Spatial and non-spatial coding of patterns by the honeybee (Apis mellifera)

Abstract

Bees separate items in the three-dimensional world visually by relative motion, by colour, by range, and also by detecting global features in two-dimensional patterns. They resolve a grating of period 3°-4° in daylight but when the locations of contrasts are made useless as cues by randomizing or alternating the positions of black areas, they cannot discriminate between many simple geometrical patterns. Although bees confuse some patterns that look very different to us, they can be trained to discriminate certain global features although local features are randomized or partially obscured. In forward vision, they behave as if they can remember only what passes a limited repertoire of large-field filters for global patterns. By 'global' I mean the properties of the target as a whole, irrespective of local detail. By 'filter' I mean a broadly tuned feature detector that responds with a graded output to the match between itself and the external pattern. A filter could be a group of neurones with large superimposed fields. Our task is to identify the features that are distinct to the bees and distinguish the filters.