See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/11547104

Self-Awareness, Probability of Improvement, and the Self-Serving Bias

Article in Journal of Personality and Social Psychology · January 2002

DOI: 10.1037/0022-3514.82.1.49 · Source: PubMed

CITATIONS READS

225 9,808

2 authors, including:

Paul J Silvia
University of North Carolina at Greensboro
388 PUBLICATIONS 29,819 CITATIONS

SEE PROFILE

All content following this page was uploaded by Paul J Silvia on 01 June 2014.

The user has requested enhancement of the downloaded file.

Journal of Personality and Social Psychology Copyright 2002 by the American Psychological Association, Inc.
2002, Vol. 82, No. 1, 49 – 61 0022-3514/02/$5.00 DOI: 10.1037//0022-3514.82.1.49

Self-Awareness, Probability of Improvement, and the Self-Serving Bias

Thomas Shelley Duval Paul J. Silvia

University of Southern California University of Kansas

Evidence for the self-serving bias (attributing success internally and failure externally) is inconsistent.
Although internal success attributions are consistently found, researchers find both internal and external
attributions for failure. The authors explain these disparate effects by considering the intersection of 2
systems, a system comparing self against standards and a causal attribution system. It was predicted that
success and failure attributions are moderated by self-awareness and by the ability to improve. When
self-focus is high (a) success is attributed internally, (b) failure is attributed internally when people can
improve, (c) failure is attributed externally when people cannot improve, and (d) these attributions affect
state self-esteem. Implications for the self-serving bias are discussed.

The self-serving attributional bias—attributing success inter- that threaten us or give us pain than do accurate perceptions. In a
nally and failure externally—appears for many psychologists to sense, they poorly serve the goals of maximizing pleasure and mini-
have achieved the status of an empirical fact (Brown & Rogers, mizing pain in the long term. (p. 234)
1991). Researchers indeed find a consistent tendency for individ-
At a more general level, people have many motives and goals
uals to attribute success to self (e.g., Miller & Ross, 1975). Yet the
simultaneously (Heider, 1958; Lewin, 1935). The presence of
literature on failure attributions shows a lot of variability. Many
contradictory effects suggests that unidentified moderators and
experiments find external attributions for failure (e.g., Snyder,
additional motivations are operating. Our goal is to develop a
Stephan, & Rosenfield, 1976, 1978). Many other studies, however,
model of self-serving attributions based on the interplay of two
find internal attributions for failure (e.g., Ames, 1975; Ross, Bier-
motivational systems, which we discuss in more detail below.
brauer, & Polly, 1974; Weary et al., 1982). In fact, Zuckerman’s
(1979) qualitative review reveals broad diversity in effects.
Among 13 studies on attributions for failed interpersonal influ- Self-Enhancement and Self-Assessment
ence, for example, 5 found a self-serving pattern, 4 found a Self-enhancement motivation engenders a preference for cogni-
non-self-serving pattern, and 4 found no effects. A recent quanti- tions and interpretations that foster a positive self-concept. This
tative review found that the distribution of effect sizes ranged general motive has broad psychological consequences (Dunning,
across positive and negative values (Campbell & Sedikides, 1999, 1999; Taylor & Brown, 1988). Research on self-understanding
p. 34); self-serving and non-self-serving effects were observed (Sedikides, 1993), for instance, shows that people seek self-
frequently. knowledge in ways designed to yield flattering results. After
The diversity of effects for failure attributions seems reasonable self-reflecting, people preferred to ask highly diagnostic questions
when we consider the costs of a dominant tendency toward defen- of themselves when the trait was highly self-relevant and posi-
siveness. As Nisbett and Ross (1980) argued tively valenced. When the trait was highly self-relevant and neg-
The costs of willy-nilly distortions in perception are simply too high atively valenced, people selected less diagnostic questions. People
to make them a cure-all for the disappointed or threatened perceiver. also generated questions judged to be more diagnostic when the
In general, misperceptions make us less able to remedy the situations trait was central to self and positive, as opposed to central and
negative. Finally, people consistently affirmed possessing posi-
tively valenced traits and denied having negative traits.
Thomas Shelley Duval, Department of Psychology, University of South- Yet self-enhancement is not the only motive associated with the
ern California; Paul J. Silvia, Department of Psychology, University of self-concept. People also want accurate information about the
Kansas. extent of their abilities and the correctness of their opinions (Fest-
Both authors contributed equally to this article. We thank Laurie inger, 1954; Trope, 1986). Self-assessment motivation promotes
O’Brien for her statistical advice. These experiments were presented at the seeking and preferring information that provides accurate knowl-
meeting of the Midwestern Psychological Association in Chicago, May edge about the self (Sedikides & Strube, 1997). Such knowledge is
2001. Preparation of this article was supported by National Science Foun- not necessarily flattering or unflattering to one’s self-image. The
dation Grant CSM 9726575.
operation of self-assessment motivation is seen when people seek
Correspondence concerning this article should be addressed to Thomas
Shelley Duval, Department of Psychology, SGM 501, University of South-
accurate feedback about their limitations and liabilities, even
ern California, Los Angeles, California 90089-1061, or to Paul J. Silvia, though this information hinders self-enhancement (Trope & Neter,
who is now at Psychology Institute II, University of Hamburg, Von-Melle- 1994; Trope & Pomerantz, 1998).
Park 5, D-20146 Hamburg, Germany. E-mail: [email protected] or In sum, people are motivated to self-enhance and to self-assess,
[email protected] and these motives may occasionally conflict. Initial research on

49
50 DUVAL AND SILVIA

self-enhancement and self-assessment tried to show that one or the sible cause. Both goal states are thus facilitated by an internal
other dominated activity. Sedikides (1993), for example, attempted attribution to self; this should bolster such attributions, as shown
to create conditions in which the positions’ predictions could be by the empirical evidence (Miller & Ross, 1975; Zuckerman,
pitted against one another; he concluded that self-assessment was 1979). But the systems’ goals conflict when self is a plausible
subordinate to self-enhancement. More recently, Sedikides and cause for failure. Attributing failure to self would result in accurate
Strube (1997) argued that accurate self-assessment occurs, but causal understanding, but it would simultaneously increase self–
always in the service of long-term self-enhancement. Other re- standard discrepancy. Conversely, an external attribution would
searchers have argued that neither motive is dominant (e.g., Dun- preserve self–standard congruity, but it would require the attribu-
ning, 1995). In this view, the task is to explore when one or the tion system to link failure with an implausible cause. How is this
other motive prevails, not to assume that one motive is inherently conflict resolved?
stronger. We use this outlook in our model of self-serving attri-
butions, which we describe in detail below. Probability of Improvement as a Moderator

