Whirr2007-03-26
Comparitive Cognition Data Dump

I was lucky enough to be in charge of co÷rdinating a workshop on Comparative Cognition today. Two of the foremost experts in non-human primate cognition--Mike Tomasello and Josep Call, from the Max Planck Institute in Leipzig--were lead the day-long discussion, and I got to sit in on it.

These people are awesome, they are both extremely smart scientists, and they have extraordinarily clever methodology. They also have access to a large number (although not that large, unfortunately) of apes--as well as dogs and cats and children--and fantastic experimental conditions. Tomasello says that he really gets bored debating methodological questions, so if someone object to one of his studies he simply designs a new one, with a completely different methodology, to test for the same thing. They often have several very different experiments to test the same idea.

In part, this is to compensate for a lack of experiment participants. Especially for exotic participants, such as chimpanzees that who were raised by humans (I think there are precisely three of these) it just isn't very easy to get access to apes to do these tasks.

The format was thus: Tomasello gave a three-hour talk about his research in social cognition, with frequent questions, tangents and discussions highly encouraged. The audience consisted of thirty professors and grad students (and me).

I'd read some of the work that Call was discussing, so I was more interested in the first half, about social cognition. Also, while I do think that the question of whether infants and chimpanzees innately prefer egocentric or allocentric frames of spatial reference, I'm more easily excited by the question of whether they possess a true theory of mind.

So here's a data dump rehashing of the coolest stuff that they talked about, and it's a little heavy on the Tomasello. And by "data dump" I mean that it might be a bit boring--it's just my notes about the cool-sounding experiments.

Also, and I doubt that this matters but I am publishing this in a public forum so I should say the following: a lot of the info in this informal talk was based on unpublished (or soon-to-be published, or submitted) work. In addition, I'm a rank layperson, and am easily confused by terminology, and might well of misheard or misremembered the exact details. So, if it looks like Mike Tomasello is claiming something truly outrageous here, it's more likely that I've just gotten it wrong.


Apropos of the conversation I had with Katya last night about whether my cat is lying to me about when she's been fed at night, Tomasello mentioned deception.

Tomasello says that you need a system of "cooperative communication", a Gricean structure, before you can have true lying. You can deceive others in three ways: inhibiting a natural response, hiding your actions, and actually lying. Chimps can inhibit their responses, so they might wait before trying to get a banana until a dominate chimpanzee left the room or something. And Tomasello told us about a study where a human would steal the banana if he saw the chimpanzee coming, so the chimpanzee had to hid behind the opaque walls, reach through an opaque tunnel, avoid a noisy door, &c. to snatch the food. He hasn't seen any examples of chimps misdirecting someone, however ( e.g., indicating that the food is over there, when it's really over here). He argues that such deception only works if you expect communication to be trustworthy in general--chimps in the wild never tell other chimps where to look for food (much better to leave it hidden and eat it yourself, later) and so it would never occur to them to pretend to do so. And if someone did try to tell a chimpanzee (truthfully or not) where the food is, they simply ignore it.

Tomasello described a test in which the chimpanzee doesn't notice a food source, until the experimenter points at it. Then the chimp follows their gaze, looks where they're pointing, sees the food and eats it. However, if the food is hidden in a bucket, the chimp looks for the food at chance, no matter whether the experimenter (or other chimpanzees, either) is pointing at the bucket--it simply never occurs to the chimpanzee that someone would be trying to communicate the location of hidden food, when they could be saving it for themselves.

Similarly, chimps could either push a button that would give them a treat, or push a different button that would give a treat to them and also to their neighbor in the cage next door. They consistently choose the first button, even though it costs nothing for them to help out their neighbor.

Which is interesting, because he also talked about a series of altruism experiments, and apes seemed interesting in assisting others, even without a reward. I'll discuss the human ones first, and then I'll describe the chimpanzee tasks.

Children from under a year old (I think) will "help out" experimenters on their own, for no reward. So in the video he showed, the experimenter "accidentally" dropped a spoon through a small hole, into a box, and then was unable to reach in to get it out. The kid (who had been allowed to play with the box previously) walked across the room, opened it, and handed the spoon to the experimenter.

