The development of vesicular-arbuscular mycorrhizal (VAM) symbiosis was monitored in Leucaena leucocephala grown in an Oxisol subjected to incremental simulated erosion. The density of VAM infective propagules in the soil diminished as the level of simulated erosion (removal of surface soil) was increased from 0 to 50 cm. The level of infection on L. leucocephala roots observed at harvest was not significantly influenced by simulated erosion unless removal of surface soil exceeded 25 cm. Inoculation of this soil and the uneroded soil with Glomus aggregatum enhanced the early onset of infection but did not significantly influence the level of infection observed at the time of harvest. Simulated erosion in excess of 7.5 cm of surface soil removal significantly delayed the development of VAM effectiveness monitored in terms of the P status of L. leucocephala subleaflets and also curtailed the level of maximum effectiveness observed. Decreases in VAM effectiveness were significantly correlated with decreases in soil chemical constituents. However, VAM effectiveness in a soil subjected to 30 cm of surface soil removal was not restored to a significant extent unless the soil was amended with P, even though other nutrients were restored to sufficiency levels. Our results demonstrate that the development of VAM effectiveness is the phase of the VAM symbiosis that is most adversely influenced by simulated erosion and that this effect appears to be caused primarily by insufficient P in the soil solution.