The biometrical genetical method of analysis of continuously varying human traits is described and shown to unify a number of common approaches to the analysis of twin data. The method, which is based on the analysis of variance, goes beyond alternative approaches by providing tests of basic assumptions and by investigating the possibility of assortative mating and dominant gene action.

Twins form an indispensable group for the application of the biometrical genetical method but the value of including other groups, often more easily collected, is demonstrated.

The techniques are illustrated through the reanalysis of data on the behavioural measures, neuroticism and intelligence quotient (I.Q.). For both measures fundamental assumptions are met. The results suggest the absence of dominant gene action for neuroticism, indicating an evolutionary history of stabilising selection, and directional dominance for high I.Q., indicating an evolutionary history of strong directional selection. Both these findings seem reasonable in view of psychological knowledge concerning the two measures.