Executive Vice President & President of Business Units, Analog Devices, Inc
Being Human | Greg Bryant | Why Do We Laugh?
Laughter is a universal human behavior. Have you ever wondered why we laugh or what it really means when we do? In this talk, Greg Bryant of UCLA explores the evolution of communication and vocal behavior, especially of spontaneous vocal expressions such as laughter.
About the speaker:
Greg Bryant received a PhD in cognitive psychology from UC Santa Cruz. He completed a postdoctoral fellowship at UCLA in biological anthropology and is currently an associate professor at UCLA in the Department of Communication Studies. He is interested broadly in the evolution of human vocal communication and...
About Gregory Bryant
Gregory Bryant, an associate professor at UCLA in the Department of Communication Studies, gave a talk in 2017 titled "Why Do We Laugh?" in which he discussed his research on the evolution of human vocal communication, focusing on laughter. He described laughter as a neuro-mechanical oscillation involving respiratory and laryngeal activity, an involuntary vocalization evolutionarily related to sounds made by other animals. Bryant stated that laughter functions as a play vocalization, similar to a dog's play bow, signaling non-threatening intent. He noted that humor involves implying things without stating them directly, and laughter can signal successful decryption of that implication. Bryant presented findings from cross-cultural research, stating that listeners across 17 societies performed above chance at distinguishing real from posed laughs, with an average accuracy of about 64%. He explained that spontaneous laughs have a higher proportion of unvoiced components, while volitional laughs are more speech-like with more voicing. Bryant also discussed laughter's role in signaling cooperative intent and building relationships, as well as its potential for social manipulation, describing it as a signal that can be used to gain trust or affiliation.
Transcript (45 segments)
G
Gregory Bryant0:12
Yes, it's my birthday. When I was asked to do this, I said, 'Well, that's my birthday.' And I thought, 'That's perfect. I'm going to do it on my birthday.' Thank you for coming. This is a cool experience for me because I never get to talk to people who've been drinking. I'm usually the one that's drunk before the talk. This is nice. Thank you to the Leakey Foundation and Arielle Johnson and everybody for coming.
All right, so I'm going to talk today about laughter. I got into studying laughter originally as a grad student. I did my dissertation on what kinds of vocal signals people use to communicate when they're being ironic. But one thing I noticed was that people laugh a lot. Later in my career, I just really took off with it. Now I've got more projects than I can count on laughter. I'm going to describe a few of them. I'm very interested in cross-cultural research and how evolutionary theory allows us to understand human vocalizations, which goes beyond a lot of the ways that traditional psychologists and linguists often think about vocal communication. I think that we're animals. Now, I know we're animals, actually, I don't just think we're animals.
Here's the example I put together... They're all kind of baring their teeth. There is what we call a phylogenetic relationship between all these animals. Evolution is conservative, meaning that not politically, it means that evolution conserved structure. It doesn't rebuild things very often, it always builds on things that already exist. So if you look at expressions across the animal kingdom, you see a lot of similarities and this is not a coincidence. Now I'm going to play a vocalization, an animal vocalization. See what you think. If I can get it to work.
Whale, bear, walrus? I liked walrus myself. Well, actually it's slowed down. Let me play it for you at normal speed. Those are 19 year old females from UC Santa Cruz laughing, not walruses. I added the ocean thing 'cause I'm tricky that way. I've been interested in slowing down animal sounds a long time, well before I became a scientist. It's partially due to my brother who's back here.
Like a few years ago, it had a way to control... it's called pitch control, just to control the speed. I was fascinated with this thing. I'm a musician, I'm interested in making experimental sounds. I also play regular instruments, but I'm really into weird sounds. I'm into sounds. One of the best features of this four track was to slow stuff down. I was just fascinated with this. In those days, you couldn't get stuff off the internet 'cause there was no internet. You had to actually buy sound effects CDs or records. I literally have sound effects LPs. I would feed them into this four track and then I would mess with the sounds.
