Artificial selection is relatively easy to achieve in the computer, and the biomorphs are a good example. It is my dream to simulate natural selection in the computer too. Ideally I'd like to set up the conditions {34} for evolutionary arms races in which ‘predators’ and ‘prey’ emerge on the screen and goad each other into progressive evolution while we sit back and watch. Unfortunately it is very difficult, for the following reason. I said that some offspring are more likely to die, and it might seem easy enough to simulate non-random death. But, in order to be a good simulation of a natural death, the demise of the computer creature must result from some interesting imperfection, like having short legs which make it run more slowly than predators. Computer biomorphs, for instance the insect-like forms in Figure 1.16, sometimes have appendages which we imagine we see as legs. But they don't use these ‘legs’ for anything, and they don't have predators. They don't have prey or food plants. There is no weather in their world and no disease. In theory we can simulate any of these hazards. But to model any one of them in isolation would be scarcely less artificial than artificial selection itself. We'd have to do something like arbitrarily decide that long, thin biomorphs can run away from predators better than short fat ones. It is not difficult to tell the computer to measure the dimensions of biomorphs and choose the lankiest for breeding. But the resulting evolution would not be very interesting. We'd just see biomorphs becoming more and more spindly as the generations go by. It is no more than we could have achieved by artificially selecting the spindliest by eye. It does not have the emergent qualities of natural selection, which a good simulation might achieve. Real-life natural selection is much subtler. It is also in one sense much more complicated though in another sense it is deeply simple. One thing to say is that improvement along any one dimension, like leg length, is only improvement within limits. In real life there is such a thing, for a leg, as being too long. Long legs are more vulnerable to breaking and to getting tangled up in the undergrowth. With a little ingenuity, we could program analogues of both breakages and entanglements into the computer. We could build in some fracture physics: find a way of representing stress lines, tensile strengths, coefficients of elasticity — anything can be simulated if you know how it works. The problem comes with all the things that we don't know about or haven't thought of, and that means almost everything. Not only is the optimal leg length influenced by innumerable effects that we haven't thought of. {35}
Worse, length is only one of countless aspects of an animals legs that interact with each other, and with lots of other things, to influence its survival. There is leg thickness, rigidity, brittleness, weight to carry around, number of leg joints, number of legs, taperingness of legs. And we've only considered legs. All the other bits of the animal interact to influence the animal's probability of surviving.
As long as we try to add up all the contributions to an animals survival theoretically, in a computer, the programmer is going to have to make arbitrary, human decisions. What we ideally should do is simulate a complete physics and a complete ecology, with simulated predators, simulated prey, simulated plants and simulated parasites. All these model creatures must themselves be capable of evolving. The easiest way to avoid having to make artificial decisions might be to burst out of the computer altogether and build our artificial creatures as three-dimensional robots, chasing each other around a three-dimensional real world. But then it might end up cheaper to scrap the computer altogether and look at real animals in the real world, thereby coming back to our starting point! This is less frivolous than it seems. I'll return to it in a later chapter. Meanwhile, there is a little more we can do in a computer, but not with biomorphs.
One of the main things that makes biomorphs so unamenable to natural selection is that they are built of fluorescent pixels on a two-dimensional screen. This two-dimensional world doesn't lend itself to the physics of real life in most respects. Quantities like sharpness of teeth in predators and strength of armour plating in prey; quantities like muscular strength to throw off a predator's attack or virulence of a poison do not emerge naturally in a world of two-dimensional pixels. Can we think of a real-life case of, say, predators and prey, which does lend itself, naturally and without contrived artificiality, to simulation on a two-dimensional screen? Fortunately we can. I've already mentioned spider webs when talking about designoid traps. Spiders have three-dimensional bodies and they live in a complex world of normal physics like most animals. But there is one particular thing about the way some spiders hunt that is peculiarly suited to simulating in two dimensions. A typical orb web is, to all intents and purposes, a two-dimensional structure. The insects that it catches move {36}
Figure 1.17 Safari park of biomorphs bred by ‘Colour Watchmaker’. The large black-and-white triangles in the background were added for purely decorative reasons.
in the third dimension, but at the moment of truth, when an insect is caught or escapes, the action is all in one two-dimension plane, the plane of the web. The spider web is as good a candidate as I can think of for an interesting simulation of natural selection on a two-dimensional computer screen. The next chapter is largely devoted to the story of spider webs, beginning with the natural history of real webs and moving on to computer models of webs and their evolution by ‘natural’ selection in the computer. {37}
<<
* * *
>>
CHAPTER 2
A GOOD WAY TO ORDER OUR UNDERSTANDING OF ANY living creature is to imagine, fancifully and with something more than poetic licence, that it (or, if you prefer, a hypothetical ‘designer’ of the creature) faces a chain of problems or tasks. First we pose the initial problem, then we think of possible solutions that might make sense. Then we look at what the creatures actually do. That often leads us to notice a new problem facing animals of this kind, and the chain continues. I did this in the second chapter of The Blind Watchmaker, with respect to bats and their sophisticated echo-ranging techniques. Here I shall follow the same strategy in this chapter on spider webs. Notice that the progression of problem leading to problem is not to be thought of as marching through one animal's lifetime. If it is a temporal progression at all the time scale is evolutionary, but it may sometimes be not a temporal but a logical progression.
