Personal Relationships, 5 (1998), 159-181. Printed in the United States of America. Copyright 0 1998 ISSPR. 1350-4126/98 $9.50

Positive illusion in close relationships

JOHN M. MARTZ,a JULIE VERETTE,” XIMENA B. ARRIAGA,b LINDA F. SLOVIK,a CHANTE L. COX: AND CARYL E. RUSBULTa a University of North Carolina at Chapel Hill; bClaremont Graduate University; and ‘Mississippi State University

Abstract The literature regarding self-other comparisons suggests that self-enhancing perceptions are prevalent, including forms of “illusion” such as excessively positive self-evaluation, unrealistic optimism, and exaggerated perceptions of control. Concepts from optimal distinctiveness theory served as the basis for two experiments examining whether illusion functions similarly when the context of evaluation involves a relationship. In both experiments participants rated themselves, the best friend, and the average other-or their own romantic relationships, the best friend’s relationship, and the relationship of the average other-using scales measuring positivity of evaluation, optimism regarding the future, and perceptions of control. In both experiments, participants exhibited centrality-based differentiation, rating targets more favorably to the degree that the target was more central to their social identity. Patterns of differentiation differed for the two contexts: In the individual context, participants differentiated themselves and their friends from the average other. In the relationship context, participants differentiated their own relationships from the relationships of friends and average others. Also, participants rated individuals as more controllable than relationships. Participants in Experiment 2 provided information regarding potential predictors of illusion. Analyses of these data suggest that favorable centrality-based differentiation may be partially accounted for by impression management, global self-esteem (particularly in the individual context), and commitment level (particularly in the relationship context).

What wild imagination one forms, where dear This research is dedicated to John Martz (1964-1995). John was an excellent scientist and an exceptional hu- self is concerned! How sure to be mistaken! man being. His enthusiasm for science and for life’s -Jane Austen other pleasures infected and inspired his friends, col- leagues, and students. We miss him very much. Whether confronted with negative or posi- The research was supported in part by a grant to tive events, whether evaluating ourselves or Caryl E. Rusbult from the NSF (No. BNS-9023817); our peers, we live in a world of illusion (e.g., preparation of this manuscript was supported by a Reynolds Research and Study Leave from the Univer- Alicke, 1985; Brown, 1986; Lipkus, Martz, sity of North Carolina and by a Visitor’s Grant from Panter, Drigotas, & Feaganes, 1993; Show- the Netherlands Organization of Scientific Research. ers, 1992; Tajfel & Turner, 1986; Taylor & We thank Abigail T. Panter and Elliot M. Cramer for Koivumaki, 1976;Van Knippenberg & Elle- their statistical advice. For their helpful comments re- mers, 1990; Weinstein, 1980). Although our garding earlier drafts of this article, we thank Christo- pher R. Agnew, Victor I. Bissonnette, Stanley 0. subjective experience of the social world in Gaines, Shelley Dean Peterson, Michael 0. Wexler, part reflects reality, perception and cogni- Jennifer Wieselquist, Arjay R. Wright, and Nancy A. tion frequently are self-enhancing, deviat- Yovetich, as well as Keith E. Davis, Garth J. 0. ing from a strictly veridical view of the Fletcher, and three anonymous reviewers. world. It has been argued that such miscon- Correspondence should be addressed to Caryl E. Rusbult, Department of Psychology, University of ceptions indicate poor adjustment and yield North Carolina at Chapel Hill, Chapel Hill, NC 27599- negative consequences (e.g., DeJoy, 1989; 3270; E-Mail: [email protected] Miller, Ashton, McHoskey, & Gimbel,

159 160 J.M. Martz et al.

