Original Paper

Folia Phoniatr Logop 1998;50:35–48

Sue Ellen Linville Acoustic Correlates of Perceived

Program in Speech Pathology and versus Actual Sexual Orientation in
Audiology, Marquette University,
Milwaukee, Wisc., USA Men’s Speech

OOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOO

Key Words
Gay speech W Gender and speech W Sexual orientation and voice W
Homosexual voice
OOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOO

Abstract
The purpose of this study was to explore possible dif-
ferences in the speech patterns of gay and straight men,
both as a function of perceived and actual sexual orien-
tation. Tape recordings of monologue readings from 5
openly gay men and 4 straight men were played to 25
listeners for judgments of perceived sexual orientation.
Monologues were analyzed in terms of /s/ duration, /s/
peak frequency, modal speaking fundamental frequen-
cy, speech rate, and long-term average speech spectra.
Listeners correctly identified the sexual orientation of
these speakers at a rate of 79.6%. Multiple regression
analysis revealed that gay judgments were significantly
associated with higher peak /s/ frequency values and
longer /s/ duration values. Acoustic cues associated
with perceived sexual orientation generally agreed
with acoustic findings as a function of actual sexual
orientation. While these results must be interpreted
cautiously, findings suggest that members of the com-
munity of openly gay men demonstrate certain speech
characteristics that are discernible to listeners.

OOOOOOOOOOOOOOOOO

Akustische Korrelate der wahrgenommenen und tatsächlichen sexuellen Orientierung

in der Sprache von Männern
Die Absicht dieser Arbeit war es, mögliche Sprachun-
terschiede zwischen homo- und heterosexuellen Män-
nern als Funktion der wahrgenommenen und der tat-
sächlichen sexuellen Orientierung zu untersuchen. Ein
Monolog wurde von 5 homosexuellen Männern, die
sich zu ihrer Homosexualität bekannten, und 4 hetero-
sexuellen Männern auf Tonband gelesen und 25 Höre-
rinnen zur Beurteilung der sexuellen Orientierung vor-
gespielt. Die Monologe wurden hinsichtlich /s/-
Dauer, /s/-Frequenzmaximum, Grundfrequenz der
modalen Sprechstimme, Sprechtempo und LTA-Spek-
tren analysiert. Die Hörerinnen identifizierten die se-
xuelle Orientierung der Sprecher korrekt in 79,6%. Die
multiple Regressionsanalyse ergab, dass die Beurtei-
lung «homosexuell» signifikant mit höheren /s/-Fre-
quenzmaxima und längerer /s/-Dauer verbunden war.
Die akustischen Merkmale der wahrgenommenen se-
xuellen Orientierung stimmten im allgemeinen mit
den akustischen Daten als Funktion der tatsächlichen
sexuellen Orientierung überein. Obwohl die Resultate
mit Vorsicht zu interpretieren sind, deuten sie darauf-
hin, dass die homosexuelle Gemeinschaft gewisse
Sprachcharakteristika aufweist, die für Hörer wahr-
nehmbar sind.

