Personal Relationships, 6 (1999), 269-289. Printed in the United States of America. Copyright 0 1999 ISSPR. 1350-4126/99$9.50

Being better than others but otherwise perfectly normal: Perceptions of uniqueness and similarity in close relationships

PAUL A. M. VAN LANGE,” CARYL E. RUSBULT? ASTRID SEMIN-GOOSSENS,L’CARIEN A. GORTS,CAND MIRJAM STALPERSc aFree University, Amsterdam; bUniversity of North Carolina, Chapel Hill; and ‘Free University, Amsterdam

Abstract The current research addresses individuals’ self-generated thoughts regarding their own and others’ relationships,examining the ways in which perceptions of uniqueness and similarity are manifested in judgments regarding own and others’ responses to dissatisfying incidents. Consistent with the uniqueness bias, participants characterized their own relationships by a greater number of constructive responses and a smaller number of destructive responses relative to characterizations of others’ relationships. Moreover, external raters judged own constructive responses to be more constructive than others’ constructive responses. Consistent with the similarity bias, external raters judged items describing others’ responses to be less frequently occurring and more extreme than their own responses. Also, this research revealed support for the claim that the similarity bias is more pronounced for destructive responses than for constructive responses. A recall task corroborated these findings, revealing very good recall for destructive responses enacted by others and poor recall for destructive responses enacted by oneself.

Virtually all relationships eventually are incidents. Broadly conceived, such re- confronted with periodic declines in satis- sponses may be construed as either con- faction-declines caused by relatively en- structive or destructive in that they either during disagreements, conflicts, or unpleas- contribute to the health and vitality of the ant events that are external to the relationship or are harmful to the well-be- relationship itself. Individuals may respond ing of the relationship (cf. Clark & Reis, in a variety of ways to such dissatisfying 1988; Holmes & Boon, 1990; Rusbult, Verette, Whitney, Slovik, & Lipkus, 1991). How do individuals perceive their construc- This research was supported in part by a grant to Paul tive and destructive responses to such inci- Van Lange from the Netherlands Organization for dents? How do individuals think about their Scientific Research (NWO, Grant No. R-57-178), and by a grant to Caryl Rusbult from the National Science own reactions in relation to the behavior of Foundation (Grant No. BNS-9023817). The authors other individuals? And given that destruc- are grateful to Bram Buunk, Jennifer Crocker, Wilma tive responses are potentially detrimental Otten, and several anonymous reviewers for their use- to the health and stability of a relationship, ful comments on an earlier draft of this article. how do individuals interpret and perhaps Correspondence concerning this article should be addressed to Paul Van Lange, Department of Social justify their destructive actions? Psychology, Free University at Amsterdam, Van der The current research seeks to understand Boechorststraat 1, 1081 BT Amsterdam, The Nether- the nature of the beliefs that individuals lands. E-mail: [email protected]. hold regarding relationship functioning, 269 270 PA.M. Van Lange et al. examining the nature of individuals’ self- 1977; Marks & Miller, 1987; Wood, 1989).* generated thoughts regarding their own and However, nearly all studies have focused on others’ constructive and destructive re- either the uniqueness bias or the similarity sponses to dissatisfying incidents. On the ba- bias, thereby largely overlooking the possi- sis of the research regarding social judgment bility that by examining such biases in con- and social comparison, we propose that indi- cert we may illuminate our understanding viduals’ beliefs regarding their own and oth- of the nature of self-other perception ers’ relationship functioning will reflect two (however, see Campbell, 1986; Marks, 1984; relatively pervasive tendencies: (a) unique- McFarland & Miller, 1990). ness bias, or a tendency to overestimate Second, prior research on close relation- one’s own uniqueness, exaggerating the su- ships has examined issues such as perceived periority of one’s own reactions and behav- (and actual) similarity and uniqueness be- iors, and (b) similarity bias, or a tendency to tween partners in a relationship (self-other overestimate the similarity between oneself perceptions within a relationship; e.g., and others, exaggerating the commonness Acitelli, Douvan, & Veroff, 1993; Ross & or “everydayness” of one’s own reactions Sicoly, 1979;for a review, see Sillars & Scott, and behaviors.This framework assumes that 1983). However, very few studies have ex- these two seemingly contradictory be- amined individuals’ beliefs regarding their liefs-uniqueness and similarity-coexist own functioning in relation to others in- and that these biases serve the function of volved in other relationships, capturing as- sustaining the general, self-enhancing belief pects of self-other perceptions among rela- that one is “uniquely better than others but tionships. We suggest that, in addition to the otherwise perfectly normal.” uniqueness bias, the similarity bias under- The current work extends prior research lies such perceptions, particularly for those in several ways. First, the uniqueness and responses that are potentially harmful to similarity biases are pervasive tendencies health and vitality of a relationship. If true, that have been demonstrated to exist in a such self-other perceptions may help un- variety of judgment domains (e.g., in re- derstand why partners may sometimes per- search on “perceived superiority” and sist in particular types of responses (e.g., a “false consensus”; Ross, Greene, & House, reluctance to talk things over), in that they may genuinely believe (and to some degree overestimate) that their own responses are 1. Throughout this article we will employ theconcepts very common, frequently occurring, and of “uniqueness bias” and “similarity bias” to refer to beliefs that are also known as “perceived supe- therefore appropriate (e.g., “Like others, I riority” (e.g., Wood, 1989) and “false consensus” do not talk about every slight problem we (Ross, Greene, & House, 1977).While common lan- have”). guage would suggest that the uniqueness bias in- Finally, as noted by Berscheid (1994, p. cludes a tendency to regard oneself as distinctly in- 84), “Relationships researchers are begin- ferior to others, it should be clear that this bias is defined as the tendency to regard oneself as dis- ning to take advantage of the theoretical tinctly superior to others. The primary reason for and empirical fruits of social cognition to employing the concepts of uniqueness bias and simi- further their understanding of interpersonal larity bias is that these concepts (a) explicate the relationships. . . . Thus the rapprochement contradictory nature of these beliefs, (b) do not that has begun between the fields of social seem to be domain-specific, and (c) have been em- ployed in prior research that has examined simulta- cognition and interpersonal relationships neously both beliefs (e.g., Campbell, 1986; Goethals has the potential to benefit both endeavors.” et al., 1991). In contrast, the concepts of superiority and consensus seem less contradictory and could be construed as relatively more domain-specific, with that are not universally appreciated or universally superiority primarily applying to attributes that are depreciated (e.g., opinions and preferences). As universally appreciated (ability, morality) or uni- will be outlined, the current research focuses on a versally depreciated (e.g., lack of ability, lack of mo- domain that presumably includes both types of at- rality), and with consensus applying to attributes tributes. Uniqueness and similarity 271

