Transcript 2 | Ofer Tchernichovski, Balanced Imitation Sustains Song Culture: Zebra Finches
[Introduction to Dr. Ofer Tchernichovski ]
Will Davis
Okay, well, let's kick things off then. So just the usual housekeeping. If everyone could remain on mute during the talk, that'd be really helpful. If you could submit any questions into the chat as we go along, Ophi's happy to be interrupted. So if there's anything that seems like it's a good question at the time, we can totally go to it. Please be aware the event is being recorded and it should be roughly an hour with more questions at the end. So now with that, please let me hand over to our chair, Diana Reiss.
Diana Reiss
Well, hi everybody. I want to welcome you all to our second talk in our Interspecies Conversations lecture series. It's an absolute pleasure for me Today to introduce Dr. Ofer Tchernichovski. It's a pleasure for several reasons. First, Ofer's an exceptionally accomplished scientist who's internationally known for his groundbreaking studies of avian vocal learning and birdsong development. Notably, Ofer's approached the study of birdsong development and vocal learning using incredibly creative approaches and techniques, which you'll be, some of which you'll be hearing about today. But on a more personal note, Ofer is also a wonderful friend and a colleague. He and I are both professors in the same department at Hunter College. We're in the Department of Psychology at Hunter College. And he's a valued and actually incredibly inspiring member of our Animal Behavior and Conservation graduate program there. And he's also a member of the faculty at CUNY's Graduate Center. Ofer did his Ph.D. at Tel Aviv University in the Department of Zoology within Galani and Yoav Benjamini as advisors. And he did his postdoc at the Rockefeller University in the lab of Fernando Nottebohm. He he's been a primary investigator or co investigator in numerous studies of birdsong published in Science, Nature, PNAS, Current Biology, Neuroscience and other journals. The majority of his papers have focused on the mechanisms underlying vocal learning and song development in birds, primarily zebra finches. If you go to Ofer's lab, you'll usually get a zebra finch on your shoulder. It's really a remarkable laboratory to go into. He studies song development across generations of birds and he's discovered vocal cultures established from this. He's watched how vocal culture gets established from the get go under controlled social conditions. In his laboratory of vocal learning at Hunter, he did pioneering research conducting one of the first quantitative analyses of bird song development and revealed how sleep affects vocal learning. And he's used many, many new techniques for tracking development, including the vocal combinatorial capacity of songbirds. And human infants. And he and his lab came up with a very novel and creative approach using a vocal robot. I'm not sure he's gonna have time to talk to us about it, but maybe we'll can ask him some questions. Using a vocal robot to study the coordination of calls and zebra finches. I also want to mention particularly to our community today that Ofer and his lab developed these techniques and gone on to share them by creating and developing a software program called Sound Analysis Pro, which is now used in many labs around the world. Ofer and his colleagues recently published a new paper and that's what he's going to be talking about today. So, Ofer, it's all yours.
[Zebra Finch song culture]
Ofer Tchernichovski
Thank you so much and thanks for coming. And I'm actually happy it's a relatively small group, so we can kind of like have some conversation on the go. So feel free to unmute and interrupt as often as you like. And I also want to say that in addition to sharing software, we're also sharing a huge library of songs from a lot of birds from the study I'm going to talk about. And this is really nice library because it includes hundreds of birds and hundreds of hours for each bird. And actually gave some of the data already to Mark Graham, who very nicely agreed to host them in the interspecies, you know, database. So I hope we can have a nice conversation about that. So let me share screen and start. I think we are good here.
Okay, you all see my slides? We are good. Yup. Yeah.
Okay, so this is a work I did with a student at Hunter. Her name is Sophie Eisenberg Edidin. She is now a PhD student at Stanford and also together with Eric Jarvis, who is at Rockefeller. And this is also the database from the Field Research center at Rockefeller. In the background, I think you may hear my canary singing in my house. You hear him? I think you may hear. But let me tell you about the song of a zebra finch. Zebra finch is the most studied songbirds in the lab. And people said that the song is not as nice as the canary that you may be hearing in the background. But I kind of disagree. I think they're very nice songs. And let's just hear one example of one bird.
So it's kind of very machine like. But of course you, you don't hear it well because it's way too fast for our ears. And the song is very repeatable, very repeated. And you can kind of like hear and also see in the sonogram image those kind of repeated motifs and Each one of those motifs is almost completely identical to the others, except that there is some structure behind the motifs to make the song more interesting that I will not talk about today. But if you just listen to a motif, it's. It sounds like this. Here is one motif, here's another motif. But they're almost the same. And of course, the phenomenon that got all of us so interested in birdsong is how accurately they learn and imitate songs from each other. In the wild, and also in the lab, people get them in families, and the young males imitate the songs often from the father, although if they have choice, they can also imitate from another. And the imitation can be very, very accurate. So here is an example of a tutor song, and here is a bird that grew up with him. So again, let me just play them again. Here's the tutor.
And you can distinguish between them, but they are very similar. And when you look at the sounds, those syllables, you can see how similar they are to each other. And it was already in the 1960s where Peter Muller, who is one of the pioneers of birdsong study, together with Thorp, figured out something very important, which is that those dyadic social influences that make the birds imitate song from each other add up to cultures that he called song dialects. And in the groundbreaking work he did at Berkeley, when he was at Berkeley, he looked at the white crowned sparrows. And here are like sonograms formal, a study from six different birds. And you can see that the songs are quite similar. They're not identical. Some of them has a tail, some of them don't have a tail, and all have an interesting thrill, kind of like in the beginning. But when this is actually in the marine area, which is kind of like north of Berkeley, the area of Pont Reyes, at Berkeley itself, when he recorded the same species, he noticed that the beginning was different, that there were like two syllables beginning as opposed to one, and even went south toward LA to Sunset Beach. Again, the songs, this is kind of the level of analysis they did. It's kind of like show the sonogram and show that they are different. And that is quite simple, right? And since the birds are imitating each other, they tend to converge locally and then diverge globally. So it's like languages, right, that locally converging and globally diverging. So this is a very nice and cool story, except that zebra finches that are most studied didn't seem to read that Mahler paper. So even though Mahler suggested, based on some species, that there is a local convergence and Global divergence. In zebra finches, we see, in a sense, quite the opposite. We see huge variability and diversity of song within a colony and very little variation that can be probably identified, but it's not very big across the colonies. And let me just play to you a few songs from the same colonies that you get a feeling of how different they are. So here is one example. Here's another one. Another one, another one.
