stress and accessibility of proximity-related thoughts: exploring the normative and intraindividual...

15
PERSONALITY PROCESSES AND INDIVIDUAL DIFFERENCES Stress and Accessibility of Proximity-Related Thoughts: Exploring the Normative and Intraindividual Components of Attachment Theory Mario Mikulincer, Gurit Birnbaum, David Woddis, and Orit Nachmias Bar-Ilan University Three studies examine the effects of stress on the accessibility of proximity-related thoughts. In all the studies, participants reported on their attachment style, and the accessibility of proximity themes and worries in a lexical decision task was assessed upon the priming of a stress or neutral word. In Study 2, the primed stress word was semantically related to attachment themes. In Study 3, lexical decisions were made under low or high cognitive load conditions. Overall, the priming of a stress word led to increased accessibility of proximity themes, regardless of attachment style. Anxious-ambivalent people also showed high accessibility to proximity themes and worries in both neutral and stress contexts. In most conditions, avoidant persons' reactions were similar to those of secure persons. However, they showed no accessibility to proximity worries even after the priming of a semantically related word and reacted with high accessibility to these worries upon the addition of cognitive load. In his classic trilogy, Bowlby (1969, 1973, 1980) developed an ethological theory about the function, antecedents, and conse- quences of the activation of the attachment system in humans. However, the vast majority of attachment studies have adopted an individual differences perspective, dealing with Bowlby's (1973, 1988) statements about the effects of a person's history of attach- ment experiences on his or her attachment style. This trend is mainly noted in adult attachment studies, which have focused on the ways people differing in attachment style differ in the way they regulate distress. Our study attempts to integrate the normative and intraindividual components of attachment theory and to examine Bowlby's hypothesis about the effect of stress arousal on the activation of the attachment system. Specifically, we examine whether a stress context cognitively activates the attachment sys- tem in most individuals while, at the same time, examining whether and how individual differences in attachment style may color this activation. Attachment, Stress Arousal, and Distress Regulation One basic statement of Bowlby's (1969, 1973, 1980) theory is that all human beings are born with an attachment system aimed at maintaining proximity to significant others (attachment figures) in times of stress. In Bowlby's terms, proximity maintenance to attachment figures helps the individual to deal with stress-inducing events and to regulate distress. These figures function as a haven Mario Mikulincer, Gurit Birnbaum, David Woddis, and Orit Nachmias, Department of Psychology, Bar-Ilan University, Ramat Gan, Israel. Correspondence concerning this article should be addressed to Mario Mikulincer, Department of Psychology, Bar-Ilan University, Ramat Gan 52900, Israel. Electronic mail may be sent to [email protected]. of safety to which the individual can retreat for comfort in times of stress as well as a "secure base" from which he or she can explore the world and develop his or her personality in a supportive atmosphere. According to Bowlby (1988), the attachment system is active over the entire life span and it is directly manifested in thoughts and behaviors related to proximity and support seeking. Along this reasoning, Bowlby (1973) also proposed that the encounter with stress automatically activates the attachment sys- tem. In his view, stress signals that something is going wrong and that some coping action should be taken. As a result, the attach- ment system is activated and the individual is driven to maintain or restore proximity to persons who can provide support in managing distress. In a smooth course of events, this activation would be manifested in actual seeking of proximity and support. However, there are cases in which these behaviors may be inhibited by the absence of attachment figures as well as by other contextual and personal factors. In this case, the attachment system may be still active at a cognitive level. That is, thoughts about proximity may still influence cognition and behavior. In social cognitive terms, stress may heighten the accessibility of proximity thoughts even when proximity-seeking behaviors are inhibited. Beyond the above ideas about the normative activation of the attachment system, Bowlby (1973, 1988) also adopted an individ- ual difference perspective. In his view, a person's attachment history might contribute to distress regulation. On the one hand, interactions with significant others who are available in times of stress and responsive to one's attachment needs may lead to the development of positive expectations about the self and the world, while encouraging the reliance on support seeking when coping with life problems. On the other hand, interactions with significant others who are unavailable and nonresponsive in times of need may elicit chronic distress and encourage the formation of other defensive strategies rather than proximity seeking. Journal of Personality and Social Psychology, 2000, Vol. 78, No. 3, 509-523 Copyright 2000 by the American Psychological Association, Inc. 0O22-3514/O0/$5.O0 DOI: 10.1037OT022-3514.78.3.509 509

Upload: idc-ac

Post on 26-Apr-2023

0 views

Category:

Documents


0 download

TRANSCRIPT

PERSONALITY PROCESSES AND INDIVIDUAL DIFFERENCES

Stress and Accessibility of Proximity-Related Thoughts: Exploring theNormative and Intraindividual Components of Attachment Theory

Mario Mikulincer, Gurit Birnbaum, David Woddis, and Orit NachmiasBar-Ilan University

Three studies examine the effects of stress on the accessibility of proximity-related thoughts. In all thestudies, participants reported on their attachment style, and the accessibility of proximity themes andworries in a lexical decision task was assessed upon the priming of a stress or neutral word. In Study 2,the primed stress word was semantically related to attachment themes. In Study 3, lexical decisions weremade under low or high cognitive load conditions. Overall, the priming of a stress word led to increasedaccessibility of proximity themes, regardless of attachment style. Anxious-ambivalent people alsoshowed high accessibility to proximity themes and worries in both neutral and stress contexts. In mostconditions, avoidant persons' reactions were similar to those of secure persons. However, they showedno accessibility to proximity worries even after the priming of a semantically related word and reactedwith high accessibility to these worries upon the addition of cognitive load.

In his classic trilogy, Bowlby (1969, 1973, 1980) developed anethological theory about the function, antecedents, and conse-quences of the activation of the attachment system in humans.However, the vast majority of attachment studies have adopted anindividual differences perspective, dealing with Bowlby's (1973,1988) statements about the effects of a person's history of attach-ment experiences on his or her attachment style. This trend ismainly noted in adult attachment studies, which have focused onthe ways people differing in attachment style differ in the way theyregulate distress. Our study attempts to integrate the normative andintraindividual components of attachment theory and to examineBowlby's hypothesis about the effect of stress arousal on theactivation of the attachment system. Specifically, we examinewhether a stress context cognitively activates the attachment sys-tem in most individuals while, at the same time, examiningwhether and how individual differences in attachment style maycolor this activation.

Attachment, Stress Arousal, and Distress Regulation

One basic statement of Bowlby's (1969, 1973, 1980) theory isthat all human beings are born with an attachment system aimed atmaintaining proximity to significant others (attachment figures) intimes of stress. In Bowlby's terms, proximity maintenance toattachment figures helps the individual to deal with stress-inducingevents and to regulate distress. These figures function as a haven

Mario Mikulincer, Gurit Birnbaum, David Woddis, and Orit Nachmias,Department of Psychology, Bar-Ilan University, Ramat Gan, Israel.

Correspondence concerning this article should be addressed to MarioMikulincer, Department of Psychology, Bar-Ilan University, Ramat Gan52900, Israel. Electronic mail may be sent to [email protected].

of safety to which the individual can retreat for comfort in times ofstress as well as a "secure base" from which he or she can explorethe world and develop his or her personality in a supportiveatmosphere. According to Bowlby (1988), the attachment systemis active over the entire life span and it is directly manifested inthoughts and behaviors related to proximity and support seeking.

Along this reasoning, Bowlby (1973) also proposed that theencounter with stress automatically activates the attachment sys-tem. In his view, stress signals that something is going wrong andthat some coping action should be taken. As a result, the attach-ment system is activated and the individual is driven to maintain orrestore proximity to persons who can provide support in managingdistress. In a smooth course of events, this activation would bemanifested in actual seeking of proximity and support. However,there are cases in which these behaviors may be inhibited by theabsence of attachment figures as well as by other contextual andpersonal factors. In this case, the attachment system may be stillactive at a cognitive level. That is, thoughts about proximity maystill influence cognition and behavior. In social cognitive terms,stress may heighten the accessibility of proximity thoughts evenwhen proximity-seeking behaviors are inhibited.

Beyond the above ideas about the normative activation of theattachment system, Bowlby (1973, 1988) also adopted an individ-ual difference perspective. In his view, a person's attachmenthistory might contribute to distress regulation. On the one hand,interactions with significant others who are available in times ofstress and responsive to one's attachment needs may lead to thedevelopment of positive expectations about the self and the world,while encouraging the reliance on support seeking when copingwith life problems. On the other hand, interactions with significantothers who are unavailable and nonresponsive in times of needmay elicit chronic distress and encourage the formation of otherdefensive strategies rather than proximity seeking.

Journal of Personality and Social Psychology, 2000, Vol. 78, No. 3, 509-523Copyright 2000 by the American Psychological Association, Inc. 0O22-3514/O0/$5.O0 DOI: 10.1037OT022-3514.78.3.509

509

510 MIKULINCER, BIRNBAUM, WODDIS, AND NACHMIAS

In examining Bowlby's ideas in adulthood, most of the studieshave focused on individual differences related to a person's at-tachment style—stable patterns of relational cognitions and behav-iors reflecting one's attachment history (Shaver & Hazan, 1993).This line of research has been based on Ainsworth, Blehar, Waters,and Wall's (1978) typology of attachment style in infancy (secure,avoidant, and anxious-ambivalent) as well as on Hazan and Shav-er's (1987) definitions of these styles in adulthood. In Hazan andShaver's terms, secure persons are more confident in others'availability in times of need, feel more comfort with closeness, andhave a more positive attachment history than avoidant andanxious-ambivalent persons. The anxious-ambivalent style is alsodefined by a desire for enmeshed relationships and the avoidantstyle by preference for self-reliance and emotional distance. Thistypology has been assessed via self-report scales and has beenvalidated in several studies (see Shaver & Hazan, 1993, for areview).

Adult attachment research has provided extensive support to thehypothesis that a person's attachment style is related to individualdifferences in distress regulation. On the one hand, secure persons,who have a history of positive attachment experiences, have beenfound to cope with stress by seeking support from others (Miku-lincer & Florian, 1995, 1998; Mikulincer, Florian, & Weller,1993). Moreover, they have been found to hold optimistic expec-tations about stress manageability (Mikulincer & Florian, 1995), astrong sense of self-efficacy (Collins & Read, 1990; Mikulincer &Florian, 1995), and trust in others' goodwill (Bartholomew &Horowitz, 1991; Collins & Read, 1990).

On the other hand, insecure persons, who have a negativehistory of attachment experiences, seem to develop alternativeways of distress regulation. The anxious-ambivalent attachmentstyle seems to consist of strategies that hyperactivate the attach-ment system and distress-related cues. Indeed, anxious-ambivalentpersons have been found to react to stress by minimizing distancefrom others and conquering their love via clinging and controllingactions (e.g., Feeney & Noller, 1990; Mikulincer, Orbach, &Iavnieli, 1998). Moreover, they have been found to excessivelyfocus on distress cues and to mentally ruminate on negativeemotions (e.g., Mikulincer & Florian, 1998; Mikulincer & Orbach,1995). The avoidant attachment style seems to consist of strategiesthat inhibit proximity-seeking behaviors, divert attention fromdistress cues, and emphasize a sense of self-reliance (e.g., Kobak,Cole, Ferenz, & Fleming, 1993). Indeed, avoidant persons havebeen found to avoid support seeking and to rely on repressivemechanisms (e.g., Fraley & Shaver, 1997; Mikulincer & Orbach,1995).

