An integrated model of the syntax and phonology of Celtic mutation

Celtic mutations are, descriptively, a set of phonological rules operating on word-initial consonants triggered by the word's morpho-syntactic environment. Mutations occur in all modern Celtic languages--as well as in other languages including Mende (West Africa), Fula (Sub-Saharan Africa), Nivkh (Siberia) and Southern Paiute (California). The phenomenon of mutations has raised major theoretical questions in both syntax and phonology because it is a phonological rule triggered by syntactic information.;The task of this dissertation is to construct a parameterized cross-linguistic model for mutation which can take a mutation derivation from underlying syntax to surface phonology. There have been numerous studies on the phonology of mutation (Sproat 1987, Lieber 1987, Kelly 1898, Ni Chiosain 1991, Paradis 1992, Gahl 1994, Grijzenhout 1995) and other studies on the "syntax" of mutation (Massam 1983, Tallerman 1990, Duffield 1991, 1996), but few which have created a model for both the syntax and phonology of mutation. This work is also unique because it discusses data from both Irish and Welsh as well as Breton, Scottish Gaelic and Cornish.;This dissertation constructs a derivational model of mutation within the framework of Distributed Morphology (Halle and Marantz 1993). Within this model, I propose that mutation occurs at the very first stage of phonology, "phonological readjustment", not in syntax or morphology per se. The trigger of these rules are morphemes bearing a morphological feature or "diacritic" which tells the phonology to perform the mutation (Hamp 1951). The stage of phonological readjustment is the same stage where word-internal, morphologically conditioned rules such as English past-tense umlaut (Halle and Marantz 1993). Mutation is only unique in that these same types of processes are occurring across a word-boundary. At this stage of phonology, the grammar can access morpho-syntactic features, surface phrasal structure, but not other sorts of syntactic information. In addition, this stage of phonology requires that all output segments conform to the phonemic inventory of the language. Hence, many repair operations are invoked to fix potentially ungrammatical output segments.;I further argue that the so-called "Lenition" mutation in the Celtic languages is actually a set of ordered rules which are triggered by morphemes specified with the Lenition diacritic. This contradicts the common assumption that Lenition is one phonological process. The fact that one mutation diacritic triggers more than one sound rule makes Lenition appear to be a unitary phonological process.