Children’s Unnatural Phonology

Most approaches to generative phonology grant a privileged status to “natural” phonological rules, those that have a clear phonetic motivation. In Optimality Theory, for example, children are typically assumed to begin language learning with markedness constraints ranked above faithfulness constraints (see Smolensky 1996). The implication — often not made explicit — is that phonetically arbitrary processes require some other apparatus in the grammar, perhaps the creation of a language-specific constraint, and consequently should pose a more difficult learning task for the child. I argue that this prediction is incorrect, based on the diversity of child phonology. The common occurrence of natural patterns can be explained without reference to grammatical constraints, while the unnatural patterns that deviate from the adult model require abandonment of such constraints (cf. Hale and Reiss 1998, 2000, Blevins 2003 for similar arguments).