BIDAR: Can Listeners Detect if a Man Is Bisexual from His Voice Alone?

ABSTRACT

Previous research has identified a range of perceptual voice and speech features that differ between gay and straight men, enabling listeners to determine if a man is gay or straight at a rate better than chance from his voice alone. To date, no published studies have examined if bisexual men’s voices differ from gay and straight men’s voices with regard to perceived masculinity-femininity – nor whether listeners can identify a bisexual man based only on his voice. In the present study, we examined if listeners could identify bisexual men’s sexual identities from voice recordings. Seventy participants (N= 70) rated 60 voice recordings of a sample of 20 gay, 20 bisexual, and 20 straight Australian men on perceived sexual orientation and degree of masculinity-femininity. Participants could correctly categorize the sexual orientations of the gay and straight speakers at rates greater than chance, but bisexual men were only identified at chance. Bisexual voices were consistently misperceived as being the most exclusively female attracted, and, contrary to expectations, were perceived as the most masculine sounding of all the speakers. Together, these findings suggest that while the voices of bisexual men in our sample were perceived as more masculine and female attracted, listeners do not associate this impression with bisexuality, and thus cannot identify bisexual men from their voices. Consequently, while bisexual men appear to be at lower risk of facing voice-based identification and discrimination than gay men, they may be often misperceived as being straight.

While societal representations of lesbian and gay individuals have been on the rise in Western cultures in the last few decades, bisexuals have arguably not enjoyed the same level of recognition. Legal theorist Yoshino (2000, p. 388) coined the term “bisexual erasure” to describe the ways in which bisexuality is rendered invisible by a pervasive tendency to view sexual orientation (and people) as either gay or straight, and identities outside of this binary as illegitimate or a phase. This erasure has arguably led to neglecting bisexuals in scientific research as well (Valentova et al. 2022). Erasure represents a pressing dilemma for bisexuals day-to-day, with bisexuals reporting less connection to other sexual minority people than their gay and lesbian peers, and also greater concealment of their sexual identity (Balsam & Mohr, 2007). Concerningly, evidence suggests that bisexuals face double-discrimination, with discrimination experienced from both heterosexual and gay/lesbian individuals (Mohr & Rochlen, 1999), suggesting that for bisexuals, recognition may be a double-edged sword. This discrimination appears particularly heightened toward bisexual men (Herek, 2002).

The present study examined whether bisexual men can be identified from their voices (i.e., whether vocal “bidar” exists) – akin to how gay men can be identified on their voice alone (i.e., commonly known as “gaydar”; e.g., Munson (2007)). If this is the case, voice may be an important target of discrimination on the one hand but may also counter the invisibility many bisexuals feel (if their bisexual identities can be apprehended by their voice alone, without explicit disclosure required). These findings may also shine light on whether bisexual male voices, like gay male voices, differ from straight voices in terms of their gender non-conformity – a question that to date has not been examined.

Verbal Communication Modalities as Social Cues of Sexual Identity

Verbal communication presents a range of modalities that convey meaning to the listener. The pitch, clarity and volume of the voice, the intonation pattern, articulatory pronunciation, and word selection, are all cues to a range of physical and/or sociocultural features of the speaker that can be implicitly decoded by the listener. For example, from the pitch of the voice listeners can decode information relating to the speakers age and gender. Studies show that listeners experience speakers with lower pitch in the respective male or female range, as more masculine and more dominant (Borkowska & Pawlowski, 2011; Puts et al., 2006). Although examining how speakers deliberately manipulate vocal cues to communicate their sexual identity (e.g., a gay male consciously deploying exaggerated gay speech stereotypes to communicate his sexuality in an interaction) – in the present context, we are interested in how perceivers/listeners use vocal cues to assess the sexual orientation of a male target.

Previous research has identified a range of speech and voice characteristics that are correlated with perceptions of gay versus straight sexual orientation in males, regardless of self-identified sexual orientation of the speakers. In studies of US (English), Italian, and German-speaking samples, higher pitch, wider pitch range, longer vowels, expanded vowel space, and more negatively skewed /s/ production has been observed (Mack & Munson, 2012; Munson, 2007; Sulpizio et al., 2015). In English-speaking samples, precise pronunciation has also been shown to be a clear speech stereotype of gay male speech (Panfili, 2011; Piccolo, 2008) whilst higher F1 and F2 values increase the likelihood that a male speaker will be perceived as sounding gay (Munson et al, 2006; Sulpizio et al., 2015) in US, Italian, and German samples. Preliminary experiments examining the language specificity of speech cues suggest that German natives also associate higher F2 with perceptions of homosexuality, and while a similar perception has been observed in Italian speakers, it has been more difficult to reproduce (Sulpizio et al., 2015).

Theorists suggest that men’s voice and speech characteristics may be influenced by social and biological factors associated with their sexual orientation, which guide listeners to infer their sexual identity. Linville (1998) proposed that gay men unconsciously acquired particular vocal conventions of the gay male community, which signaled their identification to others. In addition to associating with the gay community, Zwicky (1997) suggested that gay men are motivated to avoid the behaviors of straight individuals that they do not identify with. This, in turn, makes people interpret them as feminine for deviating from rigidly enforced masculine norms.

Alternatively, Pierrehumbert et al. (2004) suggested that biological factors, such as hypomasculinization of the male brain, thought to underlie male homosexuality, also affect language acquisition, resulting in feminized speech patterns along with other sex atypical indicators (e.g., an interest in female typical occupations). If these biological influencers work in a dose-dependent manner, then one would expect that they exert a weaker effect on the speech patterns of bisexual men than gay men, resulting in a distinct speech style intermediate between gay and straight men (Rieger et al., 2020). Regardless of etiology, both theories concluded that sexual orientation detection may be facilitated by recognizing gender-atypical/feminized male vocal patterns. Given bisexual men have been shown to demonstrate self-reported and observer reported masculinity-femininity intermediate between gay and straight men (Lippa, 2020; Rieger et al., 2020), it seems reasonable to hypothesize that bisexual men would sound more feminine than straight men, but less feminine than gay men.

