« Who's Zoomin' Who? | Gene Expression Front Page | NYC plans »
June 04, 2003


DNA testing shows that many children are not the biological offspring of their supposed fathers. The incidence of misascribed paternity in present-day western populations has often been alleged to be over 5%.

This has implications for estimates of heritability. These usually depend on a comparison of observed correlations for some trait (height, IQ, etc.) with those predicted by a genetic model. If genetic inheritance is purely additive, and there is no assortative mating, then the correlations between relatives should reflect the average proportion of genes they share by inheritance from a recent common ancestor: 50% between individual parents and offspring, 50% between full siblings, 25% between half siblings, and so on.

If there are significant levels of cuckoldry in a population, the proportion of genes shared by relatives (except those related solely through the female line) will on average be reduced. If the rate of cuckoldry is k, the average proportion of shared genes attributable to a particular line of relationship will be diluted by a factor of (1-k) for each step at which cuckoldry may have broken the link (assuming that the probability of cuckoldry in each case is independent, and that the same male is never the true father of more than one offspring of cuckoldry.) In the case of full siblings, the line of relationship through their mother is unaffected, giving them an expected correlation of .5 x .5 = .25 from this source, but the line through their father presents two opportunities for cuckoldry. The expected correlation will therefore be .25 + .25(1-k)(1-k) instead of .5. If k is 10% this gives a correlation of just over .45 instead of .5. Other cases, such as cousins, are more complicated, and each line of ancestry needs to be carefully traced.

Twins are a special case. Monozygotic twins always have the same true father. Dizygotic twins could in principle have different fathers, but this must be comparatively rare. The standard ‘twin method’ for estimating heritability is therefore not seriously affected.

The most obvious implication of cuckoldry is that we would expect correlations between offspring and their mothers to be higher than between offspring and their supposed fathers. The evidence on this is mixed. It has generally been accepted that maternal and paternal correlations are equal. This is notably the case for IQ, which has been intensively studied [ref. 1]. However, it has also been claimed that in some psychological traits mother-offspring correlations are slightly higher than father-offspring correlations, which has been ascribed to the greater role of women in child rearing [ref. 2] This overlooks the possibility of cuckoldry. Another implication is that the correlation of ordinary siblings would be lower than that of DZ twins. This has been sometimes been observed for IQ, but this has been ascribed to the greater similarity in upbringing of twins.

It may be doubted whether the incidence of misascribed paternity is really as high as often alleged. High rates have not yet been confirmed in published peer-reviewed studies. Some ten years ago a survey of the literature [ref. 5] showed a wide range of different rates in different communities. For example, in 1957 J. H. Edwards made an estimate of 5% in the UK based on blood groups. In 1963 a study in Michigan showed a rate of 1.4% for whites but over 10% for blacks. A more recent study [ref. 4] in France has found a figure of 2.8%. Over the longer term, Bryan Sykes has concluded from Y chromosome evidence that in England since the late Middle Ages cuckoldry did not average more than about 1% per generation [ref. 6]. It seems likely that the rate of cuckoldry in European and American populations in recent decades has averaged somewhere between 2% and 10%, but with much variation.

Even with a rate as high as 10%, the departure from predicted correlations would usually be small enough to fall within the range of uncertainty due to assortative mating and other complications. However, this would not apply to those cases where misascribed paternity is actually identified from DNA testing. In these cases the predicted correlation based on assumed relationships would diverge widely from those based on true genetic relationship. In the case of father-offspring, the true genetic correlation would be zero, except to the extent that assortative mating correlation between husband and wife, or between husband and true biological father, raised the expectation somewhat. I presume however that this would not raise it above, say, .2. In fact, the offspring of cuckoldry can be regarded as a natural experiment in adoption, with the difference that there is no problem of ‘selective placement’, and no ‘Pygmalion effect’, since the cuckolded fathers do not even know that the experiment is taking place. Admittedly, cases of misascribed paternity are not a random sample of the population, but they are probably closer to a random sample than many other categories studied in behaviour genetics.

All of this should be fairly obvious, but I have not seen it discussed in the literature on kinship correlations (e.g. refs. 1, 3) If anyone knows of relevant discussions that I have missed, I would be interested to know.


[1] Bouchard, T. and McGue, M. (1981). Familial studies of intelligence: a review. Science, vol. 212, 1055-1059.
[2] Boyd, R. and Richerson, P. J. (1988). Culture and the Evolutionary Process. Chicago: Chicago University Press.
[3] B. Devlin et. al. (1997) The heritability of IQ. Nature, 3 July 1997, vol. 388, 468-71.
[4] Le Roux, M-G. et al. (1992). Lancet, vol. 340, 607.
[5] Macintyre, S. and Sooman, A. (1991). Lancet, vol. 338, 869.
[6] Sykes, B. and Irven, C. (2000). Surnames and the Y Chromosome.
American Journal of Human Genetics, vol. 66, 1417-1419.

Posted by David B at 09:41 AM

One additional study from Mexico showed an inverse relationship between class and cuckoldry rates. The upper class had very high certainty of paternity and the lower class had very low certainty of paternity. Not surprisingly, there tends to be a correlation between certainty of paternity and level of paternal investment in his purported offspring. In cultures with very low rates of certainty of paternity (below 37.5%, I believe) it can make sense from a Hamilton/Trivers point of view for the mother's brother rather than the official husband to play the paternal role in a child's life.

Posted by: Steve Sailer at June 4, 2003 06:20 PM