1. The rate of learning to attack a rectangle seen at a distance was studied under various conditions.
2. Learning is rapid only if the food reward is given after the figure has been shown, and if it is near to it.
3. There was no learning to attack if the food was given in the home before showing the figure, nor if the food was given after the octopus had returned to its home.
4. There was no marked difference in rate of learning with trials at intervals of 5 and 50 min.
5. After removal of the vertical lobes learning to attack a rectangle occurred only slowly. Operated animals attacked less reliably after twenty-five trials than normal animals after five trials.
6. After being shocked three times following showings of a rectangle normal animals that had previously learned to attack made few further attacks during the next 4 days. Animals without vertical lobes began to attack again during the same day; little sign of the effects of shocks was seen after 24 hr.
7. The process of forming representations that prevent attack is estimated to be at least five times less efficient in animals without vertical lobes.
8. This was confirmed by the effect of shocks in preventing attacks at crabs. In normal octopuses attacks ceased after two or three shocks at 10 min. intervals and were only resumed 3 days later. Animals without vertical lobes attacked up to ten times in spite of shocks. A few hours later all attacked again. However, some evidence of more enduring representations in the memories was seen.
9. There was no great difference in the effect of shocks given at intervals of 5 and 50 min. in preventing subsequent attacks at crabs by animals without vertical lobes. Under some conditions the more widely spaced shocks produced the greater effect.