Whether or not people can identify a man’s bisexual identity from his voice has important social ramifications. On the one hand, being perceived as bisexual based on one’s voice could make one more vulnerable to discrimination, while on the other hand, being misconstrued as straight or gay based on one’s voice, might contribute to a sense of erasure and alienation. In gay men, Fasoli et al. (2017) demonstrated that after listening to the voices of gay and straight men, participants were more likely to hire the straight man, thought he deserved a higher salary, and preferred to converse with him over the gay man. Likewise, bisexual men may also be exposed to prejudice if their voices are judged as indicators of their sexual identity, given bisexual men are found to engender the most negative attitudes of all sexual minority groups (Herek, 2002).

Naïve Listeners Have Distinct Impressions of Straight and Gay Men’s Voices

Previous studies investigating reactions to gay and straight individuals’ voices have demonstrated that naïve listeners can distinguish gay and straight men from vocal cues above chance. In the first experiment on voice-based judgments of men’s sexual orientation, 13 participants rated voice recordings of 16 gay and straight men on 7-point response scales ranging from extremely straight to extremely gay sounding with neutral at the midpoint (Gaudio, 1994). All but one target was consistently rated by participants on the same end of the sexual continuum as their sexual identification, suggesting that listeners could accurately recognize the sexual orientation of the speakers. Twenty-five participants of a later study could correctly identify five gay and four straight targets at a rate of 68.4% and 93.5% of the time, respectively, when forced to make dichotomous judgments of the speakers (based on vocal stimuli alone) as either gay or straight (Linville, 1998).

Studies examining accuracy of voice-based judgments of gay and straight men’s voices in languages other than English have found mixed evidence for accurate gay classifications. Listeners were unable to correctly identify the gay men’s voices across multiple evaluations in a study investigating whether gay and straight men’s voices could be accurately distinguished by native Italian and German speakers (Sulpizio et al., 2015). Specifically, while straight participants were correctly identified at rates above chance during the forced-choice task (i.e., when voices were rated dichotomously as either “gay” or “straight”) in both languages, gay participants were not. However, across both languages, gay speakers were rated as more gay than straight speakers, on a 7-point response scale from completely straight to completely gay. Moreover, the sexual orientation ratings given by the listeners were positively correlated with the speaker’s self-ratings of sexual orientation. The researchers, however, were reluctant to consider this as an accuracy indicator, because mean ratings of gay speakers fell on the straight pole of the spectrum (albeit gay speakers were rated more same-sex attracted than the straight speakers). As raters were not informed of the proportion of gay versus straight voices in the sample, they may have assumed a low base rate of gay males leading to this pattern of responding. Similarly, a subsequent experiment on listener perceptions of 18 straight and 17 gay native Czech speakers demonstrated that self-ratings on a 7-point response scale from typical straight to typical gay were positively correlated with ratings from 80 Czech and American listeners (Valentova et al., 2013). While the listeners also rated gay targets as significantly more homosexual-typical than straight targets, across both studies, the mean ratings of gay and straight targets were closer to the midpoint of the sexual continuum (i.e., bisexual range), rather than the poles (i.e., exclusive heterosexual or homosexual points).

Consistent with the theory that impressions of a voice’s gender non-conformity inform sexual orientation judgments, judgment ratings of vocal Femininity-Masculinity (F-M) have shown consistent associations with sexual orientation ratings across multiple studies. In addition to rating the sexual orientation of targets on a 7-point scale (straight to gay), Gaudio’s (1994) participants rated each voice on a 7-point response scale from effeminate to masculine. The ratings were almost perfectly correlated with sexual orientation judgments, with gay targets being rated significantly more effeminate than straights. The same association was demonstrated in a subsequent study after utilizing two independent samples of raters for sexual orientation (rated dichotomously as “gay” or “straight”) and femininity-masculinity (rated dichotomously as “masculine” or “feminine”) judgments (Smyth et al., 2003).

Lastly, past studies have found that observer ratings of femininity-masculinity (F-M) are positively associated with self-ratings of F-M – suggesting that bisexual individuals’ unique self-ratings of F-M might be detectable among observers and be a clue to their bisexual identity. Evidence supporting the association between self and other rated F-M includes vocal studies such as Weirich et al. (2018), wherein self-ratings on the German version of the Personal Attributes Questionnaire (GEPAQ-F+) (assessing 7 positive attributes associated with females) and the Traditional Masculinity-Femininity scale (which measures gender-role self-concept) were found to correlate positively with listeners’ judgments of F-M. In Weirich et al. (2018), vocal F-M was measured via a forced choice paradigm, in which all combinations of voices were paired, and listeners were asked to rate which of the pair sounded “more masculine?” on a scale from 0 to 3 (“0” neither voice more masculine, to, “3” one voice much more masculine).

So how masculine or feminine will bisexual voices be relative to gay and straight voices? The best guide comes from the existing self-report and observer report literature regarding the relative masculinity-femininity of bisexual men compared to gay and straight men. Self-report studies have found mixed findings in this regard. A few self-report studies have found that bisexual men do indeed rate themselves more masculine than gay men but less masculine than straight men. For instance, Cohen (2002) found that bisexual men retrospectively reported levels of “childhood sex-atypicality,” based on a novel 24-item scale, intermediate between gay and heterosexual men. Likewise, a meta-analysis by Lippa (2005), and a subsequent empirical study by Lippa (2020), found self-rated masculinity-femininity (based on a 4-item scale assessing felt and perceived masculinity and femininity) was intermediate gay and straight men. This is consistent with how gay/lesbian and heterosexual men and women rated the relative “masculinity” and “femininity” of bisexuals on separate single item measures (Burke & LaFrance, 2016). However, other studies have failed to find an intermediate pattern of masculinity-femininity in bisexuals. In one study by Lippa (2008) bisexual men reported themselves significantly more feminine than gay men on a single item continuum measure of self-rated MF (“How masculine or feminine do you consider yourself to be mentally, compared to others of your sex and age?”), but intermediate between gay and straight men on a 10-item measure of masculine-feminine occupational preferences. Contrasting with this, Kahn and Halpern (2019) found bisexual men were more similar to heterosexual than gay men on gender-related traits, as measured by 25 items from the National Longitudinal Study of Adolescent to Adult Health found to be predictive of gender.

