Constraint-based phonology

Ph.D., University of Edinburgh
99 pages
This thesis presents the results of a study in the application of logic to phonology, the subfield of linguistics concerned with the 'sound structure' of the world's languages. The logical framework is classical first order predicate calculus with a model-theoretic semantics. Existing proposals in temporal logic (van Benthem) and feature logic (Johnson) are combined in the treatment of temporal and hierarchical organization. Phonological 'representations' are linguistic descriptions couched in a formal language. The set of utterance tokens forms the class of intended models. Some links with the sign-based view of grammatical and lexical organization are explored, with a view to ultimately supplementing sign-based linguistic theories (such as Head-Driven Phrase Structure Grammar) with phonological information. A model of feature organization based on phonological argumentation (following Sagey) and phonetic argumentation (following Browman & Goldstein) is proposed as an exemplification of the approach. The model achieves a clear distinction between articulatory and acoustic classificatory properties, lending clarity to the debate about the function of the so-called manner features, and giving content to some recent calls for a non-segmental phonology. Arising from this logical approach is a new computational metaphor for phonology, namely constraint-satisfaction. Linguistic generalizations may be stated in the declarative style, liberated from concerns about their procedural implementation in performance tasks such as generation and recognition. A working constraint-solver which interfaces to Prolog is described.
Publication Status:
Content Language:
Work Type:
Phonological theory
computational metaphors
Nature of Work:
Entry Number: