I have nothing against the mineral crystals theory, which is why I expounded it before, but what I really want to emphasise is the primacy of replication, and the strong likelihood that there was a late takeover by DNA from some forerunner. I can make the point most forcefully by deliberately switching in this book to a different particular particular theory of what that forerunner might have been. Whatever its ultimate merits as the original replicator, RNA is certainly a better candidate than DNA, and it has been cast as forerunner by a number of theorists in their so-called 'RNA World'. To introduce the RNA World theory, I need to digress on enzymes. If the replicator is the star of life's show, the enzyme is the co-star, more than just supporting cast. theory of what that forerunner might have been. Whatever its ultimate merits as the original replicator, RNA is certainly a better candidate than DNA, and it has been cast as forerunner by a number of theorists in their so-called 'RNA World'. To introduce the RNA World theory, I need to digress on enzymes. If the replicator is the star of life's show, the enzyme is the co-star, more than just supporting cast.

Life depends utterly on the virtuoso ability of enzymes to catalyse biochemical reactions in a very fussy way. When I first learned about enzymes at school, the conventional (and in my view mistaken) wisdom that science should be taught by homely example meant that we spat into water to demonstrate the power of the salivary enzyme amylase to digest starch and make sugar. From this we gained the impression that an enzyme is like a corrosive acid. Biological washing powders, which use enzymes to digest dirt out of clothes, give the same impression. But these are destructive enzymes, working to dismember large molecules into their smaller components. Constructive enzymes are involved in synthesising large molecules from smaller ingredients, and they do so by behaving as 'robotic matchmakers', as I shall explain.

The interior of a cell contains a solution of thousands of molecules, atoms and ions of many different kinds. Pairs of these could combine with each other in almost infinitely varied ways, but on the whole they do not. So there is a huge repertoire of potential chemistry waiting to happen in a cell, but most of it doesn't happen. Hold that in mind while reflecting on the following. A chemistry lab has hundreds of bottles on its shelves, all securely stoppered so their contents don't meet each other unless a chemist desires it, in which case a sample from one bottle is added to a sample from another. You could say that the shelves in a chemistry lab also house a huge repertoire of potential chemistry waiting to happen. And again most of it doesn't happen.

But imagine taking all the bottles off all the shelves and tipping them into a single vat full of water. A preposterous act of scientific vandalism, yet such a vat is pretty much what a living cell is, although admittedly with a lot of membranes that complicate the picture. The hundreds of ingredients of thousands of potential chemical reactions are not kept in separate bottles until required to react together. Instead, they are all mixed up together in the same shared s.p.a.ce, all the time. But still they wait, largely unreactive, until required to react, as though separated in virtual bottles. There are no virtual bottles, but there are enzymes working as robotic matchmakers, or we might even call them robotic lab a.s.sistants. Enzymes discriminate, much as a radio tuner does when it puts a particular wireless set in touch with a particular transmitter while ignoring the hundreds of other signals simultaneously bombarding its aerial with a babel of carrier frequencies.

Suppose there is an important chemical reaction in which ingredient A combines with ingredient B to yield product Z. In a chemistry lab we achieve this by taking the bottle labelled A off the shelf, and the bottle labelled B from another shelf, mixing their contents in a clean flask, and providing other necessary conditions, such as heat or stirring. We achieve the specific reaction we want by taking only two bottles off the shelf. In the living cell lots of A molecules and lots of B molecules are among the huge variety of molecules floating around in the water, where they may meet, but seldom combine even if they do. In any case, they are no more likely to meet than thousands of other possible combinations. Now we introduce an enzyme called abzase, which is specifically shaped to catalyse the A+B=Z reaction. There are millions of abzase molecules in the cell, each one acting as a robotic lab a.s.sistant. Each abzase lab a.s.sistant grabs one A molecule, not off a shelf but floating free in the cell. It then grabs a B molecule as it drifts by. It holds the A firmly in its grip so that it faces in a particular direction. And it holds the B equally firmly so that it abuts the A, in just the right position and orientation to bond with the A and make a Z. The enzyme may do other things too the equivalent of the human lab a.s.sistant wielding a stirrer or lighting a Bunsen burner. It may form a temporary chemical alliance with A or B, exchanging atoms or ions that will eventually be paid back, so the enzyme ends up as it started, thus qualifying as a catalyst. The result of all this is that a new Z molecule forms in the shaped 'grip' of the enzyme molecule. The lab a.s.sistant then releases the new Z into the water and waits for another A to come by, whereupon it grabs it and the cycle resumes.

If there were no robotic lab a.s.sistant, a drifting A would occasionally b.u.mp into a drifting B under the right conditions to bond. But this lucky occurrence would be rare, no more common than the occasional chance-matched encounters that either A or B might make with lots of other potential partners. A might b.u.mp into C and make Y. Or B might b.u.mp into D and make X. Small amounts of Y and X are being made all the time by lucky drift. But it is the presence of the lab a.s.sistant enzyme abzase that makes all the difference. In the presence of abzase, Z is churned out in (from the cell's point of view) industrial quant.i.ties: an enzyme typically multiplies the spontaneous rate of reaction by a factor varying between a million and a trillion. If a different enzyme, acyase, were introduced, A would be combined with C instead of B, again at racing conveyor-belt speed, to make a lavish supply of Y. It is the very same A molecules we are talking about, not confined to a bottle but free to combine with either B or C, depending on which enzyme is present to grab them.

The production rates of Z and Y will therefore depend on, among other things, how many of each of the two rival lab a.s.sistants, abzase and acyase, are floating about in the cell. And that that depends on which of two genes in the nucleus of the cell is turned on. It is, however, a little more complicated than that: even if a molecule of abzase is present, it may be inactivated. One way this can happen is that another molecule comes and sits in the active 'cavity' of the enzyme. It is as though the lab a.s.sistant's robotic arms were temporarily handcuffed. The handcuffs remind me, by the way, to issue the ritual warning that, as always with metaphors, there is a risk that 'robotic lab a.s.sistant' might mislead. An enzyme molecule doesn't actually have arms to reach out and seize ingredients such as A, let alone submit to handcuffs. Instead, it has special zones in its own surface for which A, say, has an affinity, either because of a snug physical fit to a shaped cavity, or due to some more recondite chemical property. And this affinity can be temporarily negated in ways that resemble the calculated throwing of an off-switch. depends on which of two genes in the nucleus of the cell is turned on. It is, however, a little more complicated than that: even if a molecule of abzase is present, it may be inactivated. One way this can happen is that another molecule comes and sits in the active 'cavity' of the enzyme. It is as though the lab a.s.sistant's robotic arms were temporarily handcuffed. The handcuffs remind me, by the way, to issue the ritual warning that, as always with metaphors, there is a risk that 'robotic lab a.s.sistant' might mislead. An enzyme molecule doesn't actually have arms to reach out and seize ingredients such as A, let alone submit to handcuffs. Instead, it has special zones in its own surface for which A, say, has an affinity, either because of a snug physical fit to a shaped cavity, or due to some more recondite chemical property. And this affinity can be temporarily negated in ways that resemble the calculated throwing of an off-switch.

