Previous studies have shown that tree species richness increases forest productivity by allowing for greater spatial complementarity of tree crowns (crown complementarity), which in turn results in more densely packed canopies. However, the mechanisms driving crown complementarity in tree species mixtures remain unclear. Here, we take advantage of a high-resolution, three-dimensional terrestrial laser scanning approach in the context of a large-scale biodiversity-ecosystem functioning experiment in subtropical China (BEF-China) to quantify the extent to which functional dissimilarity and divergences in branch traits between neighbouring trees affect crown complementarity at the scale of tree species pairs (i.e., two adjacent trees). Overall, we found no support that functional dissimilarity (divergence in morphological flexibility, specific leaf area and wood density) promotes crown complementarity. However, we show that the effects of functional dissimilarity (the plasticity of the outer crown structure) on crown complementarity vary in their magnitude and importance depending on branch trait divergences. Firstly, crown complementarity tended to be highest for tree species pairs that strongly differed in their functional traits, but were similar in branch density. In contrast, heterospecific pairs with a low functional trait divergence benefitted the most from a large difference in branch density compared with pairs characterised by a large functional dissimilarity. Secondly, the positive effects of increasing divergence in branching intensity (the plasticity of the inner crown structure) on crown complementarity became most important at low levels of functional dissimilarity, i.e. when species pairs were similar in their branch packing and vice versa. This suggests that species mixing allows trees to occupy canopy space more efficiently mainly due to phenotypic changes associated with crown morphology and branch plasticity. Our findings highlight the importance of considering outer and inner crown structures (e.g. branching architecture) to deepen our understanding of tree-tree interactions in mixed-species communities.