Lexical Classes in Japanese: a Reply to Rice

要旨

Phonology at Santa Cruz, Vol. 6, 1999, pp. 39-46. Lexical Classes in Japanese: a Reply to Rice Junko Ito, Armin Mester, and Jaye Padgett University of California, Santa Cruz Ito, Mester, and Padgett (1995) make an argument, based on the interaction of compound voicing and postnasal voicing in Japanese, that the [voice] specification of certain nasal- obstruent clusters, though redundant, is phonologically active. The argument presupposes a well known division of the Japanese lexicon into separate strata—native or Y AMATO , S INO - J APANESE , sound-symbolic or M IMETIC , and more recent borrowings. Recently, Rice (1997) has attempted to cast doubt on the argument for active redundant [voice], by questioning the motivation for the relevant lexical strata. The intention of this squib is to address her arguments, and show that the posited stratal divisions for Japanese are indeed well motivated. 1 Background Compound voicing (Rendaku) involves the voicing of initial obstruents in second compound members meeting the right structural conditions, as shown in (1) (see Ito, Mester, and Padgett 1995, Rice 1997, and works cited there for further examples). ori + kami 6 origami ‘paper folding’ Rendaku is blocked when the targeted word already contains a voiced obstruent, as shown in (2). • iro + tabi 6 • irotabi ‘white socks’ * • irodabi This blocking, known as Lyman’s Law, is a reflex of a more general prohibition on roots containing two voiced obstruents, such as *dabi, *baga, etc. Both the constraint on roots and Lyman’s Law are attributed to an Obligatory Contour Effect involving [voice] by Ito and Mester (1986). The argument for active redundant [voice] is based on the fact that postnasal obstruents also block Rendaku, as shown in (3). take + tombo 6 taketombo ‘bamboo dragonfly’ (a toy) *takedombo This fact is significant, because voicing in postnasal obstruents is predictable (in Y AMATO words—see below). There are no words such as *tompo or *unsari next to actual tombo ‘dragonfly’ and unzari ‘disgusted’, etc. Within traditional generative theories of underspecification (see Steriade 1995 for references and an overview), this implies that postnasal voicing is underlyingly absent and therefore phonologically inactive. In fact, the © 1999 by Junko Ito, Armin Mester, and Jaye Padgett

主題