Also, the formatting is screwed up. For now I just want the information online, but I am spending a little bit of time trying to make it readable.
The Role of /s/ Duration as a Perceptual Cue for Gay-Sounding Male Speech
by Jenny Palmer
by Jenny Palmer
A thesis submitted to the faculty of the University of North Carolina at Chapel Hill in partial fulfillment of the requirements for the degree of Master of Arts in the Department of Linguistics.
Chapel Hill 2002
Approved by:
_________________________________ Chip Gerfen, Advisor
_________________________________ Craig Melchert
_________________________________ Jennifer Smith
©2002 Jenny Palmer ALL RIGHTS RESERVED
ABSTRACT
JENNY PALMER: The Role of /s/ Duration as a Perceptual Cue for Gay-Sounding Male Speech (Under the direction of Chip Gerfen)
This thesis is an experimental analysis of the role that /s/ duration plays in how listeners perceive male sexual orientation based on speech. With listener responses measured as both a categorical (forced choice) response and a continual mean ‘gayness’ score, listeners’ perception of a man as gay increased substantially with the longer /s/ durations in word-initial, stressed /skV/ and /spV/ environments. Listener participants heard one of 3 /s/ durations of a man whose sexual orientation had been perceived as neutral. ANOVA analysis showed that listeners who heard the longer /s/ durations perceived the man as sounding "gayer." In addition, multiple regression analysis showed that listeners who heard the longer /s/ durations were significantly more likely to judge the speaker as sounding "gay."
ACKNOWLEDGEMENTS
I could not have finished this thesis without the help and support of many people. Firstly, to all my parents and family for their support, monetary and otherwise. Your encouragement throughout graduate school has meant a lot to me.
Daniel Cordaro and Kirk Baker, you provided an essential service in helping me with the planning and implementation of the /s/ duration manipulation. Thanks to Chris Weisen at the ODUM Institute for Social Science Research. I finally understand as much statistics as I need! I could not have run the SAS program without your programming and analysis skills.
Chip, qué se puede decir? Sin tu ayuda, apoyo, interés y enseñanza, no habría podido acabar este trabajo. Gracias por creer en mí. Te echaré de menos. Te prometo que nunca más tendrás que sufrirme rompiéndome a llorar en tu despacho. Y dime si me he equivocado de gramática, ya que estoy convencida que ya no puedo comunicarme en castellano!
Rusty, you have been wonderful all along the way. Thanks for all the last minute help you have given me, even after I was no longer your responsibility! To Craig and Jen, thanks for being on my committee. To Megan, because I miss you, and you supported me during some really rough times. To Randy, for his experimental design input and overall support of me during those same rough times previously mentioned. (Thank GOD they're over!)
To my friends in North Carolina, who helped me keep balance in my life so that I did not go (too) crazy: thanks for your participation, the brainpicking sessions, the coffee and the nights out dancing our cares away. I know it was a tight deadline, but I got it done, thanks to your help.
Martha, if anyone has been my anchor over the last three years, you have. You have helped keep me sane and happy, most of the time, anyway!
Last but not least, to Ols. Stephanie, if you were a man I'd marry you! I can't even begin to tell you what you mean to me. You are my best friend and I consider myself a very lucky women for it. I wish everyone could have as amazing a person as you in their lives.
CONTENTS
LIST OF TABLES
LIST OF FIGURES
INTRODUCTION
LITERATURE ON GAY-SOUNDING MALE SPEECH CHARACTERISTICS
2.1 Gay men as "effeminate"
2.2 Speech differences in men both in perceived and actual sexual orientation
3. METHODOLOGY
3.1 Experimental design and data collection
3.1.1 Speakers
3.1.2 Listener group #1
3.1.3 Manipulation, word lists and CDs
3.1.4 Listener group #2
4. RESULTS
4.1 Mean gayness ratings
4.1.1 Statistical analysis
4.1.1.1 Gay/straight variable
4.1.1.2 Distracter variables
4.2 Forced choice
4.2.1 Statistical analysis: forced choice
4.2.1.1 Gay/straight variable
4.2.1.2 Distracter variables
5. DISCUSSION
6. APPENDIX ONE: Information form: Listener group #1
7. APPENDIX TWO: Information form: Listener group #2
8. APPENDIX THREE: Information form: Speakers
9. APPENDIX FOUR: Consent form: Speakers
10. APPENDIX FIVE: Background questionnaire: Speakers
11. APPENDIX SIX: Background questionnaire: Listeners (Group #1 and #2)
12. APPENDIX SEVEN: Perception study questionnaire: Study #1
13. APPENDIX EIGHT: Perception study questionnaire: Study #2
14. APPENDIX NINE: Text
REFERENCES
LIST OF TABLES
Table 1. Mean perceived sexual orientation by listener and /s/ duration: 1=sounds gay; 6=sounds straight...........................................................................24
Table 2 Forced choice data by sex of listener......................................................25
LIST OF FIGURES
Spectrogram of /spa/ (/s/ + 0 ms.)
Spectrogram of /spa/ (/s/ + 30 ms.)
Spectrogram of /spa/ (/s/ + 60 ms.)
Percentage of Female and Male subjects, Second Listener Perception Study
Degree of gayness means
1. INTRODUCTION
When people hear a voice, they assess the speaker’s makeup in many ways. Speech allows a listener to make many judgments about the speaker’s education, origin, ethnicity and social class. Such judgments are based on both stereotypes and inferences made from past experience with individuals’ speech.
Many people claim that they can identify the sexual orientation of male speakers based on their speech. When informally probing listeners as to how they decide whether a man sounds gay, such people tend to rely on stereotypes, mentioning characteristics like “lisping” or a “sing-songy” voice. Frequently listeners feel like they know “the voice” well enough to mimic it. How does a person identify (correctly or incorrectly) the sexual orientation of a male speaker based on only his speech? Are listeners focusing on lexical items, or are there particular phonetic cues that make this possible? Do listeners use their stereotypes of men's vs. women's speech as a basis for making judgments about male sexual orientation? Does the speech of men who self-identify as gay differ in a measurable way from men who self-identify as straight? These questions are far from being answered completely, but some inroads have been made to date.
My intent for this thesis is to confirm that a longer /s/ duration provides perceptual information to listeners that in turn increases the likelihood that they will perceive a man as sounding gay based on his speech. This thesis is organized as follows:
Chapter 2 will discuss the literature that has examined the role of gay and gay-sounding speech and how it may or may not relate to women's speech. In addition, Chapter 2 provides a description of the similarities and differences between the speech of a) gay and straight men and b) gay-sounding and straight-sounding men. Chapter 3 consists of a full explanation of the experimental design and data collection phases of the research I have done to examine the role of /s/ duration as it relates to perceived sexual orientation. In Chapter 4 I report the results of that research, including a description and statistical analysis of the data. Finally, in Chapter 5 I discuss my findings and possible directions for future research.
2. LITERATURE ON GAY-SOUNDING MALE SPEECH CHARACTERISTICS1
Any understanding of phonetic characteristics of gay-sounding male speech is incipient at best. To date, there have been a few small studies that have tried to pin down phonetic markers in the speech of gay-sounding men, with varying degrees of success. The direction of research has evolved over time, from looking at gay men as displaying characteristics of women's speech, to looking at measurable speech differences between men both as a function of perceived and actual sexual orientation. This chapter attempts to give an overview of the literature that is pertinent to the discussion of /s/ duration as a perceptual cue for gay-sounding male speech.
