Abstract
A multivariate quantitative genetic model is analyzed that is based on the assumption that the genetic variation at a locus j primarily influences an underlying physiological variable yj, while influence on the genotypic values is determined by a kind of "developmental function" which is not changed by mutations at this locus. Assuming additivity among loci the developmental function becomes a linear transformation of the underlying variables y onto the genotypic values x, x = By. In this way the pleiotropic effects become constrained by the structure of the B-matrix. The equilibrium variance under mutation-stabilizing selection balance in infinite and finite populations is derived by using the house of cards approximation. The results are compared to the predictions given by M. Turelli in 1985 for pleiotropic two-character models. It is shown that the B-matrix model gives the same results as Turelli's five-allele model, suggesting that the crucial factor determining the equilibrium variance in multivariate models with pleiotropy is the assumption about constraints on the pleiotropic effects, and not the number of alleles as proposed by Turelli. Finally it is shown that under Gaussian stabilizing selection the structure of the B-matrix has effectively no influence on the mean equilibrium fitness of an infinite population. Hence the B-matrix and consequently to some extent also the structure of the genetic correlation matrix is an almost neutral character. The consequences for the evolution of genetic covariance matrices are discussed.