Kaiser, E., & Munson, B. Social Selectivity in Adults' Novel Sound

advertisement
Laboratory Phonology 11
65
Kaiser & Munson
Social Selectivity in Adults' Novel Sound Learning
Eden Kaiser* & Benjamin Munson#
*Linguistics Program, University of Minnesota; [email protected]
#Department of Speech-Language-Hearing Sciences, University of Minnesota; [email protected]
There is a good deal of evidence for the existence of social identity-based differences in speech (Labov
1972, and others), especially differences based on talker gender (Perry, Ohde, & Ashmead, 2001; Patterson
& Werker, 2002; Foulkes, Docherty, & Watt 2005, and others). However, the nature of the process by which
these differences are acquired is still unknown. The purpose of this study is to investigate sex-specific
phonetic learning in a laboratory-based novel-sound learning task. Specifically, we examined male and
female participants' tendencies to emulate phonetically distinct variants of novel sounds produced by talkers
who shared their biological sex, as opposed to the variant produced by the opposite sex. We believe that the
outcome of this endeavor will be a better understanding of how these variables might be acquired during
first-language acquisition.
To examine this, we created one set of one-syllable nonwords whose onsets contained a voiceless lateral
obstruent, and another set whose onsets contained a velar fricative. These non-English sounds were used in
order to control for any previous sex-related biases people may have for familiar English sounds. Audio
productions of these novel words were recorded by two trained linguists whose voices the authors judged as
being prototypically female and male. The novel words were presented one at a time, by either the female
voice or the male voice, and were associated with novel objects shown on a computer screen. Each
participant was trained on one onset which the man and woman produced differently (hereafter called “sexbiased”), and one onset which they produced equally variably (hereafter called “random”). The two variants
for the voiceless lateral obstruent were the voiceless lateral fricative [ɬ] and the voiceless lateral affricate [tɬ].
The two variants for the velar fricative were the voiced velar fricative [ɣ] and the voiceless velar fricative
[x]. Participants were systematically exposed to variations of onsets: half the participants heard sex-biased
variation for the lateral onset and random variation for the velar onset, while the other half heard random
variation for the lateral onset and sex-biased variation for the velar onset. Three iterations of learning phases
followed by test phases were presented to each participant, first for one onset, then for the other. During the
learning phases, participants saw a novel object on the screen and heard over headphones “This object is a
____”, spoken by either the man or the woman, and onsets varied accordingly. During the test phases,
participants were asked to name objects whose pictures appeared on the screen one at a time. Below is the
counter-balancing scheme used to systematically expose participants to both sex-biased and random
variation. There were five women and five men in each of four groups.
Group 1
random [ɬ]/[tɬ]
female-biased [ɣ] / male-biased [x]
Group 2
random [ɬ]/[tɬ]
female-biased [x] / male-biased [ɣ]
Group 3
female-biased [tɬ] / male-biased [ɬ]
random [x]/[ɣ]
Group 4
female-biased [ɬ] / male-biased [tɬ]
random [x]/[ɣ]
Table 1: Sex-biasing counter-balancing scheme
Results were coded in narrow phonetic transcription by an experienced transcriber who was blind to the
purposes of the experiment. The dependent measures were calculated by one of the authors using the
following criteria: onsets which contained a fricative were judged to be approximating the lateral fricative
[ɬ], whereas onsets which contained an obstruent were judged to be approximating the lateral affricate [tɬ];
LabPhon11 abstracts
edited by Paul Warren
Wellington, New Zealand
30 June - 2 July 2008
Abstract accepted after review
66
Laboratory Phonology 11
Kaiser & Munson
and onsets which contained a voiceless velar or voiceless glottal were judged to be approximating the
voiceless velar fricative [x], whereas all onsets which contained either a voiced velar/glottal or a rhotic were
judged to be approximating the voiced velar fricative [ɣ].
The authors found that, overall, both men and women tended to produce more approximations of the
onsets [tɬ] and [x] over their respective variants, [ɬ] and [ɣ].
Group 1
Group 2
Group 3
Group 4
Females’ use of [tɬ]
71.212
Males’ use of [tɬ]
79.059
64.583
59.722
67.677
44.556
Females’ use of [ɣ]
18.519
14.293
3.819
Males’ use of [ɣ]
6.944
4.646
4.444
Table 2: Results (in percentages)
The influence group and training condition on the percentage use of different variants did not achieve
statistical significance using conventional alpha levels. However, there were some noteworthy trends in the
data. For the lateral onset [ɬ]/[tɬ], women and men were more likely to attempt to produce [ɬ] when
presented with sex-biased variation (either biased for or against production of [ɬ]). More interestingly, men
were much more likely to produce [tɬ] than [ɬ] when presented with random variation, than when biased
against producing [tɬ]. For the velar onset [x]/[ɣ], males rarely attempted to approximate the voiced velar
fricative, opting instead for the unvoiced variant. Females were about four times as likely to attempt to
approximate the voiced velar fricative when they were presented with sex-biased variation (especially when
biased for the voiced variant). But when females were presented with random variation, their attempts to
approximate the voiced variant were on par with the men's.
Based on these results, we can conclude that women and men are sensitive to sex-biased differences when
acquiring novel sounds. In other words, people learn new phonetic variants selectively, based on the socialindexical characteristics associated with the phonetic input, namely, one's own sex and the sex of the
speaker. The next step along this line of research is to reproduce these results, removing the potentially
complicating factor of production. Currently, we are designing a perception experiment to examine whether
adults' perception is similarly biased by the perceived sex of the adults who model the novel sound. The
result of these two experiments together will be an improved understanding of the extent to which adults'
sound learning is indexed to perceived social characteristics of the people producing the sounds being
learned.
References
Foulkes, P., Docherty, G., & Watt, D. (2005). Phonological variation in child directed speech. Language, 81, 177-206.
Labov, W. (1972). Sociolinguistic Patterns. Philadelphia: University of Pennsylvania Press.
Patterson, M. L. and Werker, J. F. (2002). Infants ability to match dynamic phonetic and gender information in the face
and voice. Journal of Experimental Child Psychology, 81, 93-115.
Perry, T., Ohde, R., and Ashmead, D. (2001). The acoustic bases for gender identification from children’s voices.
JASA, 109, 2988-2998.
Download
Random flashcards
Arab people

15 Cards

Nomads

17 Cards

Ethnology

14 Cards

History of Europe

27 Cards

Create flashcards