In order to estimate top-of-atmosphere (TOA) radiative fluxes from satellite-observed radiances, angular distribution models (ADMs) that account for the angular variation in the Earth's radiation field are needed. This study introduces new ADMs for determining emitted longwave (LW) and 8-12 micron window (WN) TOA fluxes from the Clouds and Earth's Radiant Energy System (CERES) instrument. The ADMs are constructed using 8 months of coincident CERES broadband and Visible Infrared Scanner (VIRS) high-resolution spectral radiances from the Tropical Rainfall Measuring Mission (TRMM) satellite. To optimize angular sampling, CERES measurements are acquired by scanning in alongtrack and rotating azimuth plane (RAP) modes. In RAP mode, the instrument scans in elevation as it rotates in azimuth, thereby optimizing sampling in both viewing zenith angle and relative azimuth angle. The angular models are determined directly from the measurements by stratifying CERES radiances by scene type. Scene types are defined by meteorological and imager-derived cloud parameters that have strong influence on LW and WN radiance anisotropy. These include surface type, precipitable water, cloud emissivity, cloud amount, lapse rate, and surface/cloud temperature differences. The anisotropic factors generally decrease with viewing zenith angle (limb-darkening) in both LW and WN regions. The limb-darkening is more pronounced in the WN region due to the stronger sensitivity to surface temperatures (e.g. day/night differences over desert). Overall, the CERES radiance anisotropy varies smoothly with meteorological and VIRS-derived parameters. ADMs for overcast and broken cloud fields are shown to be more sensitive to changes in cloud emissivity, while cloud-free scenes are most sensitive to atmospheric lapse rate. The most pronounced changes in LW and WN anisotropy occur for thin cirrus clouds with variations in cloud emissivity.