Journal of Personality and Social Psychology Copyright 1998 by the American Psychological Association, Inc.

1998, Vol. 74, No. 2, 378-386 0022-3514/98/$3.00

The Self-Serving Bias in Relational Context

Constantine Sedikides and W. Keith Campbell Glenn D. Reeder
University of North Carolina at Chapel Hill Illinois State University

Andrew J. Elliot
University of Rochester

This article examined the impact of relationship closeness on the self-serving bias (SSB). Members
of relationally distant dyads working on interdependent-outcomes tasks manifested the SSB: They
took credit for dyadic success but blamed the partner for dyadic failure. However, members of
relationally close dyads did not manifest the SSB: They did not take more credit than their partner
for dyadic success and did not blame the partner more than the self for dyadic failure. This gracious
attributional pattern of relationally close dyad members is due, at least in part, to formation of a
favorable impression of the partner. Relationship closeness acts as a bound to an individual's self-
enhancing tendencies.

If more than one person is responsible for a miscalculation, none to protect or enhance the self. Individuals may fail to realize
will be at fault—Murphy's Law that the information available to them is incomplete, but this is
the result of an imperfect mind rather than the result of motiva-
Individuals self-enhance. They believe that they are more tion. Another explanation for the SSB advocated motivational
trustworthy, moral, and physically attractive than others and that reasons, such as the desire to think positively or avoid thinking
they are above-average teachers, managers, and leaders. One negatively of the self (Weary-Bradley, 1978). With an increased
mechanism through which individuals maintain such unduly accumulation of empirical evidence has come the understanding
positive beliefs is the self-serving bias (SSB). The SSB refers that cognitive explanations alone cannot account fully for the
to individuals taking responsibility for successful task outcomes SSB. Instead, motivational reasons, and specifically the desire
but blaming circumstances or other persons for failed task out- to enhance the positivity or diminish the negativity of one's
comes. For example, students will take credit for passing a self-concept, have emerged as the predominant (i.e., sufficient)
difficult examination but will attribute failing the examination explanation for the SSB (Zuckerman, 1979).
to its difficulty or the instructor's tough grading policy.
Explanations for the SSB have been the subject matter of
considerable debate. One early review advocated imperfect in- THE SSB IN D^iADIC INTERDEPENDENT-
formation-processing strategies such as selective attention and OUTCOMES TASKS
informational availability or accessibility in memory (Miller &
Ross, 1975). This cognitive explanation emphasizes differential The SSB has been investigated in both independent- and inter-
access to information as the leading cause of the SSB. Individu- dependent-outcomes tasks. In the former settings, participants
als manifest the SSB because they restrict their attention to the work on a task independently. The SSB in these situations is
information available to them, not because they are motivated robust (Campbell & Sedikides, 1998). In interdependent-out-
comes settings, which are of relevance to the present work,
participants collaborate on a task. Success or failure of the
dyad hinges on the joint rather than unique contribution of the
Constantine Sedikides and W. Keith Campbell, Department of Psy-
members. Given the dyad's common fate—interaction—effort
chology, University of North Carolina at Chapel Hill; Glenn D. Reeder,
Department of Psychology, Illinois State University; Andrew J. Elliot, coordination and a good working relationship between partners
Department of Psychology, University of Rochester. are crucial prerequisites for an optimal task outcome. In this
Part of this research as well as preparation of the article was supported article we are particularly concerned with whether and how
by a University of North Carolina junior faculty development award. partners' relationship affects the task outcome and, more spe-
Portions of the research were presented at the 1994 Midwestern Psycho- cifically, how it affects their attributions for the task outcome.
logical Association convention, Chicago; the 1996 Society of Southeast- We are interested in relativistic attributions (i.e., attributing the
ern Social Psychologists meeting, Virginia Beach, VA; and the 105th task outcome to the self or the partner). Furthermore, we focus
Annual Convention of the American Psychological Association, Chicago, on two types of dyadic relationships: distant (i.e., strangers)
August 1997. We thank Carrie Beck, Robert Brafford, Cameron Calo- and close (i.e., friends).
well, and Martin Curtis for their assistance.
Correspondence concerning this article should be addressed to Con-
One objective of the present research was to test whether the
stantine Sedikides, Department of Psychology, University of North Caro- SSB emerges in distant dyads. In the absence of a relational
lina at Chapel Hill, Davie Hall, CB# 3270, Chapel Hill, North Carolina bond, and in light of the robustness of the SSB in independent-
27599-3270. Electronic mail may be sent to [email protected]. outcomes tasks, we would not be surprised if distant dyads

