Speech Science Forum - Hielke Vriesendorp (Utrecht University)

Abstract:

Do we recognise a Yorkshire speaker saying ‘way’ as Yorkshire because we’ve got memories of other times we’ve heard Yorkshire speakers say ‘way’ before, or because we recognise the abstract concept of a monophthongal FACE vowel in their pronunciation of ‘way’? Work on exemplar theory suggests that the way listeners are able to recognise the social meaning of linguistic variation is because social information is a part of the detailed (episodic) memories of words - exemplars - we use to process speech (Johnson 2006, Walker & Hay 2011). However, as new innovations within the theory open up the possibility that listeners may generalise across these exemplar memories (hybrid exemplar theory; Pierrehumbert 2002, Goldinger 2007) to create more abstract linguistic representations such as phonemes, allophones, lexemes, etc.. This raises the question of whether it is still the detailed exemplars that carry social meaning, or whether they are attached to more abstract linguistic representations like allophones. I discuss two accent recognition tasks which show that it is most likely allophonic representations which carry social meaning, rather than lexical exemplars.