1. Presence-absence sampling is an important method for monitoring state and change of both individual plant species and communities. With this method only the presence or absence of the target species is recorded on plots and thus the method is straightforward to apply and less prone to surveyor judgment compared to other vegetation monitoring methods. However, in the basic setting all plots must be equally large or otherwise it is unclear how data should be analyzed. In this study we propose and evaluate five different methods for estimating plant density based on presence-absence registrations from surveys with variable plot sizes. 2. Using artificial plant population data as well as empirical data from the Swedish National Forest Inventory we evaluated the performance of the proposed methods. The main analysis was conducted through sampling simulation in the artificial populations, whereby bias and variance of density estimators for the different methods were quantified and compared. 3. Both for state and change estimation of plant density, we found that the best method to handle variable plot size was to perform generalized least squares regression, using plot size as an independent variable. Methods where plots smaller than a certain threshold were excluded or their registrations recalculated were, however, almost as good. Using all registrations as if they were obtained from plots with the nominal plot size resulted in substantial bias. 4. Our findings are important for plant population studies in a wide range of environmental monitoring programmes. In these programmes plots are typically randomly laid out and may be located across boundaries between different land use or land cover classes, resulting in subplots of variable size. Such splitting of plots is common when large plots are used, e.g. with the 100 m2 plots used in the Swedish National Forest Inventory. Our methods overcome problems to estimate plant density from presence-absence data observed in plots that vary in size.