 Okay. I'd just like to start by thanking the organizers for inviting me and to say that I found this isn't particularly my field, but I found it extremely interesting listening to the talks and I hope that you find, sorry, that I think the, here we go. And I hope you find something of interest in this talk as well. So it's not news to you of anyone here that animals rely on social information when in grouping and for foraging. But it's also become increasingly obvious or there's increasing evidence across the animal kingdom that animals also use social information or social learning to acquire new behaviors and new skills. For example, migration routes in the top right there or new tools or tool use behavior. But when we see this social information shared and maintained in population, we can also observe this emergent phenomenon that we call animal culture, which we could widely sort of define as behavior shared by members of a community that rely on social learning. And perhaps we could also say it's particularly important that this information is transmitted across age classes, is maintained in the memory of individuals and is a sort of trait based copying. So we have a sort of cultural trait that exists outside of individual behavior. And this has been perhaps best studied in the field of birdsong. So there's a very large literature and history of research on birdsong showing that it's socially learned often from fathers to sons can be extremely complex. And for example, as shown in experiments on zebra finches, it can be developed over generations. So if you isolate acoustically young zebra finches, it takes something like five generations before they get back to a sort of complex song type. However, we can also observe these sorts of cultural traits in other behavioral domains. For example, in the classic case of chimpanzee nut cracking, where we have a behavior in here, sort of percussive tool technology transmitted from older individuals to younger individuals. And in this case, we can further say that this sort of cultural behavior, which is inherited over generation, emerges from an interaction between three different traits or three different influences, the cognition of the individuals, they have to be capable of learning this behavior and learning it. They are and sort of learning it socially learning it. They also it relies on the ecology of the individual. So in the case of chimpanzee nut cracking, it appears to be particularly important in times of drought when other food sources are unavailable. And these are these nuts are and it also relies on the underlying social system of the species. In the case of chimpanzee nut cracking, this behavior appears to be extremely stable. In fact, there's archaeological evidence that they've done this for thousands of years in this site in Western Africa. However, cultural behaviors aren't actually necessarily static because we can see or study them as cultural traits. We can also study the way in which these traits change over time. And we're starting to build a sort of cultural evolution framework to study this where we take sort of the underlying or the very well understood framework of genetic evolution and look for parallels in cultural traits. So we can show that we have variation in cultural traits that can then be a selected upon inherited over generations. And then this could lead to what we might call cultural adaptation and that we can also have, for example, an influence of movement or social systems or geography on the patterning of cultural behaviors that we observe. However, if we want to study this, which I do, we need to either do it in an experimental framework or look for sort of natural experiments in the wild. And these are often fusions of new behaviors that come into populations in response to changes in the environment. And one of the most famous of these was, in fact, done by this small bird that's called the great herd. And this is it stealing milk or cream from the top of milk bottles in 1920 or in the early 1920s in England. And this behavior was first observed in the south down here, but then spread over the next 20 years across the south and was also kind of popped up again in northern England and spread from that point to with these sort of exponential uptake as you can observe on the lower part of the slide there. And this sort of these observations started really creating a whole field of thought around these diffusions of new behaviors that we might observe in animal populations. And in fact, the diffusion of innovations is a very well studied area of economics. For example, we can look at this slide of technology adoption, which is one of my favorite to show you because you can see the sort of premature death of the VCR in them. And we can take, well, we started to think we could potentially take some of these techniques to sort of understand these processes as they occurred in animals or as they potentially occurred. So I've been studying this question of the last few years in this population of tits in the UK. It's a long term field site where we have all the birds marked. And we also have the birds microchipped as well, which means that we can basically attract them to sites with antenna and then interact with them based off those microchips and the RFID interface. And this has allowed us to build social networks for this woodland based off an array of feeders that the birds come to where their pit tags are detected. And we look for spatial temporal co-occurrences in birds in these sorts of roaming fission fusion foraging flocks. So to return to this example of the milk bottle innovation in tits, we don't know for