Subtypes of homosexuality

[nordicblue]

There is a widespread belief that some men’s sexual orientation can be judged on the basis of their voice (i.e. that some men “sound gay”).  One currently untested explanation of the origins of gay sounding speech is that it acts as a social marker of membership in the gay male community.  The current study casts doubt on that hypothesis.  However, gay sounding speech was strongly related to recalled childhood gender nonconformity in both gay and heterosexual men.  This is consistent with the hypothesis that gay sounding speech emerges early in life when boys mimic and adopt certain speech patterns more typical of females.  Because gay men are significantly more likely to experience gender nonconforming childhoods, feminine speech patterns become associated with male homosexuality mainly through proxy.

Introduction

Stereotypes suggest that gay men’s speech differs from that of heterosexual men.  As one scholar noted, “A dependable wellspring of this caricature [of gay men] is popular culture, which seemingly never tires of the lisping fag, whose roller coaster intonation and high pitched shrieks mark him as an object of comedy or contempt” (Kulick, 2000, p. 260).  These stereotypes, however, are not new; case studies dating back to the nineteenth century frequently remarked on the distinctive nature of gay men’s voices (Shaw & Ferris, 1883).  Accordingly, many people believe that they can determine a person’s sexual orientation based solely on the way that he speaks.  A handful of studies investigating this issue have found that judgments of sexual orientation, based only on speech samples, are in fact usually accurate (Travis, 1981; Gaudio, 1994; Linville, 1998).  Other researchers have isolated specific acoustic cues (also characteristic of female speech) that people attend to in making these judgments (Linville, 1998; Rogers & Smyth, 2001).  Thus, the notion that gay men speak differently than heterosexual men has received some empirical support, although stereotypes may distort and exaggerate this difference.

This line of research has several implications.  It contradicts the opposing belief that sexual orientation, unlike race or gender, is largely invisible in social interactions (Frable, Platt, & Joey, 1998).  As others have noted, controversial policies such as “don’t ask, don’t tell” also rely on the assumption that sexual orientation only becomes apparent when one chooses to disclose it (Ambady, Hallahan, & Conner, 1999).  Perhaps equally important, understanding the concomitants of sexual orientation may aid in understanding the origins of sexual orientation itself.  Given the widespread belief that speech patterns are associated with sexual orientation, it is surprising that such little research has been conducted in this area.

Moreover, the existing studies in this area suffer from two major handicaps.  First, they have relied on extremely small samples to determine the accuracy of judgments of sexual orientation and to estimate the magnitude of speech differences between gay and heterosexual men.  The largest published study examining perceptual accuracy contained only 5 gay and 4 heterosexual male speakers (Linville, 1998).  Second, and more importantly, no study to date has examined the possible origins of these speech differences.  Therefore, the purposes of this paper are to replicate previous findings using a larger sample and to address the question of why some men, for lack of a better term, “sound gay.”

Current Explanations

One possible explanation is that gay sounding speech acts as a marker of membership in the gay male community and that gay men either consciously or unconsciously acquire this speech through exposure (Linville, 1998).  Others have theorized that phonemic variation in general may signal group identity (Chambers & Trudgill, 1980).  Thus, the implicit assumption is that gay sounding speech serves a functional purpose, such as uniting gay men from diverse backgrounds (Barrett, 1997).  It is also possible that gay men adopt specific speech patterns in order to identify each other in diverse social settings.  In that case, gay sounding speech would be analogous to nonverbal cues of homosexuality, such as rainbow-colored ornamentation.  All these hypotheses presume that gay sounding men are also those who self-identify as gay (Kulick, 2000).  For example, a closeted gay man may go to great lengths to conceal signs of his homosexuality rather than advertise it.  Hence, openness about one’s homosexuality, i.e. outness, is hypothesized to mediate gay sounding speech.

Although gay sounding speech may be used for functional purposes, that does not necessarily imply that it exists because of them.  There are four reasons to question the hypothesis that such a speech pattern develops in order to accomplish a particular goal, such as signaling one’s membership in the gay male community.

