Acoustic, Psychometric and Lexical Neighborhood Properties of the Spondaic Words: A Computational Analysis of Speech Discrimination Scores

Author: Ted A. Meyer, David B. Pisoni, Paul A. Luce, and Robert C. Bilger

Abstract:
Luce and Pisoni (in press) developed a model of spoken word recognition based on neighborhood probability characteristics of monosyllabic words. This model, known as the Neighborhood Activation Model (NAM), assumes that words are recognized in the context of other similar sounding patterns in the lexicon. NAM predicts that word recognition is dependent on how easy or difficult it is to confuse the stimulus word with phonetically similar words in a lexical neighborhood. Cluff and Luce (1990) demonstrated that spondaic word recognition is also dependent on the confusability of the individual syllables. Spondaic words with two easy to recognize syllables are identified more accurately than spondaic words with two hard to recognize syllables. Recently, Bilger et al. (submitted) analyzed the 36 spondaic words [Tillman et al. (1963) recording of the Hirsh et al. (1952) spondaic words] currently used to determine the speech reception threshold (SRT) (American Speech-Language-Hearing Association, 1988). They found that the thresholds and slopes of psychometric functions for the spondaic words were not equivalent, and that the spondees were not equally intelligible. In this study, we computed the lexical neighborhood characteristics (Pisoni et al., 1985) of the individual syllables in these spondaic words and compared these values to psychophysical and acoustic measures obtained from Bilger et al. Although spondaic word thresholds were not related to any of the lexical neighborhood measures of the words, the slopes of the psychometric functions were negatively correlated with the neighborhood densities of the spondaic words. This finding demonstrates that the rate at which a word becomes intelligible is inversely related to the confusability of the word with its lexical neighbors in memory. Implications for the development of new tests for speech discrimination are discussed.