A clarification is given of the differences in approaches to the estimation of F-statistics of Nei & Chesser [Ann. Hum. Genet. (1983) 47, 253–259)] and Cockerham [Genetics (1973) 74, 679–700]. The principal difference is that Nei & Chesser define quantities with respect to fixed extant populations, while Cockerham allows for evolutionary variation between populations. Weighted and unweighted analyses are compared, and a numerical example given.