The geographic mosaic theory of coevolution predicts that geographic variation in species interactions will lead to differing selective pressures on interacting species, producing geographic variation in the traits of interacting species (Thompson 2005). We supported this hypothesis in a study of the geographic variation in the interactions among Eurosta solidaginis and its natural enemies. Eurosta solidaginis is a fly (Diptera: Tephritidae) that induces galls on subspecies of tall goldenrod, Solidago altissima altissima and S. a. gilvocanescens. We measured selection on E. solidaginis gall size and shape in the prairie and forest biomes in Minnesota and North Dakota over an 11-year period. Galls were larger and more spherical in the prairie than in the forest. We supported the hypothesis that the divergence in gall morphology in the two biomes is due to different selection regimes exerted by natural enemies of E. solidaginis. Each natural enemy exerted similar selection on gall diameter in both biomes, but differences in the frequency of natural enemy attack created strong differences in overall selection between the prairie and forest. Bird predation increased with gall diameter, creating selection for smaller-diameter galls. A parasitic wasp, Eurytoma gigantea, and Mordellistena convicta, an inquiline beetle, both caused higher E. solidaginis mortality in smaller galls, exerting selection for increased gall diameter. In the forest there was stabilizing selection on gall diameter due to a combination of bird predation on larvae in large galls, and M. convicta- and E. gigantea-induced mortality on larvae in small galls. In the prairie there was directional selection for larger galls due to M. convicta and E. gigantea mortality on larvae in small galls. Mordellistena convicta-induced mortality was consistently higher in the prairie than in the forest, whereas there was no significant difference in E. gigantea-induced mortality between biomes. Bird predation was nonexistent in the prairie so the selection against large galls found in the forest was absent. We supported the hypothesis that natural enemies of E. solidaginis exerted selection for spherical galls in both biomes. In the prairie M. convicta exerts stabilizing selection to maintain spherical galls. In the forest there was directional selection for more spherical galls. Eurytoma gigantea exerted selection on gall shape in the forest in a complex manner that varied among years. We also supported the hypothesis that E. gigantea is coevolving with E. solidaginis. The parasitoid had significantly longer ovipositors in the prairie than in the forest, indicating the possibility that it has evolved in response to selection to reach larvae in the larger-diameter prairie galls.