2.1 Gay men as 'effeminate'
The study of masculine versus feminine speech is relevant to the topic of gay- sounding male speech. There is a stereotype connecting women's speech to gay-sounding male speech that speakers use to make judgments about a man's sexual orientation. Robin Lakoff (1975) observed that gay men, like women, were more likely to use specific color terms such as mauve or teal, whereas straight men tended to use less specialized terms such as purple or blue to refer to the same colors. Lakoff made her observations based on stereotypes (Bucholtz and Hall 1995), but she was not alone. The concept of gay men as exhibiting speech characteristics found in women is one that others have pursued. This “effeminate-sounding man” approach to studying the speech of gay men yielded mixed results. While some findings upheld the notion that at least some of the features found in these men's speech did in fact reflect aspects found in women’s speech (Taylor 1996), others failed to find a relationship between gay men's and women's speech (Terango 1966, Fai 1988).
Terango (1966) focused on the concept of femininity and masculinity in men’s speech with the goal of “correcting” the speech of “effeminate” sounding men. He expected to find the effeminate-sounding men to have voice characteristics similar to women’s. Terango first pre-selected his subjects by asking his colleagues to refer to him male students that they had noticed to have “especially effeminate or masculine speech” (Terango, 590). After recording the subjects, he ran a listener perception test and selected the subjects who had turned up as masculine-sounding and effeminate-sounding. Terango then studied the fundamental frequencies and some pitch range qualities of the speakers, comparing the two groups. He expected the effeminate-sounding men to have a higher fundamental frequency, similar to that of women. However, Terango found that effeminate sounding men had a lower Fo (127 Hz) than unselected American males (130 Hz). The masculine-sounding men had an Fo of 100 Hz., lower than the effeminate- sounding men. His conclusion was that even a man with a lower Fo can sound effeminate. Most of the other measurements he took, such as mean total pitch range and mean functional pitch range, also failed to differentiate the two groups. Terango did find one difference between the two groups; the effeminate-sounding group had a higher mean pitch change during inflections than did the masculine-sounding group. However, he did not show any evidence that this quality had been shown to be typical of women’s speech.
Avery and Liss (1996) also pursued characteristics of speech in men perceived as sounding effeminate, preferring the term “less-masculine-sounding.” Their hypothesis was that less-masculine-sounding (LMS) men would display some of the speech characteristics found in women. They began with 35 men who had recorded a monologue. To establish which of these men would fit the LMS and more-masculine- sounding (MMS) profile they wished to test, 35 women participated in a paired- comparison task. In this task, the women heard first four sentences from the recorded monologue of one man (“Bob”) who the two investigators had deemed as the speaker who best represented the average speech and voice qualities of the sample. After hearing Bob, the women heard the same four sentences repeated by one of the other 34 speakers and judged the new speaker’s voice along a continuum as more or less-masculine sounding than Bob. This was repeated; listeners hear Bob, listeners hear a new speaker, listeners judge new speaker as compared to Bob, until all of the speakers had been heard and judged. The 8 speakers who were chosen for the acoustic analysis were selected out of the 19 that fell closest to the poles of the continuum, after their fundamental frequencies were determined to be similar to one another.
One of the characteristics Avery and Liss examined was the acoustic profile of /s/ for the LMS and MMS subjects, based in part on prior findings that men and women can be identified by gender using isolated voiceless fricatives as a stimulus (Ingemann 1968, Schwartz 1967). Avery and Liss measured the /s/ peak frequency, /s/ skew2 and /s/ kurtosis3 in the 8 male speakers. What they found was significant differences between the LMS and MMS groups in the /s/ peak frequency, and the skew of the /s/. However, kurtosis differences were not significant.
2.2 Speech differences in men both in perceived and actual sexual orientation
Another approach has been to look at gay men as compared to straight men, with no reference to women’s speech. Lerman and Damsté (1969) investigated the speech of gay men and straight men, with the hypothesis that the fundamental frequencies in gay men would be higher than those of straight men. Their findings did not support their hypothesis. Not one of their measures showed significant differences between the average fundamental frequencies of the two groups.
Along the same lines, Gaudio (1994) published a study looking at acoustic correlates of perceived and actual sexual orientation. Eight subjects, four gay and four straight, recorded two different passages. He played both passages to 10 female and 3 male undergraduate volunteers. After hearing the stimuli, the 13 listeners rated the speakers along a continuum of straight to gay, 1 to 7. Gaudio stipulated 4 to be a neutral judgment. Overall, the listeners were able to identify the speaker’s sexual orientation based on the two recordings. Gaudio then looked at the pitch range and the pitch variability of each group along the straight/gay perceived scale and concluded that though some correlations were suggestive, neither of these acoustic measures alone was enough to make listeners perceive a man as sounding gay.
Gaudio’s research was important for the acoustic study of gay male speech in that it was the first study that looked at the perception of gay-sounding male speech, possible acoustic correlates that marked it, and its relation to actual sexual orientation. Other researchers such as Terango had approached male speech along a masculine/feminine dimension. Lerman and Damsté looked at sexual orientation alone. Gaudio’s 1994 study pioneered socio-acoustic research of gay-sounding speech in males, though the concept of research to identify effeminate-sounding speech as a pathology continued (Avery and Liss 1996).
Though Gaudio’s findings did not reach significance, they are suggestive enough to have inspired other acoustic studies. Linville (1998) also explored acoustic differences in men’s speech as a function of both perceived and actual sexual orientation. One of the acoustic measures she focused on was peak frequency and duration of /s/. She began with 9 males, 5 gay and 4 straight, who recorded a monologue from the play Torch Song Trilogy, which had been adapted to make all references to other characters gender- neutral. For her perceptual study, Linville had her listeners hear each monologue twice, in random order, after hearing a practice monologue read by a male speaker of the same approximate age of the other speakers, but who was not part of the study. Listeners made a forced choice judgment of each speaker’s sexual orientation, indicating whether the speaker had sounded gay or straight, based only on his monologue. The listener response accuracy rate was significantly higher than random guessing, with an asymmetry in this accuracy rate. Straight men were perceived as straight more frequently than gay men were perceived as gay. When Linville measured the speech of the gay-sounding participants, their average /s/ durations were significantly longer than the speakers judged as straight-sounding, and their average /s/ peak frequencies were significantly higher (Linville 1998). This shows indirect evidence that the acoustic profile of /s/ plays some role in how people perceive sexual orientation in men based on the voice.
More indirect evidence comes from Crist (1997), who looked to stereotypes of gay-sounding speech as a way to test whether onset consonants lengthened when his subjects performed what they understood to be a “queeny” voice. The subjects read a passage first in their natural voice, and then trying to imitate a gay-sounding voice. Crist then analyzed certain segments of the passage they read, comparing the tokens from the natural speech group with the tokens from the stereotyped speech. He ran two experiments, having found in the first that the /s/ in stressed /#skV/ tokens were significantly longer and that the /s/ in stressed /#spV/ tokens were suggestively longer for each of the three subjects when performing the stereotyped speech. He found no such difference in the /s/ duration of /#stV/ tokens and gave some possible explanations for these seemingly contrary results4. Crist had few tokens of the /sp/ duration type, however, so he ran a second experiment to check the pattern, rewriting the script to include more tokens under study. The /s/ duration in the stressed /#spV/ environment was significantly longer for two of the three new participants. The third participant was somewhat problematic, in that his data ran contrary to expectations for both /sk/ and /sp/ environments. However, as Zwicky (1997) points out, speakers might draw from any combination of many different cues that could identify a gay-sounding voice for production, while listeners likely recognized them all. Perhaps this man was not employing the cue in his stereotype performance that Crist was studying. The /s/ duration in the stressed /#skV/ environment was significantly longer for one of the speakers, and suggestive for another. Though only one of the participants had significant differences in the /sk/ environment in Crist’s second experiment, the first experiment’s results for /sk/ had been strong, and the second experiment had been aimed more at getting further /sp/ tokens.
