Restoring small-scale habitat heterogeneity in highly diverse systems, like tropical forests, is a conservation challenge and offers an excellent opportunity to test factors affecting community assembly. We investigated whether (1) the applied nucleation restoration strategy (planting tree islands) resulted in higher habitat heterogeneity than more homogeneous forest restoration approaches, (2) increased heterogeneity resulted in more diverse tree recruitment, and (3) the mean or coefficient of variation of habitat variables best explained tree recruitment. We measured soil nutrients, overstory and understory vegetation structure, and tree recruitment at six sites with three 5- to 7-year-old restoration treatments: control (no planting), planted tree islands, and conventional, mixed-species tree plantations. Canopy openness and soil base saturation were more variable in island treatments than in controls and plantations, whereas most soil nutrients had similar coefficients of variation across treatments, and bare ground was more variable in control plots. Seedling and sapling species density were equivalent in plantations and islands, and were substantially higher than in controls. Species spatial turnover, diversity, and richness were similar in island and plantation treatments. Mean canopy openness, rather than heterogeneity, explained the largest proportion of variance in species density. Our results show that, whereas canopy openness and soil base saturation are more heterogeneous with the applied nucleation restoration strategy, this pattern does not translate into greater tree diversity. The lack of a heterogeneity-diversity relationship is likely due to the fact that recruits respond more strongly to mean resource gradients than variability at this early stage in succession, and that seed dispersal limitation likely reduces the available species pool. Results show that planting tree islands facilitates tree recruitment to a similar degree as intensive plantation-style restoration strategies.