Title: Identifying register and speaker 'stance' from prosodic strategies, a comparison of Japanese and English parallel corpora.

Work in progress…

It is well-documented that 'register' (or 'style') has a major influence on linguistic variation (Rickford & Eckert 2001), and that intonation is one of the varying parameters (or 'factors') implicated in this variation. Recent advances in data collection and acoustic analysis facilitate accurate investigation of prosodic variation for corpora from a much wider range of situations than was possible in the past. In the process of analyzing negation strategies, we have found that an accurate study of the prosodic parameter will require analysis of a given speaker's 'stance' and 'footing' within the interaction, because the same type of 'speech act' — say 'agreement' or 'disagreement' - are not treated in the same way by speakers in different social situations.

This paper will consider a specific environment in which focal accent can be expected: not-negation and nai-negation [for English and Japanese, respectively] used in actual interactions. The paper will compare prosodic prominence on not/nai in English and Japanese. Speech researchers have concluded that pitch prominence occurs on not almost categorically in read sentences (O'Shaughnessey & Allen 1983) and in read news (Hirschberg 1990, 1993), but analysis of more interactive task oriented data has shown that, at least in English and French (Yaeger-Dror in press) prominence on negatives is not common unless the speakers are using an adversarial footing (as they do, for example, in political debates). This paper will compare English and Japanese 'CallFriend' phone conversations between friends (from the LDC; Url: www.ldc.upenn.edu) and compare those conversations with political debates in both languages. The paper will show that there is a wide range of variation in prosodic strategies used for conveying negative information and that the variation is correlated with both the language being spoken and the social situation.

All tokens of not or n't (for English) and nai (for Japanese) in the declarative sentences in the corpora were coded for linguistic environmental factors, interactive factors [e.g., whether the negative is considered to be a repair, what the relationship is between interlocutors, what kind of social situation it is], as well as for intonational characteristics [using a system based loosely on ToBI and J_ToBI]; results of the multivariate analysis of the coded data will be discussed. The evidence will demonstrate that register and interactive factors are critical for an adequate understanding of the intonation used for negatives. It will demolish the hypothesis that information which is critical to the hearer must be prosodically prominent, since in many situations interactive 'face' concerns outweigh cognitive imperatives. [Work supported by NSF, ….. and by grants from the JFK Library, and CMU.]

O'Shaughnessy and Allen (1983) Linguistic modality effects on fundamental frequency JASA 1155-1171.

Rickford and Eckert, eds. (2001) Style and Sociolinguistic Variation. Cambridge: CUP.