Most enzyme molecules are special-purpose machines which make only one product: a sugar, say, or a fat; a purine or a pyrimidine (building blocks of DNA and RNA), or an amino acid (twenty of them are building blocks of natural proteins). But some enzymes are more like programmable machine tools that take in a punched paper tape to determine what they do. Outstanding among these is the ribosome, briefly explained in Taq's Tale, a large and complicated machine tool constructed from both protein and RNA, which makes proteins themselves. Amino acids, the building blocks of proteins, have already been made by special-purpose enzymes and are floating around in the cell, available to be picked up by the ribosome. The punched paper tape is RNA, specifically 'messenger RNA' (mRNA). The messenger tape, which itself has copied its message from DNA in the genome, feeds into the ribosome and, as it pa.s.ses through the 'reading head', the appropriate amino acids are a.s.sembled into a protein chain in the order specified by the tape using the genetic code.

How this specification works is known, and it is unspeakably wonderful. There is a set of small transfer RNAs (tRNA), each about 70 building blocks long. Each of the tRNAs attaches itself selectively to one, and only one, of the twenty kinds of natural amino acids. At the other end of the tRNA molecule is an 'anti-codon', a triplet precisely complementing the short mRNA sequence (codon) that specifies the particular amino acid according to the genetic code. As the tape of mRNA moves through the reading head of the ribosome, each codon of the mRNA binds to a tRNA with the right anti-codon. This causes the amino acid dangling off the other end of the tRNA to be brought into line, in the 'matchmaking' position, to attach to the growing end of the newly forming protein. Once the amino acid is attached, the tRNA peels off in search of a new amino acid molecule of its preferred type, while the mRNA tape inches forward another notch. So the process continues and the protein chain is extruded step by step. Amazingly, one physical tape of mRNA can cope with several ribosomes at once. Each of these ribosomes moves its reading head along a different portion of the tape's length, and each extrudes its own copy of the newly minted protein chain.

As each new protein chain is completed, when the mRNA feeding its ribosome has completely gone through that ribosome's reading head, the protein detaches itself. It coils up into a complicated three-dimensional structure whose shape is determined, through the laws of chemistry, by the sequence of amino acids in the protein chain. That sequence was itself determined by the order of code symbols along the length of the mRNA. And that order was, in turn, determined by the complementary sequence of symbols along the DNA, which const.i.tutes the master database for the cell.

The coded sequence of DNA therefore controls what goes on in the cell. It specifies the sequence of amino acids in each protein, which determines the protein's three-dimensional shape, which in turn gives that protein its particular enzymatic properties. Importantly, the control may be indirect in that, as we saw in the Mouse's Tale, genes determine which other genes shall be turned on and when. Most genes in any one cell are not switched on. This is why of all the reactions that could be going on in the 'vat full of mixed ingredients', only one or two actually do go on at any one time: the ones whose specific 'lab a.s.sistants' are active in the cell.

After that digression on catalysis and enzymes, we now turn from ordinary catalysis to the special case of autocatalysis, some version of which probably played a key role in the origin of life. Think back to our hypothetical example of molecules A and B combining to make Z under the influence of the enzyme abzase. What if Z itself is its own abzase? I mean, what if the Z molecule happens to have just the right shape and chemical properties to seize one A and one B, bring them together in the correct orientation, and combine them to make a new Z, just like itself? In our previous example we could say that the amount of abzase in the solution would influence the amount of Z produced. But now, if Z actually is one and the same molecule as abzase, we need only a single molecule of Z to seed a chain reaction. The first Z grabs As and Bs and combines them to make more Zs. Then these new Zs grab more As and Bs to make still more Zs and so on. This is autocatalysis. Under the right conditions the population of Z molecules will grow exponentially explosively. This is the kind of thing that sounds promising as an ingredient for the origin of life.

But it is all hypothetical. Julius Rebek and his colleagues at the Scripps Inst.i.tute in California made it real. They explored some fascinating examples of autocatalysis in real chemistry. In one of their examples, Z was amino adenosine triacid ester (AATE), A was amino adenosine and B was pentafluorophenyl ester, and the reaction took place not in water but in chloroform. Needless to say, none of these particular chemical details, and certainly not the long names, need to be remembered. What matters is that the product of the chemical reaction is its own catalyst. The first molecule of AATE is reluctant to form but, once formed, an immediate chain reaction is set in train as more and more AATE synthesises itself by serving as its own catalyst. As if that weren't enough, this brilliant series of experiments went on to demonstrate true heredity in the sense defined here. Rebek and his team found a system in which more than one variant of the autocatalysed substance existed. Each variant catalysed the synthesis of itself, using its preferred variant of one of the ingredients. This raised the possibility of true compet.i.tion in a population of ent.i.ties showing true heredity, and an instructively rudimentary form of Darwinian selection.

Rebek's chemistry is highly artificial. Nevertheless, his story beautifully ill.u.s.trates the principle of autocatalysis, according to which the product of a chemical reaction serves as its own catalyst. It is something like autocatalysis that we need for the origin of life. Could RNA, or something like RNA under the conditions of the early Earth, have autocatalysed its own synthesis Rebek-style, and in water instead of chloroform?

The problem is a formidable one, as explained by the German n.o.bel Prize-winning chemist Manfred Eigen. He pointed out that any self-replication process is subject to degradation by copying error mutation. Imagine a population of replicating ent.i.ties in which there is a high probability of error in every copying event. If a coded message is to hold its own against the ravages of mutation, at least one member of the population in any one generation must be identical to its parent. If there are ten code units ('letters') in an RNA chain, for example, the average error rate per letter must be less than one in ten: we can then expect that at least some members of the offspring generation will have the full complement of ten correct code letters. But if the error rate is greater, there will be a relentless degradation as the generations go by, simply because of mutation alone, no matter how strong the selection pressure. This is called an error catastrophe. Error catastrophes in advanced genomes form the main theme of Mark Ridley's provocative book Mendel's Demon Mendel's Demon,3 but here we are concerned with the error catastrophe that threatened the origin of life itself. but here we are concerned with the error catastrophe that threatened the origin of life itself.

