Tempo of Spectrum Change as a Cue in Speech-Sound Discrimination by Infants

Abstract

Six- to seven-month-old infants were tested on their ability to discriminate among three speech sounds which differed on the basis of formant-transition duration, a major cue to distinctions among stop, semivowel and diphthong classes. The three speech sounds, [bε], [wε], and [uε] were produced in two different ways. The stimuli for one experiment were two-formant synthetic tokens which differed in formant-transition duration. The stimuli for a second experiment were produced with a computer-modification technique which artificially shortened or lengthened the formant-transition portion of a naturally produced [wɛ], resulting in tokens of [bɛ] and [uɛ]. The discrimination procedure involved visual reinforcement of a head-turn response following a change from a repeating background stimulus to a contrasting stimulus. Infants in both experiments discriminated [bɛ] from both [wɛ] and [uɛ]; evidence for [wε]-[uɛ] discrimination was obtained for the “computer modified” tokens only. These findings are discussed in terms of possible mechanisms underlying speech perception in infancy.