The field of social exclusion has experienced surge in research in the past 20 years, experimentally drawing conclusions on human phenomenon. Social exclusion is present throughout many cultures, and may be the most pervasive form of social punishment in humans (Williams & Sommer, 1997). Social exclusion has been shown to induce psychological pain by activating the dorsal Anterior Cingulate Cortex (dACC), the same area of the brain associated with physical pain (Eisenberger, Lieberman & Williams, 2003). This correlation highlights the evolutionary role of social exclusion as a factor in the interpretations of, and subsequent interactions with, our environment. Humans are social animals that require companionship for survival, and the punishment for being excluded from survival activities can lead to very real pain, thus encouraging the target to return to the fold (Lieberman & Eisenberger, 2005). Social exclusion is often used as a tool to encourage dissenters to return to the fold by eliciting feelings of pain and a desire to reconnect with the excluding group. In addition, physical consequences of exclusion (i.e., depriving an individual of belonging) include increased stress, a poor immune system compared to married couples, eating disorders, criminal behavior, and suicide (Baumeister & Leary et al., 1995; Williams & Zadro, 2001).
Social exclusion research dates back to the mid-twentieth century; studies conducted by Dittes (1959), Snoek (1962), and Geller et al. (1974) successfully invoked feelings of exclusion and assessed subsequent behavior (Leary, 2001). Current research focuses on exclusion’s effect on affective responses, including threats to psychosocial needs (e.g., feelings of belonging, self-esteem) and mood, as well as post-exclusion coping behavior, including effort on follow-up tasks, and aggressive acts. However, an experimental study addressing situational differences has yet to be published.
There are important nuances within the larger study of social exclusion. Often researchers use terms like ostracism and rejection interchangeably, and prior theories do not distinguish between these experiences. Ostracism is commonly defined as “being ignored and excluded, and [it] often occurs without excessive explanation or explicit negative attention (Williams, 2007 p 429). Alternatively, Rejection is “a declaration by an individual or group that they do not (or no longer) want to interact or be in the company of the individual” (Williams, 2007 p 429). Furthermore, long-term Social Exclusion is defined as “being excluded, alone, or isolated, sometimes with explicit declarations of dislike, but other times not” (Williams, 2007 p 429), indicating a potential combination of Ostracism and Rejection. This meta-analysis addresses these gaps in the literature pertaining to these differences in order to obtain a more robust understanding of social exclusion mechanisms.
Social Exclusion Theory
Many researchers have associated social exclusion, the general term for acts of ostracism and rejection, with threats to fundamental psychosocial needs including self-esteem (Leary et al., 1995; Leary, 1990), as well as feelings of belongingness (Baumeister & Leary et al., 1995), control (Bruneau, 1973; Williams & Sommer, 1997), and self-worth or a meaningful existence (Williams & Sommer, 1997; Twenge, Cantanese, Baumeister, 2002). These studies led Williams and Sommer to establish the Model of Ostracism: social exclusion prevents individuals from satisfying the fundamental psychological needs, thus resulting in worsened mood and hurt feelings, as well as behavior designed to repair and maintain the self in the short term (e.g., taking control, self affirmations, potential pro-social behavior), or long-term distancing from the group (e.g., self-imposed isolation, learned helplessness, low self-esteem) (Williams & Sommer, 1997). In this hypothesis, the effect of exclusion on mood and post-exclusion coping behavior is moderated by threats to psychosocial needs.
Conventional wisdom states that exclusion has a prominent effect on self esteem, (Sommer, 2001). It is often defined as an individual’s overall evaluation of the self (Rosenberg et al., 1995). Self-esteem is divided into trait and state conditions; trait self-esteem is considered a dimension of personality and refers to the individual’s long-standing impression of the self (i.e., high or low self-esteem), while state self-esteem refers to an individual’s current self-esteem, which is affected by daily events. Although laboratory-induced exclusion may only affect state self-esteem, it has been demonstrated that constant levels of state self-esteem may affect trait self-esteem (Zadro, Boland, & Richardson, 2006). The smallest child may get picked last for a game thus creating low state self-esteem; but this may lead to low trait self-esteem, as the anticipation of anxiety and unpleasant feelings are transferred to other activities and affect the child’s future behavior (Leary et al., 1995; Zadro, Boland, & Richardson, 2006).
Exclusion is frequently associated with low self-esteem, either directly or moderated through other psychological traits. Exclusion targets (i.e., victims of exclusion) report higher negative concepts of self, including lower self-actualization, competence, intrinsic motivation, and self-efficacy or control (Aron & Aron, 1991). Self-esteem has been shown to increase with social acceptance and decrease with rejection (Leary et al., 1995) thereby functioning as a “sociometer”; “The self-esteem system monitors others’ reactions and alerts the individual to the possibility of social exclusion” (Leary et al., 1995, p. 518). This situates self-esteem as both an outcome of the social environment and a motivator to interact with the environment, resulting in a circular process that highlights the ability of state self-esteem to affect trait self-esteem over time as the individual begins to anticipate rejection (Leary et al., 1995). Self-esteem can also act as a moderator for the interpretation of and reaction towards exclusion: high self-esteem subjects may avoid the negative implications of rejection by enhancing self concepts both internally (i.e., self-esteem) and externally (e.g., lashing out with hostility or violence), while low self-esteem subjects view the rejection as a confirmation of previously held beliefs (Sommer, 2001).
Belongingness is the fundamental motivation have to belong, which drives much of our behavior. Baumeister and Leary (1995) cite the ease with which we form social bonds, our difficulty in breaking those bonds, and the effect of belongingness on emotion as support of their Belongingness Theory. Exclusion can motivate individuals to seek belonging through behaviors like working harder in groups and conforming to the majority (Baumeister & Leary, 1995).
A sense of control over one’s environment is an essential component of mental well-being. Deci & Ryan’s Self-Determination Theory claims that “people want to feel effective in their activities (competence), to feel that their activities are self-chosen and self-endorsed (autonomy), and to feel a sense of closeness with some others (relatedness)” (Sheldon, 2001). When asked to isolate the most recent satisfying event and describe their emotions surrounding it, subjects consistently rated competence, autonomy and relatedness as important when defining events as “satisfying” (Sheldon, 2001). Humans are motivated to control their environment and experience themselves as capable and effective, and exclusion can rob an individual of this need (Bruneau, 1973) leading to disastrous results; in an interview, a woman claimed to have developed an eating disorder, after being ostracized by her mother for years, in order to “maintain some control over my life” (Williams & Sommer, 1997, p. 159). Furthermore, neutralizing threats to control has been shown to moderate aggressive responses after exclusion, indicating that individuals may use aggression as a method of revenge against their exclusion sources. (Leary, Twenge, Quinlivan, 2006).
Social inclusion is essential to ensuring a meaningful existence as perceived by the individual. Being ignored by others stimulates feelings of invisibility and can cause individuals to question their purpose in life. “Social death refers to the point at which other people cease to socially interact with the dying person” (Williams & Zadro, 2001, p 23) by refusing to acknowledge the individual’s presence and behaving as if the individual was deceased. Social exclusion is correlated with a statistically significant increase in agreement with the phrase “Life is meaningless,” engagement in self-defeating behavior (Twenge & Baumesiter, 2002), and increased rates of suicide (Sommer, 2001).
The findings around mood are not as simple. Some researchers have found that exclusion increases negative mood and decreases positive mood, while others have found no effect. In a review by Williams, “The typical effect size of ostracism on self-reported distress (as measured by moods and need threat) is high, between 1.0 and 2.0” (2007, p 434); however, these results depend on mood and needs being treated as a single construct, which eliminates the ability to differentiate exclusion effects on mood and needs. Furthermore, other studies have demonstrated that “socially excluded individuals show no signs of mood impact” (Williams, 2007, p 432), therefore, a meta-analysis is required to assess the overall effect of exclusion on mood as well as to investigate the effect of nuances of exclusion manipulations.
According to Baumeister’s Numbness Hypothesis, exclusion causes a numbing of affect, that is neither positive nor negative. Instead of eliciting the emotional distress commonly believed to be associated with exclusion, people may respond with low arousal, “empty, neutral, and even bored feelings” (Massong, Dickson, Ritzler, & Layne, 1982) in order to prevent isolating affect. According to this hypothesis, exclusion will result in a “deconstructed state,” which while somewhat vague, is operationalized as present oriented, disorder time perception, lack of meaningful existence, lethargy, lack of emotion, and an escape from self-awareness. “If social exclusion thwarts a basic human drive and challenges one’s self-worth, then people might prefer to escape self-awareness and emotional distress by hiding out in a mental state marked by numbness, lack of meaningful thought, and a narrow focus on concrete, immediate stimuli” (Twenge, Cantanese, & Baumeister, 2003). This theory, although focusing on exclusion’s numbing effect on mood, demands that exclusion affects mood, therefore the comparison between exclusion and acceptance should reveal an effect on mood independent of effects to psychosocial needs.
Individual reactions to exclusion differ drastically based on a variety of moderators including personality traits, a desire to maintain relations with the group, and gender. Levels of narcissism moderate subsequent behavior; high narcissistic subjects will respond with greater anger and aggression towards the exclusion sources (Leary, Twenge, & Quinlivan, 2006). Furthermore, Zadro, Boland, and Richardson (2006) found that high socially anxious participants neutralized threats to psychosocial needs and returned to baseline measures after being excluded slower than those that did not report high measures of social anxiety, thus demonstrating that resolving the detrimental effects of ostracism depends on personality characteristics and interpretation of the events.
