Growing evidence has revealed that ecosystem productivity depends more on the functional characteristics of species than on their number. However, just how the extent of tree diversity effects on ecosystem productivity is influenced by functional trait variability and composition has been rarely tested across and within species richness levels. Employing a meta-analysis of data from 59 global scale tree diversity experiments, we examined how functional dispersion and identity determine the outcomes of tree mixture effects on productivity, both across and at given species richness levels. We found that the positive effects of tree mixtures on productivity were strengthened by the increasing multidimensional functional dispersion and the community-weighted mean of leaf nitrogen content both across, and within, two- and four-species mixtures. Our analysis provides mechanistic insights into the potent roles of functional trait attributes in determining the magnitude (and even directionality) of the biodiversity-ecosystem functioning relationship in forest ecosystems.