© 1998 S. Karger AG, Basel Sue Ellen Linville, PhD

ABC 1021–7762/98/0501–0035$15.00/0 Program in Speech Pathology and Audiology, Marquette University
Fax + 41 61 306 12 34 PO Box 1881, Milwaukee, WI 53201-1881 (USA)
E-Mail karger@karger.ch This article is also accessible online at: Tel. (414) 288-1406, Fax (414) 288-3980
www.karger.com http://BioMedNet.com/karger E-Mail 6533 Linville@vms.csd.mu.edu
147.8.31.43 - 3/25/2017 5:18:00 PM
University of Hong Kong
Downloaded by:
OOOOOOOOOOOOOOOOO
Corrélats acoustiques des préférences sexuelles perçues et réelles du langage
masculin
Cette étude a eu pour objectif d’explorer l’existence
éventuelle de différences entre les schémas langagiers
d’hommes homosexuels et hétérosexuels en fonction
des préférences sexuelles perçues et réelles. L’enre-
gistrement d’un monologue lu par 5 homosexuels
avérés et 4 hétérosexuels a été soumis à 25 auditrices
pour l’évaluation des préférences sexuelles des locu-
teurs. L’analyse a porté sur la durée de /s/, la fréquence
maximale de /s/, la fréquence fondamentale modale de
la voix parlée, le débit et les spectres LTA. Les au-
ditrices ont correctement identifié les préférences
OOOOOOOOOOOOOOOOO
Very little is known about the influence of
sexual orientation on speech and voice, de-
spite considerable evidence that cultural stan-
dards, gender-related expectations, and per-
sonality variables influence speech patterns
and vocal quality [1–10]. If listeners perceive
differences in the speech of men who ‘sound
gay’, and if those differences correlate with
speakers’ actual sexual orientation, such in-
formation could be taken into account by
sociolinguists interested in gender-based dif-
ferences in speech. For sociolinguists to fail to
do so would be to overlook a social phenome-
non that could significantly alter dialectical
descriptions of male and female speech pat-
terns.
Several features of speech and voice have
been labeled as gender-related in previous
investigations. For instance, after administer-
ing questionnaires to 466 high school and col-
lege students, Kramer [6] reported the follow-
ing speech traits as stereotypic of females:
clear enunciation, gentle speech, fast speech,
variability in rate and pitch, and emotional
speech. Speech/voice traits labeled as male in
the Kramer [6] study included: deep voice,
demanding voice, loud speech, dominating
speech, and forceful speech. Aronovitch [11]
36 Folia Phoniatr Logop 1998;50:35–48
sexuelles des locuteurs dans 79,6% des cas. L’analyse
de régression multiple a fait apparaı̂tre que le jugement
«homosexuel» est significativement corrélé à des fré-
quences maximales de /s/ plus aiguës et à une durée de
/s/ plus longue. Les caractéristiques acoustiques asso-
ciées aux préférences sexuelles perçues ont générale-
ment été en accord avec les résultats acoustiques en
fonction des préférences sexuelles réelles. Bien qu’une
certaine prudence s’impose, ces résultats suggèrent que
les caractéristiques langagières des homosexuels mas-
culins sont perceptibles.
correlated personality judgments from speech
samples with measures of speech rate, intensi-
ty, sound-silence ratio, and mean fundamen-
tal frequency (F0). While he found significant
relationships between many of the personality
traits and the objective speech measures, the
‘most striking’ finding emerged after breaking
down the results by sex of speaker. This analy-
sis revealed that judgments of personality
from women’s voices were related to speech
time and average loudness and pitch while
judgments from men’s voices were related to
variations in loudness and pitch. Speech rate
was associated with personality judgments
from both male and female voices.
Aside from a few unsubstantiated claims
that gay men display ‘high, childish voices’
[12, p. 29], and that gay men imitate speech
traits characteristic of women [13], speech
and voice patterns in the gay population have
remained largely uninvestigated. Travis [14]
asked 20 listeners to judge spontaneous
speech and reading samples of 23 male and
female speakers who varied in sexual orienta-
tion. Judges were asked to determine the ana-
tomical sex of the speaker and to describe the
speech as ‘masculine’ or ‘feminine’. Travis
[14] then examined twelve measures related
Linville
147.8.31.43 - 3/25/2017 5:18:00 PM
University of Hong Kong
Downloaded by:
to speech rate, F0 range/speaking level, and
intensity range/speaking level in each of the
speech samples. None of the rate, F0, or inten-
sity measures were found to be significant in
differentiating male heterosexual and homo-
sexual speakers. However, when speach sam-
ples were played to listeners, both female and
male homosexual speakers were incorrectly
judged ‘masculine’ or ‘feminine’ more often
than were heterosexual speakers. Listeners
were not asked to judge speaker’s sexual
orientation directly.
The first direct evidence that gay men
demonstrate a speech pattern that is discern-
ible to listeners was obtained by Gaudio [15].
Gaudio [15] recorded 4 gay and 4 straight
men reading aloud a passage from an account-
ing text and a monologue from a gay play. He
then played samples of the readings to a panel
of 13 listeners who directly judged the speak-
ers’ sexual orientation. Listeners’ judgments
of sexual orientation were correlated signifi-
cantly with speakers’ actual sexual orienta-
tions.
If listeners are capable of perceiving differ-
ences in the speech of gay and straight men,
the presence of such cues might be explained
in two ways. One possibility is that gay and
straight men demonstrate anatomic and/or
physiologic differences that result in percepti-
ble changes in voice. A second possibility is
that gay and straight men differ systematically
in articulatory and/or suprasegmental behav-
iors.
Evidence of anatomic or physiologic dif-
ferences in gay and straight men is extremely
limited. There is some suggestion of endo-
crine system differences in male homosexuals
as compared to heterosexuals. Kolodny et al.
[16] reported lower plasma testosterone levels
in a group of 30 homosexual males in compar-
ison with 50 heterosexual controls. Although
later studies examining serum testosterone
levels in homosexual males failed to replicate
Sexual Orientation and Speech in Men
these findings [17, 18], differences in subject
selection criteria for the studies may account
for the disparate findings [17]. Lower serum
testosterone levels, if of sufficient magnitude,
could affect growth and development of the
larynx during puberty resulting in pitch and
voice quality differences [19, 20]. Specifically,
one might predict higher speaking F0 and
higher vocal tract resonances in individuals
with low testosterone levels as a result of
decreased vocal fold mass and smaller dimen-
sions of the larynx and vocal tract overall.
In only one study have possible differences
in speaking F0 as a function of sexual orienta-
tion in men been examined. Lerman and
Damsté [21] found no significant difference
in the mean speaking F0 of 13 homosexual
and 13 heterosexual men from reading sam-
ples, spontaneous speech, and sustained vow-
el productions. These authors made no effort
to control for variations in relative homo/
heteroerotic interest in their subjects, how-
ever, which could have affected their results
[17].
Interestingly, vowel formant frequencies
have been shown to play a significant role in
listener identification of speaker sex in adults,
particularly if differences in speaking F0 are
eliminated [22–28]. In addition, vowel for-
mants appear to distinguish male speakers
who differ with regard to listener perception
of perceived masculinity. Specifically, Avery
and Liss [29] found that while individual for-
mant values on four connected speech vowels
(/i/, /a/, /æ/, and /u/) varied substantially both
within and between speakers, the majority of
F1 and F2 values for less-masculine-sounding
(LMS) males were higher than values for
more-masculine-sounding (MMS) males. In
preadolescent children as well, vocal tract res-
onance properties have been found to provide
primary cues regarding sexual identification
of voice [30]. Bennett and Weinberg [30] con-
cluded that differences in vocal tract reso-
Folia Phoniatr Logop 1998;50:35–48 37
147.8.31.43 - 3/25/2017 5:18:00 PM
University of Hong Kong
Downloaded by:
nance observed in preadolescent boys and
girls might be explained either by differences
in vocal tract size related to speaker sex or by
sex-specific articulatory behaviors. It is possi-
ble, therefore, that vocal tract resonance cues,
as reflected in vowel formant frequencies,
may be significant to listeners in distinguish-
ing gay and straight speakers, if the two
groups demonstrate differing articulatory pat-
terns and/or vocal tract dimensions.
Evidence of differences in articulatory be-
haviors for gay and straight speakers also is
lacking at present. Gaudio [15] analyzed the
speech samples of gay and straight men for
overall pitch range and found no significant
differences, although differences in overall
pitch range while reading from an accounting
text were ‘suggestive’ of differences (p ! 0.10).
Gaudio [15] concluded that overall pitch