Our fundamental task is to find an efficient method of catching insects for food. One possibility is the flying swift solution. Take to the air like the prey themselves. Fly extremely fast with the mouth open, aiming accurately with keen eyes. This method works for swifts and swallows, but it absorbs costly investment in equipment for high-speed flying and manoeuvring and a high-tech guidance system. The same is true of the bat solution, which is the nocturnal equivalent using sound echoes instead of light rays for guiding the missile. {38}
A completely different possibility is the ‘sit and wait’ solution. Mantises, chameleons and certain other lizards that have evolved independently and convergently to be like chameleons make a go of this solution by being highly camouflaged and by moving in an agonizingly slow and stealthy manner until the final, explosive strike with arms or tongue. The reach of the chameleons tongue enables it to catch a fly anywhere within a radius comparable to its own body length. The reach of the mantiss grappling arms is proportionately of the same order of magnitude. You might think that this design could be improved by lengthening the radius of capture even further. But tongues and arms that were much longer than the body's own length would be prohibitively costly to build and maintain: the extra flies they'd catch wouldn't pay for them. Can we think of a cheaper way to extend the ‘reach’ or radius of capture?
Why not build a net? Nets have to be made of some material and it won't be free. But unlike a chameleons tongue the net material doesn't have to move, so doesn't need bulky muscle tissue. It can be gossamer-thin and can therefore, at low cost, be spun out to cover a much larger area. If you took the meaty protein that would otherwise have been used up in muscular arms or tongue, and reprocessed it as silk, it would
go a very long way, much further than the reach of a chameleon's tongue. There is no reason why the net should not occupy an area 100 times that of the body, yet still be economically made out of secretions from small glands in the body.
Silk is a widespread commodity among arthropods (the major division of the animal kingdom to which both insects and spiders belong). Stick caterpillars belay themselves to a tree with a single thread of the stuff. Weaver ants stitch leaves together using silk extruded by their larvae, held in their jaws as living shuttles (Figure 2.1). Many caterpillars swaddle themselves in a cocoon of silk before growing into a winged adult. Tent caterpillars smother their trees with gossamer. A single domestic silkworm spins nearly a mile of silk when it builds its cocoon. But although silkworms are the basis of our own silk industry, it is really spiders that are the virtuoso silk producers of the animal kingdom, and it is surprising that spider silk is not more used by humanity. It is used for making precision cross-hairs in {39}
Figure 2.1 Workers with silk. Weaver ants using larvae as living shuttles. Oecopbila smaragiina from Australia.
microscopes. In his beautiful book Self-Made Man, the zoologist and artist Jonathan Kingdon speculates that spider silk may have inspired human children to invent one of our most vital pieces of technology, string. Birds, too, recognize the good qualities of spider silk as a material: 165 species (belonging to twenty-three independent families, which suggests that it has been discovered many times independently) are known to incorporate spider silk into the fabric of their nests. A typical orb-weaving spider, the garden cross spider Araneus diadematus produces six different kinds of silk from its rear-end nozzles, made in separate glands in its abdomen, and it switches between the different types for different purposes. Spiders used silk long before they evolved the ability to build orb webs. Even jumping spiders, who never build webs, leap into the air with a silk safety line attached, like mountaineers roped to their most recent secure foothold.
Silk thread, then, is anciently available in the spider tool-kit, and it is eminently suited to the weaving of an insect-catching net. We can {40} think of a net as a means of being in lots of places at once. On its own scale, the spider is like a swallow with a whale's gape. Or like a chameleon with a fifty-foot tongue. A spider web is superbly economical. Whereas a chameleons muscular tongue surely accounts for a substantial fraction of its total body weight, the weight of silk in a spiders web — all twenty metres of it in a big web — is less than a thousandth part of the weight of the spiders body. Moreover, the spider recycles silk after use by eating it, so very little is wasted. But net technology raises problems of its own.
A non-trivial problem for a spider in its web is to make sure that the prey, after hurtling into the web, sticks there. There are two dangers. The insect could easily tear the web and shoot straight through. This problem could be solved by making the silk very elastic, but this aggravates the second of the two dangers: the insect now bounces straight back out of the web as if from a trampoline. The ideal silk, the fibre of a research chemist's dreams, would stretch a very long way to absorb the impact of a fast-flying insect; yet at the same time, to avoid the trampoline effect, would be gently buffered in recoil. At least some kinds of spider silk have just these properties, thanks to the remarkably complicated structure of the silk itself, elucidated by Professor Fritz Vollrath and his colleagues at Oxford, and now at Aarhus, Denmark. The silk shown enlarged in Figures 2.2 and 2.3 is actually much longer than it looks, because most of its length is coiled up inside watery beadlets. It is like a necklace whose beads contain reeled-in surplus thread. The reeling in is done by a mechanism not fully understood, but the result is not in doubt. The web threads are capable of stretching out to ten times their resting length, and they also recoil slowly enough not to bounce the prey out of the web.
