The dispersion of root injury to Zea mays L. by corn rootworm, Diabrotica spp., larvae was characterized using geostatistics to determine the appropriate sampling distance to obtain independent estimates of root injury. We also investigated the effect of the root injury levels on the spatial dependence of root injury at three different spatial scales: large-scale study with 25-m sampling distance in three, 8-ha fields; moderate-scale study with 0.2-m sampling distance in six, 2-ha fields; and small-scale study with exhaustive sampling in eight, 12-m² plots within cornfields. In the large-scale study, a nugget model best fit the semivariograms, indicating the dispersion of root injury exhibited a random pattern. In the moderate-scale study, all the semivariograms were best fit with an exponential model, and 68 to 96% of the variability was explained by spatial dependence, suggesting a high degree of spatial aggregation of corn root larval injury. The ranges of semivariograms fell within <4 m, except one field that had the range of 14.7 m. In the small-scale study, the Gaussian, exponential, or spherical models fit the semivariograms best and 50 to 89% of the variability was explained by spatial dependence, suggesting spatial aggregation of root injury. The ranges of semivariograms were from 0.35 to 1.04 m. In the small-scale study, the degree of spatial dependence increased significantly (P < 0.05) as the average root injury increased within the range of root injury observed in the study. These results suggest that the minimum distance between root samples in small-plot insecticide trials should be >1 m within a row to obtain unbiased estimates of root injury.