoidant and dismissing-avoidant attachment patterns.

 According to Bartholomew, fearfully avoidant adults,  while organizing their behavior in a defensive manner, tend to do so as an attempt to quell their insecurities. Because they are explicitly afraid of being hurt or reject rej ected,they ed,they avo avoid id ope openin ning g up to othe others rs and try to avoi avoid d becoming emotionally invested in them. The avoidant  strategies of dismissing individuals, in contrast, appear to be organized around the goal g oal of self-reliance or independence. Although dismissing individuals may avoid opening openin g up to and depen depending ding on other others, s, it is because they consciously see little need to forge close emotional bonds with others, not because they consciously fear being hurt. Theoretically, this strategy is motivated by a historyof his toryof rej reject ection ion but cons conscio ciousl uslyy it is roo rooted ted in a des desire ire to be autonomous and self-reliant. Griffin and Bartholomew (1994) argued that these theoretical “prototypes” can be arrayed along two conceptual dimensions that researchers have come to call attachment-related anxiety and attachment-related avoidance (see Fraley & Shaver S haver,, 2000). This first dimension captures captures variation in the degree to which people are vigilantly attuned to attachment-related concerns (Fraley & Shaver, 2000). A highly anxious person, for example, may worry that one’s attachment figure is unresponsive, whereas a less anxious person may feel relatively secure about attachment-related matters. The seconddimension captu captures res varia variation tion in peopl people’s e’stende tendenncies to use avoidant versus proximity-seeking strategies to regulate attachment-related behavior, thought, and feeling. People on the high end of this dimension tend to wit withdr hdraw aw fro from m clo close se rel relati ations onship hips, s, whe wherea reass peo people ple on the low end of this dimension are more comfortable opening up to others and relying on others as a secure basee (F bas (Fra rale leyy & Sh Shave averr, 20 2000) 00).. As is il illu lust stra rate ted d in Fi Figu gure re 1, the classic Hazan and Shaver categories, as well as Bartholomew’s four theoretical prototypes, can be  viewed as linear combinations of these two dimensions (see Fraley & Waller, Waller, 1998). For example, a prototypical fearful individual is relatively attentive to attachmentrelate rel ated d conc concern ernss (i. (i.e., e., is high on the anxi anxiety ety dim dimens ension ion)) and typically employs avoidant strategies to regulate h is or her feelings and behavior (i.e., is high on the avoidance dime dimension) nsion).. By contr contrast, ast, a protot prototypical ypically ly dism dismissi issive ve individual also employs avoidant strategies but is less attentive to or downplays attachment-related concerns (i.e., is low on the anxiety dimension). A prototypically  secure individual is low on both of these dimensions. 1 One of the valuable features of this two-dimensional model mod el is thatit dis distin tingui guishe shess dif differ ferent ent kin kinds ds of def defens ensive ive strategies. The classic Hazan and Shaver conception of  avoidance implied that avoidance stems from conscious insecurities, thereby conflating dismissing-avoidance  with fearful-avoidance (see Fraley & Shaver Shaver,, 2000). How-

Fraleyy, Bonanno / ATT Frale TTACHME ACHMENT NT AND LOSS

LOW AVOIDANCE SECURE

PREOCCUPIED

LOW

HIGH

ANXIETY

ANXIETY

DISMISSING-

FEARFUL-

AVOIDANT

AVOIDANT HIGH AVOIDANCE

Figure Figur e1

The two-dimens two-dimensional ional model model of individual individual dif differe ferences nces in attachment.

NOTE: This diagram illustrates the th eoretical relationships between attachment-related avoidance and anxiety and Bartholomew’s (1990) attachment prototypes.

ever, once the distinction between different forms of  avoidance is recognized, at least three alternative hypotheses about the relationship between attachment  patterns and reactions to loss can be derived from attachment theory. One hypothesis is that attachmentrelated rela ted anxi anxiety ety,, but not att attachm achment ent-re -relat lated ed avoi avoidan dance, ce, is asso as sociat ciated ed withsevere grie grief. f. The hypo hypothes thesis is is base based d on the assumption that the control systems underlying attachment-relate attachment-r elated d vigila vigilance, nce, which are tapped by the anxiety dimension, may produce maladaptive grief  responses (see Fraley & Shaver, 2000). Theoretically, highly anxious people are more vigilant and insecure regarding the psychological accessibility and availability  of their loved ones. Thus, when their loved one is missing, they should not only experience distress, but that  distress should be triggered more easily by day-to-day  reminders (both physical and psychological) of the love lo ved d on one’ e’ss ab abse senc nce. e. Ov Over er a pe peri riod od of ti time me,, th thee re repe peat ated ed activation of these stressful emotions, coupled with the failure of attachment behavior to reunite the person  with his or her loved one, may heighten feelings of distress, hopelessness, and despair. By contrast, avoidance by itself should not necessarily lead to difficulties adjusting to ber bereave eavemen ment. t. Fra Fraley ley,, Dav Davis, is, and Shav Shaver er (19 (1998) 98),, for example, have argued that avoidant strategies of emotion regulation (e.g., withdrawing from situations that 

881

make one fee make feell vul vulner nerabl able) e) can be jus justt as eff effect ectiveas iveas pro proxximity-seeking strategies (e.g., turning to others for support or comfort) in regulating attachment-related distress. According to Fraley et al. (1998), avoidant indi viduals sometimes appear vulnerable because researchers do not always distinguish between fearful-avoidance and dismissing-avoidance, thereby conflating attachment-related anxiety and avoidance. Once the distinction is made between alternative patterns of avoidance,  we should expect some avoidant individuals (i.e., those  who are fearfully avoidant) to have difficulty adjusting to loss because they are highly anxious with respect to attachment concerns, but we should expect other avoidant individuals (i.e., those who are dismissingly  avoidant) to have less difficulty adjusting to loss because they are not as anxious with respect to attachment concerns. This model implies a main effect of attachmentrelated anxiety in predicting patterns of grief but no main effect of attachment-related avoidance.  An alternative hypothesis is that both attachment  dimensions are associated with maladaptive grief, such that people who are high on both attachment-related anxiety and avoidance are more likely to experience difficulty following loss. This model holds that attachmentrelated avoidance is a maladaptive strategy, but within the context of Bartholomew’s prototypes, fearfulavoidance will be more strongly associated with psychological distress than dismissing-avoidance. This model implies main effects of both attachment-related anxiety  and avoidance in predicting patterns of adjustment to loss.  A third hypothesis is that highly secure adults will adapt most easily following loss and that insecure adults  will have an equally difficult time adjusting. According to this model, preoccupied, fearful, and dismissing adults  will experience high levels of distress following loss. This hypothesis is rooted in the assumption that dismissing adults, adult s, althou although gh defen defensive sive in many resp respects, ects, are actuall actually  y  quite vulnerable (see Dozier & Kobak, 1992; Klohnen &  John, 1998). If correct, highly dismissing adults should exhibit the same degree of maladjustment as other insecure adults. This model implies a main effect of each attachment dimension as well as a negatively weighted interaction between the two dimensions in predicting adjustment. OVERVIEW OF THE PRESENT RESEARCH

To tes testt the these se com compet peting ing hyp hypoth othese esess we exa examin mined ed pat pat-terns of bereavement in a sample of 59 bereaved adults.  At 4 and 18 months postloss we obtained o btained ratings from structured clinical interviews for symptoms of anxiety, depression, grief, and post-traumatic stress disorder