It sounded like an animal, a non-human animal. But when you slow down somebody talking, which I also would try obviously, it did not sound like an animal, it sounded like a person slowed down. I always thought this was really interesting and I didn't understand why. I eventually came around to figuring out why and I'm going to explain that now. That was slowed down people laughing. But how many people thought it was actually human slowed down? Anybody get that? Smarty pants. Oh, you got a smart guy in the back, my brother. Most people don't get that. I actually have experimental evidence showing that people cannot tell whether a slowed down laugh is human or animal.
All right, so let's just start with the basic question, the most unfunny part of what a laugh is: it's a neuro-mechanical oscillation involving respiratory and laryngeal activity. That's my definition of a laugh. It is an involuntary vocalization that is evolutionarily related to vocalizations that many other animals produce. That's what makes it such an interesting vocalization for me to study as a person who's interested in evolution and human communication. It's really one of the most perfect vocalizations, along with crying, which I'm eventually going to get to, but I have not studied crying. Pain shrieks are another one. One of my favorite topics, one that got me on this laughter kick in the first place, was orgasms.
This is a spectrogram of a laugh. You have a representation of the acoustics here. Up here is just the overall amplitude. This is the same laugh but represented two ways. Overall amplitude here, which is just collapsing all the frequencies into one representation, so the bigger the thing is, the louder it is. Down here is basically breaking it up across the frequency spectrum. You can see where the darker areas mean there's more energy at those frequencies and lighter areas means there's less energy. This line here represents the amplitude curve, which correlates exactly to this. These black lines are what's called the fundamental frequency, which correlates to our perceptions of pitch. So when I say, ha, ha ha...
They're very closely related. Let me just play it for you. Not a funny laugh but that's a standard kind of laugh. You can see that laughs have certain characteristics. They have usually an initial onset that's louder than the rest and there's a slight decay in the volume. Then you have these calls where basically there's a glottal cycle. The glottis, which is in your larynx, opens and closes. When it closes, the air gets forced through and vibrates, that makes the tone. When it opens, just air goes through and you don't hear that as a tone.
It's hard to do on purpose. When you try to laugh volitionally, when you try and laugh on purpose, it's hard to actually get it right. Some people can and I'm going to show you data about that. Laughing is very variable in how it comes across. There are lots of different kinds of laughs. Let me just play you a collection of laughs I've recorded in the lab.
That's my favorite. I don't know for sure which are real or fake. Most of those are probably fairly real, but some of them are fake. There's a lot of variation in how laughs manifest themselves, which makes them difficult to study but it's kind of paradoxical and interesting. We all know when we hear a laugh, even though they can manifest themselves in all these different ways acoustically, we all know it when someone laughs. But if I take them out of context, then they're a little harder to tell if they're a laugh. I call it squeaker laugh 'cause it's like... and when you hear it in context, it totally sounds like a laugh, but out of context, it just sounds like a weird noise. That's a standard laugh.
Let me show you some laughs across different species. This was done by my friend Marina Deauville Ross, the late Michael Overren and LK Zimmerman. They looked at the acoustic features of laughs across different primate species and then they tried to reconstruct the phylogeny of the vocal behavior in primates. Many animals laugh. We call them laughs and that's probably an anthropomorphic way of describing them, but they are play vocalizations. They are vocalizations that animals produce when they are usually juveniles, when they're doing rough and tumble play. It's a way of telling the other animal, 'I'm not threatening you now, I'm playing.' They're usually quiet in most species, which is why they're really hard to record.
But we know rats do them. Rats produce ultrasonic play vocalizations and they sound kind of like laughs. When you bring them down into our frequency range, we can't hear them because they're like 40 kilohertz, which is well beyond our hearing. But they produce these and you can tickle rats and they love it. When they play, Google it. Put it in your notes right now. Google 'rats laughing', it'll come up. Jaak Panksepp has a terrarium with rats and he's tickling them. They follow his hand around and they're laughing up a storm. When they play with each other, they're laughing too, but you have to have a special recorder designed to record bat vocalizations in order to hear them.
