Professional Documents
Culture Documents
The Double-Edged Sword of Self-Handicapping: Discounting, Augmentation, and The Protection and Enhancement of Self-Esteem
The Double-Edged Sword of Self-Handicapping: Discounting, Augmentation, and The Protection and Enhancement of Self-Esteem
The Double-Edged Sword of Self-Handicapping: Discounting, Augmentation, and The Protection and Enhancement of Self-Esteem
2, 1997
A field study was conducted to test the hypothesis that discounted and
augmented ability self-attributions mediate the interactive effects of claimed
self-handicaps and academic success and failure on self-esteem. College
students were assessed for individual differences in self-handicapping and
self-esteem at the beginning of the term and then completed a checklist of
clamed self-handicaps immediately preceding their first in-class exam. At the
following class, graded exams were returned to the students, who then
completed measures of mood, self-esteem, and performance attributions. High
self-handicappers claimed more excuses prior to the test. Among failing
students, claimed handicaps were associated with greater discounting of ability
attributions and higher self-esteem. Among successful students, claimed
handicaps were associated with augmented ability attributions and enhanced
self-esteem. However, we failed to find support for sex differences in claimed
self-handicapping. The implications of the present research with regard to the
functional utility of self-handicapping behavior are discussed.
le wish to thank David Sanbonmatsu and Brian Tschanz for their assistance with this study.
2Address all correspondence concerning this article to Frederick Rhodewalt, Department of
Psychology, University of Utah, Salt Lake City, Utah 84112.
147
OM6-7339J97JQ6ro-01«m2£<yO C 1997 Plenum Publishing Corporation
148 Feick and Rhodewalt
viduals will withdraw effort or forgo practice when uncertain about their
ability to perform successfully (Hirt, Deppe, & Gordon, 1991; Rhodewalt
& Airfield, 1991; Rhodewalt et al., 1984). They are also more likely than
low self-handicappers (LSH) to cite extenuating circumstances or obstacles
to success prior to performing (Rhodewalt & Hill's 1995; Strube, 1986,
Study 2).
METHOD
Subjects
The subjects were 121 (62 women, 49 men) University of Utah students
recruited from four upper-division undergraduate psychology courses. The
four classes had a total enrollment of 136. Ten subjects did not complete
all portions of the study and were dropped from the analyses. All students
were offered the opportunity to participate in the study or to read and
write a summary of a review chapter for extra course credit. They also had
the opportunity not to participate in either option.
The participants completed questionnaires at three different times:
during the first 2 weeks of the academic quarter, during the class period
of but prior to taking the first exam, and during the class period following
the exam after they received their exam grades.
Time 1 Measures
Time 2 Measure
Tune 3 Measures
Procedure
The study was introduced to the students during the first class meeting
by one of the two authors in each of the four upper-level classes. It was
stated that the study was an investigation of students' thoughts and feelings
about how they approach coursework. It was explained that the issues un-
der study were pertinent to topics to be discussed later in the classes. Stu-
dents were informed that (a) their participation would require completing
questionnaires three times during the first 5 weeks of the term, (b) they
would receive partial course credit for their participation, (c) they could
earn an equal amount of partial credit by reading an article and answering
some questions about it, and (d) their decision to participate in the study
would remain unknown to their instructor.
An envelope containing a cover letter, a statement of informed con-
sent, and the Time 1 questionnaires was distributed to each student by one
of the authors or the instructor. The cover letter explained the purpose of
the study and the requirements for participation. The students were in-
structed to indicate if they chose to participate in the study, to participate
in the optional reading, or to do neither extra credit activity and to sign
the consent form. If they opted to participate in the study, they were in-
structed to complete the questionnaires. All the students, regardless of
whether they were participating in the study, were instructed to place the
consent forms and questionnaires back into their envelopes and drop them
in a locked mailbox outside their classroom by the end of the second week
of the quarter.
Self-Handicapping and Attributions 155
RESULTS
Time 1 Measures
jects' actual test grades and expected test grades were unrelated to any
other Time 1 measure.
