The literature on the Supreme Court's Chevron decision is vast, perhaps excessive. How much deference courts should pay to agency interpretations of their own statutes has engaged courts, lawyers, and academics in a multi-year debate that shows no signs of abating. Strange as it may seem, this debate has been carried on without answering, without even asking, two simple questions: How should agencies go about the business of statutory interpretation? How do they? This article begins a conversation about those two questions. On the normative side it argues that the differing institutional positions of courts and agencies demand that responsible interpretive behavior diverge as well. These differences lead to an interesting question about the ongoing debate over interpretive deference: If methodology matters, deference may not be possible. But, does methodology matter? In order to answer that question we would have to know how agencies in fact go about their interpretive business. The unsurprising empirical finding is that empiricism in this context is both methodologically problematic and evidentiarily difficult. The article's ambition is to provoke some sustained inquiry into questions that are undeniably important, but that have hardly been mentioned in the literature, much less explored.