Some relevant findings are also available on the normativecomponent of the stress-attachment link in adulthood. First, ex-periences that disrupt proximity maintenance to attachment fig-ures, such as actual separation from a romantic partner or thoughtsabout the potential loss of this partner, seem to be related to stressarousal, as manifested in physiological and self-report measures(Cafferty, Davis, Medway, O'Hearn, & Chappell, 1994; Feeney &Kirkpatrick, 1996; Fraley & Shaver, 1997). Second, attachment-related regulatory strategies (e.g., deactivation or hyperactivationof the attachment system) are mainly observed under stress-eliciting conditions (Cafferty et al., 1994; Fraley & Shaver, 1998;Mikulincer, 1998a; Mikulincer et al., 1993; Mikulincer et al.,1998). Third, Fraley and Shaver (1998) observed more proximity-

seeking behaviors among couples who were separating at anairport than among couples who were not separating.

The problem with most of the above studies, with the exceptionof Fraley and Shaver's (1998) study, is that they have not beenoriginally designed to examine the normative activation of prox-imity seeking under stress contexts. Moreover, Fraley and Shav-er's (1998) study was exclusively focused on an attachment-related stressful event (separation from a partner). Thus, it couldnot test the hypothesis that stressful events that are not directlyrelated to the disruption of attachment ties would activate theattachment system. The current study follows the above line ofresearch and attempts to examine Bowlby's (1969) statementsabout the activation of the attachment system upon stress arousal.

The Current Study

The main goal of the current study is to examine the stress-attachment link in adulthood. Following Bowlby's ideas, we as-sume that this link is an automatic, inborn, universal regularity,which has been evolved throughout generations and has become anintegral part of the neural hardware of all members of our species.In evolutionary psychology terms (Cosmides & Tooby, 1987), thisregularity can be conceptualized as a basic cognitive structure thatprocesses information about stressful circumstances and activatescognitive representations of proximity to significant others, which,in turn, may be translated into behavioral intentions and actualbehavior. The attachment-stress link can be also conceptualized insocial cognitive terms. Stress arousal may activate cognitive rep-resentations of the attachment system as a ballistic missile and,then, heighten the cognitive accessibility of proximity-relatedthoughts.

Bowlby's premise about an inborn stress-attachment link im-plies that this link would be active even among insecure persons,who might have learned that proximity does not provide any relief.Of course, this is not to say that individual differences in attach-ment experiences would not affect this activation. In fact, partic-ular proximity-related thoughts and feelings reflecting a person'sattachment history may be made subsequently accessible by theactivation of the attachment system. These cognitive by-productsmay shape the way one experiences the activation of the attach-ment system and may determine the extent to which one mayengage in proximity seeking.

The stress-attachment link has been studied among infants (e.g.,Ainsworth, 1973; Brooks & Lewis, 1974; Lamb, 1976). Findingshave shown that the encounter with stressful events heightens thelikelihood of proximity-seeking behaviors. Kobak and Duemmler(1994) suggested that threatening, challenging, and conflictivesituations might lead adult persons to seek for support from others.The stress-attachment link in adulthood also received some sup-port in studies testing Schachter's (1959) fear-affiliation hypoth-esis. In these studies, fearful persons are more likely than nonfear-ful persons to affiliate with an available other while awaiting somenoxious event (see Rofe, 1984, and Shaver & Klinnert, 1982, forreviews). Although this finding has been initially interpreted interms of social comparison, more recent findings (Kirkpatrick &Shaver, 1988) suggest that stress increases proximity seeking as ameans of alleviating distress.

The main question here is how to assess the activation of theattachment system upon stress arousal. One possibility is to ob-

STRESS AND ATTACHMENT 511

serve a person's behavior and to rate the extent he or she seeks forproximity or support. This methodological strategy has been usedin studies with infants (e.g., Ainsworth, 1973; Lamb, 1976), inexperiments testing Schachter's (1959) fear-affiliation hypothesis(see Rofe, 1984, for a review), and in recent adult attachmentstudies (e.g., Fraley & Shaver, 1998; Rholes, Simpson, & Grich-Stevens, 1998; Simpson, Rholes, & Nelligan, 1992). The problemwith this strategy is that proximity-seeking behaviors may dependon a conscious deliberation about the expected utility andpersonal-cultural adequacy of these behaviors. Then, the absenceof proximity-seeking behaviors in stress contexts should not nec-essarily reflect a lack of activation of the attachment system. Infact, stress may activate this system, but an array of personal,contextual, and cultural factors may inhibit proximity seeking.

The same problem precludes the use of self-report scales. Thismethodological strategy has been mainly used in studies thatexamine the extent to which a person exposed to a stressful eventreports that he or she has sought for support paper-and-pencilcoping scales (e.g., Ways of Coping Checklist, Folkman & Laza-rus, 1985). In line with Bowlby's (1969) hypothesis, these studieshave revealed that support seeking is a frequently reported copingstrategy (e.g., Folkman & Lazarus, 1985; Mikulincer & Florian,1996). However, these studies have also revealed that self-reportsof support seeking depend on a conscious scrutiny of personal,cultural, and contextual factors (Lazarus & Folkman, 1984; Nad-ler, 1991). Moreover, adult attachment studies have shown thatself-reports of support seeking seem to be a direct manifestation ofa person's attachment style rather than a stress-tailored response(Mikulincer et al., 1993). Finally, retrospective accounts of supportseeking can be influenced by memory biases as well as by currentmood.

The stress-attachment link can be also examined by assessingthe contextual activation of proximity-related thoughts in stresscontexts. Wegner and Smart (1997) proposed that cognitive acti-vation could be assessed in two different ways. The first and mostobvious way defines a thought as "active" if it is conscious andpeople report that they are mentally occupied with the thought inquestion. That is, a thought is "active" if it is present in a person'sstream of consciousness. In this case, the length of time a thoughtis mentally rehearsed or the number of occurrences of a thought ina person's verbalization serves as a measure of cognitive activa-tion. This strategy can overcome the problems inherent in theretrospective nature of self-report scales. However, it cannot dealwith the limitations imposed by the fact that thoughts can be alsoactive at preconscious levels and may still influence cognitions andbehaviors (Wegner & Smart, 1997).

The second way of defining whether a thought is "active"focuses on its cognitive accessibility at a given moment—itsreadiness to be used in information processing. This definition isbased on the notion that a thought can be brought to mind and caninfluence mental processes before one recognizes it in one'sstream of consciousness (Wegner & Smart, 1997). In this case, theextent to which a thought influences performance in cognitivetasks can serve as a measure of cognitive activation (e.g., Bargh,Chen, & Burrows, 1996; Sherman, Mackie, & Driscoll, 1990).This strategy overcomes the problems of the above-reviewedmethodological strategies and may provide a more valid test of thestress-attachment link. That is, one can examine whether stress

arousal heightens the readiness of proximity-related thoughts toinfluence mental processes.

Along the above reasoning, we explore the effects of stress onthe accessibility of two types of thoughts. First are thoughtsdealing with proximity maintenance, support, and love. Thesethoughts have positive connotations and may by themselves reducedistress. Second are worries related to the frustration of attachmentneeds and the failure to maintain proximity, such as thoughts aboutseparation and rejection. These thoughts have negative connota-tions and may exacerbate distress. This distinction is importantbecause the activation of the attachment system may make acces-sible both positive and negative thoughts about proximity. Thismay be particularly true for insecure persons, who have experi-enced recurrent failure in maintaining proximity and hold negativebeliefs about others (Collins, 1996; Shaver, Collins, & Clark,1996). In cognitive terms, these frustrating experiences may createan associative link in the memory network between the activationof the attachment system and worries about separation and rejec-tion, so that the accessibility of these worries would increase everytime attachment needs are activated.

On this basis, we examine two main hypotheses. The firsthypothesis concerns the normative component of the stress-attachment link: Actual or symbolic stress-eliciting stimuli wouldheighten the accessibility of proximity themes, regardless of aperson's attachment style. The second hypothesis concerns theintraindividual component of this link: Actual or symbolic stress-eliciting stimuli would heighten the accessibility of proximityworries among avoidant and anxious-ambivalent. This increasedaccessibility might reflect an associative cognitive link betweenthe activation of the attachment system and insecure persons'negative expectations about proximity maintenance.

Beyond the above two hypotheses, we also explore variousalternatives related to insecure persons' ways of distress regula-tion. First, anxious-ambivalent persons' habitual hyperactivationof attachment needs and their tendency to mentally ruminate onactual or potential threats (Mikulincer & Florian, 1998) may leadto a chronic activation of proximity-related themes and worriesand may increase their accessibility even in nonstressful contexts.Second, avoidant persons' tendency to suppress attachment- anddistress-related thoughts (Fraley, Davis, & Shaver, 1998; Fraley &Shaver, 1997) may be so deep and constraining that it may inhibitthe entire stress-attachment link or, at least, the accessibility ofproximity-related worries. If this is true, further methodologicalsteps that inhibit the use of suppression mechanisms (e.g., impo-sition of cognitive pressures on a person's response) may allow usto observe the activation of the attachment system among avoidantpersons. These possibilities would be examined in the currentstudies.

In all the three reported studies, we examine cognitive accessi-bility of proximity-related themes and worries in a lexical decisiontask (Meyer & Schvaneveldt, 1971), which allows exploration ofthe cognitive basis of the stress-attachment link. In this task,participants read a string of letters and try to identify as quickly aspossible whether it is a word or a nonword, and reaction times(RTs) served as a measure of the accessibility of thoughts relatedto the target words—the quicker the RT, the higher the accessi-bility (e.g., Fischler & Bloom, 1979). In our studies, the targetwords were words related to proximity themes and worries, whichwere embedded within a stress or neutral context.

512 MIKULINCER, BIRNBAUM, WODDIS, AND NACHMIAS

Previous findings highlight the suitability of a lexical decisiontask for exploring the accessibility of proximity themes and wor-ries under a stress context. First, research found a context-relatedness effect: RTs for identifying target words are quicker ifthese words are primed by relevant contexts (e.g., Stanovich &West, 1983). Second, this RT facilitation has been found mainly ifthe target word is the most available association activated by theprimed context (e.g., Fischler & Bloom, 1979). Third, this effecthas been also found when the primed context was presentedsubliminally, so that participants could not consciously process itsconnection with the target word (Forster, 1981). Fourth, variabilityin lexical decision times has been found to reflect variability in theactivation of attachment working models (Baldwin, Fehr, Keedian,& Seidel, 1993; Mikulincer, 1998b). Thus, variations in RTs mayreflect the extent to which a target word is mentally activated by acontext, even when the context-target link is not consciouslyprocessed.

Given the above findings, if stress activates the attachmentsystem, it would heighten the readiness of proximity themes toinfluence lexical decision, which, in turn, would increase RTs forproximity words, such as love and closeness. Moreover, a stresscontext would increase insecure persons' RTs for words reflectingproximity worries.