Perception studies of gay-sounding male speech have to this point dealt mainly with perceived sexual orientation compared to actual sexual orientation. Listeners have tended to perceive a speaker’s sexual orientation accurately. For instance, in Gaudio’s 1994 perception study, all of the self-identified straight men were perceived as straight after listeners heard two different reading passages. In the same study, three of the four self-identified gay men were perceived as gay after listeners heard one passage, and all four were perceived as gay after hearing the second passage. One gay man was not perceived as gay; instead, listeners perceived his speech ambiguously. Linville (1998) showed a listener response accuracy rate of 79.6% overall, which was significantly higher than random guessing. There was an asymmetry in this accuracy rate, as straight speakers were perceived as straight 93.5 percent of the time, while gay speakers were perceived as gay with only 68% accuracy.
Though these two studies have shown that listener’s perception of sexual orientation based on speech alone has tended to agree with the speakers’ self-identifications of sexual orientation, the cues that allow the listeners to make this judgment have not been well identified. In my experiment, I aim to test directly whether perception of gay-sounding male speech changes when the /s/ duration varies.
The acoustic profile of some sibilants (especially /s/) is different in certain male populations, as shown in previous studies of men perceived to sound gay (Linville, 1998) or who were performing their idea of a “gay” stereotype (Crist, 1997). Linville found that the men perceived as gay had a significantly longer /s/ duration than the men perceived as straight. Also, Linville determined from beta weight analysis that compared to the possible acoustic cues she studied, /s/ length contributed substantially to perceived sexual orientation and that /s/ duration was the best predictor of actual sexual orientation. Linville did not control for /s/ environment. Crist’s study was not a perception study; rather, his subjects were asked to read a text in two different styles. First Crist asked his participants to record the text using their natural voice. In addition, they performed the same text using their version of stereotyped gay male speech. The stereotyped speech performed had significantly longer /s/ duration in /#skV/ and /#spV/ stressed syllables compared to the /s/ duration of the subjects’ natural speech in the same tokens. Linville showed that /s/ duration is longer in men perceived as gay, but not that the duration drives perception.
The indirect evidence found by Linville and Crist that gay-sounding men have significantly longer /s/ durations prompted me to test the hypothesis that a longer /s/ will change a listener’s percept of a man’s sexual orientation. In this paper I will show that as an /s/ duration gets longer (by 30 ms. intervals up to 60 milliseconds longer than a speaker’s natural /s/ length), listeners will perceive a man as sounding more gay, and more listeners will perceive the speaker as gay.
1 Because of the different terminology used throughout the literature, I will assume “gay- sounding”, “less-masculine sounding”, “effeminate-sounding” and “gay male stereotyped” to be analogous terms. In the same way I will assume “straight-sounding”, “more-masculine sounding”, and “masculine-sounding” also to be analogous. Any reference to gay or straight speech will refer to the speech of men who identify as gay or straight with regards to their actual sexual orientation.
2 Skew is a statistical measure of the third spectral moment of /s/ and shows the asymmetry of the distribution relative to the mean (Coughlan, Lawton and Weinreb 1993).
3 Kurtosis is a measurement of the fourth spectral moment of /s/ and shows the peakedness or flatness of the energy distribution relative to a gaussian distribution(Coughlan, Lawton and Weinreb 1993).
4 In addition to Crist's possible explanations (e.g. the possibility that an answer might come from the fact that /s/ and /t/ share a common place of articulation or that /t/ is coronal), I would suggest looking to the total duration of the onset cluster. The entire /#stV/ cluster might lengthen in the performed stereotype speech instead of just the /s/.
3. METHODOLOGY
3.1 Experimental design and data collection
To test whether as an /s/ duration gets longer listeners perceive a man's speech as sounding more gay., I have examined the possible role of /s/ duration in word-initial, stressed-syllable /spV/ and /skV/ clusters (such as spool and school), following Crist. The project is comprised of two phases: the digital recording of male speech and a two-part listener perception task. First a group of listeners identified one male speaker among eight whose speech was perceived as neither gay- nor straight-sounding (based on the mean of listener's judgments). Following the identification of a neutral-sounding subject, new listeners heard word lists extracted from the neutral speaker’s recording. Each participant heard one of 3 possible /s/ durations. One group of subjects (Group A) heard the natural speech of the neutral speaker. A second group (Group B) heard the same recordings as Group A, with one change: the duration of each word-initial /s/ was manipulated, making each /s/ sound 30 ms. longer than the original length. The recordings Group C heard differed from those of Groups A and B only in that the /s/ sounds were 60 ms. longer than the speaker’s natural /s/ durations.
3.1.1 Speakers
Nine men were recruited to participate in the study as the speakers. Before recording, I told the speakers that they would be participating in a study about how people listen to language and that listener participants would later hear their speech and judge the recorded passage for the speaker’s sociological makeup. In addition, I told speakers that their speech might be chosen for manipulation for a further study. The speakers read the information sheet (see Appendix Three) , and I asked if they consented to participate. After each participant consented, I assigned him a speaker number, which he was instructed to keep for any further correspondence with me. All speaker participants were between 19 and 25 years old and either UNC graduate students or not enrolled as students at UNC. None had any speech therapy, extensive public speaking or voice training. All were native speakers of American English. The speakers did not disclose their sexual orientation for this task (see Appendix Five).
The task given to the speakers was to read a short passage. The text (see Appendix Nine) was designed to approximate a casual speech style, and incorporated many tokens of word-initial, stressed-syllable /spV/ and /skV/ clusters (such as in the words spool and school)1. Each participant was given a few minutes to familiarize himself with the text, and then he read the passage once into a TECT UEM-801 unidirectional, cardioid microphone in a soundproofed room. The speakers were recorded in 8-bit2 monaural sound with a 22,500 Hz sampling rate, using Sound Forge 5.0. Eight of the nine subject’s recorded passages3 were then compiled on an audio CD for listeners.
After each participant recorded, I debriefed him as to the exact nature of the study, explaining that the listeners would be asked specifically how each speaker sounded with respect to age, gender, education level, socioeconomic group and sexual orientation. I explained that this deception was necessary to avoid the speakers consciously or subconsciously changing the way they speak. After I debriefed the speakers, I gave them a written consent form to sign and one to keep for their records (see Appendix Four). This was a two-stage consent process, oral consent before the recording process and written consent after the debriefing.
3.1.2 Listener group #1
Twenty participants were recruited from the undergraduate population of UNC to listen to the eight recordings. Listeners were 18-25 years old, native speakers of American English, and reported no prior hearing problems or audiological disorders (see Appendix Six).
Before participation, I advised listeners that they would be taking part in a study about how people listen to language, and that they would hear several speakers who had recorded a passage. I gave them an information sheet to read (see Appendix One) and then asked if they wished to participate. After obtaining oral consent, I then assigned each a listener number, to keep for any further correspondence with me.
