Download as pdf or txt
Download as pdf or txt
You are on page 1of 18

JOURNAL OF PERSONALITY ASSESSMENT, 76(1), 90–106

Copyright © 2001, Lawrence Erlbaum Associates, Inc.

Body-Esteem Scale for


Adolescents and Adults

Beverley K. Mendelson
Department of Psychology
Concordia University

Morton J. Mendelson
Department of Psychology
McGill University

Donna R. White
Department of Psychology
Concordia University

Body esteem (BE) refers to self-evaluations of one’s body or appearance. This article
outlines a BE questionnaire for adolescents and adults that has 3 subscales: BE–Ap-
pearance (general feelings about appearance), BE–Weight (weight satisfaction), and
BE–Attribution (evaluations attributed to others about one’s body and appearance).
The subscales have high internal consistency and 3-month test–retest reliability. Fe-
males scored lower than males on BE–Weight and BE–Appearance. BE–Weight was
the only subscale uniquely related to weight, especially in females, with heavy indi-
viduals tending to be dissatisfied with their weight. BE–Appearance was the only
subscale that consistently predicted self-esteem. BE–Appearance and BE–Weight
covaried more with Neeman and Harter’s (1986) Appearance subscale than with
other self-esteem subscales; BE–Attribution covaried more with social self-esteem
subscales than did BE–Appearance and BE–Weight.

Body esteem (BE) refers to self-evaluations of one’s body or appearance and has
typically been conceptualized as a global construct, regardless of whether it has
been measured by self-report questionnaires (Gray, 1977; Harter, 1985, 1988; B. K.
Mendelson & White, 1985), open-ended interviews (Allon, 1979), or reactions to
body-related words (Secord & Jourard, 1953). Nonetheless, some researchers have
identified what may be termed domains of BE. For example, Tucker (1981) factor
BODY-ESTEEM SCALE 91

analyzed Secord and Jourard’s (1953) Body Cathexis Scale, which was assumed to
be unidimensional, and he identified the following domains: health and physical
fitness, face and overall appearance, subordinate and independent body features,
and physique and muscular strength. However, Tucker’s limited sample size raises
questions about the stability of the results. Franzoi and Shields (1984) developed
their own measure, which was not based on the assumption of unidimensionality;
they obtained evidence that, for males, the relevant domains are physical attractive-
ness, upper body strength, and physical condition, whereas for females the domains
are sexual attractiveness, weight concern, and physical condition. However, their
measure is not appropriate for young adolescents.
B. K. Mendelson, White, and Mendelson (1996–1997) corroborated that feel-
ings about one’s weight can be differentiated from feelings about one’s general
appearance, and they further suggested that one’s own opinions may be differen-
tiated from the opinions attributed to others. However, there were limitations in
their measure; it included a dichotomous response format, which, strictly speak-
ing, raises questions about factor analyzing the data. Therefore, an expanded in-
strument was developed to tap three factors: general feelings about appearance
(e.g., “I like what I see when I look in the mirror”), weight satisfaction (e.g., “I
really like what I weigh”), and evaluations attributed to others about one’s body
and appearance (e.g., “People my own age like my looks;” B. K. Mendelson,
White, & Balfour, 1995).
We are not claiming that these are the only BE domains. Nonetheless, BE mea-
sures do not always distinguish between self-evaluations of appearance and of
weight; for example, Neeman and Harter’s (1986) Appearance subscale has both
weight and appearance items. Moreover, the issue of others’ opinions seems espe-
cially relevant to adolescents.
The purpose of this article is to outline the evolution of a BE scale suitable for
adolescents and adults from its original form as a scale for children (B. K.
Mendelson & White, 1982). The original Body-Esteem Scale for Children (B. K.
Mendelson & White, 1993–1994) taps children’s affective evaluations of their
bodies and appearance; it comprises 24 yes–no items that all have obvious face va-
lidity (e.g., “I like what I look like in pictures”). The background and data on the
reliability and validity of the scale are outlined elsewhere (B. K. Mendelson &
White, 1993–1994), and the scale was used in two studies that demonstrated good
split-half reliability (B. K. Mendelson & White, 1982, 1985). Moreover, over-
weight children had lower BE than their normal-weight peers, and this was espe-
cially true for girls; this was consistent with earlier findings (Allon, 1979; Dwyer,
Feldman, Seltzer, & Mayer, 1969; Gray, 1977; Hendry & Gillies, 1978; Miller,
Coffman, & Linke, 1980; Stunkard & Mendelson, 1967).
In a subsequent study (B. K. Mendelson et al., 1996–1997), 4 items were
dropped from the original measure, and the 20-item scale was administered to a
large number of youngsters between 8 and 15 years old, some of whom were tested
92 MENDELSON, MENDELSON, WHITE

