Analysis and critique of the concept of Natural Selection (and of the neoDarwinian theory of evolution)

Part 2. Theory’s failures to account for organic evolution

Previously published only on the author’s website. (Update March, 2006). Part 1 is published as a separate article, “Role as part of Modernism’s origination myth”.

By Stanley N. Salthe, Professor Emeritus, Brooklyn College of the City University of New York.

The concept of evolution is often conflated with the Darwinian theory widely used to explain it--we hear phrases like “natural evolution.” The following remarks apply strictly to the neoDarwinian theory.

(1) Basically, a theory of “higgledy-piggledy”
(2) Incorporates no theory of origins
(3) Fails to explain evolutionary improvement of phenotypic characters and behaviors
(4) Fails to model the evolution of more than a single phenotypic trait
(5) Internal contradiction in its major theoretical cornerstone -- Fisher’s fundamental theorem
(6) Able to explain only differences between characteristics of genealogically closely related types
(7) Unable to explain similarities between organisms and ecological systems that are not related by descent

The neoDarwinian (Synthetic) theory of organic evolution has not, after some 60 years of development, delivered a very convincing mechanism to explain the evolution of organisms (as opposed to the evolution of gene systems). It cannot explain origins, or the actual presence of forms and behaviors. It can generally explain only the evolution of adaptive differences as results of historical contingency, for only one or two traits at a time. It is limited to historical explanations, as it acknowledges no evolutionary tendencies that are not the result of accident preserved in genetic information. … I think it could be said that, were there another theory of organic evolution, the neoDarwinian one, fraught with problems as it is, would have more trouble surviving than it does. As it is, it is the “only game in town”, largely because of the competitive activities of the neoDarwinians themselves.

The following is a much-abbreviated summary by the publisher making the relevant points made in the original article more accessible to the general reader. Full transcript of the original article as a pdf file

(1) Natural selection is a mechanism consisting of accidental changes.
NeoDarwinians claim evolution is the outcome of randomness being subject to certain constraints. Take the randomness first. Can we be sure their identification of randomness reflects a true indeterminacy in the system, and not blindness on their part to some kind of order? No, we can’t.

Now the constraints, can they be identified in advance? No, they’re also somewhat hypothetical. They’re assumed to be applied to this randomness by the environment, but they usually are not identified until the outcome points to their involvement. What evolves is just what happened to happen. From what that was we deduce what in the environment could have been responsible.

A theory resting on randomness channeled by nonspecifiable constraints resists confirmation. Because it consists of accidental changes, the theory behind natural selection cannot be verified.

(2) NeoDarwinians have no adequate theory of the origin of species.
For Darwin, the origin of species was just the gradual transformation of organisms in a given population as a result of selection over a long period of time. He assumed that a naturalist examining specimens from the original population and some from the latest would judge them to be from different species. He proposed no more of a mechanism than that.

Turning Darwinism into population genetics theory does not introduce any fundamental change. It is fundamentally mathematical. In mathematics nothing new can be generated, except by way of error. The selection regime itself is just a negative, mechanical, process of culling. In the resulting creation story we are, as George Wald quipped, the products of editing, not of writing. No significant change can be created through the operation of such a process. Such a process could only generate new species if that did not involve the creation of significant change.

(3) Natural selection cannot account for evolutionary improvement in phenotypic characters and behaviors.
The Twentieth Century saw a decline in the idea of progress, including progress in evolution. Some examples of progress in evolution were recognized to have been constructed by tracing later developments backwards and then telling the story forwards as an illustration of how evolution works. “The phrase ‘the evolution of this or that,’ so common in museum displays, became oxymoronic. For example, the putative selection processes that left certain dinosaurs with feathers could not be assimilated to a story of the evolution of flight, or of the evolution of wings, except as a post hoc view from the present.”

