We examined the distribution of surface slopes of a variety of terrestrial surfaces by field measurement, representing surfaces formed by a wide range of processes, and compared the results to planetary radar data. Slope distributions of the measured surfaces differed considerably from the distributions assumed by accepted models of radar scattering. The rms slope values for the terrestrial surfaces ranged from 0¿ to 16.5¿, compared to a range of rms slope values of <1¿ to about 10¿ for planetary surfaces as inferred from radar observations. While the great majority of planetary surfaces observed by radar have rms slope estimates in the lower end of this range (<5¿), nearly all of the terrestrial surfaces we measured have rms slope values greater than 5¿. We also used Hagfors' model of radar scattering to predict the return that would be expected from surfaces where two discrete surface types were present within the radar field of view and found that the shapes of the resulting slope distributions differed from those predicted by the Hagfors model for homogeneous surfaces. Additionally, the result best fit rms slope was a nonlinear combination of those of the pure surfaces, emphasizing the smoother surface. Together, these results suggest that current methods of determining surface roughness from radar may significantly underestimate the roughness of planetary surfaces and that the derived rms slope can best be used as a qualitative guide to the physical interpretation of actual surface properties. ¿ American Geophysical Union 1993