ABSTRACT
Norvig (1987, chapter 7) started his conclusions by remarking that, “in a sense, FAUSTUS was an experiment in self-deprivation” in that it relied on only six basic inference classes, contrary to systems such as Dyer’s (1983) BORIS where “one occasionally gets the suspicion that the system designer can just add one more rule to account for each new difficulty as it arises, as long as he or she is careful about interactions with previous rules” (Norvig, 1987). Yet, I have argued in chapter 2, that all existing symbolic approaches to text understanding are postulating a priori macrostructures that constitute rules of comprehension entrenched in these models. In other words, a debate between inference-chaining and schema-matching hides, from my viewpoint, the fact that neither of these approaches offers a grounded model for linguistic comprehension. Furthermore, in surveying existing work in psycholinguistics throughout Part III of this book, I have highlighted the absence of consensus on the nature and modus operandi of these hypothesized macrostructures. As Graesser and Kreuz (1993) remarked, we still lack a complete theory of text understanding.
