Hubbell's (2001) neutral model describes how local communities are structured if population dynamics are statistically identical among species in a constant, possibly patchy, environment with random speciation. Tests of this model have been restricted largely to terrestrial communities. Here we tested the fit of this neutral model to fish, zooplankton and phytoplankton species–abundance distributions from 30 well-studied lake communities varying widely in lake size and productivity. We measured the fit of the communities to the neutral model in three ways. All but two zooplankton (7 of 9) and all but three fish (9 of 12) communities were consistent with all three measures of fit. However, all nine phytoplankton communities did not fit the neutral model by at least one measure. This result for phytoplankton communities represents to date the most consistent failure of the standard neutral model to predict the shape of species-abundance distributions.