How images may or may not represent flowers: picture-object correspondence in bumblebees (Bombus impatiens)?

Studies of bee cognition frequently use two-dimensional stimuli referred to as floral patterns, and yet how bees perceive pictorial representations is not known. An investigation of bumblebee (Bombus impatiens) picture-object correspondence was undertaken according to the theory of Fagot et al. (Picture perception in animals. Psychology Press Ltd, East Sussex, pp 295-320, 2000) that pictures and objects may be confused, perceived as independent or equivalent. In three experiments, bumblebees were given discrimination training and unrewarded testing in a radial maze. In the first experiment, preferences between artificial flowers and photographs of those flowers revealed a significant learned preference for the stimulus rewarded during training: no confusion following training. In the second experiment, bees did transfer learning from artificial flowers to photographs: some commonality between an object and photograph was perceived. In the third experiment, bees spontaneously generalized a learned preference for one artificial flower to its silhouette, but only for one of two flowers used in training. No generalization was obtained to drawn images. Some transfer between image and object is exhibited, likely by low-level feature matching, but transfer is poor with degraded images, cautioning against extrapolation of picture-based responding to natural correspondents. Despite evidence that bees exhibit some transfer while retaining discrimination, it is likely that the observed response is due to generalization more akin to confusion than true equivalence. Furthermore, although 2D patterning cues (line, edge and shade) provide discriminable cues for bees between 2D stimuli, it is not here supported that such features are perceived as equivalent to the intended floral structures.