“Natural selection can be directly demonstrated in laboratory and field experiments, and has many times been shown indirectly to (most likely) have been occurring in nature, but its connection to long term evolution is an inference only.” There are two major theoretical prongs in neoDarwinism: the Fisherian dynamical approach and the Wright-Dobzhansky kinetic approach. Neither, the author concludes, delivers real long term evolution.

(4) The theory fails to model the evolution of more than a single phenotypic trait at a time.
The Fisherian approach considered the spread through a population of only a single trait. J.B.S. Haldane noted that when the fitnesses of independent traits were considered together, this so rapidly increased the cost of natural selection in terms of survival that one could not imagine the simultaneous evolution in a single population of more than one or two phenotypic traits at a time.

How about remarkable feats being reported concerning the use of selection in computation to design shapes, robots and products using the likes of ‘genetic algorithms’?

The claim here is that, given a complex shape coded for by several to many “genes”, a selection process can be instituted to improve any function entraining that shape. The resulting shape changes are not predictable (not built into the program to begin with), nor is the trajectory taken during the improvement. In other words, this models a multigene selection process. However, it does not escape Haldane’s dilemma, because there is only one function being selected at a time -- one selection pressure. Perhaps two functions might be optimized simultaneously, given a large enough population of robots. But this is not like selection among organisms, where only fitness is maximized, not any particular function (except, perhaps rarely, in some catastrophe). Another disanalogy can be seen when we note that much, if not most, genetic information in organisms is pleiotropic (one gene involved in the definition of multiple traits). This means that not just any old change that will improve some function can be selected in organisms without consequence for other functions. The selection model seems to work better in genetic algorithms than it could in organisms!

(5) The internal contradiction in its major theoretical cornerstone -- Fisher’s fundamental theorem.
According to Fisher’s theorem, trait variation in a population gets sacrificed over the generations in exchange for a rise in overall fitness. As a result the traits most affected, because of their importance in the lives of the organisms, will end up being less variable than less important traits. So traits that have been most important in the lives of organisms up to this moment will be least likely to be able to evolve further! Hardly a recipe for continued adaptation.

(6) The theory can account for differences only between genealogically closely related types.
When we consider such distantly related forms as snails and wolves, we find ourselves considering differences, in form and behavior for example, that need not reflect differences in particular genes.  Form and behavior cannot be attributed to genes because these phenomena are the products of physical activity. And it is genetic configurations, rather than material processes, that are considered to be the locus of inheritable information. We may indeed find similar genetic forms among very distantly related organisms, as in the HOX genes, yet not infrequently these similar genes perform different functions (albeit in the same general system) in these different forms.

(7) Failure to account for similarities between organisms and ecological systems that are not related by descent.
Darwinian models are designed to account for modifications in the course of descent, but they’re less able to account for convergent similarities such as similarities in the eyes of cephalopods and vertebrates. Other examples of convergent evolution are chameleons and sea horses. “Both, of similar size, are slowly creeping, bushwhacking predators. Both have independently movable eyes and prehensile tails used to hold onto foliage. Both are camouflaged. Using different genetic materials and working from completely unlike ancestors, it is as though they have been poured into the same mold. . . . There have been some preliminary attempts to locate such similarities as results of a wholesale transference of genes, by way of viral infection, from one kind of organism to another. This would fit with neoDarwinian views, and we can expect this idea to be exploited for whatever it may be worth. In the meantime, convergent evolution tends not to get mentioned at all in important texts.”

Postscript: As added support for the viewpoint projected herein, I cite two of Richard Lewontin’s works. First, his The Genetic Basis of Evolutionary Change, (1974, Columbia University Press) has a discussion of the effects of linkage disequilibrium among genetic loci on the process of selection that makes it seem highly unlikely that selection could be very effective in improving a trait using a more realistic model of the genome than is usually used. Recently he has produced a paper for the Santa Fe Bulletin [Volume 18 (1), Winter, 2003] which raises four “complications” to the theory of natural selection that seem to me to cripple it altogether. For the mathematically inclined, there is also Fred Hoyle’s work, posthumously published as Mathematics of Evolution , Acorn Enterprises, 1999