A systematic approach to selecting task relevant neurons.

•Selecting task relevant neurons just from firing rate ignores temporal dynamics.•Modeling spiking based on spike history and stimulus includes temporal information.•Neuron selection based on likelihood ratio test between stimulus and non-stimulus models focuses on contribution of stimulus on neuron activity.•Model based selection performs better than firing rate based selection in simulations and finger movement decoding with other results being equivalent or unverifiable.•Different neurophysiological results can be seen when analyzing model selected neurons as opposed to firing rate selected neurons.