Identi¿ cation of African American Speech 267these many studies is simply that listeners are capable of accessing a variety
of cues under experimental conditions.
The fact that listeners access a particular cue during an experiment in
which their attention may be arti¿ cially drawn to that cue does not demon-
strate that they utilize that cue under more natural conditions. However, only
experiments can tease out the effects of different cues. Hence, a different kind
of experiment, one that compares the effects of cues, is called for. Purnell et
al. (1999) and Thomas and Reaser (2004b) are perhaps the only studies that
have compared the effects of different cues directly. Other studies, includ-
ing Alvarenga (1971), Lass et al. (1978), Lass et al. (1980), and Thomas and
Reaser (2004a), have done so indirectly by using such techniques as playing
stimuli backward or using lowpass-¿ ltered stimuli. Purnell et al. (1999) used
guises uttered by the same speaker as stimuli, which, while eliminating spuri-
ous factors due to individual differences in speakers’ voices, could introduce
other sources of error if the impersonator cannot control all the cues that are
relevant for identi¿ cation of ethnicity.
These issues related to ethnic identi¿ cation of voices have real implica-
tions for society. A better understanding of how listeners distinguish the eth-
nicity of voices would make it easier to demonstrate how ethnic pro¿ ling can
take place, as in the denial of appointments for jobs or residences on the basis
of the ethnic identi¿ cation over the telephone. Purnell et al. (1999) discuss
the legal implications of ethnic pro¿ ling in some detail. The ability to dis-
tinguish ethnicity can also perpetuate stereotypes, as when children identify
the voices of cartoon characters and associate them with personality traits
exhibited in the cartoon; see Lippi-Green (1997: 79–103). However, in cer-
tain cases, ethnic identi¿ cation of voices could have desirable consequences.
For example, patients who have lost the ability to speak and who are African
Americans may desire speech synthesis equipment that can produce a voice
that matches their personal identity better than the dialect-neutral synthe-
sizers now available. The social importance of the ethnic discrimination of
voices necessitates further research into the mechanisms of such distinc-
tions. In order to explore these mechanisms, we devised two experiments,
henceforth called Experiments A and B, comparing the effects of different
phonetic cues in ethnic discrimination.
For both of these experiments, the stimuli were taken from a corpus of
speech obtained from students at North Carolina State University. Thirty-
six students, divided approximately evenly between African Americans and
European Americans, were recorded in a soundproof booth performing sev-
eral tasks. They answered a few demographic questions and then recounted a
familiar children’s tale (in order to elicit spontaneous speech), read a list of