The next feature that the silk needs, in order to keep the prey from escaping, is stickiness. The substance that coats the silk in the reeling-in system we've just been talking about is not just watery. It is also sticky. One touch, and it is hard for an insect to escape. But not all spiders achieve stickiness in the same way. A different group called the cribellate spiders produce multi-stranded silk from a special silk gun called a cribellum. The spider then combs out the multi-stranded {41}
Figure 2.2 Beadlets along silk thread of spider web.
Figure 2.3 One beadlet enlarged to show coiled-up thread inside, acting as a ‘windlass’.
Figure 2.4 An alternative way for a web to be sticky: hackled thread from a cribellate spider. {42}
silk by passing it through a custom-built comb mounted on the spider's shin. Multi-stranded silk that is ‘hackled’ in this way puffs out into a tangly thicket (Figure 2.4). The entanglement is too small to see with the naked eye but it is just right for snagging insect legs. Hackled ‘cribellate’ threads behave as if they were sticky, like the gluey threads that we dealt with before. They just achieve their stickiness in a different way. In one respect, cribellate spiders have an advantage. Their threads remain sticky for longer. The non-hackling, glue-using spiders have to rebuild their gluey web anew every morning. Admittedly — and almost incredibly — this can constitute less than an hour's work, but every minute counts when you face natural selection.
But now, sticky threads pose a new and an ironic problem. Whether coated with glue or hackled into a tangle, threads sticky enough to snare an insect are tricky for a spider herself to negotiate. Spiders have no magic immunity, but evolutionary technology has come up with a mixture of partial solutions to the ‘own goal’ hazard. The legs of glue-using spiders are anointed with a special oil which provides some protection from the stickiness. This has been demonstrated by dipping spiders’ legs in ether, which strips off the oily shield and with it the protection. A second partial solution that spiders have adopted is to make some of the threads non-sticky, namely the main spokes that radiate out from the centre of the web. The spider herself runs about on these mam spokes only, using specially modified feet ending in little claws to grip the fine threads. (Male spiders build webs too. For an explanation of my sexist language, see p. 40.) She avoids the sticky spiral that winds round and round on top of the scaffolding made by the spokes. This is easy to do, because she normally sits and waits at the hub of the web, so the shortest distance to any point on the web would be along a spoke anyway.
Let's turn now to the series of problems that face a spider in actually building her web. Not all spiders are the same and, where it matters, I shall take the familiar garden spider Araneus diadematus* as {43} representative. Our — the spider's — initial problem is how to lay the first thread across the gap, say between a tree and a rock, where the web is to be sited. Once the gap has been spanned by that vital first thread, the spider can use it as a bridge. But how to build the first bridge? The pedestrian way would be to walk down, round, and all the way back up again, dragging a line. Spiders sometimes do this, but isn't there a more imaginative solution to the problem? Let's fly a kite. Couldn't we somehow exploit the light and airy properties of silk itself? Yes. Here's how a spider does it if there is enough wind. She releases a single thread with, at its tip, a tmy flattened silken sail or kite. This catches the air and floats. The kite is sticky and, if it happens to land on a firm surface the other side of the gap, it adheres. If the kite does not make a touch the spider hauls it back, recycles the precious silk by eating it, and tries again with a new kite. Eventually a serviceable bridge is thrown across the gap and the spider secures her own end of the thread by sticking it down. The bridge is now ready for crossing.
This first bridge is unlikely to be taut because the length of the thread will be whatever it chances to be: it is not tailor-made for the particular gap. The spider can now either shorten it to serve as one edge of the web; or she might drag it down into a V to form two of the major spokes of the web. The problem here is that, although it could be pulled down into a V, the V is unlikely to be deep enough to make two respectably long spokes. The spider's
own solution to this problem is not to change the bridge itself but to use it as a support while she replaces it by a new and longer thread. Here is how she does {44} it. Standing at one end of the bridge, she initiates a new line from her rear end, and fastens it down securely. Then she severs the existing bridge by biting it through, keeping hold of the cut end in her feet. She walks across, supported by the remains of the cut bridge in front, and by her new line which she pays out behind. She is a living link in her own bridge, moving steadily across its span. As for that part of the original bridge that she has already crossed, it has served its purpose, so she eats it. In this astonishing fashion, eating her old bridge as she goes along and creating a new one behind, she crosses from one side to the other. Moreover, her rear end is paying out silk at a faster rate than her front end is eating it. So the new bridge is, in a carefully controlled fashion, longer than the old one. Now securely fastened at both ends, it sags down the right distance to be pulled into a V and form the hub of the web.
Climbing Mount Improbable Page 4
