1. Octopuses will make detours to get at prey seen through a transparent barrier, even though the detours carry them temporarily out of sight of the prey. The possibility of non-visual intermediate cues can be eliminated experimentally.
2. Eight out of 29 octopuses tested made a successful detour at the first trial. The remainder had to be led through the maze by means of a crab on a string on one or more occasions before they did it spontaneously.
3. Once having completed a successful detour down a passage to get into a feeding compartment to the side of the corridor, an octopus rarely failed in subsequent trials, whether making a detour to right or to left. Eleven unoperated control animals, each tested 20 times, totalled only 10 errors, by going into the feeding compartment on the wrong side of the corridor; they failed to make a detour only 13 times in 220 trials.
4. Performance nevertheless showed some improvement with practice. The number of seconds spent attacking through the wall of the feeding compartment before entering the corridor and the number of abortive entries into the corridor both decreased as trials continued.
5. Removal of the statocysts did not stop the high proportion of correct responses. In 39 trials four animals made 6 errors, 3 of them attributable to unilateral blinding of one of the animals (see below).
6. Blinding the animals in one eye led to systematic errors in detours towards the unblinded side. In 63 trials towards the unblinded side, four octopuses made 16 errors and failed to complete the detour 28 times. In 63 trials towards the blind side there were only 4 errors and 9 failures.
7. It is concluded that orientation in detours is maintained by keeping close visual contact with the wall separating the octopus from its prey. The experimental results indicate that octopuses do not take bodily position into account in learning to make detours.
8. Animals trained to make a detour to one side only performed faultlessly when first required to make a detour in the other direction; there is no indication that octopuses are guided by external cues in these experiments.
9. These results are discussed in relation to the organization of motor and learning systems in animals generally. Creatures with hydrostatic skeletons and flexible bodies must of necessity have a decentralized motor control system, and this has important consequences on their learned behaviour.