Aim: To describe the habitat characteristics and status of the brush-tailed rabbit-rat, Conilurus penicillatus Gould, 1842, on the Tiwi Islands, northern Australia, as part of a broader programme aimed at the conservation management of this species. In addition, comparable environmental modelling is undertaken for other co-occurring small native mammals, including the black-footed tree-rat, Mesembriomys gouldii Gray, 1843, a taxonomically and ecologically related species. These objectives relate to the significance for mammal conservation of islands generally in Australia, and the recent intensification of plantation forestry on these previously little-disturbed islands. Location: Melville and Bathurst islands (Tiwi Islands), respectively, Australia's second and fifth largest islands. Methods: A systematic survey was conducted for mammals across Bathurst (115 sampled quadrats) and Melville Island (236 quadrats). A broad range of environmental variables was recorded for every quadrat. All quadrats were classified by their woody plant species composition. The relative occurrence of individual mammal species across the resulting vegetation groups was examined using Kruskal-Wallis anova. The habitat relationships of C. penicillatus and the most commonly recorded mammal species were described by generalized linear modelling, with separate models for each island, for both islands combined, for all habitats and for only those sites dominated by eucalypts. Results: Twelve small mammal species (excluding bats, macropods and feral animals) were recorded in this study. The most notable feature of this survey was the lack of records of M. gouldii from Bathurst Island. In contrast, the proportion of quadrats with C. penicillatus was not significantly different between the two islands. There was no significant tendency for these two species to co-occur in quadrats on Melville Island more or less commonly than by chance. Conilurus penicillatus was most abundant in eucalypt forest while M. gouldii showed a weak association with eucalypt forests and woodlands and shrub land. The five most commonly recorded species showed highly idiosyncratic relationships with environmental variables, with this relationship showing some variation between the two islands. None showed any significant association with floristic variation within the extensive eucalypt forests, but most showed significant associations with tree height, basal area (especially of large trees), landscape position (distance to watercourse) and fire history. Main conclusions: Conilurus penicillatus was most likely to occur in tall eucalypt forest away from watercourses. This habitat is now being targeted for clearance for the development of plantations of the exotic Acacia mangium. Seven of the 12 mammal species examined in this study (C. penicillatus, M. gouldii, Rattus tunneyi Thomas, 1904, Melomys burtoni Ramsay, 1887, Sminthopsis butleri Archer, 1979, Phascogale tapoatafa Meyer, 1793 and Petaurus breviceps Gould, 1842) were not recorded at all in plantations, and these (and other) species are likely to be severely disadvantaged by plantation development. The study also demonstrated that the two medium to large arboreal rodent species (C. penicillatus and M. gouldii) vary in environmental associations and found no evidence that C. penicillatus increased in areas unoccupied by M. gouldii.