Bristol Seely

Transcription

I'm Professor Nigel Franks and I'm based in Biology at the University of Bristol and I spend my entire working time trying to understand how ant colonies are organised. For me this is a fascinating problem because ant colonies are incredibly sophisticated. Far more interesting that anyone ever supposed, I think. In today's lecture I'm going to explain how they are organised and how we can do experiments to reveal their organisation. Ok, so this lecture is in a series of three that are under the umbrella of 'Complexity Science'. In complexity science one studies complex systems. An informal way of expressing that is, a system that is more than the sum of its parts. In other words it has emerging properties. And what we mean by that is, special things, novel characteristics bubble out the system. The parts interact and generate novel and surprising properties. One of the most impressive kinds of complex system is when we think about cognition and consciousness. This was drawn just last year but people have thought about whether there are similarities between ants and brains for a very long time. Only in the last couple of years have we made progress, scientifically, looking at this. And that progress I'd like to share with you today. That was a cartoon of an ant colony and this is a real honeybee swarm. It's probably got 50,000 bees in this living basket of bees hanging from a tree. The wonderful thing about social insects is that they are a collective organisation but they are built out of individual insects. The great thing about that is that you can take them apart and put them back together again. For me the ants, bees and wasps are the mechano or Lego or 'construction toy' of biology. The fact that we can take them apart and put them back together again in different configurations means that we can study their organisation in very nice ways. I'm going to suggest something very radical here; that we can even think of the whole colony having cognition. To do that we have to have a definition of cognition and I offer you this one; cognition is the ability to use internal representations of information, acquired in separate events, and to combine such information into novel information for application in an adaptive manner. In other words; you're internalising the information, you're coming up with novel information as a result and you can act upon that. Let me start with a very simple cartoon of ants. What ants often do is lay chemical trails. One ant can lay and trail and another ant can follow it and re-enforce it. So you can get positive feedback, with ants influencing other ants. One fabulous thing that can arise from this, and I think it's an example of collective cognition, is that the ants can solve the problem of working out which is the shorter path. I've spent hours making this animation. We have one ant running along the longer branch and one running along the shorter branch. The shorter branch ant got to the food more quickly. It will feed itself and go home more quickly, simply because it took the shorter path. As a result of the pheromones on the shorter path will build up more quickly and get re-enforced more quickly. So the entire colony will be able to choose the shorter path. The beautiful thing is that no single ant has compared those two paths. It's a colony level decision that's arisen out of positive feedback and very simple behaviour. Let's see if in real life and ants will really do this. I think this central clause; of lots of information from separate events, combined to make novel information. Here the novel information is which the shortcut path is. So, no individual has compared the paths. It's as if it's a colony, collective-level decision. That's the simplest example I can give you of collective cognition. Now what I'd like to talk to you about is an even more complicated problem. One of the main studies in my lab is 'house-hunting' in ants and other social insects. It's an important decision and I hope it's one that you could identify with. At the end of this first year you'll all be turfed out of your halls and you'll form little huddles of homeless students looking for accommodation. The ants are very good at this. It's what we do them all the time. It's an important decision. You want to live with the right people in the right place. As before we're interested in the role of the individual ant and we're interested in the roll of the entire colony. We study individuals, we paint marks on them, and we study the entire colony. We've chosen a particular species that lives down on the English Channel coast. And we break open rocks like this, very thin fissures in rocks. You've got an entire colony there. They love living by the seaside in rock nests down by the shore. If I was identifying with this, which I strongly do, this would be my perfect home. If I was an ant I'd like to live by the seashore between two enormous slabs of rock. The only drawback to this particular house is that it's on the wrong side of the channel. In the lab we offer them rather minimalistic but very modern accommodation between microscope slides. If I was to give you one of these colonies you could keep the entire colony in your wristwatch. So we can act like Olympian Gods, with an entire colony in the palm of our hands and we can study these societies in enormous detail. Sometimes we're pretty dam beastly to them; we wreck their homes so that have to find a new home. Here we've taken away the top microscope slide. What I'm going to convince you is that they go through all of these stages; It turns out that the ants are incredibly choosy in