Inoculating against the aversive effects of ostracism with acceptance: The role of attachment styles

Abstract

Ostracism's immediate impact is painful. We investigated whether affirming an accepting relationship prior to being ostracized can inoculate against this pain, and whether this approach would only be effective for those with a secure attachment style. Participants completed an attachment style measure and wrote about an unconditionally accepting or neutral relationship prior to being either ostracized or included in Cyberball, a virtual ball-toss game. Unconditional acceptance buffered the basic needs and the mood of participants with a secure attachment style, but provided no relief for those with an insecure style. Findings suggest that reminders of acceptance prior to social exclusion can be beneficial, but only for some.

Keywords::

• Ostracism
• Attachment styles
• Social pain
• Social exclusion

Ostracism, being excluded and ignored, is a ubiquitous and potent experience. The social pain produced by ostracism is immediate, strong, and registers in areas of the brain closely associated with physical pain (Eisenberger, Lieberman, & Williams, 2003). Moreover, people feel such pain under conditions that should logically preclude such reactions. For example, people experience declines in psychological well-being despite earning money for being excluded (van Beest & Williams, 2006), when they believe they are being excluded by people they despise (i.e., members of the Klu Klux Klan; Gonsalkorale & Williams, 2007), and even when they know that they are being excluded by computer-programmed characters (Zadro, Williams, & Richardson, 2004). Furthermore, previous research demonstrates that personality and individual differences, such as attachment style, loneliness, self-esteem, social anxiety, and introversion that moderate behavior in numerous social situations, show little or no relation to the immediate pain of being ostracized (Williams, 2009).

Recent research using time-sensitive measures and examining a constellation of personality traits, however, has found that personality factors can affect responses during ostracism. For example, Wesselmann, Wirth, Mroczek, and Williams (2012) measured participants' affect on a second-to-second basis during Cyberball, an online ball-toss game, and found that those high in trait loneliness experienced a steeper drop in affect after being ostracized. Wirth, Lynam, and Williams (2010) also found that participants possessing a unique combination of personality traits, specifically those descriptive of Cluster A personality disorders (i.e., paranoid, schizoid, schizotypal), associated with a strong distrust of and uneasiness with interpersonal relationships, were less likely to feel need depletion and negative affect in response to ostracism via Cyberball. Taken together, these studies suggest that specific combinations of factors can inoculate or buffer individuals from the normally potent and aversive consequences of ostracism.

When examining recovery from ostracism, individual differences appear to have a larger role. For example, the negative effects of a brief ostracism episode on basic need satisfaction were greater among those high in social anxiety than those low, even 45 min later (Zadro, Boland, & Richardson, 2006). Similarly, Oaten, Williams, Jones, and Zadro (2008) found that social anxiety moderated the impact of ostracism on individuals' ability to self-regulate when working toward nonsocial goals. Socially anxious participants were less likely to drink a healthy but unpleasant beverage after ostracism, an effect that persisted for 45 min. These findings suggest that the impact of individual differences is likely to depend on contextual factors (e.g., the passage of time), and that attempts to assuage or prevent ostracism's pain should take into account both types of factors.

Prior research, however, has yet to discover ways to intervene before ostracism to prevent its negative effects. As a whole, research demonstrating the positive impact of reminders of relationship partners during physical threats (e.g., Master et al., 2009) and after social threats (e.g., Tai, Zheng, & Narayanan, 2011) suggests that such reminders prior to being ostracized may have inoculation-type effects, but it seems likely that individual differences are likely to moderate their impact. In particular, we hypothesize that such reminders should be effective only for those comfortable relying on others for support. Thus, we investigated the interaction between attachment styles and a social acceptance prime administered before ostracism on basic need satisfaction, measured immediately following ostracism.

The Protective Effect of Acceptance

A number of studies suggest that reminding an individual of a close relationship can provide protection during a threatening experience. For instance, using a terror management theory framework, Gillath and Hart (2010) found that writing about a secure attachment (i.e., someone close, dependable, loving) after being reminded of their mortality reduced the impact of mortality salience; these participants were less likely to react defensively when reporting attitudes about foreign policy (i.e., less harsh stance toward North Korea). Likewise, a variety of manipulations designed to establish social connections (e.g., writing about a friend/family member, hugging a teddy bear) after being ostracized can help people cope with its negative effects (e.g., Tai et al., 2011). A recent study of physical pain tolerance by Master et al. (2009) found that women who focused on a picture of their romantic partner reported less physical pain during repeated exposures to a heated probe. Given the overlap between social and physical pain in the frontal cortex (Eisenberger, 2011), it seems likely that such reminders could also serve to prevent social pain.

