Domain and codomain of precedence Sample Clauses

Domain and codomain of precedence. ‌ Unlike dominance and correspondence (and alike surface correspondence in ABC), precedence is a homogeneous relation. This means that only elements from the same set 15 Dominance relations are “cross-relationally” transitive (but not themselves transitive). If a foot Fx dominates a syllable σy in ↓F-σ, and the syllable σy dominates a root node ⊙z in ↓σ-⊙, then the relation ↓σ-⊙ must exist and ⟨Fx, ⊙z⟩ ∈ ↓F-⊙. Simply put, if a foot dominates a syllable σx, then the foot also dominates all the root nodes ⊙y dominated by σx, and all the features dominated by ⊙y.

Domain and codomain of precedence. Unlike dominance and correspondence (and alike surface correspondence in ABC), precedence is a homogeneous relation. This means that only elements from the same set can participate in the relation. The set of all elements that can participate in a precedence relation is defined by the following axiom. If a pair of elements ⟨x, y⟩ are in a precedence relation, then x and y have the same derivational property and the same interpretable property. (xii) This axiom delimits the definition of ALIGN-H and CONTIGUOUS constraints, which are both crucial components of φ-Correspondence. I discuss this issue in § 2.3.2.2 and § 2.4.2.1. In the model, I do not need to impose any other restriction on the sets of elements that can participate in a precedence relation, as the choice is inconsequential. What is important for the present analysis is that precedence acts on input root nodes, output root nodes, and among like feature nodes. Precedence among root nodes is necessary for the definition of I/O constraints that refer to the order of the elements in correspondence. I/O-Correspondence occurs among root nodes, and since both CONTIGUITY and LINEARITY (XxXxxxxx & Prince 1995) refer to the order of elements in correspondence, precedence on both input and output root nodes is commonly assumed. Notice that precedence only exists among feature nodes with the same feature value, as opposed to AP where tiers may contain feature with different values. Such a relation is incompatible with the proposed model. Since segments have no internal geometry, precedence on feature nodes with different feature values (e.g., [+sib] and [−sib]) would simply replicate the precedence relation at the root node level. Most importantly, such a definition would violate Axiom III (and homogeneity), since there is no interpretable property that relates like features with different values.