range may be too crude a measure to differen-
tiate the two groups. However, Avery and Liss
[29] have reported that LMS males produce a
pattern of rising intonations followed by fall-
ing shifts more frequently and with greater
extent and slope than do MMS males.
While reporting no data, Lakoff [13] de-
scribed the speech of women (and gay men) as
characterized by a ‘lisped’ /s/. Perceptions of
‘lisping’ might be related to increased dura-
tion of fricatives or to shifts in the frequency
of the noise produced as a result of slightly
altered articulatory placement [31]. Interest-
ingly, previous research has indicated that lis-
teners are able to correctly identify the sex of
speakers from voiceless fricatives, particu-
larly /s/ and /e/ [32, 33]. In addition, Avery
and Liss [29] found that LMS males displayed
/s/ productions with higher center frequencies
than did MMS males. While Avery and Liss
[29] did not control for physical size differ-
ences in the LMS and MMS male groups, they
did not feel that the two groups were greatly
disparate in height. In addition, less vowel
reduction in connected speech was noted for
38 Folia Phoniatr Logop 1998;50:35–48
LMS speakers supporting their hypothesis
that LMS speakers produced clearer speech.
The purpose of this study is to explore pos-
sible differences in the speech patterns of gay
and straight men to determine: (1) if listeners
can accurately judge the sexual orientation of
a group of male speakers from samples of
their speech, and (2) if acoustic measures
related to /s/ duration, /s/ peak frequency,
modal speaking F0, speech rate, and long-term
average (LTA) speech spectra are associated
with perceived or actual sexual orientation.
Method
Speakers
A total of 9 men (5 gay; 4 straight) participated as
speakers in this study. All speakers spoke Standard
American English with no evidence of a regional dia-
lect. One gay speaker was an acquaintance of a re-
search assistant in the project. This gay speaker en-
listed 4 additional gay speakers. Four straight subjects
were acquaintances of either the investigator or anoth-
er female faculty/staff member in the speech pathology
department. Subjects were told only that the study was
an investigation of men’s speech.
One straight speaker was enlisted through notices
posted on public bulletin boards around the downtown
area of Milwaukee. The notices solicited subjects for a
study of ‘men’s speech’. This straight subject subse-
quently was eliminated from the study, however, be-
cause of a speech pattern judged to be disordered. Spe-
cifically, imprecision in articulation, a markedly slow
speech rate, and a monotone voice were evident. This
speaker had a history of clinical depression for which
he was receiving medication.
All speakers completed the Klein Sexual Orienta-
tion Grid [34] which applies the seven-point Kinsey
scale (0 = exclusively heterosexual, 6 exclusively ho-
mosexual) to three sexual variables (sexual attraction,
sexual behavior, sexual fantasies), two emotional/
social variables (emotional preference, social prefer-
ence) and two life-style variables (self-identification,
life-style) over three times frames (past, present, ideal).
At the time of testing (‘present’), gay speakers ranged
from 5 to 6 (mean = 5.6) on the sexual variables, from 3
to 6 on emotional/social variables (mean = 4.8), and
from 5 to 6 on life-style variables (mean = 5.7). Straight
speakers were uniformly 0 on the sexual variables and
Linville
147.8.31.43 - 3/25/2017 5:18:00 PM
University of Hong Kong
Downloaded by:
the self-identification variables. On the emotional/
social variables, straight speakers ranged from 0 to 4
(mean = 2).
Gay speakers ranged in age from 30 to 41 years
(mean = 35.8 years). Physically, gay speakers ranged in
height from 1.65 to 1.83 m (mean = 1.77 m) and in
weight from 63.5 to 99.8 kg (mean = 73.9 kg). All
5 speakers lived openly as gay, and had done so for
several years. Four of the 5 speakers had lived in Cali-
fornia, Arizona, Florida, or Pennsylvania in addition
to Wisconsin during their life. All 5 were college grad-
uates; 1 reported having attended graduate school.
Two were unemployed, 1 worked in concession sales, 1
as a caterer, and 1 as a waiter. Four of the 5 speakers
were smokers, with daily cigarette consumption rang-
ing from 1 to 2 packs. Three reported heavy alcohol
consumption, 1 reported social drinking, and 1 re-
ported no alcohol consumption. Three of the gay
speakers were HIV-positive; 2 were receiving medica-
tion for this condition (AZT, DDI and/or Bactrim). In
addition, 1 speaker was receiving medication for high
blood pressure. Three speakers reported no singing or
acting experience, 1 reported very limited acting expe-
rience, and 1 reported participation in church choir.
While 4 speakers reported no history of neurological
disorders, 1 reported a history of two concussions.
None of the gay speakers reported any endocrine disor-
ders or chronic respiratory disorders.
Straight speakers ranged in age from 24 to 43 years
(mean = 33.3 years). Straight speakers ranged in height
from 1.75 to 1.83 m (mean = 1.78 m) and in weight
from 68.0 to 77.1 kg (mean = 72.1 kg). One speaker
reported living in various locations along the east coast
for short periods of time (less than 3 months) during
his life, and 1 spent several years in Atlanta, Ga. as a
child. One speaker was a high school graduate, 2 had
undergraduate degrees, and 1 had a Master’s degree.
All 4 speakers were employed: 1 as an engineer, 1 as a
salesman, 1 as a comedian/illustrator, and 1 as a come-
dian/agent. Two speakers were nonsmokers; 2 reported
occasional smoking. All reported light to moderate
alcohol consumption. Three speakers were receiving
no medications; 1 reported receiving allergy medica-
tion. Two speakers reported no singing or acting expe-
rience; 2 reported limited professional acting/stand-up
comedy experience. Three speakers reported no neuro-
logical disorders; 1 reported a history of two concus-
sions. None of the speakers reported any endocrine
disorders. Three speakers reported no history of
chronic respiratory disorders; 1 reported a history of
hay fever.
Sexual Orientation and Speech in Men
Recording Procedures
All recording was conducted at the Marquette Uni-
versity Speech and Hearing Clinic, in a room furnished
at one end with two sofas and a coffee table; at the oth-
er end of the room was a makeshift bar facing a one-
way mirror. Each speaker was fitted with a remote tie-
clip microphone (Shure L1-V; Shure 839 W) to allow
freedom of movement during recording. The tape
recorder (Tascam 102) was located in an adjacent
room, behind the one-way mirror.
Speakers read a monologue delivered by a gay male
character in the play Torch Song Trilogy [35]. A mono-
logue was selected for the speech sample, as opposed to
a reading passage, because a play is written to be spok-
en rather than to be read. It was felt that a monologue
more nearly approximates spontaneous speech than
would a reading passage.
The setting of the monologue was a bar, and the
male character was initiating a conversation with
another patron of the bar. The name of the bar patron
was altered to be gender-neutral. The speakers were
not told the source of the passage, and the subject mat-
ter was not explicitly gay (‘Appendix A’). Prior to read-
ing aloud, speakers read the monologue silently to
familiarize themselves with the text. The monologue
was not rehearsed aloud, however; subjects’ first oral
reading of the material was recorded and analyzed.
Speakers read the passage while they stood at the end
of the ‘bar’, with an empty barstool beside them. They
were instructed to follow stage directions as they were
encountered during the monologue reading.
Acoustic Analyses and Reading Rate Measurement
From each subject’s recorded monologue reading,
the identical segment approximately 1 min 30 s in
duration was excerpted for acoustic/rate analyses and
playback to listeners. The following acoustic/rate mea-
sures were made for each speaker: (1) average /s/ dura-
tion, (b) average /s/ peak frequency, (c) LTA speech
spectra (F1, F2, F3), (d) modal speaking frequency and
(e) speech rate (syllables per second).
Prior to actual measurement of average /s/ duration
and /s/ peak frequency for these speakers, 25 /s/ pro-
ductions were identified across the written transcript
of the monologue from all word positions (12 prevo-
calic, 6 intervocalic, 7 postvocalic). These 25 /s/ pro-
ductions were measured for all 9 speakers using a DSP
5500 sonagraph (Kay Elemetrics). The DSP sonagraph
performs signal acquisition, display, and analysis of
speech signals in real-time.
Duration measurements of /s/ were made from a
spectrographic display (300 Hz filter) utilizing a sam-
pling rate of 20,480 samples/s. Time cursors were posi-
Folia Phoniatr Logop 1998;50:35–48 39
147.8.31.43 - 3/25/2017 5:18:00 PM
University of Hong Kong
Downloaded by:
tioned at the beginning and end of each fricative and
the time interval between the cursors was recorded
from the display on the screen. For each speaker, /s/
duration was the average of the 25 /s/ duration values
