The Ancestor's Tale Part 4

If only the late Miocene apes were in Africa instead of Asia, we'd have a smooth series of plausible fossils linking the modern African apes all the way back to the early Miocene and the rich proconsulid ape fauna of Africa. When molecular evidence established beyond any doubt our affinities with the African chimpanzees and gorillas, rather than with the Asian orangs, seekers of human ancestors reluctantly turned their backs on Asia. They a.s.sumed, in spite of the plausibility of the Asian apes themselves, that our ancestral line must lie in Africa right through the Miocene and concluded that, for some reason, our African ancestors had not fossilised after the early burgeoning of proconsulid apes in the early Miocene.

That's where things stood until 1998, when an ingenious piece of lateral thinking appeared in a paper called 'Primate evolution in and out of Africa' by Caro-Beth Stewart and Todd R. Disotell. This tale, of back and forth traffic between Africa and Asia, will be told by the orang utan. Its conclusion will be that Concestor 3 probably lived in Asia after all.

But never mind, for the moment, where it lived. What did Concestor 3 look like? It is the common ancestor of the orang utans and all today's African apes, so it might resemble either or both of them (see plate 5) (see plate 5). Which fossils might give us helpful clues? Well, looking at the family tree, the fossils known as Lufengpithecus, Oreopithecus, Sivapithecus, Dryopithecus Lufengpithecus, Oreopithecus, Sivapithecus, Dryopithecus and and Ouranopithecus Ouranopithecus all lived around the right time or slightly later. Our best-guess reconstruction of Concestor 3 might combine elements of all five of these Asian fossil genera but it would help if we could accept Asia as the location of the concestor. Let's listen to the Orang Utan's Tale and see what we think. all lived around the right time or slightly later. Our best-guess reconstruction of Concestor 3 might combine elements of all five of these Asian fossil genera but it would help if we could accept Asia as the location of the concestor. Let's listen to the Orang Utan's Tale and see what we think.

THE ORANG UTAN'S TALE.

Perhaps we have been too ready to a.s.sume that our links with Africa go back a very long way. What if, instead, our ancestral lineage hopped sideways out of Africa around 20 million years ago, flourished in Asia until around 10 million years ago, and then hopped back to Africa?

On this view, all the surviving apes, including the ones that ended up in Africa, are descended from a lineage that migrated out of Africa into Asia. Gibbons and orang utans are descendants of these migrants who stayed in Asia. Later descendants of the migrants returned to Africa, where the earlier Miocene apes had gone extinct. Back in their old ancestral home of Africa, these migrants then gave rise to gorillas, chimpanzees and bon.o.bos, and us.

The known facts about the drifting of the continents and the fluctuations of sea levels are compatible. There were land bridges available across Arabia at the right times. The positive evidence in favour of the theory depends upon 'parsimony': an economy of a.s.sumptions. A good theory is one that needs to postulate little, in order to explain lots. (By this criterion, as I have often remarked elsewhere, Darwin's theory of natural selection may be the best theory of all time.) Here we are talking about minimising our a.s.sumptions about migration events. The theory that our ancestors stayed in Africa all along (no migrations) seemed, on the face of it, more economical with its a.s.sumptions than the theory that our ancestors moved from Africa into Asia (a first migration) and later moved back to Africa (a second migration).

But that parsimony calculation was too narrow. It concentrated on our own lineage and neglected all the other apes, especially the many fossil species. Stewart and Disotell did a recount of the migration events, but they counted those that would be needed to explain the distribution of all the apes including fossils. In order to do this, you first have to construct a family tree on which you mark all the species about which you have sufficient information. The next step is to indicate, for each species on the family tree, whether it lived in Africa or Asia. In the diagram on the opposite page, which is taken from Stewart and Disotell's paper, Asian fossils are highlighted in black, African ones are in white. Not all the known fossils are there, but Stewart and Disotell did include all whose position on the family tree could be clearly worked out. They also drew in the Old World monkeys, who diverged from the apes around 25 million years ago (the most obvious difference between monkeys and apes, as we shall see, is that the monkeys retained their tails). Migration events are indicated by arrows.

Taking into account the fossils, the 'hop to Asia and back again' theory is now more parsimonious than the 'our ancestors were in Africa all along' theory. Leaving out the monkeys which, on both theories, account for two migration events from Africa to Asia, it need postulate only two ape migrations, as follows: [image]

In and out of Africa. Stewart and Disotell's family tree of African and Asian apes. Swollen areas represent dates known from fossils, while the lines linking these to the tree are inferred from parsimony a.n.a.lysis. Arrows represent inferred migration events. Adapted from Stewart and Disotell [ Stewart and Disotell's family tree of African and Asian apes. Swollen areas represent dates known from fossils, while the lines linking these to the tree are inferred from parsimony a.n.a.lysis. Arrows represent inferred migration events. Adapted from Stewart and Disotell [273].

1. A population of apes migrated from Africa to Asia around 20 million years ago and became all the Asian apes including the living gibbons and orang utans.2. A population of apes migrated back from Asia to Africa and became today's African apes including us.

Conversely, the 'our ancestors were in Africa all along' theory demands six migration events to account for ape distributions, all from Africa to Asia, by ancestors of the following: 1. Gibbons, around 18 million years ago2. Oreopithecus Oreopithecus, around 16 million years ago3. Lufengpithecus Lufengpithecus, around 15 million years ago4. Sivapithecus Sivapithecus and orang utans, around 14 million years ago and orang utans, around 14 million years ago5. Dryopithecus Dryopithecus, around 13 million years ago6. Ouranopithecus Ouranopithecus, around 12 million years agoOf course all these migration counts are valid only if Stewart and Disotell have got the family tree right, based on anatomical comparisons. They think, for example, that among the fossil apes, Ouranopithecus Ouranopithecus is the closest cousin to the modern African apes (its branch is the last to come off the family tree in the diagram before the African apes). The next closest cousins, according to their anatomical a.s.sessments, are all Asian ( is the closest cousin to the modern African apes (its branch is the last to come off the family tree in the diagram before the African apes). The next closest cousins, according to their anatomical a.s.sessments, are all Asian (Dryopithecus, Sivapithecus, etc.). If they have got the anatomy all wrong: if, for instance, the African fossil Kenyapithecus Kenyapithecus is actually closest to the modern African apes, then the migration counts would have to be done all over again. is actually closest to the modern African apes, then the migration counts would have to be done all over again.

The family tree was itself constructed on grounds of parsimony. But it is a different kind of parsimony. Instead of trying to minimise the number of geographical migration events we need to postulate, we forget about geography and try to minimise the number of anatomical coincidences (convergent evolution) we need to postulate. Having got our family tree without regard to geography, we then superimpose the geographical information (the black and white coding on the diagram) to count migration events. And we conclude that it is most likely that the 'recent' African apes, that is gorillas, chimpanzees and humans, arrived from Asia.

