We are conducting latent growth modeling with a data set of adolescentsí psychosocial adjustment, assessed prospectively across 5 annual waves of data. In many of our models, we are observing a significant, negative correlation between latent intercept and latent slope, suggesting regression to the mean. More problematically, some time 1 variables are associated with latent slopes in the opposite direction of the observed time 1 to time 1 relationship between the variables. (For example, time 1 body image satisfaction is inversely correlated with time depression, but time 1 body image satisfaction is positively associated with depression slope.) Does anyone have any recommendations for readings or resources that address the potential problem of regression to the mean in latent growth modeling? Thanks!