The development of human capabilities for many disadvantaged children around the world depends on growth recovery (‘catch-up growth’). Here we develop a novel framework that allows different types of catch-up growth to be classified and estimated. We distinguish between catch-up in the mean of a group toward that of a healthy reference population versus catch-up within the group. We show these different growth types can be tested in a unified setting using a latent growth framework. We apply the results to four developing countries, using longitudinal data on 7641 children collected over the period 2002–2013. The results show catch-up growth rates are generally modest but vary significantly between countries, and that local environmental factors are material to variation in child growth trajectories. The paper discusses the benefits of the new framework versus current methods, shows that the method is feasible, and suggests they call for intervention designs that are sensitive to community and country contexts.