Downy Woodpecker Foraging Behavior: Efficient Sampling in Simple Stochastic Environments

I describe a study in which free—roaming downy Woodpeckers (Picoides pubescens) were allowed to forage in three different, patchy stochastic environments where patches (small, thin logs with holes drilled into them to hold food items) contained either zero or a fixed number of food items. All the patches in an environment were, however, identical in appearance. Downy Woodpeckers systematically searched the holes of a patch for food items, and thus the foraging task for an energy—maximizing woodpecker was to determine to what extent to sample a patch without success (without finding a food item) before giving the patch up as being empty and moving on. A model is presented to determine the optimal sampling solution. Although this foraging task is relatively simple, optimal solutions to many stochastic foraging problems may be quite complex. Thus, the observed foraging behavior is compared to the model's predictions as well as to some suggestions form the literature of simple behavioral mechanisms (e.g., a fixed giving—up—time strategy) that animals may use to approximate optimal solutions. The results show that the woodpeckers clearly recognized the fact that some patches contained food items while others did not, and use this information in following a sampling strategy more sophisticated than any of the suggested simple behavioral mechanisms. The predicted sampling behavior on empty patches and that observed were in qualitative but not quantitative agreement. The model predicted that a single number of holes in a patch should be sampled without success while, in fact, a distribution of number of holes sampled was observed. The mode of the distributions, however, corresponded to the predicted value in two of the three environments. It is shown that the woodpeckers did not know with complete accuracy at least some of the model's parameters, and that this may be related to a weak counting ability. A weak counting ability, the need to sample for environmental changes, a stochastic element in a bird's behavior itself, or some combination of these factor may contribute to the observed variability in sampling behavior. Nevertheless, perhaps through some ability to estimate time, the woodpeckers exhibited a sampling strategy very similar to that predicted by a simple energy maximization model, and such sampling abilities are crucial to more complex stochastic foraging models.