Margrave is a specification language and analysis tool for access control policies with semantics based in order-sorted logic. The clear logical roots of Margrave's semantics makes policies specified in the Margrave language both machine analyzable and relatively easy for users to reason about. However, the decision conflict resolution declaration and policy set features of Margrave do not have semantics that are as cleanly rooted in order-sorted logic as Margrave policies and queries are. Additionally, the current semantics of decision conflict resolution declarations and of policy sets do not permit users to take full advantage of the multi-decision capabilities of Margrave policies. The purposes of this thesis are (i) to provide a unified extension to the semantics for policies and policy combination, (ii) to cleanly support decision conflict resolution mechanisms in a general way within those semantics and (iii) to provide insight into the properties of policy combination and decision conflict resolution for multi-decision policies in general. These goals are achieved via the realization that policy combinators may be treated as policies operating within environments extended with the results of the policies to be combined, allowing policy combinators to be defined as if they were policies. The ability to treat policy combinators as policies means that users' current understanding of policies can be easily extended to policy combinators. Additionally, the tools that Margrave has for supporting policies can be leveraged as the Margrave language and analysis tool grow to provide fuller support for policy combination and rule conflict resolution declarations. Advisors/Committee Members: Daniel J. Dougherty, Advisor, Joshua D. Guttman, Reader, Craig E. Wills, Department Head.