Non-additivity enables synchrony propagation in recurrent networks

So far, models of neural circuits typically assume that inputs to a neuron add linearly and are only nonlinearly processed thereafter. Recent experiments show that a non-additive processing acts at single neurons if excitatory inputs appear synchronously and in sufficient proximity on the dendrite. The current study now provides a systematic view on how such non-additivity impacts the collective dynamics of networks.

One main finding is that even purely random networks may exhibit robust propagation of synchronous activity (synfire chain activity) even without superimposed feed-forward structures (so far known as synfire anatomy).

Raoul-Martin Memmesheimer and Marc Timme (PLoS Comput. Biol. 8:e1002384, 2012)