論文

In natural speech, phonetic cues that distinguish lexical items can be hyperarticulated when there is a minimal pair competitor, a process known as contrastive hyperarticulation. Building upon prior work, this article examines the cue-specific nature of contrastive hyperarticulation in Japanese, focussing on stop voice onset time (VOT) using a speech corpus. We confirmed that the existence of a voicing minimal pair competitor in the lexicon triggers hyperarticulation of VOT duration on the target segment (shorter for voiced stops and longer for voiceless stops), while other contrasts (singleton vs. geminate) did not. The results also suggest that contrastive hyperarticulation (a) is more compatible with casual speech than slow/clear speech, (b) is sensitive to position in a word (greater in word-initial position than in non-initial position) and (c) applies to a greater degree in Japanese than in English due to properties of stops. This provides further evidence that the phonetic specificity of contrastive hyperarticulation is cross-linguistically relevant.

This paper offers a case study of the role of information in patterns of segmental duration, focusing on the short vs. long vowel contrast in Japanese. We confirmed that the position of vowels in linguistic units (intonation phrase/IP and word) and the length of the units affect vowel duration, but the patterns differ depending on the type of units. The results suggest that at the word level, vowel duration patterns in such a way that cues for vowel contrast are enhanced in salient (earlier) positions and units with less predictability or more lexical competitors (shorter words), while cues are relatively reduced in non-salient (later) positions and units with more predictability or less lexical competitors (longer words). In this way, the variation in vowel duration balances the successful transmission of lexical information and the cost for phonetic implementation, which supports previous findings in communication-based approaches of language.

Some Japanese predicates allow their object to be marked either with the accusative particle o, the canonical marker of an object, or with the nominative particle ga, the canonical marker of a subject. This work reports and discusses the results of corpus-based surveys on the patterns and recent trends in nominative-marking on objects. It will be demonstrated: (i) that there has been an increase in the proportion of accusative-marking over nominative-marking in recent times, (ii) that accusative-marking is more likely to be chosen in certain types of subordinate clauses than in matrix clauses, and (iii) that at least with some types of predicates, accusative-marking on the object is more likely when its referent is animate rather than inanimate. It will be argued that the effects of clause type and animacy have to do with the functional motivation to mitigate processing load and the risk of misinterpretation incurred by nominative-marking on objects.