Abstract Aim In this paper we aim to show that proportional sampling can detect species–area relationships (SARs) more effectively than uniform sampling. We tested the contribution of alpha and beta diversity in ant communities as explanations for the SAR. Location Tropical forest remnants in Viçosa, Minas Gerais, Brazil (20 °45′ S, 42 °50′ W). Methods We sampled 17 forest remnants with proportional sampling. To disentangle sampling effects from other mechanisms, species richness was fitted in a model with remnant size, number of samples (sampling effects) and an interaction term. Results A SAR was observed independent of the number of samples, discarding sampling effects. Alpha diversity was not influenced by remnant size, and beta diversity increased with remnant size; evidence to the fact that habitat diversity within remnants could be the dominant cause of the SAR. Such a relationship between beta diversity and remnant area may have also arisen due to the combined effects of territoriality and aggregation of ant species. Main conclusions The proposed model, together with proportional sampling, allowed the distinction between sampling effects and other mechanisms.