The unified neutral theory of biodiversity and biogeography has gained the status of a quantitative null model for explaining patterns in ecological (meta)communities. The theory assumes that individuals of trophically similar species are functionally equivalent. We empirically evaluate the relative contribution of neutral and deterministic processes in shaping fruit-feeding butterfly assemblages in three tropical forests in Africa, using both direct (confronting the neutral model with species abundance data) and indirect approaches (testing the predictions of neutral theory using data other than species abundance distributions). Abundance data were obtained by sampling butterflies using banana baited traps set at the forest canopy and understorey strata. Our results indicate a clear consistency in the kind of species or species groups observed at either the canopy or understorey in the three studied communities. Furthermore, we found significant correlation between some flight-related morphological traits and species abundance at the forest canopy, but not at the understorey. Neutral theory's contribution to explaining our data lies largely in identifying dispersal limitation as a key process regulating fruit-feeding butterfly community structure. Our study illustrates that using species abundance data alone in evaluating neutral theory can be informative, but is insufficient. Species-level information such as habitat preference, host plants, geographical distribution, and phylogeny is essential in elucidating the processes that regulate biodiversity community structures and patterns.