The Janzen-Connell hypothesis is one of the most important ideas to have been put forward to explain the extraordinary species diversity of trees in tropical rainforests. The hypothesis uses the idea that trees encounter high pressure from specialist natural enemies which attack neighbouring conspecifics, with the result that any species at high density is likely to suffer reduction in growth and survival. One of the expectations from this hypothesis is that adult trees will be less aggregated than juvenile trees since nearby individuals will be removed. Indeed such is the weight of force behind this idea that many ecology textbooks argue that if adults are more aggregated than juveniles then this suggests that negative density dependence is not very important in these populations. Empirical studies are equivocal as to how often adults are less aggregated than juveniles. One of the largest studies showed that out of 543 species only 84 were significantly more aggregated at the juvenile stage than at the adult stage (Condit et al. 2000). Despite this there are numerous studies that show interactions between neighbours are very important to the growth and survival of individuals of all size classes. How can these two results be reconciled? So far there has been little theory that has tackled long-term population dynamics with both size structure and neighbourhood interactions. I present a model, based on the method of moment closure, where dispersal is localised in space, and where both growth and survival are dependent on the local density of neighbours. The results show that it is entirely possible for adult trees to be more aggregated than juveniles even if the negative effects of local density are large. These results show once again how difficult it is to interpret community patterns without first considering population dynamics.