Normal Speech Development
This project is funded through a grant from the National Institutes of Health-National Institute on Deafness and Other
Communication Disorders. It seeks to examine how normally developing children learn to recover phonemic structure from the
ongoing speech signal, which lacks any explicit segmental structure, as well as how they learn to produce speech in the
precisely organized fashion of adult speech. So far, results suggest that the ability of language users to apprehend phonemic
structure in the speech they hear depends on how well they recognize relevant kinds of structure in the speech signal, and
organize that structure. There are perceptual strategies that are optimal in terms of allowing the listener to recover phonemic
structure. These optimal perceptual strategies differ across languages. We suggest that children learn these optimal strategies
through experience with their native language. This project continues to test the basic premises of this theoretical position, and
to extend our understanding of the developmental process involved in acquiring mature perceptual strategies.
Traditionally, experiments in the laboratory have been heavily dependent on labeling procedures. In these tasks, the acoustic
structure of stimuli is varied such that several properties can provide information about phonemic identity. Procedures are crafted
to allow the estimation of the relative weights assigned to each property in the labeling decision.
The Ontogeny of Segmental Speech Organization
Fricative-vowel syllables were used regularly during the early stages of theory building. In these examples, the fricative noise is a
single-pole, synthetic noise. The vowel portions are natural, with formant transitions appropriate for a preceding ‘sh’ or ‘s.’
More recently, we have been interested in how children discover the global structure of their native language. In these
experiments, we process speech to preserve some kinds of structure, and diminish other kinds. In particular, stimuli have
typically been processed to retain only amplitude envelopes or time-varying spectral structure (i.e., sine wave speech).
Examples of such processed signals are below. The sentence is Knees talk with mice.