Optimality Theory in Phonology

GENERATIVE PHONOLOGY [q.v.] aims to construct a predictive theory of natural language sound systems, rooted in a finely-detailed account of the principles defining linguistic representations and the possible relations between them. Within these broad goals, OPTIMALITY THEORY (OT) develops in the context of specific empirical theses about the way phonological systems are organized. We identify three of them here. (i) The role of output targets. Phonological representations may change from their lexical form to their surface or output forms: German /rad/, ‘wheel – lexical form’ is pronounced [rat]; Maori /inum/ ‘drink lexical form’ is pronounced [i.nu], English /lik+d/ ‘leak + past’ is pronounced [likt], and so on. (Notation: we write ‘.’ to demarcate syllables.) An important finding, announced in Kisseberth 1970 and since replicated in many areas of phonology and morphology, is that such changes are often conditioned by properties that hold of the surface or output forms in the language. In Maori, for example, no syllable is ever closed by a consonant: many lexical forms can be successfully parsed into open syllables without modification (e.g. /patu/ ‘kill’), but others, like /inum/, must accommodate to the surface regularity by doing violence to lexical specification. (ii) Intrinsic relation between change and conditions of change. Regular changes in representations take place under specifiable conditions; more subtly, the nature of the change is closely related to the nature of the conditions that provoke it. In Maori, for example, we see a change involving deletion under conditions that refer to a syllabic environment – but not just any syllabic environment is capable of forcing such deletion and not just any change may take place in that environment. The actual process must therefore have access not just to syllable structure per se, as a descriptive convenience, but to the properties of syllable structure which determine that consonant deletion, rather than (say) consonant insertion is the linguistically appropriate response. Theories of rule form in early generative phonology established little or no connection between the “structural description” or environment of a rule and the “structural change” it imposed. (iii) Universality and Difference. Phonological grammars are strongly restricted both in absolute terms – some structures and patterns must always be present, others can never arise – and relatively, in that the presence of certain patterns implies the presence of others, or in that certain effects, if present, can only occur under certain conditions, cross-linguistically limited. Yet against this background of universality, languages may still differ considerably and arbitrarily – for example, it is not the case that every language disallows closed syllables, and it is not even the case that every language disallowing closed syllables will permit deletion to eliminate problematic structures that might give rise to them.