A Dual Systems Model of Self-Serving Attributions The person’s perceived probability of improving his or her
failure should moderate failure attributions. Duval and Duval
We view self-serving attributions as the result of the interaction (1983, 1987), for example, argued that perceiving a high proba-
of two systems, a system comparing self against standards of bility of improvement will lead to an internal failure attribution.
correctness (Duval & Wicklund, 1972) and a causal attribution Granted, locating causality for failure in self should increase
system (Heider, 1944, 1958). This first system compares the incongruity between self and standards and, thus, negative affect.
present state of self against salient standards of correctness when- Yet perceiving improvement as likely should lead to the expecta-
ever attention is focused internally on the self (Duval & Silvia, tion that self–standard congruency will be quickly restored. Any
2001). The self-to-standard comparison system (SSCS) is a goal- short-term costs to self-esteem are thus offset by long-term bene-
directed system. Its preferred state is maximal similarity between fits in accurate self-assessment. But when future improvement
self and standards. If a comparison finds self to be discrepant from seems unlikely, attribution of failure to self would produce an
standards, then negative affect occurs (Ickes, Wicklund, & Ferris, irreducible discrepancy and high negative affect. In this situation,
1973; Sedikides, 1992). But if a comparison finds self to be the benefits of accurate causal knowledge seem outweighed by the
congruent with standards, then positive affect occurs (Ickes et al., long-term drop in self-esteem, given that one’s personal failing is
1973; McDonald, 1980). The person’s hedonic inclination—avoid- seen as permanent (Duval & Duval, 1987).
ing negative affect and seeking positive affect—leads to efforts at Research manipulating the perceived malleability of traits is
self-enhancement, viewed here as meeting one’s personal stan- compatible with this approach. Dunning (1995) led participants to
dards describing how self ought to be. Support for these predic- believe that they had either succeeded or failed on a task measur-
tions is reviewed in detail elsewhere (Duval & Silvia, 2001; Silvia ing an intellectual ability, which was either important or inconse-
& Duval, 2001a). quential. The ability was said to be either modifiable or unmodi-
The second system is the causal attribution system. People fiable. A modifiable trait was described as “one of the most
spontaneously strive to understand the world by making attribu- changeable, least stable, intellectual abilities around” (p. 1302) on
tions for events (Duval & Duval, 1983; Heider, 1944; Wong & which people could either “increase dramatically by practicing” or
Weiner, 1981). Attributional processes are influenced by motiva- “let it drop by ignoring it” (p. 1302). An unmodifiable trait was
tional principles. Rather than randomly attributing events to any described as “one of the most stable, least changeable, intellectual
possible cause, people instead prefer to connect effect events to a abilities— either you have it or you don’t” (p. 1302). These ma-
plausible cause. An accurate understanding of one’s abilities re- nipulations could be construed as inducing high and low proba-
quires an accurate understanding of what the self can and cannot bilities of improvement. The dependent measure was whether
bring about or influence in the environment. The attribution sys- people preferred taking a test that did or did not provide feedback.
tem, then, can also be understood as a motivational system, one People who felt deficient in an important and modifiable ability
that reflects the motive to seek accurate information about the preferred to take additional tests offering performance feedback
causal structure of the world (Heider, 1958). rather than tests not offering feedback. In contrast, people who felt
It appears, then, that the attribution system and the SSCS have deficient in an important but unmodifiable ability preferred tests
different goals. One prefers congruity between self and standards, not offering feedback. Although attributions were not measured,
whereas the other prefers an accurate causal understanding. The these findings are compatible with our general predictions. People
goals of the two systems can thus relate in three ways: They can be who believed their deficiency was malleable sought accurate in-
unrelated, in conflict, or in harmony. The systems’ goals are formation; those who believed their deficiency was stable sought
unrelated when an event occurs whose cause has no implications situations affording self-enhancement.
for self-standard congruity, such as why a neighbor’s child fell ill. People also hold different theories about the malleability of
Provided that the self did not somehow provoke the illness, any traits (Dweck, Chiu, & Hong, 1995), which seem to influence
attribution for the event does not influence the degree of incon- failure attributions in line with our predictions. Dweck and her
gruity between the self and standards. colleagues (Hong, Chiu, Dweck, Lin, & Wan, 1999) first catego-
For our purposes, situations in which the goals coincide or rized people according to implicit theories of trait malleability.
conflict are more interesting. The systems’ goals coincide when One group consisted of persons who believed personal attributes to
self is a plausible cause for success—attributing success to self be relatively fixed (entity theorists); the second group believed that
moves self toward standards and also links an event with a plau- attributes are relatively malleable (incremental theorists). The at-
 SELF-AWARENESS AND ATTRIBUTION 51

tributional consequences of this typology were explored by giving attribution or self-esteem; self-enhancement concerns are reduced
failure feedback to the different theorists and then measuring when people are not comparing the self against personal standards.
attributions to ability and effort. Incremental theorists attributed
more causality for failure to effort than did the entity theorists. Experiment 1
Contrary to predictions, the groups did not differ in their attribu-
tions to ability. Yet the results for attribution to effort are consis- Method
tent with our approach if one assumes that incremental theorists
believe that failure can be quickly improved, whereas entity the- Overview
orists believe that failure cannot be improved.
Participants worked on a self-relevant task when self-awareness was
either high or low. To create a discrepancy between self and standards, we
told all participants that they performed below the study’s standard. They
The Role of Self-Awareness
were then led to perceive either a high (98.3%) or low (2.5%) probability
Our model assumes that a second variable, level of self- of improving their poor performance in the future. Attributions for failure
were then measured.
awareness, also moderates attributions. The system comparing self
against standards only operates inasmuch as attention is directed
inward on the self (Duval & Wicklund, 1972). When self- Participants
awareness is low, the relationship between self and any given Participants were 40 male introductory psychology students who volun-
standard is indistinct and obscure to the person. This simply teered to participate for extra course credit. Each was randomly assigned to
reflects the “spotlight” function of attention (Treisman & Gelade, condition in a 2 (camera/no camera) ⫻ 2 (high/low improvement likeli-
1980): It is difficult to assess similarities and differences among hood) factorial design.
elements unless they receive attentional processing. If the person is
unaware of how self and standard match up, then any discrepancies Procedure
that might exist will not have any affective and motivational
consequences. Yet when self-awareness is heightened, the person Participants arrived individually and were seated at a table in front of a
can discriminate similarities and differences between the self’s TV monitor. The experimenter explained that the study was part of a
current state and the standard. This enables the person to feel research program sponsored by the National Institute for the Study of
positively about congruent relationships and negatively about in- Intellectual Abilities. The study’s purpose was to assess whether the
participant was deficient in the ability to make quantitative judgments in
congruent relationships. Indeed, decades of research (Carver &
three dimensions. If the person was deficient, then the assessment proce-
Scheier, 1998; Silvia & Duval, 2001a) show that the degree of
dures would indicate the degree of deficiency and evaluate the probability
self-focused attention directly affects the degree of self-to-standard that the person could improve in the future. To increase the task’s self-
comparison (e.g., Scheier & Carver, 1983). Increasing self-focus relevance, the experimenter indicated that this judgmental capacity could
beyond baseline levels intensifies the affective and motivational affect performance on several important intellectual tasks and was highly
consequences of the comparison process (Silvia & Gendolla, correlated with future career success.
2001). Likewise, decreased self-focus, such as in deindividuation Self-awareness manipulation. In the high self-awareness conditions, a
(Diener, 1979), attenuates the effects of self–standard comparison. tripod-mounted video camera was placed 3 ft away from the participant
If self-awareness is low, then the SSCS is essentially inert. As a and focused in a tight head shot (Davis & Brock, 1975; Duval, 1976;
result, it cannot interact with the attribution system; the two Duval, Duval, & Mulilis, 1992). Questionnaire responses and experimental
feedback could not be seen on the TV monitor, thus minimizing impression
systems’ goals thus cannot conflict. An absence of conflict means
management interpretations (Dana, Lalwani, & Duval, 1997; Duval &
that there is nothing for improvement beliefs to reconcile, so
Lalwani, 1999). Participants were told that the National Institute requested
perceived probability of improving failure should not have any the videotaping of a random sample of sessions, presumably to ensure
effects on attribution when self-awareness is low. But when self- standardization of testing conditions, and that the present session had been
awareness is high, the comparison system is activated. People randomly selected for videotaping. The video of the present session would
become keenly aware of self–standard discrepancies, and the po- ostensibly be briefly viewed by the experimenters and then erased. In the
tential for conflict with the attribution system arises. Probability of low self-awareness conditions, the video camera faced the wall and all
improvement is thus capable of moderating failure attributions wires were disconnected from electrical outlets. Participants were given the
because it can reconcile the competing demands of accurate self- same information but were told they had not been randomly chosen for
assessment and positive self-enhancement. videotaping.
After the self-awareness manipulation, the participant was told that he
would be assessed with regard to (a) whether he was deficient in the
Predictions domain of quantitative judgments in three dimensions, (b) the degree of
deficiency, and (c) the probability that he could improve through practice.
To summarize, when self-awareness is high and self is a plau- Assessment of these dimensions was presumably achieved by computer
sible cause for success and failure, we expect that (a) attribution of analyses of the number of correct and incorrect responses and the overall
pattern of responses on the test. Analysis of the number of correct and
success to self will be amplified because the two systems’ motives
incorrect responses would indicate whether he currently met the standard.
coincide, (b) failure will be attributed to self when people expect If he did not meet the standard, further analyses of the pattern of responses
a high probability of future improvement, (c) failure will be would yield a highly accurate prediction of the likelihood of eliminating
attributed externally when people do not expect future improve- the deficiency.
ment, and (d) attributions will influence state self-esteem. When The experimenter then said that the capacity to make quantitative judg-
self-awareness is low, improvement likelihood should not affect ments in three dimensions involved the ability to understand the solution to
52 DUVAL AND SILVIA