In addition, post-exclusion behavior often depends on whether anti-social behavior will breed further exclusion (Williams & Zadro, 2001). Subjects ostracized by faceless sources with whom they may have little future interaction (e.g., an online chat room) were openly negative in their reactions to the exclusion (Williams, Cheung, & Choi, 2000), a phenomenon referred to as “virtual bravado.” Alternatively, participants ostracized by a desired group or a group with whom future interaction is desired (e.g., work group, long term-companions) tempered their negative reactions, often choosing to disassociate from the conversation (Williams & Sommer, 1997).
Gender is also a significant moderator of post-exclusion coping behavior. Leary found that women rated themselves as less positive than men when ostracized, regardless of the reasons for exclusion (i.e., random or because of rejector preferences), suggesting that women may be more sensitive to rejection cues (1995). Furthermore, when ostracized from a work group prior to interaction, women exhibited greater effort when interacting with the group (i.e., social facilitation) whereas men withdrew and disassociated from the group (i.e., social loafing) (Williams & Sommer, 1997). Men also report a greater dislike for the group after being excluded (Dittes 1959). These differences have been attributed to the socialized desires of each gender; women worked harder in order to increase their belongingness and group status, while men separated from the conversation and played with objects in the environment in order to maintain control (Williams & Sommer, 1997). Kelly suggests that men are more likely to attribute their rejection to external factors (2001), thus taking actions to save their self-esteem that include withdrawing or disengaging from prior rejecting group members (Williams & Sommer, 1997).
Social Exclusion Conditions
There are four major methods used to induce exclusion in the laboratory: Ostracism, (Demarcated) Rejection, Hypothetical Social Exclusion (HSE or Future Rejection), and Reliving Rejection. The final condition, Reliving Rejection, asks participants to recall a time when they experience exclusion (or acceptance); this condition was eliminated from analyses due to the lack of experimental control.
Ostracism is the defined as the ignoring an individual or group by another group (Williams, 2001). Laboratory and online inductions of Ostracism exclude participants from engaging in a conversation or a ball-tossing game. In the conversation scenario, targets are actively ignored during a conversation; this paradigm usually involves three individuals: two exclusion sources and one exclusion target. In laboratory inductions of exclusion, confederates serve as the exclusion sources and ignore the target’s attempts to interact. Gardner, Pickett, and Brewer (2000) and Williams et al. (2002) successfully employed this paradigm in virtual space using “online” chat rooms, where participants were instructed to converse with other “participants” elsewhere in the building via computer, but the conversation was generated by a computer program. In the ball-tossing scenario, subjects play ball with two confederates. One of the earliest experimental paradigms to induce exclusion, the real space version of this game was first employed by Williams & Sommer (1997); a subject was placed in a room with two confederates and all were asked to remain silent for five minutes while the experimenter stepped away. One confederate would rummage through a box, “discover” a ball, and begin tossing it to the other confederate and the subject. After a few tosses, the confederates either continued to involve the subject (inclusion), or refused to throw the ball to the subject (exclusion); this method has also been replicated online using the simple program “Cyberball,” where participants toss an onscreen ball with two other individuals (computerized confederates) represented with gender and race neutral cartoon avatars. Lab inductions of Ostracism do not include explicit Rejection by the group, i.e., participants are not informed that the other players chose not to include the participant. These methods may not be applicable to the real world where Ostracism and Rejection can be co-present.
Rejection studies inform participants that they have been rejected, or not selected, by an individual or group (Leary et al., 1995; Twenge, Cantanese, & Baumeister, 2003). The manipulation is often presented as bogus ratings or poll results indicating that the other members of the group do not want to work, or engage in future interactions, with the participant. Other researchers have also effectively employed, within a controlled setting, the classic and familiar exclusion tactic of being chosen last for a team (Bourgeois & Leary, 2001). The online evolution of the selection paradigm utilizes virtual “online” chat rooms where a subject is rejected by a computer generated, opposite sex participant that he or she is conversing with; the subject is informed that their virtual partner either does or does not want to meet (Buckley, Winkle, & Leary, 2004). This finding takes social exclusion research into the world of online dating, a popular trend in recent years; despite the safeguards against actual interpersonal rejection, subjects still exhibit post exclusion behavior (e.g., reduced belongingness, reduced control, reduced meaningful existence) even when rejected by a faceless or online source.
Finally, Hypothetical Social Exclusion (HSE) informs participants that, based on a personality test, they are destined to spend a future alone without social contacts or close friends; this manipulation is also referred to as a “Future Alone” (Baumeister, Twenge, & Nuss, 2002; Baumeister, DeWall, Ciacorro, & Twenge, 2005). Some studies have attempted to increase believability by prefacing the manipulation with the appropriate diagnosis of the participants extraversion score (Baumeister, DeWall, Ciarocco, Twenge, 2005; DeWall & Baumeister, 2006). Research by Geller, Goldstein, Silver, & Sternberg (1974) successfully elicited feelings of exclusion by simply asking participants to imagine that they were excluded from a conversation, indicating that simply envisioning exclusion can elicit distress.
These scenarios and measurements focus on the immediate threat to the self as well as the short-term repercussions of exclusion, and the experimental deficits of these experiments are well documented. It is almost impossible (and unethical) to replicate laboratory studies for long-term exclusion; furthermore, although these three conditions of exclusion can be induced separately in the lab, they are often co-present in the real world. Preexisting data on this subject is difficult to quantify and often depends on diaries and self-report measures, which can be unreliable and reliving instances of social exclusion can be emotionally distressing to subjects thus complicating the analysis. Studies regarding the cognitive state of prisoners and social outcasts offer a glimpse into the psychological repercussions of long-term exclusion (Williams & Sommer, 1997), but the experiences of these individuals are inherently multifaceted and exclusion cannot be assessed independently.
Prior Social Exclusion Meta-Analyses
A recent meta-analysis by Gerber & Wheeler (2009) begins to address some of the issues still present in the literature. This article sparked a harsh response by Baumeister, DeWall, and Vohs that attacked their methods and interpretations. Despite using “rejection” to label all types of social exclusion, Gerber & Wheeler assess the differences between methods of exclusion: Ostracism, (Demarcated) Rejection, and Hypothetical Social Exclusion (HSE, or Future Rejection). Although Gerber & Wheeler note the differences between these two conditions, they do not begin to address the different mechanisms of these conditions, and their methods for quantifying threats to psychosocial needs have sparked a heated debate in the current literature.
Gerber & Wheeler isolated six major dependent variables: Self-esteem, Belonging, Control, Meaningful Existence, Mood, and Arousal. The main hypothesis of Gerber & Wheeler is that people will behave in a manner to restore belonging and control, but will prioritize belonging over control. Their analyses into post-exclusion coping behavior demonstrate that participants engage in control-enhancing and belonging-enhancing behaviors when given the opportunity. Furthermore, “when people had to be antisocial to satisfy control, they chose to be antisocial. however if control was absent or if it was possible to satisfy both belonging and control simultaneously, people were prosocial” (p 479).
In order to examine their main hypotheses addressing the relationship between Threats to Psychosocial Needs (e.g., Control, Belongingness) and post-exclusion behaviors, Gerber & Wheeler code actions for their ability to increase, or resolve, the participant’s threatened needs. Belongingness items were defined as “items that involved other people, items in which there was potential for some further interaction, or behaviors that would facilitation interaction” (e.g., Asch line task); control items were defined as those that were “indicative of people feeling powerful but not yet exercising that power” (p 474) (e.g., uncooperative turns in the prisoner’s dilemma game, pain tolerance in the cold water pressor task). Baumeister, DeWall and Vohs (2009) criticize this methodology due to the fact that no prior research establishes the relationship between several of the tasks that are coded as belongingness- or control-enhancing and their ability to resolve threats to these needs. Baumeister, DeWall, and Vohs claim that this method of coding anticipated emotions or resolution of Threats to Psychosocial Needs is flawed as people are notoriously poor at forecasting the intensity of their emotions given a specific event as well as the extent to which follow-up actions will moderate the emotional response (Wilson & Gilbert, 2003). This meta-analysis will seek to utilize more objective categorization of post-exclusion behavior
Gerber & Wheeler also found that exclusion had a moderate but significant effect on mood and concluded that “rejection makes people feel bad,” a finding contrary to Baumeister’s Numbness Hypothesis. An examination of the means by Baumeister, DeWall, and Vohs (2009) reveals that included participants report a slight increase in mood, while excluded participants report a neutral average on a scale of 1 (bad) to 7 (good), where 4 is the midpoint; “It would be hard to imagine a result that confirms the emotional numbness hypothesis better than these means of 4É Yet Gerber and Wheeler repeatedly interpreted these results as contradicting the numbness hypothesis” (p 491). Gerber & Wheeler demonstrate a interpretative bias in their reporting, assuming that a negative effect size indicates negative affect; Baumeister et al. argue that the same result may mean that inclusion increases mood in the accepted condition. The current meta-analysis investigates the situational moderators of mood in the exclusion mechanism, and addresses all possible interpretations of composite effect sizes.
Additional analyses regarding methodology are essential to the research field. Gerber & Wheeler found significant differences between the scales used to assess exclusion-induced distress (e.g., PANAS vs. BMIS) and post-exclusion behavior (e.g., behavioral measures vs. validated self-report measures), indicating that not all measures are created equal; Gerber & Wheeler demonstrated no effect of exclusion on the BMIS measure of mood and large effects when using researcher-developed items (2009). Furthermore, laboratory-induced social exclusion methods are inconsistent, and several comparison groups are employed as control conditions (e.g., acceptance, negative affect control, argument control). Acceptance was found to be the most frequent control group (a finding replicated in this meta-analysis), thus forcing the meta-analysis to utilize acceptance as the most robust design for analysis. Gerber & Wheeler’s analyses revealed that “inclusion comparison groups [do] not appear to be distorting research on rejection. In most cases, neutral comparison groups give similar results to inclusion comparison groups” (p 479). Although there is a reduced effect on mood with other non-acceptance control groups, these findings indicate that the use of acceptance groups as control is a legitimate method of analyzing exclusion.
