Jouroal of Personality and Social Psychology
1998, Vol. 75, No. 5, 1300-1320
Copyright 1998 by the American Psychological Association, Inc.
0022-3514/98/$3.00
Bright, Bad, Babyfaced Boys: Appearance Stereotypes Do Not Always
Yield Self-Fulfilling Prophecy Effects
Leslie A. Zebrowitz and Carrie Andreoletfi
Mary Ann Collins
Brandeis University
Spring Hill College
So Young Lee and Jeremy Blumenthal
Brandeis University
Three studies tested the hypothesisthat babyfaced adolescentboys would compensatefor the undesirable expectation that they will exhibit childlike traits by behaving contrary to it. Studies 1 and 2
revealed that babyfaced boys from middle- and lower class samples, including a sample of delinquents, showed higher academic achievementthan their matnre-faced peers, refuting the stereotype
of babyfaced people as intellectuallyweak. In the lower class samples, this compensationeffect was
moderated by IQ and socioeconomicstatus (SES), variables that influence the ability to overcome
low expectations. Study 3 showed that babyfacenessalso can produce negativecompensatorybehaviors. Low-SES babyfaced boys were more likely than their mature-faced peers to be delinquent,and
babyfaced delinquentscommitted more crimes, refuting the stereotype of babyfaced people as warm,
submissive, and physically weak.
Considerable research has shown a consensus in identifying
babyfaced individuals at all ages as well as a strong expectation
that babyfaced people will have childlike traits--submissive,
dependent, warm, affectionate, honest, weak, and naive (e.g.,
McArthur & Apatow, 1983-1984; Berry & McArthur, 1985;
Zebrowitz & Montepare, 1992; Zebrowitz, Montepare, & Lee,
1993; see also Montepare & Zebrowitz, 1998, for a review).
Although this babyface stereotype has been well documented,
the question of its accuracy remains. Because considerable research has shown that an individual's social environment can be
strongly influenced by his or her facial appearance (for pertinent
reviews, see Langlois, Kalakanis, Rubenstein, Larson, Hallain, & Smoot, 1996; Montepare & Zebrowitz, 1998), many
Leslie A. Zebrowitz, Carrie Andreoletti, So YoungLee, and Jeremy
Blumenthal,Department of Psychology,Brandeis University;Mary Ann
Collins, Department of Psychology, Spring Hill College.
So YoungLee is now at the Departmentof Psychology,YonseiUniversity, Seoul, South Korea. Jeremy Blumenthal is now at the Department
of Psychology, Harvard University.
This research was supported in part by National Institute of Mental
Health Grant BSR MH42684 and by a Yonsei Universitypostdoctoral
fellowship to So YoungLee. We express appreciationto the Institute of
Human Developmentat the Universityof California, Berkeley, for providing access to the IntergenerationalStudies data archives used in Study
1, and to the HenryA. Murray Center at Radcliffe College and the InterUniversityConsortium for Political and Social Research for providing
access to the Crime Causation Study data archives used in Studies 2
and 3. We also thank John Laub for his help in negotiating the Crime
Causation Study archives.
Correspondence concerningthis article should be addressed to Leslie
A. Zebrowitz, Departmentof Psychology,MS 062, Brandeis University,
Waltham, Massachusetts 02454-9110. Electronic mail may be sent to
zebrowitz@brandeis.edu.
investigators have suggested that these environmental effects
might contribute to stereotype accuracy by producing self-fulfilling prophecy effects (Adams, 1977; Jussim, Eccles, & Madon, 1996; Langlois, 1986; Langlois et al., 1996; Snyder,
Tanke, & Berscheid, 1977; Sorrell & Nowak, 1981). However,
the elusiveness of such effects in the real world suggests that
this notion may be too simplistic.
Self-fulfilling prophecies are not so readily obtained as some
have assumed because several factors in real-life social interactions work against the behavioral confirmation of others' expectations (cf. Snyder, 1992). Specifically, behavioral confirmation
may fail to occur because targets engage in self-verification,
behaving in a manner that provides corrective feedback to erroneous expectations (e.g,, S w a n n & Ely, 1984); because targets
show compensatory behavior, behaving in a manner that is opposite to disagreeable expectations (e.g., Miller & Myers, 1998;
Miller, Rothblum, Felicio, & Brand, 1995; Miller & Rudiger,
1997); or because perceivers show compensatory behavior, behaving in ways that elicit behaviors opposite to those they expect
(e.g., Bond, 1972). The latter two processes may foster selfdefeating prophecies whereby the expectancy is disconfirmed
because it initiates a chain of events that cause contradictory
behavior. Finally, even when the expectancy is confirmed, this
may reflect the accuracy of the initial expectation rather than a
self-fulfilling prophecy effect (Jussim, 1991).
The expectations elicited by a babyface may be particularly
likely to yield compensatory, disconfirming behaviors during
adolescence. As adolescents attempt to negotiate the transition
from childhood to adulthood, their primary interest is the development of autonomy. This leads them to place high value on
qualities perceived to be particularly lacking in babyfaced adolescents, such as power and competence (e.g., Adams, Montemayor, & Gullotta, 1996). However, the autonomy that adolescents seek is difficult to attain in modem societies because of
1300
1301
BRIGHT, BAD, BABYFACED BOYS
the gap between biological and social maturity. Moffitt (1993)
has argued that this maturity gap may lead adolescents to act
out in the attempt to gain adult privileges (e.g., sex, parenthood,
drinking, ownership of material goods). Because babyfaced
people are perceived as dependent and weak, babyfaced adolescents may experience the maturity gap to a greater degree than
their mature-faced peers, resulting in a stronger motivation to
compensate for the adult privileges that are denied them. Babyfaced adolescents may consequently show assertive and competent behavior that contradicts the babyface stereotype.
Consistent with the suggestion that babyfaced adolescents
will show stereotype-disconfirming behavior, Zebrowitz, Collins, and Dutta (1998) found that middle-class babyfaced adolescent boys were more assertive and hostile than their maturefaced peers according to personality assessments made by judges
who had never met the boys but who had access to data gathered
from the boys, their families, and school records. Items such as
values independence, rebellious, pushes limits, and skeptical
contributed to the assertiveness personality measure, and items
such as negativistic, self-indulgent, deceitful, condescending,
distrustful, and irritable loaded on the hostility measure. Such
traits seem aimed at compensating for the low autonomy and
low status that adolescents experience. The compensatory assertive and contentious personality traits of babyfaced adolescent boys may yield prosocial as well as antisocial self-defeating
prophecy effects. Indeed, babyfaced soldiers who served in the
military during World War II and the Korean War were more
likely to win a military award (Collins & Zebrowitz, 1995).
This outcome suggests courageous and heroic behavior that
could bridge the maturity gap. 1
Babyfaceness does not appear to increase compensatory behavior among adolescent girls, who showed no differences in
personality from their mature-faced peers (Zebrowitz et al.,
1998). Babyfaced adolescent girls may show •no stronger reactions to the maturity gap than their mature-faced peers because
the way they are treated is more congruent with a feminine
gender identity than a masculine one. Not only are perceptions
of babyfaced people and children paralleled by perceptions of
women (cf. Bakan, 1966; Friedman & Zebrowitz, 1992), but
researchers have also argued that women's self-concepts incorporate elements, such as likability and connection to others, that
are consistent with the babyface stereotype in contrast to men's
self-concepts, which emphasize power and self-sufficiency (for
a pertinent review see Cross & Madson, 1997).
The purpose of the present research was to determine whether
the previously documented finding of more assertive and contentious personalities among babyfaced adolescent boys is paralleled by achievement and antisocial behaviors that disconfirm
the babyface stereotype. We also sought to establish the generality of such compensation effects across a broad socioeconomic
range. To these ends, we examined in Study 1 the effects of
babyfaceness on years o f education in a middle-class sample of
adolescent boys and gifts. In Study 2 we investigated the effects
of babyfaceness on achievement test performance in low-income
samples of nondelinquents and delinquents. Finally, in Study 3
we examined the effects of babyfaceness on delinquent behavior
in the latter samples. Although the focus of the present research
was on babyfaceness, the effects of attractiveness, height, and
body type were also examined, because there are reasons to
expect these also to influence academic achievement and criminal behavior.
Study 1
Study 1 tested the hypothesis that the compensation effect in
personality development shown by middle-class babyfaced boys
(Zebrowitz et al., 1998) would also be manifested in their educational achievement. More specifically, we predicted that babyfaced boys would achieve more years of education than their
mature-faced peers, thereby refuting the expectation that they
are intellectually weak. Such expectations have been documented not only in the attribution of greater naivet6 to babyfaced
individuals of all ages (Zebrowitz & Montepare, 1992), but also
in the use of slower speech when explaining a game to babyfaced
children (Zebrowitz, Brownlow, & Olson, 1992) and in the
allocation of less cognitively demanding chores to babyfaced
children (Zebrowitz, Kendall-Tackett, & Fafel, 1991). We did
not expect a self-defeating prophecy effect to occur for babyfaced adolescent gifts, both because past research showed no
effects of babyfaceness on girls' personality development and
also because the theoretical basis for predicting a self-defeating
prophecy effect is more applicable to boys, whose manliness is
likely to be threatened by the babyface stereotype (Friedman &
Zebrowitz, 1992; Zebrowitz et al., 1998).
We also predicted positive effects of attractiveness on educational achievement because considerable research has shown
greater academic and occupational success among those who
are more attractive (Feingold, 1992; Langlois et al., 1996). The
effects of height and body type were examined as well. Taller
individuals were expected to show greater educational achievement, because height has been found to be related to higher IQ as
well as to greater expectations of achievement and competence
(Brackbill & Nevill, 1981; Eisenberg, Roth, Bryniarski, & Murray, 1984; Johnson, 1991). We expected more muscular boys to
show lower educational achievement, because previous research
revealed that boys who were high in muscularity during their
high school years showed lower intellectual investment as well
as lower dependability and productivity (Clausen, 1993). Finally, we investigated the interactions of facial appearance with
height, muscularity, IQ, and socioeconomic status (SES) to explore two possibilities. One possibility was that the beneficial
effects of babyfaceness and attractiveness would be most pronounced when they were supported by other factors that facilitate high achievement, such as higher IQ and SES, A second
possibility was that the beneficial effects of babyfaceness and
attractiveness would be more pronounced when they could have
a protective effect against factors that would operate against
high achievement. Such an effect would be consistent with the
finding that a girl's attractiveness had a protective effect against
paternal rejection by fathers who were stressed by economic
hardships during the depression. These fathers showed rejection
tAn alternative explanation is that this finding reflects a contrast
effect, whereby whatever courageous actions babyfaced men took received more recognition because the contrast with the perception of
babyfaced individuals as warm, dependent, and submissive made these
actions more salient.
1302
ZEBROWlTZ ET AL.
o f their daughters if they were unattractive but not if they were
attractive (Elder, Nguyen, & Caspi, 1985).
Me~od
Participants
Study 1 used data from the Intergenerational Studies of Development
and Aging (IGS), a combination of three longitudinal studies begun
between 1928 and 1933 and archived at the University of California,
Berkeley, Institute of Human Development. Three samples were used:
Berkeley Guidance, Berkeley Growth, and Oakland Growth, hereafter
referred to as Guidance, Berkeley, and Oakland, respectively. Participants
in the Guidance and Berkeley samples were born between 1928 and 1929
in Berkeley, California. Most were from White, middle-class, Protestant
families, and family educational status was above the average for the
general U.S. population (Caspi, Elder, & Bern, 1987; Eichorn, 1981).
Oakland participants, born between 1920 and 1922, were a reasonable
representation of the population attending Oakland schools where they
were enrolled at the time they were initially studied between the ages
of 10 and 12. Like the Guidance and Berkeley samples, most Oakland
participants came from White, middle-class families, although children
of blue-collar workers composed a higher percentage of the Oakland
sample (Block, 1971). (For further details about the three studies, see
Eichorn, Clansen, Haan, Honzik, & Mussen, 1981; Jones, Bayley, Macfarlane, & Honzik, 1971.) To be included in the present study, participants needed both educational outcome data and appearance data in late
adolescence. A total of 96 male and 114 female participants met this
criterion. It should be noted that these samples were not selected for their
particular appearance or personality qualities, thus minimizing possible
selection bias. Participants selected for inclusion in the present investigation comprised a subset of those included in a study investigating the
stability of appearance across the lifespan (Zebrowitz, Olson, & Hoffman, 1993), and overlapped with those included in studies investigating
the contribution of appearance to military service outcomes, to perceived
and real health, to personality, and to perceived and real honesty across
the lifespan (Collins & Zebrowitz, 1995; Kalick, Zebrowitz, Langlois, &
Johnson, 1998; Zebrowitz et al., 1998; Zebrowitz, Voinescu, & Collins,
1996).
