Having presented the classical definitions for pragmatics in Chapter 2, this chapter is devoted to demonstrating how previous criteria devised in order to distinguish between grammar and pragmatics fail. There are problems with all of the criteria offered in the literature as distinguishing between what are traditionally classified as pragmatic and grammatical phenomena. We will in turn reconsider each of the dichotomies discussed in Chapter 2. First, many of the criteria will be shown to be rather weak (left hemisphere/right hemisphere, linguistic/extralinguistic, explicit/implicit or primary/secondary, grammaticality/acceptability judgments, arbitrary/natural, competence/performance) in that they are too vague or simply don't work to delimit some relatively clear cases of what should be considered pragmatic. Second, even the relatively strong criteria (context dependence, nontruth conditionality, inferentially produced, discourse conditioned) are not consistent in delimiting what is currently taken as uncontroversially pragmatic

The bulk of the arguments we review show that (1) according to any distinction proposed, some phenomena currently classified as pragmatic manifest “grammatical” behavior, and conversely too (2) according to any distinction proposed, some phenomena classified as grammatical, manifest “pragmatic” behavior. If this is the case, the various criteria each predict a different grammar/pragmatics classification for one and the same phenomenon, which at least means that they cannot hold in conjunction with each other. The implicit assumption, of course, was that they should all converge on defining grammar on the one hand and pragmatics on the other. In the light of this, we might consider adopting just one of the criteria.