Many of us who have endured the columns of Andrew Brown at the Guardian feel that it’s time for the man to move on.  His opinions lurch all over the place (usually, however, centered on the evils of atheism and the benefits of religion), but the writing is so absolutely dreadful that it’s hard to believe that someone pays him for his prose.  When I read his latest post,”On Belief’s new look, and old arguments,” I finally decided that he needs to go.  It’s not that he attacks atheism and defends religion here—for he doesn’t—but he just rabbits on and on about evolution without seeming to understand it.

Brown praises Dawkins’s The Selfish Gene (agreed!) and argues that the idea of memes has not been fruitful (also agreed!), but then loses the plot when he gets to one of Dawkins’s best books, The Extended Phenotype.  If you’ve read that book, you’ll know that it’s about the adaptive effects that genes within an organism can have when they act outside the body of that organism.  That is, an animal’s genes can produce behaviors that enable it to modify its own environment in an adaptive way.

The classic example is beavers building dams: one could consider the dam as an aspect of the beaver’s “phenotype” (traits produced by beaver genes) in much the same way as evolution produces the beaver’s teeth and tail.  The same is true for parasites that affect the behavior of their host in adaptive way (as in the parasitic nematodes that propagate their own survival by turning their ant hosts into mock “berries”, changing the color of their black abdomens to red, making the ants raise them into the air, and weakening the connection between the abdomen and the rest of the body, all of which attracts sharp-sighted birds that spread the nematodes through their droppings).  What you see below is the extended phenotype of the nematode worm:

Fig. 1.  The ant Cephalotes atratus: uninfected (top), and infected with the nematodes (bottom)

Here’s the world’s largest extended animal phenotype: a beaver dam in Alberta that is 850 m long (2790 feet). It was discovered using Google Earth, and can be seen from space (click to enlarge):

Fig. 2.  A huge extended phenotype

The point is that genes can affect behavior, and that behavior can modify the environment in an adaptive way, whether the environment be a river surrounded by gnaw-able trees or a host that can be manipulated by parasites. (Much of Darwinian medicine, by the way, involves the idea that infectious organisms cause symptoms that help them propagate themselves:  cold viruses don’t debilitate us, but allow us to remain ambulatory so that we can sneeze, shake hands, and spread the virus, while malaria protists make us prostrate during bouts of fever, rendering us susceptible to the bites of mosquitoes who carry the vector).

Anyway, when trying to explain how “extended phenotypes” evolve, Brown ties himself up in a tangle of thought and language that completely confuses the reader:

The Extended Phenotype, however, is still worth thinking about. The idea that genes are selected because of their effects outside the body that carries them was, so far as I know, first made explicit by Dawkins, and really does make you look at the world another way. But where does it get us?

There are at least three situations in which you can talk about an extended phenotype. One is niche construction, where an animal changes the world around it to suit its purposes: birds make nests, beavers make dams, caddis larvaemake houses. To the extent that these processes are under genetic control, Dawkins can talk about these changes outside the animal as a product of its genes.

The second is seen in parasites and predators. There are all sorts of ways in which parasites and predators generally modify the behaviour of their victim species. These can be dramatic, like those parasites that lead infected ants to climb to the tops of blades of grass so that they can more easily be eaten by sheep, or almost undetectable, like the way in which mosquitoes manufacture proteins that stop mammalian blood from clotting while they drink it. Again, here, a gene is clearly being selected for its effects outside the body that carries it.

So far, so good.  But then he goes a bit off the rails:

Then there is the very interesting case of domesticated plants. These are spread around the world by humans, who breed them selectively. In the case of tobacco plants, or opium poppies, we’re breeding for the effects on our brain.

This is not a case of an extended phenotype evolving genetically in humans: it is a cultural practice favored for its usefulness.  We haven’t evolved genetic tendencies to spread or breed domesticated plants—that happened only in the last ten thousand years or so, and that’s not long enough for genes favoring a tendency to domesticate plants to spread in the human genome.  This is purely a cultural practice that’s a byproduct of our big brains. The modified plants may be seen as a “cultural extended phenotype,” but that’s just confusing.