Study 1

In Study 1, we examined the cognitive accessibility of proximitythemes and worries in a lexical decision task under a stress-inducing context. In manipulating the stress context, we used apriming procedure, in which a stress-inducing word ("failure") ora neutral word ("hat") was presented immediately before eachtarget word. This priming technique has been found to activateassociative links of the primed word and to facilitate the accessi-bility of thoughts that are related to it in the memory network (e.g.,Baldwin & Holmes, 1987; Bargh & Pietromonaco, 1982). Then, ifstress activates cognitive representations of attachment, the prim-ing of a stress word should heighten the accessibility of attachmentthemes and increase RTs for proximity-related words.

Participants previously classified according to attachment styleperformed a 120-trial lexical decision task. In each trial, weexposed participants to a word prime (stress or neutral) and thento 1 of 60 target letter strings, which could be a nonword or one ofthe following five categories of Hebrew words: proximity-relatedwords, distance-related words, positively valued words, negativelyvalued words, and neutral words. Participants were asked to decidewhether the target letter string was a word or not. For half of theparticipants, the prime was presented for 500 ms (supraliminalpresentation). For the remaining participants, the prime was pre-sented for only 20 ms (subliminal presentation). We included thismanipulation in order to examine the possible effects of the con-scious recognition of a stress word on insecure persons' RTs.

The five categories of target words were included according tothe following rationale: First, proximity and distance words werethe main target words for examining the stress-attachment link.Second, neutral words were introduced in order to control forpossible nonspecific effects of a stress context on lexical decisions.Third, positive and negative words were used to control for thepossibility that the link between a stress context and RTs forproximity and distance words is due to the hedonic affective value

of these words rather than to their attachment-related meaning. Infact, proximity words have positive affective connotations,whereas distance words have negative connotations.

On this basis, Study 1 was a mixed between-subject, within-subject 3 X 2 X 2 factorial experiment for attachment style, primepresentation (supraliminal, subliminal), and prime word (stress,neutral). The last factor was a within-subject measure. The maindependent variables were RTs for proximity and distance words.

With regard to the normative component of the stress-attachment link, we predicted that a stress word prime would leadto faster RTs for proximity words than a neutral word prime,regardless of a participant's attachment style. With regard to theintraindividual component, we predicted that a stress word primewould lead avoidant and anxious-ambivalent persons to reactquicker to distance words than a neutral word prime. A stresscontext would have no effect on secure persons' RTs to distancewords.

Method

Participants. Sixty undergraduate Social Sciences students from Bar-Ilan University (39 women and 21 men ranging in age from 18 to 32 years,median = 22) participated in the study as part of the requirements of theirBA degree. They were randomly divided into two conditions, each con-sisting of 30 participants.

Materials and procedure. In a preliminary stage, 520 students filledout two scales during lecture time in order to obtain categorical andcontinuous scores of attachment style. First, they received the Hebrewversion of Hazan and Shaver's (1987) Attachment Style Scale (Mikulincer,Florian, & Tolmacz, 1990) and classified themselves according to theirattachment style. Participants received the three prototypic descriptions offeelings and thoughts regarding attachment styles and endorsed the de-scription that best described their feelings in close relationships. Like inprevious Israeli samples (e.g., Mikulincer et al., 1990), 55% of the sampleclassified themselves as secure (n = 285), 28% as avoidant (« = 147), and17% as anxious-ambivalent (n = 88). Participants' responses in this scalewere used to sort out the independent samples of the three reported studies.In Study 1, 60 participants, 20 secure, 20 avoidant, and 20 anxious-ambivalent, were randomly selected (within each group) to participate inthe study. Refusal rate was low (5%) and did not significantly differ acrossattachment groups.1

Participants also received another 10-item attachment style scale in orderto assess two basic dimensions that seem to underlie a person's attachmentorganization: anxiety and avoidance (Brennan, Clark, & Shaver, 1998;Shaver et al., 1996). The decision to add this scale was based on Fraley andWaller's (1998) findings that categorical measures of attachment style donot provide a complete picture of the variability in attachment organization.Further, Brennan et al. (1998) showed that a two-dimensional modelunderlies most measures of adult attachment style. Specifically, a factoranalysis of the most widely used self-report instruments generated twomain factors that corresponded to the anxiety and avoidance dimensions(Brennan et al., 1998).

On this basis, we decomposed Hazan and Shaver's (1987) descriptionsof avoidant and anxious-ambivalent styles and constructed 5 items for eachdimensions (for more details see Mikulincer et al., 1990). A detailedexamination of these items revealed that our anxiety and avoidance itemscorresponded to Brennan et al.'s (1998) relevant items. Participants were

1 No gender difference in the distribution of attachment styles was foundin any of the studies. The introduction of gender as a covariate in thestatistical analyses of all the studies did not change the reported effects.

STRESS AND ATTACHMENT 513

read each item and rated the extent to which the item described them on a7-point scale, ranging from 1 (not at all) 7 (very much).

In our sample, a factor analysis with varimax rotation explained 62% ofthe 10-item scale variance. Whereas the first factor (39%) included the 5avoidance items (loading > .40), the second factor (23%) included the 5anxiety items. This factor analytic solution was indeed similar to Brennanet al.'s (1998) two-factor solution. Cronbach alphas for each factor impliedacceptable internal consistency (.73, .77). On this basis, we computed twoscores by averaging items that corresponded to each factor. Higher scoresreflect higher anxiety and high avoidance.

Two weeks after completing the attachment scales, participants wereinvited by phone to participate in a study on cognitive psychology on anindividual basis. On their arrival, the experimenter, who was blind toparticipants' attachment style, told them that they would perform a lexicaldecision task. This task was programmed based on the apparatus andprocedure used by Baldwin et al. (1993). It was run on a Pentium IBM-PC,with an SVGA color monitor. Brightness and contrasts were set somewhatlow and the primes and target letters strings were displayed in whitelettering on a black background in the middle of the monitor. Participantsworked at their own pace. They were first given 10 practice trials and thenthe 120 experimental trials. The words and nonwords in the practice trialswere different from those given in the experimental trials.

In the current study, each lexical decision trial consisted of a rapidpresentation (supraliminal or subliminal) of one of two primes (stress word,neutral word) followed, after a pause of 500 ms, by the presentation of 1of 60 target letter strings (for 1,000 ms). Participants judged as quickly aspossible whether it was a word. They initiated each trial by pressing thespace bar and responded by pressing the " 1 " on the keyboard number padif they thought that the string was a word or the "2" if they think that thestring was a nonword. Then, the string vanished and, after a 2-s pause, thenext trial began.

We randomly divided participants in each attachment style into twoconditions according to the temporal parameters of the presentation ofprimes. These parameters were similar to those used in prior studies (e.g.,Murphy & Zajonc, 1993). In the supraliminal condition, the prime waspresented for 500 ms so participants could easily read the word. They weretold that each trial will begin with an "x" in the middle of the screenfollowed by a word, which they were instructed to ignore, and then, aftera brief pause, the target letter string would be presented. In the subliminalcondition, the prime was presented for 20 ms, and participants were unableto recognize it. They were told that each trial would begin with an "x" inthe middle of the screen followed by a light flash, which they shouldignore, and then, after a brief pause, the target letter string would bepresented. It is important to note that even when a prime is presented foras little as 20 ms, the pattern may remain temporarily active in the earlyparts of the visual system (iconic afterimage). In order to avoid thisproblem, we masked the primes with a rain pattern of white dots imme-diately after their presentation. In addition, a pretest indicated that partic-ipants (n = 10) were not able to detect the subliminal primes after repeatedpresentation (less than 5% of correct detection).

The stress word prime was the Hebrew word for "failure," which waspresented in 60 trials. The neutral word was the Hebrew word for "hat,"which was primed in the remaining 60 trials. These primes were matchedin number of letters. The 60 target letter strings were constructed inaccordance to 6 categories. Proximity words were 5 Hebrew words thatconnote the attainment of proximity to others (e.g., closeness, love, hug).Distance words were 5 Hebrew words that connote interpersonal distanceor the failure to attain proximity (e.g., separation, rejection, abandonment).Positive words were 5 Hebrew words that have a positive connotation butno link to proximity themes (e.g., brightness, honesty, efficacy). Negativewords were 5 Hebrew words that have a negative connotation but no linkto proximity themes (e.g., dullness, cheat, lazy). Neutral words were 10Hebrew words that have no salient positive or negative connotations and nolink to proximity themes (e.g., office, table, boat). Nonwords were 30

nonwords that were generated by taking common Hebrew verbs and nouns(e.g., thinking) and changing one letter (e.g., tlinking). All the words andnonwords were matched for number of letters. In this way, there were 120primes-target letter strings pairs (2 X 60), which were presented once fora total of 120 trials. Trials were randomly ordered across participants.

In order to validate the categorization of the target words, a sample of 10undergraduate psychology students rated the extent to which each of theabove-described 30 words (a) reflected themes related to interpersonalrelationships and (b) had a positive affective meaning. These ratings wereperformed on a 6-point scale ranging from not at all (1) to very much (6).Ratings of each participant were averaged across word categories andanalyses of variance (ANOVAs) for repeated measures were performedexamining differences between word categories. These analyses revealedthat proximity and distance words were rated as reflecting more interper-sonal relationship themes (M = 4.12, M - 4.48) than positive, negative,and neutral words (M = 2.96, M = 2.66, M = 2.52), F(4, 36) = 25.01, p <.01. In addition, proximity and positive words were rated as having morepositive meaning (M = 4.42, M = 4.56) than neutral words (M = 3.17),which, in turn, were rated as having more positive meaning than distanceand negative words (M = 2.10, M = 2.26), F(4, 36) = 43.57, p < .01.These findings validated our categorization.

Results and Discussion

For each person, RTs (only for correct responses) were averagedaccording to type of target stimuli (proximity words, distancewords, positive words, negative words, neutral words, and non-words).2 The analyses of these RTs are presented in two sections.In the first section, we examine the hypothesized normative com-ponent of the stress-attachment link and mainly focus on our firstprediction. Specifically, we examine whether a stress word primewould lead to faster RTs for proximity words that a neutral wordprime. In the second section, we examine the intraindividualcomponent of the stress-attachment link, focusing on the role thatattachment style may play in moderating the association betweena stress context and RTs for proximity and distance words. Thisorganization of the statistical analyses will be also used in Stud-ies 2 and 3.3

Stress-eliciting events and the accessibility of proximity-relatedthoughts. In examining the hypothesized normative componentof the stress-attachment link, we conducted a two-way ANOVAfor prime presentation (supraliminal, subliminal) and primed word(stress, control) on RTs for proximity words. The last factor was awithin-subject measure. This analysis yielded a significant maineffect for primed word, F(l, 56) = 7.02, p < .05, TJ2 = .12. Othereffects were not significant. In line with our prediction, a stressword prime led to faster RTs for proximity words (M = 588.89)than a neutral word prime (M = 624.92). The lack of a significantinteraction implied that this effect of a stress word prime on theaccessibility of proximity thoughts was significant when the prime

2 Analyses in all the studies were conducted on RTs for correct re-sponses. Long RTs (greater than 1,200 ms) were considered incorrectresponses and were excluded from the analyses. The average percentage oftrials in which incorrect responses were recorded (responding to nonwordor vice versa) was low in the three studies (between 2.3% to 4.8%). Theaverage percentage of trials in which long RTs were recorded was also lowin all the studies (between 2.6% to 3.5%). Incorrect trials and long RTstrials were randomly distributed across target stimuli and priming words.

