Erica Cartmill on Animal Minds, Play and Teasing, and Comparative Cognition

Erica Cartmill

Thanks so much for having me, Sean. It's great to be here.

Erica Cartmill

Yes, you're right. Moving on. Okay, I'm just kidding. No, I think that's exactly right. There used to be this model of intelligence, but also just of evolution, called the scala naturae. The idea was that humans were on top and other animals could be arranged working up towards the perfection that was human beings. What science has done over the last many decades is try to break that apart and say it's not a linear progression towards humans. Humans are equally as evolved as every other species. The things humans have evolved to be good at help us fit into the social and ecological niches we have, but the same is true of every other species. You wouldn't expect humans to fly or echolocate or see in the dark as well as some other species. So why do we still think about cognitive capacities as a linear progression?

The way I like to think about it is that what humans have is a unique constellation of abilities. Some of those abilities are more developed than in other species, some less developed. But what we really have is this unique configuration that pulls them together in a way that helps them build on each other, and gives us something powerful that allows us to do all the things we do as a species. It just makes it hard to have a nice tidy narrative about our superiority.

Erica Cartmill

It's tricky. I started out studying language, and that was always the bulwark of humanity: humans have language, other animals don't. But when you start breaking down what we mean by language — is it syntax, reference, compositionality, the ability to put together small units into bigger units with different meanings? — for each specific feature of language you can find at least one other species that has that ability. What we can't find, and I don't think we will find, is another species that has all of those abilities. That's what I really mean when I say humans have a unique constellation: we put together all the pieces in a way that gives us something that might in fact be qualitatively different. But it isn't a single switch that gets flipped that says, "We have language, other animals don't." Other animals have pieces. They might have one thing better than we do, or partial pieces of other things. We just have all the pieces fitting together into this puzzle.

Erica Cartmill

This experiment was done in Kyoto at the primate institute. A chimpanzee named Ayumu was really the first one trained to do this. She's trained to sequence numbers. The eventual version of the task is that the numbers one through nine flash up on the screen. She sees them, and as soon as she presses the first one, they all disappear — they get masked, so she can't see where they are. So you have to look very quickly, form a mental image of where they are and the order in which you need to touch them, and keep that in what's called your working memory, your short-term memory that allows you to continue a task you're in the middle of.

Ayumu proved very good at this task, much better than untrained human adults. We run this task on the computer screens in the chimpanzee exhibit at the Indianapolis Zoo. Chris Martin, the research director there who had worked with Ayumu, installed these for zoo visitors, and it's very humbling for them. They see a chimp doing it and think, "Oh, I can beat the chimp." A human has never beaten Ayumu. However, if you take someone — it doesn't even have to be a young child — and slowly train them up, you can become an expert at this. So really it's an unfair comparison, because we're taking Ayumu, a chimpanzee expert, and having her compete against human novices.

One thing that's really interesting from a cognitive perspective is that numerosity, the ability to learn, understand, and manipulate numbers, seems to differ in particular ways between humans and other great apes. Chimpanzees can learn to recognise and sequence numbers, one through nine, and Ayumu can now go up into the teens. But teaching her a new number is just as difficult as teaching her the last one. Learning three is just as hard as learning two, learning five just as hard as learning four. That's very different from what happens with humans. In human children there's a slow, laborious period of learning one, two, and three, but then once they learn four or maybe five, they suddenly get it and realise that each new number is exactly one higher than the number before it. It's called learning the cardinal number principle. Understanding that underlying pattern and being able to apply it to new numbers is something humans figure out and have an aha moment about, and that chimpanzees don't seem to learn in the same way, even though when trained on each number they become incredibly proficient at recognising and sequencing them.

Erica Cartmill

It's interesting. This is not my particular area of expertise, but my understanding is that a number of different species can — and by species I'm including mostly mammals, but sometimes birds, and also prelinguistic human children. When we're talking about the kinds of tasks you run with animals, they're nonverbal, and we run very similar tasks with young kids. The tasks that measure addition and subtraction are not "four plus five equals nine" with manipulated symbols. It would be showing a quantity — say you have four balls and they go behind a screen, and then five balls go behind the same screen, and then the screen is raised. What you want to show in a baby or an orangutan watching this is that if there are only three balls behind the screen, it should be surprising, because that's a violation of the additive principle. And that is what you often see in the species that have been tested. There are other differences too — how far apart the numbers are before animals will recognise them seems to matter. Without being able to explicitly count, recognising seven versus eight is very difficult, because you have what's called a large approximate number system, and that's true both for young human children and for other species.