Short chains of RNA and, indeed, DNA can spontaneously self-replicate without an enzyme. But the error rate per letter is far higher than when an enzyme is present. And this means that long before a sufficient length of gene could be built up to make the protein for a working enzyme, the fledgling gene would have been destroyed by mutation. That is the Catch-22 of the origin of life. A gene big enough to specify an enzyme would be too big to replicate accurately without the aid of an enzyme of the very kind that it is trying to specify. So the system apparently cannot get started.

The solution to the Catch-22 that Eigen offers is the theory of the hypercycle. It uses the old principle of divide and rule. The coded information is subdivided into sub-units small enough to lie below the threshold for an error catastrophe. Each sub-unit is a mini replicator in its own right, and it is small enough for at least one copy to survive in each generation. All the sub-units co-operate in some important larger function, large enough to suffer an error catastrophe if catalysed by a single large chemical rather than being subdivided.

As I have so far described the theory, there is a danger that the whole system would be unstable because some sub-units would self-replicate faster than others. This is where the clever part of the theory kicks in. Each sub-unit flourishes in the presence of the others. More specifically, the production of each is catalysed by the presence of another, such that they form a cycle of dependency: a 'hypercycle'. This automatically prevents any one element from racing ahead. It cannot do so because it depends on its predecessor in the hypercycle.

John Maynard Smith pointed out the similarity of a hypercycle to an ecosystem. Fish numbers depend on the population of Daphnia Daphnia (waterfleas) on which they feed. In turn, fish numbers affect the population of fish-eating birds. The birds provide guano, which a.s.sists blooms of algae on which the (waterfleas) on which they feed. In turn, fish numbers affect the population of fish-eating birds. The birds provide guano, which a.s.sists blooms of algae on which the Daphnia Daphnia flourish. The whole cycle of dependency is a hypercycle. Eigen and his colleague Peter Schuster propose some kind of molecular hypercycle as the solution to the Catch-22 riddle of the origin of life. flourish. The whole cycle of dependency is a hypercycle. Eigen and his colleague Peter Schuster propose some kind of molecular hypercycle as the solution to the Catch-22 riddle of the origin of life.

I'm going to leave the hypercycle theory at this point and return to the suggestion, which is fully compatible with it, that RNA, in the early days when life was just beginning and proteins did not yet exist, might have served as its own catalyst. This is the RNA World theory. To see how plausible it is, we need to look at why proteins are good at being enzymes but bad at being replicators; at why DNA is good at replicating but bad at being an enzyme; and finally why RNA might just be good enough at both roles to break out of the Catch-22.

Three-dimensional shape is largely what matters for enzyme activity. Proteins are good at being enzymes because they can a.s.sume almost any shape you want in three dimensions, as an automatic consequence of their amino acid sequence in one dimension. It is the chemical affinities of amino acids for other amino acids in different parts of the chain that determine the particular knot into which the protein chain ties itself. So the three-dimensional shape of a protein molecule is specified by the one-dimensional sequence of amino acids, and that is itself specified by the one-dimensional sequence of code letters in a gene. In principle (practice is a different matter, and formidably difficult) it should be possible to write down a sequence of amino acids that would spontaneously coil itself up into almost any shape you like: not just shapes that make good enzymes, but any arbitrary shape you choose to specify.4 It is this protean talent that qualifies proteins to act as enzymes. There is a protein capable of selecting any one out of the hundreds of potential chemical reactions that could go on in a cell full of jumbled ingredients. It is this protean talent that qualifies proteins to act as enzymes. There is a protein capable of selecting any one out of the hundreds of potential chemical reactions that could go on in a cell full of jumbled ingredients.

Proteins, then, make wonderful enzymes, capable of tying themselves into knots of any desired shape (see plate 48) (see plate 48). But they are lousy replicators. Unlike DNA and RNA, whose component elements have specific pairing rules (the 'WatsonCrick pairing rules' discovered by those two inspired young men), amino acids have no such rules. DNA, by contrast, is a splendid replicator but a lousy candidate for the enzyme role in life. This is because, unlike proteins with their near infinite variety of three-dimensional shapes, DNA has only one shape, the famous double helix itself. The double helix is ideally suited to replication because the two sides of the stairway peel easily away from one another, each being then exposed as a template for new letters to join, following the WatsonCrick pairing rules. It is not much good for anything else.

RNA has some of the virtues of DNA as a replicator and some of the virtues of protein as a versatile shaper of enzymes. The four letters of RNA are sufficiently similar to the four letters of DNA that either set can serve as a template for the other. On the other hand, RNA does not easily form a long double helix, which means that it is somewhat inferior to DNA as a replicator. This is partly because the double helix system lends itself to proof-correction. When the DNA double helix splits and each single helix immediately serves as a template for its complement, errors can instantly be spotted, and corrected. Each daughter chain remains attached to its 'parent', and comparison between the two permits instant error detection. Proofreading based on this principle reduces mutation rates to the order of one in a billion, which is what makes large genomes like ours possible. RNA, lacking this kind of proofreading, has mutation rates that are thousands of times greater than DNA. This means that only simple organisms with small genomes, such as some viruses, can use RNA as their primary replicator.

But the lack of a double helix structure has its upside as well as its downside. Because the RNA chain doesn't spend all its time paired with its complementary chain but breaks away from the complement as soon as it is formed, it is free to tie itself in knots like a protein. Just as the protein does it by virtue of the chemical affinities of amino acids for other amino acids in different parts of the same chain, RNA does it using the ordinary WatsonCrick base-pairing rules, the same ones as are used to make copies of RNA. Putting it another way, lacking a partner chain to pair with in a double helix like DNA, RNA is free to 'pair' with odd bits of itself. RNA finds small stretches of itself with which it can pair, either in a miniature double helix or in some other shape. The pairing rules insist that these stretches have to be going in opposite directions. An RNA chain therefore has a tendency to fall into a series of hairpin bends.

