The Mirage of a Space between Nature and Nurture Evelyn Fox Keller
The book is to the point and reasonably well written, but the arguments are not convincing. There are two broad categories of arguments that discount behavioral genetics results (such as heritability measures from twin studies): specific methodological criticism, including questioning of modelling assumptions, and also philosophical or semantic criticisms. This book avoids getting bogged down in the small details of how behavioral genetics research is done, and concentrates on the philosophy. She does also deploy a lot of general purpose skeptical arguments, philosophical nukes that can be used to indefinitely prolong any debate.
The book mainly follows the model of philosophical argument, relying on thought experiments (intuition pumps) to create the desired understanding in the reader. Some of these, such as the height, width and area of a rectangle, will be well-known to people who follow the nature/nurture debate, but others were new to me. For example, we are told of two drummers that we hear drumming in the distance. The author argues that it is meaningless to ask whether anything about what we hear might be caused by the drum or the drummer in isolation, as opposed to the drum-and-drummer system. Read the book for how she sets up this intuition, but let me suggest an alternate intuition pump:
You are a drummer, and you are standing next to another drummer (not off in the distance). You are both drumming away and having a great time, but you start to notice that the other drum-and-drummer system happens to sound a lot better. You start to wonder if that is because you are beating on a plastic garbage can with a rolled-up newspaper, whereas he has precision-turned maple drumsticks and this really slick carbon fiber drum with a mylar head. You think: “What would happen if we switched drums?”
Such questions are the stuff of science, and also the stuff of ordinary everyday trying-to-get-by-a-little-better. Philosophy has shown us that there is no irrefutable argument from obvious principles that we could ever reliably find out the truth about anything, or that if we did grab onto one corner of a truth, there is no reason to suppose that this will remain true tomorrow or in another country.
And yet… Even before the rise of science, people have indeed discovered useful generalizations about the world, and many useful causal relationships. This in spite of the fact that causality, as an abstract philosophical concept, has proved quite resistant to a definitive definition. A powerful way to discover causal relationships is to decompose a system into sub-parts that have relatively low coupling between them, and then to try to manipulate the connections between those sub-parts.
So, the drum can be decoupled from the drummer. We switch drums, and what do we find? I don't know–you'd have to do the experiment. Surely you'd find an immediate change in your sound, but you'd have to learn a whole new technique in order to accommodate. In the end, you might decide you were a just a plastic garbage can kind of guy. That's the whole point. If you decompose a system and manipulate connections you can learn particular things that happen to be reasonably true. Since both the decoupling and the manipulation may be imperfect whatever causal generalizations we make are only true some of the time, under particular contingent circumstances which we can never precisely pin down.
One criticism that she does make of behavioral genetics methods is that it relies on statistics, and in particular, on population averages and variation about those averages. On first reading, I thought this was rather audacious and arbitrary, taking on the whole field of statistics in order to win a point. Nowadays almost all of science (especially social science) is strongly dependent on statistical methods, and this dependence extends well beyond, into economics, business planning and manufacturing. I won't try to explain here why statistics has been effective across such a wide range of human activity. Let me just assert that for many purposes, where we deal with things as an aggregate, statistics does discover useful trends and causes of variation.
But is it so arbitrary? This questioning of the use of statistics in genetics? Something I did not know when I read the book is that statistics was invented to do genetics. All of the major figures in the development of frequentist statistics (including the methods of Factor analysis, Linear regression, ANOVA and Hypothesis testing) during the critical period of about 1850 to 1930 were either geneticists or were working with geneticists and strongly interested in genetic applications. See the development of modern statistics and click through to the bios of Francis Galton, Karl Pearson, Egon Pearson, Jerzy Neyman and Ronald Fisher. Does this connection prove anything? It does prove that the application of statistics to genetics is not a coincidence, not merely yet another ill-conceived misuse. Of course it could be that (of all the places statistics is currently used), genetics just happens to be the most deceptive and ill-advised. [End of historical digression.]
After making the usual point that genetics causing variation is not a “real cause” (a cause for existence or sufficient cause, see Genetic Causes), she makes the reasonably obvious claim that statistical means and trends don't apply to any particular individual in any straightforward way. Then she says that, since statistics don't tell us anything about individuals, statistics can only tell us about groups. She doesn't directly make the accusation, but leaves the argument dangling in a way that makes quite clear her implication that anyone who has any interest in such statistics must be motivated by racism, sexism, or some other forbidden group comparison.
In my view, these philosophical arguments are a smoke-screen. It is possible that the resulting fog is an unintended consequence of the dispassionate search for philosophical clarity. Yet the effect is to make the big picture of heredity seem more complicated than it is turning out to be. One stubborn fact is that in pretty much all populations studied, for almost all measurable behavioral traits studied, identical twins are much more similar than fraternal twins (or ordinary siblings). If genetics did not matter, then there would be no difference between these two groups.
I understand that the author is not attempting to argue that genetics do not matter, but rather that genetic causes are so intertwined with other causes that it is unreasonable to suppose that it might be possible to make a separation between the two. There are subtleties of gene/gene and gene/environment interactions that could conceivably make gene/environment separation meaningless or impossible, but behavioral geneticists do know this, and argue that the effects they measure are largely additive, so can be separated. Why it should be true that genetic effects should on average behave additively is a scientific puzzle, since we also have evidence that at the micro-scale, within an organism, strong cross-coupling can exist.
The political heart of the nature/nurture dispute, which the author gets to at the end, is to what degree the sorts of differences between people that our culture considers important are shaped by conceivably controllable things, such as how parents behave, what resources they have, what teachers try to teach, and so on, and to what degree those differences may be innate, arising either from their genetic heritage, or from random fluctuation in developmental processes. For example, fingerprint patterns are clearly innate, but identical twins do not have the same fingerprints. While the number of fingerprint ridges is highly heritable, the patterns themselves are not.
I do not believe that current behavioral genetics bears in any convincing way on the usual policy debates about to what degree we should redistribute wealth or encourage people to spend time in school. But it isn't true that twin studies tell us nothing about the effect of environmental interventions. The component called “shared family environment” tells us about the typical effect of the typical variation in everything that a family shares, including family income, culture, schools and community. If, across the study population, there are rich families, poor families, good schools, bad schools, and so on, and these differences matter, then the shared environment should show a strong effect. In heritability studies it is rare for the shared environment to show more than 10% effect, in comparison to a genetic effect that tends to range from 30% up to 70% or more. You can challenge the assumptions behind shared family effect measurement, but it does claim to tell us about the typical effect of common environment variations. There is also the usual Problem of induction, that it is hard to be certain whether the trends seen in the particular people studied will still hold when applied to a larger group of people, such as the entire population of a country.
What the shared environment estimate cannot tell us is what the effect might be of interventions that currently do not exist or are rare. Population genetics also cannot tell anything about effects that currently hardly vary at all (that everyone experiences), and this may be the largest opening toward future social progress.
More philosophically, nature/nurture discussion cannot avoid considering that individual differences may arise from those individuals' free choices and the social consequences of those choices. We may hope that suitable education, incentives, consciousness raising, public information campaigns, family support, re-education, etc. can lead people to behave in constructive and pro-social ways, but they will surely remain persistently individual, and have their own ideas too. (see Determinism vs. Free Will).
— Rob MacLachlan 2014/05/16 13:32