This dissertation investigates speech processing of Hai-lu Hakka falling tones (high, low) and tone sandhi (low-rising, high-checked) by focusing on how variations influence the perception of the contrast of falling tones and the application of tone sandhi. Some recent studies (Bent 2005; Huang 2004; Wang 1995; Zhang & Lai 2010; Zhang et al 2011) indicated that tonal contrasts may not be as distinctive as assumed previously and that tone sandhi may not be fully productive, since earlier studies tended to neglect the role variations play in tonal processing. In addition, the target language Hai-lu Hakka has been understudied, and was found to undergo language attrition in recent decades (Yeh 2011; Yeh & Lu 2013; Yeh & Lin 2015). The dissertation hopes to clarify how tonal variations affect tonal contrasts and tone sandhi application by explicating Hai-lu Hakka’ tonal processing. The first experiment examines 30 Hai-lu Hakka speakers’ contrast of falling tones by tonal identification and lexical recognition tasks. These tasks require participants to identify monosyllabic stimuli’s tonal categories (either high-falling or low-falling) and to recognize their lexical meanings (either a high-falling word [ti53] ‘to know’ or a low-falling word [ti31] ‘emperor’). The second experiment examines 31 Hai-lu Hakka speakers’ processing of two sandhi rules by tonal discrimination, tonal identification, lexical recognition, and production tasks. These tasks require participants to discriminate underlying and sandhi tones, to identify their tonal categories, to recognize their corresponding meanings, and to produce tone sandhi in disyllabic compounds. The results show that (i) various lexical and phonetic/phonological factors play a role in the tonal contrast and the sandhi processes, (ii) lexical and phonetic/phonological effects can be more or less significant in different tasks, and (iii) lexical factors exhibit a more consistent pattern across different tasks than phonetic/phonological factors. These findings indicate three general implications as follows. First, the lexical and phonetic/phonological effects indicate that the contrast of two falling tones varies with both lexical and phonetic/phonological factors, so does the application of two sandhi rules. These factors may result in tonal confusion between two falling tones and lower application rates of two sandhi rules. Second, the lexical and phonetic/phonological effects indicate that variations occur frequently and extensively across different tasks in the two experiments, which calls into question those derivational theoretical models that assume a transition from variations to invariance during speech processing. Third, the results indicate that lexical effects exhibit a more consistent pattern than phonetic/phonological effects, since phonetic/phonological effects exhibit an asymmetry between perception and production patterns. These findings suggest a more crucial role of lexical influences in tonal processing, which hence favors a lexically-based model for speech processing in general. To sum up, this dissertation has three general contributions. First, relating phonetic modifications/processes to variations and phonological modifications/processes to invariance is called into questions. Second, the findings uncover specific issues such as tonal confusion and a perception-production asymmetry in Hai-lu Hakka’s tonal phonetics and phonology, supporting the crucial need to carefully examine a less-studied language like Hai-lu Hakka. Third, the findings indicate a more crucial role of lexical influences in tonal processing, suggesting that an exemplar-based processing model that argues for a non-derivational and lexically-driven account is a more reasonable approach to speech processing.
Read
