Content uploaded by Sarah Banchefsky
Author content
All content in this area was uploaded by Sarah Banchefsky on Oct 26, 2017
Content may be subject to copyright.
ORIGINAL ARTICLE
But You Don’t Look Like A Scientist!: Women Scientists
with Feminine Appearance are Deemed Less Likely
to be Scientists
Sarah Banchefsky
1
&Jacob Westfall
2
&Bernadette Park
1
&Charles M. Judd
1
#Springer Science+Business Media New York 2016
Abstract Two studies examined whether subtle variations in
feminine appearance erroneously convey a woman’s likeli-
hood of being a scientist. Eighty photos (half women) of
tenured/tenure-track science, technology, engineering, and
math (STEM) faculty at elite research universities were select-
ed from the Internet. Participants, naïve to the targets’occu-
pations, rated the photos on femininity and likelihood of being
a scientist and an early childhood educator. Linear mixed
model analysis treated both participants and stimuli as random
factors, enabling generalization to other samples of partici-
pants and other samples of stimuli. Feminine appearance af-
fected career judgments for female scientists (with increasing
femininity decreasing the perceived likelihood of being a sci-
entist and increasing the perceived likelihood of being an early
childhood educator), but had no effect on judgments of male
scientists. Study 2 replicated these findings with several key
procedural modifications: the presentation of the stimuli was
manipulated to either be blocked by gender or completely
randomized, questions pertaining to the stimuli’s appearance
were removed, and a third career judgment likelihood rating
was added to avoid tradeoffs between scientist and early child-
hood educator. In both studies, results suggest that for women
pursuing STEM, feminine appearance may erroneously signal
that they are not well suited for science.
Keywords Gendered appearance .Stereotypes .Femininity .
Face perception .Physical appearance .Science .STEM .
Sexism
In the summer of 2015, San Francisco based tech firm
OneLogin featured photos of their own employees on adver-
tising posters aimed at recruiting more engineers. One of the
featured female employees, Isis Wenger, raised doubts about
the campaign’s veracity; apparently, some people found it im-
probable that this young woman could be an engineer simply
because shedid not look like one—she was far “too attractive”
to be a “real engineer.”In response to this criticism, the
hashtag BiLookLikeAnEngineer^went viral on Twitter, with
engineers of different ages, races, and genders posting their
self-portraits in an effort to challenge notions of what engi-
neers are Bsupposed to^look like (Zamon 2015).
Women remain disconcertingly underrepresented in STEM
fields (science, technology, engineering, and math; National
Science Foundation [NSF] 2015), in part due to differential
gender roles (Eagly and Wood 2012; Eccles 1987), life goals
(Ceci and Williams 2011; Diekman et al. 2010), and gender
bias in STEM (Moss-Racusin et al. 2012). This research has
focused almost exclusively on categorical gender gaps be-
tween women’sandmen’s experiences and outcomes in
STEM. The present research, on the other hand, examines
how bias might vary within gender categories. Specifically,
we test whether real, accomplished female scientists judged
as more feminine in appearance are also deemed less likely to
be scientists.
The perceived incompatibility between femininity and sci-
ence is a recognized issue with negative consequences for
women. Over a 5-year period, 80 % of female and 72 % of
male undergraduate engineering majors surveyed agreed that
the belief that women in science or technical fields are
*Sarah Banchefsky
sarah.banchefsky@colorado.edu
1
Department of Psychology and Neuroscience, University of
Colorado Boulder, Muenzinger D244, Boulder, CO 80309-0345,
USA
2
Department of Psychology, The University of Texas at Austin, SEA
4.208, 108 E. Dean Keeton Stop A8000, Austin, TX 78712-1043,
USA
Sex Roles
DOI 10.1007/s11199-016-0586-1
unfeminine is a problem for women pursuing these careers;
indeed, the more that a woman perceived that this was a prob-
lem, the less satisfied she was in her field (Hartman and
Hartman 2008).WomeninSTEMenvironmentshavereport-
ed feeling unable to present themselves in a stereotypically
feminine manner (e.g., wearing a skirt, expressing emotions)
because they do not want to draw attention to their gender or
are apprehensive that they will seem unsuitable for a STEM
career (Hewlett et al. 2008;Proninetal.2004).
Indeed, appearance has a powerful and immediate effect on
person perception. In just fractions of a second, people come
to remarkably similar conclusions about a person’sattractive-
ness, aggression, likability, trustworthiness, and competence
(Willis and Todorov 2006). Like the major social categories of
gender and race (Ito and Urland 2003), variations in facial
appearance are automatically processed and automatically ac-
tivate stereotypes. For example, stronger stereotypic Black
features (e.g., broad nose, thick lips) activate Black stereo-
types (e.g., close with family, failing in school; Blair et al.
2002,2004a). Similarly, baby-faced features (e.g., large round
eyes and foreheads) activate youthful characteristics (e.g.,
naïve, submissive; Zebrowitz et al. 1991). These stereotypes
have potentially dramatic implications. For example, statisti-
cally controlling for the target’s perceived categorical race
(i.e., White or Black) and the seriousness of their crimes,
convicted felons with more Afrocentric features receive
harsher criminal sentences (Blair et al. 2004a), and baby-
faced individuals are seen as less suitable for leadership jobs
(Zebrowitz et al. 1991).
Research in non-STEM domains has examined the impact
of femininity on judgments. For example, more feminine-
appearing women were (accurately) judged as more likely to
be Republicans than Democrats (Carpinella and Johnson
2013); moreover, female politicians with increasingly mascu-
line facial appearance were less likely to receive votes, partic-
ularly among conservative constituents (Hehman et al. 2014).
This work suggests that, at least in the political realm, women
are rewarded for looking more feminine. The present research
examines whether naturalistic variations in feminine appear-
ance (i.e., based on physiological characteristics such as facial
bone structure, as well as gender performance such as the use
of make-up, hair style, etc.) impacts the perceived likelihood
that a woman is a scientist. We hypothesized that women
judged as increasingly feminine in appearance will also be
judged as less likely to be scientists. We did not have a strong
rationale from which to hypothesize about male targets. On
the one hand, feminine-appearing men may similarly activate
feminine gender stereotypes, thereby decreasing their judged
likelihood of being scientists. On the other hand, they may
also trigger Bnerdy^male stereotypes (Cheng 2008)thatalign
with stereotypes about the types of people who populate
STEM domains (Cheryan et al. 2011), thereby increasing their
judged likelihood of being scientists.
To rigorously test this hypothesis, we attended to the con-
ceptual and methodological issues of stimulus sampling
(Wells and Windschitl 1999) by employing a large stimulus
set and treating the stimuli as random in the analysis (Judd
et al. 2012). In traditional analyses, only participants (or more
rarely, stimuli) are treated as the sole random factor in the
analyses—variation across responses due to the other factor
are averaged and thus ignored. In contrast, the cross-random
model used in the present studies takes into account that we
have two different samples from two theoretical populations
of interest about which we would like to make inferences—a
sample of participants and a sample of stimuli.
Most of the prior research on the effects of femininity has
employed a very small sample of stimuli (e.g., one woman
dressed in a feminine manner or a neutral manner; Betz and
Sekaquaptewa 2012; six photographs, Sczesny and Kühnen
2004) that are sometimes not naturalistic (e.g., computer-
generated; Friedman and Zebrowitz 1992), raising the ques-
tion of whether results are simply due to the specific stimuli
presented. Moreover, in part due to small samples of stimuli,
previous studies have failed to examine the continuum of gen-
dered appearance (i.e., from masculine to feminine), instead
only focusing on extreme examples of masculinity and femi-
ninity on either end of the spectrum. This limited sampling not
only constricts power and the generalizability of the effects
but also is not a realistic representation of real people encoun-
tered on a daily basis.
In contrast, the present methodological approach offers
several strengths and advantages. First, we treated variations in
gendered appearance continuously rather than operationalizing
or analyzing gendered appearance in a categorical way (see
Irwin and McClelland 2003). Second, we used a large sample
of photographs of real people, specifically tenured or tenure-
track faculty members in STEM departments at elite U.S. uni-
versities. Third, we treated both stimuli (i.e., faces) and partici-
pants as random factors (Baayen et al. 2008; Judd et al. 2012).
