"Science is a way of trying not to fool yourself. The first principle is that you must not fool yourself, and you are the easiest person to fool" (Richard Feynman).
Here's a question some of my savvy students ask me: What might any scientist, perhaps one like A. J. Bateman, who did a foundational study in sexual selection, have done to save himself from his likely mistakes? That's a great question that goes right to the heart of what science is about.
Bateman might have used a control experiment to test if the assumptions of his methods were met. A simple set of observations would have told him and subsequent readers if his method was capable of answering his questions in a reliable way.
Bateman's method was seemingly very clever, even elegant. To infer who mated with whom in experimental populations, he used unique, phenotypically obvious, genetically coded "nametags" that offspring inherited from their parents. He assumed that offspring inherited the parental nametags in Mendelian frequencies. Look below at Sergio Castrezana's cartoon of Mendel's expectations when parents are each a dominant heterozygote at their unique nametag locus but wildtype at their mate's nametag locus (Bateman's method). Sergio's cartoon shows what Mendel, whose rules were discovered long before Bateman began his experiment, would have expected from any pair of Bateman's subjects, each of which was also a "heterozygote dominant" at a unique marker locus.
Remember that heterozygotes are individuals that have two different alleles (genes) at a given chromosomal (genetic) locus. If an individual is a "heterozygote dominant" at a given locus, the dominant allele is the only one that affects the individual's phenotype. Mendel's rules tell us that each one of Bateman's subjects produced two types of gametes, one with the nametag and one without, so that when two subjects mated and reproduced, they would produce four types of offspring, as Sergio Castrazana's caroon shows: One quarter would inherit a nametag from each parent and visibly express both parents' nametags so that the identity of their parents would be obvious; one quarter would get a nametag only from the mother so that observers would know who their mother was but not their father; one quarter would get a nametag only from the father so that the father's identity was obvious but not mother's; and the final one quarter would inherit only "wild type" alleles that code for normal looking flies, so that their parentage would be invisible to observers using this -- Bateman's -- method of parentage assignment.
Mendel's predictions of equal frequency of offspring phenotypes given that parents are heterozygote dominants each at a different and unique genetic locus. Image of acrylic and black ink on amate paper, with permission of artist Sergio Castrezana©.
To test the reliability of his assumption that his breeding scheme would produce offspring in Mendelian proportions, all Bateman had to do was put each type of subject two-by-two in a jar, only a single pair in "enforced monogamy," so that if those horny virgins were to mate at all, they had to mate only with each other. If offspring occurred in the expected frequencies (25 percent, 25 percent, 25 percent and 25 percent), Bateman would have known that his method was fair and would produce reliable results about the number of mates each subject had, and their number of offspring even when subjects were in populations with five males and five females.
Many investigators do monogamy trials when they wish to compare the effects of mating competition (sexual selection) between more than one male or more than one female with what happens when mating competition is impossible, as in monogamous pairs, each pair in a separate jar.
In the case of Bateman, the monogamy trials could have done double duty. If the trials showed that the method met the assumptions of Mendel about the relationship between parental genes and offspring phenotypes, one's confidence in the method would have been sealed. If Mendel's rules were violated, Bateman would have known to cut his losses, and he would have had options to ditch the planned experiment or start over with mutants that did not affect offspring survival. Moreover, it could then have provided additional data to compare with those from the experimental populations. Either way, enforced monogamy trials are a good bet, meaning they provide a high payout in confidence despite the work they involve.
My collaborators and I did enforced monogamy trials -- five for each of 25 combinations of male and female nametags that we had used in the populations like Bateman's originals.
In these monogamy trials, double-nametag offspring were missing, just as they had been in our Bateman-like populations. Proof: Bateman's method was biased. Double-nametag offspring died like flies even in the monogamy trials. Conclusion: Bateman's method was incapable of producing fair counts of the numbers of mates for each subject. It was interesting that even in the monogamy trials, fathers had more offspring than mothers, but here was proof that Bateman's method unfairly counted the number of offspring, assigning significantly fewer offspring to female subjects than to male subjects.
The monogamy trials proved that Bateman's method mismeasured both variables important to Bateman's conclusions: reproductive success and number of mates (the key variables of sexual selection). So much for "Bateman's principles." Had Bateman done monogamy trials as a control experiment, he would have known the folly of his methodology.
There seem few reasons to account for why investigators do not test the assumptions of their observations. Bateman surely knew about Mendel's laws, just as I did when I first read his paper over 30 years ago, just as most freshman biology students do today, and just as all our professors did, and yet we all overlooked the evidence in Bateman's own table 4 that the assumptions of his method were violated.
One of the big costs of our collective failures is lost time. Never mind for a moment the genuinely interesting facts that some studies in some species confirm Bateman's main conclusions. What about those other genuinely interesting studies that failed to show Bateman's predicted relationship and are facts, too? How many of these genuinely interesting facts were not published because they had a "negative" result that cut against belief? What about those investigators who just couldn't get their arms around their own observations of enhanced fitness for multiply mating females? Now I wonder if male reproductive success always increases with mate number. Hmm. And what about alternative theoretical studies showing that chance affects the number of mates for both sexes? In the past, only a very few investigators have bothered to test their results about variation in number of mates against probabalistic models. In the age of Nate Silver, many readers will realize that folly. We've lost time because we may have put too much faith in our beliefs rather than in the facts, just as some players in recent political races did.
So why did we fool ourselves? If Robert Trivers is correct about the evolutionary dynamics of deceit and self-deception, the payouts must be considerable. What's the benefit of self-deception in science? Does joining a bandwagon ease professional advancement or open doors to more rapid publication? There's a project for a historian of bandwagon science! Does gullibility discourage one from bucking a trend because it keeps one from backlash, such as failure to nab that rare post-doc or academic job? Historians of science should pick up this ball.