Erica Cartmill

I think that's fair. It might be that some great apes who have been symbol-trained can do simple arithmetic manipulating symbols, but as far as I know, that would be a real exception.

Erica Cartmill

It's difficult, because in some ways it is just storytelling. We can generate lots of hypotheses about why humans are good at particular things, why we have language, why we have particular forms of language. Ultimately it's very difficult to test those. We can ask things like: is the structure of human language linked to our ability to see hierarchy in the steps of producing a tool? There are theories about tool production and material culture, that it has nested hierarchies — you have to do this before you do this, but you have to keep the end goal in mind in order to structure the intermediate goals. That's very similar to structuring a sentence. So there are theories dating the evolution of fully structured human language, maybe connected to the emergence of certain kinds of stone tools.

Apes can make tools, but they can't make the very complex Acheulean stone tools, where it's not just flaking off a single edge but flaking a shape to make a prototype base, then making secondary flakes to give it a sharp edge on one side and not the other — modifications on top of modifications. That kind of future thinking and planning is a marker that some people think indicates a species which could also structure complete sentences. But at the end of the day, all we can really do is ask whether other great ape species alive today can understand and appreciate structural differences in human sentences. There are studies teaching great apes language and asking whether they understand the difference between "the dog chased the cat" and "the cat chased the dog." You can also ask whether they can plan for future events and order intermediate steps. That's really where a lot of the work on comparative cognition sits: not just what these species do in the wild today, but what their underlying cognitive abilities are when they're put in a similar situation with access to the same substrates and motivations. Is there structure to the way they think?

Erica Cartmill

Just as with comparing humans to other species, comparing across the great apes yields a picture of different areas of excellence. It used to be thought that chimpanzees and bonobos, the most closely related to humans, were maybe the most sophisticated cognitively. What they are is very good at particular kinds of cognitive tasks. They're the most social of the great apes, living in large multi-male, multi-female groups as humans do. And just like us, they're interested in gossip. They pay very close attention to who's doing what with whom, and when. Anything to do with others' motivations, or what information others may be keeping from you, are things chimpanzees and bonobos excel at, because those matter in their natural environments.

Orangutans, which is the species I started out studying, spend the vast majority of their time in a semi-solitary state. Males in the wild are solitary; they have large home ranges and look for females to mate with, but they don't spend social time with others outside of courting. A female will have an infant who grows into a juvenile, and they travel together for 10 to 12 years, and then when the mother has a new infant the older offspring is displaced and goes off to find their own way in life. They go to college, if you will.

Erica Cartmill

So that kind of Machiavellian intelligence — thinking about who's plotting against me, who I need to make friends with to overthrow the king — is not an area where orangutans need to use those skills in their natural environment. They are social; they can pay attention to another's goals. But they're not as intensely obsessed with others as bonobos and chimpanzees. Where orangutans really seem to excel is in material culture and material manipulation. They're very good at figuring out manual tasks.

Having worked mostly with orangutans, and now with bonobos and chimpanzees, I've seen they have very different levels of energy and very different personality types. If you give an orangutan a problem to solve, they'll sit and stare at it for a long time. They might chew on something, look at it from the side. Then eventually they'll get up, go over, and maybe solve it on the first try, maybe not. They give it some space, then give it a shot. Chimpanzees and bonobos, you present them with something and they immediately want to get their hands on it: what's this? I'm going to poke it, shove it, shake it, bang it up and down, look at me. Those differences in underlying motivation can lead to different conclusions about cognition, but they really have more to do with an orientation towards the world — maybe an orientation driven by your social world. If you're a chimpanzee or bonobo and there's an opportunity to get a piece of food, if you're not there immediately, someone's going to get to it before you. If you live in a big social world where you're not the one who gets to eat first, you'd better get in there when the getting's good. So the social and physical environments species have adapted to can really change the way they're inclined to engage with the tasks we give them.