First, gay sounding speech appears to emerge early in life, well before a specific self-identification as gay.  Two longitudinal studies examining the relationship between homosexuality and childhood gender nonconformity (discussed below) both noted feminine sounding speech in boys who disproportionately became gay in adulthood (Zuger, 1984; Green, 1987).  Zuger (1984) collapsed feminine speech into a category with feminine gestures and found that 73% of the effeminate boys displayed some feature associated with this category.  Likewise, one mother in Green’s study remarked that her son “talks like a girl, sometimes walks like a girl, acts like a girl” (1987, p. 2).  These observations are quite common; mothers of boys with gender identity problems rate their sons’ speech and motor behavior as significantly more feminine than mothers of sons in control groups (with effect sizes ranging from 0.92–4.47) (Zucker, 1992).  Thus, if gay (or feminine) sounding speech emerges early in life, this would be inconsistent with the hypothesis that such a speech pattern develops in response to something dependent on self-awareness of sexual orientation (e.g. a desire to signal other gay men).

Second, such a hypothesis fails to account for the fact that some gay men do not sound gay (Travis, 1981; Gaudio, 1994; Linville, 1998) and that, at least based on informal accounts, a few heterosexual men do.  Thus, the literature demonstrating within-sex speech differences based on sexual orientation has largely ignored within-orientation differences.  If gay sounding speech originates from a desire to unite gay men from diverse backgrounds, for example, it makes little sense why a sizable number of gay men do not possess this speech pattern and are usually misidentified as heterosexual.  Likewise, under this hypothesis, heterosexual men who sound gay (though perhaps few in number) have no reason to adopt a speech pattern widely associated with a sexual orientation discordant to their own.

Notably, there are few stereotypes about lesbian speech (Zwicky, 1997), and the limited research comparing lesbian and heterosexual women’s speech has found few actual differences (Moonwomon, 1997; Waksler, 2001; but see Travis, 1981).  At face value, though, many of the current hypotheses about the origins of gay sounding speech should apply equally well to speech differences between lesbians and heterosexual women.  However, based on current research, these differences (assuming they exist) are far less perceptually salient than comparable differences in men.  Any hypothesis attempting to explain the causes of gay sounding speech must also explain why there is such an apparent difference between gay and heterosexual men’s speech but not between lesbian and heterosexual women’s speech.

Third, none of the current explanations would predict that gay sounding speech differs from other types of speech specifically because it is shifted in a female-typical direction, which appears to be the case (Rogers & Smyth, 2001; Linville, 1998).  Indeed, some have argued that there is no a priori reason to suspect that gay sounding speech shares greater similarity with female speech, and that our perception of gay sounding speech as feminine simply reflects our bias to interpret all deviations from cultural norms as feminine (Zwicky, 1997).  For example, if the purpose of gay sounding speech is create a sense of unity among gay men (Barrett, 1997), one could just as easily imagine the adoption of some arbitrary deviation from convention—such as putting stresses on the wrong syllables or clucking in the place to pauses—to mark gay sounding speech.  In reality, though, gay sounding speech differs in a very specific, female-shifted way.

Fourth, current explanations of gay sounding speech ignore the social costs that accompany sounding gay, such as stigmatization during childhood and adolescence.  While gay male youth who are open about their sexual orientation often face harassment, the very perception of being gay (or feminine) may be an equally if not more relevant risk factor (Remafedi, Farrow, & Deisher, 1991).  Sociolinguists have idealized the causes of gay sounding speech while seemingly ignoring the practical consequences of such speech in social interactions.  Even in adulthood, most gay men prefer sex-typicality in their romantic partners (Bailey, Kim, Hills, & Linsenmeier, 1997) and frequently make proscriptions against feminine men (i.e. “No femmes” or “If I wanted to date a woman, I’d date one”).  Thus, it is unlikely that many gay men consciously choose to sound gay, given the attendant costs of such a speech pattern in the romantic marketplace.

