It is well known that the hippocampus plays a major role in the formation and retrieval of episodic memories. In rodents, hippocampal cells called place-cells exhibit activity which is selective to the spatial location of the animal in a given environment. As the animal traverses the environment, a particular pattern of place-cell activity will be generated, corresponding to the specific trajectory taken. In this way, cells with spatially adjacent place fields will also fire in temporal proximity, potentially leading to changes in their recurrent synaptic weights via spike-timing-dependent plasticity (STDP) mechanisms [1]. Here we model this process with a computational model in which we consider both pairwise and triplet STDP rules in a recurrent network of hippocampal place cells. We study how the exploration of the animal over time can lead to the formation of low-dimensional attracting manifolds in the neuronal network, each one of which encodes the memory of a specific environment. For example, the traversal of a 1D ring-like environment leads, via STDP, to a so-called ring attractor [2]. Once formed, the presence of the attractor can explain the dynamics of so-called hippocampal replay, via spontaneous re-activation of spatially localized bump states [3]. In fact, the temporal compression of replay can be related to the details of the underlying STDP rule. Interestingly, once the memories of several distinct environments have been encoded, the correlation of replay activity with any given memory can be modulated by the level of external input.

[1] G. Bi and M. Poo, J. Neurosci. 18(24):10464-10472 (1998).
[2] S. Romani and M. Tsodyks, PloS Comp. Biol. 6(8):e1000869 (2010).
[3] S. Romani and M. Tsodyks, Hippocampus 25(1):94-105 (2015).

This work was funded by a grant from the Spanish Ministry of Economics and Competitiveness BFU2012-33413.