The hippocampus, best known for its role in spatial navigation, may also generate predictions about what sound will come next in a familiar sequence, and this project will test that idea in mice. This matters because the hippocampus is thought to create a "predictive model" of the world—simulating likely future states based on past experience. While this has been studied in the context of navigation, it is unclear whether the same mechanism operates for sensory inputs like sound. Understanding how the brain predicts sensory sequences could reveal fundamental principles of memory, learning, planning, and imagination. The researchers will record neural activity from the hippocampus of head-fixed mice as they listen to repeated tone sequences. They will look for three types of predictive signals: theta sequences, replay, and omission responses (what happens when an expected tone is skipped). They will also test whether replay and omission responses depend on theta sequences during learning, and use optogenetics to probe how the hippocampus and auditory cortex communicate during these predictions. This is fundamental science with no immediate practical application. However, understanding how biological brains generate predictions could eventually inform the design of next-generation AI models that learn more efficiently and generalise better than current systems.
View original technical description
The hippocampus plays a central role in spatial navigation, forming a cognitive map that encodes a learned representation of our environment. However, recent work suggests that the hippocampus may serve a more general purpose for the brain, creating a predictive, generative model of the world that simulates likely future states based on present and past inputs. During spatial navigation, the hippocampus could use such a mechanism, for example, to determine an optimal path to a new goal or reconstruct a previously taken trajectory. Given that the hippocampus sits at the apex of the brain's sensory processing hierarchy, a similar predictive model capable of mentally constructing possible future states is likely utilized outside of spatial navigation and in the sensory domain, across a wide range of functions including planning, expectation, and even imagination. This research proposal will examine the role of the hippocampus as a predictive generative model of learned auditory stimuli. Performing Neuropixel 2.0 recordings from neuronal ensembles from intermediate/ventral hippocampus of head-fixed mice, we will examine three types of generative responses associated with an auditory stimulus (familiar sequence of repeated tones) - theta sequences, replay, and omission responses. We will next test whether replay and omission responses require theta sequences during learning, and optogenetically interrogate hippocampal-auditory cortical interactions underlying omission responses. Investigating the hippocampal mechanisms responsible for the prediction of sensory stimuli are key to understanding the mechanisms underlying memory, learning, planning and imagination, all with additional potential to advancing the design of next-generation AI models.
Plain English summaries and category classifications on this site are generated by AI and may not perfectly reflect the original research.
Is something wrong? Let us know