320 Thomas C. Purnell
4.4 Summary
Again, discriminant analyses suggest that variation by ethnically af¿ liated
dialect can occur in the interface between acoustics and perception. Spe-
ci¿ cally, we saw that the three dialects are well separated in discriminant
space, both acoustically and perceptually (ASCC is nearing 1 and the models
account for 60% to 80% of the variation). However, there was an observable
difference in the selection of measures for latent factors and in the weight of
measures across acoustic and perceptual analyses. Nevertheless, the domi-
nant factor (or factors) is the tension in the vowel space distinguishing /ܭ/
from /o/. As such, it is instructive to compare these ¿ ndings with what was
reported earlier (cf. creaky voice and vowel formant differences reported in
Thomas and Reaser, 2004, Thomas, this volume, Roeder, this volume). Pur-
nell et al. (1999) reported that F2 in /ܭ/ was important for distinguishing SAE
from AAE and ChE, and that the F0 peak also distinguished AAE from SAE
tokens. While these two reported measures are related to the ¿ ndings in the
present work, the previous work did not combine measures which might have
Figure 13.11 Vowel space for vowels in “hello” (inner ellipses for /ܭ/ and /o/ = AAE;
wider /ܭ/ ellipsis = SAE).