Huh? They wag, it's called the play bow. That is a signal a dog makes to another dog saying, 'I'm now going to attack you but I don't mean it.' It often doesn't end well because it's not clear how the signal turns off. It's not like, 'We're playing now, I'm biting and attacking you, which normally would seem aggressive but we're playing.' Then you go a little too far, which is kind of the way kids are. Kids are playing, they're laughing and suddenly somebody's crying. Laughter is a play vocalization. I think it is functionally very similar to a play bow in a dog. It allows us to play and that doesn't mean this isn't going to hurt.
Here is I think the gorilla. You hear the similarity? Now chimps are an interesting thing. They do something unique: they do not just eggressive, meaning outward, but they go in and out. The idea here is that laughter is basically labored breathing during play. This is Robert Provine, who has written about laughter quite a bit. He calls this labored breathing during play. Originally, this is what animals would do when they're playing, which is a phenomenon that facilitates that evolutionary function. This is a fairly well understood and well described process in animal signal research. This is what's happening with laughter. It's becoming an exaggerated version of this labored breathing during play.
Finally, here's some human laughs to compare. That's one of my favorite laughs of all time. I'm going to play that one again in some different ways. I slow it down and it sounds awesome. But here's another one that I think really brings out the connection between these non-human laughs and human laughs. You can hear how that's a laugh, right? But it also makes it kind of obvious that it connects to these other vocalizations.
Particularly the great apes. You can see that the last common ancestor had what they think are these certain features: long slow calls and they're noisy. It was probably the first laughs in primates went something like... something like that. They're probably breathing while they're playing. But then eventually you have things like vibration regimes, which means there's tones. Chimps evolved this alternating air flow, which makes them different. We have more regular voicing, which is important, and eggressive airflow. This is also by Marina Deauville Ross but I added my little thing here: when speech emerges, we have volitional laughter, which is where the fake laughs come in.
First one. Real laugh? You guys are smart. I make it easy at first. Actually there's a reason why it's so easy: we have evolved machinery designed to detect these kinds of things. But it can get tricky. How many people think it's real? How many people think it's fake? Oh jeez, I'd say that's almost 50/50, maybe with a slight bow on the real. That is a fake laugh.
That is actually maybe one of the best fake laughs. In my research from all over the world, that one is rated as real by people almost 80% of the time. That's real. Real? Fake. Ah, yeah. I would call that one real. That's what people think. Mind you, I didn't have wires hooked up to their larynx or anything on their brain to actually corroborate this, what people think that they're real. It was produced by a woman...
All right, last one. See what you can do here. That's fake. How many people think that one's real? All right, compare it to this one. That's super fake, right? A little better? Still fake? Okay. That is fake but what I did is I sped it up. It turns out if you speed up fake laughs, you make them sound real. I have an explanation for that. It has to do with the control of the glottis. The opening and closing of the glottis is controlled by one brain circuit and it's got an evolved efficient control over it. When we're speaking, it's harder to control that opening and closing on purpose. Even though you guys weren't tricked...
This is the spectrograms of them. One thing you'll notice is that the spontaneous or real laughs are faster generally. This one's slower, this one's a little faster. You can see it by how long these little bursts are. Oh, this is a slowed down version, sorry. You can see it here. This is the regular laugh, you can see it's pretty fast, and then this one's a little slower. Let's play this one real quick. Here's the laugh that we heard. Now, if I speed it up, you already heard that, it sounds a little more real even though you guys didn't believe it. But now check it out when I slow it down. That still sounds like a person, right? That doesn't sound like a walrus or a bear or whatever.