Okay? So. And generally, the more you go, the more diverse you find. And it's all within the same colony of birds that are supposed to be influenced from each other. So what's going on? Well, people were saying, well, zebra finches are not very accurate when they imitate. But we know that it's not really true because when you raise them one to one without a colony, they imitate very accurately. Just that once there is more complexity in the social environment, you suddenly see that. And the kind of revised view of that is that it's really arrogant of us, an alien species like us, to judge an outcome of vocal development and learning as excellent and poor. And I'm going to use this vocabulary because we use it for many years, but this is really a very limited way of thinking about it, which is mostly an indication of our own ignorance and maybe our own stupidity, thinking that we understand something that we really don't. And so perhaps high diversity of song that we see in the colony is not because learning is not good, but because this is what learning is achieving. You know, why should they want to imitate at all? I mean, it's a very interesting question, right? I mean. I mean, we all acquire language by imitation, right? And you learn from your mom and dad. But when I was a kid, the last thing I wanted to sound like is like, mom and dad, right? So you don't really want to be like them. So there are many social issues here that where the imitation is not really something that you want to achieve, but more a mean for something else you want to achieve. So we'll talk about it a little bit more in the broader context of cultural evolution. So the classical study of cultural evolution says something very simple. You have an innovation, for example, a monkey figured out, hey, I can wash the potatoes. And then everyone imitates it. And then after a while, this founder, she was a female starting to do it, and then everyone started to do it too. And everyone is washing potatoes and it keeps going like that. So in the end, we have an evolution toward a stable but boring culture. Everyone is doing the same. It may change after a while, and they Start, you know, doing other things, salting the potatoes, so forth. But. But it's kind of like all kind of like in a cohort that get, you know, monolithic. Another type of culture is what you see in, for example, in humpback whales, where, you know, a male, in this case, the males are singing, so the male come up with a cool song, and then everyone say, wow, this is such a cool song. Let's sing that song. But then they get bored of it, and the next year another male sing another song, and then they all go for the other song. Yeah. So you have kind of like a drifting culture. And of course, the dark. The culture can be very drifting or even very chaotic. But what I want to talk about is a third type of culture that is actually maybe in between those two extreme, between the chaos and uniformity, which is what I like to call a stable polymorphic culture. So here, on one hand, things are not changing much over time. For example, birdsong culture in the wild can last for hundreds of generations, are very stable, but they're not boring. They are very rich and polymorphic. And this is really what we call a cumulative culture in some sense, which is what makes humans so unique in some sense, but maybe not so unique because we see that in the bird. So I like to think about that, in a sense, when you have forces that are converging and forces that are diverging, there is a sweet spot in between where structure emerges. And this is exactly where I think some species of birdsong are. So we really need to understand the balance between the diverging and the converging forces to understand in different species, you know, what kind of culture they come up with and what does it mean to them. So what can we do? Well, the first thing that we learned when we looked at cultures and we did lab experiment of culture, so we really tried to start culture from scratch. So we would kind of raise zebra finches in isolation and let them establish colonies inside the lab. And that you go over generations. And what we see is that when you talk about the wild type zebrafin song, then you have a song that has syllables of particular types that go serially. Like you can see here, for example, a yellow syllable and a red syllable and a blue syllable, and again, yellow, red, blue. So it's like abc, abc. In other words, we'll do ABCDE and so forth. But when you look at birds that did not get any education, then don't do that. They'll do like A, A, A, B, C, C, C and stuff like that. And we've shown that when you take those kind of uneducated birds and start them as founders, then within a few generations that there is normalization to wild type cultures. If the birds have some internal compass of how it should sound like, they don't do it by not imitating. They actually do imitate, but they imitate in a biased way that we are going to talk about in much more detail today in a different context. So in this case, what we showed is that you can start from kind of like mumbo jumbo of unlearned songs and get something which is much more confined into a species typical culture. But maybe we can think about the same idea as social forces that not only kind of converge the animal into a culture, but also make sure that there is enough diversity within the culture. So can biases in the learning do that? And so we did it by looking at statistics of song learning across many birds. So we went to the Field Research center in Millbrook. I went there every week for a couple of years and just kept recording non stop a lot, a lot of birds. And today I'm going to show you analysis of 160 cases of imitation that we looked at together. So this is the Field Research center in Millbrook. This is the chalet where I did my postdoc and first faculty position there. And I'm going to show you the outcome of imitation in this colony that has a lot of families that are raised in cages. And we are kind of like looking at what happened to them one generation after the other. So the way that I'm going to show that is just by kind of like scoring imitation, you know, as kind of excellent, which is the green or poor, which is red. And that's all it tells us really is that when we run automatic similarity measurement, you know, how similar acoustically the songs are. And we are going to look at the tutor, in this case LG281, and we are going to look at seven of his male offsprings, which we call pupils. And I'm going to use the shape to show you different clutches so that you know, each clutch can be influenced. You know, the clutch member may influence each other. So it's nice to look across the clutches. So again, colors will be the accuracy of the imitation and the shapes will be the clutches. And let's look at families that had several clutches. And that's how it looks like. And you only need to take one look at it to realize that something is interesting here. Namely, if you look at the bottom there, I put families where the Outcome of imitation is almost always very inaccurate. When you look at the top you see families where the imitation is usually very accurate. And in the middle there are some cases that are more variable. But it seems that this is not a random distribution and indeed it's not a random distribution. So if you look overall a histogram of similarity scores, you can see it's a very broad distribution. But what's interesting here is how much variability in accuracy of imitation you see within a family as opposed to what you see overall. And what this figure, looking at the coefficient of variance of similarity within family shows you that if you just look at this histogram, you see that there are only say 1, 2, 3, 4 families where the coefficient of variance was higher. But in almost all the other cases you can see that the coefficient of variance was much lower than when you look across family. That doesn't mean that the outcome of imitation was similar across bird. It wasn't. What is interesting here is that how similar it is is actually conservative within the family. So some families are usually very inaccurate and some are very accurate. And we can see that very well in the data. Here is a family with an accurate imitation. So let's listen to the tutor. And here is a pupil. Here's another one. And here's another one. Now let's take a family of poor imitation. So here is the tutor. Now that's a very bizarre song. And you would think that this male is somehow defective, but he's not. He was actually a beautiful male female liked him a lot. He was actually one of the best breeders in the colony. He still alive, but his songs are very bizarre. Even though as a male he was successful and strong and everything and female liked him. The imitation of these songs were very inaccurate.
So they're not similar at all.
Vint Cerf
Wow.
[Factors influencing song imitation in Zebra Finches]
Ofer Tchernichovski
Okay, so the huge difference between those families. So why do we see those differences? Well, you can see the numbers here, but. So the point is that we spend like a decade studying good imitations. But I would say that maybe Tolstoy got it right in the opening sentence of Anna Karenina that happy families are boring, they're all the same. But unhappy families are amazing. Each one is unhappy in its own way. So we need to actually try to understand those families. So what can we really do? Well, one hypothesis is that there are genetic differences between those families. So maybe genetically this male is just genetically a poor singer and therefore genetically also his offsprings are poor singer. So that will be the first hypothesis. The second one is that there's nothing genetic here. For whatever environmental reason, one song is much more complex than the other, and then the pupils are responding to those songs differently. And in order to figure that out, all we needed to do is a cross fostering experiment, which we didn't even have to do because it happened naturally in the colony. We have cases that birds have to be moved, and when we looked at those cases, we immediately seen that across the colony over the years, imitation of a foster father was not less accurate than biological father. If anything, it was even a little bit higher. Right. So there was no significant difference here at all. So there is no support of any genetic differences, at least when we look overall, it doesn't mean that there are no genetic differences. I'm sure that they are and I'm sure that they have an effect. But at least holistically, when we look at our colony, it doesn't explain these particular effects. So what can we do? Well, if it is an environmental effect, then we should be able to measure it and it should be very easy, really. Right, because what have we?
