In connected speech, the production of a speech sound is often affected by the articulatory features of its neighboring sounds; this typically results in a different surface (phonetic) realization of the sound from its underlying form. Mismatches between surface sound and underlying sound due to contextual change in production may pose a problem in perception for listeners.
This dissertation investigates the language specificity and generality of the knowledge that allows listeners to perform surface-to-underlying sound mapping, with respect to the following aspects: 1) whether the mapping depends exclusively on language-specific knowledge, or is also facilitated by language-general articulatory knowledge; 2) whether the difficulty of mapping varies between assimilatory and dissimilatory processes, especially for naïve non-native listeners; 3) whether the difficulty of mapping is further influenced by the categoricalness/gradience of assimilatory and dissimilatory processes, for naïve non-native listeners. The above mapping issues were examined for a set of tone and tone sandhi phenomena, through a series of experiments. The results provide evidence for language specificity (in the sense of listeners’ use of native phonology in both assimilatory and dissimilatory processes) as well as for language generality (in the sense of listeners’ reliance on general knowledge of articulatory gestures in assimilatory processes) regarding surface-to-underlying sound mapping. In addition, the results suggest that language-general segmentation plays a role in the mapping for categorical dissimilatory processes (in juxtaposition to categorical assimilatory processes).
