SummaryResearch on the function of biodiversity has stimulated a renewed interest in the mechanisms underlying the invasibility of plant communities. Conflicting results provide evidence that observational correlative studies are a dead end if isolated from manipulative experiments. Using two studies in France, Portugal and Australia we illustrate how invasion research can move from pattern observation to the investigation of processes.Artificial annual communities representing a factorial combination of one to three functional groups and one to six species per group were sown in Montpellier. Natural invasibility of the plots from the seed bank showed no strong relationship with functional or species richness. The survival of individuals of two congeneric exotic species (Conyza canadensis and C. bonariensis) introduced as seedlings did not relate to these simple diversity indices. On the other hand, their adult biomass did with a decrease in vegetative performance with increasing species richness but a decrease with increasing number of functional groups. Survival and biomass of invaders were strongly influenced by the identity of the resident functional groups.In a study of the mechanisms underlying the demographic success of the pasture weed Echium plantagineum in Australia, germinations in mixtures with varying numbers, densities, and identities of co‐germinants showed no evidence that increasing diversity increased the competitive effect. Rather, significant reduction of recruitment resulted from increased densities.These examples suggest that patterns and mechan‐ isms of community invasibility are likely to be far more subtle than people have acknowledged so far. Controlled experiments coupled with theoretical work are required to advance our understanding.