1990). However, Taylor and Brown (1988) than is actually likely to be the case. A third suggest an alternative perspective (cf. type of illusion involves unrealistic opti- Scheier & Carver, 1993).These authors pro- mism about the jkture-the tendency to pose that such biased thinking to some ex- perceive the future as exceptionally benign, tent represents adaptive functioning, and particularly our own futures. that some degree of self-enhancement The illusions of positivity, optimism, and tends to be “characteristic of normal hu- control involve implicit or explicit compari- man thought” (p. 193). sons of the self to another person(s)-com- The present work explores the gener- parisons that serve to differentiate the self alizability of illusion phenomena to the from others in a favorable manner. In addi- realm of groups-specifically, to percep- tion to self-other differentiation, illusions tions of close relationships. We propose that also function to differentiate intimates parallel forms of illusion characterize cog- from other persons (e.g., Brown, 1986; Tay- nition regarding relationships, and we ex- lor & Koivumaki, 1976). Research examin- plore the nature of such cognitions by test- ing comparisons of the self, intimates, and ing hypotheses derived from Brewer’s other individuals (e.g., Campbell, 1986;Hall (1991) optimal distinctiveness theory. More & Taylor, 1976; Vallone, Griffin, Lin, & generally, we suggest that illusion in close Ross, 1990; Van Lange, 1991) has demon- relationships serves an important function, strated that individuals evaluate themselves representing a habit of thinking that sup- more positively than they evaluate their in- ports the decision to persist in a relation- timates, and they evaluate their intimates ship (cf. Rusbult & Buunk, 1993). It is easier more positively than does the average per- to sustain the energies needed to maintain son (cf. Taylor & Brown, 1988;Wood, 1989). a relationship (e.g., to accommodate, to Differentiation between an intimate and a derogate tempting alternatives) when one’s stranger may involve assimilation of the in- relationship is regarded as desirable and timate other into one’s social identity, as controllable, with a rosy future. The specific suggested by optimal distinctiveness theory. goal of the present research was to examine the existence and nature of illusion regard- Favorable Centrality-Based ing relationships; the present work merely Differentiation for individuals and Groups suggests the adaptive value of such illusion. Optimal distinctiveness theory (Brewer, 1991) suggests that self-other differentia- Favorable Self-other Differentiation tion occurs along a continuum (e.g., self > Based on an extensive review of the litera- intimate other > stranger). This theory ture regarding bias in human cognition, characterizes social identity as “a reconcili- Taylor and Brown (1988) identified three ation of opposing needs for assimilation forms of self-enhancement, which they de- and differentiation from others” (p. 475). scribe as “illusions.” This term is employed There is an optimal balance between the to represent “pervasive, enduring, and sys- need to regard oneself as part of a social tematic” cognitive misconceptions, as op- unit (social identity) and the need to regard posed to “error” or “bias,” which imply oneself as a distinct individual (personal short-lived and specific mistakes. One type identity). Individuals make favorable com- of illusion involves excessively positive self- parisons of intimates to strangers because evaluation-the tendency to perceive our- enhancing an intimate-who is relatively selves as superior to the average person, central to one’s social identity-serves to and to evaluate ourselves more positively enhance one’s social self. Thus, rather than than others evaluate us. A second type of describing the “self-enhancement” litera- illusion involves exaggerated perceptions of ture in terms of the self (i.e., in terms of control-the tendency to perceive that we favorable self-other differentiation), it may have greater control over events in our lives be appropriate to speak of centrality-based Positive illuiion in close relationships 161 differentiation-favorable differentiation about others’ relationships (Buunk, Collins, on the basis of the centrality of a target to Taylor, Van Yperen, & Dakof, 1990; Buunk one’s identity, where more central targets & Van Yperen, 1991; Hall & Taylor, 1976; are simultaneously assimilated into the self Murray & Holmes, 1993; Murray, Holmes, and differentiated from less central targets. & Griffin, 1996a, 1996b;Van Lange & Rus- Optimal distinctiveness theory also pro- bult, 1995). That is, more central targets of vides a rationale for the proposition that evaluation (i.e., one’s own grouphelation- the illusions of positivity, optimism, and ship) appear to be favorably differentiated control may color perceptions of relation- from less central targets, irrespective of the ships (i.e., dyads or groups) much as they context of evaluation (i.e., whether perceiv- color perceptions of individuals. Compari- ing individuals or relationships; cf. Brewer, sons of ingroups to outgroups follow a pat- 1991; Goethals & Darley, 1987; Levine & tern of evaluation similar to that observed Moreland, 1987; Smith, 1993). For example, in the literature on self-other compari- just as an individual may hold more be- sons-a similarity predicated by both social nevolent beliefs about her best friend than identity theory and self-categorization about the average person, she may hold theories (Tajfel & Turner, 1986; Turner, more benevolent beliefs about her best 1987). Overwhelmingly, individuals per- friend’s close relationship than about the ceive ingroups more favorably than they average close relationship. perceive outgroups (e.g., Maass & Schaller, Unfortunately, existing research on illu- 1991; Schopler & Insko, 1992; Smith, 1993). sion in the close relationships context is It should be evident that frame of reference limited in that it generally has examined plays a role in the differentiation between only one of the three illusions identified by ingroups and outgroups; ingroup members Taylor and Brown (1988)-namely tenden- are more central to the self and function as cies toward positive evaluation. The other a frame of reference for comparisons with two illusions-optimism and perceptions of outgroups. Thus, irrespective of frame of control-by and large have been ignored in reference (i.e., whether perception involves the relationships domain (but see Murray personal or social identities) and irrespec- & Holmes, 1997). Granted, it is important to tive of the specific illusion under considera- understand the factors that lead us to evalu- tion (i.e., positivity, optimism, control), the ate our relationships positively-happiness more central a target is to one’s identity, the and satisfaction are important features of more favorably it should be differentiated ongoing relationships. At the same time, the from less central targets. Just as central tar- course of an ongoing relationship may be gets should be favorably differentiated shaped by features other than positivity of from less central targets, more central dy- evaluation (cf. Rusbult, 1983). For example, ads or groups should be favorably differen- perceptions of control may be relevant to tiated from less central dyads or groups. understanding the dynamics of nonvolun- Given that the dyad formed by romantic tary dependence (e.g., remaining in an abu- partners can be one of the most central as- sive relationship; cf. Rusbult & Martz, pects of identity (e.g., Aron, Aron, Tudor, & 1995), and optimism regarding the future Nelson, 1991), it is theoretically and clini- may allow dissatisfied individuals to extend cally important to extend this line of rea- the evaluative frame of their relationships soning to the close relationships domain. into the future, thus compensating for less- Indeed, some researchers have begun to ex- than-ideal current outcomes (cf. Kelley, plore positive illusion in the context of inti- 1983). mate relationships. Early studies indicate Moreover, existing research has focused that individuals exhibit “perceived relation- on the simple differentiation between one’s ship superiority,” holding a greater number own relationship and the relationships of of positive beliefs and fewer negative be- strangers, and it has not included a more liefs about their own relationships than fine-grained analysis of centrality-based 162 J.M. Martz et al. differentiation. That is, the existing litera- land, 1987; Schaller, 1992). Because “one’s ture has not examined how perceptions of self-evaluation is to some extent anchored relationships with intermediate levels of in comparisons between oneself and others, centrality to one’s identify (e.g., the rela- and to some extent anchored in compari- tionships of close friends) may play into sons between one’s group and other optimal distinctiveness processes. Many so- groups” (Goethals & Darley, 1987, p, 34), it cial psychological theories comment on the is possible that individuals evaluate their importance of identifying the self with simi- friends more positively than they do their lar others and differentiating the self from friends’ romantic relationships, while simul- dissimilar others (e.g., social comparison taneously evaluating their own romantic re- theory, social identity theory). It may be lationships as favorably as they evaluate equally important to understand whether themselves. Thus, we advance no a priori and how individuals compare their rela- hypotheses regarding differences in cen- tionships to similar relationships and differ- trality-based differentiation for individuals entiate their relationships from dissimilar in comparison to relationships. relationships. As noted earlier, such cogni- tive maneuvers may stand as habits of Predictors of Centrality-Based thinking that support the decision to persist Differentiation in a relationship (cf. Rusbult & Buunk, 1993). Given that it is not clear how context (indi- Finally, no extant research has sought to vidual vs. relationship) and centrality of tar- determine whether substantively meaning- get may jointly influence perception, it is ful differences exist between illusory proc- desirable to consider factors that may be esses in the two contexts under considera- relevant to understanding this associa- tion-individuals versus relationships. On tion-to identify variables that may be as- average, do perceptions of individuals and sociated with centrality-based differentia- relationships differ in meaningful ways? tion, and to consider possible moderators of Some research suggests that individuals such associations. One predictor might be tend to be evaluated more positively than the truthfulness of the respondent, not only are other attitude objects, such as inanimate with himself or herself but also with others objects or groups of people (Sears, 1983). (cf. Baumeister & Hutton, 1987; Nisbett & Thus, it is possible that, on average, com- Ross, 1980; Nisbett & Wilson, 1977; Wood, pared to ratings of relationships, individuals 1989). For example, when attempting to will be evaluated more positively, perceived present themselves in a favorable manner, as more controllable, and judged to have individuals may consciously distance them- more rosy futures, irrespective of centrality selves from other individuals in order to of the target to the self. For example, just as create positive impressions (Cialdini et al., an individual may evaluate his repre- 1976; Goffman, 1959). Moreover, assuming sentative in Congress favorably than the that illusions represent deviations from re- Congress as a whole, individuals may hold ality, it is possible that there is a self-decep- more favorable beliefs about their best tive component to illusion (Paulhus, 1984). friends than about their friends’ romantic Thus, variables such as impression manage- relationships. ment and self-deception may contribute to Does the specific character of centrality- the display of centrality-based differentia- based differentiation differ for individuals tion. It would seem that self-deception and and relationships? As noted earlier, no re- desire to present oneself favorably would search has explicitly asked whether such be equally applicable to individuals and re- differentiation varies as a function of con- lationships-individuals plausibly deceive text of evaluation, so we do not know themselves and others by enhancing both whether or how centrality of target might themselves and their relationships (or by interact with context (cf. Levine & More- disparaging other individuals and others’ Positive illusion in close relationships 163 relationships). Thus, it seems plausible to specific construct, it seems probable that anticipate that the associations of self-de- it will predict self-other and self-friend ception and impression management with differentiation more powerfully than favorable centrality-based differentiation friend-other differentiation, and that it will will not differ for perceptions of individuals do so more powerfully in the relationship in comparison to perceptions of relation- context than in the individual context. ships. Thus, we suggest that context of evaluation Self-esteem is another plausible predic- will moderate the association of commit- tor of illusion, in that the tendency toward ment with centrality-based differentiation, positive illusion has been described as char- particularly for self-other and self-friend acteristic of healthy self-concept (Taylor & differentiation. Brown, 1988; Weinstein, 1984). While it seems clear that high self-esteem might in- duce more favorable ratings of the self in Research Overview and Hypotheses comparison to less central targets, to the This article presents the results of two ex- degree that social identity includes those periments, both of which included manipu- with whom we are intimate, it is also plausi- lations of context of evaluation (individual ble that high self-regard will induce more context vs. relationship context) and cen- favorable ratings of best friends in relation trality of target (self vs. best friend vs. av- to the average other. Thus, we anticipate erage other).’ In both experiments, partici- that self-esteem will be associated with fa- pants either (a) rated themselves, the best vorable centrality-based differentiation. friend, and the average other, or (b) rated However, given that conceptualizations of their own romantic relationships, the best self-esteem assume an individual context friend’s relationship, and the average (i.e., fundamentally, self-esteem concerns other’s relationship. We obtained ratings of evaluations of the self), it seems likely that positivity of evaluation, optimism regard- self-esteem will predict illusion in the indi- ing the future, and perceptions of control. vidual context more powerfully than it does Both experiments were designed to test in the context of relationships (cf. Crocker three hypotheses. First, we anticipated that, & Luhtanen, 1990; Luhtanen & Crocker, irrespective of context of evaluation, indi- 1992). Hence, we suggest that context of viduals would exhibit centrality-based dif- evaluation will moderate the association of ferentiation: self-esteem with favorable centrality-based differentiation. H1: There will be main effects of centrality Recent work suggests that perceived su- of target, such that ratings of evaluation, periority and other types of illusion may optimism, and control will be higher in the be construed as relationship maintenance self condition than in the best-friend and phenomena, arguing that such phenomena the average-other conditions, such that are driven by feelings of commitment to a ratings will be higher in the self and best- relationship (Rusbult & Buunk, 1993; Rus- friend conditions than in the average- bult, Van Lange, Verette, Yovetich, & Wild- other condition, or both. schut, 1997; Van Lange & Rusbult, 1995). Commitment represents strength of desire for a relationship to persist, and induces 1. Given the necessity of generically describing levels orientation toward the future. Individuals of centrality of target irrespective of the distinction who perceive an extended future in a re- between contexts, we employ “self” to refer to one- lationship should be motivated to think self and one’s romantic relationship, “best friend” to refer to the best friend and the best-friend’s highly of the relationship, perceive the re- romantic relationship, and “average other” to refer lationship as controllable, and feel optimis- to the average UNC student of one’s own age and tic regarding the relationship’s future. Also, sex, as well as the average-other’s romantic rela- given that commitment is a relationship- tionship. 164 J. M. Martz et al.

Second, we anticipated that centrality- individual context than in the relationship based differentiation would be evident in context. cognitions regarding individuals as well as in cognitions regarding relationships: Both experiments tested Hypotheses__ 1, L, and 3. Experiment 2 also obtained meas- H2: There will be simple effects of centrality ures of four variables that may illuminate of target for both contexts, such that rat- our knowledge of the underpinnings of in- ings of evaluation, optimism, and control dividual- and relationship-level illusion. (a) will be higher for the self than for the These measures allowed us to address a best-friend and average-other, will be fourth question: How does favorable cen- higher for the self and best-friend than for trality-based differentiation come about the average-other,or both; and (b) will be (i.e., what are some plausible predictors of higher for one’s own relationship than for such tendencies), and are associations with the best-friend’s and average-other’srela- such predictors similar for perceptions of tionships, will be higher for one’s own and individuals versus relationships (i.e., does the best-friend’s relationships than for the context moderate such associations)? average-other’s relationship, or both. Given that one important issue in under- standing perception may be the truthful- In addition, we sought to determine ness of the individual, Experiment 2 in- whether centrality-based differentiation cluded measures of both impression functions similarly in the individual context management and self-deception. We pre- and relationship context. Given that few if dicted that these variables would be posi- any studies have compared illusion regard- tively associated with favorable self-other, ing individuals and relationships (cf. Schal- self-friend, and friend-other differentia- ler, 1992), we did not advance a priori pre- tion in both the individual context and the dictions along these lines. However, one relationship context. exploratory goal of our research was to de- Experiment 2 also included a measure of termine whether illusion functions differ- self-esteem. Given that self-other differen- ently as a function of context. For example, tiation can be characterized as a continuum participants may regard themselves and of centrality-to-self, we predicted that self- their best friends equally favorably, but may esteem would be positively associated with differentiate between their own romantic relationships and the relationships of their favorable self-other, self-friend, and friends. friend-other differentiation. Also, we ex- pected that context would moderate the as- A third and somewhat tentative hy- pothesis concerned differences in ratings as sociation of self-esteem with centrality- a function of context of evaluation. Given based differentiation, such that self-esteem that previous studies have demonstrated would predict differentiation more power- that individuals tend to be regarded more fully in the individual context than in the positively than are dyads, groups, or other relationship context. sorts of attitude object (Sears, 1983), we Finally, Experiment 2 included a meas- speculated that participants might rate indi- ure of commitment to the current romantic viduals more favorably than romantic rela- relationship. Given that commitment level tionships, irrespective of centrality to the is relevant to one’s own relationship, we self. For example, we anticipated that par- predicted that commitment would be posi- ticipants might evaluate their friends more tively associated with favorable self-other favorably than their friends’ relationships. and self-friend differentiation. Also, we ex- pected that context would moderate the as- H3: There will be main effects of centrality sociation of commitment with differentia- of target, such that ratings of evaluation, tion, such that commitment would predict optimism, and control will be higher in the self-other and self-friend differentiation Positive illusion in close relationships 165 more powerfully in the relationship context ticipant was randomly assigned to one of than in the individual context. two experimental conditions. On the basis of these lines of reasoning, a fourth hypothesis was advanced: Design