While there is ample evidence that reminding individuals of romantic relationships and other types of strong attachments can have beneficial effects (Mikulincer & Shaver, 2007a), there is also evidence that these types of relationships can engender considerable ambivalence for many (e.g., Murray, Griffin, Rose, & Bellavia, 2006), making them a potentially less effective means of support prior to an ostracism episode. Therefore, we had participants recall a consistently accepting relationship in order to reduce the likelihood that participants would recall an ambivalent one, and because a highly accepting relationship is the most stable and uniformly favorable interpersonal connection people can experience, symbolizing the antithesis of ostracism. Moreover, studies in which participants are asked to visualize someone who is accepting and nonevaluative find a reduced association between rejection-related words and failure (Baldwin & Sinclair, 1996). In addition, a rich research tradition suggests the value of unconditional positive regard. Since Rogers' (1947) seminal research, many have elaborated upon the benefits of noncontingent acceptance for self-esteem, motivation, and mental health (e.g., Deci & Ryan, 1995; Kernis, 2003). To our knowledge, however, no study to date has investigated whether feelings of acceptance can impact responses to a real-time social interaction.

The Role of Attachment Styles

Although little evidence suggests that adult attachment styles moderate immediate reactions to ostracism (Williams, 2009), considerable research has demonstrated that individuals with secure and insecure attachment styles differ in how they seek and utilize social support in stressful situations (Collins, Ford, & Feeney, 2011; Mikulincer & Shaver, 2007b). Attachment theory maintains that adult behavior is influenced by a chronic set of expectancies and motives regarding close interpersonal relationships that have been shaped both by caregiver behavior in infancy and childhood and by interactions with close relationship partners in adulthood. In Bartholomew and Horowitz's (1991) classic formulation, attachment styles can be broadly described either as secure, characterized by the belief that the self is worthy of love and that others are trustworthy and available for support, or as insecure, possessing the beliefs that the self is not worthy of love (preoccupied) or that others are not trustworthy (dismissing), or a combination of both beliefs (fearful). Studies find that individuals with secure adult attachment styles are more likely to seek and use social support when facing both social threats (e.g., divorce; Birnbaum, Orr, Mikulincer, & Florian, 1997) and nonsocial threats (e.g., chronic pain; Mikulincer & Florian, 1995). Likewise, those with secure attachment styles are more likely to seek emotional support from a romantic partner when feeling anxious about an upcoming speech (Collins, Ford, Guichard, Kane, & Feeney, 2010). Moreover, securely attached adults are more likely to perceive an ambiguous response from their partner to be supportive (Collins & Feeney, 2004) and more likely to believe that social support is available (Ognibene & Collins, 1998).

These findings are consistent with the notion that securely attached individuals assume that significant others provide comfort and relief during stressful and painful experiences (Hazan & Shaver, 1987, 1990; Mikulincer, Hirschberger, Nachmias, & Gillath, 2001). Thus, reflecting upon an unconditionally accepting relationship prior to an ostracism experience may be more likely to provide comfort to those with a secure attachment style because they are more likely to believe that such relationships would be valuable and are more likely to turn to them during difficult times.

Current Study

We investigated the joint effect of primed feelings of unconditional acceptance and attachment style on participants' responses immediately after being ostracized. Unlike previous studies, we focused on the effects of reminding participants of a personal relationship prior to ostracism to examine whether it could effectively protect the basic needs of an individual who is experiencing ostracism. Because ostracism is painful to nearly everyone, and because participants with a secure attachment style are more likely to use social support in anxiety-producing situations, we predicted that when they write about an accepting relationship prior to being ostracized, they would report less social pain. By the same logic, we predicted that those with an insecure attachment style would experience no benefit from writing about an accepting relationship.

