In the past, the Callen-Callen (1965 Phys. Rev. 139 A455-71; J. Phys. Chem. Solids 27 1271-85) model has been highly successful in explaining the origin and temperature dependence of the magneto-crystalline anisotropy in many magnetic compounds. Yet, despite their high ordering temperatures of ∼650 K, the Callen-Callen model has proved insufficient for the REFe 2 compounds. In this paper, we show that it is possible to replicate the values of the phenomenological parameters K 1 , K 2 , and K 3 given by Atzmony and Dariel (1976 Phys. Rev. B 13 4006-14), by extending the CallenCallen model to second order in H CF . In particular, explanations are provided for (i) the unexpected changes in sign of K 1 and K 2 in HoFe 2 and DyFe 2 , respectively, and (ii) the origin and behaviour of the K 3 term. In addition, it is demonstrated that higher order terms are required,and that K 4 exceeds K 3 at low temperatures. Revised estimates of K 1 , K 2 , K 3 , K 4 , and K 5 are given. Finally, an alternative 'multipolar' approach to the problem of magnetic anisotropy is also provided. It is shown that the latter confers significant advantages over the older phenomenological method. In particular, all the multipolar coefficients (K N , N = 4, 6, 8, 10, 12) decrease monotonically with increasing temperature, withK N decreasing faster thanK N −2 etc. These observations are in accord with expectations based on the original Callen-Callen model.