Now here's an interesting little fact. A leading textbook of human evolution, by Richard G. Klein of Stanford University, gives a fine description of what is known of the anatomy of the main fossils. At one point Klein compares the Asian Ouranopithecus Ouranopithecus and the African and the African Kenyapithecus Kenyapithecus and asks which most resembles our own close cousin (or ancestor) and asks which most resembles our own close cousin (or ancestor) Australopithecus Australopithecus. Klein concludes that Australopithecus Australopithecus resembles resembles Ouranopithecus Ouranopithecus more than it resembles more than it resembles Kenyapithecus Kenyapithecus. He goes on to say that, if only Ouranopithecus Ouranopithecus had lived in Africa, it might even make a plausible human ancestor. 'On combined geographic-morphologic grounds', however, had lived in Africa, it might even make a plausible human ancestor. 'On combined geographic-morphologic grounds', however, Kenyapithecus Kenyapithecus is a better candidate. You see what is going on here? Klein is making the tacit a.s.sumption that African apes are unlikely to be descended from an Asian ancestor, even if the anatomical evidence suggests that they were. Geographical parsimony is being subconsciously allowed to pull rank over anatomical parsimony. Anatomical parsimony suggests that is a better candidate. You see what is going on here? Klein is making the tacit a.s.sumption that African apes are unlikely to be descended from an Asian ancestor, even if the anatomical evidence suggests that they were. Geographical parsimony is being subconsciously allowed to pull rank over anatomical parsimony. Anatomical parsimony suggests that Ouranopithecus Ouranopithecus is a closer cousin to us than is a closer cousin to us than Kenyapithecus Kenyapithecus is. But, without being explicitly so called, geographical parsimony is a.s.sumed to trump anatomical parsimony. Stewart and Disotell argue that, when you take into account the geography of is. But, without being explicitly so called, geographical parsimony is a.s.sumed to trump anatomical parsimony. Stewart and Disotell argue that, when you take into account the geography of all all the fossils, anatomical and geographical parsimony the fossils, anatomical and geographical parsimony agree agree with each other. Geography turns out to agree with Klein's initial anatomical judgement that with each other. Geography turns out to agree with Klein's initial anatomical judgement that Ouranopithecus Ouranopithecus is closer to is closer to Australopithecus Australopithecus than than Kenyapithecus Kenyapithecus is. is.

This argument may not be settled yet. It is a complicated business juggling anatomical and geographical parsimony. Stewart and Disotell's paper has unleashed a flourishing correspondence in the scientific journals, both for and against. As the available evidence stands at present, I think we should on balance prefer the 'hop to Asia and back' theory of ape evolution. Two migration events is more parsimonious than six. And there really do seem to be some telling resemblances between the late Miocene apes in Asia and our own line of African apes such as Australopithecus Australopithecus and chimpanzees. It is only a preference 'on balance', but it leads me to locate and chimpanzees. It is only a preference 'on balance', but it leads me to locate Rendezvous 3 Rendezvous 3 (and (and Rendezvous 4 Rendezvous 4) in Asia rather than Africa.

The moral of the Orang Utan's Tale is twofold. Parsimony is always in the forefront of a scientist's mind when choosing between theories, but it isn't always obvious how to judge it. And possessing a good family tree is often an essential first prerequisite to powerful further reasoning in evolutionary theory. But building a good family tree is a demanding exercise in itself. The ins and outs of it will be the concern of the gibbons, in the tale that they will tell us in melodious chorus after they join our pilgrimage at Rendezvous 4 Rendezvous 4.

Rendezvous 4.

GIBBONS.

Rendezvous 4, where we are joined by the gibbons, occurs around 18 million years ago, probably in Asia, in the warmer and more wooded world of the early Miocene. Depending on which authority you consult, there are up to twelve modern species of gibbons. All live in South East Asia, including Indonesia and Borneo. Some authorities place them all in the genus Hylobates Hylobates. The siamang used to be separated off, and people spoke of 'gibbons and siamangs'. With the realisation that they divide into four groups, not two, this distinction has become obsolete, and I shall call them all gibbons.1 Gibbons are small apes, and perhaps the finest arboreal acrobats that have ever lived. In the Miocene there were lots of small apes. Getting smaller and getting larger are easy changes to achieve in evolution. Just as Gigantopithecus Gigantopithecus and and Gorilla Gorilla got large independently of each other, plenty of apes, in the Miocene golden age of apes, got small. The pliopithecids, for instance, were small apes which flourished in Europe in the early Miocene and probably lived in a similar way to gibbons, without being ancestral to them. I suppose, for example, that they 'brachiated'. got large independently of each other, plenty of apes, in the Miocene golden age of apes, got small. The pliopithecids, for instance, were small apes which flourished in Europe in the early Miocene and probably lived in a similar way to gibbons, without being ancestral to them. I suppose, for example, that they 'brachiated'.

Brachia is the Latin for 'arm'. Brachiation means using your arms rather than your legs to get about, and gibbons are spectacularly good at it. Their big grasping hands and powerful wrists are like upside-down seven-league boots, spring-loaded to slingshot the gibbon from branch to branch and from tree to tree. A gibbon's long arms, perfectly in tune with the physics of pendulums, are capable of hurling it across a sheer ten-metre gap in the canopy. My imagination finds high-speed brachiation more exciting even than flying, and I like to dream of my ancestors enjoying what must surely have been one of the great experiences life could offer. Unfortunately, current thinking doubts that our ancestry ever went through a fully gibbon-like stage, but it is reasonable to conjecture that Concestor 4, approximately our 1-million-greats-grandparent, was a small tree-dwelling ape with at least some proficiency in brachiation. is the Latin for 'arm'. Brachiation means using your arms rather than your legs to get about, and gibbons are spectacularly good at it. Their big grasping hands and powerful wrists are like upside-down seven-league boots, spring-loaded to slingshot the gibbon from branch to branch and from tree to tree. A gibbon's long arms, perfectly in tune with the physics of pendulums, are capable of hurling it across a sheer ten-metre gap in the canopy. My imagination finds high-speed brachiation more exciting even than flying, and I like to dream of my ancestors enjoying what must surely have been one of the great experiences life could offer. Unfortunately, current thinking doubts that our ancestry ever went through a fully gibbon-like stage, but it is reasonable to conjecture that Concestor 4, approximately our 1-million-greats-grandparent, was a small tree-dwelling ape with at least some proficiency in brachiation.

