In this paper we present evidence for ternary metrical constituents in Gilbertese (Kiribati), a Micronesian language. The terminal elements of stress feet in Gilbertese are moras, not syllables. Further, the typical foot in Gilbertese contains three moras. These trimoraic constituents are units of stress in Gilbertese, and also define minimal prosodic word size, where possible. Ternary metrical constituents of the sort found in Gilbertese are quite rare cross-linguistically, and as far as we know, Gilbertese is the only language in the world reported to have a ternary constraint on prosodic word size. We present a constraint-based analysis of Gilbertese prosody that also makes use of a language-specific template-mapping algorithm.
In this paper we present evidence for ternary metrical constituents in Gilbertese, a Micronesian language spoken in Kiribati by approximately 55,000 people.2 The terminal elements of stress feet in Gilbertese are moras, not syllables. Further, the typical foot in Gilbertese contains three moras. These trimoraic constituents are units of stress in Gilbertese, and also define prosodic word size and shape. Ternary metrical constituents of the sort found in Gilbertese are quite rare cross-linguistically, and as far as we know, Gilbertese is the only language in the world reported to have a ternary constraint on prosodic word size.
This study offers two fundamental contributions to linguistic theories of stress. It provides the first in-depth study of prosody for a Micronesian language, thereby [End Page 203] contributing to the overall empirical basis of stress typology. Second, it provides a principled constraint-based analysis of Gilbertese prosody. The occurrence of trimoraic feet where possible in Gilbertese but bimoraic feet elsewhere leads us to propose that ternarity is the consequence of a grammar of ranked and violable metrical constraints. Essentially, we argue that ternarity results from the interaction of a binary requirement on foot size with a binary requirement on foot-head size. The failure of Gilbertese words to obey syllable integrity (Prince 1976; Hayes 1995), the principle that demands that a syllable never be split by foot structure, is a straightforward indication of the low ranking or absence of the syllable integrity constraint in this language.
Languages with purported ternary constituents and violations of syllable integrity are not new. Levin (1988a), Everett (1988), Dresher and Lahiri (1991), and Rice (1992) present arguments for ternary metrical constituents in Cayuvava, Pirahã, Old English, and Chuvach Alutiiq, respectively. Hayes (1995) summarizes descriptions of ternary patterns for Estonian, Finnish, Karelian, Hungarian, Sen-tani, Winnebago, and Auca as well. Data from Banawá (Buller, Buller, and Everett 1993) demonstrate violations of syllable integrity. What is unique to Gilbertese is the convergent evidence for ternary metrical constituents in the stress domain and with respect to a minimal prosodic word constraint. A novel aspect of this analysis is the account of minimal prosodic word effects. Though ultimately minimal word effects follow from the prosodic hierarchy in which word dominates foot, in Gilbertese, specific violations of the prosodic-word minimum forces one to the conclusion that a language-specific rule is responsible for bulking words of two moras up to three moras. Before presenting the Gilbertese facts in sections 3 and 4 and our analysis in section 5, we begin with a summary of earlier work on Micronesian prosody.
2. Micronesian Prosody.
Rehg (1993) is the first compilation of prosodic descriptions of Micronesian languages, descriptions that have tended to be incomplete, contradictory, and impressionistic. Nevertheless, Rehg is able to glean overarching generalizations from earlier descriptions of Pulo Annian, Puluwat, Woleaian, Ponapean, Marshallese, Gilbertese, and Kosraean. One generalization is that in all of these languages, it is the mora, not the syllable, that is the prosody-bearing unit. We will demonstrate this point for Gilbertese in section 4. Another generalization is that wherever high pitch and stress are described, they fall on adjacent moras. Based on these observations, and the reconstitution of historical final vowels, Rehg is able to reconstruct Proto-Micronesian primary accent in unmarked terminal prosodic contours. His proposal for Proto-Micronesian accent is shown in (1).
(1) Proto-Micronesian accent (Rehg 1993)
[End Page 204]
The unmarked terminal contour...