problems with only some of the relevant facts available. People were often
unaware of a deficiency because its only noticeable effect was a tendency
to make errors on problems requiring quantitative judgments, an effect
usually attributed to lapses in attention. This information was conveyed to
make the forthcoming failure feedback more plausible.
The “diagnostic test” involved deciding which of five two-dimensional
figures, when folded to make a three-dimensional figure, would match the
criterion three-dimensional figure shown in the test item. The experimenter
gave the participant two sample test items, turned on the videotape of the
test, and left the room. A total of 45 problems were presented on a TV
monitor in sets of 3 problems. After 45 s of exposure to each set, the
program continued to the next set; this procedure continued until all 45
problems had been presented.
The experimenter then reentered the room and took the participant’s
answer sheet to be analyzed. The experimenter actually went to a separate
room and picked up a computer printout that another member of the
research team had randomly designated for that participant. After 5 min, he
returned to the participant’s cubicle and opened the feedback sheet. The
experimenter had been unaware of the participant’s probability of improve-
ment up to this point.
Failure feedback and probability of improvement manipulation. After
opening the printout, the experimenter told all participants that analyses Figure 1. Failure attribution to self as a function of self-awareness and
indicated a deficiency in their ability to make quantitative judgments. improvement likelihood: Study 1.
Pointing to a statement at the bottom of the printout, he explained that the
participant “was deficient from the standard level of ability by 10%.” In the
and a significant interaction, F(1, 36) ⫽ 9.54, p ⬍ .01. When
high probability of improvement condition, he pointed out that the com-
puter analyses indicated a 98.3% probability of substantially reducing the self-awareness was high, persons expecting improvement attrib-
deficiency through practice on related problems. In the low probability of uted more causality for failure to self than did persons not expect-
improvement condition, the computer analyses indicated a 2.5% probabil- ing improvement, t(36) ⫽ 4.44, p ⬍ .001. When probability of
ity of future improvement. The experimenter then left the room to ensure improvement was high, highly self-aware persons attributed more
that his presence did not increase self-focus (Carver & Scheier, 1978; causality to self than did mildly self-aware persons, t(36) ⫽ 2.32,
Scheier, Fenigstein, & Buss, 1974). Participants completed a questionnaire p ⬍ .01. But when probability of improvement was low, highly
containing manipulation checks, dependent measures of causal attribution, self-aware persons attributed less causality to self than did mildly
and several filler items. self-aware persons, t(36) ⫽ 2.05, p ⬍ .05. Probability of improve-
ment had no effect in the no-camera conditions; t ⬍ 1.
Dependent Measures Experiment 1 thus supports our predictions; self-awareness and
Causal attributions for failure were measured with two items, “To what probability of improvement interactively determined failure attri-
extent was your performance on the test caused by factors associated with butions. When self-awareness was high, people who expected
yourself?” and “To what extent was your performance on the test caused by improvement attributed failure internally, and people who did not
factors associated with the external environment?” Each question was expect improvement attributed failure externally. Probability of
answered on a 7-point scale anchored by not at all and very much. improvement had no impact when self-awareness was low, pre-
To check the probability of improvement manipulation, we asked par- sumably because self-evaluative concerns are minimal when peo-
ticipants, “To what extent can you improve your performance on the task ple are not comparing self against standards.
in the future?” answered on a similar 7-point scale.

Experiment 2
Results and Discussion
In Experiment 2 we addressed two issues. First, we wanted to
Manipulation Check replicate our findings using different measures of attribution. In
A 2 (camera/no camera) ⫻ 2 (high/low probability of improve- Experiment 2, participants first made attributions for failure using
ment) analysis of variance (ANOVA) conducted on the improve- the four dimensions of ability, effort, task difficulty, and luck. This
ment manipulation check yielded a sole main effect for probability fourfold typology allows a more differentiated look at internal
of improvement, F(1, 36) ⫽ 91.07, p ⬍ .001. As expected, people (ability and effort) and external (task difficulty and luck) attribu-
in the high probability of improvement condition expressed a tions. We then measured attributions with the two-item internal–
greater likelihood of improving in the future relative to people in external index used in Experiment 1. Second, we wanted to ex-
the low probability of improvement condition. plore the consequences of failure attributions for state self-esteem.
Perceiving the self as the cause of failure should decrease state
self-esteem. McFarland and Ross (1982) demonstrated this effect
Attributions for Failure
by directly manipulating failure attributions; they simply told
We created a single attribution score for each person by sub- people that either internal or external factors caused their failure.
tracting external attributions from internal attributions. The pattern As expected, persons attributing failure internally expressed lower
is presented in Figure 1. An ANOVA revealed a significant main state self-esteem than did persons attributing failure externally.
effect for probability of improvement, F(1, 36) ⫽ 10.16, p ⬍ .01, These effects should hold when failure attributions occur in a more
 SELF-AWARENESS AND ATTRIBUTION 53

spontaneous fashion. Thus, our model predicts that highly self- As a check on the probability of improvement manipulation, participants
focused people expecting improvement will attribute failure to self were asked, “To what extent can you improve your performance on the task
and, as a result, experience lower state self-esteem. Highly self- in the future?” answered on a similar 7-point scale.
focused persons not expecting improvement will attribute failure
externally and, as a result, experience no change in state self- Results
esteem. This effect would be inconsistent with explanations for the
self-serving bias based on invariant motives to enhance or protect Manipulation Check
self-esteem (e.g., Snyder et al., 1978). A 2 (camera/no camera) ⫻ 2 (high/low probability of improve-
ment) ANOVA conducted on the manipulation check revealed a
Method sole main effect for probability of improvement, F(1, 36) ⫽ 48.5,
p ⬍ .001. As expected, people in the high probability of improve-
Participants ment conditions perceived a greater likelihood of improvement
than did people in the low probability of improvement conditions.
Participants were 40 male introductory psychology students who volun-
teered to participate for extra course credit. Each was randomly assigned to
condition in a 2 (camera/no camera) ⫻ 2 (high/low probability of improve- Attributing Failure to Ability, Effort, Task Difficulty,
ment) factorial design. and Luck
Ability. Failure attributions to ability are presented in Table 1.
Procedure A 2 ⫻ 2 ANOVA revealed a significant main effect for probability
of improvement, F(1, 36) ⫽ 10.56, p ⬍ .01, qualified by a
The procedure was identical to Experiment 1. Participants worked on a
significant interaction, F(1, 36) ⫽ 12.5, p ⬍ .01. When self-
task, received failure feedback, and expected either a high or low proba-
bility of improving. awareness was high, persons expecting improvement attributed
more causality to ability than did persons not expecting improve-
ment, t(36) ⫽ 4.85, p ⬍ .01. When probability of improvement
Dependent Measures was high, highly self-aware persons made more attribution to
ability than did mildly self-aware persons, t(36) ⫽ 2.97, p ⬍ .01.
Participants first completed a state self-esteem scale (McFarland & Ross,
1982) consisting of 16 adjectives (e.g., pride, worthless, shame). We used
But when probability of improvement was low, highly self-aware
this scale because it was specifically designed to measure the effects of persons attributed less causality to ability than did mildly self-
internal and external attributions on state self-esteem; McFarland and Ross aware persons, t(36) ⫽ 2.08, p ⬍ .05. Probability of improvement
reported validation evidence. It is thus more appropriate for our purposes had no effect in the no camera conditions, t ⬍ 1.
than more general self-esteem measures (e.g., Heatherton & Polivy, 1991). Effort. Failure attributions to effort are presented in Table 1. A
Two separate causal attribution measures were obtained. The first 2 ⫻ 2 ANOVA revealed a significant main effect for probability of
measure consisted of four items: “To what extent was your performance improvement, F(1, 36) ⫽ 7.24, p ⬍ .05, qualified by a significant
on the test caused by your ability?” “To what extent was your perfor- interaction, F(1, 36) ⫽ 12.76, p ⬍ .01. When self-awareness was
mance on the test caused by your effort?” “To what extent was your high, persons expecting improvement attributed more causality to
performance on the test caused by the difficulty of the test?” and “To what
effort than did persons not expecting improvement, t(36) ⫽ 4.43,
extent was your performance on the test caused by luck?” The second
attribution measure, as in Experiment 1, consisted of two items: “To what
p ⬍ .01. When probability of improvement was high, highly
extent was your performance on the test caused by factors associated with self-aware persons attributed more to effort than did mildly self-
yourself?” and “To what extent was your performance on the test caused by aware persons, t(36) ⫽ 2.9, p ⬍ .01. But when probability of
factors associated with the external environment?” Each question was improvement was low, highly self-aware persons attributed less
answered on a 7-point scale anchored by not at all and very much. causality to effort than did mildly self-aware persons, t(36) ⫽ 2.15,

Table 1
Attributions to Ability, Effort, Task Difficulty, and Luck: Experiment 2

Probability of improvement

High Low

Variable Ability Effort Task Luck Ability Effort Task Luck

Camera
M 5.82 5.43 1.64 2.36 1.55 1.44 6.24 4.84
SD 1.86 1.95 2.12 1.85 2.00 1.89 1.95 1.64
No camera
M 3.20 2.82 4.05 4.32 3.38 3.38 3.38 3.46
SD 1.97 1.85 2.01 1.69 1.70 1.96 2.07 1.77