[image]

Gibbons join. The 12 species of gibbon are now generally thought to fall into four groups. The order of branching between these four is controversial, as discussed in detail in the Gibbon's Tale. The 12 species of gibbon are now generally thought to fall into four groups. The order of branching between these four is controversial, as discussed in detail in the Gibbon's Tale.

Images, left to right: hoolock gibbon ( hoolock gibbon (Bunopithecus hoolock); agile gibbon (Hylobates agilis); siamang (Symphalangus syndactylus); golden-cheeked gibbon (Nomascus gabriellae).

Among the apes, gibbons are also second only to humans in the difficult art of walking upright. Using its hands only to steady itself, a gibbon will use bipedal walking to travel along the length of a branch, whereas it uses brachiation to travel across from branch to branch. If Concestor 4 practised the same art and pa.s.sed it on to its gibbon descendants, could some vestige of the skill have persisted in the brain of its human descendants too, waiting to resurface again in Africa? That is no more than a pleasing speculation, but it is true that apes in general have a tendency to walk bipedally from time to time. We can also only speculate on whether Concestor 4 shared the vocal virtuosity of its gibbon descendants, and whether this might have presaged the unique versatility of the human voice, in speech and in music. Then again, gibbons are faithfully monogamous, unlike the great apes which are our closer relatives. Unlike, indeed, the majority of human cultures, in which custom and in several cases religion encourages (or at least allows) polygyny. We do not know whether Concestor 4 resembled its gibbon descendants, or its great ape descendants in this respect.2 Let's summarise what we can guess about Concestor 4, making the usual weak a.s.sumption that it had a good number of the features shared by all its descendants, which means all the apes including us. It was probably more dedicated to life in the trees than Concestor 3, and smaller. If, as I suspect, it hung and swung from its arms, its arms were probably not so extremely specialised for brachiation as those of modern gibbons, and not so long. It probably had a gibbon-like face, with a short snout. It didn't have a tail. Or, to be more precise, its tail vertebrae were, as in all the apes, joined together in a short internal tail, the coccyx (p.r.o.nounced koxix).

I don't know why we apes lost our tail. It is a subject that biologists discuss surprisingly little. A recent exception is Jonathan Kingdon in Lowly Origin Lowly Origin, but even he reaches no satisfactory closure. Zoologists faced with this kind of conundrum often think comparatively. Look around the mammals, note where taillessness (or a very short tail) has independently cropped up, and try to make sense of it. I don't think anyone has done this systematically, and it would be a nice thing to undertake. Apart from apes, tail loss is found in moles, hedgehogs, the tailless tenrec Tenrec ecaudatus Tenrec ecaudatus, guinea pigs, hamsters, bears, bats, koalas, sloths, agoutis and several others. Perhaps most interesting for our purposes, there are tailless monkeys, or monkeys with a tail so short it might as well not be there, as in a Manx cat. Manx cats have a single gene that makes them tailless. It is lethal when h.o.m.ozygous (present twice) so is unlikely to spread in evolution. But it has crossed my mind to wonder whether the first apes were 'Manx monkeys'. If so, the mutation would presumably be in a Hox gene (see the Fruit Fly's Tale). My bias is against such 'hopeful monster' theories of evolution, but could this be an exception? It would be interesting to examine the skeleton of tailless mutants of normally tailed 'Manx' mammals, to see whether they 'do' taillessness in the same kind of way as apes.

The Barbary macaque Macaca sylva.n.u.s Macaca sylva.n.u.s is a tailless monkey and, perhaps in consequence, is often miscalled the Barbary ape. The 'Celebes ape' is a tailless monkey and, perhaps in consequence, is often miscalled the Barbary ape. The 'Celebes ape' Macaca nigra Macaca nigra is another tailless monkey. Jonathan Kingdon tells me it looks and walks just like a miniature chimpanzee. Madagascar has some tailless lemurs, such as the indri, and several extinct species including 'koala lemurs' ( is another tailless monkey. Jonathan Kingdon tells me it looks and walks just like a miniature chimpanzee. Madagascar has some tailless lemurs, such as the indri, and several extinct species including 'koala lemurs' (Megaladapis) and 'sloth lemurs', some of which were gorilla-sized.

Any organ which is not used will, other things being equal, shrink for reasons of economy if nothing else. Tails are used for a surprisingly wide variety of purposes among mammals. Sheep keep a fat reserve in the tail. Beavers use it as a paddle. The spider monkey tail has a h.o.r.n.y gripping pad and is used as a 'fifth limb' in the treetops of South America. The ma.s.sive tail of a kangaroo is spring loaded to a.s.sist bounding. Hoofed animals use the tail as a fly whisk. Wolves and many other mammals use it for signalling, but this is likely to be secondary 'opportunism' on natural selection's part.

But here we must be especially concerned with animals who live up trees. Squirrel tails catch the air, so a 'leap' is almost like flying. Treedwellers often have long tails as counterweights, or as rudders for leaping. Lorises and pottos, whom we shall meet at Rendezvous 8 Rendezvous 8, creep about the trees, slowly stalking their prey, and they have extremely short tails. Their relatives the bushbabies, on the other hand, are energetic leapers, and they have long feathery tails. Tree sloths are tailless, like the marsupial koalas who might be regarded as their Australian equivalents, and both move slowly in the trees like lorises.

In Borneo and Sumatra, the long-tailed macaque lives up trees, while the closely related pig-tailed macaque lives on the ground and has a short tail. Monkeys that are active in trees usually have long tails. They run along the branches on all fours, using the tail for balance. They leap from branch to branch with the body in a horizontal position and the tail held out as a balancing rudder behind. Why, then, do gibbons, who are as active in trees as any monkey, have no tail? Maybe the answer lies in the very different way in which they move. All apes, as we have seen, are occasionally bipedal, and gibbons, when not brachiating, run along branches on their hind legs, using their long arms to steady themselves. It is easy to imagine a tail being a nuisance for a bipedal walker. My colleague Desmond Morris tells me that spider monkeys sometimes walk bipedally, and the long tail is obviously a major enc.u.mbrance. And when a gibbon projects itself to a distant branch it does so from a vertically hanging position, unlike the monkey's horizontal leaping posture. Far from being a steadying rudder streaming out behind, a tail would be a positive drag for a vertical brachiator like a gibbon or, presumably, Concestor 4.

That is the best I can do. I think zoologists need to give more attention to the puzzle of why we apes lost our tail. The a posteriori a posteriori counterfactual engenders pleasing speculations. How would the tail have sat with our habit of wearing clothes, especially trousers? It gives a different urgency to the cla.s.sic tailor's question, 'Does Sir hang to the left or to the right?' counterfactual engenders pleasing speculations. How would the tail have sat with our habit of wearing clothes, especially trousers? It gives a different urgency to the cla.s.sic tailor's question, 'Does Sir hang to the left or to the right?'

