We studied a humid savanna rangeland, abandoned in 2000, where intensive cattle grazing had led to widespread encroachment by Acacia zanzibarica. We asked whether the acacia trees were able to regenerate in the absence of domestic livestock, either beneath acacia canopies or in artificial clearings. Tropical coastal Tanzania (former Mkwaja Ranch, now in Saadani National Park). We set out a total of 48 plots on four sites in November 2001, and assigned them to three treatments: trees felled (FN), trees felled and the stumps poisoned (FP) with Triclopyr, and no intervention (controls, NN). We analysed soils of plots for texture and nutrients. In two wet (July 2002 and 2003) and one dry (February 2003) seasons we assessed grass and tree leaf biomass and transpiration rates, and counted acacia seedlings and resprouts. The effects of treatments (controlled for site and other co-variables) on grass growth and acacia recruitment were determined statistically using general linear models (GLM). Acacia leaves had a much higher stomatal conductance than grasses, with the consequence that total evapotranspiration in woodland was higher than in clearings. In the wet seasons, grass biomass and seedling densities were significantly higher in clearings than in control plots, which we attributed to more favourable moisture conditions. In the dry season, by contrast, we found no differences, and all seedlings had died. On FN plots, 71% of stumps, and on FP plots, 11% resprouted (coppicing), but only a quarter of these shoots survived until July 2003. Root suckering occurred spontaneously at low densities. No root suckers or resprouts grew beyond the grass layer. Acacia woodlands do not regenerate in the absence of cattle grazing, and tree cutting – in combination with appropriate fire management – could potentially accelerate re-establishment of open grassland. However, regeneration might occur in the future due to the increasing wildlife populations within the new national park.