3 According to Cohen's (1988) tables, the estimated power for theanalyses in all the studies was .80 (alpha = .05).

514 MIKULINCER, BIRNBAUM, WODDIS, AND NACHMIAS

Table 1Means of Reaction Times According to Attachment Style, Primed Word Presentation,Primed Word, and Target Word (Study 1)

Target and primed word

ProximityStress

MSD

NeutralMSD

DistanceStress

MSD

NeutralMSD

Subliminal presentation

Secure(n = 10)

587.89b

91.03

639.07,101.89

637.36,91.30

634.63a

64.99

Avoidant(n = 10)

561.02b

91.61

658.63a

109.25

662.68,88.27

662.09a

111.58

Ambival.(n = 10)

599.95b

87.86

571.81b

64.03

581.95b

64.57

589.50b

70.48

Supraliminal presentation

Secure(« = 10)

557.51b

89.33

648.96,99.89

644.06,89.20

638.71,117.78

Avoidant(n = 10)

611.84b88.06

646.34,94.07

663.14,109.55

645.06,82.25

Ambival.(« = 10)

615.16b109.98

594.61b

79.92

602.60b

77.18

594.97b

78.75

Note. Ambival. = ambivalent. Means with different subscripts are significantly different at p < .05.

was presented either supraliminally or subliminally. Importantly,other ANO VAs found no significant effects of prime category andprime presentation on RTs for the other target stimuli (distancewords, positive words, negative words, neutral words, and non-words).4 Overall, the current findings provide support for thenormative component of the stress-attachment link.

The moderating effects of attachment style. In examining theintraindividual component of the stress-attachment link, we con-ducted three-way ANOVAs for attachment style, prime presenta-tion, and prime category on RTs for proximity and distance words.The last factor was a within-subject measure. In addition, weconducted Pearson correlations between the two continuous at-tachment scores (anxiety, avoidance) and RTs for proximity anddistance words. These correlations were separately performed foreach type of primed word and in each priming procedure (sublim-inal, supraliminal).

The three-way ANOVA performed on RTs for proximity wordsyielded the above-reported significant main effect for prime cate-gory and a significant interaction for Attachment Style X PrimeCategory, F(2, 54) = 5.89, p < .01, rf = .15. Other effects werenot significant. Tests for simple main effects indicated that in trialsin which a neutral word was primed, anxious-ambivalent personsreacted quicker to proximity words (M = 583.21) than secure(M = 639.05) and avoidant (M = 652.49) persons, F(2,54) = 6.43, p < .01. However, this difference vanished in trials inwhich a stress word was primed (A/ = 572.70 for secure, M =586.43, for avoidant, and M = 607.55 for anxious-ambivalentpersons). In these trials, all the three attachment groups showedrelatively fast RTs (see Table 1).

The three-way ANOVA performed on RTs for distance wordsyielded a significant main effect for only attachment style, F(2,54) = 3.47, p < .05, rf = .11. Scheffe" tests indicated thatanxious-ambivalent persons reacted faster to distance words (M =598.15) than both secure (M = 638.69) and avoidant (M = 658.24)persons. The lack of a significant interaction implied that thisattachment-style difference was significant upon the priming ofeither a neutral or a stress word (see Table 1).

Three-way ANOVAs conducted on RTs to attachment-unrelatedwords (positive, negative, neutral) and nonwords yielded no sig-nificant effect. This finding implies that the above-reported sig-nificant interactions were specific for attachment-related words.5

Pearson correlations indicated that attachment anxiety was sig-nificantly and inversely related to RTs for proximity words only intrials in which a neutral word was primed, either supraliminally,r(28) = - .39, p < .05, or subliminally, r(28) = - .36, p < .05.This association was not significant when a stress word wasprimed. Attachment anxiety was also significantly and inverselyassociated with RTs for distance words when a neutral or stressword was primed, either supraliminally or subliminally (rs be-tween - .38 to -.46). Fitting the ANOVAs findings, attachmentanxiety was associated with faster RTs for proximity words onlywhen a neutral word was primed as well as with faster RTs fordistance words when either a neutral or stress word was primed.No significant link was found between attachment avoidance andRTs. Moreover, no significant association was found betweenattachment scores and RTs for other attachment-unrelated stimuli.6

Conclusions. Overall, the findings for proximity words were inline with our predictions about the normative component of thestress-attachment link. All the participants, regardless of their attach-ment style, showed relatively fast RTs for these words upon thepriming of a stress-inducing word. We found this pattern of accessi-

o n all the three studies, the introduction of RTs for attachment-unrelated words (positive, negative, neutral) or nonwords as covariates inthe ANOVAs examining the effects of stress-inducing stimuli on RTs forproximity and distance words did not change the originally reportedfindings.

5 In all the three studies, the reported effects of attachment style on RTsfor proximity and distance words remained unchanged when we controlledfor RTs for attachment-unrelated words (positive, negative, neutral) ornonwords as covariates.

6 In all the three studies, the reported Pearson correlations remained thesame when we statistically controlled for RTs for attachment-unrelatedwords (positive, negative, neutral) or nonwords.

STRESS AND ATTACHMENT 515

bility in the three attachment groups and in both supraliminal andsubliminal presentations of the primes. Importantly, this effect cannotbe attributed to a nonspecific stress effect. In fact, RTs for nonwordsand attachment-unrelated words, either neutral, positive, or negative,were not affected by the priming of a stress word.

Despite the potential importance of the subliminal versus supra-liminal priming procedure, this manipulation failed to have anyimpact. One possibility is that a stress word prime had an impacton RTs for proximity words even when participants could recog-nize and consciously process the meaning of the prime. However,more research is needed before reaching such a conclusion. Re-search should use different cognitive tasks, different priming pro-cedures, and different primes and target words.

With regard to the intraindividual component of the stress-attachment link, the findings for secure persons were in line withour predictions: A stress word prime led to higher accessibility ofattachment themes than a neutral word prime. In addition, a stressword prime did not affect secure persons' RTs for distance words.In other words, secure persons' heightened accessibility of attach-ment themes under a stress context was limited to the positiveaffective connotations of these themes (e.g., love, closeness).

With regard to anxious-ambivalent persons, the findings clearlyshowed that they reacted relatively fast to proximity and distancewords even when a neutral word was primed. This heightenedaccessibility was maintained when a stress word was primed. Thispattern of findings might reflect anxious-ambivalent persons' ha-bitual hyper-activation of attachment and distress cues (Kobak etal., 1993). These persons have been found to exaggeratedly rumi-nate on attachment themes and worries and to suffer from chronicdistress (Mikulincer & Florian, 1998). As a result, they may showchronic activation of the attachment system and have relativelyhigh accessibility to attachment themes and worries even in neutralcontexts. Moreover, the tendency of these persons to appraiseenvironmental transactions in threatening terms (Mikulincer &Florian, 1995) might have led them to react to the lexical decisiontask itself with heightened distress, which, in turn, might haveincreased the accessibility of attachment themes and worries.

Interestingly, avoidant persons' RTs for proximity and distancewords resembled those of secure persons: They showed faster RTs forproximity words upon a stress than neutral word prime, but lowaccessibility to distance words even upon the priming of a stress word.Avoidant persons' low accessibility to proximity worries may reflecttheir habitual defensive suppression of distressing thoughts (Miku-lincer & Florian, 1998). Alternatively, our findings may suggest thatthe attachment schemas of avoidant persons may resemble those ofsecure persons and may be organized around positively valued cog-nitions. Unfortunately, we had no data testing these alternatives.

One way of testing the above interpretations is by priming astress word that has semantic links with proximity worries and canassociatively activate them. On the one hand, if avoidant persons'reactions result from the suppression of distressing thoughts, theywould still show low accessibility to proximity worries after thesemantic activation of these worries. On the another hand, ifavoidant persons' reactions have nothing to do with suppression,they may react, like secure persons, with increased accessibility tothe semantic activation of proximity worries. These alternativeswere examined in Study 2.

Study 2

Study 2 has two main purposes. First, it attempts to replicatefindings of Study 1 by priming a different stress word. Second, itattempts to examine avoidant persons' defensive suppression ofproximity worries by priming a stress word that has semantic linkswith these worries. For these purposes, we chose "death" as thestress word prime. Death has generally been seen as a source ofterror for most individuals (Bakan, 1971; Pollack, 1979), and it isassociated with proximity worries—the irreversible separationfrom attachment figures. In fact, death has been viewed as theultimate separation, and fear of death has been related to the fearof separation from attachment figures (Kalish, 1985; McCarthy,1981). Then, the priming of the word "death" would not onlyactivate the attachment system and heighten the accessibility ofattachment themes due to its stressful connotations, but it couldalso semantically activate proximity worries and fasten RTs fordistance words. This effect is expected to be found in mostindividuals, with the exception of avoidant persons who maymentally suppress distressing thoughts.

In Study 2, participants who were previously classified accord-ing to attachment style performed a 60-trial lexical decision task.In each trial, they decided whether a letter string was a word or notupon the subliminal priming of a neutral or stress word. For half ofthe participants, the prime was a neutral word ("hat"). For theremaining participants, the prime was the word "death." BecauseStudy 1 showed no difference between supraliminal and sublimi-nal priming, we decided to use only the subliminal priming con-dition. In addition, because Study 1 showed no significant differ-ence between neutral, positive, and negative words, we decided tocontrol only for RTs to neutral words. Then, participants receivedonly four categories of target letter strings: nonwords, neutralwords, proximity words, and distance words.

With regard to the normative component of the stress-attachment link, we predicted that the priming of the word "death"would lead to faster RTs for proximity words than the priming ofa neutral word, regardless of a person's attachment style. Withregard to the intrapersonal component, we predicted that the prim-ing of the word "death" would lead to faster RTs for distancewords than the priming of a neutral word among secure andanxious-ambivalent persons but not among avoidant persons.

Method

Participants. Seventy-two participants (24 secure, 24 avoidant, and 24anxious-ambivalent), who did not participate in Study 1, were randomlychosen from the pool of students described in Study 1. The sampleconsisted of 51 women and 21 men ranging in age from 18 to 31 (me-dian = 22). Participants in each attachment style were randomly dividedinto two experimental conditions, each consisting of 36 participants.

