Dissecting “Gaydar”: Accuracy and the Role of Masculinity–Femininity
-
Upload
independent -
Category
Documents
-
view
4 -
download
0
Transcript of Dissecting “Gaydar”: Accuracy and the Role of Masculinity–Femininity
ORIGINAL PAPER
Dissecting ‘‘Gaydar’’: Accuracy and the Roleof Masculinity–Femininity
Gerulf Rieger Æ Joan A. W. Linsenmeier ÆLorenz Gygax Æ Steven Garcia Æ J. Michael Bailey
Received: 5 November 2007 / Revised: 5 May 2008 / Accepted: 5 May 2008 / Published online: 23 September 2008
� Springer Science+Business Media, LLC 2008
Abstract ‘‘Gaydar’’ is the ability to distinguish homosex-
ual and heterosexual people using indirect cues. We inves-
tigated the accuracy of gaydar and the nature of ‘‘gaydar
signals’’ conveying information about sexual orientation.
Homosexual people tend to be more sex atypical than hetero-
sexual people in some behaviors, feelings, and interests. We
hypothesized that indicators of sex atypicality might function
as gaydar signals. In Study 1, raters judged targets’ sexual
orientation from pictures, brief videos, and sound recordings.
Sexual orientation was assessed with high, though imperfect,
accuracy. In Study 2, different raters judged targets’ sex
atypicality from the same stimuli. Ratings of sexual orien-
tation from Study 1 corresponded highly with targets’ self-
reports of sex atypicality and with observer ratings of sex
atypicality from Study 2. Thus, brief samples of sex-atypical
behavior may function as effective gaydar signals.
Keywords Sexual orientation � Person perception �Sex-typed behavior
Introduction
‘‘Gaydar’’ refers to the ability to distinguish homosexual and
heterosexual people using indirect cues rather than explicit
information about sexual orientation. Gaydar encompasses at
least two distinct phenomena. First, gaydar may reflect the
detection of intentional interpersonal signals. For example,
flirtation may be associated with unusually long eye gazes
and other nonverbal behaviors (e.g., Moore, 1985, 2002), and
the sex of the targets of such signaling may provide infor-
mation relevant to sexual orientation. Second, gaydar may
reflect the detection of stable behavioral or psychological
differences between homosexual and heterosexual people.
For example, there are stereotypes about gay men’s (and
hence heterosexual men’s) patterns of interests, movement,
and speech. To the extent that such stereotypes correspond to
actual differences between homosexual and heterosexual
people, information about a person’s behavioral patterns and
psychological traits may also provide information regarding
sexual orientation.
Little research has been done to date regarding gaydar via
intentional interpersonal signaling. Nicholas (2004) con-
ducted an ethnographic study of gaydar involving participant
observation and interviews with gay men and lesbians. Based
on this research, Nichols described the use of the ‘‘gaydar
gaze’’ to signal a homosexual identity to other homosexual
individuals. The idea that individuals from marginalized and
partly hidden minorities may attempt to find and communi-
cate to each other in this manner seems eminently plausible.
We note, however, that there is likely nothing specific about
the association of lengthy eye gazes and homosexuality. For
example, heterosexual people with a romantic or sexual
interest in a person of the other sex may also signal their
interest in this way.
The other subtype of gaydar, as noted above, may depend
in part on the validity of stereotypes regarding behavioral and
psychological differences between homosexual and hetero-
sexual people. These stereotypes include a variety of phe-
nomena that have been explored in two, almost completely
separate, research programs. They are, respectively, research
G. Rieger (&) � J. A. W. Linsenmeier � S. Garcia � J. M. Bailey
Department of Psychology, Northwestern University, 2029
Sheridan Rd., Swift Hall #102, Evanston, IL 60208, USA
e-mail: [email protected]
L. Gygax
Swiss Federal Veterinary Office, Agroscope Tanikon,
Evanston, Switzerland
123
Arch Sex Behav (2010) 39:124–140
DOI 10.1007/s10508-008-9405-2
concerning assessment of sexual orientation through quick,
initial judgments based on ‘‘thin behavioral slices’’ (e.g.,
Ambady, Hallahan, & Conner, 1999) and that concerning the
relationship of sexual orientation to masculinity and femi-
ninity (e.g., Lippa, 2005b).
Exploring the nature of gaydar has both social implica-
tions and the potential to illuminate sexual orientation and
masculinity–femininity. As Ambady, Bernieri, and Richeson
(2000) pointed out in their review on evaluations of thin slices
of behavior, initial impressions can influence subsequent
behavior towards targets. Deciding, based on gaydar signals,
that a person is probably homosexual, or probably hetero-
sexual, may have important social consequences for that
person. Understanding how observers make such judgments
may help to explain what motivates their subsequent inter-
personal behavior. Furthermore, it is not obvious that, or why,
sexual orientation should correlate with superficial aspects of
overt behavior. Exploring such correlations may enhance our
understanding of the nature and development of sexual ori-
entation, as well as other sex-typed behaviors.
Thin Slices of Sexual Orientation
Many psychological and relational characteristics can be
inferred with above-chance accuracy using information
gleaned from brief samples of expressive behavior (‘‘thin
slices’’) typically ranging from a few seconds to a few min-
utes (for reviews, see Ambady et al., 2000; Ambady &
Rosenthal, 1992). Examples include personality traits, such
as extroversion and sociability, temporary emotions, and
long-lasting emotional states, such as anxiety and depression.
Observers can also detect interaction motives, such as
deception by unknown people, at better than chance rates
from thin slices of behavior. Interpersonal relationships can
also be reliably assessed from thin slices, including romantic
involvement, dominance, friendship, and kinship.
Research has also addressed whether sexual orientation
can be assessed based on thin slices of visual information, that
is, information regarding the body and bodily movement
(Ambady et al., 1999; Johnson, Gill, Reichman, & Tassinary,
2007) and auditory information, that is, information regard-
ing speech (e.g., Bailey, 2003; Gaudio, 1994; Linville, 1998).
In other words, researchers have investigated whether as-
pects of people’s appearance, movement, and speech can
function as ‘‘gaydar signals’’ that enable others to detect their
sexual orientation.
Visual Thin Slices of Sexual Orientation
Research has generally shown that homosexual and hetero-
sexual individuals can be distinguished based on thin slices of
motor behavior. In one study, observers were able to assess a
person’s sexual orientation at better than chance rates based
on 10-s video clips, using only outlines of the targets’ body
movement (Ambady et al., 1999). Even 1-s video clips and
still photographs increased accuracy of sexual orientation
judgments compared with judgments expected by chance,
although neither conferred as much information about sexual
orientation as 10-s clips did. A subsequent study confirmed
that thin slices of motor behavior can provide useful infor-
mation for identifying sexual orientation (Johnson et al.,
2007). The researchers used a sophisticated computerized
data reduction technique to eliminate all information except
that regarding body shape and movement near body joints.
Raters viewing the resulting video files achieved greater than
chance accuracy in assessing the targets’ sexual orientation.
Auditory Thin Slices of Sexual Orientation
Gaudio (1994) examined whether listeners could differenti-
ate between speech of four heterosexual and four homosexual
men. Listeners rated voice recordings of approximately 15 s,
in which each target participant recited exactly the same
passages. Despite the small sample of targets, average ratings
were significantly related to targets’ sexual orientations.
Because the majority of targets were recruited among the
researcher’s acquaintances, it is unclear how generalizable
the findings are. However, other studies have also shown that
listeners can judge the sexual orientation of speakers better
than expected by chance (Bailey, 2003; Linville, 1998;
Smyth, Jacobs, & Rogers, 2003). Only one study included
female target participants (Sylva, Sell, & Bailey, 2007). For
male targets, results indicated that judgments of sexual ori-
entation based on speech samples were significantly related
to self-identified sexual orientation.
Individual Differences in Rating Accuracy
A common belief is that homosexual people have better
gaydar than heterosexual people. The assumption seems to be
that homosexual people have more experience interacting
with other homosexual people and are thus more familiar
with the specific behaviors and lifestyles that tend to
accompany homosexuality. In addition, they may be more
motivated to correctly identify the sexual orientation of
others. Some studies of thin slices of behavior have included
relevant data for evaluating this belief. One study found that
homosexual people were better than heterosexuals at accu-
rately judging sexual orientation from brief videotaped
interviews (Berger, Hank, Rauzi, & Simkins, 1987; results
reported in Ambady et al., 1999). Ambady et al. (1999)
found inconsistent results across two studies regarding the
relative accuracy of heterosexual and homosexual raters; in
general, results did not reveal that either was superior. Sylva
Arch Sex Behav (2010) 39:124–140 125
123
et al. (2007) found a modest significant advantage for
homosexual people in judging male (but not female) sexual
orientation based on speech samples. In general, it appears
that homosexual people may have some advantage in judging
sexual orientation using thin slices of behavior, but that the
advantage is modest and somewhat inconsistent across
studies.
Masculinity, Femininity, and Sexual Orientation
The research reviewed so far suggests that observable
behaviors can serve as gaydar signals. That is, at least some
people behave in ways that provide others with useful
information regarding their sexual orientation. However,
research presented thus far does not indicate what aspects of
these behaviors might lead observers to categorize the actors
as likely to be homosexual or heterosexual. Research
addressing the masculinity and femininity of homosexual
and heterosexual people may provide some answers.
Stereotypes about homosexual people tend to reflect the
idea that homosexual people are somewhat like heterosexu-
als of the other sex; that is, compared with other members of
their own sex, homosexual men are relatively feminine and
homosexual women relatively masculine (Kite & Deaux,
1987). Although the concepts of masculinity and femininity
have had a controversial history in scientific psychology
(Constantinople, 1973; Lippa, 2005a, Chap. 2; Spence &
Buckner, 1995), there has been recent empirical and theo-
retical progress. Masculinity and femininity can reasonably
be seen as opposite poles of a broad psychological and
behavioral dimension (e.g., Lippa, 1995a, 1995b, 2005a).