Working at their own pace in a quiet room, listeners heard the recordings using Optimus Nova-70 headphones and randomized with Microsoft® CD Player 4.0. No more than 2 listeners participated at any one time. For each voice they heard, listeners were asked to answer a questionnaire about their perceptions of the speakers, based only on the voices they heard. The questionnaire included questions about the speakers’ perceived socioeconomic status, gender, education level, sexual orientation and age (see Appendix Seven). All but the sexual orientation rating were distracters. The task was to circle the choice along a continuum (1 to 6, where 1 = gay and 6 = straight) that best represented the listener’s opinion about the speaker based on his voice. Immediately after they made each judgment, listeners also indicated how confident they were of their answer, along a continuum of 1 to 6. An answer of one meant that listeners were not at all confident in their response, whereas a 6 meant that they were very confident of their answer.
After the task was completed, I debriefed the listeners as to the true intent of the study, being careful to explain thoroughly in case any wished to drop out. Once the responses were collected, I found the mean listener response for each speaker. One voice that was perceived as neither particularly gay- nor straight-sounding (3.105, where the midpoint of the scale equaled 3.5) was manipulated for perception study #2. The next closest speaker mean was 4.315, approximately twice as far from the midpoint as the speaker chosen4.
3.1.3 Manipulation, word lists and CDs
After the neutral voice was determined, the token words were extracted from the speaker’s previously recorded text, using Cool Edit ’96. In the same program, the word initial /s/ sounds were extended to the desired length of +30 milliseconds and +60 milliseconds, by copying and pasting from the center of the natural /s/ sounds. Then, the WAV files of each token were grouped into 3 word lists of 5 words5 and joined with 2 seconds of generated silence between tokens, using Sound Forge. All sound files were converted to 16-bit depth and 44,100 Hz sample rate for CD burning. Three CDs were created, one with the word lists whose natural /s/ durations had not been lengthened by manipulation (Group A), one with the word lists whose /s/ durations were extended by 30 ms. (Group B), and one with the word lists whose /s/ durations were extended by 60 ms. (Group C). As such, there were 3 CDs each, whose content differed only in the length of the /s/ occurrences in the /#skV/ and /#spV/ stressed syllable environments. Each CD was tested and found ready for use in the next part of the listener perception study.
Figure 1. Spectrogram of /spa/ (/s/ + 0 ms.)
Figure 2. Spectrogram of /spa/ (/s/ + 30 ms.)
3.1.4 Listener Group #2
84 participants were recruited from the undergraduate population of UNC to listen to one of the 3 CDs. Listeners were 18-25 years old, native speakers of American English and reported no prior hearing problems or audiological disorders (see Appendix Six).
Figure 4. Percentage of Female and Male subjects, Second Listener Perception Study: Males:29%, Females: 71%
Before participation, I advised listeners that they would be taking part in a study about how people listen to language and that their task would be to listen to 3 tracks on a CD and report certain information about how they perceived the speaker based on the voice. I gave them an information sheet to read (see Appendix Two) and then asked if they wished to participate. After obtaining oral consent, I then assigned each a listener number, to keep for any further correspondence with me and to ensure anonymity.
Male Female
Each subject was placed in one of the 3 listener groups, A, B or C. Group A listeners heard the natural voice of the speaker. Group B listeners heard the /s/ + 30 ms. recordings, and Group C listeners heard the /s/ + 60 ms. recordings. Group assignment was based on gender of the participant, in order to maintain listener groups that had the same male/female ratios. Each group had 28 listeners, with 20 female participants and 8 male participants.
Working at their own pace, listeners heard the recordings in random order using Optimus Nova-70 headphones and Microsoft® CD Player 4.0. For each of the three word lists they heard, listeners were asked to answer a questionnaire about their perceptions of the speaker, based only on the voice in the word lists they heard. The questionnaire (see Appendix Eight) included questions about the speakers’ perceived socioeconomic status, education level, sexual orientation and age6. All ratings other than the sexual orientation rating were distracters. The task was to circle the choice along a continuum (1 to 6, where 1 = gay and 6 = straight) that best represented the listener’s opinion about the speaker based on his voice for each of the 3 tracks on the CD. Immediately after they made each judgment, listeners also indicated how confident they were of their answer, along a continuum of 1 to 6. An answer of one represented that the listeners were not at all confident in their responses, whereas a 6 meant that they were very confident of their answers.
1 The speakers reported that the content of my passage sounded "like a girl," perhaps because in writing it I let my own (female) voice through too much. I wanted to use Crist’s more neutral burning building text, but the unfortunate timing of my September 11, 2001 Institutional Review Board meeting compelled me to write my own text, as the IRB committee deemed that the content of Crist’s text could harm the participants by reminding them of the horrific events the country witnessed that day.
2 8-Bit was the default for recording in Sound Forge.
3 I excluded one of the nine speaker participants, because I recruited listeners from the Introduction to Linguistics course he taught a recitation for. I did not want to risk the listeners recognizing his voice and thus having a visual reference which could affect the results of this study.
4 The speaker's mean scores on the gay/straight continuum are as follows: Speaker 1=2, Speaker 2= 2.526316, Speaker 3= 4.526316, Speaker 4= 4.368421, Speaker 5= 4.315789, Speaker 6= 3.105263, Speaker 7= 4.684211 and Speaker 8 = 4.631579
5 Track 1 of each CD contained the following tokens, in this order: sketchy, Spanish, scam, Scott and spa. Track 2: space, skylight, spit, schedules and spare. Track 3: spicy, schedule, Spam, skeptical and speak. Each track contained both /spV/ and /skV/ tokens and were ordered so that the /spV/ and /skV/ tokens were not grouped together.
6 The male/female continuum that I used in the first listener perception test was not included in the second. I dropped this distracter because the standard ratio of distracter questions to relevant questions is 3:1.
4. RESULTS
4.1 Mean gayness ratings
Each of the eighty-four listeners made 3 judgments for each variable (gay/straight, old/young, educated/non-educated and wealthy/poor) and the confidence continuum corresponding to each variable. I took the mean of each listener’s responses for statistical analysis. The continuum had an even number of possible responses, so that no listener could choose a neutral response to any of the recordings.
Figure 5. Degree of gayness means: /s/= 4.029; /s/ +30 ms=3.941; /s/ +60 ms. = 3.345
/s/
+ 30 +60 ms. ms.
1 2 3 4 5 6
Mean "gay" rating
Listeners tended to perceive a man as sounding gayer as the /s/ duration increased. At the /s/ duration of the natural speech, listeners perceived the man’s degree of gayness at 4.02891 (where 1= most gay sounding and 6=most straight sounding.) At /s/ +30 ms., listeners perceived the speaker at a gayness degree of 3.9405. Both of these values mean that, on average, the listeners considered the speaker as sounding more straight than gay. However, by the /s/ + 60 ms. duration, the listeners’ mean gayness degree had crossed the midpoint of the continuum (3.5), reaching 3.345 (a lower number means a response closer to the gay end of the continuum). At the /s/+60ms. duration, listeners as a whole believed this speaker sounded gay. Degree of gayness rose with each longer /s/ duration, but was not significant between all groups. A Tukey's studentized range (HSD) test on the means showed that there were not significant differences between /s/ and /s/ + 30 ms. means. However, the differences between /s/ +30 ms. and /s/ +60 ms. means were significant, as were the differences between /s/ and /s/ + 60 ms. means.
4.1.1 Statistical analysis
4.1.1.1 Gay/straight variable
One-tailed ANOVA analysis of the mean gayness ratings by sex, /s/ duration and the interaction of /s/ duration and sex also reached significance. The mean gay rating decreased (i.e. the speaker sounded more gay)2 for listeners as the /s/ duration increased. The interaction of sex and duration was not significant (p=.9014), so it did not affect the final analysis. However, once that variable was eliminated, a listener’s sex did affect the mean gayness rating. Overall, men are more likely to give a gayness rating that was .-39 lower (sounds gayer) than women (p=.0429). Nonetheless, the mean gayness rating of the whole listener group did decrease significantly (t=-3.56 and p=.0003) as the /s/ duration increased.
