Abstract
The parametric character of Government Phonology (GP) and the idea that a child might be setting independent parameters lead to sharp testable predictions for cluster typology as well as for the time course of cluster acquisition. This study explores the theoretical and clinical significance of three binary GP parameters: [+/-Branching onset], [+/-Branching rhyme] and [+/-Magic empty nucleus]. Data from eight English-speaking children with phonological impairments, as well as data from other published work, support the predictions (a) that the acquisition of initial /s/-clusters is related to the acquisition of coda-onset sequences rather than to the acquisition of onset clusters, (b) that there are six developmental grammars in cluster acquisition, and (c) that initial /s/-clusters will never be acquired prior to coda-onset sequences. The three parameters are also shown to predict four stages and three different learning paths in cluster acquisition. In addition, the study offers a natural explanation of the much debated markedness status of initial /s/-clusters: Initial /s/-clusters, being coda -onset sequences with an empty nucleus, are marked, relative to the noninitial coda-onset sequences with an overt nucleus, but there is no markedness relationship between initial /s/-clusters and onset clusters. Implications for phonological intervention are discussed.