finding new nest sites. They care about whether the nest site is dark or light; they prefer dark cavities. They want a room with enough headroom in it so that they can clamber over their brood and look after it. They want a nest with a very narrow entrance that is easy to defend and only one entrance. They absolutely care about the floor area because that determines whether the whole colony can get in there. They discriminate against nests that have poor hygiene. If you put a couple of dead bodies in a nest they will reject it. They even care about the proximity of nasty neighbours. So they've got this incredibly long wish list and we've got huge numbers of experiments that show this is really the case. So here's one of the nest we can offer them; we can make it dark or light, we can change the floor area, we can change the amount of headroom, the entrance, and so on. The single most complicated thing they have to do is measure the floor area. I'd like you to identify with this problem; imagine the fire alarm goes off now and I say to 5 of you, please race out of this building and find us another lecture theatre that will be big enough for 170 students. You could count all the seats or you could try and measure the size of the lecture theatre. But imagine you are an ant doing this so you have to do it in a room of completely uncertain shape and you have to do it in complete darkness. How do you assess floor are when you have no overview whatsoever? I think floor area is the most complicated thing they have to solve. The way they do is using an algorithm that we call the Buffon's-Needle algorithm. An ant goes into a nest and it lays its own unique trail for a certain amount of time. It scribbles it all over the inside of the nest. It then goes home, comes back and it counts the number of times it crosses it's previous path. The larger the area, the fewer would those intersections be. It's the most marvellous system because it's completely tolerant of different shapes. We call this Buffon's-needle algorithm after Buffon who was a French aristocrat, who was also a zoologist. You can tell he's a zoologist because he's sitting on a dead lion and squeezing the life out of a pigeon. He was an incredible polymath and one of the things he toyed with was an experimental way of estimating pie. The idea was that you take an area with parallel straight lines drawn upon it. You throw into air a needle and you let it fall and count the number of times it touches one of those lines. There's even an argument that he did it with a baguette on the pavements of Paris. From that you can estimate area. Firstly you can estimate Pie and there's a very simple relationship between the number of intersections and Pie and Buffon came up with this formula. We were able to transform this formula into a method of estimating area. Area is equal to two times the length of the straight lines multiplied by the length of the needle you're dropping, divided by Pie, multiplied by N, the number of intersections. Now, of course the ants don't have calculators, they're not doing the maths, but this is telling us that there is a very simple inverse relationship between area and the number of intersections. As long as you hold that initial length of line fairly constant. If you think about it. If we were to drop a set of cocktail sticks on this bench and they were yellow ones and then we dropped some blue ones, completely random. In a small area they would cross over each other quite frequently. But if you did the first dropping on the Sahara desert, from a biplane and then the on the way back from Timbuktu you dropped the blue ones, hardly any of your lines would cross. So it's perfectly rational that intersection frequency should relate to area. I thought I'd show you this nice little video of an ant laying the initial trails. So, basically an ants going to come in here and we're going to indicate it's path. The BBC coloured this in three different colours. And this has yet to be shown. Basically the ant never leaves the nest and this is all a first visit. So it's laying a pheromone trail from its backside. It'll go into the entrance again and then come back in. And have another go at laying some trail and this time it'll get it all over the show. What it will do is leave, come back and then walk around and estimate the frequency at which it's crossing its previous path. That will give it an indication of nest area. It's a very beautiful system and it's an incredibly robust method that doesn't depend on shape. Here's one that the ant prepared earlier with the green initial trail of one ant and the visiting assessment trail in yellow. So, basically they can estimate area and they choose nests that are neither too small nor too large. I know this is a very complicated slide but I think 300 emigrations went into this. Essentially the ant like dark nests – all of these nests have the same area. We cut that out of the decision-making. We've offered them here a dark nest that was nice and thick, with lots of headroom and a narrow entrance. At the other end of the scale this is a nest that they dislike intensely. It was bright, it was too light, it was too thin and had a gaping wide, draughty, difficult to defend entrance. So we see that the ants like dark nests and they don't like light ones. When you do as many experiments as we've done you can begin to show that the ants are using the most sophisticated consumer strategy known to humanity. That's called a weighted additive strategy. If you were trying to buy a car for me to go collecting more ants we could say, what are the things we care about with a car? Safety, value, speed, elegance. With