Specifically, unidimensional self-ratings of masculinity–
femininity have several correlates supporting that dimen-
sion’s construct validity. These include sex-typed occupa-
tional and recreational interests (Lippa, 1991, 1995a, 1995b),
recalled childhood sex atypicality (Lippa, 2006), and sexual
orientation (Lippa, 2005b). Moreover, these variables tend to
be moderately to strongly correlated within each sex; that is,
not only with self-ratings of masculinity–femininity, but also
with each other (Lippa, 2006). Three of these variables have
been studied extensively with respect to sexual orientation:
masculinity–femininity, sex-typed occupational and recrea-
tional interests, and childhood sex atypicality.
We note that in this article, rather than using the words
masculine and feminine, we often use the term sex atypical.
This term refers to males who are relatively feminine or
relatively unmasculine compared with other males, as well as
females who are relatively masculine or relatively unfemi-
nine compared with other females. Using a single term to
refer to atypical behavior by members of both sexes simpli-
fies our presentation and also emphasizes parallels between
the two sexes.
Self-reported Masculinity–Femininity, Sex-typed Interests,
and Sexual Orientation
Lippa (2005b) meta-analyzed results from several of his
studies in which he assessed participants’ self-rated mascu-
linity–femininity and sex-typed interests. Measures were
assessed in a manner that allowed direct comparison of het-
erosexual men and heterosexual women, and of heterosexual
and homosexual individuals within each sex. The difference
between heterosexual men and women was large, especially
for sex-typed interests (d = 2.7). Furthermore, findings
suggested that sexual orientation differences within each sex
paralleled those between the sexes. That is, homosexual men
(N = 799) rated themselves as more feminine compared
with heterosexual men (N = 2,724, d = 0.6). Similarly,
homosexual women (N = 697) rated themselves as more
masculine compared with heterosexual women (N = 5,053,
d = 1.3). Effect sizes were even larger for sex-typed inter-
ests, with homosexual people indicating more sex-atypical
interests than heterosexual people; for men d = 1.3, and for
women d = 1.5. Thus, self-assessed masculinity–femininity
differentiates not only men and women, but also heterosexual
and homosexual people within each sex.
Self-reported Childhood Sex Atypicality and Sexual
Orientation
Measures of childhood sex atypicality are usually retro-
spective and assess both general sex-typed behavior (e.g., ‘‘I
was a feminine boy;’’ ‘‘I was a masculine girl’’), and more
specific aspects of such behavior (e.g., ‘‘I preferred playing
with girls;’’ ‘‘I pretended to be a boy’’). In a meta-analysis of
32 studies (Bailey & Zucker, 1995), homosexual men re-
called much more sex atypicality during their childhoods
compared with heterosexual men (d = 1.3), and in the cor-
responding meta-analysis of 16 studies of women, homo-
sexual women recalled much more childhood sex atypicality
compared with heterosexual women (d = 1.0).
Lippa (2006) has shown that recalled childhood sex
atypicality is strongly correlated with current self-rated
masculinity–femininity and with sex-typed occupational
preferences, in both male and female samples including both
homosexual and heterosexual participants. These correla-
tions are consistent with the idea that an underlying dimen-
sion of masculinity–femininity contributes to variation in all
three traits. They further suggest considerable longitudinal
stability to this dimension.
Self-reported and Observed Sex Atypicality
Only a few studies have investigated the correspondence
between self-reported sex atypicality and ratings of sex
126 Arch Sex Behav (2010) 39:124–140
123
atypicality done by other people (observers). Lippa (1998)
found that self-reported interests in sex-typed activities,
occupations, and hobbies correlated significantly with ob-
server ratings of behavioral masculinity–femininity. In an-
other study, self-reports of sex atypicality correlated with
observer ratings of sex-atypical behavior based on 10- to 30-s
videos of target participants (Rieger, Linsenmeier, Gygax, &
Bailey, 2008). This correlation was similar for heterosexual
and homosexual men and women. Thus, self-ratings of sex
atypicality seem to correspond to ‘‘thin slices’’ of sex atypi-
cality that can be observed by others.
Gaydar, Masculinity, and Femininity
Past research suggests both that sexual orientation can be
judged with some degree of accuracy using thin slices of
expressive behavior and that sexual orientation is strongly
correlated with questionnaire measures related to masculin-
ity–femininity. Furthermore, self-ratings of sex atypicality
correspond to the sex atypicality of observable behavior.
Thus, it is reasonable to raise the hypothesis that when people
judge sexual orientation based on thin slices of observable
behavior, these judgments depend on the sex atypicality of
the behavior observed. That is, expressions of masculinity in
women and of femininity in men may serve as gaydar signals,
enabling observers to make more accurate judgments of
people’s sexual orientation.
The hypothesis that a certain type of sex-atypical behavior
is a useful gaydar signal would be supported most directly if
its frequency in homosexual individuals is biased toward its
frequency in other-sex heterosexual individuals. A few
studies of motor behavior have demonstrated such differ-
ences. In one study (Bailey, 2003), raters used a rating scale
originally designed to discriminate men and women (Barlow
et al., 1979), and the ratings also demonstrated differences
between homosexual and heterosexual individuals, with
homosexual individuals showing more sex-atypical motor
behavior. Similarly, another study suggested that behavioral
displays of masculinity–femininity were related to sexual
orientation (Rieger et al., 2008). Johnson et al. (2007) fo-
cused on joint movements that are especially sexually
dimorphic in humans, namely those of the shoulder and hips.
Assessments of this movement yielded predicted homosex-
ual–heterosexual differences and also correlated with raters’
perceptions of sexual orientation.
Sometimes it is difficult to discern precisely what the
gaydar signal is, even when it is powerful. For example, Sylva
et al. (2007) found large differences in how listeners rated the
speech samples of homosexual and same-sex heterosexual
speakers. Using the same dataset, however, another research
group found relatively small differences in a particular
phonological aspect hypothesized to be important, namely
vowel production (Pierrehumbert, Bent, Munson, Bradlow,
& Bailey, 2004). Perhaps speakers’ cues regarding their
sexual orientations are perceived gestalt-like rather than with
regard to particular details. In such cases, at least two other
kinds of findings might support the notion that the gaydar
signal represents sex atypicality. First, a sex atypicality
interpretation of the gaydar signal would also be supported to
the extent that the signal is correlated with variables known to
be indicators of sex atypicality. These might include, for
example, self-rated masculinity–femininity (in the sex-
atypical direction), sex-atypical occupational interests, or
childhood sex atypicality. Second, if observers perceive the
gaydar signal as sex atypical, this would obviously support
that interpretation. One example of this kind comes from the
study of sexual orientation and speech by Gaudio (1994).
Gaudio found a very high correlation (r = .90) between
average ratings of speech samples for homosexual orienta-
tion and for femininity in a small sample of male speakers.
Overview of Studies
We conducted two studies to examine the accuracy of gaydar,
the roles of different possible gaydar signals, and the rela-
tionships among ratings of sexual orientation and ratings of
sex atypicality. We videotaped homosexual and heterosexual
men and women answering an interview question about their
interests. We then recruited two additional sets of partici-
pants to rate various aspects of brief excerpts from these
interviews. The first raters judged targets’ sexual orientations
from unedited videos and from partial information extracted
from the videos (e.g., video without sound for ratings of
movement or sound without picture for ratings of speech).
The second raters judged targets’ sex atypicality using the
same information.
The first study focused on the accuracy of gaydar, that is,
the accuracy of observer ratings of sexual orientation. We
also investigated whether certain raters (e.g., homosexual
individuals) have better gaydar than others. Ratings were
based on the targets’ overall presentation and on different
components of their behavior (e.g., movement or speech).
We further computed correlations of these different observer
ratings with targets’ self-reported masculinity–femininity to
investigate whether gaydar signals are related to indicators of
sex atypicality. The second study focused on the correspon-
dence between observer ratings of sexual orientation and
observer ratings of sex atypicality, to further investigate
whether indices of sex atypicality can function as gaydar
signals.
Arch Sex Behav (2010) 39:124–140 127
123
Study 1
Method
Participants
Targets We placed advertisements in alternative urban
newspapers and websites (The Chicago Reader,
chicago.craigslist.org), which are read by both heterosexual
and homosexual people, and in homophile publications (e.g.,
Gay Chicago). Advertisements were placed in the classified
sections under ‘‘miscellaneous job opportunities.’’ Four dif-
ferent advertisements were placed: one for ‘‘gay men,’’ one
for ‘‘lesbians,’’ one for ‘‘straight men,’’ and one for ‘‘straight
women.’’ For example, one advertisement read: ‘‘Gay men
wanted for a videotaped interview about their lifestyle and a
study about sexual orientation,’’ and analogous advertise-
ments were placed for the other groups of participants. Par-
ticipants who contacted the lab were told that we were
interested in studying whether their sexual orientation relates
to how they live their lives. They were also told that we would
like to videotape them during a casual interview. Sexual
orientation was assessed by self-labeling as either ‘‘gay’’ or
‘‘straight.’’
Recruited target participants included 25 homosexual
men, 23 homosexual women, 25 heterosexual men, and 22
heterosexual women. Mean ages (with SD) were 25.4 (3.5),
23.7 (3.1), 24.0 (2.8), and 23.6 (2.7) years, respectively. Ages
did not significantly differ between groups, p = .14, b = -
.19. (The effect size b is, throughout this article, the stan-
dardized regression coefficient as derived from multiple
regression analyses. All group differences were estimated by
calculating the main effects of sex, sexual orientation, and
their interaction effect). Forty-nine percent of the targets
were Caucasian. This proportion did not significantly differ
across groups, v2(3) \ 1. Twenty-eight percent were His-
panic and the remainder identified as African-American,
Native-American, Asian, or ‘‘other.’’
