Every now and then, the tiniest creatures show glimpses of the highest-orders of cognitive ability. Plenty of animals do very clever stuff they really shouldn't be able to do, given the way we currently conceptualise cognition. This raises obvious questions about how. But it also should make us wonder about what really makes us different from our apparently less-sophisticated brethren.
I'm not just talking about the usual comparisons between various primates and humans in their tool use or their social habits. I'm talking about the truly baffling stuff, like bees learning to count.
Bees are already a pretty big deal, cognitively speaking Bees can actually
do a great number of extraordinary things. For example they can learn from each other even very un-bee-like things, like how to play football. This kind of learning from others is called "social learning", and is something thought quite unusual in the animal world. Learning by trial and error seems pretty straightforward for most things with a brain. But social learning seems to require some understanding that other bees are different from you, with their own agency. Something children can't really do until they're about three. But this learning numbers thing is also pretty special.
Why this counting thing is extra special Numbers are abstract concepts.
Abstract concepts are basically ideas about the relationships between things that go beyond perceptual features: general rules or principles that apply across contexts and in novel circumstances. You know, for example, that a litre of water in a water bottle is the same as a litre of water frozen into ice cubes. It doesn't look the same, but it is. This is an abstract concept: equivalence. Seems trivial maybe, but I'll point out that kids have trouble doing this too.
Numbers are also abstract. 'Two' of something could be two of anything. Two books. Two days. Two ideas. Or two different things: a car and a book. The concept of number isn't really that intuitive, when you think about it.
It might be no surprise, then, that for a very long time, we've been under the impression that animals couldn't grasp abstract concepts. In 1689, John Locke wrote:
Brutes abstract not.
And until the 20th Century we just assumed that was the case. But then we started to test it and we found some surprising results.
The experiment that changed our minds The 'match-to-sample' task is a
basic test of working memory. Let's say I first show you an apple. We're going to call that 'the sample'. Then I show you an apple and a pear. We'll call these 'the alternatives'. You need to 'match the sample'—that is to say, you need to pick the alternative that is the same as the sample. In this case, you'd want to pick the apple, not the pear, or I'm going to think you have some kind of memory issue.
This test, in and of itself, is not very exciting except as a diagnostic tool. But, once you've learned how the task works, I can test you with something else—let's say a water bottle and a pot plant—and you'll be able to do the task again. I show you the water bottle (the sample), then the water bottle and the pot plant (the alternatives), and you choose (I hope) the water bottle. In doing this, you've demonstrated that you can learn an abstract concept: the concept of identity, or 'sameness'. You've learned that you need to choose the thing that is the 'same' as the sample, regardless of what that sample is.
The exploration of what kinds of things animals can learn is an entire discipline—comparative intelligence. As Roger Thomas put it, "an animal’s general cognitive ability is determined" by how many kinds of concepts an animal can learn. In the '80's, we had only got as far as thinking non-infant humans and language-trained chimpanzees were clever enough to abstract. But this was mostly because we couldn't figure out a test easy enough to train other animals on. As soon as someone figured out this extension of the match-to-sample, we piled on.
Actually, heaps of animals can learn abstract concepts For [a paper I
co-wrote in 2018]12, I did a little digging and found examples of successful tests in non-language trained rhesus and capuchin monkeys, gorillas and orangutans and baboons. But particularly clever researchers figured out how to test even non-primates. We've seen success with rats, dolphins, sea lions, corvids (the crow family), pigeons, parrots, horses, budgerigars, the short-beaked echidna, ducklings, and even goldfish.
It's no small feat, I assure you. Building something that allows sea lions to match samples and then enticing them to do it is an impressive display of persistance. In fact, many researchers concentrate on how these complex set-ups might actually not be good enough to demonstrate concept learning. Maybe some feature of the contraption could be biasing the animal in a way that simply looks like it's doing something special. I'm less interested in this. The sheer volume of the successes make this seem like a thing, even if a handful of experiments are a little dodgy.
That said, all these animals have far more sophisticated brains than the honey-bee. And yet, honey-bees can do it too.
Honey-bees really shouldn't be able to do this Typically, any kind of
complex cognition is thought to happen in the mammalian neocortex (the wrinkly bit wrapped around the outside of the brain), or its equivalent in other animals. To give you an idea of how far away honey-bees are from having anything like a neocortex, let me give you a sense of scale. When we scan brains for activity, we often use functional Magnetic Resonance Imaging (fMRI). You can use this to see where in the brain there are more active or less active neurons, and then maybe infer which regions of the brain are responsible for what. We can isolate this activity in a couple of ways to figure out what's going on—we can talk about which structures are involved (the neocortex, or maybe the basal ganglia) which gives us some information; we can talk about which regions are involved (the prefrontal cortex, for example) which gives us different information again. These things are all fairly big though, housing billions of neurons. If we want to get into the detail, we go smaller. The smallest unit of measurement we have is called a 'voxel': a tiny cube of brain between 1.5mm and 3mm cubed. Inside that voxel is between 600,000 and 1.5 million neurons, depending on where in the brain you are. That's our smallest unit of measurement in fMRI.
A honey-bee has 900,000 neurons total. Its brain is smaller than our smallest unit of measurement here. If we put it through an MRI scan, we wouldn't be able to tell anything about its brain except whether it was alive or dead. Honey-bees have various brain structures, yes. But nothing that tiny can even pretend to approach the kind of complexity the neocortex can support. That's how far away they are from having anything like a neocortex.
Now, it's true to say that they can't do the match to sample task well. Most animals can do it correctly about 90% of the time. Honey-bees max out at about 75%. But that's still far better than chance. Moreover, some computational modelling run out of the University of Sheffield demonstrates that the task doesn't require a complicated brain at all. At least, not for learning this particular abstract concept. Not for bees. Theoretically, it could be done with a circuit in the bee brain so simple you could build it in your high-school shop class.
Let's talk about what that means Following his famous claim that “brutes
abstract not”, Locke goes on to note “the having of general ideas is that which puts a perfect distinction betwixt man and brute”. The ability to abstract rules like this is one of those things that we've long thought separated us from other creatures. But here, that's not the case at all. The honey-bee bridges the divide with two simple pathways in a tiny brain.
Let's be clear, of course, just because they can solve this problem doesn't mean they can abstract in any context. Yet, the fact that they can learn 'sameness' or count—the fact that they can learn from their peers—these things challenge our traditional notions of higher-order and lower-order processes. They make us question what truly characterises sophisticated thought. To me, it's a wonderful thing. In setting aside these traditional notions, one can pose new and exciting questions. In what other ways could simple neural mechanisms come together to facilitate higher-order aptitudes? Should we keep referring to them as higher-order? And, what does it mean to be clever? It may be the case that “the having of general ideas is that which puts a perfect distinction betwixt man and brute”, but the meaning of that statement is no clearer now than in the 17th Century. These new lines of enquiry may provide the means with which to determine the answer.