By examining self-other perceptions in the adaptive to regard one’s behavior in rela- context of close relationships, we attempt to tively favorable terms by viewing one’s own enhance the linkage between the areas of actions and responses as quite positive, rea- social judgment and close relationships. sonable, and appropriate. The prominence of such favorable beliefs may promote feel- ings of self-worth and happiness, provide Perceptions of Uniqueness and Similarity validation of the quality of one’s relation- in Close Relationships ship, and enhance perceptions that presum- Past research has consistently revealed that ably are relevant-and perhaps neces- individuals are not always accurate and un- sary-to sustaining a relationship (e.g., biased in perceiving social reality, particu- feelings of trust, security, and optimism; larly when this reality has implications for Holmes & Boon, 1990; Van Lange & Rus- the self or close others. One pervasive ten- bult, 1995). dency has been termed the uniqueness bias From a functional perspective, it has (Goethals, Messick, & Allison, 1991), in ref- been argued that uniqueness and similarity erence to the inclination to overestimate biases are differentially functional depend- the superiority of one’s behavior and char- ing on whether the attribute at issue is a acteristics (e.g., perceptions of superiority universally evaluated one (that is, people regarding one’s fairness, generosity, and exhibit good agreement about whether the ability; Alicke, 1985; Brown, 1986; Messick, attribute is desirable or undesirable) or a Bloom, Boldizar, & Samuelson, 1985; Van variably evaluated one (that is, people dis- Lange, 1991). An equally pervasive ten- agree about whether the attribute is desir- dency has been termed the similarity bias, in able, and rate their own position on the reference to the inclination to overestimate attribute dimension most positively; Sher- the commonness of one’s reactions, opin- man, Chassin, Presson, & Agostinelli, 1984). ions, and beliefs (e.g., judgments of Presi- For universally evaluated attributes, dent Carter’s performance; whether to close uniqueness and similarity judgments should curtains on winter nights; Goethals, Allison, depend on the desirability of one’s own po- & Frost, 1979; Van der Pligt, 1984). As out- sition. Individuals should assume unique- lined by Ross et al. (1977, p. 280), “Individu- ness when they regard their standing on the als tend to see their own behavioral choices attribute dimension as more desirable (e.g., and judgments as relatively common and “Unlike others, I never lie to my partner”) appropriate to existing circumstances, while and should assume similarity when they re- viewing alternative responses as uncom- gard their standing as undesirable (e.g., mon, deviant or inappropriate.” While these “Like most others, I become irritated when biased perceptions typically have been I have to wait for my partner”). For variably examined in two separate lines of research, evaluated attributes, individuals should be it seems plausible that both biases-seem- motivated to seek similarity because such ingly contradictory-underlie beliefs re- beliefs provide “consensual validation,” in- garding one’s own and others’ relationships. dicating the appropriateness, “reasonable- How can it be that individuals regard them- ness,” and correctness of one’s own position selves as both unique and similar to others? (e.g., “Like most other people, we eat din- A fair amount of evidence suggests that ner around 6:OO P.M.”; Campbell, 1986; it may be functional to maintain a relatively Goethals et al., 1979). favorable view of one’s own actions. As out- With the possible exceptions of individ- lined by Taylor and Brown (1988, 1994), ual abilities and some opinions, many be- self-enhancing views may stimulate feelings haviors and attitudes relevant to under- of happiness and contentment, the ability to standing ongoing relationships may entail a care for and about others, and the capacity combination of universally and variably for creative, productive work. In the con- evaluated attributes. That is, most such be- text of relationship functioning, it may be haviors have evaluative meaning (e.g., to 272 RA.M. Van Lange et al. sacrifice for the sake of the partner is good), Study 1 but at the same time, the most extreme form Study 1 examines self-generated thoughts of the behavior presumably is not the most regarding one’s own and others’ relation- desirable form (e.g., too much sacrifice is not ships, asking participants to list constructive good), and individuals may disagree about and destructive responses that are believed the specific form of the behavior that is most to be performed relatively more frequently desirable (cf. Van Lange & Kuhlman, 1994; by either self or others. Also, the present Van Lange, Agnew, Harinck, & Steemers, study asks participants to compare them- 1997). Accordingly, we assume that re- selves with same-sex others (as opposed to sponses to dissatisfying incidents in relation- nonspecified others in Van Lange & Rus- ships have both a universally evaluated bult, 1995), forcing individuals to relate component (i.e., behaving constructively is their responses to what is believed to be an better than behaving destructively) and a important reference group in the context of variably evaluated component. close relationships (cf. Buunk & Van The above characterization of construc- Yperen, 1991; Peplau, 1983). tive and destructive responses to incidents Following the uniqueness bias, we ex- of dissatisfaction suggests the functionality pected to replicate the findings of Van of both the uniqueness bias and the similar- Lange and Rusbult (1995) and predicted ity bias. First, the uniqueness bias, or the that (a) participants will list a greater tendency to regard one’s responses as more number of constructive responses for their constructive and less destructive than those own relationships than for others’ relation- of others, resembles perceived superiority ships (“being better than others,”cf. positive regarding relationship functioning. Second, superiority; Hypothesis la); and (b) partici- the similarity bias, or regarding one’s re- pants will list fewer destructive responses sponses as common and not extreme or de- for their own relationships than for others’ viant, provides consensual validation for relationships (“being not as bad as others,” the appropriateness of own relationship cf. negative superiority; Hypothesis 1b). Fi- functioning. Third, it seems reasonable to nally, we are interested in exploring possible assume that relative to constructive re- gender effects, particularly the tendency of sponses, individuals should feel a stronger women rather than men to hold a greater need to explain or justify their destructive number of negative beliefs about others’ re- responses, in that such responses are poten- lationships,as observed in several studies re- tially harmful to the health and vitality of a ported by Van Lange and Rusbult (1995). relationship and because such responses may be discrepant with a favorable image of the self (i.e., “I do something that could Method be construed as ‘bad”’). How, then, might individuals justify such behavior? We pro- Participants. One hundred five individuals pose that individuals tend to justify their who were involved in an ongoing relation- destructive responses by viewing such re- ship participated in Study 1 (57 women, 47 sponses as relatively common and appro- men, and 1individual who failed to indicate priate to existing circumstances, a percep- gender). Participants were recruited at a va- tion that provides validation for the riety of locations on the campus of the Free correctness or appropriateness of one’s be- University at Amsterdam, including the havior (cf. Goethals et al., 1979; Marks & university cafeteria and library. Participants Miller, 1987). Given that there should be a were 23.95 years old on average (23.35 for stronger motivation to justify one’s destruc- women, 24.68 years for men), and their re- tive responses than one’s constructive re- lationships were an average of 32.46 sponses, individuals should assume greater months in duration (33.02 for women, 31.79 similarity for their destructive responses months for men). Participants were ran- than for their constructive responses. domly assigned to one of two experimental Uniqueness and similarity 273 conditions, with approximately equal num- list as many examples as possible of constructive bers of women and men in each condition. behaviors that you perform more often in your relationship than other same-sex individuals do in their relationships. Use a new line for each Procedure. The experiment was described response, and describe the behavior briefly as a study of romantic relationships. Partici- (maximum of 12 words).2 pants were told that the experiment would last about 15 minutes, and were assured Participants were given 4 minutes to list that their individual responses would re- behaviors. Following this, participants were main strictly anonymous. Participants were asked to rank-order the items in terms of asked to list the behaviors that came to the degree to which each item exemplified mind in thinking about responses to inci- the constructive responses that they per- dents of dissatisfaction in their own and form more frequently than do other indi- others’ relationships. Item Valence was ma- viduals. These ratings were used to identify nipulated by asking half of the participants “the best items” for use in Study 2, which to list constructive responses and asking the we describe later. Participants were in- other half to list destructive responses. Or- structed to rank-order the items by assign- der of Tasks was also varied: Half of the ing numbers to each behavior, assigning a participants began by listing behaviors that “1” to the best example, assigning a “2” to they performed more frequently than did the second best example, and so on. They other individuals, and then listed behaviors were asked not to assign the same rank to others enacted more frequently than they two items (i.e., no ties). Next, participants themselves did; the other half of the partici- were asked to generate as many examples pants listed “other” behaviors prior to list- as possible of constructive responses that ing “own” behaviors. In addition, Instruc- they perform less often than others-the tional Set was varied: Half of the Constructive Item Valence, Other Item Ob- participants listed responses that were ject condition. Participants also rank-or- more characteristic of their own relation- dered these items in terms of the degree to ships than that of others’ relationships (i.e., which each item exemplified the construc- focus of comparison was on self, or one’s tive responses they performed less fre- own relationship), and half listed responses quently than did other individuals. As noted that were less characteristic of others’ rela- earlier, Order of Tasks and Instructional Set tionships than that of their own relation- were counterbalanced across participants. ships (i.e., focus of comparison was on oth- In the Destructive Item Valence conditions ers). The analyses revealed no significant the instructions were identical, except for main effects or interactions involving Or- the substitution of “One may respond de- der of Tasks or Instructional Set, so these structiveZy to such situations, that is, respond variables will not be described further. in ways that are harmful to the future of the In the Constructive Item Valence, Own Item Object condition, participants read the following instructions: 2. To assist participants in generating items, the in- structions employed prompts that paralleled those Everyone experiences occasional dissatisfying used in research by Van Lange and Rusbult (1995; incidents in their relationships-times when one Experiment 1). This research revealed that the or the other partner feels unhappy, upset, or presence versus absence of such prompts (i.e., a angry about something (something the other comparison of Experiments 1 and 2) did not affect said or did, something about the relationship (a) the number of items generated, or (b) the per- itself, etc.). One may respond constructively to ceived constructiveness of the self-generated items in each category. At the same time, we cannot ex- such situations, that is, respond in ways that are clude the possibility that such prompts affect other helpful to the future of the relationship (e.g., features of the items that participants generated talking things over, changing the behavior so as (e.g., the extremity and frequency of the re- to solve the problem), On the next page please sponses). 274 PA.M. Van Lunge et al.

