Instant Genius - Why birdsong is the pop music of the natural world

Episode Date: May 17, 2026

When it comes to the sounds we encounter in the natural world, few have the beauty, elegance and variety of birdsong. But these distinctive vocalisations are not simply static calls that are common to all birds of a certain species that are passed on from one generation to the next. They vary from region to region and even evolve in single populations over time in much the same way that the style of music in the pop charts changes over time. As part of our Science of Sound miniseries, we’re joined by Dr Nilo Merino Recalde, a senior conservation scientist at the RSPB, to talk about the fascinating science behind the evolution of birdsong. He tells us about his work on tracking the evolution of birdsong as it passes from place to place and from generation to generation, how advances in AI technology are enabling birdsong researchers to learn more about this process than ever before, and what this research can tell us about the evolution of culture in the animal kingdom as a whole. Learn more about your ad choices. Visit podcastchoices.com/adchoices

I saw my friend on the other side of the street. I was heading to school with the kids. I let go of mom's hand to wave. I had already forgotten their lunches. I ran over to hug her. She came out of nowhere. And then... It stopped.

Sometimes the moments that never happen matter most. Volvo's automatic emergency braking helps ensure a safe ending for everyone. Learn more at volvocars.caps.caps.ca. This episode is brought to you by L'Oreal Group. Beauty is a powerful force that moves us. That's why L'Oreal Group has built a business that is inclusive at its heart with 100% of its brands, championing diversity. With 25,000 professional opportunities for people under 30 worldwide

and 54% of leading positions held by women, diversity is a strength that helps L'Oreal Group create the best beauty products for all people. Visit L'Oreal.com to learn more. Hey y'all, it's Kelly Clarkson with Wayfair. Ever order furniture online and wonder what if? Like, what if it doesn't hold up? That sofa was four days old. You should have ordered from Wayfair. With Wayfair, there's no what if. Just style you love and quality you can trust. Visit Wayfair. Wayfair, every style, every home. Hello and welcome to Instant Genius, a bite-sized master class in podcast form. Every Monday and Friday, you'll hear world-leading scientists and experts talking about the most fascinating ideas in science and technology today.

I'm Jason Goodyear, commissioning editor at BBC Science Focus. When it comes to the sounds we encounter in the natural world, few have the beauty, elegance and variety of birdsong. But these distinctive vocalisations are not simply static calls that are common to all birds of a certain species and are passed on from one generation to the next. They vary from region to region and even evolve in single populations over time

in much the same way that the style of music in the pop charts changes over time. In this episode, part of the Science of Sound mini-series, we're joined by Dr. Nilo Marino Ricolme, a senior conservation scientist at the RSPB to talk about the fascinating science behind the evolution of birdsong. He tells us about his work on tracking the evolution of birdsong as it passes from place to place and from generation to generation. How advances in AI technology are enabling birdsong researchers to learn more about this process than ever before. what this research can tell us about the evolution of culture in the animal kingdom as a whole. So welcome to the podcast. Thanks so much for joining us.

Thank you for having me, Jason. It's an absolute pleasure. So let's kick off then talking about one of the projects that you worked on, studying birdsong in great tits. So can you give us a brief overview of what that project was, please? Yeah, I'm really happy to do so. First sentence about the background for that work And this is something that now everyone knows is that in the same way that humans learn a lot of the information that is relevant to our lives from one another, we're born, we experienced this kind of social milieu from which we acquire lots of information that we need to exist. The same is true for lots of animal species, right? Not all the information that they need about the world comes encoded in their genes, to put it very simply. A lot of that information needs to be learned from other individuals as they grow up, as they developed.

And this is something that was recognized really early in the case of birdsong. About 40% of all birds learn the songs from other birds, right? So they are born with the innate capacity to learn songs, but the exact content of those songs varies from place to place and across time, right? So that is at the core of this work that you're referring to. Where we were trying to understand if differences in the songs, and they vary quite a bit, have to do with what we would call demographic parameters or demographic processes. And this is simply who lives, who dies, who comes and goes, right?

