The Situated Action approach to AI emphasizes the role of the environment in the generation and control of behavior; see (Norman 1993) for an introduction. Work to date has focused mainly on activity within spatially and temporally localized environments such as kitchens and video games (Agre and Chapman 1987; Agre and Horswill 1992). How useful is this perspective when larger-scale activities are considered? I attempt to answer this question by considering some issues related to navigation in urban environments. identify several constraints on the structure of street grids that make navigation much easier than arbitrary graph search. The ultimate goal is a theory of the relationship between features of an urban environment and the computational complexity of navigation. This work extends the sort of analysis advocated by (Agre and Horswill 1992; Horswill 1993).