Polycross mating systems are widely used in forest tree breeding for genetic testing. Backward selection based on polycross testing assumes equal male reproductive success and true half-sib progeny. The main objectives of this study were, firstly, to investigate the departure from these assumptions in a maritime pine polycross trial and, secondly, to evaluate the consequences for heritability and breeding values estimations. A total of 984 offspring from 98 half-sib families was genotyped with single nucleotide polymorphism markers to recover the full pedigree. Paternity was assigned successfully for 89 % of the offspring at a 99 % confidence level. We thus concluded there was an 11 % pollen contamination rate, assuming contamination when no genotype from the polymix composition could be identified as a father. The paternal contribution to the offspring varied among the males, but the departure from half-sib assumption was moderate since the average genetic correlation within the family was 0.26. Heritability and breeding values for girth at breast height and stem sweep were estimated using individual-tree mixed models with either partial or full pedigree information. The results highlighted a minor bias in heritability estimation due to unknown paternity, as well as a high correlation for estimated breeding values between the partial and full pedigree models, suggesting that the genetic merit of the parental generation for backward selection was adequately predicted using the partial pedigree model. Finally, pedigree recovery was also discussed in a perspective of forward selection.< Réduire