Current analyses of
weak crossover (WCO) focus on properties of the antecedent-variable binding relation (bijection) or of the variables themselves (homogeneity), or on the structural relation between the trace and pronoun involved (c-command). All of these theories fail in a significant case, that in which both bound positions are canonically governed. Here, there is frequently no WCO effect. This paper pursues a government-theoretic account, analyzing the presence or absence of the WCO effect in terms of an ECP which incorporates canonical government. The grammar in focus is that of Palauan, whose basic order is VOS. A principled distinction among SVO, VOS, and other grammars is provided by a
specifier parameter, which sets the specifier position in relation to that of other constituents of the phrase and determines how specifiers are governed. The distribution of WCO effects follows from the interaction of the specifier parameter and canonical government.
The Palauan data is taken from my field notes, 1981–1985; I am indebted to the Palauans in San Diego and Palau who have worked with me. Thanks also to the linguists with whom I have discussed this material: Joseph Aoun, Adriana Belletti, Noam Chomsky, Sandy Chung, Ray Freeze, Greg Carlson, Liliane Haegeman, Roberta Ishihara, S.-Y. Kuroda, Luigi Rizzi, Ian Roberts, Leslie Saxon, and Sten Vikner. Some of this material was presented at the 12th GLOW Colloquium, Utrecht, 1989, at the University of Rochester, and at the University of Massachusetts; I thank members of those audiences for their valuable comments. Finally, the paper has benefitted immensely from the critical and careful comments of Fritz Newmeyer and the NLLT reviewers.