While some gay men who do not normally sound gay may occasionally adopt such a speech pattern as a “register” in certain social contexts (Barrett, 1997), this is a temporary deviation from their natural speaking styles, and does not negate the finding that most gay men desire masculine partners.  Conversely, it is clear that some men “may have spoken in a stereotypically gay style for most or all of their lives” (Barrett, 1997, p. 194).  It is these individuals that pose the more difficult question.  What causes them to sound gay?

Alternate Explanation

There is at least one other possible explanation of the origins of gay sounding speech: gay sounding men may have been more likely during childhood to mimic and adopt the speech patterns more typical of females.  Such a method of acquisition would be similar to that of a learned dialect.  This female-shifted speech pattern, superimposed onto a male voice, produces what we perceive as gay sounding speech.

If gay sounding speech is learned, it must be acquired fairly early in life, given that some boys already begin to sound gay in childhood.  Moreover, it is likely that certain behaviors facilitate or reinforce this acquisition.  Research on the acquisition of accents may be relevant here.  Children who immigrate with their parents to a foreign country tend to pick up the accent of the new location, rather than that of their parents, because they learn from and imitate their peers (Harris, 1998).  For example, after a few months at a nursery school in California, the daughter of a British linguist began speaking “black English” (Baron, 1992).  Although not all the children attending the school were black, the children she played with were, and their speech pattern was the one she adopted.

While it is unlikely that gay sounding men developed their speech patterns by mimicking other gay sounding boys in childhood (because they were probably rare), they might have adopted the speech patterns of other girls.  The acoustic cues of gay sounding speech are also those more prevalent in female speech (Rogers & Smyth, 2001).  For example, /s/ production in gay sounding speech (the cue accounting for the most variance in voice ratings) is longer and has a higher frequency than typical male speech—but it falls in the range of typical female speech (Linville, 1998; Avery & Liss, 1996).  Given that boys and girls speak differently on average even before puberty (Bennett & Weinberg, 1979), it is possible that a small fraction of boys adopt the speech patterns more typical of girls (or women) and, as a result, sound gay.  If so, the things that promote (or at least accompany) the development of gay sounding speech in childhood might include behaviors such as preferential affiliation with females, having more friendships with females, and taking a female role during role play; affective components serving a similar function might include feeling very feminine and greater self-identification with other females.

There is, in fact, a large body of evidence documenting such behaviors in the childhoods of prehomosexual boys.  These behaviors generally fall under the label of “childhood gender nonconformity,” which takes into account behaviors such as same- versus opposite-sex peer affiliation, rough-and-tumble play, toy interests, fantasy roles, and dress-up play.  These behaviors not only differ significantly between boys and girls on average but also between prehomosexual and preheterosexual children within each sex.  As previously mentioned, two prospective studies have found that boys who exhibit marked levels of childhood gender nonconformity are much more likely to become gay in adulthood compared to controls (Zuger, 1984; Green, 1987).  While these studies included boys with levels of childhood gender nonconformity high enough to meet the Diagnostic and Statistical Manual of Mental Disorders (DSM) criteria for gender identity disorder, and hence may have limited generalizability for non-clinical populations, other studies relying on more representative samples of gay men replicate the finding that childhood gender nonconformity frequently precedes male homosexuality.

Bailey and Zucker’s (1995) meta-analysis of 41 studies demonstrated that, on average, gay men and lesbians recall significantly more gender nonconforming childhoods than their heterosexual counterparts (effect size = 1.19).  These results cannot simply be attributed to factors influencing subjective recall, e.g. that gay men internalize of stereotypes about gay men’s femininity and thus expect to remember more gender-atypical childhoods.  For example, a mother’s recall of her gay son’s gender-related childhood behavior is correlated with her son’s own recall (Bailey, Nothnagel, & Wolfe, 1995) but is uncorrelated with the extent of her knowledge about her son’s sexual orientation (Bailey, Miller, & Willerman, 1993).