Raters Twelve homosexual men and 13 homosexual
women who were recruited via the same media as target
participants, and 13 heterosexual men and 13 heterosexual
women who were undergraduate students in an introductory
psychology class, rated targets on several measures of ob-
server-rated sexual orientation. Mean ages of raters were 26.1
(5.0), 29.5 (8.6), 19.6 (1.5), and 19.5 (1.1) years, respectively.
Homosexual raters were older than the heterosexual raters,
p \ .0001, b = .58. Because judgments of the four groups of
raters correlated with each other and barely differed (see
below), differences in age did not seem to have an important
influence on ratings. Fifty percent of the raters were Cauca-
sian. This proportion did not significantly differ across the
groups, v2(3) = 5.28, p = .15. Thirty-seven percent were
Hispanic, and the remainder identified as Asian or African-
American.
Self-report Measures
Childhood sex atypicality of targets was assessed with the
Childhood Gender Nonconformity Scale (Bailey, Finkel,
Blackwelder, & Bailey, 1995; see also Rieger et al., 2008)
with seven statements for each sex. Sample items were: ‘‘As a
child, I often felt that I had more in common with [the
opposite sex]’’ or ‘‘I preferred playing with girls rather than
boys.’’ One measure of adult sex atypicality was the Con-
tinuous Gender Identity Scale (Bailey et al., 1995; Rieger
et al., 2008) with 10 items for each sex. Statements included:
‘‘I feel as if part of me is male and part of me is female’’ and ‘‘It
would be fun to go to a costume party dressed as [the opposite
sex].’’ Targets rated their agreement with each item using a
7-point scale ranging from 1 = strongly disagree to 7 =
strongly agree. Item-reliability (Cronbach’s a) of the child-
hood scale was .82 and .86 for men and women, respectively.
For the adult scale, the respective numbers were .69 and .76.
Responses to different items were averaged and higher scores
represented greater sex atypicality. Items differed somewhat
between the sexes. Hence, targets’ scores for childhood and
adult sex atypicality were standardized within sex to allow
correlational analyses across the sexes.
Sex-atypical interests in certain hobbies and occupations
in adulthood were assessed with the 134-item version of the
Gender Diagnosticity Measure (e.g., Lippa, 2000). Sample
items ask about interest in being an opera singer or a physi-
cian or in going fishing or clothes shopping. Targets used a
7-point scale, ranging from 1 = strongly dislike to 7 =
strongly like, to indicate their interest for each item. Two
research assistants rated whether each of the hobbies and
occupations was more male-typical, female-typical, or sex-
neutral. With two exceptions, they showed full agreement.
These exceptions were resolved by discussion of the two
assistants. The final numbers of items in each category were
56, 63, and 15, respectively. For both male and female targets,
item-reliability (a) exceeded .88 for the male-typical and
female-typical category.
Men’s ratings of male-typical interests and women’s rat-
ings of female-typical interests were reverse-coded so that
higher numbers now indicated less liking. For each sex, inter-
item reliability (a) across sex-atypical interests and reversed
sex-typical interests exceeded .80. Then, for each male target,
we added to his average rating of female-typical interests the
reverse of his average rating of male-typical interests. For
each female target, we added to her average rating of male-
typical interests the reverse of her average rating of female-
typical interests. Thus, for both sexes, we computed a final
average where higher scores indicated more sex-atypical
128 Arch Sex Behav (2010) 39:124–140
123
interests. For correlational analyses across the sexes, these
final scores were standardized within each sex.
Procedure
Target Interviews and Self-reports We videotaped targets
for 20 min during a casual interview about their lifestyle.
Targets sat in a chair in front of a digital camera, which
captured their entire bodies. For analyses, we used their re-
sponses to a question about their favorite interests, which we
asked towards the end of the interview. At that time, targets
were most likely to be relaxed in front of the camera, and the
majority of them used gesture and movement. We considered
their motor behavior, speech, and stated interests all to be
potential gaydar signals. After the interview, each target
stood upright so that we could take a full body picture, as
appearance was also considered a potential gaydar signal.
Then targets completed the questionnaires.
Preparation of Stimuli We used the first full sentence in
targets’ answers about their interests to prepare several types
of stimuli. One target explicitly revealed her sexual orienta-
tion in this first sentence, so, in her case, the second sentence
was taken. Selected videos of all targets were between 6 and
10 s long. This footage was used for subsequent ratings of the
target’s overall presentation. In addition, this footage was
modified in several ways to create stimuli that focused on
distinct observable components. Using Final Cut, we created
video clips, which were black and white with exaggerated
contrasts and no sound; these allowed raters to focus on the
targets’ movement (e.g., their gestures and posture). We
created sound clips by removing the video from people’s
responses in order to get ratings of the targets’ speech. For
judgments of appearance, we used the still photographs that
were taken of the targets. Finally, for ratings of stated inter-
ests, based on the brief behavioral samples, we presented
raters with written transcripts of targets’ answers.
Rating of Stimuli Equal numbers of raters from each group
(heterosexual and homosexual men and women) evaluated
each of the aforementioned stimuli sets (overall presentation,
movement, speech, appearance, and stated interests). Raters
evaluated stimuli for male targets separately from stimuli for
female targets, and across raters the order of sexes was ran-
dom. One half of raters first evaluated movement and then
speech. The other half of raters evaluated appearance first,
then interests, and then the unmodified videos for ratings of
overall presentation. Within each set, stimuli were presented
in random order.
Raters indicated their impressions of each target stimulus
on a 7-point scale. We asked them to compare each target to
the average person of his or her sex. For example, after each
video showing figural outlines of a male target, raters were
asked to indicate their impression of the target’s movement.
Specifically, they were asked ‘‘whether this man gestures,
moves, or sits in a more straight or gay manner.’’ Ratings
were made using a scale from 1 to 7, where 1 = more gay,
4 = average, and 7 = more straight. Corresponding scales
were used to rate the targets’ overall presentation (from the
unmodified videos), how targets sounded (for ratings of
speech) and looked (each photograph was shown for 3 s for
ratings of appearance), as well as whether they stated more
straight or gay interests (for ratings of the written transcripts).
Raters received no explicit information about the targets’
sexual orientation. Each rater was limited to 45 min of
evaluating stimuli.
Results
Inter-rater Consistency
Across each combination of target sex and observer-rated
measure (i.e., the targets’ overall presentation, movement,
speech, appearance, and stated interests), the ratings of the
four groups (heterosexual and homosexual men and women)
were correlated; the median r was .81, and the range was from
.51 (for heterosexual women and homosexual women rating
stated interests of women) to .94 (for homosexual men and
homosexual women rating appearance of women). Across all
raters, inter-rater reliability (Cronbach’s a) exceeded .89 for
each measure. Thus, for the majority of subsequent analyses,
we used each target’s average score (averaged across all
raters) for each of the observer-rated target measures.
How Accurate is Gaydar?
We investigated whether raters’ judgments of sexual orien-
tation were related to self-identified orientation. Ratings of
the targets’ overall presentation (from the unmodified vid-
eos) indicated that both male and female homosexual targets
were perceived as more homosexual than heterosexual tar-
gets. These effects were large; p \ .0001, d = 2.0, r = .63
for men, and p \ .0001, d = 1.8, r = .62 for women
(Fig. 1). The strength of these correlations indicated that
raters tended to be very accurate in assessing sexual
orientation.
We then used an additional approach to illustrate how
accurate raters were, on average, in judging the targets’
sexual orientation. We assumed that any target whose aver-
age rating was above 4 (that is, above average, the midpoint
of our scale) was predicted to be homosexual, and any target
whose average rating was below 4 was predicted to be het-
erosexual. Using these definitions, 87% of heterosexual tar-
gets and 75% of homosexual targets were accurately judged.
The difference in percentages was not statistically signifi-
cant, v2 = 2.4, p = .12. In total, 81% of all targets were
Arch Sex Behav (2010) 39:124–140 129
123
accurately judged, which was significantly more than 50%,
the expected chance value, v2 = 39.5, p \ .0001. These re-
sults emphasize the high accuracy of observers’ assessments.
Components of Gaydar
Next, we investigated how observer ratings of different
components of targets’ behavior related to targets’ self-
identified sexual orientation, that is, whether the individual
components functioned as useful gaydar signals. We con-
ducted a series of t-tests. Homosexual targets were generally
perceived as moving, sounding, and looking more homo-
sexual compared with heterosexual targets (Fig. 1). For
women, the correlation of sexual orientation with observer
ratings of their stated interests approached statistical signif-
icance, but for men it did not. In sum, these analyses suggest
that several observable components of behavior provided
substantial information regarding targets’ sexual ori-
entations.
The correlation of targets’ sexual orientation with ob-
server-rated sexual orientation based on movement could be
enhanced by information about targets’ appearance (e.g.,
their hairstyles or body shape). Although raters of movement
watched videos modified to minimize information about
targets’ appearance, the video modification did not com-
pletely eliminate such information. Indeed, after partialling
out the effect of appearance, and across all targets, movement
of homosexual targets was no longer perceived as
significantly more homosexual, p = .14, r = .15. Male and
female targets did not significantly differ in this partial cor-
relation, p = .41, b = .09. This effect was considerably
smaller in magnitude than the corresponding first order cor-
relation of sexual orientation with movement, p \ .0001,
r = .49. Thus, movement of homosexual targets may have
been rated as more homosexual than that of heterosexual
targets because of observed differences in appearance.
Similarly, the correlation of sexual orientation with
speech could be enhanced because raters evaluated speech
based on listening to targets talking about their interests.
However, across all targets, speech of homosexual targets
was rated as significantly more homosexual even after par-
tialling out the effect of observer-rated interests, p \ .0001,
r = .50. Male and female targets did not significantly differ
in this partial correlation, p = .23, b = -.11. Furthermore,
this effect was comparable in magnitude to the corresponding
first order correlation, p \ .0001, r = .53. Thus, speech
ratings were only marginally influenced by the content of
targets’ statements.
