The processing cost of reference-set computation: guess patterns in acquisition

Abstract

An idea which got much attention in linguistic theory in the nineties is that the well-formedness of sentences is not always determined by absolute conditions, but it may be determined by a selection of the optimal competitor out of a relevant reference-set. A restricted version of this was assumed (and gradually abandoned) at the early stages of the minimalist program, and simultaneously, it has been the central notion developed in Optimality Theory.

The type of computation required by reference-set economy is inconsistent with what is known about the processing ability of the human parser. If it is indeed vastly at work in the computational system, then processing must be guided by independent algorithms that enable skipping the actual computation, that is, special devices are needed to enable the computational system to be actually used, given the limitations on human working memory.