Language Processing and Linguistic Explanation
Max Coltheart · 2016
In this paper I will discuss a somewhat novel way of uncovering the properties of the human language comprehension system. Other work in this volume (see the chapters by Clifton, Holmes, Mitchell, and the studies reviewed in the chapter by Frazier) explore this question using a variety of experimental paradigms. This work is surely crucial, but a review of even the papers in this volume reveals that, at the moment, it is somewhat equivocal. In fact different experimental paradigms seem to favour mutually incompatible processing mechanisms. Another problem is that a large class of experimental data seems to be consistent with a wide variety of processing algorithms and even with a wide variety of mutually incompatible parsing architectures. In fact, I would argue that the main task facing the theory of human sentence processing today is to provide criteria to constrain the class of possible natural language comprehension systems and provide a much smaller range of possible parsing systems that can be subjected to experimen tal test. In this paper I will argue that paying attention to the constraints on syntactic representations proposed by linguists is essential to achieving this goal. The search for an algorithm that can retrieve previously encountered material in a way that is embeddable in those parsers can explain an important constraint on the form of grammars; the subjacency constraint on Chomsky (1973).