Time 2 Measures
The first prediction was that scores on the SHS scale would predict
the number of handicaps students would claim before taking the exam. The
number of items (including "other") on the checklist of extenuating cir-
cumstances was summed to form the measure of claimed handicaps and
entered into a multiple-regression analysis. The final regression model pre-
dicting claimed handicaps included subject sex (men = 0, women = 1),
baseline self-esteem, Self-Handicapping Scale (SHS), and their multiplica-
tive interaction terms. For all regression analyses, continuous variables were
"centered" by subtracting their means so that the main effect terms all had
means of zero (Aiken & West, 1991). The procedure of centering variables
(a) reduces potential problems of multicollinearity often introduced in the
regression analyses where interactions are present and (b) facilitates the
interpretation of regression coefficients.
Only the main effects model was significant for claimed handicaps, R2
= .14, F(3, 197) = 5.28, p < .001, with only SHS significantly and inde-
pendently predicting excuse making (b = .41). This finding replicates past
research (Rhodewalt & Hill, 1995; Strube, 1986). However, contrary to
these earlier studies, level of self-handicapping did not significantly interact
with sex in predicting claimed handicaps.
Time 3 Measures
At the first class meeting following the exam, the subjects received
their grades and immediately completed the Resultant Self-Esteem Scale,
which included subscales for State Self-Esteem, Positive Mood, and Nega-
tive Mood. Students' attributions for their test performance were scored
to form a composite, with higher scores indicating more internal-stable-
global attributions for their grades. Coefficient alpha for the three-item
composite was .66. Postfeedback mood was computed by subtracting the
sum of the Negative Mood scale from the sum of the Positive Mood scale.
Separate multi-regression analyses were conducted for each dependent
variable (ability attributions, postfeedback self-esteem, and mood). Follow-
ing the recommendation of Judd and Kenny (1981), nonsignificant and
theoretically irrelevant interaction terms were trimmed from the model. In-
cluded in the final regression model were baseline self-esteem, SHS, test
outcome, number of claimed handicaps, and six interactions: claimed handi-
Self-Handicapping and Attributions 157
caps by test outcome; SHS by claimed handicaps and by test outcome; and
the interactions of baseline self-esteem with SHS, test outcome, and
claimed handicaps. Main effect terms were entered on the first step, fol-
lowed by the two-way interactions. Preliminary analyses revealed no effect
for sex of student. In addition, the inclusion of three- or four-way higher-
order interactions did not make significant contributions to the predicted
variance of any of the dependent measures.
The test outcome variable reflected self-defined success or failure on
the examination. That is, test outcome was calculated by subtracting sub-
jects' expected grades from their actual grades. Thus, the more positive the
number, the more the outcome exceeded expectations, the more negative
the number, the more the outcome failed to meet expectations.6
Test Performance Attributions. For test attributions a significant inter-
action between claimed handicaps and test outcome was predicted such
that students would offer greater internal attributions for failure the fewer
the number of claimed handicaps. In contrast, they should make more ex-
treme internal attributions for success the greater the number of handicaps.
This prediction was supported in the overall test attribution regression
model [R2 = .30, F(12, 98) = 3.38, p < .005]. The individually significant
predictors were test outcome (b = .45, t = 4.23, p < .001), claimed handi-
caps (b = -.18, t = 1.98, p < .05) and, more importantly, the Test Outcome
x Claimed Handicaps interaction (b = .43, t = 3.95, p < .001). There was
also a significant Test Outcome x Self-Esteem interaction (b = .18, t =
1.99, p < .05).
Interactions were interpreted following the procedure outlined by
Aiken and West (1991). The predicted values for each dependent measure
were calculated by deriving the regression lines for one predictor (claimed
handicaps) at two levels of the second predictor (test outcome). The two
levels of test outcome were failure (-1 standard deviation) and success (+1
standard deviation). The regression slopes from this analysis are displayed
in Fig. la. Tests of the simple slopes (Aiken & West, 1991) indicated that
the slope for failure students was negative (-1.28) and significantly different
from 0 (t = -4.36, p < .001). That is, students' attributions for failure were
less internal the greater the number of handicaps they claimed, a pattern
which clearly supports the discounting prediction. In contrast, the slope of
6The participants reported their performance expectations both at the beginning of the
academic term and, again, at the time they received their test grades. The findings are
reported for the difference between their actual grades and their reported expectations at
the time of test feedback. This value was highly correlated with expectations stated at the
beginning of the class (r = .71). When initial expectations were substituted for expectations
stated at the time of feedback, a virtually identical pattern of significant findings obtained.
We opted to use the time of feedback measure because it reflected the sense of success or
failure the student as feeling at the time the dependent measures were collected.
