It is shown that the observations of Gyorgy, Sturge, and Van Uitert, on the anisotropy of Y3−xLuxFe5O12, are in excellent agreement with the mechanism suggested by Akselrad and Callen. The data are consistent with a tetrahedral iron mechanism, although they do not exclude an octahedral iron mechanism. The relative contributions of the two sublattices and the relevant physical considerations are discussed. The mechanism of preferential site ordering is also discussed further, with particular reference to relative site sizes.