At the turn of the last century, Constantin Stanislavski developed a new system of acting, replacing the mannered gestures and forced emotion then popular with a more natural style. The core of his system lay in having actors perform a process of scene analysis, in which an actor would flesh out the circumstances of the play so that the character's motivations and actions would follow logically. This paper is an attempt to ground Stanislavski's method of scene analysis in a formal theory of action. We discuss the relations between Stanislavskian and formal AI theories of action and planning, give a formal definition of the end product of a scene analysis, and characterize the conditions under which a scene analysis is coherent.