SCIENCE AT THE SHINE DOME canberra 2 - 4 may 2007

Symposium: Development and evolution of higher cognition in animals

Friday, 4 May 2007

Dr Nathan Emery
Royal Society University Research Fellow, Sub-Department of Animal Behaviour, University of Cambridge, UK

Nathan J. Emery is a Royal Society University Research Fellow in the Sub-Department of Animal Behaviour, University of Cambridge. He received his BSc (Hons) in Neuroscience at the University of Central Lancashire and his PhD in Neuropsychology at the University of St Andrews. He has also performed post-doctoral work at the University of California - Davis and the University of Cambridge, and was a Health Emotions Research Institute Scholar. His current research focuses on social and physical reasoning by corvids (scrub-jays, rooks and jackdaws) and apes (chimpanzees and bonobos). He is also interested in the relationship between brain and behaviour, and how our understanding of animal minds and behaviour can influence animal welfare. He uses an ecological and ethological approach to address questions of cognition based on the natural problems faced by animals. He has authored over 50 papers and book chapters, including papers in Nature and Science. He has co-edited two books; The Cognitive Neuroscience of Social Behaviour (with A. Easton; Psychology Press, 2005) and Social Intelligence: From brain to culture (with N. Clayton & C. Frith; Oxford University Press, 2007).

Apes, corvids and the evolution of cognition

(Click on image for a larger version)

When we think about what might be the best candidates for cognitive ability, whatever we call it – ‘intelligence’, ‘higher cognition’, or what I tend to choose, ‘complex cognition’ – we tend to think of the guys shown here, the great apes: the orang-utans, gorillas, bonobos or chimpanzees. That is largely because they are very closely related to us; we can construct a very simple phylogenetic tree showing our close relationship to these apes. But it is also because they behave very similarly to us. Chimpanzees, for example, have been shown to demonstrate warfare, they make and use various tools, they may look like us in some respects.

But, while apes may be perhaps the most obvious candidates for us to think about, they are not the only candidates.

(Click on image for a larger version)

What I want to talk about is another, very distantly related species, the corvids: the crows, magpies, rooks, jackdaws, jays, ravens et cetera.

What perhaps might be surprising is the number of convergences in various aspects of biology and behaviour between apes and corvids. I don’t want to go in any detail through these lists of some very strong similarities – and there are a lot more – between these two separate families, but, for example, birds and primates have very complex visual processing; they live in very complex worlds in which they have to process visual information rapidly. They have long postnatal development, they can recognise individuals, they have complex communication systems that I think Gisela Kaplan will talk about, they also live in very similar worlds and have to solve similar problems. What we have suggested is that they also have very similar psychology.

(Click on image for a larger version)

To pick one of these examples, brain size: perhaps you might be surprised to know that corvids have exactly the same sort of relative brain size, if you take away the effect of body size, as the great apes. So the corvids are shown here by the blue dots – this is a particular one, the scrub-jays, which Nicky Clayton will talk about later – but they are on the same regression line as the apes. This is as compared with other birds and mammals which have much smaller brains than we would expect for their body size.

(Click on image for a larger version)

To look at differences: they have very similar relative brain size, but very different structures of the brain. What we see here is a chimpanzee brain and a crow brain. You can get these very striking structural differences in neuroanatomy, so we have a cortex in one case and a nidopallium in the other, but we can also get resultant behaviour that is almost identical in some respects. We find stick tool use in chimpanzees and then, as we have already seen from Pat Bateson’s talk and will see in Russell Gray’s talk, stick tool use in New Caledonian crows. They are almost identical in the way that their brains are actually creating the output of behaviour.

(Click on image for a larger version)

Nicky Clayton and I have argued in a science review of a few years ago that there are probably four aspects that contribute to shared complex cognition in corvids and apes – something we call a cognitive toolkit. We suggest that there are at least four ‘tools’ in this toolkit: causal reasoning, imagination, flexibility and prospection. Causal reasoning contributes understanding about cause and effect; imagination may be thinking about things that aren’t present, which is very similar to object permanence; flexibility allows you to apply your behaviour, what you have learned in the past, in new contexts; and then prospection is thinking about the future.

(Click on image for a larger version)

Because I don’t have the time to go through all four of these cognitive tools, I want to focus on causal reasoning.

Causal reasoning has traditionally been tested by using the trap tube problem, which is represented here. A black dot represents a piece of food located in a plastic tube, and on one side of this tube there is a trap that hangs underneath the tube. An animal, typically a tool user, can use a tool to push or pull this piece of food out of the tube.

The problem here is to try and avoid the trap, because if you push one way or pull the other way the food will fall into the trap and so not be accessible any more.