obtained.
For measurement of the peak /s/ frequency, /s/ pro-
ductions first were bracketed by time cursors from the
DSP spectrographic display. Once bracketed, the aver-
age power spectrum of all points between the time cur-
sors (with 30% overlap of points) was automatically
calculated by the sonagraph and displayed on the
screen. From the averaged power spectrum, the peak
frequency was located visually by the examiner and
marked by a frequency cursor. The frequency location
of the cursor then was recorded. For each speaker, peak
/s/ frequency was the average of the 25 /s/ peak fre-
quencies obtained.
As an indication of test-retest reliability of /s/ dura-
tion and /s/ frequency measures, 3 productions were
randomly selected from each speaker (n = 27) and
remeasured by a second investigator. In all cases, /s/
duration measures were within B 1.6 ms of the record-
ed measurements. Similarly, /s/ frequency measures
were within B 80 Hz of recorded measures.
LTA speech spectra were obtained using the LTA
program of the DSP sonagraph. Specifically, a contin-
uous passage of each speaker’s monologue was input
into the DSP, starting at the beginning of the mono-
logue reading. The LTA program calculates a contin-
uous average power spectrum across voiced segments
of a passage of continuous speech for a period of up to
90 s. Once completed, the averaged power spectrum is
displayed on the screen. Frequency peaks (F1, F2, F3)
were located visually by the examiner, marked by a fre-
quency cursor, and recorded. When LTA spectrum
from 4 speakers (2 gay and 2 straight) were remeasured
by a second investigator, F1, F2 and F3 measures fell
within B 20 Hz of recorded measurements.
Measurement of modal speaking frequency was
accomplished using a Visi-Pitch (Kay Elemetrics). Spe-
cifically, four identical speech segments across each
speaker’s monologue were input to the Visi-Pitch.
Modal speaking pitch was determined from visual
examination of each trace with measurement by cursor
placement at the most frequently recurring frequency
in the trace [36]. For each speaker, modal speaking fre-
quency was the average of these four measurements.
Speech rate (syllables per second) was measured
from two separate segments (approximately 30 sylla-
bles each). One segment was near the beginning of the
expected section of the monologue, the other near the
end. Recordings of the two speech segments were input
to a strip chart recorder (Hewlett Packard 7754A). The
40 Folia Phoniatr Logop 1998;50:35–48
total duration of each segment was computed from
time markers on the tracing and the rate in syllables
per second was calculated [36]. For each speaker,
speech rate was the average of the two speech rate mea-
sures obtained.
Perceptual Tests
A listener tape was prepared containing the mono-
logue reading from each of the 9 speakers, with each
monologue appearing twice for analysis of test-retest
reliability. All items were randomized and preceded by
a practice monologue reading. The practice reading
was produced by a male speaker of the appropriate age
level who did not participate in this study.
Listeners were 25 young adult females ranging in
age from 19 to 30 years of age (mean = 23 years). All
listeners were undergraduate or graduate students in
speech pathology at Marquette University. Young
adult females were utilized as listeners for two reasons:
(1) female students at a private midwestern university
might be viewed as a fairly homogeneous population of
relatively naive listeners, and (2) female students are a
readily available population in a speech pathology pro-
gram. While use of naive listeners might tend to lead to
insignificant findings due to listener inexperience with
the gay community, obtaining significant findings un-
der such circumstances might be interpreted as a stron-
ger argument that acoustic cues to sexual orientation in
males exist.
All listeners passed a standard audiological screen-
ing at 15 dB HL for octave frequencies 250–8,000 Hz
and were native speakers of American English. Based
upon self-reports and results of the Klein Sexual Orien-
tation Grid [34], 24 of the listeners were heterosexual
with limited knowledge or experience with the gay
community. One listener reported an exclusively gay
life-style and predominantly gay sexual behavior.
The listening tape was presented to listeners in
small groups (5 of fewer). Listeners were required to
respond to each monologue by indicating on a re-
sponse sheet whether the speaker was gay or straight.
Listeners were not told what percentage of each group
was included on the tape. If unsure of their response to
a speaker, listeners were told to guess. Listeners were
instructed to listen to the entire monologue before
responding and were told not to interact with others
concerning their responses. The tape was paused brief-
ly after each speaker’s monologue to allow listeners to
respond. The listening tape was presented over a loud-
speaker, at a comfortable loudness level, in a quiet
room.
Linville
147.8.31.43 - 3/25/2017 5:18:00 PM
University of Hong Kong
Downloaded by:
 Statistical Analyses
To determine the relationship between perceived
sexual orientation, actual sexual orientation, and the
acoustic measures under study, simultaneous multiple
regression analyses were performed. Separate analyses
were completed for perceived sexual orientation and
actual sexual orientation. In the analysis on perceived
sexual orientation, the dependent variable was listen-
ers’ perceptual responses expressed on a continuous
scale as the median perceived sexual orientation score
in response to each speaker [37]. In the analysis on
actual sexual orientation, the dependent variable was
actual sexual orientation expressed as a numeric value
(1 = gay; 2 = straight). In both analyses, the indepen-
dent variables were the various acoustic measures.
Results
Perceptual Findings
As a estimate of listener reliability, per-
centages of agreement in response to the first
and second presentation of speakers’ mono-
logues were calculated. Overall, listeners dem-
onstrated 79.1% test-retest agreement.
Overall, listeners correctly identified the
sexual orientation of these speakers 79.6% of
the time from the recorded monologue pro-
ductions, an accuracy rate that is significantly
higher (Sign Test; p ! 0.01) than would be
expected from random guessing [38]. Rates of
correct identification were significantly high-
er (F = 7.51, p ! 0.05) for straight speakers
(93.5%) than for gay speakers (68.4%), al-
though accuracy rates in response to both
speaker groups are significantly higher (p !
0.01; Sign Test) than would be expected from
chance performance.
When subjected to analyses of variance
procedures, differences in listener accuracy in
response to trial 1 and trial 2 of the mono-
logue presentation were not significant, indi-
cating that listeners responded similarly to the
first and second presentation of these speak-
ers. Similarly, differences in listener accuracy
in response to trial 1 and trial 2 for both gay
and straight speaker groups individually were
Sexual Orientation and Speech in Men
not significant, indicating that listeners re-
sponded similarly to the first and second pre-
sentations of these speakers regardless of
speaker sexual orientation.
Perceptual data also were examined to de-
termine if listeners exhibited a response bias.
Listeners did demonstrate a response bias,
indicated by an incidence of straight re-
sponses (59%) that was significantly higher
(p ! 0.01; Sign Test) than 50%.
Perceptual judgments of individual speak-
ers based on their monologue reading are plot-
ted in figures 1 and 2. All 4 straight speakers
(fig. 1) were perceived correctly by listeners as
straight (p ! 0.05; Sign Test). Similarly, 3 of
the 5 gay (fig. 2) speakers (2, 4, 5) were per-
ceived accurately by listeners as gay (p ! 0.05;
Sign Test). One gay speaker (1) was perceived
as straight (p ! 0.05; Sign Test), and 1 (3) was
ambiguously perceived by listeners.
Acoustic Characteristics of Perceived
versus Actual Sexual Orientation
Simple correlations (r) among perceived
sexual orientation, actual sexual orientation,
and the acoustic measures under investiga-
tion are presented in table 1. Acoustic mea-
sures that showed a correlation (r 1 0.40) with
perceived sexual orientation included /s/ du-
ration (–0.71), /s/ frequency (–0.68), and
modal speaking frequency (0.51). Similarly,
acoustic measures correlating with actual sex-
ual orientation included /s/ duration (–0.81),
/s/ frequency (–0.59), and modal speaking fre-
quency (0.44). Actual sexual orientation cor-
related highly with perceived orientation
(0.87), a finding which is not surprising given
perceptual findings discussed above. Acoustic
measures showing moderate to high intercor-
relations (r 1 0.50) included /s/ duration and
F3 (0.77), F1 and F3 (0.58), F1 and modal
speaking frequency (0.61), and F3 and speak-
ing rate (–0.53).
Folia Phoniatr Logop 1998;50:35–48 41
147.8.31.43 - 3/25/2017 5:18:00 PM
University of Hong Kong
Downloaded by:
 Fig. 1. Listeners’ responses to
individual straight speakers.

