216 Cynthia G. Clopper
(Clopper and Pisoni 2007). Both of these effects are most likely related to
the listeners’ overall familiarity with the phonological systems of different
dialects. While some evidence suggests that dialect familiarity can develop
through exposure to national and local media (Stuart-Smith, Timmins, and
Pryce 2005), the effects of linguistic experience obtained in the vowel identi-
¿ cation and dialect classi¿ cation studies discussed previously suggest that a
listener’s personal experience with variation as a result of where he or she has
lived also plays a signi¿ cant role in dialect familiarity.
The representation of phonetic detail in memory and the role of linguistic
experience in perception and cognitive processing can both be accounted for
by exemplar-based models of spoken language. The two primary properties
of exemplar models are that every experience (e.g., utterance) is represented
and stored in long term memory and that recognition of a new stimulus results
from the comparison of that stimulus to the previously stored exemplars. The
new stimulus is identi¿ ed as being the “same” as the stored exemplar(s) it
is most similar to. Representational strength of the exemplars is assumed
to decay over time and different representations can be assigned different
weights, so that recently encountered stimuli or previous exemplars produced
by the same talker are given more weight than others in perception and rec-
ognition. Exemplar models have been proposed to account for a range of lin-
guistic phenomena, including vowel perception (Johnson 1997), lexical access
(Goldinger 1996), and phonological change (Pierrehumbert 2001, 2002).
Pierrehumbert’s (2001, 2002) model of the lexicon includes representa-
tions of lexical, phonological, and phonetic information. Representations of
individual words are linked to their constituent phonemes which, in turn,
are linked to phonetic elements. In order to account for the perception and
representation of dialect variation, several additional levels of representation
must be added. As shown in Figure 8.3, a model of speech perception that
can account for the sociophonetic research described previously would also
include representations of talker, dialect, semantic, and non-linguistic social
information. In this model, it is assumed that representations at all informa-
tion levels take the form of exemplars, but this is not a crucial assumption of
the model. Abstract, symbolic representations could also be assumed as long
as the connections between the various sources of information permitted the
same effects of phonetic detail and linguistic experience to emerge.
The core linguistic component of the model includes the phonetic, pho-
nological, lexical, and semantic representations. In general, the connections
are assumed to be bi directional. Bottom-up acoustic-phonetic information
can be used to identify phonological categories, which can then be used to
identify lexical items. On the other hand, top-down semantic information