Genetic parameters for stem diameter and wood density were compared at selection (4-5 years) and harvest (16-17 years) age in an open-pollinated progeny trial of Eucalyptus globulus in Tasmania (Australia). The study examined 514 families collected from 17 subraces of E. globulus. Wood density was assessed on a subsample of trees indirectly using pilodyn penetration at both ages and directly by core basic density at harvest age. Significant additive genetic variance and narrow-sense heritabilities (h 2 op ) were detected for all traits. Univariate and multivariate estimates of heritabilities were similar for each trait except harvest-age diameter. Comparable univariate estimates of selection-and harvest-age heritabilities for diameter masked changes in genetic architecture that occurred with stand development, whereby the loss of additive genetic variance through size-dependent mortality was countered by the accentuation of additive genetic differences among survivors with age. Regardless, the additive genetic (r a ) and subrace (r s ) correlations across ages were generally high for diameter (0.95 and 0.61, respectively) and pilodyn penetration (0.77 and 0.96), as were the correlations of harvestage core basic density with selection-and harvest-age pilodyn (r a −0.83, −0.88; r s −0.96, −0.83). While r s between diameter and pilodyn were close to zero at both ages, there was a significant change in r a from adverse at selection age (0.25) to close to zero (−0.07) at harvest age. We argue that this change in the genetic correlation reflects a decoupling of the genetic association of growth and wood density with age. This result highlights the need to validate the use of selection-age genetic parameters for predicting harvest-age breeding values.