Abstract Aim To estimate species turnover of plants on 32 small islands within a 20-year period and to assess possible changes in community composition and properties, such as species richness and factors affecting it, nestedness, species co-occurrence and overall community similarity. Additionally, to assess the possible effects of grazing, gull colonies and fire on turnover values. Location Thirty-two islets in the eastern Aegean Sea (Greece). Methods Complete sampling of plants was performed in 1974 and in 1990–94 (mostly in 1994, which was used as the reference year). Species turnover rates were estimated using both per island and per species approaches. Multiple regression was used to evaluate factors affecting species richness. Chi-square tests were applied to compare community composition among sampling periods. The effects of various factors on turnover rates and species richness were examined using one-wayanovaandancova. Mann–Whitney tests were applied in order to check for differences between frequencies of occurrence of extinct, immigrant and persisting species. Community nestedness was calculated usingbitmatnestand the C-score index for co-occurrence was estimated using EcoSim7. Species similarities among islands in each of the 1974 and 1994 data sets were assessed using Jaccard’s index and the two similarity matrices were compared using a Mantel test. Results Of 391 species recorded on the islets, 334 were present in 1974, 301 in 1994 and 244 were common to both these periods. Species richness in the 1974 and 1994 data sets was significantly correlated with elevation and area, but not with distance from the nearest large island. Richness was positively affected by grazing, but not by fire or gull colonies. The slopes of species–area and species–elevation regressions were almost identical in 1974 and 1994. Mean relative turnover was 2.06 (species per islet) and 3.26 (islets per species). Turnover was not correlated with area, elevation or distance from the nearest large island. Nestedness and co-occurrence levels were very similar. Tables of islet by islet floral similarity (Jaccard’s index) did not differ between the 1974 and 1994 data sets. Main conclusions The turnover rates found are among the highest recorded for plants; at the same time the islet communities exhibit notable stability in overall properties. Our results provide evidence for rapid shifts in species number that may nonetheless be considered as equilibrial dynamics, as these islets are able to respond rapidly to environmental change and disturbance. Human activities, notably the application of grazing, have a significant complicating effect on community dynamics, enhancing observed turnover rates.