Appetitive reversal learning differences of two honey bee subspecies with different foraging behaviors.

We aimed to examine mechanistically the observed foraging differences across two honey bee, , subspecies using the proboscis extension response assay. Specifically, we compared differences in appetitive reversal learning ability between honey bee subspecies: (Pollman), and (Skorikov) in a "common garden" apiary. It was hypothesized that specific learning differences could explain previously observed foraging behavior differences of these subspecies: switches between different flower color morphs in response to reward variability, and does not switch. We suggest that flower constancy allows reduced exposure by minimizing search and handling time, whereas plasticity is important when maximizing harvest in preparation for long winter is at a premium. In the initial or phase of the test we examined specifically discrimination learning, where bees were trained to respond to a paired conditioned stimulus with an unconditioned stimulus and not to respond to a second conditioned stimulus that is not followed by an unconditioned stimulus. We found no significant differences among the subspecies in the phase in appetitive learning. During the second, phase of the experiment, where flexibility in association was tested, the paired and unpaired conditioned stimuli were reversed. During the phase showed a reduced ability to learn the reverse association in the appetitive learning task. This observation is consistent with the hypothesis that foragers cannot change the foraging choice because of lack of flexibility in appetitive associations under changing contingencies. Interestingly, both subspecies continued responding to the previously rewarded conditioned stimulus in the reversal phase. We discuss potential ecological correlates and molecular underpinnings of these differences in learning across the two subspecies. In addition, in a supplemental experiment we demonstrated that these differences in appetitive reversal learning do not occur in other learning contexts.