Why Neurons Make Bad Coincidence Detectors But Good Periodicity Detectors


Abstract submitted to 1995 Neurosciences Meeting

Bartlett W. Mel , Laboratory for Neural Computation , USC
David Croft and Ernst Niebur, California Institute of Technology

It has been suggested that the visual system and other sensory systems use temporal codes. We study how neurons respond to temporal structure in their inputs using simulations of (1) a 164-compartment pyramidal cell with passive or active dendrites, and NMDA or non-NMDA type synaptic input, and (2) a single-compartment integrate-and-fire neuron. In each run, 100 randomly placed excitatory synapses were activated at 100 Hz and the cell's mean output firing rate was recorded. In different runs, and over a range of peak excitatory synaptic conductances, the temporal structures of input spike trains was varied along 2 continuous dimensions: synchronicity (S) and periodicity (P). S=0 meant complete asynchrony and S=1 complete synchrony among trains; P=0 meant Poisson and P=1 periodic trains. Surprisingly, only modest increases in output firing rate were seen as S increased from 0 to 1, and only for the smallest peak synaptic conductances studied. In these cases, input coincidences were required for the cell to reach firing threshold. More surprisingly, the output firing rate was dependent on P over essentially the entire range of peak synaptic conductances studied; 3-fold increases in output firing rate were common as P increased from 0 to 1. This is explained by the fact that increases in P reduced synaptic saturation (i.e. loss of driving force) due to accidental temporal overlap of EPSPs within individual input trains.

Last Edited June 5th, 1995