Simple threshold scene identification methods are developed to reduce the effects of errors in scene identification on the anisotropy of reflected and emitted radiances inferred from Earth Radiation Budget Experiment (ERBE) scanner observations. The ERBE maximum likelihood estimate (MLE) scene identification is assumed to be accurate for nadir fields of view. Various combinations of neighboring ERBE scanner fields of view at nadir are used to determine the population of cloud scene types as a function of field of view size. Longwave and shortwave thresholds are then determined for each of the ERBE solar zenith, satellite view zenith, and relative azimuth angular bins so that the population of cloud scene types at a particular satellite view zenith angle is consistent with the field of view size at the particular satellite view zenith angle. Differences between the anisotropy of reflected sunlight and emitted longwave radiation obtained using the new scene identification method and that obtained using the ERBE MLE method show that the ERBE radiative fluxes have satellite view zenith angle dependent biases. Thresholds are also developed for cloud scene identification with fields of view that are constructed to have a constant size with satellite view zenith angle. The angular dependence of reflected sunlight and emitted longwave radiation for scenes identified with these thresholds show little dependence on field of view size. This lack of dependence is a necessary condition for using scanning radiometer data to obtain radiative fluxes. ¿ American Geophysical Union 1996