Figure 1 also shows that, for some types of stimuli, ob-
server ratings of sexual orientation varied more for homo-
sexual targets than for heterosexual targets. Levene tests for
unequal variance, which compared the magnitude of absolute
residuals, indicated that this sexual orientation difference in
variance was marginally significant for overall presenta-
tion and speech, p = .08, b = .18, and p = .06, b = .20,
respectively, and was significant for movement and
obse
rver
-rat
ed fe
mal
e se
xual
orie
ntat
ion
1
4
7
l a u x e s - o m o h - o r e t e h
l a u x e s - o m o h - o r e t e h
l a u x e s - o m o h - o r e t e h
l a u x e s - o m o h - o r e t e h
obse
rver
-rat
ed m
ale
sexu
al o
rient
atio
n
1
4
7
t n e m e v o M s t s e r e t n I e c n a r a e p p A h c e e p S
* ) 4 4 . ( 2 . 1
) 8 2 . ( 7 . 0
* * * ) 4 5 . ( 4 . 1 ) 4 1 . ( 3 . 0 * * ) 7 4 . ( 2 . 1
* * * ) 6 6 . ( 3 . 2 * * ) 3 5 . ( 3 . 1 * * * ) 6 5 . ( 9 . 1
l l a r e v O
* * * ) 3 6 . ( 0 . 2
l a u x e s - o m o h - o r e t e h
* * * ) 2 6 . ( 8 . 1
Fig. 1 Observer-rated sexual orientation of 25 heterosexual and 25
homosexual male targets, and 22 heterosexual and 23 homosexual
female targets. The five measures are ratings of sexual orientation from
visual and sound recordings: overall presentation, movement, speech,
appearance, and stated interests. Points represent the scores of
individual targets, averaged across raters. Lines are 95% confidence
intervals. On the Y-axis, 1 is the most heterosexual and 7 the most
homosexual score. Numbers are effect sizes expressed as Cohen’s d’s
and, in parentheses, as correlation coefficients (rs) � p \ .10; * p \ .05;
** p \ .001; *** p \ .0001
130 Arch Sex Behav (2010) 39:124–140
123
appearance, p = .0002, b = .37, and p = .005, b = .28,
respectively. The sexes did not significantly differ for these
effects (ps between .20 and .36, bs between -.11 and .13).
That is, observer ratings varied more for homosexual targets,
regardless of their sex. We further explored this sexual ori-
entation difference in variability in Study 2.
Rater Differences in Gaydar
For each observer-rated measure, we examined whether the
four groups of raters (heterosexual and homosexual men and
women) differed in how accurately they judged the targets’
self-identified sexual orientation. We computed mixed-
effects regression analyses, treating targets as random, to
acknowledge repeated measures of targets across the four
groups of raters (Zar, 1999). The dependent variable was the
targets’ observer-rated sexual orientation, averaged within
each group of raters. Independent variables were the raters’
sex, sexual orientation, and their interaction. Further inde-
pendent variables were the targets’ sexual orientation and the
interactions of the targets’ sexual orientation with the raters’
sex, sexual orientation, and their combination. With these
interactions, we tested whether the relationship of the targets’
self-identified and observer-rated sexual orientation differed,
depending on the raters’ sex and sexual orientation. Analyses
were done separately for the male and female targets.
Some group differences in rating accuracy were found.
These depended on the measure, and they did not indicate that
any group of raters was especially accurate or inaccurate,
compared with the other groups. For example, homosexual
women were less accurate than other groups rating male
targets using movement and speech; all ps \ .01, b’s be-
tween -.09 and -.06. Male raters were more accurate
judging male targets based on appearance compared with
female raters, p = .01, b = .08. Homosexual raters were
more accurate judging female targets based on movement
compared with heterosexual raters, p = .0007, b = .09.
Finally, heterosexual male raters were less accurate rating
female targets based on speech compared with the other
groups, p \ .01, b = -.10. Thus, no clear group differences
in accuracy emerged. Moreover, the significant group dif-
ferences were relatively small in magnitude (using the cri-
teria suggested by Cohen, 1988).
Sexual Orientation and Self-reported Sex Atypicality
Before investigating the correlation of gaydar signals with
self-reported sex atypicality, we explored whether, as in
previous research, homosexual people reported more sex
atypicality than heterosexual people. On all self-report
measures, homosexual targets indicated more sex atypicality
compared with heterosexual targets, though not always sig-
nificantly so. This difference was statistically significant for
males’ and females’ reports of childhood sex atypicality and
for males’ self-reported interests, and insignificant but in the
predicted direction for the other comparisons (Fig. 2).
Because the three self-report measures were correlated for
both male targets (r’s between .43 and .56; a = .75) and
female targets (r’s between .63 and .67; a = .85), we also
computed the average score for each target. On this com-
posite of self-reported sex atypicality, both male and female
homosexual targets scored significantly higher than hetero-
sexual targets (Fig. 2).
Are Gaydar Signals Related to Self-reported Sex Atypicality?
We next investigated whether signals important for gaydar
were related to self-reported sex atypicality. We computed
correlations among the five observer-rated measures of sex-
ual orientation and the three self-reported measures of sex
atypicality (and their composite). To give our analyses more
power, we first computed correlations across all targets.
Specifically, we computed the partial correlations controlling
for both sexual orientation and sex, to show their magnitudes
independent of group. All of the partial correlations were
positive and most were significant (Table 1, below the
diagonal). For example, the partial correlation of gaydar
ratings based on overall presentation with the composite of
self-reported sex atypicality was moderate to strong in
magnitude, p \ .0001, r = .46. Thus, for both sexes and
sexual orientations, individuals who were rated as relatively
homosexual on observable components tended to report
more sex atypicality.
Table 1 also shows the intercorrelations between different
measures of observer-rated sexual orientation. Individuals
who were rated as relatively homosexual on one observable
component tended to be rated as homosexual on other com-
ponents as well.
The partial correlations in Table 1 suggest that measures
of observer-rated sexual orientation and self-reported sex
atypicality tended to be related, in general, for heterosexual
and homosexual male and female targets. It seemed possible
that these correlations would differ for the different target
groups. For example, because homosexual targets showed
somewhat more variability with respect to gaydar signals
(Fig. 1), it seemed possible that they would show stronger
correlations of gaydar signals with self-reported sex atypi-
cality. We therefore conducted a series of regression ana-
lyses. In each analysis, we computed interactions that tested
whether the relationship between two measures was moder-
ated by the targets’ sex, sexual orientation, or their combi-
nation. Four significant interactions were found, all for
relationships between two observer-rated measures of sexual
orientation. However, these interactions did not suggest
consistent sex or sexual orientation differences, and their
meaning was unclear. For the remaining correlations shown
Arch Sex Behav (2010) 39:124–140 131
123
in Table 1, heterosexual and homosexual men and women
did not significantly differ.
Discussion
Observers in Study 1 assessed sexual orientation with high
accuracy. For example, based on our operationalization of
judged sexual orientation (using the midpoint of the rating
scale as a cut-off for ‘‘gay’’ versus ‘‘straight’’ ratings), 81% of
all targets were accurately judged based on only a few
seconds of observation of each. The raters’ own sex and
sexual orientation had only minor influences on the accuracy
of their assessments. Furthermore, assessments of sexual
orientation were related to the targets’ self-reported sex
atypicality.
Several components hypothesized to be gaydar signals
seem to have helped observers to accurately judge targets’
sexual orientation: the targets’ movement, speech, appear-
ance, and, to a lesser degree, their stated interests (Fig. 1).
These results confirm previous findings, specifically the
1
4
7
1
4
7
l a u x e s - o m o h - o r e t e h
l a u x e s - o m o h - o r e t e h
l a u x e s - o m o h - o r e t e h
d o o h d l i h C
* * * ) 7 5 . ( 4 . 1
* * * ) 3 6 . ( 9 . 1
d o o h t l u d A
) 8 2 . ( 7 . 0
) 5 1 . ( 3 . 0
s t s e r e t n I
) 9 2 . ( 7 . 0
* ) 7 3 . ( 8 . 0
self-
repo
rted
fem
ale
sex
atyp
ical
ity
self-
repo
rted
mal
e se
x at
ypic
ality
e t i s o p m o C
* ) 3 4 . ( 1 . 1
* * ) 6 4 . ( 2 . 1
l a u x e s - o m o h - o r e t e h
Fig. 2 Self-reported sex atypicality of 25 heterosexual and 25 homo-
sexual male targets, and 22 heterosexual and 23 homosexual female
targets. The three self-report measures are childhood and adulthood sex
atypicality, and interests in sex-atypical hobbies and occupations. For
each measure, points represent the average scores of individual targets.
The fourth measure is a composite (average) of the three self-reported
measures. Lines are 95% confidence intervals. On the Y-axis, 1 is the
most sex-typical and 7 the most sex-atypical score. Numbers are effect
sizes expressed as Cohen’s d’s and, in parentheses, as correlation
coefficients (rs) � p \ .10; * p \ .05; ** p \ .001; *** p \ .0001
Table 1 Intercorrelations among observer-rated sexual orientation and self-reported sex atypicality across all targets (N = 95)
Measure Observer-rated Self-report
Overall Movement Speech Appearance Interests Childhood Adulthood Interests Composite
Observer-rated
Overall .67*** .78*** .70*** .41*** .56*** .50*** .47*** .56***
Movement .57*** .50*** .72*** .13 .48*** .46*** .38** .52***
Speech .69*** .35** .43*** .41*** .52*** .43*** .40*** .56***
Appearance .56*** .60*** .23* .14 .43*** .42*** .40*** .49***
Interests .38** .05 .38** .04 .16 .26* .25* .26*
Self-report
Childhood .29* .28* .29* .15 .05 .53*** .60*** .84***
Adulthood .48*** .42*** .38** .38** .23* .52*** .59*** .84***
Interests .37** .29* .38** .30* .21* .54*** .57*** .86***
Composite .46*** .40*** .43*** .34** .20* .80*** .85*** .85***
Note: Numbers above the diagonal are simple correlations. Numbers below the diagonal are partial correlations, after partialling out the effects of
sexual orientation and sex. * p \ .05; ** p \ .001; *** p \ .0001
132 Arch Sex Behav (2010) 39:124–140
123
assessment of sexual orientation based on movement
(Ambady et al., 1999; Johnson et al., 2007) and speech (e.g.,
Bailey, 2003; Gaudio, 1994; Linville, 1998). To our knowl-
edge, however, the present research is the first to investigate
these different components within a single study, and to
establish that assessments of sexual orientation based on
these different gaydar signals were generally correlated
(Table 2). That is, for example, if raters judged a target as
especially likely to be homosexual based on his or her speech,
they were also more likely to do so based on movement and
appearance. Thus, gaydar may be based on similar informa-
tion coming from different channels.