Because if you think about it, who is around you when you're learning will really have an impact in terms of what it is that you learn, right? It's a very simple idea that has really profound consequences. So we were interested in seeing whether this is something that you could observe in the songs of birds. In this case, a really common songbird that is one of the commonest in the UK. the great tit. Yeah, so was there a particular reason that you chose to study the great tit? A lot of reasons, really. One of them is very practical in nature. That is, the great deed has

become kind of de facto model animal or model system to try to understand behavior and population dynamics of birds in the wild. There's a long history of long-term research done on them in places like the Netherlands and the UK. And for that reason, we just have a lot of information about these birds in terms of their lives, the reproductive biology, and so on. Also, the songs are quite varied, so they're good learners, but at the same time, the repertoire aren't huge. You know, we're talking up to 13 different songs or so. And as I said, the songs are quite varied, but they are also fairly simple. They don't do really crazy glisandos, or they don't modulate, they sound too much. So they're the really typical kind of teacher-teacher song

that lots of people might recognize go something like, and even though there's lots of variations on that theme, the fact that they're fairly simple makes them easy to understand. So this work was part of my PhD research that I did at University of Oxford, under the supervision of Professor Ben Sheldon, and he's been leading for the last couple of decades or so. These long-term research of great deeds

that happens in Wytham Woods outside of Oxford that began in 1947. So we really have a wealth of information on the epaids. To add to that, in the 70s and 80s, there was a lot of really pioneering research that was done on songs themselves by people like John Cripps and Peter McGregor that really allowed us to understand

what songs are for a lot better. In the end, you ended up with the sort of thousands of hours of recorded data. So how did you go about recording? the songs? Did you just sort of place microphones around in the woods or, you know, were they just normal microphones? Yeah, yeah, yeah, that's spot on. Yeah, that's exactly the idea. The traditional way to do this and to these day, the best way to do this is to go around the woods, chasing the birds with good quality microphones. If you want to have really good quality audio data, that is, again, the best way to do it. if you're trying to infer things about learning, who learns from whom, and how this really kind of

hard-to-measure parameters of age and social structure and these things influence the songs, then you need a lot of data. So I have done my first share of more traditional recording with individual kind of, we'll call it the microphones, but it just simply doesn't scale. So instead, what we did is we placed around 60 acoustic loggers across the woods. Now, the good thing about great deeds is that they really like nest boxes. So if you provide nest boxes, they're readily nest and breathe within them. And what that means is that you can know where they are.

And that in turn allows us to place the recorders close to where the action happens. Yeah, so once you had, you gathered all of this, you know, vast amount of data. You actually used AI to sift through it. So, you know, without getting too technical, can you tell us, you know, some of the sort of key points about that process? Yeah, so there's a lot of hype around AI at the moment, partly understandable, mostly for marketing reasons. But it is a really important tool

when it comes to analyzing large datasets like these, and especially when it comes to less traditional media. So you're trying to understand counts of something. There's lots more types of models that you can use and so on. When it comes to things like sound or images, it's a lot harder to reduce, to simple variables that you can understand easily, right? So that's where machine learning comes into play.

So what I do specifically, and again, when I go into the technical details, and there's many of those, is I used a vision transformer, and that's a model that is kind of like chatyptipad 4 images, to try to learn something about the sound that might allow us to distinguish between individual birds, right? So it's a family of models that are not dissimilar to what you might use if you're trying to re-identify people based on the face,

And you can do this across thousands, potential millions of individuals, right? So we tried something similar for birds. And what we did is we labeled some repertoires of known birds, and we tried to see how well the model could find those birds again. And these led to two really interesting and useful outcomes. The first being that we were able to re-identify birds across years, right? So say you record the bird in 2020. If you hear that bird again in 2021, the model is going to say,

oh, that's this one bird in particular, in particular, without even having to catch it again, right? And we validated using actual recapture data from bringing the physical rings in the birds. At the same time, what this reduces is a really detailed map of cultural diversity and variation across the entire population. Because if you think about it, you take two pairs of songs and you say, okay, these two are more similar than these two. And you did this and iterate this over thousands and thousands, hundreds of thousands

of songs really. Over time, what you end up with is this map of things. cultural similarity. And this is what allowed us to infer things about who was learning from whom and about these processes of cultural variation and where that variation was coming from, which we can talk about more in detail later if you want. Are you one of those media strategy people clicking through slides, scrolling spreadsheets? Yes? Good. This is for you. Because on Spotify, there's an audience that's different. Locked in. Loyal, invested. They're called fans.

Fans don't just listen to music. They feel seen by it. like it belongs to them. So when your brand shows up on Spotify, that's who you're talking to. And you're right next to artists like me, Lizzo. So, are you ready to talk to fans? Spotify Advertising, you're among fans. You mentioned these variations in the features of the songs. What were some of the things that you were looking for there?

