Question: We studied the patterns of plant species co-occurrence on three, nested, spatial scales in semi-natural grassland communities and explored the possible ecological processes underlying the patterns. Location: Dry, semi-natural grasslands in a 4.5 km x 4.5 km area on the Baltic Island of Oland (Sweden). Methods: The study used replicated samples on three, nested, spatial scales: 50 cm x 50 cm plots (N= 516), grassland patches (N = 109) and the whole landscape (N= 6). We used a null model approach to study species co-occurrence patterns and compared the ecological amplitudes of the pairs of species contributing most to the patterns. We used linear models to search for associations between species segregation and environmental and landscape factors. Results and Conclusions: Our results support the prediction that patterns of species co-occurrence are likely to be influenced by different mechanisms on different spatial scales. On the plot scale, we interpreted the species segregation in terms of species interactions. The degree of species segregation was significantly associated with the plots'positions within the grassland patches (edge effects) and with management intensity of the grasslands - both variables can be assumed to influence species interactions. On the grassland patch scale, we interpreted the species segregation in terms of within-patch environmental heterogeneity. The degree of segregation was significantly associated with the area of the grassland patches and with management intensity - both variables that are likely to be related to environmental heterogeneity within the grasslands. Species segregation on the landscape scale was interpreted in terms of environmental heterogeneity among grassland patches and was significantly associated with land-use history