Models of Speech Processing

One of the fundamental questions about language is how listeners map the acoustic signal onto syllables, words, and sentences, resulting in understanding of speech. For normal listeners, this mapping is so effortless that one rarely stops to consider just how it takes place. However, studies of speech have shown that this acoustic signal contains a great deal of underlying complexity. A number of competing models seek to explain how these intricate processes work. Such models have often narrowed the problem to mapping the speech signal onto isolated words, setting aside the complexity of segmenting continuous speech. Continuous speech has presented a significant challenge for many models because of the high variability of the signal and the difficulties involved in resolving the signal into individual words. The importance of understanding speech becomes particularly apparent when neurological disease affects this seemingly basic ability. Lesion studies have explored impairments of speech sound processing to determine whether deficits occur in perceptual analysis of acoustic-phonetic information or in stored abstract phonological representations (e.g., Basso, Casati,& Vignolo, 1977; Blumstein, Cooper, Zurif,& Caramazza, 1977). Furthermore, researchers have attempted to determine in what ways underlying phonological/phonetic impairments may contribute to auditory comprehension deficits (Blumstein, Baker, & Goodglass, 1977). In this chapter, we discuss several psycholinguistic models of word recognition (the process of mapping the speech signal onto the lexicon), and outline how components of such models might correspond to the functional anatomy of the brain. We will also relate evidence from brain lesion and brain activation studies to components of such models. We then present some approaches that deal with speech perception more generally, and touch on a few current topics of debate.National Institutes of Health under grant NIH DC R01–3378 to the senior author (SLS