Optimal foraging theory suggests that animals have evolved to maximize their net rate of energy intake; all things being equal, they should leave a current depleting patch when an alternative patch would provide either more or sooner food. In nature, however, typically all things are not equal. For example, uncertainty about the value of alternative patches, time to travel to those patches, and potential dangers incurred in changing patches may delay leaving the depleting patch, when it would otherwise be optimal to do so. We tested the hypothesis that leaving the current patch may be delayed, by providing pigeons (Columba livia) with a continuous choice between a progressive schedule, in which each access to food could be obtained with an increasing number of pecks, and a multiple schedule, in which a colored light signaled the number of pecks required for food. The pigeons could switch from the progressive schedule to the multiple schedule at any time. We asked if pigeons would tend to switch when the signaled multiple schedule required fewer pecks than the next reinforcer provided by the progressive schedule. We found that pigeons tended to switch to the multiple schedule sooner than would have been optimal—one might say they precrastinated. We propose that, on the progressive schedule, the signal to switch was not just the number of pecks required for the next reinforcer but also the more general cue that reinforcement was becoming more difficult to obtain—a form of serial pattern learning.