【作者】 尹玉霞；

【导师】 张洪明；

【作者基本信息】 南开大学 ， 比较语言学， 2012， 博士

【摘要】 本文以汉语方言丰富复杂的连读变调现象为语言实证语料，对以规则为基础的生成音系学理论和优选论进行比较研究，检验生成音系学框架下这两种理论范式对连读变调的解释充分性。本文的目的一方面是通过这种比较，推动理论的不断改进、发展、完善，能够更好地解释自然语言中的音系现象，逼近人头脑中音系知识的本质，另一方面是通过这种比较，深入我们对汉语连读变调现象的正确认识和理解。汉语方言连读变调成因是多样的，反映历时声调演化过程的连读变调在共时平面上是任意的、没有动因的，想要弄清变化的深层原因，只能通过历史比较方法重构原始声调系统。这种由于历时声调系统演化导致的共时音系的不自然性，以规则为基础的音系理论可以通过语言特有的规则、规则的外部排序或一些词汇例外标记来分析。而优选论通过语言普遍的、自然的制约条件的交互作用解释共时音系模式，语言特有的制约条件与优选论的理念是不兼容的。这种反映历史上声调系统演变的连读变调不能通过语言普通的制约条件来解释，支持规则为基础的音系理论，表明语法的音系部分应该包含规则。反映共时平面上语音同化与异化的连读变调，优选论能够解释声调为什么变化及如何变化。形态相关的连读变调，词汇音系学在处理上强于优选论，这说明特定理论有其针对性和适用性。汉语连读变调方向的选择涉及到规则的反溃给和反阻断应用顺序导致的音系不透明现象，优选论解释音系不透明现象的最新理论－优选论候选项链（OT-CC）理论处理连读变调方向时，不能提供任何有见地、启发性的认识。OT-CC的本意是为了维护和支持OT在音系学研究中的地位，但是该理论把当初经典OT抛弃的有序推导重新引入其形式框架，严重动摇了OT当初立论的基础，消弱了经典OT在处理全局性现象上的理论优势。另外在处理汉语连读变调和其它语言不透明现象中渐渐显露的弊病，反而从侧面支持了规则为基础的音系理论。汉语连读变调辖域在句法与音系接口研究中一直处于重要位置，厦门和上海方言是最初支持传统规则框架下基于边界理论的语言证据，但是最近它的提出者Selkirk放弃了该理论，提出了优选论框架的下的匹配理论，认为该理论同样能够分析最初支持基于边界理论的语言证据，并且能够解释音系现象作用域参照句法范畴两个边界的语言现象。虽然匹配理论能够分析厦门方言，但是却不能对上海方言的连读变调辖域正确预测。优选论框架下的同界（align）和封装（wrap）理论和匹配理论都是从间接模式发展起来的句法与音系接口理论，平遥方言连读变调模式所涉及的复杂的句法条件，界定为韵律单位以后，仍然不能成功被解释，平遥方言似乎支持直接参照模式。普通话的连读变调辖域对音系与句法提出什么问题，仍然没有统一的认识，但是根据新句法与音系接口理论分析普通话时错误地预测相同句法结构的句子有相同的韵律结构，我们认为普通三声变调的辖域无需设立韵律短语层级。简言之，优选论框架下匹配理论不能取得与传统的基于单一边界理论相同的解释力。更多还原

【Abstract】 This dissertation uses rich and complicated tone sandhi across Chinese dialectsas empirical data to compare and examine the explanatory adequacy of rule-basedphonology and optimality theory with the following two aims: on the one hand topromote the reconceptualization, extension and improvement of phonological theoryto best capture the phonological phenomena of natural languages, thus makes usmore close to the nature of phonological component of grammar, on the otherhand to deepen our understanding of tone sandhi in Chinese.The causes of Chinese tone sandhi are various. Tone sandhi reflecting thehistorical changes in the process of tonal evolution is quite arbitrary andunpredictable in synchronic level. If we want to find the deep reason, historicalcomparison is needed to reconstruct the ancient tonal system. Historical tone changesresult in the unnaturalness of synchronic phonologies and those unnatural aspects ofphonologies can be accounted for in rule-based phonological theories by adoptinglanguage-specific rules, extrinsic ordering of rules or lexical marking of exceptions.Optimality theory accounts for the synchronic phonological patterns by universalconstraints interaction, however, the arbitrary, idiosyncratic tone sandhi phenomenabrought out by the historical tone change can not be explained by universalconstraints and adopting language specific constraints is not compatible with thetenets of optimality theory which is strong evidence for rule-based phonology thatphonological component of grammar should contain rules. Those tone sandhireflecting the phonetic assimilation and dissimilation can be captured in optimalitytheory. Optimality theory not only can explain why the tone change but how the tonewill change. Tone sandhi related with morphology can be better accounted for inrule-based lexical phonology than in optimality theory indicating theory has itsapplicability and pertinence.Tone sandhi directionality is phonological opaque due to the counter-feeding andcounter-bleeding rule relation. Optimality theory with candidate chain (OT-CC) designed to deal with phonological opacity fails to provide insightful and inspiringunderstanding on the reasons of governing the directionality of rule application. Inorder to increase the explanatory power for opacity, OT-CC incorporates derivationinto optimality theory which abandoned the central theoretical tenets of classicoptimality theory and weakened its advantage over rule-based phonology on globalphenomena. In addition, OT-CC fails to capture directionality of Chinese Tone sandhiand opacity in other languages, which indicates the advent of OT-CC, instead ofsupporting OT, at some extend give its support for rule-based phonology.Chinese tone sandhi domain has come to occupy a prominent position in theresearch on the syntax-phonology interaction. Xiamen and Shanghai data areoriginally taken to motivate single edge-based theory put forward by Selkirk.Recently Selkirk put forward Match theory to replace single edge-based theory.Match theory is late syntax-phonology interface theory in the framework ofoptimality theory. Selkirk claims that Match theory can not only equally consistentwith the data for single edge-based theory but also can accommodate data whichdiagnose two edges, but Match theory can account for tone sandi domain of Xiamenand fail to predict tone sandhi domain of shanghai. Align and Wrap theory and matchtheory are originated from indirect reference idea on syntax-phonology interface.Pingyao tone sandhi is sensitive to complicated syntactic constructions. The syntacticconditions even are defined as prosodic units, the tone sandhi patterns still can not beexplained very well. It is likely that Pingyao tone sandhi domain need direct referencesyntactic information. It is still debatable what kinds of questions of tone sandhidomain of Mandarin tell us about the syntax and phonology interaction. On the basisof the wrong predication that new syntax and phonology interface theories inframework of optimality theory that the sentences with the same syntactic treestructures have the same prosodic structures we argue that tone sandhi in Mandarindo not need posit phonological phrase level. In short, Match theory in framework ofoptimality theory can not achieve the equally explanatory adequacy as the rule-basedsingle edge-based theory.更多还原