Fusarium graminearum, the causal agent of Fusarium head blight (FHB) in wheat and barley, is one of the most economically destructive pathogens of these grains worldwide. Recent population genetic studies of the pathogen obtained from wheat in North America supported population subdivision in part correlated with the spectrum of trichothecene mycotoxins (chemotype) produced by individuals within each population. In contrast, a recent study of F. graminearum obtained from diseased barley in the upper Midwestern United States concluded that only a single population was present, consisting of individuals with various chemotypes. To test whether strains derived from different hosts potentially have different population dynamics, we obtained the barley strains used in the previous study and compared them with wheat strains isolated at a similar time and geographic origin. A total of 247 F. graminearum isolates from barley were assigned firmly into two clusters using a Bayesian clustering method. Subdivision within the barley population corresponded to the previously described NA1 (correlated with the 15ADON chemotype) and NA2 (correlated with the 3ADON chemotype) populations from wheat. However, in both sampling periods the barley population exhibited a higher level of genetic differentiation between NA1 and NA2 populations, fewer admixed individuals and evidence of unidirectional gene introgression (15ADON strains with NA2 genetic backgrounds). These results suggest less recombination between NA1 and NA2 populations on barley compared with wheat. The frequency of 3ADON chemotype strains in the most recently surveyed barley population suggests a latitudinal cline from the northern (49%), central (40%) to the southern (29%) sampling area. The potential to produce a novel trichothecene, 3α-acetoxy,7α,15-dihydroxy-12,13-epoxytrichothe-9-ene (NX-2), was not detected in the barley population but occurred at a low rate (2.4%) in the wheat population.