On artefactual spike train cross-correlations

Abstract Peaks in spike train cross-correlograms are usually taken as indicative of spike timing synchronization between two neurons, and the peak width is usually taken as indicative of the timescale of interaction. However, a peak merely indicates that the two neurons did not fire independently [2] , and the peak width is a function of other variables as well as of the interaction timescale [3] . Correlogram peaks generated by interactions without any spike synchronization at all can be very similar to those generated by spike synchronization [2] , and can thus be easily misinterpreted as being due to synchronization. Here, simple Poisson models generating such misinterpretable peaks are illustrated and their analytical form is presented. Methods to distinguish different types of covariations [4] are applied to spike train correlograms from a model of lateral geniculate nucleus (LGN) neurons.