Materials and procedure. Participants were invited by phone to par-ticipate in a study on cognitive psychology on an individual basis. Theexperimenter, who was blind to participants' attachment style, told themthat they would perform the lexical decision task described in Study 1.Each trial of this task consisted of a subliminal presentation (20 ms) of aprime followed, after a pause of 500 ms, by the presentation of 1 of 60target letter strings (for 1,000 ms). The 60-letter strings consisted of 10proximity words, 10 distance words, 10 neutral words, and 30 nonwords.Most of the strings were identical to those described in Study 1, with theexception of 5 new Hebrew proximity-related words (e.g., kiss, caress,affection) and 5 new Hebrew distance-related words (e.g., distance, lone-

516 MIKULINCER, BIRNBAUM, WODDIS, AND NACHMIAS

Table 2Means of Reaction Times According to Attachment Style, Primed Word, andTarget Word (Study 2)

Target word

ProximityMSD

DistanceMSD

Priming of neutral

Secure(n = 12)

698.33a

63.03

672.94a

71.59

Avoidant(n = 12)

647.81a

70.91

665.41a

85.01

word

Ambival.(n = 12)

573.25b

79.01

568.6A73.50

Priming

Securein = 12)

573.64,,63.79

562.52b

67.50

of the word

Avoidant(n = 12)

566.76b63.29

666.90a

83.85

"death"

Ambival.(n = 12)

567.36b87.38

581.25a

81.77

Note. Ambival. = ambivalent. Means with different letters are significantly different at p < .05.

liness, alone). These words were matched for number of letters with thoseused in Study 1. Five judges rated the valence of these words and relevanceto interpersonal relationships. Ratings were similar to those reported inStudy 1. For half of the participants in each attachment style, the prime inall the 60 trials was the word "death." For the remaining, the prime in allthe 60 trials was the word "hat."

Results and Discussion

Stress-eliciting events and the accessibility of proximity-relatedthoughts. In examining the normative component of the stress-attachment link, a one-way ANOVA for primed word (death,neutral) yielded a significant effect on RTs for proximity words,F(l, 70) = 14.39, p < .01, T)2 = .17. Cell means revealed that thepriming of the word "death" led to faster RTs for proximity words(M = 569.26) than the priming of a neutral word (M = 638.89).Additional ANOVAs found no significant effect of primed wordon RTs for the other target stimuli. Overall, these findings pro-vided further support for the normative component of the stress-attachment link.

The moderating effects of attachment style. In examining theintraindividual component of the stress-attachment link, a two-way ANOVA for attachment and primed word performed on RTsfor proximity words yielded the above reported significant maineffect for primed word and a significant interaction, F(2,66) = 4.20, p < .01, T)2 = .11. Tests for simple main effectsindicated that anxious-ambivalent persons reacted quicker to prox-imity words than secure and avoidant persons only when a neutralword was primed, F(2, 66) = 9.35, p < .01 (see Table 2). Thisdifference vanished when the word "death" was primed. In thiscase, all the attachment groups showed fast RTs for proximitywords.

The ANOVA for RTs to distance words yielded a significantmain effect for attachment style, F(2, 66) = 8.32, p < .01, TJ2 =.19. Scheff6 post hoc tests indicated that anxious-ambivalent per-sons reacted faster to distance words (M = 574.94) than avoidantpersons (M - 666.15). However, this effect was qualified by asignificant interaction, F(2, 66) = 4.63, p < .05, TJ2 = .11. Testsfor simple main effects indicated that anxious-ambivalent personsreacted quicker to distance words than secure and avoidant personsupon a neutral word prime, F(2, 66) = 6.86, p < .01 (see Table 2).When the word "death" was primed, both anxious-ambivalent andsecure persons reacted quicker to distance words than avoidant

persons, F(2, 66) = 6.10, p < .01 (see Table 2). In fact, thepriming of the word "death" led to faster RTs for distance wordsthan the priming of a neutral word only among secure persons,F(l, 66) = 15.11, p < .01. Anxious-ambivalent people showedfast RTs in the two priming conditions, and avoidant peopleshowed relatively slow RTs.

Importantly, ANOVAs for attachment style and primed wordconducted on RTs for neutral words and nonwords yielded nosignificant effect. This finding implies that the above-reportedsignificant interactions were specific for attachment-related words.

Pearson correlations revealed that the attachment anxiety scorewas significantly and inversely related to RTs for proximity anddistance words only in trials in which a neutral word was primed,r(34) = - .53 ,p < .01, and r{34) = -A9,p < .01. However, thesecorrelations were not significant when the "death" word wasprimed. Fitting ANOVAs findings, attachment anxiety was asso-ciated with faster RTs for proximity and distance words only whena neutral word was primed. In addition, in line with ANOVAsfindings, attachment avoidance was significantly and positivelyrelated to RTs for distance words when the "death" word wasprimed, r(34) = .41, p < .01, but not when a neutral word wasprimed. That is, under the priming of the word "death," the highera person's avoidance, the slower the RTs for distance words. Nosignificant association was found between the two attachmentscores and RTs for neutral words and nonwords.

Conclusions. Findings of Study 2 provided important infor-mation about the normative and intraindividual components of thestress-attachment link. With regard to the normative component,the findings replicated those of Study 1: All the participants,regardless of their attachment style, showed relatively fast RTs forproximity words upon the priming of a stress word. This findingindicated that the heightened accessibility of attachment themes instress contexts was not limited to the priming of a particular stressword (e.g., failure), but it could be generalized to the other stressprimes ("death").

The current findings may have some implications for TerrorManagement Theory (Greenberg, Pyszczynski, & Solomon, 1997).According to this theory, making mortality salient (e.g., the sub-liminal priming of the word "death") activates anxiety-bufferingmechanisms, which attempt to validate one's worldviews or self-esteem. In the current study, we observed that this manipulationactivated cognitions related to love and companionship, which, in

STRESS AND ATTACHMENT 517

turn, may have positive affective implications. Thus, these attach-ment cognitions may serve as alternative means for managingdeath anxiety. However, this idea is highly speculative because ourstudy was not designed within a terror management perspectiveand no evidence was collected about the defensive function ofattachment cognitions. Further research should attempt to dealwith the potential associations between terror management andattachment theories.

With regard to the intraindividual component of the stress-attachment link, the findings for anxious-ambivalent persons rep-licated those of Study 1: These persons showed relatively fast RTsto proximity and distance words upon the priming of neutral orstress words. However, the most original finding concerns thereactions of secure and avoidant persons to distance words uponthe priming of the word "death." Overall, the results were in linewith our prediction. On the one hand, secure persons were respon-sive to the semantic link between death and proximity worries and,then, reacted faster to distance words in the "death" than neutralpriming condition. On the another hand, avoidant persons showedslow RTs to distance words even upon the priming of the word"death." In this case, attachment avoidance was associated withslower RTs. These findings may reflect avoidant persons' suppres-sion of proximity worries.

Study 3

Study 3 also attempts to replicate findings of Study 1 by priminga different stress word ("illness") and to examine avoidant per-sons' defensive suppression of proximity worries. According toWegner (1994), this defensive reaction is a cognitive effortfuloperation that cannot be activated without the availability of suf-ficient cognitive resources. Findings show that tasks that drawaway the resources needed to perform mental suppression heightenthe accessibility of the to-be-suppressed material (Wegner, 1994).Thus, if suppression underlies avoidant persons' low accessibilityto proximity worries, procedures that tax on cognitive resourceswould heighten the accessibility of these worries and lead to fasterRTs for distance words. In examining this reasoning, we tax on aperson's cognitive resources through a cognitive load technique(e.g., Gilbert & Hixon, 1991; Gilbert & Osborne, 1989)—perform-ing the lexical decision task together with another effortful cogni-tive task (rehearsing a to-be-learned text). This technique has beenfound to result in increased accessibility to the to-be-suppressedmaterial (Arndt, Greenberg, Solomon, Pyszczynski, & Simon,1997; Wegner, Erber, & Zanakos, 1993).

In Study 3, participants who were classified according to attach-ment style performed a lexical decision task (see Study 1). In eachtrial, lexical decisions were made upon the subliminal priming ofeither a neutral or a stress word. Half of the participants performedthis task in the way described in Study 1 (control, nonload). Theremaining participants performed the lexical decision task togetherwith another task (load). On this basis, Study 3 consisted of a 3 X2 X 2 factorial design for attachment style, cognitive load (load,nonload), and primed word (neutral, stress). The last factor was awithin-subject measure.

With regard to the normative component of the stress-attachment link, we predicted that the priming of a stress wordwould lead to faster RTs for proximity words than the priming ofa neutral word. This effect would be found in the three attachment

styles and in both nonload and load conditions. With regard to theintraindividual component, we predicted a complex interactionfor RTs to distance words. In the nonload condition, anxious-ambivalent persons would react to a stress word prime with fasterRTs than secure and avoidant persons. In the load condition, bothanxious-ambivalent and avoidant persons would react to a stressword prime with faster RTs than secure persons.

Method

Participants. Fifty-eight participants (20 secure, 20 avoidant, and 18anxious-ambivalent) were randomly chosen from the sample described inStudy 1 (36 women and 22 men ranging in age from 20 to 28, median =21). Participants in each attachment style were randomly divided into twoconditions, each consisting of 29 participants.

Materials and procedure. Instructions and materials were similar tothose of Study 1. In Study 3, the lexical decision trials consisted of asubliminal priming (20 ms) of a stress word ("illness") or a neutral word("hat") followed, after a pause of 500 ms, by the presentation of 1 of 60target letter strings (for 1,000 ms). The 60 strings were those used inStudy 2 and consisted of 10 proximity words, 10 distance words, 10 neutralwords, and 30 nonwords. There were 120 primes-target strings pairs (2 x60), which were presented once (120 trials). Trials were randomly orderedacross participants.

Participants were randomly divided into two conditions. In the control,nonload condition, participants performed the task in the way described inStudies 1-2. In the load condition, participants were told that "to make thistask a little more difficult," they would perform another task at the sametime. This task involved hearing a tape-recorded story, repeating it in aloud voice, and trying to remember it for later. Then, participants receiveda headphone set connected to a tape recorder and performed the lexicaldecision task while hearing and repeating in a loud voice a story about aJewish community. Following the lexical decision task, participants re-ceived 5 open questions examining whether they had attended to the story.Overall, participants (in average) correctly answered 3 of the 5 questions,and no difference was found among attachment groups.

Results and Discussion

Stress-eliciting events and the accessibility of proximity-relatedthoughts. In examining the normative component of the stress-attachment link, a two-way ANOVA for cognitive load (load,nonload) and primed word (stress, control) on RTs for proximitywords, with the last factor was a within-subject measure, yielded asignificant main effect for primed word, F(l, 56) = 21.01,/? < .01,T)2 = .24. Other effects were not significant. In line with ourprediction, a stress word prime led to faster RTs for proximitywords (M = 571.04) than a neutral word prime (M = 639.84). Thelack of a significant interaction implied that this effect was foundin both load and nonload conditions. Other ANOVAs found nosignificant effects on RTs for the other target stimuli. Like inStudies 1-2, the findings provide support for the normative com-ponent of the stress-attachment link.