Method

Participants and design

A total of 211 university students (165 women; M_age = 18.8 years, SD = 0.9; 86% Caucasian) participated for extra credit. The study employed a 2 (attachment style: secure vs. insecure) ×  2 (prime: neutral vs. acceptance) ×  2 (exclusionary status: included vs. ostracized) design, but data were collected in two waves. The initial study design omitted the acceptance-inclusion condition and was based on a small sample size (N = 84). During the second data collection wave, we randomly assigned twice as many participants to this particular condition as the other three conditions. This procedure preserved random assignment to conditions and produced groups of similar sizes, while incorporating the initially omitted acceptance-inclusion condition.

Procedure

As part of a larger questionnaire,1 participants completed the Bartholomew and Horowitz (1991) four-paragraph measure of attachment style and were post-experimentally categorized as having either a secure or insecure (i.e., dismissing, preoccupied, or fearful) attachment style based on the paragraph each participant selected as self-descriptive. Past research has indicated that assessing attachment style in this fashion is valid and congruent with other measures (e.g., Brennan, Shaver, & Tobey, 1991; Sprecher & Fehr, 2011) and the percentage of participants identifying with the secure paragraph (37.4%) was similar to recent large-scale research using the same assessment (37.7%; Sprecher, 2013). Next, participants wrote either an essay used in previous research to prime feelings of unconditional acceptance or a control essay (Arndt, Schimel, Greenberg, & Pyszczynski, 2002; Baldwin & Sinclair, 1996); they wrote briefly either about a person who very clearly accepts them or about a coworker or classmate with whom they did not know well and had little interaction.

Afterward, participants played a virtual ball-toss game, Cyberball, ostensibly with two other players (Williams, Cheung, & Choi, 2000). The other players were computer-controlled confederates pre-programmed to either include or ostracize the participant. Included participants received the ball approximately one-third of the time throughout three minutes of play and approximately 30 ball tosses. Ostracized participants received the ball twice at the beginning and then never again.

Dependent measures

Immediately following Cyberball, participants completed dependent measures used in previous ostracism studies (e.g., Wirth & Williams, 2009). Participants rated 12 items designed to assess the four basic psychological needs of belonging (e.g., “disconnected”; α = .89), self-esteem (e.g., “good about myself”; α = .82), control (e.g., “powerful”; α = .71), and meaningful existence (e.g., “non-existent”; α = .88). On each item, participants rated the degree to which they felt the statement “right now” on a five-point scale (1 =  not at all; 5 =  extremely). The 12 items were averaged to form an index of basic needs satisfaction (α = .93; higher values indicate greater satisfaction). Participants also rated their mood on 12 items using the same scale. Three items each assessed happiness (α = .90), sadness (α = .85), anger (α = .89), anxiety (α = .84), and hurt (α = .85). Happiness items were reverse-coded and all were averaged to create an index of negative mood (α = .94). Participants then rated on two separate items how ignored and excluded they felt during the game (r = .93), which were averaged together to create an index of perceptions of exclusion. As a manipulation check, participants also estimated the percent of throws they received.

Participants in the second wave of data collection (32 insecure and 14 secure in the neutral condition, and 56 insecure and 26 secure in the acceptance condition) provided information about the relationship described in their essay. Participants indicated how many years they had known the person (1 =  less than 1 year, 2 = 1–3 years, 3 = 4–6 years, 4 = 7–9 years, 5 = 10 or more years) and whether the person is currently in their life or was from the past. Participants also indicated whether the person was a friend, family member, a significant other, or other (e.g., acquaintance, coworker). They indicated how difficult it was for them to choose a person to write about (1 =  not at all; 5 =  extremely). Finally, in two separate ratings, participants rated the degree to which the person was accepting and judgmental (1 =  not at all; 5 =  extremely).

Results

In order to determine whether wave of data collection impacted our results, we performed an analysis of variance (ANOVA) including wave of data collection as a variable in our original 2 (secure or insecure) ×  3 (neutral-inclusion, neutral-ostracism, or acceptance-ostracism) analyses, and we found no significant main effect (ps > .78) of collection wave or interactions with collection wave for any dependent measures. We concluded that the data collection wave had no effect and dropped it from subsequent analyses. In addition, chi-square analyses indicated that the condition added in the second wave (acceptance-inclusion) yielded a similar proportion of secure and insecure participants (χ²(1) = 1.36, p > .20) and males and females (χ²(1) = .56, p > .40) to the rest of the sample.

