Gradient Prosody in Japanese

This paper examines five independent morphological processes in contemporary Japanese: renyookei reduplication, dvandva compounds, plural reduplication, mimetic reduplication, and zuuzya-go. Common to them is that the resulting form can be divided into two portions (e.g., base and reduplicant, or two members of a compound). Some of the morphological processes exhibit prosodic augmentation through vowel lengthening or segmental epenthesis, but others do not. Among the group of those phenomena with prosodic augmentation, the augmentation is shared by the two halves of a produced word in one case but not in others. Given these two parameters, the five morphological operations display a three-way gradient pattern: (i) both prosodic augmentation and prosodic sharing occur (renyookei reduplication), (ii) neither occurs (dvandva compounds and plural reduplication), and finally, (iii) prosodic augmentation occurs, but it is not shared by the two halves of a word (mimetic reduplication and zuuzya-go). This paper has both descriptive and theoretical goals. First, I demonstrate that prosody of the five morphological constructions exhibits such a gradient pattern through a careful description. Second, on the theoretical side, I offer a detailed comprehensive analysis of the prosodic gradation within the framework of Optimality Theory (Prince and Smolensky (1993)). I argue that the notion of relativized faithfulness is essential, relativization referring to morphemes. This intra-linguistic comparative study thus lends support to Kurisu (2001b; 2005) who argues for the necessity of morpheme-specific faithfulness constraints.