Postoccupancy evaluations of large buildings often reveal significant wayfinding problems caused by poor floor-plan layout. Predicting wayfinding problems early in the design process could avoid costly remodelling and make better buildings. However, we lack formal, predictive models of human wayfinding behaviour. Computational models of wayfinding in buildings have addressed constructing a topological and geometric representations of the plan layout incrementally during exploration. The authors propose to combine this with a schema model of building memory. It is argued that people orient themselves and wayfind in new buildings using schemas, or generic expectations about building layout. In this paper the authors give their preliminary thoughts toward developing a computational model of wayfinding based on this approach