Self-Resemblance and Social Rejection

Abstract

Humans perceive and treat self-resembling others in ways that suggest that self-resemblance is a cue of kinship. However, we know little about how individuals respond to treatment by self-resembling others. Here we approach this problem by connecting facial self-resemblance to social rejection. Given that individuals should expect to cooperate with kin, we hypothesized that (1) social inclusion by perceived kin should elicit lesser feelings of rejection and (2) social exclusion by perceived kin should elicit greater feelings of rejection relative to inclusion or exclusion, respectively, by nonkin. To test these hypotheses, we recruited 90 participants to play two games of Cyberball, a virtual ball-tossing game, with separate pairs of ostensible partners. In one game, the ostensible partners were programed to fully include the participants in group play and, in the other game, they were programed to exclude the participants after a few rounds; the order of inclusion and exclusion was counterbalanced across participants. Partner faces were digitally manipulated to be either self- or nonself-resembling, and these conditions were also counterbalanced. Rejection feelings differed significantly as a function of self-resemblance between the inclusion and exclusion conditions, but only for participants who experienced inclusion first. Moreover, for these individuals, inclusion by self-resembling partners led to significantly lesser feelings of rejection than did inclusion by nonself-resembling partners. To explain this effect, we explore potential mechanisms of kin recognition and social rejection. Although nuanced, our results suggest that perceptions of kinship can moderate psychological responses to the actions of others.

Inclusive fitness theory (Hamilton, 1964) provides two broad ways in which actors can improve their genetic posterity. They can do so directly, by increasing their own ability to survive and reproduce (direct fitness), or indirectly, by helping genetic relatives—such as genealogical kin, who are more likely than chance to bear copies of the actor’s alleles identical by descent—to survive and reproduce (indirect fitness). To allocate indirect fitness benefits appropriately, many species have evolved mechanisms of kin recognition to determine genetic relatedness (Hepper, 1991; Krupp, DeBruine, & Jones, 2011; Waldman, 1987). Humans, for instance, have historically lived in communities consisting of a mixture of related and unrelated individuals with whom they would have exchanged goods and services. This situation would have encouraged the evolution of a range of kin recognition mechanisms (Cornwallis, West, & Griffin, 2009; Krupp, DeBruine, & Barclay, 2008), including those using contextual (e.g., Lieberman, 2009; Lieberman, Tooby, & Cosmides, 2007) and phenotypic cues (e.g., DeBruine, 2002; Krupp et al., 2008), to allocate resources discriminatively.

Kinship and Self-Resemblance

Among these mechanisms, humans appear to recognize kin in part by phenotype matching, a process by which an individual evaluator compares the phenotypes of social partners to information about his or her own phenotype and to the distribution of phenotypes in the larger population (Krupp et al., 2011; Krupp & Taylor, 2013). This information may develop as a consequence of genetic inheritance, learning, or both, and it also may be updated continuously throughout the life span. For instance, information about one’s own phenotype can be learned through direct experience (i.e., self-inspection) or through experience with the phenotypes of caregivers and nestmates who are typically parents and siblings. Partners whose phenotypes are relatively similar to the evaluator’s own expected phenotype are classified as kin (or “positive” relatives), and partners whose phenotypes are relatively dissimilar are classified as nonkin (or possibly even as “negative” relatives; Krupp et al., 2011; Krupp & Taylor, 2013). In general, individuals are expected to cooperate more with kin than with nonkin but to avoid them sexually, to decrease the risks of inbreeding (DeBruine, 2005).