882

PERSONALIT PERSO NALITY Y AND SOCIA SOCIAL L PSYCHO PSYCHOLOGY LOGY BULLE BULLETIN TIN

(PTSD) and ratings of participant’s overall adjustment  from fro m at lea least st two clo close se fri friend ends. s. By exa examin mining ing the pat patter tern ning of sym sympto ptoms ms and adj adjust ustmen mentt ove overr tim time, e, and the their ir relationship to individual differences in adult attachment,  we hope to take an important important step toward documenting empirically the associations between individual differences enc es in att attach achmen mentt org organi anizat zation ion and pat patter terns ns of gri grief ef as  well as clarifying the role of alternative defensive strategies in adaptation to loss. Method  PARTICIPANTS AND PROCEDURE 

There are a number of practical and ethical challenges involved in recruiting samples of bereaved indi viduals (see Bonanno & Kaltman, 1999). In the present  study,, we adopted what seemed to be the most ethically  study sound approach to obtaining research participants: We disseminated information about the study and encouraged individuals interested in participating to contact  the researchers (Penslar, 1993). Information about the study was made available to potential participants living in the Washington, D.C. metropolitan area by sending letter let terss des descri cribin bing g the stu study dy to (a) rec recent ently ly ber bereav eaved ed ind indii viduals who were listed as surviving parents or spouses in newspaper obituary notices and (b) individuals likely to have contact with bereaved individuals (e.g., medical and mental health professionals, clergy). The letters encouraged bereaved individuals who met recruitment  crit cr iter eria ia an and d who mi mightbe ghtbe in inte tere rest sted ed in jo join inin ing g th thee st stud udy  y  to contact the researchers by phone or mail. Seventyfourr ber fou bereav eaved ed ind indivi ividua duals ls con contac tacted ted the res resear earche chers rs and agreed to participate in the study. Oncee enr Onc enroll olled ed in thestudy thestudy,, ber bereav eaved ed ind indivi ividua duals ls com com-pleted plete d mailmail-in in quest questionnai ionnaires, res, inclu including ding a quest questionnai ionnaire re assessing individual differences in attachment organization (see below). Participants also took part in structured clinical interviews at approximately 4 and 18 months mon ths pos postlo tloss. ss. Par Partic ticipa ipants nts wer weree pai paid d $60 eac each h wav wavee of  data colle collection. ction. Four parti participant cipantss (5%) were unavai unavailable lable for or refused to take part in the first interview and 11 participant parti cipantss (15%) were unavailable for or refus refused ed to take part in the second interview. The final sample consisted of 59 participants, with an average age of 51 years (SD   = 7.8 years). The majority of participants were  women (64.4%). This sample did not differ on any of the measures included in the present study compared with those participants who were recruited but dropped out  by the second interview ( p s > .10). Ratings of participant’s adjustment by close friends also were available from a subsample of participants. Although we were primarily interested in studying the association between attachment patterns and clinical ratings of adjustment,  we also report ratings of participants’ adjustment pro-

 vided by their friends friends among this this subsample because we believe that it provides an additional lens by which we may view the adaptation process. MEASURES 

Structured clinical interviews . Participants were asked a series ser ies of que questi stions ons cor corres respond ponding ing to the DSM-IV (American Psychiatric Association, 1994) symptoms for Generalized ali zed Anxi Anxiety ety Dis Disord order er (9 ite items ms,, α = .78) .78),, Majo Majorr Dep Depres res-sive Disorder (8 items, α = .92), and symptoms of Posttraumatic Stress Disorder that did not overlap with Major Depr Depressio ession n (14 item items, s, α =.82).Inaddition,thefollowing eight symptoms of severe grief were defined based on previous bereavement studies (e.g., Horowitz et al., 1997): strong yearning for the deceased, preoccupreoccu pation pat ion wit with h tho thought ughtss abou aboutt the los loss, s, rec recurr urrent ent reg regret retss or self-blame about own behavior toward the deceased, recurrent regrets or blame regarding the behavior of  others toward the deceased, difficulty accepting the finality of the loss, marked loneliness or sense of emptiness, pervasive sense that life is meaningless, and unusual difficulty developing new relationships ( α = .69). Each symptom was coded as present or absent following the for format mat use used d for the Str Struct ucture ured d Cli Clinic nical al Int Intervi erview  ew  for the DSM (Spitzer, Williams, Gibbon, & First, 1990). Specif Spe cifica ically lly,, eac each h ite item m inc includ luded ed an exp explic licit it sco scorin ring g cri criteterion rio n (e. (e.g., g., mar marked kedly ly inc increa reased sed sad sadnes nesss or dis distre tress ss dur during ing the past month in situations that symbolize or remind the subject of the deceased) and a set of standard questions desig designed ned to elici elicitt informat information ion relevant relevant to the crite crite-rion (Bonanno et al., 1995; Horowitz et al., 1997). The interviewer’s decision as to whether the criterion was mett fo me forr eac each h it item em wa wass bas based ed on a co comb mbin inat atio ion n of pa part rtic iciipant report and the interviewer’s observations during the interview. The interviews were conducted by seven doctoral candidates in clinical psychology. Interviewers received extensive training in the procedures but were blind to both the goals and hypotheses of the current  study and participants’ responses to the questionnaire measur mea sures, es, inc includ luding ing the att attach achmen mentt mea measur sures. es. For com com-putation of interrater reliability, the interviews were videotaped and each interviewer coded a randomly  selected set of five additional interviews. Interrater reliability was very high (average κ  =  = .97). Adult attach attachment ment patte patterns  rns . Individual differences in adult attachment were assessed using the 30-item Relationship Scales Questionnaire (RSQ; Griffin & Bart Ba rtho holo lome mew w, 19 1994) 94).. Ea Each ch it item em wa wass ra rate ted d on a 1 (strongly  disagree ) to 7 (strongly agree ) scale. The RSQ contains items ite ms des design igned ed to tap eac each h of Bar Bartho tholom lomew’ ew’ss fou fourr pro protototype ty pess as we well ll as it item emss dr draw awn n fr from om th thee or orig igin inal al Ha Haza zan n and Shaver prototypes. Following procedures similar to those described elsewhere (e.g., Fraley, Garner, & Shaver Shav er,, 2000 2000;; Fra Fraley ley & Wa Walle llerr, 1998 1998), ), we use used d the these se ite items ms

Fraleyy, Bonanno / ATT Frale TTACHME ACHMENT NT AND LOSS to scale people on two dimensions: attachment-related anxiety anxi ety and avoi avoidan dance. ce. The These se two sca scales les wer weree weak weakly ly cor cor-related in the present sample ( r  =   = .17) and exhibited acceptable internal consistency estimates of reliability  (αs = .84 and .70 for anxiety and avoidance, respectively). The average anxiety score was 3.00 ( SD  =  = 1.03); the average avoidance score was 3.79 (SD  =  = .72).  Friend ratings . Each participant distributed rating materials to three close friends who they felt knew them  well and with whom they had relatively consistent contact. The materials asked the person to rate the participant’s adjustment along five dimensions (i.e., mental health, physical health, quality of social interactions, ability to accomplish goals, and coping ability) compare pa red d to th thee ave avera rage ge pe pers rson on of th thee sa same me age an and d ge gend nder er..  A total adjustment score was computed by averaging the five dimensions (α  = .90). To ensure confidentiality, friends returned these ratings using stamped envelopes preaddressed to the researchers. A participant’s friend data were used only when ratings from at least two friends were available. Participants with usable friend data (N  =   = 38) did not differ significantly from the remainder of the sample ( N  =  = 21) on any of the measures used us ed in cu curr rren entt st stud udyy ( p s > .15 .15). ). On av aver erage age,, fr frie iend ndss pr proo viding ratings had known the participants for 15 years (SD  =  = 13).

     0      1

883

chronic grief

     8

delayed grief      6     s     m     o      t     p      4     m     y      S

common grief

     2

     0

resilient grief

     2    -

0

5

10

15

20

Months since loss

Figure Figur e2

Prototypi Pro totypical cal patterns patterns of bereav bereavement ement..

