Second-growth tropical forests are an important global carbon sink. As current knowledge on biomass accumulation during secondary succession is heavily based on chronosequence studies, direct estimates of annual rates of biomass accumulation in monitored stands are largely unavailable. We evaluated the contributions of tree diameter increment, recruitment, and mortality to annual tree biomass change during succession for three groups of tree species: second-growth (SG) specialists, generalists, and old-growth (OG) specialists. We monitored six second-growth tropical forests that varied in stand age and two old-growth forests in northeastern Costa Rica. We monitored these over a period of 8 to 16 years. To assess rates of biomass change during secondary succession, we compared standing biomass and biomass dynamics between second-growth forest stages and old-growth forest, and evaluated the effect of stand age on standing biomass and biomass dynamics in second-growth forests. Standing tree biomass increased with stand age during succession, whereas the rate of biomass change decreased. Biomass change was largely driven by tree diameter increment and mortality, with a minor contribution from recruitment. The relative importance of these demographic drivers shifted over succession. Biomass gain due to tree diameter increment decreased with stand age, whereas biomass loss due to mortality increased. In the age range of our second-growth forests, 10-41 years, SG specialists dominated tree biomass in second-growth forests. SG specialists, and to a lesser extent generalists, also dominated stand-level biomass increase due to tree diameter increment, whereas SG specialists largely accounted for decreases in biomass due to mortality. Our results indicate that tree growth is largely driving biomass dynamics early in succession, whereas both growth and mortality are important later in succession. Biomass dynamics are largely accounted for by a few SG specialists and one generalist species, Pentaclethra macroloba. To assess the generality of our results, similar long-term studies should be compared across tropical forest landscapes.