378
 SELF-SERVING BIAS IN RELATIONSHIPS 379

manifested the SSB. That is, participants whose dyad received self-enhancement strategies. By this logic, the SSB ought to
success feedback would be likely to consider the self more be present (perhaps magnified!) in close relationships. Close
responsible than the partner for the task outcome, whereas parti- participants should take more responsibility for the dyadic suc-
cipants whose dyad received failure feedback would be likely cess than its failure.
to hold the partner more responsible than the self for the task Is there empirical evidence in the SSB literature for the rela-
outcome. The assumed motivation for the manifestation of the tionships-as-bound and relationships-as-enabler hypotheses?
SSB is the protection or enhancement of the self. Before delving into the literature, we set out to define the charac-
However, the major objective of the present research was to teristics of a stringent test for the motivationally based emer-
determine whether the SSB emerges in close dyads. Is the same gence of the SSB in dyadic interdependent settings. Such a
motivation (i.e., protection or enhancement of the self) present test ought to satisfy the following four procedural criteria: (a)
when the individual works on a task collaboratively with a close members of the dyad should be separated (i.e., placed in adja-
other? cent rooms) by the experimenter, so that they do not interact
Several theoretical perspectives offer a sound rationale for the during task completion; (b) the dyad should complete a rela-
prediction that the SSB will not be present in close relationships. tively unfamiliar, experimenter-provided task; (c) each member
We label this the relationships-as-bound hypothesis. Balance should be given fabricated success or failure feedback at the
theory (Heider, 1958) posits that attitudes toward the self extend dyadic level; and (d) each member should attribute privately
to close others. Self-expansion theory (Aron & Aron, 1997) the task outcome to either the self or the partner.
states that the self-concept expands to incorporate a close part- These procedural criteria are most likely to exclude cognitive
ner. According to interdependence theory and the communal- explanations for the obtained SSB, because (a) the dyad's attri-
exchange relationships literature, the goal in close relationships butions are based on work that was completed immediately
is to maximize outcomes for both individuals involved (Rus- before the attributions rather than in the remote past and in a
bult & Arriaga, 1997), with an accompanying genuine concern different setting, thus effectively limiting or perhaps eliminating
for the welfare of the partner (Clark & Mills, 1979). Addition- memorial biases; (b) the quality of an individual's own contri-
ally, in line with the extended self-evaluation maintenance model bution is difficult, if not impossible, to appraise unambiguously;
(Beach & Tesser, 1995), close others are motivated to protect and (c) dyad members are unaware of each other's performance.
both their own and their partners' self-concepts. Finally, as Sedi- We searched for published investigations whose experimental
kides and Strube's (1997) self-concept enhancing tactician procedures fulfilled all of the above-mentioned four critical fea-
model proposes, one function of close relationships is to place tures. We review these investigations below.
an individual's self-enhancing tendencies in check. All of these
theoretical perspectives converge in supporting the notion that,
in close dyads, one will not be more likely to protect or enhance The SSB in Interdependent-Outcomes Tasks:
the self than the partner. Participants whose dyad receives suc- When the Dyad Consists of Distant Partners
cess feedback will share the responsibility for the task outcome
Three experiments have tested whether the SSB is present in
with their partners, and so will participants whose dyad receives
dyads that are involved in interdependent-outcomes tasks and
failure feedback.
are composed of distant partners. In Johnston's (1967) experi-
On the other hand, a diverse body of literature emphasizes ment, participants engaged in a tracking task, in which they
the self-protective or self-enhancing role that close relationships manipulated a control knob in an effort to hold a moving cursor
serve. We label this the relationships-as-enabler hypothesis. In- steady at zero. Participants believed that they were working on
dividuals are not accurate in how they think they are viewed this task with a partner who was sitting in an adjacent room.
by specific related others (Kenny & DePaulo, 1993), probably In actuality, no partner was present. After three experimental
because others do not disclose their true opinion of the individual sessions, each consisting of twenty 60-s trials, participants re-
(Felson, 1993). In fact, close persons avoid evaluating each ceived either success or failure feedback. The type of feedback
other (Goffman, 1959), are more likely to discuss each other's was determined randomly. In the case of success feedback, parti-
positive than negative traits (Blumberg, 1972), and distort the cipants tended to give less credit to the self than to the partner.
communication of information so that it is more consistent with In the case of failure feedback, participants accepted more re-
each other's attitudes or self-concept (Manis, Cornell, & Moore, sponsibility for the task outcome. These patterns demonstrate a
1974). Among close persons, "good news tends to be communi- reversal of the SSB, what we call the other-serving bias (OSB).
cated more frequently, more quickly, more fully, and more spon- Participants in Wolosin, Sherman, and Till's (1973) research
taneously than bad news" (MUM effect; lesser & Rosen, 1975, read through 20 pairs of geographic locations and chose, from
p. 228). Perhaps that is why close others are trusted (Holmes & each pair, the location in which they were most likely to reunite
Rempel, 1989). In fact, at times of personal crises, close others with a friend. Participants either cooperated (Experiment 1) or
embrace the suffering individual with social support that regu- competed (Experiment 2). At the end of the task, participants
lates and repairs negative emotions (Cohen & Wills, 1985), received either success or failure feedback. Cooperative partici-
increases feelings of well-being (Cohen & Hoberman, 1983), pants took individual responsibility for the dyadic task success
and elevates the positivity of the self (Major, Testa, & Bylsma, and blamed the partner for the dyadic task failure. Clearly, the
1991). It is not surprising, then, that relationally involved per- SSB was at work. Competitive participants also accepted indi-
sons report higher self-esteem than uninvolved persons (Long, vidual responsibility for success but blamed the self and the
1983). Close others are a safety cushion, if not the springboard partner to an equal degree for failure.
or even the cheerleader, for one's initiation of self-protection or In summary, the evidence regarding the presence of the SSB
380 SEDIKIDES, CAMPBELL, REEDER, AND ELLIOT

in dyadic interdependent-outcomes tasks is weak. In two cases to a less selfish (i.e., more communal) orientation in outcome
(Wolosin et al., 1973, Experiment 1, Experiment 2, success distribution.
condition) support for the SSB was found. In another case Given that participants were unacquainted, they were unlikely
(Johnston, 1967), support for an OSB was reported. In still to expect future interactions, let alone rewarding interactions.
another case (Wolosin et al., 1973, failure condition of Experi- In fact, care was taken to ensure that no participant (a) antici-
ment 2), no support was obtained for either the SSB or the pated to interact with her or his partner after the experiment or
OSB. (b) intended to discuss the experiment with the partner in inci-
One explanation for the weak evidence for the SSB concerns dental encounters outside the laboratory. As part of the proce-
task importance. The SSB becomes stronger as task importance dure, we induced closeness for half of the participants. A vital
and an imminent threat to the self increases (Campbell & Sedi- feature in the development of a close relationship is reciprocal
kides, 1998). In fact, Johnston's (1967) tracking task—a task and escalating self-disclosure (Derlega, Metts, Petronio, & Mar-
that failed to yield an SSB—may not have been perceived by gulis, 1993). We devised a task to simulate such self-disclosure
the participants as particularly important. In our research we between participants (for a similar procedure, see Aron, Meli-
controlled for this potential problem by standardizing task im- nat, Aron, Vallone, & Bator, 1997).
portance. Specifically, we used a task that measures, in a face- On the basis of our theorizing and literature review, we felt
valid manner, creativity—an attribute that college students (our confident in making the prediction that distant partners would
sampled population) regard as important. We believed that the manifest the SSB as long as they perceived the task as important.
introduction of an important task, Coupled with the adoption of However, we were not so certain about close partners. Indeed,
the four critical procedural features described above, would the relationships-as-bound and relationships-as-enabler hypoth-
allow us to demonstrate clearly a motivationally based SSB in eses lead to incompatible derivations. The former hypothesis
dyads composed of distant partners. does not anticipate the emergence of the SSB whereas the latter
hypothesis predicts its presence.
The SSB in Interdependent-Outcomes Tasks:
When the Dyad Consists of Close Partners Method

Several studies have examined the presence of the SSB in Design and Participants
close dyads (e.g., Fincham & Bradbury, 1989; Ross & Sicoly,
1979; Thompson & Kelley, 1981). Some studies have reported The design was a balanced 2 (relationship type: close or distant) X
evidence for the SSB, others have obtained evidence for the 2 (feedback type: success or failure) X 2 (participant gender: female
or male) between-subjects factorial. We included participant gender as
OSB, and still others have obtained support for neither. We do
a variable for exploratory purposes. Past research on dyads involved in
not discuss these findings in detail, however, because two fea- interdependent-outcomes tasks has not reported gender differences in
tures of the studies fail to satisfy our review criteria. First, the SSB. In independent-outcomes tasks, however, men display the SSB
the studies tested participants' memories of past interdependent to a greater degree than women (Campbell & Sedikides, 1998). Addi-
activities rather than participants' evaluation of their currently tionally, we wanted to explore whether the two genders manifest the
completed performance. Second, the studies did not use a suc- SSB differentially as a function of closeness. Is one gender affected by
cess-failure feedback manipulation. In summary, no investiga- closeness more than the other?
tion has tested conclusively whether close participants, who Participants were 80 University of Wisconsin—Madison students.
work on dyadic interdependent tasks, display a motivationally (We excluded 4 additional dyads, because at least 1 dyad member in
based SSB. each suspected that the feedback was false.) In both Experiments 1 and
2, participants were (a) undergraduate students fulfilling an introductory
psychology course option, (b) tested in same-gender dyads, and (c)
EXPERIMENT 1: INDUCING RELATIONSHIP tested by both female and male experimenters. Also, in both experiments
CLOSENESS we used dyads whose members were unfamiliar with each other at the
start of the experiment, as verified by the experimenter.
In Experiment 1 we tested the presence of the SSB in distant
and close participants. We experimentally induced relationship
closeness, then we asked participants to complete a creativity
Procedure and Materials
test in separate rooms, provided them with bogus success or On participants' arrival at the laboratory, an experimenter placed each
failure feedback, and offered them the choice to attribute the dyad in a room and seated the 2 participants across from each other.
task outcome to the self or the partner on a continuum. Participants were informed that they would not see each other at the
We opted to induce relationship closeness (rather than use end of the experiment. Also, they were asked not to interact, or at least
close persons, i.e., friends) to bypass a potential confound, not to talk about this experiment, if they saw each other on campus.
namely the anticipation of future interactions. Friends anticipate Next, participants learned that they would engage in two short and
unrelated studies, the first of which would involve a communication
seeing each other after the experiment. Thus, friends may refrain
task. This ostensible study actually was the relationship closeness induc-
from the SSB to avoid an account of any selfish attributions tion task (RCIT), a structured self-disclosure task. The RCIT consists
about which the partner might inquire. In other words, friends of three lists of questions (which become progressively more personal)
may refrain from the SSB for the sake of relationship mainte- and instructs participants to spend 9 min mutually self-disclosing while
nance. In fact, in an influential literature on communal-exchange engaging in as natural a conversation as possible. Participants spend 1
relationships, Mills and Clark (1982) suggested that anticipa- min on List I (7 questions; e.g., "How old are you?" and "Where are
tion of future and rewarding interactions with a partner can lead you from?"), 3 min on List II (12 questions; e.g., "What are your
 SELF-SERVING BIAS IN RELATIONSHIPS 381
hobbies?" and "What would you like to do after graduating from this participant answered the two critical questions that assessed the SSB:
university?"), and 5 min on List i n (10 questions; e.g., "Is it difficult "Who was most responsible for the outcome of this test?" (1 = the
or easy for you to meet people? Why?" and "Tell me one thing about other participant, 10 = myself) and "Who made the greatest positive
yourself that most people who already know you don't know?''). Parti- contribution to this test?" (1 -the other participant, 10 = myself)."
cipants completed the RCIT in the absence of the experimenter. Next, participants completed the success and failure manipulation
After completing the RCIT, participants marked on separate sheets of checks. They responded to two questions: "How well do you think that
paper the number of questions they had asked each other from each of both you and the other participant did on this test?" (1 = not at all
the three lists. Participants asked an average of 6.46 questions from List well, 9 = very well) and "How important was the outcome of this test
I, 9.08 questions from List II, and 9.00 questions from List III. Partici- to you?" (1 = not at all important, 10 = very important). At the end,
pants also reported that they had adequate privacy, felt comfortable, and participants gave their open-ended responses to the question "What do
considered conversation a valid way to become familiar with a stranger, you think the true purpose of this experiment was?" and were debriefed