Note. n ⫽ 10 per cell. As means increase, attribution to the given dimension increases. Scale values range
from 1 to 7.
54 DUVAL AND SILVIA

p ⬍ .05. Probability of improvement had no effect in the no of improvement was high, highly self-aware persons attributed
camera conditions, t ⬍ 1. more causality to self than did mildly self-aware persons,
Task difficulty. Attributions to task difficulty are shown in t(36) ⫽ 2.25, p ⬍ .05. But when probability of improvement was
Table 1. A 2 ⫻ 2 ANOVA revealed a significant probability of low, highly self-aware persons attributed less causality to self than
improvement main effect, F(1, 36) ⫽ 10.72, p ⬍ .01, qualified by did mildly self-aware persons, t(36) ⫽ 2.08, p ⬍ .05. Probability
a significant interaction, F(1, 36) ⫽ 13.10, p ⬍ .01. When self- of improvement had no effect in the no camera conditions, t ⬍ 1.
awareness was high, persons expecting improvement attributed
less causality to task difficulty than did persons not expecting State Self-Esteem
improvement, t(36) ⫽ 4.76, p ⬍ .01. When probability of improve-
ment was high, highly self-aware persons attributed less to task A 2 ⫻ 2 ANOVA conducted on the state self-esteem data found
difficulty than did mildly self-aware persons, t(36) ⫽ 2.51, p ⬍ a significant main effect for probability of improvement, F(1,
.05. But when probability of improvement was low, highly self- 36) ⫽ 6.37, p ⬍ .01, and a significant interaction, F(1, 36) ⫽
aware persons attributed more causality to task difficulty than did 11.41, p ⬍ .01. The pattern is shown in Figure 3. When self-
mildly self-aware persons, t(36) ⫽ 2.52, p ⬍ .05. Probability of awareness was high, persons expecting improvement reported
improvement had no effect in the no camera conditions, t ⬍ 1. lower state self-esteem than did persons not expecting improve-
Luck. Attributions to luck are displayed in Table 1. A 2 ⫻ 2 ment, t(36) ⫽ 4.44, p ⬍ .01. When probability of improvement
ANOVA revealed a significant interaction, F(1, 36) ⫽ 8.33, p ⬍ was high, highly self-aware persons reported lower self-esteem
.01. When self-awareness was high, persons expecting improve- than did mildly self-aware persons, t(36) ⫽ 3.06, p ⬍ .01. No other
ment attributed less causality to luck than did persons not expect- differences were found, t ⬍ 1.
ing improvement, t(36) ⫽ 3.02, p ⬍ .01. When probability of
improvement was high, highly self-aware persons attributed less to Mediational Analyses
luck than did mildly self-aware persons, t(36) ⫽ 2.39, p ⬍ .05.
When probability of improvement was low, no significant differ- We expected that the effects of probability of improvement on
ence was found between the high and low self-awareness condi- self-esteem were mediated by attributions for failure. To assess
tions, t(36) ⫽ 1.68, p ⬎ .10. Probability of improvement had no mediation indirectly, we conducted a series of regression analyses
effects in the no camera conditions, t ⬍ 1. on the responses of persons in the high self-awareness conditions
(Baron & Kenny, 1986). Attribution was quantified by a single
Attributing Failure to Internal and External Factors attribution score [(ability ⫹ effort) ⫺ (task difficulty ⫹ luck)]
derived from the fourfold typology. Probability of improvement
We created a single attribution score for each person by sub- significantly predicted both self-esteem (␤ ⫽ .632, p ⬍ .003) and
tracting external attributions from internal attributions. The pattern failure attributions (␤ ⫽ ⫺.917, p ⬍ .001). When probability of
is presented in Figure 2. An ANOVA revealed a significant prob- improvement and attributions for failure were considered simulta-
ability of improvement main effect, F(1, 36) ⫽ 9.28, p ⬍ .01, neously, attributions significantly predicted self-esteem (␤ ⫽
qualified by a significant interaction, F(1, 36) ⫽ 11.16, p ⬍ .01. ⫺1.518, p ⬍ .001), and probability of improvement had a signif-
When self-awareness was high, persons expecting improvement icant direct effect (␤ ⫽ ⫺.761, p ⬍ .02). We will defer interpreting
attributed more causality for failure to self than did persons not this latter effect until Experiment 3, given our small sample size
expecting improvement, t(36) ⫽ 4.52, p ⬍ .01. When probability (n ⫽ 20) for this type of analysis (Cohen & Cohen, 1983, p. 59;
Pedhazur, 1982).
To assess mediation directly, we conducted Baron and Kenny’s
(1986) modified version of the Sobel (1982) test (see Kenny,
Kashy, & Bolger, 1998, p. 260), which provides a direct estimate
of whether a variable mediates between an independent and a
dependent variable. This test also revealed that the effect of im-
provement likelihood on self-esteem was significantly mediated by
failure attributions (Z ⫽ 4.52, p ⬍ .001).

Discussion
Experiment 2 replicated the attribution data obtained in the first
study: Highly self-focused people attributed failure internally
when probability of improvement was high and externally when
probability of improvement was low. Multiple attribution mea-
sures showed this pattern, which increases our confidence in the
effect’s reliability. Indeed, the effect appeared on both internal
(ability and effort) and external measures. This shows that people
were simultaneously seeing the self as less culpable for failure and
external factors as more culpable. Moving beyond replication,
Figure 2. Failure attribution to self as a function of self-awareness and Experiment 2 showed that internal and external attributions have
improvement likelihood: Study 2. corresponding effects on self-esteem. The fact that people attribute
 SELF-AWARENESS AND ATTRIBUTION 55

Experiment 3, however, we added high and low self-awareness/success

conditions. Participants in these groups learned they met the standard for
making quantitative judgments in three dimensions. Experiment 3 also
used the same probability of improvement manipulation; participants were
told they had either a high or low probability of future improvement. No
probability of improvement information, however, was given in the success
conditions. Pretesting indicated that it confused participants about their
actual level of performance.

Dependent Measures
Participants completed the state self-esteem measure used in Experi-
ment 2 (McFarland & Ross, 1982). Attributions for task performance were
measured using two items, “To what extent was your performance on the
test caused by factors associated with yourself?” and “To what extent was
your performance on the test caused by factors associated with the external
environment?” Each question was answered on a 7-point scale anchored by
not at all and very much.
As a check on the probability of improvement manipulation, participants
were asked to indicate “To what extent can you improve your performance
on the task in the future?” on a 7-point scale anchored by not at all and very
Figure 3. State self-esteem as a function of self-awareness and improve-
much. Finally, evaluations of task performance were measured with re-
ment likelihood: Study 2.
sponses to “Please rate the level of your performance on the quantitative
judgment in three dimensions task you just completed” on a similar 7-point
failure internally and experience reduced self-esteem is difficult to
scale.
reconcile with views of the self-serving bias assuming that the
self-esteem motive dominates attribution processes (Snyder et al.,
1978) or the possibility that expecting improvement leads people Results
to discount or brush off failure. Such findings are, however, fully Manipulation Checks
consistent with our dual systems model.
We should note that this study also casts doubt on a possible Failure feedback. A 2 (camera/no camera) ⫻ 3 (success/
criticism of Experiment 1. One might argue that the probability of failure– can improve/failure– can’t improve) ANOVA conducted
improvement manipulation implied that failure was due to an on responses to “Please rate the level of your performance on the
unstable factor such as effort. Because Experiment 2 found internal quantitative judgment in three dimensions task you just com-
failure attributions on Experiment 1’s internal– external index as pleted” revealed a single main effect for type of feedback, F(2,
well as on measures of both ability and effort, we view this 114) ⫽ 104.07, p ⬍ .001. Means are presented in Table 2. When
possibility as unlikely. self-awareness was high, people who succeeded rated their perfor-
mance more positively than did people who failed and expected
Experiment 3 improvement, t(114) ⫽ 5.45, p ⬍ .001, and people who failed and
did not expect improvement, t(114) ⫽ 5.71, p ⬍ .001. The same
The first two experiments have shown how self-awareness and pattern was found when self-awareness was low. People who
probability of improvement interactively determine failure attribu- succeeded viewed their performance more positively than did
tions. But our position also makes predictions about success attri- people who failed and expected improvement, t(114) ⫽ 4.94, p ⬍
butions. As noted earlier, when self is a plausible cause for .001, and people who failed and did not expect improvement,
success, an internal attribution will simultaneously enable a posi-
tive self-evaluation and an accurate causal appraisal. Self-
awareness should thus increase the tendency to attribute success
internally to self. To test this prediction, we added conditions in Table 2
which people were told they had achieved the experiment’s Perceived Quality of Task Performance: Experiment 3
standard.
Performance feedback
Method Failure: Improvement
Participants Variable Success Likely Unlikely
Participants were 120 male undergraduates who volunteered to partici-
Camera
pate for extra course credit. Participants were randomly assigned to con-
M 6.00 1.80 1.60
dition in a 2 (camera/no camera) ⫻ 3 (success/failure– can improve/ SD 2.25 2.52 2.30
failure– can’t improve) factorial design. No camera
M 5.70 1.90 2.30
Procedure SD 2.36 2.44 2.73

