Classi¿ cation of Regional Language Varieties 217can also be used to identify lexical items, which can then specify phonologi-
cal categories.
The representation of cross-dialect phonological differences requires a
match between two utterances at the lexical level, and a mismatch or ambigu-
ous or overlapping representations at the phonological level. For example, in
order for a listener to learn that a talker has a pen~pin merger, the listener
must encounter the phonological form /pԌn/ in a lexical context which unam-
biguously requires the lexical item pen. As Labov and Ash (1998) found, in
semantically predictable contexts, listeners are able to adapt to unfamiliar
vowel shifts as a result of the interaction between phonological and seman-
tic information. Without any supporting semantic information (as in Rakerd
and Plichta’s neutral contexts, 2003), the listener assumes a match between
his or her dialect and the dialect of the talker and does not adjust the mapping
between phonological and lexical representations. In the model in Figure
8.3, the dialect representations are connected to the lexical representations
Figure 8.3 Levels of representation in an exemplar-based model of the perception
and representation of dialect variation.