relationship (e.g., avoiding the problem, Study 1 participants and ask a new set of suggesting ending the relationship).” participants to rate the responses in terms of constructiveness (ratings relevant to the uniqueness bias), as well as frequency and Results and discussion extremity (ratings relevant to the similarity The number of items generated by partici- bias). Moreover, we ask participants pants was analyzed in a 2 (Item Valence: whether they perform each response more Constructive vs. Destructive responses) by or less frequently than others (these judg- 2 (Item Object: Own vs. Other responses) ments will be referred to as beliefs regard- by 2 (Gender of Participant: Women vs. ing self-other prevalence). Men) analysis of variance (ANOVA), with The goal of Study 2 is twofold. First, Item Object represented as a within-subject given that participants in Study 2 did not variable. Relevant to the predicted positive generate the items they are asked to rate, and negative superiority effects, this analy- these participants serve as external and sis revealed a significant interaction of Item blind judges of the constructive and de- Valence and Item Object, F(1,lOO) = 28.32, structive items (i.e., two levels of a factor we p < ,001. Follow-up tests of simple effects label Item Valence) generated by Study 1 revealed that (a) individuals listed a greater participants as typical of self versus others number of constructive responses for them- (i.e., two levels of a factor we label Item selves, M = 3.26, SD = 1.43, than for others, Object). Thus, these external judges serve M = 2.78, SD = 1.42; F(1,53) = 4.44, p < to illuminate the nature of the constructive .05, supporting Hypothesis la (i.e., positive and destructive responses ascribed to self superiority), and (b) individuals listed and others (i.e., for the sake of clarity, we fewer destructive responses for themselves, refer to these ratings as external judg- M = 2.58, SD = 1.49, than for others, A4 = ments). 3.96, SD = 2.58; F(1,49) = 25.39,~< .001, Second, because we examine beliefs re- supporting Hypothesis lb (i.e., negative su- garding self-other prevalence of the re- periority). There was also a significant main sponses, Study 2 allows us to identify the effect of Item Object, F(1,lOO) = 6.94, p < major predictors of such judgments (i.e., .01, with individuals listing a greater perceptions of constructiveness/destruc- number of responses for others (M = 3.35, tiveness, frequency, and extremity). Thus, SD = 1.73) than for themselves (M = 2.93, these judgments serve to identify what it SD = 1.49). No other main effects or inter- is about constructive and destructive re- actions were statistically significant. Thus, sponses that leads Study 2 participants to relative to their characterizations of others’ conclude whether they themselves or oth- relationships, participants characterized ers are more likely to engage in such their own relationship by a greater number responses (i.e., for the sake of clarity, we of constructive items and a smaller number refer to these ratings as self-relevant judg- of destructive items. This supports the ments). uniqueness bias. Hypotheses regarding the nature of con- structive and destructive responses for self and others. The first goal of Study 2 is to Study 2 assess the nature of the constructive and Study 2 is designed to provide preliminary destructive responses for self and others by evidence relevant to the claim that indi- examining “external” judgments of these viduals simultaneously exhibit uniqueness responses. On the basis of the foregoing dis- and similarity biases, and that the similarity cussion regarding uniqueness and similarity bias is relatively more pronounced for de- biases, we advanced the following hypothe- structive responses than for constructive re- ses. First, consistent with the uniqueness sponses. In Study 2 we employ a random bias, we advance two specific predictions: selection of self-generated responses from Hypothesis lc predicts that own responses Uniqueness and similarity 275 will be rated as more constructive and less perceived constructiveness will contribute destructive than other responses; and Hy- to predicting self-other prevalence (a pre- pothesis Id predicts that constructive re- diction reminiscent of Hypothesis 1);on the sponses will be rated as more frequently basis of the similarity bias, we predict that performed by the self than by others, perceived frequency and extremity will also whereas the opposite should occur for de- contribute to predicting self-other preva- structive responses. lence (a prediction reminiscent of Hy- Second, on the basis of the similarity pothesis 2). And, on the basis of the asser- bias, Hypothesis 2 predicts that own re- tion that the similarity bias is relatively sponses will be rated as more frequent and more important in accounting for less extreme than other responses (i.e., a self-other prevalence regarding destructive main effect of Item Object on ratings of responses than constructive responses, we frequency and extremity). Third, we pro- examine whether levels of perceived fre- pose that individuals will assume greater quency and extremity make greater contri- similarity for destructive responses than butions to predicting self-other prevalence for constructive responses. Accordingly, for destructive responses than for construc- Hypothesis 3 predicts that the tendency to tive responses (a prediction reminiscent of rate other responses as less frequent and Hypothesis 3). more extreme than own responses will be more pronounced for destructive re- Method sponses than for constructive responses (i.e., an interaction of Item Valence and Participants and design. Eighty-three stu- Item Object for ratings of frequency and dents at the Free University at Amsterdam extremity). participated in Study 2 (44 women, 39 men). Finally, we advance one prediction that Participants were recruited by means of an is not directly relevant to the uniqueness advertisement placed in a local university and similarity biases, but that is relevant to paper. The advertisement invited individu- the manner in which individuals process in- als who were involved in ongoing dating formation about constructive and destruc- relationships to participate in an experi- tive behaviors. Earlier, we assumed that at ment entitled “Interpersonal Relation- the global level individuals may hold rela- ships.” Participants were 22.65 years old on tively positive beliefs regarding relation- average (22.27 years for women, 23.08 for ship functioning, and that destructive re- men), and their relationships were an aver- sponses therefore should stand out as more age of 29.68 months in duration (30.14 extreme and less frequent than constructive months for women, 29.18 months for men). responses. Accordingly, Hypothesis 4 pre- Judges were paid 12.50 Dutch guilders for dicts that constructive responses will be participation in Phase 2 (approximately rated as less extreme and more frequent $7.00 in American currency). than destructive responses (i.e., a main ef- Forty items from Study 1 were randomly fect of Item Valence for ratings of fre- selected from the items that were rank-or- quency and extremity). dered as “best examples” (i.e., rated as number one) by Study 1 participants (10 Hypotheses regarding beliefs about items from each experimental condition; self-other prevalence. The second goal of see Table 1). Study 2 participants made Study 2 is to examine the extent to which judgments of these items on three dimen- perceived constructiveness, frequency, and sions (constructiveness, frequency, and ex- extremity contribute to the prediction of tremity). Study 2 employs a 2 (Item Va- beliefs regarding self-other prevalence for lence: Items describing constructive vs. constructive and destructive responses (i.e., destructive responses) by 2 (Item Object: these are self-relevant judgments). On the Responses performed by self vs. others) by basis of the uniqueness bias, we predict that 2 (Gender of Participant: Women vs. men) 276 T1A.M. Van Lange et al.

Table 1. Forty randomly selected responses to dissatisfaction

Constructive Item ValencefOwn Item Object to listen to the partner to forgive to first carefully listen, and later express your own opinion to discuss disagreements to laugh about “the problem” to consider the partner’s opinion as equally important as your own to talk about it, and to attempt to solve it to trust one another to talk to directly ask what is on his or her mind

Constructive Item Valence/Other Item Object to talk and to encourage the partner to initiate settling the argument to sit around the table, and talk about it to clearly express what one feels to ask “if there is something I can do” to make the partner feel better about himself or herself to relieve the partner to tolerate certain irritations about the partner to put yourself (and your problems) aside to avoid problems to settle the argument to be assertive Destructive Item Valence/Own Item Object to be indifferent to pretend nothing has happened to say nothing to be impatient to overgeneralize the problem to have too bad a picture of the problem: to exaggerate to be cynical or sarcastic to not talk about it to doubt the current state or quality of the relationship to scream without listening to the partner

Destructive Item Valence/Other Item Object to not be willing or able to show understanding to hit the partner to avoid all problems to fight to not adjust to the partner to avoid each other in some situations to be stubborn, to never change one’s own opinion to hit each other during fights to be jealous to keep silent until the partner starts talking