To be sure, our analyses have some limitations. For ex-
ample, although we tried to reduce information about physi-
cal appearance in the videos used to rate movement, we did
not eliminate such information completely. Thus, raters may
have relied to some extent on targets’ appearance while
making their judgments. It is unclear from our data how well
movement alone worked as a gaydar signal. A recent study,
however, found that sexual orientation could be judged from
movement alone (Johnson et al., 2007).
There was no consistent evidence that certain groups of
raters had superior gaydar. In contrast to some other studies
(Ambady et al., 1999; Berger et al., 1987; Sylva et al., 2007),
homosexual raters were not generally more accurate. As we
noted, however, the accuracy advantage of homosexual rat-
ers found in those studies was modest and somewhat incon-
sistent. If homosexual and heterosexual people have similar
levels of gaydar ability, this suggests that gaydar does not rely
much on information more available to homosexual people.
We note, however, that our heterosexual raters were college
students from a city with a large gay community. Thus, they
may have had more exposure to homosexual people com-
pared with most heterosexual people. It therefore remains
unclear how much familiarity with homosexuality enhances
gaydar.
What precise information do people rely on for gaydar? In
our study, observer-rated sexual orientation was related to
self-reported sex atypicality (Table 2), which, in turn, was
related to self-identified sexual orientation (Fig. 2). Sex
atypicality may also be expressed through people’s ob-
servable behavior (Bailey, 2003; Lippa, 1998; Rieger et al.,
2008). Thus, it may have been the sex atypicality of targets’
movement, speech, appearance, and, to a lesser extent,
interests, that enabled observers to infer their sexual orien-
tation with such high accuracy. If so, then perceived sexual
orientation should be correlated with perceived sex atypi-
cality. We assessed this relationship in Study 2.
Study 2
The second study focused on the relationship of observer-
rated sex atypicality with observer ratings of sexual orien-
tation obtained in Study 1. We employed a second group of
raters in this study, separate from those used in Study 1, in
order to avoid contamination effects. Study 2 also explored
the relationship of observer-rated sex atypicality with the
targets’ self-identified sexual orientation, as well as the
intercorrelations among different measures of sex atypi-
cality.
Method
Participants
Target participants for Study 2 were the same as in Study 1.
However, a new set of raters participated. Ten homosexual
men and 11 homosexual women who were recruited via the
same media as described in Study 1, and 16 heterosexual men
and 21 heterosexual women who were undergraduate stu-
dents in an introductory psychology class, rated targets on
several components of sex atypicality. Mean ages of raters
were 22.1 (4.2), 23.2 (3.6), 19.6 (1.3), and 19.0 (1.2) years,
respectively. Homosexual raters were older than the hetero-
sexual raters, p \ .0001, b = .57. Because judgments of the
four groups of raters correlated with each other and barely
differed (see below), differences in age did not seem to have
Table 2 Correlations of observer-rated sexual orientation with the respective measures of observer-rated sex atypicality
Measure All targets
(N = 95)
Heterosexual men
(n = 25)
Homosexual men
(n = 25)
Heterosexual women
(n = 22)
Homosexual women
(n = 23)
Overall .75*** .77*** .79*** .68** .76***
Movement .91*** .89*** .89*** .94*** .93***
Speech .77*** .81*** .94*** .57* .61*
Appearance .89*** .77*** .90*** .92*** .93***
Interests .70*** .95*** .91*** .42* .16
Note: Correlation coefficients are shown for all targets (after partialling out the effects of sexual orientation and sex), and separately for heterosexual
and homosexual men and women. * p \ .05; ** p \ .001; *** p \ .0001
Arch Sex Behav (2010) 39:124–140 133
123
an important influence on ratings. Fifty-three percent of the
raters were Caucasian. This proportion did not significantly
differ across the groups, v2(3) \ 1. Nineteen percent were
Hispanic and the remainder identified as Asian, African-
American, or ‘‘other.’’
Procedure
The procedure for Study 2 was identical to the procedure for
Study 1, except that in Study 2 raters assessed how sex typical
or sex atypical the targets were (instead of rating target’s
sexual orientation, as in Study 1). For example, after each
video showing figural outlines of a male target, raters were
asked ‘‘whether this man gestures, moves, or sits in a more
masculine or feminine manner.’’ For male targets, a score of 1
was labeled more masculine, 4 was labeled average, and 7
was labeled more feminine. A reversed scale was used for
ratings of female targets. Thus, for both sexes and across
stimuli, a score of 7 represented maximum sex atypicality.
Results and Discussion
Inter-rater Consistency
Within each target sex and for each observer-rated measure,
ratings by the four groups (heterosexual and homosexual men
and women) were correlated; the median r was .77, and the
range was from .40 (for heterosexual men and homosexual
women rating stated interests of women) to .96 (for hetero-
sexual men and heterosexual women rating the appearance of
women). Across all raters, inter-rater reliability (a) was equal
to or exceeded .80 for each measure. Thus, in subsequent
analyses, we used each target’s average score (averaged
across all raters) for each of the observer-rated measures.
Are Perceptions of Sexual Orientation and Sex Atypicality
Related?
We computed correlations of observer-rated sexual orien-
tation (the ratings analyzed in Study 1) with observer-rated
sex atypicality. Observer ratings of targets’ sexual orienta-
tion based on targets’ overall presentation were correlated
with observer ratings of sex atypicality based on overall pre-
sentation; corresponding correlations were computed for
movement, speech, appearance, and stated interests. These
correlations were computed across all targets, after partial-
ling the effects of target sexual orientation and sex, to show
their magnitude independent of target group. Correlations
were also calculated separately for heterosexual and homo-
sexual men and women. As is evident in Table 2, correla-
tions were generally positive and large. An exception was the
correlation for stated interests of women, especially homo-
sexual women, which was notably lower than the same
correlation for males. The median of the correlation coeffi-
cients calculated for each target group (.85) also indicated a
generally strong relationship.
Thus, for both unedited videos and for most components,
targets whose presentations were rated as more sex atypical
were also those who presentations were judged to be more
indicative of homosexuality. Correspondingly, those who
were seen as more sex typical were also rated as ‘‘more
straight.’’ Perhaps, then, it is the sex atypicality of targets’
behavior that enables observers to judge their sexual orien-
tation with high accuracy. This is consistent with earlier
findings that homosexual people are, on average, sex atypical
in their observable behaviors (Bailey, 2003; Johnson et al.,
2007; Rieger et al., 2008).
Are Homosexual People Perceived to be More Sex
Atypical?
The above argument suggests that self-identified homosexual
people will be perceived by others as more sex atypical. In-
deed, homosexual men and women were generally judged to
be more sex atypical than heterosexual targets, both in their
overall presentation and for several observer-rated compo-
nents (Fig. 3).
As was the case for observer-rated sexual orientation
in Study 1, homosexual targets showed more variation in
observer-rated sex atypicality, as compared to the hetero-
sexual targets (Fig. 3). The sexual orientation difference in
variance was significant for movement, speech, and appear-
ance, p = .0002, b = .38, p = .003, b = .30, and p = .04,
b = .21, respectively. The sexes did not significantly differ
for these effects (ps between .12 and .64, bs between -.10
and .14). Thus, heterosexual targets were generally rated as
sex typical (though not necessarily extremely so) with respect
to both overall presentation and the separate components, but
perceptions of homosexual targets ran the gamut from quite
sex typical to quite sex atypical.
Some other research has also found greater variability in
sex atypicality among homosexual people than among het-
erosexual people, both when measured by self-report (Lippa,
2005b) and by observer ratings (Rieger et al., 2008).
Admittedly, the present study found this pattern only for
observer ratings but not for self-report. Nevertheless, the
general finding across all these studies is that homosexual
people show more variability. Lippa (2005b) argued that both
biological and social factors might contribute to this differ-
ence. For example, the predominant pressure towards typical
sex-role socialization may make some homosexual individ-
uals behave in a sex-typical manner. At the same time, bio-
logical predispositions and counteractive influences against
stereotypical sex-typed behaviors within the gay and lesbian
community may make other individuals more likely to be-
have in a sex-atypical manner.
134 Arch Sex Behav (2010) 39:124–140
123
A comparison of Figs. 1 and 3 indicates another interest-
ing pattern. For male targets, observer-ratings of both sexual
orientation and sex atypicality showed the largest sexual
orientation differences when based on targets’ full presen-
tation, rather than on just one channel (i.e., movement, speech
patterns, appearance, or stated interests). It seems reasonable
that raters would distinguish more between the two groups of
targets when they had the most information about them. In
contrast, however, ratings of female targets based just on
appearance yielded stronger sexual orientation differences
than ratings based on the full presentation; this was true for
ratings of both sexual orientation and sex atypicality. Perhaps
appearance is the best indicator of sexual orientation and sex
atypicality in women. Information conveyed by appearance
alone can then be obscured if behavioral information is also
available, due to overweighting of the latter. Future research
should investigate this possibility.