Erica Cartmill

That's very fair.

Erica Cartmill

That's more or less what people have found. For a long time scientists argued humans were the only species that pointed, and particularly the only ape that pointed, because most comparisons about language are between humans and other apes — although there are interesting models in birds, dolphins, and whales. The pointing question changes depending on the environment the ape is in. When apes are in human care, in situations where they're prevented from reaching things due to physical barriers the humans put up, like in a zoo, they can pretty easily learn to indicate a thing that is out of reach that they'd like, and the human will give it to them.

Now, is that pointing? It certainly looks like it. There was a lot of nitty-gritty discussion about whether they're pointing with an index finger, or whether it's a modified reach. But sometimes a thing is really far away, so I'm not convinced they're just reaching — I do think they're pointing. And if you look across human societies, humans point in all kinds of ways. It's not just the extended index finger. There are places where you point with two fingers or a whole hand. People get trained in Disney parks not to point with a single finger because it's considered rude in some cultures. Some societies point with the chin, or by inclining the head — you might do the same if you were carrying something and someone asked you for directions.

Erica Cartmill

Absolutely. For all our not-great traits as a species — we have liars and cheaters — humans are fundamentally a cooperative species. We share information with others that doesn't benefit us, and sometimes disadvantages us. If someone asks what you're standing in line for, why would you tell them? That might be something you want for yourself. Or if someone asks where the best noodle place in Poughkeepsie is, why tell them, when you want more noodles for yourself? But we will stop and help people we're not related to, we share information freely, and that information is rarely a lie. People are capable of lying, but the average person in the average interaction is very trusting. That's one of the real hallmarks of our species, and it's one of the things that makes language work: both the speaker and the listener trust that the other is communicating in good faith. If you tell me something, I don't immediately say "prove it," because in order to talk about things that aren't in the here and now, you have to believe the other individual. This entire enterprise of podcasting, for example, is not something I think any other species could do.

That doesn't mean other species can't communicate at a distance. Chimpanzees and orangutans make long calls, lots of species make alarm calls, birds have territorial displays. Anything that's broadcast is out there for others to listen to, but it's often linked to specific information about the time, place, and signaller.

Erica Cartmill

This is a little off topic, but it's one of my absolute favourite studies, which I always teach in animal cognition. Song sparrows have very discrete territories. One of the neatest experiments shows that animals encode information not just about the type of call, but who is making it and from where. If you take a call made by the neighbour on the right-hand territory and play it back from the left, the bird in the middle will respond very strongly, because it's like, wait, Fred's in the wrong place — and they'll respond as if Fred's a stranger. But if you play Fred's call back from where it's supposed to be, they don't respond as strongly, because that's what they're used to hearing.

Erica Cartmill

Right. It's not just that animals recognise "this is an alarm" or "this is a food call." They're processing the information and comparing it to their memory of who is providing it and where it's coming from. So animal communication is both different from human language, but also more often more complex than we think it is.

Erica Cartmill

It differs a lot between species. More broadly across the animal kingdom, you have colony animals like meerkats or chimney swifts, where you have multiple generations in the same group. In some of those species there's a single breeding pair, and everyone else is helping, so it's similar to a eusocial insect like a honeybee with one queen. They're not as physiologically different as a queen bee, but the dominant female will suppress reproduction in the subordinate females.

In non-human primates you certainly have social hierarchies, depending on the species. Some have very strict dominance hierarchies in the males, the females, or both. In some species it's a mixed-gender hierarchy; in others all the males rank above all the females. You can have distinct rights and responsibilities afforded to the dominant individual — not just mating privileges. In some species the dominant male is allowed to mate with all the females, whereas subordinate males can only mate if the dominant male allows them — meaning he observes the male courting another female and doesn't attack him. So you can have alliances: if I'm the dominant male, maybe I let a couple of good friends mate with a few females, but maybe not my favourite female. Very complicated social rules.

