Incomplete Neutralization in Japanese Monomoraic Lengthening

Incomplete neutralization (IN) (Port et al. 1981, Fourakis & Iverson 1984, Port & O'Dell 1985) refers to cases in which two underlyingly distinct segments become nearly identical on the surface. IN has posed a challenge for traditional views of the phonetics-phonology interface. While classical modular feedforward architectures (e.g. Chomsky & Halle 1968, Bermúdez-Otero 2007) generally do not allow underlying phonological representations to directly affect phonetic realization, incompletely neutralized contrasts show subphonemic distinctions on the surface that can only be inferred from the underlying representations. We show that the combination of two independently motivated theoretical mechanisms – paradigm uniformity (Benua 1997, Steriade 2000) and weighted phonetic constraints (Legendre et al. 1990, Zsiga 2000, Flemming 2001, Pater 2009) – automatically account for the IN patterns