In an intensive mark-release-recapture study of all butterfly species in a tropical rainforest understory, 5903 individuals from 90 butterfly species (from the estimated total of 104 ᠹ species present in understory habitat) were marked, and 1308 recaptured at least once. The study proved that mark-recapture methods are feasible in tropical rainforests, but also showed its limitations, as after 232 person-days of sampling we could only characterise dispersal for one-third of the species present. The mean dispersal distance was 184 ᠴ6.1 m per species, while for six of the 14 species studied >1% of individuals were estimated to disperse 1 km or more. These parameters are, however, strongly dependent on the size and spatial configuration of the study plots, particularly in large homogeneous habitats. A new method proposed here to correct this bias revised the mean distance between two captures from 135 ᠳ3.6 to 325 ᠸ7.0 m per species. These results, in combination with data from large permanent rainforest plots, suggest that most woody plant species in tropical forests are sufficiently abundant to serve as host plant species even to monophagous Lepidoptera species.