ABSTRACT
The ongoing debate about the best human language processing model has inspired a surge of experimental work designed to address issues of cognitive architecture. Much of this work has attempted to address specifically the question of whether the language processor operates in a modular or an interactionist manner. It is recognized that, although these are two idealized extremes, the possibility also exists that reality is not quite so ideal. Nevertheless, it is a useful, and perhaps necessary, exercise for the field, in these pioneering forays, to assume an extreme hypothesis, even if that leads to the conclusion that such a hypothesis must be rejected. Therefore, for introductory purposes at least, we speak in terms of strictly modular versus freely interactive models.
