Journal of Abnormal Psychology Copyright 1997 by the American Psychological Association, Inc.

1997, Vol. 106, No. 1, 145-153 002I-843X/97/S3.00

Perfectionism, Perceived Weight Status, and Bulimic Symptoms:

Two Studies Testing a Diathesis-Stress Model
Thomas E. Joiner, Jr. Todd F. Heatherton
University of Texas Medical Branch at Galveston Dartmouth College

M. David Rudd Norman B. Schmidt

Scott and White Clinic and Hospital and Texas A & M Uniformed Services University of the Health Sciences
University Health Sciences Center

Among a total of 890 women from 2 separate nonclinical samples (1 collected in 1982 on college
undergraduates, n = 435, and 1 collected in 1992 on women who were in college in 1982, n =
455), the authors tested a diathesis-stress model of the interrelations of perfectionism, perceived
weight status, and hulimic symptoms. The authors predicted and found that perfectionism served as
a risk factor for bulimic symptoms for women who perceived themselves as overweight but did not
serve as a risk factor for those who did not perceive themselves as overweight. Perceived weight
activated perfectionism as a predictor of bulimic symptoms; actual weight did not serve the same
role. These findings are discussed in the context of recent clinical and social psychological theory
regarding development of bulimic symptoms.

Hewitt, Flett, and colleagues (e.g., Hewitt & Flett, 1991a,
1991b, 1993), as well as Frost and associates (Frost, Heimberg,
Holt, Mattia, & Neubauer, 1993), have demonstrated that perfec-
tionism appears to be reliably associated with depression. Re-
cently, investigators have elaborated on this finding, and have
proposed a diathesis-stress approach. From this perspective, per-
fectionism is viewed as a risk factor for depression, but only in
the presence, and not in the absence, of negative life stress. That
is, the statistical interaction of perfectionism and life stress is
examined as a predictor of depression (e.g., Flett, Hewitt,
Blankstein, & Mosher, 1991; Hewitt & Dyck, 1986; Joiner &
Schmidt, 1995). For example, Hewitt and Flett (1993) found
Thomas E. Joiner, Jr., Department of Psychiatry and Behavioral Sci-
ences, University of Texas Medical Branch at Galveston; Todd F. Heath-
erton, Department of Psychology, Dartmouth College; M. David Rudd,
Department of Psychiatry and Behavioral Sciences, and Scott and White
Clinic and Hospital, College of Medicine, Texas A & M University Health
Sciences Center; Norman B. Schmidt, Uniformed Services University of
the Health Sciences.
Preparation of this article was supported in part by a grant from the
Henry Murray Center at Radcliffe College and by grants from the Rad-
cliffe College Research Fund and the Boston Obesity Nutrition Research
Center (DK 42600).
The opinions or assertions contained herein are the private ones of
the authors and are not to be construed as official or reflecting the views
of the Uniformed Services University of the Health Sciences or the
Department of Defense.
We thank the Henry Murray Center of Radcliffe College for providing
access to the data from the 1982 sample and the Alumni Office of
Harvard University for supplying current addresses for participants.
Correspondence concerning this article should be addressed to
Thomas E. Joiner, Jr., Department of Psychiatry and Behavioral Sciences,
Graves Building, University of Texas Medical Branch, Galveston, Texas
77555-0425. Electronic mail may be sent via Internet to thomas.
joiner@utmb.edu.
that among depressed and other psychiatric patients, those who
were high in both perfectionism and in daily life stress were
most likely to obtain high depression scores.
Perfectionism has also been indicated as a correlate of bulimic
symptoms. For example, Rosch, Crowther, and Graham (1991)
reported that clinically bulimic university women, as compared
with general psychotherapy patients and normal controls, ob-
tained elevated perfectionism scores (for similar findings, see
Ruderman, 1986; Thompson, Berg, & Shatford, 1987). Simi-
larly, Rothenberg (1990) found that hospitalized bulimic pa-
tients differed from general psychiatric patients with respect
to perfectionism, as well as related characteristics, such as
scrupulousness.
At the same time, however, the association of perfectionism
and bulimic symptoms has been questioned. For example, Hur-
ley, Palmer, and Stretch (1990), in a study comparing eating
disordered patients with general psychiatric patients, reported
that perfectionism was not specific to eating-disordered patients.
Blouin, Bushnik, Braaten, and Blouin (1989) found that bulimic
patients were more perfectionistic than normal controls, but
similarly perfectionistic relative to diabetic patients. Newton,
Butler, and Slade (1988) reported no differences in perfec-
tionism between eating disordered patients with mild bulimic
symptoms and those with more severe symptoms.
Thus, there is some question as to the very existence of a
relationship between perfectionism and bulimic symptoms. Fur-
thermore, should the relationship exist, the form of the relation-
ship is unclear—is perfectionism in itself a risk factor for bu-
limic symptoms; or is perfectionism one factor, which, only in
combination with others, heightens risk for bulimia? The goal
of the present studies is to address these issues in the context
of a diathesis-stress framework, among large samples of women,
a subset of whom endorsed the Diagnostic and Statistical Man-
ual of Mental Disorders, 3rd edition (DSM-III; American Psy-

