What’s in a memory? Spatiotemporal dynamics in strongly coupled recurrent neuronal networks. (360G-Wellcome-214316_Z_18_Z)

Current models of memory postulate that impressions of the past are saved in synapses onto so-called engrams, groups of excitatory neurons that co-activate when a memory is recalled. However, many memories, like choreographed movements or spoken words, entail complex activation patterns that necessitate a high level of temporal control. Engram models cannot support such temporal activation patterns because they behave in a binary fashion, with neurons persistently firing at high or low rates. In a recent breakthrough we showed that rich temporal patterns of activity can be achieved when excitatory and inhibitory neurons are strongly coupled and finely tuned. In a departure from engram models, the entire network engages in "spatiotemporal" patterns of activity. Building on these results, we will investigate strongly coupled networks in which memory becomes a network-wide, interactive phenomenon, affected by network structure, sensory input and contextual state. We will build computer simulations of strongly coupled rate networks and sequentially add more layers of complexity, until we can investigate how synaptic plasticity allows the formation of new memories. My work will lead to a new theory of memory, creating biologically relevant models that can be compared to — and informed by — experimental approaches.