This paper describes howautomated deduction methods for natural language processing can be applied moreefficiently by encodingcontext in a moreelaborate way. Our workis based on formal approaches to context, and weprovide a tableau calculus for contextual reasoning. This is explained by considering an examplefrom the problemarea of presupposition projection.