Figure 2.

Schematic description of the GLM model. Each ear's input is given by a spectrotemporal representation of the sound signal. Each ear's STRF is applied to the stimulus to give a weighted sum over frequency and time. The spike history kernel is a temporal filter applied to the spiking history to take into account refractory processes. The STRF-filtered stimuli are combined with the spike history and passed through a nonlinearity to produce an instantaneous firing rate. Spikes generated probabilistically with the instantaneous firing rate give rise to a predicted PSTH (red), which is compared to the neuron's response observed in vivo (black).