and they reported frequent engagement in conversations like the one individually and thoroughly.
instigated by the RCIT.1 Most important, participants completed a ma-
nipulation check of relationship type consisting of four single-item 9- Results and Discussion
point scales that assessed closeness, similarity, degree of liking, and
likelihood of future friendship. The scales were worded as follows: Manipulation Checks
"How close do you feel to the participant with whom you are working
on this study?" (1 = not at all close, 9 = very close); "How similar Relationship Type
do you feel to the participant with whom you are working on this
study?" (1 = not at all similar, 9 = very similar); ' 'How much do you
We averaged the means on the four relationship type scales
like the participant with whom you are working on this study?" (1 = ( a = .78) to form a composite index. Participants in the close
not at all, 9 = very much); and "In the future, to what extent do you feel condition reported a higher level of relationship closeness (A/
you could befriends with the participant with whom you are working on = 5.39) than participants in the distant condition {M = 4.05),
this study?" (1 = not at all, 9 = very much).2 F( 1, 78) = 26.39, p < .0001. The RCIT was effective in induc-
Participants were subsequently informed that it was time for Study 2, ing closeness.
which involved "a test of creativity." Participants in the close condition
remained with the same partner; participants in the distant condition
Creativity Importance
were switched to a new partner, who had just completed the RCIT with
another participant. We followed this practice to ensure that participants A t test testing the significance of the overall mean (Af =
in both the close and distant conditions went through an identical rela- 7.90) against the scale midpoint (5.50) revealed that creativity
tionship induction procedure. was an important trait to participants, r(79) = 13.36, p < .0001.
The test of creativity was presented to participants as the "Lange-
Elliot Creativity Test." Participants were told that this second study
concerned the "effects of brainstorming on the creativity of dyads." Success and Failure Feedback
Brainstorming was defined as "coming up with as many uses for an
The feedback type manipulation was effective. Participants
object as you can." Participants were further informed that they would
who succeeded reported that both they and their partner per-
receive normative performance feedback.
formed better (Af = 8.60) compared to reports of participants
Participants learned that the test would consist of two segments. In
each segment the participant and her or his partner would be allotted 5
who failed (M = 4.00), F ( l , 72) = 172, p < .001, and they
min to generate as many uses as possible for an object ("brick" and considered the creativity test as more important (M = 5.13)
"candle" served as the objects for the two segments). The experimenter than did participants who failed (Af = 3.95), F ( l , 72) = 4.34,
would place the uses each participant generated in a box along with p < .04.
those the partner generated. Participants were told that the total number
of nonoverlapping uses each dyad generated would be summed to form Dependent Measures
a combined creativity score. At this point, participants completed the
creativity importance manipulation check: They rated how important the Responsibility
trait of creativity was to them (1 = not at all important, 10 = very
important). Of central interest was the Feedback Type X Relationship
Next, participants completed the creativity test, and each received Type interaction. In the case of distant relationships, we pre-
performance feedback that was determined randomly and referred to the dicted the emergence of the SSB. However, in the case of close
partners' combined performance. Success condition participants were
informed that they had scored at the 93rd percentile and were given an
explanation ("You scored better than 93% of the individuals used in our 1
We obtained identical results in Experiment 2.
normative reference sample") and an interpretation ("You did well") of 2
We also assessed participants' mood ("How do you feel right
their performance. Failure condition participants were informed that they now?"; 1 = very sad, 9 = very happy). There was no mood difference
had scored at the 31st percentile and were also given an explanation between distant and close participants.
( " \ b u scored worse than 69% of the individuals used in our normative 3
We arrived at these feedback percentile scores through pilot testing.
reference sample") and interpretation ("'You did poorly") of their 4
Dyadic data present an intradyad dependency: Responses within
performance.3
dyads may be more similar to each other than responses within other
Subsequently, participants were instructed that, because the ' 'Lange- dyads. To examine this possibility, we treated dyad as the independent
Elliot Creativity lest" was based on pooled scores, the experimenter variable, using the one-way analysis of variance method (Myers, Di-
was unable to determine which participant was more responsible for Cecco, & Lorch, 1981), and calculated an intraclass correlation. In
the overall result obtained by the dyad. To assist the experimenter with neither of the reported experiments was dyadic membership related sig-
this task, and under the guarantee of confidentiality and anonymity, each nificantly to the dependent measures.
382 SEDIKBDES, CAMPBELL, REEDER, AND ELLIOT

