A statistical analyzer for optic nerve messages.

A statistical mathematical model of the discharge in a single optic nerve fiber is proposed, based on a discharge with intervals between impulses distributed independently according to a gamma distribution ("gamma discharge"). A light stimulus distorts the time axis of this discharge according to a "frequency function" which is characteristic of the stimulus. A linear filter is described which calculates the likelihood of a certain stimulus when the nerve fiber message is fed into it. This filter forms the basis of theoretical nerve message analyzers for three visual experiments: (a) The detection of the occurrence of a flash of light of known intensity and time of occurrence, (b) the detection of the time of occurrence of a flash of known intensity, and (c) The estimation of the intensity of a flash occurring at a known time. Possible neural mechanisms in the brain for analyzing optic nerve messages, based on the above mathematical models, are suggested. Changes of excitability or discharge frequency correspond to the output of the likelihood filter. Any such mechanism must be sufficiently plastic to have a response matched to each expected stimuus for most efficient vision near threshold.