The question of whether ecological assemblages are structured by stochastic and deterministic (e.g. interspecific competition) processes is controversial, but it is difficult to design sampling regimes and experiments that can dissect the relative importance of stochastic and deterministic processes in natural assemblages. Using null models, we tested communities of arthropod decomposers in tropical epiphytes for patterns of species co-occurrence, while controlling for habitat gradients, seasonal variations and ecological succession. When environmental conditions were controlled, our analysis showed that the communities were structured stochastically. However, analysing mixed sets of communities that were deliberately created either from two distinct heights or two successional stages revealed that communities were structured deterministically. These results confirm that habitat gradients and dispersal/competition trade-offs are capable of generating non-random patterns within decomposer arthropod communities, but reveal that when such effects are accounted for, species co-occurrence is fundamentally random.