Linguistic experience and social factors in speech perception: The case of merged speakers of Mandarin sibilants

This work explores the combined effects of social expectations and a speaker's production characteristics on the perception of alveolar versus retroflex sibilants that are variably merged in Taiwan Mandarin. The variation is socially structured in that the sibilant merger is regarded as a characteristic feature of speakers from southern Taiwan. The results of an AXB discrimination task showed that although merged speakers were outperformed by their distinct counterparts, they were able to discriminate the sibilants far beyond chance level. In an identification task with social guises, participants showed a pattern reflecting the implicit bias that a southern-labeled talker is less likely to produce retroflexes, and hence use the merged form, than a northern-labeled talker. Interestingly, merged participants were again shown to be less sensitive to frication noise cues, but they more readily switched between the social and acoustic cues than distinct participants. Together, these results indicate that frequent encounters with distinct forms in a speech community with large interspeaker variation might help merged speakers remain sensitive to phonological distinctions that they do not carry. Merged speakers might have been desensitized to the acoustic cues to some degree; however, they appear to use other cues to achieve coherent speech perception whenever possible.