In a previous post I discussed the following article, which shows that in the case of extreme poverty (by US standards) the genetic heritability of intelligence is drastically reduced. It is the first study I had heard of which really showed a clear case of nonlinear response to environment (excluding cases of severe malnutrition). See related discussion of heritability and regression here.
Turkheimer, E., Haley, A., D'Onofrio, B., Waldron, M & Gottesman, I. (2003). Socioeconomic status modifies heritability of IQ in young children. Psychological Science, 14, 623-628.
In this paper he discusses interactions between genes and environment, and why they make social science hard:
Turkheimer, E. (2004). Spinach and Ice Cream: Why Social Science Is So Difficult. In. L. DiLalla (Ed). Behavior genetics principles: Perspectives in development, personality, and psychopathology. (pp. 161-189). Washington, DC, US: American Psychological Association.
Recently, he gave the following talk at Stanford, emphasizing how the problems faced by state of the art genomic science (e.g., genome wide association or gwa studies) mirror those of social science. That is, outcomes depend nonlinearly on a large number of (possibly correlated) causes. This is the "Gloomy Prospect" first referred to by psychologist Robert Plomin. I highly recommend the audio version -- Turkheimer is a good speaker and the discussion at the end is interesting.
Gloomy Prospect Wins
slides (ppt) , iTunes audio
The contemporary era has seen a convergence of genomic technology and traditional social scientific concerns with complex human individual differences. Rather than finally turning social science into a replicable hard-scientific enterprise, genomics has gotten bogged down in the long-standing frustrations of social science. A recent report of an extensive genome wide association study of human height demonstrates the profound difficulties of explaining uncontrolled human variation at a genomic level. The statistical technologies that have been brought to bear on the problem of genomic association are simply modifications of similar methods that have been used by social scientists for decades, with little success. The motivation for the statistical methods in genomics is the same as it is in traditional social science: An attempt to discern linear causation in complex systems when experimental control is not possible.
In the talk Turkheimer gives the following definition of social science, which emphasizes why it is hard:
Social science is the attempt to explain the causes of complex human behavior when:
There are a large number of potential causes.
The potential causes are non-independent.
Randomized experimentation is not possible.
He proposes that genomics will also be hard for similar reasons. Final slide:
The question is not whether there are correlations to be found between individual genes and complex behavior— of course there are—but instead whether there are domains of genetic causation in which the gloomy prospect does not prevail, allowing the little bits of correlational evidence to cohere into replicable and cumulative genetic models of development. My own prediction is that such domains will prove rare indeed, and that the likelihood of discovering them will be inversely related to the complexity of the behavior under study.