sure what happened here because it was observational, but we can hypothesize that we had a pathway which may have looked something like this. We had a novel challenge or opportunity here, a new resource leading to a behavioral innovation on the part of some number of individuals. Social transmission of that information across the population leading to a population level change in behavior. And then potentially this behavior could change over time. It didn't in this case because we redesigned milk bottles, which they were then impossible to crack open. So the first part of this, I'm going to talk about how we can possibly go from this sort of novel opportunity through to a population level change in behavior and the sort of social influences and cognitive influences that affect this transition. So I'm not going to go too much into the stage of the sort of innovation or where the behavior initially arises. But we have studied this myself and several other people for extensively at Oxford over a number of years. And these birds are both extremely innovative. And also we can pinpoint the particularly innovative sectors of the population using experiments like this. So here the social learning information is broken because each individual gets their own sort of learning program. And we can look at the rate at which they learn this in the wild, this sort of this one has to learn to press the red button to get food. And we find that young individuals tend to be the most innovative and also less competitive individuals. So those low in the dominance hierarchy. Okay, so we've got an innovation. How do we go to social learning and transmission of that innovation to others. So to look at this, I performed a two action control experiment. How we refer to it in the cognitive equality literature, which is where I gave them basically a binary choice test. So they had to slide a door, they could do that by sliding it right or left to access a food reward. And I brought birds into captivity. And I trained them to do this. And in one pop or two populations, the individuals I trained were trained to slide the door from the left towards the right to get the food. In three populations, those individuals were trained to slide the same door but from right to left. And in another three subpopulations, they weren't given any training at all. So these are controls to test the underlying rate of asocial learning. I then released these birds back into their respective subpopulations to be the initial innovator. And what did I find? Unfortunately, this video doesn't seem to be working. You think it's going to work again? We have light. So from an initial innovator trained in captivity to slide this door and access the food, we found actually that the information spread very widely. With hundreds of individuals in each population also solving. And this cage is only to prevent squirrels, birds fly straight through it. And what's important to note is that birds could equally solve any of these devices using this different alternative, equally rewarding, equally difficult technique of pushing it from the other side. So the fact we saw a bias towards the behavior that was introduced by me or by the innovator in each population was good evidence that we had social learning. And in fact, we saw an extremely strong bias towards solving variant A if variant A was introduced, or variant B if variant B was introduced, and no obvious bias in the control populations, suggesting that we had the spread of information resulting in localized traditions. The knowledge of these solutions spread with these sort of classic sigmoidal diffusion curves that again, we tend to see in economics where you see an initially so start where there's not many knowledgeable individuals and exponential spread. Once you have about equal numbers of knowledgeable and naive individuals, an eventual plateau, which here was about 80% of the population. We could also track this through the foraging networks. So here the yellow node is the innovator individual and the turning red as they acquire the behavior. And hopefully you can see that we had very strong evidence that it spread through social network ties. And in fact, we could use social network analysis to ask whether individuals were more or less likely to learn given their social position. And we found that young individuals again, were about twice as likely to learn as older birds. So that was over one winter. However, we did put these foraging puzzles out over successive winters with no further training of individuals in a subset of the populations. And we found good evidence for long term persistence of these traditions for either pushing left or pushing right in each population. And in fact, we saw it emerged about twice as fast the next year as the first year, which is basically down to the fact that we now have a larger pool of knowledgeable individuals. So that's those sort of old individuals in the population. So we have extremely stable traditions. But this seems surprising given it's a completely arbitrary choice for left or right. And it appeared that the reason we had this was because individuals work copying randomly, but we're in shed showing frequency dependent conformist copying. So they tended to disproportionately adopt the behavior in their group that was the most common behavior. And this led over time to populations converging on a single behavior and these single behaviors of choices becoming fixed in the population. So you're probably very used to seeing these sorts of positive frequency dependent curves in collective behavior, but they are much more unusual in the learning literature. And we