Loanword adaptation and the evaluation of similarity

Abstract

This paper presents a phonological analysis of loanword adaptation of English loanwords into several Polynesian languages, following and extending a model of phonological structure developed by Clements 2001. This analysis bears on a number of theoretical issues. I argue that the loanword facts presented support the conclusions of Hyman 1970, Paradis & La Charite 1997, and Jacobs & Gussenhoven 2000, who contend that the evaluation of input-output similarity may be influenced by phonological structure of the host language. This draws against several current efforts that reduce the notion of similarity to a purely phonetic-perceptual metric (Steriade 2001, Kenstowicz 2001). I show that marrying Clements' universal feature accessibility hierarchy with the system of contrastive underspecification developed by Dresher & Zhang 2000 and Dresher 2001 restricts the range of feature hierarchies that can be present crosslinguistically. Consequently, basing the evaluation of phonological similarity upon these feature hierarchies also sets limits on the range of nativization strategies we predict can appear cross-linguistically. This model of adaptation is therefore able to be both predictive and explanatory. I explore these generalisations, and further show that they are largely upheld in the data presented. I argue that Clements' accessibility model must be modified in the following ways: Firstly, following Avery & Rice 1989, Ghini 2001, redundant features may be present underlyingly in certain defined instances (implicational activation). Secondly, the accessibility of certain features in the contrastive system is conditioned by prior activation of others (contingent accessibility).