Research Focus
Our work examines how listeners recognize spoken words in fluent speech. Research in this area has always played an important role in framing central issues in cognitive science and neuroscience, including those of modularity, categorical perception, the nature of linguistic representation, and the role of innate and environmental constraints on the development and form of cognitive processes. Our goal is to provide a fresh perspective on these questions by focusing on the intersection of speech perception, spoken word recognition and phonological theory, and by developing new tools to extend our inferential reach.
Building Interdisciplinary Bridges
All of our work focuses on identifying the sources of structure that listeners leverage to recognize words given the famously invariant speech signal. This structure can be found at level of acoustic-phonetics, abstract phonological structure, and in the structure of the lexicon. Historically, different disciplines have studied each of these aspects with relative independence. We are interested in integrating the insights of acoustic-phonetics, phonology and psycholinguistic work into a common worldview. Our earliest work was concerned with examining how the patterning of stress described by metrical phonology interacts with cue salience to guide spoken word recognition. This involved the combination of careful acoustic-phonetic measurement of stress correlates of natural speech materials and behavioral techniques. This work evolved into later work on the problem of lexical segmentation used similar tactics. This work culminated in the Good Start Model of lexical segmentation, which argues that acoustic-phonetic word boundary cues enhance the salience of phonetic cues for word onsets, providing an early activation advantage for a speaker’s intended words that could be amplified by top-down lexical influences on speech perception of lexical competition dynamics.
In another line of work we have examined the perception of speech that has undergone lawful phonological modification. This work demonstrated that lawful phonological processes that other researchers had framed as an obstacle to speech perception actually facilitate it. Acoustic-phonetic analyses revealed that English place assimilation is typically non-neutralizing. Drawing on ideas borrowed from autosegmental phonology, we developed a model of feature cue parsing which explained my behavioral data through the lens of the perceptual problem of associating phonetic cues with segmental positions in stored lexical representations.