twice over 2 years. A factor analysis with oblique rotation was applied to the data,
and a three-factor solution was adopted after eliminating 2 items that did not ex-
ceed a loading of .5 on any of the factors. The analysis suggested that BE consists
of three factors: BE–Appearance (general feelings about appearance), represented
by 12 items; BE–Weight (weight satisfaction), represented by 3 items; and BE–At-
tribution (evaluations attributed to others about one’s body and appearance), rep-
resented by 3 items (see B. K. Mendelson & White, 1993–1994, for details).
However, improvements to the scale were still necessary. The yes–no re-
sponse format was originally adopted because it was suitable for children, but it
was changed to a 5-point Likert scale on which respondents rated their degree of
agreement with each statement. A 5-point scale is more robust than a dichoto-
mous scale and provides data that more fully meet the assumptions of factor
analysis. Although the BE–appearance subscale remained intact at 12 items,
both the BE–Weight and the BE–Attribution subscales were increased from 3
items to 9 items to improve their reliability. Finally, some of the items were re-
written to make them age appropriate for adolescents and adults.
The goals of this study were to test a 30-item version of the Body-Esteem Scale
for Adolescents and Adults (BESAA) with a large developmental sample, to
shorten the scale to between 20 and 24 items, and to verify the reliability and valid-
ity of the scale in a number of ways. The factor structure of the scale was examined
to confirm—in Tabachnik and Fidell’s (1989) use of the term—that the three a pri-
ori subscales applied to data from adolescents and young adults. We verified the
internal consistency of the subscales, and we examined their test–retest reliability
for a subsample of participants.
In addition, we examined how the different BE subscales relate to sex, age, and
weight. In line with previous research with different measures (B. K. Mendelson et
al., 1996–1997; M. J. Mendelson, Mendelson, & Andrews, 2000), we hypothe-
sized that females would have lower BE than males and that weight would
uniquely affect BE–Weight (weight satisfaction) more than the other subscales.
The Rosenberg Self-Esteem Scale (RSE; Rosenberg, 1979) or the Global
Self-Worth subscale (Neeman & Harter, 1986) was also administered to the partic-
ipants to enable a test of the convergent validity of the BE measures. Global
self-esteem seems to be highly related to unidimensional self-evaluations of physi-
cal appearance across the life span (Harter, 1993). We hypothesized that global
self-esteem would be more highly related to general feelings about appearance
than to weight satisfaction or to evaluations attributed to others about one’s body
and appearance.
Finally, the discriminant validity of the BE subscales was confirmed in a
subsample of participants by examining correlations between the BE measures
and a number of Neeman and Harter’s (1986) domain-specific subscales (ap-
pearance, close friendships, social acceptance, romantic relationships, and pa-
rental relationships). Neeman and Harter’s Appearance subscale includes two
BODY-ESTEEM SCALE 93

items relevant to weight and two relevant to general appearance, so BE–Appear-


ance and BE–Weight were expected to correlate more highly with it than with
the other self-esteem subscales. BE–Attribution focuses on others’ opinions of
one’s appearance, so it was expected to correlate less with Neeman and Harter’s
Appearance subscale than the other BE measures but to covary more with their
subscales relevant to social domains.

METHOD

Participants

The data analyzed here were part of data sets from other studies that addressed other
issues (Balfour, 1996; Buchholz, 1998; M. J. Mendelson et al., 2000).1 The partici-
pants were drawn from English-speaking elementary schools, high schools, univer-
sities, and a junior college in Montreal, Quebec. The total sample included 1,334
participants (763 females and 571 males) between the ages of 12 and 25 years (M =
16.8 years; see columns 5 and 9 of Table 1 for the number of females and males at
each age, respectively). A subsample of 97 junior-college students (61 women and
36 men) were retested 3 months after their initial test.

Measures

BESAA. The BESAA assesses participants’ attitudes and feelings about their
bodies and appearance. The original version of the scale contains 30 positively or
negatively worded items (e.g., “I like what I look like in pictures,” “My weight
makes me unhappy,” “My looks help me get dates”). Respondents indicate their de-
gree of agreement on a 5-point Likert scale ranging from 0 (never) to 4 (always),
and negative items are reverse scored.

Self-esteem. In two of the subsamples (Balfour, 1996; Buchholz, 1998),


self-esteem was measured with the 10-item RSE (Rosenberg, 1979). The items are
global statements (e.g., “On the whole, I am satisfied with myself”), and respon-
1The psychometric properties of the BESAA were reported in part by B. K. Mendelson, Mendelson,

Andrews, Balfour, and Buchholz (1997) at the meeting of the Society for Research in Child Develop-
ment. Otherwise, with minor exceptions, this article does not duplicate previously published informa-
tion. M. J. Mendelson et al. (2000) examined how BE variables were associated with weight, sex, and
self-esteem for a sample of junior-college students who were predominantly between 17 and 19 years
of age. However, they reported minimal information about scale validation and focused on the substan-
tive issue of the importance that individuals attribute to the BE domains.
94 MENDELSON, MENDELSON, WHITE

TABLE 1
Cronbach’s Alphas, Means, and Standard Deviations of Body-Esteem (BE)
Subscales for Male and Female Participants at Different Ages

