Natural Selection and Science as Prediction

Professor Steven Novella in his blog Neuralogica posted on a blogosphere controversy with one Egnor. Along the way he matter of factly announced that science was merely models useful for prediction, not about what's "correct," which in this context plainly meant corresponding to reality, i.e., true. This notion, in relation to natural selection, seems worthy of comment, on his comments section and here.) Post edited from the Neuralogica blog.

The principle of common descent would in another time or field of science would probably have been called a “law.” Scientific laws are commonly described as descriptions or generalizations about how things work in nature. Common descent was established on a mass of observations from morphology and embryology, and later confirmed further by genetic data. Scientific laws may be refined or possibly even wholly refuted by later observations. For instance, lateral gene transfer in bacteria seems to justify a refinement in the doctrine of common descent. There are antecedents to this principle, for instance, the cell theory, which implicitly requires that all changes in descent be physically possible for cells to carry out. And there are consequences too, such as that new species must come from old species. But is it really useful or enlightening to call these predictions?


Natural selection, on the other hand, is what some would call a theory, a scientific explanation, although I gather others would define “theory” as an algorithm(or verbal/conceptual equivalent) for generating predictions. In that case, this link [available in Prof. Novella's Neuralogica blog, July 19, 2012] says that Popper’s claim is that natural selection, equating adaptation to fitness, is non-predictive because all organisms must be adapted to their environments, else they would be extinct. The rebuttal is that 1) the unfit are extinct 2) physically impossible changes will not occur along the chain of common descent and 3) new species descend from old species.

Note that 1) is merely a restatement, not a prediction while 2) & 3) are trivial “predictions,” because they are not unique to the theory of natural selection. At this point, the link goes on to explain that fitness is defined causally, functionally, statistically a posteriori, not logically and semanticaly a priori. And besides that, fitness is not determininistic but dispositional. What is not clear is how fitness being those things makes it possible to make predictions. There have been a number of experiments that have successfully tested predictions about natural selection’s effects on gene frequencies but it seems that the difficulties in defining fitness in a way that you can predict/control makes them the exception rather than the rule in evolutionary science.

The thing worth thinking about, is that Darwin provided masses of evidence in favor of natural selection decades before experiments that teased out a prediction could be performed. Science since has provided masses more. But they weren’t predictions. [Omitted from comments to avoid pointless offense: The conclusion that Darwin wasn't doing much useful science because he didn't make many specific predictions, or propose many experiments, much less perform them, is demanded by the science as prediction perspective. Yet, it is nuts to deny Darwin was overwhelmingly right then, except for the modifications imposed later by a greater understanding of genetics. But the modern synthesis did not refute Darwin. An even greater understanding of modern genetics, as in the comprehension of the neutrality of so many mutations, is tending to modify the relative importance of natural selection in novel speciation, but is not "refuting" Darwin.]

You can use natural selection to explain vestigial organs. An organ is no longer adaptive, nature selects agains the waste of resources for it. The mechanics of genetics may not permit an easy way to simply erase the organ, but the slow increment of genetic changes diverts resources from the less fitting organ, it gets smaller and smaller, that is, vestigial. As the resources diverted become less cumbersome, however, the intensity of selection pressure becomes less and less. The vestigial organ can then survive indefintiely until the vagaries of genetic change do possibly succeed in erasing the last trace. Natural selection (particularly gene-selection) says traits are adaptive, increase fitness. We can explain fitness-decreasing vestigial organs as above, using supplementary hypotheses and contingencies that explain away the violation of this prediction.

We cannot predict which organs will become vestigial; we cannot predict which will finally disappear; we cannot predict for which a new function might be found; we can not statistically predict incidence of vestigialization, time for vestigialization, rates of vestigialization or intensity of natural selection against vestigial organs. But, rather than throw up our hands, isn’t the real clarification, not that natural selection is scientific because it is predictive, but, because it is explanatory of massive amounts of data. Charles Darwin made a convincing case for natural selection before the experiments. And the kind of evidence he presented has only been added to.

Even more to the point, if there are experiments confirming predictions of natural selection about speciation (instead of change in gene frequencies,) they are a well kept secret. I suppose it is likely that eventually science will find a way to conduct such experiments. But even if no one were ever ingeniuous enough to find the way, we already have quite a bit of evidence showing that natural selection is a major factor in novel speciation, and overwhelming evidence it is the major factor in maintaining species morphology (the forgotten aspect of speciation?) Is throwing out a lot of straw about predictions inviting the Egnors to make straw men?

[Addendum: The old canard that "natural selection is survival of the fittest" is a tautology is indeed a canard. First, technically, the whole statement is merely a definition, and definitions are themselves a kind of tautology. They are not logical arguments, but explanations. The phrase "survival of the fittest" can be interpreted as a tautology by insisting that you can't define fittest in any way other than by survival. Not true, and insofar as the link points this out, it is correct.

Second, the second problem with the capsule statement of natural selection lies in the unconscious assumption that to be a valid scientific theory, it must be predictive. And natural selection is not, repeat, not predictive in any meaningful sense. It is powerfully explanatory, as dwelt on above. Post hoc, it is in principle possible to define fitness meaningfully. An often unacknowledged consequence of this unpredictability means that there is no way to retrodict the past. All reconstructions in evolutionary history are open to the difficulties of all historical reconstructions. And all the quarrels about relative importance in principle of supplementary assumptions and ad hoc contigencies, too. As much tedious work and dispute as this enjoins on evolutionary scientists, this is far more devastating to those who would find necessity or inevitability or plan or guidance or God in evolutionary history.
All versions of theistic evolution, sophisticated or crude, are refuted by the inability of natural selection to predict. Or so I think.

Third, if you want to find a logical problem in arguments over evolutionary theory, consider the criticism that there are "gaps" in the fossil record. Defining a "gap" is as difficult as defining a "heap.
We could think of a gap as an inverted heap after all. The difficulty of defining a heap is even called a paradox! The creationists have for decades proffered a logical paradox as a criticism of legitimate science! Technically, I suppose, they could proffer simultaneously an empirical/practical definition of gap, and at least be making an honest argument. Until they do, just remember all arguments about gaps are logically flawed! (No, I am not touching the question of how useful formal logic is in doing science.)]