This study examines the relationship between poverty and health in time. Following the argument that time is significant for shaping the experience of being poor or not poor and growing evidence of heterogeneity in long-term patterns of poverty, we investigate whether different kinds of poverty have distinct consequences for long-term health. Using data from the 1968-1996 annual waves of the United States Panel Study of Income Dynamics Data, we estimate a general growth mixture model to assess the relationship between the longitudinal courses of poverty and health. The model allows us to first estimate latent poverty classes in the data and then determine their effects on latent self-rated health. Four types of long-term poverty patterns characterized as stable nonpoor, exiting poverty, entering poverty and stable poor were evident in the data. These different kinds of poverty affected self-rated health trajectories in distinct ways, but worked in concert with age, education and race to create gaps in initial health status that were constant over time.