Neurological Mechanisms of Word Formation: An Empirical Study by Collaboration of Theoretical Linguistics and Neurolinguistics

我们采用所谓的双机制模型（Dual Mechanism Model）作为我们的工作假设，该模型认为单词加工涉及记忆和计算两种不同的心理/神经机制，并进行了事件相关电位（event -相关电位）实验，从神经学层面研究日语复杂谓词的加工。我们的研究结果如下。我们使用了两种类型的使役结构，即词汇使役结构（LC）和时态使役结构（SC），并观察了对不可接受句子（ULC:^*kansei-o narab-e-ru“安排欢呼”，USC:^*ryoori-o narab-ase-ru“把盘子排成一排”）和可接受句子（ALC: ryoori-o narab-e-ru“安排盘子”，ASC: senshu-o narab-ase-ru“让玩家排成一排”）的ERP反应。ULC语句引发了典型的N400，我们将其解释为反映语义选择违规。相比之下，USC句子引发了一个an（前向否定），然后是一个N400。AN可以用来反映检测USC句子的病态性所涉及的计算过程，而下面的N400则是语义违反的反映。此外，当比较两种可接受的句子类型（ALC和ASC）时，我们发现ASC诱发了P600，我们认为这反映了处理SC双子句结构所涉及的句法计算成本。因此，我们的研究结果基本上支持使役结构的双机制方法，使役结构包括基于记忆的处理，而使役结构由句法计算处理。我们进行了另一项实验，希望积累日语语言处理的基本ERP数据，探索在违规情况下观察到哪些ERP成分。我们观察到N400违反了词汇指定的动词大小写数组（^*kodomo-ni shikaru“骂孩子”，^*hito-o butsukaru“打人”）。

We have employed the so-called Dual Mechanism Model as our working hypothesis, which claims that two different mental/neurological mechanisms of memory and computation are involved in the processing of words, and conducted an ERP (Event-Related Potentials) experiment to investigate the processing of Japanese complex predicates at the neurological level. Our findings are as follows.We used two types of causative constructions, lexical causative (LC) and sase-causative (SC), and observed the ERP responses to unacceptable sentences (ULC:^*kansei-o narab-e-ru 'arrange shouts of joy', USC:^*ryoori-o narab-ase-ru 'make dishes line up') in comparison to the acceptable counterparts (ALC: ryoori-o narab-e-ru 'arrange dishes', ASC: senshu-o narab-ase-ru 'make players line up'). The ULC sentences elicited a typical N400, which we interpret to reflect the semantic selectional violations. In contrast, the USC sentences elicited an AN (anterior negativity) followed by an N400. The AN can be taken to reflect the computational process involved in detecting the ill-formedness of the USC sentences, while the following N400 is a reflection of the semantic violation. In addition, when the two acceptable sentence types (ALC and ASC) were compared, it was observed that the ASC elicited a P600, which we interpret to reflect the cost of syntactic computation involved in processing the bi-clausal structure of SC sentences. Thus our results basically support the Dual Mechanism approach to causative constructions, LC involving memory-based processing while SC being processed by syntactic computation.We have conducted another experiment, in the hope of accumulating basic ERP data on language processing in Japanese, exploring what ERP components are observed in case violations. We observed N400 for the violation of lexically specified case array of verbs (^*kodomo-ni shikaru 'scold a child', ^*hito-o butsukaru 'hit a person').