Fig. 2. Listeners’ responses to

individual gay speakers.

42 Folia Phoniatr Logop 1998;50:35–48 Linville

147.8.31.43 - 3/25/2017 5:18:00 PM
University of Hong Kong
Downloaded by:
 Table 1. Correlations (r) among acoustic measures, speech rate, actual sexual orientation, and perceived
sexual orientation

Actual /s/ /s/ F1 F2 F3 Rate Modal

orientation duration frequency F0

/s/ duration –0.81

/s/ frequency –0.59 0.14
F1 0.10 0.26 –0.41
F2 0.12 –0.16 –0.02 –0.04
F3 –0.39 0.77 0.01 0.58 –0.01
Rate 0.10 –0.35 0.13 –0.26 0.01 –0.53
Modal F0 0.44 –0.33 –0.20 0.61 –0.41 0.08 –0.07
Perceived
orientation 0.87 –0.71 –0.68 0.24 0.14 –0.37 –0.17 0.51

The multiple regression analysis on per- orientation prediction. In contrast, modal

ceived sexual orientation included as inde-
pendent variables /s/ duration, /s/ frequency,
and modal speaking frequency. Other vari-
ables (F1, F2, F3, and speech rate) were not
included because simple correlations of these
measures with perceived sexual orientation
were low, indicating that these variables are
not particularly associated with listeners’
judgments of sexual orientation. In addition,
some of these measures were highly correlated
with other acoustic measures under investiga-
tion (table 1). Such intercorrelations can re-
sult in variables acting as suppressors in a
multiple regression analysis, causing distor-
tion of findings [39].
The multiple regression on perceived sex-
ual orientation yielded a multiple R of 0.94
(p ! 0.01). That is, 88% of the variance in lis-
teners’ perceptual judgments was accounted
for by these three acoustic measures. Exami-
nation of the Beta weights, which reflect the
independent contribution of each acoustic
measure to perceived sexual orientation, re-
vealed that /s/ duration (Beta = –0.56) and /s/
frequency (Beta = –0.56) both made a sub-
stantial contribution to perceived sexual
speaking frequency (Beta = 0.21) was not
associated with judgments of perceived sexual
orientation after interrelationships among the
acoustic measures were taken into account.
Acoustic and speech rate data for speakers
are presented in table 2. On /s/ duration, gay
speakers ranged from 0.1058 to 0.1197 s
(mean = 0.1114 s); straight speakers ranged
from 0.0854 to 0.1030 s (mean = 0.0951 s).
On /s/ frequency, gay speakers ranged from
5,882 to 7,333 Hz (mean = 6,907 Hz); straight
speakers ranged from 4,722 to 6,882 Hz
(mean = 5,768 Hz). Formant measures for gay
speakers ranged as follows: F1 467 to 540 Hz
(mean = 495 Hz); F2 1,293 to 1,440 Hz (mean
= 1,343 Hz); F3 2,293 to 2,540 Hz (mean =
2,367 Hz). For straight speakers, formant
measures varied as follows: F1 440 to 560 Hz
(mean = 502 Hz); F2 1,233 to 1,513 Hz
(mean = 1,365 Hz); F3 2,060 to 2,440 Hz
(mean = 2,263 Hz). Gay speakers demon-
strated speech rates that ranged from 3.619 to
4.923 syllables/s (mean = 4.205 syllables/s).
Straight speakers ranged from 4.206 to 4.319
syllables/s (mean = 4.277 syllables/s). Modal
speaking F0 in gay speakers ranged from 93 to

Sexual Orientation and Speech in Men Folia Phoniatr Logop 1998;50:35–48 43

147.8.31.43 - 3/25/2017 5:18:00 PM
University of Hong Kong
Downloaded by:
 Table 2. Acoustic and speech rate measures derived from monologue readings of individ-
ual speakers