The procedure was almost identical to Experiments 1 and 2. We led Note. n ⫽ 20 per cell. Higher numbers indicate higher perceived quality
some participants to believe they were discrepant from the standard. In of performance.
56 DUVAL AND SILVIA

t(114) ⫽ 4.42, p ⬍ .001. These findings suggest that the success–

failure manipulation was successful.
Probability of improvement manipulation. A 2 ⫻ 3 ANOVA
conducted on the probability of improvement manipulation check
revealed a sole main effect for probability of improvement, F(2,
114) ⫽ 27.14, p ⬍ .001. Means are presented in Table 3. No
significant differences between persons who succeeded and per-
sons who failed but could improve were found, regardless of the
level of self-awareness. However, participants in both of these
conditions perceived their chances of improving as significantly
higher relative to persons who failed but could not improve. This
pattern appeared in the high self-awareness (all ts ⬎ 4.8) and low
self-awareness (all ts ⬎ 5.5) conditions. These findings suggest
that the probability of improvement manipulation was successful.
Why persons in the success conditions also perceived their prob-
ability of improvement as high is unclear, although it may reflect
a mood-congruent judgment effect (Gendolla, 2000; Gendolla,
Abele, & Krüsken, 2001).

Figure 4. Failure attribution to self as a function of self-awareness and

Attributions for Performance
success or failure feedback: Study 3.
A 2 ⫻ 3 ANOVA performed on attributions for task perfor-
mance revealed a main effect for feedback type, F(2, 114) ⫽
19.86, p ⬍ .001, qualified by a significant interaction, F(2, 114) ⫽ failed but expected improvement, t(114) ⫽ 1.85, p ⬍ .05 (one-
7.82, p ⬍ .01. The pattern is displayed in Figure 4. tailed), and people who failed but did not expect improvement,
Failure attributions. When self-awareness was high, people t(114) ⫽ 6.25, p ⬍ .001. When self-awareness was low, the only
expecting improvement attributed more failure to self than did difference found was significantly greater self-attribution by peo-
persons not expecting improvement, t(114) ⫽ 4.43, p ⬍ .001. ple who succeeded relative to people who failed and could im-
When probability of improvement was high, highly self-aware prove, t(114) ⫽ 2.24, p ⬍ .01.
persons attributed more failure to self than did mildly self-aware
persons, t(114) ⫽ 2.43, p ⬍ .025. But when probability of im- State Self-Esteem
provement was low, highly self-aware persons attributed less fail-
A 2 ⫻ 3 ANOVA performed on the state self-esteem scale revealed
ure to self than did mildly self-aware persons, t(114) ⫽ 2.61, p ⬍
a significant main effect for the type of feedback, F(2, 114) ⫽ 6.37,
.02. Probability of improvement had no effects when self-
p ⬍ .02, qualified by a significant interaction, F(2, 114) ⫽ 7.71, p ⬍
awareness was low t ⬍ 1. The pattern of results in these four cells
.01. The pattern is shown in Figure 5. When people succeeded, high
replicates the findings obtained in Experiments 1 and 2.
self-awareness increased self-esteem, t(114) ⫽ 3.14, p ⬍ .005. When
Success attributions. Highly self-aware people attributed more
success to self than did mildly self-aware persons, t(114) ⫽ 2.01,
p ⬍ .05. This replicates past experiments that found that height-
ened self-awareness increases self-attribution of success (Federoff
& Harvey, 1976). When self-awareness was high, people who
succeeded attributed more causality to self than did people who

Table 3
Perceived Probability of Improvement: Experiment 3

Performance feedback

Failure: Improvement

Variable Success Likely Unlikely

Camera
M 4.60 5.40 2.30
SD 1.70 1.42 1.76
No camera
M 4.50 5.60 1.90
SD 1.38 1.18 1.51

Note. n ⫽ 20 per cell. Higher numbers indicate higher perceived proba- Figure 5. State self-esteen as a function of self-awareness and success or
bility of improvement. failure feedback: Study 3.
 SELF-AWARENESS AND ATTRIBUTION 57

people failed but could improve, high self-awareness decreased self- Experiment 3, however, did not find a significant unmediated path;
esteem, t(114) ⫽ 2.41, p ⬍ .025. Yet when people failed but could not failure attributions fully mediated the effects of improvement
improve, self-awareness had no effect on self-esteem, t ⬍ 1. beliefs on state self-esteem. We are inclined to place more confi-
When self-awareness was high, people who failed but expected dence in Experiment 3’s finding. Experiment 2’s small sample size
improvement had lower self-esteem than did both people who (n ⫽ 20) is not as large as one would like for multiple regression
failed but did not expect improvement, t(114) ⫽ 3.48, p ⬍ .001, analyses (Cohen & Cohen, 1983). Experiment 3, in contrast, had
and people who succeeded, t(114) ⫽ 5.12, p ⬍ .001. The success twice as many participants and thus more adequate power for
condition did not differ from the group that failed and did not testing mediational hypotheses. Nonetheless, it seems reasonable
expect improvement, t(114) ⫽ 1.63, ns. No differences were found that additional, unmeasured mediators might be relevant. Our
among the low self-awareness conditions, all ts ⬍ 1. theory would expect, for instance, that a high probability of im-
provement might influence the importance of the discrepancy,
Mediational Analyses which could also mediate its effects on state self-esteem. Either
way, both experiments found that attributions significantly medi-
As in Experiment 2, we expected that the effects of probability ated the effects of improvement beliefs on state self-esteem; it
of improvement on self-esteem were mediated by attributions for remains for future research to identify other possible mediators.
success and failure. The success and failure conditions were ana- We should note here that Experiment 3, as did the first two
lyzed separately because no improvement information was given studies, involved only male participants. The composition of the
in the success conditions. Within the success conditions, attribu- participant pool at the time made it difficult to obtain equal
tions and self-esteem were significantly correlated. As expected, numbers of men and women. We believe, though, that this does not
increased self-attribution of success was positively correlated with constrain our interpretations very much. Although men and women
higher self-esteem (r ⫽ .78, p ⬍ .001). surely differ in some of their internalized standards, self-awareness
Within the failure conditions, we conducted a series of regres- research has never found gender differences in the theory’s basic
sion analyses to assess mediation indirectly (Baron & Kenny, processes and mechanisms (Duval & Silvia, 2001). An extension
1986) on the basis of the responses of persons in the high self- of this work using almost all female participants also replicated the
awareness conditions. Probability of improvement significantly effects of self-awareness and improvement likelihood on failure
predicted both self-esteem (␤ ⫽ .629, p ⬍ .001) and failure attribution (Silvia & Duval, 2001b).
attributions (␤ ⫽ ⫺.704, p ⬍ .001). Yet when probability of
improvement and failure attributions were considered simulta- General Discussion
neously, probability of improvement no longer predicted self-
esteem significantly (␤ ⫽ .08, ns). Attributions, however, did Reviewers of the self-serving bias literature often note that the
significantly predict self-esteem (␤ ⫽ ⫺.781, p ⬍ .001). This effect is inconsistent (Campbell & Sedikides, 1999; Zuckerman,
pattern suggests, as expected, that effects of improvement likeli- 1979). Although internal success attributions are consistently
hood on self-esteem were due to attributions for failure. found (Miller & Ross, 1975), internal and external failure attribu-
To assess mediation directly, we conducted Baron and Kenny’s tions are both common findings in the empirical literature. The
(1986) modified version of the Sobel (1982) test (see Kenny et al., presence of inconsistent findings usually suggests that uncon-
1998, p. 260). This test also revealed that the effect of improve- trolled moderators are operating, and highlights the need for a new
ment likelihood on self-esteem was significantly mediated by model capable of explaining the diverging results. We have pro-
failure attributions (Z ⫽ 4.32, p ⬍ .001). posed a model including two moderators, level of self-awareness
and perceived probability of improvement, that might help resolve
Discussion some of the empirical disparities.
Our model assumes that motives associated with two systems, a
Experiment 3 tested our predictions for success attributions. As system comparing self against standards of correctness and a
expected, increasing self-awareness increased the tendency to at- causal attribution system, jointly determine success and failure
tribute success internally. Attributing success to self in turn in- attributions. The SSCS prefers identity between self and standards
creased self-esteem. Experiment 3 also replicated the first two (Duval & Silvia, 2001; Duval & Wicklund, 1972), whereas the
experiments. As before, highly self-focused people attributed fail- attribution system prefers events to be attributed to a plausible
ure internally when their probability of improvement was high, and cause (Heider, 1958). These motives are harmonious when self is
externally when their probability of improvement was low. These a plausible cause for success. Attributing success to self would
attributions had corresponding effects on state self-esteem, thus promote a positive self-evaluation, and the event would be con-
replicating Experiment 2. The manipulation check of perceived nected with a plausible cause. An attribution of success to self
performance eliminated another possible criticism of this proce- should occur because it simultaneously furthers the goals of both
dure. One might argue that improvement information influenced systems. Experiment 3 supported this hypothesis. People who had
the perception of performance. Perhaps people who expected to met the performance standard made greater internal attributions
improve did not perceive their failure as particularly severe. We than did persons who had failed; this effect was enhanced when
view this as unlikely because the high and low probability of self-focus was high.
improvement conditions rated their performance equally nega- But the systems’ goals conflict when self is a plausible cause for
tively (see Table 2). failure. In this situation, we expected that the perceived probability
As noted earlier, Experiment 2 found a significant unmediated of improving poor performance determines failure attributions;
path between probability of improvement and state self-esteem. past research shows that improvement beliefs moderate many
58 DUVAL AND SILVIA