factorial design. Item Valence and Gender Everyone experiences occasional dissatisfying are between-participant variables, and Item incidents in their relationships-times when one Object is a within-participant variable. or the other partner feels unhappy, upset, or angry about something (something the other said or did, something about the relationship Procedure. As in Study 1, the instructions itself, etc.). Individuals may differ in how they began with a brief introduction. respond to such situations. In this experiment, Ufiiqueness and similarity 277 we will present a number of such responses, and cause the current scales differ from more will ask you to make judgments with respect to commonly used measures of similarity.Most each of these responses, or behaviors. prior research has examined such judg- ments by asking participants to estimate the The 40 responses listed in Table 1 (these percentage of others who would exhibit the responses are translations from Dutch into same (versus different) preference or per- English) were randomly sampled by a non- form the same (versus different) behavior as Dutch-speaking individual who was blind to themselves (this measurement technique is the experimental purposes and hypotheses. most notable in research on “false consen- The first 10 randomly selected behaviors sus”; cf. Ross et al., 1977).Alternatively, such that fell into each of the four experimental judgments have been examined by asking conditions were included in Study 2: 10 con- individuals to rate their own position along a structive own responses, 10 constructive continuous dimension, and then to ascribe a other responses, 10 destructive own re- position to an individual target person (this sponses, and 10 destructive other responses. measurement technique is most notable in Within each of the four conditions, we sys- research on “assumed similarity” or “at- tematically controlled for gender of writer: tributive projection”; cf. Cameron & Mar- The first five items in each category were garet, 1951;Cronbach, 1955).Although such generated by men and the latter five items measures have been shown to be valid and were generated by women (see Table 1). useful (i.e., such measures exhibit consistent The order in which the items were listed was patterns across studies), we employed alter- randomized prior to the judgment tasks. native measures so as to examine the Ross et Participants were asked to make three al. (1977) claim that individuals tend to re- types of judgment with respect to each of gard their own choices as relatively com- the 40 responses. Each judgment was per- mon, while viewing alternative responses as formed using a 9-point scale. First, to assess uncommon and deviant. Thus, our measures beliefs regarding self-other prevalence, we assess judgments of how frequently each re- asked participants to estimate the relative sponse occurs in others’ relationships, as frequency with which they perform each well as the judged extremity of each re- response in comparison to same-sex others sponse. We employed two measures to tap (0 = “I do this much less often in my rela- two overlapping but somewhat different ex- tionship than do other same-sex individuals pressions of similarity. in their relationships”; 4 = “I do this about Judgments of frequency were obtained as often as do other same-sex individuals in by asking participants to rate the frequency their relationships”; 8 = “I do this much with which each response occurred in oth- more often in my relationship than do other ers’ relationships, on average (not your own same-sex individuals in their relation- relationship). We included the phrase “not ships”). Second, participants made judg- your own relationship” to discourage par- ments with respect to constructiveness ver- ticipants from basing such judgments on sus destructiveness. This task was their own experience (cf. Alicke & Largo, introduced as follows: “Behaviors differ in 1995). Concrete anchors were included for the extent to which they positively contrib- each scale point (0 = “virtually never”; 1 = ute to the future of a relationship (that is, “about once a year”; 2 = “several times a constructive behaviors) or are harmful to year”; 3 = “about once a month”; 4 = “sev- the future of a relationship (that is, destruc- eral times a month”; 5 = “about once a tive behaviors)” (0 = “extremely destruc- week”; 6 = “several times a week”; 7 = tive,” 4 = “neither destructive nor construc- “about once a day”; and 8 = “virtually al- tive,” 8 = “extremely constructive”). ways, several times a day”). Global anchors Judgments of frequency and extrem- were not employed because such scales ity-which were employed to examine the leave individuals free to employ idiosyn- similarity bias-deserve some comment be- cratic norms or base rates in their fre- 278 RA.M. Van Lunge et al. quency judgments (e.g., “hitting the part- these judgments was analyzed in a 2 (Item ner” might be judged to occur as frequently Valence: Constructive vs. Destructive re- as “initiate settling the argument” owing to sponses) by 2 (Item Object: Own vs. Other the application of idiosyncratic norms).3 responses) by 2 (Gender: Women vs. Men) Judgments of extremity were introduced analysis of variance, with Item Valence and by explaining that behaviors may differ in Item Object represented as within-partici- the extent to which they are extreme and pant variables. Mean values for each experi- salient: “That is, when you see or hear about mental condition are presented in Table 2. this behavior, that you find this behavior For ratings of constructiveness versus de- very noticeable and extreme” (0 = “not at structiveness, we predicted that “own re- all extreme or salient”; 4 = “somewhat ex- sponses” would be rated as more construc- treme or salient”; and 8 = “very extreme or tive and less destructive than “others’ salient”). We wanted to soften the poten- responses” (Hypothesis lc). Indeed, the tially negative connotation of extremity by analysis revealed a main effect for Item Ob- adding the word “salient”; we also assumed ject, F(1,79) = 61.36,~< .001, with own re- that the combination of extremity and sali- sponses (M = 4.36, SD = .41) being judged ence captures the meaning of “unusual” or to be more constructive than other re- “deviant” (i.e., Ross et al., 1977). Words sponses (M = 4.02,SD = .47).However,this such as “appropriate” or “inappropriate” main effect for Item Object was qualified by were avoided so as to discourage partici- an interaction of Item Valence and Item Ob- pants from coloring such judgments in a ject, F(1,79) = 68.56, p < .001. Follow-up self-serving manner (i.e., individuals might tests of simple effects revealed that own recall their self-other prevalence ratings, constructive responses were judged to be and extremity judgments might be colored more constructive than were others’ con- by self-presentation concerns). structive responses (i.e., positive supe- riority), F(1,80) = 133.56,~< .001, whereas one’s own destructive responses were not Results and discussion judged to be significantly less destructive Our analyses proceeded in two stages. First, than others’ destructive responses, F( 131) we examined the nature of the constructive < 1. The analysis also revealed a significant and destructive responses generated by main effect of Item Valence, F(1,79) = Study 1 participants as typical of self and 1348.01, p < .001, which reflects the trivial others (i.e., external judgments). Second, we finding that constructive responses (M = examined the extent to which self-other 6.07, SD = .58) were judged to be more con- prevalence judgments by Study 2 partici- structive-or less destructive-than were pants can be predicted by their judgments destructive responses (M = 2.31, SD = .63). of constructiveness, frequency, and extrem- Finally,there was a significant 3-factor inter- ity (i.e., self-relevant judgments). action of Item Valence, Item Object, and Gender, F( 1,79) = 6.67, p < .02. Follow-up The nature of constructive and destructive tests revealed that, compared to men, items for self and others (i.e., external judg- women tended to exhibit more positive ments). Participants rated 40 responses in judgments of their own constructive, others’ terms of self-other prevalence, construc- constructive, and own destructive responses; tiveness, frequency, and extremity. Each of women and men did not differ in their judg- ments of their own destructive responses. 3. The frequency judgments emphasize how often For ratings of frequency and extremity,we such responses occur in others’ relationships. As anticipated that items describing others’ re- noted by an anonymous reviewer, such frequency judgments can be influenced by both (a) the extent sponses would be judged to be less frequent to which dissatisfying events occur in others’ rela- and more extreme than items describing tionships, and (b) the degree to which each event own responses (Hypothesis 2); we also an- elicits this response. ticipated that these tendencies would be Uniqueness and similarity 279

Table 2. Number of items, judgments of items, and beliefs regarding self-other prevalence as a function of item valence and item object

Constructive Destructive Measure Own Other Own Other Constructivenessof Responses 6.44 5.70 2.28 2.34 (0.65) (0.65) (0.69) (0.70) Frequency of Responses 4.74 3.76 2.87 2.58 (0.89) (0.86) (1.04) (0.90) Extremity of Responses 2.36 2.82 4.57 5.00 (1.50) (1.29) (1.06) (1.05) Self-other Prevalence 5.19 4.80 3.37 2.99 (0.76) (0.71) (0.95) (0.91)

Note: Higher numbers indicate greater number of responses, greater constructiveness of responses, greater fre- quency of responses, greater extremity of responses, and greater self-other discrepancy in the frequency with which responses are enacted. Standard deviations are in parentheses. Degrees of freedom differ somewhat ow- ing to missing data for some dependent measures.

more pronounced for items describing de- pothesis 4). The main effect for Item Va- structive responses than for items describing lence indeed was significant for judgments constructive responses (Hypothesis 3). Con- of frequency, F(1,79) = 184.09,~< .001, as sistent with Hypothesis 2, the analyses re- well as for judgments of extremity, F(1,81) vealed a main effect of Item Object for the = 250.89, p < .001. As predicted by Hy- frequency measure, F(1,79) = 176.39, p < pothesis 4, constructive responses (M = .001, with own responses (M = 3.80, SD = 4.25, SD = .79) were judged to be more 72) being judged to be more frequent than frequent than were destructive responses other responses (M = 3.17, SD = .75). Simi- (M = 2.72, SD = .93), and constructive re- larly, the analyses revealed a main effect of sponses (M = 2.59, SD = 1.36) were judged Item Object for the extremity measure, to be less extreme than were destructive F(1,81) = 81.50, p < .001, with other re- responses (M = 4.78, SD = 1.00). sponses (M = 3.91,SD = 1.00) being judged For beliefs regarding self-other preva- to be more extreme than own responses (M lence, the analysis revealed a main effect of = 3.46, SD = 1.06). The results were not Item Valence, F(1,79) = 165.05, p < .001. supportive of Hypothesis 3, which would be As predicted in Hypothesis Id, in compari- revealed by a significant interaction of Item son to others, participants believed that Valence and Item Object. This 2-factor in- they exhibited a greater prevalence of con- teraction was significant for frequency judg- structive responses (M = 4.99, SD = .66) ments, F(1,79) = 40.17, p < .001, but in a than destructive responses (M = 3.18, SD = manner inconsistent with Hypothesis 3: 36). There was also a significant main effect There was a greater discrepancy in fre- of Item Object, F(1,79) = 55.64,~< .001. quency judgments for own versus other con- Individuals reported greater self-other structive responses (discrepancy M = 0.98) prevalence for “own responses” (M = 4.28, than for own versus other destructive re- SD = .46) than for “others’ responses” (M = 3.89, SD = S1). No other main effects or sponses (discrepancy M = 0.29). The inter- action of Item Valence and Item Object was interactions were statistically significant. not significant for extremity judgments. Finally, we predicted that items describ- Predicting beliefs regarding self-other ing constructive responses would be judged prevalence (i.e., self-relevant judgments). to be less extreme and more frequent than In addition to a main effect of Item Valence, items describing destructive responses (Hy- the above analyses revealed a robust main 280 RA.M. Van Lange et al.