Chimpanzees, surprisingly, are very poor at this task. A few years ago Daniel Povinelli published reports on about four chimpanzees in his lab, and in fact only one learned this task, after more than 100 trials. We tested this in rooks. Rooks are a type of social crow, but they don’t use tools – which provides us with a particular problem. So we adapted this task for non tool-using rooks.

In this case there is the same problem – the trap and tube – but in this case we included a tool in the tube itself, and two little discs surrounding the food. So now the problem is set the same, the way that they have to understand this problem is the same: how do they avoid the trap here? But in this case we take away the tool-using component of it. In this case now they have to pull either one way or the other.

Actually, rooks in this case are very good. Of our four birds that were presented with this task, three learned it very rapidly, in fact much more quickly than chimpanzees.

One problem with this task is that there are two levels of explanation of how the animals may be solving the task. One is a very high-level explanation: causal reasoning. That is, they understand the consequences of their actions on the tool, and the subsequent consequences on the food. A much lower-level explanation is that they are actually applying a very simple rule: pull away from the object hanging underneath the tube – so, always move away from the trap.

(Click on image for a larger version)

So we designed a new version to try and get away from the simpler explanation – something we called the two-trap tube task. Here is a photograph of one of our rooks performing this task.

We provided two versions of this task to our birds. In the first case we have the same configuration but we have now put in an additional trap, which is nonfunctional. So we have a functional trap; the food falls in and it is trapped. But the food can be pulled across the additional trap, because there is now a lid on it, and so become accessible.

We also gave them a second version. We have added in a second trap again, but this one is also nonfunctional because it doesn’t have a bottom to it. Pulling across this one causes the food to fall down.

The idea is that the rooks can’t use a simple rule, to pull away from a trap or a traplike object, because there are now two traps for them.

[A video showing how the two-trap tube task works was shown, with the following commentary.] This is just one of our show birds, rather than an experimental trial. But, hopefully, it will show you that actually they think about this; they do ponder, as Giorgio Vallortigara suggested with his chicks.

You would see that it is actually quite difficult to discriminate in the apparatus which side has the base to it, to work out which is the functional trap and which is nonfunctional. Sometimes they seem more interested in the stick, but eventually they will get around to taking the food.

(Click on image for a larger version)

So how did they do? In the first group of four rooks, as you can see here, we had three that would learn this very quickly – or eventually, for one of them – and one that failed. This is if they learned the first trap, Trap A. Then we gave them Trap B, and they transferred their knowledge very rapidly, within the first 10 trials. We then gave them a re-test, a memory test of the original trap, and again they seemed to remember for a certain time, for a few weeks or months afterwards.

(Click on image for a larger version)

In a second group, which now had the second trap, Trap B, all four birds learned this and again very rapidly, over about five, six or seven blocks of trials. (Each block is 10 trials.) We then gave them the second, opposite, trap and again they very rapidly transferred the information they had learned in the first session, and they also retained the information of the original trap.

So, basically, seven out of eight rooks passed both tubes A and B, transferred immediately to the other tube, and remembered the original trap when they were re-tested.

(Click on image for a larger version)

One problem with this is that are still simpler explanations for how the rooks were performing on this experiment. They can’t really use the rule of pulling away from a trap, or a traplike object, but they can discriminate between, say, a black disc at the bottom and nothing, as in example C, or, as in the case of example D, two black discs in different positions. This is a more subtle discrimination, but discrimination nonetheless.

So we decided to produce two more tubes. We had two rewarded traps, or previously positive traps. Trap C now had a lid on the top; Trap D had the lid removed. In the previous experiments, both of those traps had been equally rewarded, so they were each as positive as the other. What we did then was to make one of those traps functional again, available to trap food. To one of them we added two rubber bands, one on each side of the trap, so that pulling across would not be successful because the food would become trapped. In the second version we lowered the apparatus to a platform, so now pulling across the open-bottomed trap won’t work – the food will get trapped on the base of the apparatus. Remember, those had previously been rewarded, but now the contingencies have changed.

What we found is that out of the rooks that were tested we have one star bird. On both of these tubes, on the first trial they performed correctly. And they were successful – in one case, in nine out of 10 trials, and in another case, in 10 out of 10 trials – on these new tubes in which they had never experienced these contingencies before.

That is perhaps good evidence, positive evidence, for causal reasoning in rooks.

(Click on image for a larger version)

So what about apes? We know that apes are not very good at the original task. In fact, we tried them on the two-trap tube test and they weren’t particularly good – most of the time they tore the tube apart and broke it. So we had to design a new task that was chimp-proof, which we called the two-trap box task. This was designed to test the same principles as our rook version.

We had one version here, Trap A, where again we had a trap and the chimp would have to move the food across a platform to an exit.

