Aim This study investigated whether habitat fragmentation at the landscape level influences patch occupancy and abundance of the black-headed gull, Chroicocephalus ridibundus, and whether the response of the species to environmental factors is consistent across replicated landscape plots.Location Water bodies (habitat patches) in southern Poland.Methods Surveys were conducted in two landscape types (four plots in each):(1) more-fragmented landscape, in which habitat patches were small (mean size 2.2-6.2 ha) and far apart (mean distance 2.5-3.1 km); and (2) less-fragmented landscape, in which habitat patches were large (mean size 9.2-16.5 ha) and separated by short distances (mean 0.9-1.4 km). Observations were performed twice in 284 potential habitat patches during the 2007 breeding season.Results Colonies were significantly more frequent and larger in the lessfragmented landscapes than in the more-fragmented ones. Probability of patch occupancy and number of breeding birds were positively related with patch size and these relationships were especially strong in the more-fragmented landscapes. In the less-fragmented landscapes, the occurrence of black-headed gulls was negatively related to the distance to the nearest local population, but in the morefragmented landscapes such a relationship was not detected. As distance to the nearest habitat patch increased, the probability of the patch occupancy decreased in the more-fragmented landscapes. Moreover, abundance was negatively influenced by distance to the nearest habitat patch, especially strongly in morefragmented landscapes. Proximity of corridors (rivers) positively influenced the occupation of patches regardless of landscape type. The number of islets positively influenced occupancy and abundance of local populations, and this relationship was stronger in the more-fragmented landscapes.Main conclusions Our results are in agreement with predictions from metapopulation theory and are the first evidence that populations of blackheaded gulls may have a metapopulation structure. However, patch occupancy and abundance were differentially affected by explanatory variables in the morefragmented landscapes than in the less-fragmented ones. This implies that it is impossible to derive, a priori, predictions about presence/abundance patterns based on only a single landscape.