Three milestones in honeybee pattern recognition

The study of pattern recognition in the honeybee has seen three major milestones, or eras (Horridge, 1999; Srinivasan, 1994; Srinivasan and Zhang, 1998). The studies of the first era, by Karl Von Frisch (1915) and Hertz (1929), were carried out by cutting two or more patterns out of black paper, placing them flat on a horizontal table, and training a group of bees to collect a reward of sugar water at one of the patterns. The positions of the patterns were varied frequently to ensure that the bees learned to come to a specific pattern rather than to a specific location on the table. By analyzing the bees' performance in this task, they proposed the flicker hypothesis, which states that honeybees discriminate between visual patterns in terms of the frequency of on/off stimuli experienced by the compound eye as they cruise above the patterns (Hertz, 1933).

The next era commenced with the work of Wehner and Lindauer (1966), who began experimenting with patterns presented in the vertical plane. Although this may seem to be a simple modification at first glance, it opened up a whole new dimension in research on pattern recognition in bees. The reason was that one could now ask, for the first time, whether bees can distinguish between patterns that are rotated with respect to each other. This could not be done with patterns laid flat on a table, as the bees could approach such patterns from any direction. An interesting observation made in these studies was that during training, the bees tended to hover stably in front of the training pattern for a second or two before landing on the entrance tube leading to the reward behind the pattern (Wehner and Flatt, 1977). This led Wehner to propose the template hypothesis, according to which bees fixate the pattern and memorize it in a photographic sense. When, after training, the bee is offered a choice between the rewarded pattern and another pattern, the bee evaluates the two patterns by comparing the extent to which each pattern matches the stored template that the bee has acquired of the rewarded pattern during training (Wehner, 1967, 1981). The template theory of pattern recognition was corroborated by the work of Gould (1985, 1986), who showed that bees could be trained to distinguish rather small differences between patterns, suggesting that recognition does indeed involve a fairly precise spatial evaluation of the extent to which the black and white regions in the viewed pattern match the corresponding regions in the template.

The third era of pattern recognition in honeybees began when investigators began to wonder whether bees could learn abstract features, or properties, of patterns instead of, or in addition to, memorizing them photographically (Van Hateren et al., 1990; Wehner, 1971). This was an important step toward gaining insights into the cognitive capacities of the honeybee brain. In the next section, we outline the progress made in this era.