The distributed tactile memory system ofOctopus

Abstract

The capacity of an octopus to learn to react differently to rough and smooth objects is dispersed in the nervous system in two senses. (i) There are units in various parts of the brain capable of some change with experience. (ii) On the other hand the main tactile memory system is a complex of parts each with a distinct function. The response of drawing in a particular unfamiliar object rapidly habituates unless re-inforced by feeding. Learning consists in specific increase in the probability of taking objects rewarded by food or decrease for those that are unrewarded or are painful. Octopuses are capable of a fine degree of discrimination of degrees of roughness, mediated by feature detector cells lying probably mainly in the median inferior frontal lobe. The output of this lobe passes to the subfrontal lobe, and this inferior frontal system is responsible for 60% of the capacity for learnt discrimination. It is here that signals of taste or pain determine changes in the tendency to accept or reject an object touched. It is postulated that each feature detector and its microneurons constitutes a unit of memory or mnemon and that the accuracy of the discrimination is determined by the number of mnemons that have been appropriately altered by learning. Removal of some of the units in the subfrontal lobes leads to a corresponding reduction in accuracy of learnt discrimination. The superior frontal and vertical lobes provide further units, whose perations increase the reliability of learnt behaviour, being responsible for at least 25% of the total capacity for discrimination. In the absence of both the inferior frontal and vertical lobe systems there remains a capacity for discrimination with about 15% of the normal accuracy. Some residual capacity remains in suboesophageal regions even after the inferior frontal, vertical and subvertical lobes have been removed and the optic tract severed. All the parts that are concerned in learnt tactile discrimination contain numerous small short axon cells, which are probably the agents that give the mnemon units their capacity to change the probability of response to specific features of the input. The outputs from the various channels lead through a cascaded arrangement to the motoneurons that control response.