Soybean [Glycine max (L.) Merr.] seed composition data are becoming critical for efficient marketing and trade of soybean and soybean products. Previous reports of variation in seed protein and oil across regions within the United States focused on variation associated with state and regional boundaries. We conducted an analysis of an 8-yr survey of seed protein and oil across US states and regions, including a geostatistical approach to better characterize the continuous variation in seed composition on a regional scale. The objectives were (i) to determine the relative importance of temporal and regional variation, (ii) to explore the extent of spatial variability across years, (iii) to evaluate the temporal stability across regions, and (iv) to explore the negative correlation between protein and oil across regions and years. Our results confirmed previous findings showing higher protein concentration in southern states and regions. However, most of the observed variation occurred at scales below these political boundaries. The geostatistical approach indicated a moderate level of spatial dependency for protein and protein plus oil but low spatial autocorrelation for oil. In all cases, year-to-year variation in weather conditions modulated the expression of regional spatial patterns in composition. To fully predict seed composition at a regional scale would require additional information associated with weather and agronomic management.