Classi¿ cation of Regional Language Varieties 205on isolated vowel productions. They found that the listeners were well above
chance in the gender identi¿ cation task even when the stimulus materials were
whispered or degraded by low pass ¿ ltering. This ¿ nding suggests that acoustic
information about the talker is available in very short speech samples, even in
the absence of a laryngeal source or formant structure, and that naïve listeners
are able to access that information in an explicit gender categorization task.
Speech scientists have also examined the interaction between linguistic
and social information in spoken language processing. Mullennix and Pisoni
(1990) found evidence of interference between linguistic and social infor-
mation in a speeded classi¿ cation task with multiple talkers. In one condi-
tion, listeners were asked to categorize isolated words produced by male and
female talkers as beginning with either a /p/ or a /b/. In a second condition,
listeners were asked to categorize the same stimulus materials by the gender
of the talker (i.e., male or female). In both cases, the listeners were instructed
to ignore the unattended dimension and to respond as quickly as possible
after each stimulus was presented. When the word-initial phoneme and the
gender of the talker were uncorrelated, signi¿ cant increases in response time
were found as the variability in the unattended dimension increased. That is,
in the phoneme identi¿ cation task, the listeners were slower to respond when
the stimulus materials were produced by more different talkers. Similarly, in
the gender identi¿ cation task, the listeners were slower to respond when more
different words were presented. This result suggests that talker-speci¿ c infor-
mation such as gender is perceived and processed in parallel with linguistic
information and can also interfere with rapid phoneme processing.
More recently, Strand (1999) examined the categorical perception of the
place of articulation of voiceless fricatives. She developed several continua of
stimulus materials that ranged from /s/ to /ߑ/ by manipulating the frequency of
the fricative noise. The continua varied in terms of their perceived gender and
gender typicality, so that the “voices” were either male or female and either
prototypical or non-prototypical of their gender. The overall results replicated
the earlier categorical perception results of Liberman et al. (1957) and revealed
a sharp shift between /s/ and /ߑ/ responses near the mid point of the continuum.
However, Strand (1999) also found that the gender of the talker had a signi¿ -
cant effect on the location of the perceived category boundary. The perceived
/s/~/ߑ/ boundary was higher in frequency for the female voices than the male
voices. In addition, the perceptual boundaries for the non-prototypical voices
were closer together on the continuum than the boundaries for the prototypi-
cal voices. Thus, identical acoustic signals were judged differently depending
on the acoustic context in which they were presented, suggesting an interac-
tion between contextual information about the talker and the perception of
linguistic categories.