Phonological augmentation in prominent positions
This dissertation presents a theory of markedness constraints that apply exclusively to material in phonologically prominent or “strong” positions, called here M/str(ong) constraints. It is proposed that two substantively based restrictions hold of such constraints. The first restriction is the Prominence Condition, which states that the only legitimate M/str constraints are those whose satisfaction enhances the perceptual prominence of the strong position in question. For example, an M/str constraint demanding high-sonority nuclei in the strong position stressed syllable is legitimate, but a constraint that simply bans a typologically marked feature value in some strong position is not. The Prominence Condition correctly predicts that all M/ str constraints are prominence-enhancing or augmentation constraints.
The second restriction, the Segmental Contrast Condition, applies to M/str constraints on positions that are strong for psycholinguistic (as opposed to phonetic) reasons. This restriction has its basis in the importance of psycholinguistically strong positions for early-stage word recognition. It prohibits any M/str constraint from referring to a psycholinguistically strong position if its satisfaction would impede early-stage word recognition, such as by neutralizing segmental feature contrasts (except for those that improve left-edge demarcation, which potentially facilitates word recognition). Thus, an M/str constraint calling for high-sonority nuclei in the psycholinguistically strong position initial syllable, despite passing the Prominence Condition, will be banned by the Segmental Contrast Condition; its satisfaction neutralizes a segmental contrast that is not at the left edge.
The Prominence Condition and the Segmental Contrast Condition are formally implemented as filters on the output of generalized constraint-building schemas, determining which of the logically possible M/str constraints are actually included in the universal constraint set. In an extension of Inductive Grounding (Hayes 1999a), these and other constraint filters are viewed as the locus of functional grounding in the formal phonological system. This Schema/Filter model allows the constraint set to reflect substantive phonetic and psycholinguistic factors, while maintaining a view of phonology as a formal system that manipulates formal objects—including constraints and the basic phonological elements from which they are constructed—without direct access to fine-grained details of articulation, acoustics, perception, and processing.