The moderating effects of attachment style. In examining theintraindividual component of the stress-attachment link, we per-formed three-way ANOVAs for attachment style, cognitive load,and primed word on RTs for proximity and distance words. Thelast factor was a within-subject measure. Beyond the significantmain effect for primed word, the ANOVA performed on RTs forproximity words yielded a significant main effect for attachmentstyle, F(2, 52) = 7.73, p < .01, TJ2 = .21, and a significant

518 MIKULINCER, BIRNBAUM, WODDIS, AND NACHMIAS

Table 3Means of RTs According to Attachment Style, Cognitive Load, Primed Word, andTarget Word (Study 3)

Target and primed word

ProximityStress

MSD

NeutralMSD

DistanceStress

MSD

NeutralMSD

Control

Secure(n = 10)

563.29b

108.88

671.02a

85.30

63O.13a

74.91

687.02a

116.06

nonload condition

Avoidant(n = 10)

592.71b

67.89

694.3 la

71.88

639.24a

77.87

656.69a

54.69

Ambival.(n = 9)

539.85b

82.39

579.15b

75.23

578.03b

57.80

548.84b76.18

Cognitive load condition

Secure(n = 10)

568.25b

76.32

659.83a

92.01

667.03a

85.42

657.02a

74.66

Avoidant(n = 10)

586.14,,79.51

667.63a

89.35

588.96b

92.88

686.95a

61.54

Ambival.(n = 9)

573.04b

63.76

552.26,,69.15

574.05b

80.40

651.97a

70.57

Note. Ambival. = ambivalent. Means with different subscripts are significantly different at p < .05.

interaction for Attachment Style X Primed Word, F(2, 52) = 3.83,p < .05, rj1 = .13. Other effects were not significant. Scheffe testsrevealed that anxious-ambivalent people reacted quicker to prox-imity words (M = 561.67) than secure (M = 615.59) and avoidant(M = 635.20) people. However, tests for simple main effectsindicated that this effect was significant only when a neutral wordwas primed (M = 565.70 for anxious-ambivalents, M = 665.42for secures, M = 680.97 for avoidants), F(2, 52) = 11.98, p < .01.This difference vanished when a stress word was primed, whereinall the attachment groups showed fast RTs. The lack of signifi-cance of cognitive load implied that this pattern of effects, whichreplicated findings of Studies 1-2, was significant in both load andnonload conditions.

The three-way ANOVA performed on RTs for distance wordsyielded a significant main effect for attachment style, F(2,52) = 5.91, p < .01, T)2 = .20, and a significant three-wayinteraction, F(2, 52) = 6.19, p < .01, TJ2 = .19. Scheffe testsindicated that anxious-ambivalent persons reacted faster to dis-tance words (M = 588.22) than both secure (M = 660.30) andavoidant (M = 642.96) persons. Tests for simple main effectsindicated that this pattern of differences was significant only in thenonload condition, whether a neutral word or a stress word wasprimed, F(2, 52) = 14.25, p < .01, and F(2, 52) = 3.04, p < .05(see Table 3). That is, when no load was added, the findingsreplicated those of Study 1: Anxious-ambivalent persons showedfaster RTs for distance words than secure and avoidant persons.However, when cognitive load was added, both anxious-ambivalent and avoidant persons reacted to a stress word primewith faster RTs than secure persons, F(2, 52) = 11.37, p < .01 (seeTable 3). In fact, the addition of cognitive load, as compared withthe nonload condition, led avoidant persons to react quicker todistance words upon the priming of a stress word, F(l, 52) = 3.54,p < .05 (see Table 3). Additional ANOVAs conducted on RTs forneutral words and nonwords yielded no significant effect.

Like in Studies 1-2 and fitting the ANOVAs findings, Pearsoncorrelations indicated that attachment anxiety was significantly

related to (a) faster RTs for proximity words only when a neutralword was primed, r(21) = —.52, p < .01 for nonload condition,r(27) = - .38, p < .05 for load condition, and (b) faster RTs fordistance words in most of the conditions, r(27) = - .47, p < .05for nonload-heutral word prime, rill) = —.39, p < .05 fornonload-stress word prime, r(27) = —.47, p < .05 for load-stressword prime. Attachment avoidance was significantly related tofaster RTs for distance words only when cognitive load was addedand a stress word was primed, r(27) = —.36, p < .05. Again, likein Studies 1-2, no significant association was found between thetwo attachment scores and RTs for neutral words and nonwords.

Conclusions. The findings were in line with our predictions.With regard to the normative component of the stress-attachmentlink, the findings showed that a stress word prime led to fast RTsfor proximity words in all the attachment styles. This effect wasfound in both load and nonload conditions. Again, like in Studies1-2, a stress word prime heightened the accessibility of attachmentthemes even among insecure persons.

With regard to the intraindividual component, the findings in thenonload condition replicated those observed in Study 1. First, bothsecure and avoidant persons showed (a) faster RTs for proximitywords when a stress word was primed than when a neutral wordwas primed but (b) relatively slow RTs for distance words uponthe two primes. Second, anxious-ambivalent people showed rela-tively fast RTs to proximity and distance words when either aneutral or stress word was primed. However, the most originalfinding concerns the effects of cognitive load on avoidant persons'RTs. In this case, avoidant persons, like anxious-ambivalent per-sons, reacted to a stress word prime with faster RTs to distancewords than secure persons. That is, the specific combination ofcognitive load and a stress word prime heightened the accessibilityof avoidant persons' attachment worries. This finding is in linewith the idea that avoidant persons' low accessibility of attachmentworries observed in Studies 1-2 and in the current nonload con-dition might have resulted from a defensive suppression of theseworries.

STRESS AND ATTACHMENT 519

The current findings suggest that adding cognitive load duringthe lexical decision task, which might have reduced the ability torespond with strategic avoidance of attachment worries, may bethe best way to examine the intraindividual component of thestress-attachment link. However, one should take into account thatStudies 1-2 also provided important information about this intra-individual component. First, they consistently documentedanxious-ambivalent persons' heightened accessibility of attach-ment themes and worries in neutral conditions. Second, theyconsistently indicated that only anxious-ambivalent persons re-acted to a stress word prime with heightened accessibility ofattachment worries during the typical lexical decision task (with-out cognitive load). Third, they showed that these attachment-styledifferences could be replicated using different stress words and inboth subliminal and supraliminal priming procedures. Fourth,Study 2 showed that a stress word prime that was semanticallylinked with attachment worries heightened secure persons' acces-sibility to these worries but failed to do so among avoidant per-sons. Overall, the three studies provided compelling evidenceabout the accessibility of attachment worries in each of the attach-ment styles.

General Discussion

The current findings provide important information aboutthe cognitive makeup of the attachment system in adulthood.First, they support Bowlby's (1969) hypothesis about a stress-attachment link and suggest that the encounter with actual orsymbolic stress-eliciting stimuli activate cognitive representationsof the attachment system. That is, attachment themes and worriesmay be active and may preconsciously influence a person's judg-ments and behaviors during the encounter with stressful events,even when these events have nothing to do with interpersonalrelationships or the frustration of attachment needs. Second, thefindings delineate attachment-style differences in the hedonic va-lence of the cognitive activation of the attachment system. Third,they offer new insights on the regulatory mechanisms that underliesuch an activation. Taken as a whole, the current series of studiesopens a new avenue of research, integrating the normative andintraindividual components of the attachment theory.

The findings clearly indicate that people reacted relatively fastto words reflecting proximity, closeness, and love upon the prim-ing of a stress word (e.g., failure, illness, death). This pattern ofaccessibility has been found regardless of a person's attachmentstyle and under different priming conditions (subliminal and su-praliminal, with and without cognitive load). In addition, thispattern was not constrained to the priming of a specific stressword. Rather, it was replicated when attachment-related orattachment-unrelated stress words were primed. That is, even astressful context that is not directly related to attachment themesmay heighten the accessibility of these themes. Importantly, therelatively fast RTs for proximity words under a stress contextcannot be explained by the hedonic valence of these words or bynonspecific effects of stress-eliciting stimuli.

The above finding emphasizes the relevance of the attachmentsystem in stress contexts. Stressful stimuli may activate not onlythoughts about the event in question but also thoughts related tolove and support. That is, people exposed to a stressful event maybe consciously occupied or preoccupied with this event, but their

mental processes may be affected by highly accessible attachmentthemes. This line of thinking implies that a person's judgments andbehaviors during stressful episodes may be shaped by consciousthoughts about the impinging distress as well as by the activationof attachment themes.

Importantly, the current findings also imply that the way bywhich the attachment system may contribute to cognitive, affec-tive, and behavioral reactions to stress may depend on a person'sattachment style. Although stressful stimuli were found to activateattachment themes in most of the participants, attachment stylewas still relevant for explaining the specific attachment thoughtsthat were accessible and the regulatory mechanisms underlyingsuch activation. For example, whereas thoughts related to proxim-ity and love were activated regardless of attachment style, worriesabout separation and rejection were mainly activated among inse-cure persons. Moreover, the pattern of RTs for both proximity anddistance words among anxious-ambivalent and avoidant personsseem to reflect the strategies these persons habitually use inregulating distress (hyperactivation of the attachment system, sup-pression of worries). As a result, people differing in attachmentstyle may have accessibility to different types of attachmentthoughts, which, in turn, may have different implications formanaging distress.

Secure persons showed the expected pattern of association be-tween stress primes and the cognitive activation of the attachmentsystem. First, they reacted to stress primes with heightened acces-sibility to thoughts about proximity and love. Importantly, thispattern of accessibility was limited to stress primes. In fact, securepersons showed relatively low accessibility to proximity thoughtsupon the priming of a neutral word. This finding implies thatsecure persons' cognitive system is not chronically occupied withattachment themes and that these themes may not influence be-havior in every environmental transaction. Rather, these themesseem to be functionally activated when a transaction signals thatsome coping action should be taken. That is, the cognitive activa-tion of attachment themes among secure persons seems to fitBowlby's (1969) ideas about the regulatory function of the attach-ment system. Of course, attachment themes may be also accessiblein nonstress conditions, when contextual cues are relevant to thesethemes (e.g., reunion with a significant other). What our findingsonly imply is that secure persons' cognitions in daily life are notnecessarily guided by attachment needs.

Second, secure persons' reactions to stress primes were limitedto attachment themes that have positive affective connotations. Infact, secure people showed relatively low accessibility to proxim-ity worries under either neutral or stress contexts. This pattern ofaccessibility fits secure persons' positive working models of theself and the world (Collins & Read, 1994) and may reflect theirpositive history of attachment relationships (Shaver & Hazan,1993). It is possible that this positive history may create aninhibitory link in the semantic memory network between theactivation of the attachment system and worries about rejectionand separation. Of course, one can alternatively suggest that thefindings reflect the suppression of distressing thoughts. However,findings of Studies 2-3 reject this alternative. Secure personsshowed heightened accessibility to proximity worries upon a primethat was semantically linked with these worries. In addition, cog-nitive load had no effect on secure persons' accessibility to prox-imity worries.

520 MIKULINCER, BIRNBAUM, WODDIS, AND NACHMIAS

Overall, the findings for secure persons reflect a functionalactivation of the attachment system. First, this system seems to bemainly activated upon signals of threat for a person's well-being.Second, this activation is circumscribed to attachment themes thathave positive affective connotations and may have beneficial con-sequences for a person's well-being. In fact, thinking about loveand closeness to a significant other may lead to a state of antici-pated relief and comfort and, then, reduce the distress originallyelicited by a stressful event. Accordingly, this activation mayunderlie secure persons' optimistic and hopeful judgments andtheir tendency to seek support in times of need (Mikulincer &Florian, 1998). Of course, our findings show only that securepersons reacted to stress primes with heightened accessibility toattachment themes. However, they provided no data about theaffective outcomes of such a pattern of accessibility. Furtherresearch should attempt to examine the affective consequences ofaccessible thoughts about love.

