Free-response (FR) item formats, such as essay questions, are popular in educational assessment. The criticisms against FR items are that they are more expensive to score, take up more testing time, provide less content coverage, and are less reliable than multiple-choice (MC) items. For these reasons, FR items are often combined with MC items. Calibrating FR and MC items onto a common scale poses problems for tests constructed using item response theory (IRT), because IRT calibration may reduce the weight of the FR items below the level desired by the test developer. This paper illustrates how MC and FR items can be calibrated onto a common scale using the Multilog computer program and how an a priori weighting scheme can be incorporated within the scale. The relative worth of the MC and FR portions of the test was evaluated with respect to reliability, test information, and passing status using the Writing Skills Test of the Tests of General Educational Development (GED). Data were from a sample of 1,986 GED candidates. Results suggest that to maximize reliability using an IRT calibration model, the weights for the MC and essay sections should not be adjusted a posteriori. (Contains 5 tables, 3 figures, and 23 references.) (Author/SLD)