The year before my ‘Evolution of evolvability’ lecture, in The Blind Watchmaker I had written of ‘Stretched DC-8’ as opposed to ‘Boeing 747’ macro-mutations. The distinguished astronomer Sir Fred Hoyle (not the first or the last physicist to blunder ineptly1 into biology) expressed his scepticism of Darwinism with the image of a hurricane blowing through a junkyard and having the luck to assemble a Boeing 747. He was talking about the origin of life (abiogenesis), but his metaphor has become a favourite with creationists casting doubt on evolution itself. The point they miss, of course, is the power of cumulative natural selection, the slow climb up the gentle slopes of Mount Improbable. There’s a photo of me in the picture section standing in an aircraft graveyard, keeping a weather eye out for hurricanes that might spontaneously put together a Boeing 747.
I invoked another airliner, the Stretched DC-8, in a contrasting metaphor. This was a version of the DC-8 airliner, lengthened 11 metres by the insertion of two extra plug-ins, 6 extra metres in the forward fuselage and 5 metres in the aft fuselage. It was a DC-8 with two homeotic mutations. We can think of each row of seats in the additional portions of the fuselage, with its associated tray tables, lights, ventilators, call buttons, music ports etc. as a segment, a duplicate of the pre-mutation segments. My biological point was that while there is a fundamental objection to a radically new, complex animal, or complex organ, being produced in a single mutational leap (Hoyle’s 747), there is no principled objection to the duplication of whole segments, no matter how complex each segment might be (my DC-8). You can’t invent a vertebra from scratch. But you can make a second vertebra, given that a first already exists, in a single mutation. The embryological machinery that can make one segment can make two segments, or ten. And we now even know about the homeotic mechanism that does it.
Embryological mechanisms can also easily stretch every one of a series of segments. I’d still call the result a ‘Stretched DC-8’ even though that isn’t the way the airliner ‘mutated’. This is because it’s not a major jump in complexity, as a hypothetical ‘747 mutation’ would be. A giraffe has the same number of neck vertebrae as any ordinary mammal: seven. The giraffe’s neck achieves its great length by stretching all seven cervical vertebrae. I strongly suspect that it happened gradually, but there would be no objection, of the principled, insuperable ‘747’ type, to the proposition that the neck shot out in a single macro-mutation, affecting all seven vertebrae simultaneously. The existing embryological machinery to make neck vertebrae, with all their associated complexity of nerves, blood vessels and muscles, was all present and correct. All that was needed was a quantitative tweak in some growth field, to stretch all seven simultaneously and massively. And the same would have been true if – as in snakes – the elongation had been achieved by duplicating the vertebrae rather than by stretching each one.
The authoritarian regime in George Orwell’s 1984 prescribed a daily ‘Two Minutes Hate’ against a renegade party member called Goldstein (shades of Trotsky, or the ‘fallen angel’ myth of Satan). Substitute ‘scorn’ for ‘hate’ and you have some idea of the prevailing reaction, in the Oxford Zoology Department of my undergraduate days, to the German American geneticist Richard Goldschmidt, largely under the influence of E. B. Ford. Goldschmidt’s ‘hopeful monster’ idea of the evolutionary importance of macro-mutations is indeed misguided in the contexts in which he proposed it (for example, in the very ‘Oxford’ bailiwick of butterfly mimicry), but since he never strayed from conscionable ‘Stretched DC-8’ territory towards ‘Boeing 747’ macro-mutational fantasy, Goldschmidt was not in principle beyond the pale. And it would be hard to fault the title ‘hopeful monster’ for the first segmented animal – not that anybody has ever seen a fossil of that long-dead Model-T of morphological mass production.
