【作者】 邓劲雷；

【导师】 戴炜栋；

【作者基本信息】 上海外国语大学 ， 英语语言文学， 2009， 博士

【摘要】 语言习得研究是验证语言学和语言学习理论的重要依据,也是探索人类心智的一个重要途径。在语言学与语言习得研究中存在着两种不同的观点。第一种观点以乔氏的普遍语法和语言天赋假说为代表。该理论认为儿童的语言输入过于贫乏,缺乏天赋的语言知识,儿童将无法学会语言。天赋的语言知识由句法系统中的原则和参数组成。由于不同语言的参数设置各不相同,因而儿童在学习语言的过程中需要也仅需要设置相关的参数。在这一理论框架下,语言的产出首先通过论元结构投射产生底层结构,再经转换生成表层结构。另一种观点以认知语言学理论与基于用例的语言习得模式为代表。该理论认为语言的句法系统由抽象程度各异的多层次构式网络组成,语言学习是学习构式的过程。这一理论认为无须先天的语言特异性知识,人类也能通过通用认知能力学得语言。这一习得模式强调语言输入、语言使用以及它们的频率对语言习得的重要作用。这两种理论对特殊疑问句的句法描述各不相同。在普遍语法框架下,特殊疑问句的产出需要进行两次移位:wh-移位和I至C移位,而认知语言学理论认为语序在构式中直接指定,语言产出无须进行移位。由于这两种不同的描述,调查儿童习得特殊疑问句的发展情况在一定程度上能够检验这两种不同理论的合理性。本论文运用CHILDES数据库,对儿童在学习英语过程中特殊疑问句的发展情况进行调查。共进行四项研究。第一项研究调查特殊疑问句在语言输入中的分布情况。先前的研究发现儿向语中存在着大量的问句,但是到目前为止似乎还没有专门针对儿向语中的特殊疑问句的调查研究。因而这一项研究的第一个任务是调查特殊疑问句在儿向语中的分布情况。另外,先前的研究表明儿向语有着不同于成人语言的独特特征。因而,这一项研究的另一个任务是调查儿向语中的特殊疑问句是否也存在一些独特的特征。调查发现各个特殊疑问句的分布情况如下: what问句在特殊疑问句中的比重较高大约为60%,并且在整个语言习得过程中比重基本保持平稳; where和who问句的初始比重较高,但是随时间推移,比重逐渐下渐;how、why和when问句的情况与where和who问句刚好相反,这三类问句的初始比重较低,但随时间推移,比重逐渐增加。这一分布情况表明儿向语在一定程度上随着儿童语言能力的发展而发展,因为how、why和when问句无论从语言结构还是从所表征的概念来说都比what、where和who问句来的复杂。另外,调查发现儿向语中存在不移位的特殊疑问句,这些疑问句可能为了减轻儿童回答问题的负担而对语言形式进行调整的结果。这些现象从侧面表明儿向语不仅不是如天赋论者所认为的那样贫乏,而且从某种程度上可以说是为儿童语言习得“量身定做的”。第二项研究调查特殊疑问句在儿童语言产出中的分布情况。特殊疑问句在儿童语言中出现的顺序大体遵从以下次序:what、where、who、why、how、when。各个特殊疑问句的出现次序与它们在语言输入中的频率有着较大的相关性。what、where、who问句在早期的语言输入中频率较高,它们在儿童语言中也较早出现。why、how、when在后期的语言输入中频率较高,它们相应的在儿童语言中也出现得较晚。另外,一些特殊疑问句曾倒U形发展。如,why问句在2.25岁时大量涌现,而后使用频率又逐渐减少。本论文认为这一发展模式是由于语言使用与概念发展之间存在着密切的联系。Why问句在2.25岁时的大量出现表明了儿童在这一时期习得了因果关系这一概念。概念的习得在很大程度上受概念复杂度的影响,因而特殊疑问句的不同习得顺序还应受它们所代表概念的复杂度的影响。这一项研究表明儿童语言的习得不仅受语言输入的影响,还受概念复杂度等因素的影响。第三项研究调查儿童在语言习得过程中是否习得wh-移位规则。如果语言学习的过程是设置普遍语法中参数的过程的话,那么儿童习得特殊疑问句的顺序应与普遍语法所描述的产出特殊疑问句的顺序相一致,即儿童应先产出不移位的特殊疑问句,而后产出经wh-移位的特殊疑问句,最后产出与成人语言一致的特殊疑问句。然而先前的研究表明儿童一开始产出的特殊疑问句就将疑问词置与句首。这一现象显然与生成语法的假设不符。但是普遍语法可能是一个逐渐可及的过程,也即儿童在初始阶段所产出的语言可能不受普遍语法的制约,只有在相关的普遍语法被激活的情况下,特殊疑问句的产出顺序才跟生成语法所指定的移位顺序相一致。然而调查发现这一假设也得不到语言习得数据的支持。因而,这一研究表明儿童的特殊疑问句的产出可能并不需要进行移位操作。另外、这一项研究还发现在儿童语言中存在带what is that/this结构的非规范疑问句。这些疑问句似乎无法通过普遍语法所认为的论元结构投射生成。这些非规范疑问句的产出似乎是由于what is that/this在语言输入和输出中的高频率造成的。这些问句不仅表明普遍语法所假设的语言产出框架可能存在问题,还表明了频率在语言习得过程中起着重要作用。第四项研究调查I至C移位的习得情况。在生成语法框架下,I至C移位以语法范畴为单位进行运作,因而理论上一旦这些规则被学会,它们应当能够被应用于所有适用这些规则和语法范畴的句子当中。但是先前的研究发现在儿童语言中移位与不移位的特殊疑问句并存,这一现象显然与普遍语法的假设不符。为了解决这一问题,基与普遍语法的研究者分别从特殊疑问词差异、助动词差异等角度对这一现象进行解释。然而后继的研究从移位比率的角度对这些解释进行考察时发现这些解释都不尽如人意。与普遍语法不同,基于用例的语言习得模式认为儿童无需习得I至C移位,特殊疑问句中的主语—助动词倒置结构的习得是习得疑问语+助动词构式的过程。本论文认为I至C移位的习得还有以下两点需要进一步调查。(1)先前的研究主要从移位比率的角度进行考查,还缺乏从习得顺序的角度对上述解释进行调查的研究。(2)还缺乏对基于用例的习得模式进行验证的研究。对(1)的调查主要根据Brown(1973)提出的90%正确率为标准。调查发现系动词BE的移位最先被习得,而且表现出清晰的阶段发展趋势,助动词BE的倒置率在达到90%以后出现较明显的倒退,助动词DO的倒置率总的来说呈逐渐升高的趋势,情态动词由于样本较小,变化幅度较大。这些发现与Stromsford (1990)和Santelmann et al. (2002)所提出的假设刚好相反,他们认为在普遍语法框架下系动词BE的移位习得应当比其它助动词移位的习得更为困难。因而,这一调查再次表明了普遍语法的不合理性。对(2)的调查发现外在形式完全相同的助动词BE和系动词BE的移位习得顺序存在明显差异。虽然这一发现与基于用例的习得模式所提出的习得主语—助动词倒置结构是习得特殊疑问词+助动词构式的过程的观点不相符,但是它并不与基于用例的学习模式本身相矛盾。基于用例的学习模式强调语言学习是学习构式的过程,助动词BE和系动词BE在语言中的构式本身存在差异,因而这两个构式的习得顺序存在差异并不与基于用例的习得模式相矛盾。本论文的调查结果总的来说与基于用例的语言习得模式更为相符,而与生成语法的假设大相径庭。调查表明,儿向语并非如天赋论者所相象的那样贫乏,这一语体不仅随着儿童语言能力的变化而变化,有时甚至为了适应儿童的语言能力还会对语言形式进行调整。调查还表明语言的使用不仅受语言输入的影响而且还受认知能力等因素的影响。更为重要的是这一研究发现儿童在习得语言特殊疑问句的过程中并没有习得规则,更不用说习得普遍语法所认为的移位规则,而且儿童还产出不受题元结构制约的特殊疑问句,这些疑问句表明普遍语法所假设的语言产出框架可能也缺乏心理现实性。总的说来,调查数据表明语言学习是学习词汇、构式的过程,在语言习得过程中的频率起着重要作用。这些结果与基于用例的语言习得模式和认知语言学理论更为相符。更多还原