145
146 JOINER, HEATHEKTON, RUDD, AND SCHMIDT
chiatric Association, 1980) based criteria for a clinical diagnosis
of bulimia.
The present conceptualization views perfectionism as a risk
factor for bulimic symptoms, but only under certain conditions.
If one strives to meet high standards (e.g., perfect body weight
and shape), even if such standards are excessively high, negative
outcomes should not be expected unless the standards go unmet
(cf. Heatherton & Baumeister, 1991). Perfection, if both desired
and obtained, is a positive state of affairs. Indeed, this is pre-
cisely the rinding from the diathesis-stress work on perfec-
tionism and depression; perfectionistic people, despite their per-
fectionism, were not depressed unless things were imperfect
(i.e., unless negative life stress occurred; Hewitt & Flett, 1993;
Joiner & Schmidt, 1995). A similar result has emerged from
within the literature on negative cognitive style and depression:
People with negative attributional styles tend not to become
depressed unless they also encounter negative life stress (e.g.,
Metalsky, Joiner, Hardin, & Abramson, 1993).
According to the current conceptualization, perfectionistic
women will experience bulimic symptoms when they feel that
their standards are unmet. However, it is postulated that perfec-
tionistic women, despite their perfectionism, will not experience
bulimic symptoms when they feel that their standards have been
satisfied. This conceptualization is depicted in Figure 1.
Because there is a well-known main effect for weight percep-
tions on bulimic symptoms (e.g., Ruderman, 1986), women in
quadrants C and D of Figure 1 are predicted to have higher
bulimic symptoms than women in quadrants A and B. As was
mentioned earlier, the main effect relation of perfectionism on
bulimic symptoms is less clear, although there is at least some
evidence to support a weak main effect, such that women in
quadrants B and D may experience higher levels of bulimic
symptoms than women in quadrants A and C. More important
to the present conceptualization, women in quadrant D are pre-
dicted to experience the highest levels of bulimic symptoms,
because their high standards have been violated (i.e., they feel
overweight and expect themselves not to be overweight).
It is important to note that the lermfeel overweight has been
used, as opposed to are overweight. According to the present
view, actual weight status is less relevant to perfectionistic
High
Faals Overweight?
Figure I. Four groups created by crossing the perfectionism and per-
ceived weight variables.
women than is perceived weight status. Although both actual
and perceived weight serve as benchmarks by which standards
are evaluated, it is argued that perceived weight is more powerful
for perfectionistic women. To be sure, there is a relation between
actual and perceived weight status. But there is also disparity.
It is predicted that when perfectionistic women feel overweight,
(even if they are not; cf. quadrant D of Figure 1 ) , they will be
at high risk for bulimic symptoms, whereas when perfectionistic
women do not feel overweight, (regardless of whether they are;
(cf. quadrant B of Figure 1), they will not experience bulimic
symptoms.
To summarize, hypotheses for the present studies were as
follows;
1. Perceived weight would display a strong main effect relation
to bulimic symptoms (quadrants C and D would have higher
bulimia scores than quadrants A and B).
2. Perfectionism was more tentatively predicted to relate to bu-
limic symptoms, and in any event, the magnitude of the relation
should represent a considerably weaker effect than that of per-
ceived weight on bulimic symptoms (quadrants B and D may
have higher bulimia scores than quadrants A and C).
3. Most important to the present conceptualization, perfec-
tionism was predicted to relate strongly to bulimic symptoms
among women who felt overweight, but less strongly among
women who did not feel overweight (quadrant D would have
the highest bulimia scores). That is, perceived weight status and
perfectionism were hypothesized to interact to predict bulimic
symptoms.
4. Actual weight would display a strong main effect relation to
bulimic symptoms.
5. However, actual weight was not postulated to interact with
perfectionism to predict bulimic symptoms, and to the extent
that it did, it was predicted that it did so only because it was
highly related to perceived weight.
6. Related to Hypothesis 5, it was hypothesized that, among
perfectionistic women, those who felt overweight but who were
not overweight would experience more bulimic symptoms than
those who did not feel overweight but who were overweight.
Study 1
Method
Participants
In the spring of 1982, researchers affiliated with Radcliffe College
distributed surveys to a randomly selected sample of 800 women and
400 men who were students at Boston University. The focus of the
present study was on the women, 435 of whom responded to the ran-
domly distributed questionnaire (response rate = 54%; response rate
for men was 50%). The women were predominantly Caucasian (88%),
and the mean age was 20.11 years (SD = 1.2).
Procedure
The questionnaire included an array of items about demographic back-
ground; height and weight; concerns about dieting, eating patterns, and
body weight and shape; and eating disorder symptomatology. Of particu-
lar interest for the present study, participants completed the Perfec-
 PERFECTIONISM, PERCEIVED WEIGHT, BULIMIC SYMPTOMS
tionism and Bulimia subscales of the Eating Disorders Inventory (EDI;
Garner, Olmstead, & Polivy, 1983), responded to questions about their
weight, and responded to questions based on DSM-I1I criteria for bu-
limia nervosa, as described below.
Eating Disorders inventory and Perfectionism and Bulimia subscales.
The EDI is a frequently used 64-item self-report measure of eating-
related attitudes and traits. It yields eight subscales; Drive for Thinness,
Bulimia, Body Dissatisfaction, Ineffectiveness, Perfectionism, Interper-
sonal Distrust, Interoceptive Awareness, and Maturity Fears. The sub-
scales have shown adequate internal consistency coefficients and stable
test-retest correlations and have been extensively validated (see Garner,
Olmstead & Polivy, 1983).
The present study focused on the Perfectionism and Bulimia subscales.
The Perfectionism subscale includes six items designed to measure gen-
eral perfectionism (e.g., "I feel that I must do things perfectly, or not
do them at all" or "Only outstanding performance is good enough in
my family. 1 '). Alpha internal consistency coefficient in the present sam-
ple was .79, similar to the figure of .88 obtained among a sample of
undergraduate women (Joiner & Schmidt, 1995). The construct validity
of the scale is supported by expected correlations with depressive
(Joiner & Schmidt, 1995) and eating disorder symptoms (e.g., Rosch
et al.. 1991).
Joiner and Schmidt (1995) found that the EDI Perfectionism subscale
assesses two correlated aspects of perfectionism, self-oriented perfec-
tionism (i.e., expecting self to be perfect; three items) and socially
prescribed perfectionism (i.e., perceiving that others expect perfection;
three items). Despite the low number of items on each subscale, internal
consistency reliability in the present sample was adequate (.70 for self-
oriented; .72 for socially prescribed; Joiner & Schmidt reported .74 for
self-oriented; .76 for socially prescribed, as well as test-retest reliability
of .64 and ,68, respectively).
The Bulimia subscale includes seven items that assess bingeing and
purging (e.g., "I stuff myself with food" or "I have thought of trying
to vomit to lose weight."). Participants were asked to rate items on a
scale ranging from 1 to 6 where 1 = never, 2 = rarely, 3 = sometimes,
4 — frequently, 5 = usually, and 6 = alwavs. Coefficient alpha in the
present study was .86.
Perceived weight status. Participants were asked to categorize them-
selves as either overweight or not overweight (not vvenveight = 1;
overweight = 2). 43% of the participants (« = 187) identified them-
selves as overweight; the rest (n = 248) identified themselves as not
overweight. A separate question tapped a similar construct: whether
participants were trying to lose weight ( = 2 ) or not (= I ) . Three hundred
and twenty-six women (75%) indicated they were trying to lose weight,
and the remainder (n = 109) indicated that they were not trying to lose
weight. For both variables, these percentages are quite consistent with
those obtained by Heatherton, Nichols, Mahamedi, and Keel (1995) on
similar samples of women.
In the analyses reported below, the focus is on the perceived weight
variable. Analyses using the lose-weight variable are offered as supple-
mentary support for those involving the perceived weight variable.
Actual weight status. Participants provided their actual height and
weight. On this basis, the Body Mass Index (kg/m 2 ) was calculated.
Also, participants were categorized with regard to height/weight (very
underweight, underweight, average weight, overweight, or obese) on the
basis of standards from the National Health and Nutrition Examination
Survey (see, e.g., Kuczmarski, 1992). It is important to note that Radke-
Sharpe, Whitney-Saltiel, and Rodin (1990) demonstrated that self-mea-
surements are quite similar to measurements done by experts.
Bulimic diagnoses. Bulimic diagnoses were determined based on
participants' responses to questions based on DSM~Ifl criteria for bu-
limia nervosa. To meet criteria in the present study, participants must
have had (a) indicated recurrent experiences of at least one binge eating
episode per week; ( b ) indicated that binges consisted of large amounts
of high-calorie food (not just large meals); (c) indicated that at least
some binges occurred when participant was alone; (d) reported making
147
repeated attempts to lose weight by severely restrictive diets, self-in-
duced vomiting, or the use of cathartics or diuretics; (e) indicated that
the binges were experienced as out of control; (f ) reported often feeling
extremely guilty after overeating; and (g) indicated that weight had not
fallen more than 20 percent below normal since age 16.
These seven criteria afford good coverage of current diagnostic criteria
for bulimia nervosa in the Diagnostic and Statistical Manual of Mental
Disorders, 4th edition, (DSM-/V; American Psychiatric Association,
1994), and as shown below, their relation to the EDI bulimia subscale
(r = .56) provides some confidence in them. Furthermore, the rate of
diagnosis in the present sample (50 of 435 = 12%), although somewhat
high, is similar to estimates provided by Zuckerman, Colby, Ware, and
Lazerson (1986) and by Heatherton et al. (1995) on similar samples of
women. However, they should not be viewed as diagnoses established
through structured clinical interview. They are thus viewed cautiously
and presented as a supplement to the data on the EDI Bulimia subscale.
Results and Discussion
Means and standard deviations are presented in Table 1 (see
diagonal) and were consistent with previous reports (e.g.,
Heatherton et al., 1995). Table 1 also presents the zero-order
correlations between all measures. Of particular interest, the
Perfectionism subscale was weakly correlated with EDI bulimia
subscale (r = .21, p < .001) and not correlated at all with the
diagnostic variable (r = .07, p ~ ns). perfectionism and per-
ceived weight status were not highly correlated (r = .14, p <
.005), which suggests that multicolinearity among predictors in
the regression analyses presented below was not an issue.
As was expected, the EDI Bulimia subscale correlated
strongly with the diagnostic variable (/• = .56, p < .001), with
perceived weight status (r = .49, p < .001), and with the
Body Mass Index (r ~ .35, p < .001). Perceived weight status
correlated with the Body Mass Index (r - .55, p < .001),
consistent with the view that there is overlap and disparity be-
tween actual and perceived weight. Perceived weight status was
also correlated with the lose/gain weight variable (r - .43, p
< .001).
Prediction of Bulimic Symptoms by the Perfectionism X
Perceived Weight Status Interaction
Consistent with the recommendations of Cohen and Cohen
(1983), a setwise hierarchical multiple regression/correlation
(MRC) procedure was used to test predictions pertaining to
the Perfectionism X Perceived Weight Status interaction. EDI
Bulimia subscale scores served as the dependent variable. Be-
cause the distribution of EDI Bulimia scores was positively
skewed (as is usually the case with a symptom measure in a
population sample), EDI Bulimia scores were transformed, us-
ing the square-root transformation recommended by Cohen and
Cohen (1983, p. 252).
EDI Perfectionism subscale scores and a dichotomous vari-
able indicating perceived weight status (where 1 = do not feel
overweight and 2 = feel overweight) were entered into the
regression equation simultaneously as a set, followed at Step 2
by entry of the Perfectionism X Perceived Weight Status interac-
tion term.
As can be seen in Table 2, regression analysis indicated that,
consistent with prediction, (a) perfectionism subscale is posi-
tively and somewhat weakly related to bulimic symptoms, pr =
148 JOINER, HEATHEKTON, RUDD, AND SCHMIDT