I think that's really interesting. Yeah, yeah, no. I think it's fascinating. That's one of the reasons I do is we don't really have a good understanding. of what it is in the songs that is important. We don't have a good understanding of what it is that is kind of more constrained by biology and what is more free to vary culturally with learning, right?

Beyond the basics. I mean, we know, yeah, bigger birds, sing, deeper songs in general, things having to do with the physical properties of sound production. We understand reasonably well. In terms of what we would call song syntax in analogy with human syntax, we don't really know that much, right?

So we hope to use algorithms can infer the irrelevant variation from the data itself. So to answer your question directly, I didn't know what I was looking for. I was hoping that that would emerge as part of this process of trying to understand, right? And what you see there is that what seems to be important is the precise combination of sounds and rhythms, right, the melody and the rhythm of the song. So there are some basic elements that the birds use, some notes, if you will, and it is how exactly they arrange them that matters and that changes over time leading to this process that we call

cultural evolution. So again to draw that analogy with humans, and it is not a far-fetched one, this is like language or music change, right? Over time, people learn tunes or they learn words and, you know, kind of structure of languages from other individuals as learning, and they always introduce some differences, partly because of copying errors, partly because of directing intentional modifications that they make. And over time, it just accumulates, right? And this leads to variety and diversity and the things that we enjoy when it comes to human culture.

This is similar for the birds. With another extent to which this is particularly important to them, we know that over time they might lose the ability to kind of recognize birds as members of their own species if the songs that are vets too much. But it is the same process. Things accumulate. And exactly what it is that varies within the songs

will also depend on the species and even the region that you're studying. So let's have a look at some of the things you found there. So sticking with the cultural idea, there's differences with the songs sung by birds of different ages, which I think if you draw a parallel between humans, I think that's quite interesting because not everyone listens to the same music as their parents.

No, exactly. And again, you have to be careful with these analogies, right? You don't want to anthropomorphize too much, but also a little bit is fine because one of the issues that we've had in this field is that people have steered away of drawing parallels. Other of the fear, perhaps, of finding out that humans aren't so special after all, right? Ivan particularly encouraged these parallels if they go too far, but I think a little bit of analogy is useful to understand these kinds of things, right? So if you think about humans, there are changes in the frequency of things in our cultural world.

think of baby names, think of popular songs. All of these things change over time, and the main reason that they change is that they come in a other fashion. And why that happens has a lot to do with the frequency at which they are found in the population, right? No other extrinsic factors is just about the frequency. So thinking that are more popular, become more popular, and so on, and this kind of feedback loop that rises.

Really similar for birds. So birds have a sensitive period. when they need to learn the songs. Not all of them do, but lots of songbirds do it this way. All the birds like Starling, for example, can continue to learn through other lives, but great teachers have to learn within the first year of their life, really. They have this kind of sensitive window.

What that means is that whatever they're exposed to when they're learning becomes crystallized, and they're stuck with that for the rest of their lives. So it's a simple process if you think about it. Birds are born, they learn songs, they carry on with their lives. Most of them don't live very long, but some do. And then at some point, or the younger birds in the population, will be starting the same process.

If they have lots of old birds around, they're going to be more likely to learn the songs that existed a few years before. If they mostly have younger birds around, they learn whatever it is that the new youngs are learning. So it is a very similar process in that respect. How about sort of different, not really migration, but birds moving around in different areas, what sort of effect does that have?

So it has lots of complex. and interacting effects, but we summarized it in this research by looking at how far birds moved from the place they were born to the place where they ended up breathing. And that is a good proxy for how many other birds they encountered along the way, right? So what we see in this particular population is that birds that move a lot end up singing songs that are a lot more common. This might be slightly counterintuitive, but again, if you think about it for a second, it will become a lot more clear.

If you're born somewhere and you have, say, a set of 20 neighbors around you that you're exposed to when you're learning, you'll just learn those songs, right? If those birds themselves have not had many models to learn from many tutors, as you might call them, then they'll continue to sing the songs that are popular in that area, in that region. But they won't really know what it is that all the other birds in the population sing. Whereas if a bird moves around a lot while they're learning, they're going to be exposed to all of these different song types. So they'll have a better idea of what it is that is popular. And we think that these birds have what's called the conformist bias.

They're more likely to learn the things that are more prevalent, right? And this is a simple process that leads to more cultural kind of homogeneity where birds move more and therefore mix more. So do you think this sort of the findings and this behavior that you've discovered here can be applied to other species of birds? and perhaps even further extrapolating other species of animal. Yeah, I should say that this research is an example of something that we expect to be the case and that is not an idea I have had.

