Towards a linguistic-phonetic interpretation of neural plasticity in

the auditory pathway

 

Sarah Hawkins

University of Cambridge

Linguistics Department

Cambridge, England.

 

This talk is a first attempt to bring together a new, 'polysystemic', view of how speech units are organized with the type of neuronal plasticity demonstrated in the recent findings of Fritz, Elhilali and Shamma (2003, 2005). Standard models of speech perception and comprehension assume that the first stage of processing involves analyzing and categorizing elements of the sensory speech signal into a small number of "phonetic categories" (e.g. phonetic features, phonemes, gestures) which are devoid of sensory content. In these models, meaning is worked out only after this first transformation from physical signal to abstract sequence of phonological units has been done.

 

However, much variability in the realization of speech sounds is in fact informative, not just for enhancing phonological form, but also for directly conveying grammatical, semantic and discourse information, even when the phonemes are identical. For example, in connected speech, the way segments in /aIm/ are pronounced can indicate whether they signify the function word 'I'm' or are part of a content word like 'lime'. Support for the alternative, (polysystemic, function-oriented) theoretical approach thus comes from evidence of perceptually-salient variation in the speech signal that helps us to distinguish different meanings, but does not distinguish phonemes. In this alternative approach, speech is structured as multiple, hierarchically organized perceptual units, which, by their nature, cannot be represented independently of their broad linguistic function and context. This approach rejects the basic phonological tenet of standard perceptual models and deemphasizes the distinction between knowledge and sensation, thus encouraging re-evaluation of how sensory information is used in understanding speech. Classical and recent behavioral data are interpretable within this framework, and suggest that people listen minutely to phonetic detail to extract phonological, grammatical and semantic information in parallel, with task demands directing attention to different aspects of the signal. The anatomy and physiology of the mammalian auditory system seems more compatible with this theoretical approach, and recent experimental work on the physiology of animal hearing emphasizes rapid plasticity in, and context- and task-dependence of, auditory cortical sensitivity.