Description:

Talk by Suyeon Im (UIUC).

Evaluating the domain of f0 encoding with imitated speech
 
What is the appropriate segmentation of a phrasal f0 contour that corresponds to the domain in which intonational features are cognitively encoded? We investigate this question for American English by modeling f0 contour similarity between imitated sentences and their stimuli over domains of varying size and prosodic status, from the syllable to the prosodic phrase. Results show the greatest similarity between imitated and stimulus f0 contours for the contour modeled holistically in the domain of the entire intermediate phrase. These findings are contrary to the predictions of the compositional AM model, and are also surprising in light of claims that the prenuclear region does not encode semantic/pragmatic meaning.