More than a century ago, G. E. Moore famously attempted to refute ethical naturalism by offering the so-called open question argument (OQA), also charging that all varieties of ethical naturalism commit the naturalistic fallacy. Although there is consensus that OQA and the naturalistic-fallacy charge both fail, OQA is sometimes vindicated, but only as an argument against naturalistic semantic analyses. The naturalistic-fallacy charge, by contrast, usually finds no takers at all. This paper provides new grounds for an OQA thus restricted. But it aims chiefly at vindicating a version of the naturalistic fallacy, „the semantic-naturalist fallacy‟ (SNF), that we think defensible. We first argue that the openness of the question OQA raises against such analyses hinges on self-ascriptive, comparative judgments of content, which may be considered a priori warranted. We then provide independent reasons for the claim that the sort of mistake committed by naturalistic analyses in fact amounts to a pragmatic fallacy of a kind familiar in petitio principii and other forms of viciously circular inference. Of interest here are naturalistic analyses of ethical terms or concepts, not of properties. Our OQA (OQA*) raises an objection to the former. For no such semantic analyses can get off the ground unless moral terms are content-equivalent to purely descriptive terms, which amounts to saying that they must instantiate the same semantic types. Suppose „good‟ is the analysandum and „pleasure maximizing‟ the analysans (whichever purely descriptive term or terms would turn out to be the correct descriptive analysis of the target analysandum) of a certain semantic analysis. The claim that such terms are content-equivalent appears to be open to doubt on a priori grounds. After 1 all, whether one‟s own tokens of „good‟ and „pleasure maximizing‟ have/don‟t have the same content is a first-person, comparative judgment of content. Evaluating the proposed analysis requires, then, a self-ascriptive, comparative judgment of content.1 Judgments of this sort are thought to have a special epistemic status, since they seem grounded in neither evidence nor inference..