【Abstract】 Language acquisition provides crucial evidence to test linguistics and learning theory as well as to explore the working mechanism of human mind. In the field of linguistics and language learning, two competing paradigms coexist. One is represented by Noam Chomsky and his followers, who advocate rule-based Universal Grammar in linguistic study and language innateness hypothesis in language acquisition study. Under this paradigm, the syntax system of all languages consists of innate principles and parameters, and syntax learning requires and only requires children to set the appropriate parameters for the specific languages they are exposed to. Under this framework, language production firstly makes use of the argument structures of the main verb in a sentence to project its D-structure, which then is transformed to S-structure. The result of these two operations is the grammatical sentence. The other paradigm is represented by cognitive linguistic theory and usage-based language learning models. This paradigm believes that language system consists of networks of constructions existing at multiple levels of concreteness and abstraction, and language acquisition is a process of learning constructions. This approach, believing that general cognitive capacity is sufficient for language learning, assumes no innate language-specific knowledge. It emphasizes the importance of input, language usage and their frequency in shaping language developmental process. Wh-questions form a good testing case for these two competing approaches, for universal grammar posits that the production of wh-questions should undergo two transformations: wh-movement and I to C movement while cognitive approach, believing that word order is directly specified in constructions, assumes no transformation operation in the production of wh-questions. The different assumption about language representation and production should predict different development process in the acquisition of wh-questions, which were investigated in this study by using CHILDES database. Four analyses are conducted.The first analysis investigates the distribution of wh-questions in child-directed speech. Previous studies found that questions comprise a substantial proportion of child-directed speech. However, until now, to my knowledge, the properties of wh-questions in CDS are still left uninvestigated. Therefore, the first task of this analysis is to investigate the distributions of wh-questions in CDS. Moreover, previous studies found that CDS is a distinctive speech style, which has special features. Thus, the second task of this analysis is to investigate whether wh-questions in CDS also have some special features. The investigation found that different wh-questions have different distributional patterns. The percentage of what questions, accounting for 60% of all wh-questions, remains stable throughout the study. Where and who questions are of high proportion at the beginning but show a decrease tendency over time, while how, when and why questions show exactly opposite developmental pattern, which are of low proportion at the beginning but show increase tendency with children’s age. This distribution pattern indicates that CDS develops with the language development of children, for how, why and when questions are more complex than what, where, and who questions no matter in terms of linguistic structures or in terms of conceptual complexities. Furthermore, this analysis also found some wh-in-situ questions in CDS that is not attested in adult speech, which shows that in order to help children understand and respond to CDS, care-takers may even adjust their speech into the unconventional