Relative Size Learning in Honeybees (Apis mellifera)

Abstract

The honeybee is the only invertebrate species that has been extensively studied in learning experiments, and the results are very similar to those of vertebrates. Such similarities raise questions about the evolutionary development of learning capabilities and their underlying mechanisms. Recently, studies of learning in honeybees have shifted to more complex cognitive phenomena, such as relational learning, once thought to be exclusive to vertebrates. Honeybees are capable of solving same-different relational problems, such as oddity. In the present experiments, bees were first tested for evidence of relative size learning and then for the ability to use relative size in an oddity problem. The methodology employed a free-flying procedure in which foraging honeybees were pre-trained to shuttle between their hive and a testing window where they received a sucrose solution as a reward for choosing the correct stimulus on each trial. After drinking the sucrose, the bee returned to the hive to unload and then returned for another trial. The stimuli used were wooden blocks of four different lengths ranging from short to long. In Experiment 1, bees were tested with a simple discrimination problem using single- pair training of closely-sized stimuli to ensure that all of the stimuli used in subsequent experiments were visually discriminable by the bees. In Experiment 2, bees were presented with two blocks of different sizes on every trial. Half of the bees were in Group Large and were rewarded for choosing the larger block of the pair. The other half of the bees were in Group Small and were rewarded for choosing the smaller block of the pair. The bees in both groups learned to choose correctly, suggesting bees are capable of relative size learning. To determine that Experiment 2 had not been solved by learning the reward probabilities associated with each block size, Experiment 3 served as a novel control condition. The task could not be solved by a relational rule but only by learning reward probabilities. The bees were unable to solve this task, indicating the bees in Experiment 2 had likely solved the task using a relational rule. Given the evidence provided by Experiments 2 and 3 that bees are capable of relative size learning, Experiment 4 was conducted using a task that required the use of two relational concepts: size and same-different. Bees were presented with three blocks, two of which were the same size and one a different size. On half of the trials, the odd-sized block was larger than the other two, and on the other half, it was smaller. The bees were always rewarded for choosing the odd- sized block. The bees learned to choose the odd stimulus significantly more often than chance. These results suggest that honeybees are capable of solving a problem that requires the use of two different relational concepts simultaneously, an ability so far only demonstrated in very few vertebrate species. Such cognitive sophistication in honeybees is surprising. It may be that foraging honeybees use both size and oddity relationships in their foraging decisions. Future research should examine the flexibility and robustness of relational learning capacities in honeybees to determine to what extent this ability is similar to that of vertebrates.