Sampling of schedule parameters by pigeons: tests of optimizing theory

Abstract

In three experiments, pigeons (domesticated Columba livia) were exposed to concurrent randomratio schedules of reinforcement in a keypecking apparatus, as a simulation of foraging in an environment containing two patches of unknown food density. All reinforcements were given for pecks on a single key, but different patch densities were associated with different key colours, and the colour was changed following pecks to a second key. Patch densities varied from day to day, and no colour reliably signalled the better patch from session to session. In the first experiment, either 256 or 1024 pecks were allowed per session. As predicted by optimal sampling theory, the proportion of the first 256 pecks made in the denser patch, and the number of reinforcements gained during them, were higher in the short than in the long sessions. These results were simulated using principles derived from operant conditioning theory. In the second experiment, the number of different patch densities used was varied between conditions. Birds made fewer patch changes when the density range was more restricted. In the third experiment, the colour cues differentiating patch densities were dropped in one condition, and the birds were less successful at finding the better patch in this condition. These results show that pigeons are sensitive to aspects of a foraging situation that are predicted by optimal sampling theory (e.g. the session length), and also in ways that have not been considered so far (e.g. patch density range). But they also show that cues that are not logically necessary (e.g. patch key colour) may in practice be an essential part of an animal's mechanism for achieving efficient foraging.