The natural 15N abundance method for estimating symbiotic biological N2-fixation was tested on legume trees from two rain forests on contrasting soils (oxisols and spodosols) in French Guiana. When possible, the significance of N2-fixing species in the plant community was evaluated in terms of density, biomass and contribution of N2-fixation to the building up of the total nitrogen mass in the leaves. Of the two sites, the rain forest on spodosols was the less favourable for application of the δ15N method: the available soil nitrogen was isotopically similar to fixed-N2. Hence, the results showed that a reliable estimate of N2-fixation could not be obtained. A substantial contribution of fixed-N2 to the nitrogen nutrition of legumes was found on oxisols, with an average value of 54 % Ndfa (Nitrogen derived from the atmosphere). The contribution of the N2-fixing legumes to the biomass of the stand was estimated to be 2 t ha−1 for the leaf biomass and 136 t ha−1 for the total above-ground plant biomass. With 7.5 % of trees in the stand able to fix N2 (462 out of 6156), N2-fixation was estimated to be 7 kg ha−1 y−1. These results are the first use of the δ15N method to estimate nitrogen input by N2-fixing legumes to a natural rain forest. The inter-site variability observed in the δ15N of the non-fixing plants suggested different nitrogen-cycling processes in the two soils. The δ15N of the non-N2-fixing plants could be related to the soil nitrogen availability and be used as an indicator of efficient or non-efficient nitrogen-cycling rain forests. The spatial variability of the δ15N in the plant-available soil nitrogen pool and the nitrogen balance in tropical rain forests are discussed.