Correct assessment of the rhizodeposition of N in grassland is essential for the evaluation of biological N 2 -®xation of legumes, for the total N balance of agro-ecosystems, and for the pre-cropping value of grasslands. Using a leaf-feeding technique by which plants were 15 N labelled while growing in mezotrons in the ®eld, the rhizodeposition of N by unfertilised red clover, white clover and perennial ryegrass growing in pure stands was shown to amount to 64, 71 and 9 g N m 22 , respectively, over two complete growing seasons. The corresponding values for red clover and white clover growing in mixtures with ryegrass were 89 and 32 g N m 22 , respectively. The rhizodeposited N compounds, including ®ne roots, constituted more than 80% of the total plant-derived N in the soil, and in all cases exceeded the amount of N present in stubble. In the mixtures of red clover±ryegrass and white clover±ryegrass and the pure stands of red clover, white clover and ryegrass, respectively, the rhizodeposition constituted a 1.05, 1.52, 1.26, 2.21 and 2.77 fold increase over the total N in the shoots harvested during the two production years. In pure stands and mixtures of clover, 84 and 92%, respectively, of this N derived from biological N 2 ®xation. It is concluded that rhizodeposition provides a very substantial input of N to the legume-based grassland systems with great consequences for ecosystem N balance and turnover. Furthermore, the amount of atmospheric-derived N in the rhizodeposits may exceed that in the harvested shoots. q