In a greenhouse pot experiment we cultivated six meadow species in a replacement series design. The plants were grown at two sowing densities in monocultures and all possible species combinations. Our aim was to separate the selection effect from other diversity effects. This distinction is based on the notion that true overyielding is not a consequence of the selection effect. We suggest a hierarchical procedure, which is based on a repeated division of samples into the pots with the most productive species present and missing. Overyielding can be then demonstrated by a positive dependence of productivity on species richness in the subsets with the most productive species present. Although we found a strong dependence of biomass on species richness in the entire data set, the hierarchical method revealed no evidence of overyielding. Above-ground biomass in a monoculture was a good predictor of species success in a species mix.