Effects of context-sensitive phonetic variation and lexical structure on the uniqueness of words

Abstract

Relatively little is known about the optical phonetic speech characteristics to which speechreaders are attuned. However, it is known that phonetic context does affect visual confusability of phonemes. In study 1, behavioral experiments were performed to examine effects of context-sensitive phonetic variation on the visual confusability of consonants and vowels. Hearing and deaf participants identified consonants or vowels presented in carefully constructed sets of nonsense syllables. Vowels were presented in four different consonant–vowel–consonant environments. Consonants and frequently occurring consonant clusters were presented in 14 different disyllabic nonsense words environments. These disyllabic environments varied target consonant position (initial, medial, and final) and target consonant adjacent vowel. Vowel identification was relatively accurate and insensitive to surrounding consonantal context. Consonant identification significantly varied as a function of both position and vowel context. In study 2, computational experiments were performed to assess the importance of patterns of context-sensitive visual confusability on the uniqueness of words in the language. The computational experiments in study 2 provide evidence that choice of a segmental intelligibility model does matter for modeling the uniqueness of words. However, the distribution of words in English substantially preserves lexical uniqueness even when phonetic variability is taken into account. [Work supported by NIH/NIDCD Grant No. DC02107.]