Table 1
Descriptive Data and Intercorrelations for All Variables in Study 1 (N = 435)

Variable 1 2 3 4 5 6 7

1. EDI perfectionism 21.98

(6.46)
2. EDI bulimia .21 13.04
(5.58)
3. Bulimic diagnosis .07 .56 —
4. Perceived weight status .14 .49 .21
—
5. Trying to lose weight? .09 .36 .18 .43 —
6. Body mass index .07 .35 .18 .55 .51 21.31
(2.78)
7. Weight class .07 .33 .13 .58 .46 .90 —

Note. Numbers in parentheses represent standard deviations. Bulimic diagnosis, peceived weight status,
and trying to lose weight are dichotomous variables. Correlations s.18 are significant to the .001 level;
correlations s.13 and <,18 are significant to the .005 level; correlations ^.09 are ns. EDI - Eating
Disorders Inventory (Gamer, Olmstead, & Polivy, 1983).

.18, t(432) = 3.75, p < .05; (b) perceived weight status dis-
played a strong main effect relation to bulimic symptoms, pr =
.48, 1(432) = 11.42, p < .05; and (c) most important to the
present conceptualization, the Perfectionism X Perceived Weight
Status interaction served as a significant predictor of EDT Bu-
limia subscale scores, pr — .12, ((431) = 2.47, p < .05.
Interestingly, a similar pattern of results emerged when the
bulimic diagnosis variable was used instead of the EDI bulimia
subscale as the dependent variable (see Cohen & Cohen, 1983,
pp. 229-230, for defense of use of MRC analysis with dichoto-
mous dependent variables). These results are summarized in
Table 3. At the main effect level, perceived weight status contin-
ued to serve as a significant predictor of bulimic symptoms, pr
= .21, ((432) = 4.48, p < .05, but the Perfectionism subscale
did not, pr = .05, ((432) = 0.96, p = ns. More important, the
Perfectionism X Perceived Weight Status interaction served as
a significant predictor of the bulimic diagnosis variable, pr =
.15,r(431) = 3 . U , p < .05.
It is also of interest to note that when the perceived weight
status variable was replaced in the two sets of analyses described
above by the trying to lose weight variable, the results were
highly similar. Predicting both the EDI bulimia subscale and the
bulimia diagnosis variable, the Perfectionism X Trying to Lose
Weight interaction term was significant; respectively, pr = .11,
i(431) = 2.22, p < .05; pr = .10, t(431) = 2.02, p < .05.
Thus it appears that the study's main hypothesis (that the
Perfectionism X Perceived Weight Status interaction would be
predictive of bulimic symptoms) was supported, regardless of
whether the EDI Bulimia subscale or the diagnostic variable
was used as the dependent variable and regardless of whether
perceived weight status was assessed directly, or more obliquely
by way of the trying to lose weight question. It remains to be
demonstrated, however, that the form of this interaction is as
predicted (i.e., that perfectionistic women who perceive them-
selves as overweight will experience the highest levels of bu-
limic symptoms; see quadrant D of Figure 1).
To examine this issue, the relation between the Perfectionism
subscale and EDI Bulimic symptoms subscale was examined
among two subgroups of participants; those who did and those
who did not report feeling overweight (results using the bulimia
diagnostic variable and the trying to lose weight variable were
highly similar and are not reported; see Joiner, Metalsky, &
Wonderlich, 1995 for a similar data-analytic procedure). In line

