During 1986–1988, sampling to estimate population densities of the citrus rust mite,Phyllocoptruta oleivoraAshmead, on fruit in individual citrus trees was investigated. A sample unit was a 1-cm2surface area anywhere on a fruit except within areas receiving maximum sun exposure or at the point of the fruit closest to the trunk. Citrus rust mites were usually aggregated on and among fruit within individual citrus trees, and the degree of aggregation generally increased as levels of mites increased. Frequency distributions for the number of mites per two-sample units per fruit usually followed the negative binomial distribution, withkvalues from 0.1 to 4.2 ($\bar x$= 0.53, an estimated commonkwas 0.593). Aggregation by citrus rust mites was similar within ‘Hamlin’ (n= 231 trees) and ‘Valencia’ (n = 172 trees) across a wide range of densities of mites per square centimeter; levels of mites were low in the ‘Pineapple’ orange trees (n = 10) and ‘Marsh’ grapefruit trees (n= 30) sampled, but aggregation by citrus rust mites within these trees was similar to within ‘Hamlin’ and ‘Valencia’ trees. Fewer mites were often present in the west quadrant of trees. Average densities of mites per square centimeter on fruit in individual trees were estimated using 20 fruit samples per tree and two subsamples per fruit, with the fruit spaced evenly around the canopy at 0.3 to 2.1 m above the ground. Relative variation of estimates was large when the density of mites was small but decreased as densities increased. Optimum sample sizes per fruit and per tree therefore varied at different densities. Analyses of variance components indicated that standard errors associated with sampling 25 fruit with one subsample per fruit would be equivalent to those associated with sampling 20 fruit with two subsamples per fruit. Expected standard errors and relative variation for a wide range of sample sizes per tree were examined.