Processing social meaning in exemplar theory: Testing the role of lexical and sub-lexical representations in accent recognition

The way sociolinguistic research tends to conceptualize how social meaning is perceived is seemingly at odds with how it is conceptualized in the main cognitive speech processing theory used in sociolinguistic processing research: exemplar theory. The sociolinguistic research more broadly often focuses on how sub-lexical, often segmental variants are perceived and used to convey social meanings, while exemplar theory posits that social meaning is processed as a part of the detail in the highly detailed episodic memory traces of words, called ‘exemplars’. More recent iterations of exemplar theory, however, have opened up the possibility that representations of different sizes and levels of detail may also be used in speech processing – but it is still unclear which of these are in fact used to process social meaning. This paper presents findings from two matched accent recognition experiments, which show that listeners are able to recognize three accents (General American English, Standard Southern British English, and Yorkshire English) even in non-words, but that they perform significantly better in real words, especially if they have extensive experience with the accent they hear. This provides evidence that both sub-lexical and lexical representations play a role in sociolinguistic processing, but that lexical representations are dominant.