They did a bunch of experiments both with and without a reward. Apparently, if the kids were rewarded, they would perform the task at ceiling until they stopped getting rewarded, and then would fall off. But, if they were never rewarded they would stay at ceiling the whole time, always retrieving the dropped objects. They even tried to make it more difficult for them to help out, by making them walk across pillows in order to get to the experimenter, for example. In fact, sometimes the child had to stop playing with a fun toy in order to help out, but they still did. And, as I say, they were just as likely to compete the task with no praise, and no attention--in the video, the kid opens the box and hands the experimenter the spoon, and he takes it without looking up and just continues to stir his coffee--there's no "Thank you", no eye contact at all--no reward.

He was able to get statistically significant results on similar tasks with chimpanzees, as well (although the sample size of three is a little sketchy--he was very candid about that, and talked about the counterbalancing measures they'd tried...). If a block was "accidentally" dropped, they would pick it up (and, apparently, spend a while sniffing it and chewing on it) and then return it to the experimenter.

Another task involved one chimp who was locked out of a cage--this chimp knew that there was food inside, so he wanted to get in. A second chimp (and they took great pains to ensure that this chimp did not know about the food, so that there was no thought of reward) would unlock (which involved pulling a peg free) the door to let the first one in.

Another cool feature was that children and chimps were much more likely to help out if the object was accidentally lost. If the experimenter threw the pencil on the ground or deliberately dropped the spoon in the hole, the participants (children and, I think, chimpanzees) did not retrieve it.

Another neat experiment involved the chimpanzees completing a task for the human. They did a lot of studies to determine things that chimpanzees are not very likely to do--for example, they really like to take things apart, but they are not very likely to stack or construct things. So they had a pipe screwed into a board, and the experimenter tried to put a cup over the pipe. She tried it several times, but couldn't manage to get the cup to sit on the pipe (even though, in fact, the cup was perfectly large enough). Then she passed the board and the cup to the chimp, who instantly (and, to my eye, a little smugly) fit the cup on the pipe. In the control experiment (where the human did not try anything with the cup or pipe) the chimpanzees never put the cup on the pipe.

Why would a chimpanzee unlock a door, or place a pipe or retrieve an object without a reward, but not point out food or push the button that gave a treat to another chimp? One posited explanation is that chimps were willing to be altruistic, but only if they received a request for help--the other monkey rattled the cage door, "asking" to be let in, the human made "oh no!" noises when they dropped the piece or failed to put the cup on correctly.

They also did a bunch of tasks looking at whether chimps (and young children) understand goals. They used babies, (from six months up) and the experimenter would place a toy on her side of the table, look at it, and then pass it to the baby. After a few toys, the experimenter looks at the toy, but this time she does not pass it to the baby. The question is what goal they think the experiment has, does she want to hand over the toy, but finds herself unable to do so, or does she want to keep it for herself?

In cases where the experimenter seemed unwilling to pass the toy (they would either look at it passively, or they would tease the baby with it, or they would play with it themselves) the baby would get agitated, would reach for it, and would turn to its mother for help. In the other cases the experimenter seemed unable to pass the toy. In one case, the toy was at the bottom of a tall glass container, and the experimenter tried to reach it but couldn't; in another the experimenter kept dropping the toy, so that it rolled back away from the baby; and in a third the experimenter was distracted--talking on the telephone--and couldn't pass the toy over. If it seemed that the experimenter wanted to give the toy to the baby but couldn't, the child was fairly quiet and waited patiently.

They did this all with chimpanzees and food, too. In the video, the chimpanzee kept trying to help this poor, bumbling experimenter, who kept dropping the food before the chimp could get it. When, instead, he would offer the food and then pull it back, the chimpanzee shortly got frustrated, banged on the glass, and then just left.

They did some more theory of mind tests--knowledge vs. ignorance, and false belief. So two chimpanzees face each other, with a barrier in between them. There are three buckets--one which both can see, and two that only chimp A can see. The experimenter puts a banana in the visible bucket, and one in the hidden bucket, and then the chimps take turns choosing buckets and eating whatever is inside. Chimp B goes first, and his choice is entirely hidden from chimp A. When it's time for A to choose he knows that B has probably chosen the bucket that in which they both saw the banana being placed, and so he chooses the hidden bucket. In other words, chimp A understands chimp B's knowledge of the correct bucket and ignorance of the hidden one.