This association between childhood gender nonconformity and homosexuality, however, must be accompanied by a caveat, which is relevant to the phenomenon of gay sounding speech: not all gay men recall childhood gender nonconformity and not all heterosexual men recall childhood gender conformity.  In fact, about a third of gay men have recall profiles identical to that of heterosexual men (Bailey & Zucker, 1995).  Likewise, a boy exhibiting levels of childhood gender nonconformity comparable to those recalled by most gay men has a 49% chance of becoming heterosexual in adulthood.  Moreover, simply because childhood gender nonconformity precedes adult homosexuality, it does not necessarily follow that the former causes the latter.  The predominant biological theory of homosexuality posits that sex-atypical prenatal hormone exposure partially shifts the organization of certain brain structures toward that of the opposite sex (for a review, see Rahman & Wilson, in press).  This, in turn, affects sex-typed traits, such as sexual attraction to either males or females and certain childhood behaviors, which are usually—though not necessarily—clustered together.

If certain components of childhood gender nonconformity cause gay sounding speech, but are only imperfectly associated with adult homosexuality, this may explain why some gay men do not sound gay and why some heterosexual men do.  Likewise, because childhood gender nonconformity is significantly more predictive of homosexuality in males than in females (Bailey & Zucker, 1995), the variation in adult female speech patterns may not correspond to female sexual orientation as well as does variation in male speech patterns for male sexual orientation.  Thus, it might be the case that there are true differences between lesbian and heterosexual women’s speech, but that these differences are poor predictors of female sexual orientation and are less perceptually salient given greater overall variability in female speech.  A causal relationship between childhood gender nonconformity and gay sounding speech would also be consistent with the evidence that many prehomosexual boys already begin to sound gay in childhood.  Because this speech pattern persists into adulthood for a sizable number of men (against countervailing social forces), it appears relatively difficult to change, in much the same way that an accent is difficult to change once the critical period for language acquisition and development has passed (Munro, Flege, & MacKay, 1996; Bialystock & Hakuta, 1994).  Not surprisingly, this critical period for language overlaps with the period of sex-typed childhood behavior.

Rationale for Present Study

The primary purpose of this study is to determine if there is an association between childhood gender nonconformity and gay sounding speech.  This association may explain why judgments of sexual orientation based on speech are usually accurate, although the cues that people use to make judgments may depend more on childhood behaviors than on sexual orientation per se.  Moreover, because many of the currently proposed explanations for gay sounding speech hypothesize a linkage between speech and openness about one’s homosexuality, one simple test of this would be to examine whether degree of outness correlates with how gay a man sounds.

Methods
Participants

Participants were recruited as part of a larger study investigating various concomitants of sexual orientation; however, the results reported here are limited to those relevant to gay sounding speech.  The sample consisted of 30 gay, 4 bisexual, and 24 heterosexual males recruited primarily through networking (“snowballing”), flyers posted around the University of Texas at Austin campus and local coffeeshops, and (for a limited number of gay participants) through a posting to a local gay-oriented website.  Bisexual males were omitted in any analyses explicitly comparing groups by sexual orientation but were otherwise included.  All participants were paid $10 as compensation.  The gay men were slightly older than the heterosexual men (mean age = 24 and 21.8, respectively), but this difference was not statistically significant.  The ethnic compositions of the heterosexual and gay sample were comparable.
Measures

Sexual Orientation  Sexual orientation was assessed by both self-report (heterosexual, gay, bisexual, or other) and the Kinsey scale, which allows individuals to rate themselves along a 7-point continuum from exclusively heterosexual (0) to exclusively homosexual (6).  For participants to be considered heterosexual, they had to identify as such and score 0 or 1 on the Kinsey scale; for a male to be considered homosexual, he had to identify as such and score 5 or more.

Childhood Behavior Scale  Childhood gender nonconformity was assessed using the Recalled Childhood Gender Behavior Scale (Mitchell & Zucker, 1991).  The scale includes 18 items which assess both behavioral aspects of gender conformity (i.e. “As a child, I enjoyed playing sports such as baseball, hockey, basketball, and soccer”) as well as affective components (“As a child, I felt very masculine”).  Participants were instructed to consider only the period before 12 years of age.  Participants’ responses to individual items were averaged together, with higher scores indicating greater recalled childhood gender nonconformity.

