We examine the alternative hypotheses that on Venus (i) the spatial randomness of the crater population reflects predominantly the spatially stochastic nature of impact crater formation or (ii) a significant amount of spatial variation in crater density is due to variations in the geological rate of crater removal. Three endmember models of resurfacing are formulated: Case I assumes that a single-age production surface accounts for the spatial distribution of craters. Case II assumes that variations in spatial crater density reflect local production surface ages, while Case III assumes that individual craters are removed regionally, so that variations in spatial crater density reflect local retention ages. These endmember models are tested by examining correlations of crater properties with crater spatial density, &rgr;. We analyze the distribution of embayed and tectonized craters and the relative abundances of extended ejecta deposits (termed ''halos''). The three endmembers provide specific predictions as to the occurrence of halos as a function of &rgr;. For &rgr;?1.5 craters per 106 km2, Case III provides the best explanation for the deficiency of halos as well as the relative abundances of modified craters. For the interval 1.5≲&rgr;≲2.5, Case II is applicable but with only a modest spread in production ages. For densities >2.5, Case II is the only endmember model that matches the global observations, suggesting that regions with these high densities represent a series of distinct production surfaces with ages increasing from the planetary mean retention age. ¿ American Geophysical Union 1995