The Blind Watchmaker (56 page)

Above are seven biomorphs differing only in their ‘Number of Segments’ gene or their ‘Distance between Segments’ gene. The lefthand biomorph is the old, familiar branching tree, and the others are just serially arranged repetitive trains of the same basic tree. The simple tree, like all the original Blind Watchmaker biomorphs, is the special case of a ‘one-segment animal’.

So far, I have talked only about uniform, centipede-like segmentation. Lobster segments differ from one another in complicated ways. A simpler way in which segments can vary is through ‘gradients’. A woodlouse’s segments are more like one another than a lobster’s, but they are not as uniform as a typical millipede’s (actually some apparent ‘woodlice’ or ‘pillbugs’ are technically millipedes). A woodlouse is narrow at the front and back, and broadens in the middle. As you work your way from the front to the back of the train, the segments have a size
gradient
which peaks in the middle. Other segmented animals, such as the extinct trilobites, are widest at the front and taper to the rear. They have a simpler size gradient which peaks at one end. It was this simpler kind of gradient that I sought to imitate in my segmented biomorphs. I did it by adding a constant number (which could be a negative number) to the expressed
value
of a particular gene going from front to back. Of the following three biomorphs, the lefthand one has no gradients, the middle one has a gradient on Gene 1, and the

riir^it-l^irt\l nnfa
*\f\n f’l^rtt*\**^*

Having expanded the basic biomorphic chromosome by these two genes and the associated gradient genes, I was ready to unleash the new-style biomorphic embryology into the computer and see what it could do by way of evolution. Compare the following picture with Chapter 3’s Figure 5, all of whose biomorphs lack segmentation.

I think you will agree that a more ‘biologically interesting’ range of evolutionary versatility has now become available. The ‘invention’ of segmentation, as a new breakthrough in embryology, has opened floodgates of evolutionary potential in the land of computer biomorphs. My conjecture is that something like that happened in the origin of the vertebrates, and in the origin of the first segmented ancestors of insects, lobsters and millipedes. The invention of segmentation was a watershed event in evolution.

Symmetry was the other obvious innovation. The original Blind Watchmaker biomorphs were all constrained to be symmetrical about the midline. I introduced a new gene to make this optional. This new gene determined whether a biomorph with its original nine gene values set to those of the basic tree looked like (a) or like (b). Other genes determined whether there was symmetric reflection in the updown plane, (c), or full four-way symmetry, (d). These new genes could vary in all combinations, as in (e| and |f). When segmented animals were asymmetrical in the midline plane, I introduced a botanically inspired constraint: alternate segments should be asymmetrical in opposite directions, as in (g).

Armed with these further genes, I again set about a vigorous breeding program, to see whether the new embryology could foster a more exuberant evolution than the old. Here is a portfolio of segmented biomorphs with midline asymmetry:

(picture absent)

And here are some radially symmetrical biomorphs whose segmentation, if any, may be as cryptic as that of the adult human head:

(picture absent)

The gene for full radial symmetry tempts the selector to breed pleasing abstract designs rather than the biologically realistic ones that I had previously sought. The same is even more true of the colour version of the program that I am at present developing.

One group of animals, the echinoderms (including starfish, sea urchins, brittle stars and sea lilies), are highly unusual in being five-sidedly symmetrical. I am confident that, no matter how hard I or anybody else tries, we shall never find five-sided symmetry emerging by random mutation from the existing embryology. This would require a new ‘watershed’ innovation in biomorph embryology, and I have not attempted it. But freak starfish and sea-urchins with four or six arms rather than the usual five do sometimes turn up in nature. And in the course of exploring biomorph land I have encountered superficially starfish-like or urchin-like forms that have encouraged me to select for an increased resemblance. Here is a collection of echinoderm-like biomorphs, although none of them has the requisite five arms:

(picture absent)

As a final test of the versatility of my new biomorphic embryology, I set myself the task of breeding a biomorphic alphabet good enough to sign my own name. Every time I encountered a biomorph that resembled, however slightly, a letter of the alphabet, I bred and bred to enhance the resemblance. The verdict on this ambitious endeavour is mixed, to say the least. T and ‘N’ are well-nigh perfect. ‘A’ and ‘H’ are respectable if slightly ungainly. ‘D’ is poor, while to breed a proper ‘K’ is, I suspect, downright impossible - I had to cheat by borrowing the upright stroke of the ‘W’. Yet another gene would have to be added, I suspect, before a plausible ‘K’ could be evolved.

After my somewhat illiterate attempt to sign my own name, I had more luck with evolving that of the inspired artefact with which all this work was done:

(picture absent)

It is my strong impression, borne out, I hope, by the illustrations here, that the introduction of a few radical changes in the fundamental embryology of the biomorphs has opened up new vistas of evolutionary possibility which simply were not available to the original program described in Chapter 3. And, as I said earlier, I believe that something similar happened at various junctures in the evolution of some prominent groups of animals and plants. The invention of segmentation by our own ancestors, and separately by the ancestors of insects and crustaceans, is probably only one of several examples of ‘watershed’ events in our evolutionary history. These watershed events are, at least when seen with the wisdom of hindsight, different in kind from ordinary evolutionary changes. Our first segmented ancestors, and the first segmented ancestor of earthworms and insects, may not have been particularly good at surviving as individuals - though obviously they
did
survive as individuals, or we, their descendants, would not be here. My point now is that the invention of segmentation by these ancestors was more significant than just a new technique for surviving, like sharper teeth or keener eyes. When segmentation was added to the embryonic procedures of our ancestors, whether or not the individual animals concerned became better at surviving, the lineages to which they belonged suddenly became
better at evolving
.

Modem animals, we vertebrates and all our fellow-travellers on this planet, inherit the genes of an unbroken line of ancestors that were good at individual survival. That much I tried to make clear in
The Blind Watchmaker
. But we also inherit the embryological procedures of ancestral lineages that were good at
evolving
. There has been a kind of higher-order selection among lineages, not for their ability to survive but for their longer-term ability to evolve. We bear the accumulated improvements of a number of watershed events, of which the invention of segmentation is just one example. It is not just bodies and behaviour that have evolved in improved directions. We could even say that evolution itself has evolved. There has been a progressive evolution of evolvability.

The Macintosh version of the Blind Watchmaker program has menu options to turn on or off each of the main categories of mutation. By turning off all the new types of mutation one reverts to the earlier version of the program (or the present IBM version). Breed for a while under these conditions, and youget somefeelfor theenormous range offaunas permitted by the earlier program, but also for its limitations. If you then switch on, say, segmentation mutations, or symmetry mutations (or if you switch from IB M to a Macintosh! ), you can exult in something of the feelingofliberation that may have attended evolution’s great watershed events.

Reference Dawkins, R. (1989) The evolution of evolvability. In C.Langton (ed.)

Artificial Life
. New York: Addison-Wesley.