Theory Creep: The Quiet Shift in Evolutionary Thought
It's amazing how thinking on evolution has shifted since I started following the subject in the 1980s. Today's biologists clearly have realized that evolutionary theory must be revised to avoid conflict with genomic data, and yet they are very reluctant to say that the problems forcing the change are deep problems. A consequence of this business-as-usual approach is that young biologists may be unaware that they are inheriting a version of Darwinism that would have been considered quite peculiar only a generation ago.
One particularly telling change has to do with how sequence data are used to build phylogenetic trees -- branching diagrams that represent proposed evolutionary relationships among species. Since these trees are, in essence, nothing more than hypotheses, scientists ought to view them that way. The whole point, after all, is to see how well they stand up to critical scrutiny. Yet this is one important respect in which evolutionary thinking has shifted.
In the 1970s, before DNA sequencing became routine, scientists used various properties of proteins, including their amino-acid sequences, to construct phylogenetic trees. The data sets were much smaller then than they are now, but the basic principles and the underlying assumptions were the same. Corresponding proteins from several species were compared in order to quantify their similarities, and species trees were constructed from the results.
Although the belief that all species are related was then as sacrosanct as it is now, there was at least a general recognition then that trees are only as believable as the evidence supporting them. In particular, most biologists recognized the importance of consistency among inferred species trees, meaning that the analysis performed on corresponding genes or proteins taken from several species ought to yield the same tree regardless of which genes or proteins were used. Ellen Prager and Allan Wilson put it this way in 19761:
To test the reliability of the use of proteins for working out phylogenetic relationships, it is essential to determine whether the branching order of the evolutionary tree one obtains depends on the protein studied.
If the analysis gives different trees when applied to different proteins, then something is amiss. A decade later, David Hillis was equally emphatic2:
A primary objective of phylogenetic studies is to reconstruct the evolutionary history of a group of organisms. Because the organisms under study have a single history, systematic studies of any set of genetically determined characters should be congruent with other such studies based on different sets of characters in the same organisms. Congruence between studies is strong evidence that the underlying historical pattern has been discovered; conflict may indicate theoretical or procedural problems in one or both analyses, or it may indicate that additional data are needed to resolve the phylogenetic relationships in question.
But if we fast-forward two more decades, it becomes clear that the consistent picture that everyone expected -- all genes confirming the same pattern of species relationships -- is not to be. What we have instead is something of a mess, as James Degnan and Noah Rosenberg made clear in a paper published in 20093:
Many of the first studies to examine the conflicting signal of different genes have found considerable discordance across gene trees: studies of hominids, pines, cichlids, finches, grasshoppers and fruit flies have all detected genealogical discordance so widespread that no single tree topology predominates.
And despite consistent attempts to portray this as something less than a crisis for evolutionary theory, the news found its way into the popular press. That same year, The Telegraph jumped on the story with an article titled, "Charles Darwin's tree of life is 'wrong and misleading,' claim scientists"4.
So now that the secret is out, what are we to make of it? You would think that if "congruence between studies is strong evidence that the underlying historical pattern has been discovered", then a widespread lack of gene-tree congruence ought to prompt an acknowledgment of historical uncertainty. After all, why should we believe that gene trees reveal species histories if the genes themselves can't agree on those histories?
Indeed, it becomes so easy to construct utterly fictitious evolutionary histories when we drop the expectation of consistency that such a move ought to be viewed as undermining the whole exercise of phylogenetic reconstruction. Whisky, kerosene and milk have no common pedigree, but that wouldn't stop us from concocting one if we were to lower the standard in that way. The only prospect of elevating tree-building to something more than a game, then, is that it might uncover a strikingly consistent pattern of relationship between species. And the sobering truth is -- it doesn't.
But it's also a fact that the scientific community's commitment to naturalistic evolution of some kind runs deep. In the words of evolutionary biologist Eric Bapteste4: "If you don't have a tree of life, what does it mean for evolutionary biology? At first it's very scary--but in the past couple of years people have begun to free their minds." According to philosopher of biology John Dupré, it's all "part of a revolutionary change in biology. Our standard model of evolution is under enormous pressure. We're clearly going to see evolution as much more about mergers and collaboration than change within isolated lineages"4.
That's a start, maybe. But we have something even more revolutionary in mind.
Image credit: Cropped version of photo by Norbert Nagel, Wikimedia Commons.