To our knowledge, only two studies of observer ratings of speech or behavior in bisexual men exist to date. In the first, Rieger et al. (2020) utilized archived recordings to create brief (i.e., 10 second) audio-visual stimuli of male and female targets from across the Kinsey continuum (from exclusively heterosexual to exclusively homosexual) sitting in a chair responding to a neutral question about the weather. The experimenters utilized this stimulus to capture gender non-conformity as expressed through movement, posture, appearance, and speech. Raters were n~ = 100 heterosexual male, heterosexual female, non-heterosexual male, and non-heterosexual female students naïve to the sexual orientation of targets. This study found that bisexual men were generally intermediate between gay and straight men based on objective ratings of masculinity-femininity (rated on 7-point Likert scales of “more masculine” [1] through to “more feminine” [7]), albeit shifted closer to gay men than straight men. Although this is the most comprehensive study to date on the gender non-conformity of bisexual men – the assessment of gender non-conformity in this study was based on a gestalt judgment of body movements, posture, appearance, and speech (precluding conclusions regarding the gender non-conformity of the bisexual voice alone).

The second is the only published study (to our knowledge) that has examined the bisexual male voice specifically. Kachel et al. (2018) examined the objective and perceptual characteristics of artificially generated gay/lesbian, bisexual, and heterosexual male and female voices. In this study, vocal stimuli were created by averaging together several voices found to be the most prototypical for each sexual orientation category (Kachel et al., 2018). That is, from a larger pool of 111 speakers (57 female) from across the sexual orientation continuum, Kachel et al. (2018) selected 30 voices based on self-rated Kinsey Score (the 5 most prototypical gay/lesbian voices based on Kinsey scores (e.g., Kinsey 6ʹs), bisexual-voices-based Kinsey scores (Kinsey 3ʹs), and heterosexual voices based on scores (Kinsey 0ʹs) in both men and women) and the five most gay/lesbian, bisexual, and straight sounding speakers based on perceptual ratings by naïve raters. Voice averaging was then conducted within each group to create a single-voice-stimuli for each group (i.e., one voice stimuli for gay men, lesbian women, bisexual men, bisexual women, straight men, and straight women). Listeners rated each of these averaged voice samples on a 7-point Kinsey scale (most gay sounding [1] to most straight sounding [7]). Results demonstrated that bisexual male and female voices were perceived as straighter than gay/lesbian male and female voice averages, and less straight than heterosexual male and female voices.

However, Kachel et al.'s (2018), study although relevant to the present question, did not answer whether bisexual men can be accurately identified as bisexual based on their voice alone. First, the ecological validity of the averaging approach is low as it purposefully removes unique, individualized voice elements in creating artificial new acoustic stimulus – and thus, one’s ability to apply the findings to the real-world context is substantially diminished. Second, Kachel et al. (2018) did not examine whether listeners could accurately detect the sexual orientation of speakers, including bisexual men. Rather, they examined the perceived straightness/gayness of gay/bisexual/straight voices relative to one another. Perusal of the means indicates that lesbian/gay, bisexual, and straight voices were all rated in the bisexual range, and thus it is unknown if listeners can accurately identify bisexual voices, from gay and straight voices. Next, Kachel et al. (2018) did not examine the most likely speech element differentiating bisexual from gay versus straight voices – distinct levels of vocal masculinity-femininity. It is also notable that the study was based on German speakers and listeners – and thus whether their findings generalize to an English-speaking context is unknown. To this point, no published study has provided raw vocal stimuli of a group of gay, bisexual, and heterosexual men, and assessed whether naïve raters could accurately discern who was bisexual based on his voice alone. Moreover, no study has examined the perceived masculinity-femininity of bisexual male voices relative to gay and straight male voices. Finally, bisexuals in Kachel et al. (2018) study were identified based on their response to a 7-point Kinsey-type scale requesting participants to “self-identify their sexual orientation.” In the present study, bisexual men were categorized based on their endorsement of the discrete sexual identity options “gay,” “bisexual,” or “straight.” As such, bisexual men in our study unambiguously conceived of themselves as bisexual, unlike in the aforementioned study.

The Present Study

To date, no published study has investigated whether listeners can accurately identify whether a man identifies as bisexual from voice alone. This study addressed this gap within the field of voice-based identity inferences and examined whether distinct bisexual vocal patterns may be linked to perceived masculinity-femininity. To this end, we hypothesized the following:

1. Listeners will correctly categorize the sexual identities of gay, bisexual, and straight targets based on their voice alone (in English), at above chance rates. Previous studies have shown that the voices of gay and straight men can be distinguished at above chance rates (Gaudio, 1994; Linville, 1998). Although no prior studies have examined how accurately listeners can identify the voice recordings of bisexual men, theories of voice-based identification suggest that their voices should have a distinct signature recognizable by others.

2. On a continuum from exclusive male attraction to exclusive female attraction, bisexual targets will be rated more female attracted than gay targets, and more male attracted than straight targets based on their voices alone. While multiple studies found that listeners could distinguish gay and straight male speakers on a continuous scale, they have not always used consistent labels for these scales (Gaudio, 1994; Sulpizio et al., 2020; Valentova et al., 2011). In the present study, we reference sex preference in line with a Kinsey-type scale (Kinsey et al., 1948) i.e., “exclusively male attracted” to “exclusively female attracted” on a 9-point scale.

