Structure Preservingness, internal Merge,

The generative revolution in linguistics, started in the 1950s, culminated around 1980, when it was demonstrated that “rules of grammar” were not construction-specific, as had been believed since Antiquity, but following from the interactions of various constructionindependent principles. Perhaps most important among those principles were the locality principles, with such well-known highlights as Chomsky’s 1964 A-over-A Principle, Ross’s Constraints on Variables (1967) and Chomsky’s 1973 ideas about Subjacency and successive cyclicity. At least as important, be it relatively underexposed, was the idea of structurepreservingness. The idea of structure-preservingness has two related but different forms, which were developed in roughly the same period. The form of structure-preservingness that comes always to mind immediately, is the version proposed by Joseph Emonds in his classical dissertation of 1970. According to this form of structure-preservingness, transformational rules do not add anything new beyond what is already given in deep structure (via the phrase structure rules). The other form of structure-preservingness is trace theory. Trace theory further developed a trend existing since pre-generative structuralism, i.e., the enrichment of linguistic representations with abstract elements like zero elements and scope markers. Katz and Postal (1964) made important contributions to this effect but the development leading more directly to trace theory was the decision in Aspects (Chomsky 1965) to separate the recursive phase structure component from the lexicon. This made it possible to conceive of phrase structure as something entirely or partially unlexicalized. Trace theory added another element of structure-preservingness because it meant that deep structure information was preserved instead of being lost as the result of movement transformations. It has been my conclusion ever since that the two aspects of structurepreservingness, Emonds’s version and trace theory, have completely undermined the original idea of transformational grammar. If, with movement rules, nothing structural is gained (Emonds’s version) or nothing is lost (trace theory), it is hard to see any useful function for movement transformations at all. This judgment extends to the successors of movement transformations, such as “move alpha”, Move, and more recently, internal Merge. All these residues of transformationalism are equally superfluous, blocking the formulation of syntax in a strictly local, variable-free manner and, most important of all, obscuring the perspective on the underlying unity of grammar. By this underlying unity I mean the following. All dependencies of core grammar are local, but not all these local connections can be stated in terms of movement. In spite of occasional attempts in that direction, the antecedent-reflexive anaphor relation, for instance, cannot naturally be reformulated in terms of movement rules. For other local dependencies,