hypothesize that the function may be actually as a shortcut for learning in spatially variable environments. When individuals move between patches, they can quickly adopt what is probably the best locally adapted behavior. But the consequence is that group traditions are stable and resistant to erosion over time. So we're perhaps important for the sort of evolution of complex culture. So I haven't had time to go into all of these different influences that we see in the transmission between this sort of innovation on the part of one individual to a population level change in behavior. So you'll have to take my word for it that we see both cognitive influences, for example, conformist copying, but also a strong influence of the patterning of social interactions and population demographics. So I want to move on now to this last part of the question here, the pathway. So once we have a population level sort of behavior that's established through social information use and then transmission across networks, can we then study that behavior in itself and how it changes over time and what it changes in response to. So again, if we go back to this idea of studying it in a cultural evolution framework, I just want to, obviously I can't go through all of this, but I'm just going to pinpoint a couple of experiments we've done to look at how some of these processes might influence the eventual outcome. So in particular, I'm going to highlight how traits themselves can be under selection either through learning biases, as we've observed, we have these sort of almost stabilizing selection force of conformity, but also through an interaction with individual learning. So when we observed this behavior in particular, we observed that when we looked more closely at it, that individuals varied quite a lot in the speed in which they were able to solve the device. So this is one that we would call quite slow, an average bird and one that's much faster. And you may also notice in this video quite a lot of scrounging, and if you're interested in that, come talk to me, which is we've looked at that in a separate study. So actually, when we look over time, we see that birds reduce their speed of solving with this almost, it's not quite, but almost power law function or learning curve. And they become very, very fast over sort of hundreds of solves. And this particular one goes from about 20 seconds on its first solve down to one second once it's solved a couple of hundreds of times, hundred times. And we actually see that populations as well become faster over time and that within a single generation, birds that arrive later into the population learn later seem to benefit from this population level improvement because they learn a faster version of the trait on their first sort of go, they start lower down in this curve. So that's within one generation. What effect does this have on the across generation inheritance of behavior? So we looked here at again at populations where we had up to three generations of this. And we found that when you look at the first generation, the average time of it takes for a bird on its first attempt is about five and a half seconds, the first successful attempt. And those birds that survived the second generation, so survived two years because of average lifespan is only two years for this bird didn't differ. But birds that first learned in the second generation are now showing a much faster speed on their first attempt. So the knowledgeable birds are extremely fast reflecting that learning curve. But the naive birds also appear to benefit from the sort of cultural memory in the population that experience acquired by the older individuals. And when we look over three generations, we see this pattern continuing until it appears to sort of bottom out once the birds just can't get any faster, or at least we can't measure them being any faster. So this is actually quite similar to what's been proposed, for example, by Dora Biro as a kind of collective memory or collective cognition that you can have in groups. And we would probably call it something like a cultural legacy effect where the behavior or experience of previous generations is important for the learning experience of you. So I now want to ask what happens when the actual payoffs change in the population, whether we can see selection on cultural traits in response to that. And you can also think of this as sort of a question about adaptation. And the results that we found for this came out of a study where we were actually looking at a slightly different question, which was to ask whether traditions were vulnerable to temporal changes in the environment, in particular these sort of conformist traditions we see. So there's a big debate in this literature about whether or not we can possibly find maladaptive traditions, because we see them in humans, but we don't really have any obvious evidence in any other species. So to look at this, I did a changing environments experiment. So I went into four populations, so two where the established tradition was for pushing right, and two where the established tradition was for pushing left. And here either variant gives the same payoff. And I first varied it so that either variant gave a low payoff seed instead of worms. This was to look at the rate of individual exploration that individuals might do in response to receiving a lower payoff than they expected for an established behavior. And then I played this sort of cruel trick where I made the very rare action, give a higher rewarding payoff, the worms, and the established tradition now gave seeds or a suboptimal payoff. And so I was really depending