Exploring the use of prosody during language comprehension using the auditory moving window technique.

Researchers in psycholinguistics have speculated about the possible role of prosody in resolving syntactic ambiguities. We argue in this paper that the issue is complicated by the following considerations: first, prosody may be even more effective at conveying semantic information than syntactic structure, yet the question how prosody signals meaning is essentially unstudied. Second, the one-to-many relation between syntactic and prosodic structure leads to a great deal of variability across speakers and contexts in the way a given sentence will be produced. The parser must somehow deal with this variability. Third, resolution of architectural debates in the parsing literature requires the use of sensitive, online techniques for measuring processing load during comprehension. In the auditory domain, no optimal technique is presently available. We discuss the strengths and weaknesses of a technique we introduced in previous work, which is an analogue of the visual moving window. We present the results of an experiment demonstrating that the technique preserves some aspects of the prosody of a spoken sentence but disrupts others, and we discuss ways of dealing with this problem. We conclude that the technique is useful for studying language processing, including the use of prosody during parsing. However, we also argue that researchers should study not just the role of prosody in parsing, but also its role in establishing sentence meaning.