Soybean aphid, Aphis glycines Matsumura, is capable of reducing soybean yield by 20-40% during severe outbreaks in the North Central growing region of the US. Due to dynamic changes in soybean aphid populations within one season, repeated sampling is required to estimate aphid population densities. The objectives of this study were to describe the within-plant distribution of soybean aphid over time, define optimal sample units of varying sizes, and test the ability of selected sample units to estimate whole-plant aphid densities. In addition, within-plant and between-plant variability and associated costs of sampling soybean was used to estimate optimum number of nodes to sample per soybean plant. Within-plant distribution of soybean aphid changed significantly over time. Sample units using one- and two-nodes to estimate whole-plant aphid densities over-estimated aphid densities by 12.8 to 52.5% when using a node-average formula in open-field plots. In addition, using only the terminal node to estimate whole-plant aphid densities resulted in the poorest fit (r2=0.534) and over-estimated densities by 52.5%. However, weighted formulas that accounted for within-plant aphid distributions had slope estimates equal to 1.0 and better explained the variability for open-field plots and exclusion cages with r2 values ranging from 0.796 to 0.986. The advantages of using this sample unit are discussed within the context of further optimizing sampling plans designed to estimate aphid density for research and integrated pest management (IPM) applications.