Tone-TBU Disparity in Three Phonologies of Cantonese

I propose a restricted theory of tone-TBU association called Basic Tonal Theory (BTT). BTT includes six well-attested constraints, and a restricted GEN. The typology contains attested cases of tone overcrowding resolution such as tone deletion, mora epenthesis, and mora deletion, but also predicts unattested cases of languages that resolve tonal crowding by using a combination of the aforementioned mechanisms based on the number of the morae in the input. Cantonese has both monomaraic and bimoraic syllables, lexically specified tones, and a process of tonal suffixation that creates the context for tonal overcrowding. In an acoustic experiment, the duration and pitch of pairs of suffixed and underived forms were measured, and the analysis of the results shows that two of the three groups use a split-strategy to resolve tonal crowding. The experiment thus provides empirical evidence for the typology of BTT. Furthermore, the proposed analysis can account for a case of near mergers previously argued to be unexplainable in an optimality framework. 1 Basic Tonal Theory (BTT) 1.1 Gen BTT is a restricted theory of tone-TBU association. It generates only one syllable at a time, in which (t) is used to indicate any tone and the input and output are as follows: Input = {[t]μ , [tt]μ, [tt]μμ, [ttt]μμ} Lexical and changed tones (CVOlex, CVOtc, CVRlex, CVRtc) Output ∈{[t]μ} A sequences of 1 or more t followed by 1 or more u Correspondence: with notation: α symbol,α` output , and ↔ ‘corresponds to’ which typed as t ↔ t` and μ ↔ μ`. Each x ∈ in corresponds with 1 and only 1 xʹ ∈ out, and each xʹ ∈ out corresponds with 1 and only 1 x ∈ in. 1.2 CON The constraints (CONBTT)are: f:max-t t ∈ in ⇒∃tʹ∈ out s.t. t ↔ tʹ f:dep-t tʹ ∈ out ⇒∃t ∈ in s.t. t ↔ tʹ f:dep-μ μʹ ∈ out ⇒∃t ∈ in s.t.μ↔μʹ M:σ=μμ card {μ ∈ out} σ contains exactly two μ M:noCompC card {μ ∈ out} – 2σ contains at most two t (Yip, 2002) M:noShortT card {t ∈ out} – card {μ∈ out} for each μ in the output there is at most one t in the output (Yip, 2002) * I would like to thank Roxana Fung, Ken Cheng, Ou Jinghua and Kwok Ciu Lai for their help and support at PolyU, 赵子 at Jinan University in Guangdong for offering me access to the laboratory, finding me the speakers and for being an exquisite host, and indeed all who volunteered to take part in the experiment. I would also like to thank Paul de Lacy for enormously improving this handout with structural and substantial suggestions, Alan Prince for pointing out theoretical issues and making this handout clearer, Aaron Braver for sharing with me his expertise on incomplete neutralization. I am also grateful to Giovanna Marotta, Alan Yu, William Bennet, Eric Baković and all the attendees and organizers of Phonology 2013 for the comments and suggestions. Luca Iacoponi Tone-TBU Disparity in Three Phonologies of Cantonese 2 Other worthy-mentioning constrains that determine TBU-Tone association are M:specifyT, M:nolongT, and F:NoContour.M:specify requires at least one tone to be specified per syllable. This constraint is embedded in GEN in the current model. M:nolongT (C.F. M:noshortT) is irrelevant for tonal overcrowding: card {μ ∈ out}– card {t ∈ out}. In the theory, the morae are part of the input. However, the system is also an abstraction of models where the moraic structure is introduced in the output. The absence of a constraint on mora deletion (f:max-μ) makes the system a valid approximation, so long as segmental changes are not considered. The typology generated by this set of constraints is a sub-set of BTT’s. BTT predicts that tonal crowding can be resolved via moraic epenthesis, deletion of the root tone, faithfulness, tone shortening, or downstep (fig. 1). Those cases are well attested. The theory, though, also predicts peculiar cases where a different strategy can be chosen in a language depending on the different number of tones and TBU in the output (split strategy languages). For example, a language may choose to epenthesize a mora in monoraic domains with one extra-tone, and then leave the same tone unparsed in case of bimoraic input with two tones. These cases are hard to find evidence for, since the language has to show tonal crowding, as well as showing 1-to-1 tone-TBU association in monomoraic and bimoraic syllables. In the next sections I will show that Cantonese satisfies these requirements, and that two varieties of Cantonese are actual instantiation of split strategy languages. Figure 1: Possible resolutions of tonal crowding 2 Cantonese 2.1 Tonomatics Cantonese is a Sinitic language of the Yue family which is spoken in the East Guangdong/West Guangxi region, in Southern China, Hong Kong, Macau, and in most Chinatowns worldwide. It is the native language of approximately 60 million people (Lewis et al., 2013).It is widely documented and attested in both theoretical and descriptive literature (see Bauer and Benedict, 1997 for a review). There are nine tones of contours consist of 3 levels, 3 contours and 3 long level tones. The 3 levels are L, M, and H. The 3 contours are LM, MH, and ML. Three long level tones: HH, MM, and LL.Cantonese distinguishes between two syllable types CVO and CVR (Chao, 1974; Bauer and Benedict, 1997 and other).CVO syllables end in unreleased stops (/p, t, k/) and they are shorter than long syllables. Contours occur only in TC tones (HM). CVR syllables end in sonorants. They are longer than CVO syllables and lexically specified with a contour or a long level tone (HH, MM ... LM, MH).The TBU is the mora in Cantonese (Iacoponi 2013). Codas are moraic in CVR syllables, but non-moraic in CVO syllables. Evidence from processes that changes only one tone in the syllable (see for instance 2.3) shows that long level tones are composed of two level tones. The tonal inventory as described in table 1.The structure proposed in (7-10). 2 This number actually refers to all Yue dialects. The speakers of Cantonese stricto sensu are about 10 million. Luca Iacoponi Tone-TBU Disparity in Three Phonologies of Cantonese 3 Figure 2: Representation of an HH tone in Cantonese IPA+Chao Gloss Character Tone Letter sik “style” 式 H sik “to stab” 刺 M sik “to eat” 食 L siː “to think” 思 HH siː “history” 史 MH siː “to try” 試 MM siː “city” 市 LM siː “fact” 事 LL siː “time” 時 ML Table 1: Tones of Cantonese 2.2 High Rising Tone Change The term high raising tone change (TC) is used to describe an array of processes that due to the grammar of Cantonese all result with the same output (Iacoponi, 2013). It is one of the few tonal alternations in the language. In high raising tone change (TC) all tones change to a tone similar or identical to lexical MH (MHtc vs MHlex), with an exception of high tones, which are claimed to remain unchanged (Bauer, 1997). The array of processes that forms TC is: Red/Prefix: (1) /gwaigwaiH+dei/ → [gwaigwaidei] “It’s a bit expensive” 贵贵哋 c.f. [gwai] “expensive Suffix (voc): (2) /atsiu+H/ → [atsiu] “Chiu (voc/suffix)” 阿赵 c.f. [tsiusinsang] “Mister Chiu” 赵先生 Suffix (diminutive): (3) /douse+H/ “to pour out” → [douse] “to spill” 倒瀉 Syllable Elision: (4) /ŋɔpɛk t ͡ ʃɔligɔ/ → [ŋɔpɛkligɔ] “I chopped this (elision)” c.f. [ŋɔpɛk t ͡ ʃɔligɔ] “I chopped this (slow speech)” 3 Only H tones can trigger TC, see Iacoponi (2013) for a discussion. CVO