Speakers /s/ /s/ F1 F2 F3 Rate Modal

duration, s frequency Hz Hz Hz syllables/s frequency
Hz Hz

Gay
1 0.1134 5,882 540 1,293 NA 3.828 114
2 0.1077 7,026 467 1,300 2,293 4.923 102
3 0.1058 7,333 473 1,440 2,340 4.055 93
4 0.1106 7,298 507 1,340 2,293 4.599 97
5 0.1197 6,994 487 1,340 2,540 3.620 97
Mean 0.1114 6,907 495 1,343 2,367 4.205 101
Straight
1 0.0854 6,635 440 1,473 2,060 4.391 97
2 0.0916 6,882 527 1,240 2,320 4.206 136
3 0.1002 4,722 480 1,233 2,233 4.237 102
4 0.1030 4,832 560 1,513 2,440 4.273 111
Mean 0.0951 5,768 502 1,365 2,263 4.277 112

114 Hz (mean = 101 Hz). Values in straight
speakers ranged from 97 to 136 Hz (mean =
112 Hz).
The multiple regression analysis on actual
sexual orientation included /s/ duration, /s/
frequency, and modal speaking frequency. As
was the case for the perceived sexual orienta-
tion analysis, the remainder of the acoustic
measures (F1, F2, F3, and speech rate) were
not included because of low correlations with
actual sexual orientation and, in some cases,
fairly high intercorrelations among variables
(table 1).
The multiple regression analysis on actual
sexual orientation resulted in a multiple R of
0.95 (p ! 0.01), indicating that 90% of the
variance in actual sexual orientation was ac-
counted for by these three acoustic variables.
Examination of the Beta weights revealed that
/s/ duration (Beta = –0.71) was the best pre-
dictor of actual sexual orientation, with /s/
frequency (Beta = –0.47) also contributing
significantly to actual sexual orientation pre-
diction. As was the case for perceived sexual
orientation, modal speaking F0 (Beta = 0.11)
was not associated with actual sexual orienta-
tion once interrelationships among variables
were taken into consideration.
Discussion
Results of this investigation suggest that
listeners are capable of making accurate judg-
ments regarding male speakers’ sexual orien-
tation from speech samples. Indeed, these
findings are in agreement with those of Gau-
dio [15], who concluded that listeners general-
ly agree in their judgments of what speech
sounds ‘gay’, and that listeners’ judgments
tend to be correct with respect to identifying
male voices as belonging to gay or straight
men.
The fact that all the listeners in this study
were female could have been a factor in the
accuracy rates demonstrated. Previous re-

