ABSTRACT
The compound eye of the bee collects a composite image of the panorama in radial coordinates over about 200° horizontally (and ±90° vertically) with about 10,000 ommatidia (analogous to pixels). The photoreceptors are of three color types, with a peak in the green, blue, or ultraviolet. As each contrast in the panorama passes, small motion detectors spanning two or three ommatidia sense the angular motion relative to the eye from the sequence of intensities. The units of pattern or place recognition with excellent resolution are: a) modulation detectors one ommatidium wide; and b) edge orientation detectors three ommatidia wide. In the deep optic lobe, there are large-field neurons with tonic or phasic antagonistic inputs from two or three color types of receptors. All detectors respond to passing edges as the bee scans in the horizontal plane. Irrespective of the actual pattern, the responses of the detectors are summed by type and position to form 7 or 8 different cues in each local region of the eye. If rewarded, the cues are remembered in an order of preference, and the coincidence of different cues in a local region may be remembered as the label on a retinotopic landmark. The panorama is represented as a low-resolution sparse map of landmark labels. Bees learn a combination of one or more labels to identify a place and find the reward. To a bee, a color or pattern or even a mixture of modalities is just another landmark.