THE GIBBON'S TALE.

Written with Yan Wong.

Rendezvous 4 is the first time we greet a pilgrim band of more than a couple of already united species. Any more than that, and there can be problems with deducing relationships. These problems will become worse as our pilgrimage advances. How to solve them is the topic of the Gibbon's Tale. is the first time we greet a pilgrim band of more than a couple of already united species. Any more than that, and there can be problems with deducing relationships. These problems will become worse as our pilgrimage advances. How to solve them is the topic of the Gibbon's Tale.3 [image]

We have seen that there are 12 species of gibbons, falling into four major groups. They are Bunopithecus Bunopithecus (a group consisting of a single species, commonly known as the hoolock), (a group consisting of a single species, commonly known as the hoolock), Hylobates Hylobates (six species, of which the best-known is the white-handed gibbon (six species, of which the best-known is the white-handed gibbon Hylobates lar Hylobates lar), Symphalangus Symphalangus (the siamang), and (the siamang), and Nomascus Nomascus (four species of 'crested' gibbons). This tale explains how to build an evolutionary relationship, or phylogeny, relating the four groups. (four species of 'crested' gibbons). This tale explains how to build an evolutionary relationship, or phylogeny, relating the four groups.

Family trees can be 'rooted' or 'unrooted'. When we draw a rooted tree, we know where the ancestor is. Most of the tree diagrams in this book are rooted. Unrooted trees, by contrast, have no sense of direction. They are often called star diagrams, and there is no arrow of time. They don't start at one side of a page and end on the other. Above are three examples, which exhaust the possibilities for relating four ent.i.ties.

At every fork in a tree, it makes no difference which is the left and which the right branch. And so far (though that will change later in the tale) no information is conveyed by the lengths of the branches. A tree diagram whose branch lengths are meaningless is known as a cladogram (an unrooted cladogram in this case). The order of branching is the only information conveyed by a cladogram: the rest is cosmetic. Try, for example, rotating either of the side forks about the horizontal line in the middle. It will make no difference to the pattern of relationships.

[image]

These three unrooted cladograms represent the only possible ways of connecting four species, as long as we restrict ourselves to connections via branches that only ever split in two (dichotomies). As with rooted trees, it is conventional to discount three-way splits (trichotomies) or more (polytomies) as temporary admissions of ignorance 'unresolved'.

Any unrooted cladogram turns into a rooted one the moment we specify the oldest point (the 'root') of the tree. Certain researchers those we have relied upon for the tree at the start of this tale have suggested the rooted cladogram of gibbons shown above, on the left. However, other researchers have suggested the rooted cladogram on the right.

In the first tree the crested gibbons, Nomascus Nomascus, are distant relatives of all the other gibbons. In the second, it is the hoolock gibbon, Bunopithecus Bunopithecus, who holds this distinction. Despite their differences, both derive from the same unrooted tree (Tree A, on page 129). The cladograms differ only in their rooting. The first is found by dangling the root of Tree A off the branch leading to Nomascus Nomascus, the second by placing the root on the branch leading to Bunopithecus Bunopithecus.

How do we 'root' a tree? The usual method is to extend the tree to include at least one and preferably more than one 'outgroup': a member of a group that is universally agreed beforehand to be only distantly related to all the others. In the gibbon tree, for example, orang utans or gorillas or indeed elephants or kangaroos could do duty as the outgroup. However uncertain we may be about relationships among among gibbons, we know that the common ancestor of gibbons, we know that the common ancestor of any any gibbon with a great ape or an elephant is older than the common ancestor of any gibbon with any other gibbon: it is uncontroversial to place the root of a tree that includes the gibbons and the great apes somewhere between the two. gibbon with a great ape or an elephant is older than the common ancestor of any gibbon with any other gibbon: it is uncontroversial to place the root of a tree that includes the gibbons and the great apes somewhere between the two.

It's easy to verify that the three unrooted trees I have drawn are the only possible dichotomous trees for four groups. For five groups there are 15 possible trees. But don't try to count the number of possible trees for, say, 20 groups. It is up in the hundreds of millions of millions of millions.4 As the count of trees rises steeply with the number of groups to be cla.s.sified, even the fastest computer can take forever. In principle, however, our task is simple. Of all possible trees we must choose that which best explains the similarities and differences between our groups. As the count of trees rises steeply with the number of groups to be cla.s.sified, even the fastest computer can take forever. In principle, however, our task is simple. Of all possible trees we must choose that which best explains the similarities and differences between our groups.

How do we judge 'best explains'? Infinitely rich similarities and differences present themselves when we look at a set of animals. But they are harder to count than you might think. Often one 'feature' is an inextricable part of another. If you count them as separate, you've really counted the same one twice. As an extreme example suppose there are four millipede species, A, B, C, and D. A and B resemble each other in all respects except that A has red legs and B has blue legs. C and D are the same as each other and very different from A and B, except that C has red legs while D has blue legs. If we count leg colour as a single 'feature' we correctly group AB apart from CD. But if we naively count each of 100 legs as separate, their colours will give a hundredfold boost to the number of features supporting the alternative grouping of AC as against BD. Everyone would agree that we have spuriously counted the same feature 100 times. It is 'really' only one feature, because a single embryological 'decision' determined the colour of all 100 legs simultaneously.

The same goes for left-right symmetry: embryology works in such a way that, with few exceptions, each side of an animal is a mirror image of the other. No zoologist would count each mirrored feature twice in making a cladogram, but non-independence isn't always so obvious. A pigeon needs a deep breastbone to attach the flight muscles. A flightless bird like a kiwi does not. Do we count deep breastbone and flapping wings as two separate features by which pigeons differ from kiwis? Or do we count them as only a single feature, on the grounds that the state of one character determines the other, or at least reduces its freedom to vary? In the case of the millipedes and the mirroring, the sensible answer is pretty obvious. In the case of the breastbones it isn't. Reasonable people can be found arguing on opposite sides.

That was all about visible resemblances and differences. But visible features evolve only if they are manifestations of DNA sequences. Nowadays we can compare DNA sequences directly. As an added benefit, being long strings, DNA texts provide a lot more items to count and compare. Problems of the wing-and-breastbone variety are likely to be drowned out in the flood of data. Even better, many DNA differences will be invisible to natural selection and so provide a 'purer' signal of ancestry. As an extreme example, some DNA codes are synonymous: they specify exactly the same amino acid. A mutation that changes a DNA word to one of its synonyms is invisible to natural selection. But to a geneticist, such a mutation is no less visible than any other. The same goes for 'pseudogenes' (usually accidental duplicates of real genes) and for many other 'junk DNA' sequences, which sit in the chromosome but are never read and never used. Freedom from natural selection leaves DNA free to mutate in ways that leave highly informative traces for taxonomists. None of this alters the fact that some mutations do have real and important effects. Even if these are only the tips of icebergs, it is those tips that are visible to natural selection and account for all the visible and familiar beauties and complexities of life.