3. Bisexual men were expected to be rated as more feminine than straight men, but less feminine than gay men. This hypothesis is based on findings that voice-based perceptions of femininity-masculinity are associated with self-ratings of F-M (Weirich et al., 2018), and based on self- and other-ratings of bisexual men’s relative masculinity-femininity (Lippa, 2020; Rieger et al., 2020).

Method

Participants

Raters

Raters were recruited from The University of Sydney student community with approval of the Human Research Ethics Committee (HREC 2019/287). Sixty-three participants enrolled in either first or second-year psychology units volunteered and were rewarded with course credit for their participation. Thirteen additional male students were recruited from the university community and reimbursed $15 for their time. All listeners had identified as being heterosexual/straight (based on a mandatory pre-screen which involved the category options of “gay,” “bisexual,” straight/heterosexual,” “pansexual,” “queer,” and “other [write-in]”), above the age of 18, and as having lived in Australia for more than five years (to ensure familiarity with the Australian accent) on a mandatory pre-screen. We decided to limit our listener sample to straight/heterosexual identified individuals as we did not suspect we would have a sufficient number of non-heterosexuals to allow examination of how sexual orientation may moderate one’s ability to detect sexual orientation via voice cues. As such, we opted for a more homogeneous sample. Data from six participants were removed due to failure to follow directives. The final sample consisted of 70 usable responses (see Table 1 for relevant participant characteristics).

Table 1. Summary of participant characteristics.

		Target speakers				Listeners/Raters
	Sexual orientation	Gay (n = 20)	Bisexual (n = 20)	Straight (n = 20)	Sex	Female (n = 42)	Male (n = 28)
Age
		27.75 (9.60)	28.30 (10.14)	27.9 (9.36)		20.38 (5.7)	20.35 (1.96)
Ethnicity
	Caucasian	17	16	17		29	18
	Asian	1	2	0		8	7
	Hispanic	1	0	0		0	0
	Middle Eastern	0	0	0		1	1
	Other/Mixed	0	2	3		4	1
Religion
	Christian	5	5	5		15	2
	Judaism	5	0	1		1	0
	Buddhism	0	0	1		0	1
	Hinduism	0	0	0		2	1
	Islam	0	0	0		0	2
	Atheist/No Religion	13	12	13		22	20
Sexual orientation (9-point)	Other	28.80 (.52)	35.30 (.1.93)	01.00 (.00)		2 n/a	2 n/a

Materials

Voice Samples

Voice samples comprised a 10–15 seconds’ recording, taken via smartphone, of gay, bisexual, and heterosexual identified males speaking the first two lines of the Australian national anthem: “Australians all let us rejoice, For we are one and free; We’ve golden soil and wealth for toil; Our home is girt by sea.” We chose the Australian National anthem because it would be widely known by participants and could therefore be recited by participants from memory, without reading it from their phones. This was important because it was not guaranteed that participants would be able to view written text on their phone while recording their voice using the voice/video recorder app.

The voice sample questionnaire was positioned at the end of an online survey examining sexual orientation, gender non-conformity, and mental health of Australian gay and bisexual men (recruited via Grindr^TM) and heterosexual men (recruited via Qualtrics Panels^TM). To categorise speakers by sexual orientation, participants were asked to indicate “what sexual orientation label do you most identify with at present?” with options being “gay”, “bisexual”, “queer”, “pansexual”, “questioning”, “straight”, “other [write-in]. Participants were instructed to recite the abovementioned two lines of the Australian anthem written on the screen and instructed to “1) please speak the two lines (do not sing), 2) speak at a normal pace (do not rush), and 3) try to minimise background noise where possible.” They were then given instructions about how to upload their voice sample using the “File Upload” option in Qualtrics surveys^TM.

Silence was removed from the voice samples, and their volumes were normalized in Ableton Live 9.7. A between-subjects ANOVA comparing the mean age of the chosen twenty gay- (M = 27.75, SD = 9.60), bisexual- (M = 28.30, SD = 10.14), and straight- (M = 28.10, SD = 9.21) identified voice samples revealed no significant age differences between the groups F (2, 57) = 0.017, p = .984.

Demographics

Listeners were requested to report their ethnicity, age, religion, political orientation, present gender identity, length of time living in Australia, birth sex, and sexual orientation.

Femininity-Masculinity

Listeners recorded their ratings of the masculinity-femininity of the voices on a visual analogue scale (VAS: 0 = feminine, 100 = masculine). This scale was based on previous studies (Gaudio, 1994; Lippa, 2020).

Perceived Sexual Orientation Category

Listeners gave their impressions of voice’s sexual orientation on a 3-point nominal response scale (1 = gay, 2 = bisexual, 3 = straight).

Perceived Continuous Sexual Orientation Rating

Listeners gave Kinsey-like ratings of the voices on a 9-point response scale (1 = exclusively attracted to men, 2 = almost exclusively attracted to men, 3 = mostly attracted to men, 4 = somewhat more attracted to men, 5 = attracted to men and women equally, 6 = somewhat more attracted to women, 7 = mostly attracted to women, 8 = almost exclusively attracted to women, 9 = exclusively attracted to men). The scale was adapted from Savin-Williams (2014) who found evidence that a 9-point Kinsey-like scale may be a more sensitive measure of sexual continuity. The labels were slightly altered to ensure that they did not influence categorical judgments.

Procedure

The experiment was conducted in a standard experimental room in the School of Psychology. Listeners were given a pair of Philips SHP2000 headphones and were briefed on the task and expected duration of the experiment. The survey was hosted on Qualtrics^TM. Following the provision of participant information and the completion of an online consent form, consenting listeners undertook a calibration phase. In the calibration phase, listeners were presented with five voices (selected from a pilot study involving 10 independent raters, three female and seven male) that represented a near full range of voices from extremely feminine (0) to equally feminine and masculine (50) to extremely masculine (100).

Once the calibration phase was complete, the experiment proper commenced. Voice sample ratings were conducted in two rounds. During the first round of ratings, the 60 voice samples were presented in random order to the participants who proceeded to rate the perceived femininity-masculinity of each voice on the VAS scale.