Table 2
Perfectionism, Perceived Weight Status, and the Perfectionism Table 3
X Perceived Weight Status Interaction Predicting Scores on Perfectionism, Perceived Weight Status, and the Perfectionism
EDI Bulimia Subscale in Study 1 X Perceived Weight Status Interaction Predicting Bulimia
Diagnosis Variable in Study 1
(for Partial
Order of entry/ Ffor within-set correlation ( for Partial
predictors in set set predictors df (PRIpr) Order of entry/ Ffor within-set correlation
predictors in set set predictors df {PRIpr)
1. Main effects 80.12 2,432 .52
Perfectionism 3.75 432 .18 1. Main effects 11.34* 2,432 .22
Perceived weight status 11.42 432 .48 Perfectionism 0.96 432 .05
2. Two-way interaciton 6.10 1,431 .12 Perceived weight status 4.48* 432 .21
Perfectionism x 2. Two-way interaction 9.67* 1,431 .15
perceived weight Perfectionism X
status 2.47 431 .12 perceived weight
status 3.11* 431 .IS
Note. EDI = Eating Disorders Inventory (Garner, Olmstead, & Polivy,
1983); PR = multiple partial correlation for a set of predictors; pr - Note. PR = multiple partial correlation for a set of predictors; pr
partial correlation for within-set predictors. All effects are significant partial correlation for within-set predictors.
at the p < .05 level. *p < .05.
 PERFECTIONISM, PERCEIVED WEIGHT. BULIMIC SYMPTOMS 149

STUDY 1 STUDY 2
Perfectionism
Perfectionism
High High

Bulimia Score =10. 11 Bulimia Score = 10.76 Bulimia Score =10.45 Bulimia Score .11. 04

A B A B

Feels Feels
Overweight? Overweight'

Bull mia Score = 13.77 Bulimia Score == 16.89 Bulimia Score = 11. 73 Bulimia Score =15.03

c D C D

Compare to sample-wide Bulimia Score Mean Compare to sample-wide Bulimia Score Mean
of 1304 (§D = 5.58) of11.08(SD = 5.02)

Figure 2. Predicted scores on the Eating Disorder Inventory (Garner, Olmstead, & Polivy, 1983) Bulimia
Subscale in Study 1 (left) and Study 2 (right) for four groups created by crossing the perfectionism and
perceived weight variables.

with prediction, the Perfectionism subscale significantly pre-
dicted EDI Bulimia subscale scores among those who perceived
themselves as overweight (r = .31, p < .05), but not among
those who perceived themselves as not overweight (r = .09,
p = ns).
To graphically demonstrate these findings, following Cohen
and Cohen (1983, pp. 323 and 419), EDI Bulimia subscale
scores were computed by inserting specific values for Perfec-
tionism (i.e., 1 SD above and below the mean) and for perceived
weight status (i.e., 1 = do not feel overweight and 2 = feel
overweight) into the regression equation summarized in Table
2(see Joiner, 1995; Joiner etal., 1995 for a similar data-analytic
procedure). The results of this analysis are depicted in the left
panel of Figure 2.
As can be seen in Figure 2, consistent with prediction, women
who were both perfectionistic and reported feeling overweight
experienced the highest bulimia scores (see quadrant D of Fig-
ure 2). Interestingly, among women who did not feel over-
weight, perfectionistic participants (quadrant B) obtained simi-
lar bulimia scores as nonperfectionistic participants (quadrant
A). Women who felt overweight but were nonperfectionistic
obtained average bulimia scores. This pattern of findings is quite
consistent with the present view that perfectionism may lead to
bulimic symptoms for those who do (but not lead to bulimic
symptoms for those who do not) perceive themselves as
overweight.
predicting EDI Bulimia subscale scores are reported in Table 4
(findings predicting the bulimia diagnosis variable were similar
and are not reported). As can be seen there, consistent with
prediction, the Body Mass Index displays a main effect relation
to bulimic symptoms, pr = .37, /(432) = 8.44, p < .05, but
the Perfectionism X Body Mass Index interaction does not, pr
= .02, r(431) = 0.72, p = ns.
Second, the analyses summarized in Tables 2 and 3 were
again repeated, now with actual weight status inserted as a
covariate. That is, the Body Mass Index was entered first into
the regression equation predicting EDI Bulimia subscale scores,
followed by simultaneous entry of the Perfectionism subscale
and perceived weight status, and at the last step, by the Perfec-
tionism x Perceived Weight Status interaction. Notably, the co-
variance of the Body Mass Index had very little effect on the
results: The main effects for perceived weight status, pr = .32,
f(431) = 7.80, p < .05, and Perfectionism, pr = .16, r(431)
= 3.90, p < .05, remained, as did the significant effect for the
interaction term, pr = .12, /(430) = 2.48, p < .05.
Third, it was suggested that when perfectionistic women feel
Table 4
Perfectionism, Body Mass Index, and the Perfectionism
x Body Mass Index Interaction Predicting
EDI Bulimia Scores in Study I