relationships our theory-based derivations were divergent: The did not differ in terms of the personal responsibility they took
relationships-as-bound hypothesis did not anticipate the emer- for dyadic success and failure. In a conceptual replication of this
gence of the SSB, whereas the relationships-as-enabler hypothe- pattern, close participants claimed a less positive contribution to
sis predicted the presence of the SSB. the task outcome than did distant participants. The findings of
The interaction was significant, F ( l , 72) = 5.41, p < .02. Experiment 1 support the relationships-as-bound hypothesis
Our prediction pertaining to distant relationships was confirmed: rather than the relationships-as-enabler hypothesis.
Distant participants assumed greater personal responsibility for
the dyad's success (M = 6.20) than for its failure (M = 5.23), EXPERIMENT 2: ROLE O F PARTNER
f(38) = 2.20,p < .03, thus manifesting the SSB. It is important IMPRESSIONS
to note, however, that close participants took neither greater
Experiment 1 was the first compelling empirical demonstra-
responsibility for the dyad's success (M = 5.30) nor lower
tion of a motivationally based SSB in distant partners working
responsibility for the dyad's failure (M = 5.60), f(38) = 0.88,
jointly on an interdependent-outcomes task. More interesting is
p < .39. Close participants did not manifest the SSB, thus
that Experiment 1 also demonstrated that the SSB does not
lending support to the relationships-as-bound hypothesis.
emerge among close partners working jointly on such a task.
The Feedback Type X Participant Gender interaction was sig- Close partners appeared to hold self-oriented biases (i.e., self-
nificant, F( 1,72) = 4.21, p < .04. Men assumed greater respon- enhancement) in check. These novel findings invite replication,
sibility for the dyad's success (Af = 5.85) than for its failure which was one objective of Experiment 2.
(M = 4.95), r(38) = 2.03, p < .05, whereas women's attribu-
A more important objective of Experiment 2, however, was
tions for the dyadic success (M = 5.65) and for its failure (M
the search for mechanisms underlying the elimination of the
= 5.88) did not differ significantly, r(38) = -0.66, p < .52.
SSB in close relationships. The mechanism that this experiment
The Feedback Type X Participant Gender X Relationship Type
tested was differential impression favorability for close and dis-
interaction was not significant, F( 1,72) = 0.60, p < .44. Close-
tant partners. Do close partners refrain from the SSB because
ness did not affect the manifestation of the SSB differentially
they form more positive impressions of each other?
in women and men.
Person perception research has attested to the power of im-
pressions to affect ensuing psychological processes. A per-
Positive- Contribution Attributions ceiver's impression of another person determines to a substan-
tial extent the way in which the perceiver will think, feel, and
In this analysis, the critical question was whether participants behave toward this person (Berscheid & Walster, 1978). Do
in close (vs. distant) dyads would claim to have made a greater individuals' impressions of close versus distant others differ?
positive contribution to the outcome, relative to the partner. The Indeed, individuals evaluate close others more favorably than
feedback type main effect and the Feedback Type X Relationship they evaluate strangers. For example, compared to strangers,
Type interaction are not diagnostic of the SSB, given that the individuals consider friends more sincere, dependable, and
question probes for greater positive contribution to the test. considerate and less spiteful, rude, and superficial (Brown,
Feedback type is in no position to detect the SSB, because 1986, Experiments 2 - 3 ) .
participants can claim having made a more positive contribution
Differential favorability (i.e., more positive impressions) of
to both the success and the failure of the dyad. In contrast, the
close versus distant partners may explain the findings of Experi-
relationship type main effect is of major interest.
ment 1. Members of close dyads may form a more positive
The relationship type main effect was significant. Close parti- impression of each other, compared to members of distant dyads.
cipants reported making a less positive contribution to the test As a consequence, the former will not manifest the SSB, whereas
outcome (M - 5.27) relative to distant participants (M = 5.90), the latter will. In Experiment 2 we tested partner impressions
F ( l , 71) = 6.64, p < .01. (One participant did not answer as a mediator of the SSB.
the relevant question.) This finding parallels the results of the
Feedback Type X Relationship Type interaction pertaining to
Method
the responsibility question. The finding is consistent with the
relationships-as-bound hypothesis. Design and Participants
Men claimed a greater positive contribution to the test out- The design was a 2 (relationship type: close or distant) X 2 (feedback
come (M = 5.85) than did women (M = 5.32), participant type: success or failure) x 2 (participant gender: female or male) be-
gender main effect, F ( l , 71) = 4.70, p < .03. The Participant tween-subjects factorial. Cell sizes ranged from 12 to 16. Participants
Gender X Relationship Type interaction was not significant, were 104 students at the University of North Carolina at Chapel Hill.
F ( l , 7 1 ) = 0.11,p < .75. No participant expressed suspicion about the nature of the feedback.5