NOTE: This diagram illustrates the way in which some commonly discussed patterns of grief can be modeled using linear growth curves.

Results  MODELING PATTERNS OF B EREA EREAVEMENT  VEMENT 

Because we were able to obtain measurements across two time points, we were able ab le to conduct within-person analyses to model symptom patterns for each person, as  well as between-person analyses to model the way those patterns varied as a function of individual differences in attachment. To To model the within-person data across the two waves, we estimated the parameters of a simple linear model, Y =  a  +  +  b X, X, for each person. In this withinperson model, the parameter a  (i.e.,  (i.e., the intercept) represents the initial symptom levels exhibited by that indi vidual, the parameter  b  (i.e.,   (i.e., the slope) represents the ratee at whi rat which ch tho those se sym sympto ptoms ms cha changedfrom ngedfrom one poi point nt in time to the next, the variable Y represents the person’s adjustment levels for the two time points, and the variable X represents the amount of time (in months) since thefirstassessment(coded0forwave1and15forwave2,  which occurred 15 months after the first assessment). assessment). One of the advantages of modeling the data in this manner is that it allows us to study  patterns  of  of grief and adjustment. To highlight this idea, we have graphed in Figure 2 some configurations that are commonly discussed in the bereavement literature. Following the death of a spouse, some people may exhibit a high degreee of dis gre distre tress ss but sho show w a sub substa stanti ntial al dec decrea rease se in dis distre tress ss

over the course of a year. This prototypical pattern of  recovery, what Bonnano et al. (2002) call common grief ,  would be modeled mathematically by b y a large intercept   value and a negative slope. Other patterns of theoretical intere int erest st can be rep repres resent ented ed in thi thiss fra framew mework ork as wel well. l. For example, some people may exhibit high symptom levels initially but show little or no change in symptom levels overr tim ove time. e. Thi Thiss pat patter tern n wou would ld be ind indica icativ tivee of chro chronic nic grief  and wou would ld be model modeled ed with with a lar large ge interc intercept ept term term and a slope of zero. Other people may show no change in symptom levels over time but also may have had low distress levels initially. This pattern would be indicative of  some so me fo form rm of resilienc e (Bo (Bonan nanno, no, 2003 2003)) and and,, mat mathem hematatical ic ally ly,, wou would ld be mo mode dele led d wi with th a sm smal alll in inte terc rcep eptt te term rm and a slope of zero. Other people, although exhibiting few  symptoms initially, may show an increase in symptoms over time. As discussed earlier, this theoretical pattern, referred to as  delayed grief , is widely believed to result  from avoidant defenses early duri during ng the berea bereavement  vement  process (e.g., Middleton, Raphael, Martinek, & Misso, 1993) but has yet to be demonstrated empirically. The pattern of delayed grief would be indicated by a small intercept term and a positive slope.  Another advantage of modeling the data data in this manner is that it represents patterns of change continuously  rather than categorically. This allows us to study the

884

PERSONALIT PERSO NALITY Y AND SOCIA SOCIAL L PSYCHO PSYCHOLOGY LOGY BULLE BULLETIN TIN

TABLE TAB LE 1: Parame Parameter ter Estim Estimates ates for for Symptom Symptom Patte Patterns rns

Bereavement-Related Anxiety

M  SD  Range

Depression

Grief

P TSD  

Intercept

Slope

Intercept

Slope

Intercept

Slope

Intercept

.41 1.39 0-6

.02 (.27) .11 –.33-.47

1.69 1.96 0-8

–.03 (–0.47) .08 –.27-.13

2.47 1.86 0-8

–.08 (–1.20) .10 .1 –.33-.13

3.08 2.60 0-11

Slope  

–.06 (–.93) . 12 .1 –.33-.27

NOTE: Theavera NOTE: Theaverage ge int interc erceptvaluerepre eptvaluerepresen sents ts theaverag theaveragee sym sympto ptom m lev levelsin elsin thesampleat thefirstassess thefirstassessmen mentt poi point. nt. Theavera Theaveragesloperepre gesloperepresent sentss theaverage theavera ge amo amountof untof cha changein ngein sym sympto ptom m lev levelsper elsper mon month.The th.The val valuesliste ueslisted d in par parent enthes heses es rep repres resentthe entthe ave averag ragee amo amountof untof sym symptomchang ptomchangee between the two assessment points, computed as 15  ×  the average slope. PTSD = post-traumatic stress disorder. TABLE 2: Modeling Variation Variation in Intercepts and Slopes for Each Symptom Pattern as a Function of Individual Differences Differences in Attachment 

Bereavement-Related Anxiety Attachment Variables

 Anxiety  Avoidance  Anxiety  ×  × Avoidance 2 R 

Depression

Grief

Intercept

Slope

Intercept

Slope

Intercept

.29* .10 .15 .11*

.31* –.08 –.04 . 10* .1

.52* .08 .15 .29*

–.07 –.02 –.15 .02

.37* .18 .23† .20*

PT SD   Slope

–.12 .05 –.07 .02

Intercept

.50* .20† .23* .33*

Slope  

–.15 –.06 –.24† .07

NOTE: The coeff NOTE: coefficien icients ts are standar standardizedregressi dizedregression on weigh weights ts estim estimated ated for the fullmodel (i.e.,including anxie anxiety ty,, avoida avoidance, nce, and theirinteracti theirinteraction). on). PTSD = post-traumatic stress disorder. † p  <  < .10. * p  <  < .05.

actual patte pattern rn of sympt symptoms oms exhibi exhibited ted by a perso person n rather rather than classifying a person’s pattern into a broader category that may not describe him or her well. Finally, by  summarizing patterns with two parameters that vary  across people, we can model variation in these parameter estimates as a function of individual differences in attach att achmen ment. t. Doi Doing ng so all allows ows us to spe specif cifyy how alt altern ernati ative ve configurations of the two attachment dimensions are associated with various patterns of change.  After estimating the within-person parameters, we conducted between-subjects analyses in which we modeled the estimated intercept and slope parameters as a function funct ion of indiv individual idualdiffe difference rencess in attac attachment.Specifihment.Specifically, we estimated the parameters of two higher-order regression regre ssion equations. The firs firstt equat equation ion model modeled ed the  variation in intercept terms and the second equation modeled variation in the slopes, each as a function of  attachment-related anxiety, avoidance, and the interaction of anxiety and avoidance. NORMATIVE PATTERNS 

Tabl ablee 1 rep report ortss the des descri cripti ptive ve sta statis tistic ticss for the wit within hin-person parameters in the sample. As can be seen, there  was a general tendency for people to begin with high symptom levels 4 months postloss postloss and decrease in symptom levels over time. The only notable ex ception to this rulee was for sym rul sympto ptoms ms of anxi anxiety ety.. On aver average,there age,there was a slight increase in symptoms of anxiety over time. To help contextualize the meaning of the intercepts,  we assessed symptom levels for anxiety, depression, depression, and PTSD in an age-matched sample of nonbereaved indi-

 viduals. The average symptom levels in the nonbereaved sample were .33 (SD  =  = 1.09), .57 (SD  =  = 1.34), and 2.44 (SD  =   = 4.31) for anxiety, depression, and PTSD, respectively. When the statistics reported in Table 1 are considered in light of these averages, they indicate that  the bereaved sample had elevated symptoms on each measure. INDIVIDUAL DIFFERENCES IN ATTACHMENT  ATTACHMENT 