t for Partial
Order of entry/ Ffor within-set correlation
Role of Perceived Weight Status Versus Actual predictors in set set predictors df (PR/pr)
Weight Status
\. Main effects 47.10* 2,432 .42
Perfectionism 4.20* 432 .20
It was predicted that feeling overweight impinged more pow-
Body mass index 8.44» 432 .38
erfully on perfectionism than being overweight. This hypothesis 2. Two-way interaction 0.52 1,431 .02
was evaluated in three ways. Perfectionism X
First, the analyses summarized in Tables 2 and 3 were re- body mass index 0.72 431 .02
peated, except that actual weight status was used as a predictor
Note. EDI = Eating Disorders Inventory (Garner, Olmstead, & Polivy,
instead of perceived weight status. That is, the Perfectionism
1983); PR = multiple partial correlation for a set of predictors; pr =
subscale, the Body Mass Index, and the interaction between partial correlation for within set predictors.
them were assessed as predictors of bulimic symptoms. Results *p < .05.
150 JOINER, HEATHEKTON, RUDD, AND SCHMIDT
overweight, even if they are not, they will be at high risk for
bulimic symptoms, whereas when perfectionistic women do not
feel overweight but are, they will not experience bulimic symp-
toms. We examined this issue among perfectionistic participants
(i.e., those with EDI Perfectionism subscale scores above the
median; « = 218).
EDI Bulimia subscale scores were examined among two
groups of perfectionistic women; (a) those who felt overweight
but were not (n ~ 43), and (b) those who did not feel over-
weight but were (» = 12).' An interesting feature of this com-
parison is to orthogonalize actual and perceived weight.
If, as is predicted, perceived weight more powerfully im-
pinges on perfectionistic attitudes than does actual weight, EDI
Bulimia subscale scores should have been significantly higher
among perfectionistic women who perceived themselves as over-
weight but were not, than among perfectionistic women who
did not perceive themselves as overweight but were. This was,
in fact, the case. The mean EDI bulimia subscale score among
those who perceived themselves to be overweight but were not
overweight was 17.86; the corresponding figure among those
who did not perceive themselves to be overweight but were
overweight was 12.83. The difference between these means was
statistically significant, t(53) = 2.51, p < .01.
To summarize Study 1, results were consistent with all
hypotheses. Perfectionistic women reported bulimic symptoms,
but only when they felt that their standards were unmet (i.e.,
perceived themselves as overweight; see quadrant D of Figure
2). Perfectionistic women, despite their perfectionism, did not
experience bulimic symptoms when they perceived themselves
as not overweight. It was argued that perceived weight impinged
more forcefully on perfectionistic attitudes than actual weight,
and results were supportive of this view.
The supportive results should be viewed in the context of
some of the study's limitations. First, the response rate to the
questionnaire (54%) was not ideal. It is possible that the rela-
tively low response rate may have biased results. Second, al-
though successfully used in previous work (Heatherton et al.,
1995), the reliability of the perceived weight status variable is
difficult to formally estimate. On the other hand, the variable's
correlation with the trying to lose weight variable is encourag-
ing, as is the general finding that people's self-reports of weight-
related data are reliable (Radke-Sharpe et al., 1990). Third, the
study's findings are limited to undergraduate women, and may
not apply to women in different age categories. Fourth, although
hypotheses were supported without exception, the absolute mag-
nitude of the interaction term's effect size (e.g., pr = .12 in
Table 2) was relatively small. The relatively small effect size,
together with the questions regarding reliability and replicability,
call for replication of the present results, preferably in a sample
with a different age range and a higher response rate. Study 2
answers this call, and furthermore. Study 2 addresses a final
concern that the data from Study 1 were somewhat dated (col-
lected in 19K2 and data in Study 2 were collected in 1992).
Study 2
Method
Participants
In the spring of 1992, Heatherton and colleagues attempted to identify
and follow-up all participants from an earlier study of eating patterns
and behavior. The earlier study occurred in 1982 at Harvard University,
Participants in Study 2 differed from those of Study I in four ways.
First, the response rate in Study 2 was better (82% of the original sample
participated in 1992). Second, whereas participants in Study 1 were
students at Boston University in 1982, participants in Study 2 were
students at Harvard University in 1982. Third, whereas the data for
Study 1 were collected in 1982, those for Study 2 were collected in
1992. And fourth, whereas the average age of participants in Study I
was 20.11 years, the average age of participants in Study 2 was 29.93
(.TO = 1.81). Study 2 included 455 women.
Procedure
The procedure and materials for Study 2 were identical to those for
Study 1. As in Study 1, participants completed the perfectionism and
bulimia subscales of the EDI, responded to questions about their weight,
and responded to questions based on DSM-III criteria for bulimia ner-
vosa (see description in Study I ). Regarding the D.W-based criteria,
4% of participants (n = 19) met criteria (cf. Heatherton et al., 1995
who reported a decrease in weight- and eating-related problems from
1982 to 1992).
Results and Discussion
Means, standard deviations, and intercorrelations are pre-
sented in Table 5 and were fairly consistent with those reported
in Study 1. Interestingly, as is perhaps predictable, Harvard
alumnae in Study 2 scored higher on the perfectionism subscale
than did Boston University students in Study 1. Also, in Study
2, the correlation between perceived weight status and the Body
Mass Index was high (r = .71, p < .001), which may work
against our hypothesis that perceived weight impinges more
strongly on perfectionism than actual weight.
Prediction of Bulimic Symptoms by the Perfectionism X
Perceived Weight Status Interaction
The data-analytic approach in Study 2 was identical to that
used in Study 1. Table 6 displays the results of a regression
equation with EDI Bulimia subscale as the dependent variable,
and EDI Perfectionism subscale, perceived weight status (where
1 = do not feel overweight and 2 = feel overweight), and the
Perfectionism X Perceived Weight Status interaction term as
predictor variables.
As Table 6 shows, findings indicated that, consistent with
prediction and with the results of Study 1, (a) the Perfectionism
subscale was positively and somewhat weakly related to bulimic
symptoms, pr = .23, ((452) = 5.00, p < .05; (b) perceived
weight status displayed a fairly strong main effect relation to
bulimic symptoms, pr = .38, /(452) = 8.63, p < .05; and (c)
most important to the present conceptualization, the Perfec-
tionism X Perceived Weight Status interaction served as a sig-
nificant predictor of EDI Bulimia subscale scores, pr - .10,
/ ( 4 5 1 ) = 2.07, p < .05.
As in Study 1, a similar pattern of results emerged when the
1
Because the goal of this analysis was to separate feeling from being
overweight, participants whose perceived weight status matched their
actual weight status were excluded. Thus, those who were overweight,
and perceived themselves as such, were excluded (n = 50), as were
participants who neither viewed themselves as overweight nor were
overweight (n = I 13).
 PERFECTIONISM, PERCEIVED WEIGHT, BULIMIC SYMPTOMS 151