All further analyses utilized a 2 (attachment style: secure vs. insecure) ×  2 (prime: acceptance vs. neutral) ×  2 (exclusionary condition: ostracized vs. included) ANOVA approach unless otherwise specified.

Manipulation check

Analyses on participants' estimates of percent of throws they received indicated only the typical main effect for exclusionary condition, F(1,203) = 287.1, p <  .001, η_p² = .59. Consistent with the Cyberball manipulation, participants estimated receiving a greater percent of throws in the inclusion conditions (M = 30.54, SD = 12.80), than the ostracism conditions (M = 7.27, SD = 4.89). There were no significant effects of attachment style, prime, or interactions between any variables, Fs <  1.6, ps > .20.

Dependent measures

Basic needs

Consistent with previous research, analyses indicated that ostracized participants felt less basic need satisfaction (M = 2.44, SD = 0.75) than included participants (M = 3.74, SD = 0.68); F(1,203) = 175.5, p <  .001, η_p² = .46. Also, consistent with prior research on attachment style and subjective well-being (for a review, see Mikulincer & Shaver, 2007b), insecurely attached participants reported lower need satisfaction (M = 2.99, SD = 0.94) than securely attached participants overall (M = 3.31, SD = 0.98); F(1,203) = 8.21, p <  .01, η_p² = .04. No main effect for prime was observed, and no two-way interactions were statistically significant. Pertinent to our hypothesis, however, the two main effects were qualified by an attachment style ×  prime interaction ×  exclusionary condition; F(1,203) = 4.38, p = .04, η_p² = .02. Analyses of the individual needs showed a similar pattern as the combined needs index.

The pattern of means was consistent with our predictions that the acceptance prime would buffer the effects of ostracism only for securely attached participants (see Table 1). In particular, pairwise comparisons indicated that among securely attached participants who were ostracized, those who wrote about an accepting relationship beforehand reported greater need satisfaction (M = 2.81, SD = 0.77) than those who wrote about a neutral relationship (M = 2.31, SD = 0.79; F(1,203) = 4.81, p = .03, η_p² = .02). Meanwhile, among insecurely attached participants who were ostracized, writing an essay about an accepting relationship had no impact (p > .35). For included participants, the type of prime did not have an effect on securely (p > .50) or insecurely (p > .70) attached participants, indicating the attachment style and accepting prime effects were present only for ostracized participants. Also consistent with our hypotheses, securely attached participants who had written an acceptance essay prior to ostracism reported greater need satisfaction (M = 2.81, SD = 0.77) than insecurely attached participants in the same condition (M = 2.28, SD = 0.64; F(1,203) = 7.00, p <  .01, η_p² = .03).

Table 1 Basic needs and positive mood immediately after playing Cyberball as a function of exclusionary condition and attachment style

	Neutral prime, ostracism	Acceptance prime, ostracism	Neutral prime, inclusion	Acceptance prime, inclusion
Secure attachment style	n = 18	n = 21	n = 23	n = 17
Needs	2.31 (0.79)	2.81 (0.77)	4.03 (0.60)	3.91 (0.48)
Negative mood	2.76 (0.56)	2.28 (0.66)	1.50 (0.34)	1.64 (0.37)
Insecure attachment style	n = 34	n = 31	n = 29	n = 38
Needs	2.43 (0.76)	2.28 (0.64)	3.56 (0.77)	3.63 (0.67)
Negative mood	2.70 (0.73)	2.86 (0.71)	1.79 (0.72)	1.81 (0.63)

Notes: All values are given as mean (SD). Higher scores reflect more need fulfillment of basic needs and increased negative mood.

Although not specifically addressed by hypotheses, additional post hoc analyses on basic needs also indicated that all comparisons between the inclusion and ostracism conditions across the corresponding levels of attachment style and prime were statistically significant (all ps <  .001). In addition, secure participants who wrote the neutral essay and were included reported higher need satisfaction (M = 2.81, SD = 0.77) than insecure participants in the same condition (M = 2.28, SD = 0.64; F(1,203) = 5.78, p = .02, η_p² = .03). No other comparisons were statistically significant.