What questions remain unanswered?
In a 2007 narrative review of social exclusion literature and prior studies, Williams addresses the two main components of exclusion effects: the immediate affective responses to exclusion, or exclusion-induced distress, and post-exclusion coping behavior. This review highlights the trends in social exclusion research, including the most recent meta-analysis, and the remaining gaps.
1. Is there a difference between Ostracism, Rejection and Social Exclusion?
Prior research regarding social exclusion uses these terms interchangeably; Williams addresses this shortcoming by defining each of the terms, but then admits to also using them interchangeably in the review. Laboratory inductions of social exclusion seem to mimic these differences, but there is no direct comparison of methods. Furthermore, there is no explicit investigation of mechanisms using regression analysis in the prior research. A meta-analysis may address these differences addressed above and may provide evidence for different mechanisms, thus leading to the first hypothesis:
H1: Social Exclusion (general) encompasses three conditions: Ostracism, Rejection, and Social Exclusion. Each condition will demonstrate different effects on exclusion-induced distress and post-exclusion coping behavior.
H1A: According to the Model of Ostracism, Threats to Psychosocial Needs will moderate the relationship between Reported Exclusion and Mood, as well as post-exclusion coping behavior.
H1B: In line with the Numbness Hypothesis, Reported Exclusion will have a direct effect on Mood independent of Threats to Psychosocial Needs. Furthermore, emotional numbness should be correlated with other measures indicating a “deconstructed state,” including lethargy (or lack of follow-up effort) and Cognitive Loss.
2. What is the effect of situational differences on exclusion-induced distress and post exclusion coping behavior?
There is little research investigating the effect of situational differences on exclusion-induced distress. According to Williams, situational differences will only have an effect on post-exclusion coping behavior (2007), but no studies have directly compared different exclusion contexts (e.g., individual vs. group rejection). However, a review of the methods employed across studies reveals a variety of methodological tactics to induce exclusion, which leads to the second hypothesis:
H2: Situational differences in laboratory-inductions of exclusion will exhibit an effect on exclusion-induced distress and post-exclusion coping behavior.
H2A: Greater numbers of exclusion source(s) will result in greater effects on exclusion-induced distress and post-exclusion behavior.
H2B: The gender composition of exclusion source(s) will exhibit an effect on exclusion-induced distress and post-exclusion coping behavior.
H2C: The level at which the individual knows and interacts with their exclusion source(s) will moderate the effects on exclusion-induced distress and post-exclusion coping behavior.
Sample of Studies
An exhaustive search of articles published prior to August 2009 was conducting through PsychInfo, Ovid, and Google Scholar using the key words: social exclusion, ostracism, and rejection. The references for each of the papers along with references of other qualitative reviews were then searched for additional articles that were not found in the original database searches; no prior meta-analyses for social exclusion had been published by August 2009. This meta-analysis uses only results published in scholarly journals and does not include dissertations or unpublished results.
Inclusion Criteria for Studies
Studies were originally included if they featured normal (i.e., non-clinical) adult participants, utilized laboratory-induced exclusion, and it was possible to derive an effect size demonstrating the difference between the Exclusion condition and a Comparison Group. 148 different comparisons were available from 110 individual studies in 51 articles. Studies were selected if they employed one of the exclusion conditions of interest (Ostracism, Rejection, HSE), and utilized acceptance as their control group. This resulted in a final total of k = 85 studies from 43 articles with 239 effect sizes.
Variables coded from each research report
Several variables were coded relating to study characteristics (e.g., sample size, paradigm, publication), characteristics of the exclusion sources (e.g., number, gender, presence of exclusion sources), and the dependent variables of interest. To maximize the number of studies used for the overall analysis, studies were assigned a missing variable if appropriate results or coding information was not provided.
Sample Size and Composition: The comparison size was calculated from the reported sample sizes. The sample was split evenly when individual cell sizes were not reported. Furthermore, many studies included multiple control groups; the Study N in the Appendix reflects the total number of subjects in the study, and the Comparison N reflects the number of participants utilized in the comparison (i.e., sum N of exclusion and acceptance conditions only). The percentage of female participants and the percentage of White participants were also coded when available.
Experimental Paradigms/Conditions: Exclusion paradigms were aggregated into the three major categories reported earlier: Ostracism, Rejection, and Hypothetical Social Exclusion (HSE or Future Rejection). Eight studies utilized a Reliving Task where participants were asked to recall a time when they felt ostracized or rejected (See Gerber & Wheeler, 2009 for a review), and one study utilized a fictional narrative. These studies were eliminated from the analysis.
Comparison Group: Different studies employed different control groups. Acceptance was the most frequent comparison group, followed by Misfortune Control, and Random Exclusion Control. Due to the drastic difference in sample size, Acceptance was selected as the comparison group of interest for the remainder of the studies. Other comparison groups (e.g., gender, rejection sensitivity) were collapsed across Exclusion/Acceptance conditions.
Characteristics of Exclusion Sources
Number of Exclusion Sources: To assess the effect of the group size in exclusion, the number of exclusion sources was coded. Participants were either excluded by single individuals (1), small groups (between 2-3 other individuals), and large groups (4 or more individuals); studies that reported that participants were run in groups, the average was used as the number of exclusion sources even if this resulted in fractional values (e.g., for a study that ran participants in groups of 3 or 4, indicating 2 or 3 excluders, a value of 2.5 was given).
Gender of Exclusion Sources: When available, the gender of the exclusion source was coded as either same gender as the participant (1), mixed gender group (2), or opposite gender compared to the participant (3). Analyses assess the differences between exclusion by same gender and opposite gender sources as well as regressions utilizing the above ordinal measure to represent the changing ratio of genders during exclusion.
Length of Interaction with Exclusion Sources: When available, the length of time (in minutes) the participant engaged with the excluders was entered into the analysis. This measure was conceptually different between Ostracism and Rejection. The length of time ostracized participants engaged with their exclusion sources was the length of the ostracizing interaction; alternatively, the length of time rejected participants engaged with their exclusion sources was the length of interaction prior to rejection
Knowledge of the Exclusion Sources: Ostracism and Rejection studies were coded for the participant’s knowledge of their excluder(s). Experimental Manipulations either featured no knowledge of the exclusion source, demographic knowledge of the exclusion source (e.g., gender, college-affiliation), brief biographical knowledge delivered via written or videotaped bio-sketch, or interpersonal knowledge derived from actual interpersonal interactions with the exclusion source. The final two categories were only available in the Rejection condition; no studies that employed Ostracism included bio-sketches or interpersonal interactions. This measure was not available for HSE studies.
Presence of Exclusion Sources: Two variables were coded to assess the presence of the exclusion sources: Tangible vs. Imaginary and Online vs. Real. Studies were coded as having tangible exclusion sources if the participants physically observed an individual who would later exclude them (e.g., studies that had participants arrive in groups, but then isolated participants to their own rooms or cubicles for the remainder of the study were classified as having tangible exclusion sources). Studies were coded as having online exclusion sources if the participant was excluded via technology (e.g., chat room, Cyberball, SMS messaging) as opposed to some interpersonal interaction (e.g., real space conversation, or experimenter-informed rejection).
Dependent Variables of Interest
Several variables were originally coded from the studies; after assessing the variety of effect sizes, four major composite variables (*) were constructed and analyzed along with four other secondary variables.
Self-Reported Exclusion(*): Self-reported feelings of exclusion or inclusion were combined to create a single composite measure of Reported Exclusion. This is often a single item assessing the extent to which the participants felt excluded (e.g., “How rejected did this experience make you feel?”). This measure was only provided for Ostracism and Rejection conditions. Positive values reflect greater feelings of exclusion for excluded participants.
Self-Reported Mood(*): Self-reported Mood was coded on two dimensions: positive and negative. Studies that referred to general mood were coded as positive valanced, thus allowing an independent analysis of negative mood as well as the overall composite mood. Acceptable measures included the PANAS, PANAS-X and BMIS scales, as well as single-item measures. Gerber & Wheeler (2009) observed that short mood items developed by the researcher exhibited a greater effect size compared to the established scales, and of the established scales only the PANAS and PANAS-X scales exhibited significant effects. Positive values reflect more negative (or less positive) mood for excluded participants. Negative and positive mood were also analyzed separately in the same direction.
Self-Reported Threats to Psychosocial Needs(*): According to Baumeister, the four major Psychosocial Needs are Self-Esteem, Belongingness, Control, and Meaningful Existence; viable studies provided a general needs composite or at least one of these dimensions. When possible, these were coded independently and a Needs composite, or the average across available categories combined with general needs, was used as the main composite variable. These measures included both Baumeister’s 12-item Needs scale and specific subscales, along with researcher-generated items addressing global (e.g., “I belong.”) and contextual (i.e., “I belong to this group/in the game.”) needs. Positive values reflect greater Threats to Psychosocial Needs (i.e., less satisfaction of Psychosocial Needs) for excluded participants.