Although some studies of human phenotype matching use naturally occurring individual variation in similarity (e.g., Apicella & Marlowe, 2004; Kaminski, Dridi, Graff, & Gentaz, 2009), experimental manipulations are also possible. In this regard, the methods of choice are to either digitally “morph” or “transform” images of strangers’ faces with images of participants’ faces and ask participants to evaluate or interact with these altered images. The results of research using these images strongly suggest that judgments of facial similarity are made for the purpose of estimating kinship and are little affected by differences in age or sex (DeBruine et al., 2009; Maloney & Dal Martello, 2006). Furthermore, facial self-resemblance increases cooperative behavior, trust, and sexual aversion, interacts with other kinship cues, and is distinct from more general preferences for familiarity (Dal Martello, DeBruine, & Maloney, 2015; DeBruine, 2002, 2004, 2005; DeBruine et al., 2011; Krupp et al., 2008; Krupp, DeBruine, Jones, & Lalumière, 2012). Thus, facial self-resemblance can be thought of as a cue of kinship that influences our attributions and treatment of others. Little research, however, has examined how we respond to treatment by self-resembling others. For example, how do our reactions vary when we suffer rejection at the hands of “perceived” kin? To explore this issue, we expanded the research on facial self-resemblance by connecting it to work on social rejection.

Social Rejection

Humans are a social species, possessing psychological and behavioral adaptations designed for the purpose of forming and maintaining groups (Baumeister & Leary, 1995). The establishment of social connections ensures access to resources and support and thus is essential to maximizing inclusive fitness. Consequently, individuals have a visceral negative reaction when their sense of belonging to or acceptance in a group is threatened. Indeed, social rejection can lead to hurt, sadness, loneliness, anger, anxiety, and decreased state self-esteem (Leary & Leder, 2009; Richman & Leary, 2009).

Much of the recent research on social rejection has used Cyberball, a virtual ball-tossing game developed to study ostracism (Williams, Cheung, & Choi, 2000; Williams & Jarvis, 2006). Participants play with at least two other, ostensible players, tossing a ball back and forth among themselves. A welcome page presents a cover story that describes the game as a means of practicing “mental visualization skills,” and many aspects of the game are customizable, including the number of ball tosses, speed of each toss, number of other players, and pictures and text labels for each player. Typically, there are two conditions in the game: inclusion and exclusion. In the inclusion condition, the other, preprogramed players toss the ball to the focal participant as often as they toss it to one another (i.e., one third of the time if there are two other players). In the exclusion condition, however, the other ostensible players toss the ball twice to the participant and thereafter only to each other, ignoring the participant altogether. The most common dependent measure is a questionnaire assessing participants’ mood and need satisfaction (i.e., ratings of belonging, self-esteem, control, and meaningful existence; Zadro, Williams, & Richardson, 2004). Cyberball has been shown to reliably induce an aversive and distressing rejection experience—even when, for example, exclusion results in monetary reward (van Beest & Williams, 2006) or the participant is informed that he or she is playing with a computer rather than with real people (Zadro et al., 2004).

Of particular relevance to the current problem, several studies have also examined how various characteristics of the ostracizing players moderate people’s responses to being ostracized within the game. The results have been inconsistent for similarity or in-group–out-group status of the ostracizers, with some studies finding stronger effects of group membership on the psychological impact of rejection than others (Bernstein, Sacco, Young, Hugenberg, & Cook, 2010; Gonsalkorale & Williams, 2007; Sacco, Bernstein, Young, & Hugenberg, 2014; Williams et al., 2000).

To our knowledge, only one study has examined the effects of kinship on responses to ostracism. In a study by Sreekrishnan et al. (2014), children aged 9–12 years believed they were playing with their mother and a stranger child of their own sex, whereas mothers believed they were playing the game with their child and a stranger mother. After 108 trials of inclusion, participants were excluded for 44 trials and then rated their mood and need satisfaction. Children reported lower mood and need satisfaction than did their mothers.

