Plant diversity is generally thought to enhance productivity, driven by either (1) chance inclusion of highly productive species in more diverse communities or (2) niche-based resource acquisition with competitive interactions increasing resource use efficiency. Here, we ask whether weeding, as employed in most experiments to date, might contribute to the positive diversity-productivity relationship reported for many grasslands. Using all 82 species from our local pool, we constructed 357 experimental grassland plots (2 × 4 m each), arranged as a completely randomized experiment in an arable field prepared to minimize existing seed bank. The plots were sown to vary species richness (1, 2, 4, 6, 8, 10, 15, 20, 25, 30, 35 or 40 species) and were maintained under both ambient conditions and experimental drought. A single monoculture plot was maintained for all 82 species, and each of the other eleven species richness levels was replicated 25 times. Plots were maintained strictly without weeding, and aboveground biomass was measured at 17, 19, 27 and 29 months after the start of this experiment. No single measure of biodiversity was significantly correlated with productivity consistently across all four sampling periods. Furthermore, there were only weak overall effects of six biodiversity variables (the species richness planted, observed, and sampled; Shannon diversity, effective species richness and evenness in the sampled area) on productivity under either precipitation treatment. Regression analysis identified no equation that used a consistent subset of the biodiversity measures as predictors. In view of these transient and insubstantial effects, results from previous experiments that employed weeding treatments are suspect as tests of the hypothesis that biodiversity has positive effects on productivity.