1. English word stress and Japanese accentuation

Archibald (1997) argues that Japanese learners store the English stress lexically and do not compute metrical structures. This paper, based on the results of two stress assignment tests by university students, argues that they develop a computation system by modifying their native loanword accentuation system, which happens to be quite similar to the English stress system, i.e., no final stressing, heavy penult stressed and if penult is light, then antepenult stressed. One difference between the two systems lies in the final unit to be ignored in the computation. In the native system it is either a light syllable (below, CV, where C stands for a consonant and V stands for a vowel) or a syllable with coda nasal (below, CVN), while in English it can be a syllable with one or more coda consonants (e.g., CVO, CVN, CVNO, CVOO, where O stands for an obstruent). The difference reflects the difference of permissible coda consonants between the two languages. This paper tries to clarify how learners modify the native system to accommodate the difference in syllable structures. In this section, loanword accentuation system within Optimality Theory (Prince and Smolensky 1993) proposed in Katayama (1995) will be presented and in section 2 and 3, experimental results will be presented that lead us to a discussion of the final phenomena to be ignored in stress computation. In section 4, implications for the stress acquisition process will be discussed. On the basis of the constraints on foot formation given in (1), Katayama (1995) proposes three constraints as in (2) in the following ranking order for Japanese loanword accent as in (2): NonFinality>> Accent-to-Head>> Rightmost. (1) Constraints on moraic trochee foot formation Foot binarity: Feet must be binary at some level of analysis (Mora, Syllable). (P&S 1993:47, ref. in Katayama 1995:3) Align (PrWd, R, Ft, R): Align the right edge of every prosodic word with the right edge of some foot