Moms, booze, and why social science is so damn hard

In the past year, I've had multiple social scientists tell me that people are the hardest thing to study. Sure, you don't need a Large Hadron Collider. And the chances of suddenly requiring a HAZMAT suit are pretty slim. But people almost never give you the kind of solidly reliable data you can get out of subatomic particles or viruses. The hard part isn't doing the research. The hard part is getting trustworthy, universal answers for anything. If you want to see a good example of those problems in action, check out this great piece on drinking during pregnancy, written by Melinda Moyer.


  1. “The hard part is getting trustworthy, universal answers for anything.”

    They pretty much gave up on that once they realized that people aren’t all the same.

  2. I’ll admit, I abstained while pregnant….the research is inconclusive, and I’d had enough miscarriages that I didn’t want to do anything that *might* cause a problem.

    Having said that….here’s my go-to example on the subject: Shakespeare’s mother drank alcohol every day of her pregnancy.  Clearly it isn’t as dangerous for fetuses as we think.  The devil is in the details.

    1. Interestingly, caffeine is apparently rather risky with respect to miscarriages, but the level of public concern about caffeine during pregnancy is pretty much nil – no public information campaigns at all, pregnant ladies can drink all the coffee they want and never encounter a dirty look…

      EDIT – not that I believe pregnant ladies should be subjected to dirty looks over what they choose to consume – it’s just that the same respect for their bodily integrity does not hold true when it comes to the choice to drink alcohol. I realize that may not have been entirely clear at first.

  3. If physicists had to give every Higgs boson $10 for an hour of its time whenever they wanted to study them, they’d find their budgets wouldn’t allow for very large sample sizes, either.    

  4. Interesting bias in the article – it notes that a number of studies have found modest correlation between light or moderate drinking during pregnancy, and better behaviour and higher intelligence – then basically explores every option other than the possibility that there might be a simple causation – the observed effect being caused by the observed behaviour.  I’ve seen that bias in several articles on the topic too.

    Another complexity: most studies assess maternal drinking through interviews, and pregnant women might lie about or underestimate their consumption out of embarrassment or shame. Lower estimates can mask harmful effects if they cause light drinkers to be incorrectly categorized as abstainers and put in the comparison group. They might, however, inflate the perceived risk if heavy-drinking mothers of children with deficits get incorrectly categorized as light drinkers.

    The same difficulty understanding maternal drinking could also cause beneficial effects of light maternal drinking to be underestimated, if some of the children with the benefits are incorrectly categorized as children of abstainers – but the authors completely avoid mentioning that possibility.  They only consider the possibility that children harmed by drink are getting put in the wrong category, and ignore the possibility that there might so much as exist a category of children benefitted by drink.

      1. I think you may be missing my point.

        I’m not remarking on the existence of the uncertainty, or on their pointing out the uncertainty – it’s that they only acknowledge half the possible outcomes of the uncertainty (misclassified harms), but not the other half (misclassified benefits), even though many of the results they look at suggest there may exist benefits.

Comments are closed.