Critically, our analysis permits generalization from our specific
sample of faces to other samples of faces that we might have
used (Clark 1973; Judd et al. 2012).
Study 1
Method
Participants
Participants were 51 U.S.-based workers on Amazon.com’s
Mechanical Turk (25 men, 26 women; 78 % White, 12 %
Asian, 4 % biracial, 4 % Latino, and 2 % Black; Mean
age = 34.92, SD = 13.71, range = 18–63 years old) who were
compensated $0.75 for their time. An additional four partici-
pants failed two or more of four basic attention checks that
Sex Roles
were embedded within the survey (e.g., BIs this person’shair
blonde or brunette?^) and were excluded from the sample. An
approximate power analysis based on the calculations given
by Westfall et al. (2014), assuming a counterbalanced design
and using their default variance partitioning coefficients, sug-
gests that an experiment with 80 stimuli and 50 participants
should have 80 % power to detect an effect size as small as
Cohen’sd=.33.
Stimuli
Stimuli consisted of 80 photographs (40 men, 40 women) of
tenured/tenure-track faculty in elite STEM departments in
U.S. universities. Programs were selected according to U.S.
News and World Report’s rankings of premier graduate pro-
grams in various STEM disciplines. Our stimulus selection
rule was to select from each program’s website the first
high-quality, crisp, color photos of faculty who were smiling
and making direct eye contact with the camera (some websites
did not present photos). Moreover, in order to avoid variations
in judged likelihood of being a scientist due to perceived race,
all faculty selected appeared to be White. To ensure the faces
were naturalistic, they were not cropped (e.g., to remove hair)
and were presented in color. Example stimuli are available
from the authors upon request. In total, nine research univer-
sities (Massachusetts Institute of Technology, California
Institute of Technology, Princeton University, Stanford
University, University of Texas at Austin, University of
California Berkeley, University of Illinois, Cornell
University, and Carnegie-Mellon University), and 15 STEM
programs (aerospace engineering, astronomy, astrophysics,
bioengineering, chemical engineering, chemistry, civil engi-
neering, computer engineering, computer science, electrical
engineering, engineering, environmental engineering, math,
mathematics, and physics) were represented.
Although the majority of photographs selected were simply
the first encountered that fulfilled the preceding criteria, about
ten faces of each gender were strategically selected by the
researchers to maximize representation along the spectrum
from masculine to feminine appearance. That is, about five
highly feminine- and five highly masculine-appearing women
and men (relative to the rest of the stimulus sample) were
chosen based on a holistic first impression of gendered ap-
pearance. We made these selections because research demon-
strates that interaction effects (which we were hypothesizing)
can be difficult to detect without adequate variation in the
continuous variable of interest (i.e., gendered appearance);
we therefore made an effort to select some faces that clearly
varied in gendered appearance (McClelland and Judd 1993).
Nevertheless, the variation was naturalistic and the selected
individuals very much resemble typical people encountered in
everyday life. Specifically, subjectively masculine-appearing
individuals tended to have shorter hair and stronger facial
features (e.g., heavier jaws and brow-bones, larger noses),
whereas subjectively feminine-appearing individuals tended
to have longer hair and finer facial features (e.g., smaller jaws
and brow-bones). Some women wore jewelry (e.g., earrings, a
necklace) or subtle make-up (e.g., faint lipstick). For both
genders, although there was variation in clothing, no articles
were revealing or flashy; nearly all targets wore solid-colored
sweaters, tee-shirts or button-down shirts, or shirts with a sub-
tle pattern. Some men and women appeared to have on a
blazer, and some men wore a tie. Some of both genders wore
eyeglasses. Finally, for both genders, some photos were taken
inside offices or against a blank background, whereas others
were taken outdoors.
Importantly, the statistical model employed to analyze the
data (a mixed model treating both participants and stimuli as
random factors, described in the following) took into account
idiosyncratic differences among the stimuli, accounting for
this naturalistic variation. Put simply, the analysis enabled us
to detect whether judgments of femininity and career likeli-
hood were related over and above any unique variations be-
tween stimuli, eliminating the possibility that arbitrary varia-
tions among the photographs gave rise to the observed effects
(e.g., the relationship obtained between perceived femininity
and career likelihood; Judd et al. 2012).
Procedure
Participants were asked to evaluate 80 photographs of indi-
viduals who, as described previously, were in fact accom-
plished academic STEM scientists. Participants did not know
the targets’occupations, however, and were simply told that
the study was about first impressions, and that first impres-
sions are made very quickly and are often surprisingly accu-
rate. They were then asked to rate each photo on three 7-point
scales ranging from 1 (not at all)to7(very): masculine to
feminine, likable to unlikable, and unattractive to attractive,
in this fixed order. Note that although gendered appearance
was measured on a semantic differential scale from masculine
to feminine, we frequently refer to this variable as Bfeminine
appearance^because femininity is the primary construct of
interest.
Next, participants estimated the likelihood that the individ-
ual was a scientist, followed by the likelihood that the person
was an early childhood educator (henceforth referred to as
“teacher,”a profession that is97 % women and stereotypically
feminine; Carnevale et al. 2013), on 6-point scales ranging
from 1 (very unlikely)to6(very likely). Finally, they estimated
the age of the target, selecting one of eight 5-year ranges
starting at 25 years-old and ending at 60 and above.
Participants had as long as they desired to make the ratings
of each face before moving on. The target gender of the stim-
uli was blocked and counterbalanced (i.e., all women were
presented first or all men were presented first). Photos within
Sex Roles
each block were randomized, and each was presented on a
separate screen. Lastly, participants completed a series of de-
mographic questions.
Data Analysis
Mixed models were estimated in SAS® using Satterthwaite
approximate degrees of freedom. Initial analyses examined
whether participant’s gender or target’s gender order (i.e.,
whether participants rated men or women first) affected any
of the judgments. No effects were present so these variables
are dropped from subsequent analyses. To examine whether
gendered appearance impacted career ratings, career likeli-
hood was analyzed as a function of career (science vs. teacher,
contrast coded), face gender (male vs. female, contrast coded),
judged feminine appearance (mean-centered), and all possible
interactions. Data were analyzed using linear mixed models
with crossed random effects of participants and stimuli
(meaning that both participants and stimuli were treated as
random effects and every participant rated every stimulus;
Baayen et al. 2008;Juddetal.2012). All possible random
intercepts and slopes were estimated (Barr et al. 2013), but
the covariances between the random effects were not estimat-
ed due to convergence problems (estimating covariances in-
volved estimating an additional 81 parameters).
In addition to using mixed models in which both partici-
pants and stimuli were treated as random effects, Study 1
entailed another statistical advantage: because our experimen-
tal design involved every participant providing judgments of
gendered appearance for every face, this variable varied both
between and within participants, as well as between and with-
in faces. It was therefore possible to decompose gendered
appearance ratings into three distinct effects based on three
different sources of variance: the target effect (i.e., judgments
due to the face), the perceiver effect (i.e., judgments due to the
perceiver), and the relationship effect (i.e., judgments due to
the face and the perceiver). These effects are similar by anal-
ogy to effects estimated under the Social Relations Model
(Kenny 1994), which our analysis resembles. For the descrip-
tions that follow, let Fij denote the femininity rating given by
the i
th
participant to the j
th
face.
Case 1: Target.The target effect is the average level of
femininity that a given face elicits across all perceivers. It
asks whether certain faces are evaluated as more or less
feminine on average, relative to other faces, and how this
deviance affects judgments of career likelihood. It is com-
puted as
Fjand then mean-centered in the mixed model.
Case 2: Perceiver. The perceiver effect represents a par-
ticipant’s average rating tendency for femininity. It asks
whether certain participants on average evaluated faces as
more or less feminine, relative to other participants, and
how this deviance affects judgments of career likelihood.
It is computed as
Fiand then mean-centered in the mixed
model.
Case 3: Relationship. The relationship effect (i.e.,
Target ×Perceiver interaction) examines a perceiver’srat-
ing of a particular face, asking how much it deviates from
the face’s average femininity rating and the perceiver’s
average femininity rating tendency. It asks whether a giv-
en participant perceives a given face as more or less fem-
inine than would be expected and how this affects judg-
ments of career likelihood. It is computed as Fij−
Fj−
Fi
and then mean-centered in the mixed model (see
Raudenbush 2009; Rosnow and Rosenthal 1991).
