Experiments on Information coding in nerve impulse trains

Abstract

Shannon has provided us with a quantitative measure of information flow, channel capacity, which we apply to trains of nerve pulses. It is of course necessary to consider portions of a communication system—the coder and decoder—as well as the channel. With the help of a model, two possible codes are contrasted: a simple average-frequency code (3 bits/sec) and a complex pulse interval code (100 bits/sec). Neurophysiological experiments are described on the crayfish adversive walking response to light. Dissection of this “simple” invertebrate permits quantitative study of several single units separately and simultaneously, while the animal is in physiological condition. Critical experiments yield evidence to show that the complex code is not carrying physiological signal information: the simple 3 bit/sec code is proposed as adequate for the signal transfer requirements of the animal. The additional information (97 bits/sec) not used as meaningful signal for the crayfish may be used to give the neurophysiologist information concerning the noise processes in the coding mechanism: that is, in the process of generating nerve pulses. We hope that this article demonstrates the utility of embedding classical neurophysiological techniques within a framework of engineering systems theory: in this particular case, Shannon's mathematical theory of communication.