No initial empty CV in clusterless languages

From the inception of strict CV in Lowenstamm (1996) the question of how surface clusters are licensed in Government Phonology has been a contentious one. If constituent structure consists of a series of strictly alternating C and V positions, and empty positions are licensed by proper government, then all cluster types should be available to languages that license clusters. The initial empty CV, which though originally proposed to account for certain word-initial clitic alternations in Lowenstamm (1999), emerges as a possible solution to categorizing blatant distributional facts about clusters. Viewed as a morphological (boundary) marker imported from morphology to phonology, its presence accords with languages that show restrictions on cluster type in initial position, while its absence is consistent with languages that show free variation of clusters in initial position. In the midst of this, clusterless languages are assumed to pattern with the former group in requiring the highest restriction on clusters i.e. none at all. This paper argues that there is little ground for such an assumption and that in fact, clusterless languages have no initial empty CV, a fact that is shown to follow from the general ban on proper government in this language type. The paper proceeds as follows; §2 presents the motivations for an initial empty CV; §3 presents arguments for a parametric treatment of proper-government; §4 presents empirical evidence for the absence of an initial empty CV in clusterless languages; and §5 offers some concluding remarks.