Relationships between productivity and diversity in plant communities have been widely documented. Unimodal productivity-diversity relationships are most common along natural productivity gradients, and fertilization generally reduces diversity. Five distinct hypotheses invoke changes in competition to explain why diversity should decline from intermediate to high productivity. Because experiments measuring the effects of competition on diversity are rare, four of the five hypotheses have not been directly tested, but each hypothesis makes unique predictions that allow for indirect tests. The indirect evidence is often conflicting, and while none of the hypotheses can be rejected, only the dynamic equilibrium hypothesis is consistently supported. A new hypothesis, however, is supported by indirect evidence and may help to explain the variation in the shape of productivity-diversity relationships, as well as the most common patterns. Diversity may be high in environments that promote size symmetric competition, where soil resources limit growth and are homogeneously distributed within the soil volume explored by individual plants. Conversely, diversity may be low in environments that promote size asymmetric competition, where light is limiting, or where soil resources are limiting and are patchily distributed within rooting zones.