The brain receives a constant stream of information about the world that it must organize and encode into memory. Notably many of the episodes we experience are highly correlated, meaning features are redundant across memories and hence compressible. This proposal aims to test a theory of memory compression in the hippocampus, a structure critical for learning and memory in the mammalian brain. The model suggests that the hippocampal network can leverage the correlations between experiences in order to build compressed neural representations, which support efficient encoding of episodic memories. It provides a natural explanation for the observation of place cells in recordings of rodent hippocampal neurons, as repeated visits to a physical location will be described by highly redundant and compressible sensory and behavioral parameters. Further, the model predicts that the experimentally observed instability in hippocampal coding arises due to ongoing plasticity in the network, as it continually adjusts its internal representations to compress new inputs. Specifically, fluctuations in hippocampal coding should encode information about the recent history of experience, as they reflect the most recent synaptic changes induced by prior episodes. I will use in vivo 2-photon calcium imaging in behaving mice and computational modeling to test whether the hippocampus builds compressed representations of experience, and evaluate specific predictions of the model.
In Aim 1, I will study this question in the context of olfactory coding, to measure whether correlations between sensory stimuli affect the similarity of their representations in the hippocampus. Recent work has demonstrated that hippocampal activity also encodes information about the correlations between episodes in time.
In Aim 2, I will extend the compression model to incorporate temporal correlations, and compare simulations to neural data during a trace conditioning paradigm in order to assess whether the hippocampus adaptively compresses information about the temporal structure of experience.
In Aim 3, I will analyze variability in hippocampal place coding, in order to test the hypothesis that the fluctuations of place cells are history dependent ? that is, they can be used to decode information about episodes in the recent past. These experiments will test explicit predictions of a highly novel theory of hippocampal function. Evidence for this framework could provide a unifying account of many spatial and non-spatial coding properties of hippocampal neurons, and spur the development of new theories of hippocampal and cortical roles in memory encoding.
The hippocampus is critical for the formation and retrieval of episodic memories, but its precise role in this process remains elusive. Impaired hippocampal function in pathologies such as Alzheimer?s disease and schizophrenia have devastating impacts on quality of life. I will explore mechanisms by which the hippocampus may contribute to efficient storage of episodic memories. By advancing our basic knowledge of information processing in this structure, this proposal will facilitate identification of how these processes dysfunction during pathological brain states.