In our next set of analyses we regressed the withinperson parameters onto the two standardized attachment dimensions (i.e., attachment-related anxiety and avoidance) as well as the interaction between the two attachment dimensions. This between-subjects analysis allows us to model patterns of change as a function of  attachment style. The estimated standardized coefficients for the intercept and slope regressions are presented in Table 2. To illustrate the prototypical patterns of chan change ge imp implie lied d by the these se est estima imates tes,, we plo plotte tted d the pre pre-dicted symptom patterns for the different attachment  styles in each analysis (see Figure 3) by using the same kinds kin ds of tec techni hnique quess tha thatt are use used d to ill illust ustrat ratee int intera eracti ction on patterns in multiple regression (see Aiken & West, 1991). Because many of the theoretical issues addressed in this article are related to Bartholomew’s theoretical prototypes proto types(i.e. (i.e.,, secur secure, e, fearful fearful,, dism dismissi issing, ng, and preocc preoccuupied), we plotted the predicted patterns for each of  these four theoretical attachment patterns as derived from the two dimensions. Specifically, the pattern for security was derived by substituting values of –1 for anxi-

Fraleyy, Bonanno / ATT Frale TTACHME ACHMENT NT AND LOSS

Anxiety

885

Depression

     3

     4

F      2     s     m     o      t     p      1     m     y      S      0

F P F P S D

S D

     1    -

5

10

P

F P

S D

D S

0

5

10

Months

Months

Grief

PTSD

15

     8

F      6

F

P S D

P S D

     1      0

    s     m     o      t     p      4     m     y      S      2

F

F

P

P S D

D S

     0

0

5

10

15

Months Figure Figur e3

     1

15

     5

    s     m      3     o      t     p     m     y      2      S

    s     m     o      t     p      2     m     y      S

     0

0

     4

     3

0

5

10

15

Months

Patterns Patte rns of bereavem bereavement ent as a function function of attachme attachment nt pattern. pattern.

NOTE: Theupper NOTE: Theupper-le -left ft fig figureillus ureillustra tratesthe testhe imp implie lied d patt pattern ernss of cha changein ngein anx anxiet ietyy sym sympto ptoms ms forpeopleof vary varyingattach ingattachmen mentt patt pattern ernss (S = Sec Secure ure,, P = Preoccupied, F = Fearful, D = Dismissing). The upper-right, lower-left, and lower-right panels illustrate the implied patterns of change for depression, pressi on, grief grief,, and post-t post-traumat raumatic ic stres stresss disord disorder er sympt symptoms, oms, respe respective ctively ly,, forpeople of varyingattachmen varyingattachmentt patter patterns. ns. The dashedlines repr representthe esentthe average level of symptoms in a comparable sample of nonbereaved adults.

ety and avoidance in the estimated regression equation becaus bec ausee theproto theprototyp typica icall sec secureindiv ureindividu idual al haslow sco scores res on anxiety and avoidance. The pattern for dismissing avoidance was derived by substituting –1 for anxiety and + 1 for avoidance because the prototype of dismissingavoida avo idance nce inv involv olves es low sco scores res on anxi anxiety ety and hig high h sco scores res on avoidance (see Bartholomew & Horowitz, 1991;

Fraley et al. Fraley al.,, 1998 1998). ). (Th (Thee pro protot totype ypess are sim simple ple rot rotati ations ons of th thee tw two o di dime mens nsio ions ns so th thee re resu sult ltss il illu lust stra rate ted d in th thee fi figguress are dir ure direct ectly ly rel relate ated d to the anal analyse ysess rep report orted ed in Tabl ablee 2 [see Fraley & Waller, 1998].) Thee mo most st im impo port rtan antt fe feat atur uree to not notee ab abou outt th thee Anxiety . Th prototypical patterns for anxiety symptoms (see the

886

PERSONALIT PERSO NALITY Y AND SOCIA SOCIAL L PSYCHO PSYCHOLOGY LOGY BULLE BULLETIN TIN

upper left panel of Figure 3) is that both prototypically  secure and dismissing adults (i.e., adults with low scores on the dimension of attachment-related anxiety) exhibited a resil resilient ient pattern: Across both time point points, s, their anxiety symptom levels were lower than those for prototypica typ ically lly pre preoccu occupie pied d and fea fearful rful adu adults lts (i. (i.e., e., adu adults lts who are high in attachment-related anxiety). Also of interest  is the fact that the prototypical preoccupied and fearful patterns were associated with an increase in anxiety  sympto sym ptoms ms ove overr tim time, e, wher whereas eas the there re was no suc such h inc increa rease se associated with secure and dismissing patterns. In short, adults who were highly anxious with respect to attachattach ment exhibited higher initial anxiety symptoms and exhibited an increase in such symptoms over the course of  bereavement bereav ement.. Secur Security ity and dism dismissi issing-avoi ng-avoidance, dance, in contrast,were tra st,were ass associ ociate ated d wit with h a res resili ilientpatter entpattern n of cha change. nge.2  Depression . A similar pair of regression analyses were performed for depression scores. The standardized coefficients are presented in Table 2. To illustrate the protot pro totypi ypical cal pat patter terns ns of chan change, ge, in the upp upper er-ri -right ght pan panel el of Figure 3 we have plotted the predicted symptom patterns ter ns for peo people ple who varyin att attachm achmentstyle entstyle.. Thi Thiss fig figure ure shows that preoccupation and fearful-avoidance were associated associ ated with with higher levels levels of depr depressio ession n over time. Of  importance, attachment organization was not substantially associated with variation in slopes, implying that  highly preoccupied and fearful adults remained at ele vated symptom levels across both assessment assessment waves. It is notewor not eworthy thy tha thatt dis dismis missin sing g avoi avoidanc dancee was not ass associ ociate ated d  with a maladaptive pattern of bereavement. The prototypical dismissing adult had relatively low levels of depression initially and showed no increase in depressive symptoms over time. Similar to the prototypical secure patter pat tern, n, the dis dismis missin sing g pat patter tern n res resemb embled led a res resili ilient ent pat pat-tern of change. Grief symptoms . To study the relationship between attachment patterns and patterns of grief symptoms not  overlapping overla pping with depre depression ssion,, anxiet anxietyy, or PTSD, we again estimated the parameters of two regression equations. The standardized coefficients are presented in Table 2 and the pro protot totypi ypical cal pat patter terns ns of cha change nge are ill illust ustrat rated ed in the lower-left panel of Figure 3. This figure shows that  preoccupation and fearful-avoidance were linked to higher levels of grief over time. Consistent with the results from the depression analyses, both dismissingavoidance and security were associated with a resilient  pattern of change in grief.

stud udyy th thee re relalaPost-traumatic stress disorder symptoms . To st tionsh tio nship ip bet betwee ween n att attach achmen mentt pat patter terns ns and sym sympto ptom m pat pat-terns for PTSD, we again estimated the parameters of  two regression equations, one modeling the intercepts and the other modeling the slopes. The standardized

TABLE TABL E 3:

Modeling Variati Modeling Variation on in Intercepts Intercepts and Slopes Slopes for Friends Friends’’ Ratings of Adjustment as a Function of Individual Differences Differ ences in Attachment 

 Friends’ Ratings of Adjustment  Attachment Variables

Intercept

Slope  

 Anxiety  Avoidance  Anxiety  ×  × Avoidance 2 R 

–.13 –.14 –.09 .04

–.10 –.11 –.12 .03

NOTE:The coeff coefficien icients ts are standar standardizedregressio dizedregression n weigh weights ts estim estimated ated forthe ful fulll mod model(i.e.,includ el(i.e.,includinganxie inganxiety ty,, avoi avoidanc dance, e, and the their ir int intera eracction). N  =  = 38.