H4: Ratings of evaluation, optimism, and Each experiment was a 2 X 3 X 2 design: control will be (a) positively correlated Context of Evaluation (individual ratings with self-deception for favorable vs. relationship ratings) was a between-par- self-other, self-friend, and friend-other ticipants variable and Centrality of Target differentiation; (b) positively correlated (self vs. best friend vs. average other) was a with impression management for within-participant variable. Participant Sex self-other, self-friend, and friend-other was an additional between-participants differentiation; (c) positively correlated variable that we did not expect would influ- with self-esteem for self-other, self-friend, ence our findings. Three types of dependent and friend-other differentiation, particu- measures were assessed: evaluation, opti- larly in the individual context, and (d) mism, and control. positively correlated with commitment level for self-other and self-friend differ- Procedure entiation, particularly in the relationship context. The studies were described as computer-ad- ministered experiments in which partici- pants would be asked to answer questions Method about themselves or their relationships. The experiments were presented via personal Participants computers, linked through a server via net- Two hundred eighteen undergraduates work software. Each participant was seated (132 women, 86 men) participated in Ex- at his or her own terminal. Partitions sepa- periment 1 in partial fulfillment of the re- rated workstations and protected the ano- quirements for introductory psychology nymity of participants’ responses. Partici- courses at the University of North Carolina pants received both oral and written (UNC) at Chapel Hill. Most were Cauca- informed-consent information, and then sian (81% Caucasian; 11% African Ameri- proceeded at their own pace through a can; 5% Asian American; 3% other), most computer-administered questionnaire. At were in their first or second year of college the end of the session, participants were (30% freshmen, 45% sophomores, 16% fully debriefed and thanked for their assis- juniors, 10% seniors), and they were 19 tance. years old on average (range 17 to 44 years). One hundred fifty-eight undergraduates Independent variable manipulations and (85 women, 73 men) participated in Experi- questionnaire ment 2 in partial fulfillment of the require- ments for introductory psychology courses In both experiments, Context of Evaluation at the University of North Carolina. Most and Centrality of Target were manipulated were Caucasian (84% Caucasian, 11% Af- using parallel sets of questionnaire items. rican American, 2% Asian American, 3% For each Centrality condition, participants other), most were in their first or second completed 21 nine-point scales tapping year of college (47% freshmen, 40% sopho- three types of measure (seven items per mores, 10% juniors, 4% seniors), and they type of measure; for all items, 1 = strongly were 19 years old on average (range 18 to disagree; 9 = strongly agree)-evaluation 32 years). Participants signed up for the ex- (e.g., “My life is extremely unsatisfying” periments in same-sex groups ranging in [reverse-scored]), optimism (e.g., “No mat- size from two to eight individuals.Each par- ter what happens in my life, things will al- 166 J.M. Martz et al. ways work out for the best”), and control tions were presented in one of six random (e.g., “If we work hard enough, we can orders. make our relationship ideal”). The 21 items In Experiment 2, after completing items were randomly ordered; the same random for each target condition, the following ordering was employed across all partici- scales were presented in fixed order: Corn- pants. These items were modified to tap mitment Level (Rusbult, Verette, Whitney, each of three levels of Centrality of Tar- Slovik, & Lipkus, 1991; e.g., “Do you feel get-self, best friend, and average other for committed to maintaining your relationship two Context of Evaluation conditions-in- with your partner”; 1 = not at all; 9 = com- dividual versus relationship. Thus, there pletely); GMal Self-Esteem (HOYle, 1991; were two parallel questionnaires, one for e.g., “On the whole, 1 am satisfied with my- individual ratings and one for relationship self”; 1 = not at all like me; 9 = very much ratings; each questionnaire included 63 like me); and the Balanced Inventory of items. Desirable Responding, which measures Context of Evaluation was manipulated both Self Deception and Impression Man- by assigning participants to either the indi- agement (PaulhW 1984; e.g.9 “I sometimes vidual or the relationship ratings condition. tell lies if 1 have to”; 1 = not true; 7 = Very Participants in the individual ratings condi- true). tion completed items describing individual- level attributes (e.g., “My life just keeps Participant selection getting better every day”; “I have no con- trol over the direction my life takes” [re- In Experiment 1, in both the individual and verse-scored]). Participants in the refation- relationship ratings conditions, if a partici- ship ratings condition completed parallel pant indicated that he or she did not have a items describing relationship-level attrib- best friend, the program skipped items re- utes (e.g., “My relationship just keeps get- garding the best friend (3 individuals in the individual condition, 2 in the relationship ting better every day”; “We have no control condition). In the relationship ratings con- over the direction our relationship takes” dition, if a participant indicated that he or [reverse-scored]). Questionnaires for the she was not involved in a romantic relation- two Context conditions were identical ex- ship, the program skipped corresponding cept for the variation in referent-either items (57 individuals); if a participant indip the individual (self, best friend, or average cated that the best friend was not involved other) or the relationship (one’s own rela- in a romantic relationship, the program tionship, the best friend’s relationship, or skipped corresponding items (28 individu- the average other’s relationship). als); data for one individual were deleted Centrality Of Target was a within-Partici- owing to incomplete data. Thus, a selection pant Each participant answered bias differentiated between the Context three parallel sets of items describing the conditions: Participants in the relationship self, the best friend (the closest friend, other context were required to (a) be involved in than a romantic partner), and the average a romantic relationship and (b) have a best other (the average student at UNC of the friend who (c) was involved in a romantic Participant’s age and sex). Participants relationship (n = 37 out of 125), whereas completed self ratings (e.g.7 “I am Success- the only restriction for participants in the fU1” Or “Our relationship iS SUCCeSSfUl”), individual context was that they (b) have a best-friend ratings (e.g., “My closest friend best friend (n = 90 out of 93). is successful” or ‘‘MY closest friend’s rela- In light of the selection bias charac- tionship is successful”), and average-other terizing Experiment 1, we modified the se- ratings (e.g., ‘“fie average student is suc- lection procedure for Experiment 2. In Ex- cessful” or “The average student’s relation- periment 2 sign-up sheets indicated that, in ship is successful”). The Centrality condi- order to participate, individuals must be in- Positive illusion in close relationships 167 volved in a romantic relationship. A ques- rately for Experiment l and Experiment 2, tionnaire item was included to verify that and separately as a function of Context of each participant was in fact currently in- Evaluation and Centrality of Target. In Ex- volved. Moreover, all participants were periment 1, for items tapping self, best asked whether they had a best friend and friend, and average other, acceptable reli- whether the friend was involved in a ro- ability was observed in both the individual mantic relationship. Irrespective of the context and the relationship context for Context condition to which they were as- measures of evaluation (average alphas = signed, participants’ data were later deleted .77 and .76), optimism (average alphas = from analyses if they were not involved in a .73 and .8l), and control (average alphas = romantic relationship (0 individuals), if .51 and .63). In Experiment 2, acceptable they did not have a best friend (0 individu- reliability was observed in both the individ- als), or if the best friend was not involved in ual context and the relationship context for a romantic relationship (51 individuals; 4 measures of evaluation (average alphas = individuals failed to report whether the .83 and .84), optimism (average alphas = friend was involved); data for five individu- .78 and .80), and control (average alphas = als were deleted because of incomplete .57 and .61). In Experiment 2, reliability data. analyses also revealed good convergence To determine whether the selection pro- for items measuring each potential predic- cedure in Experiment 1 might have colored tor of centrality-based differentiation: K-R- findings for that study-particularly in the 20 coefficients were .70 for Self-Deception individual context-we performed prelimi- and .74 for Impression Management; alphas nary analyses of the Experiment 2 data, in- were .91 for Self-Esteem and .88 for Com- cluding data for those individuals whose mitment Level. data were retained for final analyses and data for those individuals whose data were later deleted on the basis of our selection Correlations among measures criteria. These preliminary analyses in- cluded a Participant Selection variable to We calculated correlations among meas- distinguish between individuals whose data ures of evaluation, optimism, and control were later retained and those whose data separately for Experiment 1 and Experi- were later deleted. (For the individual con- ment 2 and separately as a function of Con- text, this variable distinguishes between the text of Evaluation and Centrality of Target. Experiment 2 selection criteria and the Ex- In Experiment 1, examining associations periment 1 criteria.) These analyses re- among ratings of self, best friend, and aver- vealed no significant Participant Selection age other, for both the individual context main effects, no interactions of Participant and the relationship context, there were Selection with other independent variables, positive correlations of evaluation with op- and no simple effects of Participant Selec- timism (using Fisher’s r-to-z transforma- tion within levels of other independent tion, average rs = .75 and .70 both ps < variables. Accordingly, the selection proce- .Ol), evaluation with control (average rs = dures employed in Experiment 1 do not .58 and .65; both ps < .Ol), and optimism appear to have colored our findings. with control (average rs = .63 and .80; both ps < .01). In Experiment 2, there were posi- tive correlations of evaluation with opti- Results mism (average rs = .78 and 33; both ps < .Ol), evaluation with control (average rs = Reliability of measures .61 and .72; both ps < .Ol), and optimism The reliability of items designed to measure with control (average rs = .64 and .77; both each type of illusion was assessed by calcu- ps < .01). Thus, although the three types of lating alphas for each set of items, sepa- measure are associated, to some degree 168 J.M. Marrz et al. they are theoretically and empirically dis- Does centrality-based differentiation occur? tinguishable. We anticipated that, irrespective of context, targets who were more central to the self would be favorably differentiated from tar- gets who were less central to the self. Spe- Effects of Context of Evaluation and cifically, Hypothesis I predicted main ef- Centrality of Target fects of Centrality of Target, such that To examine all effects in our 2 (Context ratings of evaluation, optimism, and control of Evaluation: individual vs. relationship would be higher in the self condition than ratings) X 3 (Centrality of Target: self vs. in the best-friend and average-other condi- best friend vs. average other) X 2 (Partici- tions, and/or that ratings would be higher in pant Sex: women vs. men) design, we per- the self and best-friend conditions than in formed doubly-multivariate repeated- the average-other condition. Consistent measures analysis of variance separately with expectations, the multivariate main ef- for Experiments 1 and 2.2 This analysis is fect of Centrality of Target was significant equivalent to performing three simultane- in both Experiment 1 (Mult. F [6, 1181 = ous repeated-measures analyses of vari- 22.50,~< .Ol) and Experiment 2 (Mult. F ance, examining evaluation repeated [6,89] = 14.71,~< .01). Ignoring context of across levels of Centrality, optimism re- evaluation and type of measure, ratings peated across levels of Centrality, and con- tended to decline from the self condition to trol repeated across levels of Centrality, in the best-friend condition to the average- the context of our between-participants other condition in both Experiment 1 (see variables (Context of Evaluation, Partici- Table 1; Ms = 7.18 vs. 6.64 vs. 6.01) and pant Sex). This technique is preferable to Experiment 2 (Ms = 6.95 vs. 6.61 vs. 5.84). performing separate repeated-measures In both experiments the main effect of Cen- analyses because it minimizes possibilities trality of Target was evident for all three for capitalizing on chance. types of measure (see Table 2, “Centrality For each experiment, mean levels of of Target Main Effect” rows under heading each measure as a function of Context of “Effects for Full Design”)-positivity of Evaluation and Centrality of Target are evaluation (2 of 2 effects significant), opti- presented in Table 1. Table 2 presents an mism regarding the future (2 of 2 effects analysis of variance summary table for each significant), and perceptions of control (2 of experiment. Out of a total of 24 effects in- 2 effects significant). volving Participant Sex (12 effects for each What is the character of centrality-based experiment), only one effect was significant differentiation? Is the self differentiated in Experiment 1 and no effects were signifi- from the best friend and average other, are cant in Experiment 2. Accordingly,Tables 1 the self and best friend differentiated from and 2 do not present results as a function of the average other, or do both forms of dif- Participant Sex. ferentiation occur? Contrasts of the three target conditions for the three types’ of measure revealed that in both experiments 2. The term doubly-multivariate derives from the fact (a) the self condition was favorably differ- that the design is multivariate in two respects: (a) the presence of three dependent variables (evalu- entiated from the average other condition ation, optimism, and control) conforms to the usual (6 of 6 effects significant; see Table 2, “Self definition of multivariate analysis; and (b) al- vs. Average Other” rows under headings though the repeated measures component of the “Effects for Full Design”); (b) the self con- design (self vs. best friend vs. average other) is dition was favorably differentiated from conceptually univariate, it is preferable to repre- sent such designs as multivariate so as to directly the best-friend condition (6 of 6 effects sig- examine within-participant effects (i.e., in contrast nificant or marginal; see “Self vs. Best to a univariate or mixed-model approach; Maxwell Friend” rows); and (c) the best-friend con- & Delaney, 1990). dition was favorably differentiated from Positive illusion in close relationships 169

