ABSTRACT Severe epidemics of Fusarium head blight (FHB) caused by Fusarium graminearum, group II (teleomorph: Gibberella zeae) have been occurring on wheat crops in the northcentral United States and southern Canada. Evaluation of resistance to FHB is difficul, because resistance is partial and infection depends upon host plant maturity. Variance component analysis was conducted to determine how best to allocate resources among environments, replications, and subsamples (heads per plot) in FHB screening nurseries. Advanced breeding lines from the Ohio State University wheat-breeding program were evaluated in screening nurseries from 1995 to 1997. Nurseries were artificially inoculated and sprinkler-irrigated to induce FHB epidemics. Over 80% of the variation within an environment resulted from variation associated with subsampling individual heads within plots. The second greatest source of variation was due to genotype by replication interactions. Host plant maturity influenced disease ratings in 1997. The repeatability of genotype means was approximately 50% within environments. The greatest reduction in genotype standard errors was obtained through additional environments, and then replications. Because the cost of an additional environment was estimated at five times the cost of an additional replication, the most cost-effective improvement in precision was obtained through the addition of replications. Advanced breeding lines should be evaluated in at least four replications per environment. Segregating populations will require more replications.