Multiple optimal outputs in acquisition

The presence of intra-speaker variation in early acquisition has long presented an intriguing problem for researchers of phonological development. It would also appear to present difficulties for Optimality Theory, a theory which by its very nature states that one and only one surface output form is 'optimal'. This raises fundamental questions about the nature of children's emerging grammars with respect to the issue of 'Discontinuity' versus Strong or Weak 'Continuity'. That is, are children's grammars qualitatively different from those of adults, or are they basically the same? It also poses interesting questions for Optimality Theory regarding its ability to generalize to problems of language learning: Is Optimality Theory useful in exploring issues of early language acquisition? If so, what does it have to say about variation -or Multiple Optimal Outputs for a given target? Must some of the original assumptions about Optimality-theoretic grammars be challenged or changed to account for variation in child language? The purpose of this paper is to explore these issues more fully. First, the paper demonstrates how Optimality Theory is useful in identifying three different types of variation in acquisition, only one of which is a true case of Multiple Optimal Outputs. It then illustrates how this type of variation can be handled from an Optimality-theoretic perspective if 'partial constraint rankings', or stratified domination hierarchies, are allowed (Demuth 1995, 1997, Tesar & Smolensky 1996). The paper concludes with a discussion of what Optimality Theory can contribute to the understanding of intra-speaker variation in other domains, including second language learning, language in disordered populations, 'optional' structures in adult language, and processes of historical change.