Table 1. Mean positivity of evaluation, optimism regarding the future, and perceptions of control as a function of centrality of target and context of evaluation: Experiments 1 and 2

Context/IllusionType Self Best Friend Average Other Experiment 1 Individual Context (n = 90) Positivity of Evaluation 7.14” 7.18” 6.15h (1.16) (1.10) (0.87) Optimism Regarding the Future 6.93” 6.84” 5.82h (1.06) (1.06) (0.99) Perceptions of Control 7.31” 7.19a 6.74” (0.85) (0.89) (0.80) Relationship Context (n = 37) Positivity of Evaluation 7.76” 6.49h 6.17h (I.06) (I.61) (0.88) Optimism Regarding the Future 6.90” 5.96b 5.39h (1.23) (1.92) (1.29) Perceptions of Control 7.02” 6.1Sh 5.80” (1.22) (1.37) (I.04) Experiment 2 Individual Context (n = 50) Positivity of Evaluation 7.01” 6.84a 6.08h (1.23) (1.23) (1.11) Optimism Regarding the Future 6.74a 6.59a 5.58” (I.04) (I.31) (1.16) Perceptions of Control 7.21” 7.08a 6.64b (0.88) (1.01) (0.73) Relationship Context (n = 48) Positivity of Evaluation 7.37” 6.66b 5.86c (1.39) (1.25) (0.91) Optimism Regarding the Future 6.61” 6.23” 5.21b (1.57) (1.29) (1.05) Perceptions of Control 6.76” 6.27h 5.67‘ (I.30) (0.97) (0.94)

Note: Higher values reflect more positive evaluations, greater optimism, and stronger perceived control (range = 1 to 9). Within rows, means with different superscripts differ significantly,p < .01. the average other condition (6 of 6 effects simple effects of Centrality of Target for significant; see “Best Friend vs. Average both contexts, such that (a) participants Other” rows). Thus, ignoring context of would rate themselves more favorably than evaluation, three forms of centrality-based they rated the best friend and average differentiation are evident: self-other dif- other, and/or that they would rate them- ferentiation, self-friend differentiation, and selves and the best friend more favorably friend-other differentiation. than the average other; and (b) participants would rate their own relationships more fa- Does centrality-based differentiation occur vorably than they rated the best friend’s in both the individual context and the rela- and average other’s relationships, and/or tionship context? We anticipated that cen- that they would rate their own and the best trality-based differentiation would be evi- friend’s relationships more favorably than dent in cognitions regarding individuals as the average other’s relationship. Consistent well as in cognitions regarding relation- with expectations, in both Experiment 1 ships. Specifically, Hypothesis 2 predicted and Experiment 2, the multivariate simple 170 J.M. Martz et al.

Table 2. Analysis of variance summary table: Experiments 1 and 2

Optimism Positivity of Regarding Perceptions Effect Evaluation the Future of Control Experiment 1 Effects for Full Design Context of Evaluation Main Effect 0.01 7.02** 29.91** Context by Target Interaction 11.56** 4.70** 6.42** Centrality of Target Main Effect 47.05** 50.08** 28.00** Self vs. Average Other 100.27** 110.76** 57.81** Self vs Best Friend 6.69** 7.05** 11.25** Best Friend vs. Average Other 45.59** 42.42** 16.15** Effects Within Individual Context Centrality of Target Simple Effect 38.23** 51.68** 18.35** Self vs. Average Other 52.36** 72.73** 32.47** Self vs. Best Friend 0.08 0.59 1.57 Best Friend vs. Average Other 66.46** 70.67** 19.60** Effects Within Relationship Context Centrality of Target Simple Effect 18.80** 11.25** 12.04** Self vs. Average Other 52.01** 37.86** 25.18** Self vs. Best Friend 17.03** 7.68** 11.66** Best Friend vs. Average Other 1.27 2.43 1.94 Experiment 2 Effects for Full Design Context of Evaluation Main Effect 0.00 2.41 26.68** Context by Target Interaction 2.19 0.38 2.45t Centrality of Target Main Effect 31.79** 39.07** 24.31** Self vs. Average Other 61.59** 79.67** 47.25** Self vs. Best Friend 6.53** 2.79t 5.90* Best Friend vs. Average Other 32.85** 41.78* * 20.83 * * Effects Within Individual Context Centrality of Target Simple Effect 11.53** 26.35** 8.91** Self vs. Average Other 18.85** 50.32** 16.94** Self vs. Best Friend 0.56 0.79 0.78 Best Friend vs. Average Other 19.58** 30.18** 9.35** Effects Within Relationship Context Centrality of Target Simple Effect 21.49" * 16.27** 15.75** Self vs. Average Other 45.65** 34.71** 30.30** Self vs. Best Friend 7.85** 2.00 5.55* Best Friend vs. Average Other 14.18** 15.77** 11.49**