a weighted additive strategy you'd have a sub-total for each of these things that you care about and you'd score each car out of that sub-total. You'd add up the scores and you'd end up, buying for me please, a nice Porsche. The ants are tiny so we can get lots of them in the glove box, don't worry about that. The great thing is that you can make an assessment that takes everything into account. I would argue that the ants are doing exactly that when they are house hunting. The reason we can argue that is we got this very surprising result. I've actually told you that generally speaking they like dark nests. But 26 out of 29 colonies actually chose this nest type rather than this one. The reason they've chosen this one is that although it's bright, it's got two other characteristics that together outweigh the single most important characteristic. So what this experimental result is telling us is that the ants are taking everything into account. It's a massively important result. They're not just basing their decision on the single most important criteria, they're taking all these other things into account as well. That results in a reversal in the middle of the table. Where you can get a bright nest being preferred over a dark one because these two things together outweigh, and that's the crucial word, they outweigh the single most important criteria. This experiment, together with many others, actually suggests that they are taking everything into account in a weighted fashion and they're doing the weighted additive strategy. The ants are showing very clear preferences, consistent rankings, sensitivity; if the prefer A to B and prefer B to C, they'll always prefer A to C. And they're weighing all the attributes. These tiny ants that are 2mm long, whose brains are a fraction of the size of a pinhead. Are making incredibly well informed consumer decisions. What happens when an ant finds a new nest is it starts taking its friends there by a process called tandem running. At some point in this decision, when there are enough ants in the new nest site, they'll switch from this slow, laborious tandem running to carrying their nest mates there. If I was to pick you up and through you over my shoulder, if I was ant, I could run 3 times more quickly with you over my shoulder than if I was leading you. It's an incredible gear change in the decision and it shows complete commitment to that nest site. They've built up an assessing population and when there are enough of them that have voted with their feet they'll choose that nest site. We can show that experimentally by taking ants away from the nest site or introducing them by hand. They always switch from tandem running to carrying if there is a sufficient number of their nest mates in the new nest site. We call that quorum sensing because it tells us that there are enough ants present to proceed with the business of choosing a nest. A quorum is a term used in committees. If only half the committee shows up, you can say, if you're a pompous chair person 'oh we're not quorate, we can't possibly proceed.' So we've used that term in ant decision-making. One of the classic things in decision-making is when you see a speed accuracy trade off. It's easy to make a quick decision but it's likely to be less accurate than a slow decision. Because being accurate required lots of information and gathering information takes time. We were inspired by this horrible bit of doggerel from Ogden Nash; 'Would you be calm and placid if you were filled with Formic Acid?' These ants aren't filled with formic acid but their enemies are. If we spray an arena with the smell of their enemies they know it's an emergency and they'll get on with things. When we make harsh conditions, as though enemies are there or we have a howling gale, the ants use a much lower quorum threshold. They make many fewer forward tandem runs to they are involving less people in the decision. They are much quicker but they make much more errors. They have a much greater tendency to take some of the colony to the wrong nest. In harsh conditions they'll often make individualistic decisions. I won't wait for your opinion, or yours, I'll simply start carrying my friends there, because this is an emergency. All of the experiments I've talked to you about so far involved destroying their current home. One of the experiments that we've done is to let the nest that they're living in stay intact so they don't have to emigrate at all. But they will emigrate if you offer them an incredibly superior nest. We stumbled upon this experiment because – does anyone remember what a 3.5 inch floppy disk looks like? It's a true story but it's got sell by date because in a couple of years nobody will remember what a 3.5 inch floppy disk looks like. We had one on the lab bench and even though we lovingly look after our ants they all emigrated out of the nest that we gave them into the 3.5 inch floppy disk. That's the kind of space they like. Because I suffer from paranoia I actually thought they were in there rearranging the data. I have no evidence for that. The point is they will move if you allow them a superior nest. If anyone's brain is operating on all 8 cylinders. Can you guess what they might do with the quorum threshold? We know they lower it in harsh conditions, what would they do if they don't need to move at all? Raise it. Absolutely right. Now why should they raise it? Brilliant. That's exactly what they do. If you give them an opportunity to move to improve, they've got plenty of time so they can adjust the quorum threshold upwards. If we put them in a mediocre nest and offer them