Including each of these predictors allows an appraisal of
how femininity ratings due to stimuli, perceivers, and the re-
lationship between them each uniquely predicts judgments of
career likelihood. Importantly, decomposing the femininity
predictor in this way also avoids the problem of biased param-
eter estimates that can result from pooling together effects
from different levels of analysis (Bafumi and Gelman 2006;
Bell and Jones 2015).
We predicted that the three-way interaction of interest
(Target Gender × Career Type × Femininity) would emerge
for two of these effects: first, for the target effect, this interac-
tion would indicate that women who are judged as more fem-
inine than others, on average, are judged as less likely to be
scientists relative to teachers (Case 1, target effect); second, an
interaction involving femininity due to relationship would in-
dicate that when a given participant views a particular woman
as more feminine (over and above the participant’s typical
femininity ratings, as well as the face’s average femininity
rating), he or she also views that woman as less likely to be
a scientist relative to a teacher (Case 3, relationship effect); the
third possible effect (Case 2, perceiver effect) would mean that
those perceivers who on average see greater femininity across
all faces also judged all of the women on average as less likely
to be scientists. Although we did not hypothesize this effect,
we included it in the model so as to avoid biased parameter
estimates.
Results
Preliminary Analyses
In preliminary analyses that treated face as the unit of analysis
and averaged across participants, the male and female scien-
tists were perceived as about the same age, equally likable (the
unlikability rating was reverse-scored for interpretability in
Tab le 1), and equally attractive (see Table 1). Unsurprisingly,
female scientists were rated as significantly more feminine in
appearance than male scientists, and they also were rated as
Sex Roles
significantly more likely to be teachers. Notably, female and
male scientists were rated as equally likely to be scientists (see
Table 1). As depicted in Table 1, average judged feminine
appearance and attractiveness were highly positively correlated
for female scientists and less strongly but negatively correlated
for male scientists. Femininity was also negatively correlated
with age, but more so for male than female scientists. Due to
the high correlation between femininity and attractiveness for
female scientists (.89), and collinearity problems this
creates in the predictors, we did not include both simul-
taneously in analyses, instead analyzing their effects in
separate models. Ancillary analyses reported subsequent-
ly examined age and femininity simultaneously to con-
firm that perceived femininity affected career judgments
over and above perceived age.
Feminine Appearance and Career Likelihood
Tab le 2presents fixed effects output (effect sizes in the
form of unstandardized beta estimates are also
presented. Other effect size estimates are not presented
because there is no generally agreed upon definition of
standardized effect size estimates for mixed models;
Snijders and Bosker 1994). First, the hypothesized
three-way interaction among Target Gender, Career
Type, and Feminine Appearance was significant (target
effect), F(1, 97.9) = 6.10, p=.015. We broke down this
interaction by target gender, examining female and male
scientists separately. As hypothesized, feminine appear-
ance affected career judgments for female scientists (Career
Type × Femininity), F(1, 90.7) = 25.00, p<.001, but had no
impact on career judgments for male scientists (p=.53).
Consistent with our hypothesis, as the average rated feminine
appearance of a female scientist increased, she was judged as
significantly less likely to be a scientist, F(1, 7 8.3) = 12.67,
p<.001 (Fig. 1a), and significantly more likely to be a teacher,
F(1, 74.3) = 41.99, p<.001 (Fig. 1b). This pattern is shown by
the bold-type regression lines in each panel of Fig. 1,where
the slopes in the two panels for female scientists (but not male
scientists) are different from each other and are both signifi-
cantly different from zero.
A marginal three-way interaction emerged among
Target Gender, Career Type, and Feminine Appearance
(relationship effect), F(1, 48.5) = 3.96, p=.052. As be-
fore, the relationship effect of feminine appearance af-
fected career judgments for female scientists (Career
Type × Femininity), F(1, 66.2) = 12.60, p<.001, but
had no impact on career judgments for male scientists
(p=.80). This pattern is shown by the thin, shorter re-
gression lines in each panel of Fig. 1, which depict the
within-stimulus regressions of career likelihood on fem-
ininity (with perceiver effects removed) for each face.
Breaking this interaction down further by career indicat-
ed that the interaction for female scientists was driven
by teacher judgments: when a given participant viewed
a given woman as having more feminine appearance
than expected (based on the perceiver’s typical feminin-
ity rating and the face’s typical femininity rating), he or
she also rated her as more likely to be a teacher, F(1,
5918) = 38.81, p<.001 (Fig. 1b); however, perceiving a
given face as more feminine in appearance than expect-
ed (for that participant and for that target) did not affect
the perceived likelihood of being a scientist (p=.97,
Fig. 1a). Thus the thin regression lines in Fig. 1show
a significant positive slope on average only for teacher
ratings of the female scientists.
There were no significant effects attributable to perceiver
differences in judged feminine appearance. Lower order ef-
fects emerged in the model, but all were qualified by the two
reported three-way interactions (see Table 2).
Tabl e 1 Mean ratings and
correlations by face gender on six
face dimensions in study 1
Feature dimension Mean (SD) Correlations
Female
faces
Malefaces1. 2.3.4. 5. 6.
1. Feminine 5.11 (1.03)
a
2.85 (.78)
b
–.89** .59** −.41** −.56** .75**
2. Attractive 4.25 (.92)
a
4.04 (.63)
a
−.48** –.69** −.63** −.61** .67**
3. Likeable 3.94 (.40)
a
3.92 (.44)
a
.28 .68** –−.45** −.25 .65**
4. Age 3.11 (1.16)
a
2.80
(1.16)
a
−.69** −.08 −.11 –.31* −.31*
5. Likelihood
scientist
3.90 (.43)
a
3.96 (.53)
a
−.13 −.29 .11 .50** –−.65**
6. Likelihood
teacher
3.93 (.37)
a
3.14 (.34)
b
−.01 .54** .76** −.37* −.25 –
SD Standard deviation. Means with different subscripts are significantly different, p< .05. Age ratings were
categorical and represented ranges in years (1 = 25–29; 2 = 30–35; 3 = 36–40; 4 = 41–45; 5 = 46–50; 6 = 51–55;
7=56–60; and 8 = 61+). Correlations are based on averages for each face. Correlations for female faces are above
the diagonal. Correlations for male faces are below the diagonal. *p<.05; **p<.01
Sex Roles
Tabl e 2 Mixed-models results
for fixed effects for career
likelihood judgments in study 1
Effect Estimate SE df t p
Intercept 3.697 .075 81.3 49.090 < .0001
Main effects
Career .331 .070 112 4.690 < .0001
Gender .192 .046 95.3 4.180 < .0001
Feminine_Face (Target effect) −.011 .030 78.4 −.370 .715
Feminine_Ss (Perceiver effect) .145 .116 47.6 1.250 .218
Feminine_Rel (Relationship effect) .032 .013 43.5 2.380 .022
Two-way interactions
Career × Gender −.054 .061 95.3 −.870 .385
Career × Feminine_Face (Target effect) −.146 .043 93.3 −3.410 .001
Career × Feminine_Ss (Perceiver effect) −.121 .081 45.7 −1.490 .142
Career × Feminine_Rel (Relationship effect) −.047 .017 59.6 −2.830 .006
Gender × Feminine_Face (Target effect) .031 .032 85.8 .960 .339
Gender × Feminine_Ss (Perceiver effect) −.049 .040 46.6 −1.220 .228
Gender × Feminine_Rel (Relationship effect) .031 .013 51.7 2.270 .028
Three-way interactions
Career × Gender × Feminine_Face (Target effect)
a
−.109 .044 97.9 −2.470 .015
Career × Gender × Feminine_Ss (Perceiver effect) .064 .052 40 1.240 .221
Career × Gender × Feminine_Rel (Relationship effect)
a
−.042 .021 48.5 −1.990 .053
Estimate unstandardized beta, SE standard error, df Satterthwaite approximate degrees of freedom
a
Indicates hypothesized effect of interest
Fig. 1 Plot of mixed model
results by target gender and
career. The points in each panel
represent the mean femininity and
likelihood ratings for each
stimulus face (i.e., the target
effects), and the bold regression
line in each panel of career
likelihood ratings on femininity
ratings represents the total target
effect in that panel. The thin,
shorter regression lines passing
through each target effect
represent the within-stimulus
regressionsofcareerlikelihood
on femininity (with perceiver
effects removed; i.e., the
relationship effects), the average
of which represents the total
relationship effect in that panel.