Unfortunately, this study had several shortcomings. First, mothers and their children differed from strangers not just with respect to kinship cues but also with respect to familiarity. Moreover, mothers and their children had preexisting relationships. Consequently, we do not know whether the effects are due to kinship or to relationship closeness, which also has been shown to yield more negative responses to ostracism (Arriaga, Capezza, Reed, Wesselmann, & Williams, 2014). Second, children are dependent on their mothers for access to resources, care, and support, but mothers are not dependent on their children for these things (Sreekrishnan et al., 2014). Thus, we do not know whether the divergent responses are due to differential effects of kinship or to this power differential. Relatedly, mothers and their children differ in age, and perhaps developmental stage played a role in participants’ responses (Sreekrishnan et al., 2014). Third, the study lacked key comparison conditions, such as having some participants play only with nonkin (i.e., two strangers) to more directly compare the effects of exclusion by kin versus nonkin. Furthermore, although inclusion and exclusion were manipulated, these conditions were always presented in the same order (inclusion followed by exclusion), and participants completed the self-report measure only at the end of the exclusion trials. As a result, the effects of inclusion versus exclusion by kin—as endorsed on the self-report measure—could not be compared.

The Current Research

Given that familial networks have historically provided humans with both direct and indirect fitness benefits, we reasoned that inclusion in mutually beneficial activities by kin might elicit more positive feelings than inclusion by nonkin and, conversely, that exclusion by kin may be especially unexpected and distressing relative to exclusion by nonkin. Consequently, we sought to test the hypotheses that individuals have (1) lesser feelings of rejection when included in and (2) greater feelings of rejection when excluded from group activities by kin, relative to inclusion and exclusion, respectively, by nonkin. Recent work by van der Molen et al. (in press) provides support for the notion that individuals have different neural responses to social rejection that is expected versus unexpected. However, such neural activity does not always correlate with subjective distress (Baddam et al., 2016), which is arguably of greater practical interest. Expanding on the work of Sreekrishnan et al. (2014), we had participants play two games of Cyberball wherein we experimentally manipulated perceptions of kinship via facial self-resemblance of the other players, used a within-subjects design that counterbalanced the order of inclusion/exclusion and kinship conditions, and collected participants’ responses on a “rejection feelings” questionnaire after each game. We predicted that participants would feel less rejected following inclusion and more rejected following exclusion from a Cyberball game when their partners were self-resembling than when they were not.

Method

Participants

Undergraduate students (N = 104) enrolled in an introductory psychology course participated in this study. They received either an additional 2% toward their final exam grade or US$10 per hour if they had already received maximum course credit. The research protocols were approved by the Queen’s University General Research Ethics Board.

Following standard practice (e.g., DeBruine, 2002, 2004, 2005; DeBruine et al., 2011; Krupp et al., 2008, 2012), we used a yoking procedure to compare the responses of two different participants to inclusion or exclusion by the same set of faces, where these faces were self-resembling to one participant and nonself-resembling to the other in a given game. Hence, both participants in a yoked pair saw ostensible partners with the same faces behave in the same ways; one participant in each pair served as the “experimental” participant and the other as his or her “control,” and these roles were switched between the two games (i.e., inclusion and exclusion conditions). Consequently, any effects of the manipulation could be attributed to self-resemblance and not to some more general effect. Due to difficulties finding another participant of the same sex and self-reported racial background in time for the second session (see below), the data for 14 participants were omitted from the analyses, thus yielding a sample size of 45 yoked pairs or 90 participants (76 women, 14 men; racial background: 64 European, 17 East Asian, and 9 other groups including people who endorsed multiple racial backgrounds).

Materials

Self-resemblance manipulation

Custom-built digital transformation software (Tiddeman, Burt, & Perrett, 2001) was used to create self-resembling transforms for all participants (Krupp et al., 2012). A photograph of each participant, centered in the frame with a neutral expression, was cropped to a standard size (320 × 400 pixels). Similarly standardized photographs of strangers (base faces) were obtained from a database of past research participants, and two base faces of the same sex and racial background were selected for each participant. The shape of each participant’s face was delineated by placing 179 points around the face. After standardizing all photographs for interpupillary distance, the software computed the difference in shape between each participant’s face and an appropriate composite face (i.e., an average face of the same sex and racial background as the participant). Fifty percent of this difference was then added to the shape of each selected base face. Consequently, two same-sex, same-race transforms were generated for each participant. A 50% transformation is the strongest transformation at which participants do not consciously detect self-resemblance and are sufficient to elicit the behavioral effects previously described (e.g., greater cooperation with self-resembling others; DeBruine, 2004). The hair, clothing, jewelry, and background of the transformed photographs remained the same as those of the base faces (for photographic examples of the facial self-resemblance manipulation, see Krupp et al., 2011, 2012). The photos were resized to 200 × 250 pixels to fit into the Cyberball user interface.