What you don't have is a division of labour where some are collecting fruits and some are weaving baskets. You do have some division in terms of hunting. In species like chimpanzees that hunt, some individuals — usually males, but a few females — go on hunts together, usually for monkeys, and it tends to be the same individuals. Then they come back and share the meat with favoured individuals — their friends, or females they're courting. Meat is a precious, high-calorie resource, so it's used a bit like currency in species that hunt and food-share.

Erica Cartmill

Quite a bit. There are beautiful experiments, done originally by Robert Seyfarth and Dorothy Cheney at the University of Pennsylvania, where you can simulate hierarchy reversals by playing back simulated calls. Say you have a fight where one individual gives aggression calls, one screams victim screams, and afterwards does submissive vocalisations to show, yes, you're in charge, I yield. If you record those, you can cut them up so you have a call from individual A, an aggressive call, a call from B, and a call from C, in that order — A is dominant, B is in the middle, C is lower down.

Now, if you splice them together to create a simulated fight where C is attacking and beats A, and then play that recording back to other individuals in the group, they'll be very surprised. They'll look longer at the speaker, and then show more aggression to A afterwards, because this is a moment where they say, wait a second, A is vulnerable, and maybe this is my chance to improve my spot in the hierarchy. So they're paying attention not just to their own relationships with others, but to others' relationships to each other, because that's what you need to know if you want to get ahead. If you want mating opportunities, and you're male, it's important to make friends with other males who are going to be in power now or in the future. If you're female and want to make sure your offspring are protected from infanticide when a new male comes into power, you need strong relationships not just with other females who could come to your aid, but with males who may one day take over. It's very strategic.

Erica Cartmill

It's certainly something I've been fascinated with for a long time, and have been exploring over the last five years, trying to figure out whether great apes in particular have a sense of humour, an understanding of surprise, of these subversive moments. We've been working on an entire research line looking at humour through the lens of what I call playful teasing.

Teasing is a fascinating opportunity to look at the social minds of animals, because it's this grey area, right between aggression and play. When people have talked about teasing, and the way it's been studied in animals in the past, it's really been thought about as a form of aggression: I'm pestering someone, bothering them, trying to improve my stance in the group, a low-level form of aggression. But if you look at humans, we use teasing all the time in positive ways as well as negative ways. When people talk about teasing in humans it has a pretty bad rap, thought of as a precursor to bullying. Sometimes it is — I don't mean all teasing is positive. But some forms of teasing are actually very loving. They build relationships. They're critical in developing a complex understanding of how far another individual will go for you.

Erica Cartmill

It's hard when you just say "animals," Sean.

Erica Cartmill

I'll tell you, I think some animals have a sense of humour and some probably don't. For example, I'm doing some wildlife rehabilitation right now, and I have a possum who is unreleasable — an educational animal with a disability that means he can't be released. He's fascinating, and I love possums, but I do not think he has a sense of humour. He's very good at what he does — he kind of Pac-Mans around eating things, and he'll cuddle, he's very sweet — but I don't think there's a whole lot going on upstairs. Not to put animals on a ladder again, but that's not an ability that needed to evolve for his species to survive. Whereas we also have a cat and a dog, and I certainly think cats and dogs have a sense of humour in a way. They have things they know they're not supposed to do that they enjoy doing because it gets a reaction. That pushing of boundaries, those moments of something unexpected, a violation of a social norm or an explicit rule — I think those are at the heart of humour.

Erica Cartmill

It's a very valid question whether all humour at its heart is mean-spirited. When I started this project I thought my husband and his brother were very mean to each other. I'm an only child, and I grew up in a family that did not tease, or didn't tease to a great extent. My husband is one of five boys, and there was a lot of teasing in that family. When I saw him and his younger brother get together, it was just, how can I get under your skin until it really irritates you? And conducting this line of research in great apes has made me develop a much greater appreciation for the love that they have between them. I'm still not sure I'm going to fully participate in that level of "how can I irritate you?", but I do think it's fundamentally built on respect and trust.

