Pyrenophora semeniperda, an important pathogen in Bromus tectorum seed banks in semi-arid western North America, exhibits >4-fold variation in mycelial growth rate. Host seeds exhibit seasonal changes in dormancy that affect the risk of pathogen-caused mortality. The hypothesis tested is that contrasting seed dormancy phenotypes select for contrasting strategies for increasing pathogen fitness, and that increased fitness on nondormant seeds involves a resource trade-off between toxin production and growth. The strategy for successfully attacking rapidly germinating nondormant seeds at high inoculum loads in autumn involves increased post-infection aggressiveness to prevent seed escape through germination. An earlier study demonstrated that slow-growing strains caused higher mortality than faster-growing strains on nondormant host seeds at high inoculum loads. In this study, production of the toxin cytochalasin B was significantly higher in slower-growing strains, and was induced only in seeds or in seed-constituent-containing media. Its production was reduced in vivo by Bromus tectorum seeds, suggesting direct involvement in pathogenesis on seeds. Fast-growing strains caused significantly higher mortality than slow-growing strains at low inoculum loads on dormant seeds, which apparently have resistance that is overcome at high loads or through rapid mycelial proliferation. In a co-inoculation study, the fast-growing isolate produced 3 × more stromata than the slow-growing isolate on dormant seeds, whereas the slow-growing isolate was twice as successful on nondormant seeds. These results provide evidence that mycelial growth rate variation and associated variation in cytochalasin B production represent a trade-off maintained through temporally varying selection resulting from seasonal variation in host seed dormancy status.