A continuously grazed trial on natural and sown pastures under coconuts was reported by Watson & Whiteman (1981). This has been continued for another 3 years as a rotationally grazed trial. The paddocks were divided down the centre and three animals grazed for 28 days in one half and were transferred to the other half for 28days. The same three stocking rates of 1·5, 2·5 and 3·5 steers/ha were maintained. In the sown pasture paddocks Brachiaria decumbensand Brachiaria miliiformis were planted in rows from rooted cuttings, while Centrosema pubescens (centro) and Pueraria phaseoloides (puero) were sown, with seed at 4 and 3 kg/ha, respectively. Naturalized pastures were maintained by grazing at 2·5 steers/ha during the establishment of the sown pasture. Light transmission was 62 %. Sown grasses were lost after 4 months from the start of rotational grazing. In the sown pastures puero became dominant. Animal live-weight gain (LWG) on puero was poor in the first 140 days of grazing in each year, but after this time it increased. Centro was the dominant species in the natural pasture of 2·5 and 3·5 steers/ha. However, there was little difference in yearly LWG except in year 3 when the 2·5 steers/ha had higher gains in the natural (363 kg/ha) than in the sown pasture (250 kg/ha). Th9 mean live-weight gain (kg/ha) in the first 2 years of the rotationally grazed trial (315 kg/ha) was lower than in the continuously grazed trial (364 kg/ha) at 3·5 steers/ha. The time and efficiency of collecting fallen coconuts was lower at 1·5 steers/ha in both pastures, but not different between 2·5 and 3·5 steers/ha. In continuous and rotational grazing there is little point in planting Brachiaria species. Productive pastures in plantations oan be obtained by conti oiling weeds and sowing centro and puero into the naturalized Axonopus compressus and stocking at 2–5–3–0 steers/ha. Further research on other grasses such as Ischaemum aristatum and Stenotaphrumsecundatum may lead to better grass persistence.