4.1.1.2 Distracter variables
Again, two-tailed ANOVA analysis of the mean ratings for the distracters did not reach significance. For the old/young mean responses p=.9052, and for the educated/non- educated mean responses, p=.4190. For the wealthy/poor mean responses, p=.0619. The speaker sounded poorer to listeners when /s/ duration increased, but not to the point of significance3.
Table 1. Mean perceived sexual orientation by listener and /s/ duration: 1=sounds gay; 6=sounds straight.
Group A /s/
Group B/s/ +30 ms.
GroupC/s/+60 ms.
3 f 3 f 3.33333 f 3.33333 f 3.33333 f 3.66667 f 4f4f3f
2.333333 3 3.333333 3.333333 3.333333 3.666667
f f f f f f
1.666667 f 2.333333 f 2.333333 f 2.666667 f 3 f 3 f
4 f 4 f 4 f 4.33333 f 4.333333 f 4.333333 f 4.666667 f 4.66667 f 5 f 5f5f4f
4 f 4 f 4 f 4 f 4.333333 f 4.333333 f 4.333333 f 4.333333 f 4.666667 f
3.333333 f 3.333333 f 3.666667 f 3.666667 f 3.666667 f 3.666667 f 3.666667 f 4 f 4 f
5 f 5 f 5.33333 f 3 m 3 m 3.66667 m 4 m 4 m 4 m 4 m 5.66667 m Mean
4.02898551
5 f 5.666667 f 5.666667 f 2 m 3.333333 m 3.333333 m 3.333333 m 3.333333 m 4 m 4.333333 m 4.333333 m Mean
3.94047619
4.333333 f 4.666667 f 4.666667 f 1.666667 m 2 m 3 m 3.333333 m 3.666667 m 3.666667 m 3.666667 m 4 m Mean
3.3452381
4.2 Forced choice
Based on the mean response value, I also categorized each listener’s overall response into one of two possible forced choices, either gay (g) or straight (s). A listener response mean that was < 3.5 was assigned the value of g, whereas a listener response mean that was greater than 3.5 was assigned a value of s. None of the means equaled 3.5, as the continuum contained an even number of points.
Table 2. Forced choice data by sex of listener.
Group /s/+30m Group /s/+60m
A /s/ Group B s. C s. gfgfgf gfgfgf gfgfgf gfgfgf gfgfgf gmgmgf gmgmgf sfgmgf sfgmgf sfgmgm sfsfgm sfsfgm sfsfgm sfsfsf sfsfsf sfsfsf sfsfsf sfsfsf sfsfsf sfsfsf sfsfsf sfsfsf smsfsf smsfsf smsfsm smsmsm smsmsm smsmsm
The raw forced choice data suggest that, as hypothesized, listeners were more likely to judge a man as sounding gay as the/s/duration increased. At the/s/duration of the natural speech, 25% of listeners perceived the man as gay. At /s/ +30 ms. duration, 35.71% of listeners perceived the speaker as gay and at the /s/ + 60 ms. duration, 46.43% of listeners perceived the same voice as coming from a gay man.
Figure 5 . Percentage for Male and Female combined gay judgments (forced choice)
100 80 60 40 20 0
/s/ +30 +60 ms. ms.
4.2.1 Statistical analysis: forced choice
4.2.1.1 Gay/straight variable
I analyzed these data in two ways. I used the forced choice values to see if a listener was more likely to judge a man as sounding gay as /s/ gets longer. One-tailed multiple regression analysis4 showed that the trend was significant overall (p<.05) at p=.0488 (Z=1.66). The forced choice (gay or straight) rating was the dependent variable, whereas the independent variables were /s/ duration, sex of listener, and the interaction between sex and duration. The interaction of sex and /s/ duration did not contribute significant information to this trend (p=.2161). Neither was there a significant difference between the responses of the male and female listeners. Overall, the odds of a listener responding “gay” for each longer /s/ duration was 1.6 times that for the next shortest /s/ (odds ratio=1.6). Gay judgments
4.2.1.2 Distracter variables
Listeners made this distinction for the gay/straight variable only. Two-tailed multiple regression analyses5 on the distracter questions failed to show any significant differences. I examined the forced choice data from the three other continua, but the listeners made essentially the same judgments across the /s/ duration. Using both the old/young and educated/non-educated forced choice responses as the dependent variable, p=1. Using the wealthy/poor forced choice responses as the dependent variable, p=.1092. The fact that there were no significant results in the distracter questions supports the claim that /s/ duration is a perceptual cue for sexual orientation. Listeners attributed the difference in speech to the sexual orientation of the speaker, not his education, age or wealth.
1 The listeners in the second study perceived the speech as "straighter" than in the first study. I believe this is because the word list format lacked the extra content information that the text provided.
2 I caution readers to remember throughout this paper that a lower mean gayness rating indicates that the listener deemed the speaker as sounding more gay, as the continuum listeners used ranged from 1=gay to 6=straight. When I say that a listener’s perception of a man as gay increased, I mean that the mean gayness rating decreased.
3 This is interesting, in that a longer /s/ duration seems to relate to how poor a male speaker sounds. However, it is not significant, and so I leave further discussion of this question to others.
4 One-tailed analysis was used for both the forced choice and mean degree of gayness because I had a hypothesis about the directionality of the change.
5 I used a two-tailed analysis for both the forced choice and mean ratings for all distracters, because unlike the gay/straight question, I had no hypothesis as to directionality.
5. DISCUSSION
Results of this investigation suggest that there is at least one phonetic cue that people use to perceive a man as gay. As both a continual mean ‘gayness’ score and a categorical (forced choice) response, listeners’ perception of a man as gay increased substantially with the longer /s/ durations in word-initial, stressed /skV/ and /spV/ environments. This supports Linville’s indirect production evidence that /s/ duration is a cue in perception of sexual orientation in men.
Male listeners rated the speaker as “gayer” overall than female listeners. This could be explained by the fact that there was a smaller sample of men, so the results are less likely to be representative of all men. Previous research has pointed towards women having a higher sensitivity towards speech differences than men (Kramer 1977 and Labov 1972). However, in this particular area, men may actually be the more sensitive, tending to think another man sounds “gayer” because of their higher level of sensitivity to sexual orientation (Linville 1998). The male/female differences in forced choice categorization of sexual orientation was not significant, so the data is open to interpretation. Further studies looking at the differences in male and female judgment of sexual orientation could help us reach a conclusion.
The results of this study are surprising in light of the difficulties involved with making claims that some kind of “Uber-gay” language variety exists. I never expected results that supported my hypothesis, especially since the sample is not larger than in previous studies. Zwicky (1997) warned that identifying “the gay voice” through phonetic studies would pose problems, since it’s likely that there are only a few ways gay men might differ from straight men in their speech production, and that it’s possible that not all gay men who sound gay will employ the same cues. However, Zwicky was writing before the results from Linville and Crist suggested significant differences in /s/ duration. The literature is more conclusive now. I do not suggest, though, that we as researchers have resolved the issue with /s/ duration as a perceptual cue or production difference. On the contrary, we have merely scratched the surface and can now begin to confirm the results of the small body of literature that suggests that the role of /s/ duration is important both in the perception of men as gay and in the production differences between gay and straight men. Research involving much larger samples is a logical next step.