forms of self-enhancing activity (Dunning, 1995; Duval & Duval, bility of future improvement. Other studies induce failure by
1983, 1987; Hong et al., 1999). When highly self-focused people giving false personality feedback. The personality traits that are
feel that failure can be rapidly remedied, they will attribute failure used, such as social perceptiveness (e.g., Miller, 1976), are prob-
to self; when the likelihood of improvement seems low, however, ably interpreted by the participant as being stable and unchange-
failure will be attributed externally. All three experiments sup- able self-aspects. This will also reduce a person’s perceived prob-
ported this hypothesis. Participants who believed they could im- ability of improving.
prove their failing performance attributed more causality to self Duval and Duval (1987) argued that studies finding internal
when self-focus was high. Conversely, self-focus enhanced the failure attributions used procedures that, perhaps accidentally,
tendency to externalize failure when performance could not be produced the perception that failure could be improved. Weary et
improved. al. (1982), for example, asked participants to present a therapy
Finally, Experiments 2 and 3 showed the consequences of outline to individuals having mild phobias; some participants were
success and failure attributions for self-esteem. Mediation analyses led to believe that their performance was a failure. Approximately
found that internal success attributions increased self-esteem and one half of these persons were then asked to return in one week
internal failure attributions decreased self-esteem. Nondefensive- and re-administer the same therapy outline to their client; others
ness thus comes at a cost. Identifying times when people accept were not led to believe they would have another opportunity. From
responsibility for failure and suffer a loss of self-esteem contra- our approach, people who expected to re-administer therapy would
dicts egotism models of causal attribution (Snyder et al., 1978; at least have the opportunity to potentially improve and thus
Zuckerman, 1979) and the notion that the self-enhancement mo- perceived a reasonable probability of improving. As in the present
tives dominates the self-assessment motive (Sedikides, 1993). But studies, this promoted attributed failure to self. People who did not
this should not be surprising—internal failure attributions and expect a second opportunity essentially felt unable to improve and,
resulting self-esteem losses occur frequently in everyday life. as predicted by our theory, attributed failure to external factors.
Furthermore, as Heider (1958) argued, both self-evaluative and Ross et al.’s (1974) findings yield to a similar analysis. Expe-
informational motives influence judgment and perception. If the rienced schoolteachers and undergraduates with no training in
self-assessment motive is independent from the self-evaluation education were asked to teach an 11-year-old boy how to spell.
motive, instead of subordinate to it (Sedikides, 1993), then there One half of the participants were then told that the young boy had
should be some conditions promoting accurate informational goals done poorly on a subsequent spelling task. The schoolteachers
over self-enhancement goals. attributed causality for their failing performance to themselves, but
the novice undergraduates attributed failure externally. The pro-
Reconsidering the Self-Serving Bias Literature fessional teachers certainly anticipated having future opportunities
to improve their teaching skills and thus attributed failure to self.
How well does our model account for the self-serving bias The untrained undergraduates, on the other hand, probably did not
literature? First, the present model predicts greater self-attribution expect future improvement opportunities and thus did not expect
for success; this prediction is compatible with reviews indicating future improvement.
that self-enhancing success attributions are consistently found
(Miller & Ross, 1975). In contrast, considerable variability exists
Relations to Other Perspectives
with regard to failure attributions. Our model suggests several
reasons why both internal and external failure attributions are Perceived Probability of Improvement and Self-Efficacy
found. Many studies have used procedures that were self-
irrelevant. In a recent meta-analysis, Campbell and Sedikides Self-efficacy is the belief that one can enact a specific behavior
(1999) demonstrated that increased self-threat amplifies the self- or set of behaviors; outcome efficacy is the belief that certain
serving bias. When there was little self-threat, non-self-serving behaviors exist that if enacted will be effective in reaching a
attributions were found. In the context of our model, minimal desired goal (Bandura, 1977). Perceptions of the probability of
self-threat translates into circumstances having few implications improving incorporate both elements. Inasmuch as people believe
for how self relates to important standards. Because the costs to the they can improve, they should think that behaviors exist that if
SSCS of an internal failure attribution are low, the attribution enacted would elevate performance, as well as that they can
system will attribute failure to self. effectively carry out those actions. Inasmuch as people believe
A second reason why past studies have yielded contradictory they cannot improve, they think that no means of reducing self-
findings for failure attributions involves the confounding of failure standard discrepancies are available or that they have little capac-
feedback with improvement feedback. Zuckerman (1979) con- ity to act in ways thought to be efficacious. The effects of prob-
cluded that the achievement paradigm, in which participants work ability of improvement on attribution can be understood as
on a task and receive success or failure feedback, is the only reflecting the impact of particular combinations of self and out-
paradigm consistently finding external failure attributions. This come efficacy levels on causal attribution.
paradigm is probably so effective because popular methods of
inducing failure also engender a low likelihood of improvement. Trait Malleability, Attribution, and Self-Esteem
Most studies use novel tasks such as timed anagram unscrambling,
Mednick’s (1962) Remote Associates Task, or our own three- Earlier we briefly discussed how Dweck’s research on self-
dimensional mental rotation task, to make failure feedback plau- theories intersects with our model of self-serving attribution. It is
sible. Participants naturally do not expect ever to encounter and tempting to assume that incremental theorists generally feel able to
work on these esoteric tasks and will thus perceive a low proba- improve after failure because their implicit theory of self views
 SELF-AWARENESS AND ATTRIBUTION 59

