A Study of Generative Morphology Based on a Large-scale Corpus : Where Does a Phrasal Compound Come from?

Lexical rules generally involve no phrasal categories (No phrase constraint) and hence phrase-incorporating compounds such as *totemo kyookona soshiki zooshoku/*very solid tissue proliferation are unacceptable, while certain types of phrasal compounds are possible as in hito-no soshiki zooshoku/human tissue proliferation. The present study investigates the possible patterns of phrasal compounds in Japanese, with the aim of identifying the optimum balance and interaction between syntax and morphology. Focusing on phrasal compounds extracted from a large corpus (Balanced Corpus of Contemporary Written Japanese), their theoretical analysis is carried out within an antilexical framework. Based on this analysis, I propose that (i) NPC follows straightforwardly from the inapplicability of the morphological operation of merger, (ii) the alleviation of NPC is motivated by (a) contextual reinterpretation, (b) noun saturation, and (c) nominalization by a native root and these conditions are formalized into redundancy rules on the applicability of merger, and (iii) the easing of NPC varies cross-linguistically depending on word formation devices in individual languages and accordingly the variability in constraint relaxation can be parameterized.