Reversal learning requires an animal to learn to discriminate between two stimuli but reverse its responses to these stimuli every time it has reached a learning criterion. Thus, different from pure discrimination experiments, reversal learning experiments require the animal to respond to stimuli flexibly, and the reversal learning performance can be taken as an illustration of the animal's cognitive abilities. We herein describe a reversal learning experiment involving a simple spatial discrimination task, choosing the right or left side, with octopus. When trained with positive reinforcement alone, most octopuses did not even learn the original task. The learning behavior changed drastically when incorrect choices were indicated by a visual signal: the octopuses learned the task within a few sessions and completed several reversals thereby decreasing the number of errors needed to complete a reversal successively. A group of octopus trained with the incorrect-choice signal directly acquired the task quickly and reduced their performances over reversals. Our results indicate that octopuses are able to perform successfully in a reversal experiment based on a spatial discrimination showing progressive improvement, however, without reaching the ultimate performance. Thus, depending on the experimental context, octopus can show behavioral flexibility in a reversal learning task, which goes beyond mere discrimination learning.
Reversal of a Spatial Discrimination Task in the Common Octopus (Octopus vulgaris)