34 páginas, 4 figuras, 8 tablas. The structure and seasonal changes of earthworm communities were evaluated in a natural savanna and in a improved grass-legume pasture in a Colombian oxisol over a period of 18 months. One plot of 90×90 m was isolated in each of the systems and each month five samples of 1 m2×0.5 m and ten of 20×20×20 cm were randomly selected from a stratified block design. Species richness was similar in the two evaluated plots (seven species), whereas diversity measured by the index, H (Shannon and Weaver 1949) was clearly different, i.e. H=2.89 in natural savanna and H=1.29 in pasture. This is explained by differences in earthworm community structure. The average annual density in the savanna was 49.8, ranging from 10.8 to 135.8 individuals (ind) m–2, and biomass was 3.3 g m–2 (hand-sorting method), ranging from 0.9 to 11.5 g m–2. In the man-made pasture, density was 80.1 ind m–2 on average, ranging from 24 to 215.8 ind m–2 and biomass was more than tenfold higher, ranging from 29.2 to 110.4 g m–2. This was especially due to the presence of a large glossoscolecid anecic species, Martiodrilus carimaguensis Jiménez and Moreno, which has been greatly favoured by conversion of savanna to pasture. Endogeic species were dominant in the natural savanna whereas the anecic species accounted for 88% of total earthworm biomass in the pasture. Total earthworm density and biomass were significantly different in the two systems studied (t-test). The results indicate a clearly positive response of earthworm communities to improved pastures, a type of land use that is being increasingly adopted in moist neotropical savannas. This study was part of collaborative research between the STD-3 Macrofauna Project, an international programme supported by the European Community focused on the biology, ecology and possibilities of management of earthworm species present in natural and distributed ecosystems in the tropics, and the Tropical Lowlands Program at the International Center for Tropical Agriculture, CIAT (International Centre for Tropical Agriculture; Cali, Colombia). This study was supported in part by a research grant within the Macrofauna programme.We want to thank CIAT, especially all at the Tropical Lowlands Program for scientific, technical and human support. The invaluable help provided by field workers at Carimagua in this back-breaking work is undoubtedly greatly appreciated. Finally, the first author wishes to thank two anonymous referees for their helpful comments on the previous manuscript. Peer reviewed