Understanding how wood decomposition varies spatially at the mesoscale (between 1-ha plots) may improve carbon flux estimates in Amazonian forests. An experiment was carried out to test the influence of soil, slope, above-ground tree live biomass (biomass), fine-litter mass and characteristics of neighbouring trees on the variation of wood decomposition between 1-ha plots in four species of tropical trees that vary in wood density (Manilkara huberi – 0.86 g cm−3, Couratari guianensis – 0.54 g cm−3, Hura crepitans – 0.32 g cm−3 and Parkia pendula – 0.29 g cm−3). A wood sample from each species (30 × 5 × 2.5 cm) was placed in each of 71 plots within 64 km2 of terra firme tropical moist forest in Reserva Florestal Adolpho Ducke. One year later, samples were collected and weighed. The effects of specificity of decomposers was measured by the association of decomposition with the wood density and with the taxonomic group of the nearest tree with dbh ≥30 cm. Wood decomposition was independent of soil (texture and nutrients), slope, biomass and fine-litter mass at the mesoscale, except for C. guianensis, which showed greater decomposition in locations with greater biomass. Decomposition was also independent of wood density and taxonomic group of nearby large trees. In general, none of the variables was useful as a predictor of wood decomposition at the scale larger than 1 ha. Thus, the use of models that include soil and topography to improve estimates of carbon flux are limited because wood decomposition does not follow similar mesoscale patterns to that of biomass and fine-litter decomposition. Also, the results indicate that wood decomposition is more likely to be associated with generalist decomposers than with specialists associated with neighbouring trees.