4 Integration of the Biometrical and Path-Analytic Approaches

These different strengths and weaknesses of the traditions derived from Fisher and Wright persisted into the 1970's. The biometrical genetical approach, derived from Fisher through the ground-breaking studies of Kenneth Mather and his student John Jinks established what became known as the ``Birmingham School.'' The emphasis of this tradition was on the detailed analysis of gene action through carefully designed and properly randomized breeding studies in experimental organisms. Except where the environment could be manipulated genetically (e.g., in the study of the environmental effects of the maternal genotype), the biometrical genetical approach treated the environment as a random variable. Even though the environment might sometimes be correlated between families as a result of practical limitations on randomization, it was independent of genotype. Thus, the Birmingham School's initial treatment of the environment in human studies allowed for the partition of environmental components of variance into contributions within families (EW) and between families (EB) but was very weak in its treatment of genotype-environment correlation. Some attempt to remedy this deficiency was offered by Eaves (1976a; 1976b) in his treatment of vertical cultural transmission and sibling interaction, but the value of these models was restricted by the assumption of random mating.

The rediscovery of path analysis in a series of papers by Morton and his coworkers in the early 70's showed how many of the more realistic notions of how environmental effects were transmitted, such as those suggested by Cavalli-Sforza and Feldman (1981), could be captured much better in path models than they could by the biometrical approach. However, these early path models assumed that assortative mating to be based on homogamy for the social determinants of the phenotype. Although the actual mechanism of assortment is a matter for empirical investigation, this strong assumption, being entirely different from the mechanisms proposed by Fisher, precluded an adequate fusion of the Fisher and Wright traditions.

A crucial step was achieved in 1978 and 1979 in a series of publications describing a more general path model by Cloninger, Rice, and Reich which integrated the path model for genetic and environmental effects with a Fisherian model for the consequences of assortment based on phenotype. Since then, the approach of path analysis has been accepted (even by the descendants of the Birmingham school) as a first strategy for analyzing family resemblance, and a number of different nuances of genetic and environmental transmission and mate selection have now been translated into path models. This does not mean that the method is without limitations in capturing non-additive effects of genes and environment, but it is virtually impossible today to conceive of a strategy for the analysis of a complex human trait that does not include path analysis among the battery of techniques to be considered.