An extension of the $\lambda$-calculus is proposed, to study history-based access control. It allows for security policies with a possibly nested, local scope. We define a type and effect system that, given a program, extracts a history expression, i.e.\ a correct approximation to the set of histories obtainable at run-time. Validity of history expressions is non-regular, because the scope of policies can be nested. Nevertheless, a transformation of history expressions is presented, that makes verification possible through standard model checking techniques. A program will never fail at run-time if its history expression, extracted at compile-time, is valid.