Analyses on basic needs using the four-level attachment style measure yielded a similar pattern, including the same main effect for inclusionary status and a similar attachment style effect, F(3,195) = 3.15, p = .03, η_p² = .05 (all interactions ps > .20). Post hoc comparisons indicated that preoccupied participants had a tendency to report lower need satisfaction than secure, fearful, and dismissing participants (ps = .004, .07, .14, respectively), but that no other groups differed. Furthermore, pairwise comparisons of the effect of prime for each of the attachment groups in the ostracism condition still indicated an effect only among secures (F(1,195) = 4.76, p = .03, η_p² = .02), but no other groups (ps > .35). No significant effects emerged from the same comparisons in the inclusion condition (ps > .50).

Negative mood

A highly similar pattern as basic needs occurred for mood; analyses yielding the same main effect for inclusionary condition, F(1,203) = 113.5, p <  .001, η_p² = .36, main effect for attachment style, F(1,203) = 7.58, p <  .01, η_p² = .04, and an attachment style ×  prime ×  exclusionary condition interaction, F(1,203) = 4.24, p = .04, η_p² = .02 (see Table 1). Pairwise comparisons indicated that among securely attached participants who were ostracized, those who wrote about an accepting relationship beforehand reported less negative mood (M = 2.28, SD = 0.66) than those who wrote about a neutral relationship (M = 2.76, SD = 0.56; F(1,203) = 5.49, p = .02, η_p² = .03). Meanwhile, among insecurely attached participants who were ostracized, the type of essay had no impact (p > .30). Nor did the type of essay have an impact in the inclusion conditions for securely (p > .50) or insecurely (p > .90) attached participants. Also consistent with the basic need analyses, securely attached participants report less negative mood after writing the acceptance essay prior to ostracism (M = 2.28, SD = 0.66) than insecurely attached participants (M = 2.86, SD = 0.72; F(1,203) = 10.54, p <  .005, η_p² = .05).

Additional post hoc analyses indicated that all comparisons between the inclusion and ostracism conditions across the corresponding levels of attachment style and prime were also statistically significant (all ps <  .001). No other comparisons were statistically significant. Analyses of mood using the four-group attachment style measure also yielded a pattern highly similar to that of basic needs, but are not included for the sake of brevity.

Perceptions of exclusion

Analyses on participants' perceptions of being excluded yielded only a main effect for exclusionary condition, F(1,203) = 449.2, p <  .001, η_p² = .69, such that participants in the ostracism condition reported feeling more excluded (M = 4.40, SD = 0.86) than those in the inclusion condition (M = 1.65, SD = 0.97). No interaction effects were statistically significant (ps ranged from .11 to .59).

Analysis of essay content: ruling out alternative explanations

Although the current findings indicate that feelings of acceptance provided a buffer for participants with a secure attachment style and had little effect on those with an insecure style, an alternative explanation is possible. It may be that the prime had a differential impact because insecure participants wrote essays about relationships that were less accepting or differed in some other way. In order to rule out, as much as possible, qualitative differences between secure and insecure participants in the acceptance prime essays, we analyzed the information that participants' provided about their essay targets.

In order to examine whether insecurely and securely attached participants wrote about target persons who were equally accepting and/or judgmental, we subjected participants' ratings of their essay target to a 2 (attachment style: secure vs. insecure) ×  2 (prime: acceptance vs. neutral) ANOVA. Analyses yielded similar patterns on each rating, so the judgmental ratings were reverse-scored and the two were averaged to create an index of acceptance. Analyses revealed only the expected main effect for prime, F(1,123) = 52.7, p <  .001, η_p² = .30; participants writing an acceptance essay rated the target as more accepting (M = 4.57, SD = 0.53) than those writing a neutral essay (M = 3.60, SD = 0.90). The lack of an effect for attachment style or an interaction (ps > .50) suggests that securely and insecurely attached participants brought to mind relationships that were similar on the key construct that our prime was designed to induce.

Similarly, analyses on the difficulty with which participants generated a target person for their essay revealed only a main effect for prime, F(1,123) = 5.33, p = .02, η_p² = .04; participants writing an acceptance essay reported more difficulty identifying a target for their essay (M = 4.44, SD = 1.08) than those writing a neutral essay (M = 3.96, SD = 1.03), and no effect for attachment style or the interaction (ps > .60). Thus, it appears that it was more difficult for participants to generate an example of accepting person than an acquaintance, but that it was equally difficult for secure and insecure participants.

