UBC Theses and Dissertations
The role of spatial pattern learing in hummingbird foraging Sutherland, Glenn D.
Foraging efficiently requires that animals organize their behaviour in relation to spatial and temporal patterns of food distribution and abundance. This thesis examines how hummingbirds use information about spatial food distributions in their foraging behaviour. I used laboratory tests with large arrays of feeders to clarify the role of learning and memory for spatial location in hummingbird -foraging and to identify constraints on these abilities. The main questions were: "What characteristics of the spatial patterning of food supply constrain hummingbird foraging success?"; and "To what extent are learning and spatial memory important in determining where hummingbirds forage?". Hummingbirds learned to visit the most profitable feeders in large two dimensional arrays in which the only source of information to feeder quality was provided by spatial location. These arrays had equal numbers of profitable feeders and the same total energy available; they differed in the spatial arrangement of food. Birds learned more quickly and reached somewhat higher levels of performance on arrays of a few large groups of feeders than on those with a larger number of smaller groups. However, hummingbirds also learned to perform well even in complex arrays, but they performed consistently well only if the feeders were regularly distributed. Memory for the location of feeder groups was more important in accounting for performance in these experiments than learning individual locations or simple movement rules. In experiments in which I switched the distribution of food to its mirror-image after birds had reached asymptotic performance, all birds showed a lag in redistributing their foraging effort. There was little evidence that the birds re-learned the new distribution faster after the switch than in the initial learning phase. A detailed analysis of movements made in a complex array suggests that hummingbirds avoided visits to bad feeders by organizing their visits within and among small patches. These results are consistent with a model of hummingbird spatial information-processing in which feeder quality is "mapped" into a coarse-grained set of expectations that becomes stable and can persist in spite of changes in the underlying distribution of food. The principal effect of complexity is to reduce the rate of learning distributions of feeder quality; all of the patterns I used could be learned by the birds although birds did reach higher levels of performance on the simplest than on other patterns.