Information transmission across a dynamic synapse

Mark Goldman

Department of Brain and Cognitive Sciences
Massachusetts Institute of Technology

Many hippocampal and neocortical synapses appear to be surprisingly unreliable, with only a small fraction of presynaptic action potentials successfully triggering vesicle release. Previous work (Goldman et al., 1999) has shown that synaptic depression may remove the correlations and associated redundancy present in realistic spike trains, and has suggested that this removal of correlations may increase the information carried per vesicle release. Here we calculate the information transmitted across a depressing synapse. Information calculations on time scales typical of spike train correlations are numerically intractable. We overcome this difficulty by performing an analytic calculation. Our results show that synaptic dynamics, regardless of their form, generally increase the information transmitted across a stochastic synapse. This increase in information may occur even when the dynamics introduce correlations, and hence redundancy, into the train of synaptic transmissions. However, our results suggest that the increase in information transmission is greatest when the form of the dynamics is matched to that of the spike train correlations.