A Study of Inflection and Derivation in the Mental Lexicon

本研究旨在考虑Pinker等人提出的屈折形态中规则和不规则的明确区分是否也适用于衍生形态。根据像Pinker这样的符号学家的观点，规则的屈折形式是由规则获得的，而不规则的屈折形式则被认为是存储在联想记忆中，有时会发生新的不规则形式是通过与实际形式的类比而形成的。这一建议直接反对连接论者的主张，即规则和不规则之间没有质的区别，因此心理词汇没有规则的空间。我在这项研究中表明，与屈折词法的情况一样，衍生词法包括一些可以作为默认条件的词缀，例如英语后缀-ness和日语后缀-sa，它们都是从形容词中派生出抽象名词。本研究最密切地调查了英语和日语中代表代理人的派生名词。结果发现，英语的代理后缀-er是默认的，而日语的代理后缀中没有默认的词缀。我指出，日语中的代理名词，包括那些带有生产后缀-sha（-*）、-nin（-*）和-te（-*）之一的代理名词，都受限于它们所属的语素类（Yamato、Sino-Japanese、Foreign或Mimetic）。这四种语素类被认为至少部分地通过某些音系特性彼此区别开来。因此，我们似乎可以认为，日语中新形成的代理名词不是规则的产物，而是与存储在词汇库中的实际代理名词进行类比的产物，它们在语音上相似。这个建议将通过我计划做的一个使用无意义词语的实验来进一步证明。

The present research aims to consider whether the clear distinction between regulars and irregulars in inflectional morphology, claimed by Pinker and others, applies to derivational morphology as well. According to symbolists like Pinker, regular inflected forms are obtained by rules, whereas irregular ones are assumed to be stored in associative memory, and it sometimes happens that novel irregulars are formed by anaology with actual ones. This proposal is directly opposed to the connectionists' claim that there are no qualitative differences between ragulars and irregulars, and therefore that the mental lexicon has no room for rules.I showed in this research that, as in the case of inflectional morphology, derivational morphology includes some affixes which can be qualified as default, such as the English suffix -ness and the Japanese suffix -sa, both of which derive abstract nouns from adjectives. What this research investigated most closely is derived nouns representing agents in English and Japanese. It turned out that English has the agent suffix -er as default, while in Japanese there is no default affix among various agent suffixes. I pointed out that agent nouns in Japanese, including those which have one of the productive suffixes -sha (-*), -nin (-*) and -te (-*) attached to them, are restricted with respect to the morpheme class (Yamato, Sino-Japanese, Foreign, or Mimetic) which they belong to. These four morpheme classes are considered to be differentiated from each other at least partially by some phonological properties. It seems, therefore, that we can consider that newly formed agent nouns in Japanese are the produsts not of rules, but of analogy with the actual agent nouns stored in the lexicon to which they are phonologically similar. This proposal shuld further be proved by an experiment in which nonsense words are used which I am planning to make.