Coming back to your cat: I'll look at you, and just push it a little closer, maybe I'll wait, oh, I can push a little more, are you coming closer? One of the things that started this whole project was that we were looking for three behaviours that a psychologist at the University of Portsmouth, Vasu Reddy, had identified in human infants. She called them not humour per se, but clowning behaviours — non-verbal forms of humour. The behaviours were: offer and withdraw, so, here, would you like this thing, ha, just kidding; disrupting others' activities, so you're doing something and I jump in the middle of it or make it impossible for you to continue; and then the one with my all-time favourite label, provocative non-compliance. I think cats are masters of provocative non-compliance.

But does that mean they have a sense of humour? I'd like to leave the door open for that, but not have a strong conclusion. I'd say it suggests they have an appreciation for violating others' rules in that way, but it needs systematic exploration in a scientifically objective way. Will they do it when someone is out of the room? Are they really looking at your response? Are they trying to get a particular kind of response? If you put yourself in the role of an experimenter and said, "When they do that, I'm going to turn and walk out of the room," is that something where they're suddenly not interested? Are they doing it in order to get a response from you?

Erica Cartmill

I do think this violation of expectation is really at the heart of what a joke is. As humans we push jokes to the extreme, telling long and complicated ones, but fundamentally a joke is setting up an expectation and then violating it in some way. That was the thing that initiated this entire project for me. I completed my dissertation studying gestural communication in orangutans — I went to zoos in the UK and Europe, looking at how orangutans used gesture to communicate with each other. During that study I witnessed an interaction I had no idea how to study from a gestural perspective, but it fascinated me and stayed with me, and I came back to it years later when I had the opportunity to apply for a grant in this space.

In this interaction there was an infant orangutan hanging from a rope over her mother, with a stick — a piece of bamboo — in her hand. She reached down and offered it towards her mother. The mother, lying on her back on the ground, reached up to grab the stick, and then the infant pulled it back out of reach, and the mother put her hand down. Then the infant did it again, and the mother reached, and the infant pulled it back. At first I thought, oh, this is a cute little game, and the mother is tolerating the infant. But what I found even more interesting was that at some point the infant dropped the stick and seemed to get tired of the game, and then the mother picked it up and started doing it back to the infant.

To me that moment, that role reversal, demonstrated that it wasn't an idiosyncratic routine that just developed between that infant and that mother. It was a game. It had a structure. It had different roles that could be played. It obviously wasn't really violating an expectation, because if you understand what's going to happen, you know they're going to pull it back — the mother wasn't reaching fast to try to get it, she was playing along. But within each of those offer-and-withdrawals it has the structure of a very basic joke: a setup, the offer, and then a punchline, the violation, which is the withdrawal. Once I started to think about it as a very simple joke, it made me want to look at where else apes might create these moments of expectation violation for each other. It seemed to be something they both enjoyed, both did voluntarily, and it seemed likely you might see it more broadly if only you knew what to look for.

Erica Cartmill

Absolutely. I've been thinking a lot about the function of these teasing behaviours. It's interesting just to document them and ask whether apes do them, what they look like, when they occur, and who's doing them. But ultimately we want to know why. Is this something that has a benefit to developing a place in a social world? Does it help you understand others' minds, or develop social relationships? One of the things we've been thinking about is that these playful teasings, these proto-jokes, might serve both to build relationships and to test them, because they have this element of "how far can I push you, or how far am I willing to be pushed?" Those are things that are not transferable across individuals. I can call a good friend a name, but I'm not going to say that to my boss. Those things can strengthen a relationship.

Think about early dating — a lot of flirtation is teasing. A little nudge, a little poke, testing the waters: do you laugh it off? Do we have that kind of relationship? If we do, you can go a little further. That's true of friendships, romantic relationships, family relationships. The way teasing can serve as a test of relationship strength is seeing how much the other individual values the relationship over their momentary annoyance. It's like, ugh, fine, but I love you, so I'm not going to be annoyed. When you look at parents and infants, they put up with all kinds of things they wouldn't from others. Look at anyone in a grocery line: if a young child does something to a parent, the parent puts up with a lot more than they would if an unrelated stranger did the same thing, because it's an honest signal of how much they value that relationship.