But Brown is way off the mark when trying to explain how extended phenotypes evolve.  Really, the idea is simple: any gene will be favored if it causes a behavioral change that, by modifying the individual’s environment, gives that gene a replication advantage over other genes.  Pretty easy to envision.  And of course once the environment is modified, that sets up new selective pressures that weren’t present before. Once beavers evolved the ability to build dams, then their young are now exposed to a new environment—a cozy den inside the dam—that would favor all kinds of other behaviors, like modification of parental care, the cacheing of food in the lodges, genes favoring repair of dams when they’re damaged, and so on.

But Brown makes a mess of it:

So all these are examples of extended phenotypes, in which a gene carried in one body is transmitted as a result of its effects on another.

But where in this circuit is causation? Is it more useful to think of the gene as reaching out to change the world around its organism, or to consider the environment reaching inside to change the gene? It seems to me that causation goes very clearly in the second direction and that this is true all the way down to the DNA. After all, it is the cellular mechanisms that determine both which genes are expressed, and which bits of DNA constitute a gene.

The second alternative—the one favored by Brown—simply makes no sense.  The environment doesn’t reach inside the organism to change the gene, for what we’re trying to explain is how a genetic change alters the organism’s environment. The environment is not, in such a case, a passive interactor that forces genetic change.  True, the genetic changes must take place in a certain environment, but if one has to choose either of these explanations, I’d favor the first.

But in fact neither encompass the whole situation, for it’s the interaction between genes and the environment that favors the evolution of extended phenotypes.  Rodents able to cut down trees to build protective dens for their young were favored, but that depends both on genetic variation for tree-cutting and den-building and on a pre-existing environment that contains running water and trees.  The behavior is caused by changes in the proportion of the replicators (“alleles”, the different forms of genes), but the advantage of one allele over another depends completely on the environment in which the gene finds itself.  One needn’t choose between the primacy of gene or environment (except insofar as how the behavior is inherited): both are needed to explain the evolution of “extended phenotypes.” Brown is raising a false dichotomy and confusing the reader.

He goes on:

Even if we abstract away from the sequence, as theoretical biologists do, and consider abstract genes “for” or “against” particular behaviours it is still the concrete details of the environment that constitute the selection pressure on a gene.

Mostly true, though of course there are “internal” selective pressures as well, like the effects of a gene on embryology, physiology, and so on.

And it is selection that is the active process, not simple replication.

False. Replication of different forms of genes is also “active”, and that is what causes the extended phenotype to evolve.  In fact, in terms of things actually doing something, replication is active while the environment is passive.  And he hasn’t considered another “active” process: the mutations that produce behaviors that lead to “extended phenotypes.”

DNA sets boundaries to what the environment can achieve – you can select all you like, but you’ll never breed ravens to fly under water—

Hold on!  Isn’t that what penguins do?

Fig.3.  Birds (descended from flying birds) flying underwater

– but these constraints are the outer limits of what is possible. They don’t help much to predict what’s inside them. If you start with a wholly gene-centric model and think it carefully through, you can find you have reached a gene-peripheral one instead.

That’s such a poor piece of writing that I can’t figure out what it means. If he’s saying that organisms can only evolve within the limits of what’s genetically possible, then that’s obviously true. But we don’t know what is genetically possible: think of all the improbable species (like frogs and penguins) that one might not have though a priori possible.  But of course genes don’t predict evolution alone, because that depends on an interaction between genetic variation and environments.  But that’s obvious as well.  What he’s saying in this faux-clever language remains obscure.

Why does the Guardian retain Andrew Brown? Many have suggested it’s because his ideas are so muddled, yet expressed so forcefully, that he brings the paper lots of traffic in the form of angry readers. It’s as good a theory as any.  But what ticks me off the most is when he muddles biology.