Anxious-ambivalent persons' pattern of accessibility to attach-ment themes was opposite to that observed among secure persons.First, they showed relatively high accessibility to attachmentthemes under either stress or nonstress contexts. Second, theyshowed relatively high accessibility to thoughts about proximityand love as well as to proximity-related worries. That is, anxious-ambivalent persons' cognitive activation of the attachment systemseems to be colored by worries about separation and rejection andto spread over nonstress contexts. The high accessibility ofproximity-related worries fits anxious-ambivalent persons' nega-tive working models of the self and the world (Collins, 1996;Mikulincer, 1998a) and may reflect their habitual expectation thatthe expression of attachment needs would be followed by rejectionrather than by love (Shaver & Hazan, 1993). The high accessibilityof attachment themes in nonstress contexts may reflect anxious-ambivalent persons' chronic hyperactivation of the attachmentsystem (Mikulincer & Florian, 1998) or their appraisal of dailytransactions with the environment in threatening terms (Shaver &Hazan, 1993).

The findings for anxious-ambivalent persons reflect a dysfunc-tional activation of the attachment system. First, this system seemsto be chronically activated even when there is no signal of threatand no demand for coping actions. Second, this activation iscolored by proximity-related worries, which may have negativeconsequences for a person's well-being. In fact, thinking aboutrejection and separation may compound the distress originallyelicited by the stressful event and may underlie anxious-ambivalent persons' chronic distress (Mikulincer & Florian, 1998).Moreover, the cognitive linkage between attachment and rejectionmay detract anxious-ambivalent persons from seeking support intimes of need. Overall, anxious-ambivalent persons may react todaily events with heightened accessibility to proximity-relatedworries, which may divert cognitive resources away from task-relevant thoughts, may replete the cognitive system with distress-ing feelings, and may be a source of emotional and functioningproblems.

With regard to the above findings, one could question whetherthey reflect something unique to the anxious-ambivalent style orwhether they are an illustration of trait anxiety. In this connection,it is generally recognized that trait anxiety is manifested on auto-matic attention to threat stimuli (e.g., Mathews & MacLeod, 1994).If one assumes that the anxious-ambivalent style may be corre-

lated with trait anxiety, then one might question the discriminantvalidity of the findings, particularly those involving the anxious-ambivalent style and RTs to distance words following neutralpriming. However, a number of findings seem to provide uniquesupport for the attachment perspective. First, attachment styleswere not significantly associated with RTs to positive and negativewords that were unrelated to attachment themes. Second, theanxious-ambivalent style was related to heightened accessibilityof proximity thoughts. Third, the studies reported that stress in-creased the accessibility of proximity words, which are mostdefinitely positive. In this respect, the findings provide someevidence for the notion that stress activates cognitions related tolove and companionship.

Avoidant persons showed an interesting pattern of findings. InStudy 1, avoidant persons' pattern of accessibility to attachmentthemes resembled that observed among secure persons. That is,attachment themes were activated upon stress arousal, and thisactivation did not include attachment worries. However, Studies2-3 revealed some differences between secure and avoidant per-sons, suggesting that avoidant persons' low accessibility to attach-ment worries might have resulted from a defensive suppression ofthese worries. First, avoidant persons showed low accessibility toattachment worries even when the primed word was semanticallyassociated with these worries. Second, avoidant persons showedheightened accessibility to attachment worries when cognitive loadwas added. Overall, avoidant persons showed heightened accessi-bility to attachment themes in stress conditions and a concomitantmental suppression of attachment worries.

Avoidant persons' heightened accessibility to attachmentthemes in stress contexts provides new and important informationabout their regulatory strategies. At a first sight, our finding is atodds with findings that avoidant persons habitually deactivate theattachment system and prefer cognitive and emotional distancefrom others (Mikulincer et al., 1998; Shaver & Hazan, 1993).However, one should take into account that prior findings arebased on the scrutiny of avoidant persons' conscious self-reportsand overt behaviors. In contrast, our findings seem to reflect thecognitive activation of attachment themes that do not necessarilyreach consciousness or behavioral intention. On this basis, one canconclude that avoidant persons may preconsciously experience theactivation of attachment themes despite their conscious negation ofany need for love and support. This conclusion fits findings show-ing avoidant persons' dissociation between self-reports on the onehand and physiological and projective responses on the another(Dozier & Kobak, 1992; Mikulincer et al., 1990). It also fitsBowlby's (1988) theoretical statements on avoidant persons' in-ability and unwillingness to consciously monitor and express feel-ings of love.

The findings also indicated that avoidant persons showed amental suppression of proximity-related worries. Only in high loadconditions, which may heighten the accessibility of the to-be-suppressed material, avoidance persons showed fast RTs to dis-tance words. This finding may provide a better understanding ofprior findings on the vulnerability of avoidant persons. A numberof studies have pointed out that although avoidant persons showadequate levels of well-being and functioning in daily life, theyexhibit relatively high levels of distress and signs of maladjust-ment in severe and persistent stressful situations (e.g., Mikulincer& Florian, 1998; Mikulincer, Horesh, Eilati, & Kotler, 1999). It is

STRESS AND ATTACHMENT 521

possible that the encounter with severe and persistent stress epi-sodes may impose increasing demands on the cognitive system,which, in turn, may heighten the accessibility of proximity-relatedworries and may create a state of distress that may resemble theemotional turmoil of anxious-ambivalent persons.

Taken as a whole, the current findings refine the hypothesizedstress-attachment link. In normal life circumstances, most of theindividuals, with the exception of anxious-ambivalent persons,may have low accessibility to attachment themes. In this case, onlythe cognitive system of anxious-ambivalent persons' seems to beoccupied with attachment themes and preoccupied with proximity-related worries. However, upon stress arousal, most of the indi-viduals, regardless of their attachment style, may show an activa-tion of the attachment system and attachment themes maypreconsciously influence cognition and behavior. At this point,whereas secure persons may have accessibility only to attachmentthemes that have positive affective connotation, anxious-ambivalent persons may have accessibility also to proximity-related worries. In addition, avoidant persons may mentally sup-press proximity-related worries.

Interestingly, our findings may contribute to theory and researchon cognitive accessibility and priming. Prior studies have mainlyadopted a "semantic-specific" perspective, by which particularthemes are cognitively activated by the priming of semantic-related contexts (e.g., Fischler & Bloom, 1979; Forster, 1981). Inour case, a stress context should make accessible stress-relevantthoughts and an attachment context should make accessibleattachment-relevant thoughts. However, our findings imply thatother types of activation may accompany this semantic activation.That is, a failure in a math exam may cognitively activate attach-ment themes despite that this stressful context has no explicitsemantic association with the attachment system. Of course, fur-ther research should examine the ways by which "semantic-specific" activation and the stress-attachment link coexist togetherand jointly affect cognition and behavior.

Our findings have also important implications for a mood-congruence hypothesis of cognitive accessibility (e.g., Blaney,1986; Bower, 1992). Specifically, they point to a complex view ofaccessibility by which mood congruence may interact with theactivation of regulatory mechanisms. Whereas stress arousal mayactivate stress-related thoughts and this activation may associa-tively spread over proximity-related worries, it may also activateattachment themes that may lead people to actively cope withdistress. This view fits findings on the failure of negative moodinductions to make accessible mood-congruent negative thoughts(e.g., Ucros, 1989). It also fits ideas on the activation of copingmechanisms upon negative mood induction (Clark & Isen, 1982;Singer & Salovey, 1988). Further research should systematicallydeal with the possible interactions between mood congruence andthe activation of the attachment system in stress contexts.

Before ending this discussion, it is important to note that thecurrent study is only a first step in studying the cognitive activationof attachment themes upon stress arousal. One should take intoaccount that our studies were relatively benign and participantswere exposed to only stress-inducing words. Further researchshould examine the cognitive activation of attachment themes indifferent stress contexts as well as differences between attachment-related and attachment-unrelated stressful contexts. More researchis also needed until solid and valid conclusions could be made

about the link between semantic activation in cognitive tasks andthe kinds of deep processes postulated by attachment theorists.This line of research should examine the accessibility of attach-ment themes using other cognitive techniques (e.g., the color-naming Stroop task), generalizing our findings beyond a specificmethodology and overcoming possible limitations of the lexicaldecision task. Research should also tap other preconscious andunconscious manifestations of attachment themes. Finally, re-search should explore the effects that the cognitive activation ofattachment themes may have on behavior. In this way, our findingsopen a new avenue of research that may integrate the normativeand intraindividual components of attachment theory.

References

Ainsworth, M. D. S. (1973). The development of infant-mother attach-ment. In B. M. Caldwell & H. N. Ricciuti (Eds.), Review of childdevelopment research (Vol. 3). Chicago: University of Chicago Press.

Ainsworth, M. D. S., Blehar, M. C , Waters, E., & Wall, S. (1978). Patternsof attachment: Assessed in the strange situation and at home. Hillsdale,NJ: Erlbaum.

Arndt, J., Greenberg, J., Solomon, S., Pyszczynski, T., & Simon, L. (1997).Suppression, accessibility of death-related thoughts, and cultural world-view defense: Exploring the psychodynamics of terror management.Journal of Personality and Social Psychology, 73, 5-18.

Bakan, D. (1971). Disease, pain, and sacrifice. Boston: Beacon.Baldwin, M. W., Fehr, B., Keedian, E., & Seidel, M. (1993). An explora-

tion of the relational schemata underlying attachment styles: Self-reportand lexical decision approaches. Personality and Social PsychologyBulletin, 19, 746-754.

Baldwin, M. W., & Holmes, J. G. (1987). Salient private audiences andawareness of the self. Journal of Personality and Social Psychology, 53,1087-1098.

Bargh, J. A., Chen, M., & Burrows, L. (1996). Automaticity of socialbehavior: Direct effects of trait construct and stereotype activation onaction. Journal of Personality and Social Psychology, 71, 230-244.

Bargh, J. A., & Pietromonaco, P. (1982). Automatic information process-ing and social perception: The influence of trait information presentedoutside of conscious awareness on impression formation. Journal ofPersonality and Social Psychology, 43, 437-449.

Bartholomew, K., & Horowitz, L. M. (1991). Attachment styles amongyoung adults: A test of a four-category model. Journal of Personalityand Social Psychology, 61, 226-244.

Blaney, P. H. (1986). Affect and memory: A review. Psychological Bul-letin, 99, 229-246.

Bower, G. H. (1992). How might emotion affect learning? In S. A.Christianson (Ed.), Handbook of emotion and memory (pp. 3-32). Hills-dale, NJ: Erlbaum.

Bowlby, J. (1969). Attachment and loss: Attachment. New York: BasicBooks.

Bowlby, J. (1973). Attachment and loss: Separation, anxiety and anger.New York: Basic Books.

Bowlby, J. (1980). Attachment and loss: Sadness and depression. NewYork: Basic Books.

Bowlby, J. (1988). A secure base: Clinical applications of attachmenttheory. London: Routledge.

Brennan, K. A., Clark, C. L., & Shaver, P. R. (1998). Self report measure-ment of adult attachment: An integrative overview. In J. A. Simpson &W. S. Rholes (Eds.), Attachment theory and close relationships (pp.46-76). New York: Guilford.

Brooks, J., & Lewis, M. (1976). The effect of time on attachment asmeasured in a free-play situation. Child Development, 45, 311-316.