Inclusion/exclusion manipulation

Participant pairs played one game of Cyberball in which they were included and one game in which they were excluded, in counterbalanced order. In all cases, each participant played with two ostensible partners represented by face images on the screen. In the inclusion condition, participants received the ball exactly one third of the time. In the exclusion condition, participants received the ball twice at the beginning of the game and then never again.

Rejection feelings questionnaire

After each Cyberball game, participants completed a 30-item questionnaire designed to measure feelings of rejection (cf. Zadro et al., 2004). Items within each section were presented to participants in a randomized order. Mood was assessed using 4 items that were rated on a 9-point bipolar scale: bad/good, happy/sad, tense/relaxed, and aroused/not aroused. Four psychological needs that are known to be affected by social rejection (Williams, 1997, 2001) were assessed by 4 items each: belonging (e.g., “I felt like an outsider during the Cyberball game”), control (e.g., “I felt somewhat frustrated during the Cyberball game”), self-esteem (e.g., “During the Cyberball game, I felt good about myself”), and meaningful existence (e.g., “I felt nonexistent during the Cyberball game”). We also included 10 other questions: 3 items assessed other feelings about the game (e.g., “I enjoyed playing the Cyberball game”), 3 items were related to feeling rejected or accepted (e.g., “To what extent do you feel you were being ignored or excluded by the other players?”), and 3 items pertained to liking for or being liked by the other participants (e.g., “How much do you like the other players?”). Participants indicated to what extent each statement was true for them on a 9-point scale from 1 (not at all) to 9 (very much so). One other item was included as a manipulation check for the inclusion/exclusion manipulation (e.g., “What percent of the throws were thrown to you?”).

Although many studies examine the effects of Cyberball on five different dependent measures (mood and the four basic needs) without first examining the structure of the measures, we did not have specific predictions for how our kinship manipulation would affect, for example, need for control versus meaningfulness. Consequently, we were interested in determining the most parsimonious combination of the responses to rejection. To this end, we conducted principal axis factor analyses for the 29 items scored on the 1–9 scale using the psych package (Revelle, 2016) in R (R Development Core Team, 2008) and following the recommendations of Fabrigar, Wegener, MacCallum, and Strahan (1999). We also separately examined the items for the first and second games. In both cases, the scree plot supported a one-factor solution, whereas a parallel analysis suggested a three-factor solution, so we examined those two models for each time point. The eigenvalues for the three factors ranged from 1.29 to 13.17 for Game 1 and from 1.34 to 14.97 for the second game with the first factor having the largest eigenvalue and accounting for the most variance as is necessitated by a principal axis factor analysis. The three-factor solutions had a large number of cross loadings, which made it difficult to interpret. Thus, the more parsimonious single-factor solution was preferred and was deemed “feelings of rejection.” This one factor accounted for 48% and 59% of the variance in participants’ responses for the first and second games, respectively.

The following 3 items were excluded because they did not load substantially on this factor (loadings were less than .29 for Game 1 and less than .37 for Game 2) and yielded item–total correlations of less than .30 in subsequent item analyses for each game: the “aroused/not aroused” mood item, “Life is meaningless,” and “I am in control of my life.” After reverse scoring the appropriate items, a mean score of the remaining items was computed, such that higher scores represented greater feelings of rejection. The Cronbach’s α for this index of rejection feelings was .97 for each game, indicating excellent internal consistency.

Procedure

The study consisted of two sessions, set approximately 1 week apart. In the first session, participants were photographed and informed that their image would be used in the second session, which would involve playing a game with other participants over a computer. They also provided demographic information, which we used to help yoke them to another participant of the same sex and racial background for the second session. For the remainder of the time, they engaged in an unrelated experiment. During the week between sessions, each participant’s photograph was delineated and used to create two self-resembling transforms for him or her.

