According to the purely hereditary position with respect to the origin of differences between individuals, the correlation between the intelligence levels of single parent and offspring should at least be 0–50 when perfectly reliable tests are used. This is supposed to hold when there is random mating. In the case of assortative mating the correlation rises theoretically to 0–71. From a psychometric and theoretical point of view, parent and child should be tested at the same age with the same test. Data of 2847 father-son pairs meeting these requirements show that the correlation is 0–33. A further analysis of the data indicates that the distribution of the sons' intelligence and attainment is only weakly determined by hereditary and environmental variables. Path analyses show that almost all variance of the sons' intelligence test scores remains ‘unexplained’.