44 Folia Phoniatr Logop 1998;50:35–48 Linville

147.8.31.43 - 3/25/2017 5:18:00 PM
University of Hong Kong
Downloaded by:
search has suggested that women are more
sensitive than men to speech behavior [6, 40].
It is possible that a group of male listeners
would not have achieved the level of accuracy
observed in this study. However, it also is pos-
sible that male listeners would have demon-
strated greater accuracy than these female lis-
teners due to increased sensitivity to sexual
orientation issues in men. Gaudio [15] de-
scribed his listeners simply as ‘undergraduate
volunteers’. If one presumes that Gaudio’s
[15] group of 13 listeners consisted of both
males and females, it might be concluded that
listener gender is not a significant factor in
judgments of sexual orientation from speech.
A definitive answer to this question awaits
further research.
Listeners in this study demonstrated a bias
toward identifying speakers as straight. This
finding may reflect a reluctance on the part of
listeners to attach a label that they may have
been felt to be stigmatizing onto a speaker. It
also is possible that this finding reflects the
relative naivete of this group of listeners re-
garding gay individuals or the gay communi-
ty. It is possible that a group of gay listeners
would demonstrate different response pat-
terns, eliminating any response bias, or possi-
bly demonstrating a different bias.
The fact that 1 gay speaker in this study
was consistently identified by listeners as
straight indicates that gay men differ in the
extent to which they demonstrate markers of
‘gayness’ in their speech. This is not a particu-
larly surprising finding given what is known
concerning diversity within the gay communi-
ty. Factors such as social status, occupation,
education, frequency of bar attendance, in-
volvement in a monogamous relationship,
and apparent masculinity-femininity have
been found to vary considerably across the
male homosexual community [41–43]. Al-
though speakers in this study were not prese-
lected for the variables mentioned, and none
Sexual Orientation and Speech in Men
of these speakers presented as markedly ‘mas-
culine’ or ‘feminine’, differences in variables
such as these across the speakers might ac-
count for differences in speech patterns which
produced errors in judgments of sexual orien-
tation. Another possible explanation for a gay
speaker being identified as straight by listen-
ers is provided by Hayes [44]. In discussing
language use among gay men, Hayes [44]
noted that in some settings gay men are covert
in expressing their gay identity. In these set-
tings they are ‘painfully self-conscious about
the stigma imposed upon them by the straight
world’ [44, p. 46] and are careful to avoid ste-
reotypic language or mannerisms. One cannot
discount the possibility that some speakers in
this study may have adjusted their speech pat-
tern in response to the setting in which these
recordings took place.
Results of the acoustic analysis suggest that
significant /s/ production differences exist in
gay and straight men’s speech. Gay men dem-
onstrated peak frequencies for /s/ that aver-
aged approximately 1,100 Hz higher than
straight men. These findings parallel those of
Avery and Liss [29], who reported that /s/ cen-
ter frequencies of LMS males averaged
680 Hz higher than those of MMS males. Dif-
ferences in the magnitude of the differences
observed in the Avery and Liss [29] study and
the present investigation might be accounted
for by differences in methodology between the
two studies or by the fact that both studies
involved relatively small sample sizes.
Peak frequencies for /s/ are felt to reflect
resonances of the effective portion of the vo-
cal tract, or that portion of the vocal tract
between the point of maximum constriction
(point of articulation) and the lips [30]. Since
an inverse relationship exists between the
length of the effective portion of the vocal
tract and frequency of the peak, it appears
that gay men in this study demonstrated a
vocal tract cavity anterior to the tongue con-
Folia Phoniatr Logop 1998;50:35–48 45
147.8.31.43 - 3/25/2017 5:18:00 PM
University of Hong Kong
Downloaded by:
striction for /s/ that was shorter than the
straight men. Two possible explanations exist
for this finding: (a) gay men demonstrated
oral cavity dimensions that were physically
smaller than straight men, or (b) gay men
tended to position the tongue farther forward
in the oral cavity for production of /s/, effec-
tively shortening the resonating portion of the
vocal tract in front of the tongue constriction.
The gay men in this study did not differ
appreciably in height from the straight men, a
finding which would argue against an expla-
nation for this finding based on physical di-
mensions of the oral cavity. In addition, LTA
speech spectra did not differ significantly in
gay and straight speakers, suggesting that vo-
cal tract dimensions in the two groups were
not notably different. A more plausible expla-
nation might be found in articulatory place-
ment. An acquired pattern of speaking involv-
ing more anterior tongue placement for /s/
production could result in peak frequencies
that are substantially higher than those gener-
ally reported for male speakers. Indeed, the
average peak frequencies for /s/ observed in
the gay men in this study more closely resem-
ble those reported previously for female
speakers [31]. It might be hypothesized that
this articulatory adjustment is acquired as
part of normative expectations of appropriate
behavior in the gay male community. That is,
this articulatory feature may be acquired un-
consciously as a marker of membership in the
gay community. Chambers and Trudgill [45]
discuss use of phonemic variations as a means
of signalling group identity.
Gay speakers also tended to produce lon-
ger-duration /s/ productions than straight
men. This articulatory adjustment, too, may
serve as a marker of community membership.
By increasing the duration of a fricative, pre-
sumably the speaker would be drawing atten-
tion to it, much as increasing syllable duration
is used as a marker of stress in English [46].
46 Folia Phoniatr Logop 1998;50:35–48
In interpreting the results of this study it is
important to keep in mind that this research
focused on a small sample of openly gay men.
These findings need to be replicated on a larg-
er sample of men before firm conclusions can
be drawn regarding the entire gay male popu-
lation.
Further, conclusions from this study may
not appropriately be applied to men who are
not open about their gayness, or to bisexual
men. As noted by Gaudio [15], gay male iden-
tity involves more than sex between men. Gay
male identity is a social phenomenon involv-
ing an entire set of beliefs, attitudes, and tra-
ditions unique to the community or subcul-
ture. The individuals within the gay subcul-
ture have unique words, discourse modes, and
rules for conversational interaction [47].
While the results of this study must be
interpreted cautiously, it does appear that
members of the community of openly gay
men demonstrate a unique speech pattern
that is discernible to listeners. Future research
might elucidate additional features that dis-
tinguish gay men’s speech and/or dialectical
variations among members of the gay com-
munity.
Acknowledgments
I would like to thank Gail Kempster and Michael
Wierzbicki for their assistance with statistical analysis
for this study. Thanks as well to Candace Kirschner,
Jennifer Manek, and Teresa Joerger for their hard
work with data gathering, tape editing, and data analy-
sis. This research was supported, in part, by a faculty
development award from Marquette University.
Linville
147.8.31.43 - 3/25/2017 5:18:00 PM
University of Hong Kong
Downloaded by:
 Appendix A: Text of Monologue Read
by Speakers [35]
(Suddenly stepping back onto the stranger’s foot.)
Oh, excuse me. I’m sorry. I was just trying to duck that
pool cue. Gets pretty crowded in here on a Saturday.
Your foot O.K.? Good.
(He turns away, but then can’t help sneaking a look
at the stranger. He smiles and turns away embarrassed,
then turns to the stranger again.) No permanent dam-
age, I hope. To your foot I mean.... Good.
(He turns away again, but still tries to sneak a look
behind him. He gathers his courage and confronts the
stranger with a huge smile.) Look, the name’s Ed Reiss.
My friends call me Ed. (No response.) I’m Sagittar-
ius.... What’s so funny?.... Oh, well, some people like to
know that stuff. I don’t believe in any of it myself, but I
have done some reading about it. See, I like to know
what’s expected of me. (He stares with a huge grin.)
You have a beautiful smile..... No, really, you do. Can I
get you another beer?
Alright, one Lite coming up. (To bartender) Can I
have a Lite please?
Say, what’s your name?
Terry?
Terry. Well, nice to meet you Terry. (Handing
money to the bartender) Here you go. (Beer to Terry.)
And here you go.
Are you Italian?
Spanish?
Jewish!? I never would have guessed it. Not with
those dark romantic eyes. I don’t remember ever
seeing you here before. I don’t get in that often myself.
I teach over in Brooklyn and so I have to be up and out
pretty early.
No, I’m off tomorrow. You really do have beautiful
eyes.
So.... uh.... Are you here with a friend?
So, you’re unattached. How lucky for me. Where
you from?
Sure, I know where that is.
So... Anyway, I really live upstate. I’ve got a farm
up about an hour from Montreal. That’s my real home.
I spend the weekends there during the school year and
then all of my summer vacation. I own half of Walton
Mountain.
No, that’s really what it’s called. Most of the land is
trees, but there’s a piece of flat farming land with an
old barn and I work a half acre of that. I grow all my
own vegetables. I’ve even got a small vineyard. The
mountains have some great white water for canoeing,
and in the winter it’s a skier’s dream.
You ski?
Would you like to learn?
Alright, it’s a deal. You’ll love the house, it’s really
great. One of those old Victorian farm houses with lots
of ginger-bread and Franklin stoves. My father and I
are restoring it.
Well, my parents winter in Florida but come north
to stay with me in the warmer months. Hey, what’d’ya
say we continue this conversation in the car?

OOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOO

References
1 Colton R, Casper J: Understanding
Voice Problems. Baltimore, Wil-
liams & Wilkins, 1990.
2 Duncan S: Nonverbal communica-
tion. Psychol Bull 1969;72:118–
137.
3 Fisher H: Improving Voice and Ar-
ticulation. Boston, Houghton-Mif-
flin, 1975.
4 Hargreaves W, Starkweather J,
Blacker K: Voice quality in depres-
sion. J Abnorm Psychol 1965;70:
218–220.
5 Hunt R, Lin T: Accuracy of judg-
ments of personal attributes from
speech. J Pers Soc Psychiatry 1967;
6:450–453.
6 Kramer C: Perceptions of female
and male speech. Lang Speech 1977;
20:151–161.
7 Markel N: Relationship between
voice-quality profiles and MMPI
profiles in psychiatric patients. J Ab-
norm Psychol 1969;74:61–66.
8 Markel N, Bein M, Philips J: The
relationship between words and
tone-of-voice. Lang Speech 1973;16:
15–21.
9 Markel N, Meisels M, Houck J:
Judging personality from voice qual-
ity. J Abnorm Soc Psychol 1964;69:
458–463.
10 Moore W: Personality traits and
voice quality deficiencies. J Speech
Disord 1939;4:33–36.
11 Aronovitch C: The voice of person-
ality: Stereotyped judgements and
their relation to voice quality and
sex of speaker. J Soc Psychol 1976;
99:207–220.
12 Moses P: The Voice of Neurosis.
New York, Grune & Stratton, 1954.
13 Lakoff R: Talking Power: The Poli-
tics of Language. New York, Basic
Books, 1990.
14 Travis N: A Study of the Relation-
ship of Certain Variables to Sex
Characteristic Identification from
the Speech of Heterosexual and Ho-
mosexual Individuals; doct diss
Louisiana State University, 1981.

