One of the oldest, and most tenacious, dogmas about metaphor is the thesis of grammatical deviance (GD): the thesis that a string of expressions will be recognized and interpreted as a metaphor only if, were it interpreted literally, it would be grammatically deviant, semantically anomalous, explicitly or implicitly self-contradictory, conceptually absurd, nonsensical, a category mistake, a sortal violation, or, on the weakest version, simply false. (For the full range of formulations, see [1], [4], [20], [22], [23], [24], [25], [26], [32], [36]. Discussion here will be limited to grammatical deviance, the strongest version of the general thesis, although much of what we say equally applies to the weaker cases.) The historical origins of the thesis probably lie in the prejudice, once especially prevalent among empiricists like Locke and Hobbes, that metaphorical use is abuse of language, at best frivolous and eliminable, and at worst deceptive and dangerous. (See [17] for these and other references.) Similar attitudes were held by positivistically inclined philosophers during the earlier part of this century and, in their estimate, the deviance of metaphor could be seen in its apparent indifference to rules and its lack of function within an ideal-read: scientific-language. Hence, metaphor was to be avoided when reforming natural langauge to meet the aims of clear and unambiguous scientific discourse; instead, it was reserved for, or relegated to, poetry and other types of "non-cognitive" language. (See, e.g., [35].) Although this prejudice against metaphor is no longer shared today-at least openly-the GD-thesis has survived and prospered. Indeed, with the term 'deviant' purified of derogatory connotations in the theoretical vocabulary of generative grammar, nowadays to call any expression, or a metaphor in particular, 'grammatically deviant' is to make an explanatory claim about it.1 And in recent explanations of metaphor the GD-thesis has prominently figured, not in one but two sorts of roles.