Sex

Female Male

Subscale and Age α M SD n α M SD n

BE–Appearance
12–13 .94 2.2 1.0 42 .86 2.9 0.7 48
14 .93 2.1 0.9 85 .87 2.9 0.7 65
15 .93 2.0 1.0 110 .91 2.8 0.9 54
16 .92 2.0 0.8 98 .90 2.9 0.7 95
17 .91 2.1 0.8 92 .90 2.9 0.7 101
18 .92 2.3 0.8 115 .88 2.9 0.7 76
19 .93 2.3 0.9 88 .85 2.6 0.6 49
20–25 .91 2.5 0.7 133 .89 2.7 0.7 83
Total .93 2.2 0.9 763 .89 2.8 0.7 571
BE–Weight
12–13 .94 2.2 1.2 42 .89 3.2 0.8 48
14 .96 2.1 1.2 85 .87 3.1 0.8 65
15 .94 2.0 1.1 110 .89 2.6 0.9 54
16 .94 1.9 1.1 98 .87 3.0 0.8 95
17 .94 1.8 1.1 92 .88 3.0 0.8 101
18 .94 2.1 1.0 115 .90 2.7 0.9 76
19 .95 2.2 1.2 88 .91 2.7 0.9 49
20–25 .94 2.3 1.0 133 .86 2.7 0.8 83
Total .95 2.1 1.1 763 .87 2.9 0.8 571
BE–Attribution
12–13 .84 1.8 0.9 42 .81 2.4 0.8 48
14 .78 1.9 0.8 85 .77 2.1 0.9 65
15 .84 2.0 0.8 110 .88 2.1 1.1 54
16 .76 2.1 0.7 98 .77 2.5 0.8 95
17 .83 2.2 0.8 92 .75 2.4 0.7 101
18 .74 2.2 0.7 115 .83 2.4 0.7 76
19 .85 2.2 0.8 88 .88 2.4 0.7 49
20–25 .75 2.5 0.7 133 .85 2.4 0.8 83
Total .81 2.1 0.8 763 .81 2.3 0.8 571

dents indicate their level of agreement on a 4-point scale (1 = strongly disagree, 2 =


disagree, 3 = agree, 4 = strongly agree). Negative items are reverse scored. Like
scores for all the measures reported in this article, the RSE score was computed as
the mean of the items. For the data here, internal consistency was very good
(Cronbach’s α = .89).
In one of the subsamples (M. J. Mendelson et al., 2000), we used the Self-Per-
ception Profile for College Students (Neeman & Harter, 1986) to measure global
self-worth (SE–Global) and self-evaluations in five domains: appearance (SE–Ap-
BODY-ESTEEM SCALE 95

pearance), close friendships (SE–Friend), social acceptance (SE–Social), roman-


tic relationships (SE–Romantic), and parent relationships (SE–Parent). The
SE–Global scale includes six items, and each subscale has four items. The self-es-
teem items consist of two contrasting statements (e.g., “Some students like the
kind of person they are” and “Other students wish that they were different”). Re-
spondents are asked first to decide which statement pertains to them and then to in-
dicate whether the choice is really true or sort of true. Each item is then scored on a
4-point scale (1 = really negative, 2 = sort of negative, 3 = sort of positive, 4 = re-
ally positive). For this sample, Cronbach’s alpha coefficients indicated satisfac-
tory to good internal consistency for each of the self-esteem measures (SE–Global
α = .88, SE–Appearance α = .84, SE–Friend α = .76, SE–Social α = .73, SE–Ro-
mantic α = .83, SE–Parent α = .81).

Body build. The participants’ heights and weights were measured at the
end of each study. However, that was not convenient for 107 older adolescents,
who nonetheless self-reported their height and weight (cf. Stunkard & Albaum,
1981). An analysis of data from a similar sample indicated that self-report mea-
sures of height and weight are very highly correlated with actual measures (M. J.
Mendelson et al., 2000).
In one subsample (Balfour, 1996), the height and weight data were coded as rel-
ative weight (i.e., 100 × weight/expected weight for sex for height for age), which
provides a height-corrected percentage measure of actual to expected weight. In
the other subsamples, we computed for each participant body mass index (BMI), a
height-corrected measure of body build, as BMI = W/H2, where W refers to weight
(in kilograms) and H2 refers to height (in meters) squared (Lee, Kolonel, & Hinds,
1981). To combine the data across subsamples, it was necessary to derive a single
measure of body build. Therefore, relative weight, actual BMI, and self-reported
BMI were standardized within sex, and the standardized measures were used as the
height-corrected measure of weight.

RESULTS

BESAA: Factor Analysis

We used a factor analysis with oblique rotation to determine if the BESAA tapped
more than one construct.2 We used oblique rotation because previous work has
shown that aspects of BE were intercorrelated (B. K. Mendelson et al., 1996–1997).
2Of the 1,334 participants, 1,227 provided complete data for the original factor analysis of 30 items.

