Year-to-year stability in crop production is a crucial aspect of feeding a growing global population. Evidence from natural ecosystems shows that increasing plant diversity generally increases the temporal stability of productivity; however, we have little knowledge of the mechanisms by which diversity affects stability. In fact, understanding the drivers of stability is a major knowledge gap in our understanding of biodiversity and ecosystem function in general. We varied resource inputs into crop monocultures and intercropping of maize/pea and maize/rapeseed for 3 years in field experiments to create a wide range of values for temporal stability, complementarity effects, selection effects, competition, and facilitation. We correlated whole-system temporal stability in productivity with these values and the stability of competitively subordinate species and competitively dominant species in the intercrops. We then used structural equation modeling (SEM), which combines complex path models with latent variables, to estimate how interspecific interactions for water, nitrogen, and phosphorus affected the relationships between stability and these values. Intercropping treatments did not increase stability, but the wide range of stability created by our experiments allowed us to explore the relationship of many factors with stability. Complementarity correlated positively with the temporal stability of grain yield and aboveground biomass, suggesting that either facilitative interactions or niche partitioning shifted over time in ways that promoted stability. Furthermore, the temporal stability of total productivity of intercropping relied most on the stability of more productive species. However, facilitation tested by relative interaction index independently did not correlate with stability, but the temporal stability of the whole system increased as the competitive effects of competitively dominant species (pea and rapeseed) on competitively subordinate species (maize) decreased and was highest when these competitive effects were virtually zero. SEM indicated that as competition for soil nitrogen from competitively dominant species on competitively subordinate species decreased, the overall temporal stability of whole-system aboveground biomass increased. This stability then led to greater stability in grain production. Our findings indicate that complex shifts in complementarity and competitive intensities are likely to be key mechanisms that maintain temporal stability in species-diverse agriculture and, potentially, in natural systems.