DNA too is far from immune to the problem of multiple counting the molecular equivalent of the millipedes' legs. Sometimes a sequence is duplicated many times throughout the genome. About half of human DNA consists of multiple copies of meaningless sequences, 'transposable elements', which may be parasites that hijack the machinery of DNA replication to spread themselves about the genome. Just one of these parasitic elements, Alu Alu, is present in over a million copies in most individuals, and we shall meet it again in the Howler Monkey's Tale. Even in the case of meaningful and useful DNA, there are a few cases where genes are present in dozens of identical (or near-identical) copies. But in practice multiple counting tends not to be a problem because duplicate DNA sequences are usually easy to spot.

As a better reason for caution, extensive regions of DNA occasionally show up enigmatic resemblances between comparatively unrelated creatures. n.o.body doubts that birds are more closely related to turtles, lizards, snakes and crocodiles than to mammals (see Rendezvous 16 Rendezvous 16). Nevertheless, the DNA sequences of birds and mammals have resemblances greater than one might expect given their distant relationship. Both have an excess of G-C pairings in their non-coding DNA. The G-C pairing is chemically stronger than the A-T one, and it may be that warm-blooded species (birds and mammals) need more tightly bound DNA. Whatever the reason, we should beware of allowing this G-C bias to persuade us of a close relationship between all warm-blooded animals. DNA seems to promise a Utopia for biological systematists, but we must be aware of such dangers: there is a lot that we still don't understand about genomes.

So, having taken the necessary invocation of caution, how can we use the information present in DNA? Fascinatingly, literary scholars use the same techniques as evolutionary biologists in tracing the ancestries of texts. And almost too good to be true one of the best examples happens to be the work of the Canterbury Tales Canterbury Tales Project. Members of this international syndicate of literary scholars have used the tools of evolutionary biology to trace the history of 85 different ma.n.u.script versions of Project. Members of this international syndicate of literary scholars have used the tools of evolutionary biology to trace the history of 85 different ma.n.u.script versions of The Canterbury Tales The Canterbury Tales. These ancient ma.n.u.scripts, hand-copied before the advent of printing, are our best hope of reconstructing Chaucer's lost original. As with DNA, Chaucer's text has survived through repeated copyings, with accidental changes perpetuated in the copies. By meticulously scoring the acc.u.mulated differences, scholars can reconstruct the history of copying, the evolutionary tree for it really is an evolutionary process, consisting of a gradual acc.u.mulation of errors over successive generations. So similar are the techniques and difficulties in DNA evolution and literary text evolution, that each can be used to ill.u.s.trate the other.

So, let's temporarily turn from our gibbons to Chaucer, and in particular four of the 85 ma.n.u.script versions of The Canterbury Tales: The Canterbury Tales: the 'British Library', 'Christ Church', 'Egerton', and 'Hengwrt' versions. the 'British Library', 'Christ Church', 'Egerton', and 'Hengwrt' versions.5 Here are the first two lines of the General Prologue: Here are the first two lines of the General Prologue: BRITISH LIBRARY:.

Whan that Aprylle / wyth hys showres soote The drowhte of Marche / hath pcede to the rote CHRIST CHURCH:.

Whan that Auerell wt his shoures soote his shoures soote The droght of Marche hath pced to the roote EGERTON:.

Whan that Aprille with his showres soote The drowte of marche hath pced to the roote HENGWRT:.

Whan that Aueryll wt his shoures soote his shoures soote The droghte of March / hath pced to the roote

The first thing that we must do with either DNA or literary texts is to locate the similarities and differences. For this we have to 'align' them not always an easy task, for texts can be fragmentary or jumbled and of unequal length. A computer is a great help when the going gets tough, but we don't need it to align the first two lines of Chaucer's General Prologue, which I have highlighted at the fourteen points where the scripts disagree (see opposite page).

Two places, the second and the fifth, have three variants rather than two. That makes a total of sixteen 'differences'. Having compiled a list of differences we now work out which tree best explains them. There are many ways of doing this, and all can be used for animals as well as for literary texts. The simplest is to group the texts on the basis of overall similarity. This usually relies upon some variant on the following method. First we locate the pair of texts that are the most similar. We then treat this pair as a single averaged text, and put it alongside the remaining texts while we look for the next most similar pair. And so on, forming successive, nested groups until a tree of relationships is built up. These sorts of techniques one of the most common is known as 'neighbour-joining' are quick to calculate, but do not incorporate the logic of the evolutionary process. They are purely measures of similarity. For this reason, the 'cladist' school of taxonomy, which is deeply evolutionary in its rationale (although not all its members realise it) prefers other methods, of which the earliest to be devised was the parsimony method.

Parsimony, as we saw in the Orang Utan's Tale, here means economy of explanation. In evolution, whether of animals or ma.n.u.scripts, the most parsimonious explanation is the one that postulates the least quant.i.ty of evolutionary change. If two texts share a common feature, the parsimonious explanation is that they have jointly inherited it from a shared ancestor rather than that each evolved it independently. It is very far from an invariable rule, but it is at least more likely to be true than the opposite. The method of parsimony at least in principle looks over all possible trees and chooses the one that minimises the quant.i.ty of change.

[image]

When we are choosing trees for their parsimony, certain types of difference can't help us. Differences that are unique to a single ma.n.u.script, or a single species of animal, are uninformative uninformative. The neighbour-joining method uses them, but the method of parsimony ignores them completely. Parsimony relies upon informative informative changes: ones that are shared by more than one ma.n.u.script. The preferred tree is the one that uses shared ancestry to explain as many informative differences as possible. In our Chaucerian lines there are five informative differences to account for. Four split the ma.n.u.scripts into changes: ones that are shared by more than one ma.n.u.script. The preferred tree is the one that uses shared ancestry to explain as many informative differences as possible. In our Chaucerian lines there are five informative differences to account for. Four split the ma.n.u.scripts into {British Library plus Egerton} versus versus {Christ Church plus Hengwrt}. {Christ Church plus Hengwrt}.

These are the differences highlighted by the first, third, seventh, and eighth vertical black lines. The fifth, the virgule (diagonal stroke) highlighted by the twelfth grey line, splits the ma.n.u.scripts differently, into {British Library plus Hengwrt} versus versus {Christ Church plus Egerton}. {Christ Church plus Egerton}.These splits conflict with each other. We can draw no tree in which each change happens just once. The best we can do is the following (note that it is an unrooted tree) which minimises the conflict, requiring only the virgule to appear or disappear twice.

