Predation is a key determinant of prey community structure, but few studies have measured the effect of multiple predators on a highly diverse prey community. In this study, we asked whether the abundance, species richness, and species composition of a species-rich assemblage of termites in an Amazonian rain forest is more strongly associated with the density of predatory ants or with measures of vegetation, and soil texture and chemistry. We sampled termite assemblages with standardized hand-collecting in 30 transects arranged in a 5 km × 6 km grid in a terra firme Amazonian rain forest. For each transect, we also measured vegetation structure, soil texture, and soil phosphorus, and estimated the density of predatory ants from baits, pitfall traps, and Winkler samples. Seventy-nine termite species were recorded, and the total density of predatory ants was the strongest single predictor of local termite abundance (r = −0.66) and termite species richness (r = −0.44). In contrast, termite abundance and species richness were not strongly correlated with edaphic conditions (¦r¦ < 0.01), or with the density of non-predatory ants (rabund = −0.27; rs = −0.06). Termite species composition was correlated with soil phosphorus content (r = 0.79), clay content (r = −0.75), and tree density (r = −0.42). Assemblage patterns were consistent with the hypothesis that ants collectively behaved as generalist predators, reducing total termite abundance, and species richness. There was no evidence that ants behaved as keystone predators, or that any single termite species benefited from the reduction in the abundance of potential competitors.