coefficients are presented in Table 2 and the prototypical patterns of change are illustrated in the lowerright panel of Figure 3. Notice that preoccupation and fearful-avoidance were linked to higher levels of PTSD symptoms over time. Although the interaction between anxiety anxi ety and avo avoida idance nce was not lar large, ge, it is not notewo eworth rthyy tha that  t  fearful-avoidance was linked to initial levels of PTSD sympto sym ptoms ms mor moree strongl stronglyy tha than n wou would ld be exp expect ected ed on the basis bas is of sim simple ple lin linear ear eff effect ects. s. In oth other er wor words, ds, peo people ple who  were high in both attachment-related anxiety and avoidancee wer anc weree esp especi eciall allyy lik likely ely to exhi exhibit bit sym sympto ptoms ms of PTS PTSD. D. Consistent with the previous findings, dismissing avoidance was not associated with a maladaptive pattern of  bereavement. The prototypical dismissing adult, similar to the prototypical secure adult, had relatively low levels of PTSD symptoms initially and showed no increase in such symptoms over time. SECONDARY ANALYSES OF FRIEND RATINGS 

Ratings of participant adjustment by close friends  were available for a subsample of 38 participants. To study the relationship relationship between attachment patterns and perceptions of adjustment on the part of the participants’ friends, we estimated the parameters of two regressi gre ssion on equ equati ations ons,, one mod modeli eling ng the int interc ercept eptss and the other modeling the slopes for perceptions of adjustment. The standardized coefficients are presented in Table 3. Although the sample size for these analyses is too sma small ll to yie yield ld sta statis tistic ticall allyy sig signif nifica icant nt ass associ ociati ations ons,, it is noteworthy that the pattern of regression weights produces patterns that are in the same direction as those obtained with the clinical ratings of psychological symptoms. The pattern of coefficients for dismissing avoidance was not similar to those implied by a maladaptive pattern of adjustment.  Discussion 

Recent research on bereavement has challenged one of the fundamental assumptions of attachment theory, namely, that avoidant or defensive attachment patterns

Fraleyy, Bonanno / ATT Frale TTACHME ACHMENT NT AND LOSS should be linked to maladjustment. The primary objective of this article has been to examine the implications of such findings for attachment theor y. We have argued that the theor theoretical etical relationsh relationship ip between avoidance avoidance and adaptation to loss is not clear-cut within contemporary  attachment theory. Contemporary models of individual differences in attachment organization draw an important distinction between fearful and dismissing forms of  avoi av oida danc ncee anddeba anddebate tess ex exis istt as to wh whet ethe herr oneor bo both th of  these forms of avoidance should be associated with maladjustment. To distinguish between alternative ways of conceptualizing the relationship between attachment-related avoidance and adaptation to loss, we investigated the associations between patterns of grief and individual differences in attachment in a sample of 59 bereaved adults. Our analyses suggest that whereas attachmentrelated anxiety is associated with elevated symptoms of  grief and distress, attachment-related avoidance, for the most part, is not. More specifically, one theoretical pattern of avoidance, fearful avoidance, is associated with diffic dif ficult ultiesin iesin adju adjusti sting ng to the los losss of a lov loved ed one,where one,whereas as another pattern of avoidance, dismissing avoidance, is associated with a resilient pattern of symptoms. In the sections secti ons that follow follow,, we address address the the impli implication cationss of these findings for research and theory on attachment, as well as research on adaptation, resilience, and defense more generally. IMPLICATIONS FOR RESEARCH  AND THEORY ON ADULT ATTACHMENT  ATTACHMENT 

It is oft often en ass assume umed d tha thatt und underl erlyin ying g the def defens ensive ive qua quallities of dismissing adults are latent, unacknowledged insecurities and vulnerabilities (e.g., Onishi, Gjerde, & Block, 2001). The findings of the present study suggest  thatt dis tha dismis missin sing-av g-avoid oidant ant adu adults lts may be muc much h les lesss vul vulner ner-ablee tha abl than n has bee been n pre previo viousl uslyy bel believ ieved. ed. Whe When n fac faced ed wit with h the los losss of a lov loved ed one, dis dismis missin sing g ind indivi ividua duals ls app appear ear les lesss distre dis tresse ssed d tha than n oth others ers and and,, imp import ortant antly ly,, do not exh exhibi ibitt a resurgence of distress-related symptoms later in time. Simila Sim ilarr to sec secure ure adu adults lts,, the theyy exh exhibi ibitt a rel relati ativel velyy res resili ilient  ent  symptom pattern.  Why is it that dismissing and secure adults—both of   whom are low on the dimension of o f attachment-related anxiety—are able to weather the storm of loss better than other bereaved participants? Fraley and Shaver (2000) have argued that the dimension of attachmentrelated anxiety taps variation in the operation of a critical component in the controls systems underlying attachment attach ment behavior—a component that functions to appraise the psychological availability and responsiveness of one’s partner. Some people, based on their history of experiences in close relationships, have a lower threshold than others for deciding that the loved one is

887

unavailable or inaccessible. When the partner is perceived cei ved as unav unavail ailable able,, the ind indivi ividua duall exp experi erienc ences es anxi anxiety  ety  and turns attention away from exploratory activities (e.g., (e. g., wor work, k, pla playy, cre creati ative ve pro projec jects) ts) and tow toward ard sea search rching ing for the partner. Under normal circumstances, such attachment behavior would be abridged or terminated  when the person reestablishes a sense of “felt security” (Sroufe & Waters, 1977) by relocating the partner, receiv rec eiving ing ass assura urance nces, s, or bei being ng com comfor forted ted.. In the cont context  ext  of loss, however, the search process is likely to be sustained because the failure of these efforts to reestablish proximity is likely to continue to prime the attachment  system, further exacerbating distress and eventually  leading to feelings of despair and hopelessness when those efforts are ineffective. Importantly, this conceptualization of attachment  dynamics dynami cs sugges suggests ts that protot prototypical ypically ly secur securee and dismi dismissssing people are less likely to be b e affected by loss than otherss be er becau cause se th they ey ha have ve a hi highe gherr th thre resh shol old d th than an ot othe hers rs fo forr experiencing insecurity. Nonetheless, we suspect that  the psychological processes that facilitate this state of  affairs are quite different for highly secure and dismissing ind indivi ividua duals. ls. For sec secure ure peo people ple,, it may be thecase tha that  t   when they experience the pain of loss, they draw on positivee mem tiv memori ories es or exp experi erienc ences es wit with h the their ir par partne tnerr in a way  that brings them solace and comfort. Moreover, it may  be the case that they come to experience a “continuing bond” with the deceased individual, individual, allowing them to use the deceased individual as a secure base despite the person’ per son’ss phy physic sical al abs absenc encee (se (seee Fra Fraley ley & Shav Shaver er,, 1999 1999). ). In short, highly secure people may be able to draw on a broad array of psychological resources both to minimize the likelihood that the loss will trigger attachmentrelated distress and to effectively regulate the anxiety  that is experienced.  We suspect that the processes that facilitate resilience on the part of highly dismissing adults are quite different.. Fra ent Fraleyand leyand hiscolle hiscolleague aguess hav havee art articu iculat lated ed a mod model el of  dismis dis missin sing g def defens enses es tha thatt may hel help p to exp explai lain n the ass associ ociaation between dismissing attachment and resilience observed here (see Fraley et al., 1998; Fraley & Shaver, 1999). This model postulates two broad mechanisms, one of which involves strategies that allow dismissing adults to short-circuit thoughts and feelings that may  lead to the experience of insecurity. insecurity. Fraley et al. (1998) (199 8) argued that dismissing adults habitually redirect attention away from experiences that may threaten their sense of independence or self-worth—a process that  Bowlby (1980) referred to as defensive exclusion . By organizing their attentional resources in this manner, dismissing adults should be less likely to experience the powerful emotions that often accompany loss. Research by Fraley and Shaver (1997) is consistent with this hypothesis. They found that when dismissing adults were