[image]

Actually, in this case I haven't much confidence in our guess. Convergences or reversions are common in texts, especially when the meaning of the verse is not changed. A medieval scribe might have little compunction in changing a spelling, and even less in inserting or removing a punctuation mark such as a virgule. Better indicators of relationship would be changes such as the reordering of words. The genetic equivalents are 'rare genomic changes': events such as large insertions, deletions, or duplications of DNA. We can explicitly acknowledge these by giving more or less weight to different types of change. Changes known to be common or unreliable are down-weighted when counting up extra changes. Changes known to be rare, or reliable indicators of kinship, are given increased weighting. Heavy weighting to a change means we especially don't want to count it twice. The most parsimonious tree, then, is the one with the lowest overall weight.

The parsimony method is much used to find evolutionary trees. But if convergences or reversions are common as with many DNA sequences and also in our Chaucerian texts parsimony can be misleading. It is the notorious bugbear known as 'long branch attraction'. Here's what this means.

Cladograms, whether rooted or unrooted, convey only the order of branching. Phylograms Phylograms, or phylogenetic trees (Greek phylon phylon = race/tribe/cla.s.s), are similar but also use the length of branches to convey information. Typically branch lengths represent evolutionary distance: long branches represent a lot of change, short ones little change. The first line of = race/tribe/cla.s.s), are similar but also use the length of branches to convey information. Typically branch lengths represent evolutionary distance: long branches represent a lot of change, short ones little change. The first line of The Canterbury Tales The Canterbury Tales yields the following phylogram: yields the following phylogram: [image]

In this phylogram, the branches are not too different in length. But imagine what would happen if two of the ma.n.u.scripts changed a lot, compared to the other two. The branches leading to these two would be drawn very long. And a proportion of the changes would not be unique. They would just happen to be identical to changes elsewhere on the tree, but (and now here is the point) especially especially to those on the other long branch. This is because long branches are where the most changes are anyway. With enough evolutionary changes, the ones that spuriously link the two long branches will drown out the true signal. Based upon a simple count of the number of changes, parsimony erroneously groups together the termini of especially long branches. The method of parsimony makes long branches spuriously 'attract' one another. to those on the other long branch. This is because long branches are where the most changes are anyway. With enough evolutionary changes, the ones that spuriously link the two long branches will drown out the true signal. Based upon a simple count of the number of changes, parsimony erroneously groups together the termini of especially long branches. The method of parsimony makes long branches spuriously 'attract' one another.

The problem of long branch attraction is an important headache for biological taxonomists. It rears its head whenever convergences and reversions are common, and unfortunately we cannot hope to avoid it by looking at more text. On the contrary, the more text we look at, the more erroneous similarities we find, and the stronger our conviction in the wrong answer. Such trees are said to lie in the dangerous-sounding 'Felsenstein zone', named after the distinguished American biologist Joe Felsenstein. Unfortunately, DNA data are particularly vulnerable to long branch attraction. The main reason is that there are only four letters in the DNA code. If the majority of differences are single letter changes, independent mutation to the same letter by accident is extremely likely. This sets up a minefield of long branch attraction. Clearly we need an alternative to parsimony in these cases. It comes in the form of a technique known as likelihood a.n.a.lysis, which is increasingly favoured in biological taxonomy.

Likelihood a.n.a.lysis burns even more computer power than parsimony, because now the lengths of the branches matter. So we have vastly more trees to contend with because, in addition to looking at all possible branching patterns, we must also look at all possible branch lengths a Herculean task. This means that, despite clever shortcuts, today's computers can only cope with likelihood a.n.a.lysis involving small numbers of species.

'Likelihood' is not a vague term. On the contrary, it has a precise meaning. For a tree of a particular shape (remembering to include branch lengths), of all the possible evolutionary paths that could produce a phylogenetic tree of the same shape, only a tiny number would generate precisely those texts that we now see. The 'likelihood' of a given tree is the vanishingly small probability of ending up with the actual existing texts, rather than any of the other texts that could possibly have been generated by such a tree. Although the likelihood value for a tree is tiny, we can still compare one tiny value with another as a means of judgement.

[image]

'By me was nothyng added ne mynusshyd' (Caxton's Preface). Unrooted phylogenetic tree of the first 250 lines of 24 different ma.n.u.script versions of Unrooted phylogenetic tree of the first 250 lines of 24 different ma.n.u.script versions of The Canterbury Tales The Canterbury Tales. This represents a subset of the ma.n.u.scripts studied by the Canterbury Tales Canterbury Tales Project, whose abbreviations for the ma.n.u.scripts are used here. The tree was constructed by parsimony a.n.a.lysis, and bootstrap values are shown on the branches. The four versions discussed are named in full. Project, whose abbreviations for the ma.n.u.scripts are used here. The tree was constructed by parsimony a.n.a.lysis, and bootstrap values are shown on the branches. The four versions discussed are named in full.

Within likelihood a.n.a.lysis, there are various alternative methods of obtaining the 'best' tree. The simplest is to search for the single one that has the highest likelihood: the tree which is the most likely. Not unreasonably, this goes under the name 'maximum likelihood', but just because it is the single most likely tree doesn't mean that other possible trees aren't almost as likely. More recently it has been suggested that instead of believing in a single most likely tree, we should look at all possible trees, but give proportionally more credence to the more likely ones. This approach, an alternative to maximum likelihood, is known as Bayesian phylogenetics. If many likely trees agree on a particular branch point, then we calculate that it has a high probability of being correct. Of course, just as in maximum likelihood, we can't look at all possible trees, but there are computational shortcuts and they work pretty well.

Our confidence confidence in the tree we finally choose will depend on our certainty that its various branches are correct, and it is common to place measures of this beside each branch point. Probabilities are automatically calculated when using the Bayesian method, but for others such as parsimony or maximum likelihood, we need alternative measures. A commonly used one is the 'bootstrap' method, which resamples different parts of the data repeatedly to see how much difference it makes to the final tree how robust the tree is, in other words, to error. The higher the 'bootstrap' value, the more trustworthy the branch point, but even experts struggle to interpret exactly what a particular bootstrap value tells us. Similar methods are the 'jackknife', and the 'decay index'. All are measures of how much we should believe each branch point on the tree. in the tree we finally choose will depend on our certainty that its various branches are correct, and it is common to place measures of this beside each branch point. Probabilities are automatically calculated when using the Bayesian method, but for others such as parsimony or maximum likelihood, we need alternative measures. A commonly used one is the 'bootstrap' method, which resamples different parts of the data repeatedly to see how much difference it makes to the final tree how robust the tree is, in other words, to error. The higher the 'bootstrap' value, the more trustworthy the branch point, but even experts struggle to interpret exactly what a particular bootstrap value tells us. Similar methods are the 'jackknife', and the 'decay index'. All are measures of how much we should believe each branch point on the tree.

