Cognitive maps confer animals with flexible intelligence by representing spatial, temporal, and abstract relationships that can be used to shape thought, planning, and behavior. Cognitive maps have been observed in the hippocampus, but their algorithmic form and the processes by which they are learned remain obscure. Here, we employed large-scale, longitudinal two-photon calcium imaging to record activity from thousands of neurons in the CA1 region of the hippocampus while mice learned to efficiently collect rewards from two subtly different versions of linear tracks in virtual reality. The results provide a detailed view of the formation of a cognitive map in the hippocampus. Throughout learning, both the animal behavior and hippocampal neural activity progressed through multiple intermediate stages, gradually revealing improved task representation that mirrored improved behavioral efficiency. The learning process led to progressive decorrelations in initially similar hippocampal neural activity within and across tracks, ultimately resulting in orthogonalized representations resembling a state machine capturing the inherent struture of the task. We show that a Hidden Markov Model (HMM) and a biologically plausible recurrent neural network trained using Hebbian learning can both capture core aspects of the learning dynamics and the orthogonalized representational structure in neural activity. In contrast, we show that gradient-based learning of sequence models such as Long Short-Term Memory networks (LSTMs) and Transformers do not naturally produce such orthogonalized representations. We further demonstrate that mice exhibited adaptive behavior in novel task settings, with neural activity reflecting flexible deployment of the state machine. These findings shed light on the mathematical form of cognitive maps, the learning rules that sculpt them, and the algorithms that promote adaptive behavior in animals. The work thus charts a course toward a deeper understanding of biological intelligence and offers insights toward developing more robust learning algorithms in artificial intelligence.

When experts are immersed in a task, a natural assumption is that their brains prioritize task-related activity. Accordingly, most efforts to understand neural activity during well-learned tasks focus on cognitive computations and task-related movements. Surprisingly, we observed that during decision-making, the cortex-wide activity of multiple cell types is dominated by movements, especially “uninstructed movements”, that are spontaneously expressed. These observations argue that animals execute expert decisions while performing richly varied, uninstructed movements that profoundly shape neural activity. To understand the relationship between these movements and decision-making, we examined the movements more closely. We tested whether the magnitude or the timing of the movements was correlated with decision-making performance. To do this, we partitioned movements into two groups: task-aligned movements that were well predicted by task events (such as the onset of the sensory stimulus or choice) and task independent movement (TIM) that occurred independently of task events. TIM had a reliable, inverse correlation with performance in head-restrained mice and freely moving rats. This hinted that the timing of spontaneous movements could indicate periods of disengagement. To confirm this, we compared TIM to the latent behavioral states recovered by a hidden Markov model with Bernoulli generalized linear model observations (GLM-HMM) and found these, again, to be inversely correlated. Finally, we examined the impact of these behavioral states on neural activity. Surprisingly, we found that the same movement impacts neural activity more strongly when animals are disengaged. An intriguing possibility is that these larger movement signals disrupt cognitive computations, leading to poor decision-making performance. Taken together, these observations argue that movements and cognitionare closely intertwined, even during expert decision-making.