here on this sort of reservoir of information that was the individuals who still preferred the other side. There was always a small proportion in each population of about between two and five percent of individuals who still preferred the uncommon side. So to show you what this looks like, here's a bird that's just arrived to the sunflower side being open, the sunflower seed side being open, and it knows exactly what's going on and pushes the door the other way in order to get the better food. So what did I find? Well, here we are at the initial starting point where we have very strongly established traditions for either left or right. And we don't see that change at all in the middle stage where we provided only the lower reward. So we don't seem to see any sort of sampling behavior of the other side in response to receiving a lower award. But when we give them the unequal rewards, we do suddenly see a jump. So after a three week period, about half the population has now completely switched to preferring the alternative behavior. However, this isn't instantaneous and it's not complete. And so it was sort of a bit of a messy result and we wanted to know what was really going on here. So to do this, I went and sought some help from Richard McAworth and we built a model, a sequential learning model to look at what birds were doing at each point in the puzzle box. So here we gave them a parameter for their own memory. So what they did previously but also what they then observed when they arrived at the puzzle box at that attempt. And we found this is sort of rough, but generally that while birds relied mostly on their personal experience when choosing what to do, they also continued to incorporate social information to actually quite a high degree at each new attempt at the puzzle box. So they were still watching birds even though they knew how to solve themselves. And we separated this out and asked what form of social learning they were doing here. And we were a little bit surprised that actually it remained conformist. So we thought they might show pay-off bias copying where they were preferentially copying individuals which were receiving a better award for their behavior. But they didn't appear to be able to distinguish the awards of the birds they were watching at all. Rather, they continued to be sort of stubbornly conformist, even though we do see a bit of variation we still see this strong disproportionate tendency to copy the most common behavior in the group. So we have a paradox of conformist individuals, but a flexible population. So what's going on here? Well, it seems to be solved by an interview by a combination surprise of individual cognition and social networks. So while individuals are conformist in their social learning, they're pay-off biased in their reinforcement. So once they've had the chance to personally experience a behavior, they prefer to do that good behavior in future. So their their own memory or their own reinforcement or behavior, personal experience is pay-off biased. They get the chance to occasionally gain those experiences because the network is fission fusion and constantly remixing. So very occasionally they'll end up in a group where it just happens that most of the group is doing the uncommon high-pay-off behavior. They then get the chance to learn that. And once they've learned it, they prefer to do it themselves in future, which then leads to more social information for other individuals. And you end up with an initially slow start, but an increasingly fast snowball of this new better behavior through the network, as you can see here with the birds changing over from the initial kind of source of those green two notes, which are those maverick individuals. That said, we still didn't see all individuals change and they didn't change at the same rate. And this was because we also saw again a very strong influence of age. So this is an increasing ongoing theme through my work that we really see age-based differences and that you have to consider age structure of populations. So younger individuals were actually less likely to be conformist, so they had a lower conformity exponent. And they had they were better able to update their own behavior once they had had a chance to experience new information. So they were more flexible in their behavior. And this meant that by the end of the experiment, individuals that were one or two years old had almost all switched to the high-pay-off behavior. But individuals are sort of, you know, up to five years old, those really old individuals hadn't changed. And so again, this emphasizes that the demographics of the population are actually very influential in determining the sorts of behaviors we see in that population. And now I want to move on to looking at sort of more neutral processes, we might call it, so the influence of movement of individuals on the cultural traits we observe. So to look at this, we worked, I worked with Tamara Somerville, who led this work, and we looked at how the movement rate of individuals between patches interacted with the behaviors that we observed in each patches, even given that individuals were conformist. So we knew from previous work that when individuals moved between patches, they tended to change their behavior to match the patch they moved into. But we hypothesized that if movement rates were very high, individuals wouldn't be able to do this fast enough to still maintain different traditions in each patch. So he built a model of individuals moving between patches, and this is just the example of it based off the same woodland area, taking into account the movement rates between