888

PERSONALIT PERSO NALITY Y AND SOCIA SOCIAL L PSYCHO PSYCHOLOGY LOGY BULLE BULLETIN TIN

asked to suppress or inhibit thoughts of their partners abandoning them, they were able to lower their skinconduct cond uctanc ancee lev levels els bel below ow tho those se obs observe erved d in con contro troll conditions. In addition, the defensive strategies used by  highly dismissing adults appear to lead to the development of less complex and accessible mental representations of their relational relational experiences experiences (see Fraley et al., 2000). As a consequence, dismissing adults should be less vulnerable to the kinds of situations—such as loss— that typically activate attachment-related memories and feelin fee lings gs (se (seee Fra Fraley ley et al. al.,, 1998 1998). ). A sec second ond mec mechani hanism sm by   which dismissing adults may be able to exhibit resilience in the face of loss involves emotional attachment. Dismissing adults are less likely than others to develop a strong emotional attachment to their loved ones (see Fraley Fra ley & Dav Davis, is, 1997 1997;; Fra Fraley ley & Shav Shaver er,, 1997 1997). ). Psy Psycho cholog logiically, if their sense of self, identity identity,, and emotional security is less entwined in their close relationships, the loss of a loved one, although it is a painful experience, should sho uld be les lesss dis disori orient enting ing for the hig highly hly dis dismis missin sing g per per-son than for others. It is important to note that these two broad sets of  mechanisms are not mutually exclusive. For example, it  may be thecase thathighl thathighlyy dis dismis missin sing g ind indivi ividua duals ls areless affected by loss because (a) they actively work to inhibit  the activation of their attachment systems, (b) they are lesss emo les emotio tional nally ly inv invest ested ed in the rel relati ations onship hip,, or (c)both.  We hope that future research will identify the relative contribution of these and other processes to the way in  which dismissing adults adapt to loss. IMPLICATIONS FOR RESEARCH  AND THEORY ON RESILIENCE TO LOSS 

The findings of the current study are generally consistent with a growing body of research indicating that  there are multiple pathways to adjustment following the death of a loved one (Bonanno, Wortman, et al., 2002).  Although it has traditionally been assumed that highly  defensive defens ive people should have a diff difficult icult time recovering from loss, our data suggest that this assumption may  be warranted for some defensive people (e.g., fearfully  avoidant adults) but not necessarily for others (e.g., dismissingly avoidant adults). To fully appreciate the ways wa ys in which people adapt to loss, we believe that bereavement researchers will need to make concerted efforts to distinguish individual differences in susceptibility to distress from the defensive strategies that may be used to regulate that distress (see Fraley & Shaver Shaver,, 1999). Our findings suggest that it is necess nec essary ary to add addres resss both kin kinds ds of var variabl iables es sep separa aratel telyy to understand the ways in which psychological defenses may promote different patterns of adjustment. For example, our results indicated that individual differences in attac attachmenthment-relat related ed avoida avoidance nce (i.e. (i.e.,, variat variations ions in

theuse of def defens ensive ive str strate ategie gies) s) arerelat arerelative ively ly unr unrela elatedto tedto patterns of adaptation but that individual differences in attachmentattach ment-relat related ed anxiet anxietyy are relat related ed to adapta adaptation. tion. Because these two dimensions are only weakly correlated  with one another another,, it is possible for some highly defensive people to exhibit poor adjustment and other defensive people peo ple to exhi exhibit bit rel relianc iance. e. We sus suspec pectt tha thatt man manyy cli clinic nical al intuitions about the association between defense and adaptation are based on the assumption that defensive individuals are necessarily vulnerable, anxious, and insecure across situations. Unless the propensity to use defensive strategies is assessed separately from vulnerabilities per se, it is not possible to unconfound these two issues. We hope that our research helps to underscore this important point. ADVANTAGES ADVANT AGES AND  LIMITATIONS LIMITA TIONS OF THE CURRENT STUDY 

One of the advantages of this study is that it provides new dat dataa on the rel relati ations onship hip bet betwee ween n stra strateg tegies ies of aff affect  ect  regula reg ulatio tion n and lon long-t g-term erm adap adaptat tation ion in a sam sample ple of ind indii viduals who have experienced a nontrivial stressor—the loss of a loved one. A second advantage of this study is that the symptom outcomes we studied were not based on self-report questionnaires. Although self-report measures of adjustment can be useful, the use of such selfreports in the study of avoidance and defense raises subtlebut sub tlebut imp import ortant ant conc concern ernss abou aboutt theaccura theaccuracy cy of sel selffpresentation. The use of interviewer ratings, although not elimi eliminating nating these these problems problems altogether altogether, helps to pro vide a perspective on o n adjustment that is not exclusively  anchored in participants’ self-perceptions. Despite these advantages, there are some limitations to the present research that require our findings to be interpreted with caution. Most important, we were not  ablee to obt abl obtain ain pre prelos losss mea measur sures es of ind indivi ividua duall dif differ ferenc ences es in adu adult lt att attach achmen ment. t. Alt Althoug hough h we mea measur sured ed att attach achmen ment  t  patterns approximately 2 weeks before the first clinical assess ass essmen mentt per period iod,, it may be the cas casee tha thatt the these se mea measur sureements were distorted by the loss. 3 A second limitation of  the present research is that we had only two assessment  points poi nts.. As a con conseq sequen uence, ce, our abil ability ity to stu study dy pat patter terns ns of  change was restricted. It is possible, for example, that  some people vary in the degree to which their symptom levels oscillate over time (e.g., as might be derived from M. S. Stroebe & Schut’s, Schut’s, 1999, dual-proce dual-process ss model of bereavement). It would require a minimum of three time points poi nts to eva evalua luate te suc such h pos possib sibili ilitie ties, s, and of cou course rse,, man many  y  more to do so in a rigorous manner. A third limitation concerns potential self-selection biases. For obvious reasons, it is difficult to obtain a representative sample of  bereaved adults. It is possible that defensive individuals may be less likely to volunteer to pa rticipate in this kind of research. Although the average level of avoidance

