By taking the governing measure of a Dirichlet process as essentially the realization of a Dirichlet process itself, a hierarchical Dirichlet process emerges. This distribution enables the Bayesian treatment of two level hierarchical random effects models, with results paralleling those of Ferguson (1973).