The performance of different models for the influence of enthalpy and polar effects on radical additions is compared for the extensive data set from the Fischer group, supplemented by additional data. The best correlations result from the Fischer–Radom (FR) model, but it also contains the largest number of adjustable radical‐dependent parameters not based on physical observables. Updating the literature values of $\Delta _f H$, IP, and EA that are inputs to the FR model led to some deterioration in the quality of the correlations; this is symptomatic of remaining deficiencies in the thermochemical databases. In contrast, the Lalevee–Allonas–Fouassier (LAF) model gives poorer correlation but this is in part compensated because it uses the same inputs but with no adjustable parameters. Hammett‐type models based on polar and radical substituent constants rather than on molecular properties of the reactants perform even more poorly. In all cases, poorer correlation, as judged by increasing sd(ΔE), is accompanied by a systematic bias to over‐predict the lower E values and under‐predict the higher ones. The enthalpy contribution in the FR and LAF models is expressed as a linear Evans–Polanyi dependence of E on ΔH. Replacement by nonlinear Marcus dependences does not significantly improve performance. An attempt to significantly reduce the number of adjustable parameters in the FR model by anchoring them to a base set applicable to all radicals, which is then modulated for spin delocalization based on observable ESR hyperfine constants in the initial and adduct radicals, showed modest success. Copyright © 2008 John Wiley & Sons, Ltd.