We also had another version, Trap B, again on similar principles. They could pull the food to one side and it would fall down and become accessible, whereas if it was pulled the other way it would fall into a trap.

[A video was shown, with the following commentary.] We made this version a non tool-using version for the chimpanzees, so they had to move the food along with their fingertips, as I hope you can see in this video. This is Annette, one of our chimp subjects.

There are a series of holes; the food comes down to where one opening has the trap and the other is free. She has now been able to successfully get the food that has fallen down.

So, instead of a tool in this problem, they actually can get to manipulate and get feedback almost directly from the way that they manipulate the food in the task.

(Click on image for a larger version)

We performed the same kind of design as with the rooks, where some subjects had the first trap box, and all four subjects were successful on this one. Fewer were successful on the second version, and then two of the original chimps from the first box learned or remembered how to solve this task.

(Click on image for a larger version)

In a second group, all four again solved box B, four solved Trap A, and then some also, again, retained this information over time. So, basically, eight out of eight chimps solved Trap A, six out of eight solved Trap B, and five out of eight remembered the solution on a re-test. Those were very similar results to the results with the rooks.

(Click on image for a larger version)

In a similar way we also wanted to see how the chimpanzees could transfer the information they had learned from these boxes as a test of causal reasoning, rather than a test of visual discrimination learning.

In this case, box C now had two colourful barriers added, one that was relatively functional (it was located next to the trap) and one that the chimps could still pull towards but gain access to the food. In D the two colours were irrelevant, because they needed to pull toward the right. If they pulled toward the left, the food would still fall into the trap.

(Click on image for a larger version)

With the chimpanzees we had only one subject that could be successful on the first 10 trials, on the first trial of both C and D. One other chimp performed well on D, and another one on C. As with the rooks, there was only one subject – in this case one chimp – that could transfer information to both boxes. So again in both species only one subject was really able to pass all tests that we could suggest as being evidence of causal reasoning.

(Click on image for a larger version)

Chimpanzees are tool users, so we can build an appropriate scenario for why they should actually have this causal reasoning ability. But what about non tool-using rooks? They don’t use tools in the wild; there is no evidence that they use tools. So why would this psychological process be important to them?

You don’t get rooks in Australia – I think they were introduced into New Zealand – but one thing that maybe separates them is the fact that they are incredibly social animals and they demonstrate great flexibility in their social lives. You see here a representation of rook social organisation across a year, where different types of social organisation occur.

Rooks are a lifelong monogamous species. They spend the entirety of their existence within a pair, and in fact a lot of work in our lab has looked at the social behaviour of rooks and found that in many cases they are very primate-like in the way that they organise their social existence.

In the breeding season they are very colonial – they form very large colonies, with large interactions between individuals and with the pairs. Later on they form families, and their offspring stay with them for a short time. And then in the winter they can form very large winter flocks, sometimes of tens of thousands of individuals.

We don’t know what is going on in these large flocks, but there is at least the suggestion that social pressures and the processing of social information have given them some knock-on effect in processing information about their non-social worlds, their physical worlds.

[A video was shown, with the following commentary.] Hopefully, what I have been saying will come across in this rather cute video from a BBC program, ‘Britain’s Cleverest Animal’. This shows perhaps the flexibility of the rooks’ behaviour, in foraging or invading our environments, which is related quite strongly to their social ability as well.

This is from a motorway service station in England which has lots of rubbish bins that have black plastic bags lining them. They are filled with lots of goodies for rooks, lots of human-discarded food. So what the rooks have been shown to do, gradually over time, is to forage on these bins by pulling up the inner plastic bags to get access to the food. This is quite a well-known ability in a lot of birds – they will pull up strings in stages and hold on to the string with their foot. But what perhaps is most interesting with these birds in particular is that they will perform this in pairs, but rather than grab one piece of food and then fly away, they will actually sit there at the top of the bins and throw as much of this rubbish out onto the street as possible, for their compadres to feast on later. Of course, this really irritates the cleaners of the service station.

I think what this shows is that there is some flexibility in the way that the rooks forage – and you could possibly see the importance of causal reasoning in this sort of task – but also that they are very social and they are cooperating in these sorts of teams.

(Click on image for a larger version)

I would just like to thank a few people. The majority of this work is done in collaboration with Nicky Clayton; the chimp and rook stuff was done with Amanda Seed, and Sabine Tebbich did some of the rook work as well; and our ape work is done with Josep Call and the Max Planck Institute for Evolutionary Anthropology, in Leipzig, and the Leipzig Zoo. And then there are all of these people who give me money for some reason.