Procedure and Materials

Summary
The procedure of Experiment 2, like the procedure of Experiment 1,
In Experiment 1 we examined the manifestation of the SSB had two parts. The first part was identical to that used in Experiment 1
in dyads working on an interdependent-outcomes task. We tested
whether members of distant and close dyads display the SSB. 5
Close participants reported being in a happier mood (M = 6.60)
The SSB emerged in distant dyads but not in close dyads. That than distant participants (M = 5.88), F ( l , 102) = 6.69, p < .01.
is, distant participants took greater personal responsibility for However, we observed no changes in the reported results when we used
the dyadic success than for its failure, whereas close participants mood as a covariate in analyses of covariance (ANCOVAs).
 SELF-SERVING BIAS IN RELATIONSHIPS 383

and included the administration of the RCIT. Participants asked each dyadic failure (Af = 5.92), f(46) = -0.39, p < .70. These
other an average of 6.43 questions from List I, 8.91 questions from List patterns replicate the corresponding findings of Experiment 1.
II, and 9.17 questions from List III. The second part (i.e., creativity The Feedback Type X Participant Gender interaction was sig-
test) of Experiment 2 differed from Experiment 1 in one important way: nificant, F ( l , 96) = 4.76, p < .03. The pattern was identical
Participants completed a modified version of the Interpersonal Judgment
to that in Experiment 1: Men assumed greater responsibility for
Scale (US; Byrne, 1971) immediately before receiving test feedback.
dyadic success (Af = 6.25) than for its failure (Af = 5.12),
The modified US was designed to assess participants' impression of
their partners. Participants read: ?(48) = 2.39, p < .02, whereas women did not differ in their
responsibility attributions for dyadic success (Af = 5.38) and
While we calculate your combined score on the Lange-Elliot Cre- failure (Af = 5.56), f(52) = 0.46, p < .64. The triple interaction
ativity Test, please answer the following questions to the best of was not significant, F ( l , 96) = 1.83, p < .18.
your ability. We are interested in your impression of the participant
with whom you are working on this study. We know you may
have had only a small opportunity to form an impression of this
Positive-Contribution Attributions
participant. Please answer these questions using the little informa- As in Experiment 1, the relationship type main effect was
tion you have, "four responses will be kept confidential—they will significant. Close participants reported making a less positive
not be shown to the other participant.
contribution to the test (Af = 5.31) relative to distant partici-
pants (Af = 6.03), F ( l , 96) = 5.87, p < .02.
Participants responded to the following seven items of the modified US:
"The other participant: (a) is very intelligent; (b) is very moral; (c) Men claimed a greater positive contribution to the test out-
has a good knowledge of current events; (d) is very creative; (e) is very come (Af = 6.07) than did women (Af = 5.28), participant
well adjusted; (f) is very fair; and (g) is exactly the kind of person with gender main effect, F( 1, 96) = 7.15, p < .009. The Participant
whom I would like to work in an experiment." Participants responded to Gender X Relationship Type interaction was not significant,
these items on scales that ranged from 1 (not at all true) to 8 (very F ( l , 96) = 0.001, p < .99.
true). Next, participants completed the dependent measures.
Modified US
Results and Discussion
Responses to the seven modified US items were highly inter-
Manipulation Checks related ( a = .91). We averaged these responses to create an
index, termed partner impression. (One participant did not com-
Relationship Type plete the modified US.) We then tested the prediction that partici-
pants in the close condition formed a more favorable impression
We averaged the means on the four relationship type scales
of their partner than did participants in the distant condition, by
(a = .89) to form a composite index. Once again, the RCIT
conducting an analysis of variance (ANOV\) with relationship
was effective: Participants in the close condition reported a
type and participant gender as independent variables and partner
higher level of relationship closeness (Af = 5.67) than partici-
impression as the dependent variable. As predicted, participants
pants in the distant condition (Af = 3.77), F( 1, 102) = 48.82,
in the close condition reported a more favorable impression of
p < .0001.
their partner (Af = 5.13) than did participants in the distant
condition (Af = 4.04), F ( l , 99) = 53.16, p < .0001.
Creativity Importance
The t test that tested the significance of the overall mean (Af Mediational Analyses
= 7.84) against the scale midpoint (5.50) was significant,
Do the positive impressions that close dyad members form of
f(102) = 13.16, p < .0001. Creativity was important to
each other account for the elimination of the SSB? Statistically
participants.6
speaking, does partner impression mediate the association be-
tween relationship type and the SSB? 7
Success and Failure Feedback Responsibility attributions. The mediation concerning re-
sponsibility attributions should be interactional (Baron &
Participants who succeeded rated the test as more important
Kenny, 1986). That is, the Relationship Type X Feedback Type
(Af = 4.97) than did participants who failed (Af = 3.69), F ( l ,
interaction should predict the partner impressions, and the Part-
96) = 6.77, p < .01. (We used only this question in Experiment
ner Impressions X Feedback Type interaction, in turn, should
2.) Our success and failure feedback manipulation was effective.
predict the responsibility attributions. We carried out the media-
tional analyses in several steps.
Dependent Measures We have already established that relationship type predicts
Responsibility
6
Creativity was unexpectedly more important to close participants
The crucial Feedback Type x Relationship Type interaction
(M = 8.25) than to distant participants (M = 7.49), F( 1, 102) = 4.73,
was significant, F ( l , 96) = 4.42, p < .04. Distant participants p < .03. However, the reported results were not altered in ANCOV\s
took greater personal responsibility for the success of the dyad that used creativity ratings as a covariate.
(Af = 5.88) than for its failure (Af = 4.77), f(54) = 2.36, p 7
To simplify the presentation of the mediational analyses, we excluded
< .02. However, close participants did not take greater responsi- gender from the model. Analyses that included gender produced results
bility for dyadic success (M = 5.75) or lower responsibility for identical to the ones presented in the text.
384 SEDIKIDES, CAMPBELL, REEDER, AND ELLIOT