Erica Cartmill

Absolutely. One interesting thing about this social hypothesis-testing is that it isn't static. Once you learn the physical properties of an object, you know them; you don't have to keep testing. I don't think the same is true of social relationships, because those change in a way physical properties don't. Your relationship with your friend or your sibling today might be quite different from yesterday. It might be different because they're in a different mood, because you pushed them too far yesterday, or because something changed in their environment. So learning to predict their responses to your actions is critical in accurately anticipating and evaluating your relationship with them. I remember being a graduate student — some days you'd go into your advisor and they'd be in a great mood and say, "This is a great idea, let's go with it," and other days, "This is a horrible idea, back to the drawing board." You'd rely on the graduate students who had meetings before you to tell you whether it was a good day to pitch a new idea. Social relationships are a landscape that keeps changing over time. You don't just want to learn the baseline state, you want to improve your ability to predict, and teasing gives a low-stakes environment for practising the attribution of others' reactions to you.

Erica Cartmill

I don't think there's anything analogous to chess. But there are things analogous to tag, hide and seek, and some children's games. There's an appreciation for turn-taking that can occur with things like chase. Dogs do this all the time: I chase you, now it's your turn to chase me. And it isn't just that one dog gets tired and turns around. A dog will turn around, run a little way, look back, and if the other doesn't follow, they'll go forward, play-bow, bark, nudge them — a reminder, hey, you're supposed to be doing something, there's a game here. So there are these simple role reversals in social games that animals have.

It's also possible there can be non-social games. We've started a new project — I'm launching a canine cognition lab at IU, and our first study looks at dogs' collecting behaviour. It's a collaboration with a developmental psychologist, Martin Zetterstein, who has looked at this in children — how their collections of objects differ around the world and change over time. He's found that children basically everywhere collect at least one thing. I'm interested in whether any dogs exhibit similar behaviour: do dogs interested in a particular type of object manipulate, order, or store them in any way that has internal structure? Do they put them all in one place, rearrange them, put all the purple ones together?

I do think it might be possible that there are things we could consider essentially private games — things about setting arbitrary goals and trying to achieve them. My one example is a self-driven game from my dog Bonnie. She has a little stone turtle she's obsessed with. She actually has 11 stone turtles, but there's one in particular. They came from a garden store; we used to have them in our garden, and she loved to find them, pick them up, and bury them. There's one whose name is Darius, by the way, because every time we find him I go, "Darius!" It's really cheesy, but she likes it. In this game I roll the turtle up in towels and she has to dig through and find him. She loves this and will play it for hours. It brings her out of any bad mood, or if she's stressed after going to the vet.

What I find really interesting, and one of the things that leads me to think some animals might play games by themselves, is that she'll bury it for herself. If I've gotten tired or I'm on a Zoom call, she'll take it, put it on a towel, put another towel on top, smoosh it all up, sit there for a minute, and then try to find it again. I really think she's doing something similar to what young children do when they play. As much as you can call a game something like "I'm going to see how long I can stand on one leg" or "how many times can I spin around before I fall down" — we call these game-like play in children. They don't have structured, documented rules, but they're about setting goals for yourself that don't have any immediate benefit other than that it feels good to achieve them.

Erica Cartmill

To teasing, or to any kind of game?

Erica Cartmill

I think it feels good. One of the things we're trying to do in a different project that ties in is I'm working with a philosopher, Colin Allen, and a number of animal experimentalists and ethologists at institutions around the world, to look at joy in animals. We're looking across three very distantly related species: Kea parrots, dolphins, and great apes. What we're trying to do is develop better biometric measures of positive emotion, to look at when animals experience positive emotion, whether they communicate it to others, and how perceiving those communications impacts things like memory, attention, and pro-sociality.

With apes, we've been looking at laughter. First of all, humans are also great apes, so I've been saying "apes" just to mean non-human apes.

Erica Cartmill

Just making sure. Apes laugh just like humans do. Their laughter sounds a little different. Where human laughter is quite musical, chimpanzee and bonobo laughter is a little less musical — it still has a rhythmic panting element, but with less melodic quality. Gorillas and orangutans also laugh, but their laughter is even less musical and more breathy. All of the species laugh, and they laugh in very similar environments. It happens a lot during tickling, and a lot with infants — adults will tickle infants, and infants will laugh.