I mean, we have one song that has a lot of stuff. So for example, the accurately imitated tutor had four syllable types, you see the red, the black, the green and the yellow, whereas the other one had only one. I mean, you may think it's two, but notice that there is really no gap between those sounds. It's really segmenting into being segmented into one syllable. And I'm going to talk about this problem later on, but you can count it as one or as two, and the other one is four. And in fact, if you want to count a little bit better, you have to take into account also how common each syllable type is, because if the bird is singing something like abc, abc, abc, it's more diverse than the bird that's singing like A, A, A, A, A, A, B, C, A, A, A, A, A, A, B, C or Right. So if you take into account also the frequencies of each syllable type, then you can come up with a diversity measure. Here we use the Shannon information entropy, which is minus P log P, but that's very simple measure. You just add the syllables and multiply them by the frequency and take a log to get units of beats. So don't get nervous about this measure. It's just a diversity measure. So great. So all we needed to do is do that in a large sample and first of all just confirm that there is a correlation between the diversity of the tutor and the pupil. There has to be, because they imitate and then show that there are some you know, inconsistencies in those, correlation in the extreme. So that looked very easy, but it didn't work. When we did that, we found that there is no correlation at all, nothing between the diversity of syllables in the tutor and the pupil, which was a shocking outcome that was very counterintuitive and. And didn't really make sense. So we decided to look at it only in cases of high similarity, you know, versus low similarity. And that's what you're seeing here. And it didn't help the high similarity. We have an r square of 0.04, which is kind of nothing. So what's going on here? Well, we figured out that similarity measure is really what we are looking at is a little bit of an impoverished way of looking at comparing the similar songs. Because, for example, here you can see a case of a tutor and a pupil where everything seems to match. So you have, you know, 99.5% similarity. And also when you look at it the other way, namely how much the. So, so you can always look at it both way, namely how much the pupil imitated from the tutor, which is one issue. And the other question is you look at it the other way around and say how much of what the pupil is singing is influenced by the tutor. For example, you could imitate everything that somebody else say, but add to it stuff of your own. So the influence is telling you from what you are doing how it's similar. So you just look at it both ways. In this case it's the same. But of course, in this case you can see that the bird kind of like imitated in a way that included about only 76% of the tutor, but that's all the pupil did. So when you do it the other way around and you look at how much influence is very influenced because everything that he does came from the tutor, but he didn't imitate everything from the tutor. And of course this can also go the other way around. Namely you can have a bird that imitated everything from the tutor, but invented a lot of new stuff. In this case the similarity will be high, but the influence will be lower. So we were hoping that this will help. And I'll tell you right now, it didn't. But we are going to use it later, so don't be sad about it. So here is the outcome. So when you look at similarity from tutor to pupil and compare it to the diversity of sounds of syllables in the tutor song, we have 0r square. And when we're looking at influence Instead, they get 8% like 0.08, which is still very, very low and barely significant. So this was a complete disaster. Everything we did didn't work here. And we actually realized fairly soon why it doesn't work. And the reason that it doesn't work is because the idea of syllables, because here you can see that we are kind of segmenting the song to syllables, those colors of the tutor. And here you can see three syllables here, like the green and the yellow and the black, and you can see clusters in the feature space. It doesn't really matter, just to tell you that we can identify those guys. But when you do the same thing to the pupil, the problem is that in the pupil all of them merged into a single syllable type, which you see is a single cluster here. So in a sense, it's not a coincidence that we are failing to show similarity in the diversity of syllables, in the syllables themselves. The syllables themselves are not stable during song development. And in fact, that's kind of what we've seen in many, many times. The birds that imitate, they modify syllable boundaries when they imitate because what they call syllable is really a continuous sound. And the question is to what level this is something that the bird respects. And the answer that we got is that they don't respect it.
[Evaluating Zebra Finch song features]
Ofer Tchernichovski
So you can see like two syllables in a tutor that turn into one syllable in the pupil. You can see that the bird is continuing, it doesn't stop. And it happens a lot. And because of all of that, syllable is really not a meaningful unit to look at song learning. That's why we fail. And what is also interesting is that birds many times compress what they imitate so that they may look at something or listen to something that look distinct as four things, but they like to put it together into one. So in effect, the solution that we have for it is something that I came up with already many years ago, without even thinking about this problem. I didn't like the idea of segmenting when doing similarity scores. So the way that we developed it was to actually look continuously at two songs and look at the matrix of segments that just emerged from the analysis. Not like segmentation you do ahead of time. So when you look at the sub syllabic level, you're really looking at features continuously. Like you can look here, for example, at frequency modulation and pitch in red and wiener entropy, which is the kind of width of the power spectrum. And so you can look, you can do those graphs and what you get is something that looks quite messy. So how do you deal with that? Well, you can Deal with it when you do similarity measurement. But when you look at diversity, it's kind of like a bit too detailed. So what we figured out that we need some kind of a mesoscopic level where we can go in between. And that wasn't that hard to do. Because when we look at the distribution, for example, of different features, without segmenting just kind of continuously at all those hundreds of songs, we could see that our concentrations of features that made a lot of sense. For example, those kind of three concentrations that you see in this figure are telling us that some sounds are not very modulate. Those are, those are kind of the harmonic stack that are not modulated very much. Then there are kind of like downslopes modulation, the cluster on the left, and then the cluster on the right is the up modulation. So they kind of generates category overall. And in fact, if we add another feature called pitch, like how high is the pitch? Then in each kind of the pitch has also different kind of like categories, natural categories, and you can divide the space to low pitch and high pitch. And then you look at the same clusters in other features. So when we look at it in many dimensions, we can actually see that there are several clusters and we could capture most of what the birds were doing by 10 of those clusters. And that's a bit of an ad hoc way of doing that, but at least it didn't have the problem of segmentation. So here we can say vocal state number nine is high pitch, low modulation. And we can see here in the sonogram in yellow, all the areas that are like that. And then we look at, for example, non modulated harmonic stack, and these are the light blue and so forth. So all we did really is solve the problem of segmentation and look at it in a way that we're just looking at a different type of acoustic states. And when we do that, we can do the same trick again. We do the same feature diversity, minus P log p, the same diversity measure. And when we do that, since we have 10 vocal states, then if everything is completely widely distributed so that all the sounds are equally distributed, this is the maximum possible diversity. Then the minus P log P of 10 will give you 3.3 beats. So that's the maximum you can get. So anything between 0 and 3.32. Now when you look at the distribution of what we actually get, it's actually quite interesting because the upper bound is right here. This is like 3.3. This is the maximum you can get. At least in this categorization. Most birds are not very far from that. So if you look at the median, the median is like 3.14 pike, right? Not very far from 3.3. And very few birds go low. So it's kind of like a long tail that goes low. And now if you do the same trick and ask, okay, how is the two torsome diversity in terms of vocal state correlate with pupil song diversity? At least we get correlations. Now we have a very significant correlation. It's explaining about 20, 23% of the variance, which is not terribly a lot, but it's significant, it's real, and it's equally strong when the similarity is high or low. So overall this kind of variability is explained. But I want you to watch two things here. One of them is that the slopes is much less than 1 and that when the diversity of the tutor song is high, we're falling usually right on the diagonal. So it goes above the diagonal when the diversity is low. So that means that if you're imitating song that is high diversity, you also make high diversity. But if you imitate song that is low diversity, you will go high diversity. But of course that won't make sense across generations, right, because it's a steady state. But what we see actually is a subsets of birds with high diversity that goes down. And those that goes down, notice that they're all red, namely they're all low similarity.
