Ranking paradox in consonant voicing in Japanese

要旨

It is quite often that some phonological phenomena are found only in a certain vocabulary class but not in others within a single language. However, the basic tenet in OT, a single invariant ranking, apparently seems incompatible to those multiple vocabulary classes with inconsistent phonological phenomena. Recent OT analyses have developed useful notions in their arsenal to attack this problem. First, multiple sub-lexica are defined when their phonological properties are distinct enough. For instance, Japanese has at least four phonological sub-lexica, such as Yamato, Sino-Japanese, Mimetics, and Foreign (Ito and Mester, 1995; Ito and Mester, 1999; Fukazawa et al., 1998). Second, those sub-lexica are organized in a core-periphery structure (Ito and Mester, 1995). Generally (and historically), the native vocabulary tends to form the core part while nonnative vocabularies tend to form the periphery. A constraint-based implementation of the coreperiphery structure is to assume that the more native the sub-lexicon is, the more markedness constraints it may obey. So, for example, in the most native sub-lexicon Z, constraints *F, *G, and *H are all respected. In the least native sub-lexicon X, however, only the constraint *F is satisfied. Figure 1 shows these relations in a set of concentric ellipses. Faithfulness constraints must be ranked between any two markedness constraints for this core-periphery structure to work. For example, in the sub-lexicon X, *F is respected but *G and *H are violated. What ensures the latter two to be violated is a relevant faithfulness constraint for the sub-lexicon X. By the same token, the sub-lexicon Y has its own version of faithfulness constraint where *F and *G are respected but *H is violated. 1 The overall ranking for this example is shown in (1).

主題