Summary
The brain organizes experiences into memories that can be used to guide future behavior. Hippocampal CA1 population activity may reflect the retrieval of predictive models that contain information about future events, but little is known about how these kinds of memories develop with experience. We trained mice on a series of tone discrimination problems with or without a common statistical structure to observe how memories are formed and updated during learning. Mice that learned structured problems integrated their experiences into a predictive model that contained the solutions to upcoming novel problems. Retrieving the model during learning improved discrimination accuracy and facilitated learning by decreasing the amount of new information that needed to be acquired. Using calcium imaging to track the activity of thousands of CA1 neurons during learning on this task, we observed the emergence of a persistent hippocampal ensemble at the same time that mice formed a predictive model of their environment. This ensemble was reactivated during training and incorporated new neuronal activity patterns from each training problem. Interestingly, the degree to which mice reactivated the ensemble was related to how well their model predicted the content of the current problem, ensuring that the model was only updated with congruent information. In contrast, mice trained on unstructured problems did not form a predictive model or engage a persistent ensemble. These results show how hippocampal activity supports building predictive models by organizing newly learned information according to its congruence with existing memories.
Competing Interest Statement
The authors have declared no competing interest.