Self-Reported Concepts of Self (Self-Esteem): Threats to Self-Esteem is also investigated separately from the Needs composite. As the most frequently cited variable, Self-Esteem was also entered into the analyses to assess its effects independent of the other three needs. Although many of the analyses refer simply to the composite variable “self-esteem,” this variable includes established self-esteem scales (e.g., Rosenberg 1965) and single- or multi-item measures assessing self-ratings or concepts of self (e.g., “I feel badly about myself.”). Positive values reflect greater Threats to Self-Esteem for excluded participants.
Self-Reported Ratings of the Exclusion Source(s): Many studies asked participants to rate their exclusion source(s). These measures featured single- and multi-item measures created by the researcher and addressed general personal feelings towards the exclusion sources, as well as positive qualities like intelligence. Positive values reflect more negative ratings of the exclusion source(s) for excluded participants. The term “exclusion source” is used interchangeably with “excluders” in pooled analyses (i.e., Ostracism and Rejection); however, when the conditions are investigated separately, the terms “ostracizer” and “rejector” are used.
Aggression: This behavioral measure included any post-exclusion hostile behavior. These measures did not include any measure of control enhancement, but instead focused on the participant’s negative actions directed towards others and implicit measures of aggression or anger (e.g., hot-sauce allocation, noise blast). Positive values reflect greater Aggression for excluded participants.
Belongingness Behaviors: This behavioral measure including any positive behavior directed towards others post-exclusion (e.g., allocated money, completion of group relevant words on a fragment task). This meta-analysis did not include an explicit coding of their ability to neutralize threatened belongingness needs, but is employed to address some of Gerber & Wheeler’s (2009) findings regarding belonging-specific behavior. Negative values indicate lower belongingess behaviors.
Effort on Follow-Up Tasks(*): This behavioral measure assesses the extent to which a participant exerted effort on an individual task post-exclusion; these tasks did not affect the group or the exclusion source and included effort on Cognitive Tasks (e.g., number of math problems attempted) and Self-Regulatory Behavior (e.g., persistence on a given task, consuming a bitter beverage). Unlike Gerber & Wheeler (2009), these studies were not coded for their ability to enhance the participant’s control, but were simply defined as any task that required effort by the participant independent of a social context. Positive values reflect post-exclusion lethargy (i.e., less effort, less regulation of one’s behavior) for excluded participants.
Cognitive Loss (Diminished Cognitive Performance): This behavioral measure includes performance on various post-exclusion cognitive tasks (e.g., percentage of math problems answered correctly). Similar to Effort, these tasks have no social consequences. However, cognitive performance is conceptually different from effort; it denotes both cognitive effort and attention. Positive values indicate greater cognitive failure, or lower assessment scores for excluded participants.
Calculation of Effect Sizes
Cohen’s d was used as an effect size index; this value is the difference between two groups divided by their pooled standard deviation (SD), then corrected for a small sample bias (Hedges & Olkin, 1985). Scores were calculated using Microsoft Excel macros provided by David B. Wilson (available here: http://mason.gmu.edu/~dwilsonb/ma.html), and each d was estimated using condition means, F-score for main effects, F-score accompanied by means and standard deviations, or a Chi Squared statistic. For studies that did not supply statistics but reported no significant differences between conditions, a conservative procedure of estimating the effect size as 0 was adopted. Positive effects denote an increase in behavior in the expected direction (e.g., greater values indicate more exclusion-induced distress like threatened needs and worsened mood).
A random effects model was employed for these analyses instead of a fixed-effect model. In a fixed effect model, which assumes that there is a single effect size present within the larger population, “effect size estimates differ only as a result of sampling variability” (Hedges & Olkin, 1985, p 189); alternatively, the random effects model accounts for variations in the differences in the sampling population. This idea, that “characteristics of a study may influence the magnitude of its effect size” (Hedges & Olkin, 1985, p 189), has been investigated by several researchers. Cronbach argued that the evaluation of studies should address difference in treatment site or study (Cronbach, 1980), thereby affecting a experiment’s ability to replicate prior results. Analyses of homogeneity indicate significant differences in effect sizes across studies, thus demanding the use of a random effects model.
Effect size weights were calculated using the Standardized Mean Difference formula provided by the formula 1/se2 (Hedges). Analyses were performed using Wilson’s meta analysis macros for SPSS (available here: http://mason.gmu.edu/~dwilsonb/ma.html).
Statistical Analyses (ANOVAs and Regression)
The homogeneity statistic, Q, failed to demonstrate consistency between effect sizes, indicating that variability might be explained using the between study factors listed above. Therefore, we calculated fixed-effect categorical models to determine their influence on effect size; “grouping of studies according to their characteristics is an essential step in assessing the range of generalizability of a research finding” (Hedges & Olkin, 1985, p 148). When significant, between class effect (Qb) is reported. One-way ANOVAs investigating categorical, or class, differences and regressions between variables were conducted using a Methods of Moments Random Effects Model.
Reported Exclusion (k = 29): 34% of studies asked the participants how excluded or how included they felt. These were combined to create one composite measure of Reported Exclusion (or the negative of Reported Inclusion) which exhibited an extremely large effect of exclusion compared to acceptance (d = 2.4172). This indicates that the exclusion inductions employed by the studies successfully elicited self-reported feelings of exclusion.
Threats to Psychosocial Needs (k = 33): 39% of studies reported at least one of the Psychosocial Needs (i.e., belongingness, control, self-esteem, and meaningful existence). Across studies, exclusion exhibited a large effect on Psychosocial Needs (d = 1.3235). Furthermore, independent analyses of each individual need resulted in very large effects on control (d = 1.5195, k=11), belonging (d = 1.5186, k=12), and meaningful existence (d = 1.5186, k=11). 21% of studies reported either a single- or multi-item measure of Self-Esteem and demonstrated a large effect of exclusion (d = 0.8082).
Mood (k = 47): The most frequent dependent variable was Mood, which was reported in 54% of studies. Exclusion exhibited a moderate effect on Mood (d = 0.6651) and independent analyses revealed that social exclusion had a greater effect on positive mood (d = 0.5475, k = 36) than negative mood (d = 0.4823, k = 25), replicating Gerber & Wheeler’s (2009) findings.
Follow-Up Effort (k = 16): 19% of studies reported Effort on Follow-Up Tasks, or Follow-Up Effort. Across studies, exclusion exhibited a large effect on Follow-Up Effort (d = 1.1122), such that excluded participants demonstrated more lethargy (i.e., less effort) post-exclusion compared to included participants.
Secondary variables (i.e., k ² 15) were also investigated (See Table 1 for Mean Effect Sizes). Exclusion demonstrated moderate effects on Aggression (d = 0.7629) and on Cognitive Loss (d = 0.6783) and a large effect on Ratings of the Exclusion Source(s) (d = 0.9485. All Qs were significant, thus confirming the planned use of a Random Effects model.
Table 1: Mean Effect Differences (d)
|Dependent Variable||Effect Size|
(k = 29)
|d = 2.4172; CI = 1.85, 2.98|
(k = 46)
|d = 0.6651; CI = 0.52, 0.81|
|Threats to Psychosocial Needs
(k = 33)
|d = 1.3235; CI = 0.91, 1.74|
|Threats to Self-Esteem
(k = 18)
|d = 0.8082; CI = 0.57, 1.04|
|Negative Rating of the Exclusion Source(s)
(k = 15)
|d = 1.0605; CI = 0.66, 1.46|
(k = 11)
|d = 0.7629; CI = 0.51, 1.01|
|(Lack of) Follow-Up Effort (i.e., Lethargy)
(k = 16)
|d = 1.1122; CI = 0.70, 1.52|
(k = 12)
|d = 0.6783; CI = 0.29, 1.07|
*95% Confidence Interval [CI]
*All Qs were significant therefore homogeneity of effects was rejected (i.e., all dependent variables exhibited heterogeneity in the original analysis); this is not surprising given the varying methodologies used in this analysis.
*All ps were highly significant (p<.001).
A series of regressions were conducted to investigate the relationship between publication details (i.e., Year and Impact Factor) and exclusion-induced distress or post-exclusion coping behavior. Year was significantly correlated with Reported Exclusion, more recent studies demonstrated greater Reported Exclusion (ß = 0.3948 (p = .028) R2 = 0.1558), as well as a reduced effect on Mood (ß = .3367 (p = .0108) R2 = .1133) such that more recent journals reported greater feelings of exclusion and stronger negative moods (or less positive moods) post-exclusion. Alternatively, Impact Factor was significantly correlated to Follow-Up Effort (ß = -0.7124 (p = .012) R2 = 0.5076); more impactful journals exhibited smaller effects on post-exclusion lethargy (i.e., the Follow-Up Effort demonstrated by excluded participants was similar to included participants in more impactful journals).
Average Age of Participants: 40 studies (47%) reported the average age of their participants resulting in a total data set mean of 19.827 (SD = 2.566) indicated a generally young subject pool. The mean age of participants demonstrated no significant effects on any of the variables of interest.
Percent Female Participants: 68 studies (80%) reported the gender composition of their participants, reporting a mean of 63.9 (SD = 0.168) percent female participants. Studies that reported more female participants reported less Aggression (ß = -0.5976, p = .030, R2 = 0.356, k = 10), and smaller effects on Ratings of the Exclusion Source(s) (ß = -0.5307, p = .033, R2 = 0.2817, k = 13). The effect of percent female participants on post-exclusion Mood approached significance (ß = .2867, p = .0607, R2 = .0822, k = 41) such that a greater percentage of female participants was correlated with larger effects on Mood (i.e., larger differences between excluded and included participants).
