The status of covert movement in Universal Grammar has been a perennial source of trouble in the study of language. What kinds of structures does it derive? To what extent is it similar to overt movement? What is its place in the overall architecture of the grammar?
In this thesis I present several case studies bearing on these questions, providing new evidence for the existence of covert phrase movement. These studies contribute to the growing body of evidence that grammatical conditions hold only at interface levels [Chomsky 1993]. Further, I attempt to show that, taken together, the investigations reported here lead to a model of grammar in which the interface representations are computed cyclically, by successive applications of the basic grammatical operations merge, move and spellout on each phase of derivation.
The first studies demonstrate that covert movement licenses parasitic gaps and feeds Condition A, reversing longstanding assumptions. The apparent counterevidence that has obscured these properties of covert movement, I argue, results from a general constraint on movement (the Tucking-in condition [Richards 1997]) that prevents the formation of the required configurations in the classic experimental paradigms. In addition, the study on parasitic gaps provides evidence for the Y-model’s sequencing of overt before covert operations. However, an investigation of adjunct extraposition from NP (a report of joint work with D. Fox) yields evidence for the opposite conclusion: that a covert operation (QR) can be followed by an overt one (late adjunction to the raised NP).
Finally, I show that these conflicting results are resolved by a theory of successive-cyclic computation of structure in which spellout applies repeatedly throughout a derivation. I argue that the correct characterization of the cyclic model captures Y-model effects such as the failure of covert movement (typically) to license PGs, while allowing ‘anti-Y-model effects’ typified by extraposition. I propose a condition that limits countercyclic adjunction to the linear edge of already computed structures. This condition in turn predicts an intricate pattern of further generalizations about extraposition. The resulting theory thus unifies the overt and covert cycles in a manner consistent with the evidence for covert phrase movement.