The repertoire of three-dimensional shapes into which an RNA molecule is capable of throwing itself may not be as great as the repertoire of large protein molecules. But it is large enough to encourage the thought that RNA might furnish a versatile armoury of enzymes. And, to be sure, many RNA enzymes, called ribozymes, have been discovered. The conclusion is that RNA has some of the replicator virtues of DNA and some of the enzyme virtues of proteins. Maybe, before the coming of DNA, the arch-replicator, and before the coming of proteins, the arch-catalysts, there was a world in which RNA alone had enough of both virtues to stand in for both experts. Perhaps an RNA fire ignited itself in the original world, and then later started to make proteins that turned around and helped synthesise RNA, and later DNA too, which took over as the dominant replicator. That is the hope of the RNA World theory. It receives indirect support from a lovely series of experiments initiated by Sol Spiegelman of Columbia University, and repeated in various forms by others over the years. Spiegelman's experiments use a protein enzyme, which might be thought to be cheating, but they produce such spectacular results, illuminating such important links in the theory, that you can't help feeling it was worth it anyway.

First, the background. There is a virus called Q. It is an RNA virus, which means that, instead of DNA, its genes are entirely made of RNA. It uses an enzyme to replicate its RNA, called Q replicase. In the wild state, Q is a bacteriophage (phage for short) a parasite of bacteria, specifically of the gut bacterium Escherichia coli Escherichia coli. The bacterial cell 'thinks' the Q RNA is a piece of its own messenger RNA, and its ribosomes process it exactly as though it were, but the proteins that it manufactures are good for the virus instead of for the host bacterium. There are four such proteins: a coat protein to protect the virus; a glue protein to stick it to the bacterial cell; a so-called replication factor, which I'll mention again in a moment; and a bomb protein to destroy the bacterial cell when the virus has finished replicating, thereby releasing some tens of thousands of viruses, each to travel in its little protein coat until it b.u.mps into another bacterial cell and renews the cycle. I said I would return to the replication factor. You might think that this must be the enzyme Q replicase, but actually it is smaller and simpler. All that the little viral gene itself does is make a protein that sews together three other proteins which the bacterium is making anyway for its own (completely different) purposes. When these are st.i.tched together by the virus's own little protein, the composite so formed is the Q replicase.

Spiegelman was able to isolate from this system just two components, Q replicase and Q RNA. He put them together in water with some small-molecule raw materials the building blocks for making RNA and watched what happened. The RNA seized small molecules and built copies of itself using WatsonCrick pairing rules. It managed this feat without any bacterial host, and without the protein coat or any other part of the virus. That in itself was a nice result. Notice that protein synthesis, which is part of the normal action of this RNA in the wild, has been completely taken out of the loop. We have a stripped-down RNA replication system making copies of itself without bothering to make protein.

Then Spiegelman did something wondrous. He set a form of evolution in motion in this wholly artificial test-tube world, with no cells involved at all. Imagine his set-up as a long row of test tubes, each containing Q replicase and raw building blocks but no RNA. He seeded the first tube with a small amount of Q RNA, and it duly replicated lots more copies of itself. He then drew out a small sample of the liquid, and put a drop of it into the second tube. This seed RNA now set about replicating in the second tube, and when this had been going on for a while Spiegelman drew out a drop from the second tube and seeded the third virgin tube. And so on. This is like the spark from our fire seeding a new fire in the dry gra.s.s, and the new fire seeding another, and so on in a chain of seedings. But the result was very different. Whereas fires don't inherit any of their qualities from the seed, Spiegelman's RNA molecules did. And the consequence was ... evolution by natural selection in its most basic and stripped-down form.

Spiegelman sampled the RNA in his tubes as the 'generations' went by and monitored its properties, including its potency in infecting bacteria. What he found was fascinating. The evolving RNA became physically smaller and smaller and, at the same time, less and less infective when bacteria were offered to samples of it. After 74 generations5 the typical RNA molecule in a tube had evolved to a small fraction of the size of its 'wild ancestor'. The wild RNA had been a necklace about 3,600 'beads' long. After 74 generations of natural selection, the average inhabitant of a test tube had reduced itself down to a mere 550: no good at infecting bacteria but brilliant at infecting test tubes. What had happened was clear. Spontaneous mutations in the RNA had occurred all along the line, and the mutants that survived were well fitted to do so in the test-tube world, as opposed to the natural world of bacteria waiting to be parasitised. The main difference was presumably that the RNA in the tube world could dispense with all the coding devoted to making the four proteins needed to make the coat, the bomb and the other requirements for survival of the wild virus as a working parasite of bacteria. What was left was the bare minimum required to replicate in the featherbedded world of test tubes full of Q replicase and raw materials. the typical RNA molecule in a tube had evolved to a small fraction of the size of its 'wild ancestor'. The wild RNA had been a necklace about 3,600 'beads' long. After 74 generations of natural selection, the average inhabitant of a test tube had reduced itself down to a mere 550: no good at infecting bacteria but brilliant at infecting test tubes. What had happened was clear. Spontaneous mutations in the RNA had occurred all along the line, and the mutants that survived were well fitted to do so in the test-tube world, as opposed to the natural world of bacteria waiting to be parasitised. The main difference was presumably that the RNA in the tube world could dispense with all the coding devoted to making the four proteins needed to make the coat, the bomb and the other requirements for survival of the wild virus as a working parasite of bacteria. What was left was the bare minimum required to replicate in the featherbedded world of test tubes full of Q replicase and raw materials.

This bare minimum survivor, less than a tenth the size of its wild ancestor, has become known as Spiegelman's Monster. Being smaller, the streamlined variant reproduces more rapidly than its compet.i.tors, and therefore natural selection gradually increases its representation in the population (and population, by the way, is exactly the right word, even though we are talking about free-floating molecules, not viruses or organisms of any kind).

Amazing to relate, almost the same Spiegelman monster repeatedly evolves when the experiment is run over again. Moreover, Spiegelman and Leslie Orgel, one of the leading figures in research on the origin of life, performed further experiments in which they added a nasty substance, such as ethidium bromide, to the solution. Under these conditions, a different monster evolves, one that is resistant to ethidium bromide. Different chemical obstacle courses foster evolution towards different specialist monsters.

Spiegelman's experiments used natural 'wild type' Q RNA as a starting point. M. Sumper and R. Luce, working in the laboratory of Manfred Eigen, obtained a truly stunning result. Under some conditions, a test tube containing no RNA at all no RNA at all, just the raw materials for making RNA plus the Q replicase enzyme, can spontaneously generate self-replicating RNA which, under the right circ.u.mstances, will evolve to become similar to Spiegelman's Monster. So much, incidentally, for creationist fears (or hopes, we might rather say) that large molecules are too 'improbable' to have evolved. Such is the simple power of c.u.mulative natural selection (so far is natural selection from being a process of blind chance) Spiegelman's Monster takes only a few days to build itself up from scratch.

