Lexical Stratification and Ranking Invariance in Constraint-based Grammars*

要旨

this paper, we discuss a model of lexical stratification which fully adheres to the assumption of invariant ranking. The model, originally proposed by Fukazawa (1997), solves the conflict between language-internal variability and ranking invariance by projecting multiple sets of Input-Output faithfulness constraints. Each set of faithfulness constraints is associated with a sub-lexicon (or stratum) in the language, and all the stratum-specific constraints are ranked with other types of constraints. Such a system can account for stratum-specific phonological phenomena in a single constraint ranking. The main goal of the paper is to demonstrate the conceptual and empirical advantage of this approach, using lexical stratification in Japanese as a domain of investigation. Further, we will address two issues that arise regarding the claim that multiple projections of faithfulness constraints derive the effects of lexical stratification: i) the question of how multiple faithfulness constraints can be instantiated, and ii) implications for the fundamental organization of the lexicon with strata. This paper is organized as follows. Section 2 illustrates the superiority of the Correspondence Theoretic approach to lexical stratification to the re-ranking system. We present an analysis of hybrid words in Japanese and show that simultaneous reference to multiple sets of Input-Output faithfulnes constraints is necessary to account for the data. Section 3 provides an answer to the question: how are those multiple sets of IO constraints established? We will indicate that contradicting phonological data touch off multiplication of faithfulness constraints and ranking of constraints in the process of acquisition. Section 4 discusses the relationship between split faithfulness constraint...

主題