A Biological Interval Analyser

Abstract

A FREQUENT requirement in the analysis of biological data is the classification of time-intervals. Thus one might wish to know the most probable interval between heart-beats, or the most probable reaction time of forefinger movements in response to sound. The usual way of acquiring this sort of information is by manual measurements of relatively long biological records. This method, however, can become extremely time-consuming when the number of biological signals to be classified is large. An example may serve to illustrate our point. The neurons of the unanæsthetized, unstimulated cat's brain are in a state of continual ‘spontaneous’ activity. Their mean frequency of discharge per minute is constant (some 5–20/sec.) when observed over long periods of time, but it appears impossible to predict with certainty the exact time at which the next action potential in a series will occur. Cells in the resting visual cortex are also engaged in continual activity, but will respond to a flash of light focused upon the retina with an increased probability of discharge. One can state the relationship between stimulus and response in these circumstances only in terms of probability, since any particular flash of an identical series may or may not be followed by an action potential.