forms. These findings suggest that CDS may not be degenerated at all but rather may be fine-tuned to the language development of children.The second analysis investigates the distribution pattern of wh-questions in child speech. The acquisition of wh-questions generally conforms to the following order: what, where, who, why, how, when. This developmental pattern correlates to a substantial degree with input, in which what, where and who questions appears with high frequency at the beginning stage while why, how and when questions appears with high frequency only at later stage. Some wh-questions show U-shape developmental pattern. A conspicuous example is the development of why questions, which appears in great quantities at the age of 2.25 years old but is used less frequently afterwards. The underlying reason for this developmental pattern is that language use is closely connected with the conceptual development. Asking why questions requires the understanding of the concept of causal relationship. Different concepts are different in terms of their complexities, which require different time schedules to develop, which may in turn influence language acquisition. This analysis shows that language acquisition is influenced by input and factors such as the complexity of concept.The third analysis investigates whether children learn wh-movement rules during language learning process. If language learning is a process of setting parameters as nativists have suggested, the developmental sequences of wh-questions should mirror the production sequence specified in Universal Grammar, i.e. children should produce wh-in-situ questions first, after which, wh-questions undergoing wh-movement but not I to C movement would be produced, and only at the last stage, children could produced adult-like wh-questions. However, previous studies found that children place wh-words at the initial position of wh-questions, which is in conflict with the prediction. However, this finding is only incompatible with the assumption that Universal Grammar is available to children from very beginning. Different components of Universal Grammar could well be accessible to children at different developmental stages. If principles or parameters related to wh-questions production are not accessible to children only at very beginning, and the production of wh-questions at this stage entirely rely on memory-retrieval, it could explain why at very beginning children produce wh-questions with wh-words at the initial position. However, an investigation of this explanation shows that it is in conflict with the data. Furthermore, this analysis also found unconventional wh-questions with what is that/this construction, which is incompatible with the assumption that the production of sentences is first projected from the argument structures of verbs. They seem to be better accounted for with the frequency effect of what is that/this in input and output. Thus, these questions not only show the problematic nature of universal grammar, but also reveal that frequency plays an important role in language acquisition process.The fourth analysis investigates whether children learn I to C movement during language learning process. Under the framework of Universal Grammar, rules or principles operate on syntactic categories. Therefore, theoretically, if a child learns a rule, he should be able to apply it to all members of the category on which the rule operates. However, previous