What we've been doing is using laughter as an auditory stimulus for a set of experiments, to ask how hearing another's laughter impacts the way you think about the future, interact with others, remember things, and perceive facial expressions. Generally we've found it seems to impact them both physiologically and cognitively. We've done some things with thermal imaging, looking at changes in blood flow by measuring skin temperature. In great apes we look for a dip in nasal temperature — the nose getting cold as an indication of emotional arousal. When you're excited, blood gets drawn into your core, so your periphery gets colder. It's a fight-or-flight reflex, and it can happen when you're excited and happy, but also when you're excited and angry. So we pair it with behavioural measures, because how do we know the animal isn't just scared, or pissed off, hearing the laughter and thinking, "That's annoying, I hate that guy"?

So we pair it with a cognitive bias test, which was developed originally for animal welfare, as a way of asking whether animals become more optimistic or pessimistic following changes to their enclosure. We applied this to great apes. In the test, you first train them on two anchor stimuli — in this case a black box and a white box. The black box was always rewarded, the white box never rewarded. They press a button, and then someone holds up a box at the other side of the enclosure, and the ape has to decide, do I want to expend energy to walk over and see if there's food in it? They learn that they should always go for the black boxes and never the white — they can just press the button to skip and get a new box. So it's skip, skip, skip, ooh, this one, skip, skip, skip, that one.

In the actual test, we throw in grey boxes they've never seen before, somewhere in between the black and the white. We use that as a measure of optimism: do they treat those intermediate boxes more like the positive one, or more like the negative one? What we found is that great apes — specifically bonobos — when they've listened to laughter, are more likely to approach the grey boxes than when they've listened to a control sound. So that's one behavioural piece of evidence suggesting that listening to laughter leads them to expect more positive things.

Erica Cartmill

Absolutely. It's like a laugh track. They listen to it for a while, and it's not providing any useful information. We're interested in whether it's essentially inducing them into a better mood. If they're in a better mood, we think they should expect more positive things than if they're in a — I don't want to say negative mood, because we're not trying to put them in a bad mood, we're just playing a very neutral environmental water sound.

Erica Cartmill

One of the things people have been interested in for a long time is play signalling — communicative signals that different species use to say, "Hey, now I'm playing." This has been looked at a lot in dogs and wolves. The play-bow in dogs is the classic play signal, and what's interesting is that it seems to be species-typical. It's not learned; even young puppies do it. But it's not so much the production of it, it's the interpretation. When a dog sees another dog play-bow, they're less likely to respond to that other dog's pouncing and biting in an aggressive manner. So this is an example of a behaviour that could be creating a frame, saying, "The thing I'm about to do shouldn't be taken seriously."

Now, is that pretence? Is that imagination? Not to the extent of a four-year-old saying, "This spoon is a sceptre and I'm a princess." But it is still marking something as different from the quotidian, from the way this behaviour would normally be interpreted. I think that's also happening in teasing: I'm doing something, but it shouldn't be blown out of proportion. If I steal something from you while teasing, it shouldn't be taken as seriously as if I steal it outside of a teasing interaction.

Erica Cartmill

Exactly. It's really difficult, and a fascinating area. I think that's where philosophy can play a big role, in asking what the burden of proof would be. In a non-verbal way, how can we differentiate these things? The difficulty in looking at non-human animals is that you can't ask them, well, what did Fred mean when he did this behaviour? Why did you respond differently to Fred than to Larry when they came over and tried to bite you? In humans, even young children, you can say, oh, I know Fred was just playing. You can't do that in animals. So we need creative ways of looking either at their spontaneous behaviours — how the social dynamics and actions in the interaction differ — or developing biometric markers of emotion, ideally things you can measure externally in free-ranging animals who are naturally interacting, like skin temperature or piloerection, the hair standing on end. There are things you can see behaviourally and externally that are direct responses to internal physiological processes.

I don't think there's going to be a single answer. These are complicated questions that rely on a multi-method approach — bringing together experiments about what happens when you're presented with a certain stimulus, observations of natural behaviour, and studies of social preference: do you prefer to interact with an individual you've seen teasing someone else, or not?

