Edge detection by landing honeybees: behavioural analysis and model simulations of the underlying mechanism.

The mechanism of edge detection in the honeybee was investigated by examining the effects of combining different kinds of visual cues that define an edge. Free-flying bees were trained to land at three different types of edges which were defined by texture and relative motion cues either in isolation or in combination with each other. Bees are able to detect and land at the three types of edges, but do so with different frequencies. In contrast to the naive expectation that edges jointly defined by two cues can be detected better than those defined by a single cue in isolation, the combination of the cues does not increase and may even decrease the detectability of an edge. When bees land at an edge the orientation of their body axis is strongly affected by the visual cues defining this edge. Model simulations were performed to test whether the experimental findings can be explained on the basis of a single edge detection mechanism sensitive to both types of visual cues. In the model, the information from both types of cues is sensed by two fields of movement detectors that receive their input signals from two adjacent patches in the visual field. The output of all detectors subserving either patch is pooled by integrating cells. The signals of the two integrating cells subserving the two adjacent patches are compared at a subtraction stage. The resulting signal is then rectified and forms the output signal of the model. The model simulations closely resemble the experimental results, thus providing evidence that edge detection by the bee could be mediated by a single mechanism.