Table 5
Descriptive Data and Intercorrelatiom for All Variables in Study 2 (N = 455)

Variable 1 2 3 4 5 6 7

1 . EDI perfectionism 24.91

(5.15)
2. EDI bulimia .26 11.08
(5.02)
3. Bulimic diagnosis .23 .75 —
4. Perceived weight status .07 .41 .38
—
5. Trying to lose weight? .07 .19 .25 .28 —
6. Body mass index .01 .36 .35 .71 .26 22.01
(3.57)
7. Weight class -.05 .27 .22 .56 .31 .80 —

Note. Numbers in parentheses represent standard devialiens. Bulimic di agnosis, perceived weight status,
ana trying to lose weignt are aicnotomous vanaoies. \_orreiauons ^:. I? are signmcam 10 me ..001 level;
correlations ==.07 are ns. EDI = Eating Disorder Inventory (Garner, Olmstead, & Polivy, 1983).

bulimic diagnosis variable was used instead of the EDI bulimia predictor instead of perceived weight status, the Perfectionism
subscale as the dependent variable, and when the trying to lose X Actual Weight Status interaction did not predict bulimic
weight variable was used instead of the perceived weight status symptoms, pr = .03, t(451) = 0.73, p = ns. When actual weight
as a predictor (results available from Thomas Joiner upon status was inserted as a covariate in the regression analysis
request). described in Table 6, the Perfectionism X Perceived Weight
Again replicating the results of Study 1, the form of the Status interaction remained a significant predictor of bulimic
Perfectionism X Perceived Weight Status interaction was as pre- symptoms, pr = .10, r(450) = 2.05, p < .05.2
dicted, and the scores produced by the regression equation were In conclusion, the supportive results of Study 1 were fully
highly similar to those in Study 1 (see right panel of Figure replicated in Study 2, on a different, older, and more recent
2). Women who were both perfectionistic and reported feeling sample with a higher response rate. Perfectionistic women expe-
overweight experienced the highest bulimia scores. These re- rienced bulimic symptoms, but only when they perceived them-
sults, like those of Study 1, are quite consistent with the present selves as overweight. It appeared that perceived weight, not
model that perfectionism may lead to bulimic symptoms for actual weight, served to activate perfectionistic attitudes as a
those who do (but not lead to bulimic symptoms for those who predictor of bulimic symptoms.
do not) perceive themselves as overweight.
Genera] Discussion
Role of Perceived Weight Status Versus Actual
Across two studies (one in 1982 and one in 1992) including
Weight Status
a total of 890 women, our diathesis-stress model of the interrela-
The prediction that feeling overweight impinged more power- tions of perfectionism, perceived weight status, and bulimic
fully on perfectionism than being overweight was fully repli- symptoms received empirical support. Perfectionism appeared
cated in Study 2. When actual weight status was used as a to serve as a risk factor for bulimic symptoms for women who
perceived themselves as overweight, but not for those who did
not. Perceived weight activated perfectionism as a predictor of
Table 6 bulimic symptoms; actual weight did not serve the same role.
Perfectionism, Perceived Weight Status, and the Perfectionism Before elucidating the studies' implications, it is first im-
X Perceived Weight Status Interaction Predicting Scores on portant to consider some limitations. First, the study was not
EDI Bulimia Subscale in Study 2 prospective, and thus it cannot be argued that perfectionism has
been shown to be an antecedent of bulimic symptoms. Second,
(for Partial structured clinical interviews were not conducted, and results
Order of entry/ Ffor within-set correlation
on the diagnostic variable reported herein should be interpreted
predictors in set set predictors # (PR/pr)
cautiously, although it was reassuring that these results eon-
1 . Main effects 55.83 2,452 .45 verged with those using the EDI Bulimia subscale, without ex-
Perfectionism 5.00 452 .23 ception. Another potential measurement concern involves the
Perceived weight status 8.63 452 .38 EDI Perfectionism subscale, which measures perfectionism in
2. Two-way interaction 4.29 1,451 .10
Perfectionism x
perceived weight 2
Interestingly, the analysis comparing those with discrepant perceived
status 2.07 451 .10
and actual weight statuses was not possible, because too few perfection-
Note. EDI — Eating Disorders Inventory (Garner, Olmstead, & Polivy, istic women in 1992 reported actual weight status in the overweight
1983); PR — multiple partial correlation for a set of predictors; pr = range. This pattern mirrors the trend, reported elsewhere (Heatherton et
partial correlation for within-set predictors. All effects are significant al., 1995), of a relative decrease from 1982 to 1992 in weight and eating
at the p < .05 level. problems.
152 JOINER, HEATHERTON, RUDD, AND SCHMIDT
general, and not in weight-specific terms. In response, it should
be noted that any discrepancy between weight-specific and gen-
eral perfectionism should work against the hypotheses of the
present study, and furthermore, the present model does not sug-
gest a weight-specific altitude as a diathesis. Rather, a general
attitude (perfectionism) is postulated to be activated by a
weight-related variable to predict eating-related symptoms. It
should also be noted that Hewitt, Flett, and colleagues (Hewitt,
Flett, Turnbull-Donovan, & Mikail, 1991) have provided a so-
phisticated measure of perfectionism, the Multidimensional Per-
fectionism Scale, which would have provided a more compre-
hensive measure of Perfectionism.
The nomological status of perfectionism as a correlate of
bulimic symptoms deserves consideration. As was mentioned
earlier, the studies were not prospective, and thus the data cannot
speak to whether perfectionism precedes bulimia. At the same
time, however, the findings are not particularly consistent with
the view that perfectionism is merely a correlate, associated
feature, or consequence of bulimic symptoms. If it were, one
might expect a correlation between perfectionism and bulimia
regardless of the levels of any other variables, such as perceived
weight status. That is, if perfectionism were an associated fea-
ture or consequence of bulimia, one might expect that perfec-
tionism and bulimia would be similarly related for all women,
regardless of whether they perceive themselves as overweight.
The pattern of findings predicted and found in the current study
