Link to bioRxiv paper: http://biorxiv.org/cgi/content/short/2020.06.24.168880v1?rss=1 Authors: Taher, H., Torcini, A., Olmi, S. Abstract: A synaptic theory of Working Memory (WM) has been developed in the last decade to overcome several drawbacks related to the persistent spiking paradigm. Memory items are maintained in the synaptic facilitation and refreshed thanks to Population Bursts (PBs) promoted by synaptic depression. We have developed a neural mass model able to reproduce exactly the dynamics of heterogeneous spiking neural networks with short-term synaptic plasticity. This neural mass model gives access not only to the population firing rate, but also to the mean membrane potential. The latter is correlated to the Local Field Potentials and to electroencephalographic signals, usually measured during WM tasks to characterize the brain dynamics. The model is able to mimic several operations required by WM : memory loading, memory maintenance, competition of memory items and multi-item memory juggling. Memory access and loading is associated to self-sustained oscillations in the {beta}-{gamma} band and to transient evoked responses in the {delta} band, analougously to what observed in the prefrontal cortex of monkeys during object recognition tasks. More items can be stored in the WM by considering neural architectures composed by multiple excitatory populations and a common inhibitory pool. The items can be loaded and maintained at the same time: they are represented as trains of PBs with the same period, but delivered at evenly shifted phases. A memory competition is observable already by loading only two items. Depending on the stimulation features we can observe three different outcomes: maintenance of the first or second loaded item or their juggling in the memory. As reported in many experiments, the {gamma} power increases with the number of loaded items. The power in the -{beta} instead shows a non monotonic behaviour. Finally, we report an expression for the maximal capacity adapted to the present model and architecture. Copy rights belong to original authors. Visit the link for more info