effect of Item Object for self-other preva- dition, collapsing across the 10 responses lence, demonstrating that participants within each condition. Then we computed judged “own responses” to be more charac- the correlations among self-other preva- teristic of themselves than “others’ re- lence, judged constructiveness, judged fre- sponses.” Own responses were also judged quency, and judged extremity, for each of to be more constructive, more frequent, and the four categories of behaviors. As can be less extreme than were others’ responses. seen in Table 3, reported self-other preva- Thus, the main effect of Item Object for lence was significantly positively correlated self-other prevalence may be accounted for with judged constructiveness within all four by differences between own and others’ re- experimental conditions. More interest- sponses in constructiveness, frequency, ingly, the correlations of self-other preva- and/or extremity. We hypothesized that lence and judged frequency were marginal judged constructiveness, frequency, and ex- or nonsignificant for constructive re- tremity would independently contribute to sponses, whereas these correlations were predicting beliefs regarding self-other statistically significant for destructive re- prevalence (predictions reminiscent of Hy- sponses. Thus, individuals appear to believe potheses 1 and 2), and that judged fre- that destructive responses are more charac- quency and extremity would contribute teristic of the self to the extent that such more powerfully to predicting self-other behaviors occur more frequently. prevalence for destructive responses than Similarly, self-other prevalence was not for constructive responses (a prediction significantly correlated with judged extrem- reminiscent of Hypothesis 3). To test these ity for constructive responses, whereas hypotheses, we performed a series of corre- these correlations were statistically signifi- lational and regression analyses. cant for destructive responses. That is, de- First, for each participant we calculated structive responses were regarded as more mean values of each measure for each con- characteristic of the self to the extent that

Table 3. Correlations among self-other prevalence and judged constructiveness, frequency, and extremity, as a function of item valence and item object

Experimental Condition Constructiveness Frequency Extremity

Own Constructive Items (N= 82) Self-other Prevalence .427** .162+ - .024 Constructivenessof Response .242* - ,140 Frequency of Response -.091

Other Constructive Items (N = SO) Self-other Prevalence .451** - .031 -.134 Constructivenessof Response .053 -.159+

Frequency of Response ,161+

Own Destructive Items (N = 79) Self-other Prevalence .425** .302** -.266** Corptructiveness of Response .015 -.363** Frequency of Response .077

Other Destructive Items (N = 82) Self-other Prevalence .437** .310** - .260** Constructivenessof Response .003 -.428** Frequency of Response ,059

Note: Statistics are based on 79 to 82 individuals. Sample sizes differ slightly across analyses owing to missing data for some measures. *p < .05. **p < .01. +p < .lo, marginal. Uniqueness and similarity 281

Table 4. Multiple regression analyses predicting beliefs regarding self-other prevalence from perceived constructiveness, frequency, and extremity

Regression Coefficients Overall Regression Model

t P< R2 F df P< Own Constructive Items (N = 82) Self-other Judgments From: Constructiveness 3.93 .001 .19 5.99 3,78 .001 Frequency 0.62 .540 Extremity 0.39 ,702

Other Constructive Items (N = SO) Self-other Judgments From: Constructiveness 4.29 .001 .21 6.70 3,76 ,001 Frequency -0.43 .667 Extremity -0.53 .597

Own Destructive Items (N = 79) Self-other Judgments From: Constructiveness 3.47 ,001 .29 10.20 3,75 ,001 Frequency 3.16 .002 Extremity -1.51 .136

Other Destructive Items (N = 82) Self-other Judgments From: Constructiveness 3.69 .001 .30 10.94 3,78 .001 Frequency 3.31 ,001 Extremity -1.07 .288

Note: Statistics are based on 79 to 82 individuals. Sample sizes differ slightly across analyses owing to missing data for some measures. such behaviors are less extreme. Finally, Ta- cantly to predicting self-other judgments of ble 3 reveals that judgments of construc- destructive responses, but did not make a tiveness, frequency, and extremity were not significant contribution to predicting systematically correlated, with the excep- self-other judgments of constructive re- tion of a correlation between judged con- sponse~.~However, in none of the catego- structiveness and extremity in ratings of de- ries of behaviors did perceived extremity structive responses. Lower levels of judged contribute to predicting self-other preva- constructiveness were associated with greater judged extremity. 4. It is interesting to note that perceived constructive- Next, we conducted four simultaneous ness, frequency, and extremity accounted for about regression analyses in which self-other 20% of the variance in self-other prevalence rat- judgments were regressed onto perceived ings for constructive responses, whereas these per- constructiveness, frequency, and extremity, ceptions accounted for about 30% of the variance in self-other prevalence ratings for destructive re- one regression being conducted for each sponses (see Table 4). Thus, the two-predictor class of responses. As can be seen in Table model for self-other prevalence of destructive re- 4, consistent with Hypothesis 1, perceived sponses (i.e., perceived constructiveness and fre- constructiveness is a strong and robust pre- quency) accounts for substantially more variance than does the single-predictor model for self-other dictor Of judgments Of the beha'- prevalence of constructive responses, a finding ac- iors in each category. More counted for by the interaction of perceived fre- perceived frequency contributed signifi- quency and Item Valence. 282 PA.M. Van Lange et al. lence judgments. The absence of such ef- Study3 fects for destructive responses, at least in The goal of Study 3 is to explore individu- part, may be due to multicollinearity be- als’ recall of constructive and destructive tween perceived constructiveness and ex- responses ascribed to self or others by tremity, in that extremity exhibited substan- Study 1 participants. Such recall is impor- tial (negative) correlations with both tant because it might illuminate the proc- self-other prevalence judgments and per- esses by which individuals develop and sus- ceived constructiveness of destructive re- tain the general belief that their own sponses. These correlations suggest that ex- relationships are better than others’ rela- tremity judgments tended to have a tionships but otherwise perfectly normal. negative connotation.5 Conversely, it is plausible that such (pre- Thus, consistent with Hypotheses 1 and sumably stable) beliefs about our own and 2, self-other prevalence of destructive re- others’ relationships affect the ways in sponses can best be understood by taking which we attend to and ultimately recall into account the extent to which individu- new information about own and others’ re- als regard a given response as constructive lationships. Accordingly, we generally ex- and frequently occurring. However, pected selective recall of information re- self-other prevalence of constructive re- garding constructive and destructive sponses can be parsimoniously predicted responses that Study 1 participants listed as by level of perceived constructiveness, sup- characteristic of their own versus others’ porting Hypothesis 1 but disconfirming relationships. Hypothesis 2. The overall pattern does pro- Because negative behaviors typically vide good evidence in support of the claim tend to receive greater weight and atten- that perceived frequency makes a stronger tion than do positive behaviors, one might contribution to predicting self-other preva- assume that these responses will also lence regarding destructive responses than be better recalled. Indeed, some evidence for constructive responses (Hypothesis 3). supports this reasoning-in particular, This finding is consistent with the more there is evidence for a tendency to recall general proposition that individuals as- information that is incongruent with one’s sume uniqueness for constructive re- general beliefs or expectations (e.g., Hastie sponses, and uniqueness and similarity for & Kumar, 1979; Srull, 1981). However, evi- destructive responses. dence also indicates greater recall of posi- tive rather than negative information. Al- though several explanations of such 5. Finally, we performed simultaneous regression analyses for each item in the list, regressing findings have been advanced, it seems plau- self-other prevalence onto perceived constructive- sible that positive information may be bet- ness, frequency, and extremity. The results were in ter recalled because positive information agreement with the above findings. In regressions presumably is more self-relevant or be- predicting self-other prevalence of constructive re- sponses, perceived constructiveness exhibited sig- cause there are greater possibilities for in- nificant regression coefficients in 15 of 20 analyses terpreting such behavior in light of other (plus 3 marginal effects), perceived frequency ex- positive information (e.g., positive informa- hibited significant coefficients in 3 analyses (plus 2 tion is consistent with a favorable schema marginal effects), and perceived extremity exhib- about the self; cf. Markus, 1977; Taylor, ited no significant coefficients (only 1 effect was marginal). However, in regressions predicting 1991). self-other prevalence of destructive responses, per- A recent meta-analysis by Stangor and ceived constructiveness exhibited significant coeffi- McMillan (1992) suggests the importance cients in 9 of 20 analyses (plus 2 marginal effects), of several moderating variables, including perceived frequency exhibited significant coeffi- the strength of expectancy and the delay cients in 12 analyses (plus 1 marginal effect), and perceived extremity exhibited significant coeffi- between exposure to stimulus and recall cients in 3 analyses (plus 1marginal effect). task. For example, memory €or expectancy- Uniqueness and similarity 283

