Primitive C-command 2. Puzzles of Primitives

Work in syntactic theory has almost universally assumed that phrase structure is characterized in terms of the abstract structural relations of dominance and precedence. In this paper, we suggest instead that hierarchical structure is determined by a primitive relation of c-command. This proposal has the consequence that the range of possible phrase structures is substantially restricted in a linguistically natural way, obviating the need for certain previously stipulated restrictions. Furthermore, by understanding structure in terms of primitive c-command, we arrive at a natural understanding of the structural and derivational diierences between substitution and adjunction, without resort to the machinery of segments and categories. Any descriptively adequate grammatical theory must precisely specify the well-formed representations for the sentences of all possible human languages. Part of this task involves determining what kind of formal object constitutes the basis for grammatical representations. From the outset of generative grammar, it has been assumed almost without exception that phrase structure should be represented in terms of certain type of abstract graph, a tree, whose leaves correspond to lexical items and whose internal connectivity, what is often called constituency , reeects a range of distributional facts. Yet, the question of how tree structures are best characterized has been relatively neglected in the literature on syntactic theory. There seems to be widespread, if only implicit, adoption of the view sketched in Partee et al. (1993), inter alia, in which trees are determined by two primitive relations: dominance and precedence. Yet, so far as we are aware, there have been no empirical arguments adduced in favor of this choice. Recent work in syntactic theory, particularly under the impetus of the minimalist Thanks to the following for stimulating discussion: 2 robert frank and k. vijay-shanker program, has reopened a number of fundamental questions. It therefore seems an appropriate point to reconsider the choice of dominance and precedence as the primitives for syntactic description. In this paper, we argue that the choice of dominance and precedence in fact introduces a number of puzzling dilemmas for syntactic theory, all centering around the over-expressiveness of these relations. These dilemmas motivate our proposal that hierarchical structure be obtained in terms of a diierent primitive structural relation, namely c-command. In section 3, we investigate the formal issues involved in characterizing phrase structure in terms of c-command directly, and in section 4 we compare the expressiveness of this system to one utilizing primitive dominance and …