REMARKS AND REPLIES 639 Which Wi? Is Which? Prefixing and Suffixing in Klamath Full-Root Reduplication Michael R. Marlo Nicholas J. Pharris This article presents a new analysis of full-root reduplication in Kla- math. This process, hitherto termed ‘‘intensive’’ reduplication, has previously been analyzed as involving prefixation. Klamath prefixes normally induce vowel reduction in the following morpheme, but the intensive reduplicant does not. McCarthy and Prince (1995) attempt to account for this divergent behavior by invoking base-reduplicant faithfulness; Zoll (2002) attributes it to a stem-internal/stem-external distinction among prefixes. Based on phonological and semantic crite- ria, our analysis resolves intensive reduplication into several classes, including both prefixing and suffixing types. This analysis improves empirical coverage over previous accounts while minimizing assump- tions and demonstrates that careful morphological and phonological analysis can distinguish between prefixing and suffixing copying in cases of seemingly ambiguous total reduplication. Keywords: reduplication, phonology, morphology, Klamath 1 Introduction It is a difficult analytical issue, when dealing with full-root reduplication, to determine whether the reduplicant is prefixed or suffixed to the base. Indeed, as an anonymous reviewer aptly states, ‘‘One might naı ¨vely even suppose that no arguments could possibly discriminate between the two choices.’’ The Klamath language of southwestern Oregon provides a valuable opportunity to investigate this problem, since this language has phonological rules that apply differentially in prefixing and suffixing operations. In this article, we argue that these rules provide strong morphological and phonological evidence distinguishing prefixing and suffixing full-root redupli- cation. The addition of a prefix to a Klamath word generally causes the first vowel of the following morpheme to be reduced or deleted: wipg-a ‘escapes’, sni-w’Upg-a ‘rescues’, sni-sn-w’Upg-is ‘rescuer, savior’. 1 However, this effect is not seen in the case of the so-called ‘‘intensive’’ mor- We would like to thank Juliette Blevins, San Duanmu, Sam Epstein, Ruth Rouvier, Sally Thomason, and two anonymous LI reviewers for their helpful comments and discussion and for the encouragement to write this article. Any errors are our own. 1 The choice between reduction and deletion is automatic: reduction to schwa occurs before CC or C# and deletion occurs before a CV sequence. For the purposes of this article, we will use reduction as a cover term for both outcomes. Barker (1963) uses a to represent epenthetic [U] and [U] resulting from vowel reduction, as well as underlying /a/. Forms that have a in vowel-reducing environments in Barker 1963 will be rendered with U here. In this example and in those that follow, the hypothesized root is underlined. Consonants followed by an apostrophe (’) are glottalized, full-sized capital letters (M, N, L, W, Y) represent voiceless sonorants, and a small capital G represents an unaspirated uvular stop. All glosses are from Barker 1963, which together with Barker 1964 is the main source of data on Klamath. Linguistic Inquiry, Volume 35, Number 4, Fall 2004 639–656  2004 by the Massachusetts Institute of Technology