The influence of non-canonical vowels on phonetic parsing

Abstract

Speech segments are highly context sensitive due to coarticulation. This paper investigates phonetic context effects and how listeners parse the acoustic signal (Fowler, 1984). The previous literature reported that listeners could parse adjacent segments in a dissimilative manner (e.g., Mann, 1980), or in an assimilative manner (e.g., Fujimura et al., 1978). More recently, Rysling et al. (2019) reported that spectrally ambiguous or discontinuous transition can lead dissimilative parsing than assimilative parsing. This paper examines how non-canonical vowels affect the perception of adjacent consonant. When listeners hear non-canonical vowels, they may parse the incoming phonetic signal matching canonical exemplars, and assign any deviance in the spectral frequencies to the adjacent consonant. Hence, listeners may do assimilative parsing even with or without appropriate vowel transitions. American English listeners identified synthetic CV and VC syllables with an ambiguous fricative noise between [s] and [ʃ], and vowels including a good exemplar and poor exemplars of [i] and [u], each vowel with or without vowel transitions. Analyses are currently underway, but will illuminate the cause of assimilative parsing with respect to exemplar matching in the exemplar-based model (e.g., Johnson, 1997; Pierrehumbert, 2001).