Four ice-crystal models are tested by use of ice-cloud reflectances derived from Along Track Scanning Radiometer-2 (ATSR-2) and Polarization and Directionality of Earth's Reflectances (POLDER) radiance measurements. The analysis is based on dual-view ATSR-2 total reflectances of tropical cirrus and POLDER global-scale total and polarized reflectances of ice clouds at as many as 14 viewing directions. Adequate simulations of ATSR-2 total reflectances at 0.865 microm are obtained with model clouds consisting of moderately distorted imperfect hexagonal monocrystals (IMPs). The optically thickest clouds (tau > approximately 16) in the selected case tend to be better simulated by use of pure hexagonal monocrystals (PHMs). POLDER total reflectances at 0.670 microm are best simulated with columnar or platelike IMPs or columnar inhomogeneous hexagonal monocrystals (IHMs). Less-favorable simulations are obtained for platelike IHMs and polycrystals (POLYs). Inadequate simulations of POLDER total and polarized reflectances are obtained for model clouds consisting of PHMs. Better simulations of the POLDER polarized reflectances at 0.865 microm are obtained with IMPs, IHMs, or POLYs, although POLYs produce polarized reflectances that are systematically lower than most of the measurements. The best simulations of the polarized reflectance for the ice-crystal models assumed in this study are obtained for model clouds consisting of columnar IMPs or IHMs.