Results/Conclusions The Connell Plots have been especially useful for assessing the evidence for non-random processes that create structure in forests, particularly those that promote diversity. In this study, we used data collected over 35 years to assess the hypotheses that the degree of non-random mortality (NRM) in communities of long-lived rainforest plants is highest among the smallest size classes, and that within size classes, observed levels of NRM mortality result in significantly higher diversity among the survivors than that derived from models of mortality based on purely random expectations. We considered a cohort of 7924 plants that were alive on the northern Connell Plot in 1971, and assigned them to one of nine size classes from tiny seedlings, through tall seedlings, saplings, poles and large trees. Monte Carlo simulations were used to determine the percentage (P) of species in each size class that died non-randomly, and to determine the effect of this NRM on the species richness (S) of the survivors. P was indeed highest in the smallest size classes; 23% in seedlings to 12 cm tall, 11% in seedlings 13-24 cm tall, and 7% in seedlings 25-48 cm and 49-98 cm tall, while ≤ 5% of species died non-randomly in all larger size classes to large canopy trees. NRM had a significant but modest impact on S in the two smallest size classes only. These analyses are consistent with Connell's intuition that dynamism in the smallest size classes of plants contributes most to determining the abundance and diversity of canopy trees.