Analyses of how long participants had known the subject of their essay yielded a main effect for prime, F(1,123) = 20.8, p <  .001, η_p² = .15, and attachment style, F(1,123) = 4.53, p = .04, η_p² = .04, but no interaction (p > .25). Participants writing acceptance essays wrote about people whom they had known longer (M = 3.20, SD = 1.51) than those writing neutral essays (M = 1.85, SD = 1.01), and insecurely attached participants wrote about people they had known longer (M = 2.91, SD = 1.54) than securely attached participants (M = 2.26, SD = 1.43). This latter finding is noteworthy given that secure participants were the ones that benefited from the acceptance prime; even though their acceptance essays were about less established relationships, they provided a protective effect.

Chi-square analyses indicated that securely and insecurely attached participants, however, wrote about people currently in their life at the same rate in both the neutral (χ²(1) = .97, p > .60) and acceptance essays (χ²(1) = .59, p > .50). Consistent with the analyses on how long participants had known the target of the essay, comparisons of the prime conditions indicated that participants were more likely to write about someone currently in their life when writing about an accepting relationship (98.3%) than a neutral one (78.3%; χ²(1) = 6.83, p <  .01).

Chi-square analyses of the types of relationships participants wrote about indicated securely and insecurely attached participants identified the person they wrote about in their neutral essays in similar ways (χ²(2) = 3.81, p > .14) with 65% (n = 30) writing about a friend, 33% (n = 15) writing about an “other,” and 2% (n = 1) writing about a family member. Securely and insecurely attached participants did, however, differ in how they identified the person in their acceptance essays (χ²(2) = 9.70, p = .02). Securely attached participants were more likely to report that they wrote about a friend (44%, n = 11), significant other (36%, n = 9), and “other” (8%, n = 2) than insecurely attached participants (36%, n = 20; 25%, n = 14; 0%, N = 0, respectively), while insecurely attached participants were more likely to write about a family member (39%, n = 22) than securely attached participants (12%, n = 3).

Overall, analyses uncovered little that suggests that securely and insecurely attached participants wrote qualitatively different acceptance essays, and little that could explain the differential impact of the acceptance essay on the two groups. In fact, the only observed difference indicated that insecure participants wrote about accepting individuals with whom they had a more stable relationship.

Discussion

We investigated whether a reminder of an unconditionally accepting relationship prior to being ostracized would reduce its aversive consequences, and hypothesized that this inoculation-like effect would be moderated by participants' attachment style. In fact, we found that reflecting on an accepting relationship reduced the social pain (i.e., increased basic needs and mood) reported by those with a secure attachment style, but did not among those with an insecure style. This may explain why previous studies have not found that attachment styles were related to reactions to ostracism, namely that they only have an impact in certain contexts. Given these findings, it seems likely that other individual differences may also influence how ostracism is experienced, but the contexts in which they exert their influence have yet to be identified.

Our findings are particularly noteworthy given that prior research has indicated that individuals with an insecure attachment style are less likely to perceive that social support is available (Ognibene & Collins, 1998). In the current study, such “support” was available for all primed participants; participants wrote about an equally accepting person regardless of their attachment style. In fact, even though insecure participants were more likely to write about an accepting family member, who presumably represents a more stable relationship, they did not reap the benefits their secure counterparts did. Insecure participants may have a difficult time using support when it is present. The current data do not address why, but clearly show that such reminders can have immediate protective effect for those with a secure attachment.

Implications for experiencing ostracism

Our findings are also broadly consistent with the position that the experience of social pain is an evolutionary adaptation experienced by nearly everyone (Kerr & Levine, 2008; Williams, 2009). Even though our prime reduced the negative effects of being ostracized for securely attached participants, this group still felt worse than their included counterparts. However, our current findings join others in demonstrating that there is some flexibility in how ostracism is experienced (e.g., Wirth et al., 2010). We show that it is possible for some people to shield themselves from some of the social pain by reflecting on prior experiences of acceptance. Such reminders might change perceptions of the ostracism itself, but our analyses of participants' perceptions of exclusion yielded nothing to suggest this. Given that the prime worked only for securely attached participants, it seems more likely that reminders of acceptance can help people view rejection in the larger context of their existing supportive social relationships. For those who feel secure in those relationships, it makes it less painful, while for those who do not, it offers no comfort.

