Wednesday, October 10, 2007

Every Conceivable Thing Counts As Evidence For, No Conceivable Thing Counts As Evidence Against

Good critique.


[T]he final sentence of the paper:

The characterization of copy number variation among humans and between humans and other primates promises considerable insight into our evolutionary history.

So are these evolutionary inferences, as distinct from the actual data on comparative genetics of salivary amylases, good science? To answer this question, I ask Dr. Packer to permit me a little speculation. Imagine that the data on human and ape amylase gene copy number and variation were different from the data compiled by the authors. Of course the data could take any of a number of permutations — humans could have had fewer copies of salivary enzymes than apes, or less copy number variation, etc. What inferences could be drawn?

What if humans had low salivary amylase gene copy number, and apes had high copy number? How could the authors invoke evolutionary theory to explain these observations? The authors could assert that human vulnerability (our relative inability to digest tubers) led to a need for interdependence and socialization to ensure survival, and gave rise to human cooperation, altruism, language, and civilization. This is of course explained by natural selection. The press release: ‘We Were Weak, So Now We’re Strong: Evolutionary scientists report groundbreaking research on how an evolutionary vulnerability led to the emergence of humans...’

What if humans had low intraspecies salivary amylase copy variation, and apes had high intraspecies variation? No problem. The evolutionary theory: low human intraspecies copy variation is evidence for strong evolutionary conservation — providing further evidence that human ability to digest tubers provided enhanced energy for evolutionary adaptation or that human vulnerability (inability to digest tubers) generated a need for interdependence, gave rise to cooperation, brain growth, altruism, language, and civilization. Natural selection is consistent with either of the evolutionary hypotheses! The press release: ‘Old Genes Are Good Genes!: Scientists make groundbreaking discovery linking surfeit (or lack) of critical enzymes in saliva to human evolution and world domination...’

What if humans and apes both had high salivary amylase gene copy numbers? The evolutionary explanation: abundant salivary amylase was essential for human and ape evolution, because it allowed digestion of energy-rich tubers, thereby facilitating human and ape brain growth and giving rise to intraspecies cooperation and altruism. More evidence for natural selection! The press release: ‘Scientists Discover What Makes All Primates’ Mouths Water…”

What if humans and apes both had low salivary amylase copy numbers? The evolutionary explanation: paucity of salivary amylase is essential for human and ape evolution — lack of ability to digest energy-rich tubers caused need for interdependence gave rise to ape cooperation, and to human brain evolution, altruism, language, civilization. Clear evidence for natural selection! Press release: Hard Times Make Good Species: Could world domination have begun in the cheeks…”

Perry et al.’s inference to natural selection is irrelevant to the specific data in their paper. ‘Natural selection’ could be invoked for any permutation of their data. Yet in science, inferences must depend on data, and must be subject to falsification by data. Inferences that are independent of data, such as the inference that amylase gene copy number and variation are explained by natural selection, aren’t scientific inferences at all, because the inference to selection could be drawn from any data on the comparative genetics of salivary enzymes. Of course, individual stories as to how the adaptation arose can be adjusted to fit data, but the fundamental inference to ‘natural selection’ is untested.

At the core of Darwin’s theory of evolution are two hypotheses: heritable variation arose randomly, without teleology, and individuals that were rendered more reproductively successful by heritable variation were more reproductively successful. When applied to several-million-year-old genes for salivary enzymes, the first hypothesis — that heritable variation arose without teleology — is untested, and the second inference — that reproductively successful individuals are reproductively successful — is a tautology. The inference to ‘evolution’ in the authors’ paper is an inference to the untested and to the tautological. The authors would have us believe that their inference to evolution is cutting-edge science. Yet the synthesis of ‘untested’ and ‘tautological’ isn’t science at all.

The authors would no doubt protest that some aspects of selection can be tested, and Dr. Packer points out that ‘robust’ statistical methods were applied to questions of positive selection, negative selection, and neutral drift. Yet what do these terms mean? Positive selection means the heritable variation helped, negative selection means it hurt, and neutral drift means it didn’t matter. Any heritable variation — non-teleological or teleological — would meet one of these three criteria, so how do the ‘robust’ statistical methods provide evidence for natural selection acting on random variation? One could apply the same ‘robust’ statistical methods to genetic engineering — some inserted genes improve a function, some hinder it, and some don’t matter. Yet one could not draw the inference that Darwin’s theory was at work — on the contrary, the variation was intelligently designed. Robust statistics don’t redeem unfounded inferences.

No comments: