We used matrix models to investigate the relation between population dynamics of the liana Mansoa difficilis and environmental factors in fragmented Atlantic forest in Brazil. The fate (growth and mortality) of individuals and the number of new individuals were recorded for 3 years in 100 plots of 10 m × 10 m each. We used multinomial logistic regressions to assess the influence of environmental factors on the fate of individuals in different life stages. Adopting AIC for model selection we tested a range of models including different groups of environmental variables: soil nutrients water availability light availability and tree community structure. With the fates predicted by the best model we constructed a matrix model for each plot to calculate population growth rate (?) in that plot. The average ? was 0.962 for 2012–2013 and 0.941 for 2013–2014 both significantly lower than equilibrium (? = 1). In both periods elasticity was higher for survival of large climbers than for other fates. The best models varied between life stages and periods indicating that the impact of environmental factors on demographic rates changed through time. Model selection suggested that the influence of light availability on ? was less important than tree community structure which allowed high population growth rates only in a small part of the forest. These findings support the notion that tree community characteristics are the key to understand and predict the observed recent increase in the density of lianas in the Neotropics.