a superb one they'll use an incredibly high quorum threshold. They'll do an enormous number of tandem runs. They'll only go to that nest if they're absolutely convinced it's palatial. They can really moderate this quorum sensing to different conditions. It's even more gorgeous because by having lots of tandem runs they've informed lots of people where the new nest site is. So when they start emigrating they can get on with that really, really quickly. It's a fabulous example of speed accuracy trade off behaviour. In a lot of our experiments we can run the whole thing in a small petri dish and offer them alternative nests. But we thought we'd have some fun by titrating quality against distance. We offered them a mediocre nest that was really close by and we offered them a better nest that was a long way away. It was 9 times further away. The ants can even solve this. 16 out of 18 colonies were able to select the better nest even though it's much further away. This experiment was done with a couple of undergraduates in my lab. We kept extending the distance of the arena. The students were saying we needed a longer arena. I was thinking 'what can we do?' They looked heavenwards and said 'let's take down all the light fittings'. This is a fluorescent light fitting from my office. We put one after another until we got a hugely long arena, and the ants could still solve it. They're not just good at binary decision-making where you offer them two alternatives; they're great when you offer them one good nest in an array of poor ones. Again we could ask the same question, is this an individual decision or a colony level decision? Despite the fact that these ants are only 2mm long we can glue RFID tags on their backs. That means we can record how the ants vote with their feet. So we tag every individual. The tags are tiny, but the reader is massive. We put it above the nest entrance like this. This is a fabulous film made by Ammonite productions in Bristol. It's an incredible close up, these ants are tiny. Please flash. Every time it flashes it's reading the identity of a particular ant. The tag gets its energy from a laser beam and produces a radio signal that identifies that ant uniquely. What we can also do is what I think if one of our nicest experiments of all time. Which was, in real time, to have computer-controlled doors. What we could do is have a poor quality nest here and a high quality nest over here. Computer controlled door so that if an ant visited a poor nest it was not allowed to go into the best nest and visa versa. So we prevented all comparisons by individuals. If the colony is able to choose it's a colony level decision rather than an individual level decision. The ants were perfectly capable of solving this problem. In a sense it is a corporate level decision, not an individual decision. We can say individuals initially gather the information but it's the colony that chooses. If we return to this definition of cognition, we do have this build up of separate events into a new form of information. The new form of information is the achievement of a quorum threshold, that says yeah, a lot of ants think this nest is good and we'll commit to emigrating to it. Now I'd like to turn to house-hunting in honeybees, which in some ways of remarkably similar, in some ways it's remarkably different. A friend and colleague of mine, Tom Seeley, who works at Cornell University, recently published a book called 'Honey Bee Democracy.' About how they make decisions. I'd like to run through some of Tom's work. This is a swarm of bees and a scout has set out to find a new nest site. Remarkably if we had a swarm of honeybees in this lecture theatre. I should bring one out from under the desk and scare you all to death. Those bees could search an area that's 7km in radius. That means that bees can search an area of 150 kilometres squared. As a search engine they are amazing. This is where we are now, in the middle of Bristol, 150 square kilometres would mean that they could find a nest all the way over in Keynsham. Of course no one would really like to live in Keynsham – once a student said 'I live in Keynsham' they were almost in tears, but tough. So they can search an enormous area – sorry Keynsham, my apologies. The great thing that the bees can do is waggle dancing. The amazing thing is that they're not only searching a huge area but they're searching for a knothole in a tree with a big cavity behind it. So this bee has found a knothole in a tree and it uses a waggle dance to tell the others where it has found a nest site. It dances on the surface of the swarm with a waggle dance. The amount of waggling tells you the distance away and the angle of the waggle from the vertical tells you the angle of the nest away from the sun. So they can tell other bees the distance and direction to find the new site. The accuracy with which they do this is absolutely gobsmacking. For us the interesting issue is how to they agree? They'll dance all over the swarm and information will be coming in for all different areas around the swarm site. It's as if the swarm act like a bulletin board in the local post office where you pin up these advertisements for different houses or flats. Enormous amount of information returned to this information centre. This is a trace of the decision making of one swarm. The arrow indicates where they found the nest, the width of the arrow indicates how many bees were dancing for that particular nest site and the length of the arrow tells you how far away it was. Here they've found a nest over here but not many of them are committed to it. A few hours later they've got four