What does psychological research say about how accurately people can infer someone's sexual orientation from behavior, a

The short answer

Psychological research gives a genuinely mixed answer. Most studies find that people can guess sexual orientation from faces, voices, or body movement a bit better than chance, but “better than chance” is a low bar [64][70][71]. Average accuracy in gaydar-style studies tends to land around 62–64%, which is above a coin flip but far from reliable [70][71]. There is also a major real-world catch: because only a small percentage of the population is gay, simply guessing “straight” would often be correct even with no information at all [40][55][56].

Faces and photos

• In one large facial-image study, human judges correctly identified gay versus straight men 61% of the time and women 54% of the time from a single photo, just modestly above chance [1].
• Some earlier face-perception research found above-chance judgments from very little information, including 50-millisecond exposures, isolated facial features, one-second videos, stills, and brief “thin slices” of behavior [21][22][26][29]. People also seem able to do this from brief audio and silent video [65].
• However, face-based “gaydar” is heavily contested. A 2008 study seemed to show people could accurately judge orientation from facial photos, but later researchers argued that the effect was driven by photo-quality differences: when photo quality was controlled, participants could no longer tell who was gay and straight [52][53][54]. One line of research concludes that sexual orientation is not actually visible from the face and that face-based gaydar comes from a third-variable confound [72].
• Face judgments also depend on who is judging. In one study, only gay male raters accurately judged male faces from photos, while female raters made stereotype-based misjudgments [14]. The same study found that gay men’s faces were rated as more masculine than straight men’s faces, which goes against the simple stereotype that gay men “look feminine” [19].
• At the same time, perceived gender atypicality seems to matter: faces rated as gender-atypical are judged gay or lesbian above chance, composites of gay men’s faces tend to be rated as more feminine, and lesbian faces tend to be rated as more masculine than straight women’s faces [30][31][32].
• Computer models can pull more information from faces than human judges do, but they are also far from perfect. One algorithm reached about 81% accuracy for men and 74% for women from a single photo, rising to 91% and 83% with five photos [2][4]. Another report found 71% accuracy for women from one photo and 83% from five [11]. Reviewer commentary notes that the machine-versus-human comparison was not a fair contest, because the machine trained on thousands of images while human judges had no practice at the task [8][76][77]. Some researchers suspect humans are simply bad at combining multiple appearance cues into probability estimates [9][78][79], while sexual orientation may influence grooming and self-presentation rather than fixed facial structure [9][80].
• Methodological caveats matter: the machine-learning facial study used self-selected dating-site photos of nearly 8,000 men and nearly 7,000 women, and it has not tied actual sexual orientation to fixed facial physiognomy [10][73][81]. One reviewer also noted that prediction models using only gender information detected gay men with 57% accuracy and gay women with 58% accuracy [5].

Voice

• Voice-based judgments can also beat chance, especially for male speakers. Listeners, including women and gay men, have judged male speakers’ orientation from their voices in line with self-reported orientation [12][13][33]. Older studies such as Gaudio’s found greater-than-chance accuracy from voice [44].
• Some studies find straight speakers are identified more accurately than gay speakers [45]. A 2020 study found above-chance accuracy from voice, with sexual-minority listeners slightly more accurate than straight listeners [46].
• Listeners seem to rely heavily on perceived masculinity and femininity. Feminine-sounding male voices are more likely to be judged gay, while lower-pitched voices are perceived as more masculine and more heterosexual [16][18]. Perceived masculinity explains a large share of voice-based judgments: about 59% of variability in judgments of male targets and 87% in judgments of female targets [34].
• But actual voice pitch is not a reliable marker. One study found no significant pitch difference between gay and straight male speakers [17], and other studies also failed to find pitch as a significant predictor [47][48]. The most promising and replicable vocal dimension appears to be vocal range or dynamism, not pitch [36][37].
• Voice-based accuracy does not extend cleanly to everyone. One study found that auditory gaydar was not accurate for female speakers at all [43], and linguistic studies often find no reliable vocal differences between lesbians and straight women [49].
• Bisexual speakers are especially hard to read from voice. One Australian study found bisexual men were usually miscategorized as straight [35], and another study found bisexual voices were perceived as the most masculine-sounding of all speakers [51].
• There is also evidence that “sounding queer” is not fixed: speakers can purposely exaggerate or tone down audible cues depending on the situation [50]. Research on whether LGB people have better auditory gaydar than straight people is still limited [61].