It's a kind of a collective understanding that has been gained by a lot of people working in this field for many decades now, right? So the reason we did this research is that we wanted to test whether those ideas and those more theoretical models will hold when you try to study the process at the scale at which the learning itself happens, right? So we know that birds in different areas have different dialects, and we are convinced there is empirical evidence that this is because of this process of learning and accumulation of differences and copying errors and so on, right? So what we did here is to look at whether you could actually measure these and use these songs to infer what the population is doing in terms of, you know,

age structure and movement and so on and so forth. And this is the exciting bit we saw that that is indeed, case, right? And that opens the door to a lot of cool research. But the theory, the broader theory, is something that we think applies to any animal, any species where there is learning involved. The processes in the broader sense. In terms of what happens exactly, that is something that very much depends on the biology of the species, right? So some birds, for example, learn, as I mentioned earlier, only during the first part of their lives, or others continue to learn throughout. Some birds are more likely to adopt what,

other birds sing more commonly, others are the very opposite. They are more likely to want to prefer to have rarer songs because that plays a role in sexual attraction and things like that. So the details of these processes really depend on the species, and this is something that we saw in whales that we see in chimpanzees, it's something that we see across the, you know, across animals and obviously humans as well, but also non-human animals. But the details really change with the biology of the species. And this constraint, animal culture in a way that human culture isn't. So off the back of projects like this and others, you know, can they feed into sort of conservation

work that we can do to, you know, perhaps help species that are struggling in some way by, you know, taking a sort of cultural approach, if you can say that? Yeah, no, exactly. And I think this is something that excites a lot of people, myself included, for two reasons. One of them is there is this increasing recognition that culture is important. important for animals. They need to be able to learn. They acquired information in this way that is of kind of survival and fitness relevance. So it's not just some entirely capricious outcome of the fact that they learn, the content of this learning is particularly important.

This is probably more so the case in species where the cultural traits that we are discussing have more to do with foraging and finding food and so on and so forth, but also this is true for and behaviors that have more to do with finding mates and coordinating behavior and so on. There is one reason. This increasing recognition of the importance of animal culture is something that we're all going to see become more important part of conservation projects. At the same time, and this is closer to what I've worked on, the fact that you can infer things about populations just by listening, really, means that you can potentially monitor them at a much greater scale in a way that is more cost-effective, but also in a way that is a lot less intrusive.

So instead of having to physically catch the birds, draw a sample of blood, analyze their genomes and so on, you could potentially just place a microphone across an entire continent, really, and in for migration corridors, in for whether the populations are increasing or decreasing, how much they're mixing, all of these things that are really important for their survival, individuals and as populations, you could infer potentially from behavior. Not true for all species, but even if it's only true for a few, this opens the door to lots of really cool research and conservation outcomes. So you sort of concluded this project a little while ago now. So, you know,

what have you been working on since? What are some of the exciting things that you've been doing that you'd like to share with us? Yeah, so I've been doing lots of different things. I have many different interest and I like to keep busy. So I was working at the University of Foxford doing more kind of large-scale monitoring, not across a single species, but across all of them, to try to understand how variation in the songs relates to, you know, processes that have to do with climate change and so on. I am senior conservation scientists at the RSPB, and we've been looking at how we can work with the Big Garden Bearwoods data to infer things about, you know, population trends and how the third are in gardens and things.

things like that. And at the same time, I've been working on projects having to do with data visualization. That's something I'm really passionate about. My background is in kind of visual arts and anthropology and feels like those. So I'm really interested in visual storytelling and representing big data in the form of images and infographics. So I've been doing a lot of that, trying to see how we can simplify the output of complex research, like the one we were just describing, and things like acoustic monitoring for

conservation purposes in a way that people like museum visitors, for example, can better enjoy. Thank you for listening to this episode of Instant Genius, brought to you from the team behind BBC Science Focus. That was Dr. Nilo Marino Recalb. If you liked what you just heard, then please do consider subscribing to Instant Genius when your preferred podcast platform. If you'd like to see our guests and hosts in person, then why not check out our YouTube channel at Science Focus. The current issue of BBC Science Focus magazine is out now. Pick up a copy wherever you buy your favourite magazines or download us on your app store of choice.

You can also find us on Apple News or online at sciencefocus.com.