different nests that they're looking at. A little while later they've dropped two of those out and they're looking at these two. They're surveying 6 different nests. The next day they're beginning to build a consensus for a particular site that's down here, south of their current swarm site. That consensus builds and builds until they have achieved almost perfect unanimity and consensus over that. It's like giving a lecture in a TB hospital this, it's terrible. You poor people. Here's a simplified experiment in which we've offered them two nest sites. One we'll represent by an open zero and the other by a vertical bar. All I want you to take out of this is that they found the open nest site first – the worst one. At the end of the process they are all dancing for the bar site which is the better one. The crucial thing is that they way they've been able to achieve a consensus is that they advertise the site that they've found and then they shut up and stop advertising that site. In decision making shutting up it absolutely crucial. I would refer to the honeybee swarm as the idea committee. And I promise you that however long you live, particularly if you occupy academia, you'll never encounter the perfect committee. So all these things I'm going to show you you'll never find in real life in humanity, at least I have never done. In the perfect committee you'll have people speak up – academics are quite good at that - shut up – never – listen without prejudice and achieve a consensus or vote. That's exactly what they bees do. They advertise their site, they then stop banging on about it, they'll monitor the dances of other people that have other information, and eventually they'll achieve a consensus or vote. One of the ways they encode quality is if they find a really good site they'll do more dancing for it. If they find a really ace site they'll dance this much, go back and check it out again. Next time they come back to the swarm they'll dance this much. Back, dance and so on. But if they encounter a mediocre site they'll begin here. The higher the quality of the nest the more they'll advertise it. That's the way they bring quality into this decision-making. Eventually they'll all shut up completely. So, individual dances do not indicate the quality of a site, only its distance and direction. However, good sites are advertised for longer. In a sense the advertisement is posted on the bulletin board for longer. Doing some mathematical modelling with Nick Britain at Bath when I was there, showed that the mathematics of this decision-making is wonderful. It's so wonderful that even if news of a better site comes in really late the committee can change its mind. The bees can change their mind. After that we did some more modelling, this was lead by James Marshall, where we begin to think that the very algorithms, the very logic that the bees use is similar to the logic that we use in decision making. Here's a rather gruesome brain and here's a honeybee colony. The example from vertebrate brains that I'd like to talk to you about briefly is the example of a monkey in a box like this. It's sitting there watching a video screen and it's got to decide whether the majority of the dots are going to the left or to the right. If it decides by moving it's eyes in the right way it gets a little sip of orange juice. What I'd like us to do now, and I think I'd have to dip the lights even more, is I'd like you to engage with your monkey brain. I'm not being insulting about this, it's good that you've got a monkey brain. I've got a monkey brain; we've all got monkey brains. What I'd like you to do is put both your font paws on the desk in front of you and please look at them really hard and try and work out which is the left and the right. Are you up to that. Are you sure you know which is your left and right. I'm going to show you a pattern of moving dots and the moment the dots stop moving I'd like you to vote with your left paw or your right paw. For the direction that the majority of the dots moved in. Is that clear? Thank you Claire for nodding so beautifully. Go on, raise your paws, raise them proudly in the air. So who's voted for the right? Has anyone voted for the left? We tried this with Mark the cameraman earlier and he got it completely wrong. He's embarrassed to bits now and he won't talk to me later. Well done monkey brains. So 50% of the dots were moving in a coherent fashion to the right. Now, I'd like you to repeat the exercise with the next pattern. Put your paws on the front desk again please. If they're really hairy that would be much better because then they're much more monkey like. As soon as we stop I'd like you raise your left or right paw. If you voted right keep your hand in the air. Raise them proudly. And how many people voted the other way. So it's roughly 50 – 50. I haven't got a clue. We've got a 5% coherence but I've forgotten where I downloaded this from so I've no idea. So it was a perfect result. Terribly sorry. Anyway the point that I'm trying to make is that when people have modelled this and looked at neuronal activity. It turns out that you are actually holding a debate in your brain about which way the dots were moving. Some of your neurons were voting for the right and some were voting for the left. Not only were you holding a debate inside your cranium but there was lateral inhibition. The ones that were screaming 'it's right, I want the orange juice' will actually inhibit the other set. And it turns out that that lateral inhibition is crucial in the efficiency of this decision making. We thought, wouldn't it be marvellous if the honeybees were doing