Body movement and other behavior

• A broad line of research looks at nonverbal “gendered” cues, including body shape, movement, face, and voice, and finds these cues can signal sexual orientation above chance [23][24][25].
• Gait is one specific example: cross-gendered hip sway in men and shoulder swagger in women lead to more accurate judgments that someone is gay or lesbian [27]. Related research on judging people from body motion suggests sex categorization itself relies partly on perceived gender from walk motions such as swaying hips [85].
• Accuracy is often tied to how gender-atypical a person seems: participants judge targets more accurately when they also rate the targets as less gender-typical on a masculine–feminine dimension [28]. A separate line of research specifically examines judging sexual orientation from body motion and body morphology [84].
• In direct comparisons, face cues appear more helpful than voice cues for accurate sexual-orientation judgments [38]. In one study, judgments made from faces and judgments made from voices did not significantly match each other, suggesting the two modalities give different impressions [20].

Why “better than chance” can still be wrong

• Lab results can be statistically significant without being practically useful. One review notes that better-than-chance accuracy in a study may not be a bar people would rely on in real life [39].
• Base rates are a big reason. Because only a small minority of the population is gay, guessing “straight” in the absence of information would produce highly accurate estimates in everyday settings [40][55][56]. Even an extreme stereotypic cue, such as assuming a man in a pink shirt is gay, can lead to being wrong most of the time because of base rates [56].
• Cognitive research shows that correct “gay” judgments are not effortless. Mouse-tracking studies find people hesitate most when correctly judging a man as gay, suggesting they have to overcome a strong default toward straight [41].
• Accuracy also shifts with the judge and the situation. Thinking carefully before making a judgment makes gaydar less accurate [66]. People with anti-gay attitudes tend to perform worse, while sexual minorities and people with more familiarity tend to perform better [67]. Women’s accuracy has been found to improve during ovulation [68].
• Gaydar also struggles with bisexuality. People cannot reliably distinguish bisexual people as a separate group, and when asked to rate sexuality on a spectrum, they give gay and bisexual targets similar ratings [69].

Is “gaydar” even a valid scientific idea?

• Not all researchers accept the construct. Some academic work says the validity of gaydar remains unresolved and that it is unclear what exactly is being measured [62]. A scholarly article even describes gaydar as the “supposed ability” to detect orientation from visual cues and criticizes decontextualized measurement [63].
• The label matters psychologically. In one set of studies, people told that gaydar is real stereotyped more often than people not told that, and the researchers argued that calling stereotyping “gaydar” legitimizes it [57][58]. This is not just academic: participants in one study shocked a person more when his sexuality was implied by a stereotypic cue than when it was stated explicitly [60]. The researchers concluded that believing in gaydar can encourage harmful stereotyping [59].
• One practical recommendation from a research review is to rely on explicit, non-gendered information instead, such as what people say about their sexual orientation or who they choose as romantic partners [42].
• Beyond everyday judgments, a meta-analysis of 39 clinical judgment studies examined how client sexual orientation affects judgment accuracy and bias in clinical settings [86].

Bottom line

If you want one summary: people are not completely guessing, but they are also not mind-readers. Voices, faces, and body movement provide weak, above-chance signals, and those signals are heavily tied to perceived masculinity and femininity [16][30][33][34]. The effects are modest, some findings have been challenged by later studies, and real-world base rates make “better than chance” less impressive than it sounds [39][40][53][70][72]. The safer scientific conclusion is that unspoken appearance-based “gaydar” behaves more like stereotyping than like a reliable sixth sense [53][57][58].