This study has examined a very limited /s/ environment, based on evidence from Crist that there is a production difference in stereotype performance in the same word- initial /spV/ and /skV/ stressed syllable clusters. Linville’s results showed a longer /s/ duration between perceived and actual sexual orientation across a much wider phonetic environment. The only claim that I can make at this point is that /s/ duration is a cue for perception in a very limited phonetic environment, but it is possible that a longer /s/ serves as a perceptual cue across more, or perhaps all, phonetic environments in English. This would be an interesting question for additional study.
This study also raises the question of where on the /s/ duration continuum the cue begins to work. It seems that a longer /s/ duration in at least some phonetic environments will drive listener perception of a male’s sexual orientation. However, it remains unclear what the range of /s/ durations might be that are meaningful in this way. Presumably there is a point at which a longer /s/ just sounds unnatural. Post-hoc statistical analysis showed that there were significant differences as /s/ duration increased, but that the step from /s/ to /s/ +30 ms. was not significant. This suggests that an increase of 30 ms. is too subtle to change listener perception of male sexual orientation. Researchers could modify a study such as this one to employ a wider range of /s/ durations for listeners to judge, to try to delimit the /s/ duration that is meaningful to listeners when they make their perception judgments of sexual orientation, perhaps beginning at an /s/ +40 ms. duration.
Another question is whether there might be a different way of examining /s/ duration. I looked at absolutes, without taking into consideration the ratio of /s/ duration to some other phonetic or phonological unit. It is possible that the absolute length of /s/ is not really the cue, but rather that a certain /s/ to word or /s/ to syllable duration ratio is. This experiment used tokens that ranged from one to three syllables. Since the word lists in this study did not consider the ratio of total /s/ duration to either word or syllable duration, there is no way to tell whether the ratio has an effect on sexual orientation perception. I considered this possibility during the experimental design phase of the research, but opted to look at a range of absolute /s/ durations to make things simpler.
Though this study is about perception of male sexual orientation and was not designed to make any claim about the relationship between gay-sounding male speech as it relates to women’s speech, two ideas surfaced from my findings that hint at there being some kind of parallel.
Looking at the combined results of Schwartz (1967) , Linville (1998), Ingemann (1968), Avery and Liss (1996), and this study, I suspect that /s/ peak frequency is a perceptual cue for sexual orientation in addition to /s/ duration. Schwartz (1968) reported higher /s/ peak frequency in women, which he attributed to a smaller vocal tract. Schwartz also found that listeners were able to correctly identify speaker sex at a rate of 93% correct when the stimulus was a sustained /s/ production (Schwartz 1179). He inferred that the difference in peak frequency between men and women was the cue listeners employed to make judgments. In addition, Linville (1998) found higher /s/ peak frequency in men perceived as gay when compared to men perceived as straight1.
Avery and Liss found the same when they explored differences between less- and more- masculine sounding men. At least for peak frequency, there is a parallel between women and gay (or less- masculine sounding) men when compared to straight men, which is evidence that gay-sounding men could be mirroring women’s speech. A higher peak frequency is a valid production strategy differentiating women from men and gay men from straight men. A study similar to this one but which manipulates peak frequency instead of duration for /s/ could begin to resolve whether a higher peak frequency is a perceptual cue. If there is evidence that /s/ peak frequency is a cue in the perception of gay-sounding males, then the evidence that gay (or gay sounding) men employ at least some production strategies that are similar to those that differentiate women from straight men.
The case of /s/ peak frequency is different from the /s/ duration in only two ways. As explained above, there is no direct evidence that /s/ peak frequency is a cue for perception of a speaker’s sexual orientation. In the same way, there is no evidence that women employ a longer /s/ duration than men. Is duration a perceptual cue for sex, also? The only evidence to suggest it might be is from Ingemann's (1968) replication of Schwartz’s (1967) experiment. When her listeners reported that they were making their identifications of gender based on a certain breath quality before and after the /s/, she cut out some of the /s/ sound from each end, shortening the total duration in the process. The results were that her listeners were less able to judge speaker sex accurately with the shortened /s/. Her interpretation was that it is easier to judge a whole than a part. I wonder if in shortening the /s/, the duration length was actually part of the key information listeners needed to make a correct judgment for speaker sex2. If women do indeed have significantly longer mean /s/ durations than straight men, then it could provide further evidence that gay (or gay-sounding) men are employing phonetic cues that mirror women’s speech production patterns. Linville showed a measurable significant difference in /s/ duration between men perceived as sounding gay and men perceived as sounding straight. This study extends those findings to show that the /s/ duration is a perceptual cue, also. The /s/ peak frequency results in this field have almost mirrored the /s/ duration results. Why not see if both variables behave the same?
In short, my results provide a clearer picture of the role the duration of /s/ plays in the perception of men as sounding gay. At least for a narrow phonetic environment, a longer/s/ duration is a cue that listeners attend to when determining (consciously or subconsciously) what they believe to be the sexual orientation of a man. Further investigation will be needed to test the role of /s/ duration on a broader range of phonetic environments. It will also be of interest to extend the experimental methodology here to other potential cues involving both the acoustic profile of /s/ and other phonetic domains.
1 Linville also found that /s/ peak frequency in gay men was significantly higher than in straight men. i.e. a higher /s/ peak frequency occurred both in men perceived as gay and in men whose self-reported sexual orientation was gay.
2 Alternatively, the information lost from the deletion of part of the /s/ from the beginning and end of the /s/ could have been breath quality, as Ingemann’s subjects reported.
APPENDIX ONE. INFORMATION FORM: LISTENER GROUP #1
Introduction to the study
We are inviting you to be in a research study about how people listen to language.
Jennifer Palmer of the University of North Carolina is doing this study under Dr. Chip Gerfen of the UNC Linguistics Department.
Purpose:
The purpose of this study is to try to identify the ways some people form impressions about an individual’s speech.
What will happen during the study:
1. We will ask you to take part in 2 tasks, which should take about 30 minutes. 2. These tasks will include:
• Answering a short questionnaire about your age, native language and hearing ability • Listening to a number of speech recordings and indicating on questionnaires certain information about how you perceive the speech you hear.
3. If you have any questions or concerns about being in this study, you should email Jennifer Palmer at jlpalmer@email.unc.edu or call her at 960-7359. You may also call her faculty advisor, Chip Gerfen, at 962-4996.
Your privacy is important
We will make every effort to protect your privacy.
We will not ask your name.
We ask you to agree that we may use any information we get from this research study in any way we think is best for publication or education.
Risks and discomforts
We do not know of any risks or discomfort you will have from being in this study.
Your rights:
You decide on your own whether you want to be in this study.
You will not be punished or treated differently if you decide not to be in this study If you decide to be in the study, you will have the right to stop being in the study at any time.
Institutional Review Board Approval
The Academic Affairs Institutional Review Board (AA-IRB) of the University of North Carolina at Chapel Hill has approved this study.
If you have any concerns about your rights in this study, you may contact the Barbara Goldman, Chair of the AA-IRB at CB#4100 201 Bynum Hall, UNC-CH, Chapel Hill, NC 27599, (919) 962-7761, aa-irb@unc.edu
By filling about the questionnaire about my age, UNC student status, native language and hearing ability, I affirm that I have read the information in this information form, and that I agree to be in the study. If I wish to drop out later, I will contact Jennifer Palmer and tell her my Listener number. She will then destroy any data I have given her.
Listener # _______________ Please keep this number for reference.
