Optimality Theory and phonological acquisition

The past ten years have shown an ever-increasing revival of interest in phonological acquisition. The introduction of Optimality Theory (OT; Prince & Smolensky 1993) has played a major role in this revival. Both “markedness” and “constraints on outputs” have often figured in accounts of phonological development, and these concepts are central in OT. The most fundamental change is probably that OT has reinstated continuity: developmental grammars and final adult grammars have the same representational units and organizational principles. In the work by Jakobson (1941) and Stampe (1969) the development from an initial grammar to a final grammar had been pictured as a continuous process. For Jakobson development consisted of a gradual unfolding of a system of contrasts. For Stampe the initial grammar consisted of a set of unordered innate rules, while the final state consisted of an ordered subset of these rules. Development, then, consisted of suppressing the rules inappropriate to the adult target grammar and ordering the target-appropriate rules. In subsequent work, most notably by Smith (1973) and Kiparsky & Menn (1977) the child’s developmental grammar(s) no longer had any principled relation to the target adult grammar.1 In both works development mainly consisted of getting rid of a set of child-specific rules or strategies that filtered the target adult forms. Once this filter had been removed the target adult form and the child’s production were identical. However, the ability to produce a target-appropriate form cannot be equaled to having acquired the adult grammar. Development in this view is thus noncontinuous: the initial grammar, the “filter”, does not develop into a mature grammar, but disappears without leaving a trace. In Parameter theory (Dresher & Kaye 1990, Fikkert 1994) there is continuity between developmental grammars and mature grammars: in the initial state of the grammar parameters have a default setting, and development consists of setting the appropriate parameters to their marked values. However, the successes of Parameter theory have largely remained domain-specific, in the sense that they have been confined to syllable structure and stress. In OT the basic idea is that constraints are innate and universal and come with an initial ranking where all markedness constraints (segmental, syllabic, and prosodic well-formedness constraints) outrank all faithfulness constraints (constraints requiring similarity between underlying and surface representations). The learner needs to acquire the language-specific ranking of these constraints. By subsequent rerankings, the initial