A framework is described for exploiting auxiliary information about test items within item response theory models to enhance parameter estimates. The method also provides diagnostic information about items' operating characteristics. An empirical Bayesian estimation of Rasch item difficulty is used to illustrate the principles involved. (TJH)