Fraleyy, Bonanno / ATT Frale TTACHME ACHMENT NT AND LOSS obtained in this sample is comparable to that observed in a sa samp mple le of ma marr rrie ied d ad adul ultsof tsof th thee sa sameage meage andgen andgende derr distri dis tribut bution ion (Bo (Bonann nanno, o, 2002 2002), ), it is imp imposs ossibl iblee to rul rulee out  the possibility that the sample we studied is unrepresentative of the population of bereaved individuals. In closing, it is widely assumed that emotionally  avoidant or defensive individuals will have a difficult  time adjusting to the loss of a loved one. However, a growin gro wing g bod bodyy of res resear earch ch sug sugges gests ts tha thatt def defens ensive ive ind indivi ivid duals ua ls te tend nd to ada adapt pt qu quit itee we well ll to lo loss ss (e (e.g. .g.,, Bo Bonan nanno no et al al., ., 1995). Such findings pose a number of challenges to attachment theory—a theory that has traditionally held that emotional avoidance is indicative of poor psychologicall adjust logica adjustment. ment.4 W  Wee believe that the two-dimensional model of individual differences in attachment provides an important compromise between classic perspectives on psychological defense and grief and contemporary  research findings. According to the two-dimensional model of individual differences in attachment organization, tio n, som somee pat patter terns ns of avoi avoidanc dancee or def defens ensee (i. (i.e., e., fea fearful rful-avoidance) will be linked to distress and others will not  (i.e., dismissing-avoidance). This perspective may allow  us to retain some of o f Bowlby’s important insights regarding att attach achmen mentt and ber bereave eavemen mentt whi while le rec recogni ognizin zing g that  defense mechanisms may have adaptive consequences for some individuals and maladaptive consequences for others. NOTES 1. Although our use of terms such as “fearful avoidant” and “dismissing avoidant” may imply that we are referring to categories of  attachment, we are conceptualizing and measuring attachment patterns as continuous phenomena (see Fraley & Waller, 1998). According to contem contemporarytwo-dimens porarytwo-dimensional ional approa approaches ches to attach attachment, ment, fearful avoidance, as well as other patterns of attachment, are theoretical prototypes or configural patterns within the two-dimensional space (e.g., Fraley & Spieker Spieker,, 2003; Griffin & Bartholomew, 1994). As such, peoplecan peo plecan varyin thedegre thedegreee to whi which ch the theirpatter irpattern n of tho though ught, t, beh behavavior,, and feeling resembles these prototypes. ior 2. The predicted curves for security and dismissing-avoidance dip slightly below zero due to floor effects (i.e., positively skewed data) in the distri distributionof butionof anxiet anxietyy sympto symptoms. ms. The basic findi findings ngs repor reported ted here remain unchanged when the data are transformed to minimize the skew. 3. Bec Becausewe ausewe wer weree notableto obt obtainmeasu ainmeasuresof resof att attachm achmentprior entprior to theloss, it is poss possibl iblee tha thatt ourmeasu ourmeasurem rement entss of att attachm achmentwere entwere dis dis-torted tor ted by the los loss. s. Alt Althou hough gh we cann cannot ot def defini initiv tivelyaddre elyaddress ss thi thiss iss issue ue in the context of the current study, we believe that if the loss led to changes in attachment patterns, theoretically we should expect loss to havee exe hav exerte rted d a mai main n eff effect ect suc such h that the the sam sample ple,, on ave averag rage, e, exh exhibi ibited ted similar levels of change (e.g., heightened insecurity). If this is correct,  we would not expect the association between between attachment-related anxiety and symptoms to be distorted considerably. Barring unforeseen nonlinearities in the relationship between attachment patterns, loss, and grief, our ability to discern the associations between attachment  and patterns of grief should not be excessively compromised. 4. We should note that although Bowlby believed that avoidance  was likely to be a maladaptive response to loss, he qualified this belief in several ways. Consider the following quotations from Bowlby (1980): “Some of those who proclaim their self-sufficiency are in fact relatively  relatively  immune to loss” (p. 213) and “Individuals disposed strongly to assert 

889

their self-sufficiency fall on a continuum ranging from those whose procla pro claime imed d sel self-s f-suff uffici icienc encyy res rests ts on a pre precar cariou iouss basi basiss to tho those se in who whom m it is fir firmlyorgan mlyorganize ized” d” (p. 211 211). ). The These se and oth other er pass passage agess (se (seee Fra Fraley& ley& Shaver, 1999) suggest that although Bowlby believed that there were some indiv individualswho, idualswho, despit despitee theirdefensivemaneuver theirdefensivemaneuvers, s, wouldhave a difficult time adapting to loss, he did not believe that avoidant strategiesper se wouldbe associ associatedwith atedwith such vulne vulnerabil rabilities ities.. Frame Framed d withi within n the contex contextt of contem contemporarytwo-dimen porarytwo-dimensionalmodelsof sionalmodelsof indiv individualdifidualdifferenc fer ences es in att attachm achment ent,, Bow Bowlby lby’sideaswouldseemto ’sideaswouldseemto sug sugges gestt thatfearful indiv individuals iduals would expe experienc riencee protr protracted acted distress despite their defensive tendencies, whereas dismissing individuals would not. REFERENCES  Aiken, L. S., & West, S. G. (1991) . Multiple regression: Testing and inter-   preting i nteracti ons . Newbury Park, CA: Sage.  Ainsworth, M. D. S., B lehar, M. C., Waters, E., & Wall, S. (1978). Pat-  terns of attachment . Hillsdale, NJ: Lawrence Erlbaum.  American Psychiatr ic Associatio n. (1994).   Diagnostic and statistical    (4th ed.). Washington, DC: Author. manual of mental disorders  (4th Bartholomew, K. (1990). Avoidance of intimacy: An attachment perspective.  Journal of Social and Personal Relationships , 7 , 147-178. Bartholo Bart holomew mew,, K., & Hor Horowit owitz, z, L. M. (199 (1991). 1). Atta Attachme chment nt styl styles es among  young adults: A test of a four-category model. Journal of Personality  and Social Psychology , 61, 226-244. Bonanno, Bonan no, G. A. (200 (2002). 2). [Atta [Attachme chment nt in a non-b non-bere ereavedsample] avedsample].. Unpublished raw data. Bonanno, G. A. (2003).  Loss, trauma, and human resilience: Have we  underestimated our capacity to thrive after extremely aversive events?  Manuscript submitted for publication. Bonanno, G. A., & Field, N. P. (2001). Examining the delayed grief  hypothesis across five years of bereavement. American Behavioral  Scientist , 44 , 798-806. Bonann Bon anno, o, G. A.,Fiel A.,Field, d, N. P., Kov Kovace acevi vic, c, A.,& Kal Kaltma tman, n, S. (20 (2002 02). ). Se Selflfenhancement as a buffer against extreme adversity: Civil war in Bosnia and traumatic loss in the United States.  Personality and  Social Psychology Bulletin , 28 , 184-196. Bonanno, Bonan no, G. A., & Kalt Kaltman,S. man,S. (199 (1999). 9). Towar oward d an inte integrat grativeperspe iveperspecctive on bereavement.  Psychological Bulletin , 125 , 760-776. Bonanno, G. A., & Keltner, D. (1997). Facial expressions of emotion and the course of conjugal bereavement. Journal of Abnormal Psy-  chology , 106 , 126-137. Bonanno, G. A., Keltner, D., Holen, A., & Horowitz, M. J. (1995).  When avoiding unpleasan t emotions might not be such a bad thing: Verbal-autonomic Verbal-autonomic response dissociation and midlife con jugal bereavement.  Journal of Personality and Social Psychology , 69 , 975-989. Bonanno, G. A., Wortman, C. B., Lehman, D. R., Tweed, R. G., Haring, M., Sonnega, J., et al. (2002). Resilience to loss and chronic chro nic gri grief: ef: A pros prospect pectivestudy ivestudy frompre-lo frompre-loss ss to 18 mont months hs postloss. Journal of Personality and Social Psychology , 83 . 1150-1164. Bonanno, G. A., Znoj, H. J., Siddique, H., & Horowitz, M. J. (1999).  Verba l-a uto nom ic res pon se dis soc iat ion an d ada pta tio n to midlife conjugal loss: A follow-up at 25 months.  Cognitive Therapy  and Research , 23 , 605-624. Bowlby, J. (1980).  Attachment and loss, Vol. III. Loss: Sadness and depres-  sion . New York: Basic Books. Deutsch,H. Deut sch,H. (193 (1937).Absenceof 7).Absenceof gri grief. ef. Psychoanalytic Quarterly , 6 , 12-2 12-22. 2. Dozier,, M., & Kobak, R. R. (1992). Psychophysiology in attachment  Dozier interviews: Converging evidence for deactivating strategies.  Child   Developm ent , 63 , 1473-1480. Field Fie ld,, N. P., & Sun Sundin din,, E. C. (2 (2001 001). ). Att Attach achme ment nt sty stylein lein adj adjust ustme ment nt to conjugal bereavement. Journal of Social and Personal Relationships , 18 , 347-361. Fraley, R. C., & Davis, K. E. (1997). Attachment formation and transfer in young adults’ close friendships and romantic relationships. Personal Relationships , 4 , 131-144. Fraley Fral ey,, R. C., Davi Davis, s, K. E., & Shave Shaverr, P. R. (199 (1998). 8). Dism Dismissi issing-a ng-avoid voidance ance and the defensive organizatio organization n of emotion, cognition, and behavior.. In J. A. Simpson & W. S. Rholes (Eds.), Attachment theory and  ior   (pp. 249-279). New York: Guilford. close relationships  (pp.