Before we leave literature and return to biology, on the opposite page is a summary diagram of the evolutionary relationships between the first 250 lines of 24 Chaucer ma.n.u.scripts. It is a phylogram, in which not just the branching pattern but the lengths of the lines are meaningful. You can immediately read off which ma.n.u.scripts are minor variants of each other, which are aberrant outliers. It is unrooted it doesn't commit itself as to which of the 24 ma.n.u.scripts is closest to the 'original'.

It's time to return to our gibbons. Over the years, many people have tried to work out gibbon relationships. Parsimony suggested four groups of gibbons. On the next page is a rooted cladogram based on physical characteristics.

This cladogram shows convincingly that the Hylobates Hylobates species group together, as do the species group together, as do the Nomascus Nomascus ones. Both groupings have reasonably high bootstrap values (the numbers on the lines). But in several places the order of branching is unresolved. Even though it looks as though ones. Both groupings have reasonably high bootstrap values (the numbers on the lines). But in several places the order of branching is unresolved. Even though it looks as though Hylobates Hylobates and and Bunopithecus Bunopithecus form a group, the bootstrap value, 63, is unconvincing to those trained to read such runes. Morphological features do not suffice to resolve the tree. form a group, the bootstrap value, 63, is unconvincing to those trained to read such runes. Morphological features do not suffice to resolve the tree.

For this reason, Christian Roos and Thomas Geissmann of Germany turned to molecular genetics, specifically to a section of mitochondrial DNA called the 'control region'. Using DNA from six gibbons, they deciphered the sequences, lined them up letter-for-letter, and carried out neighbour-joining, parsimony, and maximum likelihood a.n.a.lyses on them. Maximum likelihood, which is the best of the three methods at coping with long branch attraction, gave the most convincing result. Their final verdict on the gibbons is shown on page 141, and you can see that it resolves the relationship between the four groups. The bootstrap values were enough to convince me that this was the tree to use for the phylogeny at the start of this chapter.

[image]

Rooted cladogram of gibbons, based on morphology. Adapted from Geissmann [ Adapted from Geissmann [100].

Gibbons 'speciated' branched into their separate species relatively recently. But as we look at more and more distantly related species, separated by longer and longer branches, even the sophisticated techniques of maximum likelihood and Bayesian a.n.a.lysis start to fail us. There can come a point where an unacceptably large proportion of similarities are coincidental. The DNA differences are then said to be saturated. No fancy techniques can recover the signal of ancestry, because any vestiges of relationship have been overwritten by the ravages of time. The problem is especially acute with neutral DNA differences. Strong natural selection keeps genes on the straight and narrow. In extreme cases, important functional genes can stay literally identical over hundreds of millions of years. But, for a pseudogene that never does anything, such lengths of time are enough to lead to hopeless saturation. In such cases, we need different data. The most promising idea is to use the rare genomic changes that I mentioned before changes that involve DNA reorganisation rather than single letter changes. These being rare, indeed usually unique, coincidental resemblance is much less of a problem. And once found, they can reveal remarkable relationships, as we shall learn when our swelling pilgrim band is joined by the hippo, and we are bowled over by its whale of a surprising tale.

And now, an important afterthought on evolutionary trees, drawing in lessons from Eve's Tale and the Neanderthal's Tale. We might call it the gibbon's decline and fall of the species tree. We normally a.s.sume that we can draw a single evolutionary tree for a set of species. But Eve's Tale told us that different parts of DNA (and thus different parts of an organism) can have different trees. I think this poses an inherent problem with the very idea of species trees. Species are composites of DNA from many different sources. As we saw in Eve's Tale and reiterated in the Neanderthal's Tale, each gene, in fact each DNA letter, takes its own path through history. Each piece of DNA, and each aspect of an organism, can have a different evolutionary tree.

[image]

Cladogram of gibbons, based on maximum likelihood a.n.a.lysis of DNA. Adapted from Roos and Geissmann [ Adapted from Roos and Geissmann [246].

An example of this comes up every day, but familiarity leads us to overlook its message. A Martian taxonomist shown only the genitals of a male human, a female human, and a male gibbon would have no hesitation in cla.s.sifying the two males as more closely related to each other than either is to the female. Indeed, the gene determining maleness (called SRY SRY) has never been in a female body, at least since long before we and the gibbons diverged. Traditionally, morphologists plead a special case for s.e.xual characteristics, to avoid 'nonsensical' cla.s.sifications. But identical problems arise elsewhere. We saw it previously with ABO blood groups, in Eve's Tale. My B-group gene relates me more closely to a B-group chimpanzee than an A-group human. And it is not just s.e.x genes or blood groups, but all all genes and characteristics which are susceptible to this effect, under certain circ.u.mstances. The majority of both molecular and morphological characteristics show chimps as our closest relatives. But a sizeable minority show that gorillas are instead, or that chimps are most closely related to gorillas and both are equally close to humans. genes and characteristics which are susceptible to this effect, under certain circ.u.mstances. The majority of both molecular and morphological characteristics show chimps as our closest relatives. But a sizeable minority show that gorillas are instead, or that chimps are most closely related to gorillas and both are equally close to humans.

This should not surprise us. Different genes are inherited through different routes. The population ancestral to all three species will have been diverse each gene having many different lineages. It is quite possible for a gene in humans and gorillas to be descended from one lineage, while in chimps it is descended from a more distantly related one. All that is needed is for anciently diverged genetic lineages to continue through to the chimphuman split so humans can descend from one and chimps from another.6 So we have to admit that a single tree is not the whole story. Species trees can can be drawn, but they must be considered a simplified summary of a mult.i.tude of gene trees. I can imagine interpreting a species tree in two different ways. The first is the conventional genealogical interpretation. One species is the closest relative of another if, out of all the species considered, it shares the most recent common genealogical ancestor. The second is, I suspect, the way of the future. A species tree can be seen as depicting the relationships among a democratic majority of the genome. It represents the result of a 'majority vote' among gene trees. be drawn, but they must be considered a simplified summary of a mult.i.tude of gene trees. I can imagine interpreting a species tree in two different ways. The first is the conventional genealogical interpretation. One species is the closest relative of another if, out of all the species considered, it shares the most recent common genealogical ancestor. The second is, I suspect, the way of the future. A species tree can be seen as depicting the relationships among a democratic majority of the genome. It represents the result of a 'majority vote' among gene trees.

