Acoustic cues for place of articulation in the one‐word stage of language acquisition

Abstract

The theory developed by the Haskins group claims that place of articulation is cued by context dependent, formant frequency transitions. An alternate theory [S. E. Blumstein and K. N. Stevens, J. Acoust. Soc. Am. 66, 1001–17 (1979)] proposes that children initially use innate, invarient cues to identify place of articulation and learn to use formant frequency transitions. We analyzed 30 word‐initial voiced stops of two children in the one‐word stage of speech acquisition. Onset spectra computed using LPC analysis with a 25.6 ms window were compared with the Blumstein and Stevens templates. In addition, computer editing techniques were used to produce three series of stimuli for psychoacoustic tests. Listeners were asked to identify place of articulation when they heard only the first 15, 40, or 135 ms of each utterance. Approximately 80% of the stops had invarient cues. Context dependent cues clearly are not the sole cues as the Haskins theory implies. However, place distinctions for some stops which could not be determined using the templates, were identified in the longer computer‐edited versions, where formant transitions were present. [Work supported by NIH.]