incongruent information tends to be many responses as they could remember greater when the expectancy is weak and from the rating task. As in prior research by when the time between exposure and recall Liebrand, Messick, and Wolters (1986) the is small. The current research focuses on instructions read: “During this research you responses that may be quite common over- have rated 40 responses to incidents of dis- all. The responses are generated by compa- satisfaction in relationships. We now ask rable participants (i.e., Study 1 partici- you to write down as many responses as pants), and participants presumably have you can remember from the list.” Partici- acquired a fair amount of experience with pants were given 5 minutes to complete this constructive and destructive behavior in re- task. As in several other studies (e.g., Lie- lationships. Moreover, the delay between brand et al., 1986) the lists of recalled be- exposure and recall task was small. In light haviors for each participant were catego- of these features of the current research, we rized by one of three individuals who were assume greater recall for information that is blind to the experimental hypotheses and incongruent with expectations and thus goals (i.e., each individual coded roughly propose greater recall for uncommon, in- one-third of the total number of recalled frequent responses than for common, fre- responses). The coding was not difficult, be- quent responses. cause the wording of the recalled behaviors Accordingly, we advanced three hy- was generally very similar to the original potheses that parallel earlier predictions wording. This may be so because partici- (i.e., Hypotheses 2 through 4). Specifically, pants had rated the items several times assuming that others’ responses are re- (e.g., judging constructiveness, frequency, garded as less common than one’s own re- extremity, and self-other prevalence), and sponses, we predict greater recall for other because the delay between exposure and responses than for own responses (i.e., recall task was small. Accordingly, we did paralleling Hypothesis 2, we predict a main not obtain interrater reliability information effect of Item Object). Also, assuming that for these ratings (in total, there were more differences in the commonness of own and than 1,100 recalled responses). other responses will be more pronounced for destructive responses rather than for Results and discussion constructive responses, we predict that dif- ferences in recall will be relatively more Out of 40 behaviors, participants recalled pronounced for own destructive responses an average of 14.07 behaviors. Recall of than for others’ destructive responses (i.e., items was analyzed in a 2 (Item Valence: paralleling Hypothesis 3, we predict an in- Constructive vs. Destructive responses) by teraction of Item Valence and Item Object). 2 (Item Object: Self vs. Other behaviors) by Finally, assuming that destructive responses 2 (Gender: Women vs. Men) ANOVA, with are less common than constructive re- the former two variables represented as sponses, we predict greater recall for de- within-subject factors. Consistent with Hy- structive rather than for constructive re- pothesis 2, this analysis revealed a main ef- sponses (i.e., paralleling Hypothesis 4, we fect of Item Object, F(1,81) = 62.2, p < predict a main effect of Item Valence). .001, indicating greater recall for items de- scribing others’ responses (M = 8.01, SD = 2.78) than for those describing own re- Method sponses (M = 6.06, SD = 2.39). In addition, Participants is Study 3 were the same as a main effect of Item Valence, F(1,81) = those in Study 2 (i.e., for the sake of clarity, 10.67, p < .005, revealed that individuals we present these as separate studies). After recalled more items describing constructive Study 2 participants made judgments of the responses (self plus other: M = 7.55, SD = responses, we administered a free-recall 2.92) than items describing destructive re- task in which individuals wrote down as sponses (self plus other: M = 6.52, SD = 284 €?A.M.Van Lange et al.

2.52): This finding is inconsistent with Hy- constructive responses were relatively pothesis 4, which predicted greater recall small. This latter finding is congruent with for destructive items than for constructive the broader claim that others’ responses items (we will consider this matter in the (destructive responses in particular) are un- Discussion section). common and infrequent (and therefore Moreover, there was a significant inter- memorable), whereas one’s own destruc- action of Item Valence and Item Object, tive responses are more common and fre- F(1,81) = 8.55,~< .005, revealing that par- quent (and therefore less memorable). ticipants recalled fewer destructive own re- sponses (M = 2.57, SD = 1.42) than de- General Discussion structive other responses (M = 3.95, SD = 1.60; a mean difference of 1.38), whereas The current research provides good evi- the discrepancy between recall of construc- dence in support of the claim that percep- tive own (M = 3.49, SD = 1.63) and con- tions of uniqueness and similarity underlie structive other responses (M = 4.06, SD = individuals’ beliefs regarding their re- 1.78) was less pronounced (a mean differ- sponses to incidents of dissatisfaction. Rele- ence of 0.57). This interaction is consistent vant to hypotheses involving the unique- with Hypothesis 3, which predicted that dif- ness bias, we found support for four specific ferences in recall for own and other re- predictions. Consistent with Hypotheses 1a sponses would be more prondunced for de- and lb, participants listed a greater number structive responses than for constructive of constructive responses and fewer de- responses. structive responses in describing them- Finally, an interaction of Gender and selves than in describing others. Consistent Item Valence, F(1,81) = 4.37, p < .05, re- with Hypothesis lc, although destructive vealed that the tendency to recall more items describing oneself were not rated as constructive responses than destructive re- more destructive than destructive items de- sponses was greater among women (Ms = scribing others, constructive items describ- 7.80 vs. 6.16, SDs = 3.10 and 2.75; a mean ing oneself were rated as more constructive difference of 1.64) than among men (Ms = than were constructive items describing 7.28 vs. 6.92, SDs = 2.71 and 2.19; a mean others. And in support of Hypothesis Id, difference of 0.36). One may speculate that constructive items were rated as more fre- women more than men recall constructive quently performed by oneself than by oth- responses because women are somewhat ers, whereas the opposite was observed for more likely than men to engage in construc- destructive items. tive problem-solving in ongoing relation- Consistent with the similarity bias (i.e., ships (cf. Rusbult, Johnson, & Morrow, Hypothesis 2), items describing responses 1986; Sprecher, 1992). Accordingly, con- of oneself were rated as more frequent and structive responses may be more self-de- less extreme than were items describing scriptive for women than for men. others’ responses. Also, responses describ- To summarize, recall was greater for (a) ing others were better recalled than were constructive responses than for destructive items describing one’s own responses. And responses, and (b) for items listed by Study congruent with Hypothesis 3, ratings of 1 participants as characteristic of others in constructiveness were the only solid predic- their relationships (i-e., other responses), tor of beliefs regarding self-other preva- rather than items listed by Study 1 partici- lence of constructive responses, whereas pants as characteristic of the self in their ratings of constructiveness and ratings of own relationships (i.e., own responses). frequency were significant predictors of be- Most importantly, others’ destructive re- liefs regarding self-other prevalence of de- sponses were recalled much better than structive responses. Finally, the current one’s own destructive responses, whereas work provides partial evidence in support differences between one’s own and others’ of Hypothesis 4. As predicted, constructive Uniqueness and similarity 285

responses were rated as more frequent and propriateness of one’s responses may be less extreme than were destructive re- softened or denied. Given that such infor- sponses. However, inconsistent with Hy- mation concerns the self, the individual’s pothesis 4,constructive responses were bet- ability to filter in such a manner is facili- ter recalled than were destructive tated by the possession of knowledge re- responses. Below, we briefly consider spe- garding internal processes (e.g., underlying cific findings in light of underlying mecha- intentions, effort expenditure) and prior be- nisms and alternative explanations, and we havior (Jones & Nisbett, 1971). That is, review the theoretical and practical impli- whereas one’s negative or extreme re- cations of the results. sponses may be viewed in light of benign How do we account for the simultaneous intentions or may be linked to positive re- occurrence of the uniqueness and similarity sponses in the past, such tendencies are less biases? To begin with, we propose that at feasible in processing information regard- least in part, these biases can be understood ing others’ responses. by reference to the manner in which indi- Are there alternative mechanisms that viduals process information about their may help to explain the uniqueness and own and others’ relationships. In processing similarity biases observed in the current re- information regarding others’ relationships, search? The availability of destructive re- individuals may focus on specific informa- sponses enacted by others could be guided tion that is largely negative rather than by preexisting stereotypes that come to positive (Le., the negativity effect) and may mind when thinking about destructive re- focus on information that is infrequent or sponses (i.e., beliefs regarding a specific extreme rather than common (i.e., the ex- group of real or imaginary others who be- tremity effect; e.g., Skowronski & Carlston, have quite destructively during incidents of 1989). Such information processing may re- dissatisfaction; cf. Weinstein, 1980). Such sult in a greater availability of destructive stereotypes may be further stimulated by and extreme responses for other individu- information conveyed via the mass media als. The finding that a relatively large (e.g., TV talk shows, soap operas), which number of destructive responses were gen- tend to focus on individuals who engage in erated and recalled for others is consistent rather negative and otherwise uncommon with this reasoning. Moreover, findings re- behaviors. In addition to stereotypes, it may garding self-other prevalence suggest that also be that individuals intentionally focus there is a fair amount of agreement about on others who enact destructive and ex- the types of responses that characterize treme responses (cf. downward comparison; others, but not oneself. Wills, 1991) or focus on specific domains of Although the negativity and extremity responses that allow them to maintain a effects presumably are pervasive, we pro- relatively favorable image of their own re- pose that in the processing of information sponses (cf. dimensional comparison, regarding one’s own relationship, such ten- Wood, 1989). dencies may be overshadowed by cognitive Moreover, Study 2 revealed some evi- filters that serve to sustain or improve a dence in support of the claim that individu- relatively favorable image of one’s relation- als assume greater similarity for destructive ship. Most individuals presumably have de- than for constructive responses (Hypothe- veloped fairly positive self-schemata, rela- sis 3). In particular, correlational and re- tively stable and favorable beliefs about the gression analyses revealed that perceived self, which guide the processing of incoming constructiveness alone suffices in predict- information (cf. Taylor & Brown, 1988). ing beliefs regarding self-other prevalence Through such filters, information indicative of constructive responses, whereas both of the positivity or appropriateness of one’s perceived constructiveness and frequency responses may be welcomed, whereas infor- make unique and substantial contributions mation suggesting the negativity or inap- toward predicting beliefs regarding 286 PA.M. Van Lange et al.