However, data from additional participants could be used when items were dropped for subsequent
analyses. The final factor analysis was applied to the data from 1,240 participants who answered all 23
of the relevant items. The Cronbach’s alphas, reported next, were based on varying numbers of partici-
pants who answered all the items retained for each factor (NBE–Appearance = 1,308, NBE–Weight = 1,312,
NBE–Attribution = 1,283).
96 MENDELSON, MENDELSON, WHITE

Five items were dropped because of problematic wording related to gender (e.g., “I
can wear clothes that show my figure”) or to direction of weight (e.g., “I wish I were
thinner”). The analysis yielded a four-factor solution, but it was rejected because
the fourth factor contained only two items. An analysis recomputed without the two
items yielded the three-factor solution shown in Table 2, which shows the factor
loadings (> .40) for the retained items.
The first factor, BE–Appearance (general feelings about appearance), consisted
of 10 items and accounted for 49.3% of the variance. The mean of the items was
computed to form a BE–Appearance score, which had very high internal consis-
tency: Cronbach’s α = .92, item–total rs(1,306) = .56–.79 (M = .71), ps < .01.
The second factor, BE–Attribution (evaluations attributed to others about one’s
body and appearance), consisted of five items and accounted for 10.4% of the vari-
ance. The mean of the items was computed to form a BE–Attribution score, which

TABLE 2
Factor Loadings (> .40) for Body-Esteem (BE) Items

Subscale and Item No. Factor 1 Factor 2 Factor 3

BE–Appearancea
11 .86
7 .80
9 .77
13 .76
17 .71
21 .69
6 .51
23 .50
15 .50
1 .42
BE–Attributionb
2 .83
20 .77
12 .74
5 .64
14 .61
BE–Weightc
8 .96
10 .92
16 .89
19 .77
4 .73
18 .69
22 .61
3 .58
a10 items. α = .92. b5 items. α = .81. c8 items. α = .94.
BODY-ESTEEM SCALE 97

had good internal consistency: Cronbach’s α = .81, item–total rs(1,281) = .48–.72


(M = .62), ps < .01.
The third factor, BE–Weight (weight satisfaction), consisted of eight items and
accounted for 5.9% of the variance. The mean of the items was computed to form a
BE–Weight score, which had very high internal consistency: Cronbach’s α = .94,
item–total rs(1,310) = .72–.85 (M = .78), ps < .01.

Norms

Table 1 presents the Cronbach’s alphas, means, and standard deviations for each
BE subscale for males and females at each age. The Cronbach’s alphas ranged from
.75 to .96, indicating acceptable to excellent internal consistency for the three BE
measures for both sexes at all ages.
Given that the subscale scores are averages of the appropriate items, they range
from 0 (never) to 4 (always), with higher values indicating more positive BE on a
particular dimension. Overall, the means for females tended to be slightly above 2
(sometimes), and the means for males slightly below 3 (often).

Effects of Age, Sex, and Weight

The next analyses examined the degree to which the BE measures were predicted
by main effects and interactions of sex, age, and weight. We subjected each BE
subscale to a separate hierarchical multiple regression. The three main effects were
entered as a set into the regression equation first, followed by the three two-way in-
teraction terms (Age × Sex, Age × Weight, and Sex × Weight) and then by the Age ×
Sex × Weight interaction term. Two rules were adopted for accepting a predictor:
The F for inclusion had to be significant (p < .01), and the squared part correlation
between the predictor and the criterion variable had to equal at least .035 (i.e., re-
moving the variable would result in a reduction in an R2 of at least 3.5%). The sec-
ond rule was necessary because the large sample size meant that variables or
interaction terms accounting for even less than 1% of the variance in a criterion
variable were still significant.
Table 3 shows the regression equations of the three BE subscales as well as the
zero-order correlations of each subscale with age, sex, and weight. Age did not ac-
count for variance in any of the BE measures. Sex reached the inclusion criteria for
both BE–Appearance and BE–Weight. As shown in Table 1, females tended to
have lower BE scores than males, which validates the measures. The effect of sex
on BE–Attribution reached a traditional significance level (p < .01), but the effect
size was too small to warrant consideration. It is noteworthy that the Sex × Age in-
teraction did not reach significance for any of the variables (ps > .05); examination
98 MENDELSON, MENDELSON, WHITE

TABLE 3
Regression Equations of Body-Esteem (BE) Subscales on
Main Effects and Interactions of Age, Sex, and Weight

Subscale, Step, and Variable r B SE B ∆ R2 R2

BE–Appearance
Step 1
Age .03 .02 .01
Sex .37* .33* .02
Weight –.21* –.18* .02 .18*
Step 2
Sex × Weight .07 .02
Sex × Age –.05 .01
Age × Weight .02 .01 .03
Step 3
Age × Sex × Weight .01 .01 .00 .21*
BE–Weight
Step 1
Age –.03 .01 .01
Sex .37* .41* .03
Weight –.37* –.40* .03 .28*
Step 2
Sex × Weight .22* .03
Sex × Age –.05 .01
Age × Weight .03 .01 .06*
Step 3
Age × Sex × Weight .00 .01 .00 .34*
BE–Attribution
Step 1
Age .15 .05 .01
Sex .13 .11 .02
Weight –.15 –.13 .02 .07*
Step 2
Sex × Weight .09 .02
Sex × Age –.03 .01
Age × Weight .00 .01 .02
Step 3
Age × Sex × Weight .00 .01 .00 .09

Note. Only predictors accounting for at least 3.5% of the variance in the criterion variable are
designated as significant.
*p < .001.

