Verified by 
Spirituality
The Stork-and-Baby Trap
Mere association is not the same thing as evidence for a causal connection.
Posted July 12, 2013
An extract in Huff Post Books from Top Dog: The Science of Winning and Losing, by Po Bronson and Ashley Merryman, boldly claims that “new science has revealed that the size of a woman's ring finger can predict her entrepreneurship … wired that way from the fetal stages of development." Susan Guise Sheridan posted this in the magnificent BioAnthropology News with the plea “For the love of god, make this stop!" But assuming that correlation is the same thing as causation is a cherished, age-old tradition that is not going to stop any time soon.
Correlations in Statistics
Two things may seem to be correlated (associated) when there is actually no real connection between them. This is particularly likely with changes over time. In a delightfully clear discussion of statistical aspects, mathematician Vanessa Didelez provided an entertaining example. Over the last two centuries, bread prices in Britain and sea levels in Venice increased in tandem. This simply tells us that bread prices and sea levels both rose over time. Such obvious coincidences are easily discounted, but the fundamental distinction remains crucial: correlation is just a mutual relationship between two or more things, whereas causation is a relationship in which a particular action or event can be shown to be the direct consequence of another. Two seemingly associated things may depend on some common cause that has not been considered, a statistician’s confounding factor. Ideally, we need experiments to test for cause and effect, but that is rarely possible in human biology. Lacking experiments, we must interpret indirect evidence very cautiously.
Pioneering statistician George Udny Yule, author of the seminal 1911 textbook Introduction to the Theory of Statistics, explained confounding factors with a pleasing reference to reproduction. He noted that in Alsatian villages numbers of human newborns are correlated with numbers of storks nesting locally. It is tempting to conclude that storks do actually deliver babies, but the real explanation is far more mundane. Larger villages have more houses with chimneys for storks to build nests, and more babies are of course delivered in larger villages. The confounding factor is village size.
Confounded Conception
The field of human reproduction is littered by examples of association being equated with causation. A simple misconception (no pun intended) stems from the observation that women who do not menstruate do not conceive. This led to the early belief, prevalent until the 1930s, that there is a direct causal connection, with conception resulting when semen mingles with menstrual blood. Menstruation was believed to be the fertile time in a woman’s cycle. As a result, for decades women were advised to avoid copulation during menstruation and to treat mid-cycle as the “safe period”—the exact opposite of post-1930s advice.
Another, more subtle, example comes from a study showing that contraceptive sales and human conceptions both peaked in summer. A direct interpretation would be that contraceptives actually increase the probability of conception, but this is just another example of the principle that things changing over time can show coincidental patterns.
Virtually all human populations show seasonal birth patterns that often shift over time. Numerous investigators have linked such patterns directly to annual variation in environmental factors. Temperature and rainfall are popular candidates. But annual variation in temperature or rainfall in any given region remains broadly consistent across years, so these factors cannot explain shifting birth seasons. In fact, there is an entirely different possibility. Seasonal birth patterns may have developed over evolutionary time to match average annual variation in environmental factors. So birth seasonality may be driven by internal factors rather than ambient conditions.
Meaningful Correlations
Potential confounding factors must also be considered when assessing reports of correlations between breastfeeding and infant mental development. We know that babies of higher-income women score better in mental tests than their lower-income counterparts. Well-off women are also more likely to breastfeed. Accordingly, breastfeeding may be correlated with mental test results even without a causal connection. Appropriate statistical analyses can take confounding factors into account. By 1999, twenty studies meeting that criterion had been conducted, permitting clinicial nutritionist James Anderson and colleagues to carry out a sophisticated overall analysis. They carefully considered confounding factors, such as socio-economic status and maternal education. A significant advantage of breastfeeding remained. Breastfed infants tested at 6-24 months of age consistently showed higher levels of mental function than bottlefed babies. And larger differences were found for premature babies than for infants born at term. So benefits of breast milk for mental development are even greater for preemies. In this case, then, there is good reason to believe that those initial correlations did in fact correspond to an underlying causal connection.
Link to BioAnthropology News: https://www.facebook.com/groups/BioAnthNews/permalink/
References
Anderson, J.W., Johnstone, B.M. & Remley, D.T. (1999) Breast-feeding and cognitive development: a meta-analysis. American Journal of Clinical Nutrition 70:525-535.
Cowgill, U.M. (1969) The season of birth and its implications. Journal of Reproduction and Fertility, Supplement 6:89-103.
Cummings, D.R. (2010) Human birth seasonality and sunshine. American Journal of Human Biology 22:316-324.
Didelez, V. (2007) Statistical causality. pp. 115-120 in: Consilience: Interdisciplinary Communications 2005/2006. (ed. Østreng, W.). Oslo: Centre for Advanced Study.
Hartman, C.G. (1962) Science and the Safe Period: A Compendium of Human Reproduction. Baltimore: Williams & Wilkins Co.
Roenneberg, T. & Aschoff, J. (1990) Annual rhythm of human reproduction. II. Environmental correlations. Journal of Biological Rhythms 5:217-239.