It sounds like an animal. I've got research showing that when you play spontaneous laughs, those are laughs produced in conversations between people that know each other. When you slow them down 2.6 times, people are at chance in deciding whether they're produced by a human or non-human animal. But when you play them the volitional laughs, the fake ones, then they are 65% accurate in identifying it as a human. What that suggests to me is that there are features in those acoustic laughs that are really brought out by that slowing down process, which then reveal it as an animal vocalization that shares properties with other non-human animal vocalizations. As opposed to speech...
At the ratio of how much voicing versus how much is unvoiced, where there is no tone, the higher percentage of unvoiced components per call correlates with people's judgments of whether something is real or not. I think this is a direct consequence of the machinery underlying the production of these different kinds of laughs. Spontaneous laughs are involuntarily produced and they involve certain sorts of breathing mechanisms that are really difficult to control on purpose. Laughs that are produced by the speech system are more speechy, meaning they have a higher proportion of voicing, vowel sounds. 'Ha ha ha', the ultimate fake laugh. There's no breathing in there, it's just all vowels. But if you start going... so they're a little more breathy.
Now, I'm not going to go into this in detail, but I didn't know this when I was 22 years old and slowing down chickens and seals and humans and shit. I didn't realize what I was getting into. At the same time, this guy Juergens was working on squirrel monkey vocalizations and kind of figured out that there is some special circuitry that is unique to mammals that is different than humans. You have a very simple sort of circuit here, this aqua-ductal gray region that goes straight from the amygdala to larynx. It's a direct connection. Whereas speech involves all these other connections that involve language...
Which actually allows us to imitate a great number of sounds. Humans are vocal imitators. It allows us to imitate things like crying and laughter and pain shrieks and orgasms and all these other things that can be fake. But there are ways to look at them acoustically and see the signature of human specific production. Let's think about when people are laughing, what's going on. Imagine you have this guy here, he produces this utterance X, that's a surface feature like what you actually say, but he means Y, which is the unstated implicit meaning. This is sort of like a lot of jokes, it's encryption. My friend Nicole, who's here somewhere, might ask me, 'Do you get free drinks for doing this?' And I said, 'They gave me two tickets.'
But it's encrypted that two is not enough. That's funny to her. That's not funny maybe to this guy who doesn't know that I like to drink. Two sounds like plenty, but hey, I'm taking a cab home. He goes, 'Ha ha ha, it sounds like you're trying to be funny but I don't get it.' And then I'm like, 'Yeah.' And then we have an awkward interaction. Humor is encryption. A lot of intentional humor involves me implying things that people know but you don't say the thing that's funny. That's why explaining a joke ruins it. The funny thing is when you are able to signal, 'I know what you're saying, I got ya.'
It's not about humor because somebody would say, 'Okay, I'll see you later,' and then everybody goes, 'Ha ha ha.' But what provides... I've told him this and argued with him and he probably agrees with me: you don't know why it's funny. The reason that 'I'll see you later' is funny is because next time I'm going to see you, you will have gone through that hellish experience you were just describing to me. There's something else in there about seeing you later that's not just in the surface features of the utterance. Humor is encryption and laughter can be a way to signal, 'I get the encryption, I decrypted it successfully.' If I didn't decrypt it, then I'm faking it.
All right, so I looked at this real and fake laughing across 17... this is an ongoing study. We got three more sites coming in, we'll have 20. In 20 different places in the world, they would go, 'Ha ha ha.' Some people would produce some pretty good ones like you heard. I didn't pick ones that were ridiculously bad, I picked ones that made it a challenge. We played these for people all over the world and saw if they could tell the difference. It turns out they can. Here are the results from these 17 societies. This bar here represents chance performance. Every single society is above chance overall. The average is about 64% accuracy. There's some interesting patterns. For example, we have three traditional societies: a rural Peruvian society, the Shuar people in Ecuador, and Zulu people in South Africa. They're living in fairly traditional ways.
