XCALIBRE represents a vastly improved approach to the challenge of IRT item parameter estimation. A significant amount of attention during the past several years has been directed at marginal maximum-likelihood estimation methods for obtaining IRT parameter estimates. Accurate estimation of IRT item parameters using maximum likelihood approaches, including Bayesian techniques such as those implemented in ASCAL, tend to require longer tests and larger sample sizes, generally at least 50 items and 1,000 examinees for a 3-parameter IRT model. Using marginal maximum-likelihood estimation, however, XCALIBRE provides accurate item parameter estimates with fewer examinees and shorter tests. This program: * is able to estimate both 2- and 3-parameter IRT item parameters; * works with very large data sets; * implements Bayesian prior distributions (floating priors) on the individual item parameters as ASCAL does, but these prior distributions may be updated during the estimation process; and * "fixes" common (anchor) items in a larger data set to specified parameter values (from a previous estimation run). Fixing selected item parameters: * results in automatic linking of the remaining item parameters onto the target (common) scale; * provides an evaluation of the fit of each item to the IRT model; and * works with incomplete data sets and can differentiate between items that are omitted and those that were not attempted. XCALIBRE requires a 386 processor (or higher) running DOS 3.3 (or higher), or Windows 3.1 (or higher) with 2MB RAM. A 486DX or math coprocessor is strongly recommended.