You may have seen a flurry of articles this week about bees that pull string to get a reward. This is, of course, highly entertaining and has led to a number of fun videos like this one popping up that show the bees doing this behaviour:

To ecologists or bee-enthusiasts who have spent time watching bees in their natural environment, it is perhaps not surprising that bees are capable of this task. Indeed, the manipulation of flowers if often seemingly more complex than simply pulling a bit of string. We also know that in a lab environment bees can do some pretty novel behaviour (see my article on ball-pushing bees). However, the really interesting result that comes from this research is not simply the fact that bees can pull string (adorable though this is). Instead, the research findings have implications much wider than bees and string, specifically because it is not 'intelligent' behaviour (in contrast to the claim of this video and much of the media coverage). 

Some Background

One of the key features that we associate with being human is ‘culture’. Broadly speaking, this is our ability to learn from other individuals, and then pass a particular behaviour down successive generations or throughout a group. The accumulation of many different behaviours and strategies learned from each other is what has allowed human society to advance to the current state it exists in. Any semblance of culture in other animals has, therefore, been pretty exciting to us, because it represents a means by which we can study and experimentally manipulate culture, and thus try to work out how it first arose in early humans. However, examples of culture in other animals (e.g. chimpanzees, dolphins, whales) have been difficult to experimentally manipulate in order to work out exactly the mechanism behind how the new behaviour spread to form a culture. This new study with bees demonstrates how a new behaviour can be learned by an individual and then spread through a group. This is exciting because, even though it’s not necessarily how all cultures arise, it is one way that they can arise. What’s more, it demonstrates that seemingly complex behaviour and the spread of such behaviour through a group can be explained by simple underlying rules. All the findings in this experiment come down to the fact that bees are excellent at learning associations. This statement will become clearer and more relevant as you read on. 

The Experiment

In the bee string-pulling study, recently carried out by Sylvain Alem and colleagues from Queen Mary University of London, researchers trained bees to pull string to reach an artificial flower (a flat coloured disc with a sucrose reward on it) which sat under a Plexiglas cover. To do this, the researchers had to train the bees in a step-wise procedure. First the bees were trained just to drink sucrose (sugar and water, the primary components of nectar) from the artificial flowers. They were then trained to access the flower when it was half hidden under a piece of Plexiglas. Once the bees had learned to do this, then they were trained where the flower was 75% hidden, and finally, 100% hidden, requiring the bee to pull string. This stepwise training was necessary- without it, the bees would simply try to access the flower through the Plexiglas. This first stage of the experiment shows that, yes, bees can learn a novel and seemingly complex behaviour, when trained in a series of steps.

A diagram showing how an individual bee was trained in a series of steps to pull string to gain a sucrose reward. Credit: Alem et al. 2016

This reminds me of a now-classic, slightly tongue-in-cheek study that was published in Nature in 1984 by Robert Epstein.  In 1917, a scientist claimed that chimpanzees showed ‘insight’ when they assembled a pile of boxes in order to reach a banana that was hung out of reach. In order to demonstrate that this task could be achieved without the need to invoke ‘insightful’ behaviour on the part of the animal, Epstein essentially trained pigeons to do the same thing. By training the pigeons in a series of steps, the pigeon learned to push a box into the correct position and then climb on top of it to reach a reward (which was unapologetically in the shape of a banana). If you were to just see this behaviour in its final form (can be viewed here), it appears very impressive and ‘intelligent’. However, the seemingly sophisticated behaviour can be explained by a series of simple training steps.

Going back to the bees, how was it that they learned the seemingly complex behaviour of string-pulling? One hypothesis the researchers had was that they learned that the coloured stimulus (e.g. the blue fake flower) had a nectar reward on it. This meant that the bees would get as close to the blue flower as they could and then just scrabble with their legs, resulting in the string being pulled. In the earlier stages of training, the motor action of scrabbling with their legs brought the fake flower closer towards them, and now the scrabbling action (if conducted in the correct location, on the string) would have the same effect. To test this, the researchers presented the bees with the same set-up but with the string attached only to the sucrose-containing well, with no coloured stimulus attached (see video below). This time, only 2 of the 15 bees solved the task, with most of the bees not even trying to solve it, implying that the bees needed the coloured ‘flower’ as a stimulus to either trigger the behaviour or even know that the nectar was located there. However, after being trained for an additional 48 hours, 11 of the 15 foragers solved the task without the coloured stimulus. This could be because initially when the bees were trained, their primary stimulus that predicted the sucrose reward was the blue coloured disc, however, as they gained more experience pulling string, the string became a better predictor of the sucrose reward, and thus they could solve the task without the coloured cue being there.