For both scientist and teacher
judgments, the bold lines are
significantly different from 0 for
female targets, but not for male
targets
Sex Roles
Ancillary Analyses
Because age was correlated with feminine appearance, and
because we were concerned that perhaps younger looking
individuals might be viewed as less likely to be scientists than
older ones, we also examined a model that controlled for the
target’s perceived age (again parsed into the same three
sources of variation and mean-centered). There were no sig-
nificant effects of perceived age on career judgments, and the
critical three-way interactions involving target gender, career
type, and feminine appearance remained unchanged when
controlling for age in the model.
The target’s mean attractiveness and feminine appearance
were very highly correlated (.89) for female scientists.
Because of potential collinearityproblems, we decided against
running models that included both as simultaneous predictors.
Instead, we estimated a separate model identical to that esti-
mated for feminine appearance but using attractiveness judg-
ments instead of gendered appearance judgments. This model
revealed significant two-way Career Type × Attractiveness
interactions (again for both target effect and relationship ef-
fect), indicating that more attractive scientists were seen as
less likely to be scientists and more likely to be teachers (target
effect: F(1, 105) = 28.09, p<.001; relationship effect: F(1,
55.7) = 10.86, p=.002). Unlike the effects of femininity, these
effects did not depend on target gender. These findings align
with research suggesting that people who pursue science are
stereotyped as unattractive (Hannover and Kessels 2004).
They also suggest that attractiveness is used as a cue for judg-
ing both men and women’s career likelihood, whereas only for
women is gendered appearance also used as an informative
career-likelihood cue.
Study 2
One concern in Study 1 is that asking participants to evaluate
the targets’appearance (e.g., attractiveness, femininity) may
have made these concepts especially salient or created pres-
sure to be consistent in how one related appearance judgments
and career likelihood judgments. Another potential concern
addressed in Study 2 is that the blocked presentation of the
stimuli (by target gender) produced excessive attention to
within-category variations in appearance. We hypothesized
that femininity would still be used as a cue to career-type even
when between-category differences in gender were made
more salient by presenting male and female targets inter-
spersed (see Blair et al. 2002). Finally, we were concerned
that having participants rate only two careers may have forced
participants to make a trade-off in judging career likelihood
that they would not have made if more careers had been
assessed. Study 2 addressed these concerns in addition to test-
ing the replicability of the effects in a larger sample.
Method
Participants
Because a between-subjects condition factor was added, a
larger sample of 214 people participated in the study on
Amazon’sMechanicalTurk(129women,84men;approxi-
mately 80 % White, 6 % Black, 4 % Latino, 5 % Asian, 4 %
Biracial, and 1 % Native American; Mean age = 36.27,
SD = 11.41, range = 18–68 years old). An approximate power
analysis using the same assumptions as in Study 1 (with a
slight adjustment reflecting the fact that participants in Study
2 did not make gendered appearance judgments) indicated that
these sample sizes should provide 80 % power to detect effect
sizes as small as d= .37 (Westfall et al. 2014). The survey took
about 20–30 min to complete and workers were paid $.75. No
attention checks were included, so no participants were ex-
cluded from analysis.
Design
Study 2 replicated Study 1 with the following alterations:
First, participants were randomly assigned to judge faces pre-
sented in either a blocked or mixed fashion with respect to
target gender. In the blocked condition, participants rated ei-
ther all women followed by all men or vice versa, with the
order of the target gender blocks counterbalanced, and faces
were presented in a randomized order within each block (as in
Study 1). In the mixed condition, all faces were presented in a
fully randomized order for each participant, theoretically mak-
ing the gender of the target more salient. Second, participants
only made career-likelihood judgments of each target, ensur-
ing that explicit considerations of femininity or other
appearance-related measures would not influence career-
likelihood judgments. Third, to make it less apparent that we
were examining a male stereotypic (scientist) and female ste-
reotypic career (teacher), participants first rated a relatively
gender-neutral career, journalist (64 % female; Carnevale
et al. 2013), for each target. Even in this more conservative
design, we hypothesized that perceivers would still use female
targets’feminine appearance as a career cue.
Procedure
The cover story was very similar to Study 1. Participants were
randomly assigned to judge faces either blocked by gender
(n= 103) or in a mixed presentation (n= 111). Participants
rated each face in terms of their likelihood of being a journal-
ist, scientist, and early childhood educator (teacher), in that
order, and again on 6-point scales from 1(very unlikely)to6
(very likely). Participants lastly completed the same demo-
graphics as in Study 1.
Sex Roles
Data Analysis
Career likelihood was analyzed as a function of career type
(scientist vs. non-scientist, and teacher vs. journalist, two sin-
gle degree of freedom contrasts), target gender (male vs. fe-
male, contrast coded), gendered appearance (mean-centered),
presentation (blocked vs. mixed, contrast coded), participant
gender (male vs. female, contrast coded), and all possible
interactions. The gendered appearance variable was based on
the average femininity rating for each face in Study 1. We
hypothesized that the three-way interaction of interest
(Target Gender × Career Type × Feminine Appearance) would
again emerge, and that although it might vary in strength as a
function of mixed versus blocked presentation, it would be
significant in both conditions. Data were again analyzed using
linear mixed models with crossed random effects of partici-
pants and stimuli; as before, all possible random intercepts and
slopes were estimated, but not covariances. Because per-
ceivers in Study 2 did not make gendered appearance judg-
ments, all femininity effects reported for Study 2 are Btarget
effects^(i.e., Bperceiver^and Brelationship^effects could not
be estimated).
Results
Feminine Appearance and Career Likelihood
Tab le 3presents the fixed effects of our hypothesized interac-
tions. First, the predicted significant three-way interaction
among Target Gender, Career Type (Science vs. Other), and
Feminine Appearance was found again, F(1, 81.4) = 26.94,
p<.001. As can be seen in Fig. 2, whereas feminine appear-
ance again affected career judgments for female scientists
(Science vs. Non-Scientist × Femininity), F(1, 78.5) = 43.16,
p<.001, it had no impact on career judgments for male sci-
entists (i.e., this two-way interaction was not significant for
men, p=.11). Consistent with our hypothesis, the feminine
appearance of female targets was negatively related to per-
ceived likelihood of being a scientist, F(1, 76.8) = 26.83,
p<.001, and positively related to perceived likelihood of be-
ing a non-scientist, F(1, 77.7) = 58. 37, p<.001. The lack of a
Teacher vs. Journalist × Target Gender × Feminine
Appearance interaction indicated that femininity affected ca-
reer likelihood judgments of journalist the same way that it
affected ratings of teacher.
The impact of feminine appearance was also moderated by
participant gender, F(1, 201) = 6.15, p=.01; although the crit-
ical three-way interaction was highly significant for both male
and female participants, it was stronger among female partic-
ipants (F(1, 86.4) = 32.49, p<.001) than among male partici-
pants, F(1, 94.8) = 19.54, p<.001. If anything, this suggests
that women perceivers especially may consider another
woman’s gendered appearance as a meaningful cue of her
career. However, because participants’gender did not moder-
ate the results in Study 1, this finding should be interpreted
with caution. A variety of unanticipated lower order effects
emerged, but all were importantly moderated by the predicted
three-way interaction (see Tables 4and 5).
