In “Rules, rights and promises,” Elizabeth Anscombe proposes to explain and defend Hume's claim that promises, as expressions generating a certain type of commitment to action, are “naturally unintelligible”. However, she also claims that this idea of Hume's has much wider application than he himself appreciated: indeed, she claims that it will follow from her exposition that “no language is in Hume's sense naturally intelligible.... [N]o naturalistic account of a rule, as of a promise, will work: it will follow that words and their relation to their meanings aren't ‘naturally intelligible' either.” In other words, it appears that Anscombe holds that reflection on “natural unintelligibility” will shed light on the nature of the normativity of meaning in general.

 But this is puzzling, for while she makes this remark near the beginning of her paper, she does not return to explain it later; and it is not obvious how to map her later account onto the earlier remark. For her account appears to depend on what is involved in learning the use of modal expressions, in particular what she calls “stopping modals,” expressions such as ‘You can't do that' where the very modality in question—the sense in which the learner “can't do that”, even though, in a more basic sense, the learner often plainly can do it—is precisely what is being taught. –But how can an account of this shape be deployed for the purpose of explicating the normativity of language in general? After all, it appears to depend on the acquisition of the use of a rather sophisticated linguistic device, presumably on the back of an already substantial linguistic competence.

 In the present paper I hope to shed some light on this puzzling feature of Anscombe's presentation, and in so doing, on the nature of the normativity of meaning.