254 Nancy Niedzielski
knowledge, but that this serves to create and maintain self-identity, particu-
larly in the case of the linguistically secure speaker.
- Research on implicit knowledge
At its most basic level, implicit knowledge of variation is revealed through
processes such as language accommodation. According to Communicative
Accommodation Theor y (Giles and Coupland 1991), speakers move towards
(or, in some cases, away from) their interlocutors during conversation. Early
studies revealed that this happened at broad discourse levels such as speech
volume, speech rate, and gestures, but research on speech perception in
psychology reveals that this occurs at much more ¿ ne-grained, perhaps
phonetic levels. For instance, research on shadowing (e.g., Goldinger 1998)
shows that listeners can determine with surprising accuracy whether a lexi-
cal item was produced as a “shadow” of another speaker’s production of that
item. Figure 11.1 shows that subjects are able to match the item that was
produced as a shadow at levels signi¿ cantly above chance, and they are even
more accurate if the item is produced after several repetitions of shadowing,
and if the item is a lower-frequency lexical item. Interestingly, the bottom
graph shows that the tendency of speakers to accommodate to others dimin-
ishes with time; subjects were less accurate if the shadow was delayed. This
suggests that speakers are implicitly aware of low-level features of their
interlocutors’ production, even to the extent that they mimic those features
in their own production.
Studies looking more overtly at the perception of language variation
reveal to what degree subjects are aware of the patterning of variation in their
speech communities. Hay, Warren, and Drager (2006) tested subjects’ abilities
to distinguish lexical items involved in a merger in progress in New Zealand
English (speci¿ cally the “beer/bear” merger). Subjects were asked to identify
lexical items presented aurally, and following the test, they were asked what
age they thought the speaker was. They found that subjects were more accu-
rate in distinguishing potentially merged lexical items if they thought that
the speakers were older, suggesting an implicit awareness of the fact that the
merger is correlated with age.
Koops, Gentry, and Pantos (2008) also tested implicit knowledge of the corre-
lation between mergers and age, this time in Houston, Texas. Older Anglo speakers
who are native to Houston merge high and mid front lax vowels pre-nasally, while
these vowels are unmerging in the dialect of younger Anglos. Language attitude
tests show that Houstonians correlate the merger with region, particularly urban