personal attributes as malleable (Dweck et al., 1995). Perceived Improvement Beliefs and Counterfactual Thought
probability of improvement then moderates attribution for failure.
Consistent with this notion, incremental theorists assigned slightly Counterfactual thought (Kahneman & Tversky, 1982) is also
more causal influence for failure to both ability and effort (a relevant to our model, inasmuch as the perceived ability to change
combined percentage of 57.7) than to task difficulty and luck (a an event influences self-blame. People asked to imagine experi-
combined percentage of 42.3; Hong et al., 1999). Pursuing this line encing a negative outcome that could have been changed were
of thought, entity theorists might be people doubting their capacity more likely to say they would blame self and experience negative
to improve because they construe traits as being relatively un- affect than were people who thought that a change in the negative
changeable. Believing that improvement is improbable then mod- outcome was unlikely (Boninger, Gleicher, & Strathman, 1994).
erates their attributions for failure. Yet if entity theorists actually From our perspective, the “can change” versus “cannot change”
conditions correspond to high and low probability of improvement
represent a low probability of improvement group, they fail to
manipulations. As would be expected from the dual systems ap-
show any preference for external over internal attributions (a
proach, people who imagined a malleable outcome attributed more
combined 48.3% for task difficulty ⫹ luck, and 51.7% for abil-
causality for the unpleasant event to self (assuming that self-blame
ity ⫹ effort). Entity theorists also tend to overattribute failure to
represents attribution of causality to self) and reported experienc-
ability (33.9%) relative to effort (17.8%), relative to our finding
ing higher levels of negative affect in the moment than did people
that manipulated expectancies led to equal attribution to ability and
who imagined an unchangeable outcome. This pattern of attribu-
effort (Experiment 2).
tion and affect was obtained only for people who, because of either
Given this pattern, we are disinclined to argue that individual
situational or dispositional factors, tended to focus on present
differences in the belief that personal attributes are changeable
rather than future behaviors. This result is also comprehensible
reflect cognitions that are psychologically equivalent to high and
from the dual systems perspective. On the basis of self-awareness
low probability of improvement beliefs. Indeed, to take this posi- theory (Duval & Wicklund, 1972) and research on deindividuation
tion would be tantamount to proposing that incremental and entity (Diener, 1979), people focused on current activities were probably
theorists are simply people who, for whatever reason, have differ- higher in self-focus than were people focused on the future. If high
ing estimates of self and outcome efficacy, an interpretation that self-awareness is required for these effects, as our experiments
flies in the face of ancillary evidence collected by Dweck and her suggest, then heightened self-blame and experienced negative af-
colleagues (e.g., Dweck et al., 1995; Hong et al., 1999) suggesting fect would only be expected when people focused on current
that the differing self-theories reflect distinct and complex cogni- activities.
tive frameworks and goal orientations.
Nonetheless, our results can inform some of findings in the
self-theories literature. The effects of internal failure attributions Does Self-Awareness Always Moderate Success and
on self-esteem, for example, are consistent with the behavior of Failure Attributions?
incremental and entity theorists. Incremental theorists should ex- The dual systems approach is based on the case in which the
perience lower self-esteem after an internal failure attribution causal attribution system intersects with the SSCS. These systems
(Experiments 2 and 3). But because they attributed causality to can intersect only inasmuch as they are operating simultaneously.
personal attributes seen as malleable, they should also expect A lot of research shows that a person’s level of self-focus deter-
future improvement and restoration of self-esteem. In fact, Hong et mines the degree of self–standard comparison (Duval & Silvia,
al. (1999) found that incremental theorists are more interested in 2001). When self-focus is low, people are not very concerned with
pursuing remedial activity following failure. When trait malleabil- meeting their standards (Silvia & Gendolla, 2001). This position
ity is manipulated (Dunning, 1995), persons perceiving abilities to clearly suggests that self-focus moderates attribution for events
be malleable show interest in tasks on which they would presum- that could potentially influence congruity between self and stan-
ably get positive feedback. dards. Indeed, the present results and related research (Silvia &
Entity theorists should also experience lowered self-esteem fol- Duval, 2001b) support this prediction. This suggests that partici-
lowing attribution for failure to ability. Yet they should not expect pants in previous research showing a self-serving bias were rela-
improvement and state self-esteem restoration because they see the tively high in self-focus. Cases in which no self-serving biases are
attribute as unchangeable. Both lowered self-esteem and perceiv- found, which are more numerous than might be expected (Camp-
ing events as uncontrollable are compatible with the suggestion bell & Sedikides, 1999; Zuckerman, 1979), could involve situa-
that entity theorists sometimes show learned helplessness after tions associated with low levels of self-awareness. Indeed, some of
failure (e.g., Dweck et al., 1995; Hong et al., 1999). In fact, entity the circumstances that induce self-focus—self-reflection, being
theorists actively avoid remedial situations and show little interest observed and evaluated, and social interaction (Duval & Silvia,
in challenging tasks that might improve their abilities (Hong et al., 2001)—are common elements in many social psychological ex-
1999). Furthermore, when perceived trait malleability is manipu- periments, so it is not unreasonable to suggest that self-awareness
lated (Dunning, 1995), people perceiving the ability as unmal- is an uncontrolled moderator in this literature. Further tests of this
leable avoid tasks for which performance feedback, and thus a tentative conclusion await future research.
possible reminder of personal failings, is likely. The relationship
between attribution and self-esteem found in the present experi- Coping, Control, and Attribution
ments is thus consistent with research on dispositional and manip-
ulated beliefs about the malleability of traits (Dunning, 1995; Tennen, Affleck, and Gershman (1986) suggested that self-
Hong et al., 1999). attribution for negative events— or more specifically, self-
60 DUVAL AND SILVIA

blame—increases self-efficacy. Increases in self-efficacy then en- Campbell, W. K., & Sedikides, C. (1999). Self-threat magnifies the self-
hance problem-focused coping. A path analysis found that serving bias: A meta-analytic integration. Review of General Psychol-
mothers’ self-blame for their children’s perinatal complications ogy, 3, 23– 43.
appeared to mediate level of self-efficacy; self-efficacy then influ- Carver, C. S., & Scheier, M. F. (1978). Self-focusing effects of disposi-
enced behaviors classified as effective coping strategies. Our re- tional self-consciousness, mirror presence, and audience presence. Jour-
nal of Personality and Social Psychology, 36, 324 –332.
search strongly suggests that the causal relationship between self-
Carver, C. S., & Scheier, M. F. (1998). On the self-regulation of behavior.
efficacy and self-blame is in the opposite direction. Perceived New York: Cambridge University.
capacity to improve, a variable that includes self-efficacy as a Cohen, J., & Cohen, P. (1983). Applied multiple regression/correlation
component, moderates self-attribution. Lalwani and Duval (2000), analysis for the behavioral sciences (2nd. ed). Hillsdale, NJ: Erlbaum.
for example, found that perceiving resources to be sufficient to Dana, E. R., Lalwani, N., & Duval, T. S. (1997). Objective self-awareness
deal with a potential problem (high self-efficacy) increases self- and focus of attention following awareness of self-standard discrepan-
attribution of responsibility for coping with the problem. Other cies: Changing self or changing standards of correctness. Journal of
research found that causal attribution, at least to effort, mediates Social and Clinical Psychology, 16, 359 –380.
interest in activities designed to remedy failure (Hong et al., 1999). Davis, D., & Brock, T. C. (1975). Use of first-person pronouns as a
Taken together, this research suggests, on the one hand, that the function of increased objective self-awareness and performance feed-
causal path goes from perceived probability of improvement to back. Journal of Experimental Social Psychology, 11, 381–388.
Diener, E. (1979). Deindividuation, self-awareness, and disinhibition.
causal attribution to coping behavior. On the other hand, although
Journal of Personality and Social Psychology, 37, 1160 –1171.
this pattern strikes us as more plausible, causality could be bidi-
Dunning, D. (1995). Trait importance and modifiability as factors influ-
rectional. Self-blame for a negative outcome might represent an encing self-assessment and self-enhancement motives. Personality and
acceptance of personal responsibility for that event or situation. As Social Psychology Bulletin, 21, 1297–1306.
Tennen et al. (1986) suggested, accepting responsibility could, in Dunning, D. (1999). A newer look: Motivated social cognition and the
some sense, empower the individual and lead to efforts designed to schematic representation of social concepts. Psychological Inquiry, 10,
manage the troubled person– environment interaction. Future re- 1–11.
search should try to clarify this issue. Duval, T. S. (1976). Conformity on a visual task as a function of personal
novelty on attitudinal dimensions and being reminded of object status of
self. Journal of Experimental Social Psychology, 12, 87–98.
A Matter of Time Duval, T. S., & Duval, V. H. (1983). Consistency and cognition: A theory
of causal attribution. Hillsdale, NJ: Erlbaum.
The dynamics of the dual systems approach involve processes
Duval, T. S., & Duval, V. H. (1987). Level of perceived coping ability and
unfolding over time. To the extent that individuals believe they can attribution for negative events. Journal of Social and Clinical Psychol-
improve failure, they will attribute causality to self. Although this ogy, 5, 452– 468.
results in lowered state self-esteem at Time 1, it also appears to Duval, T. S., Duval, V. H., & Mulilis, J. P. (1992). Effects of self-focus,
result in action designed to ameliorate the negative situation (Hong discrepancy between self and standard, and outcome expectancy favor-
et al., 1999). If this activity effectively resolves the problem, one ability on the tendency to match self to standard or withdraw. Journal of
would expect internal attributions for these successes. Given the Personality and Social Psychology, 62, 340 –348.
results of Experiments 2 and 3 (cf. McFarland & Ross, 1982), Duval, T. S., & Lalwani, N. (1999). Objective self-awareness and causal
those attributions should eventually increase state self-esteem to attributions for self-standard discrepancies: Changing self or changing
previous levels, or perhaps even result in a net gain. These expe- standards of correctness. Personality and Social Psychology Bulletin, 25,
riences might, in the long term, influence the development of 1220 –1229.
Duval, T. S., & Silvia, P. J. (2001). Self-awareness and causal attribution:
enduring notions of self-efficacy (Bandura, 1977) or theories of
A dual-systems theory. Boston: Kluwer Academic.
trait malleability (Dweck et al., 1995). However, as is always the
Duval, T. S., & Wicklund, R. A. (1972). A theory of objective self-
case with dynamical systems, tests of these projections require awareness. New York: Academic Press.
investigations including time as an independent variable. Dweck, C. S., Chiu, C., & Hong, Y. (1995). Implicit theories and their role
in judgments and reactions: A world from two perspectives, Psychology
Inquiry, 6, 267–285.
References
Federoff, N. A., & Harvey, J. A. (1976). Focus of attention, self-esteem and
Ames, R. (1975). Teachers’ attributions of responsibility: Some unex- attribution of causality. Journal of Research in Personality, 10, 336 –
pected nondefensive effects. Journal of Educational Psychology, 67, 345.
668 – 676. Festinger, L. (1954). A theory of social comparison processes. Human
Bandura, A. (1977). Self-efficacy: Toward a unifying theory of behavioral Relations, 7, 117–140.
change. Psychological Review, 84, 191–215. Gendolla, G. H. E. (2000). On the impact of mood on behavior: An
Baron, R. M., & Kenny, D. A. (1986). The moderator–mediator variable integrative theory and a review. Review of General Psychology, 4,
distinction in social psychological research: Conceptual, strategic, and 378 – 408.
statistical considerations. Journal of Personality and Social Psychol- Gendolla, G. H. E., Abele, A. E., & Krüsken, J. (2001). The informational
ogy, 51, 1173–1182. impact of mood on effort mobilization: A study of cardiovascular and
Boninger, D. S., Gleicher, F., & Strathman, A. (1994). Counterfactual electrodermal responses. Emotion, 1, 12–24.
thinking: From what might have been to what may be. Journal of Heatherton, T. F., & Polivy, J. (1991). Development and validation of a
Personality and Social Psychology, 67, 297–307. scale for measuring state self-esteem. Journal of Personality and Social
Brown, J. D., & Rogers, R. J. (1991). Self-serving attributions: The role of Psychology, 60, 895–910.
physiological arousal. Personality and Social Psychology Bulletin, 17, Heider, F. (1944). Social perception and phenomenal causality. Psycho-
501–506. logical Review, 51, 358 –374.
 SELF-AWARENESS AND ATTRIBUTION 61