These experiments are still not direct tests of the RNA World hypothesis of the origin of life. In particular, we still have the 'cheat' of Q replicase being present throughout. The RNA World hypothesis pins its hopes on RNA's own catalytic powers. If RNA can catalyse other reactions, as it is known to do, might it not catalyse its own synthesis? Sumper and Luce's experiment dispensed with RNA but provided the Q replicase. What we need is a new experiment that dispenses with the Q replicase too. Research continues, and I expect exciting results. But now I want to switch to a newly fashionable line of thought, fully compatible with the RNA World, and with many others among the current theories of the origin of life. What is new is the suggested location in which the crucial events first took place. Not 'warm little pond' but 'hot deep rock' an exciting theory which amounts to this: our pilgrims, to complete their journey and locate their Canterbury, are now going to have to bore deep underground, into the primordial rock. The main inspirer of the theory is another maverick, Thomas Gold, originally an astronomer but versatile enough to deserve the now-rare accolade 'general scientist', and distinguished enough to have been elected to both the Royal Society of London and the American National Academy of Sciences.

Gold believes that our emphasis on the sun as energetic prime mover of life may be misplaced. Perhaps we have yet again been misled by what happens to be familiar: yet again a.s.signed to ourselves and our kind of life a centrality in the scheme of things that we do not deserve. There was a time when textbooks a.s.serted that all life depended ultimately on sunlight. Then, in 1977, the startling discovery was made that volcanic vents on the floors of deep oceans support a strange community of creatures, living without benefit of sunlight. Heat from red-hot lava raises the water temperature to more than 100C, still well below boiling point at the colossal pressures of those depths. The surrounding water is very cold, and the temperature gradient drives various kinds of bacterial metabolism. These thermophile bacteria, including sulphur bacteria who make use of hydrogen sulphide streaming from the volcanic vents, const.i.tute the base of elaborate food chains, higher links of which include blood-red tube worms up to three metres long, limpets, mussels, starfish, barnacles, white crabs, prawns, fish and other annelid worms capable of thriving at 80C. There are bacteria, as we have seen, which can take such Hadean temperatures in their stride, but no other animal is known to do so, and these polychaete worms have accordingly been dubbed Pompeii worms. Some of the sulphur bacteria are given house room by animals, for example by mussels, and by the huge tube worms, who take special biochemical steps, using haemoglobin (hence their blood-red colour) to feed sulphide to their own bacteria. These colonies of life, based on bacterial extraction of energy from hot volcanic vents, astonished everybody, first by their very existence, and then by their abundant richness, which contrasted startlingly with the near-desert conditions of the surrounding sea bottom.

Even after this sensational discovery, most biologists continue to believe that life is centred on the sun. The creatures of the deep-sea smoker communities, fascinating though they might be, are a.s.sumed by most of us to be a rare and unrepresentative aberration. Gold believes otherwise. He thinks hot, dark, high-pressure depths are where life fundamentally belongs and where it originated. Not necessarily in the sea, but perhaps in the rocks, deep underground. We who live at the surface, in the light and the cool and the fresh air, we are the anomalous aberrations! He points out that 'hopanoids', organic molecules made in bacterial cell walls, are ubiquitous in rocks, and quotes an authoritative estimate of between 10 trillion and 100 trillion tonnes of hopanoids in the rocks of the world. This comfortably exceeds the trillion tonnes or so of organic carbon in surface-dwelling life.

Gold notes that the rocks are seamed with cracks and fissures, which, though small to our eyes, provide more than a billion trillion cubic centimetres of hot, wet s.p.a.ce suitable for life on the bacterial scale of existence. Heat energy, and the chemicals of the rocks themselves, would be enough to sustain bacteria in huge numbers. Gold notes that many bacteria thrive at temperatures up to 110C, and this would permit them to live down to depths of between 5 and 10 kilometres, a distance that would take them less than a thousand years to travel. It is impossible to verify his estimate, but he thinks the bioma.s.s of bacteria in the hot, deep rocks might exceed the bioma.s.s of the surface sun-based life with which we are familiar.

Turning to the question of the origin of life, Gold and others have pointed out that thermophily love of high temperatures is not a rare oddity among bacteria and archaeans. It is common: so common, and so widely distributed around bacterial family trees, that it might well be the primitive state from which our familiar cool forms of life have evolved. With respect to both chemistry and temperature, the conditions on the surface of the primitive Earth some scientists call it the Hadean Age were more like those in Gold's hot deep rocks than they were like today's surface conditions. A persuasive case can indeed be made that when we dig down into the rocks we are digging backwards in time, and rediscovering something like the conditions of life's scalding Canterbury.

The idea has been further championed recently by the Anglo-Australian physicist Paul Davies, whose book The Fifth Miracle The Fifth Miracle summarises new evidence discovered since Gold's paper of 1992. Various drilling samples have been found to contain hyperthermophile bacteria, alive and reproducing, amid scrupulous precautions to preclude contamination from the surface. Some of these bacteria have been successfully cultured ... in a modified pressure cooker! Davies, like Gold, believes life may have originated deep underground, and that the bacteria which still live there may be relatively unchanged relics of our remote ancestors. This idea is especially appealing for our pilgrimage because it offers us the hope of meeting something like the earliest bacteria, rather than the more familiar bacteria, modified for modern conditions of light, cold and oxygen. Having endured ridicule at first, the hot deep rock theory of the origin of life is now verging on the positively fashionable. Whether it will turn out to be right must await more research, but I confess to hoping that it will. summarises new evidence discovered since Gold's paper of 1992. Various drilling samples have been found to contain hyperthermophile bacteria, alive and reproducing, amid scrupulous precautions to preclude contamination from the surface. Some of these bacteria have been successfully cultured ... in a modified pressure cooker! Davies, like Gold, believes life may have originated deep underground, and that the bacteria which still live there may be relatively unchanged relics of our remote ancestors. This idea is especially appealing for our pilgrimage because it offers us the hope of meeting something like the earliest bacteria, rather than the more familiar bacteria, modified for modern conditions of light, cold and oxygen. Having endured ridicule at first, the hot deep rock theory of the origin of life is now verging on the positively fashionable. Whether it will turn out to be right must await more research, but I confess to hoping that it will.

There are many other theories that I have not gone into. Maybe one day we shall reach some sort of definite consensus on the origin of life. If so, I doubt if it will be supported by direct evidence because I suspect that it has all been obliterated. Rather, it will be accepted because somebody produces a theory so elegant that, as the great American physicist John Archibald Wheeler said in another context: ... we will grasp the central idea of it all as so simple, so beautiful, so compelling that we will say to each other, 'Oh, how could it have been otherwise! How could we all have been so blind for so long!'

