The co-author of IQ and the Wealth of Nations writes:
I think you have overestimated the effect of "regression toward the mean." The effects are quite small, as shown empirically in the Terman study of genius in which the IQs of the children were only about 5 points lower than those of the parents(133 vs 138, respectively). Theoretically, IQ is believed to be mostly determined by additive genes. If it were entirely determined by additive genes the average IQ of the children would be the same as that of the average of the parents. Thus Jodie Foster with her IQ of say 100 mated with a sperm donor with an IQ of 160 should have children with an average IQ of 130. Jodie Foster with her IQ of say 100 mated with Mr Average with an IQ of 100 should have children with an average IQ of 100. Regression occurs through non additive genes and environmental effects.
But wasn't the Terman study restricted to students who scored 130 and higher? William Shockley and another Nobel Prize winner famously missed the cut by a couple of points. Or perhaps I'm misinterpreting the point? Anyway, think of it this way, if your brother is 7 feet tall, how tall do you expect to be? Not 7 feet.
UPDATE: Greg Cochran fills in some details on the Terman study of highly intelligent children that began back before WWII. The average IQ of children in the study (the "Termites") was 151. The average IQ of their eventual spouses was 126. Thus, the average of the married couples was 138. The average IQ of their children was 133. This would suggest, crudely, a narrow sense heritability (parent to child) of 33/38 or 0.86, which would be fairly high even for a broad sense heritability (between identical twins). Other estimates of narrow sense heritability I've seen have been between 0.34 and 0.5, which would bring about more regression toward the mean.
Anybody have other studies on narrow-sense heritability?