Sensorimotor Learning in Bumblebees: Long-term Retention and Reversal Training

Bumblebees were trained in biologically realistic sensorimotor tasks to test how learnt information from more than a single task is organised in memory. Bees (Bombus impatiens and Bombus occidentalis) learned to collect sucrose solution from the arms of small T-mazes. The reward was offered in the right arm of a maze when the entrance was marked blue, and in the left arm when the entrance was yellow. Bees were trained on either one or both of these tasks until after they had reached saturation in terms of speed and accuracy. One group of bees (Bombus impatiens) was evaluated for long-term retention by retesting (a) on the day after training and (b) between 3 and 4 weeks after training. Performance did not decline overnight but, when bees were tested after a delay of several weeks, an increase in error scores and times taken to negotiate mazes occurred. However, bees started out at a better level than during the initial training, indicating that information acquired during initial training had not been entirely erased from memory. A second group of bees (Bombus occidentalis) was tested to see whether it was possible to reverse the association between colour and direction, so that blue entrances now meant that the reward was on the left side, and yellow entrances meant that it was on the right. Bees reached saturation on both tasks as rapidly as they had during the initial training. However, after a second reversal, bees chose directions randomly for several hundred trials, indicating that the first-learnt colour–direction associations had not been forgotten as a consequence of the first reversal: instead, information from the initial training must have been retained, but actively suppressed, causing difficulties in reestablishing the appropriate sensorimotor associations during the second reversal. Finally, a single Bombus impatiens worker was trained on only a single task, but with multiple reversals. The bee went through an intermediate phase when performance was poor until, after more than seven reversals, it required only three trials to assess that a reversal had taken place. This indicates that bees can learn actively to suppress irrelevant information, but only with an extended training schedule. Taken together, these results suggest that sensorimotor information is rarely lost from long-term memory, either during extended lay-offs or during interfering reversal training sessions. Organizing, correctly retrieving and suppressing information accumulated in memory seems to be more of a challenge than storing new information.