Heider, F. (1958). The psychology of interpersonal relations. New York: own self be better. Advances in Experimental Social Psychology, 29,
Wiley. 206 –269.
Hong, Y., Chiu, C., Dweck, C. S., Lin, D. M. S., & Wan, W. (1999). Silvia, P. J., & Duval, T. S. (2001a). Objective self-awareness theory:
Implicit theories, attributions, and coping: A meaning system approach. Recent progress and enduring problems. Personality and Social Psy-
Journal of Personality and Social Psychology, 77, 588 –599. chology Review, 5, 230 –241.
Ickes, W. J., Wicklund, R. A., & Ferris, C. B. (1973). Objective self Silvia, P. J., & Duval, T. S. (2001b). Predicting the interpersonal targets of
awareness and self-esteem. Journal of Experimental Social Psychol- self-serving attributions. Journal of Experimental Social Psychology, 37,
ogy, 9, 202–219. 333–340.
Kahneman, D., & Tversky, A. (1982). The simulation heuristic. In D. Silvia, P. J., & Gendolla, G. H. E. (2001). On introspection and self-
Kahneman, P. Slovic, & A. Tversky (Eds.), Judgment under uncertainty perception: Does self-focused attention enable accurate self-knowledge?
(pp. 201–208). New York: Cambridge University. Review of General Psychology, 5, 241–269.
Kenny, D. A., Kashy, D. A., & Bolger, N. (1998). Data analysis in social Snyder, M. L., Stephan, W. G., & Rosenfield, D. (1976). Egotism and
psychology. In D. T. Gilbert, S. T. Fiske, & G. Lindzey (Eds.), The attribution. Journal of Personality and Social Psychology, 33, 435– 441.
handbook of social psychology (4th ed., Vol. 1, pp. 233–265). Boston: Snyder, M. L., Stephan, W. G., & Rosenfield, D. (1978). Attributional
McGraw-Hill. egotism. In J. H. Harvey, W. J. Ickes, & R. F. Kidd (Eds.), New
Lalwani, N., & Duval, T. S. (2000). The moderating effects of cognitive directions in attribution research (Vol. 2, pp. 91–117). Hillsdale, NJ:
appraisal processes on self-attribution of responsibility. Journal of Ap- Erlbaum.
plied Social Psychology, 30, 2233–2245. Sobel, M. E. (1982). Asymptotic intervals for indirect effects in structural
Lewin, K. (1935). A dynamic theory of personality. New York: McGraw- equation models. In S. Leinhart (Ed.), Sociological methodology 1982
Hill. (pp. 290 –312). San Francisco: Jossey-Bass.
McDonald, P. J. (1980). Reactions to objective self-awareness. Journal of Taylor, S. E., & Brown, J. D. (1988). Illusion and well-being: A social
Research in Personality, 14, 250 –260. psychological perspective on mental health. Psychological Bulletin, 103,
McFarland, C., & Ross, M. (1982). Impact of causal attributions on 193–210.
affective reactions to success and failure. Journal of Personality and Tennen, H., Affleck, G., & Gershman, K. (1986). Self-blame among
Social Psychology, 43, 937–946. parents of infants with perinatal complications: The role of self-
Mednick, S. A. (1962). The associative basis of the creative process. protective motives. Journal of Personality and Social Psychology, 50,
Psychological Review, 26, 220 –232. 690 – 696.
Miller, D. T. (1976). Ego involvement and attributions for success and Treisman, A. M., & Gelade, G. (1980). A feature-integration theory of
failure. Journal of Personality and Social Psychology, 34, 901–906. attention. Cognitive Psychology, 12, 97–136.
Miller, D. T., & Ross, M. (1975). Self-serving biases in the attribution of Trope, Y. (1986). Self-enhancement and self-assessment in achievement
causality: Fact or fiction? Psychological Bulletin, 82, 213–225. behavior. In R. M. Sorrentino & E. T. Higgins (Eds.), Handbook of
Nisbett, R., & Ross, L. (1980). Human inference: Strategies and short- motivation and cognition (Vol. 1, pp. 350 –378). New York: Guilford.
comings of social judgment. Englewood Cliffs, NJ: Prentice-Hall. Trope, Y., & Neter, E. (1994). Reconciling competing motives in self-
Pedhazur, E. J. (1982). Multiple regression in behavioral research (2nd evaluation: The role of self-control in feedback seeking. Journal of
ed.). New York: Holt, Rinehart, & Winston. Personality and Social Psychology, 66, 646 – 657.
Ross, L., Bierbrauer, G., & Polly, S. (1974). Attribution of educational Trope, Y., & Pomerantz, E. M. (1998). Resolving conflicts among self-
outcomes by professional and nonprofessional instructors. Journal of evaluative motives: Positive experiences as a resource for overcoming
Personality and Social Psychology, 29, 609 – 618. defensiveness. Motivation and Emotion, 22, 53–72.
Scheier, M. F., & Carver, C. S. (1983). Self-directed attention and the Weary, G., Harvey, J. H., Schweiger, P., Olson, C. T., Perloff, R., &
comparison of self with standards. Journal of Experimental Social Pritchard, S. (1982). Self-presentation and the moderation of self-serving
Psychology, 19, 205–222. attributional biases. Social Cognition, 1, 140 –159.
Scheier, M. F., Fenigstein, A., & Buss, A. H. (1974). Self-awareness and Wong, P. T. P., & Weiner, B. (1981). When people ask “why” questions
physical aggression. Journal of Experimental Social Psychology, 10, and the heuristics of attributional search. Journal of Personality and
264 –273. Social Psychology, 41, 650 – 663.
Sedikides, C. (1992). Attentional effects on mood are moderated by Zuckerman, M. (1979). Attributions of success and failure revisited, or:
chronic self-conception valence. Personality and Social Psychology The motivational bias is alive and well in attribution theory. Journal of
Bulletin, 18, 580 –584. Personality, 47, 245–287.
Sedikides, C. (1993). Assessment, enhancement, and verification determi-
nants of the self-evaluation process. Journal of Personality and Social
Psychology, 65, 317–338. Received May 19, 2000
Sedikides, C., & Strube, M. J. (1997). Self-evaluation: To thine own self be Revision received May 1, 2001
good, to thine own self be sure, to thine own self be true, and to thine Accepted May 1, 2001 䡲

View publication stats