The presence of large variation in speciation and extinctin rates across the tree of life has long been hypothesized to be driven by the evolution of traits that affect diversification. To test such hypotheses, phylogenetic biologists have developed a wide class of state-dependent birth-death processes that jointly model trait evolution and the diversification process. However, it has since been shown that these models are sensitive to falsely linking traits to diversification. Here we present a Bayesian approach to test the adequacy of statedependent birth-death models by statistically checking whether they describe the variation observed in the data. Our method generates posterior predictive distributions for a suite of informative test statistics, providing a general framework for testing diversification processes and models of trait evolution. We simulate data sets under different violations of model assumptions and find that our approach successfully detects the inadequacy of the model for them. We further show that the manner in which a model fails to fit the data can reveal insights into the processes of trait evolution and diversification.