Sideward Movement

Abstract

Assuming the general framework of the Minimalist Program of Chom-sky 1995, this article argues that Move is not a primitive operation of the computational system, but rather the output of the interaction among the independent operations Copy, Merge, Form Chain, and Chain Reduction (deletion of chain links for purposes of linearization). The crucial aspect of this alternative model is that it permits constrained instances of sideward movement, whereby a given constituent "moves" from a syntactic object K to an independent syntactic object L. This version of the copy theory of movement (a) provides an explanation for why (some) traces must be deleted in the phonological component, (b) provides a cyclic analysis for standard instances of noncyclic movement, and (c) accounts for the main properties of parasitic gap and across-the-board extraction constructions.

1 Introduction

A fundamental property of human languages is that elements may be interpreted in positions different from the ones where they are phonetically realized. Within the principles-and-parameters framework (see Chomsky 1981, Chomsky and Lasnik 1993), this "displacement property" is captured by means of a movement operation relating structural positions in a phrase marker. With the recent developments of the principles-and-parameters framework that have culminated in the Minimalist Program (see Chomsky 1993, 1994, 1995), the operation Move is specifically described as follows (see Chomsky 1994:fn. 13, 1995:250): given the syntactic object Σ with constituents K and α, Move targets K, raises α, and merges α with K, forming Σ´; the operation is cyclic if Σ = K and noncyclic otherwise. Σ´ differs from Σ in that K is replaced by L = {γ, {α, K}} or L = {〈γ, γ〉, {α, K}}, depending on whether movement proceeds by substitution or adjunction. Move also forms the chain CH = (α, t), a two-element pair where t (the trace of α) is a copy of α that is deleted in the phonological component in the case of overt movement, but remains available for interpretation at LF (see Chomsky 1993:35). Under this view, the "displacement property" of human languages is thus taken to involve (a) copying, (b) merger, (c) chain formation, and (d) deletion of traces (lower copies) for PF purposes.

As discussed by Chomsky (1993), the interpretation of movement operations in terms of copying accords well with the general conceptual concerns of the Minimalist Program in that it allows binding theory to be stated solely at LF without resorting to noninterface levels, it provides the basis for the interpretation of displaced idiom chunks at LF, and it eliminates reconstruction as an additional operation of the computational system. However, the analysis of Move as an operation as complex as described above has some conceptual problems.

The most obvious one concerns the lack of motivation for deletion of traces (lower copies) in the phonological component.¹ If traces are true copies, why can they not be phonetically realized, behaving like the head of the chain? Another conceptual problem with the computational system as proposed in Chomsky 1994, 1995, is that Merge is taken to be an operation in its own right in certain cases, and a suboperation of Move in other cases. In an optimal system, we should in principle expect Merge to have the same theoretical status in every computation (see Gärtner 1997). Finally, as is emphasized by Brody (1995), who pursues a representational version of the Minimalist Program, if chain formation and Move express the same type of relation, a theory that contains both notions is redundant.

The overall complexity of the operation Move can be seen as a historical residue of the description of generalized transformations as binary or singulary operations: concatenative binary transformations such as Merge target two disconnected syntactic objects, whereas singulary transformations such as Move target two constituents of a single syntactic object (see Chomsky 1975, 1993, Kitahara 1995). Assuming the general framework outlined in Chomsky 1995, I propose here that Move should be understood not as a primitive singulary operation of the computational system, but as merely the description of the interaction of the independent operations Copy, Merge, Form Chain, and Chain Reduction (deletion of copies for purposes of linearization). Under...