Hierarchical Bayes procedures were compared for estimating item and ability parameters in item response theory. Simulated data sets from the two-parameter logistic model were analyzed using three different hierarchical Bayes procedures: (1) the joint Bayesian with known hyperparameters (JB1); (2) the joint Bayesian with information hyperpriors (JB2); and (3) the marginal Bayesian with known hyperparameters (MB). Prior and posterior distributions focusing on one- and two-stage hierarchical priors are presented, and two joint Bayesian methods that consider the specific priors are discussed. The MB procedure yielded consistently smaller root mean square differences than did either the JB1 or JB2 procedure for item and ability estimates. The maximum a posteriori estimation used along with the MB procedure yielded larger biases than did the joint Bayes model estimation in JB1 and JB2. As the sample size and test length increased, the three Bayes procedures yielded essentially the same result. Fifteen tables present study data, and there is a 39-item list of references. (SLD)