If that isn't how we finally realise we know the answer to the riddle of life's origin, I don't think we ever shall know it.

1 This was the point Darwin was making in his 'warm little pond' letter. This was the point Darwin was making in his 'warm little pond' letter.

2 Sir Peter Medawar, no slouch himself, described Haldane as the cleverest man he ever knew. Sir Peter Medawar, no slouch himself, described Haldane as the cleverest man he ever knew.

3 This excellent book has suffered the common fate of being renamed in mid-Atlantic. If you want to find it in the USA, look for This excellent book has suffered the common fate of being renamed in mid-Atlantic. If you want to find it in the USA, look for The Cooperative Gene The Cooperative Gene. Why do publishers do do this? It causes so much confusion. I hasten to say that I have nothing against this? It causes so much confusion. I hasten to say that I have nothing against The Cooperative Gene The Cooperative Gene as a t.i.tle. Genes certainly are co-operative (see as a t.i.tle. Genes certainly are co-operative (see The Selfish Gene The Selfish Gene). Mendel's Demon Mendel's Demon, too, is a fine t.i.tle, though Genetic Meltdown Genetic Meltdown might have suited the book's message even better. Matt Ridley (no relation to Mark, except as established by Y-chromosome a.n.a.lysis) tells me that although the hardback of his might have suited the book's message even better. Matt Ridley (no relation to Mark, except as established by Y-chromosome a.n.a.lysis) tells me that although the hardback of his Nature via Nurture Nature via Nurture is already so named in the USA, the paperback is to be renamed wait for it is already so named in the USA, the paperback is to be renamed wait for it The Agile Gene The Agile Gene.

4 Indeed, there are lots of different amino acid sequences that will yield the same shape, which is one reason to doubt naive calculations of the astronomical 'improbability' of a particular protein chain, obtained by raising 20 to the power of its length. Indeed, there are lots of different amino acid sequences that will yield the same shape, which is one reason to doubt naive calculations of the astronomical 'improbability' of a particular protein chain, obtained by raising 20 to the power of its length.

5 That's tube generations, of course: the number of RNA generations would be more, because RNA molecules are replicating many times within each tube generation. That's tube generations, of course: the number of RNA generations would be more, because RNA molecules are replicating many times within each tube generation.

THE HOST'S RETURN The genial host, having guided Chaucer and the other pilgrims from London to Canterbury and stood impresario to their tales, turned around and led them straight back to London. If I now return to the present, it must be alone, for to presume upon evolution's following the same forward course twice would be to deny the rationale of our backward journey. Evolution was never aimed at any particular endpoint. Our backwards pilgrimage has been a series of swelling mergers, as we were swallowed up in ever more inclusive groupings: the apes, the primates, the mammals, the vertebrates, the deuterostomes, the animals and so on back to the arch ancestor of all life. If we turn around and move forward now, we cannot retrace our steps. That would imply that evolution, were it to be rerun, would follow the same course, putting those same mergers into reverse gear in the form of splits. The stream of life would branch in all the 'right' places. Photosynthesis and an oxygen-based metabolism would be rediscovered, the eukaryotic cell would reconst.i.tute itself, cells would club together in neometazoan bodies. There would be a new split between plants on the one hand, and animals plus fungi on the other; a new split between protostomes and deuterostomes; the backbone would be rediscovered, and so would eyes, ears, limbs, nervous systems ... Eventually a swollen-brained biped would emerge, with skilled hands guided by forward-looking eyes, culminating in the proverbial cricket team to beat the Australians.

My disavowal of aimed evolution underlay my original choice to do history backwards. And yet in my opening lines I confessed to an ear for a rhyme that would lead me into cautious flirtation with recurring patterns, with lawfulness and forward directionality in evolution. So although my return as host will not be a retracing of steps, I shall be publicly wondering whether something a little bit like a retracing might not be appropriate.

Rerunning Evolution The American theoretical biologist Stuart Kauffman put the question well in a 1985 article: One way to underline our current ignorance is to ask, if evolution were to recur from the Precambrian when early eukaryotic cells had already been formed, what organisms in one or two billion years might be like. And, if the experiment were repeated myriads of times, what properties of organisms would arise repeatedly, what properties would be rare, which properties were easy for evolution to happen upon, which were hard? A central failure of our current thinking about evolution is that it has not led us to pose such questions, although the answers might in fact yield deep insight into the expected character of organisms.

I especially like Kauffman's statistical proviso. He envisages not just one thought experiment but a statistical sample of thought experiments in quest of general laws of life, as opposed to local manifestations of particular lifes. The Kauffman question is akin to the science fiction question of what life on other planets might be like except that on other planets the starting and prevailing conditions would be different. On a large planet, gravity would impose a whole new set of selection pressures. Animals the size of spiders could not have spidery limbs (they'd break under the weight) but would need the support of stout, vertical columns, like the tree trunks on which our elephants stand. Conversely, on a smaller planet, animals the size of elephants but of gossamer build could skitter and leap over the surface like jumping spiders. Those expectations about body build will apply to the whole statistical sample of high-gravity worlds and the whole statistical sample of low-gravity worlds.

Gravity is a given condition of a planet, which life cannot influence. So is its distance from its central star. So is its speed of rotation, which determines day length. So is the tilt of its axis, which, on a planet like ours with its near-circular orbit, is the main determinant of seasons. On a planet with a far-from-circular orbit like Pluto, the dramatically changing distance from the central star would be a much more significant determiner of seasonality. The presence, distance, ma.s.s and orbit of a moon or moons exert a subtle but strong influence on life via the tides. All these factors are givens, uninfluenced by life and therefore to be treated as constant in successive reruns of the Kauffman thought experiment.

Earlier generations of scientists would have treated the weather and the chemical composition of the atmosphere as givens too. Now we know that the atmosphere, especially its high oxygen and low carbon content, is conditioned by life. So our thought experiment must allow for the possibility that in successive reruns of evolution the atmosphere might vary under the influence of whatever life forms evolve. Life could thereby influence the weather, and even major climatic episodes, such as ice ages and droughts. My late colleague W. D. Hamilton, who was right too many times to be laughed away, suggested that clouds and rain are themselves adaptations manufactured by micro-organisms for their own dispersal.