158 Feick and Rhodewalt
Fig. 1. Simple regression slopes for (a) performance attributions, (b) self-esteem, and (c)
affect as a function of test outcome and level of self-handicapping.
those who met or exceeded their expectations was positive (1.59) and sig-
nificant (t = 2.24, p < .01). Success students offered more extreme internal
Self-Handicapping and Attributions 159
attributions for their success with the more handicaps they claimed. This
pattern supports the augmentation prediction.7
The Test Outcome x Self-Esteem interaction reflects the fact that high
self-esteem participants were more self-serving than lows in their attribu-
tions for failure (high y = 7.7; low y = 19.0). They did not differ in their
attributions for success (high y = 16.1; low y = 16.3).
Self-Esteem. With regard to postfeedback self-esteem, two main effects
were predicted: Self-esteem would be higher the greater the number of
handicaps and higher following success than following failure. This predic-
tion reflects the fact that within failure the self-esteem of high claimed
handicap students should be higher than low claimed handicap students
because the handicaps permit the externalization of failure. Following suc-
cess, high claimed handicap students should experience higher self-esteem
than low claimed handicap students because the success occurred despite
the handicap. Consistent with these predictions, only the main effects
model was significant [R2 = .32, F(5, 105) = 9.91, p < .001]. There were
three individually significant terms: baseline self-esteem (b = .24, t = 2.58,
p < .01), test outcome (b = .48, t = 5.64, p < .001), and claimed handicaps
(b = .38, t = 4.52, p < .001). The simple regression slopes for postfeedback
self-esteem are displayed in Fig. 1b.
The self-esteem of success students (slope = 6.85) and failure students
(slope = 4.71) was positively related to the number of handicaps claimed,
clear evidence for the buffering effects of discounting in the former case
and enhancing effects of augmentation in the latter case.
Mood. Finally, with regard to mood we expected only a main effect
for test outcome. Negative mood was subtracted from positive mood to
produce an overall mood score in order to examine this prediction. Only
the main effects model was significant [R2 = .38, F(5, 105) = 13.24, p
< .001], with only test outcome (b = M,t = 8.07, p < .001) predicting
mood. The simple regression slopes within level of claimed handicaps
displayed in Fig. 1c indicate that postfeedback affective responses were
determined largely by how well the students performed relative to their
expectations.
7Analyses of the open-ended responses the participants gave for their test performance were
also conducted. Two independent raters coded the responses for the extent to which they
attributed test performance to ability, effort, task difficulty, and luck. Interrater reliabilities
for these ratings ranged from .53 to .74. Unfortunately, a significant number of the
participants (n = 11) did not provide open-ended responses, making interpretation difficult.
However, for those that did give open-ended responses, the data showed that test outcome
(b = .26) and number of handicaps (b = .17) significantly predicted ability attributions [R2
= .12, F(1, 103) = 3.51]. Positive outcomes and the claiming of excuses predicted stronger
ability attributions. No other analyses were significant.
160 Feick and Rhodewalt
DISCUSSION
(Rhodewalt & Hill, 1995; Strube, 1986), individual differences in the ten-
dency to self-handicap as indexed by the Self-Handicapping Scale (Jones
& Rhodewalt, 1982) predicted the claiming of handicaps prior to taking
an exam. More importantly, the results provided clear evidence that self-
handicapping buffers self-esteem from experiences of failure and enhances
self-esteem for experiences of success because the handicap affords, re-
spectively, discounted and augmented ability attributions. That is, compared
to students who did not claim handicaps, students who claimed relatively
high numbers of handicaps prior to the exam made more external, specific,
and unstable attributions upon learning of a poor test grade. Upon learning
of a good test performance, these handicapping students made more in-
ternal, stable, and global attributions than did nonself-handicapping stu-
dents. As a consequence of these self-serving performance attributions, high
self-handicapping students experienced more positive self-esteem relative
to low self-handicapping students for both success and failure experiences.
