Pattern recognition in honeybees: multidimensional scaling reveals a city-block metric.

Bees were trained to discriminate ring-patterns which varied in number of rings and in size. Transfer-tests revealed size discrimination to be largely independent of pattern type and vice versa. A multidimensional scaling procedure, using Minkowski metrics as models, was applied in order to determine the bee's "perceptual metric". The city-block metric, and not the Euclidean metric, provided the best description of the data. Apparently, the bee's perceptual system derives the overall dissimilarity of complex ring-patterns additively from the component differences. These results are discussed with regard to "holistic" and "analytic" processing modes postulated for the perception of human subjects.