A Theoretical and Empirical Study on the Mental and Neural Mechanisms of Word Formation

平克等人提出的双机制模型（Dual Mechanism Model，DMM）认为，两种不同的心理机制（规则计算和联想记忆）参与了构词。这一理论认为，句子使役句（SC）是由规则加工的，而词汇使役句（LC）是由联想记忆加工的。为了验证这一假设，我们进行了一个ERP实验，研究了SC和LC加工的神经机制，刺激包括四种类型的句子：可接受和不可接受的SC句子（ASC和USC）和可接受和不可接受的LC句子（ALC和ULC）。N400和P600在ULC中出现，这表明这种偏离是由于违反了词汇语义选择。另一方面，USC在250- 300 ms内产生左前负波（LAN），在300- 500 ms内产生非偏侧前负波（AN）。这表明，加工的SC涉及句法计算。鉴于一般的理解，语义选择是一个问题的词汇网络，而句法计算是基于规则的，这些研究结果证实了DMM扩展到日本使役动词的形成的可能性。此外，ASC和ALC之间的比较显示，ASC在450- 600 ms引起可靠的正效应，其可以被解释为反映用于处理双腔结构的计算负荷将这些发现与我们以前的实验研究结果相结合，（关于使用SC和LC的失语症患者的行为实验（Sugioka等人，2001），以及关于日语名词化的研究（Hagiwara et al. 1999）），我们已经成功地表明，DMM是一个很有前途的理论，不仅可以适用于屈折变化，而且还可以用于派生，从而涵盖了整个构词范围。

The Dual Mechanism Model (DMM) as proposed by Pinker and others claims that two different mental mechanisms (computation by rule and associative memory) are involved in word formation. This theory suggests that the so-called Sase Causatives (SC) are processed by rule while Lexical Causatives (LC) are by associative memory. We conducted an ERP experiment to verify this hypothesis, by examining the neural mechanisms involved in the processing of SC and LC.The stimuli consisted of four types of sentences : Acceptable and Unacceptable SC sentences (ASC and USC) and Acceptable and Unacceptable LC sentences (ALC and ULC). N4oo and P600 were observed in ULC, which suggests that the deviance is due to the violation of lexical semantic selection. USC, on the other hand, yielded Left Anterior Negativity (LAN) in 250-300ms, and non-lateralized Anterior Negativity (AN) in 300-500ms. This suggests that the processing of SC involves syntactic computation. Given the general understanding that semantic selection is a matter of lexical network while syntactic computation is rule-based, these findings corroborate the possibility of extending the DMM to Japanese causative verb formation. Furthermore, the comparison between ASC and ALC revealed that the ASC elicited a reliable positive effect at 450-600ms, which can be interpreted as reflecting the computation load for processing the bicluasal structure (i.e. syntactic embedding) of SC.Integrating these findings with the results of our previous experimental studies (a behavioral experiment on aphasia patients on the use of SC and LC (Sugioka et al. 2001), and a study on Japanese nominalizations (Hagiwara et al. 1999)), we have succeeded in showing that the DMM is a promising theory that can be applicable not only to inflection but also to derivation, thus covering the whole range of word formation.