How Closely Related are Different Measures of Sex
Atypicality?
In Study 1, we found that the different measures of observer-
rated sexual orientation tended to be intercorrelated, as well
as related to self-reported sex atypicality (Table 1). Given the
high correlations between observer-rated sexual orientation
and observer-rated sex atypicality (Table 2) it seemed likely
that the different measures of observer-rated sex atypicality
would also be intercorrelated and, in addition, related to self-
reports. We computed the partial correlations among the five
observer-rated measures and the three self-reported measures
of sex atypicality (controlling for sexual orientation and sex).
All of the partial correlations were positive and most were
significant (Table 3, below the diagonal).
Similar to Study 1, a set of regression analyses found four
group differences for these correlations. However, these did
not demonstrate consistent sex or sexual orientation differ-
ences. For the majority of the correlations shown in Table 3,
heterosexual and homosexual men and women did not sig-
nificantly differ. In general, then, targets who were rated as
more sex atypical based on one type of stimulus (e.g., speech)
tended to be rated as more sex atypical based on the other
stimuli (e.g., movement or appearance), and further, those
who were rated by observers as more sex atypical also tended
to report more sex atypicality themselves.
General Discussion
Results of Study 1 suggest that observers can detect others’
sexual orientations with high accuracy based on short
audiovisual recordings. Thus, some aspects of people’s
behavior and appearance serve as effective gaydar signals.
These signals may include movement and speech, as well as
some appearance cues evident in still photographs. Study 2
showed that observers’ ratings of targets’ sexual orientation
1
4
7
lauxes-omoh-oreteh
lauxes-omoh-oreteh
lauxes-omoh-oreteh
lauxes-omoh-oreteh
1
4
7
tnemevoM stseretnIecnaraeppAhceepS
*)83.( 0.1
*)54.( 0.1
**)64.( 4.1 )80.( 0.0*)92.( 7.0
***)36.( 8.1*)24.( 1.1***)45.( 8.1
llarevO
***)46.( 7.1
-omoh-oreteh
***)95.( 6.1
lauxes
obse
rver
-rat
ed fe
mal
ese
x at
ypic
ality
obse
rver
-rat
ed m
ale
sex
atyp
ical
ity
Fig. 3 Observer-rated sex atypicality of 25 heterosexual and 25
homosexual male targets, and 22 heterosexual and 23 homosexual
female targets. The five measures are ratings of sex atypicality from
visual and sound recordings: overall presentation, movement, speech,
appearance, and stated interests. Points represent the scores of
individual targets, averaged across raters. Lines are 95% confidence
intervals. On the Y-axis, 1 is the most sex-typical and 7 the most sex-
atypical score. Numbers are effect sizes expressed as Cohen’s d’s and, in
parentheses, as correlation coefficients (rs) � p \ .10; * p \ .05;
** p \ .001; *** p \ .0001
Arch Sex Behav (2010) 39:124–140 135
123
corresponded closely with their ratings of targets’ sex atyp-
icality. Thus, our results were consistent with the hypothesis
that gaydar signals derive their information from sex
atypicality.
We also found that different measures of sex atypicality,
including observer ratings and self-reports, tended to be
correlated with sexual orientation (Figs. 2, 3) and with each
other (Table 3). Similar convergence of measures of sex-
typed behavior and sexual orientation has been previously
reported (Lippa, 1998; Rieger et al., 2008). This increases
confidence in the construct validity of relevant measures
because it suggests that their association is due to an under-
lying factor of sex typicality versus atypicality (alternatively,
masculinity–femininity). Lippa (2005a) has hypothesized
that there is a core to masculinity–femininity, which includes
sex-typed interests, appearance, mannerisms, nonverbal
behaviors, and possibly sexual orientation. To our knowl-
edge, ours is the first study to investigate associations among
all these variables within a single study, and one of the first to
provide direct support for the idea that gaydar signals reflect
masculinity and femininity.
Our results were consistent with some stereotypes about
differences between heterosexual and homosexual people.
For example, homosexual men are believed to have more
feminine mannerisms and ‘‘gay sounding’’ voices. Lesbians
are believed to ‘‘look butch’’ (e.g., Kite & Deaux, 1987).
Despite their support for stereotypes, our results also showed
that perceived behavior was an imperfect predictor of sexual
orientation. Most of the variation in targets’ self-identified
sexual orientation could not be explained by the raters’
judgments based on the targets’ overall presentation
(R2 = .40 for ratings of sexual orientation; R2 = .37 for
ratings of sex atypicality). Thus, although the average het-
erosexual and homosexual persons confirmed stereotypes
about differences between the groups, some individuals de-
fied the stereotypes. This was most common among homo-
sexual targets, some of whom behaved quite ‘‘straight’’ or
sex-typically.
Etiology and Development of Sexual Orientation
The general finding that homosexual orientation is associated
with sex atypicality suggests that at least some homosexual
individuals have atypical psychological sexual differentia-
tion. Much research suggests that this occurs by childhood
(e.g., Bailey & Zucker, 1995; Rieger et al., 2008). Although
in principle this could reflect either nature or nurture, or both,
we agree with those who have surveyed relevant research and
found innate explanations to be more promising (e.g., Bell,
Weinberg, & Hammersmith, 1981; Lippa, 2005a; Rahman &
Wilson, 2003). More specifically, biological factors affecting
sexual differentiation include the effects of prenatal sex
steroid hormones.
Male and female sexual orientation had similar correlates
in our study, with respect to both self-reported and observed
sex atypicality, although there were some differences. This
suggests that at least some of the causes of sexual orientation
are symmetric between men and women. Recent research has
focused attention on ways that male and female sexual ori-
entations appear quite different (e.g., Chivers, Rieger, Latty,
& Bailey, 2004; Diamond, 2008). Results such as ours sug-
gest that there may also be ways in which they are funda-
mentally similar.
Social Implications
The present findings suggest that many homosexual people
may be seen as relatively sex atypical and thus likely to be
Table 3 Intercorrelations among observer-rated and self-reported sex atypicality across all targets (N = 95)
Measure Observer-rated Self-report
Overall Movement Speech Appearance Interests Childhood Adulthood Interests Composite
Observer-rated
Overall .71*** .62*** .65*** .26* .55*** .49*** .39** .56***
Movement .63*** .48** .69*** .18 .53*** .43*** .38** .53***
Speech .50*** .37** .35** .37** .50*** .34** .39** .49***
Appearance .61*** .60*** .26* .20* .43*** .46*** .40*** .51***
Interests .16 .06 .31* .10 .30* .39** .46*** .45***
Self-report
Childhood .30* .37** .33** .25* .21* .53*** .60*** .84***
Adulthood .48*** .38** .28* .44*** .35** .52*** .59*** .84***
Interests .26* .29* .30* .34** .43*** .54*** .57*** .86***
Composite .41*** .41*** .36** .42*** .41** .80*** .85*** .85***
Note: Numbers above the diagonal are simple correlations. Numbers below the diagonal are partial correlations, after partialling out the effects of
sexual orientation and sex. * p \ .05; ** p \ .001; *** p \ .0001
136 Arch Sex Behav (2010) 39:124–140
123
homosexual. These evaluations happen quickly, within sec-
onds. To our knowledge, social consequences of these
judgments have not been systematically studied. However, it
is plausible that being initially judged as sex atypical or
‘‘gay’’ can influence how a person will be treated, because
first impressions seem important for subsequent interactions
(e.g., Jones, 1990).
Sex-atypical people suffer from social disadvantages.
Stigmatization of sex atypicality can occur at an early age.
Sex-atypical children are sometimes treated negatively
by parents and peers, probably because they violate ex-
pected gender norms (Beard & Bakeman, 2000; Carter &
McCloskey, 1983; Fagot, 1977, 1985; Landolt et al., 2004;
Maccoby, 1998; Rieger et al., 2008; Saghir & Robins, 1973;
Smith & Leaper, 2006; Tauber, 1979; Zucker, Wilson-Smith,
Kurita, & Stern, 1995). Furthermore, problems due to child-
hood sex atypicality may persist into adulthood. Homo-
sexual men who reported having been more sex atypical as
children were more likely to report dysfunctional relation-
ships with parents, particularly fathers (Beard & Bakeman,
2000; Freund & Blanchard, 1983), as well as problematic
attachments to others (Landolt et al., 2004). Furthermore,
although our results suggest that sex atypicality is especially
common among homosexual people, homosexual people
also appear to be somewhat intolerant of sex atypicality.
Homosexual men are especially likely to report that they
dislike sex-atypical people (Skidmore, Linsenmeier, &
Bailey, 2006), and that they find them less attractive as
potential romantic partners (e.g., Bailey, Kim, Hills, &
Linsenmeier, 1997; Laner, 1978; Laner & Kamel, 1977).
Given these stressors, it is plausible that sex-atypical
people experience more psychological distress. Consistent
with this hypothesis, sex-atypical homosexual men report
lower self-esteem (Harry, 1983a), more depression and
anxiety (Skidmore et al., 2006; Weinrich et al., 1992), and
higher suicidal risks (Harry, 1983b; Remafedi, French, Story,
Resnick, & Blum, 1998), compared with sex-typical homo-
sexual men. Notably, the link between sex atypicality and
distress is less robust in women than in men (Aube &
Koestner, 1992; Harry, 1983b; Skidmore et al., 2006).
Many studies of the correlates of sex atypicality involve
ongoing relationships with, for example, parents and peers
who have repeated opportunities to judge the degree to which
a person’s behavior, appearance, and interests are sex atyp-
ical. Future research should investigate whether socially
relevant consequences of sex atypicality can result from first
impressions based on ‘‘thin slices’’ of behavior.