Interestingly, the intermixed presentation of male and fe-
male stimuli enhanced categorical gender bias in career judg-
ments (regardless of feminine appearance). That is, a signifi-
cant Target Gender × Presentation × Science vs. Other
Tabl e 3 Mixed-models results
for predictors of interest fixed
effects for career likelihood
judgments in study 2
Effect Estimate SE df tp
Intercept 3.534 .052 229 68.00 <.0001
Main effects (Predictors of interest)
Scientist vs. Non-scientist (Scientist) .278 .053 100 5.23 <.0001
Teacher vs. Journalist (T vs. J) −.108 .057 115 −1.91 .058
Target gender .101 .037 91.1 2.70 .008
Femininity .064 .025 77.5 2.54 .013
Two-way interactions (Predictors of interest)
Scientist × Target gender .278 .050 81.2 −.19 .849
Scientist × Femininity −.108 .035 78.0 −2.68 .009
T vs. J × Target gender .101 .053 91.9 4.38 <.0001
T vs. J × Femininity .064 .036 82.5 1.05 .295
Target gender × Femininity .278 .026 82.7 6.49 <.0001
Three-way interactions (Predictors of interest)
Scientist × Target gender × Femininity
a
−.184 .035 81.4 −5.19 <.0001
T vs. J × Target gender × Femininity −.052 .037 86.2 −1.42 .159
Tvs. JTeacher versus Journalist, Estimate unstandardized beta, SE standard error; df Satterthwaite approximate
degrees of freedom
a
Indicates hypothesized effect of interest
Sex Roles
interaction indicated that differences in judged career likeli-
hood for male vs. female targets was stronger in the mixed
condition than in the blocked condition, F(1, 226) = 22.28,
p<.001 (see Table 4). Simple effects looking within target
revealed that male targets were judged as more likely to be
scientists (compared to other careers) in the mixed vs. blocked
presentation, F(1, 287) = 6.25, p= .013, whereas female tar-
gets were judged as less likely to be scientists (compared to
other careers) in the mixed vs. blocked presentation, F(1,
271) = 8.58, p=.004.
Ancillary Analyses
When age was included as a predictor (again using average
age ratings for each face from Study 1), the critical three-way
interaction among target, feminine appearance and career
remained highly significant. We also examined a model that
included perceived attractiveness rather than feminine appear-
ance (again using average attractiveness ratings for each face
from Study 1); consistent with Study 1, a Career Type ×
Attractiveness interaction indicated that more attractivetargets
were seen as less likely to be scientists and more likely to be
non-scientists, F(1, 99.6) = 94.28, p< .001. This did not
depend on target gender, suggesting that attractiveness affect-
ed career judgments similarly for male and female targets.
Discussion
Two studies examined how variation in judged gendered ap-
pearance of 80 real scientists related to judgments about their
likelihood of being a scientist. Participants were unaware that
the photographs they were judging were actually scientists;
rather, they were simply told that they were making first im-
pressions of individuals. Results showed that for female sci-
entists, but not male scientists, perceivers used gendered ap-
pearance as a cue about how likely they were to be scientists
(vs. early childhood educators/teachers or journalists). Study 2
demonstrated that this outcome was the case (a) regardless of
whether male and female scientists were presented in a
blocked or intermixed order; (b) when participants were not
asked to judge the person’s appearance prior to making career
judgments (i.e., when aspects of appearance were not made
salient); and (c) even when an additional, gender-neutral ca-
reer (journalist) was included in the career judgments along-
side scientist and early childhood educator. In both studies,
these results did not depend on participant gender.
Fig. 2 Plot of mixed model
results by target gender, career,
and presentation type. For female
targets, all slopes are significantly
different from 0, whereas for male
targets, all slopes are statistically
equivalent to 0
Sex Roles
Overall, the female scientist’s gendered appearance was
related to judgments about the likelihood of being in a
masculine-stereotypic career (science), a feminine-
stereotypic career (teacher), and even a relatively gender-
Tabl e 4 Fixed effects results for
presentation (blocked by gender
versus unblocked) on career
likelihood judgments, study 2
Presentation effects (Mixed vs. Blocked) Estimate SE df tp
Presentation −.067 .039 222 −1.72 .087
Presentation × Scientist −.003 .022 245 −.16 .875
Presentation × T vs. J −.039 .028 224 −1.4 .161
Presentation × Target gender .053 .015 380 3.52 .001
Presentation × Femininity −.016 .007 511 −2.08 .038
Presentation × Participant gender .004 .039 222 .09 .927
Presentation × Scientist × Target gender −.066 .014 226 −4.72 <.0001
Presentation × Scientist × Femininity .020 .009 167 2.32 .021
Presentation × Scientist × Participant gender −.017 .021 230 −.78 .433
Presentation × T vs. J × Target gender .069 .020 207 3.39 .001
Presentation × T vs. J × Femininity −.004 .011 145 −.37 .713
Presentation × T vs. J × Participant gender .048 .027 217 1.74 .084
Presentation × Target gender × Femininity −.019 .009 279 −2.1 .037
Presentation × Target gender × Participant gender −.019 .015 380 −1.27 .206
Presentation × Femininity × Participant gender −.002 .007 511 −.21 .835
a
Presentation × Scientist × Target gender × Femininity .017 .010 201 1.71 .090
Presentation × Scientist × Target gender × Participant gender −.013 .013 241 −.99 .324
Presentation × Scientist × Femininity × Participant gender .009 .008 188 1.16 .248
Presentation × T vs. J × Target gender × Femininity −.007 .013 167 −.55 .582
Presentation × T vs. J × Target gender × Participant gender .012 .020 204 .6 .550
Presentation × T vs. J × Femininity × Participant gender −.006 .011 145 −.56 .577
Presentation × Target gender × Femininity × Participant gender .000 .009 279 .03 .972
Presentation × Scientist × Target gender × Femininity ×
Participant gender
−.015 .009 203 −1.58 .116
Presentation × T vs. J × Target gender × Femininity ×
Participant gender
.004 .012 164 .36 .718
Tvs. JTeacher versus Journalist, Estimate unstandardized beta, SE standard error, df Satterthwaite approximate
degrees of freedom
a
Highest order effects of interest. Other effects of interest are in bold
Tabl e 5 Remaining fixed effects
results for gender on career
likelihood judgments, study 2
Effect Estimate SE df tp
Participant gender −.011 .039 222 −.28 .781
Participant gender × Scientist .037 .022 234 1.72 .087
Participant gender × T vs. J −.025 .028 223 −.89 .375
Participant gender × Target gender .000 .015 380 0 .996
Participant gender × Femininity .016 .007 511 2.16 .031
Participant gender × Scientist × Target gender .030 .013 234 2.26 .025
Participant gender × Science × Femininity −.022 .008 178 −2.84 .005
Participant gender × T vs. J × Target gender .003 .020 204 .17 .865
Participant gender × T vs. J × Femininity .025 .011 140 2.19 .030
Participant gender × Target gender × Femininity .013 .009 279 1.49 .138
a
Participant gender × Scientist × Target gender × Femininity −.023 .009 201 −2.48 .014
Participant gender × T vs. J × Target gender × Femininity .013 .013 163 1.07 .286
Tvs. JTeacher versus Journalist, SE standard error, df Satterthwaite approximate degrees of freedom
a
Highest order effect of interest
Sex Roles
neutral career, journalism. On the other hand, a male scien-
tists’gendered appearance was not related to career likelihood
regardless of career type. In short, a woman’sgenderedap-
pearance was used as a cue about her career in a way that a
man’s gendered appearance was not. In contrast to gendered
appearance, men and women alike who were judged as more
attractive were deemed less likely to be scientists and more
likely to be non-scientists. Aligning with previous research,
feminine appearance and attractiveness were strongly posi-
tively correlated for female targets, and less strongly but neg-
atively correlated for male targets (Perrett et al. 1998).
The methodology and statistical approach of the present
research has several important advantages. In Study 1, every
participant evaluated all 80 faces, allowing an examination of
three different sources of variation in femininity ratings—rat-
ings due to targets, perceivers, and the relationship between
the two. A good deal of research neglects to examine diver-
gent sources of variance, which can mask important relation-
ships in the data (Bell and Jones 2015;Kievitetal.2013).
Finally, the use of mixed models with crossed random effects
ensures that the results are not simply an artifact of the specific
stimuli selected for our study. Rather, the stimuli were treated
as a random factor—that is, as just one possible sample of
stimuli drawn with error from the population of interest.
Theoretically, if we were to conduct the study again with a
different stimulus set of top scientists, our results suggest that
we should expect similar estimates (Judd et al. 2012).
Although past research has suggested that femininity and
attractiveness are generally viewed as incompatible with sci-
ence (Hartman and Hartman 2008;Proninetal.2004), this is
the first research we know of to use a naturalistic, robust
stimulus set and demonstrate that subtle variations in gendered
appearance alter perceptions that a given woman is a scientist.
This research employed a more robust and sophisticated stim-
ulus set and analytic approach relative to previous research
regarding the negative implications of femininity for women
in the workplace (Sczesny and Kühnen 2004). Specifically,
past research has relied on non-naturalistic (e.g., computer-
generated, hand-drawn) stimuli, and/or a very minimal num-
ber of stimuli (e.g., using the same female Bjob applicant^
dressed in a sexy vs. non-sexy manner; see Glick et al.
