We investigate the accuracy with which the radiative forcing of a halocarbon can be estimated using a computationally efficient narrow band radiation scheme with 10 cm(-1) resolution by performing a detailed comparison with line-by-line calculations. We use CF4 (perfluoromethane) as an example, as it has unusually sharp absorption features and its absorption is more heavily overlapped by other atmospheric gases than most other halocarbons; hence it is likely to be one of the more difficult gases to assess accurately. We show that the narrow band model can reproduce the line-by-line forcing to within 10%. Given that absorption cross-sections reported in the literature rarely agree to better than this, such an agreement is deemed acceptable at present. However, degradation of the resolution of the radiation calculations beyond 10 cm(-1) can lead to rapidly growing errors. We also show that recent changes in the catalogues of spectral line parameters for the overlapping species (water vapour, methane and nitrous oxide) has led to a 5% decrease in the forcing calculated for CF4. The applicability of both the weak limit and the diffusivity factor are also investigated. Copyright (C) 1996 Elsevier Science Ltd