Appearance Predictors
Babyfaceness and attractiveness. Facial ratings were taken from a
previous study by Zebrowitz, Olson, and Hoffman (1993, Study l).
Ratings of photographs taken at ages 17, 17.5, and 18 were averaged to
create indexes of adolescent attractiveness and babyfaceness. Ratings at
three ages were used to create a more reliable measure of appearance
than any single photograph could provide. However, if a photograph at
one age was missing, the index was created by averaging ratings at the
remaining ages. Adolescent appearance was used because the criterion
variable of educational attainment is typically achieved shortly after
adolescence and well before the next appearance rating (at age 32)
available in the archive. The average alpha coefficient across age was
.86 for ratings of babyfaceness and .89 for ratings of attractiveness.
Research supporting the validity of these ratings includes significant
agreement between the rated attractiveness of girls and ratings of their
prettiness made by IGS staff in the 1960s. Previous research also has
established predictive validity of the adolescent appearance ratings: Attractiveness ratings predicted likelihood of marriage and, for those who
did marry, age of marriage (Kalick et al., 1998); babyfaceness ratings
predicted adolescent boys' personality traits and likelihood of winning
a military award (Collins & Zebrowitz, 1995; Zebrowitz et al., 1998).
Height. Height in meters at 18 years was available for most participants. In cases where these data were not available, missing data were
progressively substituted with height data at 19.5 years, 21 years, 25
years, approximately 42 years, and approximately 56 years.
Muscularity. Muscularity scores were calculated for each participant
by reversing the sign on the ponderal index, which is the ratio of each
participant's height tO the cubic root of his or her weight. Height and
weight at 18 years of age were used to compute this measure with height
data at later ages substituted as indicated above. Because weight can be
unstable across the adult years, weight data were substituted only at
19.5 years, 21 years, or 25 years. The ponderal index, which was standardized within each sex, is the most useful single index of somatotype,
because it is positively correlated with ectomorphy (linearity) and negatively correlated with mesomorphy (muscularity; Hartl, Monnelly, &
Elderkin, 1982; Sheldon, Stevens, & Tucker, 1940). Higher scores on
the reversed ponderal index signified greater muscularity. 2
Demographic Predictors
IQ. Stanford-Binet test scores at age 18 years served as the measure
of adolescent intelligence. For Berkeley and Guidance participants, missing data were replaced by substituting Wechsler-BeUevue scores from
age 18 years. For Oakland participants, missing data were replaced by
substituting earlier Stanford-Binet scores obtained when participants
were 16 or 17 years old.
SES. We calculated SES as measured by the Hollingshead Index
(Hollingsbead & Redlich, 1958) as the mean of scores collected when
participants were born and when they were approximately 18 years old.
If the score at one age was missing, the other age was used as the index
of SES.
Criterion Variable
A composite measure of educational achievement was created for
each participant. For Guidance and Oakland participants, the Eichorn
Scale of educational attainment (Hollingshead & Redlick, 1958) was
used. This scale ranges from did not complete elementary school (1)
to PhD, MD, DDS, LLB, EdD, DD (10). Data from Berkeley participants
did not include the Eichom measure, but did provide scores on the
Hollingshead Education Scale, which ranges from professional training
(1) to under 7years of school (7). Because the Eichorn Scale discriminates more between various levels of education, particularly at the high
end, the data of those participants who only had Hollingshead scores
were converted to comparable scores on the Eichorn Scale. In all but
two cases, the Hollingshead value was equivalent to a value on the
Eichorn Scale (e.g., Hollingshead value of 3 = some college = Eichorn
value of 6), and participants were simply reassigned the appropriate
value from the Eichorn Scale. Participants who had a high school graduate, some business or vocational school (4) on the Hollingshead Scale
were given a 4.5 on the new scale because they fell in between category
completed high school (4) and category high school + vocational or
business school (5) on the Eichorn Scale. Participants who received a
professional training (1) on the Hollingshead Scale were assigned a
score of college graduate + extra credential (8) on the Eichorn Scale.
As a result, these particular participants may have an underestimation
of their education on the new scale because it was impossible to discern
2 An index of obesity, the Body Mass Index (BMI; weight in kg/
height in square centimeters) was not used as a predictor in Studies 1,
2, or 3 because very few participants were overweight. In Study I, only
5 of the 96 boys and 5 of the 114 girls had BMI values above .25,
which is the 85th percentile for individuals of this age, a conventional
criterion for determining overweight (Kuczmarski, Flegal, Campbell, &
Johnson, 1994; Najjar & Rowland, 1987). Of the 967 boys in Studies
2 and 3, only 6 delinquents and 13 nondelinquents were overweight by
this criterion.
BRIGHT, BAD, BABYFACED BOYS
1303
Table 1
Correlation Matrix o f Predictor and Criterion Variables
for IGS Boys (N = 96) and Girls (N = 114)
Variable
1.
2.
3.
4.
5.
6.
7.
Babyfaceness
Attractiveness
Height
Muscularity
IQ
SES
Educational achievement
1
2
3
4
5
6
7
--.05
-.11
.01
.00
.03
.07
-.05
-.25**
-.47***
.08
.02
.08
-.27**
.02
--.33***
.25**
.11
.26**
-.05
-.21"
-.39***
--.18t
-.11
-.10
.19t
.19~
.13
.22*
-.32**
.43***
.25*
.07
.03
.18t
.33**
-.57***
.19t
.05
.14
.24*
.47***
.36***
--
Note. Boys are above the diagonal and girls are below. IGS = IntergenerationalStudies of Development
and Aging; SES = socioeconomic status.
l'P < .10 (marginally significant). *p < .05. **p < .01. ***p < .001.
whether they had completed masters or doctoral work (values of 9 and
10, respectively,on the Eiehom Scale). Nonetheless,the creation of this
new measure enabled us to retain the discrimination afforded by the
Eichorn Scale for the majority of the participants. Evidence for the
validity of this measure is provided by the findingthat it was correlated
with IQ scores (see Table 1), and that it also predicted income and
occupationalstatus when used as a control variable in a study investigating occupational outcomes at ages 50 to 60 years (Collins & Zebrowitz,
1995).
Results
Overview
Hierarchical nmltiple regression analyses were performed
separately for boys and girls to determine the influence of appearance on educational achievement. Predictor variables were
entered into the regression analyses in three blocks: Appearance
predictors (attractiveness, babyfaceness, height, muscularity)
were entered in Block 1, demographic predictors (IQ, SES)
were entered in Block 2, and interactions were entered in Block
3. Appearance predictors were entered first so that results could
be compared with past research, which typically has examined
effects of appearance without controlling for demographic variables or interaction effects. The interactions examined included
babyfaceness and attractiveness crossed with height, muscularity, IQ, and SES. We also examined interactions of height and
nmscularity with I Q a n d SES, although no predictions were
made for these exploratory analyses. Because the final models
included many nonsignificantpredictors, trimmed models were
designated by deleting all interaction predictors with ts less than
1. We centered the data for each of the continuous predictor
variables around the mean to reduce the correlation between
main effects and interaction effects (Cohen & Cohen, 1983).
Correlations among the predictor and criterion variables for
boys and girls are presented in Table 1; mean values for the two
samples are presented in Table 2.
Boys
Overall effects. Table 3 presents results of the hierarchical
regression predicting educational achievement for boys. The
block of appearance predictors accounted for 11% of the variance in educational achievement, F(4, 91) = 2.67, p < .05.
The demographic predictors contributed an additional 19% to
the variance explained, F(2, 89) = 11.78, p < .001. The
trimmed block of interactions explained an additional 4% of
the variance, F(1, 88) = 5.92, p < .05.
Babyface effects. As predicted, babyfaced boys showed
higher educational achievement than their mature-faced peers.
However, babyfaceness no longer predicted achievement once
the demographic predictors were entered into the equation. Further analyses revealed that both IQ and SES were possible third
variables accounting for the initial relationship between babyfaceness and educational achievement. Specifically, higher babyfaceness was correlated with higher levels of IQ and SES, and
IQ and SES each was correlated with higher educational
achievement (see Table 1). Regression equations to assess the
separate contributions of IQ and SES to the relationship between
babyfaceness and educational achievement revealed that the effect of babyfaceness lost significance both when tQ alone was
added to the block of appearance predictors (/~ = .13, p = .21 )
and also when SES alone was added to the equation (/~ = .15,
p = .16).
We have no explanation to offer for the surprising relationship
between babyfaceness and SES. The positive relationship of
babyfaceness to IQ was also unanticipated, and further analyses
revealed that it could not be explained by the relationship of
Table 2
Means and Standard Deviations of Predictor and Criterion
Variables f o r IGS Boys (N = 96) and Girls (N = 114)
Boys
Variable
Babyfaceness
Attractiveness
Height
Muscularity
IQ
SES
Educational achievement
M
3.84
3.39
179.17
-0.03
121.86
3.39
6.96
Girls
SD
M
SD
t(208)
0.71
4.07 0.88 2.11"
0.66
2.98 0 . 8 0 4.06***
7 . 3 2 165.21 6.34 14.62"**
1 . 0 4 -0.05 1.01 0.57
11.11 116.44 11.45 3.48**
1.16
3.25 1.12 0.86
1.69
5.77 1.41 5.50***
Note. IGS = IntergenerationalStudies of Developmentand Aging; SES
= socioeconomicstatus.
*p<.05.
**p<.01.
***p<.001.
1304
ZEBROWITZ ET AL.
Table 3
Summary of Regression Analysis Predicting Educational Achievement
for IGS Boys (N = 96) and Girls (N = 114)
Boys
Predictor variable
Block 1a
Babyfaceness
Attractiveness
Height
Muscularity
Block 2b
Babyfaceness
Attractiveness
Height
Muscularity
IQ
SES
Block 3¢
Babyfaceness
Attractiveness
Height
Muscularity
IQ
SES
Babyfaceness × IQ
Babyfaceness × SES
Babyfaceness × Height
Attractiveness × IQ
Attractiveness × Height
Attractiveness × Muscularity
Muscularity × SES
Girls
B
SE B
~
B
SE B
0.53
0.06
0.03
-0.28
.25
.26
.03
.18
.22*
.02
.13
-.18
0.17
0.02
0.06
-0.02
.15
.18
.02
.15
0.20
-0.12
0.02
-0.17
0.05
0.30
.23
.24
.02
.17
.01
.14
.09
-.05
.07
-.10
.36***
.21"
0.12
0.09
0.04
0.09
0.03
0.59
.12
.15
.02
.12
.01
.10
.08
.05
.16"
.07
.24**
.48***
0.27
-0.18
0.03
-0.10
0.05
0.25
.23
.23
.02
.16
.01
.14
.11
-.07
.12
-.06
.35***
.17]"
0.14
0.18
0.03
0.06
0.03
0.59
0.02
-0.17
0.03
-0.03
-0.03
-0.18
.12
.15
.02
.13
.01
.10
.01
.12
.02
.01
.02
.14
.09
.11
.14].
.04
.21"
.47***
.12
-.12
.12
-.17"
-.09
-.11
0.26
.11
.22*
.10
.01
.27**
-.01
Note. IGS = Intergenerational Studies of Development and Aging; SES = socioeconomic status.
a For boys, F(4, 91) = 2.67, p < .05; for girls, F(4, 109) = 2.35, p < .10. b For boys, F(6, 89) = 6.13,
p < .001; for girls, F(6, 107) = 12.82, p < .001. ¢ For boys, F(7, 88) = 6.39, p < .001; for girls, F(12,
101) = 7.52, p < .001.
].p "~ .10 (marginally significant). *p < .05. **p < .01. ***p < .001.
babyfaceness to SES. When all appearance predictors and SES
were entered into a regression equation predicting IQ, babyfaceness remained a significant predictor ( 8 = .25, p < .01). We
conducted additional analyses to see whether this relationship
reflected higher achievement motivation among more babyfaced
boys, a finding that would be consistent with a compensation
effect. We performed these analyses on a subset of 74 boys for
whom an index of motivation could be constructed from Qsortpersonality data available in the IGS archive. We created a
motivation composite by summing scores on five Q-sort items:
values intellect, ambitious, productive, self-defeating (reverse
scored), and withdraws when frustrated (reverse scored; Block,
1961/1978). Regression analyses predicting IQ from the appearance variables revealed a positive effect of babyfaceness
for this subset of boys ( 8 = .32, p < .01 ), and this effect lost
significance when controlling for motivation ( 8 = -18, p = . 10).
