1. The biodiversity-ecosystem functioning debate is a central topic in ecology. Recently, there has been a growing interest in size diversity because body size is sensitive to environmental changes and is one of the fundamental characteristics of organisms linking many ecosystem properties. However, how size diversity affects ecosystem functioning is an important yet unclear issue. 2. To fill the gap, with large-scale field data from the East China Sea, we tested the novel hypothesis that increasing zooplankton size diversity enhances top-down control on phytoplankton (H1) and compared it with five conventional hypotheses explaining the top-down control: flatter zooplankton size spectrum enhances the strength of top-down control (H2); nutrient enrichment lessens the strength of top-down control (H3); increasing zooplankton taxonomic diversity enhances the strength of top-down control (H4); increasing fish predation decreases the strength of top-down control of zooplankton on phytoplankton through trophic cascade (H5); increasing temperature intensifies the strength of top-down control (H6). 3. The results of univariate analyses support the hypotheses based on zooplankton size diversity (H1), zooplankton size spectrum (H2), nutrient (H3), and zooplankton taxonomic diversity (H4), but not the hypotheses based on fish predation (H5) and temperature (H6). More in depth analyses indicate that zooplankton size diversity is the most important factor in determining the strength of top-down control on phytoplankton in the East China Sea. 4. Our results suggest a new potential mechanism, that increasing predator size diversity enhances the strength of top-down control on prey through diet niche partitioning. This mechanism can be explained by the optimal predator-prey body-mass ratio concept. Suppose each size group of zooplankton predators has its own optimal phytoplankton prey size, increasing size diversity of zooplankton would promote diet niche partitioning of predators and thus elevates the strength of top-down control.