Chapter 16: Evolutionary Systems
Where This Leaves the Toolkit
Gather up what this chapter has found, and a single pattern emerges across three quite different mechanisms, each examined on its own terms and given every fair chance to do more than that pattern allows.
Large-scale genetic change — genes copied over and over, a jumping gene dropped in front of a silent gene — produced a dramatic, fast outcome in bacteria under steady pressure: forty-six independent successes in weeks, where Lenski's flasks had needed years. But look at what made it fast. Copying is cheap. It goes on all the time, everywhere in the genome, with nothing selecting for it; at any moment, something like one bacterium in ten carries a duplicated stretch of DNA somewhere, and the flask was already full of spare copies of citT before the starvation began. Selection didn't have to wait for a lucky event. It only had to pick the cells holding the one copy that helped. And all of that speed depended on a single condition: that citT was already there, already able to do the job under other conditions. The one strain that lacked it produced nothing at all. Copying is easy. Inventing is the hard part, and this experiment shows the difference about as sharply as it can be shown. Nothing new was made. An existing gene was multiplied and handed an existing kind of switch.
The regulatory toolkit — a shared construction program, reused with modification, deciding whether the same machinery builds an arm or a leg — is the one exception this chapter promised to take seriously, and it is a harder case than the first. Here, too, there is no new gene and no new protein. But a new structure, at the level of the whole animal, can come from a new combination of old switches, and that is a stronger claim to novelty than anything copying offered. So the regulatory toolkit is an exception in one sense: it can produce new forms. It is not an exception in the sense that matters most here. It writes no new sequence; it rearranges the use of sequence that already exists. And the price of its efficiency turned out to be steep: the whole arrangement depends on modularity, on genes properly walled off from one another's layers, and the best attempts to show how such walls could arise gradually start with an organized, working system already in hand.
Modular protein construction pushed the same question down to the smallest scale, and the shortcut turned out to be real but narrow. It needs no factory, because selection does the sorting. But it only moves domains whose connectors happen to match; it played its major part in one branch of life and one kind of protein; it finds few working combinations; it relies on point mutation to make the seams fit; and above all, it moves domains without making them.
Three mechanisms, three different scales, one recurring answer. Each can do something real, and none of them does it for free. Every one of them either rearranges material that already existed, or depends on an organized system whose own origin is at least as hard to explain as whatever it was brought in to simplify.
That is why point mutation, and the large reserve of raw sequence it works in, are still standing at the end of this chapter, not retired by anything found here. The toolkit makes evolution very good at reusing what it has: copying it, moving it, switching it on in new places, combining it in new ways. It does nothing to lower the bill for making something that wasn't there before. That bill is still paid, if it is paid at all, the old way — one letter at a time, in DNA that previously did nothing — and Chapter 8 has already shown what the arithmetic of that payment looks like. The toolkit is real. It is not, on the evidence gathered here, a substitute for the harder question this book has been asking all along.
Last revised 2026-11-19