Future adaptation to changes in the environment depends on the existence of additive genetic variances within populations. Recently, considerable attention has also been given to the non-additive component, which plays an important role in inbreeding depression and bottleneck situations. In this study, I used data from a North Carolina II crossing experiment, analysed with restricted maximum-likelihood methods, to estimate the additive and dominance genetic (co)variances for eight quantitative characters in two different-sized populations of Scabiosa canescens, a rare and threatened plant in Sweden. There was no evidence for genetic erosion in the small Hällestad population ( approximately 25 individuals) relative to the large Ahus population ( approximately 5000 individuals). In fact, slightly higher heritabilities were found in the Hällestad population. The additive genetic variance was statistically significant for all traits in both populations, but only a few additive covariances reached significance. The Hällestad population also had higher mean levels and more traits with significant dominance variance than the Ahus population. The variance attributable to maternal effects was too low to be considered significant. There was only a weak correspondence between heritabilities for each trait in the present study and previous estimates based on open-pollinated families of the same populations, but the mean heritability (over characters) was consistent between the studies.