McCulloch-Pitts strikes back: A biophysical interpretation of cortical neurons as sub-millisecond binary devices

McCulloch and Pitts originally thought that cortical neurons computed using single spikes, with temporal resolution well under a millisecond. But the most popular current simplification of those neurons is as devices which perform computations based on real-valued firing rates, averaged over many spikes and over much longer times. However, single-spike computation has many advantages over pure analog computation: (1) it is more consistent with the observed firing irregularity of cortical cells, (2) it better explains and makes use of the observed correlations in cortical firing, (3) it is more consistent with the biophysics of active cortical dendrites, and (4) it has a far higher information capacity. The only problem is that we have not yet observed it. Or have we?