Thus, consistent with the predicted compensation effect, the
higher IQ levels that accounted for babyfaced boys' higher educational attainment were due to their greater achievement
motivation.
Attractiveness effects. Contrary to predictions, there were
no effects of attractiveness on academic achievement for boys.
Other effects. There was a significant main effect for IQ,
reflecting higher educational achievement among boys with
higher IQs. There was also a tendency for boys from higher
SES backgrounds to show higher educational achievement. A
significant interaction between muscularity and SES revealed
that higher muscularity was associated with lower educational
achievement when boys were low SES, but not when they were
high SES, which suggests that a muscular body build is a risk
factor among low-SES boys (see Figure 1 ).
Girls
Overall effects. The results of the hierarchical regression
predicting academic achievement for girls are presented in Table
3. The block of appearance predictors accounted for 8% of the
variance in academic achievement, F ( 4 , 109) = 2.35, p = .06.
The block of demographic predictors accounted for an additional 34% of the variance explained, F ( 2 , 107) = 31.15, p <
.001. The block of trimmed interaction predictors explained an
additional 5% of the variance, F ( 6 , 101) = 1.71, p = .13.
Babyface effects. As expected, there were no effects of
babyfaceness on academic achievement for girls.
Attractiveness effects. Contrary to prediction, there was no
overall effect of attractiveness on girls' academic achievement.
BRIGHT, BAD, BABYFACED BOYS
1305
Figure I. Educational achievement of Intergenexational Study of Development and Aging (IGS) boys as
a function of muscularity and socioeconomic status (SES), and educational achievement of IGS girls as a
function of attractiveness and IQ. High, average, and low values of each variable represent one standard
deviation above the mean, the mean, and one standard deviation below the mean, respectively (Aiken &
West, 1991; Cohen & Cohen, 1983).
However, there was a significant interaction between attractiveness and IQ: Attractiveness was positively related to educational achievement for girls who were below average in IQ, with
a slight reversal for those who were above average in IQ (see
Figure 1).
Other effects. Consistent with past research, taller girls
showed higher academic achievement, as did girls who were
higher in IQ or SES.
Sex Differences
Comparisons between the regression coefficients for boys and
girls were made in each case in which we obtained an effect
for one sex and not the other (Cohen & Cohen, 1983, p. 111).
Although babyfaceness predicted the educational achievement
of boys but not girls, the difference in the magnitude of these
predicted results was not significant (z = 1.24, p = .11, one
tailed). Similarly, the unpredicted Attractiveness × IQ interaction effect for girls was not significantly greater than it was for
boys (z = 1.12, p = .26, two tailed). However, the effect of
SES on educational achievement was marginally greater for girls
than for boys (z = 1.69, p = .10, two tailed) and the SES ×
Muscularity interaction was significantly greater for boys than
for girls (z = 2.01, p = .04, two tailed).
Discussion
The results of Study 1 show that the previously reported
tendency for middle-class babyfaced boys to refute the babyface
stereotype with personality traits of assertiveness and hostility
extends to the academic domain, where the same boys show a
higher level of education than their mature-faced peers. This
higher achievement is partially due to the higher IQs of more
babyfaced boys, which in turn reflects their greater motivation.
For girls, on the other hand, babyfaceness had no effect on
educational achievement, just as it had shown no relationship
to personality traits in previous research (Zebrowitz et al.,
1998). As noted above, this may be because the babyface stereotype is more consistent with gender expectations for girls than
for boys (Cross & Madson, 1997; Friedman & Zebrowitz,
1992). Although it appears that a babyface motivates boys to
compensate for infantilizing expectations, but does not have this
effect for girls, further research investigating gender differences
in behavioral disconfirmation of the babyface stereotype is
needed to draw finn conclusions, because the effect for boys
was not significantly greater than the effect for girls.
Contrary to prediction, attractiveness had no effect on boys'
educational achievement although it was positively related to
their IQ scores, consistent with a recta-analysis of studies focusing on this age range (el. Langlois et al., 1996). On the other
hand, attractiveness was positively related to girls' educational
achievement only when they were below average in IQ. The
finding that attractiveness had the strongest effect for low-IQ
girls is consistent with evidence that teacher expectations predict
achievement more strongly for low than for high achievers (Madon, Jussim, & Eccles, 1997), and it suggests that the effects
of attractiveness are most pronounced when it can play a protective role in mitigating other factors that operate against high
achievement. The finding that attractiveness had an effect on
1306
ZEBROWlTZ ET AL.
girls' educational achievement but not boys' is consistent with
the meta-analytic findings of Feingold (1992), who found a
small effect of attractiveness on grades earned by girls but no
effect on grades earned by boys. Although it appears that attractiveness may have more impact on the educational achievement
"of girls than boys, further research investigating sex differences
is needed because the effect for girls was not significantly
greater than that for boys. Interestingly, SES predicted level of
education more strongly for girls than for boys. It appears that
the achievement of girls may be more influenced by external
factors, like their SES or expectations generated by their
appearance.
Study 2
Study 2 investigated the influence of babyfaceness and attractiveness on academic achievement in lower class samples of
delinquent and nondelinquent boys, with standardized test performance as the criterion variable. As in Study 1, we explored
the effects of height and muscularity and the interactions of
facial appearance with height, muscularity, IQ, and SES. Interactions with an additional variable representing family risk factors
for delinquency were also examined. This variable was included
because past research combining the delinquent and nondelinquent samples found that various family characteristics predicted not only delinquent behavior but also attitudes toward
school (Sampson & Laub, 1993). It therefore seemed plausible
that family characteristics might also explain a significant percentage of the variance in standardized tests of academic
achievement. On the other hand, it is possible that previous
findings reflected significant mean differences between the delinquent and nondelinquent samples in both family characteristics and school attitudes. We performed analyses separately for
the two groups in the present study to ensure that relationships
between the predictor and criterion variables would not simply
reflect the large mean differences between the two samples.
The separate analyses had the added advantage of providing an
opportunity to test the hypotheses across two different samples;
however, we predicted no differences in the effects of
appearance.
Although middle-class boys in Study 1 showed evidence of
compensating for the infantilizing effects of the babyface stereotype with stronger academic performance, it was less certain
that babyfaced boys in the lower class samples in Study 2 would
do so. Intellectual achievement is often less valued in lower
class samples than in middle-class samples (e.g., Dornbusch,
Herman, & Morley, 1996; Ginsburg, Bempechat, & Chung,
1992; Labov, 1972; Ogbu, 1986; Willis, 1977). Also, boys in
the Study 2 samples were less intellectually able than those in
Study 1, whose mean IQ scores were well above average (see
Table 2). These differences suggested that the effects of babyfaceness on academic achievement in the Study 2 samples might
be most pronounced for boys at the top of the distribution in
social class, ability, or both, because these boys would be more
motivated and able to compensate successfully for the babyface
stereotype in the academic domain. Although there was no effect
of attractiveness in Study 1, the documentation of a relationship
between attractiveness and various measures of intelligence and
academic achievement in a recent meta-analysis (Langlois et
al., 1996) suggested that the prediction of attractiveness effects
should still be regarded as viable. Previous research findings
also led us to continue to predict a positive relationship between
height and achievement despite the fact that this effect was
restricted to girls in Study 1 (e.g., Eisenberg et al., 1984; Johnson, 1991). Finally, previous research and the results of Study
1 yielded the prediction that more muscular boys would show
lower educational achievement, particularly when they were
from lower SES backgrounds (Clansen, 1993).
Method
Participants
Participants for this study were delinquent and nondelinquent White
men born between 1924 and 1935 drawn from the Glueck and Glueck
(1950) Crime Causation Study (CCS) archived at the Henry A. Murray
Research Center, Radcliffe College, Cambridge, Massachusetts. The delinquant sample was selected from two Massachusetts correctional
schools to which participants had been committed with an average of
three court appearances and convictions. The nondelinquentsample was
selected from the Boston public schools. The two samples were matched
on age, intelligence, ethnicity (national origin of parents), and area of
residence. All participants were followed longitudinally, with the first
wave of data collection occurring from 1940 to 1948 when they ranged
in age from 10 to 17 years (Time 1), the years that are the focus of
this study. As in Study 1, to be included in the present study, participants
needed both academic achievement data and appearance data. A total
of 461 of the 500 delinquents and 466 of the 500 nondelinquents met
this criterion. Participants included in Study 2 also served as participants
in a study investigating the contribution of appearance to parental and
peer relationships at Time 1 (Zebrowitz & Lee, in press).
Appearance Predictors
Babyfaceness and attractiveness. Slides of participants' neutral expression frontal view or profile faces, photographed when they entered
the study (M age = 14.6 years), were presented in age blocks (~=12,
13 to 14, 15 to 16, ~17), and judges were instructed to rate each boy's
appearance relative to other boys of the same age on oneof two 7-point
Likert scales (babyfaced-mature-faced or unattractive-attractive).
This was done because the social consequences of bahyfaceness or
attractiveness depend on how an individual compares with others of his
or her own age. If a 14-year-old boy is seen as more babyfaced than a
17-year-old, that has less social significance than if he is seen as more
babyfaced than other 14-year-olds. The average alpha coefficient was
.94 for ratings of frontal babyfaceness, .93 for ratings of profile babyfaceness, .89 for ratings of frontal attractiveness, and .86 for ratings of
profile attractiveness. We created composite measures of babyfaceness
and composite measures of attractiveness by standardizing and summing
judges' ratings of frontal view slides and profile view slides, which
were significantly correlated, attractiveness, r(961) = .44, p < .001;
babyfaceness, r(961 ) = .65, p < .001. These composite measures were
used to predict behavior rather than ratings based on either the frontal
or profile photos alone because they provided a more ecologically valid
indicator of social perceptions that could have instigated compensatory
behavior. Previous research has supported the validity of these appearance ratings. For example, attractiveness ratings predicted peer relations,
and babyface ratings predicted paternal discipline (Zebrowitz & Lee,
in press). For additional details regarding the procedure for collecting
appearance ratings, see Zebrowitz and Lee (in press). 3
a We also conducted analyses using frontal-view appearance ratings
as predictors, which is the measure used in past research when appearance ratings were based on photographs rather than actual exposure to
the individual. No additional effects emerged in these analyses.
BRIGHT, BAD, BABYFACED BOYS
1307
Height. A measure of height for each participant taken at Time
1 was extracted from the CCS archival data. We standardized each
participant's height in inches at the time of the facial photograph within
each age group so that height was relative to others of the same age.
Muscularity. As in Study 1, we calculated muscularity scores for
each participant by reversing the sign of the ponderal index, which is
the ratio of each participant's height in inches at Time 1 to the cubic
root of his weight in pounds,
Demographic Predictors
;I
Age. We used nge at time of entrance into the study as a control
predictor in the present investigation to ensure that the effects of appearance predictors were not confounded by age. Age was recorded in the
CCS archives on a 12-point scale (1 = under 11 years; 2 through 11
= successive 6 month increments from 11 to 16.5 years; 12 = over 16.5
I
. . ~ @ . . . ~
II
. . . .
I I I
years).
IQ. The total full-scale score from the Wechsler-Bellevue Scale,
extracted from the archival data, served as the measure of intelligence•
Because delinquents and nondelinquents were matched in IQ, the mean
difference between the groups was small, albeit significant.
Family risk composite. We converted three variables to z scores and
then summed them to form the family risk composite: family size (number of children), father's deviance, and mother's deviance (measures of
parental criminality and alcoholism ranging from 0 to 4). We expected
this composite to account for significant variance in academic achievement because each of the component variables had previously been
found tO predict attitudes toward school in research combining the delinquent and nondelinquent samples (Sampson & Laub, 1993).
SES. We converted two variables to z scores and then summed them
to form the index of SES. One variable, a standardized scale extracted
from the archives, was composed of the average weekly income of the
family and a measure of the family's reliance on outside aid. Ratings
made by CCS researchers were recoded so that participants were given a
score of 4 if their family was rated as living in comfortable circumstances
(having enough savings to cover 4 months of financial stress); 3 for
marginal circumstances (little or no savings but only occasional dependence on outside aid); and 2 for dependent circumstances (continuous
receipt of outside aid for support). (Scores of 1 in the original archive
signified missing data.) The second variable was a measure of parents'
education level on a 9-point scale ranging from 1 (both illiterate) to 9
I
I I I I
li
. . . . . .
I l l l l l
~
~
I I I I
I
I
II
I
II
~@~@
.~
.
I
.~
I
.
.
.
li
??o?oooo
I'll
. . . . .
. . . . .
(both attended college).
r~
e~
I
I
. . . ~
II
. . . .