In the second session, the experimenter informed participants that they were to play two online games with students from other universities. Following the procedure used by Krupp, DeBruine, and Barclay (2008), the experimenter told participants to begin by reading the on-screen instructions while she left the room to make a mock phone call to a laboratory at another university, on the pretense of ensuring that the other players were ready to begin. After the experimenter finished the call, she told the participants to begin playing the game when they were ready. Both participants in each yoked pair were randomly assigned to the same order of rejection conditions (inclusion followed by exclusion or exclusion followed by inclusion). Each yoked pair was also randomly assigned to an order of photo presentation for the two Cyberball games, such that one participant’s self-resembling transforms were used to represent the two other players in the first game and the other participant’s self-resembling transforms were used to represent the two other players in the second game. In this way, one participant would play the first game with his or her perceived kin while the other participant served as his or her yoked control. In the second game, the kinship condition for each participant was switched: The participant who served as the yoked control in the first game now played with perceived kin whereas the other participant now served as his or her yoked control.

Once participants completed the rejection feelings questionnaire that followed the first game, the experimenter reentered the room to load the second game and to inform the participants that they would be playing with two new partners. As outlined above, each participant played the second game in the rejection and kinship conditions opposite to the conditions of the first game (e.g., if a given participant was excluded by perceived kin in the first game, he or she was included by nonkin in the second game). After completing the rejection feelings questionnaire for a second time, participants answered questions about the experiment itself to probe for suspicion. Participants did not appear to detect the self-resemblance manipulation: When asked what the experiment was about or if anything seemed unusual, no participants made any reference to the other players resembling themselves or otherwise appearing unusual. Finally, participants were debriefed, compensated, and thanked for their time.

Analysis of Self-Resemblance Bias

The procedure for computing a measure of self-resemblance bias was adapted from DeBruine (2005) and Krupp et al. (2008). For each game played, the mean rejection feelings score for the participant in the control (i.e., nonself-resembling) condition was subtracted from that of the yoked participant in the experimental (i.e., self-resembling) condition. In this way, a self-resemblance bias score was computed for each game played, allowing us to determine the effect of self-resemblance on rejection feelings in response to inclusion or exclusion.

A difference score of zero would indicate that participants in a yoked pair responded in the same way to the same faces, regardless of self-resemblance. In this case, the participants’ responses would more likely have resulted from something inherent about the faces being viewed (e.g., their attractiveness) or to being included or excluded in general. A nonzero difference score would indicate that participants in a yoked pair responded differently to the same faces, implicating self-resemblance as the cause of the difference. We predicted that participants would exhibit a negative self-resemblance bias when included, meaning that those in the experimental condition would experience lesser feelings of rejection than those in the control condition. Conversely, we predicted that participants would exhibit a positive self-resemblance bias when excluded, meaning that those in the experimental condition would experience greater feelings of rejection than those in the control condition.

Results

A 2 (inclusion/exclusion) × 2 (order) mixed-design analysis of variance (ANOVA) was conducted to evaluate the effect of inclusion/exclusion condition, as well as the order of these conditions, on feelings of rejection experienced by each participant. As this analysis used individual participants as the unit of analysis, the data for all participants (N = 104) were used, with feelings of rejection as the dependent variable. The main effect of inclusion/exclusion condition was significant, F(1, 102) = 359.13, p < .001, partial η² = .78. The main effect of order was also significant, F(1, 102) = 4.05, p = .047, partial η² = .04. These main effects, however, were qualified by a significant interaction between rejection condition and order, F(1, 102) = 8.26, p = .005, partial η² = .08.

