Abstract
When driving, we use mirrors to localize objects that would otherwise be invisible to us because they lie outside our field of view. Various species of vertebrates can learn to use a mirror to localize objects hidden from their view (monkeys: Anderson and Gallup 2011; chimpanzees, Menzel et al. 1985; a gorilla, Nicholson and Gould 1995; elephants, Povinelli 1989; pigs, Howell and Bennett 2011; African gray parrots, Pepperberg et al. 1995; crows, Medina et al. 2011). Octopuses are highly capable visual hunters who prey on live crabs. In the present study, we tested the hypothesis that Octopus bimaculoides could learn to use the mirror image of a visual scene to localize a predictor of food reward.
Three octopuses were tested in the present study. At the beginning of each trial, the animal was placed in an opaque box facing a mirror. A virtual crab was projected on a back screen hidden from view of the subject, but its image was reflected in the mirror. The animal’s task was to move out of the box, turn around and go to the side where the virtual crab was projected. This required using the mirror as a tool to locate the side of the projected but 'hidden' crab (Supplementary Figure 1, Panel A).
We found that octopuses made the correct choice significantly more often than the incorrect choice (O1: 9/10, O2: 6/8, O3:6/9; Fisher: chisq=15.68, df=6; p<.05). A simulation of this experiment with 250,000 iterations revealed that the frequency of correct choices was not due to chance (p= 0.002524, Supplementary Figure 1, Panel B).
In conclusion, we showed that octopuses are capable of learning to utilize a mirror to infer where their prey was located. This requires the cognitive capacity to use a complex visual representation of the environment to drive goal-oriented behavior.