** p < .01. * p < .05. +p< .lo. Note: Table values are F's from analyses of variance performed on data from Experiments 1 and 2. Positive illusion in close relationships 171 effect of Centrality of Target was significant participants rated themselves more favor- in both the individual context (Experiment ably than the average other (6 of 6 effects; 1 Mult. F [6, 831 = 18.84,~< .01; Experi- see Table 2, rows under headings “Effects ment 2 Mult. F [6,43] = 10.31,~< .01) and Within Individual Context”); (b) partici- the relationship context (Experiment 1 pants did not rate themselves more favor- Mult. F [6,30] = 8.79,~< .01; Experiment ably than the best friend (0 of 6 effects); and 2 Mult. F [6,41] = 8.00,~< .01). (c) participants rated the best friend more Ignoring type of measure, in the individ- favorably than the average other (6 of 6 ual context, ratings tended to decline from effects). the self condition to the best-friend condi- In contrast, in the relationship context, tion to the average-other condition in both (a) participants rated their own relation- Experiment 1(see Table 1;Ms = 7.13 vs.7.07 ships more favorably than the average- vs. 6.24) and Experiment 2 (Ms = 6.99 vs. other’s relationship (6 of 6 effects; see rows 6.84 vs. 6.10). Ignoring type of measure, in under headings “Effects Within Relation- the relationship context, ratings tended to ship Context”); (b) participants rated their decline from the self condition to the best- own relationships more favorably than the friend condition to the average-other condi- best-friend’s relationship (5 of 6 effects); tion in both Experiment 1(see Table 1;Ms = and (c) participants rated the best-friend’s 7.23 vs. 6.21 vs. 5.79) and Experiment 2 (Ms relationship more favorably than the aver- = 6.91 vs. 6.39 vs. 5.58). In both experiments, age-other’s relationship in Experiment 2 (3 the simple effect of Centrality of Target was of 3 effects), but not in Experiment 1 (0 of evident in both contexts for all three types of 3 effects). Thus, in the individual context, measure: (a) The simple effect of Centrality participants exhibited self-other and friend in the individual context was significant for -other differentiation, but did not exhibit evaluation, optimism, and control (6 of 6 ef- self-friend differentiation (i.e., they did not fects significant; see Table 2, “Centrality of differentiate between themselves and their Target Simple Effect” rows under headings friends). In the relationship context, partici- “Effects Within Individual Context”); and pants exhibited self-friend and self-other (b) the simple effect of Centrality in the re- differentiation, but did not exhibit reliable lationship context was significant for evalu- friend-other differentiation (i.e., they did ation, optimism, and control (6 of 6 effects differentiate between their own relation- significant; see rows under headings “Ef- ships and the friend’s relationship). fects Within Relationship Context”). Does centrality-based differentiation Are individuals and relationships evaluated function similarly in the individual context differently? We did not hold strong expec- and relationship context? We did not ad- tations about differences in ratings as a func- vance a priori predictions along these lines, tion of context. At the same time, some evi- but one exploratory goal of our research dence suggests that, on average, individuals was to determine whether illusion functions tend to be regarded more positively than are differently as a function of context. In Ex- dyads or groups. Thus, in a tentative vein periment 1, the interaction of Context of Hypothesis 3 predicted main effects of Con- Evaluation with Centrality of Target was text of Evaluation, such that ratings of significant for all three measures; in Experi- evaluation, optimism, and control would be ment 2 this interaction was marginal for higher for individuals than for relationships. one of three measures (see Table 2, “Con- Consistent with expectations, the multivari- text by Target Interaction” rows under “Ef- ate main effect of Context of Evaluation was fects for Full Design” headings). We per- significant in both Experiment 1 (Mult. F[3, formed contrasts of the three target 1211 = 15.65, p < .Ol) and Experiment 2 conditions for the three types of measure (Mult. F [2,92] = 16.58,~< .01). Ignoring separately for the individual and relation- centrality of target and type of measure, rat- ship contexts. In the individual context, (a) ings of individuals tended to be somewhat 172 J. M. Martz et al. more favorable than ratings of relationships Hypothesis 4 predicted that centrality- in both Experiment 1(see Table 1;Ms = 6.81 based differentiation would be positively vs. 6.41) and Experiment 2 (Ms = 6.64 vs. associated with Self-Deception, Impression 6.29). The main effect of Context was evi- Management, Self-Esteem (particularly in dent for perceptions of control in both ex- the individual context), and Commitment periments (see Table 2, “Context of Evalu- Level (particularly for self-other and ation Main Effect” rows under headings self-friend differentiation in the relation- “Effects for Full Design”), was evident for ship context). Preliminary analyses examin- optimism regarding the future in Experi- ing possible interactions with context ment 1but not in Experiment 2, and was not revealed that the associations of centrality- evident for positivity of evaluation in either based differentiation with Self-Deception experiment. and Impression Management do not differ for the individual context and relationship Associations of potential predictors with context; however, the associations of cen- favorable centrality-based differentiation trality-based differentiation with Self-Es- teem and Commitment Level do differ for Experiment 2 obtained measures of four the individual context and relationship con- variables that we hypothesized would be text. That is, it is suitable to examine main associated with centrality-based differen- effects only for Self-Deception and Impres- tiation (data for 16 individuals were deleted sion Management, but to examine interac- owing to incomplete data). Our next task tions with Context for both Self-Esteem was to explore whether and how these vari- and Commitment Level. ables relate to favorable centrality-based Accordingly, we performed nine 7-factor differentiation. To evaluate the associations regression analyses, regressing each of of Self-Deception, Impression Manage- three differentiation scores (self-other, ment, Self-Esteem, and Commitment Level self-friend, friend-other) for each of three with centrality-based differentiation, for types of measure (positivity, optimism, con- ratings of evaluation, optimism, and con- trol) simultaneously onto (a) the main ef- trol, we calculated three signed discrepancy fect of Self-Deception; (b) the main effect scores (e.g., positive evaluation for the self of Impression Management; (c) the main minus positive evaluation for the friend) to effect of Self-Esteem; (d) the interaction of develop measures of favorable self-other Self-Esteem with Context; (e) the main ef- differentiation, favorable self-friend differ- fect of Commitment Level; (f) the interac- entiation, and favorable friend-other differ- tion of Commitment with Context; and (g) entiation .3 the main effect of Context of Evaluation (0 = individual context; 1 = relationship con- ~~~~~~~ text). Table 3 presents beta values for each 3. The use of discrepancy scores has been criticized on effect for each of nine 7-factor analyses. the grounds that it is unclear whether associations with such scores are attributable (a) to links with the For example, in the 7-factor analysis pre- discrepancy per se, or (b) to links with one or the dicting Self-other Differentiation for Posi- other component of the discrepancy (cf. Cohen & tivity of Evaluation, the regression coeffi- Cohen, 1983). However, in the current context the cient for Self-Deception was .13 (see Table use of discrepancy scores seems suitable in that illu- 3, “Self-other Differentiation” row under sion can be discerned only in the differentiation be- tween ratings of central targets and noncentral tar- heading “Self-Deception Main Effect,” gets. Moreover, in the present work we do not seek beta under column for “Positivity of Evalu- to determine whether differentiation occurs ation”). Table 3 presents beta values for the through enhancement of central targets, derogation interactions of Context with Self-Esteem of noncentral targets, or both. Finally, these meas- and Commitment Level; these coefficients ures do constitute a decomposition of the Centrality of Target main effect into three within-participant tell us whether the association of centrality- contrasts-contrasts that have been rather thor- based differentiation with Self-Esteem or oughly examined in the preceding analyses. Commitment differs for the two contexts Positive illusion in close relationships 173

Table 3. Regression of favorable self-other, self-friend, and friend-other differentiation onto self deception, impression management, self esteem, commitment level, and context, including interactions of self-esteem and commitment with context: Experiment 2

Optimism Positivity of Regarding Perceptions Predictor Evaluation the Future of Control Self-Deception Main Effect Self-other Differentiation .13 .15 .17t Self-Friend Differentiation .10 .08 .13 Friend-Other Differentiation .01 .04 .05 Impression Management Main Effect Self-other Differentiation -.13 .O1 -.lo Self-Friend Differentiation -.28** - .22* -.31** Friend-Other Differentiation .21t .23* .22t Self Esteem by Context Interaction Self-other Differentiation -0.39 - 0.02 -0.12 Self-Friend Differentiation -0.89" -I .07** -0.26 Friend-Other Differentiation 0.66 1.03* 0.15 Self-Esteem-Individual Context Self-other Differentiation .37** .28* .26? Self-Friend Differentiation .35** .29* .ll+ Friend-Other Differentiation - .02 - .04 .17 Self-Esteem-Relationship Context Self-other Differentiation .18 .27+ .20 Self-Friend Differentiation - .08 -.21 - .O1 Friend-Other Differentiation .29 .45** .24 Commitment by Context Interaction Self-other Differentiation 1.13** 1.18** 1.04** Self-Friend Differentiation 0.78* 0.94** 0.77* Friend-Other Differentiation 0.30 0.10 0.32 Commitment-Individual Context Self-other Differentiation - .22 - .24 -.14 Self-Friend Differentiation .12 .16 .22 Friend-Other Differentiation - .38* - .37* - .39* Commitment-Relationship Context Self-other Differentiation .49** SO** .51** Self-Friend Differentiation .60** .75** .70** Friend-Other Differentiation - .20 -.31* -.19 Context Main Effect-Individual vs. Relationship Self-other Differentiation - .43 - 38' - .57 Self-Friend Differentiation .28 .23 - .29 Friend-Other Differentiation - .82 - .99+ - .32

Note: Table values are betas for each predictor variable from nine simultaneous 7-factor regression analyses, predicting self-other, self-friend, and friend-other differentiation for ratings of evaluation, optimism, and con- trol. ** p < .01. *p< .05. tp < .lo. 174 J.M. Martz et al.

