A central goal of community ecology is to understand the forces that maintain species diversity within communities. The traditional niche-assembly theory asserts that species co-occur in a community only when they differ from one another in resource use. But this theory has some difficulties to explaining the diversity often observed in species-rich communities such as tropical forests. As an alternative to niche theory, Hubbell and other ecologists introduced a neutral model. Hubbell argues that the number of species in a community is controlled by species extinction and immigration and speciation of new species. Assuming that all individuals of all species in a trophically similar community are ecologically equivalent, Hubbell's neutral theory predicts two important statistical distributions. One is the asymptotic log-series distribution for the metacommunity under point mutation speciation, and the other is the zero-sum multinomial distribution for both local community under dispersal limitation and metacommunity under random fission speciation. Unlike the niche-assembly theory, the neutral theory takes similarity in species and individuals as a departure for investigating species diversity. Based on the fundamental processes of birth, death, dispersal and speciation, the neutral theory first presented a mechanism that generates species abundance distributions remarkably similar to those observed in nature. Since the publication of the neutral theory, there has been much discussion of it, pro and con. In this paper, we summarize new progresses in research on assumption, prediction and speciation mode of neutral theory, including progress in the theory itself and tests about the theory's assumption, prediction and speciation mode at metacommunity level. We also suggest that the most important task in the future is to bridge the niche-assembly theory and the neutral theory, and to add niche-differences in neutral theory and more stochasticity into niche theory.