Penn Arts & Sciences Logo


Friday, September 20, 2013 - 12:00pm

IRCS Conference Room

William Warren
Cognitive, Linguistic, and Psychological Sciences
Brown University

What is a cognitive map? and how could you build one?

It is often assumed that “cognitive maps in rats and men” have a Euclidean geometric structure. Such spatial knowledge could be built up from path integration and would support accurate shortcuts. However, apparently Euclidean behavior may also result from weaker knowledge together with adaptive navigation strategies. I triangulate on this question in a series of experiments using ambulatory virtual reality to create non-Euclidean environments. Humans have poor path integration, make unreliable shortcuts, rely heavily on visual landmarks, ignore radical violations of Euclidean structure, and do not build a globally consistent cognitive map -- yet they can still take approximate shortcuts if necessary. I will argue that spatial knowledge can be characterized as a labeled graph, a primarily topological structure augmented by local metric information, which can be used to generate rough shortcuts on the fly.