To evaluate the nature of this interaction, four t-tests for use with mixed designs (Toothaker, 1991) were conducted with a corrected α of .013. The means and standard deviations used for these pairwise comparisons are presented in Table 1. As expected, compared to the inclusion condition, participants experienced significantly greater feelings of rejection in the exclusion condition both when exclusion preceded inclusion, t(102) = 15.33, p < .001, and when it followed inclusion, t(102) = 11.08, p < .001. Within the exclusion condition, feelings of rejection did not significantly differ based on order, t(102) = 0.55, p = .580. Within the inclusion condition, however, participants experienced significantly greater feelings of rejection when included in the first game than when included in the second game, t(102) = −3.45, p < .001.

Table 1.

Descriptive Statistics for Feelings of Rejection Within Inclusion/Exclusion Condition and Order.

Rejection Condition	Order	M	SD
Inclusion	First game	3.74	1.04
	Second game	3.02	0.75
Exclusion	First game	6.21	1.12
	Second game	6.10	1.31

Self-Resemblance Bias

To test for an effect of self-resemblance within each rejection condition, two single-sample t-tests were conducted. The dependent variable for both tests was self-resemblance bias, computed as outlined above. These tests used yoked pairs as the unit of analysis, so only the data for yoked participants (N = 45 yoked pairs) were used. As predicted, the mean self-resemblance bias score in the inclusion condition (M = −0.24, SD = 1.20) was negative, indicating that participants who were included by self-resembling partners experienced lesser feelings of rejection than did their yoked controls; however, this score did not significantly differ from 0, t(44) = −1.36, p = .182. Also as predicted, the mean self-resemblance bias score in the exclusion condition (M = 0.09, SD = 1.78) was positive, indicating that participants who were excluded by self-resembling partners experienced greater feelings of rejection than did their yoked controls who were excluded by nonself-resembling partners; once again, however, this score was not significantly different from the test value of 0, t(44) = 0.33, p = .747.

To determine whether the effect of self-resemblance differed across rejection conditions or as a function of order, a 2 (inclusion/exclusion) × 2 (order) mixed-design ANOVA was conducted. This test also used yoked pairs as the unit of analysis, with self-resemblance bias as the dependent variable. The main effects of rejection condition and order were both nonsignificant, ps > .29, but the interaction between rejection condition and order was marginally significant, F(1, 43) = 3.87, p = .056, partial η² = .08. The nature of this marginal interaction was explored further with four t-tests and a corrected α of .013, as above. The mean self-resemblance bias scores for each condition are shown in Figure 1. When inclusion was followed by exclusion, the self-resemblance bias differed significantly between the inclusion condition for which the difference score was negative, and exclusion condition for which the bias score was positive, t(43) = 2.17, p = .035. All other pairwise comparisons were nonsignificant, ps > .013.

Figure 1.

Self-resemblance bias for each inclusion/exclusion condition within each order of conditions. A positive self-resemblance bias score indicates greater feelings of rejection when playing with self-resembling partners than with nonself-resembling partners, and a negative self-resemblance bias score indicates lesser feelings of rejection when playing with self-resembling partners relative to the nonself-resembling partners. Error bars represent standard deviations.

To further clarify the effect and more directly test our predictions, we used one-sample t-tests to compare the self-resemblance bias means to 0. When excluded first, the self-resembling bias mean was not significantly different from 0 when participants were either included, t(22) = 0.48, p = .638, or excluded, t(22) = 0.53, p = .604. When included first, as predicted, participants felt less rejected when included by self-resembling than by nonself-resembling others, t(21) = 2.05, p = .053; however, participants did not differ significantly in their responses to rejection when excluded by self-resembling versus nonself-resembling others, t(21) = 1.09, p = .290.

Discussion

As expected, given extensive previous research, participants experienced significantly greater feelings of rejection when excluded than when included in games of Cyberball. Furthermore, exclusion was approximately equally hurtful when it was experienced prior to or following inclusion by a different set of partners. However, game order affected the experience of inclusion such that participants felt less rejected when they had been excluded in a previous game. This suggests that inclusion feels particularly rewarding when it follows such a powerfully distressing social event. Participants who were excluded prior to inclusion might, for instance, have felt that they had overcome an obstacle to group membership, making such inclusion feel especially satisfying.

