Internal structure of phonetic representations: Non-linear analyses of error data from patients with apraxia of speech.

点击翻译按钮获取中文摘要

This proposal is aimed at disclosing the internal structure of phonetic plans or speech motor programs. According to a theory developed by Levelt, Roelofs & Meyer (1999), the input to the motor execution component of speech production consists of linear strings of holistic, syllable-sized programs, but there is only little empirical support to this hypothesis. In our project we pursue a neurolinguistic approach which is based on the assumption that the representations of the phonetic encoding component of speech production are corrupted by a clinical condition termed apraxia of speech. Since the occurrence of apraxic speech errors is systematically influenced by the phonological / phonetic make-up of words, a word¿s sensitivity to apraxic failure can be assumed to mirror the complexity of its underlying phonetic representation. In our approach we estimate error-probabilities for a representative sample of words with different phonological structures and model these data by variables describing the formal properties of the target words. Estimates of error probabilities will be obtained from the speech samples of a large population of apraxic speakers, and a novel method will be used to predict these data by a non-linear metrical model of phonological form. The shape of the resulting model can be interpreted to reflect the internal structure of phonetic representations. The specificity of this result for the phonetic encoding stage will be demonstrated by inclusion of a control sample of patients with phonological impairments. The speech samples and perceptual data collected in this project will be made accessible, in a large database, to other researchers in this field.