Agroecological theory suggests that environmental impacts of agriculture can be reduced while maintaining yields by increasing diversity in agroecosystems. This increased diversity may support enhanced ecosystem services, reducing the need for inputs such as pesticides and fertilizers. Agricultural researchers have begun to test in agroecosystems what is already well established in unmanaged ecosystems; that more diverse communities, especially those with greater functional trait diversity, are more productive and tend to be higher functioning. Using data collected in the field of cover crop mixtures and a meta-analysis approach, I set out to test the effect of different levels of diversity (intraspecific, interspecific, and functional group), on a set of ecosystem services such as yield, weed suppression, and nitrogen fixation among others. Additionally, I attempted to evaluate the effect of environment on the relationship between diversity and ecosystem service performance, again using multiple methods including a mother-daughter design on working farms. The meta-analysis of cultivar mixtures showed an overall 2.2% relative yield benefit in food crops, however increased intraspecific diversity had only a minimal effect in cover crop mixtures in the two year field experiment at the research farm or on farms. There was a significant reduction in weed biomass in cultivar mixtures overall. Ultimately, there was no negative impact on any ecosystem service. Cover crop mixtures of multiple species, especially those with both grass and legume functional groups, often had better biomass production compared to monocultures. The grass/legume mixtures buffered the effects of soil nitrogen fertility much better than either functional group alone for some services. In both the meta-analysis and field trials, more stressful environmental conditions tended to accentuate the diversity effect. In other words, the benefit of more diverse communities was greater in higher stress environments, lending some support to the stress-gradient hypothesis. The on-farm results showed that while functional group mixtures were always as good as or better than monocultures, the precise relationship between these different levels of diversity for the different services varied depending on the location. Overall, my results indicate benefits from all levels of diversity, with more functionally diverse mixtures delivering greater ecosystem service performance, especially in higher stress environments.