Although small-scale agroforestry systems (swiddens, complex and single-crop-dominated agroforests, and homegardens) form a diverse and important tropical land use, there has been no attempt to collate information on their value for biodiversity. This paper reviews 52 published studies that compared species richness and/or abundance between agroforests and primary forest, and 27 studies that compared biodiversity parameters across agroforests. The former covered a broad range of taxa and geographical areas, but few focused on homegardens, while those comparing across agroforestry systems were biased towards studies of plants (21 studies) and homegardens (13 of 27). Of 43 studies comparing species richness or diversity across habitats, 34 reported lower richness in agroforests than in adjacent forest. There was also high β diversity between primary forests and agroforests. Patterns of abundance shifts were less straightforward, with many species traits (for example diets) being generally poor indicators of response to agricultural disturbance. Among the few trends identified, restricted-range or rare species, and terrestrial and some understorey vertebrates tended to decline most, and open country species, granivores and generalists increased most in agroforests. Variability in biodiversity retention across systems has been linked most strongly to economic function, management intensity and extent of remnant forest within the landscape, as well as more subtle cultural influences. Species richness and abundance generally decrease with increasing prevalence of crop species, more intensive management, decreasing stratum richness and shortening of cultivation cycles. Increasing holding size did not necessarily reduce α diversity. Knowledge of the general effects of small-scale agroforestry on biodiversity is substantial, but the great diversity of systems and species responses mean that it is difficult to accurately predict biodiversity losses and gains at a local level. Further work is required on the influence of spatial and temporal structure of agricultural holdings on biodiversity retention across agriculture/succession/forest mosaics, how β diversity across individual holdings influences biodiversity across landscapes, and ultimately on how agricultural intensification can be best managed to minimize future losses of biodiversity from tropical landscapes.