Elusive contrasts in the intonational tunes of American English

Jennifer Cole, Northwestern University

It is well-known that phrasal intonation in English plays an important role in conveying pragmatic meaning. Yet despite a rich literature offering diverse accounts of form-meaning associations, there remain many gaps in our understanding. Progress is stymied by a lack of consensus about the sound encoding of intonation (the units of representation) and difficulty in identifying acoustic correlates of phonological form in dynamic acoustic signals (F0 trajectories), in addition to challenges in developing experimental paradigms to test intonational meaning. In this talk I address the first two of these challenges, focusing on the sound encoding of intonation, and specifically, on the encoding of the phrase-final F0 trajectories (or nuclear tunes)in American English, where distinctions in the dynamics of rising, falling and flat pitch trajectories are associated with distinctions in pragmatic meaning related to information structure (focus, givenness) and speech act (e.g., assertion, question). I present findings from a series of experiments testing the production and perception of nuclear tunes. In the production experiments, nuclear tunes are elicited using an imitation paradigm where participants hear a tune on a model utterance and reproduce it on a new sentence. Distinctions among the pitch trajectories of imitated productions are modeled in two ways: using clustering analyses on unlabeled F0 trajectories, and using targeted F0 measures with pitch trajectories associated with phonological tune labels (based on the ToBI annotation system). The results reveal a robust primary distinction between high-rising and non-high-rising tunes, with finer and more variable secondary distinctions within each of these tune classes. A perception experiment using the AX discrimination paradigm with the same stimuli used in the production experiment reveals parallel findings. I discuss the challenges these data present for the Autosegmental-Metrical (AM) model of Pierrehumbert (1980, and later work), in the elusive contrast and fine-grained phonetic variation of nuclear tunes. The talk ends with a pointer to our new work that integrates the phonetics and phonology of intonation through dynamical systems modeling of pitch. 

This work is joint with Jeremy Steffman (Northwestern University).


An in-person viewing will be held in the Linguistics Department seminar room; the event will be followed by a reception from 5-6pm in the Linguistics Department Library.