Outness Inventory  Participants identifying as gay completed the Outness Inventory, which assesses the degree to which a person is open about his or her homosexuality (Mohr & Fassinger, 2000).  Overall outness is derived from averaging together the inventory’s subscales (outness to family, religion, and world).

Recording Procedures

All recordings were obtained with a microphone headset (Labtec Axis-521) attached to a computer.  The use of a headset standardized the distance from the speaker’s mouth to the microphone.  Voice samples were digitally recorded at 44.100 kHz using the 1st Sound Recorder software with noise cancellation activated.

Before the recording, I explicitly requested that participants try their best not to alter their voice in any way from natural speech.  In accord with Gaudio (1994) and Linville (1998), participants read an excerpt from a play (Torch Song Trilogy), first silently and then aloud using their natural speaking voices.  The dialogue was a conversation between two individuals in a bar, but participants read the part of only one speaker.  The advantage to having all participants read the same dialogue is that it controls for any differences in what participants say, focusing instead on how they say it.  The obvious drawback to this method is that reading text aloud is an imperfect representation of a person’s actual, spontaneous speech.  However, prior research found no interaction between the type of speech sample acquired (spontaneous versus reading) and sexual orientation on voice ratings of sounding gay (Travis 1981).  In other words, although there may be perceptible differences between spontaneous and read speech, these differences are not immediately relevant to the phenomenon of gay sounding speech.

Participant Ratings

An excerpt (approximately 30 seconds long) was then extracted from the middle of the voice samples for playback.  Each participant rating the voices (the listener) wore earphones.  Presentation order of the samples was randomized and listeners had control over when to advance to the next voice.  Listeners rated each voice on a 7-point scale for how gay it sounded.  In a pilot study of 10 listeners and a partial sample of the voices, interrater reliability for the voice ratings was high (Cronbach’s α = 0.91).  The results reported here are based on 4 listeners (not involved in any other part of the experiment) who rated all 58 voices.  The reliability of these listeners’ voice ratings was also high (Cronbach’s α = 0.84).

Results

To determine whether or not listeners are accurate in judging sexual orientation from voice samples, previous studies have typically employed forced-choice items in which participants must classify each voice as belonging to either a heterosexual or gay male.  However, the problem with this approach is that results may be distorted by listeners who are overly conservative or overly liberal in their judgments.  For instance, a listener who rates the vast majority of voices as belonging to heterosexual males will have a very poor accuracy index.  To avoid this problem, the present study used discriminant function analysis.  In discriminant analysis, the independent variables are the predictors (the voice ratings of how gay a voice sounded) and the dependent variables are the groups (gay or heterosexual).  Thus, based on all four listeners’ ratings of how gay a voice sounded, a prediction was made regarding the sexual orientation of each speaker and compared with that speaker’s actual sexual orientation.  The overall accuracy of these predictions was 68.5%, which is significantly greater than what would be expected by chance.  Broken down by sexual orientation, 75% of the heterosexual speakers were correctly classified, while 63% of the gay speakers were correctly classified.

A one-tailed t test revealed that gay men were significantly more gay sounding than the heterosexual men (t (52) = -3.964, p = .0002).  The effect size for this difference was large (Cohen’s d = 1.12).  Moreover, how gay a man’s voice sounded correlated with childhood gender nonconformity (full sample, r = .544, p < .0001) in both the gay (r = .429, p = .009) and heterosexual men (r = .365, p = .040).  Figure 1 shows the relationship between these variables in the full sample.

[MrMazda]

People actually spend money on studies like this? That's kind of sad really. It makes you wonder if maybe they have better things to be spending that money on.

[cureton]

@MrMazda:

People actually spend money on studies like this? That's kind of sad really. It makes you wonder if maybe they have better things to be spending that money on.

more budget in health care and less in this stupid studies.

[HerrSamsa]

Basically it's the straight homophobes that talk funny because they forcibly dissuade themselves from adopting speech or vocal patterns utilized by half the population