How Prosodic Phrasing Conveys Syntactic Structure: Evidence from Production and Perception Experiments

Abstract: Prosodic cues encode information about syntactic bracketing and can–at at least for some structures and at least under certain circumstances– disambiguate attachment ambiguities (Schafer et al. 2000, Snedeker & Truswell 2003, Kraljic & Brennan 2005, i.a.). To what extent prosodic cues reliably encode syntax and how exactly they are used in disambiguation is still controversial. Clifton, Carlson and Frazier (2002) propose the ‘informative prosodic boundary hypothesis’ according to which prosodic boundaries are interpreted relative to earlier boundaries. This talk reports production and perception results that support this relative view of boundary strength and show the relevance of quantitative differences in pre-boundary lengthening. The results suggest that prosodic boundaries later in the utterance are scaled relative to earlier ones in a gradient way, thus enabling speakers to encode low or high attachment without having to commit early in the utterance to an entire syntactic parse. The results throw some new light on incremental production and on why certain types of structural ambiguities may not be prosodically differentiated.

Bio: Michael Wagner received his Ph.D. in Linguistics in 2005 at MIT. After a postdoc in the Department of Brain Cognitive Sciences at MIT, he worked for 2 years as an Assistant Professor at Cornell University. In 2008, he joined the faculty of the Linguistics Department at McGill and has been Canada Research Chair in Speech and Language Processing since 2009. His current research explores how prosody is used to encode and retrieve syntactic and semantic/pragmatic information.