Evaluation of the constant ratio rule for consonant confusions

Abstract

The constant-ratio rule for consonant confusions asserts that the ratio of errors of any consonant to any others remains constant regardless of the size of the matrix from which the cells were chosen. In this study, log-linear models were applied to test the constant-ratio rule. Normal-hearing young adults were presented various closed-set arrangements of syllable tokens, spoken by a male and female speaker, and selected from a set of 14 VCs each paired with the vowel /ɑ/. Stop consonants and fricatives, both voiced and voiceless, were presented at three presentation levels as isolated sets (by feature) and in combination. The relationships among voiceless stop consonants remained constant irrespective of the size and nature of the response foil. Similarly, voiceless fricatives were unaffected by the nature of the closed set. In contrast, patterns of errors among voiced stops were dependent on the set of alternatives, as were voiced fricatives. Speaker differences, individual differences among listeners, and implications relating to the generalizability of confusion data collected in small closed-set arrangements are also discussed. [Work supported by a grant from NINCDS.]