A Reader in Sociophonetics

Identi¿ cation of African American Speech 281

West Virginians are not exposed on a daily basis to North Carolina vowels,
and hence they are not sensitive to variants—for example, /o/ with a fronted
glide—that are common in North Carolina but not in West Virginia. At the
same time, the low proportion of African Americans in West Virginia, par-
ticularly in northern West Virginia, where the university is located, may
have made the West Virginia subjects less sensitive to vowel variants such
as raised /æ/ that typify African American English. At any rate, the fact that
the West Virginians were not accessing vowel variation to any signi¿ cant
degree probably explains their poor performance in Experiment A, which
put a heavy emphasis on vowel variation.

4. Conclusion

The results of the two experiments described here paint a more complicated
picture of ethnic identi¿ cation than previous work. First, it appears that no
single cue serves as the silver bullet for listeners. Instead, listeners rely on
several cues, including cues from segmental quality, prosody, and voice qual-
ity. In fact, what cue is most important may not be the appropriate question to
ask. Listeners exhibit a degree of À exibility. Both experiments, though espe-
cially Experiment A, as well as the earlier results from Thomas and Reaser
(2004b), suggest that listeners shift the cues they use depending on the sex of
the speaker. Furthermore, as Experiment B suggested, they may shift the cues
they use depending on what information is available to them in the speech
that they hear.

Table 12.4 Logistic Regression Results by Listener Group and Independent Variable

Independent

variable

North Carolina

African American

listeners (n=11)

North Carolina

European

American

listeners (n=24)

West Vi rg i n ia

European

American

listeners (n=25)

presence/absence of

diagnostic vowels

Ȥ^2 =18.57

p<.0001

Ȥ^2 =33.35

p<.0001

Ȥ^2 =0.48

p<.4900

natural/swapped

prosody

Ȥ^2 =12.58

p<.0004

Ȥ^2 =17.63

p<.0001

Ȥ^2 =8.81

p<.0030

sex of speaker Ȥ^2 =25.44

p<.0001

Ȥ^2 =0.93

p<.3351

Ȥ^2 =3.79

p<.0515

ethnicity of speaker Ȥ^2 =13.83

p<.0002

Ȥ^2 =2.45

p<.1177

Ȥ^2 =0.33

p<.5664