One often observes a familial resemblance of risk factors of disease. The following question then arises: How much familial aggregation of cases of disease would be expected because of this resemblance? This problem is attacked through a particular model where the risk is supposed to depend exponentially on the risk factors. Only pairs of relatives (father/son) are considered. The calculations are performed both with normally distributed risk factors and with particular skewed distributions. An application to coronary heart disease is given.