She pressed her fingertips together.

'What struck us right away was that the cells don't merely have receptors, they have pairs of receptors. We racked our brains trying to figure out why, and then we cracked it: it's about ensuring the collective stays healthy. We labelled the receptors according to their function. The universal receptor says, I am the yrr. The special receptor says, I am a fully functioning healthy yrr-amoeba with intact DNA, worthy of being part of the collective and ready to take part in the pow-wow.'

'But couldn't you achieve that through a single receptor?' asked Shankar, with a frown.

'No. Probably not,' said Oliviera. 'It's actually an ingenious system. According to our model, a yrr-amoeba is rather like a military camp fenced in by a wall. Any soldier approaching from the outside is identified by a universal marker his uniform. The uniform tells the other soldiers in the camp, I'm one of you. But those of you who've seen your Michael Caine war movies will know that uniforms are sometimes a disguise. Once your camp's been infiltrated by an outsider, your lives are in danger. So if Michael Caine's to be admitted, he has to know the special signal too. He needs the password. How am I doing from a military point of view, Sal?'

Peak gave a nod. 'Absolutely right.'

'Thank goodness for that. So, when the yrr join together, the following occurs: yrr that have already aggregated produce a scent molecule, a pheromone. The pheromone reaches the other cells' universal receptors and initiates the primary connections; I am the yrr. The first part of the identification process has taken place. The second step requires the special receptors to receive the message, I am a healthy yrr. Well, that's all very well, but some yrr-cells aren't fully operational or healthy. In other words, they've got defective DNA. Since our adversary exists in swarms of billions and seems able to evolve continually, it has to weed out any yrr-cells that aren't capable of further development. The trick seems to be that while every amoeba has a universal receptor, only healthy ones capable of development are in possession of the special receptor. Defective cells don't have them. And now comes the really surprising bit, the bit that should make us afraid. Defective yrr don't know the password. They're excluded from the aggregation. But that's not enough. Yrr are amoebas, and like all amoebas they reproduce by cell division. A species that's continually learning and evolving obviously can't allow a second, defective, population to come into being, so it has to act quickly to stop faulty cells reproducing. That's when the pheromone reveals its dual purpose. In the event that a defective yrr is rejected, the pheromone clings to the amoeba's universal receptor and serves as a fast-acting toxin. It induces programmed cell death, a phenomenon otherwise unheard of in single-cell organisms. The faulty yrr-cell dies at once.'

'How can you tell it's dead?' asked Peak.

'Easy. Its metabolism stops. Besides, you can recognise a dead yrr because it stops glowing. For yrr, luminescing is a biochemical necessity. The best-known example of marine bioluminescence is probably the Aequorea, a hydromedusa from the South Seas. It glows when it produces a pheromone. A similar process is going on here. Pheromones are released by the yrr, causing them to glow. The flashes of light are a sign of particularly intense biochemical activity within the aggregated cells. When yrr luminesce, they're communicating and thinking. When they die, the light goes out.'

Oliviera looked at the others. 'So here's why we need to be afraid. The yrr use basic means to run a complex system of selection. If a yrr-amoeba is healthy and has a fully functioning pair of receptors, the pheromone triggers aggregation. But if that amoeba lacks a special receptor, the pheromone takes its deadly toll. The point is, a species that works like this has a very different perspective on death. Death in yrr society is vital. It would never occur to the yrr to spare a defective yrr-cell. To them it would seem absurd – stupid, even. It's imperative for them to destroy the threat to their own evolution. Whenever the collective is threatened, the yrr respond with the logic of death. It's no good pleading for mercy or expecting compassion. The logic of death doesn't make exceptions, and it's not about brutality. Such thoughts are alien to the yrr and, as such, they'll never understand why they should spare us – given that we're a concrete threat.'

'Uh-huh. So their biochemistry imposes a different morality,' said Li.

'Well, I dare say that's very interesting,' interrupted Vanderbilt, 'but what does it matter if they all use Chanel No. 5 or whatever? I mean, what's the point of knowing that? We could all go and aggregate with them. Yeah, that's it, I'll club together with some yrr.'

Crowe gave him a withering look. 'Like they'd let you.'

'Oh, screw you, Crowe.'

'You guys can keep fighting if you like,' said Anawak, 'but Karen and I have an idea about how yrr cognition might work. We've got Sigur, Mick and Sue pulling their hair out over it. Biologically, it's nonsense – but it would answer a whole heap of questions.'

Weaver took over. 'We programmed our virtual amoebas with electronic DNA, and set it up to keep mutating. In other words, the DNA was learning. All of a sudden we found ourselves back where we'd started – with a functioning neural network computer. We'd originally split our electronic brain into its smallest programmable units and tried to put them back together again as a thinking whole. It didn't work, or at least not until the individual cells were capable of learning. But the only way that a biological cell could learn is through mutations in its DNA, and that's unheard of- but it's exactly what we told our virtual cells to do. We used a scent, like Sue described.'

'The thing is,' Anawak continued, 'we didn't just get our fully functioning neural network computer back: we had yrr operating within their natural habitat. Our version of the network came with a few added extras – we allowed the cells to move through three-dimensional space. It replicated deep-sea conditions, with pressure, currents, friction and so on. First, we had to answer the question as to how members of a collective are able to recognise each other. The pheromone is only half of the story. The rest involves limiting the size of the collective. And that's where Sue and Sigur's discovery comes into play. They found that yrr amplicons differ from each other in small, hypervariable sections, so, as we said before, the cells would have to change their DNA after they came into being. Well, we think that's exactly what happens, and that these hypervariable sections serve as a code for them to recognise each other and to know which collective they belong to.'

'Yrr-amoebas with the same coding recognise each other, and small collectives can aggregate with larger ones,' said Li.

'That's right.' Weaver nodded. 'So we coded our virtual cells too. Each cell already had basic information about its habitat, but some cells were given additional information that the others didn't have. As you'd expect, the first cells to aggregate were the ones that shared the same coding. Then we tried a different tack, and attempted to join two collectives with non-identical coding. It worked, and the unthinkable happened: the cells not only succeeded in aggregating; they managed to exchange their individual coding and mutually update each other. They programmed themselves to share the same standard code, thereby attaining a new state of knowledge. The two collectives merged into one, which joined with a third, and that, too, gave rise to something new.'

'Next we wanted to examine their learning strategies,' said Anawak. 'Once again we created two collectives, each with different coding. We gave one information about a specific experience – an enemy attack. It's not especially original, I know, but we decided to use a shark. We programmed it to take a big bite out of the collective, then we showed the collective how to dodge it. The second collective wasn't taught the trick, and it got bitten. Then we aggregated the collectives and sent in the shark – the new conglomerate dodged it. The whole mass of cells had learned what to do. Finally, we divided the collective into smaller groups, and all of them knew how to dodge a shark.'