One thing we haven't looked at yet, but thinking about the origins of human humour: one of the things that should be more funny is bringing down those in power — punching up the hierarchy. I wouldn't be surprised if we saw that same thing in apes. If you watched a video where a dominant male slipped and fell in the mud, or fell off a branch, as opposed to a subordinate or juvenile — or you watch a juvenile teasing an adult male versus an adult male teasing a juvenile — I think it should be more interesting, and potentially funnier, to see the mighty brought low.

Erica Cartmill

In thinking about the comparative approach, the intuitions we have about what humour is like in humans, we obviously have to ask those questions in humans as well, to make sure we're not simply building on our assumptions based on our own past experience.

Erica Cartmill

This is potentially, to date myself, the $64,000 question. As someone who has built a career studying communication and comparative cognition, the AI question is a fascinating but also irritating one — and I say irritating tongue-in-cheek. AI can mean many different things, but for a lot of people, when it comes to asking what AI understands, what it can do, whether it can pass this or that psychological test, we're really talking about LLMs, things like ChatGPT.

I think we're in a critical moment where it could go one of two ways. For a long time, language was seen as the defining line between humans and other animals, and even today there are some kinds of questions that some scientists aren't happy with people asking in a non-verbal way. There's still debate about whether animals have things like episodic memory — can they remember themselves in the past, from a first-person, or first-animal, perspective? The way you ask about that in humans is through self-report. There are behavioural tasks where you can ask whether animals can use their memory of something in the past to plan for the future, and some people argue that demonstrates episodic memory, others argue it doesn't. Those arguments are difficult to resolve given that you can't ask most animals verbally. You can ask an LLM, and it will give you an answer, and it will often be the same answer a human would give you. But they also have access to essentially all of human language, all the documented data points and next-word predictions of human history on the current internet.

The big question is: does an LLM understand anything? This is something we've done some work on, advocating for applying more guidelines and methods from animal communication and animal cognition to the study of AI and LLMs specifically. A lot of the time, because of the methods we use to study cognition in humans, when we get verbal report from humans we assume it tells us something about what's going on inside. I don't think the same should be true for an LLM. We can't just take them at their word. What we need to do is develop much more thoughtful, systematic ways of probing their levels of understanding — by manipulating the way you ask questions, changing the modality or the situation you pose to them.

This is an area where comparative cognition researchers, people working in animal behaviour and cognition, and developmental psychologists have a lot of experience, because you can't just ask a one-year-old or a two-year-old, "Why did you think that?" You can't ask a magpie or a chimpanzee, "Why did you do this?" You can ask an LLM, and it will give you an answer, but the question is whether that answer maps onto why it did that. Coming back to trusting language, this is a question even with human adults: if you ask someone why they gave that answer, the reason we think we did something might not actually be the reason we did it. So even with human adults, you have to be somewhat careful in taking verbal self-report as the final word. To understand what's going on under the hood, we need to bring a multitude of approaches to bear.

Erica Cartmill

I think some of them are. I don't think any of them are as much as they should. This is really an area where I think we'll see more collaboration and development in the years to come, when people hit walls and want to understand why they're getting a certain error, why the model isn't doing the thing they expected because they thought it was doing it the way humans do. I don't think these are going to be problems you can engineer your way out of just by improving the algorithm. They're things where you really have to find ways of assessing what the LLM is telling you about how it's doing the thing.

Erica Cartmill

Thanks so much for asking. One of the things I'd love people to check out is that we're actually asking people to share their own experiences with animals teasing. This could be your dog or your cat, two crows over a field, or something you saw at the zoo. We're collecting any kind of animal experiences of teasing you might have witnessed. The website is observinganimals.org — we have several public science studies up there, and one is on teasing. If anyone listening has reflections about animals they've seen teasing, we'd love to hear about them.

More broadly, if people are interested in the work my lab is doing at IU, we're starting a new research centre called the Center for Possible Minds — that's possibleminds.org. We'd love to have folks reach out if they're interested in getting involved. This is a new initiative launched by myself and Jacob Foster, and we also run the Diverse Intelligences Summer Institute, which you can find out more about through our centre website.

Erica Cartmill

Thanks again, Sean.