A few days ago The New York Times had a blog post up which addressed the relationship between genes & environment in shaping our behavior & choices (see Genetic Future). One of the authors even posted a follow up comment where they evinced some surprise at the bile of the responses. I have to say that some people are naive; statistical sciences are a good reflection of the tenor of society. If you say a trait is 50% heritable, that is a statement of fact, but individuals will “spin it” however they want to based on their own outlook and the preferences of their target audience. Years ago Steven Pinker recounted to Robert Wright that when he states that a trait is 50% heritable he is often accused of being a genetic determinist, even though it is a logical implication of his assertion that 1/2 of the variation in the population is due to non-heritable factors. In fact a regular reader of this weblog labeled me as a “genetic determinist” years ago (on his deleted weblog, so I can’t link to the exchange) when I suggested that only 50% of variation in religiosity was due to environment (since he held that 100% was due to environment, he was of course an environmentalist).
There are obvious gross social and political reasons why people are uncomfortable with the idea that a large amount of psychological variation, and therefore a large portion of life outcomes which we implicitly use to measure the worth of a man, are due to variation in genes. Steven Pinker covers the topic in depth in The Blank Slate, so I won’t belabor it. But a more fundamental disquiet probably has to do with our intuition that Free Will exists, and that there is an animating ghost in the machine which has final executive control (even if you are an atheist who rejects the ghost in the machine intellectually, on a deep cognitive level you probably don’t). Of course models predicated only on environmental inputs can remove Free Will from the equation as well, so the issue is less about the heritability of a trait than the fact that one might be able to predict outcomes with a high degree of certitude.
These debates always hinge around the preconception of “Nature vs. Nurture.” This is rather like beginning a discussion of physics by sagely noting the theories propounded by Aristotle. In my 7 years of blogging it is obvious that it is almost impossible to erase from the minds of most that 50% heritability entails 50% of “it” is due to genetics and 50% of “it” is due to environment. 50% heritability simply means that one can adduce that 1/2 of the variation of the trait within the population is controlled by variation of genes, while the other 1/2 is not.* The standard intuition which humans begin with is simply not even wrong.
Terms like heritability only make sense in a specific environmental context. The reason that height is more heritable in the West than in Africa is not because of some different of genetic architecture, rather, the environmental component of variation (nutritional inputs) that is irrelevant in the West remains relevant in Africa. One could imagine that total phenotypic variance in a deprived Third World country would be greater (because of a larger number of individuals who are physically stunted), but the proportion of the variance that was genetic, and heritable, would be lower. But obviously this does not speak to any given person, we are addressing the level of a population (though population level analyses do say something about individuals at the end of the day). Even if someone is stunted because they are starved, a necessary precondition to their growth are a set of genes which express and trigger a sequence of development. Conversely, even in environments of nutritional plenty obviously growth is contingent upon eating food, even if the variation of that consumption is now no longer of any importance in prediction of the variation of the trait.
Though all this is straightforward, sometimes the concepts of gene-environment interaction and gene-environment correlation are not. The easy one first:
You see here two things:
1) The same genotypes in both environments
2) Which react differently to the two environments
If all genotypes react in the same way to an environment shift then there isn’t gene-environment interaction. If you can for example predict the quantitative position of a trait with respect to the mean value across as you shift the environment equally there isn’t an interaction parameter you need to worry about. As an example, imagine that a mean height for males is 68 inches with 3,000 calories per day with a normal distribution (+/- 2.5 standard deviation), and 60 inches with 1,000 calories (+/- 2.5 standard deviation), and all of the average genotypic values remain the same in terms of standard deviation units from the mean.
That is rather straightforward, a confused scramble, but an honest one. Gene-environment correlation is more interesting (to me) and perhaps more sinister. Gene-environment interaction can be thought of as a wild card, putting a cloak of mystery over ultimate outcomes due to the vicissitudes of the universe. It is in this way an aid to the idea of Free Will, as the variables which load the die are harder to tease apart. In contrast, gene-environment correlation exhibits a more regular pattern, tending to take a propensity and amplifying it. In other words where genes determine the direction of the vector the correlation can amplify the magnitude.
Imagine for example a child who is mildly above average in athletic ability and one who is mildly below. Assume that these differences, due to variation in eye-hand coordination, quickness, body fat percentage, etc., are greatly controlled by genes. Nevertheless the traits are only of modest difference in absolute terms during their early years. But as these individuals grow and mature, and select from a range of extracurricular activities, one would presume that they would seek those tasks to which they are suited by their ability to attain virtuosity and avoid those which they are not. Athletic endeavors are such that practice improves skills over time, and the cumulative effect is definitely significant. Environment, learning, practice, etc. does make a difference, but the choices one makes are often strongly conditioned by one’s genetic makeup, leading to a correlation which can multiply small initial differences and eventually lead to a yawning chasm.
So why is this sinister? From The New York Times:
The interaction between genetic tendencies and life experiences may explain another puzzling finding: the heritability of many psychological traits — from intelligence to anxiety — increases as people mature. This result seems odd at first glance, since genes are most important in brain development in babies and children. But children also have less control over their environment than adults. As people get older, they become more able to determine their own circumstances, and they may be able to choose environments that reinforce their natural personality tendencies. Apparently those of us who suspect we are turning into our parents as we get older may have a valid point.
One finding from behavior genetics is that adoptive parents can make a difference, but that the difference often diminishes a great deal after a child leaves the home (the extent of this varies by trait). This makes natural sense. Once the guardrails are gone one’s natural inclinations come to the fore. The sinister part is that this sort of dynamic makes a social engineer’s life rather difficult. To shift the life choices of a subculture one may have to apply coercive or incentive pressure indefinitely barring a change in the underlying genotype. In the most extreme cases of psycopaths and sociopaths punitive measures (specifically their threat) may be the only way society can change behavior or outcome, but behavioral economics and its allied fields tell us that the modal human also has a strong propensity for irrational or self-destructive behavior.
Utopian attempts to change behavior or eliminate social differences have often failed. Most recently in the Kibbutz movement, more generally in Utopian movements as a whole. On the other hand there have been some modest successes, e.g., the reduction in drinking in Scandinavia and the United States since the 19th century through a combination of regulation, taxation and moral suasion. Though there are limits, there are also possibilities.
Like biology, it seems likely that in the social sciences history has not ended. Our common biological heritage is certainly of interest, but individual differences are also important. We may pretend that our liberal order is predicated on some non-aggression principle where we do as we wish so long as there are no victims, but that is an obvious fiction, social norms remain strong, and some are enforced by legal sanction. The gene-environment correlation means that guardrails will be necessary, and that a particular segment of the population will enshrine their own dispositions into social norms and laws which shape the contours of those guardrails for the poor souls whose nature wars against the will of the deciders. Science may not only strip away our belief in Free Will, but may render naked the conceit that man is governed by a Social Contract contingent upon a universal human nature.
* I oversimplify here, but I would rather not get into narrow vs. broad sense, or whether epistatic variance is collapsed into environmental.