A method for the detection of segregating major genes based on the analysis of estimated marginal posterior major gene variance density was examined. The properties of the method were investigated using data sets simulated for a real population of laying hens consisting of eleven generations. Marginal posterior densities of model parameters were estimated by the Gibbs sampling approach proposed by Janss et al. (1995). With the data of about 4000 observations it was possible to detect a major gene responsible for one third of the genetic variance and one tenth of the phenotypic variance, irrespectively of the degree of dominance at the major locus. The inference based on the posterior marginal major gene variance can be sensitive to skewness of the data. It was shown that skewness of 0.2 can lead to a false detection of a major gene. The method is robust against a non-genetic mixture of normal distributions.