Fraud in the Imputation of Fraud
The Mis-measure of Stephen Jay Gould
Posted October 4, 2012
Many of us theoretical biologists who knew Stephen Jay Gould personally thought he was something of an intellectual fraud because he had a talent for coining terms that promised more than they could deliver, while claiming exactly the opposite. One example should suffice—the notion of “punctuated equilibria”—which simply asserted that rates of (morphological) evolution were not constant, but varied over time, often with periods of long stasis interspersed with periods of rapid change. All of this was well known from the time of Darwin. The classic example were bats. They apparently evolved very quickly from small non-flying mammals (perhaps less than 20 million years) but then stayed relatively unchanged once they reached the bat phenotype we are all familiar with today (about 50 million years ago). Nothing very surprising here, intermediate forms were apt to be neither very good classic mammals, nor good flying ones either, so natural selection pushed them rapidly through the relevant evolutionary space.
But Steve wanted to turn this into something grander, a justification for replacing natural selection (favoring individual reproductive success) with something called species selection. Since one could easily imagine that there was rapid turnover of species during periods of intense selection and morphological change, one might expect species selection to be more intense, while during the rest of the equilibrium stabilizing selection would rule throughout. But rate of species turnover has nothing to do with the traits within species—only with the relative frequency of species showing these traits. As would prove usual, Steve missed the larger interesting science by embracing a self-serving fantasy. Species selection today is a small but interesting topic in evolutionary theory, not some grand principle emerging from paleontological patterns.
Recently something brand new has emerged about Steve that is astonishing. In his own empirical work attacking others for biased data analysis in the service of political ideology—it is he who is guilty of bias in service of political ideology (Jason Lewis and colleagues 2011). What is worse—and more shocking—is that Steve’s errors are very extensive and the bias very serious. A careful reanalysis shows that his target is unblemished while his own attack is biased in all the ways Gould attributes to his victim. His most celebrated book starts with a takedown of Samuel George Morton (The Mismeasure of Man ). Morton was a scientist in the early 19th Century who devoted himself to measuring the human cranium, especially the volume of the inside, a rough estimate of the size of the enclosed brain. He did so meticulously by pouring first seeds and then ball bearings into skulls until they were full and then pouring them out and measuring them. He was a pure empiricist. He knew brain size was an important variable but very little about the details (indeed, we do not know much more today). He thought his data would bear on whether we were one species or several, but in any case he was busy creating a vast trove of true and useful facts.
I love these people—they work for the future and gather data whose logic later generations will make real. Precisely because they have no axes to grind or hypotheses to prove, their data are apt to be more reliable than the first wave after a new theory. I have benefitted from them in my own life, most memorably when I was shown a large and accurate literature on ratios of investment in 20 ant species, gathered long before anyone appreciated why these facts might be of some interest.
In any case, Morton grouped his data by population according to best estimates of gross relatedness, Amerindians with Amerindians, Africans with Africans, Nordic Europeans with Nordic, and so on. It is here, Gould alleged, that all sorts of errors were made that supported preconceived notions that among the stupider (and therefore smaller cranial capacity) peoples would be Amerindians and Africans. For example, Gould claimed that Morton made more subgroups among Nordic people than tropical ones, thus permitting more of them to be above norm, but in fact, the opposite is true. Morton reported more Amerindian subsamples than European and routinely pointed out when particular Amerindian subsamples were as high or higher than the European mean, facts that Gould claimed Morton hid.
In other cases, Gould eliminates all samples with less than four individuals in order to reduce the number of sub-samples with only one sex—a statistically meaningless goal but one that happened to be biased in his favor and permitted him to make additional errors by arbitrarily eliminating some skulls while including others. If you are comparing group means, you may not wish to use means of less than four, but if you are adding up sub-samples to produce a larger sample, there is no reason not to aggregate all data. Morton is made to look careless and incorrect when it is really Steve who is arbitrarily biasing things in his own favor.
There is an additional contrast between Morton and Gould worth noting. To conjure up Morton’s mistakes, Gould lovingly describes the action of unconscious bias at work: “Morton, measuring by seed, picks up a threateningly large black skull, fills it lightly and gives a few desultory shakes. Next, he takes a distressingly small Caucasian skull, shakes hard, and pushes mightily at the foramen magnum with his thumb. It is easily done, without conscious motivation; expectation is a powerful guide to action.” Indeed it is, but careful re-measures show that Morton never made this particular mistake—only three skulls were mis-measured as being larger than they were and these were all either Amerindian or African.
The same can not be said of Gould. He came across distressingly objective data of Morton, and by introducing biased procedures (no sample size below four) he was able to get appropriately biased results. And by misrepresenting the frequency of Nordic vs Amerindian subpopulations, he was able to create an illusion of bias where none existed, by mere emphatic assertion (no one bothered to check).
Where are the unconscious processes at work here? Is Steve flying upside-down on auto-pilot, unconsciously looking for the actions (substitute Nordic for Tropical, delete all samples smaller than four) that will invite the results he wants (while hiding his bias)? Is the conscious organism really completely in the dark while all of this is going on? Hard to imagine—but at the end the organism appears to be in full self-deception mode—a blow-hard fraudulently imputing fraud, with righteous indignation, coupled with magnanimous forgiveness for the frailties of self-deception in others.
In response to the criticism of Lewis et al, the keeper of Gould's Tomb—his longtime editor at Natural History, Richard Milner—had some choice comments in defense of Gould (William 2012). Gould acted with "complete conviction and integrity" (that is, with full self-deception). "He was a tireless crusader against racism in any form" (In what way is misrepresenting the true facts about population differences—and then hiding this misrepresentation—a contribution to anti-racism?) And then, fully in flight, he says that any bias was "on the side of the angels". Who of us is in any position to say what is on the side of the angels? We barely know what is in our own self-interest.
A general point is that it is often very hard to draw the line between conscious and unconscious deception—or to define the precise mixture of the two. In The Folly of Fools (p 288) I discuss linguistic analysis by Hancock and colleagues (2010) suggesting that the architects of the U.S. 2003 war on Iraq were speaking deceptively when they warned that Saddam Hussein caused 9/11 and Iraq possessed WMDs. I naively suggest that this analysis shows conscious deception but I no longer agree with myself—unconscious deception could cause the same symptoms—reduced use of the word “I”, less qualifiers, and so on.
Lewis, J.E. et al. 2011. The mismeasure of science: Stephen Jay Gould versus Samuel George Morton on skulls and bias. PLoS Biol 9(6): e1001071. doi:10.1371/journal.pbio.1001071.
William, S. 2012. The Mismeasure of Stephen Jay Gould. Discover (Jan/Feb), pp 66-67.
Hancock, J. et al. 2010. Language, lies and politics: a linguistic analysis of the justifications for the Iraq war. Manuscript: Cornell University.