268 Erik R. Thomas, Norman J. Lass, and Jeannine Carpenter
sentences, read a dialog, and read a list of words. The sentences were designed
so that they either highlighted or avoided speci¿ c items, such as a particular
vowel. The stimuli for both of the experiments described here came from the
recorded sentences.
An earlier experiment (Thomas and Reaser 2004b) was also based on
this corpus, but included stimuli from the speakers’ spontaneous speech and
the dialog. In that experiment, the stimuli were subjected to three treatments:
they were lef t u n modi¿ ed, they were monotonized, and they were low pass-¿ l-
tered at 660 Hz. Monotonization eliminates F0-dependent information, such
as jitter and most aspects of intonation. The lowpass ¿ ltering was designed
to eliminate most of the discernable aspects of vowel quality differences that
are correlated with ethnic differences, since those differences depend largely
upon F2 and lowpass ¿ ltering at 660 Hz nearly always removes F2. A variety
of acoustic measurements were obtained on the stimuli. The results showed
that vowel quality differences—mostly those having to do with /o/, /u/, or
/æ/ (as in bat)—were important cues for the ethnicity of all speakers. Of the
vowel quality differences that were analyzed, the glide of /o/ showed the high-
est level of signi¿ cance. However, the sex of the speaker made a difference
for other cues. Phonation emerged as the most important cue for the ethnicity
of male speakers, with breathier speech associated with African Americans.
However, phonation showed no correlation at all with identi¿ cations of eth-
nicity for female speakers. Conversely, F0-dependent factors proved signi¿ -
cant for female speakers but not for male speakers.
- Experiment A
Experiment A was designed to compare the effects of /o/ and /æ/ against each
other and against certain features of voice quality and prosody. Numerous
studies have found that European Americans, both inside and outside the
South, tend to show more fronted variants of /o/ than African Americans do
(Hall 1976; Graff et al. 1986; Thomas 1989a, b, 2001; Thomas and Bailey
1998; Fridland 2003). In fact, the fronting of /o/ represents part of a general
f r o nt i n g of /o /, / u /, a n d /ࡱ/ (as in good) that, as a whole, is more advanced
among European Americans than among African Americans. A smaller num-
ber of studies have reported a tendency for the front vowels /æ/, /͑/ (as in bet),
and /Ԍ/ (as in bit) to be higher in African American English than in European
American English (see Thomas 2001). An advantage of focusing on /o/ and
/æ/ is that they represent the back vowel subsystem and the front lax vowel
subsystems, respectively, and thus are relatively independent of each other.