抄録
As for the morphosyntactic size of a compound, it has occasionally been suggested that some of the N-N compounds can be larger than derived words but are smaller than phrases (Allen (1978), Giegerich (2005)) or that some of the V-V compound are words, while others are phrases (Kageyama (1993, 2001), Nishiyama (1998)). However, exactly how large each compound is remains controversial, partly because their nature is synchronically variable in terms of phonology, morphology, syntax, and/or semantics. In relation to this problem, there has been a long-standing issue of which of morphology and syntax should deal with the internal structure of these two types of compound and others. Here arises a set of reciprocative discussions between the lexicalists and anti-lexicalists over the data that belong to morphosyntax, and yet no settlement has been reached so far, because both types of approach have as much defects as merits. With these problems recalcitrant to a synchronic analysis in mind, I will shed a diachronic perspective on them. More specifically, this article launches a simple hypothesis that the morphosyntactic size of a compound tends to be diachronically enlarged from the domain of morphology to that of syntax, as is known by the names such as demorphologization and/or constructionalization. I will collect relevant data from the Corpus of Historical American English (COHA) and the literature on the traditional Japanese linguistics. Then, I will provide a morphosyntactic analysis of the diachronic generalization, in terms of two outstanding syntactic theories: Distributed Morphology (Halle and Marantz (1993), Marantz (1997)) and Cartography (Cinque (2003, 2006)). Three kinds of data presented in support of the above hypothesis are: (i) the demorphologization of many combining forms including -phobia, -holic, psycho-, techno-, their reanalysis as independent words, and their development as N-N compounds, (ii) the emergence of the resultative construction from the corresponding V-A form in English, and (iii) the development of the syntactic V-V compounds from the lexical V-V compounds such as kami-kiru `bite-cut' and yomi-kiru `read-cut' in the history of Japanese (Aoki (2010)). I will argue that these three types of diachronic changes are the instances of what I call ``syntactic constructionalization'' at the so-called ``word'' level, the VP/vP-level, and the AspectP-level, respectively.
