On Internal Merge

Abstract

Abstract The rule MOVE, used in various forms in generative grammars to capture displacement or discontinuous constituency, has recently been talked of as an “internal” version of MERGE, the operation of simple node- or set-formation. Internal merge “reconstructs” the displaced element in its original argument-structural position at the level of logical form via a “copy”, to which it has been identical throughout the derivation. Reducing MOVE to MERGE seems to be on the side of simplifying the theory of grammar, potentially eliminating the need for constraints on movement in order to limit overgeneration. The paper addresses the question of how internal merge should be defined in formal terms. An account of discontinuity is proposed in which copies originate in the lexicon, as seems to be required by a strict interpretations of the Inclusiveness Condition of Chomsky (1995b), where they can be thought of as binders and variables in lexical logical form (lf). Merger is defined via a small number of type-dependent combinatory rules, which apply to strictly string-adjacent categories to monotonically project from the lexical array varieties of discontinuous dependencies that have been described in terms of various forms of movement, including “A”, “Ā”, “remnant”, “head”, “parallel”, “sideward”, “covert”, “roll-up”, and “late merge”, without any attendant “constraints on movement” other than those projected from lexical types. The analysis extends to a plethora of other discontinuous operations that have been proposed in addition to or instead of MOVE, including AGREE, LABEL, TRANSFER, and DELETE, all of which are replaced by synchronous monotonic lf and pf merger of contiguous categories. The result is to eliminate structure-dependence and action-at-a-distance of all kinds from syntactic rules.