A Minimal Triplet in Altaic: Round Licensing, Harmony, and Bisyllabic Triggers*

This paper presents an optimality-theoretic comparison of three round vowel patterns in the Altaic family. At the core is an analysis of bisyllabic trigger round harmony, a pattern uncovered in recent investigation of Classical Manchu and Oroqen (Tungusic; Zhang 1996). In these languages round spreading takes place only when the first two syllables of a word are round. This study isolates two separate properties of round vowel distribution in bisyllabic trigger patterns: round licensing ([+round] must be linked to the initial syllable) and round spreading. Bisyllabic triggers are derived through a constraint interaction explored here, called ÔParasitic Constraint SatisfactionÕ, in which a given constraint loses to satisfaction of a dominating constraint except when violation of that dominating constraint is independently induced by a third higherranked constraint. In the case of bisyllabic trigger round harmony, this kind of interaction is achieved by ranking licensing and spreading separately with respect to a tautosyllabicity constraint on features. More detailed examination of the spreading and licensing requirements finds that they can be analyzed solely in terms of markedness and faith constraints. Positional identity (Beckman 1997) and markedness of feature cooccurrences play a critical role, characterizing the special status of the initial syllable and realizing height stratification of rounding restrictions. The related patterns of simple round licensing and simple round harmony, exhibited in two other Altaic languages (Classical Mongolian and Ulcha), are produced by minimally distinct constraint hierarchies, bearing out the predictions of factorial ranking. This paper presents a comparison of three round vowel patterns in the Altaic family. At the core is an analysis of a pattern of round harmony uncovered in recent investigation of Classical Manchu and Oroqen (Tungusic; also known as Manchu-Tungus). These languages complicate the usual pattern of Tungusic harmony by requiring round vowels in the first two syllables of a word to initiate round spreading (Zhang 1996; Zhang and Dresher 1996). The two-syllable trigger requirement simply expresses a descriptive generalization; the challenge for phonological analysis is understanding what underlies this outcome without stipulating a numeric threshold condition. This paper aims to derive the bisyllabic trigger effect through the ranking of optimality-theoretic constraints (Prince and Smolensky 1993). The analysis draws on familiar phonological demands in Altaic languages, such as round licensing (the requirement that [+round] have a link to the initial syllable) and round spreading, which produce multiple-linking of vowel features. In the bisyllabic trigger languages, both licensing and spreading are visibly active. These demands are shown to each conflict with a constraint prohibiting cross-syllable feature linkage. The innovative theoretical result is the parasitic constraint satisfaction that arises when the featural tautosyllabicity constraint is interleaved between them: the lower-ranked spreading constraint is satisfied only when it can feed off the tautosyllabicity violations produced by round licensing. This parasitic effect realizes the two-syllable trigger condition as a consequence Ñ no reference is made to the number two. The second theoretical goal of this study is to employ a minimal set of well-motivated constraints and explore how different rankings produce three Altaic round vowel patterns: round licensing, round harmony, and bisyllabic triggers. Only three kinds of constraints are drawn on in this analysis: faith, markedness, and edge constraints. The edge-based constraint is one of a * For comments on this work I am grateful to Jill Beckman, Diamandis Gafos, Junko It™, John McCarthy, Armin Mester, Jaye Padgett, Geoff Pullum, and Cathie Ringen, also audience members at the LSA and CLA, and participants in phonology reading groups at the University of Massachusetts, Amherst and the University of California, Santa Cruz. This research was supported by SSHRC fellowship 752-93-2397 and NSF grant SBR9510868 to Junko It™ and Armin Mester and it benefitted from ideas discussed in meetings for NSF grant SBR9420424 to John McCarthy. Any errors are my own.