self-other prevalence of destructive re- own destructive responses and viewing sponses. This finding suggests that individu- them as fairly appropriate) or could be less als justify or discount their own destructive feasible in interpreting others’ responses behaviors by believing that such responses (e.g., information regarding the context in are common and occur frequently in oth- which others’ destructive and constructive ers’ relationships as well, thereby underlin- responses take place may be limited). ing the appropriateness and reasonableness Finally, Study 3 revealed greater recall of such undesirable behaviors (cf. Camp- for others’ responses than for one’s own re- bell, 1986; Goethals, 1986; Sherman et al., sponses, along with greater recall for one’s 1984). In contrast, the destructive responses own constructive responses than for one’s typical of others are perceived as infre- own destructive responses. In accounting quent (and extreme); such views may be for these findings, it is useful to assume that strengthened by some of the specific several specific mechanisms may underlie mechanisms outlined above (e.g., the ex- recall abilities (cf. Stangor & McMillan, tremity effect, stereotype salience, down- 1992). First, greater recall for others’ re- ward comparison, and dimensional com- sponses than for one’s own responses may parison). be attributed to the fact that others’ re- It is conceivable that, in addition to con- sponses were rated as relatively less com- sensual validation, alternative mechanisms mon and more extreme (cf. Skowronski & assumed to underlie the similarity bias can Carlston, 1989). However, recall also may account for some of the current find- have been influenced by individuals’ glob- ings-the general occurrence of the similar- ally benign views of relationship function- ity bias could follow from selective expo- ing, such that individuals remember those sure, cognitive availability, focus of responses that are congruent with global ex- attention, and situational attribution (cf. pectations and beliefs (cf. Markus, 1977; Marks & Miller, 1987). However, the tradi- Stangor & McMillan, 1992). In particular, tional formulations of such processes seem this mechanism may account for the finding less capable of explaining (a) why the simi- that individuals recalled a moderate larity and extremity biases coexist, and (b) number of their own constructive responses. why the similarity bias is relatively more Do those two mechanisms account for the pronounced for destructive responses. Nev- finding that others’ constructive behaviors ertheless, these processes may be valid in were recalled at a level similar to that ob- the context of constructive and destructive served for one’s own constructive behaviors responses in relationships, if one no longer or for others’ destructive behaviors? assumes that these mechanisms necessarily Perhaps recall is also influenced by a so- lead to assumed similarity. Specifically,indi- cial comparison activity that has been viduals may be selectively exposed to infor- termed self-improvement, in reference to mation about responses of others that are the tendency to compare oneself to others negative and otherwise extreme (e.g., selec- who teach one how to perform better or tive exposure through mass media influ- who motivate one to improve (cf. Wood, ences); instances of destructive responses 1989). Participants may have assigned enacted by others may come to mind rela- greater attention to others’ constructive re- tively easily (i.e., cognitive availability does sponses because doing so would teach them not necessarily imply that instances of simi- how to deal with future incidents of dissat- lar responses are most salient); a focus on isfaction. The poor recall of one’s own de- self could also serve as an anchor with structive responses may be explained by which information about others is to be the fact that-relative to others’ destructive contrasted (i.e., noticing self-enhancing dif- responses-these responses were viewed as ferences in addition to similarities); and, fi- common and somewhat self-descriptive. nally, situational attributions could be made Accordingly, these responses may be easily in a self-enhancing manner (e.g., justifying forgotten because they do not stand out as Uniqueness and similarity 287

exceptional or uncommon and because (e.g., other partners do exactly the same), these somewhat self-descriptive responses whereas their partners are less likely to sof- are inconsistent with positive self-sche- ten or discount such responses. A conceiv- mata. Finally, inspection of Table 1suggests able result of such discrepant interpreta- that there may be somewhat more overlap tions is a metaconflict, or conflict regarding among the destructive responses (e.g., ag- the conflict process itself, which may fur- gressive acts, ignoring the problem) than ther contribute to the dissatisfaction. In- among the constructive responses, which deed, metaconflicts have been argued to be may have reduced individuals’ ability to re- an important source of contention and call destructive responses. stress in close relationships (cf. Braiker & Two final issues deserve some brief dis- Kelley, 1979). cussion. First, consistent with predictions, We wish to outline briefly some destructive responses were rated as less fre- strengths and limitations of the current re- quent and more extreme than were con- search. First, one limitation is that the cur- structive responses. This finding provides rent work provides little insight into pre- evidence in support of the basic assump- cisely how and why individuals assume tions underlying negativity and extremity both uniqueness and similarity in their re- effects (e.g., Reeder & Brewer, 1979; Skow- sponses to dissatisfaction. Clearly, further ronski & Carlston, 1989), and demonstrates research is needed to examine directly the their robustness in the context of relation- roles of information-processing styles (e.g., ships. Second, throughout this article we in regard to the negativity and extremity employed the term “bias” to describe per- effects as well as several social comparison ceptions of uniqueness and similarity.How- activities). Also, we cannot rule out alterna- ever, most of the obtained findings can in tive explanations that emphasize the roles fact be explained by assuming that indi- of person positivity (e.g., the self-as-a-per- viduals form and maintain fairly accurate son is compared to “a group” of same-sex beliefs (cf. Colvin & Block, 1994; Dawes, others; cf. Sears, 1983) and social desirabil- 1989). For example, it is not necessarily er- ity. However, social desirability is unlikely roneous for most individuals to believe that to play an exceptionally important role, in they “hit the partner” less frequently than that recent research has revealed that ten- did others, assuming that some small minor- dencies toward perceived superiority are ity of people hit their partners; assuming not correlated with instruments assessing this to be the case, most people may accu- tendencies toward socially desirable re- rately claim that they engage in such de- sponding (Rusbult, Van Lange, Verette, & structive responses less frequently than do Yovetich, 1996). others. An interesting implication following Second, the direction of causality be- from this line of reasoning is that individu- tween beliefs regarding self-other preva- als may maintain relatively benign views of lence and perceived constructiveness and their relationships while simultaneously be- frequency is not clear. For example, do we ing fairly accurate. At least in part, incom- believe that we perform destructive re- ing information concerning relatively un- sponses less frequently than do others be- common and extreme forms of destructive cause such behaviors are destructive and behaviors may help individuals to maintain uncommon, or do we believe that such re- a feeling of superiority, whereas informa- sponses are destructive and uncommon be- tion concerning more ordinary or common cause we do not perform such behaviors as forms of destructive behaviors may provide frequently as others? Third, we have as- consensual validation for their own de- sumed that responses to dissatisfaction in- structive responses. A potential threat to volve both universally and variably evalu- the well-being of relationships is that indi- ated dimensions. While this assumption viduals may soften or to some degree “dis- seems plausible, this claim remained un- count” their own destructive responses tested in the current research. 288 PA.M. Van Lange et al.

Finally, our method of measuring simi- viduals’ understanding of their behavior larity bias is somewhat unusual, and the ab- than are items generated by the experi- sence of more commonly used measures menter. Moreover, by using Study 2 partici- does not facilitate careful comparison with pants as external judges, we obtained a prior work on similarity. Moreover, the fre- fairly unobtrusive measure of uniqueness quency judgments may be guided by (a) the and similarity; the recall measure seems frequency with which dissatisfying events similarly unobtrusive. Also, it is worth not- occur, and (b) the extent to which partici- ing that these findings were observed pants believe that others responded con- among Dutch participants, thereby comple- structively or destructively to a given dissat- menting prior research on social-cognitive isfying event (see note 3).At the same time, processes in relationships, which by and we believe that research on the similarity large has been conducted in the United bias may profit from less commonly used States. operationalizations that focus on individu- als’ perceptions of commonness, deviance, Conclusion and appropriateness-constructs delineat- ing judgments that are central in the Ross The current work illustrates that two seem- et al. (1977) definition of “false consensus.” ingly contradictory beliefs-uniqueness One important strength of the current and similarity biases-coexist, enabling in- research concerns the multifaceted meth- dividuals to sustain the general, self-en- odology that included the generation of hancing belief that one is “uniquely better items (Study l), judgment of items (Study than others but otherwise perfectly nor- 2), and recall of items (Study 3) describing mal.” The similarity bias (“being perfectly constructive and destructive responses be- normal”) seems pertinent in understanding lieved to be performed by oneself or others. how individuals may regard and justify be- One advantage of this approach is that such haviors that are potentially harmful to the items are more likely to be central to indi- well-being of a relationship.

