eight observations on “biology” and social science

By this point, I am a longtime observer/participant in the effort to integrate genetic data into social science. So Andy’s post stirred up various convictions that I have. Not so much directed at Andy, but at lines of argument suggested by his post that are familiar to me by now. So, eight responses:

1. Add Health deserves enormous credit for its pioneering role in making genotypic data available to investigators. Nevertheless: the existing body of published candidate gene or candidate gene-environment studies based on Add Health mostly follows a logic that we now know has resulted in voluminous false positives in medical genetics. I say: if it didn’t work there, why would it work here? The studies typically employ no power analyses and are only significant because of effect sizes larger than anyone in the age of GWAS has reason to believe are actually plausible. I could go on about particulars of the voting study–nothing personal against Fowler or Dawes or any of the other smart people who’ve worked in this area–but suffice it to say that I am extremely skeptical.

If interested, check out this meta-analysis for an idea of how one might remain gloomy about the entirety of candidate gene-environment studies in psychiatric genetics. Also, consider various social science studies in light of this policy of the journal Behavior Genetics, borne of exasperation with false positives, that requires replication before publication.

2. Work on quantitative behavioral genetics in political science (e.g., twin stuff) has become more sophisticated than what is being done in sociology. This is partly due to some social network issues that are too inside baseball for me to go into, but there is also a basic disciplinary disadvantage: in political science, anything that causes a political outcome is in the jurisdiction of political science, so their literature has spent a fair amount of time contemplating the meaning and import of genetic “main effects.” Whatever else, sociology regards genetic main effects on any outcome as “not sociology,” unless the purpose is debunking them, and any positive discussion has to be couched in terms of gene-environment interaction/correlation.

3. The twin study work in political science is substantially more credible than the candidate gene work in political science. Puzzle to ponder: as far as I can tell, in political science, some of the people who involved in twin study work are highly productive folks who have access to molecular genetic data, and yet they don’t seem to be cranking out candidate gene findings of their own. Hmm.

4. Heritability estimates are commonly misinterpreted and I have made a variety of statements in print to that effect (e.g., here, here, here). It’s a big leap from that to saying that heritability estimation is useless, especially since that kind of argument is often a prelude to people making arguments that seem like they miss the point of heritability estimation or want to sweep it under the rug (the point being to link genetic variation to outcome variation). Shared human capacities (e.g., having two eyes) are entirely beside the point–the point is accounting for variation.

5. My problem with “All complex behaviors are 100% genetic and 100% environmental” is that it trivializes what are some very important questions about human development and the intergenerational reproduction of behavior. Does one think that differences in behavior between persons are 100% genetic? Eep, then: what about differences in behavior between groups? There’s a large educational gradient in voting, is that 100% genetic? Is it genetic at all? Or is the claim just that everybody who votes has a human genome and is not, say, a dog?

I mean, I believe very strongly that all genetic causation of behavioral outcomes occurs within socially and historically contingent fields. Indeed, I think greater appreciation of that would be one of the best contributions that sociology could make to behavioral genetics. But there actually is much at stake in the question of how much and how exactly genetic variation in a population is causally related to outcomes like educational attainment or voting.

6. One variety of sociological complaint goes that social reality is incredibly complicated and that “biological” explanations are “simplistic.” And yet, sometimes the same people who make this argument also make arguments about “social” causes that are very simple. Indeed, there is a whole line of sociological argument that contends that reality is quite simple and that “biology” distracts us from that fact. For example, check out Troy Duster’s discussion of alcoholism on page 3 of his ASA Presidential Address.

7. Sociologists often have a dual consciousness where, if pressed, they acknowledge that the influence of genetic variation in some outcome is “nonzero,” but otherwise they will talk about the cause of the outcome in every way as though the influence of genetic variation is, in fact, zero.

This is most irksome with respect to methodology. There are entire literatures in sociology predicated on premises like, if the number of books in the parents’ house is associated with the academic performance of children, then that’s all you need to draw vast conclusions about environmental causality. Sure, the conclusions could well be true, but anybody who thinks that sort of evidence suffices is either a zero/nonzero dualist or a causal inference pushover. Don’t pretend otherwise.

8. What does it mean to be anti-science? Does it mean to wish ultimately that inquiry would not proceed on a given topic, or to cheer for work on a topic to be so mired by complexity as to fail to make any progress? If so, then, yes, I do think there is a portion of sociological commentary on the integration of genetics into social science that is anti-science.

Or, maybe it’s better to say that some sociologists are pro-mystery. Set folks in a room with two buttons. Press the red button, and various issues about the underlying causes of human behavioral variation remain equally as mysterious thirty years hence as they are now. Press the green button, and all these standing puzzles about why you have substantial heritabilities for things like voting and educational attainment and the like are resolved, although nobody can control in advance exactly what the answers are (because that’s why one goes to the trouble of doing science). I don’t think one should underestimate the strength of the red button contingent in sociology.

