25 Jan Why Marital Status is Heritable
I often feel as though I am swimming upstream when I make the case that behavioral genetics is not headed for a golden future in which the genetics of intelligence and schizophrenia are understood all the way down to the genome. Why not? They are heritable, and now that we have our hands on the DNA, we can identify more SNPs every year, and we can use those SNPs to build better and better GPS. Easy for me to be a skeptic now, people say, but just wait a few years when the GPS for IQ accounts for half the variance, and pathway models of the genomics of schizophrenia lead to early identification of risk and life-saving treatments. What will I say then?
It isn’t going to happen. Or, at the very least, it isn’t necessarily going to happen, you can’t count on the inevitable progress of science to get us from here to there. My argument (which, at the prompting of Charles Murray, I recently suggesting naming “soft genomics”) is that everything we know about the genetics of human behavior can be true– heritability, GWAS, GPS, GREML– without leading to either a golden age of scientific genomics or a dark age of eugenics, however you want to look at it.
As I have suggested many times before, consider marital status. Marital status, or divorce if you like to think about negative outcomes, is heritable. I will presume everyone agrees with me that if anyone bothered to try, large enough GWAS would produce significant SNP hits, which could be combined into functioning GPS, or GREML analyses, and whatever else. But somehow, the common intuition about intelligence– We’re just a few years away from mechanistic bottom-up neurogenetic understanding!– doesn’t extend naturally to marital status. Let’s think for a second about what kind of biological process might cause something as obviously social and psychological as marital status to be heritable.
A diagram would be useful here but I don’t have a good way to produce one, so bear with me. Imagine human development as a long, nearly infinite, branching pathway from genotype to phenotype. Starting with fertilization, cells start to divide, and right from the onset there are bifurcations that will eventually have implications for marital status. Maybe something can happen that has an effect on your body size, or your health, and as the embryo continues to develop, there are branches that have an effect on your neurological development, your temperament, your taste for alcohol, your physical attractiveness and your sexual inclinations. All that before birth. Once you are born you begin a relationship with your parents, and there are forks in the road relevant to attachment, sibling relationships, and things like that. Biological choice points keep occurring all the while down in the cells. Then you have your first peer relationships, first romance…. you get the idea. At each fork in the road you can either take the path more likely to lead to a lifetime stable relationship, or to D-I-V-O-R-C-E.
Here is the key point- at every fork in the road, from when you are a pinhead-sized bunch of cells until you are reacting to your first date, at every single choice point, your genome has its thumb on the scales, making it slightly more likely that you will develop in one direction than another. But equally important, it’s not the same genes having an effect at every choice point. The system doesn’t know that at the end of the game we are going to be checking your marital status. So the GPS, if we can imagine one, that is having a cellular effect on your facial symmetry and thus how good-looking you are going to be doesn’t have anything to with the relevant genome at the choice point that has an effect on your taste for alcohol or sexual orientation or how you are going to feel when your crush turns you down for a date. But some weighted function of your genotype is going to nudge you in one direction or another every time.
You would think that such a complicated process would completely wash out any identifiable effect of genotype by the time you reach the end of the story, but two things turn out to be true:
First, across the whole hyper-complex developmental pathway, people who start more genomically similar wind up, on average, more phenotypically similar, for marital status and everything else. Notice that this observation is independent of which genes nudged you in which direction at all they myriad choice points. Whatever the process, on average if you start out more similar you end more similar. And MZ twins, who start out genetically identical, wind up way more similar than anyone else for most phenotypes. Having complete genomic identity is the one thing that can reliably serve to overcome all that developmental complexity as you grow up.
Second, even though it’s a completely different set of genes pushing your probabilities in one direction or another at all the biological and psychological choice points, when all is said and done, with a big enough sample it is still possible to find a few SNPs are correlated with the phenotype across the vast space of the developmental network.
All of this produces the First Law of Behavior Genetics: Everything is heritable. Even things that have no “biological” edges at all, things that are just regions of multivariate phenotypic space, are heritable. And given enough sample size, things that are heritable will produce significant GWAS hits, and significant GWAS hits will produce functioning GPS. What are some consequences of this model?
- It is only in the weakest (softest!) possible sense that you can refer to these genes as “genes for” the phenotype, or as “causal variants” for a phenotype, because they don’t actually have anything to do with the phenotype, with divorce, per se. The are cellular level body size effects, or appearance effects, or impulsivity, or alcohol or ginger hair effects, and even listing them like that probably makes their effects much more discrete and concrete than they actually are. If I had to guess most of the genetic action is at the cellular level, so causally distant from how you feel after a fight with your spouse that there is no point in even talking about it.
- Heritability in and of itself has no particular relevance to questions about whether things like intelligence or divorce are biological as opposed to social constructions. Social constructions are heritable.
- There is no “divorce” to be found anywhere in the model. The twin heritability of divorce and the imagined GWAS and GPS are just sums over some hyper-complex pathway that, with large enough samples, preserves some of the probabilities of winding up in one region or another of phenotype space given a genotype. You wouldn’t say that the GPS is a measure of the genetic capacity or predisposition to fail at marriage.
- If all you want to do is to use a cotwin or a GPS to control for genetic pathways between divorced parents and outcomes in kids, none of this is a problem because you don’t have to care about the absence of a meaningful etiological story. All the genetic control is doing (and it’s a lot) is providing a probabilistic counterfactual to the outcomes of children from divorced families. How are they likely to have turned out if their parents had not been divorced? You don’t have to commit to an essentialist, hard-wired view of the genetics of marital status for such a comparison to be worthwhile
- You would expect replication to be very difficult for anything resulting from a model of this kind. Each fork in the developmental pathway between gene-action and marital failure introduces another contingency, another ginger child, in between the gene and the outcome. Yes, I know that some SNPs for intelligence have “replicated” in that they have accounted for a tiny but significant portion of the variance more than once, all within broadly European samples. Simple statistical replication is not the ultimate goal. The goal (coming back to PvS, if you think human genomics is going to produce biogenetic explanations of behavior) is to link the genetic variation with the outcome in a way that transcends all of the developmental contingencies. Otherwise all you are doing is noticing that good old-fashioned psychological development produces heritabilities, GWAS hits and GPS correlations.