A field experiment was established at Port-Bouët in Ivory Coast (West Africa) on quaternary coastal sandy soils to evaluate the percentage of atmospheric nitrogen fixed by Acacia mangium in sylvicultural conditions. Three inoculation treatments with rhizobium, replicated in three blocks of a complete randomized block design were tested: a first treatment was composed of trees. inoculated with Aust13c, a Bradyrhizobium sp. strain previously selected on the basis of its high efficiency on A. mangium; a second one consisted of trees inoculated with CB756, a Bradyrhizobium sp. strain which was shown to be poorly efficient on A. mangium whereas the third one was the control treatment including the uninoculated trees. Twenty-one months after planting, we observed a positive effect of the inoculation with strain Austl3c on tree growth with a significant increase of 10% in height and 15% in diameter compared to the trees inoculated with CB756 or to the uninoculated control trees, the two latter treatments being not significantly different. The percentage of nitrogen derived from atmospheric N2 (Ndfa%) measured by the natural 1 5N abundance (a 15N) method in the leaves of 21 month-old A. mangium varied greatly and ranged from 19 to 91% depending on the block and on the inoculation treatments, with a major effect of the block factor. Ndfa% reached only 27% (for all inoculation treatments combined) in trees located in the most fertile block (block I) which was covered before planting by the perennial herbaceous legume Pueraria phaseoloides. In contrast, with a total mean of 65%, Ndfa% was 2.4 times higher in trees established in the blocks of low fertility (blocks II and III) initially covered by the graminaceous weed Imperata cylindrica.