of the means in Table 1 for each variable across the eight age groups reveals a rela-
tively constant sex difference from early adolescence to young adulthood.
Weight also reached the inclusion criteria for both BE–Appearance and
BE–Weight, and again, the effect of weight on BE–Attribution reached signifi-
cance (p < .01), but the effect size was small. Heavier individuals tended to have
relatively low BE. The main effect of weight on BE–Weight was qualified by
BODY-ESTEEM SCALE 99

the significant Sex × Weight interaction. We computed follow-up partial corre-


lations between BE–Weight and weight, controlling for age. For males, the cor-
relation between BE–Weight and weight, although significant, was small,
rpartial(506) = –.16, p < .001; indeed, weight accounted for less than 3% of the
variance in BE–Weight. In contrast, for females, weight and BE–Weight were
highly correlated, rpartial(667) = –.54, p < .001; indeed, weight accounted for
more than 10 times as much variance in females’ BE–Weight as it did in males’
BE–Weight. Independent of age, heavy females tended to be dissatisfied with
their weight.
Although weight inversely predicted the three BE subscales (ps < .001), it ac-
counted for more variance in BE–Weight than in BE–Appearance or BE–Attri-
bution. A question remained, however, whether the variance in the three BE
subscales accounted for by weight was shared or independent. Therefore, with
further analyses, we examined the partial correlation between each BE subscale
and weight, controlling for the other two subscales. Because of the significant
Sex × Weight interaction for BE–Weight, we computed the partial correlations
separately for males and females. For males, weight accounted for only a mini-
mal amount of unique variance in each BE subscale (BE–Weight = 1.1%,
BE–Appearance = 0.2%, BE–Attribution = 0.2%). For females, in contrast,
weight accounted for fully 25.5% of unique variance in BE–Weight, whereas it
accounted for only 5.3% in BE–Appearance and only 2.1% in BE–Attribution.
These results support the construct and convergent validity of the measures.

Test–Retest Reliability

The BESAA was given to a subsample 3 months after their initial test. The test–re-
test correlations were high: BE–Appearance r(95) = .89, p < .001; BE–Weight
r(95) = . 92, p < .001; and BE–Attribution r(95) = . 83, p < .001, which supports the
reliability of the measures.

BE and Self-Esteem

To further examine the validity of the BE measures, we computed partial correla-


tions between each BE subscale and either the RSE or SE–Global, partialing out the
other BE subscales (Table 4). For both males and females, at almost every age,
BE–Appearance was correlated with global self-esteem independent of the other
two BE subscales, whereas the same was not consistently true of BE–Weight and
BE–Attribution.
Table 4 also shows the means and standard deviations of the correlation coeffi-
cients for males and females for each BE subscale as well as the results of t tests
that were computed to compare the mean correlations with zero. On average, the
100 MENDELSON, MENDELSON, WHITE

correlations between self-esteem and BE–Appearance were significantly positive,


whereas the correlations between self-esteem and the other two BE measures did
not differ significantly from zero; that is, males and females who highly evaluated
their appearance tended to have high self-esteem.
We conducted analyses to verify the discriminant validity of the BE mea-
sures. We computed correlations for each sex among the three BE measures and
the five domain-specific self-esteem subscales, partialing out SE–Global (i.e.,
controlling for a general effect of overall self-esteem; see Table 5). As expected,
BE–Appearance and BE–Weight covaried more highly with SE–Appearance
than with the other self-esteem subscales for both females and males. Also as

TABLE 4
Partial Correlations Between Each BE Subscale With an SE Measure,
Controlling for Other BE Subscales, for Male and Female Participants

Sex and Age SE Measure df BE–Appearance BE–Weight BE–Attribution

Male
13 RSE 43 .31* –.04 .09
14 RSE 61 .38** .09 .04
15 RSE 50 .39** .05 .06
16 RSE 89 .44** .11 .06
17 RSE 68 .28* .35** .42**
17 SE–Global 28 .52** .03 .03
18 SE–Global 36 .61** .09 –.05
19 SE–Global 18 .71** –.16 –.32
20–25 RSE 50 –.17 .07 –.08
Mr .39 .07 .03
SD r .25 .14 .19
t(8) 4.62** 1.45 0.43
Female
13 RSE 39 .63** .36* –.06
14 RSE 82 .47** –.04 .30**
15 RSE 105 .54** –.02 –.01
16 RSE 91 .51** .23* –.06
17 RSE 51 .53** .00 .11
17 SE–Global 27 .43* .18 .08
18 SE–Global 44 .69** –.19 .12
19 SE–Global 23 .76** .09 –.38
19 RSE 18 –.21 .32 .11
20–25 RSE 94 –.17 –.03 .19
Mr .39 .06 .05
SD r .35 .16 .19
t(9) 3.41** 1.13 0.80

Note. Partial correlations were negative in some cases, but the zero-order correlations were all
positive. BE = body esteem; SE = self-esteem; RSE = Rosenburg Self-Esteem Scale.
*p < .05. **p < .01.
BODY-ESTEEM SCALE 101

TABLE 5
Partial Correlations Between BE Measures and Domain-Specific
SE Subscales, Controlling for Global SE, for Each Sex

Sex and Measure SE–Appearance SE–Friend SE–Social SE–Romantic SE–Parent

Femalea
BE–Appearance .49** .08 .30** .23* .12
BE–Weight .79** –.14 –.06 .07 –.04
BE–Attribution .21* .21* .27** .45** –.07
Maleb
BE–Appearance .42** .07 .14 .13 .07
BE–Weight .58** –.06 .10 .14 –.01
BE–Attribution .10 .11 .24* .24* .03

Note. BE = body-esteem; SE = self-esteem.


an = 110. bn = 107.

