Abstract:

Knowing the perceived similarity between different consonants and consonant clusters in different syllable positions is important for models of word recognition that involve activation of neighborhoods of similar-sounding words. Similarity data may also help in understanding whether clusters are single entities, or are decompositional. Furthermore, the extent to which similarity is dependent on the particular talker is relevant for issues of talker normalization. In this study, participants rated the similarity of pairs of consonant--vowel and consonant cluster--vowel syllables in two different voices (one male, one female) over the course of five sessions. The similarity spaces will be discussed in relation to the issues described above. In general, there was no evidence for the decompositionality of consonant clusters. There was a high correlation (r=0.81) in the similarity spaces between the two voices. However, the results from a subset of the items presented in a third voice showed much lower correlations than did the results from the original two voices. This highlights the need to control factors such as the corpus of items, voices, and context when examining similarity scaling data. [Work supported by NIDCD Grant Nos. R01 DC00219 and R01 DC0265801-A1 to SUNY at Buffalo.]