APPENDIX TWO INFORMATION FORM: LISTENER GROUP #2
Introduction to the study
We are inviting you to be in a research study about how people listen to language.
Jennifer Palmer of the University of North Carolina is doing this study under Dr. Chip Gerfen of the UNC Linguistics Department.
Purpose:
The purpose of this study is to try to identify the ways in which people form opinions about an individual’s speech.
What will happen during the study:
1. We will ask you to take part in 2 tasks, which should take about 30 minutes. 2. These tasks will include:
• Answering a short questionnaire about your age, native language, and hearing ability. • Listening to a speech recording and indicating on questionnaires certain information about how you perceive the speech you hear.
3. If you have any questions or concerns about being in this study, you should email Jennifer Palmer at jlpalmer@email.unc.edu or call her at 960-7359. You may also call her faculty advisor, Chip Gerfen, at 962-4996.
Your privacy is important
We will make every effort to protect your privacy.
We will not ask your name. We ask you to agree that we may use any information we get from this research
study in any way we think is best for publication or education.
Risks and discomforts
We do not know of any risks or discomfort you will have from being in this study.
Your rights:
You decide on your own whether you want to be in this study. You will not be punished or treated differently if you decide not to be in this study
If you decide to be in the study, you will have the right to stop being in the study at any time.
Institutional Review Board Approval
The Academic Affairs Institutional Review Board (AA-IRB) of the University of North Carolina at Chapel Hill has approved this study.
If you have any concerns about your rights in this study, you may contact the Barbara Goldman, Chair of the AA-IRB at CB#4100 201 Bynum Hall, UNC-CH, Chapel Hill, NC 27599, (919) 962-7761; aa-irb@unc.edu
By completing the questionnaire about my age, UNC student status, native language and hearing ability, I affirm that I have read the information in this consent form, and that I agree to be in the study. If I wish to drop out later, I will contact Jennifer Palmer and tell her my Listener number. She will then destroy any data I have given her
Listener # __________ Please keep this number for reference.
APPENDIX THREE. INFORMATION FORM: SPEAKERS
Introduction to the study
We are inviting you to be in a research study about how people listen to language.
Jennifer Palmer of the University of North Carolina is doing this study under Dr. Chip Gerfen of the UNC Linguistics Department.
Purpose:
The purpose of this study is to try to identify the ways in which people form opinions about an individual’s speech.
What will happen during the study:
1. 2.
We will ask you to take part in 2 tasks, which should take about 30 minutes. These tasks will include:
• Answering a short questionnaire about your age, native language, and speech background. • Recording a passage you will read that will later be played to a group of listeners and then possibly manipulated digitally and played to another group of listeners. Both groups will be asked to make judgments based on their perception of your speech.
If you have any questions or concerns about being in this study, you should email Jennifer Palmer at jlpalmer@email.unc.edu or call her at 960-7359. You may also call her faculty advisor, Chip Gerfen, at 962-4996.
Your privacy is important
We will make every effort to protect your privacy.
Your name will not be linked in any way to the data you give us.`
We ask you to agree that we may use any information we get from this research study in any way we think is best for publication or education.
Risks and discomforts
The only risk we anticipate is that your voice may be recognized by one of the listeners. However, we have attempted to eliminate this risk by drawing listeners from the undergraduate population.
Your rights:
You decide on your own whether you want to be in this study. You will not be punished or treated differently if you decide not to be in this study
If you decide to be in the study, you will have the right to stop being in the study at any time.
Institutional Review Board Approval
The Academic Affairs Institutional Review Board (AA-IRB) of the University of North Carolina at Chapel Hill has approved this study.
If you have any concerns about your rights in this study, you may contact Barbara Goldman, Chair of the AA-IRB at CB#4100 201 Bynum Hall, UNC-CH, Chapel Hill, NC 27599, (919) 962-7761; aa-irb@unc.edu
By accepting this written document, I affirm that I have read the information in this form, and that I agree to be in the study. If I wish to drop out later, I will contact Jennifer Palmer and tell her my Speaker number. She will then destroy any data I have given her and erase my recording.
Speaker # __________________ Please keep this number for reference.
APPENDIX FOUR CONSENT FORM: SPEAKERS
Introduction to the study
We are inviting you to be in a research study about how people listen to language.
Jennifer Palmer of the University of North Carolina is doing this study under Dr. Chip Gerfen of the UNC Linguistics Department.
Purpose:
The purpose of this study is to try to identify the ways in which people form opinions about an individual’s speech.
What will happen during the study:
1. 2.
We will ask you to take part in 2 tasks, which should take about 30 minutes. These tasks will include:
• Answering a short questionnaire about your age, native language, and speech background. • Recording a passage you will read that will later be played to a group of listeners. Based on the outcome of the first group of listeners, this speech may then be altered so to make some of the s- sounds longer, and then played to a second group of listeners. Both groups will be asked to make judgments about your sociological makeup, based on their perception of your speech. Listeners will be asked about things like your gender, age, sexual orientation, socioeconomic group, and education level.
If you have any questions or concerns about being in this study, you should email Jennifer Palmer at jlpalmer@email.unc.edu or call her at 960-7359. You may also call her faculty advisor, Chip Gerfen, at 962-4996.
Your privacy is important
We will make every effort to protect your privacy.
Your name will not be linked in any way to the data you give us. We will not use your name in any area of the research.
We ask you to agree that we may use any information we get from this research study in any way we think is best for publication or education.
Risks and discomforts
The only risk we anticipate is that your voice may be recognized by one of the listeners. However, we have attempted to eliminate this risk by drawing listeners from the undergraduate and non-UNC population.
Your rights:
You decide on your own whether you want to be in this study. You will not be punished or treated differently if you decide not to be in this study
If you decide to be in the study, you will have the right to stop being in the study at any time.
Institutional Review Board Approval
The Academic Affairs Institutional Review Board (AA-IRB) of the University of North Carolina at Chapel Hill has approved this study.
If you have any concerns about your rights in this study, you may contact Barbara Goldman, Chair of the AA-IRB at CB#4100 201 Bynum Hall, UNC-CH, Chapel Hill, NC 27599, (919) 962-7761; aa-irb@unc.edu
I have had the chance to ask any questions I have about the study and they have been answered for me. By signing this written form, I affirm that I have read the information in this consent form, and that I agree to be in the study. My Speaker Number and any data I provide are not linked to this consent form or my name. If I wish to drop out later, I will contact Jennifer Palmer and tell her my Speaker Number. She will then destroy any data I have given her and erase my recording.
Participant signature: ____________________________
APPENDIX FIVE BACKGROUND QUESTIONNAIRE: SPEAKERS
Speaker Number ______________ 1. How old are you? _________ 2. Are you a UNC student? __________ Yes _________ No 3. If yes, are you an undergraduate or a graduate student? ____Undergrad ____Grad 4. Are you a native speaker of American English? ____ Yes _____ No 5. Have you ever had any speech therapy? If so, explain.
6. Have you had extensive acting and/or singing experience? If so, explain.
APPENDIX SIX BACKGROUND QUESTIONNAIRE: LISTENERS (GROUP #1 AND #2)
Listener Number ______________ 1. How old are you? _______ 2. Are you a UNC student? __________ Yes _________ No 3. If yes, are you an undergraduate or a graduate student?
____Undergrad _____Grad
4. Are you a native speaker of American English? _____Yes ______ No 5. Have you had hearing problems, ear surgery or other audiological disorders? If so,
explain.
