Abstract For many tree species, growth patterns derived from tree-ring time series have been shown to be good indicators of tree mortality. Although tree rings of common beech (Fagus sylvatica) have been widely used to answer complex questions of forest ecology, there are only few studies using growth characteristics, such as growth decline or growth variability, as informative predictors of tree mortality in old-growth beech forests. To do this, we used dendrochronological data of living and dead trees from a nature reserve in eastern Germany. Using a logistic regression model, we predicted tree mortality based on growth characteristics (basal area increment; variance, autocorrelation, mean sensitivity of ring widths; growth trends) over different time horizons. Beech mortality could reliably be predicted up to 20 years in advance on the basis of relative basal area growth. Trees that grew slower in relative basal area than 0.95 cm2/cm y on average over a 20-year-period had a much higher mortality risk than faster-growing trees. Other statistical characteristics of the dendrochronological time series, such as the variance or autocorrelation of ring widths, the mean sensitivity or growth trends, did not convey significant additional information and did not lead to better mortality predictions.