A general approach to estimating the sampling variance of the correlation coefficients estimated in genetic experiments from the analyses of variance and covariance is outlined. This is an extension of a method proposed by Tallis (Aust. Jour. Stat. 1959). A general solution for the sampling variance of genetic, environmental and phenotypic correlation coefficients from the one-way design and from sire components in the nested design is given in detail. It is exphasized that the use of this type sampling error for tests of significance assuming a normal distribution are not necessarily appropriate. This is so since the distribution of [image]. is unknown.