Macro-mutations (mutations of large effect) do occur. There is no principled objection to a macro-mutation being incorporated into a gene pool as the norm, although it seldom happens. My principled objection is to the idea of a macro-mutation putting together a brand new, complex, functioning organ or system, with many parts whose simultaneous combination would be too much of a coincidence: something like an eye, with its retina, lens, focusing muscles, aperture-controlling machinery and so on. There is no principled objection to the idea that the ‘four-eyed fish’, Bathylychnops, could have acquired its two extra eyes in a single macro-mutation. Indeed, that is probably how it did happen, a lovely example of Stretched DC-8 evolution, by homeotic mutation. The embryonic machinery of the pre-mutated ancestor already ‘knew’ how to make an eye. But any one of those eyes, or indeed any vertebrate eye, could not have been made from scratch in a single mutational step: such ‘747 evolution’ would be inadmissibly miraculous. The machinery of a vertebrate eye had originally to be built up gradually, step by step.
Hereabouts, by the way, lies the answer to the silly claim, originating with Stephen Gould and often reiterated, that Darwin, as a ‘gradualist’, would have been opposed to so-called ‘punctuated’ evolution. Darwin was a ‘gradualist’ only in the sense that he would have had no truck with 747 macro-mutations. Although Darwin obviously didn’t use airliner terminology, the nature of his objections was such as to preclude only macro-mutations of the 747 kind, not the Stretched DC-8 variety.
The evolution of language could be an interesting test case for discussion. Could the ability to speak have arisen in a single macro-mutation? As I mentioned on page 290, the main qualitative feature that separates human language from all other animal communication is syntax: hierarchical embedment of relative clauses, prepositional clauses etc. The software trick that makes this possible, at least in computer languages and presumably in human language too, is the recursive subroutine. A subroutine is a piece of code that, when called, remembers whence it was called and returns there when it is finished. A recursive subroutine has the additional ability to call itself and then return to an outer (more global) version of itself. I went into this in detail in Appetite for Wonder, so will here content myself with the summary diagram below. The sentence was composed by a computer program that I wrote, capable of generating an infinite number of perfectly grammatical (if lacking in semantic content) sentences, recognizable as syntactically correct by any native speaker of English. I have parsed this particular sentence using brackets and a typeface which shrinks with the depth of embedment. Notice how the subsidiary clauses embed themselves within the main sentence, rather than being tacked on at the end.
It takes almost no effort to write a program capable of generating any number of grammatically correct (though semantically empty) sentences of this kind. But only if your computer language allows recursive subroutines. It could not, for example, have been written in the original IBM Fortran language, or any of its contemporary rivals. I wrote it in the only slightly younger language, Algol 60, and it could easily be written in any of the more modern programming languages developed after the ‘macro-mutation’ of recursive subroutines was introduced.
It looks as though the human brain must possess something equivalent to recursive subroutines, and it’s not totally implausible that such a faculty might have come about in a single mutation, which we should probably call a macro-mutation. There’s even some suggestive evidence that a particular gene called Fox P2 might be involved, since those rare individuals with a mutated version of this gene can’t talk properly. More tellingly, this is one of the minority of regions of the genome where humans are unique among the Great Apes. However, the evidence on Fox P2 is unclear and controversial and I won’t discuss it further. The reason I am prepared to contemplate macro-mutation in this case is a logical one. Just as you can’t have half a segment, there are no intermediates between a recursive and a non-recursive subroutine. Computer languages either allow recursion or they don’t. There’s no such thing as half-recursion. It’s an all or nothing software trick. And once that trick has been implemented, hierarchically embedded syntax immediately becomes possible and capable of generating indefini
tely extended sentences. The macro-mutation seems complex and ‘747-ish’ but it really isn’t. It’s a simple addition – a ‘stretched DC-8 mutation’ – to the software, which abruptly generates huge, runaway complexity as an emergent property. ‘Emergent’: important word, that.
If a mutant human was born, suddenly capable of true hierarchical syntax, you might well ask who she could talk to. Wouldn’t she have been awfully lonely? If the hypothetical ‘recursion gene’ was dominant, this would mean that our first mutant individual would express it and so would 50 per cent of her offspring. Was there a First Linguistic Family? Is it significant that Fox P2 actually does happen to be a genetic dominant? On the other hand, it’s hard to imagine how, even if a parent and half her children did share the software apparatus for syntax, they could immediately start using it to communicate.