In the second round of ratings, the same 60 voices were randomly played one at a time (once each), to increase the likelihood that the listeners were unaware of their previous ratings and not tempted to match their sexual orientation ratings to their previous femininity-masculinity ratings. For each voice, listeners judged the sexual orientation on two questions. The first, a categorical item asked participants to judge if the speaker was gay, or bisexual, or straight. The second item, directly below, asked listeners to rate the speaker on a Kinsey-like scale from exclusively attracted to men (1), through equally attracted to men and women (5), to exclusively attracted to women (9). To counter low base-rate biasing observer ratings toward heterosexuality, prior to ratings voices in round two, listeners were informed that there were 20 gay, 20 bisexual, and 20 straight male voices in the study so that they should not assume that most of the voices were straight. After judging the voices, listeners completed the demographic questions.

Data Analysis Plan

Outliers were determined a priori to be values which fell > |3.29| standard deviations beyond the mean, as suggested by Tabachnick and Fidell (2001). The 3.29 SD is a common value used for outlier’s detection (e.g., Tabachnick & Fidell, 2001) based on the premise that only .05% of the individuals should report such extreme values on a variable that is considered to have an underlying normal distribution. With a sample such as ours, it should be statistically unlikely to have such a extreme true response. To investigate our first hypothesis, a series of chi-square goodness-of-fit tests comparing the observed frequencies of correct sexual orientation judgments to the expected frequency of correct identifications due to chance. Our remaining hypotheses were assessed using repeated measures ANOVAs, followed by planned Bonferroni-adjusted contrasts examining the proportion of correct identifications for each sexual orientation. Lastly, the continuous sexual attraction ratings and the femininity-masculinity ratings of each sexual orientation were assessed using ANOVAs.

Results

Data Preparation

Missing value analyses revealed that less than 0.001% of total data were missing. Following guidelines for handling missing values (Schlomer et al., 2010), pairwise deletion was implemented, meaning that neither mean substitution or imputation for missing data was conducted (rather a particular data point, if missing on a critical variable, would result in that case being omitted from the relevant analysis). To assess outliers within the data, we standardized scores and looked for values that fell beyond the > |3.29| standard deviation cut off suggested by Tabachnick and Fidell (2001); none was found. The normality of the dependent variables across all levels of our independent variables was demonstrated for all experiments by nonsignificant Shapiro-Wilk tests (ps > .05). Mauchly’s test of sphericity failed to reach significance (ps > .05) and sphericity was assumed, except where otherwise stated.

Can Sexual Orientation Identity Be Identified at Above Chance Rates?

With 70 listeners rating 20 voices of each sexual orientation, we expected that, at chance, 466.7 out of 1,400 ratings of each specific orientation would be correct. Of the 1,400 judgments of gay voices, 539 correctly identified the speaker’s sexual orientation, significantly exceeding the expected chance rate, χ²(1, N = 70) = 16.82, p< .001. Similarly, participants also correctly categorized the voices belonging to the straight men at greater-than-chance rates, χ²(1, N = 70) = 106.86, p< .001. Conversely, the number of bisexual male voices which were accurately judged as belonging to bisexual men did not differ from what we expected due to chance, χ²(1, N = 70) = 0.01, p= .985. (See Figure 1 and Table 2).

Figure 1. Illustration of the number of correctly identified gay, bisexual, and straight voices out of 1,400 judgments. The horizontal line indicates the frequency expected due to chance.

Table 2. Mean proportion (and standard deviations) of correct sexual orientation judgments by sexual orientation of stimulus and distribution of voice ratings.

Speaker sexual orientation	Gay (n = 20)	Bisexual (n = 20)	Straight (n = 20)
Proportion correctly identified	.385ab (.106)	.334ac (.102)	.464bc (.115)
Distribution of voice ratings
Rated gay	508	283	299
Rated bisexual	479	460	451
Rated straight	401	657	650

Different subscripts in a row indicate significant difference, p < .001.

Did Continuous Sexual Attraction Ratings Reflect Target Sexual Preferences?

To further examine how participants perceived the gender preferences of target voices, we employed a repeated measures ANOVA focusing on how listeners’ Kinsey-like ratings differed according to the sexual orientation of the speakers. The omnibus test revealed a significant main effect, such that the Kinsey-like ratings ascribed by the listeners changed according to the sexual orientation of the speaker, F(2, 138) = 139.675, p < .001, partial η² = .669. Planned contrasts were carried out utilizing the Bonferroni-corrected p value of .0167 to control the family-wise error rate at α ≤ .05. In support of our hypothesis, bisexual voices (M = 6.16, SD = 0.62) were rated significantly more female attracted than gay voices (M = 4.62, SD = 0.64) on the Kinsey scale, F(1, 69) = 227.804, 95%CI [1.29, 1.80], p < .001 partial η² = .768. Furthermore, straight voices (M= 5.9, SD = 0.67) were also rated significantly more female attracted than the gay voices F(1, 69) = 165.24, 95%CI [−1.53, −1.04], p < .001 η² = .705. Surprisingly, bisexual voices were perceived as being more female-attracted than the straight voices, F(1, 69) = 7.66, 95%CI [0.03, 0.49], p = .022, partial η² = .10. See, Figure 2.

Figure 2. Mean Kinsey ratings of gay, bisexual, and straight voices. Note due to the within-subjects nature of the design, the 95% CI should be interpreted with caution. See Franz & Loftus (2012) for review.

Were the Bisexual Voices Perceived as More Masculine or Feminine than Gay or Straight Voices?