Sexual Orientation and Speech in Men Folia Phoniatr Logop 1998;50:35–48 47

147.8.31.43 - 3/25/2017 5:18:00 PM
University of Hong Kong
Downloaded by:
15 Gaudio R: Sounding gay: Pitch
properties in the speech of gay and
straight men. New Ways of Analyz-
ing Variation in English Conf,
Georgetown University, Washing-
ton, 1991.
16 Kolodny R, Masters W, Hendryx B,
Toro G: Plasma testosterone and se-
men analysis in male homosexuals.
N Engl J Med 1971;285:1170–
1174.
17 Birk L, Williams G, Chasin M, Rose
L: Serum testosterone levels in ho-
mosexual men. N Engl J Med 1973;
289:1236–1239.
18 Tourney G, Hatfield L: Androgen
metabolism in schizophrenics, ho-
mosexuals, and normal controls.
Biol Psychiatry 1973;6:23–36.
19 Hoffman W, Supal C, Tosi O: Com-
puter analyses of acoustical parame-
ters in hypopituitary children before
and after growth hormone treat-
ment. Int J Pediatr Otorhinolaryn-
gol 1984;7:1–9.
20 Maceri D: Head and neck manifes-
tations of endocrine disease. Otolar-
yngol Clin North Am 1986;19:171–
180.
21 Lerman J, Damsté P: Voice pitch of
homosexuals. Folia Phoniatr 1969;
21:340–346.
22 Brown W, Feinstein S: Speaker sex
identification utilizing a constant la-
ryngeal source. Folia Phoniatr 1977;
29:240–248.
23 Coleman R: Male and female voice
quality and its relationship to vowel
formant frequencies. J Speech Hear
Res 1971;19:168–180.
24 Coleman R: Speaker identification
in the absence of inter-subject differ-
ences in glottal source characteris-
tics. J Acoust Soc Am 1973;53:
1741–1743.
48 Folia Phoniatr Logop 1998;50:35–48
25 Coleman R: A comparison of the
contributions of two voice quality
characteristics to the perception of
maleness and femaleness in the
voice. J Speech Hear Res 1976;19:
168–180.
26 Günzbürger D, de Vries M: How do
minor acoustical cues affect male
and female voice quality? Proc Eu-
rospeech 89 Conf on Speech Com-
mun Technol 1989, vol 2, pp 143–
145.
27 Lass N, Hughes K, Bowyer M, Wa-
ters L, Bourne V: Speaker sex identi-
fication from voiced, whispered,
and filtered isolated vowels. J
Acoust Soc Am 1976;59:675–678.
28 Weinberg B, Bennett S: Speaker sex
recognition of 5- and 6-year-old chil-
dren’s voices. J Acoust Soc Am
1971;50:1210–1213.
29 Avery J, Liss J: Acoustic characteris-
tics of less-masculine-sounding male
speech. J Acoust Soc Am 1996;99:
3738–3748.
30 Bennett S, Weinberg B: Sexual char-
acteristics of preadolescent chil-
dren’s voices. J Acoust Soc Am
1979;65:179–189.
31 Hughes G, Halle M: Spectral prop-
erties of fricative consonants. J
Acoust Soc Am 1956;28:303–310.
32 Schwartz M: Identification of speak-
er sex from isolated, voiceless frica-
tives. J Acoust Soc Am 1967;43:
1178–1179.
33 Ingemann F: Identification of the
speaker’s sex from voiceless frica-
tives. J Acoust Soc Am 1968;44:
1142–1143.
34 Klein F: Sexual orientation as a
multivariate dynamic process; in
McWhirter D, Sanders S, Reinisch J
(eds): Homosexuality/Heterosexu-
ality Concepts of Sexual Orienta-
tion. Kinsey Inst Ser. New York,
Oxford University Press, 1990.
35 Fierstein H: Torch Song Trilogy.
New York, Villard Books, 1979.
36 Linville SE, Skarin B, Fornatto E:
The interrelationship of measures
related to vocal function, speech
rate, and laryngeal appearance in el-
derly women. J Speech Hear Res
1989;32:323–330.
37 Linville SE, Fisher H: Acoustic char-
acteristics of perceived versus actual
vocal age in controlled phonation by
adult females. J Acoust Soc Am
1985;78:40–48.
38 Silverman F: Research Design in
Speech Pathology and Audiology.
Englewood Cliffs, Prentice-Hall,
1977.
39 Cohen J, Cohen P: Applied Multiple
Regression/Correlation Analysis for
the Behavioral Sciences. New York,
Wiley, 1975.
40 Labov W: Sociolinguistic Patterns.
Philadelphia, University of Phila-
delphia Press, 1972.
41 Harry J: Urbanization and the gay
life. J Sex Res 1974;10:238–247.
42 Harry J, DeVall W: Age and sex-
ual culture among homosexually
oriented males. Arch Sex Behav
1978;7:199–209.
43 Weinberg M, Williams C: Male ho-
mosexuals: Their Problems and Ad-
aptations. New York, Oxford Press,
1974.
44 Hayes J: Gayspeak; in Cheesbrov
(ed): Gayspeak. Cleveland, Pilgrim
Press, 1981.
45 Chambers J, Trudgill P: Dialectolo-
gy. New York, Cambridge Universi-
ty Press, 1980.
46 Borden G, Harris K: Speech Science
Primer. Baltimore, Williams & Wil-
kins, 1984.
47 Moran J: Language use and social
function in the gay community;
Master’s thesis Georgetown Univer-
sity, Washington, 1990.
Linville
147.8.31.43 - 3/25/2017 5:18:00 PM
University of Hong Kong
Downloaded by: