An analytical study shows that the ratio of backscatter from ice particles at two CO2 lidar wavelengths is substantially different from the ratio from water drops. This forms the basis for a new method to discriminate between ice, water, and mixed-phase clouds. (The polarization technique often used by lidan operating in or near the visible part of the spectrum is not effective for C02 lidars, because depolarization from ice particles is usually very small at its infrared wavelength.) The effect of particle size distribution, differential attenuation in the cloud and clear air and other sources of uncertainty on the accuracy of the two-wavelength method are evaluated. Five wavelength pairs are examined to establish criteria for designing an optimum lidar.