One of the fundamental claims of Optimality Theory is that by varying the rankings of universal constraints, different grammars result (Prince & Smolensky 1993). Just as the ranking A » B should define an occurring language, so should the ranking B ≫ A. In this paper, we examine this claim in the domain of tongue root harmony systems, specifically with respect to the behaviour of low vowels. We examine cases where the relative ranking of faithfulness conditions and alignment conditions is varied with respect to substantive conditions governing low vowels. Our primary conclusions are twofold. First, we find that the types of typological variation expected to occur do occur; six different types of harmony patterns are presented. Second, we note that a large degree of variation is attested in a very narrowly defined area of the phonology. This paper begins by a basic discussion of the formal constraints assumed to govern vowel harmony, followed by a discussion of a case where low vowels harmonise in a manner comparable to other vowels (Degema). We then turn to six cases (five languages) where we observe asymmetric behaviour. First, we discuss cases involving constraints against feature "insertion" and feature "spreading ", constraints of the faithfulness family (Yoruba, Konni, Ngbaka-Ma'bo). Second, we turn to cases involving constraints of the alignment family, cases where harmony exhibits directional asymmetries (Ngbaka-Ma'bo, Emalhe, Maasai).
展开▼