Furthermore, our findings are the first demonstration of an “inoculation” effect in a real-time experience of ostracism. Previous research had shown that reminders of a romantic partner could lessen physical pain (e.g., Master et al., 2009), but the current study shows that social pain can also be alleviated for some by reminders of an accepting relationship. Likewise, many studies have shown that relationship reminders after the fact can lessen ostracism's impact, but our findings suggest that they can prevent pain for some. Although a conservative interpretation of our data is that our priming manipulation facilitated immediate coping with the effects of ostracism, it may have also served to buffer the immediate pain. Additional studies can shed light on the timing of these effects and whether there are interventions that might be effective regardless of attachment style.

Future directions

Our findings also offer new perspective on research investigating the effects of priming feelings of security (for reviews, see Gillath, Selcuk, & Shaver, 2008; Mikulincer & Shaver, 2007a). Experiments using security-related words, pictures of attachment figures, recall of positive attachment experiences, and other similar prompts have found that such priming has a positive impact on all individuals, regardless of attachment style. For example, Mikulincer and Shaver (2001) found that when participants were prompted to visualize an attachment figure, “ a person who accepts and loves you and helps you in times of need” (p. 105), all participants, regardless of attachment style, reported less anxiety and more favorable attitudes toward a commonly derogated ethnic out-group. Our prime, however, only had a positive impact on those with a secure attachment style, suggesting that attachment security and acceptance are not equivalent constructs. While some of our participants wrote about important attachment figures who provided long-standing support, many wrote about friends, old and new, who were simply very accepting but may not represent “true” attachment figures. It may be that accepting relationships may provide a valuable sense of social inclusion, but do not afford the involved, loving support that attachment relationships do. Our findings suggest that insecure individuals may have difficulty capitalizing on relationships that are merely accepting, and point to the need for future research to more clearly delineate when a warm and supportive adult relationship represents an attachment and when it does not.

Furthermore, this study suggests that priming participants with an accepting relationship can affect how they respond to a real-time social interaction, something that neither security priming nor unconditional positive regard studies have done. Many studies on adult attachment styles have examined their influence on in vivo interactions (e.g., Fraley & Shaver, 1998) but these have largely focused on interactions within couples (see Mikulincer & Shaver, 2007b). More recently, researchers investigated the degree to which the affective and behavior dynamics associated with attachment style generalize to nonattachment relationships (e.g., Sadikaj, Moskowitz, & Zuroff, 2011). Our study also contributes to this growing body of research demonstrating that attachments styles can also influence reactions to the behavior of strangers.

Future studies may benefit from assessing attachment anxiety and attachment avoidance independently. Although our analyses did not indicate differences among the three insecure attachment subtypes, a more powerful regression-based approach using individual difference measures (e.g., Fraley, Waller, & Brennan, 2000) may shed light on the degree to which anxiety and avoidance may be independently associated with our effects. For example, those high in avoidance may be relatively indifferent to acceptance and it might have little impact. Then again, they may value people who accept their avoidant behavior without criticism, so it might be rather palliative. Likewise, acceptance may soothe those high in anxiety, or conversely, they may have difficulty recalling a convincing example of acceptance and respond with more anxiety (cf. Hermann, Leonardelli, & Arkin, 2002).

In summary, the current study provides preliminary evidence that it is possible to preemptively limit the aversive consequences associated with being ostracized, at least for some individuals. In addition, we offer new perspective on the conditions under which attachment styles are related to reactions to ostracism. Our findings emphasize the value of examining the joint influence of individual differences and situational factors in understanding how people react to and cope with ostracism.

Acknowledgements

We thank Kristen Burkart and Anna Murr for their assistance with conducting this research and, Jean Clore, Stephenie Chaudior, Lane Beckes, and David Schmitt for their comments on earlier drafts.

Notes

¹ Also included were measures of self-esteem (Rosenberg, 1965) and approach and avoidance motivation (Carver & White, 1994). Analyses showed that these did not interact with the experimental conditions on the dependent measures, so they were dropped. This does, however, suggest that any attachment effects cannot be explained by these constructs.