So far as we know, the innermost workings of the Earth remain unaffected by the froth of life on its surface. But thought experiments in the rerunning of evolution should acknowledge possible differences in the course of tectonic events, and hence the histories of continental positions. It is an interesting question whether episodes of volcanism and earthquakes, and bombardments from outer s.p.a.ce, should be a.s.sumed to be the same on successive Kauffman reruns. It is probably wise to treat tectonics and celestial collisions as important variables that can be averaged out if we imagine a sufficiently large statistical sample of reruns.

How shall we set about answering the Kauffman question? What would life be like if the 'tape' were rerun a statistical number of times? Immediately we can recognise a whole family of Kauffman questions, of steadily increasing difficulty. Kauffman chose to reset the clock at the moment when the eukaryotic cell was a.s.sembled from its bacterial components. But we could imagine restarting the process two or three eons earlier, with the origin of life itself. Or, at the other extreme, we can restart the clock much later, say, at Concestor 1, our split from the chimpanzees, and ask whether the hominids would, in a statistically significant number of reruns given that life had reached Concestor 1, have evolved bipedality, brain enlargement, language, civilisation and baseball. In between, there is a Kauffman question for the origin of the mammals, for the origin of the vertebrates, and any number of other Kauffman questions.

Short of pure speculation, does the history of life, as it actually happened, provide anything approaching a natural Kauffman experiment to guide us? Yes it does. We met several natural experiments throughout our pilgrimage. By happy accidents of prolonged geographical isolation, Australia, New Zealand, Madagascar, South America, even Africa, furnish us with approximate reruns of major episodes of evolution.

These landma.s.ses were isolated from each other, and from the rest of the world, for significant parts of the period after the dinosaurs disappeared, when the mammal group displayed most of its evolutionary creativity. The isolation was not total, but was sufficient to foster the lemurs in Madagascar, and the ancient and diverse radiation of Afrotheria in Africa. In the case of South America, we have distinguished three separate foundations of mammals, with long periods of isolation in between. Australinea provides the most perfect conditions for this kind of natural experiment its isolation was nearly perfect for much of the period in question, and it began with a very small, possibly single, inoculum of marsupials. New Zealand is an exception, for alone among these revealing natural experiments it found itself without mammals during the period in question.

As I look at these natural experiments, mostly I am impressed by how similarly evolution turns out when it is allowed to run twice. We have seen how similar Thylacinus Thylacinus is to a dog, is to a dog, Notoryctes Notoryctes to a mole, to a mole, Petaurus Petaurus to flying squirrels, to flying squirrels, Thylacosmilus Thylacosmilus to the sabretooths (and to various 'false sabretooths' among the placental carnivores). The differences are instructive too. Kangaroos are hopping antelope-subst.i.tutes. Bipedal hopping, when perfected at the end of a line of evolutionary progression, may be as impressively fast as quadrupedal galloping. But the two gaits are radically different from each other, in ways that have wrought major changes in the whole anatomy. Presumably, at some ancestral parting of the ways, either of the two 'experimental' lineages could have followed the route of perfecting bipedal hopping, and either could have perfected quadrupedal galloping. As it happens possibly for almost accidental reasons originally the kangaroos hopped one way and the antelopes galloped the other. We now marvel at the downstream divergences between the end-products. to the sabretooths (and to various 'false sabretooths' among the placental carnivores). The differences are instructive too. Kangaroos are hopping antelope-subst.i.tutes. Bipedal hopping, when perfected at the end of a line of evolutionary progression, may be as impressively fast as quadrupedal galloping. But the two gaits are radically different from each other, in ways that have wrought major changes in the whole anatomy. Presumably, at some ancestral parting of the ways, either of the two 'experimental' lineages could have followed the route of perfecting bipedal hopping, and either could have perfected quadrupedal galloping. As it happens possibly for almost accidental reasons originally the kangaroos hopped one way and the antelopes galloped the other. We now marvel at the downstream divergences between the end-products.

The mammals underwent their disparate evolutionary radiations at roughly the same time as each other, on different landma.s.ses. The vacuum left by the dinosaurs freed them to do so. But the dinosaurs in their time had similar evolutionary radiations, although with notable omissions for example, I can't get an answer to my question of why there seem to have been no dinosaur 'moles'. And before the dinosaurs there were yet other multiple parallels, notably among the mammal-like reptiles, and these too culminated in similar ranges of types.

When I give public lectures I always try to answer questions at the end. The commonest question by far is, 'What might humans evolve into next?' My interlocutor always seems touchingly to imagine it is a fresh and original question, and my heart sinks every time. For it is a question that any prudent evolutionist will evade. You cannot, in detail, forecast the future evolution of any species, except to say that statistically the great majority of species have gone extinct. But although we cannot forecast the future of any species, say, 20 million years hence, we can forecast the general range of ecological types that will be around. There will be herbivores and carnivores, grazers and browsers, meat eaters, fish eaters and insect eaters. These dietary forecasts themselves presuppose that in 20 million years there will still be foods corresponding to the definitions. Browsers presuppose the continued existence of trees. Insectivores presuppose insects, or anyway small, leggy invertebrates doodoos, to employ that useful technical term from Africa. Within each category, herbivores, carnivores and so on, there will be a range of sizes. There will be runners, fliers, swimmers, climbers and burrowers. The species won't be exactly the same as the ones we see today, or the parallel ones that evolved in Australia or South America, or the dinosaur equivalents, or the mammal-like reptile equivalents. But there will be a similar range of types, making their livings in a similar range of ways.

If, during the next 20 million years, there is a major catastrophe and a ma.s.s extinction comparable to the end of the dinosaurs, we can expect the range of ecotypes to be drawn from new ancestral starting points, and notwithstanding my speculation about rodents at Rendezvous 10 Rendezvous 10 it might be quite hard to guess which of today's animals will provide those starting points. A Victorian cartoon it might be quite hard to guess which of today's animals will provide those starting points. A Victorian cartoon (see plate 49) (see plate 49) shows Professor Ichthyosaurus discoursing upon a human skull from some remote recycling past. If, in the time of the dinosaurs, Professor Ichthyosaurus had mooted their catastrophic end, it would have been quite hard for him to forecast that their place would be taken by the descendants of the mammals, which were then small, insignificant, nocturnal insectivores. shows Professor Ichthyosaurus discoursing upon a human skull from some remote recycling past. If, in the time of the dinosaurs, Professor Ichthyosaurus had mooted their catastrophic end, it would have been quite hard for him to forecast that their place would be taken by the descendants of the mammals, which were then small, insignificant, nocturnal insectivores.

