Remember the joke about the drunk man searching unsuccessfully for his keys under the streetlight?
After a few minutes the policeman asks if he is sure he lost them here, and the drunk replies, no, and that he lost them in the park. The policeman asks why he is searching here, and the drunk replies, “this is where the light is.”
We see this behavior all to often in science. Researchers interrogate the data that they have to answer a specific question when the answer is nowhere to be found in that data. Yesterday’s piece in the NYTimes, Does Breast Milk Have a Gender Bias?, illustrates both the “drunkard’s search” as well as the shoddy nature of most breastfeeding research.
This is a “drunkard’s search.” The authors used the data they had instead of the data that is required.
According to Nancy Segal and Satoshi Kanazawa:
In 1973, the biologist Robert Trivers and the computer scientist Dan Willard made a striking prediction about parents and their offspring. According to the principles of evolutionary theory, they argued, the male-to-female ratio of offspring should not be 50-50 (as chance would dictate), but rather should vary as a function of how good (or bad) the conditions are in which the parents find themselves…
In short: If things are good, you have more boys, and give them more stuff. If things are bad, you have more girls, and give more of your stuff to them.
Is this hypothesis correct? In new research of ours, to be published in the April issue of The Journal of Experimental Child Psychology, we suggest that in the case of breast-feeding, at least, it appears to be.
The first red flag is the journal. What is a paper on breastmilk composition doing in The Journal of Experimental Child Psychology when the issue under investigation has nothing to do with child psychology? Almost certainly because the authors couldn’t get it into one of the relevant journals.
The second, glaring red flag is the nature of the study itself. The authors were trying to discern if breastmilk produced for male babies has a different composition or amount than breastmilk produced for female babies. Obviously, the only way to determine that is to compare the composition and volume of the two types of breastmilk.
That’s where the principle of the drunkard’s search comes in. The authors are psychologists, not chemists and they can’t investigate the actual composition of breastmilk. They do have a giant data base of information on adolescents, so they decide to look there.
They offer a Rube Goldberg version of reasoning to explain what they did.
The aim of our research was to figure out how to study the divergent effects of breast milk on sons and daughters on a much larger scale. Our solution: to look at twins.
In light of the emerging evidence of sex-specific variation in breast milk quality, the Trivers-Willard hypothesis implies that breast-feeding may be less beneficial for opposite-sex twins than for same-sex twins. Same-sex twins, after all, can benefit from their mother’s sex-tailored breast milk just as non-twins can. However, the breast milk of mothers of opposite-sex twins cannot be tailored for either sex. Perhaps it is selectively tailored for the wrong sex part of the time, or for neither sex all of the time.
Thus, the Trivers-Willard hypothesis led us to surmise that opposite-sex twins would be disadvantaged for health and growth. This would be reflected in their being shorter and weighing less than same-sex twins.
What??!!
There are so many unwarranted assumptions here that it’s hard to keep track.
Unwarranted assumption #1:Assuming gender specific breastmilk confers benefits. This is a version of the white hat bias that plagues all breastfeeding research: if it’s in breastmilk it must provide a benefit. But that doesn’t have to be true. If breastmilk plays a role in regulating gender composition of a population isn’t doesn’t necessarily have to provide gender specific benefits. It’s only necessary to provide benefits to one gender.
Unwarranted assumption #2: Assuming that looking at adolescent height and weight should tell us whether breastmilk is gender specific. This assumption is just bizarre. If the point of gender specific variations in breastmilk is to regulate the gender composition of the population, we would expect to see the differences in infancy and childhood, not adolescence.
Unwarranted assumption #3: Assuming that the benefits of gender specific breastmilk would be expressed in physical characteristics? Why? The point is supposed to be a survival advantage for one gender over another. What would make height and weight the relevant variables to look at. Wouldn’t survival rates be far more important?
Unwarranted assumption #4: Assuming that it makes any sense at all to look at anything besides the chemical composition and volume of breastmilk to determine whether there are gender specific differences in breastmilk. The authors undertook a drunkard’s search of the data in their possession instead of searching for the answer in the only place it can possibly be: in the composition and volume of breastmilk itself.
And if all that isn’t bad enough, the authors looked at only one twin in each pair, breastfeeding history depended on maternal recall more than a decade after the fact, and zygosity of same sex twins was determined by the mother’s impression instead of genetic testing.
What did they find?
The Wave I sample included 779 twins for whom the sex of their co-twin could be determined; of these, 546 respondents (277 females and 269 males) had same-sex twins and 233 respondents (120 females and 113 males) had opposite-sex twins…
The results showed that, among the ever-breastfed twins, same-sex twins were marginally significantly taller, F(1, 176) = 3.784, p = .053, significantly heavier, F(1, 170) = 5.400, p = .021, and significantly larger (either taller or heavier), F(1, 165) = 5.900, p = .016, than opposite-sex twins. In sharp contrast, among never-breastfed twins, same-sex/opposite-sex twin status was not significantly associated with height, F(1, 275) = 0.228, p = .633, weight, F(1, 266) = 0.417, p = .519, or body size, F(1, 259) = 0.798, p = .373.
The authors’ reasoning is extremely tortured:
The analysis of the Add Health data showed that breastfed same-sex twins were consistently taller and heavier than their opposite-sex counterparts throughout adolescence and early adulthood (from Age 16 to 29). In contrast, never-breastfed same-sex twins were no taller or heavier, and might even have tended to be shorter and lighter, than their opposite-sex counterparts. Our results are consistent with the recent findings that human breast milk may be tailored for each sex to facilitate its health and growth most efficiently. Same-sex twins can benefit from such sex-tailored breast milk, just as singletons can, but opposite-sex twins cannot do so and, therefore, may be at a disadvantage.
That reasoning would only apply if all four of their unwarranted assumptions were true and there’s no evidence that any of them are.
This is shoddy research, not surprising when you consider that it is a drunkard’s search. The authors used the data they had instead of the data that is required. The authors presume to draw conclusions about gender specificity of breastmilk without looking at breastmilk and without looking at infant and child outcomes. There is simply no reason for believing that adolescent height and weight tell us anything about breastmilk and it is irresponsible for the authors to imoly that there is.
