The conventional way to drive modifications in old forest tree seed orchards is to establish progeny trials involving each parent tree and then evaluate its contribution to the performance of the progeny by estimating its general and specific combining ability (GCA and SCA). In this work, we successfully applied an alternative parent selection tactic based on paternity testing of superior offspring derived from a hybrid seed orchard established with a single Eucalyptus grandis seed parents and six E. urophylla pollen parents. A battery of 14 microsatellite markers was used to carry out parentage tests of 256 progeny individuals including two independent samples of selected trees and one control unselected sample, all derived from 6-year-old forest stands in eastern Brazil. Paternity determination was carried out for all progeny individuals by a sequential paternity exclusion procedure. Exclusion was declared only when the obligatory paternal allele in the progeny tree was not present in the alleged parent tree for at least four independent markers to avoid false exclusions due to mutation or null alleles. After maternity checks to identify seed mixtures and selfed individuals, the paternity tests revealed that approximately 29% of the offspring was sired by pollen parents outside the orchard. No selfed progeny were found in the selected samples. Three pollen parents were found to have sired essentially all of the offspring in the samples of selected and non-selected progeny individuals. One of these three parents sired significantly more selected progeny than unselected ones ( P< or =0.0002 in a Fisher exact test). Based on these results, low-reproductive-successful parents were culled from the orchard, and management procedures were adopted to minimize external pollen contamination. A significant difference ( P<0.01) in mean annual increment was observed between forest stands produced with seed from the orchard before and after selection of parents and revitalization of the orchard. An average realized gain of 24.3% in volume growth was obtained from the selection of parents as measured in forest stands at age 2-4 years. The marker-assisted tree-breeding tactic presented herein efficiently identified top parents in a seed orchard and resulted in an improved seed variety. It should be applicable for rapidly improving the output quality of seed orchards, especially when an emergency demand for improved seed is faced by the breeder.