We used multiple regression modeling to investigate the numerical response by the predatory insects Hippodamia convergens Guérin-Méneville, H. parenthesis (Say), and C. septempunctata L. (Coleoptera: Coccinellidae), Chrysoperla plorabunda (Fitch) (Neuroptera: Chrysopidae), and Nabis americoferus Carayon (Hemiptera: Nabidae) to aphids during 5 yr in three geographically separated alfalfa fields in eastern South Dakota. Regression models for abundance of adults of all species were significant. Regression models for immature H. convergens, H. parenthesis, and C. septempunctata were significant, but regression models for immature C. plorabunda and N. americoferus were not significant. Regression parameters differed among the three fields for most predator species, indicating that the numerical response was dependent on geographical location. To obtain insight into why the numerical response by predators differed among fields we determined how the abundance of predators in alfalfa fields was influenced by the landscape surrounding a field and the vegetation in it. Variables describing the complexity of the landscape surrounding alfalfa fields and the plant community in the fields entered into regression models for predator abundance and explained a greater proportion of the variance in predator abundance than aphid abundance did. We conclude that the structure of the landscape matrix plays an important role in determining the abundance of aphid predators in alfalfa fields, as does the plant community in a field. These effects can sometimes overshadow the direct numerical response by predators to aphids.