Reliabiity of Spike Encoding in Cortical Neurons

There have been many recent articles on the coding properties of nerve cells, with emphasis on the precision and reliability of the processes involved 1−7. In our investigations of stochastic effects in nonlinear biological neuron models , we have examined especially some results obtained on the firing properties of certain neocortical cells 3 in which various currents were injected repetitively and the accompanying sequences of spike times were recorded. An interesting and seemingly paradoxical result was obtained: when a constant current was injected into the cell, a spike train was obtained which became more variable with time, whereas when a variable current waveform was repeatedly injected, practically identical spike trains were obtained in each trial. It thus seemed that variability somehow led to greater reliability because a constant input current (here called case A) led to a noisy output whereas a noisy input (here called case B) led to a fixed output pattern.