something similar. That they were holding a debate with lateral inhibition. It would be terribly disappointing if they didn't do this, but they do. And earlier this year, so this is red hot, brand new science, we showed that they do exactly that. This was lead by Tom Seeley, with some other people including myself and James Marshall. We showed that the logic of the ways bees do it is extremely similar to the way that vertebrate brains do it. I've already introduced Tom Seeley, this is Kirk Bishop, who's wearing a beard of bees. You might ask why would anyone in their right mind wear a beard of bees. I don't have any answer for that. This is Thomas Schlegel who did some of the empirical work I'll show you in a minute. And this is Patrick Hogan who did some of the modelling. James Marshall who also did some of the modelling. That's our crew. I'll show you this video. This bee here is dancing for a site. It's marked yellow. And this one. The yellow bee has been to a yellow site and we've put yellow paint on it. The pink bee is committed to the pink site. What's remarkable is that the yellow bee is doing this perfectly honest waggle dance advertisement. The pink bee goes up to it, head buts it, goes 'bee'. And shuts it up and it probably causes it physical pain. So essentially you've got a real contra-lateral inhibition. Let me show you that again. It's a strange tape because it's from Appledore Island and there are seabirds screeching in the background. If you listen really carefully and watch the pink bee it will headbut the yellow bee, really savagely, in its midriff and go 'bee'. 'Bee, bee.' That's Appledore Island, Thomas Schlegal doing the work with Tom Seeley and Kirk Bishop. So the two hive sites which we'll call yellow and ceres Whenever a bee turns up at the yellow site we put a blob of yellow paint on its back. We call this 'abduction by aliens'. We were able to show that it's the yellow bees that are being beeped by the pink bees. And the pink bees are being beeped by the yellow bees. What we are seeing is this cross-inhibition, of the bees arguing about who should do the signalling. Tom Seeley called this a bee democracy. Let me try and illustrate the logic of this with the House of Commons. Nobody has every used those words in conjunction before. We'd better go back to Victorian times; if the different parties were voting on different options, call them pink and yellow. What they would do is file though the lobbies. If there was a 3 line whip the majority side would be victorious and their policy would be adopted. If they are just filling through it's the absolute difference between the two groups that will count at the end of the day. If the bees were doing that you'd end up with this tiny number of committed bees at the end of it. If they are shouting at each other before they file through the lobbies and changing each other's minds. You can get a much larger difference accruing from this kind of thing. It's as if the two armies are attacking each other and you can end up with this huge difference between the sides. Without cross-inhibition you get just a few bees that remain committed. With this sort of battle before the vote you can get a much larger number of bees committed to the better site. This is an argument for confrontational politics. Instead of immediately voting if you allow the debate to go on. This is a bit like Prime Ministers question time. 'You silly sausage, that's a ridiculous point of view'. 'You insufferable oaf, why don't you shut up.' So they are really trying to change each other's minds in a very gentlemanly fashion. This is what's going on inside your monkey brain when you're making this decision, there are these arguments going on. The joy of this is that it can lead to beautifully clear decision-making. We have all of the other attributes that I've talked to you about before but they've got this procedure that will prevent deadlock. If we offer them two identical site and it might go on for every trying to decide. And if there's just by chance a few more choosing one over the other. Using this differencing procedure they can make a fairly rapid decision. So it prevents deadlock and it's the kind of this that's actually going on inside your own brain when you're decision making. So let's summarise. This was only found out last year and honeybees have been studied for 200 years. The secrets that they're showing us and the exquisite similarities between this and what's going on in our brains is absolutely wonderful. When they've finally decides some of the bees come back to the swarm and they'll do these piping signals that tell the rest of the bees to take off. We'll watch this happen now. This is on the veranda of a log cabin on Shoals Island off the coast of Maine. That's Tom Seeley's hand reaching in. There's a microphone picking up the beeping signals. The swarm gets incredibly excited and in moments they all take off. You'll have the cameraman uttering screams and stumbling down the stairs. The great thing about the bees when they're doing this is that they don't think about anything else. You can have 20,000 bees in mid air and off they'll go to the new nest site. Ouch, ouch, ouch. OK, there we go. They will all go to the best site that's offered to them that they've chosen. This leaves us with a question; at the end of this process we've got a few bees that know where the new nest site is. If you do the sums only about 1% of the colony knows where the new nest site is. So they have to lead the swarm to that site. Could it be that they are like squadron leaders in a group of aircraft? We could think it's a bit like this. A few years