suggests either a diathesis-stress view, wherein perfectionism
precedes bulimia for some women but not others or, alternatively,
a subset of women with bulimic symptoms (i.e., those who also
perceive themselves as overweight) for whom perfectionism is
an associated feature or consequence of bulimia. A prospective
study could address this issue more definitively.
It is interesting to note that Hewitt, Flett and colleagues (e.g.,
Hewitt & Flett, 1991a) have emphasized the difference between
socially prescribed perfectionism (perceiving that others expect
perfection from you) and self-oriented perfectionism (expecting
oneself to be perfect). Some "imperfections" may be expected
to impinge more powerfully on self-oriented versus socially
prescribed perfectionism. However, perceived weight status may
not operate in this way, since women may expect themselves to
be thin, and may expect others to expect them to be thin. In
accord with this view, current results were similar when the
EDI Perfectionism subscale was divided into self-oriented and
socially prescribed items (see Joiner & Schmidt, 1995 for dis-
cussion of use of subset of EDI Perfectionism subscale as mea-
sures of socially prescribed vs. self-oriented perfectionism).
Our findings are consistent with Heatherton and Baumeister's
(1991) theory that binge eating may serve as an escape from
painful self-awareness. These theorists hypothesized that binge
eaters are acutely sensitive to standards, and when they fall
short of standards, binge eaters view themselves negatively and
assume that others do as well (cf. self-oriented vs. socially
prescribed perfectionism). These painful self-perceptions, ac-
companied by attendant emotional distress, cause an escape
response, characterized by disinhibition and, thus, binge eating.
In terms of the present study, it is interesting to note that Heath-
erton and Baumeister's theory proposes a mechanism (self-view
and emotional distress) through which perfectionism may affect
eating behavior among women who perceive that standards have
gone unmet. Empirical scrutiny of whether changes in self-
view and in emotional distress mediate the relation between
perfectionism and bulimic symptoms represents an interesting
avenue for future research.
The present results are quite consistent with self-discrepancy
theory (Higgins, 1987) and research (e.g., Strauman, Vookles,
Berenstein, Chaiken, & Higgins, 1991). Strauman and col-
leagues theorized that girls are, in general, socialized to be
overcontrolled and, thus, may be more likely to suffer from
overcontrolled symptoms and syndromes, such as eating disor-
ders, whereas boys are socialized to be less controlled and, thus,
may be more likely to experience discontrolled symptoms and
syndromes (e.g., conduct disorder). Consistent with this view,
these researchers found that a discrepancy between ideal and
actual views of the self was highly related to bulimotypic symp-
toms, such as body dissatisfaction, among undergraduate
women. The present findings were very similar; for women
whose view of actual self (i.e., perceived weight) was discrep-
ant from the ideal (i.e., perfectionism), bulimic symptoms were
more likely.
Strauman et al.'s (1991) theory that overcontrolled symptoms
are more likely in women than in men is interesting in fight of
Rothenberg's (1990) hypothesis that many eating-disordered
symptoms are manifestations of obsessive-compulsive pathol-
ogy. In defense of this position, Rothenberg found that eating-
disordered patients, when compared with general psychiatric
patients, displayed higher levels of rumination, rituals, and ex-
cessive cleanliness and orderliness. It is also noteworthy that
both bulimia (Hudson, Laffer, & Pope, 1982) and obsessive-
compulsive disorder (Benkert, Wetzel, & Szegedi, 1993) have
been postulated to be manifestations of mood disorder. One
could interpret these syndromal interrelations as supportive of
Strauman et al.'s "overcontrolled" hypothesis, although one
would need to grapple with the lack of gender differences in
obsessive—compulsive disorder.
Although our findings were based on nonclinical samples,
they may have implications for the psychotherapy of bulimic
symptomatology (both cognitive—behavioral and interpersonal
therapies have received at least preliminary support as effective
treatments for bulimia; Fairburn, Marcus, & Wilson, 1993;
Fairburn et al., 1995). Levels of perfectionism should be as-
sessed (Hewitt et al.'s [1991] Multidimensional Perfectionism
Scale is a well-researched device) and treatment tailored accord-
ingly. For example, in the context of a cognitive-behavioral
approach, high levels of perfectionism should be an early and
constant focus, especially insofar as reanalysis of the National
Institute of Mental Health Treatment of Depression Collabora-
tive Research program suggests that perfectionism levels among
depressed people remain high despite treatment and that perfec-
tionism predicts low functioning at follow-up (Blatt, Quinlan,
Pilkonis, & Shea, 1995). From an interpersonal perspective, it
is important to note that perfectionism has been demonstrated
to affect quality of relationships generally (Hewitt, Flett, &
Mikail, in press), as well as the quality of therapeutic alliance
(Blatt ct al., 1995). The dimensions of socially prescribed per-
fectionism, as well as other-oriented perfectionism (expecting
others to be perfect), appear particularly relevant to interper-
sonal approaches.
In summary, a diathesis-stress model was proposed, wherein
general perfectionism was framed as a risk factor for bulimia,
but only when standards were perceived to have gone unmet.
 PERFECTIONISM, PERCEIVED WEIGHT, BULIMIC SYMPTOMS
Perceived weight status was argued to be more important than
actual weight status. Results from two large, nonclinical samples
of women were consistent with the model.
References
American Psychiatric Association. (1980). Diagnostic and statistical
manual of mental disorders (3rd ed.). Washington, DC: Author.
American Psychiatric Association. (1994). Diagnostic and statistical
manual of mental disorders (4th ed.). Washington, DC: Author.
Benkert, O., Wetzel, H., & Szegedi, A. (1993). Serotonin dysfunction
syndromes: A functional common denominator for classification of
depression, anxiety, and obsessive-compulsive disorders. International
Clinical Psychopharmacology, S(Suppl. 1), 3-14.
Blatt, S. J., Quinlan, D., Pilkonis, P., & Shea, M. T. (1995). Impact of
