In Exemplar Theory, speech perception is informed by any sorts of frequently experienced co-variance between functionally relevant nodes, including those formed through socially-conditioned lexical use. Accordingly, previous research demonstrated that words associated with a particular age group are recognized faster when they are produced by talkers from that age group. While such effects arose from multiple socially-indexed sub-phonemic cues (e.g, voice quality, pronunciation styles), a phonetic priming experiment reported here tested whether exposure to a single phonetic variant is sufficient to guide probabilistic inference for socially-conditioned words. In each trial, participants (N=40) heard two female talkers (age: 40, 41) producing a priming word, the initial stop of which was acoustically manipulated in VOT and F0 dimensions into either of two age-related realizations (i.e., young vs. old guises) based on an ongoing sound change of phrase-initial stops in Seoul Korean. The priming word was followed by a lexical decision task in which listeners pressed a button indicating whether the auditory stimulus produced in the same voice was a real word. As a result, reaction times for young-associated words (e.g., notap, ‘no solution’) were shorter when the target appeared after the prime word with the young guise than old guise, and vice versa for old-associated words (emem, ‘mother’). The results suggest that phonetic, lexical, and social cues are so closely tied as storage information within the lexicon that a subtle auditory input that may not explicitly activate listeners’ awareness of talker age difference can directly index words associated with similar social information.
