Copying Contiguous Gestures: An Articulatory Account of Bella Coola Reduplication

Reduplication in Bella Coola (or Nuxalk) has been described as highly irregular (e.g. by Newman 1971 and Nater 1984), involving different prefexing and infixing patterns, and a variety of phonemic alternations, as well as expressing a range of grammatical functions. Analyses by Bagemihl (1991) and Carlson (1997) have identified some of the factors that determine shape and position of the reduplicant, but have continued to rely on lexical specification for many of the apparent irregularities. Using articulatory gestures as the basic unit of analysis (following Browman & Goldstein 1992), I argue in this paper that reduplication in Bella Coola is best analyzed as the copying of a contiguous sequence of overlapping gestures.

In particular, I show that several generalizations emerge on the gestural level which cannot be captured on the segmental level. For example, the reduplicant always contains a vocalic gesture that is coordinated with a limited number of preceding and following consonantal gestures. As a result of this limitation, multiply articulated segments (with non-simultaneous component gestures) may reduplicate only partially. Reduplicated ejectives lose their glottal gesture post-vocalically, but their oral gesture pre-vocalically, so that nik' 'to cut' reduplicates as nix 'to cut continuously', but k'nc 'sperm whale' as ÷nk'nc 'small sperm whale' (C' denotes glottalization). Note that an independent lenition process causes coda stops to be reduplicated as homorganic fricatives (see Carlson 1997), a fact that can be ascribed to the familiar emergence of the unmarked ranking schema (McCarthy & Prince 1994). Given a coordination pattern in which the onset of the oral gesture precedes that of the glottal gesture (see e.g. Kingston 1985, Steriade 1997), the copied gesture in each case can be identified as the one which is temporally closer to the vocalic gesture, i.e. the reduplicated gestures are contiguous in the base. Proposing a constraint I-CONTIG-G that requires the copying of a contiguous string of gestures in the base (adapting the notion of segmental contiguity of McCarthy & Prince,1995), and a constraint *COMPLEX-G that prohibits sequences of consonantal gestures, the correct candidates can be derived. This is shown in tableaux (1) and (2), with segments representing the different gestural constellations (thus [k'] denotes a sequence of an oral tongue dorsum gesture followed by a glottal constriction gesture). Lower ranked constraints violated by the optimal candidate (e.g. B-R correspondence constraints) are not included in these tableaux.

(1) Partial copying of post-vocalic ejectives

(2) Partial copying of pre-vocalic ejectives

In my paper I account for further generalizations, including the non-occurrence of coda consonants after sonorant nuclei in the reduplicant, and phonemic alternations undergone by other consonants in the base. For example, sonorants consist of both a vocalic and a consonantal gesture, e.g. /l/ consists of a vocalic tongue dorsum retraction gesture and a consonantal tongue tip closing gesture. Given the presence of a consonantal gesture in the sonorant, *COMPLEX-G prevents the copying of a subsequent consonantal gesture. These generalizations with respect to the reduplicant's shape can only be stated at the level of gestures. It follows that Bella Coola reduplication requires reference to a level of representation below the segment, namely, gestures and their timing relations. In sum, along with other recent work (e.g. by Sproat & Fujimura 1993, Silverman 1997, Steriade 1997, and Gafos, in print) the core argument of this paper provides converging evidence for the relevance of gestural timing in phonology.