Limitations
The present study has some important methodological limi-
tations. These include the correlational nature of the data; the
inability to establish clear correlations between some
observer-rated measures; the brevity of video clips, possibly
leading to unreliable observer ratings; the operationalization
of judged sexual orientation for one of our analyses; and
possible volunteer bias of both targets and raters due to small,
self-selected, and non-representative study samples.
Correlational Results
One interpretation of our results is that raters first judged
some targets as sex atypical and consequently classified them
as likely to be gay. Given that these results were based on
correlations, there is no evidence that the raters’ assessments
did indeed follow this process. It is possible, for example, that
raters judged some targets to be sex atypical because they
appeared to be gay rather then the reverse. Or, they may have
rated some targets as homosexual for other reasons than their
apparent sex atypicality. For example, homosexual targets
may have been, on average, more comfortable participating
in a study on sexual orientation than heterosexual targets and,
thus, more relaxed and expressive in front of the camera. Our
scales may have forced raters to judge the more expressive
style of homosexual targets as ‘‘gay’’ or sex atypical because
these were the only choices available. To test this assump-
tion, an independent group of raters judged the expressive-
ness of targets. Ratings of expressiveness were unrelated to
ratings of sexual orientation and sex atypicality; r’s = .03
and .04. Thus, raters did not seem to be influenced by the
targets’ expressive style. Still, future research should inves-
tigate whether there is a causal link between judgments of sex
atypicality and of sexual orientation. Future research should
also examine the extent to which observer ratings of other
personality traits (e.g., anxiety and extroversion) relate to
self-reported and observer-rated sexual orientation.
Correlations Among Measures
The correlations between some of the observer-rated mea-
sures could have been enhanced by overlapping cues. In
particular, as noted before, the correlation of observer-rated
movement and appearance could have been enhanced be-
cause raters of movement had some information on appear-
ance. This weakens our interpretation that the association of
these measures reflects an underlying factor of sex atypical-
ity. We note, however, that for the correlations between other
measures, in particular for speech ratings with ratings of
movement and appearance, no such confounds were possible.
Raters of speech had no information on the other components
and vice versa, but the correlations between these measures
were significant (Tables 1 and 3). In addition, observer rat-
ings were correlated with self-reports of sex atypicality. This
increases our confidence in the conclusion that there is a core
factor of sex atypicality. Nevertheless, future research should
Arch Sex Behav (2010) 39:124–140 137
123
employ more precise methodologies for establishing these
intercorrelations.
Brief Behavioral Samples
We used video clips that were at most 10 seconds long to
assess sex atypicality. Ratings based on such short samples
are likely to contain measurement error. Thus, the true dif-
ferences in presentation between heterosexual and homo-
sexual people may be larger than those reported here.
Furthermore, the correlations between several observer-
rated measures may be underestimated by our method.
Operationalization of Judged Sexual Orientation
For one analysis of rater accuracy, we used the midpoint of
the 7-point rating scale as a cut-off to classify judgments as
‘‘gay’’ versus ‘‘straight.’’ This approach ignores the fact that
raters were somewhat uncertain about many targets and thus
judged them, for example, ‘‘somewhat more gay’’ than
average rather than simply ‘‘gay.’’ Different operationaliza-
tions (e.g., requiring raters to make discrete guesses about
sexual orientation) might yield different results.
We also note that unlike the present research, some
researchers have assessed masculinity and femininity as
separate dimensions (e.g., Bem, 1974; Spence & Helmreich,
1978). The use of a two-dimensional conceptualization of
masculinity and femininity, like the use of a different ap-
proach to assessing perceived sexual orientation, might have
yielded somewhat different results than those reported here.
However, we believe that a unidimensional model of mas-
culinity–femininity is currently the most scientifically
promising (see Lippa, 2005a).
Representativeness of Samples
Targets were a self-selected group, which might have re-
sulted in some sampling bias. For example, the majority of
homosexual targets were openly homosexual. It is possible
that more sex-atypical homosexual people are more likely to
be ‘‘out,’’ perhaps because their presentation makes it harder
for them to be closeted about their sexual orientation. If this is
true, our sample of homosexual people may have been gen-
erally biased towards sex atypicality.
Some evidence consistent with this possibility comes from
inspection of effect sizes in our data. For self-reported
childhood sex atypicality (Fig. 2), the effect sizes of sexual
orientation for both male targets (d = 1.9) and female targets
(d = 1.4) were somewhat larger than the effect sizes reported
in the meta-analysis by Bailey and Zucker (1995; respective
d’s = 1.3 and 1.0). If the homosexual targets in the current
study recalled higher levels of childhood sex atypicality than
most homosexual people report, this would make our sample
unusual. It is also possible, however, that homosexual par-
ticipants have become more willing to report higher levels of
childhood sex atypicality since the meta-analysis was pub-
lished in 1995. Some indication for this comes from a recent
study on twins, including thousands of participants (Bailey,
Dunne, & Martin, 2000). In this study, the respective d’s were
approximately 1.8 for both men and women.
Finally, both targets and raters came from a large, North
American urban area with a noticeable gay community. It is
possible that under such conditions, people with the same
sexual orientation tend to display certain behaviors because
of a shared social environment. Furthermore, increased
familiarity with homosexuality may enable both heterosex-
ual and homosexual people alike to more accurately judge
sexual orientation. In other words, both heterosexual and
homosexual people may learn how a typical person of their
sexual orientation behaves in their environment, and this may
make them more likely to behave in such ways themselves.
Both may also learn what behaviors are typical for those of
the other orientation. This knowledge could then enhance
their ability to identify people from either group. We note,
however, that even if our participants were atypical because
they had unusual knowledge about sexual orientation, this
would still imply that there is validity to the notion that
gaydar signals distinguish homosexual and heterosexual
individuals.
Future Directions
As pointed out earlier, future research should investigate
whether social consequences can result from initial impres-
sions that a person is ‘‘gay’’ or sex atypical. For example,
people may be rejected or judged as unattractive based on
‘‘thin slices’’ of behavior indicating sex atypicality. Future
studies should also look at whether such initial evaluations
are related to subsequent social interactions and to the level of
psychological distress experienced by the person who was
judged.
Further studies should also explore the considerable var-
iation in sex-atypical behavior of homosexual people. Some
homosexual people appear quite sex typical and are thus ‘‘off
the gaydar.’’ One possible explanation is that due to stig-
matization, some homosexual people may conceal behaviors
that signal their sexual orientation. Studies could investigate
how well homosexual people can conceal their sex atypical
behavior and if they can more effectively control the degree
of sex-atypicality displayed through some channels (e.g.,
movement) than through others (e.g., voice). Alternatively,
differences in sex atypicality may reflect different pathways
to becoming homosexual. Developmental research might
address this hypothesis.
Ability to conceal cues to sexual orientation might depend
not only on the channel involved, but also on the context. If
138 Arch Sex Behav (2010) 39:124–140
123
concealment of cues to sexual orientation is a possible but
effortful process, then ability to conceal should be decreased
when people are engaged in other cognitively demanding
tasks (e.g., Gilbert & Osborne, 1989). This, too, might be
addressed by future research.
Finally, we propose a cross-cultural study on the accuracy
of gaydar and the nature of gaydar signals. Raters of different
cultures could judge the sexual orientation and sex atypicality
of targets of their own and other cultures. At least three
questions could be addressed with this design. First, is gaydar
based on the same underlying dimension, that is, sex atypi-
cality, across the different cultures? Second, are the exact
gaydar signals similar across cultures? For example, do
similar ways of moving distinguish heterosexual from
homosexual people? Third, do the cultural backgrounds of
either targets or raters influence the accuracy of judgments of
sexual orientation?
Such a study could provide insight into how social factors
influence the codevelopment of sex atypicality and sexual
orientation. For example, if cultures differ substantially in
which behaviors let observers distinguish heterosexual from
homosexual people, and to what extent, then the development
of these behaviors and their association with sexual orienta-
tion may have social influences. If, however, sex-atypical
behaviors and their relation with sexual orientation are sim-
ilar across cultures, other factors may influence their code-
velopment. These factors may, for example, be genetic
predispositions or early exposure to different levels of go-
nadal hormones (for reviews on this topic, see Rahman &
Wilson, 2003; Wilson & Rahman, 2005).
Similarly, investigating the ability to differentiate be-
tween heterosexual and homosexual people across different
cultures may indicate to what degree judgments of sexual
orientation are influenced by social factors. It is possible, for
example, that observers can judge sexual orientation best if
they evaluate people of their own culture, with whose
behaviors they are most familiar. However, it is also possible
that similar behaviors are relevant across cultures, again
suggesting biological influences. Such research might also
provide insight into the origins of stereotypes of heterosexual
and homosexual people.
Further research on gaydar and the roles of sex-typed
behaviors may, thus, illuminate the determinants and con-
sequences of being judged as gay or sex atypical. It may also
have implications for our understanding of the co-develop-
ment of masculinity–femininity and sexual orientation.
References
Ambady, N., Bernieri, F. J., & Richeson, J. A. (2000). Toward a
histology of social behavior: Judgmental accuracy from thin slices
of the behavioral stream. Advances in Experimental SocialPsychology, 32, 201–271.
Ambady, N., Hallahan, M., & Conner, B. (1999). Accuracy of
judgments of sexual orientation from thin slices of behavior.
Journal of Personality and Social Psychology, 77, 538–547.
Ambady, N., & Rosenthal, R. (1992). Thin slices of expressive behavior
as predictors of interpersonal consequences: A meta-analysis.
Psychological Bulletin, 111, 256–274.
Aube, J., & Koestner, R. (1992). Gender characteristics and adjustment:
A longitudinal study. Journal of Personality and Social Psychol-ogy, 63, 485–493.
Bailey, J. M. (2003). The man who would be queen: The science ofgender-bending and transsexualism. Washington, DC: Joseph
Henry Press.
Bailey, J. M., Dunne, M. P., & Martin, N. G. (2000). Genetic and
environmental influences on sexual orientation and its correlates in
an Australian twin sample. Journal of Personality and SocialPsychology, 78, 524–536.
