Assessing the recovery of species diversity and composition after major disturbance is key to understanding the resilience of tropical forests through successional processes, and its importance for biodiversity conservation. Despite the specifi c abiotic environment and ecological processes of tropical dry forests, secondary succession has received less attention in this biome than others and changes in species diversity and composition have never been synthesised in a systematic and quantitative review. Th is study aims to assess in tropical dry forests 1) the directionality of change in species richness and evenness during secondary succession, 2) the convergence of species composition towards that of old-growth forest and 3) the importance of the previous land use, precipitation regime and water availability in infl uencing the direction and rate of change. We conducted meta-analyses of the rate of change in species richness, evenness and composition indices with succession in 13 tropical dry forest chronosequences. Species richness increased with succession, showing a gradual accumulation of species, as did Shannon evenness index. Th e similarity in species composition of successional forests with old-growth forests increased with succession, yet at a low rate. Tropical dry forests therefore do show resilience of species composition but it may never reach that of old-growth forests. We found no signifi cant diff erences in rates of change between diff erent previous land uses, precipitation regimes or water availability. Our results show high resilience of tropical dry forests in term of species richness but a slow recovery of species composition. Th ey highlight the need for further research on secondary succession in this biome and better understanding of impacts of previous land-use and landscape-scale patterns.Succession has been a major focus of plant community ecology for more than a century (McIntosh 1999), yet the processes underlying assembly of secondary forests continue to be actively researched (Norden et al. 2015). Th e early view of Clements (1916) (monoclimax hypothesis) and initial fl oristic composition model of Egler (1954) were further formalized by the three models of Connell and Slatyer (1977). Th ese deterministic models are all based on the idea that tradeoff s between traits promote success in diff erent stages of succession (Huston and Smith 1987). However, none of these models accounts for stochastic events or historical contingencies (Young et al. 2005). Lawton (1987) proposed a model of succession based only on random survival of established species and colonization by new species, paving the way for the neutral theory of community assembly (Hubbell 2001). Although deterministic niche-based and neutral models of succession have often been treated as mutually exclusive explanations for empirical patterns, a growing body of literature evidences the importance of the integration of the two (Chave 2004, Gravel et al. 2006, Tilman 2004. A major focus of research is now to assess the res...