Classi¿ cation of Regional Language Varieties 207varieties of American English (see e.g., Labov, Ash, and Boberg 2005; Thomas
2001). As a result, most of the cross-dialect perception research dealing with
linguistic categories has focused on vowel perception. Using both natural and
synthetic stimulus materials, vowel perception has been examined in vowel
identi¿ cation and word recognition tasks. The word recognition tasks have
typically included multiple conditions, in which the words were presented in
isolation, in semantically neutral sentence contexts, and/or in semantically
predictable sentence contexts. In addition, vowel identi¿ cation performance
by local listeners has been compared to responses to the same set of stimulus
materials by non-local listeners. The results of these studies provide insights
into the perception of well-documented regional vowel shifts by naïve listen-
ers, the role of linguistic context in producing cross-dialect lexical interfer-
ence, and the role of familiarity with a given variety in perception.
In one set of studies, Labov and Ash (1998) used two different tasks to
obtain converging evidence for the effects of linguistic context and the lis-
tener’s region of origin on vowel perception. First, they asked naïve listeners
in Birmingham, Philadelphia, and Chicago to identify naturally produced
Birmingham vowels in kVd utterances in an open-set identi¿ cation task.
They found that the Birmingham listeners were more accurate overall than
the other two listener groups, although the same vowels which were dif-
¿ cult for the Philadelphia and Chicago listeners were also dif¿ cult for the
Birmingham listeners.
Labov and Ash (1998) then conducted a word recognition task, in which
they played progressively longer samples of speech in semantically predict-
able contexts and asked listeners to identify the target word. The speech sam-
ples were again produced by Birmingham speakers and the listeners were
from Birmingham, Philadelphia, and Chicago. The region of origin of the lis-
teners signi¿ cantly affected word recognition accuracy for the phrase-length
utterances. The Birmingham listeners were again more accurate than the non-
local listeners, although overall performance varied greatly depending on the
target vowel. Region of origin was not a signi¿ cant factor in determining per-
formance for the word-length and sentence-length utterances due primarily
to À oor and ceiling effects, respectively. Thus, Labov and Ash (1998) found
evidence of an interaction between linguistic context and region of origin of
the listener in the perception of local vowels.
Using a classic categorical perception paradigm, Rakerd and Plichta
(2003) asked naïve listeners to identify a series of synthetic vowel stimuli as
/ľ/ or /æ/. The stimulus materials were constructed such that the middle part
of the continuum was ambiguous between a fronted /ľ/ characteristic of the
Northern Cities Chain Shift (NCCS) and an unshifted /æ/ typical of many