perfectionism and need for approval on the treatment of depression:
The National Institute of Mental Health treatment of depression col-
laborative research program revisited. Journal of Consulting and
Clinical Psychology, 63, 125-132.
Blouin, A. G., Bushnik, T, Braaten, J., & Blouin, J. H. (1989). Bulimia
and diabetes: Distinct psychosocial profiles. International Journal of
Eating Disorders, 8, 93-100.
Cohen, J., & Cohen, P. (1983). Applied multiple regression/correlation
analysis for the behavioral sciences (2nd ed.). Hillsdale, NJ: Erlbaum.
Fairburn, C. G., Marcus, M. D., & Wilson, G. T. (1993). Cognitive-
behavioral therapy for binge-eating and bulimia nervosa: A compre-
hensive treatment manual. In C. G. Fairburn & G. T. Wilson (Eds.),
Binge eating: Nature, assessment, and treatment (pp. 361-404). New
York: Guilford Press.
Fairbum, C. G., Norman, P. A., Welch, S. L., O'Connor, M. E., Doll,
H. A., & Peveler, R. C. (1995). A prospective study of outcome in
bulimia nervosa and the long-term effects of three psychological treat-
ments. Archives of General Psychiatry. 52. 304-312.
Flett, G. L., Hewitt, P. L., Blankstein, K. R., & Mosher, S. W. (1991).
Perfectionism, life events, and depression: Testing a diathesis-stress
model. Canadian Psychology, 32, 311.
Flett, G. L., Hewitt, PL., Blankstein, K. R., & O'Brien, S. (1991).
Perfectionism and learned resourcefulness in depression and self-es-
teem. Personality and Individual Differences, 12, 61 -68.
Frost, R. O., Heimberg, R. G., Holt, C. S., Mania, J. I, & Neubauer, A.
(1993). A comparison of two measures of perfectionism. Personality
and Individual Differences, 14, 119-126.
Gainer, D. M., Olmstead, M. P., &Polivy,J. (1983). The Eating Disorder
Inventory: A measure of cognitive-behavioral dimensions of anorexia
nervosa and bulimia. In P. L. Darby, P. E. Garfinkel, D. M. Garner, &
D. V. Coscina (Eds.), Anorexia nervosa: Recent developments in re-
search (pp. 173-184). New York: Alan R. Liss.
Heatherton, T. F., & flaumeister, R. F. (1991). Binge eating as escape
from self-awareness. Psychological Bulletin, 110, 86-108.
Heatherton, T. F., Nichols, P., Mahamedi, F., & Keel, P. (1995). Body
weight, dieting, and eating disorder symptoms among college students,
1982 to 1992. American Journal of Psychiatry. 152, 1623-1629.
Hewitt, PL., &Dyck, D. G. (1986). Perfectionism, stress, and vulnera-
bility to depression. Cognitive Therapy and Research, 10, 137-142.
Hewitt, P. L., & Flett, G. L. (199la). Dimensions of perfectionism in
unipolar depression. Journal of Abnormal Psychology, 100, 98-101.
Hewitt, PL., & Flett, G. L. (1991b). Perfectionism in the self and
social contexts: Conceptualization, assessment, and association with
psychopathology. Journal of Personality and Social Psychology, 60,
456-470.
Hewitt, P. L., & Flett, G. L. (1993). Dimensions of perfectionism, daily
stress, and depression: A test of the specific vulnerability hypothesis.
Journal of Abnormal Psychology, 102, 58—65.
153
Hewitt, P. L., Flett, G. L., & Mikail, S. F. (in press). Perfectionism and
relationship maladjustment in chronic pain patients and their spouses.
Journal of Family Psychology.
Hewitt, P. L., Flett, G. L., lurnbull-Donovan, W., & Mikail, S. F. (1991).
The Multidimensional Perfectionism Scale: Reliability, validity, and
psychometric properties in psychiatric samples. Psychological Assess-
ment: A Journal of Consulting and Clinical Psychology, .?, 464-468.
Higgins, E. T. (1987). Self-discrepancy: A theory relating self and af-
fect. Psychological Review, 94, 319-340.
Hudson, J. I., Laffer, P. S., & Pope, H. G., Jr. (1982). Bulimia related
to affective disorder by family history and response to the dexametha-
sone suppression test. American Journal of Psychiatry, 139, 685-
687.
Hurley, J. B., Palmer, R. L., & Stretch, D. (1990). The specificity of the
Eating Disorders Inventory: A reappraisal. International Journal of
Eating Disorders, 9, 419-424.
Joiner, T. E., Jr. (1995). The price of soliciting and receiving negative
feedback: Self-verification theory as a vulnerability to depression the-
ory. Journal of Abnormal Psychology, 104, 364-372.
Joiner, T. E. Jr., Metalsky, G. I., & Wonderlich, S. A. (1995). Bulimic
symptoms and the development of depressed symptoms: The moderat-
ing role of attributional style. Cognitive Therapy and Research, 19,
651-666.
Joiner. T. E., Jr., & Schmidt, N. B. (1995). Dimensions of perfectionism,
life stress, and depressed and anxious symptoms: Prospective support
for diathesis-stress but not specific vulnerability among male under-
graduates. Journal of Social and Clinical Psychology, 14, 165-183.
Kuczmarski, R. J. (1992). Prevalence of overweight and weight gain in
the United States. American Journal of Clinical Nutrition, 55, 495S-
502S.
Metalsky, G. I., Joiner, T. E., Jr., Hardin, T. S., & Abramson, L. Y.
(1993). Depressive reactions to failure in a naturalistic setting: A test
of the hopelessness and self-esteem theories of depression. Journal
of Abnormal Psychology, 102, 101-109.
Newton, T., Butler, N., & Slade, P. D. (1988). Denial of symptoms and
self-report in eating disorders. British Review of Bulimia and Anorexia
Nervosa, 2, 55-59.
Radke-Sharpe, N., Whitney-Saltiel, D., & Rodin, J. (1990). Fat distribu-
tion as a risk factor for weight and eating concerns. International
Journal of Eating Disorders, 9, 27—36.
Rosch, D. S., Crowther, J. H., & Graham, J. R. (1991). MMPI-derived
personality description and personality subtypes in an undergraduate
bulimic population. Psychology of Addictive Behaviors, 5, 15-22.
Rothenberg, A. (1990). Adolescence and eating disorder: The obsessive-
compulsive syndrome. Psychiatric Clinics of North America, 13, 469-
488.
Ruderman, A. (1986). Bulimia and irrational beliefs. Behaviour Re-
search and Therapy, 24, 193-197.
Strauman, T. J.. Vookles, J., Berenstein, V., Chaiken, S., & Higgins, E. T.
(1991). Self-discrepancies and vulnerability to body dissatisfaction
and disordered eating. Journal of Personality and Social Psychology,
61, 946-956.
Thompson, D. A., Berg, K. M., & Shatford, L. A. (1987). The heteroge-
neity of bulimic symptomatology: Cognitive and behavioral dimen-
sions. International Journal of Eating Disorders, 6. 215-234.
Zuckerman, D. M., Colby, A., Ware, N. C., & Lazerson, J. S. (1986).
The prevalence of bulimia among college students. American Journal
of Public Health, 76, 1135-1137.
Received February 7, 1996
Revision received June 21, 1996
Accepted June 21, 1996 •