I’ll briefly mention again the possibility, which I discussed in An Appetite for Wonder, that this recursive software might have been used for some pre-linguistic function such as planning a hunt for an antelope or a battle against a neighbouring tribe. Each phase of a cheetah’s hunt has a series of appetitive routines, calling subordinate routines, each one terminated by a ‘stopping rule’ which signals a return to the point in the superior program from which the subroutine was called. Could this subroutine-based software have paved the way for linguistic syntax, just waiting for the last macro-mutation to fall into place, the mutation that allowed a subroutine to call itself – recursion?
Noam Chomsky is the genius mainly responsible for our understanding of hierarchically nested grammar, as well as other linguistic principles. He believes that human children, unlike the young of any other species, are born with a genetically implanted language-learning apparatus in the brain. The child learns the particular language of her tribe or nation, of course, but it is easy for her to do so because she is simply fleshing out what her brain already ‘knows’ about language, using her inherited language machine. Hereditarian tendencies in intellectuals nowadays (though not always in the past) tend to be associated with the political right, and Chomsky, to put it mildly, hails from the opposite pole of the political spectrum. This disjunct has sometimes struck observers as paradoxical. But Chomsky’s hereditarian position in this one instance makes sense and, more to the point, interesting sense. The origin of language may represent a rare example of the ‘hopeful monster’ theory of evolution.
Less dramatic than hopeful monsters such as might have initiated segmentation or, arguably, language, there could have been lots of embryological innovations which, while not conferring dramatic survival advantages on their first individual possessors, opened floodgates to future evolution. And so we return to the evolution of evolvability. In coining this phrase at the Los Alamos conference, I meant to include a kind of higher-order natural selection which we notice with hindsight. A new innovation, whether or not it directly improves the survival of the individual in the short term, leads to multiple evolutionary branchings such that its descendants inherit the earth. Segmentation was my primary example, and language might be an especially dramatic one, but there are others. The early adaptations that enabled fish to leave the water and invade the land didn’t just help those pioneers to a new source of food, or a new way to escape marine predators. They pioneered new living environments, not just for individual survival in the short term but for clades blossoming through future ages. Just as Darwinian selection favours adaptations that help individuals to survive, so there can be a higher-order, non-Darwinian selection (or Darwinian only in a vague and arguably confusing sense) among lineages for the quality of evolvability. Such was the point I made in my ‘evolution of evolvability’ lecture at the Los Alamos conference, and I illustrated it with my computer biomorphs and the new vistas of evolution that were opened up to them when I rewrote the program with new genes for segmentation, and symmetry in various planes.
In the question session after my lecture (sympathetically chaired by the distinguished theoretical biologist Stuart Kauffman) somebody jokingly asked whether my biomorph program, in addition to breeding an alphabet, could breed money. In a flash, I was able to bring up on the screen a passable dollar sign (see the ‘s’ in my signature on page 375), and thus my talk came to its end in good-humoured laughter.
Kaleidoscopic embryos
Although my Los Alamos talk was called ‘The evolution of evolvability’, I didn’t at that stage carry the theme as far as I might have. The chapter of Climbing Mount Improbable called ‘Kaleidoscopic embryos’ went further, and in a direction that I find rather satisfying. I’ve already mentioned the ‘mirror genes’ that I introduced in one of the later versions of my biomorph program. The genes that control animal symmetry in various planes can also be thought of as inserting ‘mirrors’ in the embryo, like the mirrors in a kaleidoscope. Most, but not all, animals have this kind of mirror running down the midline, which makes them symmetrical in the left/right direction. A mutation in the third leg of an insect might theoretically affect the right side only, but actually it is mirrored on the left side too. Technically this mirroring is a constraint, for it restricts freedom to evolve: without it, perfect symmetry could still be achieved – ‘contrived’ might be more apt – by separate mutations on the two sides, along with a whole lot of exotic asymmetries. But if we assume (as is plausible for reasons I discussed in Climbing Mount Improbable) that there is some more global benefit to left/right symmetry itself, evolutionary improvement is speeded up if mutations are automatically mirrored on both sides. Therefore, rather than being seen as a constraint (which it strictly is), the imposition of symmetry (a midline ‘mirror’ in the embryonic kaleidoscope) can be seen as the opposite – as an evolutionary advance in evolvability.
