Phonological Occurrences: Relations and Copying

We formally and empirically examine the phonological relation immediate precedence, its properties, and further relations that can be derived from it. When each occurrence of a phonological segment in a nonlinear multiprecedence structure is reconcatenated into a linear sequence of copies, reduplication results (Raimy 2000a). A number of consequences arise in the representation of affixes, in modelling allomorphy in Hindi and Turkish echo reduplication, and in explaining patterns of multiple reduplication in Lushootseed (Salish). We offer directions for further analysis of reduplication that distinguish conditions on multiprecedence from other wellformedness modules without “the emergence of the unmarked” as a central tenet. 1 The Fundamental Role of Immediate Precedence Raimy (2000a:12) observes that, “there are non-trivial and non-derivable ordering relationships between segments in phonology.” These relationships are non-trivial in at least two ways: (a) there are no palindromic or anagrammatic languages where, e.g., [kæt]=[tæk], and (b) phonological rules, processes, and constraints must have access to ordering information. A rule like A⇒B/C D can only apply if the information is available that C immediately precedes A and A immediately precedes D. Crucial OT Constraints like Linearity and Contiguity also make explicit use of immediate precedence information. Raimy proposed that these crucial precedence relations be explicitly represented in phonology: A→B is read ‘A immediately precedes B.’ In this system, (1a) and (1b) both represent what is traditionally represented as [kæt]. (1) a. start → k → æ → t → end b. end ← t ← æ ← k ← start The beginning (start) and end (end) junctures are needed here in order to be fully explicit about order. These are the only symbols that are not both immediately preceded and immediately followed by at least one segment. If these symbols were excluded, some other convention would have to be adopted to determine the first and last segment in a precedence representation. start and end are further motivated by the need ∗Many thanks to audiences at the 2nd North American Phonology Conference, the 27th Penn Linguistics Colloquium, and participants of the 10th EGG Summer School course “Multiple Introductions: Precedence for Syntacticians, Dominance for Phonologists,” where portions of this work have been presented. We have benefited immensely from discussions with Klaus Abels, David Adger, Eugene Buckley, John Frampton, Morris Halle, Aaron Iba, Victor Manfredi, Alec Marantz, Eric Raimy, Charles Reiss, and Bert Vaux. Only the authors are responsible for the facts, opinions and mistakes contained here. Partial support for this work was provided by a National Science Foundation Graduate Research Fellowship. For the moment, immediate precedence relations only seem motivated between X-slots (or a CV skeleton) on the timing tier. Thus the representations in (1) are shorthand for (i).