European rabbit Oryctolagus cuniculus abundance was evaluated relative to habitat variables within two zones in Cadiz Province (south-east Spain), 2 years after the spread of rabbit viral haemorrhagic disease (RVHD). The study areas were low-lying regions of mixed pasture/agriculture, Mediterranean forest and matorral, west (Zone A) and east (Zone B) of the Sierra de Cadiz mountain chain. A total of 111 sites was selected to sample all existing habitat types. Relative rabbit density in each site was estimated using a pellet count method in one 50×2 m transect. Mean pellet densities were 21.05 pellet m-2 in Zone B and significantly lower, 5.85 pellets m-2 in Zone A. Physical variables, such as vegetation characteristics (density, height and patchiness of cover-layer vegetation, etc.), topographical characteristics and human influences, cover-layer vegetation species (woody stemmed shrub species or trees forming dense stands), and ground-layer vegetation species (herbaceous species forming a low-lying layer) were determined within transects. In addition to these variables, soil types, rock type and land form, as well as the presence of aquifers, irrigation, and the proximity of rivers, were determined for each site. Significantly higher pellet densities were found in mixed grassland/matorral and cork oak/matorral than in all other habitats. Pellet density was negatively correlated with density of ground-layer vegetation, while high pellet densities were associated with medium-height ground-layer vegetation (50–100 cm). The presence of Entisols was negatively associated with pellet density. The cover-layer species Pistacia lentiscus, Quercus coccifera, Quercus suber, Opuntia ficus-indica and the ground-layer grasses all showed positive associations with pellet density, while the ground-layer species Echium gaditana and Scorpiurus vermiculatus were negatively associated with pellet density. Multivariate analyses identified the mixed grassland/matorral and cork oak forest/matorral habitats as explaining most of the variation in pellet density. The height of cover-layer vegetation was negatively associated, while Opuntia and table-and/calcareous sandstone were positively associated, with mixed grassland/matorral. Distance from habitation and human impact were negatively associated and Quercus suber was positively associated with cork oak forest/matorral. Comparison with previous studies suggest that the low pellet counts in this study represent very low rabbit densities and the population as a whole appears depleted and fragmented.