Sunday, May 20, 2012

Part 2: "How Do Phenotypes Get To Be How They Get To Be? (or) Is Natural Selection Biology's Phlogiston?"

I submitted the previous blog post to Bjørn Østman's Carnival of Evolution. My submission did not please Mr. Østman, and he expressed his displeasure through comments posted on this blog. His comments were similar to those that the post received on www.thescienceforum.com. Both sources of commentary accused me of mischaracterizing the theory of natural selection. But unlike those on the forum, the comments of Mr. Østman mostly were informed and thoughtful.
Mr. Østman explained that I had failed to distinguish between natural selection and genetic drift. These processes participate in distributing genes, but they are distinct mechanisms, according to the normal view. He offered some details about how researchers distinguish the twothat is, about how they decide when to attribute the distribution of a gene (or allele) or a phenotypic trait primarily to drift or to selection. Despite his clarifications, and I appreciate the time he took, the criteria apparently typically used to distinguish the two mechanisms seem not really to be able to do so. The argument:
Two mechanisms are proposed as the primary causal trendsetters in gene (allele) and trait distribution. These mechanisms are genetic drift, a random process, and  natural selection, a nonrandom process. The latter explains the advance and retreat of traits in populations in terms of adaptation and fitness. The former explains the ebb and flow of traits in terms of chance. Given shifting trait frequencies through generations (in a local population of a given species), how does one distinguish outcomes due to natural selection from outcomes due to genetic drift?

Apparently one does so by determining whether a particular genotype or phenotype is more fit than a competing genotype or phenotype (in the local population of the given species). One does this by comparing the number of fertile progeny left behind by individuals with the one genotype or phenotype with the number of fertile progeny left behind by individuals with the other genotype or phenotype. The ratio is a measure of fitness of the types relative to one another. Phenotypes with similar traits also can be grouped, and averages taken.

Genotype/phenotype variants associated with a small fitness advantage (slightly better reproductive success) will tend to spread more rapidly in a small population but more slowly in a large population, due to greater opportunity for random mixing, or drift, in the large population. So, if a spreading gene or trait is associated with even a small fitness advantage, then the spread of the gene or trait in a small population can be taken to be a case natural selection. But in a large population, the advantage must be large for selection to be given credit for the spread of a gene or trait.

In other words, when the ratio of the fitness advantage to the population size crosses a threshold, then natural selection gets credit for the spread of the gene or trait, otherwise the spread is attributed to drift, or some other mechanism.

Whatever the merits of such a threshold-crossing formula, it cannot distinguish among mechanisms. Like the difference between a sluggish economy and a recession, or between the outbreak of a disease and an epidemic, or between a planet and a Pluto, the distinction between genetic drift and natural selection rests on an arbitrary threshold. It's a taxonomic distinction based on a variance from a statistical baseline. It's just a convenient way to label distribution ranges.

The differential labeling in the case of drift vs. selection is supposed to imply that a qualitative distinction is being used to distinguish two mechanisms, but nothing about the formula or its application entails that there be a qualitative distinction between or among proposed mechanisms. A difference of mechanism is implied, but there's no there there. In short, the formulas of population genetics can distinguish statistically among outcome distributions and assign labels to various distribution ranges, but that's about it.

To claim that such a formula justifies a theory of how phenotypes get to be how they get to be due to adaptation and fitness—that the formula informs us as to when changes in phenotypes are due to selection or to drift—really does beg the question, because it assumes that, among differing genotypes and phenotypes, differences in reproductive success (when they cross a statistical threshold) are DUE TO the differences among the genotypes and phenotypes of previous generations, specifically due to their relative adaptiveness. But to draw causal inferences from statistical results is some kind of logical fallacy.

I can't even say that it's assuming causation from correlation, because there's no correlation. We can't say that certain outcomes are correlated with drift and other outcomes are correlated with selection and that therefore each proposed mechanism causes the correlated outcome. The problem is that the item in question—natural selection—is taken from the outset as being a mechanism available for explaining gene distributions. Wouldn't one have to establish first that it's actually available, before one moves on to the question of which criteria distinguish it from other mechanisms? My argument is that it's not there; it's not available, unless it's just another name for the observation that some creatures enjoy more reproductive success than others, which we know is the primary cause of gene and trait distributions.

So, if natural selection cannot be the primary mechanism of macroevolution, what might be? Unlike various mechanisms that have been proposed, such as the ongoing and contingent interplay of endogenous and environmental variables called natural history ("just one damn thing after another"—this solution is proposed by Fodor and Piattelli-Palmarini in "What Darwin Got Wrong", from which I appropriated the phrase, how phenotypes get to be how they get to be), or the deliberate intervention of alien intelligences, or one of the other proposals listed HERE, the star larvae hypothesis proposes a mechanism that not only agrees with the latest discoveries in genetics, epigenetics, and the growing sense among researchers that environmental influences should be downplayed and endogenous factors emphasized in accounting for phenotypes (this is the gist of the papers collected in "Evolution, The Extended Synthesis"), but also accounts for evolution's apparent directionality.

The star larvae hypothesis proposes that evolution comprises stages of a developmental life cycle.