How do you know that's fake or real? I actually don't. I'm just using the context as my criterion. But it turns out that a lot of our real laughs in conversation are conversational laughs that are produced by the speech system, so fake in that sense. That doesn't mean they're trying to deceive each other. Two friends laughing in conversation can have a lot of functions, like 'I get what you're saying' or 'It's your turn to talk' or 'I think that's funny but it really didn't trigger a real humorous reaction in me.' The answer is probably that more of these laughs were fake than real overall. I think it's interesting that these traditional people are actually probably a little more accurate. Other places have this bias to say that laughs are real.
This is a good demonstration that no matter where you go in the world, people can hear laughs produced by UC Santa Cruz 18 year olds and decide whether they are real or fake with factually reasonable accuracy. All right, let's go back to our guys here and check out another situation. We have the same guy producing some sort of utterance, the surface feature denotes Y, which is the implied meaning, and these guys all get it. They're in on a joke and then they all laugh. Now we have a group of people laughing. Whereas this guy is like, 'I don't get it.' But he knows something's funny. 'I know these guys have some idea of what this is about...'
That's encrypted and who doesn't. You can use the laugh as a way to decide who's friends and who's not in this group, who has the information. These kinds of inside jokes can really help a crowd figure out who knows who. Here's a good example of this from 'Breaking Bad'. Hank, what a tool. This is where he realizes...
Friends or strangers? Friends? How many people think strangers? How many people think friends? Wow, okay. I'll explain that in a minute. All right, let's try another one. Friends? Now you guys are afraid. Strangers? Now I got you all confused. Those are friends. Strangers? Now, here's what I did, I'm tricking you. The first one is actually I constructed two individual laughs from people that never interacted. They were men. But that last one was the same laugh except I changed the pitch to make them sound like females. Check it out again. That's men, now I made them into women. If this experiment works well, damn you. This is my first try of a new theory based on some other data I'm going to show you: people think that women are more likely to be friends than men when they hear them laugh just at baseline. I know that's true.
When you hear the same laughs but then they sound like women, you're going to think they're more likely to be friends, even though it's the same laugh. I'm trying to separate the assumption that women are friends when they're laughing together versus the acoustic features of the laughs, which I'm very interested in. Let's try a couple more. Friends or strangers? Friends? Strangers? You guys, I got y'all fucked up from that first one. They're friends. Stoner or not stoner? Santa Cruz, baby. I actually am pretty sure they were stoned based on the conversation.
In this study, they hear a one second clip of laughter out of context and they have to decide: are these people friends or strangers? I did this in many of the same places I did it before: LA, a couple of Peruvian societies, Brazil, all the way down to Japan, Korea, and China. The indigenous places we got here... there's a bunch. Here's what happened. Basically everywhere we went, people were able to tell the difference. What was striking was that in every single place, two females were judged most accurately. The American participants were 90% accurate in getting the right answer when they hear two female friends. They nailed it.
No, they all heard the same laugh set. Yes, thank you for clarifying that. They all heard 48 clips of two people laughing. Half of them knew each other and half of them didn't. They were all UC Santa Cruz students who were in conversations. Half the people had been friends for an average of about two years and half the people had just met the day that they had this conversation. They either knew each other or they didn't. They were either in female pairs, male pairs, or mixed pairs. Thank you for asking that.
What you can see is that there are some similar patterns, but there are some variations as well. Sometimes people were good at getting the strangers that were male and other times they weren't. There were certain biases that people had. People tend to think when they hear two females laughing together, they're more likely to say they're friends. In fact, it's really striking. Here you see the response bias: just the likelihood of saying friends.
That's right. All the laughs come from a corpus of conversations that I recorded at UC Santa Cruz. There are times when people would laugh at the same time and they were on average one second long. These are very brief clips of people laughing at the same time and they either knew each other or they didn't. Everywhere around the world, they heard the exact same stimulus set in the exact same condition. It was a computerized experiment where they wore headphones. Even in Tanzania or Kenya, the people would go out there and give them the experiment. We had people from all walks of life.
