 The predictive map hypothesis suggests that the hippocampus is responsible for encoding the expected state occupancy of locations in the environment. This hypothesis is supported by behavioral and electrophysiological evidence, and can be explained by a model based on error-driven temporal difference learning. This model utilizes spiking neurons modulated by theta-band oscillations, diffuse and overlap in place cell-like state representations, and experimentally matched parameters. It also explains the topographical ordering of place field sizes along the dorsal ventral axis, which prevents the detrimental mixing of larger place fields, which encode longer time scales, with more fine-grained predictions of spatial location. This article was authored by Tom M. George, William DeCothe, Kimberley L. Stachenfeld, and others.