The representation of laryngeal-source contrasts in Japanese


Kuniya NASUKAWA

Abstract

Most studies investigating the phonological phenomena of Japanese have traditionally employed the bivalent feature [±voice] to represent the two-way source contrast between 'voiced' and 'voiceless'. This implies that at least three types of process should be found in the system: [+voice] is active; [-voice] is active; and both [+voice] and [-voice] are simultaneously active in processes. However, the processes involving laryngeal-source contrasts in Japanese do not employ these three possibilities in equal measure: [+voice] is traditionally said to trigger most dynamic processes such as postnasal voicing and compounding; [-voice] rarely triggers processes except for high vowel devoicing between voiceless consonants; and no simultaneous participation of both features is attested. Furthermore, in the context of a rule-based multi-stratal model, the bivalent format substantially over-generates the number of unattested processes. 
	In addition, employing the feature [±voice] universally to classify all two-way source contrasts ignores the variation that can be observed in laryngeal-source contrasts across different languages, as reflected in VOT measurement. For example, this assumes that Japanese (showing a contrast between zero and negative VOT) and English (showing a contrast between zero and positive VOT) are identical because they exploit the same lexical source categories at an underlying level. It further assumes that the differences in VOT are derived at some more superficial level where the lexical values of [±voice] are translated into phonetic categories using language-specific rules (Keating 1990). 
	Bearing in mind generative restrictiveness and the significance of cross-language variation in source contrasts, this paper identifies those phonological primes responsible for creating the laryngeal-source contrasts of Japanese. The arguments will be based on Element Theory (Kay, Lowenstamm & Vergnaud 1990; Harris 1994; Harris & Lindsey 1995, 2000), a theory of melodic representation in which (unlike SPE-type distinctive feature theories) primes are privative and can be interpreted separately without needing to be combined with other primes. The theory admits only two autonomous melodic categories for cross-linguistic source contrasts: one contributes aspiration and the other prevoicing. This paper claims that Japanese exploits only the prevoicing element in the representation of phonation-type contrasts, this position being supported by evidence from assimilatory processes, early language acquisition and aphasia. The argument leads to the further claim that vowel devoicing is not a process triggered by the laryngeal element for aspiration, but by a manner element called 'noise'.