Serial reversal learning in nectar-feeding bats.

We explored the behavioral flexibility of Commissaris's long-tongued bats through a spatial serial reversal foraging task. Bats kept in captivity for short periods were trained to obtain nectar rewards from two artificial flowers. At any given time, only one of the flowers provided rewards and these reward contingencies reversed in successive blocks of 50 flower visits. All bats detected and responded to reversals by making most of their visits to the currently active flower. As the bats experienced repeated reversals, their preference re-adjusted faster. Although the flower state reversals were theoretically predictable, we did not detect anticipatory behavior, that is, frequency of visits to the alternative flower did not increase within each block as the programmed reversal approached. The net balance of these changes was a progressive improvement in performance in terms of the total proportion of visits allocated to the active flower. The results are compatible with, but do not depend on, the bats displaying an ability to 'learn to learn' and show that the dynamics of allocation of effort between food sources can change flexibly according to circumstances.

Discrimination of inclined path segments by the desert ant Cataglyphis fortis.

Desert ants (Cataglyphis fortis) navigate by means of path integration, and perform accurately even in undulating terrain. They are able to correctly calculate the ground distance between nest and feeder even if their foraging excursion leads them over corrugated surfaces. To compute the respective ground projection when walking over an inclined surface, ants must measure its slope with sufficient accuracy--but how they do so is still not understood. Using a new behavioural assay that included a negative reinforcement, we investigated how well different slopes are discriminated by the ants. Ants were trained to visit an elevated feeder, via a ramp of fixed inclination (five training inclinations were used: 0°, 15°, 30°, 45°, 60°). The ants discriminated a steeper test slope that differed from the training slope by 12.5°. This discrimination performance was found to be constant for training slopes between 0° and 45°. Ants trained on a 60° slope, however, did not discriminate all steeper slopes, up to a vertical ascent, from the 60° inclination. The consequences of this discrimination accuracy for errors in the path integration process are discussed.