Author: jeremy

I am the Ethel and John Lindgren Professor of Sociology and a Faculty Fellow in the Institute for Policy Research at Northwestern University.

11 thoughts on “eight observations on “biology” and social science”

  1. Great. Now there is going to be a whole “red button” campaign to raise awareness for the saving of mystery in sociology. Thanks a lot, Jeremy.


  2. I agree with most of the points here. My point is not to argue that the *concept* of heritability is useless, but rather that the twin model fails to capture a theoretically robust concept of heritability because variation is chained/sequenced and therefore not truly partitionable. My example here explains why, but here’s another version for the specifically political-science crowd. Voting could be simply GxE: a gene “for” prosocial behavior and the presence of an election. It could be GxGxE: a gene for prosocial behavior, a gene for misunderstanding the likelihood of one’s vote mattering, and the presence of an election. Or, it could be the result of a gene for prosocial behavior, a gene for *understanding* the likelihood of making a difference, a cultural disposition not to care about whether it makes a difference, and the presence of an election (GxGxExE, if you’re keeping score, and the second G just changed direction from the previous version). In each of these cases, we might find main effects using twin studies, but we would know little or nothing about what these actually mean.

    My contention, given this claim about the underlying mechanics, is that the confidence interval around a twin-study-based heritability estimate for a complex behavioral trait should be considered very, very wide. In general, a significant heritability estimate can be interpreted as meaning that true heritability is nonzero, but probably not much more than that. I suspect that that’s *why* the twin-study work in poli sci hasn’t yielded much in the way of convincing candidate-gene studies: there’s not actually much heritability to explain, and too many candidate genes to explain it with!

    I don’t see myself as “cheering” for the science to be too complex to interpret, but rather as urging caution in matching empirical methods to underlying ontological claims.

    Finally: of course our field is full of plenty of bad science that isn’t biosocial in character, and plenty that is anti-bio-social. But that’s no reason to excuse bad science that is biosocial in character!


  3. Thank you Jeremy for posting this incredibly thoughtful post. I wonder how much of sociologists’ reactions (I guess I should say “our” reactions) result from a reaction against eugenic arguments that throws the baby (useful genetic research) out with the bathwater (morally repugnant eugenics research). Given this not-too-distant history, I think it is incumbent on researchers and editors is also to be sure that results are not sensationalized (e.g., “Genes predict voting” — while technically true, somewhat misleading with respect to the magnitude of influences of various causes). Beyond the moral issues implied by a modern-day eugenics (which none of the authors argue, but is present in the world — see: Murray, Charles), the real potential consequence of such sensationalized reports of findings is for the public and Congress to simply look for genetic answers is a serious problem if those headlines are taken at face value.

    In the meantime Red Button for GOP Nominee 2012!


    1. I think “racist” is probably the more accurate word than “eugenics.” Or at least that would characterize my own worries better. I think eugenics is only one part of how genetic arguments have been used in the service of racial injustice.

      My own personal conviction is that good science about the genetics of behavior and traits will result in conclusions that strengthen a broad anti-racist position. So then, for me, the problems are more about worrying about bad science and, as you highlight, bad communication about the character of the science to the public.

      Another part of this is that genetics is a vast world and stuff is coming whether sociologists participate or not and, for that matter, whether Americans participate or not. My understanding is that the largest site of actual sequencing right now is in Shenzhen. Another major vendor is in Essex. The most important stuff on GWAS methods as they might relate to social science has been out of Brisbane. The most important gene-environment study for psychiatric genetics is based in Dunedin. Maybe the most important medical genetic data is out of Reykjavik. I would like to think that if American social science ends up having any impact on how this plays out, it will be to the good.


      1. Thanks, Jeremy! This makes a lot of sense and I agree. I wish that I understood more about this, because it sounds really, really interesting.

        I also think American sociologists could do a lot of good explaining the social construction of race to medical doctors, many of whom end up (either through assumption or training) thinking of race as a biological factor. I think that pointing out where genetics certainly does *not* help our understanding of racial differences, if I interpret what you are saying correctly, could certainly help with that.


  4. Amen Jeremy. Can we think of some predictions unambiguous enough that we can make bets (for real money) with the red-button crowd?


  5. (because that’s why one goes to the trouble of doing science) because we are so scattered out and that time continues to march on, isn’t social science actually a good thing ? . Also as the globe continues to spiral smaller, doesn’t social science apply and contribute to this? We expanded outwards only to contract inwards.


Leave a Reply

Please log in using one of these methods to post your comment:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s