Our causal judgements seem to be sensitive to normative features of cases. This result has been tested empirically, and turns out to be remarkably stable across a wide range of cases, including ones involving statistical norms, functional norms and scientific norms. I criticize a number of attempts to account for this data, including appeals to pragmatics, appeals to ambiguity, non-normative contrastive accounts, and the idea that causal claims are evaluated with respect to 'modal surrogates'. In their place, I suggest a simple revision to the counterfactual account - x causes y iff x and y both occur and y wouldn't have occurred if a default event had occurred, where a default event is a highest-ranked member of a set of pair-wise incompatible events which includes x. I argue that this account explains the semantic data, captures interesting elements of actual scientific explanatory practice, and has independent theoretical and genealogical motivations. I close by using the account raise doubts about attempts to 'naturalize' norms in causal terms.