individuals. You can't see the network, unfortunately. And in this case, it's where we have very high rates of movement and no conformity, and you get a complete mix of behaviors across the population. If we have very relatively high rates of conformity to movement, we end up with a domination of one behavior across the population. And in fact, to see the patterning of local traditions, it depends very much on the sort of balance between movement rates and conformity. So when conformity is strong relative to movement rates, but movement rates are still sort of if they're very high, you end up with domination. But if they're kind of quite low, you end up with this patterning of traditions. And in fact, the further you fragment habitat, the more cultural behaviors you see across the landscape. So we have an interesting lesson in how much in considering geography and movement as well as just individual cognition, we're looking at group behavior. So the final thing I want to talk about is this question of increasing complexity or accumulation of modifications. And this is something which has been a big source of debate because humans are very good at it, leading to incredibly complex cultural traits like international space stations. But we don't obviously have evidence for that and other animals and so it's sort of an ongoing argument whether or not we still might observe some change or increasing complexity or multi component traditions over time in other species. So to look at this or how we might observe this arising. I introduced a sort of different puzzle. So this is what I called the dial task or a telephone task. So you have to turn this in order to get the food reward. And this when I introduced this into the populations, I observed very similar sigmoidal diffusion curves in the experimental sites as in my initial experiment. I then entered the sort of step two of this experiment which I called the training phase or the recombination of traits. So here what I did is I combined the sliding door task with the dial task. And so I had some individuals who knew both because they were old, they'd done this in previous years and now they had just learned this. I had some individuals which had only were only familiar with this. And I gave them an incentive to combine the tasks because I gave them a better reward if they did both. So here you can see a bird that's dialed and then also slid the door as well in order to get the better award. And in fact, I saw that if birds were aware had learned both of these components, they were able to put them together very successfully to do this multi step task in either order. So some slid and then dialed. But yeah, they were able to solve it. And if they were able to they had learned one of the components, they were slower but they were still able to learn the entire multi component task. I then gave this to the second generation. So I put it out the following winter. We know for the training of any kind depending on the old individuals that had knowledge of the whole multi component task to be a source of information for younger individuals or those new naive juveniles in the population. And I also only gave them a reward if they solved the entire task. So they had to learn a full multi step two component task. And I found that old individuals were very successful at doing this. They already had knowledge from the previous year which they retained and learned again very fast and younger individuals. Oh, I should also say in control population we saw absolutely nothing. So there seemed to be basically impossible for them to innovate the solution to without any social information. But with younger individuals, I saw that they could learn the task. But I found really interesting the diffusion curves had completely changed. So instead of this sort of classic S shape we saw before, we saw a very slow start where no individuals appeared to learn before this exponential takeoff but a plateau at a much lower level in the population. So it seemed something more like only 30 percent of individuals were actually able to acquire this behavior, even with the social social information available to them. So this seems to be right on the edge of what they're capable of doing. So I had to think of it as this particular task was perhaps less of a radio example and more like the airplane where we have very complex tasks that appears to be very difficult to them to learn and that changes the way in which it diffuses through populations. We have to do some more work on this, but I think it's really interesting to speculate how this would also change how it could be maintained in populations over generations. So just to conclude, we saw the diffusion of innovations through social networks in birds leading to arbitrary traditions that persisted over generations and the dynamics and passion of this behavior was strongly influenced by cognitive and social factors and the interaction between them. Individuals and populations can refine learned behavior over time through an interaction between individual learning and social learning and in fact can transmit this refined generation to the next refined brain generation to the next refined behavior to the next generation and modify it to form multi-component traditions. But even so, even given how kind of persistent and strong this behavior is in populations, it's still actually very flexible under environmental change and can show evolution or cultural evolution over time. And with that, I'd just like to thank all my collaborators that have worked on the project and the funders. And thank you for listening. Thank you very much, Lucy. Time for questions.