Mario DiPoppa

Control of persistent state through spatial correlations in background
activity and applications in a working memory network


Recordings in primates performing delayed response tasks show that persistent neuronal activity in the prefrontal cortex underlies working memory (WM)(e.g., Fuster and Jervey, 1981, Science: 212, 952-955). Three distinct effects are required for these tasks in addition to activity maintenance: that item-related activity should be turned on rapidly (read-in), that such activity is protected from distractors, and that it is rapidly turned off at the response (clear-out). We suggest that the effects of spike-time correlations in neural activity during the WM task may provide a unified mechanism for all three required phenomena.


We implement a discrete working memory model in which neurons are connected by fast and excitatory AMPA-like synapses and receive excitatory random background activity (noise).   In this network we demonstrate that synchronization of spike times can be achieved by increased spatial correlation in the noise and also by transient excitatory input. Synchronization can extinguish the persistent state as was shown by Gutkin et al. (2001 J Comp Neurosci: 11, 121-134) in a spatial working memory model.  We show that it is possible to turn-off the persistent state by projecting the transient "clear" signal to only the activated item-network. We find that the excitatory switch on and switch off signals differ only in temporal and not in spatial statistics of spikes.


Furthermore we propose a novel mechanism preventing the activation of a persistent state by distracting stimuli. We find that increased spatial correlation in background activity prevents stable activation of a persistent state.  Hence distractors can neither be loaded into their item-networks nor spuriously activated when background correlations in these networks are enhanced. This finding can form the basis for a new paradigm about how the global time scale of a working memory can be modulated.