So another thing that we've seen is that if you try to predict how rich the song of a bird would be one of the best predictor is how well the bird imitated. It's not a trivial thing. It's telling us that the birds that did not imitate well, they could improvise, right? They could do other things. But in effect they had songs of lower diversity. And in a sense we find that really if you try to predict song diversity of a pupil, those are the two things that we care about. One of them is how high is the diversity of the tutor, and the other one is how good is the imitation, so to speak. And those together explain about 30% of the variance. So, okay, so we got all of that. But now what happens when you try to predict similarity? And we already told you that the tutoral song diversity is not a good measure of that. And in fact, when we do that on the level of vocal states, it still doesn't work. So when we look at the diversity of the total song, it doesn't predict how very well. At least only 8% of how much the pupil will be similar, only in Some cases. But remember that I told you about the idea of influence. Then you're looking at the imitation the other way and looking at how much the bird is improvising. And here when we do that, and we're looking at not exactly how much was imitated, namely how much it took from the tutorial, but how much he was influenced, namely how much the influence is telling us how much the pupil really added to the tutor. And here when we do that, then we go up from 8% to 25%. And it's very significant. What it's really telling us is very simple. He's telling us that pupil of tutor with impoverished low diversity song, they're not necessarily not imitating them. Sometimes they don't, but quite often they do. But what they always do much more likely is add stuff to that. So the improvise, the level of improvisation is really triggered by how much they can get from the tutor, which makes a lot of sense. Another thing that I want you to realize is that even though we showed you there is a significant correlation between the diversity of tutor and pupil, when you look at the population level and you try to look, okay, this is the tutor and this is the diversity of the song of the tutor and what will happen to the pupil then? As I told you before, tutor that had low song diversity had pupil with higher song diversity. So those red areas are all going up. So that's what I showed you before. But you can also see clearly that it also go the other way. So that there is very little inertia across generations. The steady state doesn't change. But most of the birds that had low diversity had pupils that improved. But then you have another group from the high diversity that actually went worse. So it's kind of like seems to be balancing both ways. And to understand better what happens, let's actually look at multi generations. So for example, if you look at the bottom, you have on the bottom Left you have HP10 tutor and almost all of his pupils were not imitating well, four out of five and even the fifth one wasn't great. But notice that when you look at the next generation, you can see that it gets from red to green. And the same thing you see in DG4, all of them are red, but quite a few of them go to green or to yellow. So what happens here? Well, in fact, those two cases are very interesting because we can understand what happened there. Here is the song of the grand tutor and here is the song of the first generation. And you can see that the song look Much richer. And therefore the similarity score also look lower for all kind of reasons. But then in the second generation it's actually become very similar, as if this poor imitation is not really poor imitation, it's a correction. And you see something similar also quite here, for example, here the tutor has a lot of harmonic stocks and what the pupil did, he imitated quite a few of them, but kind of like made other sounds more frequent. And therefore the next generation kept doing that. So you can see mechanism of self correction across the generations.
It's not always like this. For example, here you can see poor imitation that is really that the bird just didn't imitate all of the tutorial. So what we call poor limitation can be different things. And you can see actually how can understand that. So here I'm looking at, I simplified the vocal set a little bit to make it more clear. So I took it down from 10 to 5 by just joining them according to pitch and modulation. And you can see here a tutor song and a pupil song. But what is really important is looking at this pie chart, because this pie chart should tell you how much for each state you see in the song. And you can see that here half of it is the high pitch, but in the pupil only about a third of it. And the same thing you see here, like half of it is high pitch, but here only about a third of it. You can see that the pie, the slices in the pie become more evenly distributed. And you can see three generations. And again here is the other way around. Here the non modulated harmonic stacks are very high, here they go lower. And here you can see additional balancing happening in this area. So what you can see here is a natural tendency to diversify the song, not necessarily by not imitating, but kind of changing the balance of different sounds. And a more quantitative way of looking at that would be to looking at the abundance of each vocal states in each bird, tutor versus pupil. And we are looking at it in just dividing the birds to excellent imitation and poor imitation on the right and the middle one in the middle. So those are the four quartiles. And what I want you to notice is that the points are going below the diagonal in the high, so that when the abundance in the tutor song is high, it's going lower in the pupil. And you can see it regardless of how good is the imitation. In fact, the strength of this effect, namely that they amplify the low ones and reduce the high ones, appears to be very, very similar regardless of how well the bird imitated in terms of accuracy. So that tell us that when you have low similarity doesn't mean that the bird is not equally affected, that the bird is doing the same kind of process either way. And this is just a way of quantifying and showing you that the gain versus decline is different is similar across the imitation quartile. So what we can learn from that is that overall what you are having here is a process of fixing the tutor. And this is like showing you overall across all birds, the abundances of vocal state in tutor versus pupils. And let me actually flip the axis and I'm just going back and forth, flipping the axis and if you look at that, you can actually see very clearly what happens, right? You can see that all those points that goes down, you know, and again this is all the same distribution, but you see that they go up. So there's asymmetry here. And we can easily quantify the dynamics of this process and actually compare it to what we did in the past when we look at isolate songs. And in the end it's a very similar story. Namely, when the isolate songs are normalized across generations to normal songs, we can see frequency dependent improvement. And we can see the same thing in the wild type birds, even though here it's happening on a very, very small scale. So they keep kind of balancing in a very, very small scales and increasing the diversity. Finally, I want to make one more point which is that when you look at the distribution of vocal states in birds of high diversity and birds with low song diversity, they're not the same. If you look at high diversity, almost by design, the distribution of vocal state has to be similar, which is what you're seeing in that line right here. But when you look at the population of birds that had low song diversity, as a population, individually of course they were biased, but also as a population they were biased and they were biased to do a lot of high pitched sounds, which is actually very much like what we see in isolators. And of course, when we are trying to look at what happened in the outcome, then we get the same thing. Namely, the most influential songs were the one that were balanced more or less in the acoustic states in the frequencies of them, Whereas