(see headings “Self-Esteem by Context In- text Interaction”). Consistent with expecta- teraction” and “Commitment by Context tions, in the individual context Self-Esteem Interaction”). Table 3 also presents beta val- was positively associated with self-other ues calculated separately for the individual differentiation (3 of 3 effects; see rows un- context and relationship context; these co- der heading “Self-Esteem-Individual efficients represent the actual associations Context”) and self-friend differentiation (3 of centrality-based differentiation with of 3 effects); coefficients for friend-other Self-Esteem and Commitment in the two differentiation were nonsignificant (0 of 3 contexts (e.g., see heading “Self-Es- effects). However, Self-Esteem was also teem-Individual Context”).4 weakly associated with centrality-based dif- The analyses revealed no support for ferentiation in the relationship con- Hypothesis 4u. The multivariate main effect text-for both selfiother differentiation (1 of Self-Deception was not significant (Mult. of 3 effects; see heading “Self-Esteem-Re- F [6,69]= 0.55,~< .77), and the regression lationship Context”) and friend-other dif- coefficients for this variable were nonsigni- ferentiation (1 of 3 effects). ficant in all nine univariate analyses (one We also examined both the main effect effect was marginal; see Table 3, rows under of Commitment and the interaction of this heading “Self-Deception Main Effect”). variable with Context. Consistent with Hy- The analyses revealed some support for pothesis 44 the multivariate interaction of Hypothesis 4b. The multivariate main effect Commitment with Context was significant of Impression Management was marginally (Mulr.F[6,69] = 3.28,~< .Ol);in univariate significant (Mult. F [6,69] = 2.04,~< .07), analyses the interaction of Commitment and coefficients for this variable were sig- with Context was significant in six of nine nificant or marginal in six of nine analyses. instances (see rows under heading “Com- Consistent with expectations, Impression mitment by Context Interaction” in Table Management was positively associated with 3).Consistent with expectations, in the rela- favorable friend-other differentiation (3 of tionship context Commitment was posi- 3 effects significant or marginal); inconsis- tively associated with self-other differentia- tent with expectations, Impression Manage- tion (3 of 3 effects; see rows under heading ment was negatively associated with “Commitment-Relationship Context”) self-friend differentiation (3 of 3 effects). and self-friend differentiation (3 of 3 ef- In these analyses we examined both the fects); Commitment was unreliably nega- main effect of Self-Esteem and the interac- tively associated with friend-other differen- tion of this variable with Context. In weak tiation (1 of 3 effects). Surprisingly, in the support of Hypothesis 4c, the multivariate individual context Commitment was nega- interaction of Self-Esteem with Context tively associated with friend-other differen- was marginally significant (Mult. F [6,69] = tiation (3 of 3 effects; see rows under head- 2.07, p < .07); in univariate analyses, the in- ing “Commitment-Individual Context”); teraction of Self-Esteem with Context was coefficients for self-other and self-friend significant in three of nine instances (see differentiation were nonsignificant. rows under heading “Self-Esteem by Con- The multivariate main effect of Context of Evaluation was not significant (Mult. F 4. The self-friend and friend-other differentiation [6,69] = 1.26, p < .29), and the regression measures fully describe the Centrality of Target coefficients for this variable were nonsigni- effect represented by our three Centrality condi- ficant in all nine univariate analyses (2 ef- tions: multivariate effects are based on these two fects were marginal). criteria. However, it is instructive to examine uni- variate effects for all three types of centrality-based differentiation. Therefore, we performed univari- Discussion ate analyses for all nine criteria, examining self-friend, friend-ther, and self-other differen- Consistent with Hypothesis I, the present tiation for evaluation, optimism, and control. research revealed that the more central a Positive illusion in close relationships 175 target is to the individual’s identity, the respective of context. We discovered an in- greater is the individual’s tendency to per- teresting difference between patterns of ceive the target favorably. Ignoring context differentiation for individuals and groups: of evaluation, individuals exhibited three In the individual context, participants ex- forms of centrality-based differentiation: hibited reliable self-other and friend-other (a) Ratings were higher in the self condi- differentiation, but did not exhibit tion than in the average-other condition; self-friend differentiation (i.e., they did nut (b) ratings were higher in the self condition differentiate between themselves and their than in the best-friend condition; and (c) friends). In the relationship context, partici- ratings were higher in the best-friend con- pants exhibited reliable self-other and dition than in the average-other condition. self-friend differentiation, but did not ex- These findings support the line of reasoning hibit reliable friend-other differentiation outlined in the introduction: Consistent (i.e., they did differentiate between their with optimal distinctiveness theory (cf. own relationships and the friend’s relation- Brewer, 1991), it appears that it is suitable ship). Thus, in perceiving individuals, to speak of “centrality-based differentia- friends are incorporated into the self; in tion,” whereby relatively central targets are perceiving relationships, friends’ relation- simultaneously assimilated into the self and ships are contrasted from one’s own rela- differentiated from less central targets. tionship. Furthermore-and consistent with pre- This finding may be explained through vious research-individuals engage in at two lines of reasoning. One line of reason- least three forms of centrality-based differ- ing suggests that two close individuals read- entiation, exhibiting illusion with respect to ily form a natural, identity-defining social positivity of evaluation, optimism regarding unit, whereas larger groups are less auto- the future, and perceptions of control (cf. matically unitlike in their relation to social Taylor, & Brown, 1988). These three types identity. For example, the process of assimi- of illusion appear to function similarly, be- lation involved in self-other merger may be ing influenced in like manner by the vari- easier for an individual “other” than for an ables examined in our work. For example, “other” composed of a dyad or group (cf. evaluation, optimism, and control were Aron & Aron, 1997).Assuming this to be so, similarly affected by variations in centrality it would make sense that, in our work, two of target, and they exhibited parallel asso- close individuals-either oneself and one’s ciations with predictors of centrality-based best friend, or the dyad comprised of one- differentiation. Also, the three types of self and one’s romantic partner-formed a measure were positively correlated. necessary and sufficient identity-defining Does centrality-based differentiation oc- unit. In the context of relationship percep- cur in cognition regarding relationships as tion, if one’s two-person romantic involve- well as in cognition regarding individuals? ment is bolstered, it is unnecessary to bol- In support of Hypothesis 2, the simple effect ster identity further by bolstering a friend’s of centrality of target was significant in involvement; in individual perception, both the individual context and the rela- greater potential exists for bolstering iden- tionship context. Moreover, the effects of tity “with a little help from [one’s] friends.” centrality were significant for evaluation, Indeed, relationships appear to stand as optimism, and control. These findings do natural perceptual categories, playing a role much to extend the literature regarding in “organizing spontaneous mental repre- positive illusion, demonstrating the exist- sentations” of the social world (Sedikides, ence of centrality-based differentiation for Olsen, & Reis, 1993, p. 71). Also, research all three types of illusion, for both individu- comparing interindividual to intergroup in- als and relationships. teractions consistently has demonstrated Our work also asked whether centrality- that one-on-one interactions are perceived based differentiation functions similarly, ir- as more congenial and cooperative than are 176 J. M. Martz et al. group-on-group interactions (e.g., Insko, ship into one’s identity, and it is relatively Schopler, Hoyle, Dardis, & Graetz, 1990; easy to incorporate one’s best friend into Schopler, Insko, Graetz, Drigotas, & Smith, one’s identity. Successfully linking a best- 1991). This body of research argues that a friend’s relationship to one’s own relation- qualitative discontinuity exists between ship may be a far more complex interde- one-to-one perception and group-to-group pendence phenomenon. perception. The present work also asked whether, on According to a second line of reasoning, average, individuals perceive relationships the realities of interdependence may ex- as favorably as they perceive individuals. plain why an individual is more readily in- Consistent with Hypothesis 3, main effects corporated into one’s identity than a group. of context-individual versus relation- In short, compared to two persons (i.e., my ship-were observed for perceptions of best friend and I), four persons (i.e., my control. Why are individuals regarded as partner, my best friend, my friend’s partner, more controllable than relationships? and I) face more daunting problems of in- Whereas control in the individual context terdependence. For example, couple-to- depends on individual-level attributes (e.g., couple interactions may be complicated by skill, knowledge), control in the relation- romantic jealousy, envy, or competition, and ship context also rests on the partner’s members of larger groups necessarily will characteristics and emergent features of the find it difficult to judge one another’s pref- relationship-the course of a relationship is erences, coordinate action, and achieve de- partially determined by relatively complex sirable outcomes for all parties. Also, the forms of interdependence (cf. Kelley & odds that four individuals will actively en- Thibaut, 1978; Rusbult & Van Lange, 1996; joy one another’s company is exponentially Thibaut & Kelley, 1959). Thus, whereas it smaller than the odds that two people will may not be unreasonable to perceive that find one another congenial. Similarly, it is individuals exert some degree of control easier to locate a one-person social unit over events in their lives (cf. Langer, 1989; whom one is willing to bring to one’s bosom Lerner & Miller, 1978), control over events in an identity-defining manner; in the case in an ongoing relationship clearly is more of a three-person social unit, just one bad egg can obliterate desire to link one’s iden- complex and multifaceted (cf. Jones & tity to the unit. In her novel Heartburn, Davis, 1965). In light of previous findings regarding Nora Ephron (1983, p. 101) describes a suc- cessful instance of dual-relationship inter- the person-positivity bias (e.g., Sears, 1983), dependence (i.e., couple-to-couple dating): it is curious that in our work the main effect of context was not significant for positivity of evaluation. This may be due to methodo- We saw each other every Saturday night and logxal differences between our work and every Sunday night, and we had a standing en- gagement for New Year’s Eve. Our marriages traditional studies of person-positivity: were tied together. We went to Italy, we went to Rather than comparing evaluations of “my Ireland, we went to St. Martin.. . . In some fun- congresswoman” to evaluations of “Con- damental sense, we were always on the road, gress,” as in prototypic person-positivity re- merrily on our way to nowhere in particular. search, we compared evaluations of indi- Two of us liked dark meat and two of us liked viduals (e.g., my best friend) to evaluations light meat and together we made a chicken. of a substantially smaller group-a roman- tic dyad (e.g., my best friend and her ro- The rarity of such congenial, identity-defin- mantic partner). Thus, it is possible that in ing foursomes pays testimony to the com- the present work, traditional person-posi- plexities of interdependence in larger tivity was overridden by the tendency to groups. In short, it is relatively easy to incor- view relationships as a reflection of the in- porate one’s romantic partner and relation- dividual partner’s social identities, and vice Positive illusion in close relationships 177 versa (cf. Cooper, 1981; Tajfel & Tbrner, Consistent with Hypothesis 4d, commit- 1986). ment level was associated with centrality- Relevant to Hypothesis 4b, Experiment based differentiation in the relationship 2 revealed that favorable centrality-based context. Also, in the relationship context, differentiation may be partially accounted commitment was associated with self-other for by impression management. It is inter- and self-friend differentiation but was esting that impression management ap- unrelated to friend-other differentiation. pears to promote enhancement of one’s As was true for self-esteem, these findings best friend. Impression management was support the construct validity of commit- negatively associated with favorable ment, demonstrating that commitment is self-friend differentiation and positively as- uniquely tied to one’s own relationship (cf. sociated with friend-other differentiation. Rusbult & Buunk, 1993). Recent research Thus, the process of impression manage- suggests that the link between commitment ment may be relatively sophisticated and and centrality-based differentiation may be indirect. Individuals may believe that it is motivational: Manipulations of psychologi- ultimately less effective to enhance them- cal threat enhance the commitment-illusion selves publicly than it is to enhance iden- link, whereas injunctions to be accurate tity-defining intimates (e.g., parents brag weaken this link (Rusbult et al., 1997). One about their children, not themselves). (We finding from the present work may be con- should also note that this research was con- sistent with the claim that commitment pos- ducted in the South, where modesty is a sesses motivational properties: In the indi- virtue and self-aggrandizement is regarded vidual context, commitment was negatively as inexcusably gauche.) associated with friend-other differentia- Hypothesis 4u received virtually no sup- tion, suggesting that the derogation of port. A priori, it seemed reasonable to (threatening) others in relation to (trusted) speculate that self-deception might be friends may be one consequence of strong linked with inclinations toward illusion. It commitment. However, on the basis of the appears that deceiving others (i.e., engaging current findings this line of reasoning is in impression management)-particularly highly speculative. in the indirect fashion evident in the pre- sent work-is a more reliable predictor of Limitations and directions for future work illusion than is deceiving oneself. Our findings for global self-esteem were As with any experimental procedure, the generally congruent with Hypothesis 4c, present methodology is not without limita- suggesting that this variable may be more tions. First, although in some respects strongly associated with differentiation in within-participant designs are preferable to the individual context than in the relation- between-participant designs (e.g., greater ship context. Also, self-esteem was consis- information per participant, increased tently associated with self-other and power), there are drawbacks to this meth- self-friend differentiation but was unre- odology, such as the inability to control for lated to friend-other differentiation. These potentially persistent carryover effects results provide evidence for the construct (Maxwell & Delaney, 1990). In future re- validity of self-esteem, suggesting that self- search it would be useful to employ be- esteem is uniquely tied to individual iden- tween-participant manipulations to exam- tity. Future work designed to explore the ine the phenomenon of centrality-based relationship between individual and group differentiation. functioning would benefit from exploring A second limitation of the present work the differential utility of individual and col- centers on participant loss, which occurred lective self-esteem constructs (cf. Brewer, when participants reported that they did 1991; Crocker & Luhtanen, 1990;Luhtanen not have a best friend, or when either the & Crocker, 1992). participant or the best friend was not cur- 178 J.M. Martz et al. rently involved in a romantic relationship. for a variety of reasons, it might be useful to Participant loss was particularly problem- study related phenomena using behavioral atic in Experiment 1, where differential loss measures, or to investigate illusory proc- across conditions yielded an interaction of esses in more naturalistic settings (e.g., us- selection bias with context of evaluation. In ing thought-listing techniques; using diary Experiment 2, steps were taken to guaran- methodologies to study illusion in everyday tee equivalent samples in the individual perception; cf. Reis & Wheeler, 1991;Webb, and relationship contexts. At first glance, Campbell, Schwartz, & Sechrest, 1966). one might question the validity of findings Finally, the reader should take note of based on a subset of the larger sample. At the fact that, in describing the correlational the same time, it should be clear that the results of Experiment 2, we have referred hypotheses and design necessitated a sam- to associations with illusion. It is important ple in which both participants and their to acknowledge that our results regarding friends were romantically involved. While it the predictors of centrality-based differen- would be possible to “eliminate” selectivity tiation clearly represent plausibility tests, by specifying multiple requirements for and that no direct inferences regarding participation on sign-up sheets, it should be cause and effect can be formed. More gen- clear that such a procedure merely shifts erally, it is important to note that we cannot selectivity from the hands of the experi- infer causation in describing the effects of menter into the hands of the participant centrality of target, in that targets were not pool. At the very least, the current method randomly assigned to participants-indeed, provides an estimate of the number of indi- participants themselves decided which par- viduals who would have been excluded had ticular friend and average other they would such restrictions been imposed. Moreover, bring to mind and describe. preliminary analyses of the Experiment 2 While many extensions of the present data allowed us to examine the effects of work seem promising, space permits con- our selection procedures; those analyses re- sideration of just a few fruitful directions vealed no significant differences between for future research: First, do partners in a participants whose data were retained and given relationship exhibit parallel forms of participants whose data were later deleted illusion in perceiving their relationship (i.e., from analyses. do partners exhibit folie u deux)? It would Third, research utilizing self-report can not be surprising if “the bonds of intimacy be challenged on the grounds that the ob- bring with them a large degree of cognitive tained results reflect self-report bias (e.g., interdependence, a tendency for individu- socially desirable responding, consistency als’ thought processes and structures to be bias). Seeking to translate this liability into mutually determined” (Wegner, 1987, p. an asset, in Experiment 2 we asked partici- 199). Observing the extent to which part- pants to complete the Balanced Inventory ners exhibit parallel illusion-in both de- of Desirable Responding (Paulhus, 1984), gree and form-may aid in explaining the which allowed us to examine the associa- development, maintenance, and functions tions of centrality-based differentiation of illusion. with self-deception and impression man- Second, how does illusion relate not only agement. Self-deception was unrelated to to individual well-being but also to rela- response tendencies, and associations with tionship well-being? While some degree of impression management were complex (i.e., illusion may promote adjustment, extreme inconsistent with straightforward self-en- levels might well be detrimental (cf. Taylor hancement). Thus, our findings cannot read- & Brown, 1994). Where lies the point of ily be accounted for by self-report artifacts maximum yield? One approach to explor- such as desire to present oneself favorably. ing such issues would be to examine the Nevertheless, given that participants may link between illusion and later breakup, or tailor their responses in complex ways and with measures of couple adjustment such as Positive illusion in close relationships 179 the Dyadic Adjustment Scale (Spanier, tiation in perceiving relationships much as 1976;cf. Murray & Holmes, 1997;Rusbult et they do in perceiving individuals. Tenden- al., 1997). Another approach would be to cies toward positive evaluation, optimism investigate whether partners employ illu- regarding the future, and perceptions of sion to mitigate internal or external threats control characterize cognition regarding to a relationship (e.g., Murray & Holmes, both individuals and relationships. Al- 1993;cf. Brickman, Dunkel-Schetter, & Ab- though individual-level and relationship- bey, 1987). level illusion are parallel in many respects, Third, it would be interesting to explore various substantively meaningful differ- how illusion changes over time, if at all. For ences exist in the nature of centrality-based example, when and how does illusion differentiation for the two contexts; also, the emerge? How and why does illusion grow predictors of such differentiation appear to stronger, and how and why does it deterio- differ for the two contexts. These findings rate? Consistent with the logic underlying have important implications for our knowl- the present work, Murray and Holmes edge of the social nature of the self, reveal- (1997) recently demonstrated that the ro- ing differences in the nature of identity mantic relationships of individuals with across interpersonal domains. More broadly, greater levels of illusion are more likely to this research suggests that it may be fruitful persist over time. Given that illusion affects to conceive of self-enhancement in terms of both individual well-being and couple well- centrality-based differentiation, demon- being, the ability to predict change in illu- strating that such an approach may illumi- sion might allow us to inoculate healthy re- nate our understanding of the nature of indi- lationships in preparation for turbulent vidual and social identity. periods. Alternatively, it may be possible to Finally, the present results may encour- shape relationship-specific illusions in such age the sharing of concepts across typically a manner as to enhance functioning in dis- segregated literatures (e.g., social cognition, tressed relationships. personality, ingroupoutgroup relations, close relationships). Exploring such proc- ess-oriented portability of concepts surely Conclusions will lead to an increasingly coherent ap- The present research demonstrated that in- proach to our understanding of human so- dividuals exhibit centrality-based differen- cial behavior.

