The effect of plant species diversity on productivity and competitive ability was studied in an experiment carried out simultaneously in five European countries: Czech Republic (CZ), the Netherlands (NL), Sweden (SE), Spain (SP), and United Kingdom (UK). The aim was to separate the 'chance' or 'sampling effect' (increasing the number of sown species increases the probability that a species able 'to do a job' will be included) from the complementarity effect (species-rich communities are better able to exploit resources and to take care of ecosystem functions than species-poor communities). In the experiment, low diversity (LD) and high diversity (HD) mixtures of grassland species were sown into fields taken out of arable cultivation. The HD mixture consisted of five grass species, five legumes and five other forbs. The LD mixtures consisted of two grasses, one legume and one other forb, with different plant species combinations in each replicate block. The design of the experiment was constructed in such a way that the total number of seeds of each species over all the replications was exactly the same in HD and LD treatments, and the total number of grass seeds, leguminous seeds and other forb seeds were the same in both LD and HD. The responses measured were the total above-ground biomass las a measure of productivity) and the average number of naturally establishing species in a plot las a measure of the competitive ability of the mixture), both measured in the third year of the experiment. The results show that, on average, the HD plots performed better (i.e., attained higher biomass, had better weed suppression), but that the best LD mixture was as good as the best HD mixture. On the contrary, the worst LD mixture was always less successful than the worst HD replicate. The performance of particular species in the HD mixtures was a good predictor of the success of a certain species combination in a LD mixture (explaining 61% of variability between particular LD mixtures). In all sites, the LD mixture composed of species which were the most abundant in HD mixtures was as efficient in suppressing weeds as the HD mixture itself. It is argued that the performance of a species assemblage is influenced mostly by the identity of species and the diversity effect is mainly due to the 'chance' or 'sampling' effect with increasing number of species the probability that an important species will be included in the mixture increases. Caution is urged in interpreting experiments with manipulated diversity and the possible limitations of such experiments are discussed.