A new approach to constructing angular distribution models (ADMs) from satellite data has been developed. The ADMs model the radiance anisotropy and are used to convert satellite measured radiance to flux at the top of the atmosphere. The radiance pairs method (RPM) processes radiance pairs that view approximately the same area at the same time. By ratioing the paired radiances, the flux or field strength is eliminated, producing ratios of anisotropies which are taken as the data source for the ADMs. The ADMs are modeled as random functions, and the RPM estimates the mean of the ADM. The RPM is compared to the standard method of sorting by angular bins (SAB) and is shown to remove questionable assumptions, converge faster, and give better accuracy than the SAB method. Both methods were applied to the same Nimbus 7 ERB data and resulted in statistically different longwave ADMs. ¿ American Geophysical Union 1996