2005; using the same female role model and having her dress
in feminine or non-feminine clothing; Betz and Sekaquaptewa
2012). In these studies, the generalizability of the findings is
limited because the results could simply be due to the specific
(and potentially extreme) stimuli selected (Wells and
Windschitl 1999).
By using a large number of naturalistic photographs (of
scientists) and appropriately treating stimulus as a random
factor in the analysis (i.e., as just one selection of stimuli from
the theoretical population of interest), the present studies have
a number of strengths. First, given that all targets were indeed
scientists, we can rule out that participants were responding to
a real relationship between appearance and the likelihood of
being a scientist. Second, gendered appearance varied along a
continuum—as it does in real life—rather than only
representing extremes. Third, our statistical analysis supports
the idea that idiosyncratic differences between the individual
photographs did not give rise to the results. Finally, the statis-
tical analysis supports the idea that a different selection of
photographs should theoretically obtain the same results.
Practice Implications
Our work has a number of implications. First, we would rec-
ommend that scientists who are already established within
STEM fields strive to celebrate and highlight existing diver-
sity within STEM—both between social categories (e.g., dif-
ferent genders or racial groups) but also within social catego-
ries (see Galinsky et al. 2015). People are drawn to fields
where they feel they would belong and be similar to others
(Hannover and Kessels 2004). In addition to being discour-
aged by male-dominated STEM environments (Murphy et al.
2007) or those populated by male stereotypic objects
(Cheryan et al. 2009), women’s interest in STEM may also
be thwarted by the undue perception that women scientists
cannot express femininity. The #iLookLikeAnEngineer cam-
paign exemplifies a marketing strategy that challenges stereo-
types about what engineers look like. Already, this hashtag is
being touted by several companies and universities to display
the variety of individuals within engineering. Given that ex-
posure to counter-stereotypic STEM role models has been
shown to increase men and women’s interest in STEM, such
a strategy should benefit men and women and boys and girls
alike (Cheryan et al. 2012).
Such campaigns may also alleviate pressure on women in
STEM to suppress their femininity. Indeed, research has found
that some women in STEM not only minimize feminine ap-
pearance (e.g., avoid wearing make-up) but also eschew femi-
nine traits, behaviors, and goals (e.g., being emotional, leaving
work to raise children; Pronin et al. 2004). Problematically,
cultures that devalue femininity can also lead women to dis-
tance themselves from and criticize other women (Ellemers
et al. 2004), especially feminine women (Rhoton 2011). Such
practices reinforce the perceived incompatibility between fem-
ininity and STEM, bolster the status quo, minimize the diver-
sity that women have to offer to STEM fields, and are harmful
to the women involved—leading to isolation, dissatisfaction
and potential abandonment with their field (Hartman and
Hartman 2008; Hewlett et al. 2008).
To counteract pressure to assimilate and become Bone of
the guys,^people in male-dominated or male stereotypic
fields should strive to cultivate an environment that celebrates
diversity and where individuals feel as though they can pres-
ent themselves in whatever way they choose (Galinsky et al.
2015). Indeed, research shows that racial minorities in
Sex Roles
companies felt more engaged in their work (BI am proud to tell
others I work [for this organization]^) to the extent that their
White colleagues endorsed a multicultural perspective that
celebrated and recognized diversity (e.g., BEmployees should
recognize and celebrate racial and ethnic differences^)rather
than an assimilationist perspective that maintained minorities
should strive to be more like Whites (e.g., BEmployees should
downplay their racial and ethnic differences^; Plaut et al.
2009, pp. 444). Ideologies about how best to approach group
differences also exist concerning gender (Hahn et al. 2015).
Given that such ideological perspectives are malleable
(Wolsko et al. 2000) and shift between various workplace
environments (Plaut et al. 2009), STEM environments might
aim to embrace an ideological approach that is more welcom-
ing to women and people who generally do not fit the stereo-
typical STEM mold. Indeed, when racial minorities rated their
interest in a company where they would be numerically un-
derrepresented (compared to one where they would be more
equally represented), they preferred a company with a multi-
cultural perspective rather than a colorblind perspective. This
is likely because a colorblind company, coupled with numeric
underrepresentation of one’s social group, implied that their
racial identity was not valued or welcomed and that they
would be expected to minimize their social identity and as-
similate to the predominant group (Purdie-Vaughns et al.
2008).
At the very least, we should not conclude that feminine or
Bgirly images^of women in STEM are uniformly harmful to
fostering women’s interest in STEM (Betz 2012; Betz and
Sekaquaptewa 2012). In our opinion, there is an important
distinction between portraying naturalistic variation in
women’s gendered appearance in STEM versus extreme, ob-
jectified or sexualized portrayals of feminine women scien-
tists. The latter approach to fostering women’s interest in
STEM has proven to be ineffective. For example, the
European Commission launched a campaign entitled
BScience, it’s a girl thing^in an effort to convey that science
can be feminine. However, their promotional video, which
was ultimately withdrawn due to criticism, featured young
women strutting around the lab in high-heels and mini-skirts,
playing with make-up and blowing kisses at test-tubes as a
male scientist observed them (Khazan 2012). Although such
depictions of feminine women in science are clearly problem-
atic, it remains unclear how exposure to naturalistic variation
in gendered appearance in STEM might cultivate greater in-
terest in STEM fields. It is our hope that future research will
further explore the conditions under which feminine
appearing women in STEM inspire and motivate others.
Limitations and Future Directions
Despite the strengths of our research, several questions remain
to be addressed in the future. One clear limitation of the
present research is that we intentionally used only White
men and women scientists to avoid the possibility of arousing
intersecting race or ethnic biases. For example, common race-
based stereotypes maintain that Asians are better at math than
Whites (Aronson et al. 1999) and that Blacks are less academ-
ically capable than Whites (Steele and Aronson 1995). Such
stereotypes may have affected the perceived likelihood of be-
ing a scientist for both male and female targets, and in poten-
tially different ways (e.g., for Black men vs. women; Shields
2008). Because our primary question concerned differential
use of gendered appearance of femininity as a career cue for
male vs. female targets, we presented individuals who all ap-
peared to be White. That said, future research should examine
whether our findings extend to people of other apparent races
and ethnicities.
Another interesting and important issue is precisely which
aspects of appearance participants were using to make judg-
ments about femininity (e.g., inherent facial structure or facial
features vs. performed femininity such as hairstyle and make-
up), and how each of these might differentially contribute to
inferences, attributions, and career judgments about a target.
For example, performed femininity, such as wearing make-up,
in contrast to femininity in facial structure, may be viewed as
particularly incompatible with STEM careers because it sug-
gests that a woman puts too much effort or time into her
appearance.
Some readers may wonder whether participants were pick-
ing up on a real phenomenon whereby women in STEM are
really less feminine than other women. This asks an empirical
question that remains to be addressed—might women in
STEM actually be objectively less feminine in appearance
on average (see Carpinella and Johnson 2013)? Although fu-
ture research is needed to examine this issue (and to under-
stand why this may or may not matter), the present research
demonstrates that regardless of whether this is the case, there
is certainly variation in gendered appearance among scientists,
and this variation is used as a cue of a woman’s—but not a
man’s—likelihood of being a scientist. Whether or not women
in STEM are less feminine-appearing on average than women
outside STEM, there clearly is variation in appearance within
STEM fields in terms of gendered appearance and attractive-
ness, and this variation is sufficient to elicit bias (see Hewlett
et al. 2008; Seymour and Hewitt 1997).
Moreover, although career likelihood may seem like a
somewhat innocuous outcome compared to, for example, de-
cisions to hire an individual, it a) represents a validated phe-
nomenon wherein feminine women in STEM fields report
encountering doubt about their likelihood of being a member
of a STEM field (e.g., “But you don’t look like a program-
mer!”; Hewlett et al. 2008; Zamon 2015) and b) is likely
related to potentially more blatant forms of bias. For example,
role incongruity theory maintains that perceiving that a person
does not Bfit^within a career (i.e., that they are unlikely to be
Sex Roles
in that career) due to a mismatch between that person’sgender
role and the career role elicits prejudice towards the individual
(Eagly and Karau 2002;Heilman2012). Thus, a woman who
is more feminine in appearance than other women will elicit
stronger perceived role incongruity and will therefore experi-
ence more prejudice than their less feminine counterparts
(Eagly and Karau 2002;Heilman2012). Future research is
required to validate other forms of bias that feminine women
in STEM might encounter.
