Metaphors are dangerous things. On the one hand, it seems pretty much impossible to avoid using them, especially in rather abstract fields like philosophy and science. On the other hand, they are well known to trick one’s mind into taking the metaphor too literally, thereby creating problems that are not actually reflective of the reality of the natural world, but are only perverse constructs of our own warped understanding of it.

Take the metaphor of living organisms as analogous to complex artifacts, which led William Paley to articulate the most famous argument in favor of Intelligent Design -- an argument that, incidentally, has not changed in its broad philosophical outline since the early 18th century. David Hume -- rather presciently, since he wrote before Paley -- pointed out that the metaphor is flawed. Hume argued that living organisms are not like watches, to use Paley’s analogy. They are not machines that are assembled, but organic beings that develop gradually over time. Accordingly, the ID argument of “irreducible complexity,” as it is known nowadays, doesn’t make sense because it is based on the machine metaphor. To put it another way, biochemist Michael Behe doesn’t understand how the bacterial flagellum could have evolved because he doesn’t understand evolution and insists in thinking of the flagellum as a “propulsion engine” analogous to those built by Evinrude.

Yet, even serious biologists (i.e., unlike Behe) have been guilty of enthusiastically pushing what is clearly a flawed metaphor: the idea that the DNA sequence of an organism’s genome is analogous to a computer “program,” and that it provides the “blueprint” for building said organism. Hence the wild (and, as it turns out, completely unfounded) claim a few years ago that sequencing the human genome will tell us everything there is to know about “making” a human being. The human genome has been sequenced, and what we have found is that genes, though playing a crucial causal role in development, are just one piece of a vast and as yet largely unresolved puzzle.

Ironically, the harbinger of the demise of the genetic program-blueprint metaphor is the serious study of genomics itself. A recent article by Tanguy Chouard in Nature (20 November 2008) explains why. Researchers are finding out that what matters is not so much individual genes, but the way networks of genes function together. Take the example of the Bicoid gene in Drosophila: it was thought to be essential in establishing the form of the body in all insects, based on its effects on the development of body shape in fruit flies. No such thing, as it turns out. Once scientists looked for Bicoid-like genes in other insects they simply did not find them! Turns out that Drosophilais an exception (ah, the perils of “model” organisms), and that in species from wasps to beetles the job carried out by Bicoid is achieved by minor rearrangements of a large regulatory network encompassing a myriad of other genes.

According to the same article, biologists are even beginning to document how evolution transitions from one regulatory network to another. Phylogenetically informed investigations conducted with different species of yeast, for instance, show nice intermediate stages from one arrangement to another, demonstrating that major changes at the genetic level can occur with minimal disruption of physiological or other phenotypic features. Finally, researchers have modeled the evolution of large regulatory networks and found that any particular phenotype can be underlined by a huge number of functionally equivalent genotypes, which implies that evolution of genetic networks can often be semi-neutral with respect to the organism’s fitness.

All of this should dispatch once and for all ideas like “genetic program” and “genetic blueprint,” thereby also dramatically undercutting any claim to genetic determinism (as opposed to more mild “genetic causalism,” for lack of a better term). There is no program or blueprint because the developmentally-relevant information is distributed among different levels of organization, including but not limited to the level of gene networks. So phenotypes are truly emergent properties of gene-gene (and of course, gene-environment) interactions. This is completely different from the case of human-made programs and blueprints, which feature a relationship between input and output that is much closer to a simple one-to-one mapping.

This is why living beings can evolve one complex feature after another without having to be “redesigned” from scratch. Hume was right: machines are simply not good metaphors for organisms, and it is time for stubbornly reductionist biologists to move on and search for better metaphors.