*p < .05. **p < .01.

expected, BE–Attribution correlated less with SE–Appearance than did BE–Ap-


pearance and BE–Weight; moreover, BE–Attribution correlated more highly
with social self-esteem subscales than did BE–Appearance and BE–Weight. The
data provide evidence for the discriminant validity of the BE measures.

DISCUSSION

The 23-item BESAA is an easy-to-administer, psychometrically sound instrument


that taps three aspects of BE in adolescents and adults: (a) general feelings about
appearance, (b) weight satisfaction, and (c) attributions of positive evaluations
about one’s body and appearance to others. As demonstrated in this article, the
scale is valid and reliable over a wide age range, so it can be given confidently to
children as young as 12 and to individuals well into adulthood.
J. K. Thompson, Heinberg, Altabe, and Tantleff-Dunn (1999) extensively out-
lined various definitions and measures of body image. Techniques to assess body
image have burgeoned in recent years. The most widely used measures are those
related to satisfaction, esteem, appearance evaluation, distress, and anxiety. It is
clear that investigators who study what falls under the general heading of body im-
age can choose from a wide variety of measures. The choice of which one to use
should be based most importantly on the specific construct of interest as well as on
practical considerations.
One approach uses silhouettes of different sizes, from underweight to over-
weight, to tap the discrepancy between people’s conception of their current figure
and their “ideal” figure (e.g., M. A. Thompson & Gray, 1995) or between their
conception of their current figure and their “affective” evaluation of it (i.e., how
102 MENDELSON, MENDELSON, WHITE

they feel they look; J. K. Thompson, 1990). Although this type of assessment ap-
pears to be quite concrete, it requires respondents to make sophisticated perceptual
or evaluative judgments of their bodies, something that may be difficult for adoles-
cents whose bodies are changing significantly.
Another approach involves interview methods, developed for the field of eat-
ing disorders, that may help our understanding of body image disturbance (e.g.,
Fairburn & Cooper, 1993). However, individuals must be well trained in admin-
istering the interviews, and interviews are impractical for large-scale data
collection.
A more common approach uses questionnaires, usually to measure some
combination of satisfaction, evaluation, and anxiety related to one’s body (e.g.,
Garner, 1991). Some advantages to questionnaires, which apply to the scale pre-
sented in this article, are that they can be administered with minimum difficulty,
that they are practical for large samples or multiple testing, and as a result, that
they can be tested psychometrically and modified if necessary. Nonetheless,
questionnaires must be screened for length, especially if they are to be used with
other instruments, and for the specific construct being measured. The question-
naire presented here taps three distinct constructs with short subscales that can
be used separately or in combination.
Researchers have recently adopted other approaches to assess body image.
Investigators have increasingly focused on sociocultural issues, such as the in-
fluence of the media and society on one’s perception of body image (Cusumano
& Thompson, 1997; Delaney, O’Keefe, & Skene, 1997). Others have examined
social–cognitive constructs, such as the degree of attention one pays to body
shape (Beebe, 1995), the importance of BE (M. J. Mendelson et al., 2000), so-
cial comparison (e.g., J. K. Thompson, Heinberg, & Tantleff, 1991), and feed-
back from others (Tantleff-Dunn, Thompson, & Dunn, 1995). Such factors are
particularly interesting because they may mediate or moderate body image ef-
fects. For example, in a previous study, high BMI predicted low global self-es-
teem and low self-evaluations of appearance but only for individuals who
regarded weight as important, not for those who regarded it as relatively unim-
portant (M. J. Mendelson et al., 2000).
Because of the number of measures available to assess aspects of body im-
age, the question may arise as to the need for another measure. The major contri-
bution of the scale presented in this article is that it differentiates feelings about
appearance from feelings about weight. Many measures focus too narrowly on
weight as the primary determinant of BE and, ultimately, of self-esteem. Our
measure also expands on typical self-evaluations of the body by assessing the
degree to which individuals attribute positive evaluations of their bodies and ap-
pearances to others.
There was good evidence for the discriminant validity of the three BE
subscales. The factor analysis differentiated them, and the BE–Appearance and
BODY-ESTEEM SCALE 103

BE–Weight subscales were more closely related to Neeman and Harter’s (1986)
Appearance subscale than was the BE–Attribution subscale, whereas the BE–At-
tribution subscale correlated more highly with Neeman and Harter’s social
subscales. Thus, the BE–Attribution subscale is distinct from self-evaluations of
weight and appearance. The attributional aspect of BE may be critical in body
image research, because how people look to others, or at least how people as-
sume that they look to others, may help form their opinions about themselves
(see J. K. Thompson et al., 1999, for a complete review of this topic).
Females had lower scores on BE–Appearance and BE–Weight than did males,
and the difference appeared to be constant across the age range. This sex difference
in BE corroborates earlier findings (Feingold & Mazzella, 1996; Harter, 1985; B. K.
Mendelson & White, 1985; B. K. Mendelson et al., 1996–1997; M. J. Mendelson et
al., 2000) and thereby supports the construct validity of the measures.
It is particularly disturbing that the difference in BE between males and fe-
males has apparently widened over the last 50 years, especially in adolescents
(Feingold & Mazzella, 1996). Starting early in life, girls in North America may
incorporate unattainable stereotypes about an ideal appearance and weight and
may consequently develop a general dislike of their body. Indeed, our previous
research has suggested that the sex difference in self-evaluations of appearance
emerges between 8 and 13 years of age and is attributable to a decline in
self-evaluations over that period for girls but not for boys (B. K. Mendelson et
al., 1996–1997).
Neither age nor interactions with age accounted for variance in any of the BE
subscales. We previously reported limited age differences in self-evaluations of
weight and appearance within childhood as well as age differences in self-evalua-
tions of appearance from childhood to adolescence and within early adolescence
(B. K. Mendelson et al., 1996–1997). However, the measure in that study used a
dichotomous response format, which could have exaggerated what may be small,
inconsequential differences in BE during adolescence. The lack of developmental
differences in the sample in this study may mean that the aspects of BE we mea-
sured are developed by 12 or 13 years of age and are stable for a decade beyond
that. Of course, such a conclusion would be best supported by longitudinal rather
than cross-sectional data.
Further validation studies of our measure should include older adults. BE
may change in women beyond their early 20s, a time when other developmental
tasks, such as pregnancy and childbirth, put the body in a different perspective.
As well, men may evaluate themselves differently when, for example, aging re-
duces their advantage in strength. In addition, work on the importance of the BE
subscales, which has been conducted with 17- to 19-year-olds (M. J. Mendelson
et al., 2000), should be extended to younger age groups to determine when the
aspects tapped by the subscales become salient to adolescents. Moreover, the
measure needs to be tested with clinical populations, such as individuals suffer-
104 MENDELSON, MENDELSON, WHITE

ing from eating disorders, for which the BE–Attribution subscale may be partic-
ularly important.
The results of this study also clearly demonstrate the unique relation between
global self-esteem and feelings about appearance, independent of weight satisfac-
tion and attributions of positive evaluations to others. Most important, the results
reinforce the conclusion that, regardless of actual weight, global self-esteem is
uniquely related to positive feelings about appearance (cf. B. K. Mendelson et al.,
1996–1997). The clinical implication of this finding is that perhaps the best route
to good self-esteem is through improving feelings about appearance. Rather than
focusing on changes in weight satisfaction through dieting, which has high failure
and recidivism rates (cf. Brownell & Rodin, 1994), the emphasis might better be
placed on appearance-altering strategies such as changing personal grooming, in-
corporating aerobic exercise into daily routines, and changing notions about ideal
weight to accommodate a healthy weight for body type and age. This might insu-
late youngsters, particularly females, from the prevailing norm of the “body beau-
tiful” as the cornerstone of global self-worth.

ACKNOWLEDGMENTS

This research was supported by Les Fonds pour la Formation des Chercheurs et
l’Aide à la Recherche, Gouvernement du Québec.
We thank Louise Balfour and Annick Buchholz, who graciously provided us
the use of their data relevant to our measure.
Donna White died on May 8, 1999, shortly before this article was initially sub-
mitted. She will be remembered by many as a fine teacher, colleague, and friend,
and she will be missed by all who knew her.

REFERENCES

Allon, N. (1979). Self-perceptions of the stigma of overweight in relationship to weight losing patterns.
American Journal of Clinical Nutrition, 32, 470–480.
Balfour, L. (1996). Examining how avoidant coping and anger suppression relate to emotional eating in
young women. Unpublished doctoral dissertation, Concordia University, Montreal, Canada.
Beebe, D. W. (1995). The Attention to Body Shape Scale: A new measure of body focus. Journal of Per-
sonality Assessment, 65, 486–501.
Brownell, K. D., & Rodin, J. (1994). The dieting maelstrom: Is it possible and advisable to lose weight?
American Psychologist, 49, 781–791.
Buchholz, A. (1998). The role of female emotional development in relation to disordered eating and
binge eating in adolescent girls. Unpublished doctoral dissertation, Concordia University, Mon-
treal, Canada.
Cusumano, D. L., & Thompson, J. K. (1997). Body image and body shape ideals in magazines: Expo-
sure, awareness, and internalization. Sex Roles, 37, 701–721.
BODY-ESTEEM SCALE 105