earlier we'd actually modelled this and had our paper published on the cover of nature. The lead author was Ian Cousin who was a PhD student of mine. The bees, as with fish, obey interesting rules. If they are too far apart from another bee. If I was a bee I'd come towards you because I'm attracted to you. I wouldn't get too close. If I was reasonably close and not too far away I would orientate in the same direction as you are. They come together, but not too close. If they're in the intermediate space they'll all start orientating in the same direction? Its as if their behaviour glues them together. What we were able to show is that in this system leaders don't need to be identified. It's enough that some of them know the way. Those ones with a sense of purpose kind of suck the rest along in their wake. I know the way, you're all glued to me and you're glued to one another. We were able to show that you can have a very accurate swarm with only 0.05 of the bees, as a proportion, knowing which way to go. You don't need many bees to be knowledgeable and you get a beautifully elongated swarm at this point. 5% would be fantastic. They don't need to identify themselves producing pheromones of anything else. All they need to do is influence the other bees. The problem is with the bees is that it might be only 1%. Even with the best computers at Princeton we're only able to do this with fairly small numbers. As the numbers increase the better the system becomes. So it could be argued that with a huge swarm 1% would be sufficient. What's really nifty is that the bees supplement the fact that they are only 1% by changing their behaviour. What they actually do is that they'll pass through the swarm, sucking the other bees along with them in their wake. To represent 5% rather than 1% they'll zip round the back and fly through it again. Stirring it up in the right way, running to the back and stirring in the right way. Can anyone guess where the bees should fly? They don't want them to be followed on the way back. The bees swam is in the air. If you were trying to sneak back to the back, where would you fly to be beneath the radar? Any ideas? Around the edge? You certainly don't want to do through the middle, but can you think of somewhere where you'd be least visible? Yes Down? Brilliant. He said it. They fly underneath the swarm against the background of the earth. So they are invisible. If they flew through the air they'd be a clear black spot in the air. So they typically fly along the ground, underneath the swarm so they are invisible and then fly through it again. So 1% of the bees can act as if it is 5% and effectively they can suck the whole swam along in their wake. The other thing that's extraordinary about this, and it could be Nigel going completely potty. Is that for me this looks a bit like a waggle dance on a huge scale. Whether the waggle dance evolved from this kind of behaviour or not, I don't think anybody other than me has been daft enough to suggest. A cartoon; sure I follow the herd, not out of brainless obedience but out of a deep and abiding respect for the concept of community. In the next lecture Ganesh will be talking about the mathematics of swarming. Let me summarise. What I've hoped to show you is that the social insects, the ant and the bees in particular, are marvellous material for doing experiments. Don't let anyone else try to convince you of anything other than the fact that what really matters in science is experiments. Mathematical models are important, ideas are important, but the cutting edge, the thing that always decides whether it's right or wrong, are experiments. I know models are fashionable but it's experimentation that really matters. If we do the right experiments what we see in the ants and bees is that they've got very simple rules. That we can explore that underlie the complex problems that they are able to solve. I've spent 30 years at this game, man and boy. What I think is remarkable over my career is we've been able to unearth these rules. But we find that there are so many more than we could ever have imagined. The ants and bees are so much more sophisticated in the number of rules that they have at their disposal and the way those rules interact with one another. You've got relatively simple rules but a surprising large number of such rules in multi-layered systems to solve complex problems. Finally I think there are incredibly deep parallels between ants and bee societies and brains. In a way the social insects are tractable brains in the sense that we can really understand what's going on in them. Here's an idea to leave you with. You've all heard of the idea of convergent evolution. We've got an Ichthyosaur, a reptile that looked like a dolphin that looked like a shark. We refer to this as convergent evolution because the ancestor of the Ichthyosaur was some type of terrestrial reptile. The ancestor of the dolphin was some kind of terrestrial mammal and an ancestor of the shark was some sort of fishy thingy. You've got this wonderful convergent evolution. Different starting points coming up with similar solutions. What we're seeing in our studies of social insects is we're actually going beyond convergent evolution at the morphological level to show convergent evolution at the algorithmic level. The deep logic of how these different systems solve problems is remarkably similar. It applies to social insects and brains but maybe it even applies to our own social constructs like adversarial politics. If you can't bear to stop at this point you could watch some of this on an Horizon programme. If you put into Youtube, 'Horizon out of