You can imagine it's a pretty limited demographic, but it makes it even more striking because these are people that are completely foreign. If you go to Tanzania to the Hadza people, they have very low rates of encountering people like that. Even they can do it really successfully. In New Guinea, we had people in New Ireland. Here, our subject is Corrina Apicella, one of my collaborators. She works with the Hadza. This is the experimental context you see. Here is in Papua New Guinea, where they would do the experiment with everybody around them. I think only she could hear it.
The instructions are all translated in the experiment, but sometimes they had to be read to the participants and then the experimenter would enter the answer. In this case, I was testing whether people... my idea is that laughter potentially in a group can signal something about the affiliation between the people that are laughing together. I'm testing how sensitive listeners are in picking up the information about whether two people know each other, using a paradigm we call thin slice. Thin slice is usually thought to be like a minute or so. My thin slices are one second. In one second, can you get enough information to make an accurate judgment?
That's right. Potentially, but what my data suggests is that if you played people recordings of those people laughing together, they'd be able to tell the difference between people that know each other and people that don't. It comes down to certain acoustic features. Laugh speed matters. Faster laughs, which is probably correlated with the arousal associated with that emotional vocal system distinct from the speech system, they detect that. There are these acoustic features in the individual laughs that actually give it away. Even if people that don't know each other are laughing together and it feels authentic, it could actually still be different and detectable in one second to somebody in New Guinea.
You can actually transform the pitch values, normalize them, and then you can actually see what's the effect of pitch independent of whether it's... so it's not just that females have higher pitched voices, you can control that effect. That's a good question. The laughers? Did I mention that these were recorded at UC Santa Cruz? That's a great question. I don't have data on that. My suspicion is that yes, for the most part, anywhere you go... I do have recordings of people laughing at the experiment. People do this experiment, they're laughing along with it. The laughs sound like laughs.
It's true and it has to do with the intonation contour. French and German have slightly different rise-fall patterns. Baby cries from babies that are a couple months old or less have those particular intonation patterns already in their cries. At a glance, they sound like babies crying, but if you do the acoustic analysis, they're a little bit different. I bet that laughter also has some things that are variable depending on the language they speak. A fake laugh would be even more, yeah, exactly.
All right, so just the final thing: why do we laugh? One reason I think is that we're signaling cooperative intent. When we're laughing with one another, we are giving some information that's similar to a play bow. When humans laugh together, they're also engaging in these potentially positive signals that are helping you develop a relationship for later. Continuing to do it strengthens emotional connections. That's why you laugh best with some of your oldest friends and you're strengthening these bonds. It's also a way to signal decryption of indirect language. You can actually reveal that you have some information that is implied, and that helps you recognize who is in your group and who isn't, or who has certain information and who doesn't. It also can signal affiliation. When a group of people are laughing at the bar, you all know they're buddies. That might help you assess them as a group of people you might have to deal with later.
They have some connection and they're laughing about something, which is an honest signal that they're sharing information at some level. It also plays a role in conversational turn-taking and coordination. I have other work showing that people who laugh more in conversation end up coordinating their speech rates more, which is associated with a greater likelihood to cooperate in a behavioral economic game. When people are talking at the same speech rate, if they converge in their speech rate, they're more likely to actually cooperate when you say, 'Here's a situation where you can give money or take money.' Laughing plays a role in that. Finally, there's the evil side: social manipulation. People can try to laugh and try to gain your trust, like the car salesman.
We're manipulating each other. That's the rule of animal signaling. That's what signals are. They are designed to manipulate the behavior of other organisms. That's the general definition of a signal and laughter is no exception. We are manipulating one another, sometimes for our own mutual benefit and sometimes not. All right, that is the end. I think we're going to have time for questions. Thank you for your attention. Here's my list of collaborators. This was good and I liked the questions too.