Mauchly’s test of sphericity indicated that the assumption of sphericity was violated, χ²(2, N = 70) = 41.001, p < .001. Therefore, the Greenhouse-Geisser correction (ε = .76) was interpreted. The main effect of sexual orientation showed a statistically significant difference in mean masculinity ratings, F(1.53, 105.26) = 174.29. p < .001, η² = .716. Planned contrasts utilizing the Bonferroni-corrected p value of .0167 were performed to examine perceived masculinity differences between the gay, bisexual, and straight voices. Both bisexual voices (M = 70.00, SD = 7.86) and straight voices (M = 67.31, SD = 8.5) were rated as more masculine than gay voices (M = 59.50, SD = 11.23), F(1, 69) = 237.92, 95%CI [6.25, 9.375], p < .001, partial η² = .775, and F(1, 69) = 149.95, p < .001, 95%CI [−9.38, −6.25], partial η² = .685, respectively. Bisexual voices were judged significantly more masculine than straight voices, F(1, 69) = 46.98, p < .001, 95%CI [1.73, 3.65], partial η² = .405. See Figure 3.

Figure 3. Ratings of femininity-masculinity by sexual orientation of speaker. Note due to the within-subjects nature of the design, the 95% CI should be interpreted with caution. See Franz and Loftus (2012) for review.

Are All Bisexual Men in Our Sample More Masculine and Female-attracted Sounding than Straight Men or Just Some?

As past research suggests that bisexual men may be a heterogeneous group, with some more similar to gay men in their sexual arousal patterns, and others more similar to heterosexual men (Rieger et al., 2005; Slettevold et al., 2019), we decided to compare subgroups of bisexual men on key variables of interest. Four bisexual men reported being gynephilic-leaning (“exclusively attracted to women” [1] to “somewhat more attracted to women” [4]), seven equally gynephilic and androphilic (i.e., [5]), and eight being androphilic-leaning (“exclusively attracted to men” [6] to “somewhat more attracted to men” [9]). We then ran repeated measures ANOVAS on perceived Kinsey score ratings and perceived masculinity-femininity ratings, with the within subject factor being sexual orientation categories, including three subtypes of bisexuals: androphilic-learning bisexuals, equally andro/gynephilic bisexuals, gynephilic-learning bisexuals, as well as the original gay and straight identified men.

Kinsey Scores

Mauchly’s test of sphericity indicated that the assumption of sphericity was violated, χ²(9, N = 70) = 50.744, p < .001, for the repeated measures ANOVA on perceived Kinsey scores, so the Greenhouse-Geisser correction was interpreted (ε = .76). A main effect of sexual orientation was observed for perceived Kinsey score between the five groups, F(3.068, 211.673) = 57.901. p < .001, η² = .456. Bonferroni-corrected contrasts were performed to compare Kinsey ratings between the gay (M = 4.62, SD = .64), androphilic-leaning bisexual (M = 5.60, SD =.97), equally andro/gynephilic bisexual (M = 6.90, SD = .98), gynephilic-leaning bisexual (M = 6.01, SD = 1.27), and straight voices (M = 5.90, SD = .67). These contrasts revealed that androphilic-leaning (p < .001, 95% C.I. = [−1.35, −.62]), equally andro/gynephilic (p < .001, 95% C.I. = [−2.71, −1.87]), and gynephilic-leaning bisexuals (p < .001, 95% C.I. = [−1.88, −.91]) were all perceived as more female attracted than the gay voices (see Figure 4). Equally, andro-gynephilic bisexuals were rated as more female attracted than straight voices (p < .001, 95% C.I. = [−1.329, −.679]); however, androphilic-leaning (p = .299, C.I. = [−.09, .68]) and gynephilic-leaning (p = 1.00, C.I. = [−.626, .398]) bisexual voices were not rated significantly more female attracted than straight voices. Finally, equally andro/gynephilic bisexuals were rated as more female attracted than androphilic (p < .001, C.I. = [.81, 1.78]) and gynephilic-leaning bisexuals (p < .001, C.I. = [.355, 1.425]), who did not differ from one another (p = .317, C.I. = [−1.33, .954]). See Figure 4.

Figure 4. Mean Kinsey rating of gay and straight voices, as well as the three subtypes of bisexual voices. Note due to the within-subjects nature of the design, the 95% CI should be interpreted with caution. See Franz & Loftus (2012) for review.

Masculinity-Femininity

For repeated measures ANOVAs on perceived masculinity-femininity, Mauchly’s test of sphericity indicated that the assumption of sphericity was violated, χ²(9, N = 70) = 25.335, p < .001, so the Greenhouse-Geisser correction was interpreted (ε = .76). As in the previous analysis, a main effect of sexual orientation was observed for masculinity ratings between the five groups, F(3.047, 210.219) = 103.633. p < .001, η² = .600. Bonferroni-corrected contrasts were performed to compare masculinity ratings between the gay (M = 59.50, SD = 11.23), androphilic-leaning bisexual (M = 67.92, SD = 8.86), equally andro/gynephilic bisexual (M = 73.04, SD = 7.66), gynephilic-leaning bisexual (M = 69.06, SD = 9.13), and straight voices (M = 67.31, SD = 8.50). These contrasts revealed that androphilic-leaning (p < .001, 95% C.I. = [−10.34, −6.51]), equally andro/gynephilic (p < .001, 95% C.I. = [−16.26, −10.81]), and gynephilic-leaning bisexuals (p < .001, 95% C.I. = [−11.71, −7.41]) were all perceived as more masculine than the gay voices (see Figure 5). Equally, andro/gynephilic bisexuals were rated as more masculine than straight voices (p < .001, 95% C.I. = [−7.443, −4.008]); however, androphilic-leaning (p = .211, C.I. = [−1.583, .356]) and gynephilic-leaning (p = .057, C.I. = [−3.532, .030]) bisexual voices were not rated significantly more masculine than straight voices. Finally, equally andro/gyne-philic bisexuals were rated as more masculine than androphilic (p < .001, C.I. = [3.085, 7.139]) and gynephilic-leaning bisexuals (p < .001, C.I. = [1.832, 6.115]), who did not differ from one another (p = 1.00, C.I. = [−2.002, .975]). See, Figure 5.

Figure 5. Ratings of femininity-masculinity by sexual orientation of speaker (inclusive of gay, straight, and three subtypes of bisexual voices). Note due to the within-subjects nature of the design, the 95% CI should be interpreted with caution. See Franz and Loftus (2012) for review.

