Bees live in a fantastically complex visual world, visiting luscious flowers and navigating through meadows and woods. But it wasn't clear how bees process complex images from their cluttered visual environment. Adrian Dyer, from RMIT University, Australia, explains that there were two competing theories of how bees process visual images.

In the first, the bees simply learn the shape of an image, such as a flower, on the retina and then recognise it by matching their current view with the image that they learned previously. However, the second theory suggested that bees' brains are capable of more complex visual processing: that they might be able to learn the relative configuration of features in an image so that they can still recognise the object even when approaching from different angles. With evidence stacking up for both theories, Dyer wondered whether the bees might actually use both tactics. Intrigued, he developed two different training procedures to find out which mechanism they use and when (p. 397).

First, Dyer painstakingly taught individual foraging bees to visit a specially designed Y-shaped maze where the bee could choose to visit one of two images. Then, he trained them to see if they could learn a memory of a simple image by rewarding them with sucrose solution when they visited a target four-block image and punishing them with a bitter taste of quinine when they visited a second four-block image (the distractor) that he did not want them to learn. Finally, he removed the sucrose and quinine and tested whether the bees would keep returning to the first block image. They did, and when he tried to confuse them by placing the image on a tree background, they still recognised it.

But the real test was to see whether the bees would recognise the four-block image when it was enlarged. If they did not, then they must be using the image matching visual processing mechanism. Scaling up the image, Dyer realised that the bees no longer recognised it. The bees were using the retina image mechanism because of the simple training regime. But which mechanism would the bees use after more complex training?

To test this, he showed the bees a series of four-block images that all shared the same overall layout but varied subtly. Using the same training procedure – where visits to the target images in the maze were rewarded with sucrose, while visits to the distractor were punished with a dose of quinine – Dyer tested whether the bees could learn the shape of an image based on the configuration of key features by showing them an enlarged version of the four-block image.

This time, the bees passed the enlarged image test; they had no problem picking the correct image. They were using the more sophisticated configuration mechanism to recognise the image.

So the bees were using both mechanisms to recognise images depending on how they had been trained and Dyer adds that this makes perfect ecological sense. ‘A bee lives in a tree of a certain shape. As it is flying it sees the tree from a variety of different distances and then it is valuable to use the configural mechanism. But if it is foraging from flowers and always views the same flowers from 5 cm, then maybe it just wants to use a retinotopic match because it is much faster and easier to learn’, he says.

A. G.
D. W.
Seeing near and seeing far; behavioural evidence for dual mechanisms of pattern vision in the honeybee (Apis mellifera)
J. Exp. Biol.