【作者】 刘艳；

【导师】 倪传斌；

【作者基本信息】 南京师范大学 ， 英语语言文学， 2020， 博士

【摘要】 在句法与语义研究领域,生成语言学和认知语言学两大阵营早就存在着纷争。自20世纪50年代以来,Chomsky“句法自治”理论(Autonomy of Syntax)伴随其生成语法理论经历了“语义缺席--解释语义--语义紧缩--模块论阐释--方法论自然主义阐释”五个阶段的发展轨迹。无论哪个阶段,乔氏所坚持的基本观点仍然不变:句法是一个独立的(self-contained)、自治的(autonomous)形式系统。语言中合乎语法的句子是一套形式化的符号并基于句法规则生成,随后获得语义解释。不同于乔氏的句法驱动生成理论,以Lakoff、Langacker和Talmy为代表的认知语言学反对用数理公式进行形式化运算和表征抽象化符号系统的研究方法,认为语言具有“体验性”,是人们对现实世界进行“互动体验”和“认知加工”的结果,句法离不开语义,整体意义也不是部分意义的简单组合,因此倡导从心智角度对语言进行研究。Goldberg等构式语法学者也反对句法自治的观点。他们认为语义才是句法产生的动机。一种语言的词汇携带了丰富的句法语义信息,句子规范与否是语义和语用共同作用的结果。句子的句法语义结构在很大程度上是由句子中所包含的关键词语的类属信息决定的。尽管基于印欧语系的句法--语义理论历经变迁,但理论间的冲突使得一些关键问题依然悬而未决,突出地表现在:在语言理解中,句法和语义之间的关系;在语言加工的具体时间进程中,句法与语义何时出现交互;二者之间的关系会受到哪些因素影响。基于印欧语系下的这些理论争端,本研究希冀进行跨语系的实证探讨,借助汉语理解和加工任务,反观句法驱动的生成理论。具体而言,本研究以汉语特有的、区别于印欧语系的量词为切入点,以汉语母语者为被试,借助时间分辨率较高、生态效度较好的眼动阅读技术,辅以离线的接受度判断和回溯访谈,探讨以下三个研究问题:第一,在语言理解中,句法是否是一个独立于语义的自治系统?第二,在语言加工的时间进程中,句法和语义存在着怎样的关系?句法是否先于语义被激活?二者是独立加工抑或是交互作用?如果存在交互,体现在哪些语言层面?第三,在语言理解中,句法与语义的关系是基于语言系统内部,抑或会受到其他因素的影响?希冀借助语言理解反观句法驱动的生成理论,厘清各理论所争议的焦点,为现存的基于印欧语语系的句法与语义理论提供跨语系的实证参考。具体而言,本研究在汉语DP、TP和CP等三个层面展开。实验一:限定词短语层面(DP)。操纵一个句法变量和一个语义变量,构成3(量词槽中词汇的词性:量词、名词、动词)×3(量词与名词的语义关联度:相关、无关、控制)的实验设计,探讨句法与语义在限定词短语理解和加工时间进程中的互动以及相关的影响因素。实验二:TP(tense phrase)层面。操纵一个句法变量和两个语义变量,构成2(句序:“一+量+名”、“名+一+量”)×2(名词的抽象度:抽象、具体)×2(量词的新奇度:普通、新奇)的实验设计,探讨句法与语义在TP理解和加工时间进程中的互动以及相关的影响因素。实验三:内嵌从句的CP(complementizer phrase)层面。操纵两个句法变量和一个语义变量,构成2(关系从句类型:主语关系从句、宾语关系从句)×2(“一+量”的位置分布类型:关系从句在“一+量”内的名词短语、关系从句在“一+量”外的名词短语)×2(关系从句中中心名词的抽象度:抽象、具体)的实验设计,探讨句法与语义在CP理解和加工时间进程中的互动以及相关的影响因素。本研究借助以上三项实验,得到如下结论。对于句法与语义的关系,本研究探明了句法与语义的双向互动关系。在限定词短语理解中,语义可以消解句法上的冲突,二者相互依存;在TP层面,小句中语言成分的语义特征对所在小句的句法表征产生制约作用;在CP层面,关系从句的句法表征受到关系从句中心名词语义抽象度的影响。对于加工时间进程中句法与语义的交互关系,本研究发现在语言加工的早期,在DP层面和CP层面,句法与语义均已显著激活,且存在显著的交互关系;在语言加工的晚期,在DP、TP和CP层面,句法与语义均存在显著的交互效应。基于此,本文建构了基于时间进程的并行加工扩展模型(stage-based extended Concurrent Model),用以阐释汉语语言加工进程中句法与语义的互动关系;此外,本研究还探明了影响句法与语义交互关系的七大主要因素,涉及语言系统内部的因素,诸如语义真值性,词汇间语义的兼容度,以及语言系统外部的因素,诸如语用频率,语言使用者的个人因素,如文学阅读量,识解方式及心理完形(mental gestalt)等。这驳斥了语言研究“从句法内部寻求解释”的生成观,证实了语言“体验性”。本研究的意义体现在以下几个方面:从研究对象来看,本研究借助汉语语料和汉语母语受试,为基于印欧语系构建的句法语义理论提供较为翔实的跨语系证据,丰富这些理论对人类语言实时加工机制的普适性;从研究内容来看,本研究立足于当前的三大理论争议,操纵句法和语义两大变量,全面地考察句法与语义在三大语言学层面的互动机制,为后续研究提供可资借鉴的研究框架及维度;从研究方法来看,本研究结合了时间分辨率较高、生态效度较好的眼动阅读技术,离线的接受度判断和回溯访谈任务,借助量化的接受度指标,眼动指标以及质性数据,全面地勾勒出句法---语义交互的连续统及其影响因素,丰富了句法语义理论的心理现实性;从理论贡献来看,本研究对基于印欧语系的“句法自治观”和“模块加工理论”提出理论挑战,构建了基于时间进程和影响因素的并行加工扩展模型,并从算子理论(operator theory)、结构距离假设(Structural Distance Hypothesis)、语义真值性、语义兼容度等视角对实验结论进行阐述;最后,从量词本体研究来看,本研究探明了:第一,量词的通达方式受量词新奇度的影响;“名+数+量”的重心与焦点受到数词槽“一”的语义真值性影响;“一+量”在关系从句中的位置分布偏好受从句中心名词抽象度影响。此外,除了表量的本体功能,本研究还探明了量词的修辞和韵律等衍生功能。以上发现可为汉语量词本体研究和汉语二语学习提供些许参考。更多还原

【Abstract】 There has been a long-standing controversy and dispute over the relationship between syntax and semantics in sentence comprehension and processing in the fields of generative linguistics and cognitive linguistics.Structuralism maintained that the meaning of a linguistic form is determined by the language system itself.The extra-linguistic factors,such as the world out there,how people interact with it,how they perceive and conceptualize it,do not impinge on the language system itself and language remains a self-contained,autonomous system,with its own structure,its own constitutive principles,and its own dynamics.