The democratic idea the genetic vote is the one that I prefer. In this book, all relationships between species should be interpreted in this way. All the phylogenetic trees I present should be viewed in this spirit of genetic democracy, from the relationships between apes to the relationships between the animals, plants, fungi and bacteria.

1 Siamangs were separated off because they are larger, and they have a throat sac for amplifying their calls. Siamangs were separated off because they are larger, and they have a throat sac for amplifying their calls.

2 Perhaps the good old-fashioned family values of the gibbons, and the pious hope that our evolutionary ancestors once shared them, should be drawn to the attention of the right-wing 'moral majority', whose ignorant and single-minded opposition to the teaching of evolution endangers educational standards in several backward North American States. Of course, to draw any moral would be to commit the 'naturalistic fallacy' but fallacies are what these people do best. Perhaps the good old-fashioned family values of the gibbons, and the pious hope that our evolutionary ancestors once shared them, should be drawn to the attention of the right-wing 'moral majority', whose ignorant and single-minded opposition to the teaching of evolution endangers educational standards in several backward North American States. Of course, to draw any moral would be to commit the 'naturalistic fallacy' but fallacies are what these people do best.

3 The subject matter of this tale inevitably makes it tougher than other parts of the book. Readers should either don thinking caps for the next thirteen pages, or skip now to page 143 and return to the tale when they want their neurons exercised. Incidentally, I have often wondered what a 'thinking cap' actually is. I wish I had one. My benefactor Charles Simonyi, one of the world's greatest computer programmers, is said to wear a special 'debugging suit' which may help to account for his formidable success. The subject matter of this tale inevitably makes it tougher than other parts of the book. Readers should either don thinking caps for the next thirteen pages, or skip now to page 143 and return to the tale when they want their neurons exercised. Incidentally, I have often wondered what a 'thinking cap' actually is. I wish I had one. My benefactor Charles Simonyi, one of the world's greatest computer programmers, is said to wear a special 'debugging suit' which may help to account for his formidable success.

4 The actual number is (325)(425)(525)...( The actual number is (325)(425)(525)...(n25) where n n is the number of groups. is the number of groups.

5 The 'British Library' ma.n.u.script belonged to Henry Dene, Archbishop of Canterbury in 1501, and, together with the Egerton ma.n.u.script and others, is now kept at the British Library in London. The 'Christ Church' ma.n.u.script now resides close to where I am writing, in the library of Christ Church, Oxford. The earliest record of the 'Hengwrt' ma.n.u.script shows it belonging to Fulke Dutton in 1537. Damaged by rats gnawing at the sheepskin on which it is written, it is now in the National Library of Wales. The 'British Library' ma.n.u.script belonged to Henry Dene, Archbishop of Canterbury in 1501, and, together with the Egerton ma.n.u.script and others, is now kept at the British Library in London. The 'Christ Church' ma.n.u.script now resides close to where I am writing, in the library of Christ Church, Oxford. The earliest record of the 'Hengwrt' ma.n.u.script shows it belonging to Fulke Dutton in 1537. Damaged by rats gnawing at the sheepskin on which it is written, it is now in the National Library of Wales.

6 The longer the time between species splits (or the smaller the population size), the more ancestral lineages are lost by genetic drift. So tidy-minded taxonomists, who hope that species trees coincide with gene trees, will find it easier to deal with animals whose divergences are well s.p.a.ced out in time, unlike African apes. But there are always genes, such as The longer the time between species splits (or the smaller the population size), the more ancestral lineages are lost by genetic drift. So tidy-minded taxonomists, who hope that species trees coincide with gene trees, will find it easier to deal with animals whose divergences are well s.p.a.ced out in time, unlike African apes. But there are always genes, such as SRY SRY, for which separate lineages are systematically maintained by natural selection over huge spans of time.

Rendezvous 5.

OLD WORLD MONKEYS.

As we near this rendezvous and prepare to greet Concestor 5 approximately our 1.5-million-greats-grandparent we cross a momentous (if somewhat arbitrary) boundary. For the first time in our journey we leave one geological period, the Neogene, to enter an earlier one, the Palaeogene. The next time we do this will be to burst into the Cretaceous world of the dinosaurs. Rendezvous 5 Rendezvous 5 is scheduled at about 25 million years ago, in the Palaeogene. More specifically it is in the Oligocene Epoch of that Period, the last stop on our backward journey when the climate and vegetation of the world are recognisably similar to today's. Much further back, and we shall not find any evidence of the open gra.s.slands that so typify our own Neogene Period, or the wandering herds of grazers that accompanied their spread. Twenty-five million years ago, Africa was completely isolated from the rest of the world, separated from the nearest piece of land Spain by a sea as wide as that which separates it from Madagascar today. It is on that gigantic island of Africa that our pilgrimage is about to be invigorated by a new influx of spirited and resourceful recruits, the Old World monkeys the first pilgrims to arrive bearing tails. is scheduled at about 25 million years ago, in the Palaeogene. More specifically it is in the Oligocene Epoch of that Period, the last stop on our backward journey when the climate and vegetation of the world are recognisably similar to today's. Much further back, and we shall not find any evidence of the open gra.s.slands that so typify our own Neogene Period, or the wandering herds of grazers that accompanied their spread. Twenty-five million years ago, Africa was completely isolated from the rest of the world, separated from the nearest piece of land Spain by a sea as wide as that which separates it from Madagascar today. It is on that gigantic island of Africa that our pilgrimage is about to be invigorated by a new influx of spirited and resourceful recruits, the Old World monkeys the first pilgrims to arrive bearing tails.

Today, the Old World monkeys number just under 100 species, some of which have migrated out of their mother continent into Asia (see the Orang Utan's Tale). They are divided into two main groups: on the one hand are the colobus monkeys of Africa together with the langurs and proboscis monkeys of Asia; on the other hand are the mostly Asian macaques plus the baboons and guenons, etc. of Africa.

The last common ancestor of all surviving Old World monkeys lived some 11 million years later than Concestor 5, probably around 14 million years ago. The most helpful fossil genus for illuminating the period is Victoriapithecus Victoriapithecus, which is now known from more than a thousand fragments, including a splendid skull, from Maboko Island in Lake Victoria. All the Old World monkey pilgrims join hands around 14 million years ago to greet their own concestor, perhaps Victoriapithecus Victoriapithecus itself, or something like it. They then march on backwards to join the ape pilgrims at our own Concestor 5, 25 million years ago. itself, or something like it. They then march on backwards to join the ape pilgrims at our own Concestor 5, 25 million years ago.

And what was Conces