Identi¿ cation of African American Speech 279similarly. For the North Carolinians, accuracy rates were quite high for stim-
uli with both diagnostic vowels and natural prosody. Accuracy dropped only
slightly when either the diagnostic vowels were absent or the prosody was
swapped, but it dropped steeply when listeners lacked both diagnostic vowels
and natural prosody to serve as cues. This result suggests that listeners rely on
a suite of cues, and if one is missing, they can rely on another, but the absence
of multiple cues is what creates dif¿ culty.
Figure 12.3 shows the interaction of sex of speaker with natural/swapped
prosody. In this comparison, the two European American listener groups
behave similarly. For the European American listeners, the prosody swapping
decreases accuracy rates by about 8% for male speakers but makes virtually no
difference for female speakers. For African American listeners, however, pros-
ody swapping decreases accuracy rates for both male and female speakers.
In Experiment B, accuracy of identi¿ cation, rather than correlation with
phonetic measurements taken on the stimuli, was the object of the statistical
Figure 12.2 Interaction between presence/absence of diagnostic vowels and natural/
swapped prosody in Experiment B.