APPENDIX SEVEN. PERCEPTION STUDY QUESTIONNAIRE: STUDY #1
Listener # ____________ Speaker # ____________
You will hear each passage once in random order. Please listen and report how you perceive the speaker and the confidence you have in your response, based only on the speech you will hear. You will be offered a continuum from 1 to 6, for two responses: your perception of certain sociological aspects of the speaker, and how confident you are in your response. We encourage you to answer all the questions. Circle the best response, and when you go on to the next speaker, use a new sheet.
Examples:
1. You listen to the voice and it sounds like a very, very old person to you. Your response is to circle 1. You are very confident in your response. You circle 6 under confidence rating.
2. The voice also sounds slightly non-educated. Y ou circle 4. Y ou are rather unsure of your answer, so you circle 2 under confidence rating.
Does the speaker sound: Confidence rating (1= not confident, 6= very confident)
wealthy 123456 poor 123456 old 123456 young 123456 male 123456 female 123456 gay 123456 straight 123456 educated 123456 non-educated 123456
APPENDIX EIGHT PERCEPTION STUDY QUESTIONNAIRE: STUDY #2
Listener # ____________ Track # ____________
You will hear 3 tracks on a CD. Each track is a word list approximately 11-13 seconds long. Please listen and report how you perceive the speech you hear and the confidence you have in your response, based only on each word list. You will be offered a continuum from 1 to 6, for two responses: your perception of certain sociological aspects of the speaker, and how confident you are in your response. We encourage you to answer all the questions. Circle the best response, and when you go on to the next track, use a new sheet.
Examples:
1. You listen to the voice and it sounds like a very, very old person to you. Your response is to circle 1. You are very confident in your response. You circle 6 under confidence rating.
2. The voice also sounds slightly non-educated. Y ou circle 4. Y ou are rather unsure of your answer, so you circle 2 under confidence rating.
Does the speaker sound: Confidence rating (1= not confident, 6= very confident)
wealthy 123456 poor 123456 old 123456 young 123456 gay 123456 straight 123456 educated 123456 non-educated 123456
APPENDIX NINE TEXT
You won’t believe what happened to me the other day. You know Scott, that guy I know from my Spanish class last year? He knows I am trying to get a good job as a chef, right? So I run into him at school, when we were both skipping class. He mentions that his cousin works at this resort called Skylight, outside the city. He told me they were looking for a chef, and gave me a number to call the owner. I called her and we figured out a time that worked with both of our schedules, and set up an interview. From the way she described it, it really wouldn’t be a big deal. I just had to prepare a sample menu for her and then afterwards I’d spend the whole day looking over the space they’d provide for me, and etcetera. She even told me I could bring my mom along so she could take advantage of the day spa the restaurant was attached to. I’m glad Mom skipped out of that. She’d have been mortified. But I’ll get to that in a sec.
So I planned a linguine dish, with a garlic, spinach and clam sauce. Then for the second course I made a rub and roasted some spare ribs until the meat was falling off the bone. I thought about serving it with those rosemary potatoes I make, but since I already had pasta, I decided that too many starches would not be a good idea. So instead I went for a mango chutney to accompany it. Anyway, that’s what I had planned. It sure didn’t work out though. When I got there, I was a little skeptical, cause it was just some house in the ‘burbs, but they just said they had a schedule conflict at the restaurant, so they had just set things up here at the owner’s house. Why I believed that, I do not know. Anyway, they gave me some set ingredients to work with. I was not allowed to use the stuff I’d brought. So much for my menu! I thought at first, ok, this is a real test. I can do it. Then she showed me what I had to work with. Any spice I wanted, which was fine, but the meat she gave me? Spam. I kid you not. And as a side dish? Instant mashed potatoes. I was so furious I was about to spit. It was a total scam—that loser Scott had set me up! I can’t believe I didn’t know something sketchy was going on from the first. I always thought Scott didn’t like me very much. Why he went to those extremes, I have no idea. Man! I will never speak to him again! I spent a long time getting that menu just right! Oh well, chalk it up to experience, I guess.
References
Anshen, F. (1978) Statistics for Linguists. Rowley, MA: Newbury House Publishers, Inc.
Avery, J. and J. Liss (1996) Acoustic Characteristics of Less-Masculine-Sounding Male
Speech. Journal of the Acoustical Society of America 99.6:3738-48.
Bucholtz, M. and K. Hall (1995) Introduction: Twenty years after Language and Women's Place. In K. Hall and M. Bucholtz (eds) Gender Articulated: language and the socially constructed self New York: Routledge 1-24.
Coughlan, J., S. Lawton and G. Weinreb (1993) SoundScope User's Manual. GW Instruments. Somerville, Massachusetts.
Crist, S. (1997) Duration of Onset Consonants in Gay Male Stereotyped Speech. University of Pennsylvania Working Papers in Linguistics 4:3:53-70.
Fai, D. (1988) A Study of the (-ING) Variable in the Ottawa Male Community. Actes du colloque: Tendances actuelles de la recherche sur la langue parlée. Centre international de recherche sur le bilinguisme/International Center for Research on Bilingualism, Quebec, 35-40.
Gaudio, R. (1994) Sounding Gay: Pitch Properties in the Speech of Gay and Straight Men. American Speech 69.1:30-57.
Ingemann, F. (1968) Identification of the Speaker's Sex from Voiceless Fricatives. Journal of the Acoustical Society of America 44(4), 1142-1144.
Jacobs, G. (1996) Lesbian and Gay Male Language Use: A Critical Review of the Literature. American Speech 71:49-71.
Kramer, C. (1977) Perceptions of Female and Male Speech. Language and Speech 20:151-161.
Labov, W. (1972) Sociolinguistic Patterns. Philadelphia: University of Pennsylvania Press.
Lakoff, R. (1975) Language and Woman's place. New-York: Harper and Row.
Leap, W. (1995) Beyond the Lavender Lexicon: Authenticity, Imagination and Appropriation in Lesbian and Gay Languages. New York: Gordon and Breach Publishers.
Leap, W. (1996) Word's Out: Gay Men's English. Minneapolis: University of Minnesota Press.
Lerman, J. and P. Damsté (1969) Voice Pitch of Homosexuals. Folia phoniat. 21:340-46.
Linville, S. (1998) Acoustic Correlates of Perceived versus Actual Sexual Orientation in Men's Speech. Folia Phoniatrica et Logopaedica 50.1:35-48.
Livia, A. and K. Hall, eds. (1997) Queerly Phrased: Language, Gender, and Sexuality. New York: Oxford University Press.
Moonwomon, B. (1997) Towards the Study of Lesbian Speech. In Anna Livia and Kira Hall (eds.), Queerly Phrased: Language, Gender, and Sexuality. New York: Oxford University Press.
Schwartz, M. (1967) Identification of speaker sex from isolated voiceless fricatives. Journal of the Acoustical Society of America. 43(5), 1178-1179.
Taylor, B. (1996) Gay Men, Femininity, and /t/ in New Zealand English. In Janet Holmes (ed.), Wellington Working Papers in Linguistics. 8:70-92.
Terango, L. (1966) Pitch and duration characteristics of the oral reading of males on a masculinity-femininity dimension. Journal of Speech and Hearing research. 9, 590-95.
Ward, G. and G. Greenman II (1998) Studies on Gay & Lesbian Language: A Partial Bibliography. Located on the internet at: http://www.msu.edu/~greenm14/outil/gaybib.html.
Zwicky, A. (1997) Two Lavender Issues for Linguists. In Anna Livia and Kira Hall (eds.), Queerly Phrased: Language, Gender, and Sexuality. New York: Oxford University Press. 25-45.