II
lii
II
II
Criterion Variable
We used scores on the Stanford Achievement tests administered by
CCS psychologists to derive a composite measure of academic achievement, which was the sum of each boy's "reading quotient" and "arithmetic quotient" z scores. These scores were significantly correlated,
r(925) = .62, p < .001. Evidence for the validity of these measures is
provided by the fact that they significantly differentiated delinquents
from nondelinquents (Glueck & Glueck, 1950). Evidence for the validity
of the composite measure is provided by the finding that it was strongly
correlated with IQ within each sample (see Table 4).
I I I I
q ~ .
I
Results
Overview
We conducted hierarchical multiple regression analyses to
determine the influence o f appearance on academic achievement. The analyses were identical to Study 1 except that age
and family risk were included as additional demographic predictors in Block 2 and family risk was crossed with appearance
~'~ ~_'~
~,~ o - ~
z~-.
1308
ZEBROWITZ ET AL.
Table 5
Means and Standard Deviations of Predictor and Criterion
Variables for Nondelinquents and Delinquents
Nondelinquents
Variable
M
Babyfaceness
0.13
Attractiveness
0.01
Height
0.09
Muscularity
-13.16
IQ
94.15
SES
0.30
Age
14.42
Family risk
-0.55
Academic achievement
0.44
Early onset
Time 1 charges
Time 2 charges
Time 3 charges
SD
1.82
1.73
1.05
0.58
11.89
1.31
1.36
1.37
1.79
Delinquents
M
SD
t(925)
-0.14
1.80
2.32*
-0.04
1.66
0.51
-0.10 0.93
2.90**
-12.93 0.41
6.89***
91.85 12.96
2.81"*
-0.36
1.46
7.25***
14.70 1.51 -2.99**
0.56 1 . 3 6 ~-12.34"**
-0.45
1.69
7.81"**
0.41 0.60
6.44 3.52
5.94 5.39
3.49 4.93
Note. Age was originally on a 12-point scale; means were interpolated
to give value in years. For nondelinquents, n = 466; for delinquents, n
= 461 for the first nine variables, n = 464 for early onset and Time 1
changes, n = 435 for Time 2 changes, and n = 418 for Time 3 changes.
SES = socioeconomic status.
*p---.05. **p <.01. ***p <.001.
predictors in Block 3. Correlations among the predictor and
criterion variables for delinquents and nondelinquents are presented in Table 4, and mean values for the two samples are
presented in Table 5.
Overall Effects
Results of the hierarchical regression predicting academic
achievement for nondelinquents and delinquents are presented in
Table 6. For nondelinquents, the b!ock of appearance predictors
accounted for 3% of the variance in academic achievement,
F (4, 461 ) = 3.51, p < .01. The block of demographic predictors
added an additional 55% to the variance explained, F ( 4 , 457)
= 151.79, p < .001. The trimmed block of interactions added
an additional 1% to the explained variance, F ( 3 , 454) = 3.77,
p = .01. For delinquents, the block of appearance predictors
accounted for 3% of the variance in academic achievement,
F ( 4 , 456) = 3.58, p < .01. The block of demographic predictors
added an additional 46% to the variance explained, F ( 4 , 452)
= 101.61, p < .001. The trimmed block of interactions added
an additional 2% to the explained variance, F ( 7 , 445) = 2.80,
p < .01.
Babyface Effects
There was a marginally significant Babyfaceness × SES interaction effect for nondelinquents (see Figure 2). Consistent with
the evidence for a self-defeating prophecy effect of babyfaceness
for middle-class boys in Study 1, babyfaceness showed a small
positive relationship to academic achievement for boys who
were above average in SES in this lower class sample of nondelinquents. However, babyfaceness was negatively related to academic achievement for boys who were below average in SES,
Figure 2. Academic achievement as a function of babyfaceness and
socioeconomic status (SES) for nondelinquents, babyfaceness and IQ
for delinquents, and attractiveness and family risk for nondelinquents.
High, average, and low values of each variable represent one standard
deviation above the mean, the mean, and one standard deviation below
the mean, respectively (Aiken & West, 1991; Cohen & Cohen, 1983).
BRIGHT, BAD, BABYFACED BOYS
1309
Table 6
Summary of Regression Analyses Predicting Academic Achievement for Nondelinquents
(N = 466) and Delinquents (N = 461)
Nondelinquents
Predictor variable
Block 1a
Babyfaceness
Attractiveness
Height
Muscularity
Block 2b
Babyfaceness
Attractiveness
Height
Muscularity
IQ
SES
Age
Family risk
Block 3¢
Babyfaeeness
Ata:activeness
Height
Muscularity
IQ
SES
Age
Family risk
Babyfaceness × IQ
Babyfaceness x SES
Babyfaceness × Family Risk
Attractiveness × Family Risk
Attractiveness × Muscularity
Height × IQ
Muscularity x SES
Muscularity × IQ
Delinquents
B
SE B
~
B
SE B
0.06
0.05
0.35
0.36
.06
.05
.10
.16
.06
.05
.20***
.12"
0.11
0.07
0.31
-0.06
.06
.05
.12
.23
.12"
.07
.17"*
-.02
-0.02
-0.06
0.08
0.17
0.09
0.02
-0.28
-0.08
.04
.03
.07
.11
.00
.04
.02
.04
-.02
-.061"
.05
.06
.61"**
.01
-.42***
-.06t
0.02
0.00
0.10
-0.02
0.09
0.04
-0.10
-0.05
.04
.03
.09
.17
.00
.04
.02
.04
.02
.00
.06
-.01
.67***
.04
-.17"**
-.04
-0.02
-0.06
0.10
0.16
0.09
0.03
-0.28
-0.07
.04
.03
.07
.11
.00
.04
.02
.04
-.02
-.06*
.06
.05
.61"**
.02
-.43***
-.05
0.02
0.01
0.13
-0.08
0.09
0.03
-0.10
-0.05
0.01
.04
.04
.09
.18
.00
.04
.02
.04
.00
.03
.01
.07
-.02
.68***
.03
-.19"**
-.04
.10"
0.04
.02
.05*
0.04
.02
.06*
0.01
.00
.07*
0.02
-0.03
-0.08
0.02
0.11
0.02
.02
.02
.09
.01
.11
.01
.04
-.05
-.03
.19"**
.03
.05
Note. SES = socioeconomic status.
=For nondelinquents, F(4, 461) = 3.51, p < .01; for delinquents, F(4, 456) = 3.58, p < .01. b F o r
nondelinquents, F(8, 457) --- 79.95, p < .001; for delinquents, F(8, 452) = 54.17, p < .001. c For
nondelinquents,F(ll, 454) -- 60.23, p < .001; for delinquents,/7(15, 445) = 31.01, p < .001.
*p -< .10 (marginally significant). *p < .05. **p < .01. ***p < .001.
suggesting a self-fulfilling prophecy effect in this group. Among
delinquents, there was a significant main effect for babyfaceness
such that more babyfaced delinquents showed higher achievement than their mature-faced peers. Although this effect was
eliminated when the block of demographic predictors was entered into the equation, a significant interaction between babyfaceness and IQ remained in the final equation (see Figure 2).
As predicted, babyfaced delinquents who were above average
in IQ showed higher academic achievement than their maturefaced peers, consistent with a compensatory self-defeating
prophecy effect. On the other hand, babyfaceness was negatively
related to academic achievement for boys who were below average in IQ, consistent with a self-fulfilling prophecy effect.
Attractiveness Effects
Contrary to prediction, more attractive nondelinquents
showed significantly lower academic achievement. This e f f e c t
was qualified by a significant Attractiveness × Family Risk
interaction, which revealed that the negative effect of attractiveness on academic achievement occurred only for boys who
had low or average levels of family risk and not for those at
high risk, Who showed low achievement regardless of their level
of attractiveness (see Figure 2). For the delinquent sample,
there were no significant effects of attractiveness on academic
achievement.
Other Effects.
Age was negatively related to academic achievement in both
groups, suggesting that academic performance on age-graded
standardized tests declined significantly over time in these highrisk, lower SES samples of nondelinquents and delinquents.
Higher IQ was positively related to academic achievement for
both nondelinquents and delinquents. Significant Height × IQ
interactions for both nondelinquents and delinquents revealed
that being taller was positively related to academic achievement
1310
ZEBROWITZETAL.
Figure 3. Academic achievement of nondelinquents and delinquents as a function of height and IQ. High,
average, and low values of each variable represent one standard deviation above the mean, the mean, and
one standard deviation below the mean, respectively (Aiken & West, 1991; Cohen & Cohen, 1983).
only for those boys who were above average in IQ (see Figure 3).
Sample Differences
Although we predicted no differences in the effects of appearance on the academic achievement of delinquents and nondelinquents, we compared the regression coefficients for delinquents
and nondelinquents in each case in which an effect was obtained
for one group and not the other (Cohen & Cohen, 1983, p. 111 ).
The results revealed that the Babyface × IQ interaction effect
was significantly greater for delinquents than nondelinquents (z
= 2.25, p = .02) whereas the Attractiveness × Risk effect was
greater for the nondelinquents (z = 2.09, p = .04).
Discussion
The results of Study 2 reveal that babyfaceness can affect
academic performance even in lower class samples. However,
in these samples, unlike the high-IQ, middle-class sample examined in Study 1, babyfaceness appears to foster compensatory
high achievement only for those individuals who are relatively
more advantaged than their peers either in SES or intelligence.
Babyfaceness was positively related to academic achievement
for delinquents who were above average in IQ and for nondelinquents who were above average in SES for this sample. On
the other hand, delinquents who w e r e below average IQ and
nondelinquents who were below average in SES for this sample
provided evidence consistent with a self-fulfilling prophecy effect: The more babyfaced they were, the lower their academic
achievement. It thus appears that babyfaced boys will show
positive compensation effects consistent with a self-defeating
prophecy when they are able to do so either by virtue of their
intelligence or a minimally favorable socioeconomic environment. However, a babyface can lead to negative self-fulfilling
prophecy effects in the domain of academic performance when
boys are low in intelligence or subject to unfavorable environmental conditions (cf. Labov, 1972; Steele & Aronson, 1995;
Willis, 1977).
As in Study 1, attractiveness had no effect on the academic
achievement of boys in the delinquent sample. Moreover, contrary to the predicted self-fulfilling prophecy effect, attractiveness was negatively related to achievement for nondelinquent
boys who came from low- or average-risk families, whereas
those from high-risk families showed low achievement regardless of appearance. The unexpected finding that attractiveness
undermined the achievement advantages of coming from a lowrisk family may reflect the social activities of more attractive
boys. Past research has indicated that attractive individuals are
more popular (Feingold, 1992; Langlois et al., 1996). The
greater popularity of more attractive boys may have distracted
them from their schoolwork or strengthened peer group influences, which could work against high academic achievement in
this Iow-SES sample (e.g., Dornbusch et al., 1996; Labov, 1972;
Ogbu, 1986; Willis, 1977).
Although no a priori predictions were made regarding differences in the effects of appearance on the academic achievement
of delinquents versus nondelinquents, we did observe some differences. These results should be interpreted with caution because they were unpredicted as well as weak. The stronger
Attractiveness × Risk interaction among nondelinquents may
reflect the fact that delinquents with even low family risk were
at a risk level comparable to that of the high-risk nondelinquents,
the group that showed no effect of attractiveness. A possible
explanation for the stronger Babyface X IQ interaction among
BRIGHT, BAD, BABYFACED BOYS
delinquents is that boys who are at greater risk for delinquent
behavior respond more strongly to the babyface stereotype. This
seems plausible in the context of the maturity gap proposed by
Moffit (1993). Those adolescents who are motivated by the
maturity gap to engage in delinquent behavior may also be
more motivated than nondelinquents to compensate for negative
expectations associated with the babyface stereotype.
Study 3
Study 3 investigated the relationship between appearance and
criminal behavior. Such a connection is ubiquitous in writings
of fiction and is also found in the tenets of criminal anthropologists of the late 19th century. In particular, Lombroso argued
that criminals could be recognized by their morphological resemblance to apes (Lombroso & Ferrero, 1895). Although
Lombroso's biological theory of morphology and criminality
holds little credibility today, modern theorists have suggested
that appearance may influence antisocial behavior by means of
self-fulfilling prophecy effects. Indeed, intervention programs
in prisons have provided plastic surgery to criminals on the
assumption that a more attractive appearance would reduce recidivism (Pick, 1948; Thompson, 1990). The rationale for such
programs has been that the negative expectancies held for unattractive individuals coupled with their unfavorable social outcomes may foster antisocial behavior. Consistent with this suggestion, a recent recta-analysis revealed significantly better psychological adjustment among more attractive children and
adolescents, based on studies that included measures of social
skills, depression, anxiety, and freedom from juvenile delinquency (Langlois et al., 1996). More directly pertinent to delinquency, Cavior and Howard (1973) found that both Black and
White delinquents were significantly lower in facial attractiveness than nondellnquents of the same race, and Agnew
(1984) also found a negative correlation between attractiveness
and delinquency. However, methodological shortcomings in
these studies make it difficult to draw confident conclusions.
For example, attractiveness ratings were based on lower quality
photographs for delinquents in the study by Cavior and Howard
(1973), and attractiveness ratings in the Agnew (1984) study
were made at the end of a 2-hr interview and could have been
influenced by the contents of the interview. The present study
addressed the shortcomings of past research by eliminating confounding variables in the assessment of facial appearance.
Whereas stereotypes of unattractive people may contribute to
antisocial behavior by means of self-fulfilling prophecy effects,
stereotypes of babyfaced people would diminish such behavior
if they too are self-fulfilling. The traits of warmth, nalvett,
and honesty that are attributed to babyfaced individuals are
not conducive to criminal behavior. However, recent theories of
delinquency provide reason to expect the babyface stereotype
to produce self-defeating rather than self-fulfilling prophecy effects. Moffitt (1993) distinguished between two categories of
delinquents: "life-course persistent" and "adolescent limited."
For life-course persistent delinquents, antisocial behavior
emerges at an early age and reflects the early interaction of
neuropsychological problems and a pathological environment.
For adolescent-limited delinquents, antisocial behavior is due to
a maturity gap that, as mentioned above, can cause adolescents
1311
to covet what they perceive to be adult privileges (e.g., sex,
drinking, ownership of material goods). Because these privileges are not readily attainable, many adolescents resort to delinquent behavior to attain them. It seems reasonable to propose
that babyfaced boys would experience the maturity gap more
keenly than their mature-faced peers, thus putting them at greater
risk for antisocial behavior. Higher rates of antisocial behavior
among those who are more babyfaced also can b e predicted
f r o m other theories of delinquency. In particular, Emler and
Reicher (1995) argued that delinquency derives from a desire
to promote the reputation of being tough and hard, something
that would be more motivating to babyfaced boys, who are seen
as submissive and weak.
Drawing on the CCS archives that were used in Study 2, we
examined the effects of appearance on two indexes of criminality: (a) the likelihood of being a delinquent versus a nondelinquent, and (b) the frequency of criminal behavior among those
who were classified as delinquent. Because we expected appearance to have the strongest influence on the criminal behavior
of adolescent onset delinquents, we included early onset of criminal behavior as a control variable in the analysis predicting the
frequency of criminal behavior among the delinquents. We also
investigated effects of appearance on the likelihood of being a
delinquent not only for all delinquents but also for a subset of
adolescent onset delinquents.
The frequency of criminal behavior shown by delinquents had
been previously assessed at three time periods: from birth to age
17 (Time 1 ), from ages 17 to 25 (Time 2), and from ages 25
to 32 (Time 3). Predictions regarding the effects of appearance
varied across these three time periods. At Time 1, delinquents
were relatively homogeneous in their level of criminal activity.
All met the criterion of being inmates at a state correctional
school, and the majority had two to four criminal convictions
when they entered the study. This homogeneity was appropriate
for the original aims of the Glueck and Glueck (1950) study,
which compared delinquents to nondelinquents. Howeva; it yields
a restricted range of criminal behavior within the dehnquent sampie; we therefore expected that appearance effects would be
weak, even with delinquency onset controlled. The effects of
appearance were expected to be more marked at Time 2, when
greater variability in criminal charges was possible and when the
effects of appearance on responses to the maturity gap would
still be operative. The possibility of greater variability in criminal
charges would continue into Time 3. However, we expected the
tendency for babyfaceness to exacerbate criminal behavior in
response to the maturity gap to be less apparent at Time 3 because
participants were well beyond the adolescent deprivation of adult
privileges. Low attractiveness may continue to predict greater
criminal behavior at Tune 3, because it can have negative social
and psychological consequences quite apart from the maturity
gap. Finally, we expected onset of delinquency to be a stronger
predictor of criminal behavior at Time 3 than at the earlier time
periods, because many adolescent onset delinquents should have
desisted from criminal activity by Time 3, whereas early onset
delinquents would persist.
We predicted that boys who were highly babyfaced or unattractive would be more prevalent in the group of delinquents
than in the nondelinquent control group. We also predicted that
delinquents who were more babyfaced or less attractive would
1312
ZEBROWITZ ET AL.
show a higher frequency of criminal behavior. For the reasons
outlined above, the latter effects were expected to be most
marked at Time 2. On the basis of previous results for this
sample as well as of other studies, we also predicted that muscular boys would be more likely to be delinquent and, if delinquent,
would show a higher rate of criminal behavior (Glueck &
Glueck, 1950, 1968; Hartl et al., 1982). Although no predictions
were made for the effects of height, we included this variable
for comparability to the other studies. Finally, the interactions
of facial appearance with height, muscularity, IQ, family risk,
and SES were examined, as in Studies 1 and 2. We also investigated interactions of height and muscularity with the demographic predictors, although no predictions were made for these
exploratory analyses.
Method
The methodology for Study 3 was identical to that used in Study 2
with two exceptions. First, the criterion variables were delinquency
status and the number of criminal charges accrued by delinquents at
Time 1, Time 2, and Time 3. Second, we included onset of delinquency
as an additional predictor variable in the analysis of crime frequency.
In addition to these changes in the predictor and criterion variables, it
should be noted that we used appearance ratings at Time 1 to predict
behavior at Times 2 and 3 because no appearance data were available
at the later ages• Justification for this procedure is provided by evidence
that the babyfaceness and attractiveness of men in the IGS sample relative
to their peers remained stable from childhood through the early 30s
(Zebrowitz, Olson, et al., 1993). Moreover, recent research has revealed
that appearance assessed at Time 1 in the CCS sample continued to
influence outcomes at Time 3. In particular, Time 1 babyface ratings
predicted which delinquents remain unmarried at Time 3 (Lee & Zebrowitz, 1998).
Table 7
Summary of Logistic Regression Analysis Predicting
Delinquency Status (N = 961)
Predictor variable
Block 1
Babyfaceness
Attractiveness
Height
Muscularity
Block 2
Babyfaceness
Attractiveness
Height
Muscularity
IQ
SES
Age
Family risk
Block 3
Babyfaceness
Attractiveness
Height
Muscularity
IQ
SES
Age
Family risk
Babyfaceness x SES
Attractiveness X Muscularity
Attractiveness X Height
Attractiveness × SES
Attractiveness X Family Risk
Height x SES
Height x Family Risk
Muscularity x Family Risk
Muscularity X SES
B
SE B
Wald's X2(1)
- 0.10
0.03
- 0.16
0.83
.05
.04
.09
.17
4.65*
0.42
3.20#
24•35***
-0.02
0.06
0.10
0.95
-0.01
-0.21
0.04
0.56
.05
.04
.10
.18
.01
.05
.03
.06
0.18
1.64
0.91
28.07***
1.41
15.16"**
2.70t
99•25***
-0.01
0.05
0.11
1.12
-0.01
.05
.04
-0.19
0.04
0.58
-0.07
0.24
0.06
0.05
0.05
-0.09
0.10
0.24
-0.20
• 10
.19
.01
.06
.03
.06
.04
•I0
.05
.03
.03
.08
.06
.13
.14
0.04
1.33
1.05
33.36***
1•31
11•67***
2.00
98.80***
3.09t
5.80*
1.51
2.72t
2.08
1.33
2.61
3.42#
1•94
Note. SES = socioeconomic status•
Predictor Variable
Following the procedure used by Sampson and Laub (1993), we
classified delinquents as early onset if they had a record of persistent
antisocial behavior prior to 8 years of age according to self-reports,
parent reports, or both extracted from the CCS archives and as adolescent onset if they had no record of early antisocial behavior. This yielded
a total of 160 early onset and 304 adolescent onset delinquents with
photos in the archives. The analysis of crime frequency included delinquency onset as a predictor with a value of 1 assigned to participants
who showed early onset of delinquent behavior and a value of 0 assigned
to participants who did not.
Criterion Variables
Delinquency status. Delinquency status was designated by assigning
a value of I to participants in the delinquent sample and a value of 0
to participants in the nondelinquent sample.
Crime frequency. We created a measure of crime frequency from
an itemization of all charges compiled by Sampson and Laub (1993)
from the criminal history data of each delinquent from birth to age 17
(Time 1), from ages 17 to 25 (Time 2), and from ages 25 to 32 (Time
3). This measure was not available for nondelinquents.
Results
t P ~ .10 (marginally significant).
*p < .05. ***p < .001.
predictors entered in three blocks, as in Study 2. Means and
standard deviations of criterion variables for nondelinquents and
delinquents are presented in Table 5. We performed the analysis
of delinquency status with all delinquents included as well as
with only adolescent onset delinquents included. Only the former analysis will be reported because the results were virtually
identical, probably because the large majority of delinquents
were in the late onset group.
Overall effects. The results of the hierarchical logistic regression predicting delinquency status are presented in Table 7.
The block of appearance predictors accounted for 5% of the
variance in delinquency status and represented a significant improvement in prediction above the base rate, X2(4, N = 961 )
= 55.85, p < .001. The demographic predictors accounted for an
additional 14% of the variance explained and added significant
improvement above that of the appearance predictors, X2(4, N
= 961 ) = 161.01, p < .001. The block of interactions explained
an additional 2% of the variance and added significant improvement above that of the appearance and demographic predictors,
X2(9, N = 961) = 18.08, p < .05. 4
Delinquency Status
To assess the contribution of appearance to delinquency stares, we conducted a hierarchical logistic regression analysis with
4 The percentage of variance accounted for by the model for delinquency status is a pseudo-R 2, calculated according to the formula sug-
BRIGHT, BAD, BABYFACED BOYS
1313
Figure 4. Likelihood of being a delinquent as a function of babyfaceness and socioeconomic status (SES),
attractiveness and SES, attractiveness and muscularity, and muscularity and family risk. High, average, and
low values of each variable represent one standard deviation above the mean, the mean, and one standard
deviation below the mean, respectively (Aiken & West, 1991; Cohen & Cohen, 1983).
Babyface effects. Although there was an effect of babyfaceness when the block of appearance predictors was entered alone,
this tendency for babyfaced boys to be less likely to have delinquent status was eliminated when the demographic predictors
and interaction effects were entered into the equation. However,
gested by Aldrich and Nelson ( 1984): R 2 = c/(N + c), where c is the
chi-square statistic for overall goodness of fit.
a Babyface × SES interaction remained in the equation (see
Figure 4). Babyfaced boys who were below average in SES for
this sample were more likely to be delinquent than their maturefaced peers, consistent with a compensation effect. However,
babyfaced boys who were above average in SES were less likely
to be delinquent, consistent with a self-fulfilling prophecy effect.
Attractiveness effects. There was no overall effect of attractiveness on delinquent status when the block of appearance predictors was entered alone or when the other predictors were
1314
ZEBROWITZ ET AL.
added to the equation. However, there was an Attractiveness x
SES effect and an Attractiveness x Muscularity effect. As
shown in Figure 4, at/ractiveness was positively related to delinquent status for boys who were high in SES, with a slight
reversal for those who were low, and it was positively related
to delinquent status for boys who were high in muscularity, with
a slight reversal for those who were low.
Other effects. I n a replication o f past findings, there was a
significant effect of muscularity, with more muscular boys being
more likely to be in the delinquent group. Lower SES boys and
boys from higher risk families were also more likely to be in
the delinquent group. A Muscularity x Family Risk interaction
revealed that the greater tendency for more muscular boys to be
in the delinquent group was weakest when family risk was low
(see Figure 4 ) .
in Table 4. The results of the hierarchical regression analysis
predicting crime frequency for delinquents at Time 1, Time 2,
and Time 3 are presented in Table 8. The block o f appearance
predictors accounted for less than 1% of the explained variance
at Time 1 ( F < 1), 2% of the explained variance at Time 2,
F ( 4 , 430) = 2.38, p = .05, and l % of the variance at Time 3,
F ( 4 , 413 ) = 1.26, p = .28. The block of demographic predictors
accounted for an additional 7% of the variance explained at
Time 1, F ( 5 , 4 5 4 ) = 6.92, p < .001, an additional 3% of the
variance at Time 2, F ( 5 , 4 2 5 ) = 2 . 6 0 , p < .05, and an additional
4 % of the variance at Time 3, F ( 5 , 4 0 8 ) = 3.92, p < .01.
The block of trimmed interaction predictors accounted for an
additional 1% of the variance at Time 1, F ( 3 , 450) = 1.55, p
= .20, an additional 2% of the variance at Time 2, F ( 3 , 422)
= 3.15, p < .01, and an additional 2% of the variance at Time
3, F ( 4 , 4 0 4 ) = 1.78, p = .13.
Babyface effects. Consistent with predictions, babyfaceness
had no effect at Time 1 or Time 3 when either the block of
Crime Frequency
Overall effects. Correlations among the predictor and criterion variables at Time 1, Time 2, and Time 3 are presented
Table 8
Summary of Regression Analyses Predicting Number of Criminal Charges
for Delinquents at lime 1, Time 2, and Time 3
Time 1
Predictor variable
Block 1a
Babyfaceness
Attractiveness
Height
Muscularity
Block 2 b
Babyfaceness
Attractiveness
Height
Muscularity
IQ
SES
Age
Family risk
Early onset
Block 3c
Babyfaceness
Attractiveness
Height
Muscularity
IQ
SES
Age
Family risk
Early onset
Babyfaceness X IQ
Babyfaceness × Height
Attractiveness × Height
Attractiveness × Muscularity
Height × IQ
Height x Family Risk
Muscularity × IQ
Time 2
B
SE B
B
SE B
#
B
SE B
-0.06
0.10
-0.23
-0.07
.11
.10
.24
.42
-.03
.05
-.06
-.01
0.31
0.06
-0.13
1.09
.18
.16
.39
:74
.10t
.02
-.02
.08
0.03
0.00
-0.38
0.65
.17
.15
.37
.70
.01
.00
-.07
.05
-0.08
0.09
-0.10
-0.08
0.00
-0.08
-0.19
0.49
0.18
.11
.10
.24
.41
.01
.11
.06
.12
.28
-.04
0.38
.04
0.02
-.03
-0.04
-.01
1.34
.00
0.01
-.03
0.32
-.17"**
0.14
.19"**
0.46
.03
0.72
.18
.15
.39
.74
.02
.18
.09
.20
.46
.12"
0.09
.01 -0.04
-.01
-0.22
.10t
0.91
.02
0.01
.09t
0.07
.08
0.08
.11"
0.44
.08
1.45
.17
.15
.37
.70
.02
.17
.09
.18
.43
.03
-.01
-.04
.08
.03
.02
.05
.12"
.17"**
-0.06
0.09
-0.08
-0.04
0.00
-0.07
-0.19
0.51
0.22
0.01
.11
.10
.24
.41
.01
.11
.06
.12
.28
.01
.18
.16
.39
.79
.02
.18
.09
.20
.45
.17
.15
.37
.71
.02
.17
.09
~18
.43
.01
.05
-.01
-.03
.09
.04
.02
.05
.13"
.17"**
.09
.15
.10t
0.14
-.01
-0.04
-.02
-0.15
.06
1.12
.02
0.01
.07
0.08
.07
0.08
.10"
0.47
.08
1.45
0.02
-.08t
-0.11
.10
-.03
0.31
.04
-0.03
-.02
-0.10
-.01
0.82
.00
0.01
-.03
0.27
-.16"**
0.12
.19"**
0.42
.04
0.74
.04
-0.25
-.05
-0.95
.42
-.12"
.03
.19
.06
.16"
.06
.14"
0.21
.13
#
Time 3
.08t
0.08
.05
.07
0.06
0.22
0.13
#
Note. For Time 1, n = 464; for Time 2, n = 435; for Time 3, n = 418. SES = socioeconomic status.
a For Time 1, F(4, 459) = 0.54; for Time 2, F(4, 430) = 2.38, p < .05; for Time 3, F(4, 413) = 1.26.
b For Time 1, F(9, 454) = 4.10, p < .001; for Time 2, 17(9, 425) = 2.52, p < .01; for Time 3, F(9, 408)
- 2.76, p < .01. CForTimel, F(12,451)=3.47, p<.OO1;forTime2, F(12,422)=2.71, p<.Ol;
for Time 3, F(13, 404) = 2.47, p < .01.
"~p < .10 (marginally significant). * p < .05. ***p < .001.
BRIGHT, BAD, BABYFACED BOYS
appearance predictors was entered alone or in the final equation
with demographic variables and interaction effects included. On
the other hand, babyfaced delinquents had more criminal
charges than their more mature-faced peers at Time 2 in all
three blocks. However, this effect was qualified by a Babyface
× Height interaction in the final equation. As shown in Figure
5, the tendency for babyfaced boys to accrue more criminal
charges was most pronounced when they were also short.
Attractiveness effects. There was no overall effect of attractiveness on criminal behavior at Time 1, Time 2, or Time 3
when either the block of appearance predictors was entered
alone or in the final equations. However there was a significant
Attractiveness × Muscularity interaction at Time 2. As shown
in Figure 5, less attractive boys committed more crimes when
the boys were highly muscular but fewer crimes when the boys
were low in nmscularity.
Other effects. A younger age of entry into the study predicted more criminal charges at Time 1. This finding probably
reflects the selection criteria in the study rather than anything
of theoretical interest. In particular, boys who entered the study
at 12 or 13 years of age with three criminal convictions would
be likely to accrue more total charges in the Time 1 period than
those who entered at age 15 or 16 with the same number of
convictions. Consistent with this methodological explanation,
there was no association between age of entry into the study and
criminal behavior at Time 2 or Time 3. As expected, delinquency
onset predicted the frequency of criminal behavior at Time 3,
when life-course persistent delinquents were expected to continue criminal activity and adolescent onset delinquents were
expected to desist. On the other hand, delinquency onset was
not a significant predictor of criminal behavior at Time 1 or at
1315
Time 2, when both adolescent onset delinquents and life-course
persistent delinquents were expected to engage in criminal behavior. Higher family risk factors also predicted more criminal
behavior at Time 1, Time 2, and Time 3.
Discussion
Study 3 revealed that babyfaceness can produce negative
compensatory behaviors as well as positive ones. The stereotype
of babyfaced individuals as weak and nonthreatening was disconfirmed by babyfaced boys from lower SES backgrounds,
who were more likely than their mature-faced peers to be in the
delinquent group. On the other hand, babyfaced boys from
higher SES backgrounds tended to confirm the benign stereotype, showing less likelihood of being a delinquent than their
mature-faced peers. These findings provide an interesting complement to the results of Study 2, where it was shown that higher
SES babyfaced nondelinquents show slightly higher academic
achievement, suggesting that they compensate for the babyface
stereotype in prosocial ways. Babyfaceness influenced not only
boys' delinquency status, but also the frequency of criminal
behavior among those who were in the delinquent group. The
effect of a babyface on delinquent status and crime frequency
is consistent with the suggestion that delinquency derives from
the desire to promote the reputation that one is tough (Emler &
Reicher, 1995) as well as with Moffit's (1993) hypothesis that
delinquency may derive from a maturity gap that causes adolescents to engage in behaviors that will provide them with the
adult privileges they crave. Insofar as such privileges are less
available to babyfaced than to mature-faced adolescents, the
former may suffer the maturity gap more acutely and show more
Figure 5. Criminal charges incurred by delinquents at Time 2 as a function of babyfaceness and height
and attractiveness and muscularity. High, average, and low values of each variable represent one standard
deviation above the mean, the mean, and one standard deviation below the mean, respectively (Aiken &
West, 1991; Cohen & Cohen, 1983).
1316
ZEBROWrrz ET AL.
antisocial behavior in an effort to bridge that gap. The fact that
babyfaceness was a stronger predictor of criminal behavior at
Time 2 among boys who were short supports this argument,
because short stature is likely to exacerbate the maturity gap.
That babyfaceness no longer predicted criminal behavior at
Time 3, when those in the delinquent group were well beyond
adolescence, also supports this argument.
Contrary to self-fulfilling prophecy theory, attractiveness was
positively related to delinquency among boys who were above
average in muscularity or SES for this sample, with a slight
negative relationship among boys who were low in rrmscularity
or SES. Although these results are inconsistent with previous
evidence that more attractive adolescents are less likely to be
delinquents (e.g., Agnew, 1984; Cavior & Howard, 1973), it
has been noted that there are serious methodological shortcomings in that research. Moreover, the positive relationship between
attractiveness and delinquent status is consistent with the finding
that more attractive delinquent boys in the present sample were
more likely to belong to a gang (Zebrowitz & Lee, in press).
The influence of attractiveness on gang membership is in accord
with the considerable evidence that attractive people are more
popular with their peers (e.g., Feingold, 1992; Langlois et al.,
1996). Thus, in a sample of adolescents where popularity can
yield acceptance into a gang, attractiveness serves to increase
the likelihood that a boy who is relatively high in nmscularity
or SES will become a delinquent. Why attractiveness was a risk
factor for delinquency among boys who were high in SES for
this sample but not for those who were low is uncertain. Perhaps
factors other than facial appearance play a more significant
role in peer acceptance among lower SES boys. The fact that
attractiveness was a risk factor for delinquency among boys
who were highly muscular but not for those who were low in
muscularity is not surprising, because popularity is likely to be
higher for boys who are both attractive and muscular, particularly in a Iow-SES sample (Clausen, 1975; Labov, 1972; Willis,
1977). Although the combination of high attractiveness and muscularity increased the likelihood of being a delinquent, these
two appearance qualities had a different effect on the number
of criminal charges incurred at Time 2 by those who were
delinquent. It was the low-attractive, high-rrmscular delinquents
, who had the most criminal charges. Thus, whereas high attractiveness was a risk factor for highly muscular boys becoming
delinquent, low attractiveness was a risk factor for high levels
of criminal behavior among highly muscular boys who were
delinquent. This may reflect a selection effect. If attractive boys
are more successful at gaining access to delinquent groups, the
unattractive boys who do become delinquent may need to commit more crimes to prove themselves among their peers, and
those who are muscular are most able to do so.
General Discussion
The three present studies provide support for the hypothesis
that babyfaced adolescent boys compensate for the expectation
that they will exhibit childlike warmth and weakness in the
physical, social, and intellectual domains by behaving contrary
to these undesirable expectations. Rather than showing behavior
consistent with a self-fulfilling prophecy effect, babyfaced boys
defeated the prophecy that they would be intellectually weak
by showing higher academic achievement than their maturefaced peers. They also defeated the prophecy that they would
be warm and physically and socially weak by showing more
antisocial behavior than their mature-faced peers. The likelihood
of showing these compensation effects was moderated by other
variables influencing the boys' ability or motivation to do so.
In particular, compensation in the achievement domain was
shown by middle-class babyfaced boys, who were higher in
intelligence than their mature-faced peers because they were
more motivated, and it was shown by lower class babyfaced
boys who were at the top of their group in SES or intelligence.
Babyfaced boys who lacked the social background or intelligence to compensate for the expectation that they would be
intellectually weak fulfilled those expectations with lower academic performance than their mature-faced peers. Similarly,
babyfaced boys compensated by showing a greater likelihood
of becoming delinquent only when their Iow-SES status already
pushed them in this direction. Once they were in the delinquent
group, babyfaced boys accrued more criminal charges, particularly if they were short, further supporting the compensation
hypothesis. On the other hand, babyfaced boys from higher SES
backgrounds that are less conducive to delinquent behavior were
less likely than their mature-faced peers to be delinquent, thus
fulfilling stereotyped expectations about babyfaced individuals.
As expected, there was no evidence to indicate that babyfaced
adolescent girls compensate for the expectation that they will
exhibit childlike intellectual weakness, which is consistent with
the suggestion that girls are less rankled by such expectations.
However, firm conclusions regarding this sex difference will
require additional research with more statistical power, because
the effect for boys was not significantly greater than the effect
for girls.
The stereotype disconfirming behaviors shown by babyfaced
adolescent boys in the present studies and the stereotype inconsistent personality traits found by Zebrowitz et al. (1998) contrast with some other evidence for behavioral confirmation of
the babyface stereotype. In particular, babyfaced college men
have been found to score lower on a personality test of aggression (Berry, 1990, 1991; Berry & Brownlow, 1989), a finding
that is consistent with the perception of babyfaced individuals
as nonthreatening (Berry & McArthur, 1986). Babyfaced college men also have reported less control in their social interactions, higher self-disclosure and greater intimacy, findings that
are consistent with the stereotype that babyfaced individuals
are submissive, dependent, and warm (Berry & Landry, 1997).
Babyfaced college women have also shown more behaviors
indicative of warmth when interacting with partners of the same
sex, such as positive facial expressions and closer interpersonal
distances, and fewer behaviors indicative of dominance, such as
direct gaze and gesturing (Hansen & Berry, 1995).
There are several possible explanations for the inconsistent
effects of babyfaceness. One possibility is that they reflect age
differences in the various samples. Perhaps the tendency f o r
babyfaced adolescents to compensate for the maturity gap with
stereotype disconfirming behavior is replaced with confirmatory
behavior by the time they are of college age. However, this
possibility seems unlikely because the measures of educational
attainment in Study 1 and criminal charges at Time 2 in Study
3 were assessed when the boys were of college age. Another
BRIGHT, BAD, BABYFACED BOYS
possible explanation for the inconsistent effects of babyfaceness
is cohort differences. Perhaps the tendency for babyfaced adolescents to compensate for the maturity gap with stereotype
disconfirming behavior was true for cohorts who were adolescents in the 1940s, like the IGS and CCS samples, but not for
more recent cohorts of adolescents, who have shown stereotype
confirming behavior. This possibility seems unlikely because
researchers have used the suggestion that the adolescent maturity
gap contributes to compensatory delinquent behavior to explain
delinquency in recent cohorts (Moffitt, 1993). Moreover, research has shown that babyfaced individuals in current cohorts
experience social outcomes that assail their comt, ztence and
autonomy (for a pertinent review, see Montepare & Zebrowitz,
1998).
A plausible explanation for the divergent effects of babyfaceness is differences in the particular behaviors assessed. Stereotype-consistent behavior may be more likely to occur when
babyfaced individuals are directly influenced by the expectations
of those with whom they are interacting. For example, babyfaced
men may report that they feel less control in their social interactions (Berry & Landry, 1997) not because they are particularly
submissive but because those with whom they interact see them
as submissive and attempt to dominate them. Indeed, research
has shown that highly assertive men and women, who tend to
dominate others, prefer to date babyfaced rather than maturefaced people (Hadden & Brownlow, 1991). Events such as low
levels of interpersonal control are consistent with a situationally
circumscribed self-fulfilling prophecy effect, but these very
events may in fact contribute to global self-defeating prophecy
effects (cf. Zebrowitz, 1997). In particular, the perception of
low control in social interactions may, in the long run, motivate
behavior in babyfaced men like that observed in the present
studies: high educational achievement or criminal behavior that
can compensate for the low power babyfaced individuals experience in their social interactions. Whereas differences in the impact of direct social influence may account for some of the
divergent effects of babyfaceness, they cannot account for the
finding that more babyfaced college men scored lower on personality tests of aggression (e.g., Berry, 1990, 1991; Berry &
Brownlow, 1989), whereas more babyfaced delinquents showed
more criminal behavior in Study 3. What may account for these
discrepancies are sample differences. In particular, babyfaced
men who compensate with antisocial behavior may not make it
to college. Consistent with this suggestion, the Iow-SES babyfaced boys who showed compensation with a higher likelihood
of delinquency in Study 3 failed to compensate with achievement in Study 2, showing instead stereotype-consistent low academic achievement.
Like the effects of babyfaceness in the present studies, the
effects of attractiveness did not always support the expectation
of a self-fulfilling prophecy effect, whereby more attractive individuals would show more socially desirable behaviors. More
attractive girls did show higher educational achievement than
their less attractive peers, albeit only when considering girls
who were below average in IQ. Low attractiveness not only
worked against high achievement by low-IQ girls, but also fos~
tered higher rates of criminal behavior among highly muscular
delinquent boys, consistent with a self-fulfilling prophecy effect.
On the other hand, there was also evidence consistent with a
1317
self-defeating prophecy effect of attractiveness on both achievement and antisocial behavior. More attractive lower class boys
showed lower academic achievement and more attractive lower
class boys were more likely to be delinquents, albeit only when
considering boys who were above average for their group in
SES or muscularity. As noted earlier, the negative impact of
attractiveness on academic achievement may reflect adverse social influences on attractive adolescents in a lower class sample.
Similarly, the tendency for attractiveness to increase the likelihood of being a delinquent may reflect the positive influence
that attractiveness had on gang membership (Zebrowitz & Lee,
in press). It thus appears that the popularity produced by attractiveness can foster behavior inconsistent with the attractiveness
halo effect when such behavior is encouraged by peers. As noted
earlier, peer pressures may also account for the finding that
unattractive muscular boys who do become delinquent commit
more crimes, because they may need to do so to prove themselves among their peers.
Although the present research was concerned primarily with
the effects of facial appearance, some effects of height and
muscularity are worth mentioning because they shed light on
the process of expectancy confirmation; Like the self-defeating
prophecy effect on achievement shown by babyfaced boys, the
self-fulfilling prophecy effects shown by taller individuals were
confined to those whose IQ levels were conducive to high performance in the academic domain. The finding that educational
achievement was lower for highly muscular boys at the lower
end of the SES spectrum in the IGS sample and that delinquency
was more likely among highly muscular boys from the CCS
sample suggests that lower SES boys who are muscular may
confirm the "tough guy" and "dumb j o c k " stereotypes.
It should be acknowledged that the significant effects of babyfaceness and other appearance factors on boys' academic
achievement and criminal behavior were small. However, they
are notable when one considers the myriad factors that can
influence academic achievement and criminal behavior in the
real world. Moreover, the relationship between babyfaceness
and achievement was not trivial. In Study 1, the average boy in
the top quartile of babyfaceness had a mean education level
that included schooling beyond college graduation, whereas the
average boy in the bottom quartile did not complete his college
education. This comparison highlights one reason why effect
sizes were smaller in the present research than is typically the
case in experimental studies. Whereas the latter examine effects
for individuals representing the extremes of an appearance dimension, the present research examined effects for individuals
representing the entire range of babyfaceness and attractiveness.
The fact that this minimal manipulation of appearance accounted for significant variance makes a small effect more impressive (Prentice & Miller, 1992), as does the fact that it should
be difficult for variables like appearance to influence the dependent variables of academic achievement mad criminal behavior
that were the focus of this research. Indeed, the impact of apt
pearance becomes more impressive when placed in the context
of other rational predictors, which it sometimes surpassed. The
Babyface × IQ effect on delinquents' academic achievement in
Study 2 was stronger than the effect of SES or family risk
factors, and the effect of babyfaceness on criminal charges at
Time 2 in Study 3 was equal to the effect of family risk factors,
1318
ZEBROWITZ ET AL.
and stronger than the effect of SES or an early history of delinquent behavior.
Although the higher academic performance and greater criminal behavior shown by babyfaced boys in the present studies
have been construed as reflecting compensation for disagreeable
expectations, other possible explanations merit consideration.
One possible explanation for the observed correlations between
appearance and behavior is suggested by recent evidence that
babyfaced delinquents have poorer relationships with their families, receiving less supervision from their mothers, poorer quality discipline from their fathers, and experiencing weaker parent-child attachment (Zebrowitz & Lee, in press). One could
make the reasonable argument that these experiences contribute
to the higher levels of criminal behavior among babyfaced delinquents. However, it is difficult to argue that they also contribute
to the higher levels of academic achievement among these boys.
Another possible explanation for the correlations between
appearance and behavior that were observed in the present studies is that they reflect biological rather than social influences.
In particular, Arcus and Kagan (1995) have found that children
who are biologically predisposed to develop a shy, inhibited
temperament have narrower upper faces than those who are
biologically predisposed to be more extroverted. It is likely that
broader faced children are perceived as more babyfaced, in
which case it may be that it is the biologically based temperament of babyfaced individuals that predisposes them to assert
themselves by showing higher academic performance and
greater criminal behavior. Assessing the viability of this explanation would require longitudinal data that permit an investigation of the relationship between babyfaceness and behavior at
early ages as well as the lagged effects across time.
It should be noted that a biological explanation and a compensation, self-defeating prophecy explanation for the present findings are not mutually exclusive. It is possible that there is a
biological link between babyfaceness and assertiveness that can
foster achievement or antisocial behavior and also that babyfaced boys are likely to react negatively to the babyface stereotype, particularly given such a basic temperament. Consistent
with this reasoning, previous research that has used the IGS
middle-class sample found that more babyfaced boys as young
as 10 years old engaged in more negativity, lying, and quarreling
than their more mature-faced peers, suggesting the possibility
of a biological contribution to such behavior. At the same time,
babyfaceness in childhood predicted more assertive and hostile
behavior in adolescence even when controlling for childhood
levels of such behavior, suggesting compensatory behavior consistent with a self-defeating prophecy effect (Zebrowitz et al.,
1998).
Clearly, it i s important for future research to more definitively
establish the mechanism by which facial appearance is related
to behavior. The current literature is too piecemeal to accomplish
this important goal. With one exception (Snyder et al., 1977),
studies have examined only one link in the chain from appearance to behavior. Some, like the present study, have examined
the link between appearance and behavior. Others have examined the link between appearance and possible mediators of
behavior, such as social outcomes and self-concept. Still others
have examined the link between possible mediating variables
and behavior (see Zebrowitz, 1997, for a pertinent review).
What is needed is research that examines all of these links
within a single sample of people who are followed from an
early age. Although an experimental research paradigm is often
viewed as the best way to uncover causal mechanisms, longitudinal data will prove more useful 'for unpacking self-fulfilling and
self-defeating prophecy effects for two reasons. First, if begun
at a sufficiently early age, it will permit an assessment of the
relationship between appearance and behavior before social influences can play a significant role. Second, it will capture
effects of social influences that may take considerable time to
develop, as is likely in the case of self-fulfilling or self-defeating
prophecy effects of appearance on stable dispositions as opposed to their effects on situationally circumscribed behaviors.
Such a longitudinal research paradigm is equally important for
studying the contribution of other expectancies to self-fulfilling
prophecy effects (cf. Jussim, 1991).
In conclusion, the present study illustrates that the relationship of appearance and behavior is more complex than suggested in past theoretical emphases on self-fulfilling prophecy
effects. Although babyfaced boys who were at the low end of
the distribution in SES or IQ did confirm the expectation that
they would be intellectually weaker than their mature-faced
peers, such effects were not shown for more advantaged boys,
whose high academic performance was consistent with a selfdefeating prophecy effect. Also consistent with a self-defeating
prophecy effect, low-SES babyfaced boys were more likely than
their mature-faced peers to be delinquent, and babyfaced delinquents tended to show more criminal behavior than their maturefaced peers. Thus, babyfaced young men may compensate for
expectations that run counter to masculine gender norms, showing behavior that refutes those expectations. These results extend
past evidence that babyfaced adolescent boys are more assertive
and contentious than their mature-faced peers by demonstrating
that these stereotype-inconsistent personality traits are paralleled by positive achievement behavior as well as negative criminal behavior. They also establish the generality of such compensation effects across a broad socioeconomic range. The effects
of attractiveness in the present studies also suggest caveats to
past emphases on self-fulfilling prophecy effects. Positive effects of attractiveness on educational achievement may be limited to cases in which it can mitigate other factors that work
against high achievement, like low IQ. Moreover, attractiveness
may have negative effects on educational achievement when its
positive effects on popularity work at cross purposes to academic performance, as it may in a lower class sample. Finally,
rather than uniformly promoting the development of prosocial
behavior, attractiveness may promote delinquent behavior when
its usual effect on popularity increases the likelihood of gang
membership. Clearly, appearance stereotypes do not always
yield self-fulfilling prophecy effects, as has been commonly
assumed.
References
Adams, G. R. (1977). Physical attractiveness research: Toward a developmental social psychology of beauty. Human Development, 20, 217239.
Adams, G. R., Montemayor, R., & Gullotta, T. P. (Eds.). (1996). Psychosocial development during adolescence. Thousand Oaks, CA:
Sage.
BRIGHT, BAD, BABYFACED BOYS
Agnew, R. (1984). Appearance and delinquency. Criminology: An Interdisciplinary Journal 22, 421-440.
Aiken, L. S., & West, S. G. (1991). Multiple regression: Testing and
interpreting interactions. Newbury Park, CA: Sage.
Aldrich, J. H., & Nelson, E D. (1984). Linear probability, Iogit, and
probit models. Beverly Hills, CA: Sage.
Arcus, D., & Kagan, J. (I995). Temperament and craniofacial variation
in the first two years. Child Development, 66, 1529-1540.
Bakan, D. (1966). The duality of human existence: Isolation and communion in the Western man. Boston: Beacon Press.
Berry, D. S. (1990). Taking people at face value: Evidence for the kernel
of truth hypothesis. Social Cognition, 8, 343-361.
Berry, D.S. (1991). Accuracy in social perception: Contributions of
facial and vocal information. Journal of Personality and Social Psychology, 61, 298-308.
Berry, D. S., & Brownlow, S. (1989). Were the physiognomists right?
Personality correlates of facial babyishness. Personality and Social
Psychology Bulletin, 15, 266-279.
