We study the intergenerational transmission of cognitive and non-cognitive abilities between parents and sons using population-wide enlistment data. Conscripts are evaluated at the same age and with comparable methods across cohorts, and we correct for measurement error bias in fathers' ability measures by using their brothers' abilities as instruments. The uncle instrument is supported by a host of validity tests. This strategy also enables us to predict mothers' abilities. Our results suggest that previous estimates of intergenerational ability correlations are biased downwards; in particular for non-cognitive skills. When this bias is corrected for the non-cognitive correlation is close to that of cognitive abilities. Using predicted abilities, we further find the mother-son cognitive ability correlation to be stronger than the father-son correlation. Finally, educational attainment and labor market outcomes of both sons and daughters are found to be strongly related to both parents' cognitive and non-cognitive abilities.