We used logistic and Poisson regression models to determine factors of forest and landscape structure that influence the presence and abundance of rodent species in the rain forest of Ranomafana National Park in southeastern Madagascar. Rodents were collected using live-traps along a gradient of human disturbance. All five endemic rodent species (Nesomys rufus, N. audeberti, Eliurus tanala, E. minor and E. webbi) and the introduced rat Rattus rattus were captured in both secondary and primary forests, but the introduced Mus musculus was only trapped in secondary forest. The abundance of R. rattus increased with the level of habitat disturbance, and it was most common in the heavily logged secondary forest. Furthermore, the probability of the presence of R. rattus increased with decreasing distance from forest edge and decreasing canopy cover, while the probability of presence increased with increasing herbaceous cover, altitude and overstory tree height. The species was never observed farther than 500 m away from human habitation or camp-site. N. rufus prefered selectively-logged forest at altitudes above 900 m a.s.l. Its probability of presence increased with increasing canopy cover, herbaceous cover and distance from forest edge, and with decreasing density of fallen logs, overstory tree height and distance from human habitation. N. audeberti prefered heavily-logged areas, while E. tanala was the only species occurring along the entire range of forest disturbance. We suggest that in the Ranomafana National Park the spread of R. rattus is associated with deforestation.