Discussion

Although several studies have investigated whether gay and straight men’s voices can be identified above chance (Gaudio, 1994; Linville, 1998; Sulpizio et al., 2015; Valentova et al., 2013) and one other has examined objective and perceptual correlates of artificially generated bisexual male voices (Kachel et al., 2018), no previous published study has investigated if bisexual male voices can be identified at a rate above chance, nor have previous studies examined the perceived femininity-masculinity of bisexual male voices (related to gay and straight male voices). Current participants rated the voices of gay, bisexual, and straight men on multiple indices to investigate how listeners judged the voices of men of different sexual identities. Whereas the sexual identities of gay and straight men were identified at rates greater than chance, bisexual men were not. Additionally, each sexual orientation was considered to have a distinct sexual partner preference, but bisexual men were perceived as being more exclusively female attracted than both gay and straight men. Furthermore, straight men were rated more masculine than gay men, but contrary to our hypothesis that bisexual men would be intermediate between gay and straight men on masculinity, they were judged more masculine than both.

Bisexual Men Were Not Identified

While these observations corroborate the results of previous studies (Gaudio, 1994; Linville, 1998; Valentova et al., 2013) that demonstrated that listeners can distinguish the voices of straight and gay men, they suggest this does not extend to the voices of bisexual men. Rather, it seems that naïve interlocutors can do no better than to guess whether a man identifies as bisexual based on his voice. There are many possibilities as to why bisexual men are difficult to recognize via their voices. One might be that they actively mask their true speech patterns more so than gay or straight men. Previous research has suggested that gay men modulate their voices to conceal their sexual identity according to how comfortable they feel revealing their sexual orientation to their audience (Daniele et al. 2020). Given the evidence that bisexual individuals report feeling less connected to the LGBTQIA+ community and are less open to revealing their sexual orientation than their gay peers (Paul et al., 2014), they may consciously modify their vocal patterns so as to not sound same-sex attracted. While speakers were recruited from a same-sex dating app, Grindr does not require users to disclose their sexual orientation on their profiles and facilitates discrete interactions through label and search functions, thereby allowing bisexual men to keep their sexual identity hidden (Wilson et al., 2020). Another possibility is that distinct bisexual vocal cues do exist, but they have not yet been enregistered by the wider public, including by speakers and listeners in our study, and thus were not perceived (Johnstone, 2016). For ways of speaking to become associated with the communication of, and perception of, certain social identities, such as bisexual identities in men, a given community must develop norms of speech over time, and the broader culture must have sufficient exposure to the group in question to enregister these vocal cues. Bisexual men continue to be the least prevalent sexual minority demographic, comprising less than 1% of males (Wilson et al., 2020), the most negatively viewed by heterosexual and gay/lesbian individuals alike (Herek, 2002), and consequently the least likely to disclose their sexual identity or affiliate with LGBTQIA+ communities (Mohr et al., 2017). Such cultural conditions may contribute to a failure to enregister bisexual male speech in the wider society, and a failure to identify bisexual voices in the present study.

Bisexual Men Were Considered the Most Exclusively Female Attracted

The prediction that gay and straight men would be distinguished as male and female attracted, respectively, with bisexuals falling between the two, was partially supported. Whereas straight men were rated as being more exclusively female attracted than gay men, bisexual men, at least those equally attracted to men and women, were judged more exclusively female attracted than both (in comparison, androphilic and gynephilic-leaning bisexuals were rated as just as female attracted as straight speakers). This finding supports previous studies which also found that straight targets are appraised to be more exclusively female attracted than gay targets (Gaudio, 1994; Sulpizio et al., 2015; Valentova et al., 2013). Based on Kinsey ratings of voices, our data suggest that bisexual men are most typically rated as straight. Consequently, it seems that in voice only conversations (e.g., a phone call) bisexual men are likely to be perceived as straight men rather than bisexual or gay men. Despite some bisexual targets being considered the most exclusively female-attracted, all the group averages fell closer to the midpoint of the Kinsey spectrum than the gender exclusive poles. This tendency toward the midpoint of the Kinsey scale is consistent with previous studies (Sulpizio et al., 2015; Valentova et al., 2013) and may suggest that there is a reporting bias when participants rate others when using this scale or that the variability in judgments may be a good element for further investigations (e.g., are there individuals who are consistently good at identifying sexual orientations from voice alone?).

Bisexual Men Sounded the Most Masculine

Along with the finding that bisexual men were judged to be the most exclusively female attracted, they were also judged to be the most masculine. These findings provide further support that perceptions of male vocal masculinity and heterosexuality go hand-in-hand regardless of the speaker’s sexual orientation. Contrary to expectations, the voices of bisexual targets were not only rated as more masculine sounding than those of gay men, but either equal to or more masculine than those of straight men. A number of factors may explain this misjudgment. Following Zwicky’s (1997) social theory of sexuality speech signaling, bisexual men may engage in excessively heteronormative behavior due to identity confusion concerning their sexuality, self-stigma, or lack of connection to the queer community (Balsam & Mohr, 2007; Herek et al., 2009). Alternatively, Grindr users’ preferences may have motivated bisexual speakers to emphasize the masculinity of their voices. Multiple lab and field studies demonstrate a demand bias for masculine over feminine partners amongst men seeking men (Bailey et al., 1997; Cascalheira & Smith, 2020; Reynolds, 2015; Sánchez & Vilain, 2012; Sarson, 2020; Zheng & Zheng, 2016) and gay men deem voices they consider more masculine to also be more attractive (Valentová et al., 2013). That said, any motivation to appear more masculine and less feminine would apply equally to our gay speakers who were judged the most feminine sounding. Perhaps bisexual men, who may be more practiced in masculine presentation, are more successful in employing this hypermasculine mating strategy than gay men.

Do These Findings Relate to All Bisexual Men?