Although several laboratory studies (Rhodewalt et al., 1991) and a
field study (Rhodewalt & Hill, 1995) have provided support for the dis-
counting effects of self-handicapping, evidence for the augmentation effects
of self handicapping has been limited to those with high self-esteem in a
few highly circumscribed laboratory conditions (Rhodewalt et al., 1991;
Tice, 1991). The present results are the first to demonstrate both discount-
ing and augmentation effects in a naturalistic setting. However, the present
study did not find evidence that augmented ability attributions and bol-
stered self-esteem was the exclusive province of high self-esteem individuals
as has been the case in past laboratory research (Rhodewalt et al., 1991;
Tice, 1991). This fact may point to an important difference between the
field and laboratory settings. In the laboratory studies, handicapping was
done in public view of the experimenter. If, as Tice (1991) argues, differ-
ences in self-esteem reflect differences in self-presentational motives, these
motives should be relevant to the public, laboratory setting. In contrast,
the present field study went to great lengths to make the students' re-
sponses anonymous and private. Here self-presentational motives were pre-
sumably less active and, thus, self-esteem did not moderate the effect of
handicaps and success on postfeedback attributions and self-esteem.
A second contribution of the present research is that it provides fur-
ther support for Weiner's (1986) distinction between outcome-generated
affect and attribution-dependent affect. Weiner's model of achievement
motivation and emotion proposes that outcome-generated affect is the gen-
eralized affective reaction a person experiences in response to positive or
negative performance feedback. Attribution-dependent affect is a secon-
dary affective response generated as a result of the attributions persons
make for their performance and is similar to feelings of self-worth. Our
162 Feick and Rhodewalt
data indicated that postfeedback mood was simply a function of the test
outcome (outcome generated) and not related to test performance attri-
butions. However, postfeedback self-esteem was attributionally mediated.
In summary, this investigation provided support for Jones and Berglas'
(1978) formulation of self-handicapping as an attributionally mediated
strategy for the protection as well as the enhancement of self-esteem. As
we have noted in our previous work (Rhodewalt, 1994; Rhodewalt & Hill,
1995), self-handicapping is a "Catch 22" in that handicappers take a hand
in bringing on an unwanted outcome in order to protect a desired self-im-
age. To casual observers such acts appear to be self-defeating and, at times,
pathological (see Berglas, 1988). We agree that certain chronic and extreme
forms of self-handicapping are indeed potentially self-injurious. However,
the present findings in conjunction with our earlier research (Mayerson &
Rhodewalt, 1988; Rhodewalt & Hill, 1995; Rhodewalt et al., 1991) indicate
that self-handicaps, at least claimed ones, are short-term functional strate-
gies that permit one to maintain a positive self-image in the face of po-
tentially contradictory evidence. Our findings are limited to claims of
obstacles to success. Future research will need to address whether behav-
ioral handicaps provide the same rewards.
REFERENCES
Aiken, L. S., & West, S. G. (1991). Multiple regression: Testing and interpreting interactions.
Newbury Park, CA: Sage.
Berglas, S. (1988). The three faces of self-handicapping: Protective self-presentation, a strategy
for self-esteem enhancement, and a character disorder. In S. L. Zelen (Ed.),
Self-representation: The second attribution-personality conference, CSPP-LA, 1986 (pp.
133-169). New York: Springer-Verlag.
Berglas, S., & Jones E. E. (1978). Drug choice as a self-handicapping strategy in response to
noncontingent success. Journal of Personality and Social Psychology, 36, 405-417.
Fleming, J. S., & Courtney, B. E. (1984). The dimensionality of self-esteem. II. Hierarchical
facet model for revised measurement scales. Journal of Personality and Social Psychology,
46, 404-421.
Higgins, R. L., & Harris, R. N (1988). Strategic "alcohol" use: Drinking to self-handicap.
Journal of Social and Clinical Psychology, 6, 191-202.
Hirt, E. R., Deppe, R. K., & Gordon, L. J. (1991). Self-reported versus behavioral self-handicapping:
Empirical evidence for a theoretical distinction. Journal of Personality and Social
Psychology, 61, 981-991.
Jones, E. E., & Berglas, S. (1978). Control of attributions about the self through
self-handicapping strategies: The appeal of alcohol and the role of underachievement.
Personality and Social Psychology Bulletin, 4, 200-206.
Jones, E. E., & Rhodewalt, F. (1982). The Self-Handicapping Scale. Unpublished manuscript
(Available from F. Rhodewalt, Department of Psychology, University of Utah).
Judd, C., & Kenny, D. A. (1981). Estimating the effects of social interventions. Cambridge,
England: Cambridge University Press.
Kelley, H. H. (1972). Causal schemata and the attribution process. Morristown, NJ: General
Learning Press.
Self-Handicapping and Attributions 163