Percent White Participants: 27 studies (32%) reported the racial composition of their participants, reporting mean of 0.713 (SD = 0.110) percent White participants. The percentage of White participants in a given study was correlated with Aggression (ß = 0.8257 (p = .004) R2 = 0.6817, k = 5) and Cognitive Loss (ß = 0.4376 (p = .011) R2 = 0.1915, k = 7); studies that reported a high percentage of White participants also reported larger effects on Aggression and greater detriments to Cognitive Loss.
H1: Social Exclusion (general) encompasses three conditions: Ostracism, Rejection, and Social Exclusion. Each condition will demonstrate different effects on exclusion-induced distress and post-exclusion coping behavior.
Table 2, a one-way ANOVA by condition, provides the mean effect sizes for each of the methodological groupings by dependent variable of interest. A review of the means reveals major differences between these groups. It is important to note that this grouping by condition (i.e., Ostracism, Rejection, and HSE) achieved homogeneity for almost all of the variables; only Ostracism’s effect on Mood continues to demonstrate heterogeneity (p = .0482).
Table 2: One-Way ANOVA by Exclusion Condition
|Hypothetical Social Exclusion (HSE) k=31 (36.5%)|
|Reported Exclusion (p = .0643)||d = 2.7473 ( k = 21) pQ = .8289||d = 1.5535 (k = 8) pQ = .5604||No available studies|
|Neg Mood (p = .0744)||d = 0.8186 (k = 16) pQ = .0482||d = 0.4172 (k = 13) pQ = .9116||d = 0.7044 (k = 17) pQ = .9891|
|Threats to Psychosocial Needs (p = .0812)||d = 1.6316 (k =22) pQ = .7530||d = 0.8274 (k = 8) pQ = .7151||ns (d = 0.3661) (k = 3) pQ = .9427|
|Threats to Self-Esteem (p = .1911)||d = 0.9817 (k = 10) pQ = .4981||d = 0.6976, (k = 6) pQ = .5290||ns (d = 0.2484) (k = 2) pQ = .5618|
|Neg Rating of Exclusion Source(s) (p = .0084)||d = 0.6431 (k = 8)||d = 1.5444 (k = 7)||No available studies|
|Aggression (p = .4255)||d = 0.8320 (k = 3)||d = 1.4500 (k=1)||d = 0.6847 (k = 7)|
(p = .1523)
|ns (d = 0.5629)
(k = 2)
|d = 1.1567 (k = 6)||d = 1.8700 (k = 1)|
|Lack of Effort/Lethargy
(p = .2865)
|d = 1.6598 (k = 3) pQ = .1040||ns (d = 0.7992) (k = 2) pQ = .7428||d = 1.0067 (k = 11) pQ = .9249|
(p = .0011)
|ns (d = 0.0580) (k=1)||ns (d = 0.2306) (k = 2)||d = 0.9771 (k = 9)|
Planned comparisons (see Table 3) revealed several significant difference between the conditions. Ostracism elicited a significantly greater effects on Reported Exclusion when compared to Rejection (no HSE studies reported a manipulation check), and greater Threats to Psychosocial Needs when compared to Rejection and HSE combined (p = .0291). Rejection elicited significantly more negative Ratings of the Exclusion Source(s) compared to Ostracism (p = .0084). However, Rejection elicited a significantly smaller effect on Mood compared to the other two conditions. It has been shown that attributions for the exclusion can moderate its effects (Snoek, 1962), but Ratings of the Exclusion Source(s) did not moderate the effect on Mood. HSE demonstrated a significantly greater effect on Cognitive Loss compared to Ostracism and Rejection ( p= .0001); however the small number of studies employing Ostracism or Rejection that included this variable (k = 3) makes these findings difficult to interpret.
Table 3: Planned Comparisons
|DVs||Ostracism vs (Rejection & HSE)||Rejection vs (Ostracism & HSE)||HSE vs (Ostracism & Rejection)|
|Reported Exclusion*||Q(1,27) = 3.4227, p = .0643||Q(1,27) = 3.4227, p = .0643||ns|
|Negative Mood||Q(1,44) = 2.8907, p = .0891||Q(1,44) = 4.8856, p = .0271||ns|
|Threats to Psychosocial Needs||Q(1,31) = 4.7600, p = .0291||ns||ns|
|Threats to Self-Esteem||ns||ns||ns|
|Neg Rating of Exclusion Source(s)*||Q(1,13) = 6.9422, p = .0084||Q(1,13) = 6.9422, p = .0084||ns|
|Lack of Effort (Lethargy)||ns||ns||ns|
|Cognitive Loss||ns||Q(1,10) = 7.4700, p = .0063||Q(1,10) = 15.3696, p=.0001|
*Reported Exclusion and Rating of the Exclusion Source were only present in the Ostracism and Rejection conditions.
Furthermore, a one-way ANOVA between conditions on the Threats to individual Psychosocial Needs demonstrated a significant difference for Meaningful Existence only. Ostracism reported a significantly greater effect on Meaningful Existence (d = 1.7378 [CI] = 1.1863, 2.2893, k = 9) than Rejection (d = 0.5477, [CI] = -0.6063, 1.7017, k = 2), Q(1,9) = 3.326, p = .0682. This may be related to the acknowledgment associated with Rejection compared to the complete disregard for the participants presence required by Ostracism. This difference will be addressed further in the discussion section.
Many of these analyses suffer from an imbalance of reporting across studies; Mood was reported by 46 studies and evenly distributed across the three manipulation groups, allowing for a more robust analysis. The difference between these three groups approached significance (p = .0744) and planned comparisons in Table 3 revealed that Rejection elicited a significantly smaller effect on Mood compared to Ostracism and HSE combined (p = .0271); Ostracism elicited the greatest effect on Mood, and this difference approached significance (p = .0891). For HSE, all variables were regressed to Mood, the most frequent dependent variable available in this condition (k = 17), but no significant predictors emerged.
H1A: According to the Model of Ostracism, Threats to Psychosocial Needs will moderate the relationship between Reported Exclusion and Mood, as well as post-exclusion coping behavior.
Across studies, regressions revealed that Reported Exclusion predicted a large percentage of the variance in Threats to Psychosocial Needs (ß = 0.8486, p = .000, k = 20, R2 = .7201) including Self-Esteem (ß = 0.4511, p = .0530, k = 14, R2 = .2035), and Mood (ß = 0.6428, p = .0050, k = 16, R2 = .4133). Furthermore, Threats to Psychosocial Needs was significantly correlated with Mood (ß = 0.4375, p = .0457, k = 21, R2 = .1914) and the predictive effect of Reported Exclusion on Mood became non-significant (p = .3409) when controlling for Threats to Psychosocial Needs (ß = 0.6195, p = .0065, k = 13). This provides evidence for the Model of Ostracism (Williams & Sommer, 1997), where Mood is an outcome of exclusion’s Threats to Psychosocial Needs.
Also in line with the Model of Ostracism, the effect of Threats to Psychosocial Needs approached significance for Follow-Up Effort (ß = 0.8792, p = .0650, k = 3, R2 = .7730), such that greater Threats to Psychosocial Needs was correlated with less Follow-Up Effort. The effect of Self-Esteem on Rating of the Exclusion Source(s) also approached significance (ß = 0.5854, p = .0978, k = 7, R2 = .3426) such that greater negative effects on Self-Esteem were correlated with greater negative Ratings of the Exclusion Source(s), but Reported Exclusion did not have a direct effect on these post-exclusion coping behaviors. Together, these findings reveal that Threats to Psychosocial Needs affects Follow-Up Effort or post-exclusion lethargy, but an insufficient k made it impossible to assess whether Threats to Psychosocial Needs moderates the relationship between Reported Exclusion and Follow-up Effort.
When the conditions were assessed independently, the aforementioned moderation was only significant for the Ostracism condition. Threats to Psychosocial Needs moderated the effect of Ostracism-induced Reported Exclusion on Mood. Rejection studies continued to exhibit an effect on Mood even when controlling for Threats to Psychosocial Needs (this will be explored further in H1B). Furthermore, the relationship between Needs and Follow-Up Effort as well as Self-Esteem and Ratings of the Exclusion Source(s) were no longer significant for studies that employed Ostracism. However, Reported Exclusion exhibited a significant effect on Ratings of the Exclusion Source(s) in Ostracism studies (ßRExc = 0.9246, p = .0013, k = 4, R2 = .8549), but this effect was no longer significant when Threats to Psychosocial Needs was entered with Reported Exclusion.
H1B: In line with the Numbness Hypothesis, Reported Exclusion will have a direct effect on Mood independent of Threats to Psychosocial Needs, and Mood will moderate exclusion’s effect on a “deconstructed state.”
As mentioned above, there was no effect of Reported Exclusion on Mood across studies when controlling for Threats to Psychosocial Needs; however, when split by condition (i.e., Ostracism vs. Rejection (vs. HSE)), Rejection-induced Reported Exclusion significantly predicted Mood (ß = 0.9835, p = .000, k = 5, R2 = .9673), and this effect was even stronger when controlling for Threats to Psychosocial Needs (ß = 1.0036, p = .0000, k = 5, R2 = .9720). These analyses reveal that Rejection has a significant effect on Mood independent of Threats to Psychosocial Needs, a relationship not present in Ostracism studies. These findings may be interpreted as exclusion causes an overall negative effect on Mood, or that acceptance causes an overall positive effect on Mood; this will be addressed further in the Discussion section. In addition, Ratings of the Rejector(s) (i.e., exclusion source for the rejection condition) were only predicted by Threats to Self-Esteem (ßSE = 0.8043 p = .0207, k = 5, R2 = .6469), not Reported Exclusion as seen in the Ostracism condition. These findings indicate that Rejection is a special case of Exclusion, where Reported Exclusion has a significant effect on Mood even after controlling for Threats to Psychosocial Needs, but Threats to Psychosocial Needs (specifically Self-Esteem) moderate the effect of Reported Exclusion on post-exclusion Ratings.
