The ability to predict reliable thermochemical propel-ties of molecules and ions has led to an ever increasing application of ab initio molecular orbital theory. Methods such as G2 theory have been shown to generally give accurate heats of formation (Delta(j)H) for closed-shell molecules and ions. Open-shell systems have been less thoroughly examined to date and the present paper attempts to redress this situation through a detailed assessment of the performance of a variety of levels of theory in calculating Delta(j)H values for foe radicals. Representatives of three families of theoretical procedures have been studied: the infinite basis set extrapolation techniques of Martin, the CBS procedures of Petersson et al., and the G2 methods of Pople et al. Among the specific influences investigated are choice of geometry, zero-point vibrational energy, high level electron correlation treatment and basis set size. We recommend a new procedure called CBS-RAD for the treatment of free radicals. CBS-RAD is a modification of the CBS-Q method in which the geometry and zero-point energies are obtained at the QCISD/6-31G(d) level, and coupled-cluster theory is used in place of quadratic configuration interaction in single-point energy calculations. We find that for free radicals with low spin contamination G2 theory also performs adequately, but as (S-2) increases the results of 02 calculations can become increasingly unsatisfactory. The recommended CBS-RAD procedure should yield more reliable results over a broader range of free radicals.