Computer-controlled learning in a simple system.

Abstract

An insect preparation, in which the frequency of discharge of an identi­fied tonic motoneurone can be modified by operant-conditioning, is recommended as a model for a study of cellular neuronal events underlying learning. To facilitate the study of the underlying events, we have developed computer programs for on-line control of training. The dis­charge pattern of the neurone is analysed continually, trends in mean frequency being recognized automatically. Experiments were conducted on a large number of locusts with the perfected computer programs. Statistically significant shifts are used to trigger the delivery of aversive electric shocks to the preparation. In the increase mode , shocks were correlated with spontaneous de­creases in frequency. These were followed, in 68% of preparations, by a stable increase in mean frequency (up-learning) that averaged 140.0%. In the decrease mode , shocks were correlated with spontaneous increases in frequency. These were followed, in 64% of preparations, by a decrease in frequency (down-learning) that averaged 62%. Experimentally raised or lowered frequencies were maintained for periods up to a few hours without, and longer with, reinforcement. The trends could be reversed by applying the reciprocal computer program.