the one that were less influential of the pupils were the one that were less balanced, very much like what we see here. And that can be translated directly to the overall, much more easier to understand mean features of the song. So, for example, think about a song that is high pitch and song that is low pitch on average. And we are trying to try to relate that to how influential they were. Well, if the pitches are very high, then the song is not very balanced. It's only going to be balanced in the center. And indeed you can see that if you look at the distribution, you can see a much broader distribution of pitches in the bottom influence and a much narrower distribution in the top influence. Namely, influential songs were moderate, the extreme songs were less influential. I wish it was true to people right in our culture, it's not like that. So they have a mechanism to get rid of extremism. And you can see that in all the features that you always see that the moderate, the influential is always narrower distribution and the less influential are broader distribution. Furthermore, if you go across different colonies, these are four different colonies, and you look at the influential zone versus the non influential zone, even without knowing anything about what happened in those colonies, we see the distribution of pitch is very much in that high influence that we found in our colony. And the same will be true to other features. So in a sense, what we are finding here is exactly this kind of question that bring us back to what we talked about before. So let me summarize and then we can have some discussion. We found that the acoustic structure of the tutor song explained variability in imitation outcome across the families. We found that tutor with more acoustically diverse songs have greater influence over the pupil song. We found that pupils imitate vocal states in a very balanced manner. So that even small deviation that makes the song more extreme, even just a little bit more extreme, they immediately correct it. And if there's something that's a little bit more rare, they immediately correct it. So it's a very, very sensitive process in which they kind of change that. We show that it's an active process and that it's by and it's really related to very, very small fluctuation in the abundance that we see. So in a sense this is bringing us back to the idea of polymorphic song culture. And what I would like to end by saying that the bottom line that we are suggesting is that we are seeing a frequency depending balance process where each bird, when they imitate, they do those corrections. And those correction is really not poor imitation. And those corrections are doing two things. In one hand, they keep the birds close to the center so that we don't have too much drift across colonies. So it's kind of preventing chaos from happening and keeping a species specific repertoire. But on the other hand, the balance is keeping the repertoire big enough so that on one hand you have a culture that is rich, but it's not chaotic. So that you have like exactly some kind of a sweet spot that exists that allow the culture to persist. And of course, in different species, we may see that these spots are different from each other. So that's it. I just want to mention again, Sophie Eisenberg Edidin, who helped me and did a lot of the analysis, Eric Jarvis from Rockefeller, and also the bird keepers at Rockefeller Field Center, Lotem and Gillian. And of course, I want to thank the NIH and the HHMI grants that supported us. So that's it. So let's have some discussion.
[Q&A]
Will Davis
Thank you so much. We have a question from Francine. What is the difference between those pupils who add to the imitated song from those who don't?
Ofer Tchernichovski
Well, I mean, as I said, you know, what seems to be driving the addition is the properties of the tutor song that if the tutor song is impoverished, they tend to imitate more tweaks. So it's necessarily a factor that is taken into account by the pupil that they kind of like seems to be wired to have something balanced. The question is, how come that we have unbalanced songs in the colony and how come that those songs persist and those birds are doing quite well? This is to me, the kind of. And it seems that those birds come because they themselves were not capable of imitating accurately. So there are kind of two issues here. One of them is a true imitation failure, and the other one is a bias. So the true imitation failures are generating birds that are, you know, have poor song diversity. And it could be that those birds have high testosterone because testosterone really makes you stupid. And so we know that. We know that from our own experience, and we know it also mechanistically that high testosterone can close the learning period. So it could be that that's kind of like what drives it.
Will Davis
Perfect con. Did you have a question?
Con Slobodchikoff
Yes, actually, I did have a question. If we go to look at how humans learn in school, there are some teachers which are very good. They are very patient with their students. They want their students to learn. They repeat things a lot and the students learn a lot. Then there are other students or, sorry, teachers which are very impatient and are very strict and don't repeat stuff, and their students don't learn things and instead go their own way in terms of making up information. I wonder, do you have any behavioral data on the tutors to indicate that maybe the tutors that are really good at teaching their pupils have behavioral differences that sort of reflect this kind of patience and care that we see in human.
Ofer Tchernichovski
That's an excellent question. So we are having behavioral data that we are actually only beginning to analyze. Overall, it seems to us that as opposed to what we see in humans, at least in schools, the birds don't need the tutor to actively train them that much. They can learn very well from playbacks. That doesn't mean that there's no variability here, but the variability that we see in the attention is not necessarily correlated with what we see as a good outcome, which is not necessarily even a good way of looking at it, because we don't understand what a good outcome is. And I think that this is the main issue, that even though we have the behavioral data, we find it difficult to interpret them because we don't understand how to judge. We are an alien species. We are looking. You know, I always thought about it that if an alien species kind of landed here and seeing people in the subway with those headphones on and try to analyze all those things that happening, they'll never understand what's going on. They won't even have a concept of what music is. And the fact that music is actually meaningless in any way you think about it, except that it's magical in some other ways. So again, I think that the main issue here is our own limitation, and this is a very good forum to talk about this, that we have to appreciate our own limitations and our own stupidity when we are trying to deal with an alien species.
Will Davis
Vint, did you have a question?
Vint Cerf
Having to play games, finding the unmute button, I can't tell you how fascinating it was to receive this fire hose of information. I've never had data from fly at me at quite this.
Ofer Tchernichovski
I'm sorry.
Vint Cerf
Oh, no, don't be sorry. It was one of those. Wow, this is really cool. It must be like a dog sticking its head out the window and getting acceleration of all the odors and smells, saying, whoa, what a universe out there. So I'm sitting here thinking about the variations that you cataloged, and now I'm starting to wonder, okay, we know that I guess between species there are very distinguishable bird songs. And I assume that that's important because the species don't want to be confused for each other and the males want to attract the right flavor of females and all that stuff. Within a species, though, is there any value at all in either evolutionary value or some other value in variation? You know, could it be that it's a way for within species, for different groups to distinguish themselves? And why is that valuable? So can you. Is that first, I don't know if that's a well formed question.
