A parsing explanation of linear movement asymmetry in natural language

Natural language sentences are made up of words. Groups of words which can act together as a unit are called constituents. Movement is a syntactic process which displaces a constituent from its canonical position. The moved constituent is called a filler. The position from which a filler is displaced is termed a gap. A gap is a particular type of empty category.;Interestingly, a filler always appears to the left of its associated gap. My thesis is that this linearity constraint on filler-gap dependencies is not a syntactic constraint but rather a product of the parsing mechanism.;I develop a parsing algorithm which can process filler-gap dependencies only if the filler is identified before the gap. What makes this possible in general is that I treat the problem of licensing and identification of empty categories during on-line processing as a case of ambiguity resolution. I develop a fine-grained typology of nominals (both overt and empty) which supports the use of underspecified representations. In turn, this permits the parser to resolve empty category ambiguities incrementally as parsing proceeds.;The main results of the dissertation are as follows. The parsing algorithm presented herein offers an explanation for the curious fact that overt movement is leftward. It also explains why, in a language like Italian, postverbal DPs are preferentially interpreted as objects rather than subjects when they appear after optionally transitive verbs. There are also several predictions which arise from this work. The operation of the algorithm predicts that a prenominal relative clause is possible only in a language which permits personal pronouns to act as relative pronouns. There is also predicted to be a filled-gap effect with structures (such as left dislocation in English) for which there is an ambiguity as to whether movement has taken place. Finally, this research supports the view that the human language processing mechanism constructs a single underspecified representation, and resolves ambiguities incrementally as the parse progresses.