Control', you'll get a very detailed programme about brains and ants and stuff like that. That's it, we're all done, thank you very much indeed. What is the application for these kinds of experiments. I don't think of myself as an applied biologist. To be absolutely candid by own work is driven by curiosity and I'm completely unapologetic about that. I just think it's fun to find out how the natural world works. If you're desperate to have an application, some of the work that I spoke about earlier, particularly using shortcuts. Actually as fed into a field called ant colony optimisation. When you speak into your mobile phone, the routing algorithms that that message takes. It doesn't all go to the same mast and on, it goes by different routes. The way they calculate which route is best to go by, some of that work has been influenced by work on ant colonies. So there are applications, in distributed robotics and so on. But as I said I'm perfectly happy to do pure curiosity driven research. Where do you get the funding? We struggle for funding because we're not doing applied work. The wonderful thing about my work is that it's relatively inexpensive. I'd rather do low cost work and satisfy my curiosity than chase funding. Going back to the bees. Do you have a timeframe for when they do from speaking up and then shutting up to then beeping? Is that hours or is it weeks? It's hours. Almost all honeybee swarms will be able to find a new nest within about 3 days. They are very time limited so they have to make a decision in a relatively short time. Swarms often happen in spring and they could be decimated by the frost. Actually they can put up with one frost because they all huddle together and they generate so much warmth. Somebody once put a swarm in a polystyrene box in the back of van and it melted its way out of the box. So they are quite good at generating heat, but that takes a lot of energy. They do no foraging when they are a swarm in a tree. So they have an incredible dilemma. They have to find a really good nest site, because they could be there for the next 3 years. And they've got to do it relatively quickly. So they have to have this incredibly powerful search engine. Another application of this might be, thinking about it, search engines. Looking at the ant and when you give them a nest and their nest has not been destroyed, so they were doing more tandem runs. Does it take into account other colonies and will other colonies go and take the nest they've found? Yes, that's a really good point and that's how we know to avoid nasty neighbours. If you offer them two identical nests and there's a hostile colony near one of those nests they'll choose the one away from it. So they are incredibly responsive to their environment. Going back to the ants, you say that no individual ant could hold all the information but there's also 4 or 5 different factors that they have to integrate. Does one any manage to integrate all the factors when it explores the nest? That's a really interesting point. A lot of people say could you have an ant that just does floor are, that specialises in the door. That would actually make it a very fragile system. Because if you lost the door measurer you wouldn't have any information at all on that. I think our data suggests that all of the ants can do all of these things. The reason I focus on the floor area is that's the tricky thing. In terms of headroom they just raise their antennae and touch the roof. If it's the entrance they want one where they can touch both sides. If it's light they can work out how much light's coming into their eyes. Many of the other things are relatively easy. It could be that if the nest stinks of dead bodies they just have a queasy feeling about it and that might dampen their enthusiasm. So I think individuals are actually integrating all of that themselves. You said even if no ant could do to both nests they'd still be able to reach a communal decision. How would they be communicating that? If it's a better nest they'll start recruiting to it earlier. It's a bit like that shortcuts story. The better nest gets a snowballing effect. 'I like it, I'll get somebody else. We both like it, we'll both get somebody.' So, 2 becomes 4, becomes 8, becomes 16. Do they do a deliberation process; 'ok I like this nest but it's not great so I'll wait a couple of minutes'. Basically they just engage in the process of taking other ants their by quorum running. The quorum is sort of the deliberation because they're building up to a certain threshold. The most difficult term that I introduced in that process was deliberation. What it means is that the jury goes out and deliberates over a verdict. It means you're not just having a knee jerk reaction, you're building up a sense of; 'oh yeah, lots of us agree on this, let's go ahead and deliver our verdict'. The logic is very much like a shortcut selection. Back to the bees. You spoke a lot about the scouting and how they deliver the information. How is it that they reach a consensus? There is some evidence that the bees are doing quorum sensing as well, that they're somehow assessing the abundance of their nest mates in a new nest site. If that gets to a certain level they'll some back and start piping. Essentially that process of building a quorum will be interfered with by this cross-inhibition. A really good site will accrue enough honey bees to throw the switch to say; 'ok, enough of us are here, we're all agreed, it's good, let's go back, tell everybody in the swarm, we'll take to the air and we'll lead them to the promised land. On that Biblical note we'll stop. Thank you very much indeed.