Ofer Tchernichovski
No, it's a very well formed question and so people will give you a different answer. I can give you my own feeling about it. It's not supported by data, but my feeling is that birds usually don't have a problem identifying their own species that much. That that's a relatively easy problem when you want. When you walk in the street and you see a chimp or a gorilla or a human, you usually know which one to have sex with. But most people maybe, and I think that that's true also too. Birds are very smart and are very well imprinted. So I think that the within species is really what matters. What we are sensitive, of course, is a cross species because we are alien species. But for them it's really trying to find a good mate within their species. That is the difficult problem. That's a difficult problem in our population too. We have like 50% failure. We think that we are very smart, but we have 50% failure even when we get married. So understanding this variability within a species is essential. And again, I talked about many things that I think relates to human experience. For example, the idea of how extreme you are going to go. Like, I mean, Vint of course knows a lot about the Internet, right? And about what's happening on the Internet. We are doing all this amazing communication in the Internet without any shield of evolution or adaptation. I mean those birds have their own Internet but they over many millions of years developed a shield that allowed this system to give good information, which we didn't manage to do yet online. So we can actually, instead of reinventing the wheel, we should think about what nature already come up with to solve those problems.
Vint Cerf
So let me make an invitation to you then to write an essay called what the Birds Can Teach Us about Misinformation and Disinformation.
Ofer Tchernichovski
I already did. There is a paper I wrote in Ion a couple of years ago.
Vint Cerf
Send us a copy, send us a copy. I'd love to read.
Will Davis
Yeah, perfect. Question from Irene. Do you see students actively seeking out multiple tutors?
Ofer Tchernichovski
Excellent question, excellent question. In this experiment, this is not even an experiment, it's just a colony where we really didn't give the bird much opportunity to do that at all. And in fact everything was based on the birds not having that opportunity, which makes the study somewhat limited. The problem, of course, when you look at the colony, when bird really wander around and choose tutors, then it's very, very difficult to understand what's going on without also following the interactions from moment to moment. And several labs are now actually looking from moment to moment in those interactions between pupils and tutors and try to understand. There's another issue, of course, how they coordinate their vocalization, which is also an amazing issue. I would say that right now we are in a stage where the technology to follow and track birds dynamically and understand interactions from moment to moment over development is very, very strongly improved. But we didn't become any smarter trying to interpret what happened, try to even interpret where the song came from, where a syllable or not even a syllable, where its sounds came from. The more you let them freedom to do that, the more difficult it's become to understand their choices. And I think that something very similar happened in neuroscience that we develop amazing techniques to now do entire brain cellular imaging and people showing those amazing flashing neurons in an entire brain. But that doesn't mean that they understand what's going on. And so it's a very difficult question. I would say that the choice of a tutorial is a very, very important question that we really don't understand. And this choice is really kind of like a process that has to do with the social network that the bird is in. Because if you think about what the birds are doing, I mean, they try to come up with a song that works for them. It's a song that will make them popular, perhaps a song that will attract mates, a song that won't get them into too much trouble. Right. They want to have individual identity, so they want to be themselves. They don't want to be like anyone else, but they also want to belong to the group. They want to be able to influence the behavior of others. And all of that is happening as you develop. And you are kind of like, hey, here is this tutor and here is that tutor, and then what will happen in the next generation? So to me, this is the most amazing question. You know, how those birds make those decisions. You know, think about yourself as you grew up. How did you kind of like decide where you're going to be? So from that point of view, this system of looking at those little cute creatures developing is very, very important to take our understanding to the next level. And I hope that we are going to see that happening in the next few years.
Will Davis
Question from Stuart. Does this constrain either the size of the colony or the area it can occupy? Is there a relationship between polymorphic diversity and spatial distribution?
Ofer Tchernichovski
Yes. So again, amazing question, and I regret that the answer will be very short, which is that we don't know. But I think that, you know, in some sense you may think that the balancing process that you're seeing here is making the birds maybe more robust to bottlenecks. So even if there's a small, strong bottleneck, they can actually resist. And I think there are studies about evolution, culture of song, evolution in the lab shows that they can very, very quickly get rid of those bottlenecks. But of course, when you have environment that is richer, when you have more things to learn from, there's a study published recently showing that, for example, big cities have a lot of innovations. Even if you correct to a lot of things, resources, stuff like that, it's still that big cities have more innovations than somewhat smaller cities. So there is something about concentrations that somehow get people over the mark and somehow manage to do collective creation. But again, in the birds, we are not yet at the level that we have an answer to, that we have some results in the wild. It shows that when there's a lot of disruptions, where you have a lot of ecological problems, and when the niche is kind of like getting fragmented into small, many, many small problems, so many small areas, then you know, you have consequences in song imitation that at least seems to not be so good. But there are very, very few results that we have on that.
Will Davis
Diana, you had a question?
Diana Reiss
Oh, hi. Yeah, hi, Ofer. So first of all, thanks for an absolutely stunning talk. I was fascinated when I read it the first time and it was great hearing you talk about it in person. So when you were talking, you know, you sort of started talking about this already, about what attracts these birds to learn what's salient in the signal and, you know. Or who's cool is another way of saying it. Who should I follow? When you're talking about influence, do you think that pitch characteristics might be important there? Because when you were talking about the high diversity song, I don't know if you said maybe I misunderstood. But did you. Do they then tend to have lower pitch or they don't have the higher pitches of the low diversity? And I was wondering if somehow that conveys fitness to females and it's been selected for and it's more attractive. Could you just talk about that a little bit?
Ofer Tchernichovski
Yeah, yeah, of course. So first of all, this study is limited to low level features like pitch and frequency modulation and wiener entropy, which is also the width of the power spectrum. So there's another dimension of the song that we call rhythm that we do not include in this study. And in fact, I think that rhythm probably even more important in zebra finches than pitch. And I hope in the future to be able to look at it a little bit more. The low diversity song had quite often very high pitch, which is indicative of isolate song with a little influence which probably not working so well on females. But as I said, when we are looking at female preference and have a lot of data now trying to see how female respond not to the song but to the males that sings that song, you can see that whatever my opinion is about which songs are amazing and which songs are terrible has nothing to do with what the females think. And they are fairly consistent in what they think. They just don't think like I think. And the problem is that people always miss a very important issue which is that what you sing matters in the context of who you are. If you happen to be a gentleman, then maybe one song will work well for you. If you happen to be a bully, maybe another song will work well for you. So what you really need to understand is the interaction between what message come from the song in terms of think about song as a music. The song is affecting your internal states, right? You hear music and suddenly you feel that you want to walk or that you want to dance or no, that you want to cry and all those kind of feeling that the music is evoking in a mechanism that we kind of somewhat beginning to understand should somehow work with who the bird is, right? If you are trying to get a mate by showing how gentle and kind you are, maybe making the female want to dance is not exactly what you want. If you're a bully, maybe that's what you want. So I think that the problem that we have here is that we need to put a lot of work in each bird. That's what people don't like to do. Particularly there are old style ecologists that always want let's look at the entire population and do averages and stuff like that. My approach to that. No, we need to understand every bird very well. We need to know the birds, we need to get intimate with each one of them, understand who they are and in that content look at the song. Not let's say, oh, maybe the pitch, maybe the no. I think that if there was an easy solution we already would have found it. There are hundred years of birdsong research which made zero progress almost in understanding what song mean. We understand a lot about how they make the song, how they learn the song, how understand almost nothing about what it means. And I think it will require a very, very different approach and much more respect to the individual elements.
