A methodology is developed to apply a parameterization of radiative transfer calculations to satellite analyses of cirrus clouds. Cloud heights and optical depths are derived from visible and infrared window measurements taken during the First ISCCP (International Satellite Cloud Climatology Project) Regional Experiment (FIRE) when cirrus clouds were present. Geostationary satellite retrievals are compared to lidar-derived cloud heights and retrievals from a polar-orbiting satellite taken at different angles to determine which theoretical models of scattering phase function and single-scattering albedo best represent actual cirrus clouds. Models using small hexagonal ice crystals with a diameter of 20 μm (C20) and a size distribution of slightly larger hexagonal ice crystals representing a cirrostratus (CS) cloud produce the best results. The resulting mean cloud heights are within ±0.3 km of the lidar results and have instantaneous uncertainties of ±1.3 km. Mean cloud heights derived using a model based on water droplets with a 10-μm effective radius (ID) and a model based on a distribution of large ice crystals (cirrus uncinus, CU) are 1.3 km less than the lidar heights. The cloud height biases are due to overestimates of the cloud optical depths that are as much as 1.7 times greater than the C20 values. Reflectance patterns computed with the ice crystal models are consistent with the dual-satellite, multiangle observations of optical depth. Of the three ice-crystal models, the C20 model produced the least bias (3%), while the CU model yielded the greatest (12%). The ID-model optical depths derived using the geostationary satellite were 67% less than those from the polar-orbiting satellite. It is concluded that interpretation of cirrus reflectance with water droplet models leads to biased results. This finding has important implications for the cirrus cloud properties derived by the ISCCP. The cloud-height and optical depth biases can be minimized with the use of the C20 or CS models.