Paper short abstract:

This paper attempts at a re-evaluation of the Japanese word class system with focus on so-called no-modifiers by defining syntactic criteria for a subcategorization into subgroups of different nominal and adjectival status. The theoretical and empirical adequacy is shown based on a corpus study.

Paper long abstract:

In this paper, I attempt at a (re-) evaluation of the Japanese word class system, focusing on those modifiers that appear with the element -no in attributive position, in classical Japanese linguistics simply labeled as nouns. I show that what I label `no-modifiers’ should make up a category in its own rights, and that this category is more diverse, but also of greater significance for the Japanese word class system than previously assumed. I put forth a taxonomy of no-modifiers based on syntactic criteria and a division into dedicated subgroups which show a gradual change from nominal to adjectival status. Then I show how this process allows us to sort relevant lexemes into these subgroups which I exemplify based on a corpus study.

The array of possible parts of speech that appear with -no in attributive position ranges not only from nouns to numeral classifiers to modifiers sometimes called no-adjectives (Muraki 2012). In fact, most elements appearing with -no show mixed behavior. We find among others variation in the elements appearing in attributive position (-no vs. -na) or in contextual behavior (referential vs. modificational). It is generally argued that such variation, which with different exponents is observable in other languages as well, does not arise as an individual phenomenon but is related to the general logic and properties of individual grammatical systems and that for each language, specific criteria for class membership need to be put forward (McNally / Swart 2015). I argue that the category of no-modifiers is made up of several, I could establish seven, subgroups which can be ordered from more adjectival to more nominal. Besides the criteria mentioned above, elements in attributive position, and Reference vs. Property, I put forth the criteria Gradability and Nominalization as determining factors.

These criteria do not only make theoretical predictions, they have also proven empirically adequate and testable. I have performed a corpus study in the Balanced Corpus of Contemporary Japanese (BCCWJ) by testing relevant lexemes appearing with -no (#=2300) and show how this subcategorization allows us to re-analyze and sort relevant lexemes into these subgroups.