Date: Friday, April 22, 2011
Time: 3:30-5:00 PM
Place: 32-141
Speaker: Junko Ito, University of California, Santa Cruz
Title: On the sources of (un)accentedness
ABSTRACT
Recent work on the distribution of unaccentedness in the lexicon of Tokyo Japanese has found a concentration of unaccented words in very specific areas, which are to a large extent defined in prosodic terms. Thus 4-mora words are in their majority unaccented in both native words and loans. On the other hand, 1-2 mora words and 5-mora words are in the majority accented in both native words and loans. For 3-mora words we find a split: They are in the majority unaccented in native words, but accented in loans. What is relatively clear is that unaccentedness is some kind of default for words of these specific shapes. What is less clear is the prosodic rationale of the particular distribution of (un)accentedness. The goal of the talk is to investigate the underlying structural reason, and to develop a formal account in optimality-theoretic phonology.
The regular pitch accent pattern of Tokyo Japanese (i.e., the pattern that is not lexically marked on a morpheme-by-morpheme basis, but emerges, e.g., in loanwords) is characterized by two constraints widely seen at work in other languages: Rightmostness (if the word has an accent, it should fall on the last foot in the word) and Nonfinality (the accent should not fall on a subconstituent that is final in the word). There is a clear tension between these two constraints, which is in many languages resolved by priority ranking (Nonfinality beats Rightmostness in cases of conflict). Seen in this light, unaccentedness is another way of dealing with the conflict: By not assigning an accent, the conflict disappears.
More precisely, the guiding idea is the following: For words of specific prosodic profiles (such as LLLL), unaccentedness is a default because it fulfills both Nonfinality and Rightmostness (no accent means no violation of either constraint). For other shapes, such as LLL and LLLLL, both constraints can be fulfilled even by accented words, so unaccentedness brings no prosodic benefit and is avoided since it violates the ?-Accent constraint requiring words to be accented. For short (one- to two-mora) words, a constraint calling for the construction of the “perfect prosodic word” (a word coextensive with a single foot) plays an important role. Other cases of perfect word effects to be presented include Serbian and Danish.