It was impossible to investigate the Numbness Hypothesis with HSE as none of these studies that employed HSE assessed (self) Reported Exclusion. All of the available variables were regressed to Mood for HSE, but no significant relationships emerged.
H2: Situational differences in exclusion will effect exclusion-induced distress and post-exclusion coping behavior.
Ostracism studies excluded participants via a conversation or an interactive ball tossing game and both of these methods were employed with real (individuals who excluded the participant) and computerized confederates (participants were excluded by a computer program emulating a conversation or ball tossing game). There were no differences between studies that employed an interactive game and those that employed a conversation scenario on any of the major DVs; however, a significant difference emerged for post-exclusion Rating of the Exclusion Source(s). Ostracism studies that employed a conversation scenario demonstrated significantly more negative ratings of the ostracizer(s) (d = 1.0292, k = 3) compared to studies that employed the ball toss scenario (d = 0.2886, k = 3), Q(1,4) = 20.7921, p = .0000. It is important to note that the conversation scenario elicited a greater threat to Self-Esteem (d = 1.2886) compared to the interactive game (d = 0.8033), similar to the findings of Gerber & Wheeler (2009) but this difference was non-significant (p = .1793).
Rejection studies excluded participants by informing them that they were not selected by an individual or group. These analyses will be further investigated in H2A.
Hypothetical Social Exclusion (HSE) studies induced exclusion by informing the participants that their personality type made them prone to a future without close social ties; some studies provided this information immediately after the participant completed the bogus personality test, while other studies offered the participant a correct diagnosis of their extroversion style in order to increase believability of the deception. Planned comparisons revealed no differences between these two methods on the major DVs. However, a significant difference emerged in Cognitive Loss (Q(1,7) = 7.2937, p = .0069); studies that employed an extroversion-specific manipulation exhibited smaller effects on Cognitive Loss (i.e., excluded participants performed similarly to accepted participants; d = 0.0690, ns, k = 1) compared to studies that did not employ the participant’s extroversion score (d = 1.091, k = 8).
H2A: Greater numbers of exclusion sources will result in greater effects on exclusion-induced distress and post-exclusion behavior.
For studies that employed Rejection, significant differences in Aggression emerged between group-rejection and individual-rejection; studies utilizing group rejection reported significantly more post-exclusion Aggression compared to studies where participants were rejected by a single individual (p = .0023) (See Table 4). The effect on Rating of the Rejector(s) was reversed, individual rejection elicited more negative Ratings of the Rejector(s), and this difference approached significance (p = .0966). This one-way ANOVA could not be conducted with the other two conditions, Ostracism and HSE; Ostracism requires the interaction of others to induce feelings of exclusion, and therefore cannot be induced with a single exclusion source, and HSE visualizes larger social exclusions due to individual characteristics.
Table 4: Individual vs. Group Rejection
|DVs||Individual Rejection||Group Rejection|
|Aggression Q(1,5) = 9.3014,
p = .0023
|d = .4297 (k = 4)||d=1.0199 (k = 3)|
|Rating of the Rejector Q(1,5) = 1.7603 p = .0966||d = 1.8541 (k = 5)||d=.7359 (k = 2)|
Across Ostracism and Rejection studies, regression analyses revealed that the raw number of exclusion sources exhibited a significant effect on the participants’ Ratings of the Exclusion Source(s) (ß = 0.4441, p = .052, k = 15, R2 = .1973), such that more exclusion sources elicited less negative post-exclusion Ratings (this relationship exhibited less significance (p = .054) when the number of exclusion sources was aggregated into groups). The raw number of sources predicted Follow-Up Effort, and this relationship approached significance (ß(1,3) = 0.7531, p = .0692, R2 = .5672); more exclusion sources was correlated with less lethargy (i.e., greater effort on follow-up tasks), indicating that exclusion by larger groups may encourage the exclusion target to engage in more effortful behavior. This effect became non-significant when controlling for the percent of female participants, meaning that this social facilitation elicited by large group exclusion may be a female-specific method of reengaging with the group, as demonstrated in earlier research (Williams & Sommer, 1997). Too few studies assessed Rating of the Exclusion Source(s) and Follow-Up Effort to conduct the necessary moderation analyses. In addition, the size of the group (i.e., aggregated number of exclusion) predicted Aggression across studies, and this relationship approached significance (ß(1,8) = 0.5710, p = .063, R2 = .3261) such that more exclusion sources resulted in more Aggression, thus duplicating the findings of individual vs. group rejection. This effect was not present in the regression analysis using the raw number of exclusion sources.
Figure 2 demonstrates a trend emerging across variables for small groups. Effects on Reported Exclusion and Psychosocial Needs peaked at 2-3 exclusion sources, and planned comparisons revealed that this peak approached significance (p = .0611, .0851). Follow-Up Effort was only reported for studies that employed Individual and Small Group Exclusion and is not featured in Figure 2.
H2B: The gender composition of the exclusion source(s) will exhibit an effect on exclusion-induced distress and post-exclusion coping behavior.
27 studies (31.6%) reported the gender of the exclusion source(s); these were coded as either same gender (k = 18), mixed gender (k = 6), and opposite gender (k = 3). 9 studies (10.6%) did not report the gender of the exclusion source, and the remaining 49 studies (57.6%) employed imaginary gender sources; these included studies that employed chat rooms and Cyberball where the identifiers were gender-neutral (e.g., initials, animated characters).
Due to this bias across studies, drawing conclusions from the results was difficult. Mood was the only dependent variable that demonstrated a main effect of gender; studies that featured opposite sex exclusion sources elicited significantly greater negative Mood compared to studies where participants were rejected by members of the same sex (Q(1,10) = 7.5533, p = .006) (See Table 5). This finding indicates that perhaps same-gender exclusion sources offer a more “clean” measure of exclusion effects, thereby further reducing the effect of exclusion on Mood. Furthermore, studies that featured opposite gender exclusion sources reported more positive Ratings of the Exclusion Source(s) (d = -0.4059, k = 2) compared to studies that excluded participants with members of their own gender (d = 0.9206, k = 5), although these analyses suffered from a comparatively low k.
Regressions employing gender composition as the independent variable also revealed a significant effect on Mood (ß = -0.7234; R2 = .5233 (p = .001) k = 14); more members of the opposite sex in the excluding group led to greater negative affect (See Table 5).
Table 5: Mood by Gender Composition of Exclusion Source(s)
|Same Gender||Mixed Gender||Opposite Gender|
|Mood||d = .3729 [CI] = -.6180, -.1279, k = 10||d = -.9544 [CI] = 1.4965, -.4124, k = 2||d = 1.2122 [CI] = -1.7437, -.6787, k = 2|
H2C: The level at which the individual knows and interacts with their exclusion source(s) will moderate the effects on exclusion-induced distress and post-exclusion coping behavior.
Three study characteristics were employed to investigate this hypothesis: (1) interaction time, the length of time the participant interacted with the rejectors, on the variables of interest, (2) types of interactions prior to or during exclusion, and (3) presence of the exclusion source.
(1) Interaction Time: Ostracism and Rejection studies were assessed separately. The interaction time for Ostracism studies indicated the length of time that the participant was excluded (i.e., minutes spent engaging with, or excluded from, an interactive game or conversation) (M = 5.4375, SD = 2.412) and was not significantly correlated with any of the dependent variables . The interaction time for Rejection studies indicated the length of time that the participant interacted with the rejectors prior to being rejected (M = 11.6, SD = 6.85) and was not significantly correlated with any of the dependent variables.
(2) Knowledge of the exclusion source was grouped into four categories: (1) no knowledge of the source, (2) demographic information about the source (e.g., gender, university or political affiliation), (3) brief written or videotaped biographical sketch (i.e., no interaction but increased details about the source), and (4) actual interaction, which included both interpersonal interaction, and chat rooms. The average standard mean differences are plotted in Figure 3.
Aggression was significantly different between groups (Q(3,6) = 19.8678, p = .0002, k = 10); no knowledge of the exclusion source and interpersonal interaction with the source elicited more Aggression (d = 1.1486, 1.0991 respectively) than demographic or bio-sketch knowledge (d = 0.2911, 0.5573 respectively). Rating of the Exclusion Source was also significant between groups (Q(3,11) = 8.6863, p = .0338, k = 15); no knowledge of the exclusion source elicited significantly less negative ratings (d = 0.4122) compared to demographic, bio-sketch, or interpersonal interactions with the source (d = 1.2095, 1.9444, 1.1007).
(3) Presence of the exclusion source was measured using two categorical variables: online vs. real exclusion sources and imaginary vs. tangible exclusion sources), and emerged as a significant moderator of several dependent variables. Online studies that excluded participants through technology (e.g., chat room, virtual game, text messages) reported greater effects on Reported Exclusion (p = .0486) and greater Threats to Psychosocial Needs (p = .0336) compared to studies that excluded participants via some interpersonal interaction (e.g., participant is informed about rejection by a the experimenter) (See Table 6). Online studies also elicited greater Aggression, and this relationship approached significance (p = .0877). Studies conducted through technology seemed to elicit greater exclusion-induced distress, and this was not moderated by the average age of the participants, despite a young overall subject pool.