Cafferty, T. P., Davis, K. E., Medway, F. J., O'Hearn, R. E., & Chappell,

522 MIKULINCER, BIRNBAUM, WODDIS, AND NACHMIAS

K. D. (1994). Reunion dynamics among couples separated during Op-eration Desert Storm: An attachment theory analysis. In K. Bartholomew& D. Perlman (Eds.), Attachment processes in adulthood (pp. 309-330).London: Jessica Kingsley.

Clark, M. S., & Isen, A. M. (1982). Toward understanding the relationshipbetween feelings states and social behavior. In A. Hastorf & A. M. Isen(Eds.), Cognitive social psychology (pp. 73-108). New York: Elsevier.

Cohen, J. (1988). Statistical power analysis for the behavioral sciences(2nd ed.). Hillsdale, NJ: Erlbaum.

Collins, N. L. (1996). Working models of attachment: Implications forexplanation, emotion, and behavior. Journal of Personality and SocialPsychology, 71, 810-832.

Collins, N. L., & Read, S. J. (1990). Adult attachment, working models,and relationship quality in dating couples. Journal of Personality andSocial Psychology, 58, 644-663.

Collins, N. L., & Read, S. J. (1994). Cognitive representations of attach-ment: The structure and function of working models. In K. Bartholomew& D. Perlman (Eds.), Attachment processes in adulthood (pp. 53-92).London: Jessica Kingsley.

Cosmides, L., & Tooby, J. (1987). From evolution to behavior: Evolution-ary psychology as the missing link. In J. Dupre (Ed.), The latest on thebest: Essays on evolution and optimality (pp. 277-306). Cambridge,MA: MIT Press.

Dozier, M., & Kobak, R. R. (1992). Psychophysiology in attachmentinterviews: Converging evidence for deactivating strategies. Child De-velopment, 63, 1473-1480.

Feeney, B. C , & Kirkpatrick, L. A. (1996). Effects of adult attachment andpresence of romantic partners on physiological responses to stress.Journal of Personality and Social Psychology, 70, 255—270.

Feeney, J. A., & Noller, P. (1990). Attachment style as a predictor of adultromantic relationships. Journal of Personality and Social Psychol-ogy, 58, 281-291.

Fischler, I. S., & Bloom, P. A. (1979). Automatic and attentional processesin the effects of sentence contexts on word recognition. Journal ofVerbal Learning and Verbal Behavior, 18, 1—20.

Folkman, S., & Lazarus, R. S. (1985). If it changes it must be a process:Study of emotion and coping during three stages of a college examina-tion. Journal of Personality and Social Psychology, 48, 150-170.

Forster, K. I. (1981). Priming and the effects of sentence and lexicalcontexts on naming time: Evidence for autonomous lexical processing.Quarterly Journal of Experimental Psychology, 33, 465-495.

Fraley, R. C , Davis, K. E., & Shaver, P. R. (1998). Dismissing avoidanceand the defensive organization of emotion, cognition, and behavior. InJ. A. Simpson & W. S. Rholes (Eds.), Attachment theory and closerelationships (pp. 201-232). New York: Guilford.

Fraley, R. C , & Shaver, P. R. (1997). Adult attachment and the suppres-sion of unwanted thoughts. Journal of Personality and Social Psychol-ogy, 73, 1080-1091.

Fraley, R. C , & Shaver, P. R. (1998). Airport separations: A naturalisticstudy of adult attachment dynamics in separating couples. Journal ofPersonality and Social Psychology, 75, 1198-1212.

Fraley, R. C , & Waller, N. G. (1998). Adult attachment patterns: A test oftypological model. In J. A. Simpson & W. S. Rholes (Eds.), Attachmenttheory and close relationships (pp. 77-114). New York: Guilford.

Gilbert, D. T., & Hixon, J. G. (1991). The trouble of thinking: Activationand application of stereotypic beliefs. Journal of Personality and SocialPsychology, 60, 509-517.

Gilbert, D. T., & Osborne, R. E. (1989). Thinking backward: Some curableand incurable consequences of cognitive busyness. Journal of Person-ality and Social Psychology, 57, 940-949.

Greenberg, J., Pyszczynski, T., & Solomon, S. (1997). Terror managementtheory of self-esteem and cultural worldviews: Empirical assessmentsand conceptual refinements. In P. M. Zanna (Ed.), Advances in experi-

mental social psychology, (Vol. 29, pp. 61-141). San Diego: AcademicPress.

Hazan, C , & Shaver, P. R. (1987). Romantic love conceptualized as anattachment process. Journal of Personality and Social Psychology, 52,511-524.

Kalish, R. A. (1985). Death, grief, and caring relationships. New York:Cole.

Kirkpatrick, L. A., & Shaver, P. R. (1988). Fear and affiliation reconsid-ered from a stress and coping perspective. Journal of Social and ClinicalPsychology, 7, 214-233.

Kobak, R. R., Cole, H. E., Ferenz, G. R., & Fleming, W. S. (1993).Attachment and emotion regulation during mother-teen problem solv-ing: A control theory analysis. Child Development, 64, 231-245.

Kobak, R. R., & Duemmler, S. (1994). Attachment and conversation:Toward a discourse analysis of adolescent and adult security. In K.Bartholomew & D. Perlman (Eds.), Attachment processes in adulthood(pp. 121—149). London: Jessica Kingsley.

Lamb, M. E. (1976). Twelve-month-olds and their parents: Interaction in alaboratory playroom. Developmental Psychology, 12, 237-244.

Lazarus, R. S., & Folkman, S. (1984). Stress, appraisal and coping. NewYork: Springer.

Mathews, A., & MacLeod, C. (1994). Cognitive approaches to emotion andemotional disorders. Annual Review of Psychology, 45, 25-50.

McCarthy, J. (1981). Death and anxiety. New York: Academic Press.Meyer, D., & Schvaneveldt, R. W. (1971). Facilitation in recognizing pairs

of words: Evidence of a dependence between retrieval operations. Jour-nal of Experimental Psychology, 90, 227-234.

Mikulincer, M. (1998a). Adult attachment style and affect regulation:Strategic variations in self-appraisals. Journal of Personality and SocialPsychology, 75, 420-435.

Mikulincer, M. (1998b). Attachment working models and the sense oftrust: An exploration of interaction goals and affect regulation. Journalof Personality and Social Psychology, 74, 1209-1224.

Mikulincer, M., & Florian, V. (1995). Appraisal and coping with a real-lifestressful situation: The contribution of attachment styles. Personalityand Social Psychology Bulletin, 21, 408-416.

Mikulincer, M., & Florian, V. (1996). Coping and adaptation to trauma andloss. In M. Zeidner & N. Endler (Eds.), Handbook of coping (pp.554-572). New York: Wiley.

Mikulincer, M., & Florian, V. (1998). The relationship between adultattachment styles and emotional and cognitive reactions to stressfulevents. In J. A. Simpson & W. S. Rholes (Eds.), Attachment theory andclose relationships (pp. 143-165). New York: Guilford.

Mikulincer, M., Florian, V., & Tolmacz, R. (1990). Attachment styles andfear of personal death: A case study of affect regulation. Journal ofPersonality and Social Psychology, 58, 273-280.

Mikulincer, M., Florian, V., & Weller, A. (1993). Attachment strategies,and posttraumatic psychological distress: The impact of the Gulf war inIsrael. Journal of Personality and Social Psychology, 64, 817-826.

Mikulincer, M., Horesh, N., Eilati, I., & Kotler, M. (1999). The associationbetween adult attachment style and mental health in extreme life endan-gering conditions. Personality and Individual Differences, 27, 831-842.

Mikulincer, M., & Orbach, I. (1995). Attachment styles and repressivedefensiveness: The accessibility and architecture of affective memories.Journal of Personality and Social Psychology, 68, 917-925.

Mikulincer, M., Orbach, I., & Iavnieli, D. (1998). Adult attachment styleand affect regulation: Strategic variations in subjective self-other sim-ilarity. Journal of Personality and Social Psychology, 75, 436-448.

Murphy, S. T., & Zajonc, R. B. (1993). Affect, cognition, and awareness:Affective priming with optimal and suboptimal stimulus exposures.Journal of Personality and Social Psychology, 64, 723-739.

Nadler, A. (1991). Help-seeking behavior: Psychological costs and instru-mental benefits. In M. S. Clark (Ed.), Review of personality and social

STRESS AND ATTACHMENT 523

psychology: Prosocial behavior (Vol. 12, pp. 290-311). Newbury Park,CA: Sage.

Pollack, J. M. (1979). Correlates of death anxiety: A review of empiricalstudies. Omega, 10, 97-121.

Rholes, W. S., Simpson, J. A., & Grich-Stevens, J. (1998). Attachmentorientations, social support, and conflict resolution in close relation-ships. In J. A. Simpson & W. S. Rholes (Eds.), Attachment theory andclose relationships (pp. 166-188). New York: Guilford.

Rofe, Y. (1984). Stress and affiliation: A utility theory. PsychologicalReview, 91, 235-250.

Schachter, S. (1959). The psychology of affiliation. Stanford, CA: StanfordUniversity Press.

Shaver, P. R., Collins, N. L., & Clark, C. L. (1996). Attachment styles andinternal working models of self and relationship partners. In G. Fletcher& J. Fitness (Eds.), Knowledge structures in close relationships: A socialpsychological approach. Hillsdale, NJ: Erlbaum.

Shaver, P. R., & Hazan, C. (1993). Adult romantic attachment: Theory andevidence. In D. Perlman & W. Jones (Eds.), Advances in personalrelationships (pp. 29-70). London: Jessica Kingsley.

Shaver, P. R., & Klinnert, M. (1982). Schachter's theories of affiliation andemotions: Implications of developmental research. In L. Wheeler (Ed.),Review of personality and social psychology (Vol. 3, pp. 37-71). Bev-erly Hills, CA: Sage.

Sherman, S. J., Mackie, D. M., & Driscoll, D. M. (1990). Priming and the

differential use of dimensions in evaluation. Personality and SocialPsychology Bulletin, 16, 405-418.

Simpson, J. A., Rholes, W. S., & Nelligan, J. S. (1992). Support seekingand support giving within couples in an anxiety-provoking situation: Therole of attachment styles. Journal of Personality and Social Psychol-ogy, 62, 434-446.

Singer, J. A., & Salovey, P. (1988). Mood and memory: Evaluating thenetwork theory of affect. Clinical Psychology Review, 8, 211-251.

Stanovich, K. E., & West, R. F. (1983). On priming by a sentence context.Journal of Experimental Psychology: General, 112, 1-36.

Ucros, C. G. (1989). Mood state-dependent memory: A meta-analysis.Cognition and Emotion, 3, 139-167.

Wegner, D. M. (1994). Ironic processes of mental control. PsychologicalReview, 101, 34-52.

Wegner, D. M., Erber, R., & Zanakos, S. (1993). Ironic processes in themental control of mood and mood-related thought. Journal of Person-ality and Social Psychology, 63, 903-912.

Wegner, D. M., & Smart, L. (1997). Deep cognitive activation: A newapproach to the unconscious. Journal of Consulting and Clinical Psy-chology, 65, 984-995.

Received January 19, 1999Revision received June 24, 1999

Accepted July 12, 1999 •