Legume-based intercropping is an interesting practice to increase nitrogen (N) content in the soil and associated crop, and so to reduce the use of fertilizers. The aim of this study was to test how useful the natural 15N abundance method is for estimating this increment. A greenhouse trial was carried out in pots with canavalia-banana association, in order to evaluate N transfer via exudates and root decomposition after cutting the aboveground parts. Estimations were recorded by a box model including isotopic fractionation in the soil and banana plants. The contribution of the total N taken up by banana was: 5 % from canavalia exudates, 53 % released from root decomposition and 42 % from the soil. Root exudate contribution was relatively low due to a strong competition between canavalia and banana, mainly for the available soil phosphorus. 15N content of the youngest banana half-leaf showed a high correlation with 15N content of the entire plant. This indicated that the youngest leaf may be a useful index to estimate leguminous N transfer. Both, N transfer determination along canavalia growing cycle and the evaluation of the actual N transfer rate need to be calibrated under field conditions for each type of soils.