“Autonomy of Syntax” goes through the five stages of Transformational-generative grammar,from Classical Theory,Standard Theory,Extended Standard Theory,the Theory of Government and Binding to Minimalist Program.The cognitive linguistic paradigm,however,maintains that language is not an autonomous cognitive faculty and ‘grammar is conceptualization’.Knowledge of language emerges from language use.They take many beyond-language factors into consideration when examining meanings.Although syntax--semantics theories in Indo-European frameworks have upgraded from generation to generation,these disputes were still in suspense,which embodied in three aspects including the relation of syntax and semantics in language comprehension,syntax-semantics interaction in the time course of language on-line processing as well as the factors influencing their relation and interaction.In view of this,this study conducted three experiments with Chinese specific and unique classifiers as the stimuli and Chinese native speakers as the participants,in the hope of examining these conflicting syntax-semantics theories and thus pushing the dense fog aside with cross-language evidences.To be specific,the study,in the hope of reflecting the conflicting theories concerning language generation through examining language comprehension and processing,was designed to deal with three main research questions.First,what is the relationship between syntax and semantics in language comprehension? Is syntax separate,autonomous,self-contained,or semantically constrained in language comprehension? Second,from the time course of processing stages,at which processing stage and on which linguistic level will syntax and semantics interact with each other and how? Are the syntactic information and the semantic information of language processed in a modular,interactive or concurrent way in language processing? Third,are the interaction of syntax and the semantics determined by the internal language system itself,or influenced by other external,beyond-language factors? How will the internal and external factors influence syntax-semantics relationship and interaction?Given all the aforementioned,the present study manipulated semantic and syntactic variables to explore the three research questions by combining the approach of psycholinguistic on-line eye-tracking experiments,acceptability judgment tasks and a retrospection interview task at Chinese classifiers(CL)--based determiner phrase(DP)level,TP level and relative clause-embedded CP level.To be specific,Experiment 1 focused on syntax-semantics interaction at DP Level.In this experiment,a 3(syntactic variable: part of speeches of the Chinese characters in the CL slot of "yi + CL + N")* 3(semantic variable: the semantic relatedness between the CL and N)ANOVAs for Repeated Measures design was adopted.Experiment 2 focused on the syntax-semantics interaction at TP level,in which a 2(syntactic order variable: "yi + CL + N" and "N + yi + CL")* 2(semantic variable: the abstractness of N)* 2(semantic variable: the abstractness of CL)ANOVAs for Repeated Measures design was adopted.Experiment 3 focused on syntax-semantics interaction at RC-embedded CP level,in which a 2(syntactic variable: relative clause types,i.e.subject-extraction relative clause--SRC and object-extraction relative clause--ORC)* 2(syntactic variable: the distribution of NCLs--Numeral-Classifier sequence,i.e.inner modifier nominal--IMN or outer modifier nominal--OMN)* 2(semantic variable: the abstractness of head noun,i.e.abstract noun--AN and concrete noun--CN)ANOVAs for Repeated Measures design was established.Taken together,the research findings could be summarized as follows.As for the relationship between