Admittedly, all this concerns quite recent evolution, not so prolonged a rerun as Kauffman imagined. But these recent reruns can surely teach us some lessons about the inherent reproducibility of evolution. If early evolution ran along similar lines to later evolution, those lessons might amount to general principles. My hunch is that the principles we learn from recent evolution since the decease of the dinosaurs probably hold good at least back to the Cambrian, and probably back to the origin of the eukaryotic cell. I have a hunch that the parallelism of mammal radiations in Australia, Madagascar, South America, Africa and Asia may provide a sort of template for answering Kauffman questions for much older starting points, such as the one he chose, the origin of the eukaryotic cell. Earlier than that landmark event, confidence evaporates. My colleague Mark Ridley, in Mendel's Demon Mendel's Demon, suspects that the origin of eukaryotic complexity was a ma.s.sively improbable event, perhaps even more improbable than the origin of life itself. Influenced by Ridley, my bet is that most rerun thought experiments that start with the origin of life will not make it into the eukaryocracy.

[image]

The Fortyfold Path to Enlightenment. Landscape of eye evolution by Michael Land. Landscape of eye evolution by Michael Land.

We don't have to rely on geographical separation as in the Australian natural experiment to study convergence. We can think of the experiment of evolution being rerun, not from the same starting point in different geographical areas, but from different starting points very possibly in the same geographical area: convergence in animals so unrelated to each other that what they tell us has nothing to do with geographical separation. It has been estimated that 'the eye' has evolved independently between 40 and 60 times around the animal kingdom. This inspired my chapter called 'The Fortyfold Path to Enlightenment' in Climbing Mount Improbable Climbing Mount Improbable, so I won't repeat myself here, except to say that Professor Michael Land of Suss.e.x University, our leading expert on the comparative zoology of eyes, recognises nine independent principles of optical mechanism, each of which has evolved more than once. He was kind enough to prepare for that book the landscape reprinted opposite, in which separate peaks represent independent evolutions of eyes.

It seems that life, at least as we know it on this planet, is almost indecently eager to evolve eyes (see plate 50) (see plate 50). We can confidently predict that a statistical sample of Kauffman reruns would culminate in eyes. And not just eyes, but compound eyes like those of an insect, a prawn or a trilobite, and camera eyes like ours or a squid's, with colour vision and with mechanisms for fine-tuning the focus and the aperture. Also very probably parabolic reflector eyes like those of a limpet, and pinhole eyes like those of Nautilus Nautilus, the latter-day ammonite-like mollusc in its floating coiled sh.e.l.l, whom we met at Rendezvous 26 Rendezvous 26. And if there is life on other planets around the universe, it is a good bet that there will also be eyes, based on the same range of optical principles as we know on this planet. There are only so many ways to make an eye, and life as we know it may well have found them all.

We can do the same kind of count for other adaptations. Echo-location the trick of emitting sound pulses and navigating by accurate timing of the echoes has evolved at least four times: in bats, toothed whales, oilbirds and cave swiftlets. Not as many times as the eye, but still often enough to make us think it not too unlikely that, if the conditions are right, it will evolve. Very probably, too, reruns of evolution would rediscover the same specific principles: the same tricks for confronting difficulties. Once again, I shall not repeat my exposition from a previous book,1 but will simply summarise what we might predict for reruns of evolution. Echolocation should repeatedly evolve using very high-pitched cries (for better resolution of detail than low-pitched). The cries in at least some species are likely to be frequency-modulated, sweeping down or up in pitch during the course of each cry (accuracy is improved because early parts of each echo are distinguishable from late parts by their pitch). The computational apparatus used to a.n.a.lyse the echoes might very well make (subconscious) calculations based on Doppler shifts in frequency of echoes, for the Doppler effect is certainly universally present on any planet where there is sound, and bats make sophisticated use of it. but will simply summarise what we might predict for reruns of evolution. Echolocation should repeatedly evolve using very high-pitched cries (for better resolution of detail than low-pitched). The cries in at least some species are likely to be frequency-modulated, sweeping down or up in pitch during the course of each cry (accuracy is improved because early parts of each echo are distinguishable from late parts by their pitch). The computational apparatus used to a.n.a.lyse the echoes might very well make (subconscious) calculations based on Doppler shifts in frequency of echoes, for the Doppler effect is certainly universally present on any planet where there is sound, and bats make sophisticated use of it.

How do we know that something like the eye or echolocation has evolved independently? By looking at the family tree. Relatives of oilbirds and of cave swiftlets don't do echolocation. Oilbirds and cave swiftlets have separately taken up life in caves. We know they have evolved the technology independently of bats and whales, since nothing else in the surrounding family tree does it. Different groups of bats may have evolved echolocation more than once independently. We don't know how many more times echolocation evolved. Some shrews and seals have a rudimentary form of the skill (and some blind human individuals have learned it). Did pterodactyls do it? Since there is a good living to be made flying at night, and since bats weren't around in those days, it is not unlikely. The same goes for ichthyosaurs. They looked very like dolphins, and presumably made their living in a similar way. Since dolphins make heavy use of echolocation, it is reasonable to wonder whether ichthyosaurs did too, in the days before dolphins. There is no direct evidence, and we must remain open-minded. One point against: ichthyosaurs had extraordinarily big eyes it is one of their most conspicuous features which might suggest that they relied upon vision instead of echolocation. Dolphins have relatively small eyes, and one of their their most conspicuous features, the rounded b.u.mp or 'melon' above the beak, acts as an acoustic 'lens', focusing sound into a narrow beam projected in front of the animal like a searchlight. most conspicuous features, the rounded b.u.mp or 'melon' above the beak, acts as an acoustic 'lens', focusing sound into a narrow beam projected in front of the animal like a searchlight.

Like any zoologist, I can search my mental database of the animal kingdom and come up with an estimated answer to questions of the form: 'How many times has X evolved independently?' It would make a good research project, to do the counts more systematically. Presumably some Xs will come up with a 'many times' answer, as with eyes, or 'several times', as with echolocation. Others 'only once' or even 'never', although I have to say it is surprisingly difficult to find examples of these. And the differences could be interesting. I suspect that we'd find certain potential evolutionary pathways which life is 'eager' to go down. Other pathways have more 'resistance'. In Climbing Mount Improbable Climbing Mount Improbable, I developed the a.n.a.logy of a huge museum of all life, both real and conceivable, with corridors going off in many dimensions and representing evolutionary change, again both real and conceivable. Some of these corridors were wide open, almost beckoning. Others were blocked off by barriers that were hard or even impossible to surmount. Evolution repeatedly races down the easy corridors,