Classi¿ cation of Regional Language Varieties 219judgments typically reÀ ect social stereotypes, they are also connected to
other semantic representations, as well as individual talker representations.
The model shown in Figure 8.3 supplements a relatively traditional model of
phonology (the black ovals) with representations of some of the kinds of social
information (the white ovals) that have been shown to affect speech perception
and spoken language processing. In particular, the model suggests mechanisms
for the interference between linguistic and social information in speech percep-
tion, the poorly de¿ ned dialect representations exhibited by naïve listeners in
the dialect classi¿ cation tasks, and the role of acoustic and semantic context in
the perception of local and non-local vowel systems. The bidirectional connec-
tions between the nodes account for perceptual learning of linguistic and social
categories as well as later recognition of those categories. The model can also be
interpreted with respect to second dialect acquisition and speech production more
generally, where similar kinds of relationships would be predicted to exist.
The model in Figure 8.3 also suggests several future directions for socio-
phonetic perception research. First, the nature of the connections between
the phonological information and the dialect representations must be more
fully described. The speci¿ c phonological variants that are salient to naïve
listeners have not been identi¿ ed and the role of mediation by semantic infor-
mation and the lexicon in constructing dialect representations must be inves-
tigated in much greater detail. Second, the role of dialect mismatch between
the talker and the listener should be further explored at the level of word
recognition and spoken language processing. Most of the existing research
on dialect perception is focused on the explicit identi¿ cation of either pho-
nological or social categories, but cross-dialect interference in speech per-
ception may have deeper implications for sociolinguistic interaction, models
of language processing, and listener populations with speech, language, or
hearing impairments. Finally, we have only scratched the surface of questions
relating to linguistic experience and the ability of naïve listeners to access
speci¿ c dialect representations in both speech perception and production. In
all three of these research areas, speech science can provide baseline methods
and results on which to build a research program in sociophonetics.
Acknowledgments
This work was supported by grants from the National Institutes of Health to Indiana
University (T32 DC00012 and R01 DC00111) and to Northwestern University (F32
DC007237).