Noise intensity and spectrum cues of spoken fricatives

Abstract

The noise of utterance initial/f/ and /θ/ in naturally spoken syllables /fi/ and /θi/ was replaced by splicing-in both /s/ and /f/ noise cut away from spoken /si/ and /∫i/ syllables. The intensity of the replacing noise had been attenuated to the level of the original /f/ and /θ/ noise. A natural sounding /f/ or /θ/ resulted depending upon whether the original fricative had been /f/ or /θ/, respectively. This procedure was previously utilized by the author in relation to intervocalic fricatives (J. Acoust. Soc. Am. 63, S21(A) (1978)] with similar results. These results suggest that the lower noise intensity of spoken /f/ or /θ/ distinguishes them from /s/ and /∫/. The range of noise spectrum variation which is physically manifest in spoken /f, θ, s, ∫/appears to have no perceptual significance for /f/ and /θ/, which are distinguished from each other by the formant transition cues of the vocalic portion of the syllables. However, noise spectrum differences are important cues in the distinction between /s/ and /∫/. This evidence favors the categorization of both /f/ and /θ/ as nonstrident in distinctive feature terminology.