responsibility attributions (Relationship Type X Feedback "type the partner impression main effect was significant, F ( l , 100)
interaction, F[l, 100] = 3.78, p < .05). Furthermore, we have = 4.30, p < .04.
established that relationship type predicts partner impression In conclusion, partner impression mediated the effect of rela-
(relationship type main effect, F[l, 101] = 54.81, p < .0001). tionship type on positive-contribution attributions. Compared to
The next step involved testing whether partner impression distant participants, close participants claimed a less positive
predicts responsibility attributions. More formally, does the contribution to the test outcome because of their more favorable
Partner Impression X Feedback Type interaction predict respon- impression of their partner.11
sibility attributions? We conducted a median split on partner
impression (Mdn = 4.19), and performed an ANOVA with part- Summary
ner impression and feedback type as independent variables and
responsibility attributions as the dependent variable. The inter- Experiment 2 established partner impression as a mediator
of the SSB. Members of relationally close dyads working on
action was significant, F( 1, 99) = 3.03, p < .05. When partners
interdependent-outcomes tasks refrained from the SSB, in part
formed a positive impression of each other, they did not manifest
because they formed a positive impression of the partner. As
the SSB (Success M = 5.72; Failure M = 5.80; f[52] = -0.19,
was the case with Experiment 1, the results were consistent with
p < .85). However, when partners formed a negative impression
the relationships-as-bound hypothesis.
of each other, they did manifest the SSB (Success M = 6.00;
Failure M = 4.88; ;[47] = -2.35, p < .02). 8
Last, we examined the complete model with (a) relationship GENERAL DISCUSSION
type and feedback type as independent variables, (b) responsi- Social and personality psychology has maintained a persistent
bility attributions as the dependent variable, and (c) partner interest in viewing the self in relational context (Borden &
impression as well as the Partner Impression X Feedback Type Levinger, 1991; Holmes & Murray, 1996). The present investi-
interaction as covariates. If an interactional mediation exists, gation reflects this interest. The investigation was fueled primar-
two conditions will need to be met. First, the Relationship Type ily by two incompatible hypotheses. The relationships-as-bound
X Feedback Type interaction should become nonsignificant. hypothesis posits that close relationships place limits on an
Second, the Partner Impression X Feedback Type interaction individual's self-enhancement tendencies. The individual is dis-
should remain significant. Both conditions were met: The Rela- couraged from using the relationship for his or her own benefit.
tionship Type X Feedback Type interaction became nonsignifi- This hypothesis does not anticipate the emergence of the SSB
cant, F ( l , 97) = 0.08, p < .78, and the Partner Impression X in close relationships. In contrast, the relationships-as-enabler
Feedback Type interaction remained significant, F ( l , 99) = hypothesis proposes that relationships act as enablers that pro-
5.72, p < .02. tect and even enhance the individual's self-concept. Close rela-
In conclusion, partner impression mediated the effect of rela- tionships support, sustain, and even encourage an individual's
tionship type on responsibility attributions. Compared to distant self-enhancement tendencies. By implication, the individual
participants, close participants refrained from the SSB because should feel free to use the relationship for her or his own benefit.
of their more favorable impression of their partner.9 This hypothesis predicts the emergence (or even magnification)
Positive-contribution attributions. We followed a similar of the SSB in close relationships.
strategy in examining the mediational role of partner impression Members of either distant or close dyads worked on an inter-
in the effect of relationship type on the positive-contribution dependent-outcomes task (a creativity test), received bogus suc-
attributions. A conceptual replication of the responsibility attri- cess or failure feedback at the dyadic level, and attributed the
bution mediational results would be as follows: Relationship dyad's performance to either the self or the partner. Participants
type would predict partner impressions, which in turn would manifested the SSB when the partner was a distant other,
predict the positive-contribution attributions. whereas they refrained from the SSB when the partner was a
We have already shown that relationship type predicted posi- close other. Also, compared to distant participants, close partici-
tive-contribution attributions, F ( l , 102) = 5.23, p < .02, and pants claimed that they made a less positive contribution to the
that relationship type predicted partner impression. An ANO\A task outcome. These findings are consistent with the relation-
also demonstrated that partner impression (Mdn = 4.29) pre-
dicted positive contribution attributions. Participants were more 8
For reasons of presentational clarity and consistency, we opted to
likely to display the SSB when they had an unfavorable impres- report in the text the ANOV\ results rather than the more powerful
sion of the partner (M = 5.33) than when they had a favorable regression results. The beta for the Partner Impression X Feedback Type
impression of the partner (M = 6.10), F ( l , 101) = 6.60, p interaction was -1.65, f(99) = -3.40, p < .001.
9
< .01.'° We repeated the steps of the interactional mediation analyses twice.
Last, we examined the complete model with (a) relationship Specifically, we assessed the mediational role of (a) liking for the partner
type and feedback type as independent variables, (b) positive- (Questions 2, 5, 6, and 7 on the modified US) and (b) perceptions of
partner ability (Questions 1, 3, and 4 ) . In both cases the results of the
contribution attributions as the dependent variable, and (c) part-
mediational analyses were identical to the ones described in the text.
ner impression as a covariate. A mediational relation will be 10
The beta for the partner impression main effect was -0.30, t( 101)
established if (a) the relationship type main effect becomes = -3.13, p < .002.
nonsignificant and (b) the partner impression effect remains " We repeated the results of the mediational analyses twice, as we
significant. This was indeed the case. The relationship type main had done for the responsibility question. The results were identical to
effect was not significant, F( 1, 100) = 0.55, p < .46, whereas those reported.
 SELF-SERVING BIAS IN RELATIONSHIPS 385

