A Reader in Sociophonetics

270 Erik R. Thomas, Norman J. Lass, and Jeannine Carpenter

Virginia University. Subjects were asked to circle, for each stimulus, either

African American or European American on an answer sheet. Results from

subjects with hearing impairments, from those whose ¿ rst language was not

English, and from those of ethnicities other than African American or Euro-

pean American were excluded from the numerical analyses.

2.2 Results and discussion

Levels of accuracy of ethnic identi¿ cations by listener group, experimental

treatment, and sex of speaker are given in Table 12.1. Three general trends are

evident in Table 12.1. First, the West Virginia listeners showed lower accuracy

than the two North Carolina groups of listeners. The speakers who produced

the stimuli were North Carolinians, and the West Virginians often associated

Southern dialect features by some European American speakers with African

American identity. However, they probably also had less daily contact with

African Americans than North Carolinians do and thus may be less familiar

with African American speech. Second, while the monotonized stimuli were

identi¿ ed with only slightly less accuracy than the unmodi¿ ed stimuli, the

schwa-converted stimuli were identi¿ ed with considerably less accuracy. This

¿ nding suggests that vowel quality information is more important than F0-de-

pendent information, but the noise that the synthesizer created on some of the

schwa-converted stimuli may have distracted listeners. Third, female speakers

were consistently identi¿ ed with slightly lower accuracy than male speakers.

Table 12.1 Overall Accuracy Levels in Experiment A

Treatment

North Carolina

African American

listeners (n=11)

North Carolina

European American

listeners (n=33)

West Vi rg i n ia

European American

listeners (n=39)

male

speakers

female

speakers

male

speakers

female

speakers

male

speakers

female

speakers

unmodi¿ ed 97.0% 95.5% 97.5% 94.8% 85.7% 80.7%

monotonized 91.3% 89.0% 94.1% 91.0% 79.0% 77.1%

schwa-

converted

78.0% 62.3% 81.9% 64.3% 73.7% 60.2%

Table 12.2 shows the same data broken down into stimuli featuring /æ/

prominently, stimuli featuring /o/ prominently, and control stimuli. It can be

observed that, in most cases, the three types of stimuli do not differ much in