Discussion

Question 1: To what extent do you think the causal reasoning that the rooks are displaying is domain-specific? Is there any data to suggest that they might be able to do the same kind of reasoning in non-spatial domains? The reasoning task that they are set, the same as for chimpanzees, is processing spatial information.

Nathan Emery: I’m not sure it is a spatial problem, in a way. We change the sides on which the trap is placed and where the food is located, so it is not really a particularly spatial problem. Why I think this is non domain-specific is that they don’t use tools, and obviously the suggestion is that you are only going to get physical reasoning, physical cognition, in a species that actually requires something in a physical domain, meaning a tool user. So in a way we think this is actually very good evidence for non domain-specific cognition.

Question 1 (cont.): It could be that in their normal foraging there are situations in which they need to solve problems of that kind, so that is what has driven the evolution of this.

Nathan Emery: They are caching animals, and particularly striking as a caching animal. Whereas most caching birds will have the food in their beak and then probe that deep into the ground, rooks will place the food on the ground and then actually dig a very deep hole and make a big song and dance about caching. Perhaps they need to understand something about gravity – or there are lots of cognitive abilities that you could suggest for just hiding food and retrieving it that might be related to this sort of task.

Question 2: The nest-building must be a spatially related activity.

Nathan Emery: Yes.

Question 2 (cont.): Is this cognition necessary for nest-building, or is that something separate?

Nathan Emery: That is a very good question, and I think it is something that hasn’t really been addressed by comparative psychologists. Rooks go back to the same nest each year, so there is obviously a spatial ability to find the same nest. But also they repair their nests if they are damaged; obviously, there is a great cost to having a bad nest. There are some birds – pigeons, for example – that are notorious for making very poor nests. They will lay their eggs, which will fall straight through the nest. Nest-building has always been seen as an instinctual behaviour but I think there has to be some sort of learning involved in making a good nest. If you are 20 metres up on the top of a tree that is swaying a lot in high winds, you want a nest that is not flung off; you want one that stays there and is actually going to keep your offspring safe. I think nest-building and cognition is a very interesting area that nobody has really looked at before.

Question 2 (cont.): Pulling twigs in, stealing things would seem to me something which rooks ought to be good at.

Nathan Emery: Yes, and putting them in together so that they are actually going to form a bridge, definitely.

Question 3: The young birds must learn from their parents, or get feedback by themselves.

Nathan Emery: This is a completely a man-made, novel task that we are giving them, so we don’t know about the effects of social learning in this task. It is something we want to do.

As regards other forms of their behaviour, with the bin anecdote we don’t know how that information is transferred. We know something about social learning, but not from parents to offspring. Russell Gray will talk about how New Caledonian crows develop tool use from observing their parents, but in rooks we don’t know anything. And these are all hand-raised birds, so in a way we don’t know anything much about how they would interact with their parents anyway, because we are mum and dad.

Question 4: An emerging picture seems to be, with both primates and corvids, one of striking individual variation in some of the more complex cognition tasks. Why do you think this is? What is driving the big individual differences?

Nathan Emery: The short answer is that I don’t know. The second answer is that I am sure that, as with us, there are differences in temperament or attraction to particular objects – neophobia, being scared of new things, or neophilia, being attracted to them – something during development, obviously. The close association with a large social group perhaps gives you more opportunities to learn things. How that then affects your individual differences I think it is too early to say, but I think it is largely to do with temperament and how you approach the world. But that is a biological trait, a biological phenomenon, rather than a directly cognitive one.

The fact that we have one individual in both species that performs this does show that there is the capacity, but obviously we are not bringing it out in the rest of the individuals that we test. It is not that they are all dumb and this one is particularly smart; I think it is just that there is something more that is not necessarily cognitive. There might be some biological issues. But it is too early to say, really.

Question 5: My question is whether you have recognised in your work the fact that the rook has been in direct opposition to the cognition of humans. Rooks are regarded as vermin. I am a farmer’s son from England, and I made it my challenge when I was a small boy to kill every rook I could reach. The spatial problem they had to solve was building nests that I couldn’t get to. They also had to get away from anybody who carried a gun, and they can detect the difference between somebody who carries a stick and somebody who carries a gun. My question is whether you have taken this into account.

Nathan Emery: Yes, very much so. Obviously, there is another challenge, in terms of our being able to investigate their cognition, that we have to be competing with them to find the right experiments to get it out. But we haven’t looked at specific questions related to interactions with man. It is certainly something that we could do. Certainly nobody has really looked at the discrimination between sticks and guns as an experiment, and I think it’s quite a nice one. But if you are interacting very closely with an animal that you have hand-raised, you don’t want to introduce something that means they are suddenly not going to go anywhere near you. But I think that in terms of discriminations it makes a lot of sense that they are very much in tune with very subtle differences between things in the world.