ships-as-bound hypothesis: Closely related individuals refrain Blumberg, H. H. (1972). Communication of interpersonal evaluations.
from self-enhancement. Relationship closeness is an effective Journal of Personality and Social Psychology, 23, 157-162.
prescription for modesty. Borden, V. M. H., & Levinger, G. (1991). Interpersonal transformations
Why do close partners refrain from the SSB? In Experiment in intimate relationships. Advances in Personal Relationships, 2, 3 5 -
2 we tested an explanation based on favorable partner impres- 56.
sions: Close participants refrain from the SSB because they Brown, J. D. (1986). Evaluations of self and others: Self-enhancement
form a favorable impression of each other. Future research biases in social judgments. Social Cognition, 4, 353-376.
should focus on additional mechanisms in an effort to elucidate Byrne, D. (1971). The attraction paradigm. New "York: Academic Press.
more precisely the relationships-as-bound hypothesis. Campbell, W. K., & Sedikides, C. (1998). Self-threat magnifies the self-
serving bias: A meta-analytic integration. Unpublished manuscript,
The results revealed reliable gender differences in the mani-
University of North Carolina at Chapel Hill.
festation of the SSB. Men were more likely than women to
Clark, M. S., & Mills, J. (1979). Interpersonal attraction in exchange
display the SSB. This result is consistent with literature that
and communal relationships. Journal of Personality and Social Psy-
reports gender differences in the manifestation of the SSB in chology, 37, 12-22.
independent-outcomes settings (Campbell & Sedikides, 1998).
Cohen, S., & Hoberman, H. (1983). Positive events and social supports
This gender difference has been attributed to men having higher as buffers of life change stress. Journal of Applied Social Psychology,
success expectancies (Rosenfield & Stephan, 1978) and to men 13, 99-125.
having higher global self-esteem (Harter, 1993) than women. Cohen, S., & Wills, T. A. (1985). Stress, social support, and the en-
On a related note, the two genders were not affected differently ablering hypothesis. Psychological Bulletin, 98, 310-357.
by closeness. Derlega, V. J., Metts, S., Petronio, S., & Margulis, S. T. (1993). Self-
A remaining issue concerns the limitations in our measure of disclosure. Newbury Park, CA: Sage.
the SSB. Our research participants were constrained to allocate Felson, R. B. (1993). The (somewhat) social self: How others affect
responsibility for the joint outcome between themselves and self-appraisals. In J. Suls (Ed.), Psychological perspectives on the
their partner. We selected this measure because we regarded it self (Vol. 4, pp. 1-26). Hillsdale, NJ: Erlbaum.
as the most direct test of the two hypotheses involved. On this Fincham, F. D., & Bradbury, T. N. (1989). Perceived responsibility for
measure, members of close dyads were less inclined to self- marital events: Egocentric or partner-centric bias? Journal of Mar-
enhance compared to members of distant dyads. However, on riage and the Family, 51, 27-35.
many tasks in everyday life, individuals' attributional choices Goffman, E. (1959). The presentation of self in everyday life. New
are less constrained. In a tennis doubles match, for example, York: Doubleday.
players can make attributions to their team (as a unit), to the Harter, S. (1993). Causes and consequences of low self-esteem in chil-
referees, or to the quality of the opposing team. As a result, dren and adolescents. In R. Baumeister (Ed.), Self-esteem: The puzzle
self-serving attributions for the team's outcome, or for one's of low self-regard (pp. 87-116). New 'York: Plenum.
own contribution to that outcome, need not involve belittling Heider, F. (1958). The psychology of interpersonal relations. New York:
Wiley.
one's teammates. Future studies will need to develop additional
measurement strategies to test the external validity of our Holmes, J. G., & Murray, S. L. (1996). Conflict in close relationships.
In E. T. Higgins & A. W. Kruglanski (Eds.), Social psychology: Hand-
findings.
book of basic principles (pp. 622-654). New York: Guilford Press.
We opened this article with one of Murphy's laws. In light Holmes, J. G., & Rempel, J. K. (1989). Trust in close relationships. In
of our empirical findings, we are obliged to conclude with a C. Hendrick (Ed.), Review of personality and social psychology (Vol.
revision of that law: "If more than one person is responsible 10, pp. 187-220). London: Sage.
for a miscalculation, and the persons are close, both will be at Johnston, W. A. (1967). Individual performance and self-evaluation in
fault." a simulated team. Organizational Behavior and Human Performance,
2, 309-328.
References Kenny, D. A., & DePaulo, B. M. (1993). Do people know how others
view them? An empirical and theoretical account. Psychological Bul-
Aron, A., & Aron, E. N. (1997). Self-expansion motivation and includ- letin, 114, 145-161.
ing other in the self. In S. Duck (Ed.), Handbook ofpersonal relation-
Long, B. H. (1983). Heterosexual involvement of unmarried undergrad-
ships: Theory, research and intervention (2nd ed., pp. 251-270).
uate females in relation to self-evaluations. Journal of Youth and
Chichester, England: Wiley.
Adolescence, 18, 489-500.
Aron, A., Melinat, E., Aron, E. N., Vallone, R., & Bator, R. (1997). The
Major, B., Testa, M., & Bylsma, W. H. (1991). Response to upward and
experimental generation of interpersonal closeness: A procedure and
downward comparisons: The impact of esteem relevance and per-
some preliminary findings. Personality and Social Psychology Bulle-
ceived control. In J. Suls & T. A. Wills (Eds.), Social comparison:
tin, 23, 363-377.
Contemporary theory and research (pp. 237-260). Hillsdale, NJ:
Baron, R. M., & Kenny, D. A. (1986). The moderator-mediator variable
Erlbaum.
distinction in social psychological research: Conceptual, strategic, and
statistical considerations. Journal of Personality and Social Psychol- Manis, M., Cornell, S. D., & Moore, J. C. (1974). The transmission of
ogy, 51, 1173-1182. attitude-relevant information through a communication chain. Journal
Beach, S. R. H., & Tesser, A. (1995). Self-esteem and the extended self- of Personality and Social Psychology, 30, 81 —94.
evaluation maintenance model: The self in social context. In M. Kernis Miller, D. T, & Ross, M. (1975). Self-serving biases in attribution of
(Ed.), Efficacy, agency, and self-esteem (pp. 145-170). New 'York: causality: Fact or fiction? Psychological Bulletin, 82, 213-225.
Plenum. Mills, J., & Clark, M. S. (1982). Communal and exchange relationships.
Berscheid, E., & Walster, E. (1978). Interpersonal attraction (2nd ed.). In L. Wheeler (Ed.), Review of personality and social psychology
Reading, MA: Addison-Wesley. (pp. 121-144). Beverly Hills, CA: Sage.
386 SEDIKIDES, CAMPBELL, REEDER, AND ELLIOT

Myers, J. L., DiCecco, J. V, & Lorch, R. R, Jr. (1981). Group dynamics In L. Berkowitz (Ed.), Advances in experimental social psychology
and individual performances: Pseudogroup and quasi-F analyses. (Vol. 8, pp. 193-232). New York: Academic Press.
Journal of Personality and Social Psychology, 40, 86-98. Thompson, S. C , & Kelley, H. H. (1981). Judgments of responsibility
Rosenfield, D., & Stephan, W. G. (1978). Sex differences in attributions for activities in close relationships. Journal of Personality and Social
for sex-typed tasks. Journal of Personality, 46, 244-259. Psychology, 41, 469-477.
Ross, M., & Sicoly, R (1979). Egocentric biases in availability and Weary-Bradley, G. (1978). Self-serving biases in the attribution process:
attribution. Journal of Personality and Social Psychology, 37, 322- A reexamination of the fact or fiction question. Journal of Personality
336. and Social Psychology, 36, 56-71.
Rusbult, C. E., & Arriaga, X. B. (1997). Interdependence theory. In S. Wolosin, R. J., Sherman, S. J., & Till, A. (1973). Effects of cooperation
and competition on responsibility attribution after success and failure.
Duck (Ed.), Handbook of personal relationships: Theory, research
Journal of Experimental Social Psychology, 9, 220-235.
and intervention (2nd ed., pp. 221-250). Chichester, England: Wiley.
Zuckerman, M. (1979). Attribution of success and failure revisited, or:
Sedikides, C , & Strube, M. J. (1997). Self-evaluation: To thine own
The motivational bias is alive and well in attribution theory. Journal
self be good, to thine own self be sure, to thine own self be true,
of Personality, 47, 245-287.
and to thine own self be better. In M. P. Zanna (Ed.), Advances in
experimental social psychology (Vol. 29, pp. 209-269). New \brk: Received July 29, 1997
Academic Press. Revision received October 22, 1997
Tesser, A., & Rosen, S. (1975). The reluctance to transmit bad news. Accepted October 22, 1997 •