Identi¿ cation of African American Speech 2692.1 Methods
Of the 36 spea kers who were pa r t of the cor pus of recordi ngs, 12 were selected
for this experiment. They were selected on the basis of ratings they were given
by a panel of sociolinguists as to how African American or European Ameri-
can they sounded. The three speakers of each ethnicity/sex combination (i.e.,
African American females, etc.) who were rated as sounding most typical of
their ethnicity were selected. Readings of six sentences were taken for each
of the selected speakers. The ¿ rst two sentences feature /æ/ prominently, the
next two feature /o/ prominently, and the last two were intended as control
sentences because they do not include /æ/, /o/, /ai/ (as in sight or side), /au/, or
/u/, all of which are known to serve as ethnic markers:
After that, Hattie got sad and came back to the pad.
Pat sat on a hat, a cat, a bat, and a tack. (Apologies to Dr. Seuss.)
Joe hoped he could go shop for a stove.
Hoke showed up when smoke came in his Geo.
She got up early and went in ¿ rst.
He dropped three books on Ted’s front seat.The stimuli were subjected to three treatments. One was to leave them
unmodi¿ ed. The second was monotonization. The third was conversion of
all vowels to schwa. Both the monotonization and the conversion to schwa
were conducted using the Kay Analysis Synthesis Laboratory (ASL), which
is a linear predictive coding synthesizer. For monotonization, F0 was set at
120 Hz for male speakers and 200 Hz for female speakers. For schwa conver-
sion, formant values were set at F1=500 Hz, F2=1500 Hz, and F3=2500 Hz
for male speakers and F1= 600 Hz, F2= 1800 Hz, and F3= 3000 Hz for female
speakers. The ASL performed monotonization quite well, but left many of the
schwa-converted stimuli with extraneous noise.
All listeners who served as subjects heard all the stimuli: schwa-converted
stimuli in the ¿ rst section of the experimental recording, monotonized stimuli
in the second section, and unmodi¿ ed stimuli in the ¿ nal section. Within each
section, the order of stimuli was randomized and stimuli were presented to
subjects in sets of ¿ ve. A voice announced the beginning of each new sec-
tion and set. Several stimuli at the beginning of each section served to accli-
mate listeners to the task and responses to them were excluded from analysis.
Three types of listeners served as subjects: European Americans in Raleigh,
North Carolina (mostly undergraduate students at North Carolina State Uni-
versity), African Americans in Raleigh, and European Americans at West