We quantify tree dynamics over a century of free development in a small broad-leaved forest dominated by Fraxinus excelsior and Ulmus glabra. What are the internal and external factors driving the changes, and how predictable are they? What were the time scale and effects of the spread of Dutch elm disease (DED)?Vårdsätra, eastern central Sweden.The survival, growth and recruitment of all trees (≥12 cm girth) were monitored in 1912, 1967, 1988 and 2013 (more often for a part of the forest). Woody species in the field and shrub layers were surveyed in permanent plots in 1976 and 2012. We used transition matrix models to project changes in population sizes and species composition within the century and for 2050.The results indicate that the forest was in a successional development during the first period. The species composition had stabilized by 1967, except for an expansion of Acer platanoides and the drastic effect of DED that struck the forest around 2000. It took only a decade to kill virtually all large elms in the forest, leading to strong decrease in stem density and basal area. The evidence for effects of DED is still weak, but there has been an increase in saplings, notably of Fraxinus, Prunus padus, Ulmus, and of shoots of Corylus avellana. Several species that are abundant in the vicinity and as seeds fail to establish (Picea abies, Betula spp., Quercus robur, Populus tremula). Projections for 2050 based on the third period (1988–2013) are probably unrealistic since Fraxinus may also disappear because of the recent arrival of ash dieback.Slow dynamics in forests that could follow from climate change will locally probably be overruled by unforeseen catastrophes, such as invasions of forest pathogens. These initiate changes with long lag phases that are difficult to quantify. Still, a dense deciduous forest can resist invasion of colonist species and of regionally dominant conifers; the reason being unfavourable conditions for establishment rather than dispersal limitation.