Abstract
I will present an extension of the sheaf theoretic framework for contextuality (Abramsky and Brandenburger, 2011) by introducing a causal structure which takes the form of a partial order on the set of measurements. Once care is taken to ensure that sheaf theoretic restrictions respect the partial order, this leads naturally to a general notion of causal empirical models as families of compatible probability distributions taking values in a certain presheaf. Both contextuality and violations of macroscopic realism, in the sense of the Leggett-Garg theorem, are characterized in a unified way by the absence of global sections. I will illustrate this with the example of the double slit experiment which exhibits a logical violation of macrorealism.