Another lingering question is the extent to which using
femininity in social judgments occurs automatically and, re-
latedly, whether it could be controlled. The robust correlation
found between feminine appearance and career likelihood is
particularly disconcerting because unlike sensitivity to poten-
tial categorical biases (e.g., being sexist or racist), people tend
to be less aware and capable of controlling biases based on
within-category variations, despite being clearly informed
about how such biases operate and asked not to use them
(Blair et al. 2004b; Sczesny and Kühnen 2004). This suggests
that even when evaluating only women for a position or con-
scientiously combating gender bias, feminine women may
nevertheless evoke more negative judgments. Indeed, re-
search regarding Afrocentricity suggests that providing clear
diagnostic information about an individual still did not over-
ride the influence of within-category appearance cues on judg-
ments about that individual (Blair et al. 2005). Future research
is warranted to examine whether people are aware that they
are using feminine appearance in making judgments and
whether they can overcome such responses.
Finally, a further exploration of how these processes affect
women is warranted. What happens to a woman when she is
explicitly told, or signaled in some way, that it does not look as
though she belongs in a given field? How do such interactions
affect women across their lifetime? For example, before
choosing science, are feminine girls and women—because
they don’tBlook^like scientists—treated differently by par-
ents, teachers, and others (Tenenbaum and Leaper 2003)?
What about relatively masculine girls and women who pursue
stereotypically feminine careers—do they encounter addition-
al hurdles simply based on their appearance? Such interactions
may elicit a cascade of inferences that not only guide the
perceiver’s behavior, but in turn affect the self-perceptions
and behavior of the girls and women themselves (Snyder
et al. 1977).
Conclusions
The present paper opened with a story about Isis Wenger, a
woman whose legitimacy as a computer engineer was
contested when her photograph was featured in a recruiting
advertisement for her company. Our results suggest that her
story is not an isolated event—in our studies, men and women
alike used women’s gendered appearance, but not men’s, as an
indication that they were less likely to be scientists (and more
likely to be teachers and journalists). This work empirically
validates claims made by some women in STEM that their
belonging or aptitude in their career has been doubted simply
due to their feminine appearance, and itcontributes to research
suggesting that appearance is more valued, scrutinized, and
consequential for women than men (Bar-Tal and Saxe 1976;
Feingold 1990;Hehmanetal.2014). Documenting what may
be a novel type of gender bias, the present work indicated that
gendered appearance was uniquely used as a cue to a women’s
career but not a man’s career. Overall, our findings suggest
that for women, within-category variation in feminine appear-
ance has the potential to negatively impact the current national
strategic goal of creating a diverse, welcoming, and egalitarian
STEM workforce.
Compliance with Ethical Standards The University of Colorado
Boulder’s Institutional Review Board (IRB) approved the research pre-
sented in the manuscript. This research did not receive external funding.
Conflict of Interest There were no conflicts of interest in conducting
this research.
References
Aronson, J., Lustina, M. J., Good, C., Keough, K., Steele, C. M., &
Brown, J. (1999). When white men can’t do math: Necessary and
sufficient factors in stereotype threat. Journal of Experimental
Social Psychology, 35(1), 29–46.
Baayen, R. H., Davidson, D. J., & Bates, D. M. (2008). Mixed-effects
modeling with crossed random effects for subjects and items.
Journal of Memory and Language, 59(4), 390–412.
Bafumi, J., & Gelman, A. E. (2006). Fitting multilevel models when
predictors and group effects correlate. Retrieved from http://
academiccommons.columbia.edu/item/ac:125243
Barr, D. J., Levy, R., Scheepers, C., & Tily, H. J. (2013). Random effects
structure for confirmatory hypothesis testing: Keep it maximal.
Journal of Memory and Language, 68(3), 255–278.
Bar-Tal, D., & Saxe, L. (1976). Physical attractiveness and its relationship
to sex-role stereotyping. Sex Roles, 2(2), 123–133.
Bell, A., & Jones, K. (2015). Explaining fixed effects: Random effects
modeling of time-series cross-sectional and panel data. Political
Science Research and Methods, 3(1), 133–153.
Betz, D. (2012, September 11). The trouble with Barbie science. Scientific
American. Retrieved from scientificamerican.com.
Betz, D. E., & Sekaquaptewa, D. (2012). My fair physicist? Feminine
math and science role models demotivate young girls. Social
Psychological and Personality Science, 3(6), 738–746.
Blair, I. V., Judd, C. M., Sadler, M. S., & Jenkins, C. (2002). The role of
Afrocentric features in person perception: Judging by features and
categories. Journal of Personality and Social Psychology, 83(1), 5–
25.
Blair, I. V., Judd, C. M., & Chapleau, K. M. (2004a). The influence of
Afrocentric facial features in criminal sentence. Psychological
Science, 15(10), 674–679.
Sex Roles
Blair, I. V., Judd, C. M., & Fallman, J. L. (2004b). The automaticity of
race and Afrocentric facial features in social judgments. Journal of
Personality and Social Psychology, 87(6), 763–778.
Blair, I. V., Chapleau, K. M., & Judd, C. M. (2005). The use of
Afrocentric features as cues for judgment in the presence of diag-
nostic information. European Journal of Social Psychology, 35(1),
59–68.
Carnevale, A. P., Strohl, J., & Melton, M. (2013). What’s it worth?: The
economic value of college majors. Georgetown University
Institutional Report. Retrieved from https://cew.georgetown.edu/
wp-content/uploads/2014/11/whatsitworth-complete.pdf
Carpinella, C. M., & Johnson, K. L. (2013). Politics of the face: The role
of sex-typicality in trait assessments of politicians. Social Cognition,
31(6), 770–779.
Ceci, S. J., & Williams, W. M. (2011). Understanding current causes of
women’s underrepresentation in science. Proceedings of the
National Academy of Sciences, 108(8), 3157–3162.
Cheng, C. (2008). Marginalized masculinities and hegemonic masculin-
ity: An introduction. The Journal of Men’sStudies,7(3), 295–315.
Cheryan, S., Plaut, V. C., Davies, P. G., & Steele, C. M. (2009). Ambient
belonging: How stereotypical cues impact gender participation in
computer science. Journal of Personality and Social Psychology,
97(6), 1045–1060.
Cheryan, S., Siy, J. O., Vichayapai, M., Drury, B. J., & Kim, S. (2011). Do
female and male role models who embody STEMstereotypes hinder
women’s anticipated success in STEM? Social Psychological and
Personality Science, 2(6), 656–664.
Cheryan, S., Drury, B. J., & Vichayapai, M. (2012). Enduring influence of
stereotypical computer science role models on women’sacademic
aspirations. Psychology of Women Quarterly, 37(1), 72–79.
Clark, H. H. (1973). The language-as-fixed-effect fallacy: A critique of
language statistics in psychological research. Journal of Verbal
Learning and Verbal Behavior, 12(4), 335–359.
Diekman, A. B., Brown, E. R., Johnston, A. M., & Clark, E. K. (2010).
Seeking congruity between goals and roles: A new look at why
women opt out of science, technology, engineering, and mathemat-
ics careers. Psychological Science, 21(8), 1051–1057.
Eagly, A. H., & Karau, S. J. (2002). Role congruity theory of prejudice
toward female leaders. Psychological Review, 109(3), 573–598.
Eagly, A. H., & Wood, W. (2012). Social role theory. In P. van Lange, A.
Kruglanski, & E. T. Higgins (Eds.), Handbook of theories in social
psychology (pp. 458–476). Thousand Oaks: Sage Publications.
Eccles, J. S. (1987). Gender roles and women’s achievement-related de-
cisions. Psychology of Women Quarterly, 11(2), 135–172.
Ellemers, N., Van den Heuvel, H., Gilder, D., Maass, A., & Bonvini, A.
(2004). The underrepresentation of women in science: Differential
commitment or the queen bee syndrome? British Journal of Social
Psychology, 43(3), 315–338.
