14th International Congress of Phonetic Sciences (ICPhS-14)

San Francisco, CA, USA
August 1-7, 1999

Failing to Parse Hz in Perceiving [voice] and [continuant]

John Kingston

Linguistics Department, University of Massachusetts, Amherst, USA

Perceptual theories such direct realism in which the objects of in a sequence depends on their identification of a neighboring speech perception are acoustic properties' articulatory causes rather segment, but that this “context-bias” is not influenced by the than their auditory consequences predict that listeners can keep acoustic properties of the segments. The results presented here test separate perceptually the effects of different articulations on a single the generality of this result by seeing whether listeners' acoustic property's values. This paper reports Dutch and English identification of one feature of a segment is influenced by their listeners' failure to “parse” of the articulatory contributions to F1 at identification of another of its features and whether that “segmentthe edges of flanking vowels in deciding whether an intervocalic bias” is uninfluenced by the segment's acoustic properties; cf. [1] consonant was /b/, /p/, /v/, or /f/. Because F1 is cut back more in vowels flanking [-voice] than [+voice] consonants but its offset frequency is lower next to [-continuant] than [+continuant] consonants, listeners should easily parse the effects of the laryngeal and oral contrasts. Multinomial logit models show unmistakably that they do not: disconfirming this prediction of articulatory theories. The results also show that listeners perceive these phonemes as unanalyzed wholes rather than as combinations of their distinctive feature values.

Full Paper

Bibliographic reference.  Kingston, John (1999): "Failing to parse Hz in perceiving [voice] and [continuant]", In ICPhS-14, 869-872.