The same is true of other planes of symmetry, although these are less common in real biology. On the left in the illustration on the following page is a computer biomorph with four-way symmetry (two ‘kaleidoscopic mirrors’ at right angles). In the middle is the skeleton of a radiolarian (an exquisite microscopic single-celled creature) and on the right is a stalked jellyfish (obviously not to the same scale). All of them have ‘two mirrors’ at right angles, buried deep in their embryology. In the case of the biomorph, I know this is true because I wrote the embryology’s software. In the case of the two real animals, I don’t know for sure but I would bet my shirt that the four-way symmetry is a default constraint in the embryology. My conjecture is that whatever was the innovation in the fundamental embryology that set up this kaleidoscopic constraint, it had an advantage; and I would want to call that innovation an evolutionary advance in evolvability.
Echinoderms (starfish, sea urchins, brittle stars etc.) mostly have five-way symmetry. Once again, it seems to me almost obvious that the relevant symmetry rule is lodged deep within the embryology, such that a mutation in a detail at the tip of one arm of a starfish, say, is mirrored in all five arms (this generalization is not negated by the fact that occasionally starfish with more than five arms turn up). And once again, given that the symmetry is for some reason a good thing for a starfish to have, ‘mirroring’ the mutations is a short cut (when compared with piecemeal change in each arm separately) to achieving change without departing from five-way symmetry. It therefore deserves to be considered under the heading of ‘evolution of evolvability’. And it is significant that all my efforts to breed five-way symmetrical biomorphs on the computer screen failed. It’s almost obvious. Five-way symmetry could be achieved only by a radical rewrite of the embryology routine – which again makes the point that we really are talking about the evolution of evolvability. The ‘echinoderm’ biomorphs that I managed to breed on the screen are all ‘cheats’ (see illustration opposite). They look superficially like a sand dollar, a sea lily, a sea urchin, a brittle star and two starfish respectively, but not one of them is five-way symmetrical.
At the time of the Los Alamos conference there were no colour Macs. When I finally got one, the obvious next step in expanding the genome of my biomorphs
was to add a new suite of genes for colour. At the same time I added genes to modify the lines with which the basic trees of the embryology algorithm were drawn. Simple lines were still allowed, but I introduced a new gene to change their thickness, and other new genes to change them from simple lines to rectangles or ovals; to control whether these shapes were filled or open; and to control the colour of the lines and of the fill. These extra genes opened up new floodgates of evolution, tempting the selecting human to breed biomorphs that looked more and more like exotic flower designs, table mats and butterflies. I had a fancy to take the computer out into the garden and offer real bees and butterflies the chance to choose the ‘flowers’ and ‘butterflies’ on the screen. I hoped that real insects would breed simulacra of real species of flowers, starting from non-flower-like beginnings. Unfortunately, it turned out – as I should have anticipated – that the bright daylight that brings the insects out to forage is the very same bright daylight that makes the screen difficult to see. As so often happens with seemingly bright ideas, I shelved the project and never came back to it. Perhaps night-flying moths? Would a modified version of a touch-sensitive screen like an iPad’s respond directly to the buffeting of a moth?
I was creating colour biomorphs at about the time I met Lalla. Embroidery is one of her many talents – at that time she hadn’t moved on to mosaics, painting ceramics or (her current art forms) weaving and drawing with a sewing machine – and she was inspired by the coloured four-way symmetrical biomorphs to embroider cushions and chair covers in which the stitches of the embroidery corresponded exactly to pixels on the computer screen (see picture section). They are still much admired twenty years later.
Brief Candle in the Dark Page 37