890

PERSONALIT PERSO NALITY Y AND SOCIA SOCIAL L PSYCHO PSYCHOLOGY LOGY BULLE BULLETIN TIN

Fraley, R. C., Garner, J. P., & Shaver, P. R. (2000). Adult attachment  and the defensive regulation of attention and memory: The role of pree preempti mptive ve and poste postempti mptive ve proc processe esses. s. Journal of Personali ty and  Social Psychology , 79 , 816-826. Fraley, R. C., & Shaver, P. R. (1997). Adult attachment and the suppressionof pre ssionof unwan unwanted ted thoug thoughts. hts. Journal of Personalit y and Social Psy-  chology , 73 , 1080-1091. Fraley, R. C., & Shaver, P. R. (1999). Loss and bereavement: Attachment theory and recent controversies concerning “grief work” and the nature of detachment. In J. Cassidy & P. R. Shaver (Eds.), Handbook of attachment: Theory, research, and clinical applications  (pp. 735-759). New York: Guilford. Fraley,, R. C., & Shaver Fraley Shaver,, P. R. (2000). Adult romantic attachment: TheTheoretical developments, emerging controversies, and unanswered questions. Review of General Psychology , 4 , 132-154. Fraley Fral ey,, R. C., & Spie Spieker ker,, S. J. (200 (2003). 3). Whatare the diff differe erencesbetwee ncesbetween n dimensional and categorical models of individual differences in attachment? Reply to Cassidy (2003), Cummings (2003), Sroufe (2003), and Waters and Beauchaine Beauchaine (2003). Developmental Psychol-  ogy , 39 , 423-429. Fraley Fral ey,, R. C., & Wal Waller ler,, N. G. (199 (1998). 8). Adul Adultt attac attachmen hmentt patte patterns:A rns:A test  of the typological model. In J. A. Simpson & W. S. Rholes (Eds.),   (pp. 77-114). New York: Attachment theory and close relationships  (pp. Guilford. Griffin, D. W., & Bartholome Bartholomew, w, K. (1994). The metaphysics of measurement: The case of adult attachment. In K. Bartholomew & D. Perlman (Eds.), Advances in personal relationships: Vol. 5. Attach-    (pp. 17-52). London: Jessica Kingsley. ment processes in adulthood  (pp. Hazan, Haza n, C., & Shave Shaverr, P. R. (198 (1987). 7). Roma Romanticlove nticlove conce conceptual ptualize ized d as an attachment process. Journal of Personality and Social Psychology , 59 , 511-524. Hazan, C., & Shaver, P. R. (1992). Broken attachments: Relationship loss from the perspective of attachment theory. In T. L. Orbuch Closee rela relations tionship hip loss:Theore loss:Theoreticalapproach ticalapproaches  es (pp.90-107). (Ed.), Clos ( pp.90-107). New   York: Springer-Verlag. Horowitz, M. J., Siegel, B., Holen, A., Bonanno, G. A., Milbrath, C., & Stinson, C. (1997). Diagnostic criteria for Complicated Grief Disorder: An empirical evaluation. American Journal of Psychiatry , 154 , 904-910. Klohnen, E. C., & John, O. P. P. (1998). Working models of attachment:  A theory-ba sed prototype approach . In J. Simpson & S. Rholes (Eds.), Attachment theory and close relationships  (pp. 115-140). New   York: Guilford. Lazare, A. (1989). Bereavement and unresolved grief. In A. Lazare (Ed.),   Outpatient psychiatry: Diagnosis and treatment   (2nd ed., pp. 381-397). Baltimore: Williams & Wilkins. Lindemann, E. (1944). Symptomatology and management of acute grief.  American Journal of Psychiatry , 101 , 1141-1148. Main, M., Kaplan, N., & Cassidy, J. (1985). Security in infancy, childhood, and adulthood: A move to the level of representation. In I. Bretherton & E. Waters (Eds.), Growing points of attachment  theory and research. Monographs of the Society for Research in Child  (1-2, Serial No. 209), 66-104.  Developm ent , 50 (1-2,

Middleton, W., Raphael, B., Martinek, N., & Misso, V. (1993). Pathological grief reactions. In M. S. Stroebe, W. Stroebe, & R. O. Hansson (Eds.), Handbook of bereavement: Theory, research, and inter-    (pp. 44-61). New York: Cambridge. vention  (pp. Nolen-Hoeksema, Nolen-Hoe ksema, S., & Larson, J. (1999). Coping with loss . Mahwah, NJ: Lawrence Erlbaum. Onishi, M., Gjerde, P. F., & Block, J. (2001). Personality implications of romantic attachment patterns in young adults: A multimethod, multi-informant study.   Personality and Social Psychology  Bulletin , 27 , 1097-1110. Osterweis, M., Solomon, F., & Green, M. (1984).  Bereavement: Reac - -  tions, consequences, and anger . Washington, DC: National Academy  Press. Parkes, C. M., & Weiss, R. (1983).  Recovery from bereavement . New York: Basic Books. Protectin ectingg humansubject humansubjects: s: Insti Institutio tutional nal revi review ew boar board  d  Penslar Pens lar,, R. L. (199 (1993). 3). Prot Rockvil ville, le, MD: Offi Office ce for Hum Human an Rese ResearchProtect archProtections ions,, guidebook . Rock U.S. Department of Health and Human Services. Raphael,B. Rapha el,B. (198 (1983). 3). Theanatomy of bere NewYork:: Basi Basicc Book Books. s. bereaveme avement  nt . NewYork Sanders, C. M. (1993). Risk factors in bereavement bereavement outcome. In M. S. Stroebe, W. Stroebe, & R. O. Hansson (Eds.), Handbook of bereave-    (pp. 255-267). New York: ment: Theory, research, and intervention  (pp. Cambridge University Press. Shaver, P. R., & Tancredy, C. M. (2001). Emotion, attachment and bereavement: A conceptual commentary. In M. S. Stroebe,  W. Stroebe, R. O. Hansson, & H. Schut (Eds.), Handbook of bereave-   (pp. 63-88). Washington, DC: ment: Consequences, coping, and care  (pp.  American Ps ychological As sociation. Spitzer, R. L., Williams, J. B. W., Gibbon, M., & First, M. B. (1990). User’s guide for the Str uctured Clinical Interview for DSM-III-R . Washington, DC: American Psychiatric Press. Sroufe, L. A., & Waters, E. (1977). Attachment as an organizational construct. Child Development , 48 , 1184-1199. Stroe Str oebe be,, M. S.,& Sc Schut hut,, H. (19 (1999) 99).. Thedualproce Thedualprocess ss mod modelof elof cop coping ing  with bereavement: Rationale and description.  Death Studies ,  23 , 197-224. Stroebe, W., & Stroebe, M. (1987).  Bereavement and health: The psy-  chological and physical consequences of partner loss . New York: Cambridge.  Wayment, H. A., & Vierthaler, J. (2002). Attachmen t style and bereavement reaction.  Journal of Loss and Trauma , 7 , 129-149.  Weinberger,, D. A., Schwartz, G. E., & Davidson, J. R. (1979). Low Weinberger anxious and repressive coping styles: Psychometric patterns of  behavioral and physiological responses to stress. Journal of Abnor-  mal Psychology , 88 , 369-380.  Wortman, C. B., & Silver, Silver, R. C. (1989). The myths of coping with loss.  Journal o f Consulti ng and Cli nical Ps ychology , 57 , 349-357. Received April 9, 2003 Revision accepted September 26, 2003