References Acitelli, L. K., Douvan, E., & Veroff, J. (1993). Percep- Campbell, J. D. (1986). Similarity and uniqueness: The tions of conflict in the first year of marriage: How effects of attribute type, relevance, and individual important are similarity and understanding? Jour- differences in self-esteem and depression. Journal nal of Social and Personal Relationships, 10, 5-19. of Personality and Social Psychology, 50, 281-294. Alicke, M. E. (1985). Global self-evaluation as deter- Clark, M. S., & Reis, H. T. (1988). Interpersonal proc- mined by the desirability and controllability of trait esses in close relationships. Annual Review of Psy- adjectives. .Journal of Personality and Social Psy- chology, 39, 609472. chology, 49, 1621-1630. Colvin, C. R., & Block, J. (1994). Do positive illusions Alicke, M. E., & Largo, E. (1995). The role of the self foster mental health? An examination of the Taylor in the false consensus effect. Journal of Experimen- and Brown formulation. Psychological Bulletin, tal Social Psychology, 31, 28-47. 116,3-20. Berscheid, E. (1994). Interpersonal relationships. An- Cronbach, L. J. (1955). Processes affecting scores on nual Review of Psychology, 45, 79-129. “understanding of others” and “assumed similar- Braiker, H. B., & Kelley, H. H. (1979). Conflict in the ity.” Psychological Bulletin, 52, 177-193. development of relationships. In R. L. Burgess & Dawes, R. M. (1989). Statistical criteria for establishing T. L. Huston (Eds.), Social exchange in developing a truly false consensus effect. Journal of Experi- relationships (pp. 135-168). New York: Academic mental Social Psychology, 25, 1-17. Press. Goethals, G. R. (1986). Fabricating and ignoring social Brown, J. D. (1986). Evaluations of self and others: reality: Self-serving estimates of consensus. In J. M. Self-enhancement biases in social judgments. So- Olson, C. P. Herman, & M. P. Zanna (Eds.), Social cial Cognition, 4, 353-376. comparison and relative deprivation: The Ontario Buunk, B. P., & Van Yperen, N. W. (1991). Referential symposium (Vol. 4, pp. 135-157). Hillsdale, NJ: comparisons, relational comparisons, and exchange Erlbaum. orientation: Their relation to marital satisfaction. Goethals, G. R., Allison, S. J., & Frost, M. (1979). Per- Personality and Social Psychology Bulletin, 17, ceptions of the magnitude and diversity of social 709-717. support. Journal of Experimental Social Psychol- Cameron, N., & Margaret, A. (1951). Behavior pathol- Oxy, 15, 570-581. ogy. Boston: Houghton Mifflin. Goethals, G. R., Messick. D. M., & Allison, S. T. (1991). Uniqueness and similarity 289

The uniqueness bias: Studies of constructive social Sears, D. 0. (1983). The person-positivity bias. Journal comparison. In J. Suls & T. A. Wills (Eds.), Social of Personality and Social Psychology, 44, 233-240. comparison: Contemporary theory and research Sillars, A. L., & Scott, M. D. (1983). Interpersonal per- (pp. 149-176). Hillsdale NJ: Erlbaum. ception between intimates: An integrated review. Hastie, R., & Kumar, P. (1979). Person memory: Per- Human Communication Research, 10, 153-176. sonality traits as organizing principles in memory Sherman, S. J., Chassin, L., Presson, C. C., & for behaviors. Journal of Personality and Social Agostinelli (1984). The role of the evaluation and Psychology, 37, 25-38. similarity principles in the false consensus effect. Holmes, J. G., & Boon, S. D. (19YO). Developments in Joiirniil of Personality and Social Psychology, 47, the field of close relationships: Creating founda- 1244-1262. tions for intervention strategies. Personality and Skowronski. J. J., & Carlston, D. E. (1989). Negativity Social Psychology Bulletin, 16, 23-41. and extremity biases in impression formation. Psy- Jones, E. E.. & Nisbett, R. E. (1971).The actor and the chological Bulletin, 105, 131-142. observer: Divergent perceptions of the causes of Sprecher. S. (1992). Social exchange perspectives to behavior. In E. E. Jones, D. E. Kanouse, H. H. the dissolution of close relationships. In T. L. Or- Kelley, R. E. Nisbett, S. Valins, & B. Weiner (Eds.), buch (Ed.), Close relationship loss: Theoretical ap- Attribution: Perceiving the causes of behavior (pp. proaches (pp. 47-66). New York: Springer. 79-94). Morristown, NJ: General Learning Press. Srull, T. K. (1981). Person memory: Some tests of asso- Liebrand, W. B. G.. Messick, D. M.. & Wolters, F. J. M. ciative storage and retrieval models. Journul of Ex- (1986). Why we are fairer than others: A cross-cul- perimental Psychology: Human Learning and tural replication and extension. Journal of Experi- Memory, 7, 440-463. mental Sociul Psychology, 22, 590-604. Stangor, C., & McMillan, D. (1992). Memory for expec- Marks, G. (1984). Thinking one’s abilities are unique tancy-congruent and expectancy incongruent in- and one’s opinions are common. Personality and formation: A review of the social and social devel- Sociul Psychology Bulletin, 10, 203-208. opmental literatures. Psychological Bulletin, 111, Marks, G., & Miller, N. (1987). Ten years of research on 42-61. the false consensus effect. Psychological Bulletin, Taylor, S. E. (1991). Asymmetrical effects of positive 102, 72-90. and negative events: The mobilization-minimiza- Markus, H. (1977). Self-schemata and processing infor- tion hypothesis. Psychological Bulletin, 110, 67-85. mation about the self. Journal of Personality and Taylor, S. E., & Brown. J. D. (1988). Illusion and well- Social Psychology, 35, 63-78. being: A social psychological perspective on men- McFarland, C.. & Miller. D. T. (1990). Judgments of tal health. Psychological Bulletin, 103, 193-210. self-other similarity: Just like other people, only Taylor, S. E., & Brown, J. D. (1994). Positive illusions more so. Personality and Social Psychology Bulle- and well-being revisited: Separating fact from fic- tin, 16, 475-484. tion. Psychological Bulletin, 116, 21-27. Messick, D. M., Bloom, S., Boldizar, J. P., & Samuelson, Van der Pligt, J. (1984). Attributions, false consensus, C. D. (1985). Why we are fairer than others. Jour- and valence: Two field studies. Journal of Personrrl- nal ofExperimenta1 Social Psychology, 21,480-500. ity and Social Psychology, 46, 5748. Peplau, L. A. (1983). Roles and gender. In H. H. Kelley, Van Lange, P. A. M. (1991). Being better but not E. Berxheid, A. Chistensen, J. H. Harvey, T. L. smarter than others: The Muhammad Ali effect at Huston, G. Levinger, E. McClintock, L. A. Peplau. work in interpersonal situations. Personality and & D. R. Peterson (Eds.), Close relationships (pp. Social Psychology Bulletin, 17, 690494. 220-264). New York: W. H. Freeman. Van Lange. €? A.M., & Kuhlman, D. M. (1994). Social Reeder, G. D., & Brewer, M. B. (1979). A schematic value orientations and impressions of partner’s model of dispositional attribution in interpersonal honesty and intelligence: A test of the might versus perception. Psychological Review, 86, 61-79. morality effect. Journal of Personality and Social Ross, L.. Greene, D.. & House, F! (1977). The “false Psychology, 67, 126-141. consensus effect”: An egocentric bias in social per- Van Lange. P. A. M., & Rusbult, C. E. (1995). My ception and attribution processes. Journal of Ex- relationship is better than-and not as bad perimental Social Psychology, 13, 279-301. as-yours is: The perception of superiority in close Ross, M., & Sicoly,F. (1979). Egocentric biases in avail- relationships. Personality and Social Psychology ability and attribution. Journal of Personality and Bulletin, 21, 3244. Social Psychology, 37, 322-337. Van Lange, I? A. M., Agnew, C. R., Harinck, F., & Rusbult, C. E., Johnson, D. J., & Morrow, G. D. (1986). Steemers, G. E. M. (1997). From game theory to Impact of couple patterns on problem solving on real life: How social value orientation affects will- distress and nondistress in dating relationships. ingness to sacrifice in ongoing relationships. .lour- Journal of Personality und Social Psychology, 60, nal of Personality and Social Psychology, 73,1330- 543-578. 1334. Rusbult, C. E., Van Lange, P A. M., Verette, J., & Weinstein, N. D. (1980). Unrealistic optimism about Yovetich, N. A. (1996). A functional analysis of per- future life events. Journal of Personality and Social ceived superiority in close relationships. Unpub- Psychology, 39, 806-820. lished manuscript, University of North Carolina, Wills, T. A. (1991). Similarity and self-esteem in down- Chapel Hill. ward comparison. In J. Suls & T. A. Wills (Eds.), Rusbult, C. E.,Verette, J., Whitney, G. A,, Slovik, L. F., Social comparison: Contemporary theory and re- & Lipkus, I. (1991). Accommodation processes in search (pp. 51-78). Hillsdale NJ: Erlbaum. close relationships: Theory and preliminary empiri- Wood, J. V. (1989). Theory and research concerning cal evidence. Journal of Personality and Social Psy- social comparisons of personal attributes. Psycho- chology, 60,53-78. logical Bulletin, 106, 23 1-248.