Camellia rosthorniana is a long-lived evergreen shrub, which is widely distributed in subtropical areas of China and occurs in many types of communities on Mt. Jinyun (29°50′ N, 106°26′ E). In the present study, reproductive allocation and fruit set (number of mature fruits/number of flower buds) of C. rosthorniana populations at three successional stages (Giant bamboo forest, mixed coniferous broadleaved forest and evergreen broadleaved forest) with different population structures and densities were compared. One 2 000 m 2 plot was set up in each of the three C. rosthorniana populations. Field surveys showed that the basal diameters of the largest individuals of C. rosthorniana in the giant bamboo community, mixed coniferous broadleaved community and evergreen broadleaved community were 3.52, 5.37 and 6.44 cm, respectively. Population densities (Flowering plants) in the plots were 167, 222 and 621 for giant bamboo community, mixed coniferous broadleaved community and evergreen broadleaved community, respectively. It was also found that individuals of C. rosthorniana began to flower only after it reached a basal diameter of 1 cm. All individuals in the 1.0-1.5 cm (Basal diameter) size class in the giant bamboo community flowered, but only a proportion of individuals in the same size class flowered in the mixed coniferous broadleaved community and evergreen broadleaved community. Six reproductive individuals were randomly sampled from each size class (with 0.5cm as an interval) in each population. Reproductive allocation, based on annual production, was estimated for each size class and population. Fruit set was also estimated for each population from ten randomly selected reproductive individuals. Three hypotheses were evaluated with regard to the observed patterns of reproductive allocation. One hypothesis, that reproductive allocation patterns were a function of differences in resource availability, did not provide an explanation for our results. A second hypothesis, that reproductive allocation was negatively correlated with successional maturity of the habitat could explain the observed patterns of reproductive allocation but could not explain the differences in size structures of the three populations. The third, life history hypothesis which assumed the existence of a trade-off between current reproduction and vegetative growth and/or survival rate, could explain the differences in size structures among the three populations. Reproductive allocation increased monotonically within a range of basal area sizes only, and a parabolic, not linear, model could best delineate the relationship between reproductive allocation and plant size (indicated by basal diameter). No significant differences in fruit set among populations were detected.