J. Walton and R. Orlikoff

For a fresh perspective on binary differentiation from acoustic and perceptual cues let us turn to Walton and Orlikoff’s study of black v. white prisoners in Mississippi. (See also Jacobs et al. on gay v. straight speakers, Szakay on Maori v. Pakeha speakers, and several studies entirely or in part devoted to differences between male and female speakers.) One-second samples were collected from sustained /a/ vowels produced by black and white inmates (mean age 29) who were speakers of Southern American English varieties commonly spoken by blacks and whites in the region (e.g. African American Vernacular English, Gulf Southern). Both expert and naïve listeners were able to judge the race (or at least the dialect) of speakers of randomly paired black and white tokens with 60 percent accuracy.

While black speakers have been reported to speak with lower average F0, and expert listeners have suggested that they can tell black from white speakers on the basis of “identifiable differences in nasality, pitch, length, and rate”, Lass et al. (1979) found that judgments could be accurately made on the basis of less than .5 sec voice samples presumably devoid of all this information.

Walton and Orlikoff asked listeners to identify the order samples were played in (white-black or black-white) to elicit comparisons between voices, not judgments from memory (a technique I would do well to adopt). The mean F0 of the black and white voices was not statistically different, nor was formant structure. Black voices, however, were distinguished by larger jitter and shimmer and lower harmonic ratio relative to noise (and white voices vice versa). Walton and Orlikoff conclude that the statistically significant level of correct judgments could have been as a result of these phenomena. They speculate on possible physiological, paralinguistic, and lifestyle causes and confounds.

In considering the implications of this study for my own work on gender, I find that not only should I look for material to give me a sense of jitter, shimmer, and harmonic-noise ratio differences by gender, in English or crosslinguistically, but that it will be necessary to consider the dialectal background of my subjects in greater detail than I had anticipated to determine whether similar isolated factors (e.g., to invent a finding, greater breathiness among Indo-Canadian women or higher F0 among First Nations men, even where L1 is English and an individual does not speak a dialect identified with a particular socioethnic group) may complicate a study of creak, where like Walton and Orlikoff’s study, vocal rather than vowel characteristics are expected to differentiate groups. Journal of Speech and Hearing Research 37/4.… (más)