Nine varieties of dry beans representing five market classes were grown in three locations (Maryland, Michigan, and Nebraska), and subsamples were collected for each variety (row composites from each plot). Aqueous methanol extracts of ground beans were analyzed in triplicate by UV spectrophotometry. Analysis of variance-principal component analysis was used to quantify the relative variance arising from location, variety, between rows of plants, and analytical uncertainty and to test the significance of differences in the chemical composition. Statistically significant differences were observed between all three locations, between all nine varieties, and between rows for each variety. PCA score plots placed the nine varieties in four categories that corresponded with known taxonomic groupings: (1) black beans (cv. Jaguar and cv. T-39), (2) pinto beans (cv. Buster and cv. Othello), (3) small red beans (cv. Merlot), and (4) great northern (cv. Matterhorn and cv. Weihing) and navy (cv. Seahawk and cv. Vista) beans. The relative plant-to-plant variance, estimated from the between row variance, was 71-79% for 25-40 plants per row.