Delaney, M. E., O’Keefe, L. D., & Skene, K. M. L. (1997). Development of a sociocultural measure of
young women’s experiences with body weight and shape. Journal of Personality Assessment, 69,
63–80.
Dwyer, J. R., Feldman, J. J., Seltzer, C. C., & Mayer, J. (1969). Body image in adolescents: Attitudes to-
ward weight and perception of appearances. Journal of Nutrition Education, 1, 14–19.
Fairburn, C. G., & Cooper, Z. (1993). The Eating Disorder Examination (12th ed.). In C. G. Fairburn & G.
T. Wilson (Eds.), Binge eating: Nature, assessment, and treatment (pp. 3–14). New York: Guilford.
Feingold, A., & Mazzella, R. (1996). Gender differences in body image are increasing. Psychological
Science, 9, 191–195.
Franzoi, S. L., & Shields, S. A. (1984). The Body-Esteem Scale: Multidimensional structure and sex dif-
ferences in a college population. Journal of Personality Assessment, 48, 173–178.
Garner, D. M. (1991). Eating Disorder Inventory–2: Professional manual. Odessa, FL: Psychological
Assessment Resources.
Gray, H. (1977). Social aspects of body image: Perception of normalcy of weight and affect of college
undergraduates. Perceptual & Motor Skills, 45, 1035–1040.
Harter, S. (1985). Manual for the Self-Perception Profile for Children. Denver, CO: University of Den-
ver Press.
Harter, S. (1988). Manual for the Self-Perception Profile for Adolescents. Denver, CO: University of
Denver Press.
Harter, S. (1993). Causes and consequences of low self-esteem in children and adolescents. In R. F.
Baumeister (Ed.), Self-esteem: The puzzle of low self-regard (pp. 87–116). New York: Plenum.
Hendry, L. B., & Gillies, P. (1978). Body type, body-esteem, school and leisure: A study of overweight,
average and underweight adolescents. Journal of Youth and Adolescence, 7, 181–195.
Lee, J., Kolonel, L. N., & Hinds, M. W. (1981). Relative merits of the weight-corrected-for-height indi-
ces. American Journal of Clinical Nutrition, 34, 2521–2529.
Mendelson, B. K., Mendelson, M. J., Andrews, J., Balfour, L., & Buchholz, A. (1997, April). Three aspects
of body esteem in adolescents and young adults: Differential relations with weight and self-esteem.
Poster presented at the meeting of the Society for Research in Child Development, Washington, DC.
Mendelson, B. K., & White, D. R. (1982). Relation between body-esteem and self-esteem of obese and
normal children. Perceptual & Motor Skills, 54, 899–905.
Mendelson, B. K., & White, D. R. (1985). Development of self–body-esteem in overweight youngsters.
Developmental Psychology, 21, 90–96.
Mendelson, B. K., & White, D. R. (1993–1994). Manual for the Body-Esteem Scale for Children. Con-
cordia University Research Bulletin, 12(2), 1–10.
Mendelson, B. K., White, D. R., & Balfour, L. (1995). The Body-Esteem Scale for Adolescents and
Adults. Unpublished manuscript, Concordia University, Montreal, Canada.
Mendelson, B. K., White, D. R., & Mendelson, M. J. (1996–1997). Self-esteem and body-esteem: Ef-
fects of sex, age, and weight. Journal of Applied Developmental Psychology, 17, 321–346.
Mendelson, M. J., Mendelson, B. K., & Andrews, J. (2000). Self-esteem, body esteem, and body-mass
in late adolescence: Is a Competence × Importance model needed? Journal of Applied Developmen-
tal Psychology, 21, 249–266.
Miller, T. M., Coffman, J. G., & Linke, R. A. (1980). Survey on body image, weight and diet of college
students. Journal of the American Dietetic Association, 77, 561–566.
Neeman, J., & Harter, S. (1986). Manual for the Self-Perception Profile for College Students. Denver,
CO: University of Denver Press.
Rosenberg, M. (1979). Conceiving the self. New York: Basic Books.
Secord, P. F., & Jourard, S. M. (1953). The appraisal of body-cathexis: Body cathexis and self. Journal
of Consulting Psychology, 17, 343–347.
Stunkard, A. J., & Albaum, J. M. (1981). The accuracy of self-reported weights. American Journal of
Clinical Nutrition, 34, 1593–1599.
106 MENDELSON, MENDELSON, WHITE

Stunkard, A., & Mendelson, M. (1967). Obesity and body image: I. Characteristics of disturbances in the
body image of some obese persons. American Journal of Psychiatry, 123, 1296–1300.
Tabachnik, B. G., & Fidell, L. S. (1989). Using multivariate statistics. New York: Harper & Row.
Tantleff-Dunn, S., Thompson, J. K., & Dunn, M. E. (1995). The Feedback on Physical Appearance
Scale (FOPAS): Questionnaire development and psychometric evaluation. Eating Disorders: The
Journal of Treatment and Prevention, 3, 332–341.
Thompson, J. K. (1990). Body image disturbance: Assessment and treatment. Elmsford, NY: Pergamon.
Thompson, J. K., Heinberg, L. J., Altabe, M., & Tantleff-Dunn, S. (1999). Exacting beauty. Washing-
ton, DC: American Psychological Association.
Thompson, J. K., Heinberg, L. J., & Tantleff, S. (1991). The Physical Appearance Comparison Scale
(PACS). Behavior Therapist, 14, 174.
Thompson, M. A., & Gray, J. J. (1995). Development and validation of a new body-image assessment
scale. Journal of Personality Assessment, 64, 263.
Tucker, L. A. (1981). Internal structure, factor satisfaction, and reliability of the Body Cathexis Scale.
Perceptual & Motor Skills, 53, 891–896.

Morton J. Mendelson
Department of Psychology
McGill University
1205 Doctor Penfield Avenue
Montreal, Quebec H3A 1B1
Canada
E-mail: mmendelson@psych.mcgill.ca

Received June 2, 1999


Revised May 30, 2000

You might also like