Table 6 : Online vs. Other Exclusion Sources (ANOVA)
Q(1,27) = 3.8905, p = .0486
|d = 2.8998, k=17||d = 1.7350, k=12|
|Satisfaction of Needs
Q(1,31) = 4.5168, p = .0336
|d = 1.6872, k=19||d = 0.8256, k=14|
Q(1,9) = 2.9163, p = .0877
|d = 1.3210, k=1||d = 0.6790, k=10|
Similar to the effect of online exclusion, studies that utilized imaginary exclusion source(s) (i.e., the participant had no physical interaction with, or observation of, an individual/group who would later exclude them) reported greater effects on Reported Exclusion (p = .0004) and Mood (p = .0669) compared to tangible exclusion source(s) (i.e., the participant had some physical interaction with an individual/group who would later exclude them), indicating that imaginary exclusion sources (presented via written or filmed biographical sketches) elicited greater self-reported feelings of exclusion (See Table 7). However, this trend is not consistent with respect to the Rating of the Exclusion Source(s); tangible exclusion source(s) elicited stronger negative ratings, and this difference approached significance (p = .0783).
Table 7: Imaginary vs. Tangible Exclusion Sources (ANOVA)
|DVs||Imaginary Exclusion Source(s)||Tangible Exclusion Source(s)|
|Reported Exclusion Q(1,27) = 10.5811, p=.0011||d = 3.0368, k = 19||d = 1.2412, k = 10|
|Mood Q(1,44) = 3.3575, p = .0669||d = 0.7583, k = 32||d =0 .4679, k = 24|
|Cognitive Performance Q(1,10) = 15.3696 p = .0001||d = 0.9694, k = 9||ns (d = -0.133), k = 3|
|Rating of the Exclusion Source(s) Q(1,13) = 4.8890 p= .0783||d = 0.7002, k = 7||d = 1.3794, k = 8|
Differences between Ostracism, Rejection and HSE
The findings reveal that there is a difference between Ostracism, Rejection, and Hypothetical long-term Social Exclusion (HSE, or Future Rejection) in terms of their effects and mechanisms. A review of the means across dependent variables reveals differences in effect size; three of the four major composite dependent variables exhibited differences between conditions that approached significance (p².08). Ostracism elicited large effects on Self-Reported Exclusion, negative Mood post-exclusion, and greater Threats to Psychosocial Needs. This large effect of Ostracism compared to the other conditions may be explained by the greater salience of the exclusion in this condition; participants are forced to endure their exclusion over a period of time and remain in the presence of their excluders. The Rejection condition simply informs participants of their rejection; some studies have assessed the effect of expecting a follow-up interaction with the exclusion source (Buckley, Winkel, & Leary, 2004), but no studies actually force the participant to experience this rejection over time. In real interpersonal interactions, ostracism and rejection are rarely mutually exclusive. Further analyses are required to compare the two conditions and assess their combined effect.
The secondary dependent variables demonstrate more nuanced differences between the groups. Rejection elicited greater negative effects on the Ratings of the Exclusion Source(s), despite the overall trend of Ostracism as the most distressing condition. In Rejection studies, participants may be retaliating against an individual who previously excluded them.
Across studies, there is evidence for Williams & Sommer’s Model of Ostracism (1997), which states that Threats to Psychosocial Needs moderate the effect of Reported Exclusion on exclusion-induced distress, Mood is considered a dependent variable in this model, and post-exclusion coping behavior. Reported Exclusion was highly correlated with Threats to Psychosocial Needs, and Threats to Psychosocial Needs predicted post-exclusion coping behavior including Follow-Up Effort and Ratings of the Exclusion Source(s). However, when the conditions are investigated separately, this relationship was only significant for studies that employ Ostracism. Furthermore, when investigating Ostracism studies, Threats to Psychosocial Needs moderated the relationship between Reported Exclusion and Mood. These findings support the Model of Ostracism, but demonstrate that it may not be applicable to the mechanisms of Rejection.
Alternatively, Twenge, Cantanese, & Baumeister (2003) propose the Numbness Hypothesis, which states that exclusion induces a “deconstructed state” characterized by diminished arousal, a numbing of emotions, both positive and negative, lethargic behavior, and reduced cognitive functioning. This theory implies that exclusion has an effect on Mood, and therefore the comparison of exclusion to acceptance should reveal an effect independent of Psychosocial Needs. Although the effect of Reported Exclusion on Mood was not significant across studies after controlling for Threats to Psychosocial Needs, an analysis of Rejection studies tells a different story. For studies that employed Rejection, Reported Exclusion predicted Mood, and the effect was even stronger when controlling for Threats to Psychosocial Needs. Furthermore, the regression analyses in this meta-analysis reveal that the effect of Mood (independent of Threats to Psychosocial Needs i.e., controlling for Threats to Psychosocial Needs) is specific to Rejection only, indicating that Rejection causes a negative shift in Mood (Gerber & Wheeler, 2009), or inclusion causes a positive shift in Mood (Baumeister, DeWall, Vohs, 2009). Similar analyses could not be conducted on the HSE condition; none of these studies provided a measure of Reported Exclusion thus making it impossible to assess the role of Mood when thinking about a future alone. This is particularly troubling as Baumeister insists that the method of HSE or Future Rejection will induce mood numbness. However, none of the available variables demonstrated a correlation with Mood. Although there is evidence that exclusion may cause a “deconstructed state,” further research is required to assess the mechanisms within this black box definition. The current analyses reveal that Rejection has a direct effect on Mood that is not present in Ostracism, but baseline mood measures are required to properly investigate the Numbness Hypothesis.
In addition, Baumeister et al. claim that this deconstructed state also affects other variables including post-exclusion effort and cognitive functioning, independent of the effect on Mood. Although the Numbness Hypothesis is supported by these results, post-exclusion behavior may not be due to the deconstructed state alone; Threats to Self-Esteem moderated the effect of Reported Exclusion on Ratings of the Exclusion Source, indicating a more complex mechanism present in Rejection’s effect on post-exclusion behavior.
The conflicting mechanisms of Ostracism and Rejection demonstrate that, although exclusion can be felt through several methods, different experiences elicit different dimensions of exclusion-induced distress. These differences indicate that “Reported Exclusion” is a general term that measures different constructs in each of these conditions. Ostracism makes the existing group dynamic salient by forcing the participant to watch the group’s interaction, while the studies in the Rejection condition only allude to a group that is never available to the participant. Despite Ostracism’s relationship between Reported Exclusion and Threats to Psychosocial Needs, only the immediate Reported Exclusion variable predicted variation in the Ratings of the Ostracizers. This result may be attributable to the lack of interaction between the participant and the exclusion sources; thus causing the exclusion experience to have a direct effect on the Rating of the Exclusion Source(s). However, Figure 3 reveals that Ratings of the Exclusion Source increase with greater knowledge, from no knowledge to bio-sketch/video, but drops with interpersonal interaction. Unfortunately, the mechanisms of HSE are harder to analyze without the explicit manipulation check of Reported Exclusion. Further research is required to address the mechanisms and internal effects of a future alone.
Situational Differences in Exclusion
According to Williams’ 2007 review of Social Exclusion, situational differences have a significant effect on post-exclusion coping behavior, but very little research has investigated the role of situational differences on exclusion-induced distress and post-exclusion coping behavior. This meta-analysis reveals that being ostracized from a conversation elicits stronger negative effects on Ratings of the Ostracizers compared to an interactive game; this may also be related to increased knowledge of the exclusion sources acquired while listening to the conversation.
The size of the excluding group is an important component of exclusion-induced distress and coping behaviors; however, no single study was found that directly compared individual and group-exclusion. Across studies, Aggression was the only dependent variable that exhibited a linear relationship with the number of exclusion sources; more exclusion sources result in greater post exclusion aggression. However, the effect on Rating of Exclusion Source(s) was reversed; in the Rejection condition, participants rated an individual rejector as more negative than a group (raw number of exclusion sources was non-significant). These disparate effects indicate that these items may not be related. Aggression may have both targeted (i.e., inflicting pain on the Exclusion Source) and global (e.g., trait-displaced aggression) effects. Furthermore, research has shown that “people aggress to regain some sense of control (or the power to control)” (Gerber & Wheeler, 2009 p 471); if aggression is interpreted as behavior designed to enhance control, the results replicate the findings of Gerber & Wheeler (2009).
There is also an decrease in lethargy as the number of exclusion sources increases, and this effect is moderated by the percentage of female participants in a given study. As mentioned in the results, this may indicate that exclusion by larger groups encourages the targets of the exclusion to socially facilitate, but this effect may be specific to women. Furthermore, behaviors coded as Follow-Up Effort include measures that Gerber & Wheeler (2009) identify as behaviors designed to increase belongingness and control post-exclusion (e.g., mimicry and consuming an unpleasant beverage); therefore these increases may be related to a specific psychosocial need. Additional coding is required to separate these behaviors and their relationship to psychosocial needs; additional analyses may reveal more effort for behaviors specifically designed to increase belongingness and control.
