Molecular markers have been previously found useful for increasing genetic gain in maize (Zea mays L.). The use of doubled haploids (DHs) instead of F2 plants (or, equivalently, F3 families) may permit a better estimation of marker-trait associations. Our objective was to determine the usefulness of DH versus F2 populations in marker-assisted recurrent selection (MARS) and genomewide selection. We simulated testcrosses from a DH population and an F2 population from the same cross between two inbreds and studied genetic models defi ned by the number of quantitative trait loci (QTL) and trait heritability (H). Equal-time comparisons of selection response were between Cycle 3 with an F2 population and Cycle 2 with a DH population. For the genetic model of 100 QTL, H = 0.20 and a population size of N = 100, the ratio of response to selection in DH versus F2 populations was RDH:F2 = 109% for genomewide selection and 128% for MARS. For the genetic model of 20 QTL, H = 0.80 and N = 100 these values decreased to 99% for genomewide selection and 109% for MARS. Although genomewide selection was superior to MARS for a given type of population, the advantage of using DH instead of F2 populations was greater in MARS than in genomewide selection. We concluded that DH populations are most useful in genomewide selection and MARS when many QTL control the trait, H is low, and N is small.