syntax and semantics in language comprehension,this study confirmed the two-way interaction between syntax and semantics.At DP level,semantics played a significant role in dissolving the syntactic conflict and mismatch,implying the interdependence of syntax and semantics in construal.At TP level,the abstractness of the nouns in clauses,as well as the novelty of CLs both determined the syntactic representations of the construction they were in.At CP level,syntax of sentences was influenced by the semantics of the head nouns in RCs,and the influence was conversely embodied by the syntactic order.As for the time course when syntax and semantics interacted with each other,in the early processing stage,syntax and semantics of language were both activated and interacted with each other at DP level and CP level;in the late processing stage,syntax and semantics of language all interacted with each other at DP level,TP level and CP level.Accordingly,a stage-based extended Concurrent Model was constructed to account for syntax-semantics interaction in the time course of Chinese processing.As for the influence factors,seven factors were identified to influence syntax-semantics processing and interactions including factors in language internal system and external,beyond-language factors.The contributions of this study,at last,could be summarized as follows.First,research participants.The previous syntax-semantics interaction mechanism models were mainly drawn based on Indo-European languages,while this study designed Chinese stimuli for Chinese native speakers to evaluate the explanatory power of theories in Chinese context,and thus to explore a more comprehensive insight into the operation mechanism of language processors in human brain in real-time.Second,research framework.This study made a comprehensive exploration by conducting research at three linguistic levels,manipulating two kinds of variables to deal with three theoretical disputes concerning syntax and semantics,which will provide a ready-made framework to fertilize the research dimensions for the subsequent studies.Third,research methods.This study combined well-acknowledged syntactic acceptability task,the “temporally sensitive and ecologically valid” eye-tracking task as well as the retrospection interview task to sketch the interaction continuum of syntax and semantics in the processing stages at the level of milliseconds to enrich the mental/psychological reality of syntax-semantics theoretical accounts systematically and multi-dimensionally.Fourth,theoretical contributions.This study posed threat to the existing syntax-semantics theories including “the autonomy of syntax” and modular "syntax-first" processing models.New stage-based extended Concurrent Model was constructed concerning the syntax-semantics interaction and influence factors.Finally,ontological study for Chinese classifiers.This study also obtained some theoretical findings concerning Chinese classifiers.First,the Access Mechanism of CLs was influenced by the semantic factor.Second,the focus of NCL sequences,i.e."N + Num + CL",was not syntactically dependent as claimed by end-weight principle,but semantically constrained,i.e.determined by the truth-value semantics of NCLs.Third,the distribution preference of NCLs in RCs was not syntax-autonomous,but affected by the semantic factor.Finally,the derivational functions of NCLs,such as the categorization function,rhetorical function and prosody function,were identified.更多还原