But what if chimp A gets to choose first? The smart move is to eat the non-hidden banana, so that chimp B will have to guess, and then maybe chimp A will be able to choose the hidden banana and get twice as much food. Chimps can't do this, but children (six-year-olds, I think) can.

To test false belief, they showed both chimps where the first banana was, but then showed only chimp A that they were moving the bucket. Humans know that person B thinks the banana is still in the original place, and make their choice accordingly. Chimpanzees do not seem to be able to do this--although there are a lot of confounding factors. Tomasello repeatedly cautioned against drawing conclusions from negative responses.

There was another neat experiment in which a dominant chimpanzee and a sub-dominant were competing for bananas. They both see a room with a banana in an obvious location, but there is a hidden banana that only the sub-dominant chimp can see. The sub-dominant chimp is let in first--he knows that the dominant chimp is coming right behind him, and he knows that there will be a fight for the obvious banana. He also knows that the hidden banana will be uncontested, so he goes for that one first.



Call had a series of interesting experiments about logical inference. Two cups--one has a treat in it, and one is empty. The experimenter either 1) lifts the empty cup, but rattles the treat cup, 2) lifts the treat cup but shakes the empty cup, or 3) lifts both cups. If one cup rattles, the chimps choose that one (as do dogs). If neither cup is shaken, they choose at random. If one cup has a treat, and the empty one is shaken, the chimps know to choose the other one (dogs don't--they're distracted by the shaking cup and choose that one, regardless).

Two cups are placed on a seesaw, and 100% of the apes chose the bottom cup--they knew that the cup with a treat would be heavier, and that it would weigh more. For the control for this, they used a wedge-shaped platform instead of a see-saw--to find out whether apes just preferred cups on the bottom. But no, 50% of them chose the bottom (correct) cup, but 50% of them chose the cup on the top. Many folks pointed out that this was probably the only time in his career that he would get exactly perfect results like that. (There were 8 participants in the study, which is certainly more numerous than other studies they discussed).

Another interesting study--a grape was either hidden beneath a flat board, or beneath a wedge-shaped board. The apes could see that one board was propped up at 45 degrees by the grape while the other was flat, so they chose the correct one. The wedge-shapes are already at 45 degree angles, so they chose at random.

The really interesting bit, though, is this: when offered a choice between carrots and bananas, they always choose bananas. Apparently, chimpanzees really do love bananas, and they would even choose a very small banana in preference to a very large carrot. However, if a very small banana and a very large carrot were hidden under flat pieces of wood, both would be propped up but the large carrot would propped up the wood at a higher angle. Personally, I'd want to pick the flatter wood, because it would have the smaller (but tastier) treat. Chimpanzees don't, though, they always choose the carrot.

Let's see... he did some really cool stuff with tool use, as well. Offered a choice between a plastic stick, a paper stick and a piece of rope (the task at hand involved pulling a banana-on-a-string up to their cage) the chimpanzees know that the rope is useless. They usually avoid the paper, too (although one of them picked up both the plastic and the paper, and then ate the paper one).

In another test, the chimpanzee was given a long tube with a peanut at the bottom. They first tried to rip the tube off the wall, then they tried to break it, and then they tried to eat it. After 300 seconds, the chimpanzee took a mouthful of water of spat it into the tube. A few seconds later they took another mouthful, and then a final one a few seconds later, at which point the peanut was floating at the top of the tube. That was trial one, in all subsequent trials the chimpanzees didn't bother with the eating or breaking, they filled the tube with water after only a few seconds. In other words, they had learned how to solved the task, and didn't bother with stuff that they knew wouldn't help. The plan is to test them again in five year (or, more likely, one year). Call predicts that they will remember the solution, and will use it quickly.

In conclusion, monkeys are very cool. Sometimes they are smart, and they are maybe more smart than you might think. They aren't very smart at all about other things, though. Some of them are shaggy.

Comments