In addition, Figure 2 reveals a peak in exclusion-induced distress in the small group condition (i.e., when small groups are operationalized as 2-3 exclusion sources) on Reported Exclusion and Threats to Psychosocial Needs. Coding behaviors for their ability to increase belongingness and control may produce a different pattern than the peaks seen in Figure 2. Prior research by Asch demonstrated that at least three confederates were required to elicit stable conforming behavior, thereby establishing a social precedent, or rule of three. Asch’s research implies that a plateau should appear between small and large groups, indicating stable threats to belongingness needs for groups. Additional analyses using a more specific coding of behavior may reveal that belongingness behaviors exhibit no difference between small and large groups, while control behaviors may increase linearly similar to aggression. Ê
Not surprisingly, exclusion by an opposite gender source elicits significantly greater negative Mood. However, the drastic difference between the two classes is surprising; exclusion by a same sex source had a very small effect on Mood (d = 0.3722, k = 18), while the effect of exclusion by an opposite sex source was extremely large (d = 1.7569, k = 3). When these two classes are pooled, the effect on Mood is large (d = 0.6651), but this is deceptive. If exclusion by a same gender source is a clean measure of exclusion, then exclusion has very small effect on Mood, thus supporting Baumeister’s Numbing Hypothesis. Additional analyses using an aggregated percentage of gender composition (Same Gender, Mixed Gender, Opposite Gender) significantly predicted Mood, such that greater percentages of opposite sex exclusion sources elicit more negative Mood. Due to the fact that the research almost exclusively uses same-gender exclusion sources, these findings are tenuous, thus demanding additional research in this area.
No studies compared the effect of being excluded by a same gender or an opposite gender individual, and the large difference presented in Table 9 demonstrate that the use of same-gender exclusion sources may be dampening the effect of exclusion on mood. In addition, mixed-gender groups require more attention if laboratory studies are to be generalized to real life. It is also important to note that 2 of the 3 studies that employed opposite gender exclusion sources featured all female participants facing a job interview situation with a male interviewer (Stroebe et al. 2009). The third study (Ayduk, Gyurak, & Luerssen, 2008) featured two same-sex participants “competing” for the attention or selection of a potential opposite sex partner, thus confounding the effects of exclusion. Gender must be analyzed in a less confounding environment (e.g., Cyberball with gender-specific names for the confederates).
Interaction time was not a significant predictor of exclusion-induced distress or post-exclusion coping behavior in either Ostracism, where interaction time indicated the length of exclusion experience, or Rejection, where the interaction time indicated the extent of interaction prior to rejection. Although all of the studies used in this meta-analysis feature laboratory studies where the participants had no prior relationship with their excluders, it was hypothesized that different levels of interaction with the exclusion source(s) would demonstrate an effect on exclusion-induced distress and post-exclusion coping behavior. Aggregating studies into different levels of knowledge regarding the exclusion source(s) revealed interesting trends, but only Aggression and Rating of the Exclusion Source(s) demonstrated significant differences between groups; demographic knowledge of the exclusion source(s) elicited the least Aggression while biographical sketches of the exclusion source(s) elicited the most negative Ratings post-exclusion. Unfortunately, no studies feature exclusion from a group to which the participant reported a prior relationship. Some studies attempt to manipulate in- and out-group using demographic cues like smoker/non-smoker, Mac/PC user, and musical preferences, but the manipulation of group identification has demonstrated inconsistent results. These studies employ superficial groups, and additional research is required to assess the effect of prior identification with an excluding group.
“Leary and colleagues (e.g., Buckley et al, in press) have found mood effects after social rejection, but in their methods the participants do not meet their rejecters in person; this may make the experience less traumatic and thus not produce the defensive response of numbness” (Twenge, Cantanese, & Baumeister, 2003 p 414).
The results presented here demonstrate that this is not the case. Participants reported larger effects on Mood and Cognitive Loss when excluded by imaginary sources. However, these results may indicate that acceptance by imaginary sources has a greater positive effect on Mood. This may indicate that the interpersonal interaction that confounds tangible exclusion sources may moderate the effect of exclusion, or that exclusion by imaginary sources may be more distressing due to the participant’s imagination of the exclusion source. This is evident in the analysis provided in Figure 3, where the most minimal knowledge of the exclusion source results in the greatest effects on Reported Exclusion, while knowing more personal knowledge about the exclusion source (e.g., video/written biographical sketch) elicited the most negative post-exclusion ratings.
Overall, participants report greater Reported Exclusion and exclusion-induced distress when the exclusion sources are not present (i.e., imaginary and online exclusion sources). However, different results emerge when assessing post-exclusion coping behavior. Tangible exclusion sources elicit greater effects on negative Ratings of the Exclusion Source, indicating greater externalization of negative affect when there is a viable target. Alternatively, online studies elicit greater Aggression compared to studies where the participant was excluded through an interpersonal interaction. This may be a phenomenon associated with technology. Prior research has demonstrated that individuals are more likely to aggress in online environments due to the safety of their anonymity (Williams, Cheung, & Choi, 2000) and this finding may be further evidence of “virtual bravado.”
This meta-analysis sheds light on the different experiences of exclusion. Results demonstrate that Ostracism and Rejection may activate different psychological mechanisms related to Mood and Psychsocial Needs, resulting in different distress and coping behavior. Furthermore, situational differences (e.g., group size, gender composition of exclusion source) demonstrate effects on the exclusion experience, despite prior theories by Williams, which reside on major gaps in the research.
Additional methodological tactics must be addressed; these include unpublished data sets, alternative outlier procedures, and subjective coding.. The current meta analysis utilized only published studies, thus potentially biasing the results towards significant findings. Efforts must be made to collect and include unpublished data sets. Further analyses regarding the effect of the file drawer problem in social exclusion research. Outliers were winsorzied in this meta-analysis, but winsorizing with weights must be employed prior to finalizing these results including.
This analysis includes belongingness, control, and meaningful existence as self-report measures of Threats to Psychosocial Needs, and behaviors are coded as either Effort on an individual task (i.e., the task does not affect the group or the exclusion source, e.g., number of math problems attempted, ability to regulate desires), Aggression or aggressive acts towards the exclusion source, Cognitive Performance (i.e., performance on a cognitive task, independent of effort, e.g., percentage of math problems answered correctly), and Rating of the Exclusion Source (i.e., the participant’s opinions of their excluder). We believe that utilizing these categories over the subjective coding of control-restoring behaviors will provide a more objective analysis of exclusion-induced distress and post-exclusion coping behavior due to a concrete definition of behaviors and their correlates instead of depending on measures that have been shown to exhibit human error.
These analyses begin to address the findings of Gerber & Wheeler (2009) regarding mood, effort, and the alleviation of threatened needs. Regression analyses examined the effect of self reported threats to control and belonging on aggression and effort, but the Ns were not sufficient for the necessary analyses. Further coding is required to replicate Gerber & Wheeler’s theory that threats to control have a strong effect on aggression, or antisocial behavior, and that desire to mitigate belonging threats moderate these behaviors. This meta-analysis found that as studies increased the number of exclusion sources, participants demonstrated less lethargy, and this may be related to alleviating threats to belongingness needs. The number of studies that employed belongingness behaviors was not sufficient for many of the analyses and therefore cannot be discussed at length, although exclusion exhibited a significant effect on the overall composite variable. In order to replicating the findings of Gerber & Wheeler (2009), individual actions and behaviors coded in this meta-analysis as aggression, belongingness behaviors, and effort, must be recoded for the extent to which they allow the participant to regain feelings of control and belongingness.
However, as argued by Baumeister, DeWall, and Vohs (2009), this method is inherently flawed due to issues of affective forecasting; in general, people inaccurately predict the intensity and duration of emotions when asked to consider future events (Gilbert & Wilson, 2003). Although not included in the current analysis, 10 coders were asked to rate their emotional responses to the methods utilized in exclusion studies. These coders exhibited very high interrater reliability (ICC= .978); however, these coders rated interpersonal rejection as inducing the greatest reported exclusion, and Cyberball as the least distressing. Analysis of the effect sizes reveals the exact opposite trend, indicating that the error of emotional forecasting may have an effect here. Furthermore, Baumeister, DeWall, and Vohs (2009) also demonstrate faulty emotional forecasting in their assumption that imaginary rejectors will be less traumatic, indicating the need for additional work regarding the situational effects on exclusion-induced distress and post-exclusion coping behavior.
Future Directions for Researchers
Most importantly, the lack of consistent dependent variables across studies makes robust analyses difficult. The most glaring omission is the complete failure of studies that utilize Hypothetical Social Rejection (or Future Rejection) manipulations to include a measure of Reported Exclusion. This failure is also highlighted in Gerber & Wheeler’s (2009) meta-analysis and this variable is necessary to assess the theories regarding social exclusion across different conditions. Furthermore, a better understanding of self-reported exclusion may aid in investigating the method of Relived Rejection by providing a standard measure for memories that are difficult to quantify.
In order to properly test Baumeister’s Numbness Hypothesis, baseline mood and arousal measures are required. In the studies utilized in this survey, only one featured baseline and follow-up mood measures; Krill, Platek, and Wathne (2008) found that although subjects reported feeling angry while playing the Cyberball game, they demonstrated no change between baseline and follow-up mood measures, regardless of condition. These methods should be adopted by all exclusion researchers to assess changes in mood and arousal due to exclusion.
This meta-analysis demonstrates the differences between Ostracism and Rejection on exclusion-induced distress and post-exclusion coping behavior, but future studies must manipulate the method of exclusion to better understand the mechanisms of exclusion outside of the laboratory. Despite discussion of the differences between Ostracism, Rejection, and Hypothetical Social exclusion, no studies compare different exclusion manipulations. Rarely is exclusion felt by previously unknown individuals, and often Rejection and Ostracism are co-present. Excluding groups can be comprised of same-gender or opposite exclusion sources, or race congruent and incongruent members. Furthermore, as we become a more mediated society, a greater understanding of online exclusion and acceptance is required. Future studies must compare the effects of ostracism, rejection, and combined exclusion, as well as the multitude of situational differences present in exclusion experiences.