Michael Wagner will travel to Cornell University this week for the dissertation defense of Masayuki Gibson. The defense, titled “A Rise Is Not a Rise Is Not a Rise: The interaction of lexical tone and sentential intonation”, will take place Tuesday, December 6 at 4:00pm. The dissertation abstract is below.

There is still much to be learned regarding the nature of the interaction between lexical tone and sentence-level intonation.  Previous studies in individual languages tend to be too narrow, focusing on ways to model the final F0 output without regard to cross-linguistic implications; studies mainly concerned with phonological patterns across languages tend to over-generalize, missing or glossing over many language-specific and category-specific phenomena.  This dissertation attempts to address the gap left by these previous studies.

The first part of the dissertation presents results from a series of production and perception experiments conducted for a handful of tone languages, including Standard Mandarin, Henan Mandarin, Hong Kong Cantonese, North Kyeongsang Korean, and Kansai Japanese (a family of dialects including Shiga, Kyoto, Osaka, and Kobe Japanese).   The production experiments were designed to elicit multiple renditions of various lexical tones in declarative and echo question contexts, and the perceptual experiments were designed to test the degree of recoverability for each communicative function (tone and intonation) in the various conditions.

The second part of the dissertation considers the implications of the experimental results for building a comprehensive model of speech melody.  First, by examining the behavior of intonation across tonal categories within each language, I show that there is evidence for unpredictable tone-dependent intonation implementation, suggesting that our model must allow for some interaction between the two at some level before phonetic implementation.  In addition, I assess the results cross-linguistically, characterizing the ways in which the model must be parameterized.  Finally, I propose a model that meets both of the above demands.  The phonological component of this model includes an autosegmental geometry that captures tone in languages like Mandarin and Cantonese as tones associated with syllables and so-called “accentual melodies” in languages like Kansai Japanese and NKK as tones that associate with words.