Feingold, A. (1990). Gender differences in effects of physical attractive-
ness on romantic attraction: A comparison across five research par-
adigms. Journal of Personality and Social Psychology, 59(5), 981–
993.
Friedman, H., & Zebrowitz, L. A. (1992). The contribution of typical sex
differences in facial maturity to sex role stereotypes. Personality and
Social Psychology Bulletin, 18(4), 430–438.
Galinsky, A. D., Todd, A. R., Homan, A. C., Phillips, K. W., Apfelbaum,
E. P., Sasaki, S. J., …Maddux, W. W. (2015). Maximizing the gains
and minimizing the pains of diversity: A policy perspective.
Perspectives on Psychological Science, 10(6), 742–748.
Glick, P., Larsen, S., Johnson, C., & Branstiter, H. (2005). Evaluations of
sexy women in low and high status jobs. Psychology of Women
Quarterly, 29(4), 389–395.
Hahn, A., Banchefsky, S., Park, B., & Judd, C. M. (2015). Measuring
intergroup ideologies positive and negative aspects of emphasizing
versus looking beyond group differences. Personality and Social
Psychology Bulletin, 41(12), 1646–1664.
Hannover, B., & Kessels, U. (2004). Self-to-prototype matching as a
strategy for making academic choices: Why high school students
do not like math and science. Learning and Instruction, 14,51–67.
Hartman, H., & Hartman, M. (2008). How undergraduate engineering
students perceive women’s(andmen’s) problems in science, math
and engineering. Sex Roles, 58,251–265.
Hehman, E., Carpinella, C. M., Johnson, K. L., Leitner, J. B., & Freeman,
J. B. (2014). Early processing of gendered facial cues predicts the
electoral success of female politicians. Social Psychological and
Personality Science, 5(7), 815–824.
Heilman, M. E. (2012). Gender stereotypes and workplace bias. Research
in Organizational Behavior, 32,113–135.
Hewlett, S., Luce, C., Servon, L., Sherbin, L., Shiller, P., Sosnovich, E.,
…Sumberg, K (2008). The Athena factor: Reversing the brain
drain in science, engineering, and technology. Wate r t own :
Harvard Business School.
Irwin, J. R., & McClelland, G. H. (2003). Negative consequences of
dichotomizing continuous predictor variables. Journal of
Marketing Research, 40(3), 366–371.
Ito, T. A., & Urland, G. R. (2003). Race and gender on the brain:
Electrocortical measures of attention to the race and gender of mul-
tiply categorizable individuals. Journal of Personality and Social
Psychology, 85(4), 616–626.
Judd, C. M., Westfall, J., & Kenny, D. A. (2012). Treating stimuli as a
random factor in social psychology: A new and comprehensive so-
lution to a pervasive but largely ignored problem. Journal of
Personality and Social Psychology, 103(1), 54–69.
Kenny, D. A. (1994). Interpersonal perception: A social relations
analysis. New York: Guilford Press.
Khazan, O. (2012, June 22). E.U.’s‘Science, it’s a girl thing’campaign
sparks a backlash. The Washington Post. Retrieved from www.
washingtonpost.com
Kievit, R., Frankenhuis, W. E., Waldorp, L., & Borsboom, D. (2013).
Simpson’s paradox in psychological science: A practical guide.
Quantitative Psychology and Measurement, 4,1–14.
McClelland, G. H., & Judd, C. M. (1993). Statistical difficulties of de-
tecting interactions and moderator effects. Psychological Bulletin,
114 (2), 376–390.
Moss-Racusin, C. A., Dovidio, J. F., Brescoll, V. L., Graham, M. J., &
Handelsman, J. (2012). Science faculty’s subtle gender biases favor
male students. Proceedings of the National Academy of Sciences,
109(41), 16474–16479.
Murphy, M. C., Steele, C. M., & Gross, J. J. (2007). Signaling threat:
How situational cues affect women in math, science, and engineer-
ing settings. Psychological Science, 18(10), 879–885.
National Science Foundation, National Center for Science and
Engineering Statistics. (2015). Women, minorities, and persons with
disabilities in science and engineering: 2015. Special Report NSF
15–311. Arlington, VA. Retrieved from http://www.nsf.gov/
statistics/wmpd/
Perrett, D. I., Lee, K. J., Penton-Voak, I., Rowland, D., Yoshikawa, S.,
Burt, D. M., …Akamatsu, S. (1998). Effects of sexual dimorphism
on facial attractiveness. Nature, 394(6696), 884–887.
Plaut, V. C., Thomas, K. M., & Goren, M. J. (2009). Is multiculturalism or
color blindness better for minorities? Psychological Science, 20(4),
444–446.
Pronin, E., Steele, C. M., & Ross, L. (2004). Identity bifurcation in re-
sponse to stereotype threat: Women and mathematics. Journal of
Experimental Social Psychology, 40(2), 152–168.
Purdie-Vaughns, V., Steele, C. M.,Davies, P. G., Ditlmann, R., & Crosby,
J. R. (2008). Social identity contingencies: How diversity cues sig-
nal threat or safety for African Americans in mainstream institutions.
Journal of Personality and Social Psychology, 94(4), 615–630.
Raudenbush, S. W. (2009). Adaptive centering with random effects: An
alternative to the fixed effects model for studying time-varying
Sex Roles
treatments in school settings. Education Finance and Policy, 4(4),
468–491.
Rhoton, L. A. (2011). Distancing as a gendered barrier understanding
women scientists’gender practices. Gender and Society, 25(6),
696–716.
Rosnow, R. L., & Rosenthal, R. (1991). If you’re looking at the cell
means, you’re not looking at only the interaction (unless all main
effects are zero). Psychological Bulletin, 110(3), 574–576.
Sczesny, S., & Kühnen, U. (2004). Meta-cognition about biological sex
and gender-stereotypic physical appearance: Consequences for the
assessment of leadership competence. Personality and Social
Psychology Bulletin, 30(1), 13–21.
Seymour, E., & Hewitt, N. M. (1997). Talking about leaving: Why un-
dergraduates leave the sciences. Boulder: Westview.
Shields, S.A. (2008). Gender: An intersectionality perspective. Sex Roles,
59(5), 301–311.
Snijders, T. A., & Bosker, R. J. (1994). Modeled variance in two-level
models. Sociological Methods & Research, 22(3), 342–363.
Snyder, M., Tanke, E. D., & Berscheid, E. (1977). Social perception and
interpersonal behavior: On the self-fulfilling nature of social stereo-
types. Journal of Personality and Social Psychology, 35(9), 656–
666.
Steele, C. M., & Aronson, J. (1995). Stereotype threat and the intellectual
test performance of African Americans. Journal of Personality and
Social Psychology, 69(5), 797–811.
Tenenbaum, R., & Leaper, C. (2003). Parent–child conversations about
science: The socialization of gender inequities? Developmental
Psychology, 39(1), 34–47.
Wells, G., & Windschitl, P. (1999). Stimulus sampling and social psycho-
logical experimentation. Personality and Social Psychology
Bulletin, 25(9), 1115–1125.
Westfall, J., Kenny, D. A., & Judd, C. M. (2014). Statistical power and
optimal design in experiments in which samples of participants re-
spond to samples of stimuli. Journal of Experimental Psychology:
General, 143(5), 2020–2045.
Willis, J., & Todorov, A. (2006). First impressions: Making up your mind
after a 100-ms exposure to a face. Psychological Science, 17(7),
592–598.
Wolsko, C., Park, B., Judd, C. M., & Wittenbrink, B. (2000). Framing
interethnic ideology: Effects of multicultural and color-blind per-
spectives on judgments of groups and individuals. Journal of
Personality and Social Psychology, 78(4), 635–654.
Zamon, R. (2015, August 4). #ILookLikeAnEngineer reminds us that
anyone can (and should) be an engineer. Retrieved from http://
www.huffingtonpost.ca/2015/08/04/ilooklikeanengineer_n_
7934098.html
Zebrowitz, L. A., Tenenbaum, D. R., & Goldstein, L. H. (1991). The
impact of job applicants’facial maturity, gender, and academic
achievement on hiring recommendations. Journal of Applied
Social Psychology, 21(7), 525–548.
Sex Roles
A preview of this full-text is provided by Springer Nature.
Content available from Sex Roles
This content is subject to copyright. Terms and conditions apply.
