Problems encountered in defining or applying a label like "irrealis" arise when a discussion fails to distinguish among three kinds of uses to which different scholars happen to put the label. It may be language specific, i.e., a linguist may use it as a label for a grammatical category in the description of a particular language. Here, we may not be told on what basis the speakers' uses and translations were judged, and therefore analyzed, as referring to events that are not realis or, at least, not real. And, further, we may not be able to judge from the description whether a category called irrealis in the language is best analyzed as simply the unmarked member in opposition to one marked "realis." The resulting inconsistencies prompt some scholars to counsel avoidance of the term. Irrealis may also be a label in our linguistic metalanguage. Here, it represents a vague concept for universal description of languages whose precise definition we are still seeking. Or, irrealis may be attributed to the meaning of a category in a language because the category can be understood to refer to "unreal" events, in a confusion of the unreality of the referents with the category's meaning. The latter, however, may allow such reference on other grounds. Reconciliation of these uses will be attempted using marking theory and pragmatic inference and illustrated with a tense-mood-aspect marker in Newari (Tibeto-Burman) that is a candidate for the label "irrealis," even though it can also refer to actions we would want to judge "real."