Diana Reiss
Ofer. Could I just tag one quick thing on what you were talking about with segmentation and not looking at that. I mean, I think that's such a huge contribution because we just found, and we have a paper in Frontiers coming out about with human whistlers, where we can decode that it's the same thing. What we think should be the segmentation is not holding up with human whistle language. So that's it. So I'm thrilled to hear you say this.
Ofer Tchernichovski
I mean, yeah, this is a painful issue. The thing is that the syllabic model for human language didn't last long at all because people immediately understood that it's wrong. The syllabic model for birdsong learning is lasting for 100 years now because we are stupid. We don't know that it's wrong. And, you know, the people are, you know, just. And many times I see it like people are giving a talk other whales or a bird or whatever. And then they kind of say, yeah, there are 253 syllable types. And I raise my hand and say, how do you know that you should look at syllables and why do you put all the eggs in this basket? And I never got positive respond when I asked that. So I stopped asking. But I feeling that, you know, you're taking this butcher knife and you cut the behavior based on something that is just, you know, you're just becoming a butcher. And it's like the first anatomy lesson that I took in as an undergrad. The professor said something that I remember, that dissection is separating the animal to its organs, not cutting it to pieces. And so when you, when you do dissection of songs, that's what people do. They take a butcher knife and cut the song to pieces instead of separating it to organs. Yeah, and so that's a big problem.
Con Slobodchikoff
When you look at, or when I look at your syllables, I can see a lot of little variation in each syllable that you call a single syllable. Have you thought about breaking down or expanding these syllables to look at more fine scale variation and how that might code with either what the tutor is doing or what the context is of the song?
Ofer Tchernichovski
Well, I think that's what we are trying to do. Yeah, we are trying to go on a finer variation, but we are not committed to syllables as unit for the reason that the birds don't seem to be committed to doing that. So my kind of instinct is not to segment, but find time events. Time events is much safer thing to do. So, you know, even if you get some time events right or wrong, you're fine, but you have a segmenting wrong, then you're really kind of like making, you know, disgusting stuff that doesn't exist. So the fine scale analysis should be based on time events, continuous. And also think about, like, what would make a song appealing. It's not necessarily the sounds themselves. I can sit on my piano and make different sounds and there's a piano behind me, right? I can make different sounds and it won't mean anything to you. I mean, there's no meaning to the high pitch keys or the low pitch keys. I mean, what makes music is the transition and the dynamics and all those kind of hierarchies and groupings that happen in all kind of levels. And in order to understand it, you have to understand, you know, what kind of expectations are formed. When you listen, you can hear one sound and another one, and you always try to predict what will come next, right? So we can create expectations, we can validate them, we can break them, we can make surprises. We can make something that is very athletic or make something that is very fame that has. So we should think about song in a similar way we think about music rather than try to use a blunt instrument to say what's going on. If there's one thing you remember from this talk should be this.
Will Davis
Darcy has a question. And then after Darcy, Irene has a question.
Darcy
So actually I want to ask another question. But, you know, the famous thing about speech is that there are no silences between words, right?
Ofer Tchernichovski
So in fact, our concept of a word would, you know, an alien from space, just segmenting it on the grounds of intensity as a function of time, would not be able to come up with a correct vocabulary. So in some ways, although we've accepted the idea of these syllables for so long, if you think about it from the human perspective, it doesn't make that much sense. And then my other comment is, you know, really, you know, to whom are you singing? Right. So you're gonna. One has to therefore decide that, you know, either you're looking at sexual selection and you're looking at mate choice, or you're looking at some kind of male dominance hierarchy in which the one that ends up singing the most, you know, gets the mate. And is there any data on that?
Well, there's also a third thing you can think about is song as a friendly behavior. And in zebra finches, we have some evidence that song can be maybe friendly males like to hear songs. We know that the songs are reinforcing to them, even though they're not supposed to be the subject, it's supposed to be the females, but the dopamine level in the brain and the motivation to hear song is much higher in males than in female zebra finches. So I think that even here we should break this notion of that it's all about sex and violence, and there is more in life than sex and violence, even though both are very cool. Right. There's more. And we have to really think at the subtle contexts of to whom the birds sing, not individually. Also, what about the bird singing to himself? You know, I mean, many times you talk to your. I talk to myself all the time. Right. So.
Darcy
Oh, you sound like Fernando.
Ofer Tchernichovski
Is that a compliment or an insult?
Darcy
It's neither. It's an observation. Thanks. It was a great talk. I really enjoyed it.
Will Davis
Irene, you had a question?
Irene Pepperberg
I mean, this is really interesting because what we're learning from the parrots, learning human speech, is they seem to be very sensitive. Not so. I mean, somewhat at the syllable, but more at the morpheme. Okay. And the point is Griffin trying to learn how to say too. And for years he'd go, ooh. And finally he's going t oo okay. With the space between it. We're trying to get together. But the point is. And Alex was learning, you know, spool, and he went wool with a big space, and then it was swole and then, you know, spool or one for seven. Okay. And then to get to seven. So I think there's a real cultural, you know, issue in there that we really have to be, you know, that you're talking about, but is really very clear when we're seeing, when we're trying to cross the species barrier, that there is this cultural learning as well as to what is important within the culture of what's being learned and what's important for, you know, in both directions and how you're going to pull that apart.
Ofer Tchernichovski
Yeah, yeah. I mean, I can't agree more. I mean, understanding culture is a very interesting challenge across species. And I feel very privileged that I had a chance in my life to study that a little bit because, you know, culture is such a crazy thing. It's who we are, you know, as a human being, you just don't. Can't even understand what it means to exist without a culture. You won't even be able to talk. Right. And understanding, you know, how an individual, a parrot, like Griffin, fortunate to meet in your loved ones. Right. I mean, it's kind of like interesting to see how those animals individually. I remember that I once seen that he come up with new words. Right. Like the cork nut. Right. That was Never told anything about it, but. But he came up with that just based on his life experience. And then, you know, when I heard him saying cork nut, I couldn't stop laughing. You know, it was so amazing that he came up with that. Yeah, the almond. Not the almond. The almond in the shell. So he got an almond in a shell, but you know that you discovered it.
Irene Pepperberg
Or “baneri". Okay, so banana cherry. The elision for banana cherry for an apple.
Ofer Tchernichovski
Yeah, yeah.
Irene Pepperberg
My wildest dreams. I would never come up with that myself, but, you know, Alex came up with things like that, so.