As genetic relatedness tends to foster cooperation, we hypothesized that humans would have evolved positive sentiments to inclusion and negative sentiments to exclusion from kin networks. However, in the omnibus model, self-resemblance did not cause significantly (1) lesser feelings of rejection when participants were included in group play or (2) greater feelings of rejection when participants were excluded from group play. Nevertheless, closer inspection revealed an interesting effect of the order in which the games were played: The predicted pattern of self-resemblance effects did emerge when participants experienced inclusion first and exclusion second but only for the inclusion condition. Therefore, we conclude that participants found inclusion by self-resembling partners to be more rewarding than inclusion by nonself-resembling partners in the first game. As outlined previously, we believe this is due to the adaptive benefits (e.g., resource sharing) of being included in kin networks.

Although the mean was in the anticipated direction for exclusion by self-resembling partners to be more painful than exclusion by nonself-resembling partners in the second game, the difference was not significantly different from 0. We discuss reasons for this in more detail below. Moreover, we did not find any significant differences for participants who experienced exclusion first. In other research using Cyberball, the effect of exclusion on affect has generally been quite powerful, and for many years, researchers did not believe it was possible to find moderation of its effects because it was such a strong situation (McDonald & Donnellan, 2012). Thus, we suspect that our results for self-resemblance bias were weaker for the exclusion condition than we expected because exposure to this condition obscured or interacted with other events or processes in the conditions that followed it.

Indeed, there may be carryover effects of exclusion on subsequent interactions. For instance, participants may be emotionally overwhelmed—to the point of distraction—by the experience of being excluded, making them less likely to detect self-resemblance in a later game. If not impairing attention in general, exclusion may shift one’s focus of attention to cues other than self-resemblance, such as the number of tosses received, in ensuing interactions. Another possibility is that prior exclusion by self-resembling partners yields a more positive interpretation of inclusion by nonself-resembling partners than is typical, perhaps because exclusion signals that the parties are in fact unrelated. Alternatively, perhaps the critical difference lies not in this order of inclusion/exclusion conditions, but in the opposite: Exclusion by self-resembling partners in the second game may be even more unexpected and distressing to individuals who have previously been included by nonself-resembling others and who thereafter would understandably expect—even if unconsciously—to be accepted by perceived kin.

Limitations and Future Directions

A potential limitation of our study is that the self-resemblance manipulation was too subtle when presented within the Cyberball interface. If true, this may explain the nonsignificant effect of self-resemblance on social rejection when exclusion was followed by inclusion: Participants may have been more easily distracted by exclusion when the self-resemblance manipulation was relatively weak. Future Cyberball studies might try using larger facial images or stronger than 50% transformations to determine whether these changes make self-resemblance more salient and therefore increase the magnitude of self-resemblance effects.

It may also be that the crossover effect we find here is an artifact of a within-subjects design. Consequently, it would be interesting to compare the results of within-subjects and between-subjects designs. For instance, participants might play a single game of Cyberball, wherein half of the participants are included and the other half excluded by their partners. Moreover, if using a within-subjects design, it might be interesting to provide a neutral task between Cyberball games to determine whether any carryover effects of exclusion or inclusion persist.

Conclusions

Some researchers have contended that social rejection, particularly in the form of ostracism, is so aversive a phenomenon that moderation is difficult, if not impossible (e.g., McDonald & Donnellan, 2012). However, others have challenged that claim (Sacco et al., 2014; Williams, 2009). The current study adds further support to the latter position, showing that even a subtle manipulation—exposure to self-resembling partners—can moderate the powerful psychological impact of social exclusion, if only under specific conditions. Specifically, our study has shown that self-resemblance can alter the subjective experience of social inclusion and exclusion, at least when exclusion follows inclusion. Future research should establish an effect of self-resemblance on responses to inclusion and exclusion separately, before further investigation into potential order effects. In addition, we expect that more salient self-resemblance manipulations will yield clearer results and stronger effects. In any case, we believe that our findings justify further forays into the emotional responses that individuals have to the actions of their putative kin, and how these responses might interact with other psychological processes.