Spaceborne millimeter-wave radar has been identified as a possible instrument to make global measurements in ice clouds, which have an important but poorly understood role in the earth's radiation budget. In this paper, the authors explore the potential of a dual-frequency spaceborne radar to estimate crystal size in cirrus clouds and, hence, determine ice water content and the shortwave extinction coefficient more accurately than would be possible using a single radar. Calculations show that gaseous attenuation is not a serious problem for a nadir-pointing radar measuring down to cirrus altitudes at frequencies between 35 and 215 GI Iz, provided the frequencies are chosen to lie in the window regions of the atmospheric absorption spectrum. This enables one to exploit the significant benefits of using frequencies too high to be operated from the ground. Radar reflectivity at 35, 79, 94, 140, and 215 GHz has been calculated from aircraft ice particle size spectra obtained during the European Cloud Radiation Experiment (EUCREX) and the Central Equatorial Pacific Experiment (CEPEX), and it is shown that overall the most promising dual-wavelength combination for measuring crystal size and ice water content is 79 and 215 GHz. For a minimum radar sensitivity of -30 dBZ, this combination can measure ice water content and median volume diameter with errors of between 10% and 30% when the reflectivity is greater than -15 dBZ (equivalent to an ice water content of around 0.015 g m(-3)). If only a single wavelength radar were affordable, then, for estimating ice water content, 215 GHz would be the preferred choice. Since the two radars would be likely to use the same antenna, the authors also consider the effect of cloud inhomogeneities to introduce a random error into the reflectivity ratio because of the different beamwidths at each frequency. It is found, using data from the cloud radars at Chilbolton, England, that this is more than 0.2 dB for frequency pairings that include 35 GHz but for all other combinations is less than 0.1 dB, which is comparable to the other errors in the system and much smaller than the typical values being measured. Nonspherical crystals are shown to have a significant effect on the size measured by a nadir-pointing dual-wavelength radar, but the authors present evidence that this can be largely eliminated by viewing at 45 degrees from nadir.