Ofer Tchernichovski
Yeah, yeah. So the thing is that there's one thing about an animal figuring something out, there's another thing about how the culture is evolved on that. Of course. I mean, in the parrots, it's almost impossible, right, because you can barely figure out what one of them. But in the smaller birds, we may do some of it. There was a question from Ted, I think.
Ted
Yes, hi. Just a observation. So I do a lot of work with kids, hearing deaf kids of hearing parents who have learning deficits because the parents don't teach them sign and because they can't learn English because they can't hear it. But one of the things we've seen in the literature is that these kids, if the parents just try to learn the vocabulary, just give them something to start with, the kids will actually start filling in the rest of asl. Sort of like the diversity you're seeing with the, you know, from the tutor to the pupil, they sort of make up for the lack of that diversity. I don't know where I'm going with this, but it seems like there might be…
Irene Pepperberg
the Nicaraguan sign language studies.
Ofer Tchernichovski
Yeah, that's what I wanted to say. I mean, this was studied very nicely in Nicaragua, where you could see the innovation, but in this case, it was the innovation of the very young kids that really regularized the language in the next generation. So. So we could refer you to some very nice studies about this process, which people actually compared to what we found in the birds. Yeah.
Irene Pepperberg
Susan, Golden Meadow.
Ted
Oh, yeah, yeah, yeah, you're right.
Will Davis
I think. Vint, you had a question.
Vint Cerf
Two of them, actually. I'm a know nothing in this space, so I get to ask really naive questions. The first one is that since the birds are often in heavy foliage, they probably can't actually see each other. Sound carries, obviously. So isn't part of the singing an important part of localizing another member of the same species?
Ofer Tchernichovski
Not in zebra finches. Zebra finches sing from very, very close distances when they see each other. So it really depends on the species. In some species, it has a lot to do with localization because they sing from very large distances and they don't tolerate each other, so they're territorial. Zebra finches are not territorial. But it's not only zebra finches. Starling also are not territorial and sing from very short distances. But definitely in some species, localization is important. But of course, in this case, the female kind of like go to the forest and she listens to songs coming from different directions. I think it's not very hard for her to know where the songs are coming from. And it's really a game of looking at the quality of the territories, look at the quality of the song, and which basket should I put my eggs in? Right.
Irene Pepperberg
Literally.
Vint Cerf
The second question has to do with an interspecies matter. Is there any indication that between species there's any useful signaling? We've seen within species, all kinds of warning signals like the ones that describes for prairie dogs. Is there any indication that some species can communicate with others in the form of buzz off or, you know, I'm defending territory not only from similar species, same species, but also others? I have no idea. But I'm curious to know if there's an interspecies element here.
Ofer Tchernichovski
I'm probably not the right person to ask this question. I think that there's a lot of data that suggests that animal communication goes across species quite a bit more than people thought in the past. Those are not my studies, so I should not talk much about them. But I think that Irene and other people can tell you a lot about me.
Vint Cerf
Irene looks like she's getting repaired.
Irene Pepperberg
Yeah, there's a lot of mobbing calls that are interspecies. I mean, you walk into a forest and you'll hear the squirrels starting to chatter when they see you. And then the blue jays will pick up and start calling. And then all of a sudden the forest becomes quiet because everybody knows that you're not, you know, that you're. You, this intruder are there. And there's a paper that I had to learn all this stuff to do a paper for a review paper. And so there's lots of these cross. Cross species where there's actually a lot of collaboration between species that one species will argue that there's, you know, will alarm for the ground predator and then other species will alarm for the. So it's myna birds and monkeys and the myna birds will alarm for the hawks that are coming over and the, you know, the monkeys Will alarm for the, you know, pythons or something, which are both predators for both species. Okay. And they collaborate on things like that. So there's a lot of. If you're interested, let me know and I'll see if I can dig up the paper.
Vint Cerf
I'd love to see that. If you can share that, shoot me.
Irene Pepperberg
Send me an email so I don't forget.
Vint Cerf
Yeah, we'll do.
Ofer Tchernichovski
This is actually very early in evolution that communication across species evolved and you can see it even in bacteria that they do a lot of coordination and communication across species of bacteria to coordinate an attack or to coordinate turning light on so that cuttlefish can see where they go and they can take some crumbs from the meals. So this idea of trading, you know, information and in fact it has to do a lot with, you know, the evolution of cooperation and mating systems that really co evolved, you know, sharing genesis and sharing cooperation and information that really joined together very, very early stages, not only within species.
Will Davis
Con?
Con Slobodchikoff
As you know, there's a lot of information being discussed now about different personalities in animals where some animals are bold, some animals are shy and so on. Is there any indication that this is the case in zebra finches? And maybe the bold ones are the ones who copy imperfectly by making up new syllables, whereas the shy ones are the more traditional ones who stick with what the tutor is teaching and never extend themselves. And if so, if that is the case, would there be any fitness value to being bold versus shy in zebra finches?
Ofer Tchernichovski
Yeah. So I suspect that it's actually the other way around, that the bold one are more limited in their imitation and the shy one are more creative, maybe. And I think that those relationships are really multidimensional. And if anything, we are underestimating how complex animals are, you know, in terms of their behavior. People were trying for many years to talk about behavioral syndrome and personalities. But, I don't think that they did good enough job yet. I think that just like humans are complex, animals are complex. And you know, people always remember, once I've seen, I think in the movie Kinsey, who was of course famous for his work on sex, but he was a zoologist at Harvard and he was looking at the microscopic parasitic wasps. And in the scene, I don't know if it's really happened like that, but he's looking under the microscope of how those little microscopic parasitic wasps behaves and mate, and he tells the student, look, everyone is different. They're not the same. Everyone is so different. And I know that it's true even when you look lower in smaller creatures, that. And of course, when you talk about animals as complex as a zebra finch, you know, there is a really complex array of what makes the animal what it is, not just. You cannot really say, oh, yeah, this is my personality. You know, it can be summarized in two words or three sentences. Like people doing in dating sites, right? Like I'm like this, I'm like this and like that. And then maybe we are a good fit. And I think that as we already learned that it doesn't work so well in website, in. In dating sites, that in a sense, we ended up using things like Tinder much more than we're using the more complex ones, because the complex ones are not so helpful. Just look at the faces. You can do equally well or equally badly. And so my kind of feeling is that, yes, there's a lot of exclamation mark. We need to understand the animals and, and not to reduce the behavior to a few categories, but try to look just like not cutting with a butcher knife. The songs don't count what animals are with a butcher knife. And don't decide what are categories ahead of time. You know, it requires a lot of work and it's hard and maybe not a good career as a scientist, but if you really want something that adds signal to the noise, it needs to be done.
Will Davis
Yeah, perfect. I think that's it for the questions, unless anyone has anything. Well, if not, thank you so much, opa. That was absolutely fascinating. I think everyone really, really enjoyed that. Thank you. And I hope all of you will be able to join us next month for our conference. And final details will be going out at the beginning of July, so please keep an eye out for that.
