Many prominent cognitive neuroscientists argue that the present categories of cognitive science are badly misguided, and that neuroscientific work will radically revise our 'cognitive ontology.' Implicit in these debates is the assumption that cognitive science must aim at unification, in part because cognitive models of different domains must ultimately interact with one another. That can be denied, and has been denied by a number of so-called contextualists, including (apparently) me. Having suggested this a while back, it still sounds right to me.  In this talk I double down. I present an account ("The Rosencrantz and Guildenstern Model")  on which different cognitive domains can interact without falling under a unified model, and suggest some computational reasons why we might expect such a result.