Introduction

In the previous two chapters I examined the development and content of various theories of phonological representation, segmental and non-segmental, procedural and declarative. I argued that segmental theories of representation require the introduction of excessively powerful, procedural, rewriting rules, whereas a non-segmental approach to representation can avoid the need for rules of such a powerful kind, by expressing phonological relations in the representations, rather than in the rules. I also made some more specific proposals regarding a non-segmental, non-procedural approach to the definition of phonological representations. In chapter 3 I presented a general method for representing ‘rewrite rules’ in phrase-markers, with special cases for (strictly) context-sensitive rules and the well-known case of phrase-markers for context-free rules. Each derivation (or rather, each class of equivalent derivations) can then be represented as the join (unification) of a set of local graphs, each of which represents a rule. Because unification is associative, the notion of ‘order of rule application’ becomes rather meaningless. In this way, I showed that any grammar can be interpreted declaratively, rather than procedurally.

In chapter 4 I demonstrated the great value of Unification-based Phrase-Structure Grammars as a common formalism for integrating the key concepts of Autosegmental and Metrical Phonology. I showed that the focus and spreading of autosegments and the prosodic hierarchies of Metrical Phonology can both be formalised using only context-free immediate dominance and linear precedence constraints.