Additive and digenic variances for dry matter yield were estimated from a population of F1 lines of alfalfa (Medicago sativa L.). The F1 lines and their polycross (PX) progenies were both evaluated over three harvests in a simulated sward in a growth room. From the analyses of variance, the genetic variances of the F1 and PX lines were estimated and the covariance between the F1 and PX performance calculated for dry matter yield. Correlations between relatives were used to partition these variances into additive and digenic effects. Narrow-sense heritabilities and the expected response to selection were estimated for the F1 population and broad-sense heritabilities were calculated from the F1 and PX trials. The pooled estimates of additive variance were significantly greater than zero for the second harvest and total yield while all estimates of digenic variance were nonsignificant. This indicated that the genetic variance for dry matter yield was mainly additive. The expected response to selection for total dry matter yield, if the top 10% of the F1 lines were recombined, was 23% of the F1 population mean. Based on these results, it was concluded that selection gains for yield could be made in this population of alfalfa.