Bumblebees flexibly adjust learning and decision strategies to sensory information content in a foraging task

Author's response

Anna Stöckl shared

You use the color orange and yellow alternatingly in some assays. Could you elaborate on why in some assays you chose orange, and why in some you chose yellow? You mention that bees naively prefer blue over orange and teal over blue—do you know where yellow fits in on this hierarchy (if indeed it’s a hierarchy and teal is then therefore preferred to orange)?

The genesis of these colours resulted from the fact that we actually started this study as student laboratory projects. In courses at Würzburg University, B.Sc. and M.Sc. students would perform pilot experiments that eventually led to this study. We started using out blue and yellow as a colour combination, because this is a typical combination used in many learning paradigms with honeybees and bumblebees. When we obtained the first results which showed that bumblebees indeed did not learn the pattern and shape features simultaneously with this perceptually distant colour combination, we tried a close colour combination, using the yellow we already had and adding orange. After observing that bumblebees with this close colour pair would learn both pattern and shape features and rely on these for their foraging decisions in addition to the colour features, we developed the final design, adding cyan to end up with two close and two distant colour combinations, to ensure that the observations we report do not depend on the wavelength of the stimuli, but indeed on their perceptual distances in the animals’ colour space.

Since we tested individual bees only with one colour combination each, we cannot establish a general hierarchy of colour preferences from our results. The preferences are only valid within the respective pairs. Colour preferences could also have been influenced by the shapes or patterns they were associated with. And indeed, while we found a preference for yellow over blue in the bees tested with patterns, there was none when presented together with shapes. Consistent across all experiments was the preference for blue over orange, and teal over blue. Yet, this particular stimulus design was not set up to resolve colour preferences or their hierarchies, and therefore we caution against interpreting the preferences we observed as general colour preferences of bumblebees.

Would a full blocking test (e.g., train configural cue first, then add color) help verify the associative model?

It would definitely be an interesting follow-up experiment to use a time-separated blocking test, a) training colour first, and then add the configural cue, to substantiate that learning colour first blocks learning of the configural cue and training configural cues first and then add colour, to see if configural cues, when learned first, would be equally potent at blocking colour learning.

Potential differences in learning outcomes between these sequences would be extremely interested in terms of weighting of cues and the potential to form associations. So there are many more exciting experiments to be performed to better understand how multiple cues are processed and learned in bumblebees in particular, and insects in general.

Do you have any ideas on what neural circuits might support this cue weighting in bees? Mushroom body? Lateral horn? Or perhaps somewhere earlier in the visual system itself?

The short answer is: we do not know. The more speculative answer is: colour learning has been shown to be linked to the mushroom bodies in insects (see beautiful work by Kinoshita et al. in butterflies https://doi.org/10.1016/j.cub.2021.12.032, and Paulk et al. in bumblebees https://doi.org/10.1016/j.asd.2008.03.002 and Vogt et al. in fruit flies https://elifesciences.org/articles/02395 ). But then, Drosophila has also been shown to learn some aspects of colour without the mushroom bodies (https://pmc.ncbi.nlm.nih.gov/articles/PMC311233/) – so it might also depend on the context as well where this information is processed. Where pattern and shape memories are formed, to our knowledge, is currently unknown, but the mushroom bodies are definitely the first place to look.

Moreover, the mushroom bodies are required for learning more complex sequences in bees https://doi.org/10.1007/s00422-008-0241-1  and would provide the right synaptic architecture to implement blocking of new associations by previously formed ones 10.1016/j.celrep.2023.112974, https://doi.org/10.1073/pnas.1508422112.