This study investigated different techniques of grassland restoration to overcome dispersal or establishment limitation, which are key processes influencing early-successional plant community assembly. A fully randomized in situ experiment was set up in a former arable land in a floodplain along the Garonne river (south-western France) to test for the effect of (1) the type and completeness of the biological input (any biological material—seed, hay, and soil—transferred to the site under restoration), (2) soil disturbance by deep tillage, and (3) their interaction on plant community dynamics for 5 years. All inputs influenced the plant community structure and composition. The effects depended on the type of the input but not necessarily in link with its completeness. High density and diversity seed mixture led to high levels of richness and relative abundance of target species like soil blocks, the input considered as the most complete. During the first year, hay transfer mostly influenced community assembly through negative litter effects. Delayed germination of several species contributed to buffer these early effects. This study supports the importance of dispersal limitation during early succession in degraded grassland ecosystems whereas soil disturbance had only subtle effects on the seed bank and standing plant community, indicating that competition and establishment limitation were of secondary importance in our study system. Our results suggest that even low-frequency immigration events (e.g. when a few seeds are transferred within hay) can be of great importance providing that a sufficiently long time period is considered for grassland restoration.< Réduire