Four latosolic soils, varying in total nitrogen content from 0.050 to 0.098 per cent and in organic matter from 0.95 to 1.56 per cent, were collected from the farm of University College, Ibadan, Nigeria. No change in nitrogen content was observed when the soils were kept moist in the greenhouse for four months. Where a legume (Centrosema pubescens) was grown gains of nitrogen in the soil—plant system equivalent to from 100 to 200 pounds nitrogen per acre — 6 inches deep were obtained. Where a grass (Eleusine coracana) was grown gains of nitrogen in the soil—plant system equivalent to from 100 to 130 lb. nitrogen per acre — 6 inches deep occurred in the two soils with low initial nitrogen levels. The possible origin of this nitrogen accretion is discussed. The data support, but do not prove, the hypothesis that it is due to non-symbiotic fixation in the soil. In one of the soils showing a significant nitrogen gain under grass the presence of Azotobacter was not demonstrated and Beijerinckia occurred in only one of 24 samples examined. Thus, if non-symbiotic biological nitrogen fixation did occur, Clostridium and other micro-organisms not belonging to the Azotobacteriaceae must have been the agents of fixation involved. On the other hand, the presence of Azotobacter in the two soils in which no fixation occurred could easily be demonstrated. It is suggested that non-symbiotic nitrogen fixation by soil micro-organisms under grass cover in the humid tropics may be of greater significance than has been believed in the past.