208 Cynthia G. Clopper
varieties of American English. The vowel stimuli were embedded in mono-
syllabic words which were presented in isolation and at the end of seman-
tically neutral sentences. Half of the carrier sentences contained Northern
Cities shifted vowels and the other half contained unshifted vowels. Rakerd
and Plichta (2003) found that two factors interacted to cause a perceptual cat-
egory boundary shift: region of origin of the listener and preceding linguistic
context. Listeners from the upper peninsula of Michigan, where the NCCS is
not common, perceived the /ľ/~/æ/ category boundary at the same point on
the continuum regardless of the context of the stimulus item.
However, listeners from the Detroit area, where the NCCS is common
among local white speakers, perceived the boundary at a more fronted loca-
tion along the continuum when the stimulus item was preceded by a carrier
phrase containing Northern Cities shifted vowels, but not in isolation or the
unshifted context. Like the results of Labov and Ash’s (1998) word recogni-
tion study, these ¿ ndings suggest that both the listener’s region of origin and
the linguistic context of the utterance can affect the perception of local vowel
category boundaries.
Niedzielski (1999) also used continua of synthetic vowel stimuli to exam-
ine the perception of the NCCS in Detroit. In her task, she asked listeners
to match the vowel sound in a target word to one token from a set of syn-
thetic vowel stimulus items. While all of her target items were produced by a
single female speaker with Northern Cities shifted vowels, her listeners did
not always select the most acoustically-similar synthetic vowels as the best
match. In particular, Niedzielski (1999) manipulated the biases of her listen-
ers by telling half of them that the talker was from Detroit. The other half of
her listeners were told than the talker was from Canada. Niedzielski (1999)
found that the listeners who believed the talker was from Detroit tended to
select canonical unshifted vowels as the best match, whereas the listeners
who believed the talker was from Canada tended to select the most acousti-
cally-similar vowels as the best match. As in the studies by Strand (1999) and
Mullennix and Pisoni (1990), in which linguistic and social information in
the speech signal were found to interact in perception, the results of this study
suggest that the listener’s beliefs about a talker can have a signi¿ cant impact
on the perception of the linguistic information in the speech of that talker.
Taken together, these studies have applied methods developed by speech
scientists to explore important research questions in sociophonetics and have
extended our understanding of the perception of vowel categories by naïve lis-
teners. In particular, these ¿ ndings have shown that vowel perception is more
complex than a simple mapping between the acoustic signal and a phono-
logical vowel category representation. The listener’s experience with speci¿ c