Results/Conclusions   We present a generalization of Hubbell's Unified Neutral Theory of Biodiversity and Biogeography. The original neutral model has demonstrated its ability to reproduce many widespread ecological patterns such as locally sampled species abundance distributions and species area curves, but it has mostly failed to combine this with convincing timescales for the turnover of species amongst other things. The existing work attempting to solve this problem introduces new modes of speciation for the neutral model. However, all of these either fail to fit species abundance distributions, or have no known solution for the species abundance distribution in a local community. We introduce a new mode of speciation, which unlike the alternatives, is a gradual process rather than an instantaneous event. The new model is simple enough to remain tractable, both analytically and computationally. We show that it resolves several timescale related problems without sacrificing any of the existing achievements of the original model. Comparisons with empirical data of tropical trees, corals and reef fish show that the new model fits species abundances at least as well as the classic neutral model, but also makes much more reasonable predictions for speciation rate and species lifetimes. Neutral theory can thus be defended against many long-standing criticisms and now makes new predictions about species abundance distributions at large scales. We conclude that it is both necessary and straightforward to model speciation as a gradual process in future studies of neutral models.