Past research suggests that bisexual males may be a particularly heterogeneous group (Klein, 2014; Rieger et al., 2005; Slettevold et al., 2019; Weinrich & Klein, 2002; Weinrich et al., 2014). In particular, some bisexuals may be close to gay men in their sexual orientations, whereas others may be closer to straight men. To determine if more masculine and heterosexual sounding voices were consistent of bisexual men as a group, or rather, were only true of some bisexual men, we compared androphilic-leaning, gynephilic-leaning, and equally andro/gynephilic bisexuals on variables of interest. A clear pattern emerged wherein all bisexual voices were more female attracted/masculine sounding than gay voices, but only equally andro/gynephilic bisexuals were rated as more female attracted/masculine sound than straight men. This finding is counter-intuitive, and contrary to past evidence that masculinity in bisexual males increases the closer they are in attraction pattern to heterosexual men (i.e., gynephilic-leaning bisexual men would be expected to demonstrate the closest match on perceived masculinity/Kinsey score to straight men; Rieger et al., 2020). Potential explanations for this unexpected effect may fall in one of two categories. The first is that bisexual men generally, and particularly those who are equally male/female attracted, possess stable traits that lead them to sound particularly masculine/heterosexual. Some research has found that bisexual men demonstrate higher sex drive, socio-sexuality, and lower conscientiousness (Lippa, 2020; Rieger et al., 2013; Stief et al., 2014), and lower socio-economic-status (Ross et al., 2016). Future research should examine whether these personality and demographic characteristics contribute to the perceived masculinity/heterosexuality of voices. Another possibility is that a social-desirability bias both explained why bisexual men in our sample labeled themselves equally male/female attracted AND demonstrated the most masculine/heterosexual sounding voices. As bisexuality is often viewed as a transitional phase to coming out as gay, it is possible that a subset of our bisexual males, who were perhaps sensitive to being viewed as gay, picked the prototypical bisexual attraction pattern (equally male/female attracted) and modulated their voice away from the “gay” direction, in order to be perceived as more legitimately bisexual. Lastly, our findings of within-group variability in vocal masculinity-femininity and perceived sexual orientation among bisexual men need to be interpreted with caution owing to the small number of bisexual men in each subgroup and the particular context in which they were recruited (e.g., from a men who have sex with men dating app).

Limitations and Future Research

This is the first study to investigate the accuracy with which people categorize the sexual orientation of bisexual men based solely on their voice and speech patterns. Speakers were instructed to record themselves on their smartphone devices at their leisure without any supervision. This meant that some confounds may not have been controlled for during testing. Audible background noise was an exclusion criterion for the voice sample to be considered viable; however, the recording environment was not controlled for. Recordings made in echoic spaces (e.g., bathrooms) or outside would have different signal transmission and capture due to atmospheric conditions. Microphone to mouth distance was not controlled for which is also another factor that can affect signal quality. Another limitation of recording voices on smartphones is that they are limited to transmitting only a narrow band of frequencies, which may exclude important spectral properties of the speaker’s voices. However, the use of mobile phone recordings improved the ecological validity of the study by replicating the authentic sound of a mobile phone conversation, the most likely of the disembodied voices one experiences regularly. Despite this lack of control, the convenience with which participants could record their voices allowed us to not only recruit from difficult to contact groups, but also provided us with the largest sample size of target voices utilized in any sexual orientation voice perception study to date.

In addition to the voice recording methods, there are issues concerning our measurements. First, the recited text was chosen for the sake of familiarity and brevity, not phonetic content and thus may be restricted in speech characteristics. Given the text of the national anthem has no real-life communicative meaning, it is possible that the voice recordings are not representative of the speech and voice characteristics of the speakers in everyday conversation. Finally, sexual identities are more diverse than the three included in the present study, and the social perceptions of additional identities, such as pansexual or asexual, may be markedly different from those we examined.

Future studies should strive to understand the complex relationship between sexual identity and speech patterns. While the present study suggests that the way bisexual men speak is received as equal to, or more masculine than, the speech of straight and gay men, we have yet to determine what factors influenced the judgments of listeners. Future acoustic analyses of similar samples may aid in identifying unique acoustic cues differentiating bisexual voices. Future studies should also consider how different ways of operationalizing and measuring bisexuality (e.g., based on self-identity, attraction pattern, or behavior) might influence objective and perceptual features of bisexual voices. For instance, in this study bisexuality was measured via a three-option categorical forced choice item and a 9-point Kinsey type measure of sexual attraction. However, no measure of current or past sexual behavior was assessed (see Klein et al. (1985)), or romantic attraction for that matter, which could potentially garner different results. Finally, vocal communication usually occurs in an interpersonal context – whereas in the present study speakers were not communicating with an interlocutor but reciting a section of the Australian national anthem. A past study found that gay men modulated their voice to sound more gay when speaking to people to whom they were “out” (Daniele et al., 2020). It is possible that similar context-specific vocal modulating could occur in bisexual men as well (albeit playing out in an even more complex manner); thus future studies should examine vocal cues in bisexual men when communicating with different types of interlocutors (i.e., with gay men, with other bisexual men, with heterosexual people) as well as to interlocutors to whom the bisexual men is out versus passing as straight or gay.

Conclusion

The present study provided mixed preliminary indications of a distinct vocal pattern unique to bisexual-identified men. The current study adds support to previous studies that found that naïve listeners can distinguish the voices of gay and straight men at rates greater than chance rates, although this was not the case for bisexual men. Interestingly, listeners also rated the voices of a subset of bisexual men to be the most exclusively female-attracted and masculine, suggesting that their voices may indeed impart a unique impression, however, an impression that listeners do not associate with bisexual men. These findings contribute to the understanding of social perceptions of bisexual men and the relationship between speech-based sexuality judgments, and impressions of vocal femininity-masculinity.

Disclosure Statement

No potential conflict of interest was reported by the authors.

Supplementary Material

Supplemental data for this article can be accessed online at https://doi.org/10.1080/00224499.2023.2182267