References

Alicke, M. D. (1985). Global self-evaluation as deter- Brickman, P., Dunkel-Schetter, C., & Abbey, A. (1987). mined by the desirability and controllability of trait The development of commitment. In P. Brickman adjectives. Journal of Personality and Social Psy- (Ed.), Commitment, conflict, and caring (pp. chology, 49,1621-1630. 145-221). Englewood Cliffs, NJ: Prentice-Hall. Aron, A., & Aron, E. N. (1997). Self-expansionmotiva- Brown, J. D. (1986). Evaluations of self and others: tion and including other in the self. In S. Duck Self-enhancement biases in social judgement. So- (Ed.), Handbook of personal relationships: Theory, cial Cognition, 4,353-376. research, and interventions (2nd ed., pp. 251-270). Buunk, B. P., Collins, R. L., Taylor, S. E., Van Yperen, Chichester: Wiley. N. W., & Dakof, G. A. (1990). The affective conse- Aron, A., Aron, E. N., Tudor, M., & Nelson, G. (1991). quences of social comparison: Either direction has Close relationships as including other in the self. its ups and downs. Journal of Personality and Social Journal of Personality and Social Psychology, 60, Psychology, 59,1238-1249. 241-253. Buunk, B. P., & Van Yperen, N. W. (1991). Referential Baumeister, R. F., & Hutton, D. G. (1987). Self-presen- comparisons,relational comparisons,and exchange tation theory: Self-construction and audience context: Their relation to marital satisfaction. Per- pleasing. In B. Mullen & G. R. Goethals (Eds.), sonality and Social Psychology Bulletin, 17, Theories of group behavior (pp. 72-87). New York: 709-717. Springer-Verlag. Campbell, J. D. (1986). Similarity and uniqueness: The Brewer, M. B. (1991). The social self: On being the effects of attribute type, relevance, and individual same and different at the same time. Personality differences in self-esteem and depression. Journal and Social Psychology Bulletin, 17,475-482. of Personality and Social Psychology, 50,281-294. 180 J.M. Martz et al.

