A final brief in the case against agrammatism: the role of theory in the selection of data.

In an analysis of the assumptions behind the use of patient groupings in neurolinguistic and cognitive neuropsychological research (Badecker & Caramazza, 1985), we have argued that clinical categories in general, and agrammatism in particular, allow no valid means for reasoning from patterns of impaired performance to an understanding of the processing mechanisms underlying normal linguistic performance. On the basis of these arguments, we proposed that categories of this sort serve no useful function and ought to be dispensed with in this research. In his response to our paper, Caplan (1986) claims that there are indeed reasons for maintaining agrammatism as an object of study, and he suggests that "one can approach agrammatism in a way that renders it an appropriate category for aphasia, and [this conclusion] leads to a number of interesting research topics regarding this category" (p. 263). We maintain that this can only be true if the goals of Caplan's enterprise are radically different from what we take to be those of modern cognitive neuropsychology. We hope to make this clear in the space of a few pages.