Berry, D. S., & Landry, J. C. (1997). Facial maturity and dally social
interactions. Journal of Personality and Social Psychology, 72, 570580.
Berry, D. S., & McArthur, L. Z. (1985). Some components and consequences of a babyface. Journal of Personality and Social Psychology,
48, 312-323.
Berry, D. S., & McArthur, L. Z. (1986). Perceiving character in faces:
The impact of age-related craniofacial changes on social perception.
Psychological Bulletin, 100, 3-18.
Block, J. (1971). Lives through time. Berkeley, CA: Bancroft Books.
Block, J. (1978). The Q-sort method in personality assessment and
psychiatric research. Palo Alto, CA: Consulting Psychologists Press.~
(Original work published 1961 )
Bond, M. H. (1972). Effect of an impression set on subsequent behavior.
Journal of Personality and Social Psychology, 24, 301-305.
Brackbill, Y., & NeviU, D. D. ( 1981 ). Parental expectations of achievemant as affected by children's height. Merrill-Palmer Quarterly, 27,
429-441.
Caspi, A., Elder, G. H., & Bern, D. (1987). Moving against the world:
Life-course patterns of explosive children. Developmental Psychology,
23, 308-313.
Cavior, N., & Howard, L. (1973). Facial attractiveness and juvenile
delinquency. Journal of Abnormal Child Psychology, 1, 202-213.
Clausen, J. A. (1975). The social meaning of differential physical and
sexual maturation. In S. E. Dragastin & G. H. Elder, Jr. (Eds.)Adolescence in the life cycle (pp. 25-47). New York: Wiley.
Clausen, J. A. (1993). American lives: Looking back at the children of
the Great Depression. New York: Free Press.
Cohen, J., & Cohen, P. (1983). Applied multiple regression/correlation
analysis for the behavioral sciences ( 2nd ed.). Hillsdale, NJ: Erlbaum.
Collins, M. A., & Zebrowitz, L: A. (1995). The contributions of appearance to occupations outcomes in civilian and military settings. Journal
of Applied Social Psychology, 25, 129-163.
Cross, S. E., & Madson, L. (1997). Models of self: Self construals and
gender. Psychological Bulletin, 122, 5-37.
Dornbusch, S. M., Herman, M. R., & Morley, J. A. (1996). Domains of
adolescent achievement. In G.R. Adams, R. Montemayor, & T.P.
GuUotta (Eds.), Psychosocial development during adolescence (pp.
181-231 ). Thousand Oaks, CA: Sage.
Eichom, D. H. (1981). Samples and procedures. In D. H. Eichorn, J. A.
Clausen, N. Haan, M. P. Honzik, & P. H. Mussen (Eds.), Present and
past in middle life (pp. 33-51). New York: Academic Press.
Eichorn, D. H., Clausen, J. A., Haan, N., Honzik, M. P., & Mussen, P. H.
(Eds.). ( 1981 ). Present andpast in middle life. New York: Academic
Press.
Eisenberg, N., Roth, K., Bryniarski, K.A., & Murray, E. (1984). Sex
1319
differences in the relationship of height to children's actual and attributed social and cognitive competencies. Sex Roles, II, 719-734.
Elder, G.H., Jr., Ngnyen, T.V., & Caspi, A. (1985). Linking family
hardship to children's lives. Child Development, 56, 361-375.
Emler, N., & Reicher, S. (1995). Adolescence and delinquency. Oxford,
England: Blackwell.
Feingold, A. (1992). Good looking people are not what we think. Psychological Bulletin, 111, 304-341.
Friedman, H., & Zebrowitz, L. (1992). The contribution of typical sex
differences in facial maturity to sex role stereotypes. Personality and
Social Psychology Bulletin, 18, 430-438.
Ginsburg, G. S., Bempechat, J., & Chung, Y. E. (1992). Parent influences on children's mathematics. In T. G. Sticht, M. J. Beeler, & B. A.
McDonald (Eds.), The intergenerational transfer of cognitive skills
(Vol. 2, pp. 91-121). Norwood, NJ: Ablex.
Glueck, S., & Glueck, E. (1950). Unraveling juvenile delinquency. New
York: The Commonwealth Fund.
Glueck, S., & Glueck, E. (1968). Delinquents and nondelinquents in
perspective. Cambridge, MA: Harvard University Press.
Hadden, S. B., & Brownlow, S. ( 1991, March). The impact of facial
structure and assertiveness on dating choice. Poster presented at the
annual meeting of the Southeastern Psychological Association, New
Orleans, LA.
Hansen, J. S., & Berry, D. S. (1995). Facial maturity and nonverbal
behavior in same-sex dyads. Unpublished manuscript. Southern Methodist University, Dallas, TX.
Hartl, E. M., Monnelly, E. P., & Elderkin, R. D. (1982). Physique and
delinquent behavior: A"thirty-year follow-up of William H. Sheldon's
Varieties of Delinquent Youth. New York: Academic Press.
Hollingshead, A.B., & Redlich, E (1958). SOcial class and mental
illness. New York: Wiley.
Johnson, E W. ( 1991 ). Biological factors and psychometric intelligence:
A review. Genetic, Social and General Psychology Monographs, 117,
313-357.
Jones, M. C., Bayley, N., Macfarlane, J.W., & Honzik, M. P. (Eds.).
(1971). The course of human development. New York: Wiley.
Jussim, L. (1991). Social perception and social reality: A reflectionconstruction model. Psychological Review, 98, 54-73.
Jussim, L., Eccles, J., & Madon, S. (1996). Social perception, social
stereotypes, and teacher expectations: Accuracy and the quest for the
powerful self-fulfilling prophecy. In M. P. Zarma (Ed.), Advances in
experimental social psychology (Vol. 28, pp. 281-388). San Diego,
CA: Academic Press.
Kalick, S.M., Zebrowitz, L.A., Langlois, J.H., & Johnson, R.M.
(1998). Does human facial attractiveness honestly advertise health?
Psychological Science, 9, 8-13.
Kuczmarski, R.J., Flegal, K.M., Campbell, S.M., & Johnson, C.L.
(1994). Increasing prevalence of overweight among U.S. adults: The
national health and nutrition examination surveys 1960-1991. Journal
of the American Medical Association, 272, 205-211.
Labov, W. (1972). Language in the inner city: Studies in the Black
English vernacular. Philadelphia: University of Pennsylvania Press.
Langlois, J.H. (1986). From the eye of the beholder to behavioral
reality: The development of social behavior and social relations as a
function of physical attractiveness. In C. E Herman, M. R Zanna, &
E. T. Higgins (Eds.), Physical appearance, stigma, and social behavior: The Ontario symposium (Vol. 3, pp. 23-51). Hillsdale, NJ:
Erlbanm.
Langlois, J., Kalakanis, L., Rubenstein, A., Larson, A., Hallam, M., &
Smoot, M. (1996, July). The myths of beauty: A meta-analytic review.
Paper presented at the annual meeting of the American Psychological
Society, San Francisco, CA.
Lee, S. Y., & Zebrowitz, L. A. (1998). [Appearance and marital status].
Unpublished research. Brandeis University.
1320
ZEBROWlTZETAL.
Lombroso, C., & Ferrero, W. (1895). The female offender. Littleton,
CO: Fred B. Rothman.
Madon, S., Jussim, L., & Eccles, J. (1997). In search of the powerful
self-fulfilling prophecy. Journal of Personality and Social Psychology,
72, 791-809.
McArthur, L. Z., & Apatow, K. ( 1983-1984). Impressions of babyfaced
adults. Social Cognition, 2, 315-342.
Miller, C. T., & Myers, A. M. (1998). Compensating for prejudice: How
heavyweight people (and others) control outcomes despite prejudice.
In J. K. Swim & C. Stangor (Eds.) Prejudice: The target's perspective
(pp. 191,218). San Diego, CA: Academic Press.
Miller, C. T., Rothblum, E. D., Felicio, D., & Brand, P. (1995). Compensating for stigma: Obese and nonobese women's reactions to being
visible. Personality and Social Psychology Bulletin, 21, 1093-1106.
Miller, C. T. & Rudiger, L. (1997, October). Compensation for prejudice: How the stigmatized obtain desired outcomes despite prejudice.
Paper presented at the annual meeting of the Society of Experimental
Social Psychology, Toronto, Ontario, Canada.
Moffitt, T. (1993). Adolescence-limited and life-course-persistent antisocial behavior: A developmental taxonomy. Psychological Review, 100,
674-701.
Montepare, J. M., & Zebrowitz, L. A. (1998). Person perception comes
of age: The salience and significance of age in social judgment. In
M. P. Zanna (Ed.) Advances in experimental social psychology (Vol.
30, pp. 93-161). New York: Academic Press.
Najjar, M. E, & Rowland, M. (1987). Anthropometric reference data
and prevalence of overweight: United States, 1976-1980. Htal health
statistics (Series 11, No. 238). Washington, DC: U.S. Department of
Health, Education, and Welfare.
Ogbu, J. U. (1986). The consequences of the American caste system.
In U. Neisser (Ed.), The school achievement of minority children (pp.
19-56). Hillsdale, NJ: Erlbaum.
Pick, J. E (1948). Ten years of plastic surgery in a penal institution:
Preliminary report. Journal of the International College of Surgeons,
11, 315-319.
Prentice, D. A., & Miller, D. T. (1992). When small effects are impressive. Psychological Bulletin, 112, 160-164.
Sampson, R. J., & Laub, J. H. (1993). Crime in the making: Pathways
and turning points through life. Cambridge, MA: Harvard University
Press.
Sheldon, W. H., Stevens, S. S., & "lhcker, W.B. (1940). The varieties
of human physique: An introduction to constitutional psychology. New
York: Harper.
Snyder, M. (1992). Motivational foundations of behavioral confirmation. In M. P. Zarma (Ed.), Advances in experimental social psychology (Vol. 25, pp. 67-114). San Diego, CA: Academic Press.
Snyder, M., Tanke, E., & Berscheid, E. (1977). Social perception and
interpersonal behavior: On the self-fulfilling nature of social stereotypes. Journal of Personality and Social Psychology, 35, 656-666.
Sorrell, G. T. & Nowak, C. A. ( 1981 ). The role of physical attractiveness
as a contributor to individual development. In R. M. Lemer & N. A.
Bnsch-Rossnagel (Eds.), Individuals as producers of their development: A life span perspective (pp. 389-446). London: Academic
Press.
Steele, C. M., & Aronson, J. (1995). Stereotype threat and the intellectual test performance of African Americans. Journal of Personality
and Social Psychology, 53, 797-811.
Swarm, W. B., Jr., & Ely, R. (1984). A battle of wills: Self-verification
versus behavioral confirmations. Journal of Personality and Social
Psychology, 53, 1287-1302.
Thompson, K.M. (1990). Refacing inmates: A critical appraisal of
plastic surgery programs in prison. Criminal Justice and Behavior,
17, 448-466.
Willis, P. (1977). Learning to labor. New York: Columbia University
Press.
Zebrowitz, L. A. (1997). Reading faces: Window to the soul? Boulder,
CO: Westview Press.
Zebrowitz, L. A., Brownlow, S., & Olson, K. (1992). Babytalk to the
babyfaced. Journal of Nonverbal Behavior, 16, 143-158.
Zebrowitz, L. A., Collins, M. A., & Dutta, R. (1998). The relationship
between appearance and personality across the lifespan. Personality
and Social Psychology Bulletin, 24, 736-749.
Zebrowitz, L. A., Kendall-Tackett, K., & Fafel, J. ( 1991 ). The influence
of children's facial maturity on parental expectations and punishments.
Journal of Experimental Child Psychology, 52, 221-238.
Zebrowitz, L. A., & Lee, S. Y. (in press). Appearance, stereotype incongruent behavior, and social relationships. Personality and Social Psychology Bulletin.
Zebrowitz, L. A., & Montepare, J. M. (1992). Impressions ofbabyfaced
individuals across the lifespan. Developmental Psychology, 28, 11431152.
Zebrowitz, L. A., Montepare, J. M., & Lee, H. K. (1993). They don't
all look alike: Differentiating same versus other race individuals. Journal of Personality and Social Psychology, 65, 85-101.
Zebrowitz, L. A., Olson, K., & Hoffman, K. (1993). The stability of
babyfaceness and attractiveness across the lifespan. Journal of Personality and Social Psychology, 64, 453-466.
Zebrowitz, L. A., Voinescu, L., & Collins, M. A. (1996). 'Wide-eyed'
and 'crooked-faced': Determinants of perceived and real honesty
across the lifespan. Personality and Social Psychology Bulletin, 22,
1258-1269.
Received December 17, 1997
Revision received June 16, 1998
Accepted June 29, 1998 •