Descriptive notions vs. grammatical categories: unrealized states of affairs and `irrealis'

Over the past decades, there has been extensive debate about whether or not `irrealis', in the sense of a general notion of unrealized state of affairs, plays a role in the grammatical organization of the world's languages. This paper discusses various facts pertaining to the encoding of unrealized states of affairs cross-linguistically, including facts about person marking, switch-reference, negation, and the encoding of states of affairs that failed to take place in the past or may possibly take place in the future. These facts show that there are three issues that have generally failed to be kept distinct in the debate on `irrealis': the fact that the notion of unrealized state of affairs can be used to describe particular grammatical patterns, the fact that this notion plays a role in the mechanisms that bring about these patterns, and the fact that this notion corresponds to a grammatical category of particular languages. This problem is representative of a widespread approach in linguistic analysis, one where the description of observed grammatical patterns and the formulation of explanatory hypotheses about these patterns are combined together. These two levels are independent of each other, and mixing the two may lead linguists to address the wrong theoretical issues