study found that in child speech, inverted and non-inverted wh-questions coexist, which suggest that I to C movements are not uniformly applied to all cases that should undergo movement. This finding is incongruent with rule-based learning theory. To remedy the theory, different properties among wh-words and auxiliaries are invoked to explain this divergence. Unfortunately, studies afterwards show that these explanations are also untenable when the inversion rates of different combinations of wh-words and auxiliaries are examined. However, inversion rates may not be a good index to examine these explanations, for these explanations may predict difference in acquisition orders rather than in inversion rates. Therefore, the first task of this analysis is to investigate the acquisition order of subject auxiliary inversion, and it is investigated against 90% correctness standard proposed by Brown (1973). The investigation found that the inversion of copula BE is acquired the first, and exhibits stage-like development; the auxiliary BE achieves a 90% inversion rate half a year later than copula BE, but it shows serious regression afterwards; the inversion rate of auxiliary DO increases gradually; and the inversion rate of modals fluctuates a lot due to the scarcity of related samples. These findings are exactly the opposite to Stromsford (1990) and Santelmann et al. (2002)’s proposal that inversion of copula BE is more difficult to acquire than the inversion of other auxiliaries under the framework of Universal Grammar. Cognitive linguistics believes that the coexistence of inverted and non-inverted wh-questions is better accounted for under the usage-based language learning models, which posits that subject auxiliary inversion is acquired by learning specific combinations of wh-word + auxiliary rather than by learning the rule of I to C movement. An investigation of this proposal shows that the inversion behaviors of copula BE and auxiliary BE with the same overt forms are different from each other. Although this finding is incompatible with the proposal, it is not in conflict with the usage-based language learning model itself. Under the usage-based model, a fundamental task of language acquisition is learning constructions. Copula BE and auxiliary BE appear in different constructions. Therefore, the different inversion behavior between copula and auxiliary BE is not unexpected for usage-based learning model.The results of these four analyses are more congruent with the usage-based language learning models than with rule-based Universal Grammar and language innateness hypothesis. This study found that CDS is not as degenerated as nativists have proposed. Rather it seems to be attuned to the language development of children, and sometimes care-takers even use unconventional expressions to help children understand and respond to the speech. Furthermore, this study did not find any evidences suggesting that children learn rules, let alone the rules of wh-movement or I to C movement, during language acquisition process. Moreover, the unconventional wh-questions discovered in this study suggest that sentence may not be the result of projection from the argument structure of verbs, which further calls into questions the sentence production framework of Universal Grammar. All in all, this study found that language learning process is a process of learning lexical frames or constructions, within which frequency effect plays a vital role.更多还原