The predictive structure of navigation determines mouse CA1 representational structure in a multicompartment environment

Abstract

Predictive theories of cognitive mapping propose that these representations encode the predictive relationships among contents as experienced by the navigator. One hallmark of these theories is that representational structure in complex environments can be predicted from the behavioral history of the navigator. Here, we image neural activity in hippocampal CA1 as initially naive mice repeatedly navigate a multicompartment environment to test whether representational structure in this subregion is consistent with these predictions. We find that different mice instantiate different patterns of remapping across identically shaped compartments. Within mouse, compartments with more similar predictive navigational histories on a particular spatiotemporal scale are represented more similarly, accounting for these individual differences. Manipulating navigational options induces reorganization of the CA1 structure which specifically resembles the new predictive navigational structure on this scale. Through computational modeling we show that a combination of predictive encoding and geometrically structured inputs can uniquely account for this pattern of results. Together, these results demonstrate that the structure of CA1 representations in complex environments can be predicted from the behavioral history of the navigator, consistent with predictive theories of cognitive mapping.