Language Processing, Linguistics and Constraints

Abstract

Recently, Gibson and Pearlmutter (1998, 2000) and Lewis (2000a, 2000b), among others, have debated whether the initial spark that gets language comprehension started is deterministically created in the same narrow place, always, or, on the contrary, whether that ignition may come from a variety of different places, a broad base. One may schematically conceptualise these two opposite approaches to initial parsing as either an inverted or a non-inverted pyramid respectively. In this paper I will argue for a broad-base, non-inverted pyramid view of the ignition problem. In support of this I will rely primarily on the strength of recent psycholinguistic evidence, as exemplified through the extensively studied [Complex NP+Relative Clause] construction. A key issue will be whether it makes sense to assume that the processing of linguistic reality should be any easier than the (notoriously complex) linguistic reality itself. This point makes sense against the background of a series of well-known formalist accounts of parsing which dominated psycholinguistic research in the 80s and early 90s by appealing deterministically, to merely two or three kinds of syntactic geometry.