In a dynamic social or biological environment, the interactions between the underlying actors can undergo large and systematic changes. Each actor in the networks can assume multiple related roles and their affiliation to each role as determined by the dynamic links will also exhibit rich temporal phenomenon. We propose a state space mixed membership stochastic blockmodel which captures the dependency between these multiple correlated roles, and enables us to track the mixed membership of each actor in the latent space across time. We derived efficient approximate learning and inference algorithms for our model, and applied the learned models to analyze an email network in Enron Corp., and a rewiring gene interaction network of yeast collected during its full cell cycle. In both cases, our model reveals interesting patterns of the dynamic roles of the actors.