Bailey, J. M., Finkel, E., Blackwelder, K., & Bailey, T. (1995).
Masculinity, femininity, and sexual orientation. Unpublished
manuscript.
Bailey, J. M., Kim, P. Y., Hills, A., & Linsenmeier, J. A. W. (1997).
Butch, femme, or straight acting? Partner preferences of gay men
and lesbians. Journal of Personality and Social Psychology, 73,
960–973.
Bailey, J. M., & Zucker, K. J. (1995). Childhood sex-typed behavior and
sexual orientation: A conceptual analysis and quantitative review.
Developmental Psychology, 31, 43–55.
Barlow, D. H., Hayes, S. C., Nelson, R. O., Steele, D. L., Meeler, M. E.,
& Mills, J. R. (1979). Sex role motor behavior: A behavioral
checklist. Behavioral Assessment, 1, 119–138.
Beard, A., & Bakeman, R. (2000). Boyhood gender nonconformity:
Reported parental behavior and the development of narcissistic
issues. Journal of Gay & Lesbian Psychotherapy, 4, 81–97.
Bell, A., Weinberg, M. S., & Hammersmith, S. K. (1981). Sexualpreference: Its development in men and women. Bloomington, IN:
Indiana University Press.
Bem, S. L. (1974). The measurement of psychological androgyny.
Journal of Consulting and Clinical Psychology, 42, 155–162.
Berger, G., Hank, L., Rauzi, T., & Simkins, L. (1987). Detection of
sexual orientation by heterosexuals and homosexuals. Journal ofHomosexuality, 13, 83–100.
Carter, D., & McCloskey, L. A. (1983). Peers and the maintenance of
sex-typed behavior: The development of children’s conceptions of
cross-gender behavior in their peers. Social Cognition, 2, 294–314.
Chivers, M. L., Rieger, G., Latty, E., & Bailey, J. M. (2004). A sex
difference in the specificity of sexual arousal. PsychologicalScience, 15, 736–744.
Cohen, J. (1988). Statistical power analysis for the behavioral sciences(2nd ed.). Hillsdale, NJ: Lawrence Erlbaum Associates.
Constantinople, A. (1973). Masculinity–femininity: An exception to a
famous dictum? Psychological Bulletin, 80, 389–407.
Diamond, L. M. (2008). Sexual fluidity: Understanding women’s loveand desire. Cambridge, MA: Harvard University Press.
Fagot, B. I. (1977). Consequences of moderate cross-gender behavior in
preschool children. Child Development, 48, 902–907.
Fagot, B. I. (1985). Beyond the reinforcement principle: Another step
toward understanding sex role development. DevelopmentalPsychology, 21, 1097–1104.
Freund, K., & Blanchard, R. (1983). Is the distant relationship of fathers
and homosexual sons related to the sons’ erotic preference for male
partners, or to the sons’ atypical gender identity, or to both? Journalof Homosexuality, 9, 7–25.
Gaudio, R. (1994). Sounding gay: Pitch properties in the speech of gayand straight men. American Speech, 69, 30–57.
Arch Sex Behav (2010) 39:124–140 139
123
Gilbert, D. T., & Osborne, R. E. (1989). Thinking backward: Some
curable and incurable consequences of cognitive busyness. Jour-nal of Personality and Social Psychology, 57, 940–949.
Harry, J. (1983a). Defeminization and adult psychological well-being
among male homosexuals. Archives of Sexual Behavior, 12, 1–19.
Harry, J. (1983b). Parasuicide, gender, and gender deviance. Journal ofHealth & Social Behavior, 24, 350–361.
Johnson, K. L., Gill, S., Reichman, V., & Tassinary, L. G. (2007).
Swagger, sway, and sexuality: Judging sexual orientation from
body motion and morphology. Journal of Personality and SocialPsychology, 93, 321–334.
Jones, E. E. (1990). Interpersonal perception. New York: W.H.
Freeman.
Kite, M. E., & Deaux, K. (1987). Gender belief systems: Homosexuality
and the implicit inversion theory. Psychology of Women Quarterly,11, 83–96.
Landolt, M. A., Bartholomew, K., Saffrey, C., Oram, D., Perlman, D., &
Bartholomew, K. (2004). Gender nonconformity, childhood
rejection, and adult attachment: A study of gay men. Archives ofSexual Behavior, 33, 117–128.
Laner, M. R. (1978). Media mating II: ‘‘Personals’’ advertisements of
lesbian women. Journal of Homosexuality, 4, 41–61.
Laner, M. R., & Kamel, G. L. (1977). Media mating I: Newspaper
‘‘personals’’ ads of homosexual men. Journal of Homosexuality, 3,
149–162.
Linville, S. E. (1998). Acoustic correlates of perceived versus actual
sexual orientation in men’s speech. Folia Phoniatrica et Logo-paedica, 50, 35–48.
Lippa, R. (1991). Some psychometric characteristics of gender diag-
nosticity measures: Reliability, validity, consistency across
domains, and relationship to the big five. Journal of Personalityand Social Psychology, 61, 1000–1011.
Lippa, R. (1995a). Do sex differences define gender-related individual
differences within the sexes? Evidence from three studies.
Personality and Social Psychology Bulletin, 21, 349–355.
Lippa, R. (1995b). Gender-related individual differences and psycho-
logical adjustment in terms of the big five and circumplex models.
Journal of Personality and Social Psychology, 69, 1184–1202.
Lippa, R. (1998). The nonverbal display and judgment of extraversion,
masculinity, femininity, and gender diagnosticity: A lens model
analysis. Journal of Research in Personality, 32, 80–107.
Lippa, R. A. (2000). Gender-related traits in gay men, lesbian women,
and heterosexual men and women: The virtual identity of
homosexual–heterosexual diagnosticity and gender diagnosticity.
Journal of Personality, 68, 899–926.
Lippa, R. A. (2005a). Gender, nature, and nurture (2nd ed.). Mahwah,
NJ: Lawrence Erlbaum.
Lippa, R. A. (2005b). Sexual orientation and personality. Annual Reviewof Sex Research, 16, 119–153.
Lippa, R. A. (2006, April). Childhood gender nonconformity and adultpersonality in heterosexual and homosexual men and women.Paper presented at the Gender Development Conference, San
Francisco, CA.
Maccoby, E. E. (1998). The two sexes: Growing up apart, comingtogether. Cambridge, MA: Belknap Press.
Moore, M. M. (1985). Nonverbal courtship patterns in women: Context
and consequences. Ethology & Sociobiology, 6, 237–247.
Moore, M. M. (2002). Courtship communication and perception.
Perceptual and Motor Skills, 94, 97–105.
Nicholas, C. L. (2004). Gaydar: Eye-gaze as identity recognition among
gay men and lesbians. Sexuality & Culture, 8, 60–86.
Pierrehumbert, J. B., Bent, T., Munson, B., Bradlow, A. R., & Bailey, J.
M. (2004). The influence of sexual orientation on vowel produc-
tion. Journal of the Acoustical Society of America, 116, 1905–
1908.
Rahman, Q., & Wilson, G. D. (2003). Born gay? The psychobiology of
human sexual orientation. Personality and Individual Differences,34, 1337–1382.
Remafedi, G., French, S., Story, M., Resnick, M. D., & Blum, R. (1998).
The relationship between suicide risk and sexual orientation:
Results of a population-based study. American Journal of PublicHealth, 88, 57–60.
Rieger, G., Linsenmeier, J. A. W., Gygax, L., & Bailey, J. M. (2008).
Sexual orientation and childhood gender nonconformity: Evidence
from home videos. Developmental Psychology, 44, 46–58.
Saghir, M. T., & Robins, E. (1973). Male and female homosexuality: Acomprehensive investigation. Baltimore: Williams & Wilkins.
Skidmore, C. W., Linsenmeier, J. A. W., & Bailey, J. M. (2006). Gender
nonconformity and psychological distress in lesbians and gay men.
Archives of Sexual Behavior, 35, 685–697.
Smith, T. E., & Leaper, C. (2006). Self-perceived gender typicality and
the peer context during adolescence. Journal of Research onAdolescence, 16, 91–103.
Smyth, R., Jacobs, G., & Rogers, H. (2003). Male voices and perceived
sexual orientation: An experimental and theoretical approach.
Language in Society, 32, 329–350.
Spence, J. T., & Buckner, C. (1995). Masculinity and femininity:
Defining the undefinable. In P. J. Kalbfleisch & M. J. Cody (Eds.),
Gender, power, and communication in human relationships (pp.
105–138). Hillsdale, NJ: Lawrence Erlbaum Associates.
Spence, J. T., & Helmreich, R. (1978). Masculinity and femininity: Theirpsychological dimensions, correlates and antecedents. Austin:
University of Texas Press.
Sylva, D., Sell, L., & Bailey, J. M. (2007). An audible minority:Homosexuality and speech. Unpublished manuscript.
Tauber, M. A. (1979). Sex differences in parent–child interaction styles
during a free-play session. Child Development, 50, 981–988.
Weinrich, J. D., Grant, I., Jacobson, D. L., Robinson, S. R., McCutchan,
J. A., & HNRC. (1992). Effects of recalled childhood gender
nonconformity on adult genitoerotic role and AIDS exposure.
Archives of Sexual Behavior, 21, 559–585.
Wilson, G. D., & Rahman, Q. (2005). Born gay? The psychobiology ofsex orientation. London: Peter Owen.
Zar, J. H. (1999). Biostatistical analysis (4th ed.). Englewood Cliffs, NJ:
Prentice-Hall.
Zucker, K. J., Wilson-Smith, D. N., Kurita, J. A., & Stern, A. (1995).
Children’s appraisals of sex-typed behavior in their peers. SexRoles, 33, 703–725.
140 Arch Sex Behav (2010) 39:124–140
123