Cialdini, R. B.,Borden, R. J.,Thorne, A., Walker, M. R., Personality and Social Psychology Bulletin, 18, Freeman, S., & Sloan, L. R. (1976). Basking in re- 302-318. flected glory: Three (football) field studies.Journal Maass, A,,& Schaller,M. (1991). Intergroup biases and of Personality and Social Psychology, 34,366-375. the cognitive dynamics of stereotype formation. In Cohen, J., & Cohen, P. (1983). Applied multiple regres- W. Stroebe & M. Hewstone (Eds.), European re- sionlcorrelation analysis for the behavioral sciences. view of social psychology (Vol. 2, pp. 189-209). Hillsdale, NJ Erlbaum. Chichester: Wiley. Cooper, W. H. (1981). Ubiquitous halo. Psychological Maxwell, S. E., & Delaney, H. D. (1990). Designing Bulletin, 90,218-224. experiments and analyzing data: A model compari- Crocker, J., & Luhtanen, R. (1990). Collective self-es- son perspective. Belmont, CAWadsworth. teem and ingroup bias. Journal of Personality and Miller, A. G., Ashton, W. A,, McHoskey, J.W., & Gim- Social Psychology, 58,6047. bel, J. (1990). What price attractiveness? Stereotype DeJoy, D. M. (1989). The optimism bias and traffic and risk factors in suntanning behavior. Journal of accident risk perception. Accident Analysis and Applied Social Psychology, 20,1272-1300. Prevention, 21,333-340. Murray, S. L., & Holmes, J. G. (1993). Seeing virtues in Ephron, N. (1983). Heartburn. New York: Knopf. faults: Negativity and the transformation of inter- Goethals, G. R., & Darley, J. M. (1987). Social compari- personal narratives in close relationships Journal son theory: Self-evaluation and group life. In B. of Personality and Social Psychology, 65,707-722. Mullen & G. R. Goethals (Eds.), Theories of group Murray, S. L., & Holmes, J. G. (1997). A leap of faith?: behavior (pp. 21-47). New York Springer-Verlag. Positive illusions in romantic relationships. Person- Goffman, E. (1959). The presentation of self in every- ality and Social Psychology Bulletin, 23, 586-604. day life. Garden City, NY Doubleday. Murray, S. L., Holmes, J. G., & Griffin, D. W. (1996a). Hall, J., & Taylor, S. E. (1976). When love is blind The benefits of positive illusions: Idealization and Maintaining idealized images of one’s spouse. Hu- the construction of satisfaction in close relation- man Relations, 29,751-761. ships.Journal of Personality and Social Psychology, Hoyle, R. H. (1991). Evaluating measurement models 70, 79-98. in clinical research: Covariance structure analysis Murray, S. L., Holmes, J. G., & Griffin, D. W. (1996b). of latent variable models of self-conception.Jour- The self-fulfilling nature of positive illusions in ro- nal of Consulting and Clinical Psychology, 59, mantic relationships: Love is not blind, but pre- 67-76. scient. Journal of Personality and Social Psychol- Insko, C. A., Schopler, J., Hoyle, R. H., Dardis, G. J., & ogy, 71, 1155-1180. Graetz, K. A. (1990). Individual-group discontinu- Nisbett, R. E., & Ross, L. (1980). Human inference: Strategies and shortcomings of social judgement. ity as a function of fear and greed. Journal of Per- sonality and Social Psychology, 58,68-79. Englewood Cliffs, NJ Prentice-Hall. Nisbett,R.E.,& Wilson,T.D. (1977).Tellingmore than Jones, E. E., & Davis, K. E. (1965). From acts to dispo- sitions: The attribution process in person percep- we can know: Verbal reports on mental processes. Psychological Review, 84, 231-259. tion. In L. Berkowitz (Ed.), Advances in experi- Paulhus, D. L. (1984). Two-component models of so- mental social psychology (Vol. 2, pp. 283-329). New cially desirable responding. Journal Personality York: Academic Press. of and Social Psychology, 46, 598-609. Kelley, H. H. (1983). Love and commitment. In H. H. Reis, H. T., & Wheeler, L. (1991). Studying social inter- Kelley, E. Berscheid, A. Christensen, J. H. Harvey, action with the Rochester Interaction Record. In T. L. Huston, G. Levinger, et al. (Eds.), Close rela- M. P. Zanna (Ed.), Advances in experimental social tionships (pp. 265-314). New York: W. H. Freeman. psychology (Vol. 24, pp. 269-318). San Diego: Aca- Kelley, H. H., & Thibaut, J. W. (1978). Interpersonal demic Press. relations: A theory of interdependence. New York: Rusbult, C. E. (1983). A longitudinal test of the invest- Wiley. ment model The development (and deterioration) Langer, E. J. (1989). Minding matters: The conse- of satisfaction and commitment in heterosexual in- quences of mindlessness-mindfulness. In L. Berk- volvements. Journal of Personality and Social Psy- owitz (Ed.), Advances in experimental social psy- chology, 45, 101-117. chology (Vol. 22, pp. 137-173). New York: Rusbult, C. E., & Buunk, B. I? (1993). Commitment Academic Press. processes in close relationships: An interdepend- Lerner,M. J., & Miller, D. T. (1978). Just world research ence analysis. Journal of Social and Personal Rela- and the attribution process: Looking back and tionships, 10, 175-204. ahead. Psychological Bulletin, 83,1030-1051. Rusbult, C. E., & Martz, J. M. (1995). Remaining in an Levine, J. M., & Moreland, R. L. (1987). Social com- abusive relationship: An investment model analy- parison and outcome evaluation in group contexts. sis of nonvoluntary commitment. Personality and In J. C. Masters & W. I? Smith (Eds.), Social com- Social Psychology Bulletin, 21, 558-571. parison, social justice, and relative deprivation: Rusbult, C. E., & Van Lange, I? A. M. (1996). Interde- Theoretical, empirical, and policy perspectives (pp. pendence processes. In E. T. Higgins & A. Kruglan- 105-127). Hillsdale, NJ: Erlbaum. ski (Eds), Social psychology: Handbook of basic Lipkus, I., Martz, J. M., Panter, A. T., Drigotas, S. M., & principles (pp. 564-596). New York: Guilford Press. Feaganes, J. R. (1993). Do optimists distort their Rusbult, C. E., Van Lange, I? A. M., Verette, J., predictions of future positive and negative events? Yovetich, N. A., & Wildschut, R. T. (1997). A func- Journal of Personality and Individual Differences, tional analysis ofperceived superiority in close rela- 15,577-589. tionships. Unpublished manuscript, University of Luhtanen, R., & Crocker, J. (1992). A collective self-es- North Carolina at Chapel Hill, Chapel Hill, NC. teem scale: Self-evaluation of one’s social identity. Rusbult, C. E., Verette, J., Whitney, G. A., Slovik, L. F., Positive illusion in close relationships 181

& Lipkus, I. (1991). Accommodation processes in being: A social psychological perspective on men- close relationships:Theory and preliminary empiri- tal health. Psychological Bulletin, 103,193-210. cal evidence. Journal of Personality and Social Psy- Taylor, S. E., & Brown, J. D. (1994). Positive illusions chology, 60, 53-78. and well-being revisited: Separating fact from fic- Schaller, M. (1992). In-group favoritism and statistical tion. Psychological Bulletin, 116,21-27. reasoning in social inference: Implications for for- Taylor, S. E., & Koivumaki, J. H. (1976). The perception mation and maintenance of group stereotypes. of self and others: Acquaintanceship, affect, and Journal of Personality and Social Psychology, 63, actor-observer differences. Journal of Personality 61-74. and Social Psychology, 33,403408. Scheier, M. F., & Carver, C. S. (1993). On the power of Thibaut, J. W., & Kelley, H. H. (1959). The socialpsy- positive thinking: The benefits of being optimistic. chology of groups. New York: Wiley. Current Directions in Psychological Science, 2 (l), Turner, J. C. (1987). Rediscovering the social group: A 26-30. self-categorization theory. Oxford: Blackwell. Schopler, J., & Insko, C. A. (1992). The discontinuity Vallone, R. P., Griffin, D. W., Lin, S.,& Ross, L. (1990). effect in interpersonal and intergroup relations: Overconfident prediction of future actions and Generality and mediation. In W. Stroebe & M. outcomes by self and others. Journal of Personality Hewstone (Eds), European review of social psy- and Social Psychology, 58,582-592. chology (Vol. 3, pp. 121-151). Chichester: Wiley. Van Knippenberg, A., & Ellemers, N. (1990). Social Schopler, J., Insko, C. A., Graetz, K. A., Drigotas, S. M., identity and intergroup differentiation processes. & Smith, V. A. (1991).The generality of the individ- In W. Stroebe & M. Hewstone (Eds.), European ual-group discontinuity effect: Variations in posi- review of social psychology (Vol. 1, pp. 137-169). tivity-negativity of outcomes, players’ relative Chichester: Wiley. power, and magnitude of outcomes. Personality Van Lange, P. A. M. (1991). Being better but not and Social Psychology Bulletin, 17, 612-624. smarter than others: The Muhammad Ali effect at Sears, D. 0. (1983). The person-positivity bias. Journal work in interpersonal situations. Personality and of Personality and Social Psychology, 44, 233-250. Social Psychology Bulletin, 17,689-693. Sedikides, C., Olsen, N., & Reis, H. T. (1993). Relation- Van Lange, P. A. M., & Rusbult, C. E. (1995). My ships as natural categories. Journal of Personality relationship is better than-and not as bad and Social Psychology, 64, 71-82. as-yours is: The perception of superiority in close Showers, C. (1992). Compartmentalization of positive relationships. Personality and Social Psychology and negative self-knowledge: Keeping bad apples Bulletin,21, 324. out of the bunch. Journal of Personality and Social Webb, E. J., Campbell, D. T., Schwartz, R. D., & Psychology, 62, 10361049. Sechrest, L. (1966). Unobtrusive measures: Nonre- Smith, E. R. (1993). Social identity and social emo- active research in the social sciences. Chicago: Rand tions: Toward new conceptualizations of prejudice. McNall y. In D. M. Mackie & D. L. Hamilton (Eds.), Affect, Wegner, D. M. (1987). Transactive memory: A contem- cognition, and stereotyping: Interactive processes in porary analysis of the group mind. In B. Mullen & group perception (pp. 297-315). San Diego: Aca- G. R. Goethals (Eds.), Theories of group behavior demic Press. (pp. 185-208). New York: Springer-Verlag. Spanier, G. B. (1976). Measuring dyadic adjustment: Weinstein, N. D. (1980). Unrealistic optimism about New scales for assessing the quality of marriage future events. Journal of Personality and Social and similar dyads. Journal of Marriage and the Psychology, 39,806-820. Family, 38,15-28. Weinstein, N. D. (1984). Why it won’t happen to me: Tajfel, H., & Turner, J. C. (1986). The social identity Perceptions of risk factors and susceptibility. theory of intergroup behavior. In S. Worchel & W. Health Psychology, 3,431-457. Austin (Eds.) Psychology of intergroup relations Wood, J. V. (1989). Theory and research concerning (pp. 7-24). Chicago: Nelson-Hall. social comparisons of personal attributes. Psycho- Taylor S. E., & Brown, J. D. (1988). Illusion and well- logical Bulletin, 106,231-248.