A video of the 2 bees that solved the task without the coloured stimulus present: (the bee finally solves the problem at around 4 ½ minutes in)

Could any bees solve the task without this extensive training? Well, yes, but only two of 50 tested. These two seemed to be particularly motivated individuals, and seemed to stumble across the answer by chance. Watching this video of one of these persistent bees is either rather painful, or pretty wonderful, depending on how you get your kicks.

Now, the really interesting part of this experiment was that other ‘observer’ bees could learn to string-pull from the trained (‘demonstrator’) bees, without having had any training themselves. 25 bees watched demonstrator bees pull string to get a reward 10 times. When given a chance to pull the string themselves, 15 of these observer bees were able to do it (although it took them a lot longer, about 3 minutes. However, if a human pulled the string, and then let a demonstrator bee land on the flower and drink from it, the observer bee didn’t learn. Thus it seems that the observer bees didn’t actually learn the string-pulling itself from the other bees or anything to do with the physics of moving the object, but instead learned from them where to go and scrabble to gain the sucrose reward. Indeed, observer bees were seen to fly to the area where the demonstrator bee had been, spending more time exploring that area. What’s more, if the experimenter rotated the flower such that the string protruded in a different location, the bee was never able to solve how to pull it. Having said that, the observer bees did spend more time on the side of the Plexiglas where the string was. Therefore, overall, it seems that the observer bees paid most attention to the location of the demonstrator bees, using this as their primary stimulus, and paid less (but still some) attention to the location of the string itself.

An observer bee (previously watching from inside the box) solves the task herself after a bit of scrabbling around. 

Now, the really exciting part of this research is what happened next. Using the novel string-pulling behaviour that the researchers had trained bees to do, they then went on to see whether this behaviour could spread through the colony. To do this, the researchers first trained a single individual in each of three colonies to string-pull. They then let foragers out in pairs to engage with the task. As a bee learned the task from the trained demonstrator bee, so other bees then learned it from her, and so the behaviour spread through the foraging bees within the colony. By watching videos of the bees interacting, the researchers determined that the bees generally learned from each other in a series of steps. Initially, they just scrounged, landing on the flower once the trained bee had pulled it out and was already drinking from it. The scrounging bee, at this point, had learned that the demonstrator bee generally led to a reward. At this point, the scrounger tended to follow the trained bee around, staying close to her. This meant that the scrounger would often end up near and on the artificial flower to get the reward, and thus learn that the artificial flower predicted a reward. This could be seen as the bee trying to reach the artificial flower, by scrabbling at it through the Plexiglas, without pulling the string. After much scrabbling near the string, next to the trained bee who was pulling the string, and ineffectively scrabbling at the string but not pulling it, the bee would finally learn to pull the string herself. Those observer bees that learned to string-pull then essentially became demonstrators themselves, as other naïve bees learned string-pulling from them. Thus the behaviour was able to spread throughout the group, and bees learned it who had never had contact with the originally-trained demonstrator bee. Indeed, in one colony the original demonstrator bee actually died, but the behaviour still continued to spread throughout the group via the bees that had learned from her.

Here is a video of how the behaviour spread through the group. The originally trained, demonstrator bee is shown in yellow, and all the other bees are represented by the other numbers (or ‘nodes’). An interaction between two bees is shown by a line between them, meaning that the more interactions they have with each other, the thicker the line becomes. The total number of interactions a given bee has with any bee is shown by the size of her node (bigger nodes= a bee who interacts with a greater number of bees). For more information see Alem et al. (2016).

Therefore, what is wonderful about this study is that it demonstrates how this behaviour (that appears complex) can spread throughout a group and establish a ‘culture’, while only needing to invoke the ability to learn a series of simple associations (and that a particular motor behaviour leads to a reward). This demonstrates that animals do not need sophisticated cognition in order to have culture, and that in many cases where we see culture in other animals, it may be explained by simple mechanisms. This study is one of a handful of recent studies that has demonstrated that seemingly sophisticated behaviour can be explained by simple underlying mechanisms. Much like a magic trick, discovering the mechanism behind a behaviour can take the mystery out of it for some people and make it seem less impressive. However, I think the opposite it true, that the truly exciting thing about sophisticated behaviour is understanding exactly what mechanisms underlie it.



Main reference: Alem, S., Perry, C. J., Zhu, X., Loukola, O. J., Ingraham, T., Søvik, E., & Chittka, L. (2016). Associative Mechanisms Allow for Social Learning and Cultural Transmission of String Pulling in an Insect. PLoS Biol14(10), e1002564. (link)

Other references:

Epstein, R., Kirshnit, C. E., Lanza, R. P., & Rubin, L. C. (1984). "Insight" in the pigeon: Antecedents and determinants of an intelligent performance. Nature308, 61-62. (link)