The effect of the use of different doses of the fungicide fenpropimorph on populations of barley powdery mildew Blumeria (Erysiphe) graminis f. sp. hordei was investigated in a laboratory selection experiment. A sample from the Danish aerial population of powdery mildew was split into populations, and these were kept separately for 31 generations on susceptible barley seedlings treated with fungicide at two concentrations, as well as on a control. Samples from these populations were tested for their resistance to fenpropimorph and their virulence spectra. There was a large amount of environmental variation in the ED50 values used to measure fungicide resistance. In both treated populations, the average level of fungicide resistance increased, this increase being faster and greater in the population treated with the high dose. The diversity of pathotypes of the treated populations decreased, with the decline being more rapid in the population treated with the high dose, where one pathotype dominated the population after 31 generations. This pathotype was apparently not the fittest in the population treated with the low dose. This implies that knowledge of ED50 is not sufficient to predict pathotype evolution under different fungicide treatments. The dominant pathotype in the high-dose treatment may not have been clonal, as there was evidence of two levels of fungicide resistance. The large environmental variation observed in estimated ED50 values for resistance towards fenpropimorph may help to explain why this resistance has evolved at a slower rate than resistance towards other fungicides.