Research ReportVariability of spike firing during theta-coupled replay of memories in a simulated attractor network

Simulation work has recently shown that attractor networks can reproduce Poisson-like variability of single cell spiking, with coefficient of variation (Cv2) around unity, consistent with cortical data. However, the use of local variability (Lv) measures has revealed area- and layer-specific deviations from Poisson-like firing. In order to test these findings in silico we used a biophysically detailed attractor network model. We show that Lv well above 1, specifically found in superficial cortical layers and prefrontal areas, can indeed be reproduced in such networks and is consistent with periodic replay rather than persistent firing. The memory replay at the theta time scale provides a framework for a multi-item memory storage in the model.This article is part of a Special Issue entitled Neural Coding.

► We analysed spike train variability (Lv) in a cortical attractor memory model. ► The analysis revealed two Lv clusters consistent with firing patterns found in vivo. ► Likely-random firing pattern was associated with persistent attractor activity. ► Clumpy-bursty firing was consistent with periodic replay of memories at theta scale. ► The study supports memory reactivation in layer 2/3 of higher-order cortical areas.